Cell - 28 March 2013

Leading Edge
In This Issue
Mobile Elements Mobilize Oncogenes
PAGE 101
Shukla et al. show that L1 retrotransposons, common in the human genome, can mobilize and activate oncogenes in the livers of individuals infected with hepatitis B or C virus, promoting development of hepatocellular carcinoma. Specically, insertions in the mutated in colorectal cancer gene activate oncogenic Wnt signaling, whereas other insertions interrupt a negative feedback loop regulating the tumor suppressor ST18. These genes present new options for cancer screening and intervention and suggest that retrotransposon activation may be an important tumorigenic mechanism.
Oncogenic Remodeler
PAGE 71
Translocation
Remodels
Chromatin
A translocation that fuses the SS18 gene on chromosome 18 to a related gene on X (SSX) is the likely initiating event in human synovial sarcoma, a soft-tissue malignancy. Kadoch and Crabtree now elucidate the transformation mechanism by showing that SS18 is a subunit of the chromatin-remodeling complex BAF. The SS18-SSX fusion competes with wild-type SS18, creating an altered BAF complex that lacks the tumor suppressor BAF47 and hence activatesrather than suppressesSox2, driving proliferation.
Neurotransmitter Signaling Drives Cancer Progression

PAGE 86
Interstitial uid pressure and glutamate receptors have been independently implicated in tumor progression. Li and Hanahan now demonstrate that pressure-driven ow conditions activate an autocrine glutamate-to-NMDA receptor (NMDAR) signaling circuit in cancer cells, involving the CaMK and MEK pathways, thereby enhancing invasiveness and proliferation. Elevated expression of NMDAR plus glutamate transporters correlates with poor prognosis in certain human cancers, and inhibition of NMDAR signaling in a mouse model impairs tumor progression.
Lipid Protector Battles Inuenza

PAGE 112
There is a need for new and more effective drugs for treating inuenza, given the potential for future lethal pandemics. Morita et al. report that the lipid mediator protectin D1 (PD1) suppresses inuenza replication by impeding the nuclear export of viral RNAs. Administration of PD1 protects mice against the lethal 2009 H1N1 inuenza even at later stages of infection when current antiviral therapies are ineffective.
Antibody Mutations that Are Worth the Wait

PAGE 126
Antibodies typically accumulate somatic hypermutations in the complementarity determining regions (CDR). CDRs are scaffolded by canonical framework regions (FWRs) that are less tolerant of mutations. Yet, Klein et al. show that FWR mutations play a crucial role in antibodies with broad and potent HIV-1 neutralization activity (bNAbs). This may explain the long latency period required for the appearance of bNAbs in HIV-1-infected individuals and should be considered in HIV-1 vaccine design.
Achilles Heel for Glioblastoma in GSC-Derived Vasculature

PAGE 139
Glioblastomas are highly vascular, lethal brain tumors that contain tumorigenic self-renewing glioma stem cells (GSCs). Using in vivo lineage tracing, Cheng et al. now show that GSCs give rise to vascular pericytes, cells that support blood vessel function. Eliminating GSC-derived pericytes disrupts the neovasculature and inhibits tumor growth, suggesting that targeting of pericytes may be an effective approach that could complement current anti-angiogenic therapies.
Y an RNA Reshapes Enzyme Function

PAGE 166
In animal cells and some bacteria, the Ro protein binds noncoding RNAs of unknown function called Y RNAs. Chen et al. show that a bacterial Y RNA tethers Ro protein to the exoribonuclease polynucleotide phosphorylase (PNPase), altering its specicity such that it specializes in breaking down structured RNAs. This work shows that by tethering a protein cofactor, a noncoding RNA can alter the substrate specicity of an enzyme.
Cell 153, March 28, 2013 2013 Elsevier Inc. 1
Mitosis Reshufes the DNA Deck at the Nuclear Lamina

PAGE 178
The nuclear lamina interacts with hundreds of large genomic regions termed lamina-associated domains (LADs). Using a new visualization technique, Kind et al. show that in each nucleus only a third of LADs are positioned at the periphery and that upon mitosis, LAD positioning is not detectably inherited but instead is stochastically reshufed. In single cells, contact between the nuclear lamina and individual LADs is linked to transcriptional repression and H3K9 dimethylation.
Prions Foster Budding Friendships

PAGE 153
Holmes et al. show that a prion formed by the Mot3 transcription factor governs the acquisition of facultative multicellularity in the budding yeast. The prions were induced by ethanol and eliminated by hypoxiaconditions that occur sequentially in the natural respiro-fermentative cycles of yeast populations. Their ndings demonstrate that prions can act as environmentally responsive molecular determinants of multicellularity and contribute to the natural morphological diversity of budding yeast.
Horses for Courses in DNA Methylation

PAGE 193
DNA methylation regulates genes and transposable elements across eukaryotes. Zemach et al. study the RNA-dependent DNA methylation (RdDM) pathway and the nucleosome remodeler DDM1, which are both required for DNA methylation in Arabidopsis. They show that RdDm and DDM1 act separately: DDM1 makes heterochromatic regions bound by linker histone H1 accessible to methyltransferases, whereas RdDM methylates less heterochromatic regions and relies on a different nucleosome remodeler, DRD1. The work demonstrates that chromatin structure is a key regulator of DNA methylation.
Speedy Ligase ExchangeYes, We Cand1!

PAGE 206
SCF complexes comprise a large family of ubiquitin ligases that share core components yet have different substrate receptor proteins. Combining a real-time FRET assay and quantitative proteomics, Pierce et al. show that Cand1 functions as an exchange factor for SCF complexes by accelerating their dissociation by one-million-fold. Cand1 thus ensures rapid equilibration of cullin complexes with a large network of substrate-binding modules.
Golgi Lipids Drive Hypertrophy

PAGE 216
A major feature of cardiac remodeling in response to stress is hypertrophic growth. Zhang et al. now show that conditional deletion of PLC in mouse cardiac myocytes protects against stress-induced hypertrophy. Surprisingly, PLC acts on PI4P in the Golgi rather than the canonical plasma membrane PLC substrate PIP2. In response to hypertrophic signals, PLC signaling drives the expression of hypertrophy genes.
Life-Extending Drug Acts through Microbes

PAGE 228
The drug metformin is widely prescribed to treat type 2 diabetes and can also slow aging in animal models, but its mechanism of action remains unclear. Cabreiro et al. provide evidence that the life-extending effect of metformin in C. elegans stems from its impact on host-microbe interactions. Metformin alters microbial folate and methionine metabolism, inducing methionine restriction in the host, thus slowing aging.
Lifestyle Choice Affected by Diet

PAGE 240
MacNeil et al. show that signals from C. elegans bacterial food source affect life-history traits such as developmental rate, reproductive timing, and lifespan. Relative to the standard laboratory diet of E. coli, a diet that includes Comamonas bacteria, even in small amounts, accelerates worm development by altering the larval molting program via a mechanism that is independent of TOR and insulin signaling but likely involves nuclear hormone receptors. These ndings show how diet can impact lifestyle strategies.
Dietary Response Networks

PAGE 253
Watson et al. use a combination of forward and reverse genetics to identify a network of 184 C. elegans genes that affect its response to a diet of Comamonas bacteria. Many of these genes encode mitochondrial enzymes involved in amino acid metabolism and the TCA cycle, and many correspond to genes implicated in inborn errors of metabolism in humans. Furthermore, the results reveal extensive communication between mitochondrial metabolic networks and nuclear gene regulatory networks controlling dietary responses.
Leading Edge
Select
Linking Long Noncoding RNAs to Function
Although the vast majority of the human genome is devoid of protein-coding capacity, we now know that amidst the dark matter are an estimated 10,000 to 200,000 long noncoding RNAs (lncRNAs). This Select highlights recent insights into the regulation of lncRNAs expression and their unique roles.
Where Do Phenotypes eXIST?

Xist encodes a lncRNA that is responsible for mediating X chromosome inactivation (XCI) to achieve dosage compensation in mammalian females. Previous work has indicated that binding of pluripotency factors, such as Oct4, Sox2, and Nanog, to the rst intron of the Xist gene (intron 1) is critical for repression of Xist expression in undifferentiated mouse embryonic stem cells (ESCs). Thus, the ndings reported by Minkovsky et al. (2013) that intron 1 appears dispensable for Xist-expression regulation and developmental dynamics of XCI are surprising. Normal growth and reproductive phenotypes displaced in intron-1-null mice and the random XCI observed in heterozygous females support their conclusion. The authors further rule out the requirement of intron 1 for the downregulation of Xist and subsequent loss of XCI during reprogramming to pluripotency. However, the authors also describe a mild skewing effect of XCI toward intron-1-depleted allele and a slight upregulation of Xist clouds in differMouse embryonic broblasts isolated from entiating intron 1 and Tsix double-depleted male mouse ESCs. The females lacking Xist intron 1 display normal discrepancy between the normally occurring in vivo XCI choice and the Xist RNA coating. Image courtesy of K. Plath. ex vivo XCI-skewing phenotype in the absence of intron 1 suggests a more robust Xist regulatory mechanism in vivo than in vitro. Functional dissection of the genetic and molecular bases of this intricate network, which is composed of multifaceted cis elements and trans-factors, controlling Xist expression warrants further investigation. Minkovsky, A., et al. (2013). Cell Rep., Published online March 21, 2013. http://dx.doi.org/10.1016/j.celrep.2013.02.008.
Transcription versus Transcript

It has been speculated that some lncRNAs merely represent by-products of transcription and therefore have limited specic biological functions. The catalogs of lncRNAs in human ESCs presented by Sigova et al. (2013) elucidate their genomic origin and regulation. The majority (89%) of Airn lncRNA is expressed paternally and lncRNAs species are associated with the promoters, enhancers, and silences Igf2r in cis. Image courtesy of F. Pauler bodies of protein-coding genes, among which a surprisingly large fraction and D. Barlow. (65%) originate divergently from regions near the promoters of active coding genes. In addition, changes in transcription of lncRNAs and their neighboring coding gene pairs tend to be coordinately regulated upon ESC differentiation. Together, these results suggest that divergent transcription regulates lncRNA and protein-coding gene pairs throughout the genome. However, whether a lncRNA exerts its function through the RNA transcript itself or the act of the transcriptional activity of the lncRNA gene remains unclear. Latos et al. (2013) provide mechanistic insights into how a paternally expressed 118 kb lncRNA, Airn, silences in cis the paternal allele of Igf2r. By genetically shortening Airn transcripts and repositioning the Airn promoter, the authors assign the Airn repressor activity to a 4 kb region overlapping the Igf2r promoter. Strikingly, replacing this 4 kb region with a Tk-neo reporter gene that lacks any nucleotide or structural similarity also preserves the imprinting. These data support the transcriptional interference model whereby Airn transcription per se, but not Airn lncRNA products, suppresses the Igf2r promoter by reducing the recruitment of RNA polymerase II. Interestingly, silencing of Slc22a3, which is restricted to the placenta, in this imprinted cluster depends on the Airn lncRNA transcript recruiting a histone methyltransferase, revealing that one lncRNA has evolved diverse mechanisms to regulating gene expression. Sigova, A.A., et al. (2013). Proc. Natl. Acad. Sci. USA 110, 28762881. Latos, P.A., et al. (2013). Science 338, 14691472.
Lncing Yin and Yang

A growing body of evidence in both plants and animals supports the notion of target mimicry, known as the competing endogenous RNA theory, in which mRNAs, transcribed pseudogenes, and lncRNAs inuence each others level by competing for a limited pool of microRNAs. For example, posttranscriptional expression of the tumor suppressor PTEN is regulated by a PTEN pseudogene (PTENpg1) through its microRNA sponging activity. Johnsson et al. (2013) identify two PTENpg1 antisense (as)RNA isoforms, a and b, that regulate PTEN with opposite effects. The cytosolic b asRNA interacts with and stabilizes the PTENpg1 sense transcript, further enhances microRNA sequestration, and ultimately upregulates translated PTEN. In contrast, the a asRNA stays in the nucleus and acts in trans at the PTEN promoter, resulting in the recruitment of chromatin regulators and the epigenetic suppression of PTEN transcription. Thus, these ndings establish a bimodal mechanism in which a pseudogene regulates its protein-coding gene counterpart via the interplay between sense RNA, PTENpg1 antisense RNA nuclear a isoform asRNA, and microRNA at both transcription and posttranscription stages. suppresses PTEN transcription, whereas the Examining the hypothesis that emerged from this study, whereby the cytosolic b isoform upregulates PTEN translaswitch or crosstalk between a and b asRNA isoforms balances the function via the PTENpg1 sense transcript. Image tional output, will shed light on the intricate PTEN regulatory network and courtesy of K. Morris. tumor biology. Johnsson, P., et al. (2013). Nat. Struct. Mol. Biol., Published online February 24, 2013. http://dx.doi.org/10.1038/nsmb. 2516.
Super Sequestration
Both pseudogene transcripts and lncRNAs have been previously reported to sequester microRNAs. A pair of papers now add circular (circ) RNAs to this list. A circRNA, named ciRS-7 by Hansen et al. (2013) and CDR1as by Memczak et al. (2013), is expressed in neuronal tissues, harbors approximately 70 microRNA-7 (miR-7)-binding sites, and provides a platform for trapping the miR-7/Argonaute complex. Its super sponging capacity antagonizes miR-7 activity. As shown by Memczak et al., the heterologous expression of ciRS-7/CDR1as in zebrash embryos recapitulates the miR-7 depletion defects in midbrain. Interestingly, miR-671 interacts with ciRS-7/CDR1as, and unlike miR-7, the interaction induces the endonucleolytic cleavage of ciRS-7/CDR1as, offering a potential destruction mechanism to trigger the release of the miR-7/Argonaute complex from ciRS-7 sequesters and inactivates miR-7 but circRNAs. Hansen et al. further report that another known circRNA, which is resistant to miR-7 regulation due to its originates from testis-specic Sry gene, can act as a miR-138 decoy, circular nature, whereas miR-671 cleaves suggesting the sequester effects achieved by circRNA are a general ciRS-7, recovering the function of miR-7. Image phenomenon. Indeed, comprehensive transcriptomic analyses by Mem- courtesy of J. Kjems. czak et al. reveal that thousands of circRNAs are derived from worm to human genomes. Their cell-type- and developmental-stage-specic expression patterns and sequence conservation suggest that circRNAs form a novel class of posttranscriptional regulators to compete with other RNAs for microRNA and protein binding. Hansen., T.B., et al., Nature (2013). Published online February 27, 2013. http://dx.doi.org/10.1038/nature11993. Memczak., S., et al., Nature (2013). Published online February 27, 2013. http://dx.doi.org/10.1038/nature11928.
Steve Mao
Leading Edge
Voices
Noncoding RNAs and Cancer
Knockout Punch for Cancer The Promise of Small RNAs Genomic Dark Matter
Judy Lieberman
Boston Childrens Hospital
Frank Slack
Yale University
Pier Paolo Pandol

Beth Israel Deaconess Medical Center, Harvard University
A decade after RNAi was found in mammals, clinical studies using microRNA antagonists and lipid-encapsulated siRNAs are showing promise for inhibiting hepatitis C virus replication and knocking down genes in the liver. The biggest challenge to RNA-based cancer therapeutics remains delivery into cells outside of the liver, which is critical for treating disseminated cancer. Attractive approaches to targeted delivery being developed avoid liposomes and use RNAs covalently linked to small-molecule conjugates or RNA aptamers. siRNA cocktails can sidestep the dangers of cancer escape mutations, since multiple mutations generally compromise tness. The choice of siRNA targets may need to be customized based on each tumors unique dependencies; siRNA drugs are ideally suited to personalized therapy. siRNA cocktails could be selected, based on a tumors molecular ngerprint, from a siRNA pharmacy of the future containing hundreds of RNAs. In addition, mimics of tumor suppressor microRNAs (which have the same delivery issues as siRNAs) could provide a way to manipulate large networks of genes to inhibit tumor proliferation or metastasis or induce differentiation. It will be difcult for tumors to escape from these broadly acting RNAs. It is becoming clear that long noncoding RNAs regulate transcription, maintain heterochromatin, promote enhancer function, and help to repair DNA damageall processes central to malignant transformation and cancer treatment. RNA-based drugs that manipulate these newer noncoding RNA functions provide a therapeutic opportunity worth exploring.
From the initial reports of microRNAs and RNAi in C. elegans, it is now clear that small RNAs play big regulatory roles in almost all species and regulated processes. Given their roles in regulating key human disease genes, they now show promise as therapeutics for a whole range of human ills. Our understanding of microRNAs in cancer is best developed, and even though the rst human microRNA was only discovered in 2000, the eld has already moved past the discovery stage into the translational phase. Because certain microRNAs are overexpressed in cancer and have oncogenic properties (oncomiRs), whereas others are lost in cancer and have tumor suppressor (TS) properties, the eld has a two-pronged strategy: inhibit the oncomiRs (therapeutic target strategy) and restore the TS microRNAs (targeted therapeutic strategy). Both strategies show promise alone in mouse models of cancer, but it is likely that a combination of the two will be required for efcacy in the heterogeneous background of human cancer. This nascent eld has beneted immensely from the prior knowledge and chemical innovations developed for oligo and siRNA therapeutics, and it suffers from some of the same drawbacks and challenges. Enhancing effective delivery and stability while eliminating potential side effects are perennial issues being addressed with novel nanoparticle strategies. The ingenuity of nanoengineering is highly likely to make this promise a reality for patients.
I believe that cancer research, diagnosis, and care will be revolutionized by the impact of the noncoding RNA dimension. However, despite intensive research, the functions of the noncoding portion of the genome have largely remained uncharted territory. We discovered that pseudogenes, noncoding RNA species, and, importantly, proteincoding mRNAs can function as competitive endogenous RNAs (ceRNAs) and compete for a limited pool of microRNAs. Such competition leads to crosstalk that can be predicted based on number and identity of shared microRNA response elements (MREs), the letters of the MRE language, and allows the formation of complex regulatory networks. Such networks are crucial in lending robustness to evolutionarily conserved biological systems while permitting speciation and diversication but also promote disease when their balance is perturbed. The identication of the ceRNA dimension will allow us to unravel and functionalize large portions of the dark matter of the genome. Once such knowledge has been integrated with our model of the protein-coding world, we will be one step closer to understanding the biology of cells. We are currently approaching this by interrogating one noncoding RNA at a time in cells and animal models. I believe, however, that bringing together systems biologists, mouse modelers, bioinformaticians, and molecular biologists will more rapidly identify the critical nodes in these complex networks and their contribution to disease. This, in turn, will provide novel potential drug candidates that can be explored for therapeutic purposes.
lncRNAs Complicate Cancer
Assessing RNA Therapies
Revealing ncRNA Disease Roles
Arul Chinnaiyan
University of Michigan
Reuven Agami
The Netherlands Cancer Institute
Joshua T. Mendell
UT Southwestern Medical Center
Emerging evidence suggests that the vast, often uncharted, noncoding landscape of the human genome plays an important role in biology and disease progression. Small ncRNAs, such as microRNAs, have well-established differential patterns of expression in cancer and function as tumor suppressors or oncogenes by silencing target gene expression. Much less is known about their longer cousins, long noncoding RNAs (lncRNAs), subsets of which have been characterized as epigenetic factors, enhancers, antisense transcripts, and pseudogenes. The popularization of RNAseq technology to interrogate the transcriptome has revealed the pervasive expression of lncRNAs in cancer progression. While the research community is still in the early days of characterizing lncRNAs across tumor types, lineagespecic and sometimes cancer-specic patterns of expression are emerging. The function of most lncRNAs in cancer remains a mystery. Prototypical lncRNAs have been shown to be overexpressed in subsets of cancers, interact with epigenetic complexes, be recruited to DNA sequences as enhancers, and function in RNA-RNA hybrids. A major challenge in understanding the function of a specic lncRNA in cancer is deciphering its interactomewhether it be with specic protein complexes, regions of DNA, or other RNAs. Several high-throughput technologies are now emerging to address this gap. Ultimately, lncRNAs will likely have implications in cancer management, including the development of more specic cancer biomarkers, due to their lineage- and cancer-specic properties.
The past decade has witnessed a swift advance in the identication of microRNAs with disease-specic expression patterns and those with causative roles in disease development andin the case of cancerin resistance to therapy. This has led to the initiation of clinical trials testing oligonucleotide-based therapies inhibiting or mimicking microRNAs. With their simple design, scalability, and systemic delivery to the body, oligonucleotide-based therapy is easier to develop, improve, and put to practice than geneand drug-based therapies. The outcome of these trials will determine how valuable microRNAs are as therapeutic targets. Concerning cancer, I thus believe that the greatest challenge ahead is to convincingly demonstrate efcient oligonucleotide delivery accompanied by long-lasting effects on tumor cells in patients. Particularly, specicity and safety of treatment should be demonstrated, as some microRNAs can function as both oncogenes and tumor suppressors, depending on the targeted tissue. In contrast to microRNAs, the research eld of lncRNA lags far behind. Their recent identication, long size, low expression level, and excessive structure delay progress. Although they are as powerful as microRNAs in controlling essential cellular processes, dedicated preclinical investigation of lncRNAs is required to decode the language that they talk and the words they use in order to discover the best ways to control their function. Only when such a pronounced fundamental understanding is achieved can we start implementing lncRNAs in therapy.
Over the last decade, numerous studies have documented the potent pro- and antitumorigenic functions of noncoding RNAs (ncRNAs) and, particularly, microRNAs (miRNAs). Understanding how aberrant ncRNA activity mechanistically contributes to tumorigenesis and predicting the sequelae of targeting ncRNAs therapeutically require a deeper understanding of the roles of ncRNAs in normal physiology in vivo. Functional studies of miRNAs and other ncRNAs in knockout and transgenic mice have repeatedly revealed surprisingly subtle phenotypes that can be enhanced by various forms of stress, including acute injury and chronic disease. Based on this emerging paradigm, animals with ncRNA knockout or overexpression should be exposed to broad panels of perturbations to reveal defective pathways that might not be apparent in controlled laboratory environments. Mechanistically dissecting these phenotypes in whole animals poses a daunting challenge. It is often stated that miRNAs ne-tune the expression of many messenger RNAs simultaneously. Likewise, long ncRNAs are often components of broadly acting chromatin-modifying complexes. Subtle regulation of a large number of targets may, in aggregate, result in profound effects on cellular behavior. Yet it remains possible that many ncRNA-driven phenotypes result from the regulation of a few key targets hidden within these extensive networks. Uncovering these critical nodes, should they exist, will be essential for understanding ncRNA functions in physiology and pathophysiology. Moreover, these crucial effectors of ncRNA activity may themselves represent druggable targets, potentially offering greater specicity and ease of targeting than ncRNAs themselves.
10 Cell 153, March 28, 2013 2013 Elsevier Inc.
Leading Edge
Previews
Synovial Sarcoma Mechanisms: A Series of Unfortunate Events
Jesper Q. Svejstrup1,*
1Cancer Research UK, London Research Institute, Clare Hall Laboratories, Blanche Lane, South Mimms, Hertfordshire EN6 3LD, UK *Correspondence: j.svejstrup@cancer.org.uk http://dx.doi.org/10.1016/j.cell.2013.03.015
Human synovial sarcoma is caused by a chromosome translocation, which fuses DNA encoding SSX to that encoding the SS18 protein. Kadoch and Crabtree now show that the resulting cellular transformation stems from disruption of the normal architecture and function of the human SWI/SNF (BAF) complex.
Damage to our chromosomes is an ines- transforming fusions are often between recognition sites for the DNA targeting capable consequence of cell metabolism a protein with a general function in tran- domain could be inappropriately activated and division. DNA damage can take scription or in chromatin modication (or, alternatively, repressed), resulting in many forms, but because their repair is and a targeting domain in another cellular transformation. A much studied not always precise, double-strand DNA protein (a DNA-binding domain, for example is the fusion of the DNA-binding (dsDNA) breaks are among the most example). It is easy to imagine how such domain of MLL to subunits of the so-called dangerous. Indeed, if a dsDNA break is re- protein fusions might lead to cellular super elongation complex (Smith et al., paired by nonhomologous end-joining deregulation: genes in the vicinity of the 2011). As with mixed-linage and acute rather than homologous myeloid leukemias (MLL and recombination, it can potenAML, respectively), human sytially result in the fusion of novial sarcomas are caused DNA ends that were not supby translocations generating posed to be joined. Over the new fusion proteins. The last couple of decades, it translocation event observed has become clear that chroin this type of cancer fuses mosome breakage is not the SS18 gene on chromocompletely stochastic; cersome 18 and one of three tain regions are much more SSX genes (SSX1, SSX2, and likely to break than others, SSX4) found in a cluster on and moreover, certain specifchromosome X, which generic chromosome fusion events ates a stable SS18-SSX fusion are also more likely than protein (Figure 1, upper). others. In some cases, chroAlthough earlier data sugmosome translocations even gested that SS18 interacts fuse protein-coding genes in with chromatin-remodeling frame so that a specic, novel complexes (Thaete et al., protein species arises. The 1999; Nagai et al., 2001), the existence of recurring fusion molecular consequence of events has become clear fusion protein expression from the study of human canhave been unclear. cers, including several types In a detailed and compelling of leukemia, and the examstudy published in this ples of oncogenic transforissue of Cell, the mechanism mations correlating with the of cellular transformation in generation of fusion proteins human synovial sarcoma is Figure 1. Generation and Effect of an SS18-SSX Fusion Protein are unexpectedly plentiful (renow unveiled. Kadoch and (Upper) Chromosomal translocation t(X;18)(p11.2;q11.2) fuses the SS18 gene viewed by Saha and Jones, Crabtree (2013) initially pro(on chromosome 18) to SSX1, SSX2, or SSX4 (on the X chromosome), resulting in a fusion protein in which the eight C-terminal amino acids of SS18 are re2005; Taki and Taniwaki, vide evidence that SS18 is placed with 78 amino acids from the relevant SSX C terminus. (Lower) The 2006). Interestingly, in the a hitherto overlooked, integral SS18-SSX fusion protein (green-red) is incorporated into the multisubunit BAF last couple of decades subunit of the human SWI/SNF complex (right), replacing the SS18 protein (green) of normal BAF complexes (left). This results in eviction and degradation of the BAF47 subunit (right). research has shown that the (BAF) chromatin-remodeling
complex. It is tightly associated with the catalytic Brg subunit, dissociating from the multisubunit complex at a much higher urea concentration than the well-known BAF47/hSNF5/INI1 or BAF250/ARID1 subunits, for example. Importantly, the SS18-SSX fusion protein becomes incorporated into the BAF complex in place of SS18, and this in turn results in the eviction, and subsequently proteasomal degradation, of the BAF47 subunit (Figure 1, lower). BAF47 is already a well-established tumor suppressor. For example, loss of the BAF47 gene causes extremely aggressive malignant rhabdoid tumors (MRTs), and its re-expression in MRT cells stops their proliferation (Kia et al., 2008). It might therefore be expected that eviction of BAF47 also plays an important role in human synovial sarcoma tumorigenesis. In agreement with this idea, the altered BAF complex binds the Sox2 locus and reverses polycomb-mediated repression, resulting in activation of this pluripotency gene. Sox2 is uniformly expressed in human synovial sarcoma tumors and is
essential for their proliferation, so its anomalous activation may well be transformative. It is intriguing that eviction of BAF47, and thus transformation, depends on only two amino acids of the SSX protein, explaining why SSX1, SSX2, and SSX4, but not SSX3, are observed in synovial sarcoma fusion proteins: SSX3 has methionine-isoleucine in place of the evicting lysine-arginine amino acid pair found in the otherwise highly conserved SSX homologs. Altogether, this fascinating story of a unique oncogenic transformation mechanism underscores the frustratingly random nature of human cancer: if it invariably elicits efcient programs to drive cellular transformation, even an exceedingly rare and unlikely event like that in human synovial sarcoma may become a recurring human health issue. Encouragingly, the ndings of Kadoch and Crabtree indicate potential avenues of therapeutic intervention. As the authors point out, iffor examplea decoy molecule could be developed that causes the BAF47-evicting amino acids of the transformative SSX molecules to
resemble the corresponding surface of the benign SSX3 protein it would offer some hope for the development of a new treatment that builds on understanding the fusion proteins unusual mechanism of action.
REFERENCES Kadoch, C., and Crabtree, G.R. (2013). Cell 153, this issue, 7185. Kia, S.K., Gorski, M.M., Giannakopoulos, S., and Verrijzer, C.P. (2008). Mol. Cell. Biol. 28, 3457 3464. Nagai, M., Tanaka, S., Tsuda, M., Endo, S., Kato, H., Sonobe, H., Minami, A., Hiraga, H., Nishihara, H., Sawa, H., and Nagashima, K. (2001). Proc. Natl. Acad. Sci. USA 98, 38433848. Saha, S., and Jones, L.K. (2005). Fusion proteins and diseases (Chichester, UK: John Wiley & sons). Smith, E., Lin, C., and Shilatifard, A. (2011). Genes Dev. 25, 661672. Taki, T., and Taniwaki, M. (2006). Curr. Opin. Oncol. 18, 6268. Thaete, C., Brett, D., Monaghan, P., Whitehouse, S., Rennie, G., Rayner, E., Cooper, C.S., and Goodwin, G. (1999). Hum. Mol. Genet. 8, 585591.
Ring around the Ro-sie: RNA-Mediated Alterations of PNPase Activity

Brian J. Geiss1,* and Jeffrey Wilusz1,*
1Department of Microbiology, Immunology and Pathology, Colorado State University, Fort Collins, CO 80523, USA *Correspondence: brian.geiss@colostate.edu (B.J.G.), jeffrey.wilusz@colostate.edu (J.W.) http://dx.doi.org/10.1016/j.cell.2013.03.005
Chen et al. demonstrate a new way by which noncoding RNAs tailor the function of multicomponent complexes. They show that a noncoding RNA interacts with an exoribonuclease, altering its substrate specicity and enzymatic activity by serving as a ribonucleoprotein scaffold and, perhaps, a gate for entry of the RNA substrate.
Multiprotein complexes are the workhorses of the cell and provide critical functions that are necessary for cellular growth and viability by merging related activities into compact molecular machines. Protein-protein interactions are well known to be involved in allosteric regulation, altering substrate specicity and localization of enzymatic function to specic subcellular compartments. Several RNAs that serve as scaffolds for such molecular machines have been described, including yeast TLC1 RNA and telomerase (Lebo and Zappulla, 2012), pRNA and the 29 DNApackaging motor (Harjes et al., 2012), and IRES elements and translation factors. The ability of RNAs to scaffold molecular machines is also being investigated for synthetic biology applications (Delebecque et al., 2012). Given the
complex. It is tightly associated with the catalytic Brg subunit, dissociating from the multisubunit complex at a much higher urea concentration than the well-known BAF47/hSNF5/INI1 or BAF250/ARID1 subunits, for example. Importantly, the SS18-SSX fusion protein becomes incorporated into the BAF complex in place of SS18, and this in turn results in the eviction, and subsequently proteasomal degradation, of the BAF47 subunit (Figure 1, lower). BAF47 is already a well-established tumor suppressor. For example, loss of the BAF47 gene causes extremely aggressive malignant rhabdoid tumors (MRTs), and its re-expression in MRT cells stops their proliferation (Kia et al., 2008). It might therefore be expected that eviction of BAF47 also plays an important role in human synovial sarcoma tumorigenesis. In agreement with this idea, the altered BAF complex binds the Sox2 locus and reverses polycomb-mediated repression, resulting in activation of this pluripotency gene. Sox2 is uniformly expressed in human synovial sarcoma tumors and is
essential for their proliferation, so its anomalous activation may well be transformative. It is intriguing that eviction of BAF47, and thus transformation, depends on only two amino acids of the SSX protein, explaining why SSX1, SSX2, and SSX4, but not SSX3, are observed in synovial sarcoma fusion proteins: SSX3 has methionine-isoleucine in place of the evicting lysine-arginine amino acid pair found in the otherwise highly conserved SSX homologs. Altogether, this fascinating story of a unique oncogenic transformation mechanism underscores the frustratingly random nature of human cancer: if it invariably elicits efcient programs to drive cellular transformation, even an exceedingly rare and unlikely event like that in human synovial sarcoma may become a recurring human health issue. Encouragingly, the ndings of Kadoch and Crabtree indicate potential avenues of therapeutic intervention. As the authors point out, iffor examplea decoy molecule could be developed that causes the BAF47-evicting amino acids of the transformative SSX molecules to
resemble the corresponding surface of the benign SSX3 protein it would offer some hope for the development of a new treatment that builds on understanding the fusion proteins unusual mechanism of action.
REFERENCES Kadoch, C., and Crabtree, G.R. (2013). Cell 153, this issue, 7185. Kia, S.K., Gorski, M.M., Giannakopoulos, S., and Verrijzer, C.P. (2008). Mol. Cell. Biol. 28, 3457 3464. Nagai, M., Tanaka, S., Tsuda, M., Endo, S., Kato, H., Sonobe, H., Minami, A., Hiraga, H., Nishihara, H., Sawa, H., and Nagashima, K. (2001). Proc. Natl. Acad. Sci. USA 98, 38433848. Saha, S., and Jones, L.K. (2005). Fusion proteins and diseases (Chichester, UK: John Wiley & sons). Smith, E., Lin, C., and Shilatifard, A. (2011). Genes Dev. 25, 661672. Taki, T., and Taniwaki, M. (2006). Curr. Opin. Oncol. 18, 6268. Thaete, C., Brett, D., Monaghan, P., Whitehouse, S., Rennie, G., Rayner, E., Cooper, C.S., and Goodwin, G. (1999). Hum. Mol. Genet. 8, 585591.
Ring around the Ro-sie: RNA-Mediated Alterations of PNPase Activity

Brian J. Geiss1,* and Jeffrey Wilusz1,*
1Department of Microbiology, Immunology and Pathology, Colorado State University, Fort Collins, CO 80523, USA *Correspondence: brian.geiss@colostate.edu (B.J.G.), jeffrey.wilusz@colostate.edu (J.W.) http://dx.doi.org/10.1016/j.cell.2013.03.005
Chen et al. demonstrate a new way by which noncoding RNAs tailor the function of multicomponent complexes. They show that a noncoding RNA interacts with an exoribonuclease, altering its substrate specicity and enzymatic activity by serving as a ribonucleoprotein scaffold and, perhaps, a gate for entry of the RNA substrate.
Multiprotein complexes are the workhorses of the cell and provide critical functions that are necessary for cellular growth and viability by merging related activities into compact molecular machines. Protein-protein interactions are well known to be involved in allosteric regulation, altering substrate specicity and localization of enzymatic function to specic subcellular compartments. Several RNAs that serve as scaffolds for such molecular machines have been described, including yeast TLC1 RNA and telomerase (Lebo and Zappulla, 2012), pRNA and the 29 DNApackaging motor (Harjes et al., 2012), and IRES elements and translation factors. The ability of RNAs to scaffold molecular machines is also being investigated for synthetic biology applications (Delebecque et al., 2012). Given the
abundance of noncoding that serve as the assembly RNAs (ncRNAs) in cells, site for the PNPase, perhaps could they act as dynamic suggesting that additional scaffolds for ribonucleoproproteins may be regulated tein (RNP) complex forin a similar fashion. Moreover, mation, altering enzymatic there is no a priori reason why functions and regulating other ncRNAs could not funccellular processes? In this tion in a similar fashion in issue of Cell, Chen et al. nd other RNP machines. Thus, that this may be the case for ncRNAs could be used to the Ro protein-Y RNA comselect different protein pairplex (Chen et al., 2013). ings and to provide altered The authors demonstrate RNP functions. Indeed, direct Figure 1. Regulation of Protein Function by RNA Cofactors that, in the extremophile Deprotein-protein interactions Noncoding RNAs can change the perspective of protein enzymes through inococcus radiodurans, the may be only a small part the formation of ribonucleoprotein complexes. On the left, a PNPase (and perhaps other cellular ribonucleases) has little ability to discriminate between noncoding Y RNA acts as of the puzzle for how environdifferent types of RNA substrates. On the right, after formation of an RNP an adaptor between the mentally responsive macrostructure in which the noncoding Y RNA bridges an interaction between the Ro protein ortholog Rsr (an molecular machines are Ro protein (the lens of the monocle) and PNPase, the enzyme is able to distinguish structured RNAs more clearly and is much more selective RNA-binding protein) and formed and regulated. Y regarding the type of RNA substrate that it prefers. the polynucleotide phosRNAs, for example, are known phorylase (PNPase, a 30 -50 to interact with at least ve exoribonuclease) to produce an RNA conserved in Salmonella typhimurium, other proteins (RoBP1, hnRNP I, hnRNP degradation machine with altered sub- indicating that the components are K, nucleolin, and ZBP1), and it will be strate preference and enzymatic part of an evolutionarily conserved RNP interesting to see whether such RNA-profunction (Figure 1). Previous data from system. tein interactions also affect these cellular In addition to the PNPase regulation factors. the Wolin laboratory have shown that Rsr associates with Y RNA and that in- described above, ribonuclease activities Another intriguing hypothesis is that alteractions between Rsr and the PNPase in cells generally appear to be tightly terations in environmental conditions are important for RNA degradation regulated in macromolecular complexes. might change ncRNA expression, folding, (Wurtmann and Wolin, 2010). How this The dual endo/exoribonucleases J1 and or general availability and might drive the process actually works, however, has J2 in B. subtilis form a complex that formation of RNP complexes with been unclear. Based on biochemical an- regulates their enzymatic activity and sub- enhanced properties to help the cell adapt alyses, electron microscopic image strate specicity (Mathy et al., 2010). In eu- to its new environment. Along these lines, reconstruction, and modeling to a known karyotic cells, the poly(A)-specic exonu- we note that, whereas Rsr interacts with Ro protein-Y-RNA fragment, the authors cleases CCR4 and CAF1 localize RNase II and RNase PH during heat stress present a new model in which a dual together in a complex that is assembled to help mature rRNA (Chen et al., 2007), ring structure channels RNA substrates around the NOT1 scaffold (Petit et al., it interacts with the PNPase during the into the PNPase enzymatic cavity. In 2012). Rrp6 and isoforms of the Dis3 30 -50 stationary phase to degrade misfolded the EM reconstructions based on previ- exonucleases are sequestered and func- RNA (Wurtmann and Wolin, 2010). It will ous crystallographic structures, Y RNA tion as part of a large exosome complex be interesting to see whether changes ts within a narrow density between Rsr (Drazkowska et al., 2013). As these in scaffolding ncRNAs under these and the PNPase. Interestingly, no con- macromolecular complexes utilize pro- different conditions allow Rsr to form tacts between the proteins are observed. tein-protein interactions to regulate new RNP structures or alter the subThis, along with the analysis of the inter- ribonuclease function, it is particularly cellular localization of an RNP. actions of puried factors, suggests that interesting that, in this example, an RNA In closing, this study emphasizes the Y RNA is responsible for holding the that would ultimately be a substrate for potential for ncRNAs to adapt protein complex together. Y RNA not only serves the enzyme has been chosen by the modules to varied functions. The compoas the backbone of the complex, but cell to regulate the function of a powerful nents of RNPs, therefore, may be easily also blocks the KH/S1 domain of ribonuclease. changed by mixing and matching different A major implication of the Chen et al. parts, making them more like the beloved PNPase, reducing the enzymes ability to interact with single-stranded RNA study is that enzymatic function/protein classic Mr. Potato Head toy than previsubstrates. Thus, the Rsr-Y RNA- associations can be dynamically con- ously thought. PNPase machine is more active on trolled by the level and type of the ncRNA. structured RNA substrates and less Most organisms contain more than one ACKNOWLEDGMENTS active against single-stranded RNA sub- Y RNA species with a Ro protein-binding strates than the free PNPase enzyme. stem and signicant variations in their B.J.G. is supported by NIH grants R01 AI046435 Interestingly, this Y-RNA-assembled de- loop structures (Sim and Wolin, 2011). and U54 AI065357. J.W. is supported by NIH gradation machine appears to be Interestingly, it is the loop structures grants R01 GM072481 and U54 AI065357.
REFERENCES Chen, X., Wurtmann, E.J., Van Batavia, J., Zybailov, B., Washburn, M.P., and Wolin, S.L. (2007). Genes Dev. 21, 13281339. Chen, X., Taylor, D.W., Fowler, C.C., Galan, J.E., Wang, H.-W., and Wolin, S.L. (2013). Cell 153, this issue, 166177. Delebecque, C.J., Silver, P.A., and Lindner, A.B. (2012). Nat. Protoc. 7, 17971807.
Drazkowska, K., Tomecki, R., Stodus, K., Kowalska, K., Czarnocki-Cieciura, M., and Dziembowski, A. (2013). Nucleic Acids Res. Published online February 12, 2013. http://dx.doi.org/10.1093/nar/ gkt060. Harjes, E., Kitamura, A., Zhao, W., Morais, M.C., Jardine, P.J., Grimes, S., and Matsuo, H. (2012). Nucleic Acids Res. 40, 99539963. Lebo, K.J., and Zappulla, D.C. (2012). RNA 18, 16661678.
bert, A., Mervelet, P., Be nard, L., Mathy, N., He ans, A., Li de la Sierra-Gallay, I., Noirot, P., Dorle Putzer, H., and Condon, C. (2010). Mol. Microbiol. 75, 489498. Petit, A.-P., Wohlbold, L., Bawankar, P., Huntzinger, E., Schmidt, S., Izaurralde, E., and Weichenrieder, O. (2012). Nucleic Acids Res. 40, 1105811072. Sim, S., and Wolin, S.L. (2011). Wiley Interdiscip. Rev. RNA 2, 686699. Wurtmann, E.J., and Wolin, S.L. (2010). Proc. Natl. Acad. Sci. USA 107, 40224027.
Cullins Getting Undressed by the Protein Exchange Factor Cand1

Michael H. Olma1 and Ivan Dikic1,*
1Institute of Biochemistry 2 and Buchmann Institute for Molecular Life Sciences, Goethe University School of Medicine, Frankfurt 60590, Germany *Correspondence: ivan.dikic@biochem2.de http://dx.doi.org/10.1016/j.cell.2013.03.014
Cullin-RING ubiquitin ligase complexes (CRLs) rely on a vast array of adaptor proteins to recognize their substrates. Pierce et al. and related papers from Zemla et al. and Wu et al. in Nature Communications show that Cand1 promotes exchange of adaptor proteins to regulate the CRL repertoire.
At some point during the lifetime of a cell, virtually all of its proteins have had direct contact with the small protein ubiquitin (Grabbe et al., 2011). Some become temporarily modied as a signaling event. Most of them experience it during their last minutes as a destruction signal preceding their degradation. This vast array of targets requires a very versatile cellular system of ubiquitin ligases that attach ubiquitin specically to the intended proteins. The largest group of this enzyme class is the modular cullin-RING ubiquitin ligases (CRLs) with their hundreds of different substrate receptors (Petroski and Deshaies, 2005). However, it is largely unknown how cells adapt their huge repertoire of CRLs to immediate needs for a specic ligase. There have been reports that substrates can activate their specic ligase, but how the substrate adaptor is integrated into a ligase is not known. Now, in this issue of Cell, Pierce et al. (2013), together with works by Wu et al. (2013) and Zemla et al. (2013), shed light on how cells can make use of substrate receptors. All three studies focus on the best-understood CRL, the SCF (Skp1, cullin, and F box) complex, comprised of the scaffold Cul1 and one of a family of adaptors, the F box proteins (FBPs, Figure 1A). Like all CRLs, SCF complexes become activated with the attachment of Nedd8 in a process called neddylation (Duda et al., 2008). One other regulator crucial for SCF activity is Cand1. Genetic studies in various organisms have clearly shown that Cand1 functions as an activator of CRLs (e.g., Bosu et al., 2010). However, in vitro evidence pointed toward an inhibitory role of Cand1 on CRLs (Liu et al., 2002). The crystal structure of Cand1 bound to Cul1 especially emphasized its negative impact on CRL assembly whereby Cand1 blocks both the FBPbinding site and the neddylation site (Goldenberg et al., 2004). To better understand the relationship between Cand1, the FBPs, and Cul1, Raymond Deshaies and colleagues apply FRET-based real-time assays to measure the association and disassociation kinetics of Cand1 and FBPs with Cul1. They conrm previous reports that the FBP, in complex with the small adaptor Skp1, binds tightly to Cul1 and that the addition of Cand1 displaces the FBP/ Skp1-dimer (Figures 1B1D). By measuring the kinetics of this disassembly they show that Cand1 drastically accelerates the dissociation of FBPs from Cul1. Importantly, the effect of Cand1 is limited to the disassembly; the kinetics of the SCF complex assembly remained unchanged. Intriguingly, the same effect of FBP/Skp1 is detected for the Cand1Cul1 interaction: stably bound Cand1 is
REFERENCES Chen, X., Wurtmann, E.J., Van Batavia, J., Zybailov, B., Washburn, M.P., and Wolin, S.L. (2007). Genes Dev. 21, 13281339. Chen, X., Taylor, D.W., Fowler, C.C., Galan, J.E., Wang, H.-W., and Wolin, S.L. (2013). Cell 153, this issue, 166177. Delebecque, C.J., Silver, P.A., and Lindner, A.B. (2012). Nat. Protoc. 7, 17971807.
Drazkowska, K., Tomecki, R., Stodus, K., Kowalska, K., Czarnocki-Cieciura, M., and Dziembowski, A. (2013). Nucleic Acids Res. Published online February 12, 2013. http://dx.doi.org/10.1093/nar/ gkt060. Harjes, E., Kitamura, A., Zhao, W., Morais, M.C., Jardine, P.J., Grimes, S., and Matsuo, H. (2012). Nucleic Acids Res. 40, 99539963. Lebo, K.J., and Zappulla, D.C. (2012). RNA 18, 16661678.
bert, A., Mervelet, P., Be nard, L., Mathy, N., He ans, A., Li de la Sierra-Gallay, I., Noirot, P., Dorle Putzer, H., and Condon, C. (2010). Mol. Microbiol. 75, 489498. Petit, A.-P., Wohlbold, L., Bawankar, P., Huntzinger, E., Schmidt, S., Izaurralde, E., and Weichenrieder, O. (2012). Nucleic Acids Res. 40, 1105811072. Sim, S., and Wolin, S.L. (2011). Wiley Interdiscip. Rev. RNA 2, 686699. Wurtmann, E.J., and Wolin, S.L. (2010). Proc. Natl. Acad. Sci. USA 107, 40224027.
Cullins Getting Undressed by the Protein Exchange Factor Cand1

Michael H. Olma1 and Ivan Dikic1,*
1Institute of Biochemistry 2 and Buchmann Institute for Molecular Life Sciences, Goethe University School of Medicine, Frankfurt 60590, Germany *Correspondence: ivan.dikic@biochem2.de http://dx.doi.org/10.1016/j.cell.2013.03.014
Cullin-RING ubiquitin ligase complexes (CRLs) rely on a vast array of adaptor proteins to recognize their substrates. Pierce et al. and related papers from Zemla et al. and Wu et al. in Nature Communications show that Cand1 promotes exchange of adaptor proteins to regulate the CRL repertoire.
At some point during the lifetime of a cell, virtually all of its proteins have had direct contact with the small protein ubiquitin (Grabbe et al., 2011). Some become temporarily modied as a signaling event. Most of them experience it during their last minutes as a destruction signal preceding their degradation. This vast array of targets requires a very versatile cellular system of ubiquitin ligases that attach ubiquitin specically to the intended proteins. The largest group of this enzyme class is the modular cullin-RING ubiquitin ligases (CRLs) with their hundreds of different substrate receptors (Petroski and Deshaies, 2005). However, it is largely unknown how cells adapt their huge repertoire of CRLs to immediate needs for a specic ligase. There have been reports that substrates can activate their specic ligase, but how the substrate adaptor is integrated into a ligase is not known. Now, in this issue of Cell, Pierce et al. (2013), together with works by Wu et al. (2013) and Zemla et al. (2013), shed light on how cells can make use of substrate receptors. All three studies focus on the best-understood CRL, the SCF (Skp1, cullin, and F box) complex, comprised of the scaffold Cul1 and one of a family of adaptors, the F box proteins (FBPs, Figure 1A). Like all CRLs, SCF complexes become activated with the attachment of Nedd8 in a process called neddylation (Duda et al., 2008). One other regulator crucial for SCF activity is Cand1. Genetic studies in various organisms have clearly shown that Cand1 functions as an activator of CRLs (e.g., Bosu et al., 2010). However, in vitro evidence pointed toward an inhibitory role of Cand1 on CRLs (Liu et al., 2002). The crystal structure of Cand1 bound to Cul1 especially emphasized its negative impact on CRL assembly whereby Cand1 blocks both the FBPbinding site and the neddylation site (Goldenberg et al., 2004). To better understand the relationship between Cand1, the FBPs, and Cul1, Raymond Deshaies and colleagues apply FRET-based real-time assays to measure the association and disassociation kinetics of Cand1 and FBPs with Cul1. They conrm previous reports that the FBP, in complex with the small adaptor Skp1, binds tightly to Cul1 and that the addition of Cand1 displaces the FBP/ Skp1-dimer (Figures 1B1D). By measuring the kinetics of this disassembly they show that Cand1 drastically accelerates the dissociation of FBPs from Cul1. Importantly, the effect of Cand1 is limited to the disassembly; the kinetics of the SCF complex assembly remained unchanged. Intriguingly, the same effect of FBP/Skp1 is detected for the Cand1Cul1 interaction: stably bound Cand1 is
Figure 1. Substrate Receptor Exchange Mediated by Cand1

(AF) Cand1s main function is to catalyze the exchange of substrate receptors of CRLs. For Cand1 to work efciently, Nedd8 (N8) has to be removed from the cullin (A and B). Cand1 then binds, generating a transient complex with both Cand1 and Skp1 (C). The competition for partially shared binding sites on Cul1 releases the F-Box-protein (FBP)/Skp1-dimer and allows a new substrate receptor to bind (D and E). This new CRL can then be activated again by Neddylation (F). Substrate receptor features (such as expression level) inuence which receptor will get integrated into a ligase, in contrast to the substrate availability regulating the activation status of the CRL.
easily removed from Cul1 by FBP/Skp1 (Figures 1D1E). These opposing ndings lead the authors to the hypothesis that Cand1 is neither a competitive nor an allosteric inhibitor but that it acts as a novel type of exchange factor. Such functionality is known for guanine nucleotide exchange factors (GEFs), which reactivate GTPases by promoting the exchange of GDP for GTP (Bos et al., 2007). Pierce et al. (2013) propose that Cand1 does something similar, only for proteins instead of nucleotides. As such, the main function of Cand1 as a substrate receptor exchange factor (SREF) should be that it facilitates the exchange of FBPs. The authors verify this hypothesis in vitro. In an elegantly designed experiment, Cul1, preloaded with an unrelated FBP, cannot use another FBP for the ubiquitination of its corresponding substrate. After addition of Cand1, an FBP switch can take place, and the substrate is ubiquitinated. Besides addressing the molecular function of Cand1, the authors also analyze the in vivo consequences of Cand1 on the large variety of SCF complexes. Utilizing quantitative mass spectrometry experiments, the authors simultaneously detect the abundance of up to 34 different, immunopuried SCF complexes. Upon depletion of Cand1, the abundances of specic SCF complexes are changed; some are up- and some downregulated in comparison to controls. This suggests that Cand1 generates an
equilibrium of different SCF complexes and without this Cand1 function newly synthesized FBPs cannot engage the SCF complexes efciently. Two other recent manuscripts study the functionality of the yeast counterpart of Cand1. One study directly analyzes the FBP incorporation into the SCF complex (Wu et al., 2013), whereas the other analyzes the adaptation of the SCF complex to carbon source switching (Zemla et al., 2013). Both conclude that yeast Cand1 also has the same exchange functionality as in mammalian cells. Recent reports found that substrate availability induces the activation of CRLs. Based on this, Pierce et al. propose a model whereby two factors dene the CRL-adaptor repertoire (Figure 1). If a CRL encounters a substrate, neddylation is induced and the substrate adaptor is xed in place since Cand1 can no longer properly bind. The CRL becomes deneddylated if all substrate is processed. If Cand1 encounters this CRL, a transient complex forms, with Cand1 and Skp1 competing for an overlapping binding site. This results in the release of the FBP and consequently the exchange of the adaptor. Such a model suggests that both the substrate availability and the adaptor itself, with its expression level, its turnover rate, and its modication state and localization, dene the CRL landscape. As Cand1 has been reported to bind to other cullins besides Cul1, Cand1 might
act as a general SREF for all CRLs. Especially for cullin 4-based ligases (CRL4s) it will be interesting to analyze the impact of Cand1. CRL4s bind the substrate adaptors only indirectly via the large linker protein Ddb1, leaving the question: Is there also an exchange factor for adaptors on Ddb1? For the Cul1-FBP interaction there might even be an additional SREF, as one FBP, Fbxw2, is largely insensitive to Cand1 function (Pierce et al., 2013). But also the direct regulation of Cand1 function at the CRL by posttranslational modications seems possible. Perhaps there are also other processes which require an SREF function to generate a balance between similar binding partners on a scaffold. A detailed analysis of the binding kinetics of factors with overlapping binding sites could expand this new group of protein exchange factors to a wide variety of processes, from DNA replication to transcription to vesicle trafcking. A lot to nd out..
REFERENCES Bos, J.L., Rehmann, H., and Wittinghofer, A. (2007). Cell 129, 865877. Bosu, D.R., Feng, H., Min, K., Kim, Y., Wallenfang, M.R., and Kipreos, E.T. (2010). Dev. Biol. 346, 113126. Duda, D.M., Borg, L.A., Scott, D.C., Hunt, H.W., Hammel, M., and Schulman, B.A. (2008). Cell 134, 9951006. Goldenberg, S.J., Cascio, T.C., Shumway, S.D., Garbutt, K.C., Liu, J., Xiong, Y., and Zheng, N. (2004). Cell 119, 517528.
Grabbe, C., Husnjak, K., and Dikic, I. (2011). Nat. Rev. Mol. Cell Biol. 12, 295307. Liu, J., Furukawa, M., Matsumoto, T., and Xiong, Y. (2002). Mol. Cell 10, 15111518. Petroski, M.D., and Deshaies, R.J. (2005). Nat. Rev. Mol. Cell Biol. 6, 920.
Pierce, N., Lee, E., Liu, X., Sweredoski, M.J., Graham, R., Larimore, E.A., Rome, M., Zheng, N., Clurman, B., Hess, S., et al. (2013). Cell 153, this issue, 206215. Wu, S., Zhu, W., Nhan, T., Toth, J.I., Petroski, M.D., and Wolf, D.A. (2013). Nat. Commun. Published
online March 27, 2013. http://dx.doi.org/10.1038/ ncomms2636. Zemla, A., Thomas, Y., Kedziora, S., Knebel, A., Wood, N.T., Rabut, G., and Kurz, T. (2013). Nat. Commun. Published online March 27, 2013. http://dx.doi.org/10.1038/ncomms2628.
Leading Edge
Review
Lessons from the Cancer Genome
Levi A. Garraway1,2,4 and Eric S. Lander3,4,5,*
of Medical Oncology and Center for Cancer Genome Discovery, Dana-Farber Cancer Institute, Boston, MA 02215, USA of Medicine, Brigham and Womens Hospital 3Department of Systems Biology Harvard Medical School, Boston, MA 02115, USA 4The Broad Institute of Harvard and MIT, Cambridge, MA 02142, USA 5Department of Biology, Massachusetts Institute of Technology, Cambridge, MA 02139, USA *Correspondence: lander@broadinstitute.org http://dx.doi.org/10.1016/j.cell.2013.03.002
2Department 1Department
Systematic studies of the cancer genome have exploded in recent years. These studies have revealed scores of new cancer genes, including many in processes not previously known to be causal targets in cancer. The genes affect cell signaling, chromatin, and epigenomic regulation; RNA splicing; protein homeostasis; metabolism; and lineage maturation. Still, cancer genomics is in its infancy. Much work remains to complete the mutational catalog in primary tumors and across the natural history of cancer, to connect recurrent genomic alterations to altered pathways and acquired cellular vulnerabilities, and to use this information to guide the development and application of therapies.
Introduction More than a century ago, Theodor Boveri proposed that cancer is caused by chromosomal derangements that cause cells to divide uncontrollably (Boveri, 2008)that, in modern terms, cancer is a disease of the genome. It took 70 years for molecular biologists to prove this concept by showing the existence of mutated cancer-causing genes (Stehelin et al., 1976; Tabin et al., 1982). By the mid-1980s, researchers had established two main types of cancer-causing genes (oncogenes and tumor suppressor genes) and had dened the genomic alterations that give rise to them (e.g., nucleotide substitutions, chromosomal copy number alterations, and DNA rearrangements; reviewed in Macconaill and Garraway [2010]). These studies also began to suggest considerable complexity in the mutational origins of cancer, with cancer-causing genes varying across and within tumor types and with multiple genes contributing to tumorigenesis. In an inuential commentary in 1986, Renato Dulbecco argued that the complete sequence of the human genome would be an essential tool for systematically discovering the genes that drive cancer (Dulbecco, 1986). If we wish to learn more about cancer, we must now concentrate on the cellular genome, he wrote. We have two options: either to try to discover the genes important in malignancy by a piecemeal approach, or to sequence the whole genome.it will be far more useful to begin by sequencing the cellular genome. Responding to this and other calls, the Human Genome Project (HGP) was launched in 1990. A draft sequence was completed by 2000 (Lander et al., 2001; Venter et al., 2001) and a near-complete sequence by 2003 (IHGSC, 2004). With the availability of the genome sequence, cancer researchers rapidly began to develop a new eld of cancer genomics. Cancer genomics involves systematic studies of (some or all of) the genome to nd sites of recurrent derangement in specic cancer types. Pioneering genomic studies at the Sanger Institute and Johns Hopkins uncovered genes mutated frequently in melanoma and colon cancer, respectively (Davies et al., 2002; Samuels et al., 2004). Studies by several groups in Boston and New York then discovered frequent activating mutations in lung cancer, which largely explained patient response to a drug (Lynch et al., 2004; Paez et al., 2004; Pao et al., 2004). Soon thereafter, a working group of the U.S. National Cancer Institute proposed a Human Cancer Genome Project (see http://www.genome.gov/Pages/About/ NACHGR/May2005NACHGRAgenda/ReportoftheWorkingGroup onBiomedicalTechnology.pdf), which came to be called The Cancer Genome Atlas (TCGA). The NCI launched a pilot project for TCGA in 2006 and a full project in 2009. In parallel, an International Cancer Genome Consortium was launched and has grown to involve researchers in more than 15 countries (Hudson et al., 2010). The notion of taking a genomic approach to characterizing cancer was not universally endorsed, as reected in the title of one commentary: Human Cancer Genome Project: Another Misstep in the War on Cancer (Gabor Miklos, 2005). Some thoughtful critics felt that hypothesis-driven research was the best way to study cancer and worried that systematic studies were so expensive that they would drive out focused investigation (Weinberg, 2010). Proponents argued that science requires investment in both hypothesis generation and hypothesis testing and that unbiased genomic studies were an excellent way to nd surprises. They also expected the cost of genomic studies to plummet, with new technologies just over the horizon. Some scientists were skeptical because they believed that there were few cancer-related genes left to discover, whereas others thought that cancers were too hopelessly complicated to yield
to systematic analysis. This open debate helped to shape the design of cancer genome projects. In the end, however, the questions could only be answered with data. With cancer genome projects now underway for several years (Table 1), the time is right to assess the early returns and to consider next steps for the eld. (As a complement to this Review, we recommend an earlier review by Stratton et al. [2009], which describes many foundational aspects of cancer genomics.) Here, we describe the remarkable tapestry of biological, evolutionary, and therapeutic insights that have emerged from systematic cancer genome characterization. At the end, we suggest the next steps for cancer genomics. Technology Revolution Initial cancer genome projects had to be carried out with what today seem like primitive technologies. Mutations were identied by traditional capillary-based sequencing in which each exon to be studied was amplied and sequenced individually, and chromosome copy number alterations were surveyed with DNA microarrays. DNA rearrangements could hardly be cataloged at all. The high cost and extensive infrastructure needed for large-scale DNA sequencing placed tight constraints on the amount of data that could be collected. Exome-scale projects blom could only be carried out on small numbers of samples (Sjo et al., 2006; Wood et al., 2007); thus, much effort was spent developing lists of candidate gene for sequencing based on a priori notions of cancer mechanisms or therapeutic targets (Ding et al., 2008; Greenman et al., 2007; CGARN, 2008). The emergence of massively parallel sequencing (MPS) revolutionized the entire enterprise (Bentley et al., 2008; Margulies et al., 2005). Initially, MPS made it possible to sequence nearly 1 billion bases (1 gigabase [Gb]) in a single run; this number grew to >600 Gb/run by 2012. In parallel, methods were developed that employ hybridization to oligonucleotide baits in aqueous solution to capture specic portions of the genome most importantly, the 2% of genomic DNA that contains known exons (the exome) (Gnirke et al., 2009; Hodges et al., 2007). MPS also made it possible to use a single technology platform for all categories of genome analysis (discovering point mutations, assessing copy number alterations and translocations, measuring transcript levels, identifying alternative splicing, detecting DNA methylation, and mapping chromatin structure). The rst whole cancer genome sequenced by MPS was reported in 2008 (Ley et al., 2008). Whereas initial studies were conned to single samples (Pleasance et al., 2010a, 2010b), studies of hundreds of samples have quickly become the norm. Plummeting costs have propelled an unprecedented explosion of sequence data. For example, >16,000 cancer samples had been subjected to genome or exome sequencing by late 2012 just at our institution alone (Broad Institute). With the MPS data came a need for completely new analytic tools. The rst challenge was to accurately determine the sequence in individual tumor and normal samples from the raw sequence data. Each type of alteration in the DNA and RNA required a specialized detection method, including for single nucleotide variants, small insertions/deletions, chromosomal rearrangements, gene fusions, alternatively spliced transcripts, chromosomal copy number alterations, and detection
Table 1. Current Large-Scale Cancer Genome Projectsa Anatomic Site Brain/Central nervous system Tumor Type glioblastoma multiforme low-grade glioma pediatric: medulloblastoma pediatric: pilocytic astrocytoma Head and neck Thoracic Breast head/neck squamous cell cancer thyroid carcinoma lung adenocarcinoma lung squamous cell carcinoma breast lobular carcinoma breast ductal carcinoma breast triple-negative breast HER-2 positive breast ER positive vs. negative Gastrointestinal esophageal adenocarcinoma esophageal squamous carcinoma gastric adenocarcinoma gastric (intestinal/diffuse) hepatocellular (alcohol/adiposity) hepatocellular (virus) hepatocellular (general) pancreatic adenocarcinoma colorectal adenocarcinoma colon cancer (non-Western) Gynecologic ovarian serous cystadenocarcinoma endometrial carcinoma cervical cancer (squamous + adeno) Urologic renal: clear cell carcinoma renal: papillary carcinoma renal: chromophobe carcinoma bladder cancer prostate adenocarcinoma prostate adenocarcinoma, early onset Skin Soft tissue (Sarcoma) melanoma, cutaneous solitary brous tumors desmoid tumors angiorsarcomas leiomyosarcomas extraskeletal myxoid chondrosarcomas Hematologic acute myeloid leukemia lymphoma: chronic lymphocytic leuk. lymphoma: germinal B cell lymphoma: diffuse large B cell chronic myeloid disorders
a
In conjunction with The Cancer Genome Atlas, International Cancer Genome Consortium, and Slim Initiative for Genomic Medicine.
of foreign DNA (such as from viruses) (Beroukhim et al., 2007; Chen et al., 2009; Cibulskis et al., 2013; Dees et al., 2012; Kim and Salzberg, 2011; Kostic et al., 2011; Trapnell et al., 2009). Algorithms employ probabilistic methods to identify mutations
or rearrangements based on their presence in multiple tumor sequence reads and absence in the paired normal DNA sequence (Meyerson et al., 2010). Detecting mutations with high accuracy turned out to be surprisingly tricky. Because somatic point mutations in cancer are so infrequent (1/Mb), the background error rate must be an order of magnitude lower to ensure that most apparent events are true positives. Many false positives initially arose from sequencing errors and inaccurate alignment of reads to the human genome. In addition, false negatives may arise from admixture of noncancer cells (tumor purity), copy number variations inherent in cancer genomes (ploidy), and the presence of variant subclones within the cancer cell population (heterogeneity) (Carter et al., 2012). Increasing the sequencing depth (average number of reads per base) was found to improve both the specicity and sensitivity of mutation calling. Currently, tumor sequencing is performed with 100- to 150-fold coverage for whole-exome analysis and 30- to 60-fold coverage for whole-genome analysis. (The whole-genome sequence should ideally be deeper but is currently limited by cost.) Obtaining accurate mutation calls for a collection of individual samples is only the rst step. The harder challenge is to distinguish between driver events that are causally related to the development of cancer and random passenger events that have simply accumulated over the course of development and cell growth. This requires determining which genes show significantly more mutations than random expectation. Sophisticated mathematical methods are needed to ensure that the random expectation properly accounts for (1) variation in background mutation rates across the genome, (2) variation across tumors, and (3) variation in purity and heterogeneity (Chapman et al., 2011a; Dees et al., 2012; Hodis et al., 2012; M.S. Lawrence, personal communication). Without such corrections, genes may be spuriously declared to be drivers, with the problem growing worse as sample size grows (because even modest deviations from expectation will appear to be signicant). Recent studies have highlighted some likely spurious results and have developed solutions to eliminate them (G. Getz, personal communication). Perfecting these algorithms remains an area of active research. Other algorithms have been developed to study the structure of amplications and deletions to detect the possibility of multiple target genes within a given locus (Beroukhim et al., 2007; Beroukhim et al., 2010). Mutational Mechanisms The explosion of genomic data quickly shed light on the mutational processes of cancer, revealing an unexpected richness of mechanisms. These insights may propel a deeper understanding of factors governing genome integrity and tumor evolution. Mutation Rates Initial plans for cancer genome analysis assumed a single uniform background mutation rate (1/Mb). In fact, cancer mutation rates turned out to be much more variable, ranging from as low as one base substitution per exome (<0.1/Mb) in some pediatric cancers to thousands of mutations per exome (100/Mb) in certain mutagen-induced malignancies (such as lung cancer and melanoma). Moreover, mutation rates were
found to vary substantially across the genome, governed by processes such as transcription-coupled repair and replication timing. Mutational Spectra Cancer genome sequencing has also revealed a wide array of mutational patterns both across and within individual tumor types. Their distinctive characteristics may reect extrinsic factors (e.g., UV light or tobacco smoke) or intrinsic patterns such as DNA repair deciencies. For example, a recent study (Nik-Zainal et al., 2012a) pointed to at least ve distinct nucleotide substitution patterns, most of which occur by as-yet unknown mechanisms. One such process, which produces C > A, C > G, or C > T substitutions at TpCpX trinucleotides, appeared to underpin most nucleotide substitutions in 10% of ER-positive breast tumors. These studies also discovered a new regional hypermutation mechanism characterized by multiple base mutations that occur in cis near rearrangement breakpoints. Termed kataegis (Greek, kataegis = shower or thunderstom), this process likely involves the activationinduced deaminase (AID) and apolipoprotein B mRNA-editing enzyme catalytic polypeptide-like (APOBEC) protein families. New mutation patterns (in this case, A > C transversions at AA dinucleotides) have also been discovered in esophageal cancers by large-scale sequencing (A.M. Dulak, personal communication). Chromosomal Gains and Losses Although tumor cell aneuploidy has long been recognized, global cancer genome studies have yielded a systematic assessment of large-scale (whole chromosome or chromosome arm) and focal copy number aberrations. The typical cancer cell exhibits large-scale gains or losses involving a quarter of its genome and carries focal events affecting 10% (Beroukhim et al., 2010). Based on current sample collections, many focal amplications and deletions have been localized to peak regions containing a median of 67 genes (although the number is 150200 in some cases). For the majority of focal events, the driver gene(s) still cannot be assigned denitively. Chromosomal Shattering One of the most striking mutational patterns unveiled by whole-genome sequencing studies consists of a catastrophic phenomenon that produces dozens or even hundreds of rearrangements. The resulting disarray is distinctive for two reasons: it is typically localized within one or a few chromosomes, and it usually involves only two distinct copy number states (Stephens et al., 2011). Termed chromothripsis (Greek, thripsis = shattering), this process occurs in 2%3% of human cancers, with an elevated prevalence in bone cancers, pediatric medulloblastoma, and neuroblastoma (Molenaar et al., 2012; Rausch et al., 2012). Genomic shattering appears to develop as a result of erroneous chromosome segregation during mitosis and the subsequent entrapment of individual chromosomes within micronuclei (Crasta et al., 2012). Micronuclei have a tendency toward premature chromosome condensation, which may result in pulverization of chromosomal segments. Chromosomes that survive this process, having undergone aberrant reassembly through nonhomologous end-joining, emerge with dense rearrangements that may sometimes dysregulate cancer genes (Stephens et al., 2011).
Chromosomal Chains A whole-genome sequencing study of primary human prostate cancer (Berger et al., 2011) uncovered a distinct category of complex chromosomal rearrangements. Prostate cancer genomes often exhibit chains of copy-neutral rearrangements that consist of 412 distinct breakpoint junctions distributed across multiple chromosomes, with the breakpoints forming a closed chain (A to B, B to C, C to D, and nally back to A) that distinguishes the process from chromothripsis or other complex rearrangements. Closed chain rearrangement breakpoints tend to occur near open chromatin (that is, transcriptionally active chromatin) in prostate cancer genomes harboring ETS transcription factor rearrangements but near closed chromatin in certain ETS-negative prostate cancers. These chains have recently been termed chromoplexy (Greek, plexy = weave or braid) (Baca et al., 2013). Additional Processes Other complex DNA rearrangements seem to arise through errors in DNA replication (Liu et al., 2011). These may include fork-stalling and template-switching events that trigger microhomology-dependent DNA priming, duplications, and DNA template insertions (for a recent review, see Holland and Cleveland [2012]). Interestingly, these replication-dependent rearrangements show a strong correlation with TP53 mutations in subtypes of medulloblastoma (Rausch et al., 2012). Thus, somatic alterations in DNA-damage-sensing pathways may render tumor progenitor cells vulnerable to ensuing catastrophic genomic events. Insights into mutational patterns may bring a deeper understanding of tumor evolution. In contrast to a simple gradualist notion in which somatic mutations accumulate steadily, tumor evolution can be punctuated by various types of catastrophic events (Baca et al., 2013). A fuller knowledge of mutational processesparticularly those that preferentially enact cancer genesmay help to identify driver mechanisms in tumors. New Cancer Genes A key question is whether cancer genomics has led to the discovery of new genes and, ideally, to new classes of genes not previously known to play a causal role in cancer. Although much work still lies ahead, the answer is clear. The trickle of biological discoveries from early studies has become a wave, implicating a wide range of cellular processes in cancer. Whereas some of the new cancer genes encode classical signaling proteins, most populate new and sometimes surprising categories, such as metabolism, epigenetics, chromatin biology, splicing, protein homeostasis, and cell differentiation (Table 2). The insights from these studies are already guiding hypothesis-driven cancer research ranging from basic cell and molecular biology to novel therapeutics. Signal Transduction Pathways Studies from the 1980s and 1990s revealed that signaling pathways linked to proliferation and survival played a crucial role in many cancers. Mutations were discovered in key genes that encode members (or regulators) of receptor tyrosine kinase (RTK)-signaling pathways (HER-2, c-KIT, ABL, RAS, NF1, NF2, MET, PTEN), the WNT/b-catenin pathway (APC), and the TGF b pathway (SMAD2 and SMAD4), among others. Moreover, the
pharmaceutical industry showed that drugs could be developed to inhibit protein kinases. The poster child was imatinib (Gleevec) against the ABL and KIT kinases, which proved remarkably effective in treating malignancies driven by activating mutations in these oncoproteins (chronic myelogenous leukemia [CML] and gastrointestinal stromal tumors [GIST]) (Demetri et al., 2002; Druker et al., 2001). Recognition of the importance and druggability of RTKs motivated the rst unbiased sequencing surveys in the early 2000s, which employed Sanger sequencing to examine dozens of genes in dozens of patients. The studies quickly hit pay dirt, with the nding of mutations in BRAF in 50% of melanomas, PIK3CA in 25%30% of breast and colorectal cancers, EFGR in 10%15% of non-small cell lung cancers, FGFR2 in 15%20% of endometrial cancers, and JAK2 in myeloproliferative diseases (Davies et al., 2002; Dutt et al., 2008; Kralovics et al., 2005; Levine et al., 2005; Lynch et al., 2004; Paez et al., 2004; Pao et al., 2004; Pollock et al., 2007; Samuels et al., 2004). (Some of the ndings came to have a major impact on drug development and clinical treatment, including the development of selective RAF and MEK inhibitors that have produced dramatic remissions in melanoma and the ability to target the use of EGFR inhibitors to the subset of lung cancer patients who derive benet.) In turn, these successes led researchers to scale up sequencing surveys to discover additional candidate genes in signaling pathways and eventually to all genes. Recurrent mutations were found in genes involved in severalsometimes surprisingpathways not previously suspected to drive cancer. These included the MAP3K1 and MAP2K4 genes in breast cancer (encoding serine/threonine kinases involved in the P38JNK signaling pathway) (Banerji et al., 2012; Ellis et al., 2012; Stephens et al., 2012; CGAN, 2012), RAC1 in melanoma (a GTPase involved in the RAC/PAK-signaling module involved in focal adhesion) (Hodis et al., 2012; Krauthammer et al., 2012), ELMO1 and DOCK2 in esophageal cancer (two genes that activate RAC/PAK signaling) (A.M. Dulak, personal communication), MYD88 in diffuse large B cell lymphoma (activates NFkB signaling) (Ngo et al., 2011), and PREX2 in melanoma (a guanine nucleotide exchange factor that controls RAC/PAK and PI3K signaling) (Berger et al., 2012). Remarkably, a pathway involved in axon guidance in neurons (the ROBO/SLIT pathway) turned out to be a target of mutations in 20% of pancreatic adenocarcinomas (Biankin et al., 2012). And, a pathway that governs the oxidative stress response in all cells (the KEAP1/ NRF2-signaling pathway) is activated by mutation in >30% of squamous lung cancers (Hammerman et al., 2012; Shibata et al., 2008). Many of these results would have eluded hypothesis-based investigation. In addition, genome-wide studies of copy number alterations based on DNA microarrays revealed recurrently amplied genes in signaling and cell survival pathways. MCL1 and BCL2L1, which encode anti-apoptotic proteins that are critical regulators of tumor cell survival, were found to be amplied in a wide range of cancers, including breast, lung, colorectal, melanoma, and glioblastoma (Beroukhim et al., 2010). FGFR1 was found to be amplied in >20% of lung squamous cancer (Weiss et al., 2010) and in 10% of breast cancers (Chin et al., 2006).
Table 2. Discoveries from Cancer Genome Characterization Cellular Process Altered by Genomic Alterations RTK signaling MAPK signaling (oncogenes) MAPK signaling (TSG) PI3K signaling (oncogenes) PI3K signaling (TSG) Notch signaling (oncogene or TSG) TOR signaling (TSG) Wnt/b-catenin signaling (TSG) TGF-b signaling (TSG) NF-kB signaling (oncogene) Other signaling Epigenetics DNA methylation Epigenetics DNA hydroxymethylation Chromatin histone methyltransferases Chromatin histone demethylases Chromatin histone acetyltransferases Chromatin SWI/SNF complex Chromatin other Transcription factor lineage dependency or oncogene Transcription factor other Splicing RNA abundance Translation/protein homeostasis/ubiquitination Metabolism Genome integrity Telomere stability Cell cycle (oncogene) Cell cycle (TSG) Apoptosis regulation
a b
Examples of Cancer Genes Discovered (or Extended to New Cancers*) by Genomics EGFR,a ERBB2,*,a MET,*,a ALK,*,a JAK2,a RET,*,a ROS,*,a FGFR1,*,a FGFR2,a PDGFRA,*,a and CRKLa KRAS,*,a NRAS,*,a BRAF, a and MAP2K1a NF1*,b PIK3CA,a AKT1,a and AKT3a PTEN*,b and PIK3R1b NOTCH1,c NOTCH2,c and NOTCH3b STK11,*,b TSC1,*,b and TSC2*,b APC*,b and CTNNB1*,a SMAD2,*,b SMAD4,*,b and TGFBR2b MYD88a RAC1,a RAC2,a CDC42,a KEAP1,b MAP3K1,b MAP2K4,b ROBO1,b ROBO2,b SLIT2,b SEMA3A,b SEMA3E,b ELMO1,d and DOCK2d DNMT3Ab TET2b MLL,*,b MLL2,b MLL3,b EZH2,c NSD1,b and NSD3b JARID1A,b UTX,b KDM5A,b and KDM5Cb CREBP b and EP300b SMARCA1,*,b SMARCA4,b ARID1A,b ARID2,b ARID1B,b and PBRM1b CHD1,b CHD2,b and CHD4b MITF,a NKX2-1,a SOX-2,a ERG,a ETV1,a and CDX2a MYC,*,a RUNX1,b GATA3,b FOXA1,b NKX3.1,b SOX9,a NFE2L2,a and MED12d SF3B1,d U2AF1,d SFRS1,d SFRS7,d SF3A1,d ZRSR2,b SRSF2,d U2AF2,d and PRPF40Bd DIS3d SPOP,d FBXW7,*,b WWP1,*,b FAM46C,d and XBP1d IDH1a and IDH2 a TP53,*,b MDM2,a MSH,*,b MLH,*,b and ATM*,b TERT promoter mutationsa CCND1*,a and CCNE1*,a CDKN2A,*,b CDKN2B,*,b and CDKN1Bb MCL1,a BCL2A1,a and BCL2L1a
Activating mutation or amplication. Inactivating mutation or deletion. c Both activating and inactivating genomic events observed. d Effect of mutations on protein function unknown.
CRKL, which encodes a signaling adaptor protein, was found amplied in a subset of lung cancers (Kim et al., 2010). Despite their successes, these studies were sobering in revealing that the early promise of using a single kinase inhibitor (imatinib) to treat prevalent oncoprotein mutations (as in CML and GIST) was not going to be widely generalizable: most cancers lacked a highly recurrent mutation in genes encoding kinases (or other readily druggable targets) (Greenman et al., 2007). This underscored the importance of more deeply probing the cancer genome. Metabolism If there were any doubt that genomic approaches would reveal surprises, they should have been put to rest by a pioneering study in 2008. In this paper (which predated the maturation of
MPS technology), Vogelstein and colleagues employed an impressive brute force approach to PCR amplify and sequence 175,471 exons from 20,661 genes (Parsons et al., 2008). They were rewarded with the discovery of highly recurrent mutations in the IDH1 gene, which encodes the cytoplasmic metabolic enzyme isocitrate dehydrogenase, a seemingly unlikely candidate for a cancer gene (Figure 1); the mutations affected a single amino acid in the active site. Subsequent studies found that specic mutations in IDH1 and IDH2 (IDH1s mitochondrial homolog) occurred in >70% of secondary glioblastomas, oligodendrogliomas, and high-grade astrocytomas (Parsons et al., 2008; Yan et al., 2009) and in 15%30% of acute myelogenous leukemias (AML) (Mardis et al., 2009). Because isocitrate dehydrogenases convert isocitrate to
Figure 1. Somatic IDH1/2 Mutations Produce the Oncometabolite 2HG

Oncogenic effects of 2HG include generation of a CIMP-like phenotype and inhibition of a-ketoglutarate-dependent enzymes such as histone methyltransferases (KMT), histone demethylases (KDM), and prolyl hydroxylases (EGLN). TET2 mutations are mutually exclusive with IDH1/2 mutations in leukemias and may exert common downstream effects on DNA methylation. Mutant IDH1/2 proteins are the targets of emerging drug discovery efforts (boxed).
a-ketoglutarate (a-KG) in the tricarboxylic acid (TCA) cycle, the observation suggested a previously unrecognized link between cell metabolism and cancer. It soon became clear that the mutations caused a gain-of-function (or neomorphic) activity, whereby isocitrate was converted to a distinct metabolite: the R-enantiomer of 2-hydroxyglutarate (2HG; Figure 1) (Dang et al., 2009; Ward et al., 2010). How this oncometabolite might drive cancer, however, remained a mystery. The answer emerged from a different type of genomic analysis: genome-wide surveys of DNA methylation. The methylation studies revealed that a subset of glioblastomas (the proneural subtype) showed a DNA methylation pattern that strongly resembled the CpG island methylator phenotype (CIMP) originally described in colorectal cancer (Noushmehr et al., 2010). Remarkably, the CIMP-like phenotype was tightly correlated with the presence of IDH1 mutations (Figure 1). A follow-up study conrmed that introduction of mutant IDH actually caused the CIMP-like phenotype (Turcan et al., 2012). Unexpectedly, the mechanism was claried by yet another genomic survey, this time involving acute myelogenous leukemia (AML). This large-scale study showed that IDH1/IDH2 mutations were mutually exclusive with inactivating TET2 mutations (Figueroa et al., 2010), suggesting that the two types of mutations had similar effects and were thus functionally redundant. The TET2 protein catalyzes 5-methylcytosine hydroxylation in
a a-KG-dependent manner, and loss of TET2 produces a CIMP-like phenotype. Studies then showed that 2HG appears to inhibit several a-KG-dependent enzymes (Xu et al., 2011), including Jumonji-C domain histone demethylases that affect gene expression (Lu et al., 2012) and prolyl-4-hydroxylases (EGLN1/2/3) that regulate hypoxia inducible factor (HIF), which is involved in certain cancers (Koivunen et al., 2012) (Figure 1). The surprising discoveries about IDH1/IDH2 have helped to spark enormous interest in cancer metabolism. They have also spawned new areas for cancer drug discovery that had little precedent prior to these cancer genome studies. Lineage Survival Oncogene Transcription Factors Another important discovery concerned master lineagespecic transcription factors (TFs). Because such TFs are typically involved in terminal differentiation of cell types, the prevailing hypothesis was that overexpression would suppress cancer by promoting lineage maturation and cell-cycle arrest. Surprisingly, however, an integrative analysis of genome-wide copy number and transcription showed that MITF, which encodes the master transcription factor that regulates melanocyte survival and differentiation, underwent gene amplication in a subset of metastatic melanomas (Garraway et al., 2005). MITF thus served as a prototype for a new category of cancer genes termed lineage survival oncogenes. Systematic genomic studies subsequently uncovered several additional lineage survival oncogene TFs. Examples include NKX2.1 in lung adenocarcinoma, SOX2 in esophageal cancer, and CDX2 in colorectal cancer (Bass et al., 2009; Salari et al., 2012; Weir et al., 2007). In hindsight, these TFs are analogous to the androgen receptor (AR), a nuclear hormone TF that plays crucial roles in proliferation and survival of normal and malignant prostate epithelia and is frequently amplied or mutated during tumor progression (Taplin et al., 1995). Exome-sequencing studies of castration-resistant prostate cancer have recently identied somatic mutations in both AR and several key coregulators (Grasso et al., 2012). Epigenomics One of the most far-reaching discoveries from genomic studies has been the critical role of epigenomic changes in tumorigenesis, which in turn has unleashed a torrent of hypothesis-driven studies and drug discovery efforts. Abnormal DNA methylation and chromatin structure were known to be common in cancers, but it was unclear whether these epigenomic changes played a causal role in cancer or were simply a noncausal correlate of the cancerous state. The question was settled with the recognition that 40 genes encoding epigenomic regulators show highly recurrent somatic alterations across a wide range of cancer types (reviewed in Dawson and Kouzarides, 2012). Mutations that affect the epigenome would seem like a highly efcient mechanism to rewire cellular circuitry because they provide a way to affect multiple target genes simultaneously. The next several sections discuss various epigenomic processes related to chromatin and DNA methylation that are affected by mutations in cancer (Figure 2). Chromatin: Histone Modication Genomic studies have provided clear genetic evidence that dysregulation of chromatin modiers drives many types of cancer. Recurrent mutations were found in genes encoding enzymes
Figure 2. Genes Encoding Epigenetic and Chromatin Regulators Are Frequent Targets of Mutations in Cancer
The enzymes DNMT3A and TET2 regulate 5-methylcytosine and 5-hydroxymethylcytosine production in genomic DNA; the genes encoding these enzymes are frequently mutated in leukemias. The histone H3 component of the nucleosome undergoes extensive modications involving its lysine (K)-rich tail. Genes encoding enzymes that read, produce, or interpret these modications are frequently mutated in cancer. Examples include histone lysine methyltransferases (KMTs), histone lysine demethylases (KDMs), and histone acetyltransferases (HATs). Genes encoding components of the SWI/SNF chromatin-remodeling complex are also recurrently mutated in cancer. Novel therapeutics targeting chromatin and epigenetic mechanisms have entered clinical use or are in development (boxed).
that add, subtract, or interpret posttranslational modications to histone H3. These enzymes include histone (lysine) methyltransferases (KMTs) and histone (lysine) demethylases (KDM), which activate or repress genes by modifying specic lysine residues; histone acetyltransferases (HATs), which regulate transcription by adding acetyl groups to the histone H3 tail; and histone readers, which bind various histone modications and recruit additional protein complexes to carry out specic effector functions (Figure 2). Among the KMTs, mutations affect the MLL subfamily, which acts on lysine 4 of H3 (e.g., H3K4); the NSD subfamily, which acts on H3K36 (Dolnik et al., 2012); and EZH2, which methylates H3K27 (Morin et al., 2010). Among the KDMs, mutations affect JARID1A and UTX, which demethylate H3K4 and H3K27, respectively. Among the HATs, mutations affect CREBP and EP300 (Gui et al., 2011; Morin et al., 2011; Peifer et al., 2012).
The genes encoding histone-modifying enzymes typically exhibit lineage-restricted mutational patterns. For example, the NSD1 and NSD3 KMTs have so far been found rearranged only in AML (Jaju et al., 2001; Rosati et al., 2002), and mutations affecting the histone demethylases (HDMs) KDM5A and KDM5C appear to occur exclusively in AML and renal cell cancer, respectively. However, some genes show a much broader mutational distribution. Although MLL is named for its association with a particular leukemia subtype (mixed lineage leukemia), cancer sequencing studies have found recurrent MLL gene mutations in a variety of hematologic and solid tumors, including smallcell lung cancer (Peifer et al., 2012), lung squamous cancer (Hammerman et al., 2012), gastric cancer (Zang et al., 2012), head and neck cancer (Stransky et al., 2011), and prostate cancer (Barbieri et al., 2012; Grasso et al., 2012). Interestingly, the DOT1L KMT, which methylates H3K79, is not itself mutated but becomes essential in MLL-translocated leukemias (Bernt et al., 2011). HAT mutations are found in B cell lymphomas (Pasqualucci et al., 2011a), small-cell lung cancers (Peifer et al., 2012), and medulloblastoma (Robinson et al., 2012). These distinctive patterns suggest that mutations affecting chromatin-modifying enzymes contribute to cancer by disrupting expression of specic target genes that play critical roles in particular cell types. However, the identities of these target genes remain unknown, and we lack systematic methods for identifying them. Affected tumors are typically heterozygous for apparent lossof-function alleles, indicating that haploinsufciency for these chromatin-modifying enzymes propels cancer and that complete loss is cell lethal. (An exception is EZH2, in which gain-of-function mutants are observed in follicular lymphoma [Morin et al., 2010], whereas loss-of-function events are seen in myeloid cancers [Jankowska et al., 2011; Makishima et al., 2010]). This makes chromatin-modifying enzymes attractive targets for anticancer drugs because cancer cells carrying only one functional gene could be uniquely sensitive to inhibitors (a synergy termed synthetic lethality). Vigorous drug discovery efforts are currently underway against many of the enzymes. So far, the only drugs targeting histone-modifying enzymes in clinical use are the histone deacetylase (HDAC) inhibitors. Ironically, the HDAC inhibitor vorinostat is approved for treatment of myelodysplastic syndromes and cutaneous T cell lymphomas, although HDAC genes have not been found mutated in these (or any other) malignancies. Chromatin: Nucleosome Remodeling Another major mutational target affecting chromatin biology is the SWI/SNF complexes (Figure 2), which regulate chromatin structure through ATP-dependent nucleosome remodeling (for a recent review, see Wilson and Roberts [2011]). The importance of these complexes in tumor biology was initially suggested by the discovery of biallelic deletions involving SNF5 (a core SWI/ SNF protein) in malignant rhabdoid tumors (an aggressive pediatric cancer). Multiple cancer sequencing surveys then revealed that the class of genes encoding SWI/SNF factors is one of the most commonly mutated targets in cancer. In renal cell cancer, 41% of tumors harbor mutations in PBRM1, which encodes BAF180, a histone acetylation reader and integral component of the so-called BAF SWI/SNF complex) (Varela et al., 2011);
only the VHL tumor suppressor is mutated more commonly in this malignancy. Similarly, >50% of ovarian clear cell carcinomas carry inactivating mutations in ARID1A, which encodes another BAF protein (Jones et al., 2010; Wiegand et al., 2010). Frequent ARID1A mutations have since been observed in many other cancer types, including up to 30% of hepatocellular carcinomas (Fujimoto et al., 2012; Huang et al., 2012a), 34% of bladder cancers, and 21% of endometrioid cancers. Its homologs ARID1B or ARID2 (a component of the PBAF SWI/SNF complex) harbor recurrent mutations in melanoma (Hodis et al., 2012; Krauthammer et al., 2012), hepatocellular (Fujimoto et al., 2012; Li et al., 2011), and pancreatic cancers (Biankin et al., 2012). As with other histone-modifying proteins, the SWI/SNF gene mutations are typically loss-of-function alleles; they often exhibit biallelic inactivation or loss of protein expression, consistent with a tumor suppressor mechanism. Chromatin: Compaction Another unexpected chromatin-related target is the chromodomain-helicase-DNA-binding (CHD) gene family. CHD proteins regulate chromatin compaction during stem cell differentiation and may also promote genome stability (Ho and Crabtree, 2010) (Figure 2). Inactivating CHD1 mutations and deletions comprise likely founder events (together with SPOP mutations) in a newly recognized ETS-negative genetic subtype of prostate cancer (Barbieri et al., 2012), where they appear to confer distinct patterns of genome derangement (Huang et al., 2012b). The homolog CHD4 is frequently deleted in endometrial cancers (Le Gallo et al., 2012). Histone H3.3 itself contains highly recurrent hot spot mutations in pediatric astrocytoma and a subtype of medulloblastoma (Robinson et al., 2012; Schwartzentruber et al., 2012). Overall, the discovery of extensive chromatin and epigenetic mutations by unbiased cancer genome characterization has opened up vast new areas of basic and clinical discovery. DNA Methylation Systematic surveys have revealed that DNA methylation also plays a critical role in shaping the cancer genome (Figure 2). In particular, some cancers show a clear CpG island methylator phenotype (CIMP). The notion that DNA hypermethylation might dene a biologically important cancer subtype in colorectal cancer (CRC) originated from focused studies of individual genes (Toyota et al., 1999), but other reports challenged its existence or at least its biological relevance. Systematic interrogation of all available methylation markers (at that time) across >100 CRC samples provided denitive evidence for CIMP in CRC (CRCCIMP) (Weisenberger et al., 2006). Most CRC-CIMP tumors show high microsatellite instability (MSI) (Ogino et al., 2006; Weisenberger et al., 2006); this is likely due to the fact that such tumors typically have hypermethylation (and hence repression) of the MLH locus, whose loss of expression results in MSI. The etiology of CIMP in CRC remains mysterious, with these tumors showing few mutations in the DNA methylation machinery. Subsets of glioblastoma and AML were also found to have CIMP-like patterns (as described above). In these cases, the phenomenon is likely due, in part, to 2HG generated from mutant IDH1/2 proteins (Noushmehr et al., 2010), as described above. DNA hypomethylation also plays an important role in some cancers. A whole-genome sequencing survey revealed that
25% of AMLs carry inactivating mutations in DNMT3A (Ley et al., 2010), an enzyme that catalyzes the addition of methyl groups to CpG dinucleotides. AML cells with DNMT3A mutations show reduced DNA methylation at the promoter of many genes jkova et al., 2012); these mutations correinvolved in cancer (Ha late with poorer overall survival (Ley et al., 2010). Subsequently, recurrent DNMT3A mutations were also found in the myelodysplastic syndrome (MDS) (Walter et al., 2011), a neoplastic condition that often progresses to AML. The recognition of the key role of DNA methylation has galvanized interest in drugs that inhibit this process, such as 5-azacitidine and decitabine. Conceivably, these drugs may act in a synthetic lethal manner against tumors carrying mutations in DNMT3A and other genes affecting DNA methylation. Azacitidine has proved especially intriguing: it is the rst drug to improve the survival of patients with myelodysplastic syndrome (MDS) and has also shown promising efcacy in AML (reviewed in Estey [2007]). DNMT3A mutations or other altered methylation phenotypes may dene leukemic patient subpopulations that are more likely to benet from these drugs (Marcucci et al., 2012). As with chromatin dysregulation, the critical genes affected by aberrant DNA methylation remain unclear. DNA Hydroxyl Methylation Genomic studies have uncovered a link between a novel epigenetic modication and cancer. In 2009, biochemical studies identied a new type of DNA modication: the conversion of 5-methylcytosine (5mC) at CpG islands to a hydroxylated variant called 5-hydroxymethylcytosine (5hmC) by the ten/ eleven translocation (TET) family of DNA hydroxylases (Kriaucionis and Heintz, 2009; Tahiliani et al., 2009) (Figure 2). Soon thereafter, genomic surveys found that a family member TET2 shows recurrent inactivating mutations in AML, MDS, and other myeloproliferative disorders (Delhommeau et al., 2009; Langemeijer et al., 2009). As noted above, the TET enzymes require a-ketoglutarate for their activity and are inhibited by the 2HG oncometabolite product of mutant IDH1/2. TET2 and IDH1/2 mutations thus act, at least in part, through a common mechanism; as would be expected, these mutations rarely co-occur in AML. Interestingly, however, TET2 and DNMT3A mutations frequently co-occur in MDS, pointing to an as-yet unexplained cooperativity between dysregulation of 5mC and 5hmC in leukemogenesis. RNA Splicing Complementing the targets above affecting RNA transcription, cancer sequencing uncovered other important targets involved in RNA splicing (Figure 3). Though it had long been known that cancers showed aberrant splicing patterns, it was impossible to know whether these events played a causal role in cancer or were simply an effect of cancer. The answer became clear with exome-sequencing studies in chronic myelogenous leukemia (CLL) and myelodysplastic syndromes (MDS). In CLL, the spliceosome gene SF3B1 is mutated in 10%15% of cases, and other spliceosomal genes, such as SFRS1, SFRS7, and U2AF2, are also mutated at lower frequencies (Puente et al., 2011; Quesada et al., 2012; Wang et al., 2011a). In MDS, the spectrum is even more striking: 45%85% of cases harbor mutations in a spliceosome gene, with SF3B1 and U2AF1 being the most common and other
Figure 3. Cancer-Associated Mutations in the RNA-Splicing Machinery

Genes encoding spliceosomal components are recurrently mutated in both hematologic malignancies and solid tumors. Drugs that target SF3B1 have entered clinical trials.
genes (such as SF3A1, ZRSR2, SRSF2, and U2AF2) occurring at lower frequencies (Papaemmanuil et al., 2011; Yoshida et al., 2011). Spliceosomal genes have also been found signicantly mutated in solid tumorsmost notably, U2AF1, SF3B1, U2AF2, and PRPF40B mutations in lung adenocarcinomas (Imielinski et al., 2012). SF3B1 is also recurrently mutated in breast cancer (Ellis et al., 2012) and pancreatic cancer (Biankin et al., 2012). The pattern of mutations in the spliceosomal genes contains important clues about their function. First, the mutations tend to occur in a mutually exclusive fashion in all tumor types examined, suggesting that they play similar roles and are thus functionally redundant with respect to causing cancer (for a recent review, see Lindsley and Ebert [2013]). Second, several of the genes carry heterozygous missense mutations affecting specic protein domains, suggesting that they confer a gain of function. SF3B1 (encoding a member of the splicing factor 3b complex, which interacts with SF3A proteins and a snRNA species to form the U2 small nuclear ribonucleoprotein [snRNP]) has mutations affecting the carboxy-terminal HEAT domains. U2AF1 (encoding a member of the U2 snRNP auxiliary factor, a spliceosomal component that binds the 30 splice acceptor site within target pre-mRNAs) has mutations affecting conserved zinc nger domains. SRSF2 (a serine-arginine-rich protein that mediates U2 snRNP assembly through binding of exon-splicing enhancer elements within pre-mRNA species) also has distinct codon localizations (Yoshida et al., 2011). In contrast, the ZRSR2 gene (encoding a spliceosomal adaptor protein) has mutations distributed throughout its open reading frame and has frequent nonsense mutations; the pattern is indicative of loss-of-function mutations. What is missing, of course, is knowledge of the specic aberrant cancer-related splicing events caused by these mutations. Genotype-phenotype connections offer some additional clues. In MDS, SF3B1 mutations occur primarily in subtypes associated with ring sideroblasts (Papaemmanuil et al., 2011; Yoshida et al., 2011), whose presence signies defective erythrocyte maturation. This observation raises the possibility that mutated SF3B1 may cause ring sideroblast formation, at least in some MDS subtypes, by governing splicing of a key erythroid lineage differentiation factor. Mutations in several splicing factors carry prognostic information that might inuence clinical management. For example, U2AF1 mutations have been linked to increased progression from MDS to AML, and SRSF2 mutations correlate with the socalled chronic myelomonocytic leukemia (CMML) subtype of MDS. In CLL, SF3B1 mutations correlate with more rapid disease progression and lower overall survival (Quesada et al., 2012; Wang et al., 2011a). U2AF1 mutations were associated
with poor progression-free survival in lung adenocarcinoma (Imielinski et al., 2012). Splicing factors were not previously considered attractive targets for anticancer therapies, but that assessment is changing (Figure 3). Indeed, several small molecules and natural products known to target the spliceosome have been reported, including spliceostatin A (SSA), a metabolite derived from Pseudomonas that inhibits the SF3b complex and suppresses splicing in vitro, and pladienolide, a compound produced by Streptomyces platensis that inhibits the SF3B1 protein directly (Kaida et al., 2007). A derivative of pladienolide called E7107 has entered phase I clinical trials and shows moderate activity in thyroid cancer (Folco et al., 2011). Protein Homeostasis Genome-wide and exome-wide sequencing in multiple myeloma (MM) suggested an unexpected (and still unexplained) role of protein synthesis and degradation. In MM, mutations were found at high frequencies in DIS3, FAM46C, and XBP1 (Chapman et al., 2011a). DIS3 is an RNA exonuclease that regulates RNA abundance through the exosome complex. FAM46C is a protein whose function remains unknown but whose expression pattern is nearly perfectly correlated with that of genes encoding ribosomal proteins, eukaryotic initiation factors, and translation elongation factors (Chapman et al., 2011a). XBP1 encodes a factor involved in the unfolded protein response; mutations in the mouse homolog cause a myeloma-like condition. Unbiased genomic studies have also uncovered unexpected roles for the ubiquitination machinery in cancer. In prostate and endometrial cancers, mutations in SPOP have been observed in 8%14% of cases (Barbieri et al., 2012; Kan et al., 2010; Le Gallo et al., 2012). SPOP encodes the substrate recognition component of an E3 ubiquitin ligase complex. In prostate cancer, the SPOP mutations affect highly conserved amino acid residues situated within the substrate-binding motif (MATH domain), suggesting that they abrogate normal ligase/ substrate interactions. These mutations are mutually exclusive with ETS rearrangements, thereby dening a distinct genetic subtype of prostate cancer (Barbieri et al., 2012). In endometrial cancer, SPOP mutations also occur in the MATH domain but involve different amino acid residues than those seen in prostate cancer (Le Gallo et al., 2012). The distinct pattern of mutations in these two cancers suggests loss of recognition for distinct substrates, leading to their accumulation. The ubiquitin ligase gene FBXW7 shows recurrent mutations in endometrial, head/neck, bladder, and GI cancers but only rarely shows recurrent mutations in prostate cancer. In contrast to SPOP, the mutations appear to be simple loss-of-function events. The ubiquitin ligase gene WWP1 thus far only shows recurrent mutations in liver cancer (Fujimoto et al., 2012). The
Figure 4. Genetic Alterations Affecting Lineage Specication Are Common in Squamous Tumors
NOTCH and several other lineage regulatory factors are disrupted by genomic alterations in lung, cervical, head/neck, and cutaneous squamous carcinomas. SOX2 is also a lineage survival oncogene that regulates squamous maturation. Genes that encode proteins shaded in red undergo mutational activation or amplication; those shaded gray undergo mutational inactivation or deletion.
distinct spectra of cancers seem likely to result from insufcient degradation of different proteins that are critical for different cell types. Finding the protein targets is a high priority. Squamous Differentiation Exome-sequencing studies in head and neck squamous cell carcinoma (HNSCC) revealed unexpected roles for pathways involved in squamous cell differentiation (Agrawal et al., 2011; Stransky et al., 2011). The studies found mutations in NOTCH1 in 15% of cases, as well as mutations and focal copy number alterations of NOTCH2 and NOTCH3 in an additional 11% (Stransky et al., 2011) (Figure 4). Whereas activating NOTCH1/ 2 mutations had been reported in various blood cancers (Lohr et al., 2012; Pasqualucci et al., 2011b; Puente et al., 2011; Weng et al., 2004), the NOTCH mutations in HNSCC were clearly loss-of-function events. Parallel studies in myeloid leukemia also identied recurrent loss-of-function NOTCH mutations (Klinakis et al., 2011). The NOTCH mutations turned out to be just a part of the story. A more sophisticated analysis (of gene sets with recurrent mutations) pointed to genes known to be involved in epidermal development and squamous differentiation in HNSCC (Stransky et al., 2011) (Figure 4). Additional genes mutated in HNSCC (such as RIPK2, EZH2, and DICER1) were linked to the squamous differentiation program based on results from genetically engineered mice. Two further genes (SYNE1 and SYNE2, mutated in 20% and 8% of cases, respectively) were also implicated; these genes encode proteins that control nuclear polarity and spindle orientation, which stand upstream of NOTCH signaling in squamous lineage development (Williams et al., 2011). In all, nearly one-third of HNSCC tumors appeared to harbor at least one mutation predicted to affect squamous differentiation. Comprehensive genomic studies soon demonstrated the importance of dysregulated squamous differentiation in other tumor types. For example, inactivating NOTCH1/2 mutations occur in >75% of cutaneous squamous cell carcinomas (Wang et al., 2011b). Moreover, a study of squamous lung cancer revealed that 44% of cases harbored mutations in genes that
Figure 5. Genetic Alterations Disrupt Multiple Cellular Processes

Alterations in a range of cellular processes presumably contribute to cancer through their action on one or more target genes, mRNAs, or proteins, although the precise targets remain unknown in many cases (illustrated by shaded ovals). Even in advance of such knowledge, many mutations suggest potential targets for therapeutic development and allow stratication for clinical trials of targeted drugs.
regulate squamous differentiation (Hammerman et al., 2012). The loss of function in squamous differentiation contrasts with the SOX2 lineage survival TF oncogene, which undergoes frequent amplication in squamous lung cancer, HNSCC, and cervical squamous cancers (http://www.cbioportal.org/ public-portal/index.do). Connecting the Dots: From Cancer Genes to Cancer Processes Hanahan and Weinberg have proposed hallmark processes that must become dysregulated in tumorigenesis and metastasis (Hanahan and Weinberg, 2000, 2011). These processes include genome instability, unlimited cell division, sustained proliferative signaling, evasion of growth suppression, cellular energetics, and resisting apoptosis. Many classical cancer genes encode proteins that mediate or control such processes: for example, mutations in receptor tyrosine kinases or cell-cycle inhibitors can be directly understood in terms of jamming the accelerator pedal or eliminating the brakes on cell growth. By contrast, many of the newly discovered cancer genes affect global processes whose precise connection to cancer remains obscure. These cancer genes act by deranging gene expression (through changes to chromatin and DNA methylation), RNA splicing, protein synthesis and degradation, and cellular metabolism (Figure 5). Presumably, these global changes propel cancer by affecting one or more specic targets involved in cancer processesactivating or repressing specic
genes, altering the isoforms of specic mRNAs, and increasing or decreasing steady-state levels of specic proteins. The key targets are likely cell type specic, accounting for the presence of specic subsets of driver genes in particular cancer types. For the most part, we are ignorant of the precise targetsor whether we are looking for single targets or multiple targets. Indeed, mutations affecting global processes seemingly provide an efcient mechanism by which multiple coregulated targets might be affected. In some respects, the situation may be analogous to amplication and deletion of chromosome arms, which may provide a similarly efcient means to dysregulate multiple targets. In each case, identifying the full range of target genes will likely require unbiased genomic surveys at the DNA, RNA, and protein levels to generate hypotheses, as well as focused experiments to prove them. Connecting the new cancer genes to known (or as-yet unknown) cancer processes will surely accelerate efforts to understand and treat cancer. Of course, therapeutic progress can be made even without a full understanding of their action. For example, inhibitors of the neomorphic IDH1/2 enzymes or perturbed splicing factors may prove valuable even without understanding the full range of enzymes affected by 2HG or SF3B1. Moreover, the set of cancer genes mutated in a tumor provides a powerful classication tool, identifying natural subtypes that can be studied in both preclinical and clinical investigation to detect distinct vulnerabilities and correlate outcomes. Completing the Picture: Long Tails, Dark Matter, Heterogeneity, and Heredity Genomic studies have denitely shown that our previous inventory of cancer genes was far from complete. The question now is do we nally have a near-comprehensive catalog? The honest answer: we dont know. Long Tails For many cancer types, a handful of cancer genes are mutated at high frequency, but many more cancer-related genes are found mutated at much lower frequencies. For example, a recent genomic study of breast cancer reported 40 loci that were mutated at statistically signicant rates (Stephens et al., 2012); of these, 53% of the apparent driver mutations or focal copy number alterations were concentrated in six genes (TP53, PIK3CA, ERBB2, FGFR1/ZNF703, and GATA3), and the remainder were dispersed across 34 genes. Only eight of the genes were mutated in at least 10% of breast cancers. Many tumor types exhibit similar long tail distributions. Some of the genes found mutated at low frequencies in some cancers are more commonly (and signicantly) mutated in other cancers. In the breast cancer example mentioned above, long tail genes that are signicantly mutated in other cancers include the SWI/SNF complex genes ARID1A and ARID1B, the KMT-encoding genes MLL2 and MLL3, and KRAS. This nding might suggest that the discovery of new driver genes is approaching a plateau. On the other hand, the fact that so many driver genes occur at lower frequencies raises the possibility that many such genes may yet remain undiscovered. Moreover, some tumors (e.g., some primary prostate cancers) appear to lack even a single mutation in a proven driver gene.
The problem is due, in part, to the fact that most studies to date have been insufciently poweredlacking adequate sample size to detect low-frequency events and/or adequate depth of sequence coverage to overcome impurity due to stromal contamination. Fortunately, it should be feasible to enumerate all genes carrying nonsynonymous coding mutations in at least 2% of tumors of every cancer type by sequencing a sufciently large number of tumor-normal pairs. (Roughly 950 pairs will be needed per tumor type if the background mutation rate in the cancer is 2 mutations per Mb and 2,500 pairs if the rate is 10/Mb.) This scale seems readily achievable for many tumor types over the next several years. Dark Matter In contrast to point mutations in coding regions, our ability to discover and understand other types of driver mutations is still distressingly limited. Many more important cancer drivers may be lurking in the places that we cannot currently interpret. These include copy number alterations, chromosomal rearrangements, and noncoding regions. As noted above, gains and losses spanning whole chromosome arms occur commonly in most types of cancer, but it is difcult to pinpoint the key genes for which the presence of a few extra copies contributes to cancer. Even for focal amplications or deletions, nding the target genes can be difcult. A study of copy number alterations across cancer types found that proven cancer genes were known for less than half of recurrent focal amplications and an even smaller proportion of recurrent focal deletions (Beroukhim et al., 2010). Incorporating sample-matched data sets can help to suggest candidates for functional validation. For example, a study in glioblastoma showed that gain of extra copies of chromosome 7 was associated with dysregulation of the HGF-MET axis (Beroukhim et al., 2007); pharmacologic experiments showed that cell lines carrying nonfocal chromosome 7 gains together with HGF and MET overexpression were preferentially dependent on MET signaling. Chromosomal rearrangements are also pervasive in many cancers, but our ability to characterize and interpret their impact has been limited. Whereas basic cancer genome analyses can be accomplished by mapping short DNA sequences to a xed reference sequence, comprehensive study of rearrangements requires obtaining larger-scale linking information to reconstruct unexpected genomic junctions and performing transcriptome sequencing to detect expressed fusion genes. These efforts have been aided by recent computational advances, such as algorithms that reconstruct transcriptomes without the need for an underlying reference genome (Grabherr et al., 2011). Most rearrangements may be random passenger events, but some clearly disrupt cancer genes by creating fusion proteins or by subjecting a gene to new regulation. Genome analysis has identied several new fusions involving known cancer genes, including RAS, RAF, ERG, and PTEN, in prostate cancer and in other malignancies (L.A.G. and E.S.L, unpublished data; Palanisamy et al., 2010; Wang et al., 2011c) and NOTCH genes in breast cancers (Robinson et al., 2011). Although relatively few instances of recurrent rearrangements implicating new cancer genes have emerged (possibly owing to limited sample sizes and the challenge of interpreting these events),
those that have been discovered may implicate new biological processes. Examples include MAST kinases in breast cancer (Robinson et al., 2011) and R-spondin family members in 10% of colon cancers (Seshagiri et al., 2012). The great uncharted frontier is the >98% of the human genome that does not encode proteins. Our ignorance is due to two factors. First, we have lacked adequate data because cancer genome studies to date have largely focused on the exome rather than on the whole genome for reasons of cost. Second, we lack adequate analytical techniques to recognize recurrent mutations in nongenic territory. To detect a cancerassociated target, one must aggregate mutations across a dened region to test whether the rate is sufciently elevated above background. This is straightforward for protein-coding regions, where one can aggregate nonsynonymous mutations across thousands of bases. But it is more challenging for the rest of the genome. Unbiased searches require scanning millions of small regions across the genome to nd those with an unusually high mutation rate. If one searches with a small window, the mutational signal will be weak (unless the mutation frequency is very high) and detection will require large sample numbers. If one uses a large window size, the signal may be obscured by random noise in the surroundings. At present, the best approach may be to focus on regions dened by features corresponding to known biological functions, such as promoters, evolutionary conservation, and epigenomic modication. A recent study of regulatory regions in melanoma has conrmed that important mutations may be lurking in noncoding regions (Huang et al., 2013). Whole-genome sequencing revealed the presence of highly recurrent somatic mutations at two specic nucleotides situated within the promoter of the TERT gene, which encodes a reverse transcriptase component of the telomerase enzyme. Both of these mutations are cytidine-to-thymidine transitions that generate a de novo binding site for the ETS transcription factor. These sites increase expression from the TERT promoter in reporter assays. The mutations occur in >70% of melanomas and 16% of other tumor types examined, including bladder and hepatocellular carcinomas. Heterogeneity Cancer genome analyses have largely focused on tumors as a whole. Yet it has been clear for decades that tumors show extensive cellular and molecular heterogeneity. Indeed, heterogeneity was inherent in Nowells original clonal model for tumor evolution (Nowell, 1976). Some early genomic studies have begun to come to grips with tumor heterogeneity. Initial forays have documented subclonal variation across distinct geographic regions of a primary tumor (Gerlinger et al., 2012) and within hematopoietic malignant populations (Ding et al., 2012). Studies of heterogeneity are beginning to provide fascinating glimpses into paths of tumor evolution. For example, a study of 21 breast cancers showed that the most recent common ancestral tumor cellwhich contains the full complement of mutations common to all tumor cellsarose remarkably early in molecular time (Nik-Zainal et al., 2012b). The precursor cell typically gives rise to a dominant subclone that represents at least 50% of all cells in the primary tumor. Knowledge of intratumoral heterogeneity has also revealed instances of conver28 Cell 153, March 28, 2013 2013 Elsevier Inc.
gent evolution. In one study of renal cancer, only 30%35% of somatic mutations were concordant across multiple primary and metastatic sites sampled (Gerlinger et al., 2012); however, several cancer genes contained distinct genomic alterations that had arisen in geographically disparate regions of the primary tumor. This observation thus revealed a remarkable mutational consolidation that engaged critical pathways linked to chromatin regulation (SETD2, KDM5C) or signal transduction (PTEN, mTOR). Tumor heterogeneity could have important implications for precision cancer medicine. Some subclones may contain pre-existing mutations that confer drug resistance or accelerate tumor relapse in cancers that show poor clinical responses to targeted inhibitors. Studies that seek to stratify patients for clinical trials of targeted agents based on specic actionable mutations may be confounded if a biopsy sample is not representative of the whole tumor. On the other hand, the ability to identify driver or resistance mutations within subclonal populations may allow improved prediction of clinical outcomes (Landau et al., 2013). The growing understanding of intratumoral heterogeneity may inform the design of clinical studies that account for this process (e.g., by following the therapeutic response of the biopsied lesion in addition to the overall tumor burden) and circumvent its subversive effects (e.g., by developing therapeutic combinations directed against major and minor subclones). Studies of cancer heterogeneity will be accelerated by recent genomic advances enabling single-cell sequencing (Navin et al., 2011). Whole-exome sequences have been produced from single cells in both hematologic neoplasms and solid tumors (Hou et al., 2012; Xu et al., 2012). Moreover, new protocols that yield more uniform and accurate whole-genome amplication have been developed (Zong et al., 2012). Single-cell analyses have already provided new insights into the evolutionary history of tumors within individual patients and have revealed functional differences across individual tumor cells (Kreso et al., 2013). In the future, these advances may enable detailed genomic studies of circulating tumor cells, thereby providing high-resolution monitoring of therapeutic responses or emerging resistance mechanisms and facilitating detection of aggressive tumor subclones. Heredity Although many of the genetic factors that drive a cancer are acquired through somatic mutation, some are inherited at birth. Epidemiological studies have long noted an increased risk of cancer in relatives of affected individuals (Pomerantz and Freedman, 2011). Genomics has revealed many genes that inuence predisposition to cancer, although the picture remains far from complete. Our focus in this Review is on somatic mutations, but we briey summarize the current state of progress for inherited variation (see recent reviews by Hindorff et al. [2011] and Chung and Chanock [2011]). One method to identify genes that confer predisposition to cancer is to study rare, highly penetrant Mendelian cancer syndromes. These syndromes arise when mutant alleles confer such a high increased risk (>10-fold) that it is straightforward to trace their transmission in families by linkage analysis. More than 100 genes underlying such cancer syndromes have been
identied, including those underlying retinoblastoma (RB1), breast cancer (BRCA1, BRCA2), and colon cancer (APC, MUTYH and the mismatch repair genes MLH1, MSH1, MSH6, and PMS2). Such genes have been deeply informative about cancer biology but together account for <5% of the estimated heritability of cancer (Cazier and Tomlinson, 2010). To identify cancer genes that confer more modest risks, it is necessary to use population-based association studies rather than family-based linkage studies. The methodology for association studies depends on whether one wishes to study common (>1%) or rare (<1%) variants. Common variants are frequent enough that they can be tested for their individual effects on cancer risk by genotyping of millions of variants in cases and controls in genome-wide association studies (GWAS) (Altshuler et al., 2008). Rare variants must be combined together for analysis: studies examine the aggregate frequency of rare coding variants in each gene to look for an elevated frequency in cases versus controls. More than 150 cancer risk loci have been identied thus far, with most having been found through GWAS (Chung and Chanock, 2011; Hindorff et al., 2011). The common alleles appear to include many regulatory variants and to confer a lower increased risk (<30%), whereas the rare alleles affect coding regions of known cancer genes (such as ATM, BRIP1, CHEK2, PALB2, and RAD51C in breast cancer) and tend to have higher risk (2- to 3-fold). The relative roles of the two classes vary among cancer types. Importantly, the risk factors identied to date explain only a fraction of the heritability of cancer (Hindorff et al., 2011). Genomic studies with much larger samples will be needed to obtain a fuller picture of the inherited basis of cancer risk. Understanding the mechanisms by which common inherited genetic variants predispose to cancer will require integrative genomic analysis, which will likely yield important biological insights. One instructive case is found in a 500 kb gene desert in chromosome 8q24. Whereas most cancer-associated loci are tumor-type specic, this region contains variants that affect risk of prostate, colon, esophagus, head/neck, breast, and pancreas cancers (reviewed in Hindorff et al. [2011]). Epigenetic and chromosome conformation studies in human and genetic engineering studies in mouse suggest that the variants alter distal regulatory sequences controlling the MYC locus, which lies telomeric to the region (Ahmadiyeh et al., 2010; Pomerantz et al., 2009; Sur et al., 2012; Tuupanen et al., 2009). A similar situation occurs in a 500 kb region in 9p21, where different variants affect multiple types of cancer (including breast cancer, melanoma, glioma, and leukemia) as well as noncancer-related diseases such as type 2 diabetes and myocardial infarction; these variants likely alter regulation of the cell-cycle genes CDKN2A/CDKN2B. The observation that a number of additional cancer-related loci also affect diabetes (or insulin dysregulation) suggests an important role for metabolic processes in cancer (Dupuis et al., 2010; Pal et al., 2012). Finally, understanding the inherited factors may help to explain some disparities among ethnic groups. For example, a proportion of the higher risk of prostate cancer in African Americans and other men of African descent may be due, in part, to allele frequency differences at chromosome 8q24 (Haiman et al., 2011; Murphy et al., 2012).
Applying the Knowledge: Diagnostics and Therapeutics The ultimate test of cancer genomics will be its ability to improve diagnostics and therapeutics. Academic centers are already beginning to adopt rst-generation genome proling platforms to guide cancer treatment (Dias-Santagata et al., 2010; MacConaill et al., 2009; Thomas et al., 2007; Wagle et al., 2012) These platforms involve testing a few hundred specic cancer-associated mutations or performing full sequencing of a limited set of cancer-associated genes (Lipson et al., 2012; Wagle et al., 2012). The early returns suggest that, in 40% 60% of cases for many common solid tumors, the information points to at least one alteration that might inuence therapeutic decision-making or might suggest enrollment in a particular clinical trial (Beltran et al., 2012; Hammerman et al., 2012; CGAN, 2012). As sequencing costs fall, diagnostics may move to whole-exome or whole-genome sequencing. The challenge will be to lter and annotate the results for oncologists, based on a constantly changing landscape of scientic knowledge. Eventually, genomic analysis will likely become part of the standard of care for cancer patients. Cancer genomics will also become a key component in the design, execution, and interpretation of clinical trials. Investigators are already using genomic information for retrospective clinical analyses that correlate treatment response with specic genomic features. There is growing interest in using deep genomic characterization of exceptional cases, such as rare tumors that show a complete clinical response to a particular anticancer regimen. For example, a recent tumor genomesequencing study of a bladder cancer patient who experienced a complete response to a TOR inhibitor (everolimus) identied two distinct cancer gene mutations (TSC1 and NF2) predicted to affect oncogenic TOR signaling (Iyer et al., 2012). Sequencing of additional everolimus-treated tumors conrmed that TSC1 mutations correlate with clinical response. The prospective use of genomic information may substantially transform trial design. Cancer trials have traditionally selected patients based on histologic tumor subtypes. However, it makes more sense to test targeted therapeutics on the subset of patients carrying the relevant genetic lesions; by selecting the patients most likely to benet, one decreases sample size, cost, and unjustied harm. In some cases, it will make sense to enroll patients carrying the same genetic alteration across a wide range of tumor types (for example, a trial led by investigators at Memorial Sloan-Kettering Cancer Center in which BRAFV600 mutant tumors from colon, thyroid, lung, and other organ sites are treated with a selective RAF or MEK inhibitor). Moreover, novel designs are becoming possible in which one simultaneously tests multiple drugs or drug combinations. In these basket trials, patients are assigned to different therapeutic regimens based on the specic genetic proles in their tumor (Kim et al., 2011). Basket trials may employ an adaptive design, allowing real-time adjustments if hints of specic genotype-driven responses are detected (Berry, 2011). Trials can further be shaped by genomic analysis from serial biopsies to assess pharmacodynamics response and to characterize the presence of resistance mechanisms. It may be useful to create a worldwide clearinghouse mechanism that connects patients to trials based on their genotype, especially to obtain
a large enough sample to evaluate responses in tumors with rarer genetic features. Cancer drug discovery efforts are already being shaped by the ndings from genomic studies. In some cases, the product of the mutated gene may be an appropriate drug target. In many other cases, mutations may confer specic vulnerabilities on the cancer cell that can be discovered through functional genomic studies, such as comprehensive gene inhibition screens with RNA interference across large numbers of cancer cell lines with varying genotypes. Beyond the development of specic drugs, knowledge of the cancer genome will be critical to design combination therapies, which will be essential for conquering cancer. Most tumors eventually develop resistance to single-agent therapeutics (reviewed nne [2012]). For example, the use of RAF in Garraway and Ja and MEK inhibitors in BRAF mutant melanomas leads to spectacular responses, but tumors reappear within a year (Chapman et al., 2011b; Flaherty et al., 2012a, 2012b; Sosman et al., 2012). Multiple genetic mechanisms of resistance have been described (oncogenic NRAS mutations, COT/MAP3K8 gains, BRAF amplication, activating MEK1 mutations, and NF1 loss), each of which produces sustained MAP kinase (ERK) activity in the presence of drug (Emery et al., 2009; Johannessen et al., 2010; Nazarian et al., 2010; Poulikakos et al., 2011; Wagle et al., 2011; Whittaker et al., 2013). These ndings raise the possibility that adding an ERK inhibitor to existing RAF/MEK inhibitor regimens could provide an additional clinical benet (Whittaker et al., 2013). Systematic preclinical studies may make it possible to anticipate the mechanisms of resistance, allowing therapeutic scientists to plan for resistance long before it arises in the clinic. For example, a recent study performed large-scale screens (using RNAi knockdown and ORF overexpression) to identify genes whose loss or amplication can confer resistance to RAF inhibition in a melanoma cell line; the results were conrmed by clinical observations in patients tumors (Johannessen et al., 2010; Whittaker et al., 2013). Another group systematically screened stromal cell lines to identify those that secrete factors that confer resistance on adjacent cancer cells; the screen revealed that hepatocyte growth factor confers resistance to RAF inhibition (Straussman et al., 2012). Such approaches may make it possible to formulate rational combination therapies even before the results of single-agent clinical trials are known. In the end, combination therapy depends on shifting the odds of resistance. There is cause for optimism: mathematical modeling suggests that resistance may often be due to pre-existing mutations in the tumor cell population (Michor et al., 2005). If so, it should be possible to prevent recurrence by treating simultaneously with drugs directed against several independent targets so that the chance of a single cell carrying all the necessary resistance mutations is vanishingly small. This is, of course, the basis for the successful triple-drug combinations against HIV. Ultimately, cancer genomics should aim to provide a comprehensive roadmap for selecting rational, multidrug combinations for anticancer therapy. Next Steps for Cancer Genomics The early fruits of cancer genome studies have conrmed Renato Dulbeccos prediction about the value of complement30 Cell 153, March 28, 2013 2013 Elsevier Inc.
ing piecemeal approaches with systematic genome-wide studies. The results have already opened new frontiers in basic, translational, and clinical investigation. Still, current studies have only scratched the surface of what can be learned from comprehensive study of the cancer genome. Cancer genomics has largely focused on documenting the mutations in primary tumors. Over the coming years, the eld should expand its focus to gather systematic information to inform a wider range of biological and clinical questions. Below, we suggest four important components for the next phase of cancer genomics. Complete the Mutational Atlas of Primary Tumors A straightforward but critical component is to nish compiling the catalog of signicantly mutated genes in primary tumors of every feasible cancer type. Given the long tail of cancer genes and the variable background mutation rates, such studies will require thousands of tumor-normal pairs. Why bother to press for completeness? Scientically, because the low-frequency drivers may in aggregate make a substantial contribution and because they are likely to harbor further surprises. Medically, because physicians will want to be able to recognize all driver mutations in each patient to optimize therapy. Fortunately, these efforts should become increasingly feasible and affordable given the decreasing costs of sequencing and the increasing ability to analyze small amounts of starting material from formalin-xed, parafn-embedded archival samples. The analysis must expand beyond the exome to include the whole genome (including longrange links to detect translocations), the transcriptome, and the epigenome (at least the methylome and key chromatin modications). Improved laboratory and analytical methods will be needed to discern the targets of nonfocal chromosome copy number aberrations, epigenomic modications, and nongenic translocations. In addition, the genomic information should be thoroughly mined to identify germline variants that contribute to cancer risk. Expand the Mutational Atlas beyond Primary Tumors The second component is to systematically expand the atlas beyond primary tumors to include the natural history of human cancer, as well as the homology to cancer in key model systems. A mutational atlas of the natural history of cancer would involve comprehensive genomic analysis of preneoplastic lesions, metastases from various organ sites, and tumors that show different types of responses to therapies, including extreme response, intrinsic resistance, and acquired resistance. Ideally, all clinical trials in oncology would be subject to such analysis. Genomic characterization should also be applied to animal models of cancer so that we can better connect these to human cancers based on mechanism. In addition to genetically engineered mouse models, intensive studies of naturally occurring cancers in large animals, especially dogs (Karlsson and Lindblad-Toh, 2008), may provide both insights and important preclinical models for drug testing. Create a Functional Encyclopedia of Altered Pathways and Acquired Vulnerabilities Though a mutational catalog will provide a comprehensive picture of cancer genomes, this catalog alone is not enough. We need to produce a functional encyclopedia of altered cellular pathways and acquired vulnerabilities that correspond to each
cancer genome. Genomic approaches can propel systematic functional studies, just as they have propelled comprehensive structural studies. Building a functional encyclopedia will involve (1) creating tractable models representing the full range of cancer genotypes and (2) characterizing these models with respect to their genomic alterations, essential pathways, and therapeutic vulnerabilities. Already, ongoing projects are assembling large collections of cancer cell lines; dening their genomic changes; characterizing their cellular states at the RNA, protein, and posttranslational levels; and determining their sensitivities to anticancer drugs, RNAi-based inhibition of every gene, and microenvironmental interactions (Barretina et al., 2012; Garnett et al., 2012). With a sufciently large collection of cell models, one can correlate pathways and vulnerabilities with specic genetic lesions, providing invaluable insights into cancer biology, markers for patient selection in clinical trials, and potential new targets for cancer drug development. One limitation has been that current cancer cell lines represent a biased sampling of cancer and cancer genotypes, owing to differences in the ability to derive cell lines. However, new methods (such as Rho kinase inhibitor-treated feeder layers and organoid culture systems) appear poised to greatly expand the repertoire of available cancer models (Huch et al., 2013; Liu et al., 2012). Patient-derived xenografts can also play a key role in preclinical studies of new therapeutics. Enable and Promote Sharing of Cancer Genomic Information Finally, there is one critical component that is an essential foundation for the others: widespread information sharing. Cancer genome information will grow exponentially in the years ahead as genome analysis moves from the research lab to routine clinical care for millions of patients around the world. If it were possible to share and analyze this torrent of genomic information together with associated clinical outcome data, it could signicantly accelerate the understanding and treatment of cancer. The information would speed not only the identication of cancer genes, but also the correlation of therapeutic responses to specic tumor genotype, including dramatic responses to new targeted agents seen in some patients and more modest responses to different regimens. In effect, it would connect cancer care around the world into a laboratory for continuous improvement. Making this world a reality will require coordinated efforts by researchers, hospitals, and patient groups to accomplish two goals: (1) creating the computational infrastructure to enable sharing and (2) promoting a culture of sharing. It is easy to imagine an alternative future in which cancer genomic information cannot be aggregated because it is stored in inaccessible sites and incompatible formats, much as is the case with electronic medical records in the U.S. To avoid this outcome, it will be necessary to have common or interoperable standards for data and analysis, cloud-based storage solutions to ensure data security, and rigorous systems to enforce patients instructions concerning their data. But technology platforms alone will not sufce. Clinicians, hospitals, and healthcare networks will need to become engaged in collecting and sharing clinical outcome data. Pharmaceutical companies and others will need to share data from completed clinical trials. Ultimately, patient advocacy groups may provide the impetus for cultural change,
as happened with AIDS. Though it must be up to each patient to decide whether to share his or her data, we suspect that most cancer patients will actively want to allow their information to be appropriately aggregated and shared (with appropriate rules and technology to protect privacy) to accelerate progress for this and future generations of patients. We must ensure that patients have the right and ability to contribute their information to a global ght against cancer. Conclusions Genomics has become a powerful tool for cancer research, yielding important biological surprises and enabling systematic classication based on cellular mechanism. Cancer genomics is just now emerging from its rst phase, which has been largely focused on creating basic mutational catalogs in primary tumors. To fulll its full promise, the eld will need to deepen the structural characterization of cancer genomes, complement it with comprehensive functional characterization of cancer cells, and enable and promote information sharing across the world. Ultimately, cancer genomics is about fully knowing the enemy. While not alone a guarantee of victory, it is an essential part of any overall plan of attack.
ACKNOWLEDGMENTS We thank Todd Golub, Matthew Meyerson, and Michael Stratton for reviewing the manuscript and for valuable comments. We thank Gad Getz, Charles Roberts, and Matthew Freedman for helpful discussions. The design of gures was aided by materials from ScienceSlides (http://www.visiscience.com). L.A.G. and E.S.L. are supported by the U.S. National Human Genome Research Institute, the U.S. National Cancer Institute, the Slim Initiative for Genomic Medicine, and the Broad Institute. L.A.G. also receives support from the Department of Defense, the Starr Cancer Consortium, the Prostate Cancer Foundation, the Melanoma Research Alliance, and the Adelson Medical Research Foundation. L.A.G. and E.S.L. are consultants and equity holders in Foundation Medicine, Inc.
REFERENCES Agrawal, N., Frederick, M.J., Pickering, C.R., Bettegowda, C., Chang, K., Li, R.J., Fakhry, C., Xie, T.X., Zhang, J., Wang, J., et al. (2011). Exome sequencing of head and neck squamous cell carcinoma reveals inactivating mutations in NOTCH1. Science 333, 11541157. Ahmadiyeh, N., Pomerantz, M.M., Grisanzio, C., Herman, P., Jia, L., Almendro, V., He, H.H., Brown, M., Liu, X.S., Davis, M., et al. (2010). 8q24 prostate, breast, and colon cancer risk loci show tissue-specic long-range interaction with MYC. Proc. Natl. Acad. Sci. USA 107, 97429746. Altshuler, D., Daly, M.J., and Lander, E.S. (2008). Genetic mapping in human disease. Science 322, 881888. Baca, S.C., Prandi, D., Lawrence, M.S., Mosquera, J.M., Romanel, A., Drier, Y., Park, K., Kitabayashi, N., MacDonald, T.Y., Ghandi, M., et al. (2013). Punctuated evolution of prostate cancer genomes. Cell 153. Published online April 25, 2013. http://dx.doi.org/10.1016/j.cell.2013.03.021. Banerji, S., Cibulskis, K., Rangel-Escareno, C., Brown, K.K., Carter, S.L., Frederick, A.M., Lawrence, M.S., Sivachenko, A.Y., Sougnez, C., Zou, L., et al. (2012). Sequence analysis of mutations and translocations across breast cancer subtypes. Nature 486, 405409. Barbieri, C.E., Baca, S.C., Lawrence, M.S., Demichelis, F., Blattner, M., Theurillat, J.P., White, T.A., Stojanov, P., Van Allen, E., Stransky, N., et al. (2012). Exome sequencing identies recurrent SPOP, FOXA1 and MED12 mutations in prostate cancer. Nat. Genet. 44, 685689.
Barretina, J., Caponigro, G., Stransky, N., Venkatesan, K., Margolin, A.A., Kim, r, J., Kryukov, G.V., Sonkin, D., et al. (2012). The Cancer S., Wilson, C.J., Leha Cell Line Encyclopedia enables predictive modelling of anticancer drug sensitivity. Nature 483, 603607. Bass, A.J., Watanabe, H., Mermel, C.H., Yu, S., Perner, S., Verhaak, R.G., Kim, S.Y., Wardwell, L., Tamayo, P., Gat-Viks, I., et al. (2009). SOX2 is an amplied lineage-survival oncogene in lung and esophageal squamous cell carcinomas. Nat. Genet. 41, 12381242. Beltran, H., Yelensky, R., Frampton, G.M., Park, K., Downing, S.R., Macdonald, T.Y., Jarosz, M., Lipson, D., Tagawa, S.T., Nanus, D.M., et al. (2012). Targeted next-generation sequencing of advanced prostate cancer identies potential therapeutic targets and disease heterogeneity. Eur. Urol. Published online September 5, 2012. http://dx.doi.org/10.1016/j.eururo.2012.08.053. Bentley, D.R., Balasubramanian, S., Swerdlow, H.P., Smith, G.P., Milton, J., Brown, C.G., Hall, K.P., Evers, D.J., Barnes, C.L., Bignell, H.R., et al. (2008). Accurate whole human genome sequencing using reversible terminator chemistry. Nature 456, 5359. Berger, M.F., Lawrence, M.S., Demichelis, F., Drier, Y., Cibulskis, K., Sivachenko, A.Y., Sboner, A., Esgueva, R., Pueger, D., Sougnez, C., et al. (2011). The genomic complexity of primary human prostate cancer. Nature 470, 214220. Berger, M.F., Hodis, E., Heffernan, T.P., Deribe, Y.L., Lawrence, M.S., Protopopov, A., Ivanova, E., Watson, I.R., Nickerson, E., Ghosh, P., et al. (2012). Melanoma genome sequencing reveals frequent PREX2 mutations. Nature 485, 502506. Bernt, K.M., Zhu, N., Sinha, A.U., Vempati, S., Faber, J., Krivtsov, A.V., Feng, Z., Punt, N., Daigle, A., Bullinger, L., et al. (2011). MLL-rearranged leukemia is dependent on aberrant H3K79 methylation by DOT1L. Cancer Cell 20, 6678. Beroukhim, R., Getz, G., Nghiemphu, L., Barretina, J., Hsueh, T., Linhart, D., Vivanco, I., Lee, J.C., Huang, J.H., Alexander, S., et al. (2007). Assessing the signicance of chromosomal aberrations in cancer: methodology and application to glioma. Proc. Natl. Acad. Sci. USA 104, 2000720012. Beroukhim, R., Mermel, C.H., Porter, D., Wei, G., Raychaudhuri, S., Donovan, J., Barretina, J., Boehm, J.S., Dobson, J., Urashima, M., et al. (2010). The landscape of somatic copy-number alteration across human cancers. Nature 463, 899905. Berry, D.A. (2011). Adaptive clinical trials in oncology. Nat. Rev. Clin. Oncol. 9, 199207. Biankin, A.V., Waddell, N., Kassahn, K.S., Gingras, M.C., Muthuswamy, L.B., Johns, A.L., Miller, D.K., Wilson, P.J., Patch, A.M., Wu, J., et al.; Australian Pancreatic Cancer Genome Initiative. (2012). Pancreatic cancer genomes reveal aberrations in axon guidance pathway genes. Nature 491, 399405. Boveri, T. (2008). Concerning the origin of malignant tumours by Theodor Boveri. Translated and annotated by Henry Harris. J. Cell Sci. 121(Suppl 1), 184. CGAN (Cancer Genome Atlas Network). (2012). Comprehensive molecular portraits of human breast tumours. Nature 490, 6170. CGARN (Cancer Genome Atlas Research Network). (2008). Comprehensive genomic characterization denes human glioblastoma genes and core pathways. Nature 455, 10611068. Carter, S.L., Cibulskis, K., Helman, E., McKenna, A., Shen, H., Zack, T., Laird, P.W., Onofrio, R.C., Winckler, W., Weir, B.A., et al. (2012). Absolute quantication of somatic DNA alterations in human cancer. Nat. Biotechnol. 30, 413421. Cazier, J.B., and Tomlinson, I. (2010). General lessons from large-scale studies to identify human cancer predisposition genes. J. Pathol. 220, 255262. Chapman, M.A., Lawrence, M.S., Keats, J.J., Cibulskis, K., Sougnez, C., Schinzel, A.C., Harview, C.L., Brunet, J.P., Ahmann, G.J., Adli, M., et al. (2011a). Initial genome sequencing and analysis of multiple myeloma. Nature 471, 467472. Chapman, P.B., Hauschild, A., Robert, C., Haanen, J.B., Ascierto, P., Larkin, J., Dummer, R., Garbe, C., Testori, A., Maio, M., et al.; BRIM-3 Study Group. (2011b). Improved survival with vemurafenib in melanoma with BRAF V600E mutation. N. Engl. J. Med. 364, 25072516.
Chen, K., Wallis, J.W., McLellan, M.D., Larson, D.E., Kalicki, J.M., Pohl, C.S., McGrath, S.D., Wendl, M.C., Zhang, Q., Locke, D.P., et al. (2009). BreakDancer: an algorithm for high-resolution mapping of genomic structural variation. Nat. Methods 6, 677681. Chin, K., DeVries, S., Fridlyand, J., Spellman, P.T., Roydasgupta, R., Kuo, W.L., Lapuk, A., Neve, R.M., Qian, Z., Ryder, T., et al. (2006). Genomic and transcriptional aberrations linked to breast cancer pathophysiologies. Cancer Cell 10, 529541. Chung, C.C., and Chanock, S.J. (2011). Current status of genome-wide association studies in cancer. Hum. Genet. 130, 5978. Cibulskis, K., Lawrence, M.S., Carter, S.L., Sivachenko, A., Jaffe, D., Sougnez, C., Gabriel, S., Meyerson, M., Lander, E.S., and Getz, G. (2013). Sensitive detection of somatic point mutations in impure and heterogeneous cancer samples. Nat. Biotechnol. Published online February 20, 2013. http://dx.doi. org/10.1038/nbt.2514. Crasta, K., Ganem, N.J., Dagher, R., Lantermann, A.B., Ivanova, E.V., Pan, Y., Nezi, L., Protopopov, A., Chowdhury, D., and Pellman, D. (2012). DNA breaks and chromosome pulverization from errors in mitosis. Nature 482, 5358. Dang, L., White, D.W., Gross, S., Bennett, B.D., Bittinger, M.A., Driggers, E.M., Fantin, V.R., Jang, H.G., Jin, S., Keenan, M.C., et al. (2009). Cancer-associated IDH1 mutations produce 2-hydroxyglutarate. Nature 462, 739744. Davies, H., Bignell, G.R., Cox, C., Stephens, P., Edkins, S., Clegg, S., Teague, J., Woffendin, H., Garnett, M.J., Bottomley, W., et al. (2002). Mutations of the BRAF gene in human cancer. Nature 417, 949954. Dawson, M.A., and Kouzarides, T. (2012). Cancer epigenetics: from mechanism to therapy. Cell 150, 1227. Dees, N.D., Zhang, Q., Kandoth, C., Wendl, M.C., Schierding, W., Koboldt, D.C., Mooney, T.B., Callaway, M.B., Dooling, D., Mardis, E.R., et al. (2012). MuSiC: identifying mutational signicance in cancer genomes. Genome Res. 22, 15891598. , A., Delhommeau, F., Dupont, S., Della Valle, V., James, C., Trannoy, S., Masse Kosmider, O., Le Couedic, J.P., Robert, F., Alberdi, A., et al. (2009). Mutation in TET2 in myeloid cancers. N. Engl. J. Med. 360, 22892301. Demetri, G.D., von Mehren, M., Blanke, C.D., Van den Abbeele, A.D., Eisenberg, B., Roberts, P.J., Heinrich, M.C., Tuveson, D.A., Singer, S., Janicek, M., et al. (2002). Efcacy and safety of imatinib mesylate in advanced gastrointestinal stromal tumors. N. Engl. J. Med. 347, 472480. Dias-Santagata, D., Akhavanfard, S., David, S.S., Vernovsky, K., Kuhlmann, G., Boisvert, S.L., Stubbs, H., McDermott, U., Settleman, J., Kwak, E.L., et al. (2010). Rapid targeted mutational analysis of human tumours: a clinical platform to guide personalized cancer medicine. EMBO Mol. Med. 2, 146158. Ding, L., Getz, G., Wheeler, D.A., Mardis, E.R., McLellan, M.D., Cibulskis, K., Sougnez, C., Greulich, H., Muzny, D.M., Morgan, M.B., et al. (2008). Somatic mutations affect key pathways in lung adenocarcinoma. Nature 455, 1069 1075. Ding, L., Ley, T.J., Larson, D.E., Miller, C.A., Koboldt, D.C., Welch, J.S., Ritchey, J.K., Young, M.A., Lamprecht, T., McLellan, M.D., et al. (2012). Clonal evolution in relapsed acute myeloid leukaemia revealed by whole-genome sequencing. Nature 481, 506510. Dolnik, A., Engelmann, J.C., Scharfenberger-Schmeer, M., Mauch, J., nke, J., Ku hn, M.W., Kelkenberg-Schade, S., Haldemann, B., Fries, T., Kro Paschka, P., et al. (2012). Commonly altered genomic regions in acute myeloid leukemia are enriched for somatic mutations involved in chromatin remodeling and splicing. Blood 120, e83e92. Druker, B.J., Talpaz, M., Resta, D.J., Peng, B., Buchdunger, E., Ford, J.M., Lydon, N.B., Kantarjian, H., Capdeville, R., Ohno-Jones, S., and Sawyers, C.L. (2001). Efcacy and safety of a specic inhibitor of the BCR-ABL tyrosine kinase in chronic myeloid leukemia. N. Engl. J. Med. 344, 10311037. Dulbecco, R. (1986). A turning point in cancer research: sequencing the human genome. Science 231, 10551056. Dupuis, J., Langenberg, C., Prokopenko, I., Saxena, R., Soranzo, N., Jackson, A.U., Wheeler, E., Glazer, N.L., Bouatia-Naji, N., Gloyn, A.L., et al.; DIAGRAM Consortium; GIANT Consortium; Global BPgen Consortium; Anders Hamsten
on behalf of Procardis Consortium; MAGIC investigators. (2010). New genetic loci implicated in fasting glucose homeostasis and their impact on type 2 diabetes risk. Nat. Genet. 42, 105116. Dutt, A., Salvesen, H.B., Chen, T.H., Ramos, A.H., Onofrio, R.C., Hatton, C., Nicoletti, R., Winckler, W., Grewal, R., Hanna, M., et al. (2008). Drug-sensitive FGFR2 mutations in endometrial carcinoma. Proc. Natl. Acad. Sci. USA 105, 87138717. Ellis, M.J., Ding, L., Shen, D., Luo, J., Suman, V.J., Wallis, J.W., Van Tine, B.A., Hoog, J., Goiffon, R.J., Goldstein, T.C., et al. (2012). Whole-genome analysis informs breast cancer response to aromatase inhibition. Nature 486, 353360. Emery, C.M., Vijayendran, K.G., Zipser, M.C., Sawyer, A.M., Niu, L., Kim, J.J., Hatton, C., Chopra, R., Oberholzer, P.A., Karpova, M.B., et al. (2009). MEK1 mutations confer resistance to MEK and B-RAF inhibition. Proc. Natl. Acad. Sci. USA 106, 2041120416. Estey, E. (2007). Acute myeloid leukemia and myelodysplastic syndromes in older patients. J. Clin. Oncol. 25, 19081915. Figueroa, M.E., Abdel-Wahab, O., Lu, C., Ward, P.S., Patel, J., Shih, A., Li, Y., Bhagwat, N., Vasanthakumar, A., Fernandez, H.F., et al. (2010). Leukemic IDH1 and IDH2 mutations result in a hypermethylation phenotype, disrupt TET2 function, and impair hematopoietic differentiation. Cancer Cell 18, 553567. Flaherty, K.T., Infante, J.R., Daud, A., Gonzalez, R., Kefford, R.F., Sosman, J., Hamid, O., Schuchter, L., Cebon, J., Ibrahim, N., et al. (2012a). Combined BRAF and MEK inhibition in melanoma with BRAF V600 mutations. N. Engl. J. Med. 367, 16941703. Flaherty, K.T., Robert, C., Hersey, P., Nathan, P., Garbe, C., Milhem, M., Demidov, L.V., Hassel, J.C., Rutkowski, P., Mohr, P., et al.; METRIC Study Group. (2012b). Improved survival with MEK inhibition in BRAF-mutated melanoma. N. Engl. J. Med. 367, 107114. Folco, E.G., Coil, K.E., and Reed, R. (2011). The anti-tumor drug E7107 reveals an essential role for SF3b in remodeling U2 snRNP to expose the branch pointbinding region. Genes Dev. 25, 440444. Fujimoto, A., Totoki, Y., Abe, T., Boroevich, K.A., Hosoda, F., Nguyen, H.H., Aoki, M., Hosono, N., Kubo, M., Miya, F., et al. (2012). Whole-genome sequencing of liver cancers identies etiological inuences on mutation patterns and recurrent mutations in chromatin regulators. Nat. Genet. 44, 760764. Gabor Miklos, G.L. (2005). The human cancer genome projectone more misstep in the war on cancer. Nat. Biotechnol. 23, 535537. Garnett, M.J., Edelman, E.J., Heidorn, S.J., Greenman, C.D., Dastur, A., Lau, K.W., Greninger, P., Thompson, I.R., Luo, X., Soares, J., et al. (2012). Systematic identication of genomic markers of drug sensitivity in cancer cells. Nature 483, 570575. nne, P.A. (2012). Circumventing cancer drug resistance Garraway, L.A., and Ja in the era of personalized medicine. Cancer Discov. 2, 214226. Garraway, L.A., Widlund, H.R., Rubin, M.A., Getz, G., Berger, A.J., Ramaswamy, S., Beroukhim, R., Milner, D.A., Granter, S.R., Du, J., et al. (2005). Integrative genomic analyses identify MITF as a lineage survival oncogene amplied in malignant melanoma. Nature 436, 117122. Gerlinger, M., Rowan, A.J., Horswell, S., Larkin, J., Endesfelder, D., Gronroos, E., Martinez, P., Matthews, N., Stewart, A., Tarpey, P., et al. (2012). Intratumor heterogeneity and branched evolution revealed by multiregion sequencing. N. Engl. J. Med. 366, 883892. Gnirke, A., Melnikov, A., Maguire, J., Rogov, P., LeProust, E.M., Brockman, W., Fennell, T., Giannoukos, G., Fisher, S., Russ, C., et al. (2009). Solution hybrid selection with ultra-long oligonucleotides for massively parallel targeted sequencing. Nat. Biotechnol. 27, 182189. Grabherr, M.G., Haas, B.J., Yassour, M., Levin, J.Z., Thompson, D.A., Amit, I., Adiconis, X., Fan, L., Raychowdhury, R., Zeng, Q., et al. (2011). Full-length transcriptome assembly from RNA-Seq data without a reference genome. Nat. Biotechnol. 29, 644652. Grasso, C.S., Wu, Y.M., Robinson, D.R., Cao, X., Dhanasekaran, S.M., Khan, A.P., Quist, M.J., Jing, X., Lonigro, R.J., Brenner, J.C., et al. (2012). The muta-
tional landscape of lethal castration-resistant prostate cancer. Nature 487, 239243. Greenman, C., Stephens, P., Smith, R., Dalgliesh, G.L., Hunter, C., Bignell, G., Davies, H., Teague, J., Butler, A., Stevens, C., et al. (2007). Patterns of somatic mutation in human cancer genomes. Nature 446, 153158. Gui, Y., Guo, G., Huang, Y., Hu, X., Tang, A., Gao, S., Wu, R., Chen, C., Li, X., Zhou, L., et al. (2011). Frequent mutations of chromatin remodeling genes in transitional cell carcinoma of the bladder. Nat. Genet. 43, 875878. Haiman, C.A., Chen, G.K., Blot, W.J., Strom, S.S., Berndt, S.I., Kittles, R.A., Rybicki, B.A., Isaacs, W.B., Ingles, S.A., Stanford, J.L., et al. (2011). Characterizing genetic risk at known prostate cancer susceptibility loci in African Americans. PLoS Genet. 7, e1001387. ka, A., jkova , H., Markova , J., Ha rova , I., Fuchs, O., Kostec Ha skovec, C., Sa , K., and Schwarz, J. (2012). Decreased DNA meth , P., Michalova Cetkovsky ylation in acute myeloid leukemia patients with DNMT3A mutations and prognostic implications of DNA methylation. Leuk. Res. 36, 11281133. Hammerman, P.S., Hayes, D.N., Wilkerson, M.D., Schultz, N., Bose, R., Chu, A., Collisson, E.A., Cope, L., Creighton, C.J., Getz, G., et al.; Cancer Genome Atlas Research Network. (2012). Comprehensive genomic characterization of squamous cell lung cancers. Nature 489, 519525. Hanahan, D., and Weinberg, R.A. (2000). The hallmarks of cancer. Cell 100, 5770. Hanahan, D., and Weinberg, R.A. (2011). Hallmarks of cancer: the next generation. Cell 144, 646674. Hindorff, L.A., Gillanders, E.M., and Manolio, T.A. (2011). Genetic architecture of cancer and other complex diseases: lessons learned and future directions. Carcinogenesis 32, 945954. Ho, L., and Crabtree, G.R. (2010). Chromatin remodelling during development. Nature 463, 474484. Hodges, E., Xuan, Z., Balija, V., Kramer, M., Molla, M.N., Smith, S.W., Middle, C.M., Rodesch, M.J., Albert, T.J., Hannon, G.J., and McCombie, W.R. (2007). Genome-wide in situ exon capture for selective resequencing. Nat. Genet. 39, 15221527. Hodis, E., Watson, I.R., Kryukov, G.V., Arold, S.T., Imielinski, M., Theurillat, J.P., Nickerson, E., Auclair, D., Li, L., Place, C., et al. (2012). A landscape of driver mutations in melanoma. Cell 150, 251263. Holland, A.J., and Cleveland, D.W. (2012). Chromoanagenesis and cancer: mechanisms and consequences of localized, complex chromosomal rearrangements. Nat. Med. 18, 16301638. Hou, Y., Song, L., Zhu, P., Zhang, B., Tao, Y., Xu, X., Li, F., Wu, K., Liang, J., Shao, D., et al. (2012). Single-cell exome sequencing and monoclonal evolution of a JAK2-negative myeloproliferative neoplasm. Cell 148, 873885. Huang, J., Deng, Q., Wang, Q., Li, K.Y., Dai, J.H., Li, N., Zhu, Z.D., Zhou, B., Liu, X.Y., Liu, R.F., et al. (2012a). Exome sequencing of hepatitis B virus-associated hepatocellular carcinoma. Nat. Genet. 44, 11171121. Huang, S., Gulzar, Z.G., Salari, K., Lapointe, J., Brooks, J.D., and Pollack, J.R. (2012b). Recurrent deletion of CHD1 in prostate cancer with relevance to cell invasiveness. Oncogene 31, 41644170. Huang, F.W., Hodis, E., Xu, M.J., Kryukov, G.V., Chin, L., and Garraway, L.A. (2013). Highly recurrent TERT promoter mutations in human melanoma. Science 339, 957959. Huch, M., Dorrell, C., Boj, S.F., van Es, J.H., Li, V.S., van de Wetering, M., Sato, T., Hamer, K., Sasaki, N., Finegold, M.J., et al. (2013). In vitro expansion of single Lgr5+ liver stem cells induced by Wnt-driven regeneration. Nature 494, 247250. , R.R., Hudson, T.J., Anderson, W., Artez, A., Barker, A.D., Bell, C., Bernabe Bhan, M.K., Calvo, F., Eerola, I., Gerhard, D.S., et al.; International Cancer Genome Consortium. (2010). International network of cancer genome projects. Nature 464, 993998. Imielinski, M., Berger, A.H., Hammerman, P.S., Hernandez, B., Pugh, T.J., Hodis, E., Cho, J., Suh, J., Capelletti, M., Sivachenko, A., et al. (2012). Mapping the hallmarks of lung adenocarcinoma with massively parallel sequencing. Cell 150, 11071120.
International Human Genome Sequencing Consortium. (2004). Finishing the euchromatic sequence of the human genome. Nature 431, 931945. Iyer, G., Hanrahan, A.J., Milowsky, M.I., Al-Ahmadie, H., Scott, S.N., Janakiraman, M., Pirun, M., Sander, C., Socci, N.D., Ostrovnaya, I., et al. (2012). Genome sequencing identies a basis for everolimus sensitivity. Science 338, 221. Jaju, R.J., Fidler, C., Haas, O.A., Strickson, A.J., Watkins, F., Clark, K., Cross, N.C., Cheng, J.F., Aplan, P.D., Kearney, L., et al. (2001). A novel gene, NSD1, is fused to NUP98 in the t(5;11)(q35;p15.5) in de novo childhood acute myeloid leukemia. Blood 98, 12641267. Jankowska, A.M., Makishima, H., Tiu, R.V., Szpurka, H., Huang, Y., Traina, F., Visconte, V., Sugimoto, Y., Prince, C., OKeefe, C., et al. (2011). Mutational spectrum analysis of chronic myelomonocytic leukemia includes genes associated with epigenetic regulation: UTX, EZH2, and DNMT3A. Blood 118, 3932 3941. Johannessen, C.M., Boehm, J.S., Kim, S.Y., Thomas, S.R., Wardwell, L., Johnson, L.A., Emery, C.M., Stransky, N., Cogdill, A.P., Barretina, J., et al. (2010). COT drives resistance to RAF inhibition through MAP kinase pathway reactivation. Nature 468, 968972. Jones, S., Wang, T.L., Shih, IeM., Mao, T.L., Nakayama, K., Roden, R., Glas, R., Slamon, D., Diaz, L.A., Jr., Vogelstein, B., et al. (2010). Frequent mutations of chromatin remodeling gene ARID1A in ovarian clear cell carcinoma. Science 330, 228231. Kaida, D., Motoyoshi, H., Tashiro, E., Nojima, T., Hagiwara, M., Ishigami, K., Watanabe, H., Kitahara, T., Yoshida, T., Nakajima, H., et al. (2007). Spliceostatin A targets SF3b and inhibits both splicing and nuclear retention of premRNA. Nat. Chem. Biol. 3, 576583. Kan, Z., Jaiswal, B.S., Stinson, J., Janakiraman, V., Bhatt, D., Stern, H.M., Yue, P., Haverty, P.M., Bourgon, R., Zheng, J., et al. (2010). Diverse somatic mutation patterns and pathway alterations in human cancers. Nature 466, 869873. Karlsson, E.K., and Lindblad-Toh, K. (2008). Leader of the pack: gene mapping in dogs and other model organisms. Nat. Rev. Genet. 9, 713725. Kim, D., and Salzberg, S.L. (2011). TopHat-Fusion: an algorithm for discovery of novel fusion transcripts. Genome Biol. 12, R72. Kim, Y.H., Kwei, K.A., Girard, L., Salari, K., Kao, J., Pacyna-Gengelbach, M., Wang, P., Hernandez-Boussard, T., Gazdar, A.F., Petersen, I., et al. (2010). Genomic and functional analysis identies CRKL as an oncogene amplied in lung cancer. Oncogene 29, 14211430. Kim, E.S., Herbst, R.S., Wistuba, I.I., Lee, J.J., Blumenschein, G.R., Jr., Tsao, A., Stewart, D.J., Hicks, M.E., Erasmus, J., Jr., Gupta, S., et al. (2011). The BATTLE trial: personalizing therapy for lung cancer. Cancer Discov. 1, 4453. Klinakis, A., Lobry, C., Abdel-Wahab, O., Oh, P., Haeno, H., Buonamici, S., van De Walle, I., Cathelin, S., Trimarchi, T., Araldi, E., et al. (2011). A novel tumoursuppressor function for the Notch pathway in myeloid leukaemia. Nature 473, 230233. Koivunen, P., Lee, S., Duncan, C.G., Lopez, G., Lu, G., Ramkissoon, S., Losman, J.A., Joensuu, P., Bergmann, U., Gross, S., et al. (2012). Transformation by the (R)-enantiomer of 2-hydroxyglutarate linked to EGLN activation. Nature 483, 484488. Kostic, A.D., Ojesina, A.I., Pedamallu, C.S., Jung, J., Verhaak, R.G., Getz, G., and Meyerson, M. (2011). PathSeq: software to identify or discover microbes by deep sequencing of human tissue. Nat. Biotechnol. 29, 393396. Kralovics, R., Passamonti, F., Buser, A.S., Teo, S.S., Tiedt, R., Passweg, J.R., Tichelli, A., Cazzola, M., and Skoda, R.C. (2005). A gain-of-function mutation of JAK2 in myeloproliferative disorders. N. Engl. J. Med. 352, 17791790. Krauthammer, M., Kong, Y., Ha, B.H., Evans, P., Bacchiocchi, A., McCusker, J.P., Cheng, E., Davis, M.J., Goh, G., Choi, M., et al. (2012). Exome sequencing identies recurrent somatic RAC1 mutations in melanoma. Nat. Genet. 44, 10061014. Kreso, A., OBrien, C.A., van Galen, P., Gan, O.I., Notta, F., Brown, A.M., Ng, K., Ma, J., Wienholds, E., Dunant, C., et al. (2013). Variable clonal repopulation dynamics inuence chemotherapy response in colorectal cancer. Science 339, 543548.
Kriaucionis, S., and Heintz, N. (2009). The nuclear DNA base 5-hydroxymethylcytosine is present in Purkinje neurons and the brain. Science 324, 929930. Landau, D.A., Carter, S.L., Stojanov, P., McKenna, A., Stevenson, K., Lawrence, M.S., Sougnez, C., Stewart, C., Sivachenko, A., Wang, L., et al. (2013). Evolution and impact of subclonal mutations in chronic lymphocytic leukemia. Cell 152, 714726. Lander, E.S., Linton, L.M., Birren, B., Nusbaum, C., Zody, M.C., Baldwin, J., Devon, K., Dewar, K., Doyle, M., FitzHugh, W., et al.; International Human Genome Sequencing Consortium. (2001). Initial sequencing and analysis of the human genome. Nature 409, 860921. Langemeijer, S.M., Kuiper, R.P., Berends, M., Knops, R., Aslanyan, M.G., Massop, M., Stevens-Linders, E., van Hoogen, P., van Kessel, A.G., Raymakers, R.A., et al. (2009). Acquired mutations in TET2 are common in myelodysplastic syndromes. Nat. Genet. 41, 838842. Le Gallo, M., OHara, A.J., Rudd, M.L., Urick, M.E., Hansen, N.F., ONeil, N.J., Price, J.C., Zhang, S., England, B.M., Godwin, A.K., et al.; NIH Intramural Sequencing Center (NISC) Comparative Sequencing Program. (2012). Exome sequencing of serous endometrial tumors identies recurrent somatic mutations in chromatin-remodeling and ubiquitin ligase complex genes. Nat. Genet. 44, 13101315. Levine, R.L., Wadleigh, M., Cools, J., Ebert, B.L., Wernig, G., Huntly, B.J., Boggon, T.J., Wlodarska, I., Clark, J.J., Moore, S., et al. (2005). Activating mutation in the tyrosine kinase JAK2 in polycythemia vera, essential thrombocythemia, and myeloid metaplasia with myelobrosis. Cancer Cell 7, 387397. Ley, T.J., Mardis, E.R., Ding, L., Fulton, B., McLellan, M.D., Chen, K., Dooling, D., Dunford-Shore, B.H., McGrath, S., Hickenbotham, M., et al. (2008). DNA sequencing of a cytogenetically normal acute myeloid leukaemia genome. Nature 456, 6672. Ley, T.J., Ding, L., Walter, M.J., McLellan, M.D., Lamprecht, T., Larson, D.E., Kandoth, C., Payton, J.E., Baty, J., Welch, J., et al. (2010). DNMT3A mutations in acute myeloid leukemia. N. Engl. J. Med. 363, 24242433. Li, M., Zhao, H., Zhang, X., Wood, L.D., Anders, R.A., Choti, M.A., Pawlik, T.M., Daniel, H.D., Kannangai, R., Offerhaus, G.J., et al. (2011). Inactivating mutations of the chromatin remodeling gene ARID2 in hepatocellular carcinoma. Nat. Genet. 43, 828829. Lindsley, R.C., and Ebert, B.L. (2013). Molecular Pathophysiology of Myelodysplastic Syndromes. Annu. Rev. Pathol. 8, 2147. Published online August 28, 2012. http://dx.doi.org/10.1146/annurev-pathol-011811-132436. Lipson, D., Capelletti, M., Yelensky, R., Otto, G., Parker, A., Jarosz, M., Curran, J.A., Balasubramanian, S., Bloom, T., Brennan, K.W., et al. (2012). Identication of new ALK and RET gene fusions from colorectal and lung cancer biopsies. Nat. Med. 18, 382384. Liu, P., Erez, A., Nagamani, S.C., Dhar, S.U., Ko1odziejska, K.E., Dharmadhikari, A.V., Cooper, M.L., Wiszniewska, J., Zhang, F., Withers, M.A., et al. (2011). Chromosome catastrophes involve replication mechanisms generating complex genomic rearrangements. Cell 146, 889903. Liu, X., Ory, V., Chapman, S., Yuan, H., Albanese, C., Kallakury, B., Timofeeva, O.A., Nealon, C., Dakic, A., Simic, V., et al. (2012). ROCK inhibitor and feeder cells induce the conditional reprogramming of epithelial cells. Am. J. Pathol. 180, 599607. Lohr, J.G., Stojanov, P., Lawrence, M.S., Auclair, D., Chapuy, B., Sougnez, C., Cruz-Gordillo, P., Knoechel, B., Asmann, Y.W., Slager, S.L., et al. (2012). Discovery and prioritization of somatic mutations in diffuse large B-cell lymphoma (DLBCL) by whole-exome sequencing. Proc. Natl. Acad. Sci. USA 109, 38793884. Lu, C., Ward, P.S., Kapoor, G.S., Rohle, D., Turcan, S., Abdel-Wahab, O., Edwards, C.R., Khanin, R., Figueroa, M.E., Melnick, A., et al. (2012). IDH mutation impairs histone demethylation and results in a block to cell differentiation. Nature 483, 474478. Lynch, T.J., Bell, D.W., Sordella, R., Gurubhagavatula, S., Okimoto, R.A., Brannigan, B.W., Harris, P.L., Haserlat, S.M., Supko, J.G., Haluska, F.G., et al. (2004). Activating mutations in the epidermal growth factor receptor
underlying responsiveness of non-small-cell lung cancer to getinib. N. Engl. J. Med. 350, 21292139. Macconaill, L.E., and Garraway, L.A. (2010). Clinical implications of the cancer genome. J. Clin. Oncol. 28, 52195228. MacConaill, L.E., Campbell, C.D., Kehoe, S.M., Bass, A.J., Hatton, C., Niu, L., Davis, M., Yao, K., Hanna, M., Mondal, C., et al. (2009). Proling critical cancer gene mutations in clinical tumor samples. PLoS ONE 4, e7887. Makishima, H., Jankowska, A.M., Tiu, R.V., Szpurka, H., Sugimoto, Y., Hu, Z., Saunthararajah, Y., Guinta, K., Keddache, M.A., Putnam, P., et al. (2010). Novel homo- and hemizygous mutations in EZH2 in myeloid malignancies. Leukemia: ofcial journal of the Leukemia Society of America. Leukemia 24, 17991804. zek, K., Marcucci, G., Metzeler, K.H., Schwind, S., Becker, H., Maharry, K., Mro Radmacher, M.D., Kohlschmidt, J., Nicolet, D., Whitman, S.P., et al. (2012). Age-related prognostic impact of different types of DNMT3A mutations in adults with primary cytogenetically normal acute myeloid leukemia. J. Clin. Oncol. 30, 742750. Mardis, E.R., Ding, L., Dooling, D.J., Larson, D.E., McLellan, M.D., Chen, K., Koboldt, D.C., Fulton, R.S., Delehaunty, K.D., McGrath, S.D., et al. (2009). Recurring mutations found by sequencing an acute myeloid leukemia genome. N. Engl. J. Med. 361, 10581066. Margulies, M., Egholm, M., Altman, W.E., Attiya, S., Bader, J.S., Bemben, L.A., Berka, J., Braverman, M.S., Chen, Y.J., Chen, Z., et al. (2005). Genome sequencing in microfabricated high-density picolitre reactors. Nature 437, 376380. Meyerson, M., Gabriel, S., and Getz, G. (2010). Advances in understanding cancer genomes through second-generation sequencing. Nat. Rev. Genet. 11, 685696. Michor, F., Hughes, T.P., Iwasa, Y., Branford, S., Shah, N.P., Sawyers, C.L., and Nowak, M.A. (2005). Dynamics of chronic myeloid leukaemia. Nature 435, 12671270. Molenaar, J.J., Koster, J., Zwijnenburg, D.A., van Sluis, P., Valentijn, L.J., van der Ploeg, I., Hamdi, M., van Nes, J., Westerman, B.A., van Arkel, J., et al. (2012). Sequencing of neuroblastoma identies chromothripsis and defects in neuritogenesis genes. Nature 483, 589593. Morin, R.D., Johnson, N.A., Severson, T.M., Mungall, A.J., An, J., Goya, R., Paul, J.E., Boyle, M., Woolcock, B.W., Kuchenbauer, F., et al. (2010). Somatic mutations altering EZH2 (Tyr641) in follicular and diffuse large B-cell lymphomas of germinal-center origin. Nat. Genet. 42, 181185. Morin, R.D., Mendez-Lago, M., Mungall, A.J., Goya, R., Mungall, K.L., Corbett, R.D., Johnson, N.A., Severson, T.M., Chiu, R., Field, M., et al. (2011). Frequent mutation of histone-modifying genes in non-Hodgkin lymphoma. Nature 476, 298303. Murphy, A.B., Ukoli, F., Freeman, V., Bennett, F., Aiken, W., Tulloch, T., Coard, K., Angwafo, F., and Kittles, R.A. (2012). 8q24 risk alleles in West African and Caribbean men. Prostate 72, 13661373. Navin, N., Kendall, J., Troge, J., Andrews, P., Rodgers, L., McIndoo, J., Cook, K., Stepansky, A., Levy, D., Esposito, D., et al. (2011). Tumour evolution inferred by single-cell sequencing. Nature 472, 9094. Nazarian, R., Shi, H., Wang, Q., Kong, X., Koya, R.C., Lee, H., Chen, Z., Lee, M.K., Attar, N., Sazegar, H., et al. (2010). Melanomas acquire resistance to B-RAF(V600E) inhibition by RTK or N-RAS upregulation. Nature 468, 973977. Ngo, V.N., Young, R.M., Schmitz, R., Jhavar, S., Xiao, W., Lim, K.H., Kohlhammer, H., Xu, W., Yang, Y., Zhao, H., et al. (2011). Oncogenically active MYD88 mutations in human lymphoma. Nature 470, 115119. Nik-Zainal, S., Alexandrov, L.B., Wedge, D.C., Van Loo, P., Greenman, C.D., Raine, K., Jones, D., Hinton, J., Marshall, J., Stebbings, L.A., et al.; Breast Cancer Working Group of the International Cancer Genome Consortium. (2012a). Mutational processes molding the genomes of 21 breast cancers. Cell 149, 979993. Nik-Zainal, S., Van Loo, P., Wedge, D.C., Alexandrov, L.B., Greenman, C.D., Lau, K.W., Raine, K., Jones, D., Marshall, J., Ramakrishna, M., et al.; Breast
Cancer Working Group of the International Cancer Genome Consortium. (2012b). The life history of 21 breast cancers. Cell 149, 9941007. Noushmehr, H., Weisenberger, D.J., Diefes, K., Phillips, H.S., Pujara, K., Berman, B.P., Pan, F., Pelloski, C.E., Sulman, E.P., Bhat, K.P., et al.; Cancer Genome Atlas Research Network. (2010). Identication of a CpG island methylator phenotype that denes a distinct subgroup of glioma. Cancer Cell 17, 510522. Nowell, P.C. (1976). The clonal evolution of tumor cell populations. Science 194, 2328. Ogino, S., Cantor, M., Kawasaki, T., Brahmandam, M., Kirkner, G.J., Weisenberger, D.J., Campan, M., Laird, P.W., Loda, M., and Fuchs, C.S. (2006). CpG island methylator phenotype (CIMP) of colorectal cancer is best characterised by quantitative DNA methylation analysis and prospective cohort studies. Gut 55, 10001006. nne, P.A., Lee, J.C., Tracy, S., Greulich, H., Gabriel, S., Herman, Paez, J.G., Ja P., Kaye, F.J., Lindeman, N., Boggon, T.J., et al. (2004). EGFR mutations in lung cancer: correlation with clinical response to getinib therapy. Science 304, 14971500. Pal, A., Barber, T.M., Van de Bunt, M., Rudge, S.A., Zhang, Q., Lachlan, K.L., Cooper, N.S., Linden, H., Levy, J.C., Wakelam, M.J., et al. (2012). PTEN mutations as a cause of constitutive insulin sensitivity and obesity. N. Engl. J. Med. 367, 10021011. Palanisamy, N., Ateeq, B., Kalyana-Sundaram, S., Pueger, D., Ramnarayanan, K., Shankar, S., Han, B., Cao, Q., Cao, X., Suleman, K., et al. (2010). Rearrangements of the RAF kinase pathway in prostate cancer, gastric cancer and melanoma. Nat. Med. 16, 793798. Pao, W., Miller, V., Zakowski, M., Doherty, J., Politi, K., Sarkaria, I., Singh, B., Heelan, R., Rusch, V., Fulton, L., et al. (2004). EGF receptor gene mutations are common in lung cancers from never smokers and are associated with sensitivity of tumors to getinib and erlotinib. Proc. Natl. Acad. Sci. USA 101, 1330613311. Papaemmanuil, E., Cazzola, M., Boultwood, J., Malcovati, L., Vyas, P., Bowen, D., Pellagatti, A., Wainscoat, J.S., Hellstrom-Lindberg, E., Gambacorti-Passerini, C., et al.; Chronic Myeloid Disorders Working Group of the International Cancer Genome Consortium. (2011). Somatic SF3B1 mutation in myelodysplasia with ring sideroblasts. N. Engl. J. Med. 365, 13841395. Parsons, D.W., Jones, S., Zhang, X., Lin, J.C., Leary, R.J., Angenendt, P., Mankoo, P., Carter, H., Siu, I.M., Gallia, G.L., et al. (2008). An integrated genomic analysis of human glioblastoma multiforme. Science 321, 1807 1812. Pasqualucci, L., Dominguez-Sola, D., Chiarenza, A., Fabbri, G., Grunn, A., Trifonov, V., Kasper, L.H., Lerach, S., Tang, H., Ma, J., et al. (2011a). Inactivating mutations of acetyltransferase genes in B-cell lymphoma. Nature 471, 189195. Pasqualucci, L., Trifonov, V., Fabbri, G., Ma, J., Rossi, D., Chiarenza, A., Wells, V.A., Grunn, A., Messina, M., Elliot, O., et al. (2011b). Analysis of the coding genome of diffuse large B-cell lymphoma. Nat. Genet. 43, 830837. ndez-Cuesta, L., Sos, M.L., George, J., Seidel, D., Kasper, Peifer, M., Ferna L.H., Plenker, D., Leenders, F., Sun, R., Zander, T., et al. (2012). Integrative genome analyses identify key somatic driver mutations of small-cell lung cancer. Nat. Genet. 44, 11041110. Pleasance, E.D., Cheetham, R.K., Stephens, P.J., McBride, D.J., Humphray, n ez, G.R., Bignell, G.R., S.J., Greenman, C.D., Varela, I., Lin, M.L., Ordo et al. (2010a). A comprehensive catalogue of somatic mutations from a human cancer genome. Nature 463, 191196. Published online December 16, 2009. http://dx.doi.org/10.1038/nature08658. Pleasance, E.D., Stephens, P.J., OMeara, S., McBride, D.J., Meynert, A., Jones, D., Lin, M.L., Beare, D., Lau, K.W., Greenman, C., et al. (2010b). A small-cell lung cancer genome with complex signatures of tobacco exposure. Nature 463, 184190. Published online December 16, 2009. http://dx.doi.org/ 10.1038/nature08629. Pollock, P.M., Gartside, M.G., Dejeza, L.C., Powell, M.A., Mallon, M.A., Davies, H., Mohammadi, M., Futreal, P.A., Stratton, M.R., Trent, J.M., and
Goodfellow, P.J. (2007). Frequent activating FGFR2 mutations in endometrial carcinomas parallel germline mutations associated with craniosynostosis and skeletal dysplasia syndromes. Oncogene 26, 71587162. Pomerantz, M.M., and Freedman, M.L. (2011). The genetics of cancer risk. Cancer J. 17, 416422. Pomerantz, M.M., Ahmadiyeh, N., Jia, L., Herman, P., Verzi, M.P., Doddapaneni, H., Beckwith, C.A., Chan, J.A., Hills, A., Davis, M., et al. (2009). The 8q24 cancer risk variant rs6983267 shows long-range interaction with MYC in colorectal cancer. Nat. Genet. 41, 882884. Poulikakos, P.I., Persaud, Y., Janakiraman, M., Kong, X., Ng, C., Moriceau, G., Shi, H., Ate, M., Titz, B., Gabay, M.T., et al. (2011). RAF inhibitor resistance is mediated by dimerization of aberrantly spliced BRAF(V600E). Nature 480, 387390. n ez, G.R., Villamor, N., Puente, X.S., Pinyol, M., Quesada, V., Conde, L., Ordo ` , S., Gonza lez-D az, M., et al. (2011). WholeEscaramis, G., Jares, P., Bea genome sequencing identies recurrent mutations in chronic lymphocytic leukaemia. Nature 475, 101105. n ez, G.R., Jares, P., Bassaganyas, Quesada, V., Conde, L., Villamor, N., Ordo ` , S., Pinyol, M., Mart nez-Trillos, A., et al. (2012). Exome L., Ramsay, A.J., Bea sequencing identies recurrent mutations of the splicing factor SF3B1 gene in chronic lymphocytic leukemia. Nat. Genet. 44, 4752. tz, A.M., Zichner, T., Weischenfeldt, Rausch, T., Jones, D.T., Zapatka, M., Stu ger, N., Remke, M., Shih, D., Northcott, P.A., et al. (2012). Genome J., Ja sequencing of pediatric medulloblastoma links catastrophic DNA rearrangements with TP53 mutations. Cell 148, 5971. Robinson, D.R., Kalyana-Sundaram, S., Wu, Y.M., Shankar, S., Cao, X., Ateeq, B., Asangani, I.A., Iyer, M., Maher, C.A., Grasso, C.S., et al. (2011). Functionally recurrent rearrangements of the MAST kinase and Notch gene families in breast cancer. Nat. Med. 17, 16461651. Robinson, G., Parker, M., Kranenburg, T.A., Lu, C., Chen, X., Ding, L., Phoenix, T.N., Hedlund, E., Wei, L., Zhu, X., et al. (2012). Novel mutations target distinct subgroups of medulloblastoma. Nature 488, 4348. Rosati, R., La Starza, R., Veronese, A., Aventin, A., Schwienbacher, C., Vallespi, T., Negrini, M., Martelli, M.F., and Mecucci, C. (2002). NUP98 is fused to the NSD3 gene in acute myeloid leukemia associated with t(8;11)(p11.2;p15). Blood 99, 38573860. Salari, K., Spulak, M.E., Cuff, J., Forster, A.D., Giacomini, C.P., Huang, S., Ko, M.E., Lin, A.Y., van de Rijn, M., and Pollack, J.R. (2012). CDX2 is an amplied lineage-survival oncogene in colorectal cancer. Proc. Natl. Acad. Sci. USA 109, E3196E3205. Samuels, Y., Wang, Z., Bardelli, A., Silliman, N., Ptak, J., Szabo, S., Yan, H., Gazdar, A., Powell, S.M., Riggins, G.J., et al. (2004). High frequency of mutations of the PIK3CA gene in human cancers. Science 304, 554. Schwartzentruber, J., Korshunov, A., Liu, X.Y., Jones, D.T., Pfaff, E., Jacob, K., njes, M., et al. (2012). Driver Sturm, D., Fontebasso, A.M., Quang, D.A., To mutations in histone H3.3 and chromatin remodelling genes in paediatric glioblastoma. Nature 482, 226231. Seshagiri, S., Stawiski, E.W., Durinck, S., Modrusan, Z., Storm, E.E., Conboy, C.B., Chaudhuri, S., Guan, Y., Janakiraman, V., Jaiswal, B.S., et al. (2012). Recurrent R-spondin fusions in colon cancer. Nature 488, 660664. Shibata, T., Ohta, T., Tong, K.I., Kokubu, A., Odogawa, R., Tsuta, K., Asamura, H., Yamamoto, M., and Hirohashi, S. (2008). Cancer related mutations in NRF2 impair its recognition by Keap1-Cul3 E3 ligase and promote malignancy. Proc. Natl. Acad. Sci. USA 105, 1356813573. blom, T., Jones, S., Wood, L.D., Parsons, D.W., Lin, J., Barber, T.D., Sjo Mandelker, D., Leary, R.J., Ptak, J., Silliman, N., et al. (2006). The consensus coding sequences of human breast and colorectal cancers. Science 314, 268274. Sosman, J.A., Kim, K.B., Schuchter, L., Gonzalez, R., Pavlick, A.C., Weber, J.S., McArthur, G.A., Hutson, T.E., Moschos, S.J., Flaherty, K.T., et al. (2012). Survival in BRAF V600-mutant advanced melanoma treated with vemurafenib. N. Engl. J. Med. 366, 707714.
Stehelin, D., Varmus, H.E., Bishop, J.M., and Vogt, P.K. (1976). DNA related to the transforming gene(s) of avian sarcoma viruses is present in normal avian DNA. Nature 260, 170173. Stephens, P.J., Greenman, C.D., Fu, B., Yang, F., Bignell, G.R., Mudie, L.J., Pleasance, E.D., Lau, K.W., Beare, D., Stebbings, L.A., et al. (2011). Massive genomic rearrangement acquired in a single catastrophic event during cancer development. Cell 144, 2740. Stephens, P.J., Tarpey, P.S., Davies, H., Van Loo, P., Greenman, C., Wedge, D.C., Nik-Zainal, S., Martin, S., Varela, I., Bignell, G.R., et al.; Oslo Breast Cancer Consortium (OSBREAC). (2012). The landscape of cancer genes and mutational processes in breast cancer. Nature 486, 400404. Stransky, N., Egloff, A.M., Tward, A.D., Kostic, A.D., Cibulskis, K., Sivachenko, A., Kryukov, G.V., Lawrence, M.S., Sougnez, C., McKenna, A., et al. (2011). The mutational landscape of head and neck squamous cell carcinoma. Science 333, 11571160. Stratton, M.R., Campbell, P.J., and Futreal, P.A. (2009). The cancer genome. Nature 458, 719724. Straussman, R., Morikawa, T., Shee, K., Barzily-Rokni, M., Qian, Z.R., Du, J., Davis, A., Mongare, M.M., Gould, J., Frederick, D.T., et al. (2012). Tumour micro-environment elicits innate resistance to RAF inhibitors through HGF secretion. Nature 487, 500504. ha rautio, A., Yan, J., Turunen, M., Enge, M., Taipale, Sur, I.K., Hallikas, O., Va M., Karhu, A., Aaltonen, L.A., and Taipale, J. (2012). Mice lacking a Myc enhancer that includes human SNP rs6983267 are resistant to intestinal tumors. Science 338, 13601363. Tabin, C.J., Bradley, S.M., Bargmann, C.I., Weinberg, R.A., Papageorge, A.G., Scolnick, E.M., Dhar, R., Lowy, D.R., and Chang, E.H. (1982). Mechanism of activation of a human oncogene. Nature 300, 143149. Tahiliani, M., Koh, K.P., Shen, Y., Pastor, W.A., Bandukwala, H., Brudno, Y., Agarwal, S., Iyer, L.M., Liu, D.R., Aravind, L., et al. (2009). Conversion of 5-methylcytosine to 5-hydroxymethylcytosine in mammalian DNA by MLL partner TET1. Science 324, 930935. Taplin, M.E., Bubley, G.J., Shuster, T.D., Frantz, M.E., Spooner, A.E., Ogata, G.K., Keer, H.N., and Balk, S.P. (1995). Mutation of the androgen-receptor gene in metastatic androgen-independent prostate cancer. N. Engl. J. Med. 332, 13931398. Thomas, R.K., Baker, A.C., Debiasi, R.M., Winckler, W., Laframboise, T., Lin, W.M., Wang, M., Feng, W., Zander, T., MacConaill, L., et al. (2007). Highthroughput oncogene mutation proling in human cancer. Nat. Genet. 39, 347351. Toyota, M., Ahuja, N., Ohe-Toyota, M., Herman, J.G., Baylin, S.B., and Issa, J.P. (1999). CpG island methylator phenotype in colorectal cancer. Proc. Natl. Acad. Sci. USA 96, 86818686. Trapnell, C., Pachter, L., and Salzberg, S.L. (2009). TopHat: discovering splice junctions with RNA-Seq. Bioinformatics 25, 11051111. Turcan, S., Rohle, D., Goenka, A., Walsh, L.A., Fang, F., Yilmaz, E., Campos, C., Fabius, A.W., Lu, C., Ward, P.S., et al. (2012). IDH1 mutation is sufcient to establish the glioma hypermethylator phenotype. Nature 483, 479483. Tuupanen, S., Turunen, M., Lehtonen, R., Hallikas, O., Vanharanta, S., Kivioja, rklund, M., Wei, G., Yan, J., Niittyma ki, I., et al. (2009). The common T., Bjo colorectal cancer predisposition SNP rs6983267 at chromosome 8q24 confers potential to enhanced Wnt signaling. Nat. Genet. 41, 885890. Varela, I., Tarpey, P., Raine, K., Huang, D., Ong, C.K., Stephens, P., Davies, H., Jones, D., Lin, M.L., Teague, J., et al. (2011). Exome sequencing identies frequent mutation of the SWI/SNF complex gene PBRM1 in renal carcinoma. Nature 469, 539542. Venter, J.C., Adams, M.D., Myers, E.W., Li, P.W., Mural, R.J., Sutton, G.G., Smith, H.O., Yandell, M., Evans, C.A., Holt, R.A., et al. (2001). The sequence of the human genome. Science 291, 13041351. Wagle, N., Emery, C., Berger, M.F., Davis, M.J., Sawyer, A., Pochanard, P., Kehoe, S.M., Johannessen, C.M., Macconaill, L.E., Hahn, W.C., et al. (2011). Dissecting therapeutic resistance to RAF inhibition in melanoma by tumor genomic proling. J. Clin. Oncol. 29, 30853096.
Wagle, N., Berger, M.F., Davis, M.J., Blumenstiel, B., Defelice, M., Pochanard, P., Ducar, M., Van Hummelen, P., Macconaill, L.E., Hahn, W.C., et al. (2012). High-throughput detection of actionable genomic alterations in clinical tumor samples by targeted, massively parallel sequencing. Cancer Discov. 2, 8293. Walter, M.J., Ding, L., Shen, D., Shao, J., Grillot, M., McLellan, M., Fulton, R., Schmidt, H., Kalicki-Veizer, J., OLaughlin, M., et al. (2011). Recurrent DNMT3A mutations in patients with myelodysplastic syndromes. Leukemia: ofcial journal of the Leukemia Society of America. Leukemia 25, 11531158. Wang, L., Lawrence, M.S., Wan, Y., Stojanov, P., Sougnez, C., Stevenson, K., Werner, L., Sivachenko, A., DeLuca, D.S., Zhang, L., et al. (2011a). SF3B1 and other novel cancer genes in chronic lymphocytic leukemia. N. Engl. J. Med. 365, 24972506. Wang, N.J., Sanborn, Z., Arnett, K.L., Bayston, L.J., Liao, W., Proby, C.M., Leigh, I.M., Collisson, E.A., Gordon, P.B., Jakkula, L., et al. (2011b). Loss-offunction mutations in Notch receptors in cutaneous and lung squamous cell carcinoma. Proc. Natl. Acad. Sci. USA 108, 1776117766. Wang, X.S., Shankar, S., Dhanasekaran, S.M., Ateeq, B., Sasaki, A.T., Jing, X., Robinson, D., Cao, Q., Prensner, J.R., Yocum, A.K., et al. (2011c). Characterization of KRAS rearrangements in metastatic prostate cancer. Cancer Discov. 1, 3543. Ward, P.S., Patel, J., Wise, D.R., Abdel-Wahab, O., Bennett, B.D., Coller, H.A., Cross, J.R., Fantin, V.R., Hedvat, C.V., Perl, A.E., et al. (2010). The common feature of leukemia-associated IDH1 and IDH2 mutations is a neomorphic enzyme activity converting alpha-ketoglutarate to 2-hydroxyglutarate. Cancer Cell 17, 225234. Weinberg, R. (2010). Point: Hypotheses rst. Nature 464, 678. Weir, B.A., Woo, M.S., Getz, G., Perner, S., Ding, L., Beroukhim, R., Lin, W.M., Province, M.A., Kraja, A., Johnson, L.A., et al. (2007). Characterizing the cancer genome in lung adenocarcinoma. Nature 450, 893898. Weisenberger, D.J., Siegmund, K.D., Campan, M., Young, J., Long, T.I., Faasse, M.A., Kang, G.H., Widschwendter, M., Weener, D., Buchanan, D., et al. (2006). CpG island methylator phenotype underlies sporadic microsatellite instability and is tightly associated with BRAF mutation in colorectal cancer. Nat. Genet. 38, 787793. Weiss, J., Sos, M.L., Seidel, D., Peifer, M., Zander, T., Heuckmann, J.M., Ullrich, R.T., Menon, R., Maier, S., Soltermann, A., et al. (2010). Frequent and focal FGFR1 amplication associates with therapeutically tractable FGFR1 dependency in squamous cell lung cancer. Sci. Transl. Med. 2, 62ra93. Weng, A.P., Ferrando, A.A., Lee, W., Morris, J.P., 4th, Silverman, L.B., Sanchez-Irizarry, C., Blacklow, S.C., Look, A.T., and Aster, J.C. (2004). Acti-
vating mutations of NOTCH1 in human T cell acute lymphoblastic leukemia. Science 306, 269271. Whittaker, S.R., Theurillat, J.P., Van Allen, E., Wagle, N., Hsiao, J., Cowley, G.S., Schadendorf, D., Root, D.E., and Garraway, L.A. (2013). A genome-scale RNA interference screen implicates NF1 loss in resistance to RAF inhibition. Cancer Discov. Published online January 3, 2013. http://dx.doi.org/10.1158/ 2159-8290. Wiegand, K.C., Shah, S.P., Al-Agha, O.M., Zhao, Y., Tse, K., Zeng, T., Senz, J., McConechy, M.K., Anglesio, M.S., Kalloger, S.E., et al. (2010). ARID1A mutations in endometriosis-associated ovarian carcinomas. N. Engl. J. Med. 363, 15321543. Williams, S.E., Beronja, S., Pasolli, H.A., and Fuchs, E. (2011). Asymmetric cell divisions promote Notch-dependent epidermal differentiation. Nature 470, 353358. Wilson, B.G., and Roberts, C.W. (2011). SWI/SNF nucleosome remodellers and cancer. Nat. Rev. Cancer 11, 481492. blom, T., Leary, R.J., Shen, Wood, L.D., Parsons, D.W., Jones, S., Lin, J., Sjo D., Boca, S.M., Barber, T., Ptak, J., et al. (2007). The genomic landscapes of human breast and colorectal cancers. Science 318, 11081113. Xu, W., Yang, H., Liu, Y., Wang, P., Kim, S.H., Ito, S., Yang, C., Wang, P., Xiao, M.T., Liu, L.X., et al. (2011). Oncometabolite 2-hydroxyglutarate is a competitive inhibitor of a-ketoglutarate-dependent dioxygenases. Cancer Cell 19, 1730. Xu, X., Hou, Y., Yin, X., Bao, L., Tang, A., Song, L., Li, F., Tsang, S., Wu, K., Wu, H., et al. (2012). Single-cell exome sequencing reveals single-nucleotide mutation characteristics of a kidney tumor. Cell 148, 886895. Yan, H., Parsons, D.W., Jin, G., McLendon, R., Rasheed, B.A., Yuan, W., Kos, I., Batinic-Haberle, I., Jones, S., Riggins, G.J., et al. (2009). IDH1 and IDH2 mutations in gliomas. N. Engl. J. Med. 360, 765773. Yoshida, K., Sanada, M., Shiraishi, Y., Nowak, D., Nagata, Y., Yamamoto, R., Sato, Y., Sato-Otsubo, A., Kon, A., Nagasaki, M., et al. (2011). Frequent pathway mutations of splicing machinery in myelodysplasia. Nature 478, 6469. Zang, Z.J., Cutcutache, I., Poon, S.L., Zhang, S.L., McPherson, J.R., Tao, J., Rajasegaran, V., Heng, H.L., Deng, N., Gan, A., et al. (2012). Exome sequencing of gastric adenocarcinoma identies recurrent somatic mutations in cell adhesion and chromatin remodeling genes. Nat. Genet. 44, 570574. Zong, C., Lu, S., Chapman, A.R., and Xie, X.S. (2012). Genome-wide detection of single-nucleotide and copy-number variations of a single human cell. Science 338, 16221626.
Review
Interplay between the Cancer Genome and Epigenome
Hui Shen1 and Peter W. Laird1,*
1USC Epigenome Center, University of Southern California, Room G511B, 1450 Biggy Street, Los Angeles, CA 90089-9061, USA *Correspondence: plaird@usc.edu http://dx.doi.org/10.1016/j.cell.2013.03.008
Leading Edge
Cancer arises as a consequence of cumulative disruptions to cellular growth control with Darwinian selection for those heritable changes that provide the greatest clonal advantage. These traits can be acquired and stably maintained by either genetic or epigenetic means. Here, we explore the ways in which alterations in the genome and epigenome inuence each other and cooperate to promote oncogenic transformation. Disruption of epigenomic control is pervasive in malignancy and can be classied as an enabling characteristic of cancer cells, akin to genome instability and mutation.
Introduction Cancer develops through successive disruptions to the controls of cellular proliferation, immortality, angiogenesis, cell death, invasion, and metastasis. This evolutionary process requires new malignant traits to be stably encoded so that oncogenic events can accumulate in clonal lineages. Genetic mechanisms of mutation, copy number alteration, insertion, deletion, and recombination are particularly well suited as vehicles of persistent phenotypic change. For this reason, cancer has long been viewed as a disease based principally on genetics. Nevertheless, genetic events occur at low frequency and are thus not a particularly efcient means for malignant transformation. Some cancer cells overcome this bottleneck by acquiring DNA repair defects, thus boosting the mutation rate. Mechanisms of epigenetic control offer an alternative path to acquiring stable oncogenic traits. Epigenetic states are exible yet persist through multiple cell divisions and exert powerful effects on cellular phenotype. Although cancer cells have long been known to undergo epigenetic changes, genome-scale genomic and epigenomic analyses have only recently revealed the widespread occurrence of mutations in epigenetic regulators and the breadth of alterations to the epigenome in cancer cells (You and Jones, 2012). It is now clear that genetic and epigenetic mechanisms inuence each other and work cooperatively to enable the acquisition of the hallmarks of cancer (Hanahan and Weinberg, 2011). Shaping the Epigenome Epigenetic mechanisms allow genetically identical cells to achieve diverse stable phenotypes by controlling the transcriptional availability of various parts of the genome through differential chromatin marking and packaging. These embellishments include direct DNA modications, primarily CpG cytosine-5 methylation (Jones, 2012), but also hydroxylation, formylation, and carboxylation (Ito et al., 2011), as well as nucleosome occupancy and positioning (Gaffney et al., 2012; Valouev et al., 2011), histone variants, and dozens of different histone modications (Tan et al., 2011b), interacting proteins (Ram et al., 2011), and noncoding RNAs (Fabbri and Calin, 2010; Lee, 2012). These
epigenetic marks do not act in isolation but form a network of mutually reinforcing or counteracting signals. Genome-scale projects charting the human epigenome are rapidly extending our understanding of epigenetic marks and how they interact (Adams et al., 2012; Dunham et al., 2012; Ernst et al., 2011). A key facet of epigenetics is that these marks can be stably maintained yet adapt to changing developmental or environmental needs. This delicate task is accomplished by initiators, such as long noncoding RNAs, writers, which establish the epigenetic marks, readers, which interpret the epigenetic marks, erasers, which remove the epigenetic marks, remodelers, which can reposition nucleosomes, and insulators, which form boundaries between epigenetic domains. Epigenetic writers are directed to their target locations by sequence context, existing chromatin marks and bound proteins, noncoding RNAs, and/or nuclear architecture. Those marks are then recognized by reader proteins to convey information for various cellular functions. The establishment, maintenance, and change of epigenetic marks are intricately regulated, with crosstalk among the marks and writers to help guide changes to the epigenetic landscape. DNA Methylation De novo methylation of DNA is catalyzed by the enzymes DNMT3A and DNMT3B and is then maintained by the major DNA methyltransferase DNMT1, with participation from DNMT3A and DNMT3B (Jones and Liang, 2009). DNA methylation patterns are guided in part by primary DNA sequence context (Cedar and Bergman, 2012; Lienert et al., 2011) and are inuenced by germline variation (Gertz et al., 2011; Kerkel et al., 2008). Much of the mammalian genome consists of vast oceans of DNA sequence containing sparsely distributed but heavily methylated CpG dinucleotides, punctuated by short regions with unmethylated CpGs occurring at higher density, forming distinct islands in the genome (Bird et al., 1985). These CpG islands (CGIs) are protected from DNA methylation in part by guanine-cytosine (GC) strand asymmetry and accompanying R loop formation (Ginno et al., 2012) and possibly also by active demethylation mediated by the TET family members (Williams et al., 2012). The unmethylated state of CpG islands in the germline, along with biased gene
Figure 1. Representative Epigenetic States

Examples of representative epigenetic states are shown for several typical categories of genes and in different cellular contexts. (A) CpG-poor promoters are often tissue specic and/or reside in inducible genes that can be readily turned on or off. Transcription factor (TF) binding initiates nucleosome-depleted regions (NDR) at regulatory elements and at the promoter. (B) Many genes with CpG island promoters are constitutively expressed housekeeping genes. (C) Some genes with CpG island promoters, such as TF master regulators of differentiation and development, are repressed by the Polycomb complexes in stem cells and are kept in a bivalent state with both active and repressive marks. (D) Polycomb targets in stem cells are predisposed to cancer-associated promoter hypermethylation. Repressive marks are shown in red, and active marks are shown in blue.
conversion, helps to preserve CpG islands despite ongoing attrition of methylated CpG dinucleotides by cytosine deamination throughout most of the genome (Cohen et al., 2011). Transition zones between CpG islands and CpG oceans are called CpG shores and display more tissue-specic variation in DNA methylation (Irizarry et al., 2009). CpG islands span the transcription start sites of about half of the genes in the human genome, largely representing genes that are either actively expressed or poised for transcription (Figure 1). Methylated DNA is recognized by methyl-CpG binding domains (MBD) or C2H2 zinc ngers. The MBD-containing DNA methylation readers include MBD1, MBD2, MBD4, and MeCP2, whereas Kaiso (ZBTB33), ZBTB4, and ZBTB38 proteins use zinc ngers to bind methylated DNA. MBDs and Kaiso are believed to participate in DNA methylation-mediated transcrip-
tional repression of tumor suppressor genes with promoter DNA methylation. Histone Modications Posttranslational modications of histones are coordinated by counteracting histone methyltransferases (HMTs) and demethylases (e.g., KDMs), histone acetyltransferases (HATs) and deacetylases (HDACs), and writers and erasers of phosphorylation, as well as many other modications (Chi et al., 2010; Tan et al., 2011b). These histone modiers generally act in complexes, such as the repressive Polycomb (PcG) and activating Trithorax (TrxG) group complexes, which counterbalance each other in the regulation of genes important for development but which have also been implicated in cancer (Mills, 2010). Polycomb repressive complexes (PRCs) are guided to their targets in part by intrinsic signals in the genome sequence (Ku et al., 2008; Tanay
Figure 2. Histone H3 Lysine Writers, Erasers, and Readers

Although many other important histone modications also occur, only major histone H3 lysine modications (Ac: Acetylation; me1: monomethylation; me3: trimethylation) with well-dened functions are shown above a representative gene. The distribution of the marks is shown as colored bars and wedges to indicate approximate abundance. Repressive marks are shown in red, and active marks are shown in blue. Epigenetic regulators are listed to the right of each mark. Acetylation across different lysines shares writers and erasers, whereas methylation usually has dedicated enzymes. Readers (which can also be writers and erasers themselves) recognize different chromatin states and propagate the signal in various ways, including self-reinforcement or crosstalk, transcriptional activation or repression, or DNA repair. Crosstalk can also occur between histone modication and DNA methylation because DNMT3A, DNMT3L, and UHRF1 all contain reader domains for chromatin states.
et al., 2007). The histone H3K27me3 mark deposited by Polycomb repressive complex 2 (PRC2) provides docking sites for PRC1, whose enzymatic core unit RING1B monoubiquitinylates histone H2A at lysine 119 (H2AK119ub1), thereby blocking RNA polymerase II elongation. The Trithorax group complex, containing MLL, which lays down the H3K4 methylation mark, counteracts Polycomb function. The transcription factors encoding master regulators of differentiation and development are targeted by PRC2 in embryonic stem cells and are held in a bivalent chromatin state poised for transcription, with both the activating H3K4me3 and the repressive H3K27me3 (Bernstein et al., 2006) (Figure 1). During differentiation, the Trithorax demethylase KDM6A/UTX removes the repressive H3K27me3 mark, allowing transcription elongation to proceed for genes required in that particular lineage, whereas genes not required in that cell type lose the H3K4me3 active mark and undergo spreading of the
H3K27me3 repressive mark (Hawkins et al., 2010). Other histone marks have various readers with binding motifs, including bromodomain, PHD domain, chromodomain, and tudor domain (Musselman et al., 2012) (Figure 2). Trithorax and Polycomb complexes recruit HATs and HDACs, respectively, to counteract each other, and the establishment of histone acetylation can block Polycomb binding (Mills, 2010). Histone Variants Histone variants provide an additional layer of regulation. The main histone genes have multiple copies in the genome and are expressed during S phase. Single-copy variants are also expressed at other phases of the cell cycle and have distinct functions and/or locations. H2A has the largest number of variants, including H2A.Z, MacroH2A, H2A-Bbd, H2AvD, and H2A.X (Kamakaka and Biggins, 2005). The H3 variants include H3.3 and centromeric H3 (CenH3 or CENP-A), as well as a mammalian
testis-specic histone H3 variant called H3.4. Nucleosomes containing H3.3 and H2A.Z are located at dynamic regions requiring nucleosome mobility and exchange, such as at actively expressed gene promoters (Jin et al., 2009). Wide presence of H2A.Z in embryonic stem cells (Zhu et al., 2013) suggests prevalent chromatin exchange, which is consistent with the emerging idea that the genome of ESC is generally kept highly accessible. During differentiation, H2A.Z quickly redistributes. The mechanisms of recruitment have not been fully delineated, but various chromatin remodeler complexes and/or chaperones have been shown to be involved. For example, SRCAP is involved in H2A.Z loading into promoter/TSS, whereas H3.3 is loaded to telomeric/pericentric regions by the ATRX/DAXX complex and promoter/TSS by HIRA (Boyarchuk et al., 2011). Nucleosome Positioning and Remodeling The positioning of nucleosomes displays a weak 10 bp periodicity associated with minor sequence composition uctuations in phase with the DNA helical repeat. Some nucleosomes are more consistently positioned in phased arrays anchored by sequence-specic binding of proteins such as CTCF or adjacent to nucleosome-free regions at transcription start sites (Gaffney et al., 2012; Valouev et al., 2011). CpG islands have been associated with transcription-independent nucleosome depletion at mammalian promoters (Fenouil et al., 2012). ATP-dependent chromatin-remodeling complexes are responsible for sliding of the nucleosomes, as well as insertion and ejection of histone octamers, processes that are important for transcriptional repression and activation, and other important cellular functions such as DNA replication and repair. The remodeling complexes can be divided into four families: SWI/SNF, CHD (chromodomain and helicase-like domain), ISWI, and INO80 (including SWR1, or SRCAP in mammals). Insulators The CCCTC-binding factor CTCF and its paralog CTCFL/BORIS (expressed in the germline) are the only insulator proteins that have been identied so far in vertebrates. CTCF has a strong binding motif, and there is extensive overlap of the occupied CTCF-binding sites among different cell types (Kim et al., 2007). CTCF binds to enhancer blocking elements to prevent enhancer interactions with unintended promoters (enhancer blocking insulator) and also demarcates active and repressive chromatin domains (barrier insulator). Nuclear Architecture The genome can be compartmentalized based on nuclear architecture and associated genomic features into mostly heterochromatic late-replicating regions attached to the nuclear lamina at the nuclear periphery and more gene-rich early-replicating regions closer to the nuclear interior (Dunham et al., 2012; Meuleman et al., 2013). Lamina-associated sequences (LASs) enriched for a GAGA motif are bound by transcriptional repressors and appear to contribute to the establishment of lamina-associated domains (LADs) in the mammalian genome (Zullo et al., 2012). Maintaining the Epigenetic State The persistence of epigenetic traits in a growing tumor requires that the epigenome be faithfully copied during cell division. The chromatin structure is dismantled for passage of the replication fork (RF). Newly synthesized DNA and histone octamers are then assembled at the RF by chromatin assembly factor I (CAF1),
which is tethered to the RF by PCNA. Similarly, the dedicated maintenance DNA methyltransferase DNMT1 and the euchromatic H3K9 methyltransferase G9a, among other epigenetic maintainers, are loaded to RFs and copy the epigenetic marks. The Trithorax and Polycomb complexes are recruited prior to replication and are distributed evenly to the mother and daughter strands at the RF, and they restore the correct marks on the daughter molecules during G1 (Petruk et al., 2012). The histone marks are self-reinforcing and self-propagating, as PcG, SUV39H1/2, SETDB1, and TrxG all bind to the marks that they are responsible for catalyzing via an intrinsic reader domain or by interacting with a reader protein, thus helping to maintain the epigenetic state. Nucleosomes containing methylated DNA also stabilize DNMT3A/3B, which is a self-reinforcing mechanism for DNA methylation maintenance (Sharma et al., 2011). Disruption of Epigenetic Control in Cancer Most studies of cancer epigenetics have focused on DNA methylation as the epigenetic mark that most easily survives various forms of sample processing, including DNA extraction, and even formalin xation and parafn embedding (Laird, 2010). However, other epigenetic marks also undergo broad changes, including long noncoding RNAs and miRNAs (Baer et al., 2013; Baylin and Jones, 2011; Dawson and Kouzarides, 2012; Sandoval and Esteller, 2012) and histones, including loss of K16 acetylation and K20 trimethylation at histone H4 (Fraga et al., 2005; Hon et al., 2012; Kondo et al., 2008; Seligson et al., 2005; Yamazaki et al., 2013). Loss of 5-methylcytosine in cancer cells was discussed more than three decades ago (Ehrlich and Wang, 1981), with global DNA hypomethylation reported in cancer cell lines (Diala and Hoffman, 1982; Ehrlich et al., 1982) and reduced levels of DNA methylation found at selected genes in primary human tumors compared to normal tissues (Feinberg and Vogelstein, 1983). The widespread loss of DNA methylation contrasted starkly with the subsequent nding of hypermethylation of CpG islands in cancer (Baylin et al., 1986), including of promoter CpG islands of tumor-suppressor genes (Jones and Baylin, 2002). These seemingly contradictory ndings have been widely reported for many types of cancer (Baylin and Jones, 2011). The causal relevance of epigenetic changes in cancer was initially questioned, but this concern has now largely been laid to rest. First, many known tumor-suppressor genes have been shown to be silenced by promoter CpG island hypermethylation (Jones and Baylin, 2002). Importantly, the nding that these silencing events are mutually exclusive with structural or mutational inactivation of the same gene, such as the case for BRCA1 in ovarian cancer (Cancer Genome Atlas Research Network, 2011) and for CDKN2A in squamous cell lung cancer (Cancer Genome Atlas Research Network, 2012a), reinforces the concept that epigenetic silencing can serve as an alternative mechanism in Knudsons two-hit hypothesis (Jones and Laird, 1999). Second, mouse models of cancer have been shown to require epigenetic writers and readers for tumor development (Laird et al., 1995; Prokhortchouk et al., 2006; Sansom et al., 2003). Third, some DNA methylation changes appear to be essential for cancer cell survival, suggesting an acquired addiction to epigenetic alterations (De Carvalho et al., 2012). Finally, a plethora of signicantly mutated epigenetic regulators have
now been reported for many types of human cancer, as discussed further below. Long-Range Coordinated Disruptions and Nuclear Architecture The genome of undifferentiated embryonic stem cells is uniformly heavily methylated across CpG oceans, punctuated by unmethylated CpG islands. As stem cells differentiate and proliferate, the late-replicating lamin-associated domains (LADs) undergo progressive loss of DNA methylation within CpG oceans, and the LADs become recognizable as long partially methylated domains (PMDs), which become even more strikingly demarcated as hypomethylated domains in cancer cells (Berman et al., 2012; Hansen et al., 2011; Hon et al., 2012; Lister et al., 2009). This loss of DNA methylation is associated with an increase of repressive chromatin with large organized chromatin-lysine-(K) modication regions (LOCKs) (Hansen et al., 2011; Hon et al., 2012; Lister et al., 2009). CpG island hypermethylation is enriched in the hypomethylated domains, suggesting that these two events may be mechanistically linked but conned to distinct areas of the genome near the nuclear periphery (Berman et al., 2012). These long regions of DNA hypomethylation and repressive chromatin are consistent with prior reports of coordinated epigenetic silencing events located across megabase distances, a phenomenon termed long-range epigenetic silencing (LRES) (Clark, 2007; Coolen et al., 2010). It is noteworthy that the euchromatic part of the genome associated with the interior of the nucleus is generally much more epigenetically stable during cell differentiation, aging, and malignant transformation. However, loss of the DNA methyltransferase Dnmt3a can promote tumor progression with uniform hypomethylation across the genome and moderate deregulation of genes in euchromatic regions (Raddatz et al., 2012). Disruption of Differentiation and Development Differences between cell types are guided by the expression of tissue-specic transcription factors and consolidation of associated epigenetic states. Therefore, the epigenome of a cancer cell is determined in part by the cell of origin for that cancer and includes passenger hypermethylation events at genes not required in that particular lineage (Sproul et al., 2012). Epithelial to mesenchymal transition (EMT) of cancer cells is partly under reversible epigenetic control (De Craene and Berx, 2013). For example, primary breast tumors display heterogeneous and unstable silencing of the CDH1 (E-cadherin) gene, which facilitates the plasticity required during extravasation, metastasis, and establishment of a solid tumor at the metastatic site (Graff et al., 2000). It has long been debated whether cancer cells arise by dedifferentiation or instead originate from stem cells or early progenitors by a differentiation block. Polycomb repressors mark genes in stem cells encoding master regulators of differentiation and development, poised to either be turned on to coordinate differentiation of a lineage or to be fully repressed if it is not needed in that particular lineage (Bernstein et al., 2006). These genes occupied by Polycomb repressors in stem cells are particularly prone to acquiring CpG island hypermethylation during cell proliferation, aging, and particularly malignant transformation (Ohm et al., 2007; Schlesinger et al., 2007; Teschendorff et al., 2010; Widschwendter et al., 2007) (Figure 1). Although the genes affected by this process are primarily those not required or
expressed in that particular cell lineage, cancer cells do also show evidence of silencing of genes essential for differentiation of their cell of origin (Berman et al., 2012; Easwaran et al., 2012; Gal-Yam et al., 2008; Mohn et al., 2008; Teschendorff et al., 2010). This predisposition of Polycomb target genes to aberrant permanent epigenetic silencing is consistent with a model in which stem cells slowly acquire irreversible silencing of poised master regulators required for successful differentiation. As a consequence, some stem cells lose their ability to properly differentiate while retaining their self-renewal capabilities and become attractive candidates for malignant transformation by subsequent genetic and epigenetic events. One provocative implication of this model is that the rst steps of oncogenesis may in some cases be an epigenetic defect affecting the differentiation capabilities of stem cells, as opposed to a gatekeeper mutation. Hematopoietic cell lineages and their corresponding malignancies also offer insights into the role of epigenetics in differentiation and transformation. For example, the DNMT3A gene is commonly mutated in human cases of acute myeloid leukemia (AML) (Ley et al., 2010; Yan et al., 2011), whereas loss of Dnmt3a in mice progressively impairs hematopoietic stem cell differentiation (Challen et al., 2012), suggesting that epigenetic perturbation can lead to differentiation block and subsequent malignant transformation. CpG Island Methylator Phenotypes Aberrant DNA methylation of promoter CpG islands in cancer was initially viewed as a spontaneous or stochastic event with selection for functionally relevant silencing events. However, the discovery of cases of colorectal cancer with an exceptionally high frequency of CpG island hypermethylation suggested a coordinated event, possibly attributable to an epigenetic control defect. This phenomenon was referred to as a CpG island methylator phenotype (CIMP) (Toyota et al., 1999), analogous to the mutator phenotypes observed in mismatch repair-decient cancers. Although the existence of CIMP subsets of cancer was initially disputed (Yamashita et al., 2003), more recent genome-scale analyses have unambiguously documented distinct epigenetic subtypes for some types of cancer, such as colorectal cancer (Hinoue et al., 2012; Cancer Genome Atlas Network, 2012b) and glioblastoma (Noushmehr et al., 2010), and not for others, such as serous ovarian cancer (Cancer Genome Atlas Research Network, 2011). The most distinct examples of CIMP show exceptionally strong associations with other molecular or pathological features of the tumors, lending further validity to the biological relevance to this classication. For example, colorectal CIMP is very tightly associated with the V600E mutation of the BRAF oncogene (Weisenberger et al., 2006), whereas glioma CIMP (G-CIMP) is exceptionally tightly associated with mutation of the IDH1 gene (Noushmehr et al., 2010). In the case of G-CIMP, IDH1 mutation appears to be a causal contributor to the phenotype (Turcan et al., 2012), whereas BRAF mutation does not appear to be directly implicated in colorectal CIMP (Hinoue et al., 2009). The affected gene subsets differ between colorectal CIMP and glioblastoma G-CIMP, and their predisposition to aberrant methylation appears to be distinct from the susceptibility of stem cell polycomb targets in lamin-attachment domains (Hinoue et al.,
2012), which is generally not restricted to cancer subtypes. Despite a clear rationale for the association of IDH1 mutation with G-CIMP, the mechanistic basis for the coordinated hypermethylation events in most cases of CIMP is unknown and will remain an active area of investigation. Epigenetic Inuences on Genomic Integrity Mutation rates vary strikingly across the genome, with strong local inuences of base composition on single nucleotide variation (SNV) and regional effects of sequence composition, chromatin structure, replication timing, transcription, and nuclear architecture, among others, on both SNVs and structural alterations (Hodgkinson and Eyre-Walker, 2011). Despite widespread misuse of the term in the literature, it should be recognized that mutation rates of a tumor cannot be inferred directly from observed mutation numbers or frequencies in a tumor without consideration of the number of cell divisions that have occurred since a shared reference genome, although comparisons across ck uctuation the genome obviate the need for Luria-Delbru modeling and analysis. Epigenetic mechanisms can inuence both the rates at which lesions arise and the rates at which they are repaired. For example, the epigenetic mark 5-methylcytosine undergoes spontaneous deamination at higher rates than do unmethylated cytosines (Wang et al., 1982), whereas epigenetic silencing of the MLH1 mismatch repair gene increases mutation frequencies by several orders of magnitude, providing an adaptive advantage to mismatch repair-decient cancer cells. Unmethylated and methylated cytosine residues both undergo spontaneous hydrolytic deamination but yield uracil and thymine, respectively. Uracil is not a normal constituent base in DNA and is repaired much more efciently than thymine in a mismatch with guanine. As a consequence, the rate of C-to-T mutations in the context of CpG dinucleotides, most of which contain methylated cytosines, is about 10-fold higher than any other SNV in the human genome (Hodgkinson and Eyre-Walker, 2011). This effect is particularly pronounced in highly proliferative tissues because deamination of 5-methylcytosine in the parent strand just prior to DNA replication results in a full T:A base substitution that is not recognizable as a lesion for repair. Approximately a quarter of all TP53 mutations in human cancer are thus attributable to this epigenetic mark (Olivier et al., 2010). Regional Effects of Chromatin Organization Chromatin regulators play a role in maintaining genomic integrity (Papamichos-Chronakis and Peterson, 2013), and regional chromatin structure has a major impact on mutation frequencies. Megabase regions of repressive chromatin, represented by the H3K9me3 mark, are positively correlated with single-nucleotide ckler and Lehner, 2012), variations in cancer (Schuster-Bo whereas open chromatin associated with DNase I hypersensitive sites (DHS) have a lower inferred mutation rate, but this is partly due to evolutionary constraints on this compartment (Hodgkinson and Eyre-Walker, 2011). Transcription-coupled repair may also play a role in suppressing observed mutation frequencies in gene-rich euchromatic regions. Other types of mutation and structural change also appear to be associated with chromatin states. For example, retrotransposition occurs more frequently in hypomethylated regions (Lee et al., 2012). Genes resistant to cancer-associated hypermethy-
lation are more likely to have SINE and LINE retrotransposons near their transcription start sites than methylation-prone genes cio et al., 2010). Severe hypomethylation appears to be (Este associated with genomic instability. Mouse models of DNA methyltransferase deciency display chromosomal instability (Eden et al., 2003), and germline mutations of the DNMT3B gene cause ICF syndrome, characterized by centromeric instability (Okano et al., 1999). Indeed, areas of hypomethylation in the human germline showed higher frequencies of structural mutability (Li et al., 2012). DNA breakpoints associated with somatic copy number alterations are also enriched in hypomethylated domains (De and Michor, 2011). Epigenetic Inuences on DNA Repair Depletion of DNA methyltransferases causes increased microsatellite instability (Guo et al., 2004; Kim et al., 2004), destabilization of repeats (Dion et al., 2008), and dramatically increased telomere length, telomeric recombination, and alternative telomere lengthening (Gonzalo et al., 2006). These effects of DNA methyltransferase depletion appear to be mediated in part by a drop in DNA repair proteins as part of DNA damage response (Loughery et al., 2011). The Dnmt1 protein has also been shown to be recruited to areas of irradiation-induced DNA damage, possibly to facilitate repair of epigenetic information following DNA repair (Mortusewicz et al., 2005). It is increasingly appreciated that chromatin can serve as a cellular sensor for DNA damage and other genomic events (Johnson and Dent, 2013). Epigenetic silencing of DNA repair genes such as MLH1, MGMT, BRCA1, WRN, FANCF, and CHFR can boost mutation rates and promote genomic instability in cancer cells (Toyota and Suzuki, 2010). Familial cases of tumors with microsatellite instability (MSI) in Lynch syndrome result from germline mutations in mismatch repair genes, primarily MSH2 and MLH1. However, most MSI-high tumors arise from an epigenetic defect in sporadic cases of cancer. Approximately 15% of sporadic cases of colorectal cancer display MSI as a consequence of epigenetic silencing of the MLH1 mismatch repair gene by promoter CpG island hypermethylation (Herman et al., 1998) in the context of CIMP (Toyota et al., 1999; Weisenberger et al., 2006). MSI caused by epigenetic silencing of MLH1 has also been reported in other types of cancer, including about a quarter of sporadic endometrial cancers (Simpkins et al., 1999). Germline variants of MLH1 and MSH2 can predispose to extensive somatic epigenetic silencing of these genes and thereby increase cancer risk (Hitchins et al., 2011; Ligtenberg et al., 2009). Such familial cases of systemic epigenetic abnormalities can masquerade as germline transmission of epigenetic defects. True transgenerational epigenetic inheritance is evident in genomic imprinting and in mouse models but has been difcult to demonstrate directly in human populations, although there is indirect evidence for its existence (Daxinger and Whitelaw, 2012). The O6-Methylguanine DNA methyltransferase (MGMT) enzyme repairs O6-alkylated guanine residues in genomic DNA. O6-methylguanine pairs with thymine and would lead to a G-to-A transition during DNA replication if left unrepaired. MGMT promoter methylation in colorectal cancer is associated with G-to-A mutations in KRAS (Esteller et al., 2000b) and in TP53 (Esteller et al., 2001). Alkylating agents such as temozolomide are the current standard of care for malignant glioblastoma
Figure 3. Genetic Alterations in Epigenetic Regulators

Mutations and other genetic alterations reported for selected epigenetic regulators are shown for various types of human cancer in a heatmap. Malignancies are grouped by epithelial, hematological, and other cancers. Mutations, represented by colored cells, are deemed loss of function (blue) unless evidence for gain of function (either hypermorphic or neomorphic, red) has been shown. Other genetic alterations are plotted with different symbols, with a slash indicating translocation events and a dot indicating copy number alterations. Translocations that generate oncogenic fusion proteins are represented in red as well. The mutation
(legend continued on next page)
(GBM) but are counteracted by MGMT-mediated repair of the alkylation damage. Epigenetic silencing of MGMT by promoter CpG island hypermethylation inactivates this repair pathway and renders the tumor more sensitive to the temozolomide treatment (Esteller et al., 2000a; Hegi et al., 2005). A Genetic Basis for Epigenetic Disruption in Cancer The discovery of mutations in SMARCB1/SNF5 driving malignant rhabdoid tumors rst introduced genetic disruption of epigenetic control as a mechanism of oncogenesis (Versteege et al., 1998). Mutations in epigenetic regulators continued to emerge from subsequent cancer studies and have surged in recent large-scale sequencing efforts (Figure 3). Epigenetic control genes are mutated in about half of hepatocellular carcinomas (Fujimoto et al., 2012) and bladder cancer (Gui et al., 2011) and represent 6 of the 12 most signicantly mutated genes in medulloblastoma (Pugh et al., 2012). It is conceivable that disruption of epigenetic control by mutation of a key regulator has the capacity to cause widespread changes to the transcriptome, multiplying the effect of the single genetic alteration. It should be recognized that some of the mutations reported for epigenetic regulators may be passenger events, particularly in tumors with high background mutation rates. Therefore, we have emphasized hot spot mutations and genes recurrently mutated at signicant frequencies. We focus here on somatic mutations, but germline variations have also been shown to play a role in cancer. For example, germline mutations in BAP1 have been found to be linked to a tumor predisposition syndrome characterized by melanocytic tumors, mesothelioma, and uveal melanoma (Testa et al., 2011; Wiesner et al., 2011), and rare germline allelic forms of PRDM9 have been found to be associated with childhood leukemia (Hussin et al., 2013). DNA Methylation Writers and Erasers The DNA methyltransferase DNMT3A is recurrently mutated in AML and other myeloid malignancies (Ley et al., 2010; Yan et al., et al., 2011), as well as T cell lymphoma (Couronne 2012). The mutations often occur at a R882 hot spot but nevertheless likely reect loss of function of DNMT3A. Mutations in the DNA methylation eraser TET2 have also been identied in the same cancer types (Abdel-Wahab et al., 2009; Langemeijer et al., 2009; Quivoron et al., 2011), and bone marrow from patients with TET2 mutations shows reduced levels of 5hmC (Ko et al., 2010). The isocitrate dehydrogenases IDH1 and IDH2 are also recurrently mutated in AML. IDH1 enzymes with the R132 hot spot mutation and IDH2 enzymes containing R140 or R172 mutations have lost the ability to produce a-ketoglutarate (a-KG) but instead convert a-KG to an aberrant metabolite 2-hydroxyglutarate (2-HG), a competitive inhibitor of a-KG-dependent dioxygenases, such as the TETs and JmjC-domain-containing histone demethylases (Lu et al., 2012; Xu et al., 2011). IDH1/2 mutations are mutually exclusive with
TET2 mutations in AML, which is consistent with the inhibitory effect of 2-HG on TETs as a mediator of the effects of IDH1/2 mutations (Figueroa et al., 2010; Weissmann et al., 2012). The same hot spot mutation for IDH1 and, less often, IDH2, is also found in gliomas and glioblastomas. Both glioblastomas with IDH1 mutations (Noushmehr et al., 2010) and cases of AML with mutation of IDH1 or IDH2 (Figueroa et al., 2010) display CpG island methylator phenotypes. Histone Gene Mutations Mutations in histone variants H3.3 (H3F3A) or sometimes H3.1 (HIST1H3B) have been found in pediatric (Schwartzentruber et al., 2012; Wu et al., 2012) and adult brain tumors (Sturm et al., 2012) with K27M and G34R or G34V mutation hot spots. Tumors with G34 mutations display extensive DNA hypomethylation, particularly in subtelomeric regions (Sturm et al., 2012), perhaps contributing to alternative lengthening of telomeres (ALT) (Schwartzentruber et al., 2012). Mutations were also observed in the ATRX and DAXX genes, encoding proteins responsible for loading of the H3.3 variant into the telomere region (Schwartzentruber et al., 2012). Pancreatic neuroendocrine tumors (PanNETs) with ATRX and DAXX mutations also exhibit ALT (Heaphy et al., 2011). Because this phenotypic effect is associated with an H3.3 loading defect, the G34 mutations may also interfere with H3.3 loading. In contrast, tumors with K27M mutations did not display ALT, and these mutations may instead mimic dimethylated lysine 27, a repressive Polycomb mark, given that methionine is a natural mimic of this epigenetic mark (Hyland et al., 2011). H3.3 G34R mutations have also been reported in primitive neuroectodermal tumors of the central nervous system (CNS) (Gessi et al., 2013), mirroring the defect in the ATRX-DAXX-H3.3 axis in other brain tumors and PanNETs. Mutations in HIST1H3B and HIST1H1C have been found in diffuse large B cell lymphoma (DLBCL), although the mutations do not occur in clusters (Lohr et al., 2012; Morin et al., 2011). These might be functionally different from the hot spot mutations seen in brain tumors. Focal deletion of a histone gene cluster at 6p22 is seen in near-haploid cases of acute lymphoblastic leukemia (Holmfeldt et al., 2013). Histone Methylation Writers The MLL gene, encoding one of the H3K4 methyltransferases, has more than 50 translocation fusion partners in different lineages of leukemia. These rearrangements account for 80% of the cases of infant leukemia and 5%10% of adult leukemia cases and are generally associated with poor prognosis (Tan et al., 2011a). The primary mechanism has been attributed to the recruitment of inappropriate epigenetic factors to MLL targets by fusions between recruitment proteins and the DNAbinding N terminus of MLL. Target genes for these recruited complexes include the HOX genes, particularly HOXA9, whose upregulation is a key feature of MLL leukemia. MLL regulates the expression of HOX genes in normal pluripotent cells, but the oncogenic fusion proteins keep them from being turned off
frequencies, represented by the darkness of the shade, are based on recent whole-genome/exome studies, with adjustments made where whole-genome/ exome studies are not available or have a small sample size. Different subtypes of lung cancers are combined without adjusting for subtype prevalence, and certain mutations may only represent one subtype. Cells showing no entry may represent false negatives in our curation or in the literature. Cancer types highly covered with whole-genome/exome studies (e.g., breast cancer) might have fewer false negatives than those that are not. MSS/MSI, microsatellite stable/ instable (MSI CRCs are excluded due to the high background mutation rate); DLBCL, diffuse large B cell lymphoma; FL, follicular lymphoma.
during differentiation and therefore impart stem-cell-like properties. Targeted therapeutic strategies are emerging for AML with MLL fusions, including inhibition of menin (encoded by MEN1), DOT1L, PRMT1, the histone acetylation reader BRD4, and LSD1 (Zeisig et al., 2012). In addition to the translocations, loss-of-function mutations of MLL-MLL3 have been reported in many different types of cancer, including AMLpossibly another way of disturbing the temporal control at promoters associated with pluripotency. MLL2 is mutated at very high frequency in B cell follicular lymphoma and diffuse large B cell lymphoma, which is consistent with the gain-of-function mutations of EZH2 in the same tumor types. Although menin is critical to the oncogenic effects of MLL fusion proteins in AML (Yokoyama and Cleary, 2008), loss-offunction mutations have been found in PanNETs (Jiao et al., 2011), which is consistent with a tumor-suppressor role, suggesting that cellular context is important. The recurrent t(5;11)(q35;p15.5) translocation in AML results in the fusion of the H3K36 methyltransferase NSD1 to nucleoporin98 (NUP98), with elevated levels of H3K36me3 levels at HOXA genes and accompanying transcriptional activation. Translocations involving another dedicated H3K36 methyltransferase WHSC1/MMSET/NSD2 are seen in 20% of multiple myelomas. Another H3K36 methyltransferase, SETD2, is recurrently mutated in clear cell renal cell carcinomas (ccRCC) (Dalgliesh et al., 2010). A recent study reconstructed the phylogenetic structure of molecular events in ccRCC with multiple spatially separated samples from the same tumors (Gerlinger et al., 2012). In both of the two patients studied, distinct SETD2-inactivating mutations were found in different parts of the same tumor. Immunohistochemistry staining conrmed H3K36me3 loss in all the mutant tumors. This convergent somatic evolution indicates that failure to establish H3K36 methylation marks provides a strong selective advantage relatively late in ccRCC progression. A similar molecular convergence was found for KDM5C, an H3K4 demethylase, in one of the two patients. This, together with recurrent mutations in other epigenetic regulators, shows that epigenetic dysregulation, often mediated by genetic events, is important in advanced ccRCCs. EZH2, the writer for the H3K27 methylation mark associated with Polycomb repression, has long been viewed as an oncogene in cancer. Indeed, gain-of-function mutations are seen in lymphomas. However, loss-of-function mutations in this gene have recently been described in other cancers. We discuss these divergent effects of EZH2 mutations and other alterations to this pathway in more detail later (Figure 4). Histone Methylation Erasers Consistent with EZH2 overexpression in various solid tumors, the corresponding eraser KDM6A/UTX is mutated in more than a dozen tumor types, with the highest frequency in bladder (Gui et al., 2011; van Haaften et al., 2009). The H3K9 demethylase KDM4C/GASC1 is amplied in breast cancer and has been shown to drive transformation (Liu et al., 2009; Rui et al., 2010). Ectopic expression of this putative oncogene in vitro causes an efcient decrease of H3K9me3 (Cloos et al., 2006). Its coamplication with JAK2 (Rui et al., 2010)which phosphorylates H3Y41 and prevents binding of H3K9 methylation reader HP1 to the H3K9 methylation markin lymphoma makes for an inter46 Cell 153, March 28, 2013 2013 Elsevier Inc.
esting example of a single genetic event hitting two possible epigenetic regulators. Inhibiting the two coamplied and cooperating gene products is efcient at killing these lymphoma cells. Histone Acetylation Writers and Erasers The counteracting HATs and HDACs are considered to be promiscuous and often have important nonhistone substrates such as TP53. There are three major families of HATs, namely the CBP/P300, GNAT, and the MYST families. CREBBP is mutated at high frequency in follicular lymphoma and DLBCL (Morin et al., 2011; Pasqualucci et al., 2011) and in ALL (Mullighan et al., 2011), particularly relapsed hyperdiploid ALL (Inthal et al., 2012). Its paralog EP300 also undergoes frequent mutation (Gayther et al., 2000; Gui et al., 2011; Pasqualucci et al., 2011; Zhang et al., 2012) and loss of heterozygosity (LOH) in many different epithelial cancers. HATs have also been implicated in gene fusions. The t(8;16)(p11;p13) translocation in AML fuses the N-terminal part of MOZ, the founding member of the MYST HAT family, to the major part of the CBP gene containing the acetylase domain. MOZ has also been found to be involved in fusions with EP300 (Yang, 2004). These translocations generating chimeric oncoproteins with the DNA-binding domain of MOZ and the transcription-activating domain of another coactivator are associated with AML M5/M4. These results suggest that both disruption and redirection of HAT could contribute to cancer. Reports of mutations in HDACs are rare. Rather, HDACs are often co-opted by other genetic alterations. A prime example is the PML-RARa translocation, which is responsible for 95% of the AML FAB-M3 (APL, acute promyelocytic leukemia) cases. The leukemogenetic effect of this translocation is primarily mediated through aberrant recruitment of N-CoR/HDAC repressor complexes (Minucci and Pelicci, 2006). The retinoic acid receptor-a (RARa) part binds to retinoic acid-responsive elements (RAREs), whereas the PML moiety recruits the HDAC-containing repressive complex. All-trans retinoic acid (ATRA) targets RARa, dissociates these repressor complexes, and effectively induces differentiation of the leukemic promyelocytes. Combination therapy of ATRA and arsenic trioxide showed excellent clinical response and turned APL into a highly curable disease (Wang and Chen, 2008). Another fusion protein, PLZFRARa, blocks differentiation by a similar mechanism. However, APL with this translocation is ATRA resistant due to the higher afnity of the PLZF moiety to the N-CoR complex, but a combination of ATRA with HDAC inhibitors can fully reverse the transcriptional repression and induce terminal differentiation for this type of AML (Wang and Chen, 2008). Similarly, the RUNX1-ETO fusion, the AML1-ETO fusion, and the CBF-MYH11 protein from inv(16) all recruit HDACs, and efcacy of HDAC inhibitors has been demonstrated for all three (Zeisig et al., 2012). Epigenetic Readers The epigenetic readers add another layer of control to the epigenetic state by serving as interpreters of the epigenetic state and relaying epigenetic signals. Many of the epigenetic writer/eraser/ remodelers have intrinsic reader domains or interact with dedicated readers to sense the presence or absence of particular epigenetic marks. Translocations joining BRD4 or occasionally BRD3both readers of the BET bromodomain-containing familyto almost the entire length of the NUT gene dene
Figure 4. Genetic Disruption of Epigenetic Control at H3K27 in Cancer

The counteracting writer EZH2 and eraser KDM6A/UTX form a pair in regulating an important epigenetic mark, methylation at H3 lysine 27. EZH2 catalyzes the methylation process with help from other components in PRC2, whereas KDM6A, part of the Trithorax complex, removes this repressive mark. The K27me3 mark attracts another Polycomb complex, PRC1, which ubiquitinates H2AK119, and thereby blocks PolII elongation. Another Polycomb complex, PR-DUB, is also critical to the maintenance of the repression at a subset of the Polycomb genes, although it removes the H2AK119ub mark and thus counteracts PRC1 in that regard. Mutations and genetic alterations spanning a wide spectrum of human cancers hit this epigenetic pathway. Solid tumors show possibly neomorphic histone K27 mutations (mimicking H3K27me2), UTX mutation, EZH2 amplication, and/or overexpression due to genomic loss of the repressive microRNA miR101, as well as amplication/overexpression of the PRC1 member BMI1, and lymphoma exhibits gain-of-function mutations of EZH2, which is consistent with a gain of Polycomb repression (red boxes) in the affected malignancies. In contrast, myeloid malignancies and ALL, particularly early T cell precursor ALL, show mutations that could sabotage Polycomb repression (blue boxes). Mouse models show that loss of BAP1, the enzymatic unit of PR-DUB, leads to myeloid transformation, although BAP1 mutation in MDS is rare. Gray boxes indicate that the effect on H3K27me3 is not clear.
a lethal, poorly differentiated pediatric tumor, NUT midline carcinoma (NMC) (French et al., 2008). The BET family members targetable by BET inhibitors (Filippakopoulos et al., 2010) are lysine acetylation readers that bind transcriptionally active chromatin as acetylated lysine readers and are targetable by BET inhibitors (Filippakopoulos et al., 2010). Functional studies show that the BRD-NUT fusion oncoprotein binds avidly to acetylated histones, resulting in a differentiation block potentially by interfering with transcriptional programs driving differentiation (French et al., 2008). BRD3 is signicantly mutated in lung adenocarcinomas (Imielinski et al., 2012), and BRD8 is mutated in liver cancer (Fujimoto et al., 2012). In addition, the plant homeodomain (PHD)-domain containing gene PHF6 is recurrently mutated in AML (Van Vlierberghe et al., 2011), and overall loss of this gene (mutation and/or deletion) is observed in T-ALL (Van Vlierberghe et al., 2010).
Chromatin Remodelers A large number of SWI/SNF complexes exist in mammals and contribute to lineage- and tissue-specic gene expression (Wilson and Roberts, 2011). The two major types of SWI/SNF complexes are BAF (BRM containing, or SWI/SNF-A) and PBAF (BRG1 containing, or SWI/SNF-B), dened by a core enzymatic unit being either SMARCA2 (BRM) or SMARCA4 (BRG1). Those complexes also contain other core units, such as SMARCB1/SNF5 and ARID1A/Bwhich are unique to BAF and PBRM1 and BRD7which are unique to PBAF. Truncating mutations in the SMARCB1 gene are very common in malignant rhabdoid tumors (RTs) (Versteege et al., 1998), a rare yet lethal tumor diagnosed in children. The biallelic nature of the inactivation ts with a tumor suppressor role for SMARCB1. Familial cases of RTs are associated with inheritance of one defective SMARCB1 allele. SMARCB1 is also mutated in a few other
cancers (Figure 1). SMARCA4 mutation is also seen in familial cases of RT (Schneppenheim et al., 2010), indicating that it is indeed SWI/SNF dysfunction that is responsible for RT development. SMARCA4 is also mutated Burkitts lymphoma in a mutually exclusive manner with ARID1A mutations (Love et al., 2012), again suggesting a driver role for SWI/SNF mutations. Recurrent SMARCA4 mutation is also seen in lung cancer (Imielinski et al., 2012) and medulloblastoma (especially the WNT subtype) (Jones et al., 2012; Parsons et al., 2011; Pugh et al., 2012; Robinson et al., 2012). ARID1A mutations have been found in more than ten different tumor types, with the highest rate in the clear cell subtype of ovarian cancer, where it is mutated in more than half of the tumors (Jones et al., 2010; Wiegand et al., 2010). ARID1A is also mutated in the endometrioid subtypes of ovarian (Wiegand et al., 2010) and endometrial cancers (Guan et al., 2011). ARID1B and ARID2 mutations are also seen in various cancer types, including liver cancer (Fujimoto et al., 2012) and melanoma (Hodis et al., 2012), among others. In addition, the polybromocontaining PBRM1 in the PBAF complex was recently found to be the second most mutated gene in clear cell renal cell carcinomas (Varela et al., 2011). Another SWI/SNF gene, SMARCE1, is highly recurrently mutated in clear cell meningiomas (Smith et al., 2013). The exceptionally high mutation rate of SWI/SNF member in clear cell tumors from different tissues (ovary, kidney, and meninges) highlights an interesting possible link between clear cell tumors and SWI/SNF dysfunction. ATRX, responsible for H3.3 incorporation at telomeres and pericentric heterochromatin, is often mutated in PanNETs, the second most common malignancy of the pancreas (Jiao et al., 2011). Interestingly, there are also recurrent mutations in the associated chaperone DAXX in the same cancer type, and the two mutations are mutually exclusive. Mutations of these two genes are also found in GBM, where they are mutually exclusive with the H3F3A mutations described earlier, with any of these three genes mutated in almost half of the tumors studied (Schwartzentruber et al., 2012). These mutations all lead to alternative lengthening of telomeres associated with increased genomic instability (Heaphy et al., 2011; Schwartzentruber et al., 2012). With the possible exception of ATRX and DAXX, most of the mutations in the SWI/SNF family members are not associated with genomic instability (Wilson and Roberts, 2011). Rather, perturbed differentiation may be the major mechanism, as cells from different lineages coexist within an individual rhabdoid tumor. The CHD family chromatin remodelers can be divided into three classes: class I (CHD1/2), class II (CHD3/4, in the NuRD/ Mi-2/CHD complex), and class III (CHD59). The NuRD/Mi-2/ CHD complexes are unique in that they have core enzymatic subunits with at least two distinct functionsATP-dependent remodeling (CHD3 and CHD4), as well as histone deacetylase (HDAC1 and HDAC2) functions (Lai and Wade, 2011)and therefore couple two epigenetic processes in one complex for transcriptional repression. Their MBD and MTA subunits target the complex to different parts of the genome by binding methylated DNA (MBD) or other transcription factors (MTA) in physiological and pathological conditions. For example, the MTA-2-containing NuRD complex associates with TWIST in
breast cancer, represses genes such as E-cadherin (CDH1), and contributes to EMT. CHD4 is mutated in 17% of serous endometrial cancer (Le Gallo et al., 2012). Of the other CHDs, CHD1 is the second most frequently deleted gene in prostate cancer, dening an ETS-negative subtype (Grasso et al., 2012), with mutations reported as well (Berger et al., 2011). Epigenetic Insulators CTCF is located in 16q22.1 with LOH in breast and prostate cancers (Filippova et al., 1998) and Wilms tumors (Mummert et al., 2005). CTCF mutation has also been reported in breast cancer (Filippova et al., 2002), and this mutation is found to be signicant in a cohort of 510 tumors (Cancer Genome Atlas Network, 2012c). Rare mutations in this gene have also been reported for prostate cancer (Filippova et al., 2002), Wilms tumor (Filippova et al., 2002), AML (Dolnik et al., 2012), ALL (Mullighan et al., 2011; Zhang et al., 2012), and endometrial cancers (Le Gallo et al., 2012). The functional implications of CTCF deletion/mutation, especially given the low frequency, have not been fully delineated yet, but abrogation of proper insulation might be one mechanism. Genetic Disruption of a Central Epigenetic Control Circuit Figure 4 illustrates the diverse ways in which a central epigenetic control circuit can be impacted in cancer. EZH2 catalyzes methylation at H3K27, as part of PRC2. Two other core subunits of PRC2 are EED and SUZ12, and other components such as JARID2 can be part of a PRC2 complex too. EZH2 has long been thought to be oncogenic because it is overexpressed as a result of amplication of EZH2 in breast, bladder, and other cancers (Bracken et al., 2003), as well as genetic loss of miR101, which represses EZH2 in prostate cancer (Varambally et al., 2008). In line with this view, gain-of-function hot spot mutations (Y641 and A677) in the SET domain of EZH2 have been found in a signicant portion of lymphomas (Lohr et al., 2012; Morin et al., 2010, 2011; Pasqualucci et al., 2011). More convincingly, EZH2 amplication and overexpression in two of the largest subgroups of medulloblastoma are mutually exclusive with mutation of the H3K27 demethylase KDM6A, suggesting that accumulation of H3K27me3 is a key step in these tumors (Robinson et al., 2012). On the other hand, loss-of-function mutations of EZH2 have also been found in a series of myeloid malignancies, including myelodysplastic syndrome (MDS), multiple myeloma, myeloproliferative neoplasms (MPN), and myelodysplastic/myeloproliferative neoplasms (MDS/MPN), as well as in head-and-neck squamous cell carcinomas (HNSCCs) (Ernst et al., 2010; Nikoloski et al., 2010; Stransky et al., 2011), suggesting that PRC2 can also act as a tumor suppressor. Mutually exclusive recurrent deletion and loss-of-function mutations of EZH2 and SUZ12 in T lineage acute lymphoblastic leukemia (T-ALL) (Ntziachristos et al., 2012) and of all three PRC2 subunits in early T cellprecursor ALL (Zhang et al., 2012) further substantiate a tumor suppressor role for PRC2 function. Indeed, disruption of EZH2 is sufcient to induce T-ALL in mice (Simon et al., 2012). PRC2 component mutations are much more common in early T cell precursor ALL, a lymphoblastic leukemia with myeloid features. The Polycomb repressive deubiquitinase (PR-DUB) component ASXL1 is also mutated in myeloid malignancies, and ASXL1
Figure 5. Interplay between the Cancer Genome and Epigenome

The genome and epigenome inuence each other, as the genome provides the primary sequence information and encodes regulators of epigenetic states, whereas the epigenome controls the accessibility and interpretation of the genome. Changes in one can inuence the other, forming a partnership in producing genetically or epigenetically encoded phenotypic variation subject to Darwinian selection for growth advantage and thus eventually achieving the hallmarks of cancer (Hanahan and Weinberg, 2011). Genetic instability and mutation and epigenomic disruption can be considered enabling characteristics of cancer cells.
mutation mediates myeloid transformation through loss of PRC2 repression (Abdel-Wahab et al., 2012), which is in line with the observed loss of function of PRC2 members in myeloid disorders. Mouse models also show that loss of BAP1, the enzymatic unit of PR-DUB, leads to myeloid transformation (Dey et al., 2012). This, together with a BAP1 catalytic mutation found in a MDS patient lacking other MDS mutations (Dey et al., 2012), further lends credibility to the idea that loss of Polycomb repression drives myeloid disorders, whereas in B cell lymphoma and solid tumors, gain of Polycomb repression seems important. EZH2 also exhibits PRC2-independent oncogenic activities. For example, in castration-resistant prostate cancer, Akt-mediated phosphorylation of EZH2 at S21 can shift EZH2 from PRC2-dependent promoters to EZH2 solo promoters. This
EZH2 activity is often associated with androgen receptor (AR) and activates gene expression at these loci (Xu et al., 2012). These complex ways in which the H3K27me3 axis is disrupted in cancer suggest that differential therapeutic approaches should be developed for (1) myeloid malignancies and early T cell ALL, (2) lymphomas and some solid tumors, and (3) possibly hormone-associated cancers. In particular, caution should be used when considering EZH2 inhibitors for the rst group of tumors, which features genetic lesions leading to loss of PRC2 repression. Conclusions It is clear that the cancer genome and epigenome inuence each other in a multitude of ways (Figure 5). They offer complementary
mechanisms to achieve similar results, such as the inactivation of tumor-suppressor genes, including BRCA1, CDKN2A, VHL, and RB1, by either deletion or epigenetic silencing, and they can work cooperatively as in the case of CIMP and BRAF mutation in colorectal cancer, where CIMP appears to create a permissive context for BRAF mutation as early as in the precursor lesion (Hinoue et al., 2009; Yamamoto et al., 2012). Many questions and challenges remain. For example, the explosion in the number of epigenetic regulator mutations identied in human cancer has underscored the importance of epigenetic control in tumor suppression, but the phenotypic consequences of these mutations remain largely uncharacterized. However, this plethora of newly identied mutations in epigenetic regulators opens entirely new avenues of therapeutic attack (Dawson and Kouzarides, 2012). Another major question remaining in the eld is the mechanistic basis for some well-recognized examples of disruption of epigenetic control, such as CIMP in colorectal cancer. Even in the case of G-CIMP in gliomas and its very tight association with IDH1 mutation, it is not clear what confers the gene specicity of the hypermethylation events. Correlated events such as CIMP create other challenges as well. In contrast to most mutations, CIMP-associated DNA methylation events are highly correlated, with a large number of recurrent alterations that appear to be passenger events without functional contribution to the cancer process. This high degree of correlation precludes the straightforward use of recurrence frequency among different tumors as a main lter criterion in the identication of functionally relevant epigenetic driver events. Therefore, the identication of epigenetic drivers must rely more on the analysis of transcriptional consequences, mutual exclusivity with other events in the same pathway within a tumor, complementary mechanisms of inactivation of the same gene in other tumors, and, most importantly, functional experimental validation of an impact of the epigenetic gene inactivation on cellular proliferation, immortality, angiogenesis, cell death, invasion, or metastasis. Cancer epigenetics and genetics may inform each other. Genetics can shed light on the identity of epigenetic drivers by revealing mutual exclusivity with genetic aberrations in the same gene or pathway. Epigenetics may also provide insight into genetic drivers in a similar fashion. In addition, the understanding of epigenetic networks provides a framework to interpret the functional signicance of lower-frequency drivers in the same pathway. The high frequency of epigenetic regulator mutations seen in various cancers, the hot spot nature of some mutations found, mutual exclusivity between different mechanisms affecting the same genes/pathways, clonal analysis highlighting convergent evolution, and validations in experimental systems all attest to the importance of mutations in epigenetic regulators in cancer and strengthen the concept that disruption of epigenetic control is a common enabling characteristic of cancer cells.
ACKNOWLEDGMENTS This work was supported by NIH grants R01-CA170550, R01-CA157918, R01-DA030325, and U24 CA143882 (P.W.L.). We would like to thank Peter Jones, Benjamin Berman, and Timothy Triche, Jr. for helpful discussions.
REFERENCES Abdel-Wahab, O., Mullally, A., Hedvat, C., Garcia-Manero, G., Patel, J., Wadleigh, M., Malinge, S., Yao, J., Kilpivaara, O., Bhat, R., et al. (2009). Genetic characterization of TET1, TET2, and TET3 alterations in myeloid malignancies. Blood 114, 144147. Abdel-Wahab, O., Adli, M., LaFave, L.M., Gao, J., Hricik, T., Shih, A.H., Pandey, S., Patel, J.P., Chung, Y.R., Koche, R., et al. (2012). ASXL1 mutations promote myeloid transformation through loss of PRC2-mediated gene repression. Cancer Cell 22, 180193. Adams, D., Altucci, L., Antonarakis, S.E., Ballesteros, J., Beck, S., Bird, A., Bock, C., Boehm, B., Campo, E., Caricasole, A., et al. (2012). BLUEPRINT to decode the epigenetic signature written in blood. Nat. Biotechnol. 30, 224226. Baer, C., Claus, R., and Plass, C. (2013). Genome-wide epigenetic regulation of miRNAs in cancer. Cancer Res. 73, 473477. Baylin, S.B., and Jones, P.A. (2011). A decade of exploring the cancer epigenome - biological and translational implications. Nat. Rev. Cancer 11, 726734. ppener, J.W., de Bustros, A., Steenbergh, P.H., Lips, C.J., and Baylin, S.B., Ho Nelkin, B.D. (1986). DNA methylation patterns of the calcitonin gene in human lung cancers and lymphomas. Cancer Res. 46, 29172922. Berger, M.F., Lawrence, M.S., Demichelis, F., Drier, Y., Cibulskis, K., Sivachenko, A.Y., Sboner, A., Esgueva, R., Pueger, D., Sougnez, C., et al. (2011). The genomic complexity of primary human prostate cancer. Nature 470, 214220. Berman, B.P., Weisenberger, D.J., Aman, J.F., Hinoue, T., Ramjan, Z., Liu, Y., Noushmehr, H., Lange, C.P., van Dijk, C.M., Tollenaar, R.A., et al. (2012). Regions of focal DNA hypermethylation and long-range hypomethylation in colorectal cancer coincide with nuclear lamina-associated domains. Nat. Genet. 44, 4046. Bernstein, B.E., Mikkelsen, T.S., Xie, X., Kamal, M., Huebert, D.J., Cuff, J., Fry, B., Meissner, A., Wernig, M., Plath, K., et al. (2006). A bivalent chromatin structure marks key developmental genes in embryonic stem cells. Cell 125, 315326. Bird, A., Taggart, M., Frommer, M., Miller, O.J., and Macleod, D. (1985). A fraction of the mouse genome that is derived from islands of nonmethylated, CpG-rich DNA. Cell 40, 9199. Boyarchuk, E., Montes de Oca, R., and Almouzni, G. (2011). Cell cycle dynamics of histone variants at the centromere, a model for chromosomal landmarks. Curr. Opin. Cell Biol. 23, 266276. Bracken, A.P., Pasini, D., Capra, M., Prosperini, E., Colli, E., and Helin, K. (2003). EZH2 is downstream of the pRB-E2F pathway, essential for proliferation and amplied in cancer. EMBO J. 22, 53235335. Cancer Genome Atlas Research Network. (2011). Integrated genomic analyses of ovarian carcinoma. Nature 474, 609615. Cancer Genome Atlas Research Network. (2012a). Comprehensive genomic characterization of squamous cell lung cancers. Nature 489, 519525. Cancer Genome Atlas Network. (2012b). Comprehensive molecular characterization of human colon and rectal cancer. Nature 487, 330337. Cancer Genome Atlas Network. (2012c). Comprehensive molecular portraits of human breast tumours. Nature 490, 6170. Cedar, H., and Bergman, Y. (2012). Programming of DNA methylation patterns. Annu. Rev. Biochem. 81, 97117. Challen, G.A., Sun, D., Jeong, M., Luo, M., Jelinek, J., Berg, J.S., Bock, C., Vasanthakumar, A., Gu, H., Xi, Y., et al. (2012). Dnmt3a is essential for hematopoietic stem cell differentiation. Nat. Genet. 44, 2331. Chi, P., Allis, C.D., and Wang, G.G. (2010). Covalent histone modications miswritten, misinterpreted and mis-erased in human cancers. Nat. Rev. Cancer 10, 457469. Clark, S.J. (2007). Action at a distance: epigenetic silencing of large chromosomal regions in carcinogenesis. Hum. Mol. Genet. 16(Spec No 1), R88R95.
Cloos, P.A., Christensen, J., Agger, K., Maiolica, A., Rappsilber, J., Antal, T., Hansen, K.H., and Helin, K. (2006). The putative oncogene GASC1 demethylates tri- and dimethylated lysine 9 on histone H3. Nature 442, 307311. Cohen, N.M., Kenigsberg, E., and Tanay, A. (2011). Primate CpG islands are maintained by heterogeneous evolutionary regimes involving minimal selection. Cell 145, 773786. Coolen, M.W., Stirzaker, C., Song, J.Z., Statham, A.L., Kassir, Z., Moreno, C.S., Young, A.N., Varma, V., Speed, T.P., Cowley, M., et al. (2010). Consolidation of the cancer genome into domains of repressive chromatin by longrange epigenetic silencing (LRES) reduces transcriptional plasticity. Nat. Cell Biol. 12, 235246. , L., Bastard, C., and Bernard, O.A. (2012). TET2 and DNMT3A mutaCouronne tions in human T-cell lymphoma. N. Engl. J. Med. 366, 9596. Dalgliesh, G.L., Furge, K., Greenman, C., Chen, L., Bignell, G., Butler, A., Davies, H., Edkins, S., Hardy, C., Latimer, C., et al. (2010). Systematic sequencing of renal carcinoma reveals inactivation of histone modifying genes. Nature 463, 360363. Dawson, M.A., and Kouzarides, T. (2012). Cancer epigenetics: from mechanism to therapy. Cell 150, 1227. Daxinger, L., and Whitelaw, E. (2012). Understanding transgenerational epigenetic inheritance via the gametes in mammals. Nat. Rev. Genet. 13, 153162. De, S., and Michor, F. (2011). DNA secondary structures and epigenetic determinants of cancer genome evolution. Nat. Struct. Mol. Biol. 18, 950955. De Carvalho, D.D., Sharma, S., You, J.S., Su, S.F., Taberlay, P.C., Kelly, T.K., Yang, X., Liang, G., and Jones, P.A. (2012). DNA methylation screening identies driver epigenetic events of cancer cell survival. Cancer Cell 21, 655667. De Craene, B., and Berx, G. (2013). Regulatory networks dening EMT during cancer initiation and progression. Nat. Rev. Cancer 13, 97110. Dey, A., Seshasayee, D., Noubade, R., French, D.M., Liu, J., Chaurushiya, M.S., Kirkpatrick, D.S., Pham, V.C., Lill, J.R., Bakalarski, C.E., et al. (2012). Loss of the tumor suppressor BAP1 causes myeloid transformation. Science 337, 15411546. Diala, E.S., and Hoffman, R.M. (1982). Hypomethylation of HeLa cell DNA and the absence of 5-methylcytosine in SV40 and adenovirus (type 2) DNA: analysis by HPLC. Biochem. Biophys. Res. Commun. 107, 1926. Dion, V., Lin, Y., Hubert, L., Jr., Waterland, R.A., and Wilson, J.H. (2008). Dnmt1 deciency promotes CAG repeat expansion in the mouse germline. Hum. Mol. Genet. 17, 13061317. Dolnik, A., Engelmann, J.C., Scharfenberger-Schmeer, M., Mauch, J., Kelken nke, J., Ku hn, M.W., Paschka, berg-Schade, S., Haldemann, B., Fries, T., Kro P., et al. (2012). Commonly altered genomic regions in acute myeloid leukemia are enriched for somatic mutations involved in chromatin remodeling and splicing. Blood 120, e83e92. Dunham, I., Kundaje, A., Aldred, S.F., Collins, P.J., Davis, C.A., Doyle, F., Epstein, C.B., Frietze, S., Harrow, J., Kaul, R., et al.; ENCODE Project Consortium. (2012). An integrated encyclopedia of DNA elements in the human genome. Nature 489, 5774. Easwaran, H., Johnstone, S.E., Van Neste, L., Ohm, J., Mosbruger, T., Wang, Q., Aryee, M.J., Joyce, P., Ahuja, N., Weisenberger, D., et al. (2012). A DNA hypermethylation module for the stem/progenitor cell signature of cancer. Genome Res. 22, 837849. Eden, A., Gaudet, F., Waghmare, A., and Jaenisch, R. (2003). Chromosomal instability and tumors promoted by DNA hypomethylation. Science 300, 455. Ehrlich, M., and Wang, R.Y. (1981). 5-Methylcytosine in eukaryotic DNA. Science 212, 13501357. Ehrlich, M., Gama-Sosa, M.A., Huang, L.H., Midgett, R.M., Kuo, K.C., McCune, R.A., and Gehrke, C. (1982). Amount and distribution of 5-methylcytosine in human DNA from different types of tissues of cells. Nucleic Acids Res. 10, 27092721. Ernst, T., Chase, A.J., Score, J., Hidalgo-Curtis, C.E., Bryant, C., Jones, A.V., Waghorn, K., Zoi, K., Ross, F.M., Reiter, A., et al. (2010). Inactivating mutations of the histone methyltransferase gene EZH2 in myeloid disorders. Nat. Genet. 42, 722726.
Ernst, J., Kheradpour, P., Mikkelsen, T.S., Shoresh, N., Ward, L.D., Epstein, C.B., Zhang, X., Wang, L., Issner, R., Coyne, M., et al. (2011). Mapping and analysis of chromatin state dynamics in nine human cell types. Nature 473, 4349. cio, M.R., Gallegos, J., Vallot, C., Castoro, R.J., Chung, W., Maegawa, S., Este Oki, Y., Kondo, Y., Jelinek, J., Shen, L., et al. (2010). Genome architecture marked by retrotransposons modulates predisposition to DNA methylation in cancer. Genome Res. 20, 13691382. Esteller, M., Garcia-Foncillas, J., Andion, E., Goodman, S.N., Hidalgo, O.F., Vanaclocha, V., Baylin, S.B., and Herman, J.G. (2000a). Inactivation of the DNA-repair gene MGMT and the clinical response of gliomas to alkylating agents. N. Engl. J. Med. 343, 13501354. Esteller, M., Toyota, M., Sanchez-Cespedes, M., Capella, G., Peinado, M.A., Watkins, D.N., Issa, J.P., Sidransky, D., Baylin, S.B., and Herman, J.G. (2000b). Inactivation of the DNA repair gene O6-methylguanine-DNA methyltransferase by promoter hypermethylation is associated with G to A mutations in K-ras in colorectal tumorigenesis. Cancer Res. 60, 23682371. Esteller, M., Risques, R.A., Toyota, M., Capella, G., Moreno, V., Peinado, M.A., Baylin, S.B., and Herman, J.G. (2001). Promoter hypermethylation of the DNA repair gene O(6)-methylguanine-DNA methyltransferase is associated with the presence of G:C to A:T transition mutations in p53 in human colorectal tumorigenesis. Cancer Res. 61, 46894692. Fabbri, M., and Calin, G.A. (2010). Epigenetics and miRNAs in human cancer. Adv. Genet. 70, 8799. Feinberg, A.P., and Vogelstein, B. (1983). Hypomethylation distinguishes genes of some human cancers from their normal counterparts. Nature 301, 8992. Fenouil, R., Cauchy, P., Koch, F., Descostes, N., Cabeza, J.Z., Innocenti, C., Ferrier, P., Spicuglia, S., Gut, M., Gut, I., and Andrau, J.C. (2012). CpG islands and GC content dictate nucleosome depletion in a transcription-independent manner at mammalian promoters. Genome Res. 22, 23992408. Figueroa, M.E., Abdel-Wahab, O., Lu, C., Ward, P.S., Patel, J., Shih, A., Li, Y., Bhagwat, N., Vasanthakumar, A., Fernandez, H.F., et al. (2010). Leukemic IDH1 and IDH2 mutations result in a hypermethylation phenotype, disrupt TET2 function, and impair hematopoietic differentiation. Cancer Cell 18, 553567. Filippakopoulos, P., Qi, J., Picaud, S., Shen, Y., Smith, W.B., Fedorov, O., Morse, E.M., Keates, T., Hickman, T.T., Felletar, I., et al. (2010). Selective inhibition of BET bromodomains. Nature 468, 10671073. Filippova, G.N., Lindblom, A., Meincke, L.J., Klenova, E.M., Neiman, P.E., Collins, S.J., Doggett, N.A., and Lobanenkov, V.V. (1998). A widely expressed transcription factor with multiple DNA sequence specicity, CTCF, is localized at chromosome segment 16q22.1 within one of the smallest regions of overlap for common deletions in breast and prostate cancers. Genes Chromosomes Cancer 22, 2636. Filippova, G.N., Qi, C.F., Ulmer, J.E., Moore, J.M., Ward, M.D., Hu, Y.J., Loukinov, D.I., Pugacheva, E.M., Klenova, E.M., Grundy, P.E., et al. (2002). Tumorassociated zinc nger mutations in the CTCF transcription factor selectively alter tts DNA-binding specicity. Cancer Res. 62, 4852. Fraga, M.F., Ballestar, E., Villar-Garea, A., Boix-Chornet, M., Espada, J., Schotta, G., Bonaldi, T., Haydon, C., Ropero, S., Petrie, K., et al. (2005). Loss of acetylation at Lys16 and trimethylation at Lys20 of histone H4 is a common hallmark of human cancer. Nat. Genet. 37, 391400. French, C.A., Ramirez, C.L., Kolmakova, J., Hickman, T.T., Cameron, M.J., Thyne, M.E., Kutok, J.L., Toretsky, J.A., Tadavarthy, A.K., Kees, U.R., et al. (2008). BRD-NUT oncoproteins: a family of closely related nuclear proteins that block epithelial differentiation and maintain the growth of carcinoma cells. Oncogene 27, 22372242. Fujimoto, A., Totoki, Y., Abe, T., Boroevich, K.A., Hosoda, F., Nguyen, H.H., Aoki, M., Hosono, N., Kubo, M., Miya, F., et al. (2012). Whole-genome sequencing of liver cancers identies etiological inuences on mutation patterns and recurrent mutations in chromatin regulators. Nat. Genet. 44, 760764.
Gaffney, D.J., McVicker, G., Pai, A.A., Fondufe-Mittendorf, Y.N., Lewellen, N., Michelini, K., Widom, J., Gilad, Y., and Pritchard, J.K. (2012). Controls of nucleosome positioning in the human genome. PLoS Genet. 8, e1003036. Gal-Yam, E.N., Egger, G., Iniguez, L., Holster, H., Einarsson, S., Zhang, X., Lin, J.C., Liang, G., Jones, P.A., and Tanay, A. (2008). Frequent switching of Polycomb repressive marks and DNA hypermethylation in the PC3 prostate cancer cell line. Proc. Natl. Acad. Sci. USA 105, 1297912984. Gayther, S.A., Batley, S.J., Linger, L., Bannister, A., Thorpe, K., Chin, S.F., Daigo, Y., Russell, P., Wilson, A., Sowter, H.M., et al. (2000). Mutations truncating the EP300 acetylase in human cancers. Nat. Genet. 24, 300303. Gerlinger, M., Rowan, A.J., Horswell, S., Larkin, J., Endesfelder, D., Gronroos, E., Martinez, P., Matthews, N., Stewart, A., Tarpey, P., et al. (2012). Intratumor heterogeneity and branched evolution revealed by multiregion sequencing. N. Engl. J. Med. 366, 883892. Gertz, J., Varley, K.E., Reddy, T.E., Bowling, K.M., Pauli, F., Parker, S.L., Kucera, K.S., Willard, H.F., and Myers, R.M. (2011). Analysis of DNA methylation in a three-generation family reveals widespread genetic inuence on epigenetic regulation. PLoS Genet. 7, e1002228. rner, E., Mu hlen, A.Z., Waha, A., and Gessi, M., Gielen, G.H., Hammes, J., Do Pietsch, T. (2013). H3.3 G34R mutations in pediatric primitive neuroectodermal tumors of central nervous system (CNS-PNET) and pediatric glioblastomas: possible diagnostic and therapeutic implications? J. Neurooncol. 112, 6772. din, F. (2012). R-loop Ginno, P.A., Lott, P.L., Christensen, H.C., Korf, I., and Che formation is a distinctive characteristic of unmethylated human CpG island promoters. Mol. Cell 45, 814825. Gonzalo, S., Jaco, I., Fraga, M.F., Chen, T., Li, E., Esteller, M., and Blasco, M.A. (2006). DNA methyltransferases control telomere length and telomere recombination in mammalian cells. Nat. Cell Biol. 8, 416424. Graff, J.R., Gabrielson, E., Fujii, H., Baylin, S.B., and Herman, J.G. (2000). Methylation patterns of the E-cadherin 50 CpG island are unstable and reect the dynamic, heterogeneous loss of E-cadherin expression during metastatic progression. J. Biol. Chem. 275, 27272732. Grasso, C.S., Wu, Y.M., Robinson, D.R., Cao, X., Dhanasekaran, S.M., Khan, A.P., Quist, M.J., Jing, X., Lonigro, R.J., Brenner, J.C., et al. (2012). The mutational landscape of lethal castration-resistant prostate cancer. Nature 487, 239243. Guan, B., Mao, T.L., Panuganti, P.K., Kuhn, E., Kurman, R.J., Maeda, D., Chen, E., Jeng, Y.M., Wang, T.L., and Shih, IeM. (2011). Mutation and loss of expression of ARID1A in uterine low-grade endometrioid carcinoma. Am. J. Surg. Pathol. 35, 625632. Gui, Y., Guo, G., Huang, Y., Hu, X., Tang, A., Gao, S., Wu, R., Chen, C., Li, X., Zhou, L., et al. (2011). Frequent mutations of chromatin remodeling genes in transitional cell carcinoma of the bladder. Nat. Genet. 43, 875878. Guo, G., Wang, W., and Bradley, A. (2004). Mismatch repair genes identied using genetic screens in Blm-decient embryonic stem cells. Nature 429, 891895. Hanahan, D., and Weinberg, R.A. (2011). Hallmarks of cancer: the next generation. Cell 144, 646674. Hansen, K.D., Timp, W., Bravo, H.C., Sabunciyan, S., Langmead, B., McDonald, O.G., Wen, B., Wu, H., Liu, Y., Diep, D., et al. (2011). Increased methylation variation in epigenetic domains across cancer types. Nat. Genet. 43, 768775. Hawkins, R.D., Hon, G.C., Lee, L.K., Ngo, Q., Lister, R., Pelizzola, M., Edsall, L.E., Kuan, S., Luu, Y., Klugman, S., et al. (2010). Distinct epigenomic landscapes of pluripotent and lineage-committed human cells. Cell Stem Cell 6, 479491. Heaphy, C.M., de Wilde, R.F., Jiao, Y., Klein, A.P., Edil, B.H., Shi, C., Bettegowda, C., Rodriguez, F.J., Eberhart, C.G., Hebbar, S., et al. (2011). Altered telomeres in tumors with ATRX and DAXX mutations. Science 333, 425. Hegi, M.E., Diserens, A.C., Gorlia, T., Hamou, M.F., de Tribolet, N., Weller, M., Kros, J.M., Hainfellner, J.A., Mason, W., Mariani, L., et al. (2005). MGMT gene silencing and benet from temozolomide in glioblastoma. N. Engl. J. Med. 352, 9971003.
Herman, J.G., Umar, A., Polyak, K., Graff, J.R., Ahuja, N., Issa, J.P., Markowitz, S., Willson, J.K., Hamilton, S.R., Kinzler, K.W., et al. (1998). Incidence and functional consequences of hMLH1 promoter hypermethylation in colorectal carcinoma. Proc. Natl. Acad. Sci. USA 95, 68706875. Hinoue, T., Weisenberger, D.J., Pan, F., Campan, M., Kim, M., Young, J., Whitehall, V.L., Leggett, B.A., and Laird, P.W. (2009). Analysis of the association between CIMP and BRAF in colorectal cancer by DNA methylation proling. PLoS ONE 4, e8357. Hinoue, T., Weisenberger, D.J., Lange, C.P., Shen, H., Byun, H.M., Van Den Berg, D., Malik, S., Pan, F., Noushmehr, H., van Dijk, C.M., et al. (2012). Genome-scale analysis of aberrant DNA methylation in colorectal cancer. Genome Res. 22, 271282. Hitchins, M.P., Rapkins, R.W., Kwok, C.T., Srivastava, S., Wong, J.J., Khachigian, L.M., Polly, P., Goldblatt, J., and Ward, R.L. (2011). Dominantly inherited constitutional epigenetic silencing of MLH1 in a cancer-affected family is linked to a single nucleotide variant within the 5UTR. Cancer Cell 20, 200213. Hodgkinson, A., and Eyre-Walker, A. (2011). Variation in the mutation rate across mammalian genomes. Nat. Rev. Genet. 12, 756766. Hodis, E., Watson, I.R., Kryukov, G.V., Arold, S.T., Imielinski, M., Theurillat, J.P., Nickerson, E., Auclair, D., Li, L., Place, C., et al. (2012). A landscape of driver mutations in melanoma. Cell 150, 251263. Holmfeldt, L., Wei, L., Diaz-Flores, E., Walsh, M., Zhang, J., Ding, L., PayneTurner, D., Churchman, M., Andersson, A., Chen, S.C., et al. (2013). The genomic landscape of hypodiploid acute lymphoblastic leukemia. Nat. Genet. 45, 242252. Hon, G.C., Hawkins, R.D., Caballero, O.L., Lo, C., Lister, R., Pelizzola, M., Valsesia, A., Ye, Z., Kuan, S., Edsall, L.E., et al. (2012). Global DNA hypomethylation coupled to repressive chromatin domain formation and gene silencing in breast cancer. Genome Res. 22, 246258. Hussin, J., Sinnett, D., Casals, F., Idaghdour, Y., Bruat, V., Saillour, V., Healy, J., Grenier, J.C., de Malliard, T., Busche, S., et al. (2013). Rare allelic forms of PRDM9 associated with childhood leukemogenesis. Genome Res. 23, 419 430. Hyland, E.M., Molina, H., Poorey, K., Jie, C., Xie, Z., Dai, J., Qian, J., Bekiranov, S., Auble, D.T., Pandey, A., and Boeke, J.D. (2011). An evolutionarily young lysine residue in histone H3 attenuates transcriptional output in Saccharomyces cerevisiae. Genes Dev. 25, 13061319. Imielinski, M., Berger, A.H., Hammerman, P.S., Hernandez, B., Pugh, T.J., Hodis, E., Cho, J., Suh, J., Capelletti, M., Sivachenko, A., et al. (2012). Mapping the hallmarks of lung adenocarcinoma with massively parallel sequencing. Cell 150, 11071120. Inthal, A., Zeitlhofer, P., Zeginigg, M., Morak, M., Grausenburger, R., Fron mayer, R. kova, E., Fahrner, B., Mann, G., Haas, O.A., and Panzer-Gru (2012). CREBBP HAT domain mutations prevail in relapse cases of high hyperdiploid childhood acute lymphoblastic leukemia. Leukemia 26, 17971803. Irizarry, R.A., Ladd-Acosta, C., Wen, B., Wu, Z., Montano, C., Onyango, P., Cui, H., Gabo, K., Rongione, M., Webster, M., et al. (2009). The human colon cancer methylome shows similar hypo- and hypermethylation at conserved tissue-specic CpG island shores. Nat. Genet. 41, 178186. Ito, S., Shen, L., Dai, Q., Wu, S.C., Collins, L.B., Swenberg, J.A., He, C., and Zhang, Y. (2011). Tet proteins can convert 5-methylcytosine to 5-formylcytosine and 5-carboxylcytosine. Science 333, 13001303. Jiao, Y., Shi, C., Edil, B.H., de Wilde, R.F., Klimstra, D.S., Maitra, A., Schulick, R.D., Tang, L.H., Wolfgang, C.L., Choti, M.A., et al. (2011). DAXX/ATRX, MEN1, and mTOR pathway genes are frequently altered in pancreatic neuroendocrine tumors. Science 331, 11991203. Jin, C., Zang, C., Wei, G., Cui, K., Peng, W., Zhao, K., and Felsenfeld, G. (2009). H3.3/H2A.Z double variant-containing nucleosomes mark nucleosome-free regions of active promoters and other regulatory regions. Nat. Genet. 41, 941945. Johnson, D.G., and Dent, S.Y.R. (2013). Chromatin: receiver and quarterback for cellular signals. Cell 152, 685689.
Jones, P.A. (2012). Functions of DNA methylation: islands, start sites, gene bodies and beyond. Nat. Rev. Genet. 13, 484492. Jones, P.A., and Laird, P.W. (1999). Cancer epigenetics comes of age. Nat. Genet. 21, 163167. Jones, P.A., and Baylin, S.B. (2002). The fundamental role of epigenetic events in cancer. Nat. Rev. Genet. 3, 415428. Jones, P.A., and Liang, G. (2009). Rethinking how DNA methylation patterns are maintained. Nat. Rev. Genet. 10, 805811. Jones, S., Wang, T.L., Shih, IeM., Mao, T.L., Nakayama, K., Roden, R., Glas, R., Slamon, D., Diaz, L.A., Jr., Vogelstein, B., et al. (2010). Frequent mutations of chromatin remodeling gene ARID1A in ovarian clear cell carcinoma. Science 330, 228231. ger, N., Kool, M., Zichner, T., Hutter, B., Sultan, M., Cho, Y.J., Jones, D.T., Ja tz, A.M., et al. (2012). Dissecting the genomic Pugh, T.J., Hovestadt, V., Stu complexity underlying medulloblastoma. Nature 488, 100105. Kamakaka, R.T., and Biggins, S. (2005). Histone variants: deviants? Genes Dev. 19, 295310. Kerkel, K., Spadola, A., Yuan, E., Kosek, J., Jiang, L., Hod, E., Li, K., Murty, V.V., Schupf, N., Vilain, E., et al. (2008). Genomic surveys by methylationsensitive SNP analysis identify sequence-dependent allele-specic DNA methylation. Nat. Genet. 40, 904908. Kim, M., Trinh, B.N., Long, T.I., Oghamian, S., and Laird, P.W. (2004). Dnmt1 deciency leads to enhanced microsatellite instability in mouse embryonic stem cells. Nucleic Acids Res. 32, 57425749. Kim, T.H., Abdullaev, Z.K., Smith, A.D., Ching, K.A., Loukinov, D.I., Green, R.D., Zhang, M.Q., Lobanenkov, V.V., and Ren, B. (2007). Analysis of the vertebrate insulator protein CTCF-binding sites in the human genome. Cell 128, 12311245. Ko, M., Huang, Y., Jankowska, A.M., Pape, U.J., Tahiliani, M., Bandukwala, H.S., An, J., Lamperti, E.D., Koh, K.P., Ganetzky, R., et al. (2010). Impaired hydroxylation of 5-methylcytosine in myeloid cancers with mutant TET2. Nature 468, 839843. Kondo, Y., Shen, L., Cheng, A.S., Ahmed, S., Boumber, Y., Charo, C., Yamochi, T., Urano, T., Furukawa, K., Kwabi-Addo, B., et al. (2008). Gene silencing in cancer by histone H3 lysine 27 trimethylation independent of promoter DNA methylation. Nat. Genet. 40, 741750. Ku, M., Koche, R.P., Rheinbay, E., Mendenhall, E.M., Endoh, M., Mikkelsen, T.S., Presser, A., Nusbaum, C., Xie, X., Chi, A.S., et al. (2008). Genomewide analysis of PRC1 and PRC2 occupancy identies two classes of bivalent domains. PLoS Genet. 4, e1000242. Lai, A.Y., and Wade, P.A. (2011). Cancer biology and NuRD: a multifaceted chromatin remodelling complex. Nat. Rev. Cancer 11, 588596. Laird, P.W. (2010). Principles and challenges of genomewide DNA methylation analysis. Nat. Rev. Genet. 11, 191203. Laird, P.W., Jackson-Grusby, L., Fazeli, A., Dickinson, S.L., Jung, W.E., Li, E., Weinberg, R.A., and Jaenisch, R. (1995). Suppression of intestinal neoplasia by DNA hypomethylation. Cell 81, 197205. Langemeijer, S.M., Kuiper, R.P., Berends, M., Knops, R., Aslanyan, M.G., Massop, M., Stevens-Linders, E., van Hoogen, P., van Kessel, A.G., Raymakers, R.A., et al. (2009). Acquired mutations in TET2 are common in myelodysplastic syndromes. Nat. Genet. 41, 838842. Le Gallo, M., OHara, A.J., Rudd, M.L., Urick, M.E., Hansen, N.F., ONeil, N.J., Price, J.C., Zhang, S., England, B.M., Godwin, A.K., et al.; NIH Intramural Sequencing Center (NISC) Comparative Sequencing Program. (2012). Exome sequencing of serous endometrial tumors identies recurrent somatic mutations in chromatin-remodeling and ubiquitin ligase complex genes. Nat. Genet. 44, 13101315. Lee, J.T. (2012). Epigenetic regulation by long noncoding RNAs. Science 338, 14351439. Lee, E., Iskow, R., Yang, L., Gokcumen, O., Haseley, P., Luquette, L.J., 3rd, Lohr, J.G., Harris, C.C., Ding, L., Wilson, R.K., et al.; Cancer Genome Atlas Research Network. (2012). Landscape of somatic retrotransposition in human cancers. Science 337, 967971.
Ley, T.J., Ding, L., Walter, M.J., McLellan, M.D., Lamprecht, T., Larson, D.E., Kandoth, C., Payton, J.E., Baty, J., Welch, J., et al. (2010). DNMT3A mutations in acute myeloid leukemia. N. Engl. J. Med. 363, 24242433. Li, J., Harris, R.A., Cheung, S.W., Coarfa, C., Jeong, M., Goodell, M.A., White, L.D., Patel, A., Kang, S.H., Shaw, C., et al. (2012). Genomic hypomethylation in the human germline associates with selective structural mutability in the human genome. PLoS Genet. 8, e1002692. beler, D. Lienert, F., Wirbelauer, C., Som, I., Dean, A., Mohn, F., and Schu (2011). Identication of genetic elements that autonomously determine DNA methylation states. Nat. Genet. 43, 10911097. Ligtenberg, M.J., Kuiper, R.P., Chan, T.L., Goossens, M., Hebeda, K.M., Voorendt, M., Lee, T.Y., Bodmer, D., Hoenselaar, E., Hendriks-Cornelissen, S.J., et al. (2009). Heritable somatic methylation and inactivation of MSH2 in families with Lynch syndrome due to deletion of the 30 exons of TACSTD1. Nat. Genet. 41, 112117. Lister, R., Pelizzola, M., Dowen, R.H., Hawkins, R.D., Hon, G., Tonti-Filippini, J., Nery, J.R., Lee, L., Ye, Z., Ngo, Q.M., et al. (2009). Human DNA methylomes at base resolution show widespread epigenomic differences. Nature 462, 315322. Liu, G., Bollig-Fischer, A., Kreike, B., van de Vijver, M.J., Abrams, J., Ethier, S.P., and Yang, Z.Q. (2009). Genomic amplication and oncogenic properties of the GASC1 histone demethylase gene in breast cancer. Oncogene 28, 44914500. Lohr, J.G., Stojanov, P., Lawrence, M.S., Auclair, D., Chapuy, B., Sougnez, C., Cruz-Gordillo, P., Knoechel, B., Asmann, Y.W., Slager, S.L., et al. (2012). Discovery and prioritization of somatic mutations in diffuse large B-cell lymphoma (DLBCL) by whole-exome sequencing. Proc. Natl. Acad. Sci. USA 109, 38793884. Loughery, J.E., Dunne, P.D., ONeill, K.M., Meehan, R.R., McDaid, J.R., and Walsh, C.P. (2011). DNMT1 deciency triggers mismatch repair defects in human cells through depletion of repair protein levels in a process involving the DNA damage response. Hum. Mol. Genet. 20, 32413255. Love, C., Sun, Z., Jima, D., Li, G., Zhang, J., Miles, R., Richards, K.L., Dunphy, C.H., Choi, W.W., Srivastava, G., et al. (2012). The genetic landscape of mutations in Burkitt lymphoma. Nat. Genet. 44, 13211325. Lu, C., Ward, P.S., Kapoor, G.S., Rohle, D., Turcan, S., Abdel-Wahab, O., Edwards, C.R., Khanin, R., Figueroa, M.E., Melnick, A., et al. (2012). IDH mutation impairs histone demethylation and results in a block to cell differentiation. Nature 483, 474478. Meuleman, W., Peric-Hupkes, D., Kind, J., Beaudry, J.B., Pagie, L., Kellis, M., Reinders, M., Wessels, L., and van Steensel, B. (2013). Constitutive nuclear lamina-genome interactions are highly conserved and associated with A/Trich sequence. Genome Res. 23, 270280. Mills, A.A. (2010). Throwing the cancer switch: reciprocal roles of polycomb and trithorax proteins. Nat. Rev. Cancer 10, 669682. Minucci, S., and Pelicci, P.G. (2006). Histone deacetylase inhibitors and the promise of epigenetic (and more) treatments for cancer. Nat. Rev. Cancer 6, 3851. Mohn, F., Weber, M., Rebhan, M., Roloff, T.C., Richter, J., Stadler, M.B., Bibel, beler, D. (2008). Lineage-specic polycomb targets and de novo M., and Schu DNA methylation dene restriction and potential of neuronal progenitors. Mol. Cell 30, 755766. Morin, R.D., Johnson, N.A., Severson, T.M., Mungall, A.J., An, J., Goya, R., Paul, J.E., Boyle, M., Woolcock, B.W., Kuchenbauer, F., et al. (2010). Somatic mutations altering EZH2 (Tyr641) in follicular and diffuse large B-cell lymphomas of germinal-center origin. Nat. Genet. 42, 181185. Morin, R.D., Mendez-Lago, M., Mungall, A.J., Goya, R., Mungall, K.L., Corbett, R.D., Johnson, N.A., Severson, T.M., Chiu, R., Field, M., et al. (2011). Frequent mutation of histone-modifying genes in non-Hodgkin lymphoma. Nature 476, 298303. Mortusewicz, O., Schermelleh, L., Walter, J., Cardoso, M.C., and Leonhardt, H. (2005). Recruitment of DNA methyltransferase I to DNA repair sites. Proc. Natl. Acad. Sci. USA 102, 89058909.
Mullighan, C.G., Zhang, J., Kasper, L.H., Lerach, S., Payne-Turner, D., Phillips, L.A., Heatley, S.L., Holmfeldt, L., Collins-Underwood, J.R., Ma, J., et al. (2011). CREBBP mutations in relapsed acute lymphoblastic leukaemia. Nature 471, 235239. Mummert, S.K., Lobanenkov, V.A., and Feinberg, A.P. (2005). Association of chromosome arm 16q loss with loss of imprinting of insulin-like growth factor-II in Wilms tumor. Genes Chromosomes Cancer 43, 155161. te , J., and Kutateladze, T.G. (2012). Musselman, C.A., Lalonde, M.E., Co Perceiving the epigenetic landscape through histone readers. Nat. Struct. Mol. Biol. 19, 12181227. nnisNikoloski, G., Langemeijer, S.M., Kuiper, R.P., Knops, R., Massop, M., To sen, E.R., van der Heijden, A., Scheele, T.N., Vandenberghe, P., de Witte, T., et al. (2010). Somatic mutations of the histone methyltransferase gene EZH2 in myelodysplastic syndromes. Nat. Genet. 42, 665667. Noushmehr, H., Weisenberger, D.J., Diefes, K., Phillips, H.S., Pujara, K., Berman, B.P., Pan, F., Pelloski, C.E., Sulman, E.P., Bhat, K.P., et al.; Cancer Genome Atlas Research Network. (2010). Identication of a CpG island methylator phenotype that denes a distinct subgroup of glioma. Cancer Cell 17, 510522. Ntziachristos, P., Tsirigos, A., Van Vlierberghe, P., Nedjic, J., Trimarchi, T., Flaherty, M.S., Ferres-Marco, D., da Ros, V., Tang, Z., Siegle, J., et al. (2012). Genetic inactivation of the polycomb repressive complex 2 in T cell acute lymphoblastic leukemia. Nat. Med. 18, 298301. Ohm, J.E., McGarvey, K.M., Yu, X., Cheng, L., Schuebel, K.E., Cope, L., Mohammad, H.P., Chen, W., Daniel, V.C., Yu, W., et al. (2007). A stem celllike chromatin pattern may predispose tumor suppressor genes to DNA hypermethylation and heritable silencing. Nat. Genet. 39, 237242. Okano, M., Bell, D.W., Haber, D.A., and Li, E. (1999). DNA methyltransferases Dnmt3a and Dnmt3b are essential for de novo methylation and mammalian development. Cell 99, 247257. Olivier, M., Hollstein, M., and Hainaut, P. (2010). TP53 mutations in human cancers: origins, consequences, and clinical use. Cold Spring Harb. Perspect. Biol. 2, a001008. Papamichos-Chronakis, M., and Peterson, C.L. (2013). Chromatin and the genome integrity network. Nat. Rev. Genet. 14, 6275. Parsons, D.W., Li, M., Zhang, X., Jones, S., Leary, R.J., Lin, J.C., Boca, S.M., Carter, H., Samayoa, J., Bettegowda, C., et al. (2011). The genetic landscape of the childhood cancer medulloblastoma. Science 331, 435439. Pasqualucci, L., Trifonov, V., Fabbri, G., Ma, J., Rossi, D., Chiarenza, A., Wells, V.A., Grunn, A., Messina, M., Elliot, O., et al. (2011). Analysis of the coding genome of diffuse large B-cell lymphoma. Nat. Genet. 43, 830837. Petruk, S., Sedkov, Y., Johnston, D.M., Hodgson, J.W., Black, K.L., Kovermann, S.K., Beck, S., Canaani, E., Brock, H.W., and Mazo, A. (2012). TrxG and PcG proteins but not methylated histones remain associated with DNA through replication. Cell 150, 922933. Prokhortchouk, A., Sansom, O., Selfridge, J., Caballero, I.M., Salozhin, S., Aithozhina, D., Cerchietti, L., Meng, F.G., Augenlicht, L.H., Mariadason, J.M., et al. (2006). Kaiso-decient mice show resistance to intestinal cancer. Mol. Cell. Biol. 26, 199208. Pugh, T.J., Weeraratne, S.D., Archer, T.C., Pomeranz Krummel, D.A., Auclair, D., Bochicchio, J., Carneiro, M.O., Carter, S.L., Cibulskis, K., Erlich, R.L., et al. (2012). Medulloblastoma exome sequencing uncovers subtype-specic somatic mutations. Nature 488, 106110. , L., Della Valle, V., Lopez, C.K., Plo, I., Wagner-Ballon, Quivoron, C., Couronne O., Do Cruzeiro, M., Delhommeau, F., Arnulf, B., Stern, M.H., et al. (2011). TET2 inactivation results in pleiotropic hematopoietic abnormalities in mouse and is a recurrent event during human lymphomagenesis. Cancer Cell 20, 2538. Raddatz, G., Gao, Q., Bender, S., Jaenisch, R., and Lyko, F. (2012). Dnmt3a protects active chromosome domains against cancer-associated hypomethylation. PLoS Genet. 8, e1003146. Ram, O., Goren, A., Amit, I., Shoresh, N., Yosef, N., Ernst, J., Kellis, M., Gymrek, M., Issner, R., Coyne, M., et al. (2011). Combinatorial patterning of
chromatin regulators uncovered by genome-wide location analysis in human cells. Cell 147, 16281639. Robinson, G., Parker, M., Kranenburg, T.A., Lu, C., Chen, X., Ding, L., Phoenix, T.N., Hedlund, E., Wei, L., Zhu, X., et al. (2012). Novel mutations target distinct subgroups of medulloblastoma. Nature 488, 4348. Rui, L., Emre, N.C., Kruhlak, M.J., Chung, H.J., Steidl, C., Slack, G., Wright, G.W., Lenz, G., Ngo, V.N., Shaffer, A.L., et al. (2010). Cooperative epigenetic modulation by cancer amplicon genes. Cancer Cell 18, 590605. Sandoval, J., and Esteller, M. (2012). Cancer epigenomics: beyond genomics. Curr. Opin. Genet. Dev. 22, 5055. Sansom, O.J., Berger, J., Bishop, S.M., Hendrich, B., Bird, A., and Clarke, A.R. (2003). Deciency of Mbd2 suppresses intestinal tumorigenesis. Nat. Genet. 34, 145147. Schlesinger, Y., Straussman, R., Keshet, I., Farkash, S., Hecht, M., Zimmerman, J., Eden, E., Yakhini, Z., Ben-Shushan, E., Reubinoff, B.E., et al. (2007). Polycomb-mediated methylation on Lys27 of histone H3 pre-marks genes for de novo methylation in cancer. Nat. Genet. 39, 232236. hwald, M.C., Gesk, S., Hasselblatt, M., Jeibmann, A., Schneppenheim, R., Fru Kordes, U., Kreuz, M., Leuschner, I., Martin Subero, J.I., Obser, T., et al. (2010). Germline nonsense mutation and somatic inactivation of SMARCA4/BRG1 in a family with rhabdoid tumor predisposition syndrome. Am. J. Hum. Genet. 86, 279284. ckler, B., and Lehner, B. (2012). Chromatin organization is a major Schuster-Bo inuence on regional mutation rates in human cancer cells. Nature 488, 504507. Schwartzentruber, J., Korshunov, A., Liu, X.Y., Jones, D.T., Pfaff, E., Jacob, K., njes, M., et al. (2012). Driver Sturm, D., Fontebasso, A.M., Quang, D.A., To mutations in histone H3.3 and chromatin remodelling genes in paediatric glioblastoma. Nature 482, 226231. Seligson, D.B., Horvath, S., Shi, T., Yu, H., Tze, S., Grunstein, M., and Kurdistani, S.K. (2005). Global histone modication patterns predict risk of prostate cancer recurrence. Nature 435, 12621266. Sharma, S., De Carvalho, D.D., Jeong, S., Jones, P.A., and Liang, G. (2011). Nucleosomes containing methylated DNA stabilize DNA methyltransferases 3A/3B and ensure faithful epigenetic inheritance. PLoS Genet. 7, e1001286. Simon, C., Chagraoui, J., Krosl, J., Gendron, P., Wilhelm, B., Lemieux, S., Boucher, G., Chagnon, P., Drouin, S., Lambert, R., et al. (2012). A key role for EZH2 and associated genes in mouse and human adult T-cell acute leukemia. Genes Dev. 26, 651656. Simpkins, S.B., Bocker, T., Swisher, E.M., Mutch, D.G., Gersell, D.J., Kovatich, A.J., Palazzo, J.P., Fishel, R., and Goodfellow, P.J. (1999). MLH1 promoter methylation and gene silencing is the primary cause of microsatellite instability in sporadic endometrial cancers. Hum. Mol. Genet. 8, 661666. Smith, M.J., OSullivan, J., Bhaskar, S.S., Hadeld, K.D., Poke, G., Caird, J., Sharif, S., Eccles, D., Fitzpatrick, D., Rawluk, D., et al. (2013). Loss-of-function mutations in SMARCE1 cause an inherited disorder of multiple spinal meningiomas. Nat. Genet. 45, 295298. Sproul, D., Kitchen, R.R., Nestor, C.E., Dixon, J.M., Sims, A.H., Harrison, D.J., Ramsahoye, B.H., and Meehan, R.R. (2012). Tissue of origin determines cancer-associated CpG island promoter hypermethylation patterns. Genome Biol. 13, R84. Stransky, N., Egloff, A.M., Tward, A.D., Kostic, A.D., Cibulskis, K., Sivachenko, A., Kryukov, G.V., Lawrence, M.S., Sougnez, C., McKenna, A., et al. (2011). The mutational landscape of head and neck squamous cell carcinoma. Science 333, 11571160. Sturm, D., Witt, H., Hovestadt, V., Khuong-Quang, D.A., Jones, D.T., Koner njes, M., Sill, M., Bender, S., et al. (2012). Hotspot mutamann, C., Pfaff, E., To tions in H3F3A and IDH1 dene distinct epigenetic and biological subgroups of glioblastoma. Cancer Cell 22, 425437. Tan, J., Jones, M., Koseki, H., Nakayama, M., Muntean, A.G., Maillard, I., and Hess, J.L. (2011a). CBX8, a polycomb group protein, is essential for MLL-AF9induced leukemogenesis. Cancer Cell 20, 563575.
Tan, M., Luo, H., Lee, S., Jin, F., Yang, J.S., Montellier, E., Buchou, T., Cheng, Z., Rousseaux, S., Rajagopal, N., et al. (2011b). Identication of 67 histone marks and histone lysine crotonylation as a new type of histone modication. Cell 146, 10161028. Tanay, A., ODonnell, A.H., Damelin, M., and Bestor, T.H. (2007). Hyperconserved CpG domains underlie Polycomb-binding sites. Proc. Natl. Acad. Sci. USA 104, 55215526. Teschendorff, A.E., Menon, U., Gentry-Maharaj, A., Ramus, S.J., Weisenberger, D.J., Shen, H., Campan, M., Noushmehr, H., Bell, C.G., Maxwell, A.P., et al. (2010). Age-dependent DNA methylation of genes that are suppressed in stem cells is a hallmark of cancer. Genome Res. 20, 440446. Testa, J.R., Cheung, M., Pei, J., Below, J.E., Tan, Y., Sementino, E., Cox, N.J., Dogan, A.U., Pass, H.I., Trusa, S., et al. (2011). Germline BAP1 mutations predispose to malignant mesothelioma. Nat. Genet. 43, 10221025. Toyota, M., and Suzuki, H. (2010). Epigenetic drivers of genetic alterations. Adv. Genet. 70, 309323. Toyota, M., Ahuja, N., Ohe-Toyota, M., Herman, J.G., Baylin, S.B., and Issa, J.P. (1999). CpG island methylator phenotype in colorectal cancer. Proc. Natl. Acad. Sci. USA 96, 86818686. Turcan, S., Rohle, D., Goenka, A., Walsh, L.A., Fang, F., Yilmaz, E., Campos, C., Fabius, A.W., Lu, C., Ward, P.S., et al. (2012). IDH1 mutation is sufcient to establish the glioma hypermethylator phenotype. Nature 483, 479483. Valouev, A., Johnson, S.M., Boyd, S.D., Smith, C.L., Fire, A.Z., and Sidow, A. (2011). Determinants of nucleosome organization in primary human cells. Nature 474, 516520. van Haaften, G., Dalgliesh, G.L., Davies, H., Chen, L., Bignell, G., Greenman, C., Edkins, S., Hardy, C., OMeara, S., Teague, J., et al. (2009). Somatic mutations of the histone H3K27 demethylase gene UTX in human cancer. Nat. Genet. 41, 521523. Van Vlierberghe, P., Palomero, T., Khiabanian, H., Van der Meulen, J., Castillo, , J., Gonza lez-Garc a, S., Toribio, M., Van Roy, N., De Moerloose, B., Philippe M.L., et al. (2010). PHF6 mutations in T-cell acute lymphoblastic leukemia. Nat. Genet. 42, 338342. Van Vlierberghe, P., Patel, J., Abdel-Wahab, O., Lobry, C., Hedvat, C.V., Balbin, M., Nicolas, C., Payer, A.R., Fernandez, H.F., Tallman, M.S., et al. (2011). PHF6 mutations in adult acute myeloid leukemia. Leukemia 25, 130134. Varambally, S., Cao, Q., Mani, R.S., Shankar, S., Wang, X., Ateeq, B., Laxman, B., Cao, X., Jing, X., Ramnarayanan, K., et al. (2008). Genomic loss of microRNA-101 leads to overexpression of histone methyltransferase EZH2 in cancer. Science 322, 16951699. Varela, I., Tarpey, P., Raine, K., Huang, D., Ong, C.K., Stephens, P., Davies, H., Jones, D., Lin, M.L., Teague, J., et al. (2011). Exome sequencing identies frequent mutation of the SWI/SNF complex gene PBRM1 in renal carcinoma. Nature 469, 539542. venet, N., Lange, J., Rousseau-Merck, M.F., Ambros, P., Versteege, I., Se Handgretinger, R., Aurias, A., and Delattre, O. (1998). Truncating mutations of hSNF5/INI1 in aggressive paediatric cancer. Nature 394, 203206. Wang, Z.Y., and Chen, Z. (2008). Acute promyelocytic leukemia: from highly fatal to highly curable. Blood 111, 25052515. Wang, R.Y., Kuo, K.C., Gehrke, C.W., Huang, L.H., and Ehrlich, M. (1982). Heat- and alkali-induced deamination of 5-methylcytosine and cytosine residues in DNA. Biochim. Biophys. Acta 697, 371377. Weisenberger, D.J., Siegmund, K.D., Campan, M., Young, J., Long, T.I., Faasse, M.A., Kang, G.H., Widschwendter, M., Weener, D., Buchanan, D., et al. (2006). CpG island methylator phenotype underlies sporadic microsatellite instability and is tightly associated with BRAF mutation in colorectal cancer. Nat. Genet. 38, 787793. Weissmann, S., Alpermann, T., Grossmann, V., Kowarsch, A., Nadarajah, N., Eder, C., Dicker, F., Fasan, A., Haferlach, C., Haferlach, T., et al. (2012). Landscape of TET2 mutations in acute myeloid leukemia. Leukemia 26, 934942.
Widschwendter, M., Fiegl, H., Egle, D., Mueller-Holzner, E., Spizzo, G., Marth, C., Weisenberger, D.J., Campan, M., Young, J., Jacobs, I., and Laird, P.W. (2007). Epigenetic stem cell signature in cancer. Nat. Genet. 39, 157158. Wiegand, K.C., Shah, S.P., Al-Agha, O.M., Zhao, Y., Tse, K., Zeng, T., Senz, J., McConechy, M.K., Anglesio, M.S., Kalloger, S.E., et al. (2010). ARID1A mutations in endometriosis-associated ovarian carcinomas. N. Engl. J. Med. 363, 15321543. Wiesner, T., Obenauf, A.C., Murali, R., Fried, I., Griewank, K.G., Ulz, P., Windpassinger, C., Wackernagel, W., Loy, S., Wolf, I., et al. (2011). Germline mutations in BAP1 predispose to melanocytic tumors. Nat. Genet. 43, 10181021. Williams, K., Christensen, J., and Helin, K. (2012). DNA methylation: TET proteins-guardians of CpG islands? EMBO Rep. 13, 2835. Wilson, B.G., and Roberts, C.W. (2011). SWI/SNF nucleosome remodellers and cancer. Nat. Rev. Cancer 11, 481492. Wu, G., Broniscer, A., McEachron, T.A., Lu, C., Paugh, B.S., Becksfort, J., Qu, C., Ding, L., Huether, R., Parker, M., et al.; St. Jude Childrens Research HospitalWashington University Pediatric Cancer Genome Project. (2012). Somatic histone H3 alterations in pediatric diffuse intrinsic pontine gliomas and non-brainstem glioblastomas. Nat. Genet. 44, 251253. Xu, W., Yang, H., Liu, Y., Yang, Y., Wang, P., Kim, S.H., Ito, S., Yang, C., Wang, P., Xiao, M.T., et al. (2011). Oncometabolite 2-hydroxyglutarate is a competitive inhibitor of a-ketoglutarate-dependent dioxygenases. Cancer Cell 19, 1730. Xu, K., Wu, Z.J., Groner, A.C., He, H.H., Cai, C., Lis, R.T., Wu, X., Stack, E.C., Loda, M., Liu, T., et al. (2012). EZH2 oncogenic activity in castration-resistant prostate cancer cells is Polycomb-independent. Science 338, 14651469. Yamamoto, E., Suzuki, H., Yamano, H.O., Maruyama, R., Nojima, M., Kamimae, S., Sawada, T., Ashida, M., Yoshikawa, K., Kimura, T., et al. (2012). Molecular dissection of premalignant colorectal lesions reveals early onset of the CpG island methylator phenotype. Am. J. Pathol. 181, 18471861. Yamashita, K., Dai, T., Dai, Y., Yamamoto, F., and Perucho, M. (2003). Genetics supersedes epigenetics in colon cancer phenotype. Cancer Cell 4, 121131. Yamazaki, J., Estecio, M.R., Lu, Y., Long, H., Malouf, G.G., Graber, D., Huo, Y., Ramagli, L., Liang, S., Kornblau, S.M., et al. (2013). The epigenome of AML stem and progenitor cells. Epigenetics 8, 92104. Yan, X.J., Xu, J., Gu, Z.H., Pan, C.M., Lu, G., Shen, Y., Shi, J.Y., Zhu, Y.M., Tang, L., Zhang, X.W., et al. (2011). Exome sequencing identies somatic mutations of DNA methyltransferase gene DNMT3A in acute monocytic leukemia. Nat. Genet. 43, 309315. Yang, X.J. (2004). The diverse superfamily of lysine acetyltransferases and their roles in leukemia and other diseases. Nucleic Acids Res. 32, 959976. Yokoyama, A., and Cleary, M.L. (2008). Menin critically links MLL proteins with LEDGF on cancer-associated target genes. Cancer Cell 14, 3646. You, J.S., and Jones, P.A. (2012). Cancer genetics and epigenetics: two sides of the same coin? Cancer Cell 22, 920. Zeisig, B.B., Kulasekararaj, A.G., Mufti, G.J., and So, C.W. (2012). SnapShot: acute myeloid leukemia. Cancer Cell 22, 698. Zhang, J., Ding, L., Holmfeldt, L., Wu, G., Heatley, S.L., Payne-Turner, D., Easton, J., Chen, X., Wang, J., Rusch, M., et al. (2012). The genetic basis of early T-cell precursor acute lymphoblastic leukaemia. Nature 481, 157163. Zhu, J., Adli, M., Zou, J.Y., Verstappen, G., Coyne, M., Zhang, X., Durham, T., Miri, M., Deshpande, V., De Jager, P.L., et al. (2013). Genome-wide chromatin state transitions associated with developmental and environmental cues. Cell 152, 642654. -Regi, R., Gaffney, D.J., Epstein, C.B., Zullo, J.M., Demarco, I.A., Pique Spooner, C.J., Luperchio, T.R., Bernstein, B.E., Pritchard, J.K., Reddy, K.L., and Singh, H. (2012). DNA sequence-dependent compartmentalization and silencing of chromatin at the nuclear lamina. Cell 149, 14741487.
Review
Inuence of Metabolism on Epigenetics and Disease
William G. Kaelin, Jr.1,2,* and Steven L. McKnight3,*
Leading Edge
1Department of Medical Oncology, Dana-Farber Cancer Institute and Brigham and Womens Hospital, Harvard Medical School, Boston, MA 02215, USA 2Howard Hughes Medical Institute, Chevy Chase, MD 20815, USA 3Department of Biochemistry, University of Texas Southwestern Medical Center, Dallas, TX 75390, USA *Correspondence: william_kaelin@dfci.harvard.edu (W.G.K.), steven.mcknight@utsouthwestern.edu (S.L.M.) http://dx.doi.org/10.1016/j.cell.2013.03.004
Chemical modications of histones and DNA, such as histone methylation, histone acetylation, and DNA methylation, play critical roles in epigenetic gene regulation. Many of the enzymes that add or remove such chemical modications are known, or might be suspected, to be sensitive to changes in intracellular metabolism. This knowledge provides a conceptual foundation for understanding how mutations in the metabolic enzymes SDH, FH, and IDH can result in cancer and, more broadly, for how alterations in metabolism and nutrition might contribute to disease. Here, we review literature pertinent to hypothetical connections between metabolic and epigenetic states in eukaryotic cells.
Introduction Central to the many denitions of epigenetics is the knowledge that genes contain regulatory information beyond their nucleotide sequences. This information can either be dynamic and transitory in nature or be relatively stable, capable of being passed on to somatic daughter cells, as occurs during lineage commitment, and in some cases to offspring via the germline, as occurs with parentally imprinted genes. The most thoroughly understood epigenetic mechanisms inuence gene expression and do so as a result of changes in chemical modications of the DNA (for example, methylation of CpG dinucleotides within gene promoters) or the physical accessibility of the DNA by virtue of its association with histones, nonhistone proteins, or noncoding RNAs (for example, XIST). The basic building block of chromatin is the nucleosome core particle, consisting of approximately 147 base pairs of DNA wrapped around a histone octamer that contains two copies each of histones 2A, 2B, 3, and 4. The tails of histones H3 and H4 are subject to a variety of posttranslational modications including acetylation, methylation, phosphorylation, sumoylation, and ubiquitylation. In general, histone acetylation is associated with a more open chromatin conguration (euchromatin) that is permissive for transcription. Histone deacetylation is usually associated with condensed, compacted chromatin (heterochromatin) and transcriptional repression. The positions of nucleosomes relative to the DNA strand also inuence which genes are capable of being transcribed and are regulated by chromatin remodeling complexes such as SWI/SNF complexes. Histone acetylation leads to an increased negative charge, which loosens the interaction between the histone and the negatively charged DNA. In addition, acetylated histones recruit specic chromatin-associated proteins that contain bromodomains. Histone methylation, by contrast, does not alter histone
charge but instead creates a docking site for chromatin-associated proteins that contain specic methyl histone-binding domains, such as plant homeodomain (PHD) domains, tudor domains, or chromodomains. These chromatin-associated reader proteins often recruit other proteins that contain additional chromatin-modifying activities (including writers and erasers that add or remove specic histone posttranslational modications, respectively). The consequences of histone methylation are inuenced by the specic histone residue that is modied, the number of methyl groups added (mono-, di-, or trimethylation), and other contextual factors. Methylation of H3K4, H3K36, and H3K79 is often associated with transcriptionally active euchromatin. By contrast, methylation of H3K9, H3K27, and H4K20 helps specify transcriptionally repressed heterochromatin. Histone methylation can also inuence DNA methylation and vice versa. Specic methyltransferase enzymes are involved in de novo and maintenance DNA methylation. Methylation of CpG dinucleotides in promoter regions typically inhibits transcription. DNA methylation tends to be a more stable modication than histone methylation but can undergo changes during embryogenesis and aging. It has been appreciated for many years that cancers can display global DNA hypomethylation while, at the same time, exhibiting hypermethylation of genomic regions responsible for the expression of tumor suppressor genes. Many enzymes that play important roles in epigenetic gene regulation utilize cosubstrates generated by cellular metabolism, thereby providing a potential link between nutrition, metabolism, and gene regulation. In this review, we describe examples of such enzymes as well as evidence that altered metabolism, through altered epigenetics, can contribute to disease. As this topic has recently also been reviewed by others (e.g., Dawson
Figure 1. Metabolism and Acetylation/Deacetylation

Histone acetylases use acetyl-CoA (Ac-CoA) as an acetyl donor, whose synthesis requires coenzyme A (CoA). Ac-CoA can be regenerated in chemical reactions involving pyruvate, citrate, acetate, and various amino acids such as threonine and by fatty acid beta oxidation. Deacetylation by Sirtuin family histone deacetylases requires NAD+, leading to the generation of O-acetylADP ribose and nicotinamide (NAM). NAD+ is produced from NMN (nicotinamide mononucleotide), which can be salvaged from NAM or produced de novo from tryptophan. For simplicity, enzymes catalyzing the various reactions are not shown.
and Kouzarides, 2012; Lu and Thompson, 2012; Teperino et al., 2010) we focus particular attention on pertinent studies that might have been overlooked. We have also purposely been provocative by raising questions and, in some cases, challenging existing dogma in the eld. Acetyl-CoA: Activated Acetate and Histone Acetylation Acetyl-CoA fuels the TCA cycle for the production of ATP under aerobic conditions and is a critical building block for cholesterol, lipids, amino acids, and other components required for cell growth. It was discovered as the activated form of acetate, so named because of its favorable energetic state for twocarbon donation in anabolic biochemistry (Lipmann and Kaplan, 1946). Acetyl-CoA is also the substrate used by histone acetyl transferase (HAT) enzymes to modify histone tails as an integral determinant of the epigenetic state of chromatin in eukaryotic cells (Figure 1) (Lee and Workman, 2007; Shahbazian and Grunstein, 2007). Acetyl-CoA Generation Acetyl-CoA can be produced through a variety of metabolic pathways, both catabolic and anabolic (Figure 1). Principal among these is the conversion of pyruvate into acetyl-CoA via the mitochondrial pyruvate dehydrogenase complex late during the oxidation of glucose and the b-oxidation of fatty acids. Given its polarity and relative chemical complexity, acetyl-CoA does not readily diffuse across membranes. As such, metazoan cells
have evolved the malate-citrate antiporter system to move mitochondrial citrate to the cytoplasm, where it can combine with ATP and CoA to be converted to acetyl-CoA and oxaloacetate via the ATP citrate lyase enzyme, thereby affording a cytoplasmic pool of acetyl-CoA for lipid biosynthesis. Acetyl-CoA can also be produced catabolically from threonine via a mitochondrial threonine dehydrogenase enzyme uniquely expressed in mouse embryonic stem cells; anabolically from acetate, ATP, and CoA via acetyl-CoA synthase enzymes localized in mitochondrial, cytoplasmic, or nuclear compartments of mammalian cells; and via an anaplerotic pathway through reductive carboxylation of a-ketoglutarate. Acetyl-CoA Fluctuation A recurring theme woven throughout this review asks the question of how the levels of consumable nutrients, enzyme substrates, and even molecular oxygen are sensed by cells as a means of adaptation. More specically, we focus on the possibility that gene expression is modulated via epigenetic modication of chromosomal proteins in accord with the abundance of essential metabolites. The activity of almost all enzymes involved in intermediary metabolism is regulated as a function of the abundance of both enzyme substrate and product. By contrast, many enzymes involved in key aspects of intracellular signaling are not regulated in this way. Take, for example, the hundreds of protein kinase enzymes that, using ATP as a substrate, modify target proteins by phosphorylation. With the exception of the adenosine monophosphate-regulated protein kinase (Hardie, 2011), almost no protein kinase enzymes are built to sense the level of cellular ATP in the context of their regulatory function. Unless a cell is in a deathly sick state of ATP under abundance, protein kinase enzymes are capable of functioning perfectly well irrespective of ATP levels. This obviously results from the fact that protein kinases are endowed with afnity for substrate that is considerably more avid than the ambient levels of intracellular ATP. Turning to enzymes involved in epigenetic regulation, starting with HATs, we ask the following: are we to consider them as being analogous to substrate-limited metabolic enzymes, or are they instead protein kinase-like in having evolved properties that shield themselves from uctuation in the level of intracellular acetyl-CoA? Before considering whether acetylase enzymes might be substrate regulated, it is rst important to ask whether the intracellular levels of acetyl-CoA uctuate in biological settings. Studies of two different eukaryotic cells have indeed given evidence of signicant uctuation in acetyl-CoA. For example, prototrophic (wild-type) strains of yeast grown in the nutrient-limiting environment of a chemostat spontaneously enter a synchronous and highly robust metabolic cycle (Klevecz et al., 2004; Tu et al., 2005; Tu and McKnight, 2006, 2009; Tu et al., 2007). Over a 45 hr cycle the cells rhythmically oscillate between oxidative and reductive growth (Figure 2). Mitochondrial respiration during the oxidative phase of this yeast metabolic cycle (YMC) helps accumulate appropriate levels of energetically valuable building blocks required for transition into a reductive, glycolytic phase wherein the cells commit to DNA synthesis and cell division. Acetyl-CoA levels uctuate dramatically as a function of the YMC. Acetyl-CoA levels peak at a 6-fold-higher level at the transition of the oxidative (Ox)
Figure 2. Evidence of Transient Acetylation of Histone H3 Only during the Oxidative Phase of the Yeast Metabolic Cycle
(A) Periodic uctuation in oxygen levels in a chemostat growing prototrophic yeast. The yeast metabolic cycle (YMC) is roughly 5 hours in duration and dened by sequentially repeating oxidative (Ox), reductive building (RB), and reductive charging (RC) metabolic phases (adapted from Tu et al., 2005). (B) Quantitative measurement of acetyl-CoA levels over the YMC reveal elevated levels of the metabolite during the Ox phase of the YMC. Western blot measurements of H3K9 acetylation over the YMC reveal dynamic acetylation temporally correlate with the peak abundance of acetyl-CoA. (C) ChIP-seq analysis of H3K9 acetylation on the promoter of the gene encoding the RPS7B ribosomal protein reveals modication limited to the Ox phase of the YMC. (D) Transcript abundance of the RPS7B mRNA peaks during the Ox phase of the YMC precisely when acetyl-CoA levels are of highest abundance and when the promoter of the RPS7B gene is modied by H3K9 acetylation.
phase to the reductive building (RB) phase relative to the metabolically quiescent, reductive charging (RC) phase of the YMC (Cai et al., 2011; Tu et al., 2007). One likely means by which intra58 Cell 153, March 28, 2013 2013 Elsevier Inc.
cellular levels of acetyl-CoA peak at the Ox/RB boundary is the coordinated induction of all enzymes required to convert ethanol into acetylaldehyde, acetylaldehyde into acetate, and acetate into acetyl-CoA at this precise temporal window of the YMC. In this way, ethanol fermented via the consumption of glucose during the RB phase of the YMC and accumulated in the extracellular reservoir of the chemostat can be retrieved and rebuilt into a valuable cellular building block. That the recycled hydrocarbon of ethanol enzymatically converted into acetyl-CoA might be directly relevant to the epigenetic state of yeast cells is strongly hinted by the fact that the terminal enzyme in the pathway, acetyl-CoA synthase, has been shown to be localized to the nucleus (Takahashi et al., 2006). Signicant uctuation in the abundance of acetyl-CoA has also been observed as a function of the differentiation of mouse embryonic stem cells (ESCs). Undifferentiated ESCs contain signicantly higher levels of acetyl-CoA than the embryoid body (EB) cells induced to differentiate by the combined withdrawal of leukemia-inhibitory factor (LIF) and application of retinoic acid (RA) (Wang et al., 2009a, 2011). The observed uctuation of acetyl-CoA levels correlates with dramatic changes in the expression of the threonine dehydrogenase (TDH) enzyme, which is rate limiting for the conversion of threonine into glycine and acetyl-CoA. Undifferentiated mouse ESCs express levels of TDH upward of 1,000-fold higher relative to any other source of mouse cells or tissues, and the gene encoding TDH is stringently repressed immediately upon induction of ESC differentiation in response to LIF withdrawal and RA administration. When undifferentiated ESCs are exposed to a specic chemical inhibitor of the TDH enzyme, intracellular levels of acetyl-CoA drop precipitously (Alexander et al., 2011). As will be discussed subsequently, it has been hypothesized that the ability of the TDH enzyme to convert threonine into glycine and acetyl-CoA may not only fuel mouse ESCs in a specialized manner but also help dictate an equally specialized epigenetic state. Control of Gene Expression by Acetyl-CoA Do the unusually high levels of acetyl-CoA present in undifferentiated mouse ESCs, or yeast cells poised at the Ox/RB boundary of the YMC, play a determinative role in epigenetic regulation of gene expression? Where this has only been hypothesized for mouse ESCs, data have been gathered to afrmatively answer this question in yeast cells (Figure 2). As described by Tu and colleagues (Cai et al., 2011; Cai and Tu, 2011, 2012), numerous acetylation marks on the K9, K14, K23, and K27 residues of histone H3 and the K5, K8, and K12 residues on histone H4 only appear over a 3045 min window corresponding exactly with the Ox/RB boundary that is coincident with the peak abundance of intracellular acetyl-CoA. The fact that these acetylation marks peak at the Ox/RB boundary gives evidence that this form of epigenetic regulation is unusually dynamic, consistent with reports that the half-life of histone acetylation may be as short as 3 min (Waterborg, 2002). Further DNA microarray and ChIP-seq experiments led to the identication of roughly 1,000 growth genes selectively acetylated and activated only when intracellular levels of acetyl-CoA peak (Cai et al., 2011, Tu et al., 2005). These include genes encoding ribosomal components, translation factors, and the regulatory D1 cyclin, corresponding precisely to the set of genes known to gate entry
of yeast cells into the cell division cycle (Jorgensen et al., 2002). The precision of temporal induction of these 1,000+ genes is astounding; the entire gene set is coordinately induced within a single-digit number of minutes within the 45 hr YMC (Rowicka et al., 2007). Acetyl-CoA and Histone Acetyltransferase Enzymes The GCN5 histone acetylase enzyme of the SAGA complex has been identied as the critical enzyme responsible for transient acetylation of growth genes at the Ox/RB boundary (Cai et al., 2011). This conclusion derives from ChIP-seq experiments showing the selective association of the SAGA complex with the promoters of the entire battery of growth genes only during the window of peak acetyl-CoA accumulation, along with genetic experiments wherein it has been demonstrated that catalytically active GCN5 is critically required for YMC oscillation (Cai et al., 2011). What properties of the SAGA complex and GCN5 might uniquely qualify it as an acetyl-CoA sensor? Several regulatory subunits of the SAGA complex are themselves transiently acetylated only during the Ox/RB window, raising the possibility that a complex pathway of allosteric regulation is at the heart of the sensing ability of SAGA (Cai et al., 2011). More simplistically, it is possible that GCN5 requires high levels of acetylCoA and that the enzyme is less active in other phases of the YMC relative to the Ox/RB window wherein acetyl-CoA levels peak. In this regard it may be notable that the off-rate of acetyl-CoA binding to the yeast GCN5 enzyme is more than an order of magnitude more rapid than that of human p300/CBP HAT enzyme, human GCN5, or Tetrahymena GCN5 (Langer et al., 2002). Cancer Connection The genes whose promoter regions and chromatin are differentially acetylated exactly when acetyl-CoA levels peak during the YMC encode precisely those protein and RNA products required to enable cell growth. This yeast growth gene battery matches closely with the genes induced by the c-Myc oncoprotein in mammalian cells (Ji et al., 2011), which have been reported to be codependent upon c-Myc and GCN5/SAGA (McMahon et al., 1998, 2000). This precisely orchestrated pattern of yeast growth gene induction in response to ambient levels of intracellular acetyl-CoA probably represents an evolutionarily ancient regulatory pathway allowing cells to properly link the commitment of cell growth and division to nutritional state. Future studies will help assess whether this same pathway is employed by mammalian cells, especially the nutrient-limited cells of solid tumors. In this vein, it is noteworthy that the production of acetyl-CoA in HeLa cells necessary to drive histone acetylation has been attributed to the enzymatic conversion of citrate into oxaloacetate and acetyl-CoA via the ATP citrate lyase enzyme (Wellen et al., 2009). Whereas mammalian cells contain three paralogous enzymes capable of converting acetate into acetyl-CoA, the latter study provides evidence that cancer cells make acetyl-CoA via a fundamentally different pathway than prototrophic yeast. The observations of Wellen and colleagues do conclude, however, that the GCN5 histone acetyltransferase enzyme of the SAGA complex is of critical importance for histone acetylation in response to the combined provision of glucose and growth factors to otherwise quiescent cells. As such, both yeast and human cancer cells
may employ similar strategies to couple nutrient availability to the control of gene expression. NAD+ and Deacetylation The burning of metabolic fuels uses molecular oxygen as the ultimate electron acceptor. Instead of being directly transferred to O2, electrons evolving from oxidative reactions use pyridine nucleotides as specialized carriers, with the reduced forms of these carriers then being able to transfer electrons to molecular oxygen. Nicotinamide adenine dinucleotide (NAD) is a key electron carrier in the oxidation of hydrocarbon fuels. The nicotinate moiety of NAD (niacin or vitamin B6) is derived from tryptophan and combines with 5-phosphoribosyl-1-pyrophosphate (PRPP) to yield nicotinate ribonucleotide and inorganic pyrophosphate. Desamido-NAD is then formed via the transfer of an AMP moiety from ATP to nicotinate ribonucleotide, with the nal step in the synthesis of NAD involving the transfer of the ammonia generated from the amide group of glutamine to the nicotinate carboxyl group. NADP, the related, phosphorylated derivative of NAD, is made via the transfer of a phosphoryl group from ATP to the 20 -hydroxyl group via an NAD kinase enzyme. Upon electron acceptance, NAD+ and NADP+ are converted to the reduced forms of these pyridine nucleotides. The ambient intracellular ratio of NAD+/NADH is roughly 100:1, whereas the ratio of NADP+/NADPH is 1:100. These ratios reect the evolved necessity for NAD+ to function primarily as an electron acceptor in the burning of hydrocarbon fuels and the necessity for NADPH to fulll anabolic biosynthetic reactions including the synthesis of cholesterol, bile acids, steroid hormones, and triglycerides. Considerable attention has been paid to the hypothetical role of uctuating NAD+ levels as a function of nutritional state and the activity of deacetylase enzymes. These enzymes come in two avors, those that catalyze deacetylation in an NAD+-independent manner, yielding the deacetylated substrate and free acetate as products; and those that are NAD+-dependent, yielding O-acetyl-ADP-ribose, the deacetylated substrate, and nicotinamide as products (Denu, 2005; Feldman et al., 2012; Haberland et al., 2009; Sauve et al., 2006). The latter proteins are members of the sirtuin family of deacetylases (Figure 1), which include two isoforms that are primarily housed in the nuclei of mammalian cells (SIRT6 and SIRT7), three that are localized to mitochondria (SIRT3, SIRT4, and SIRT5), and two that are found in both cytoplasmic and nuclear compartments (SIRT1 and SIRT2) (Finkel et al., 2009; Guarente, 2011a; Haigis and Sinclair, 2010; Verdin et al., 2010). NAD+ and Sirtuin Deacetylases For the purpose of simplicity, one can consider the action of sirtuin deacetylase enzymes as being a counterbalance to nutrientdriven protein acetylation. On a more microscopic level, the involvement of the mitochondrial SIRT3 enzyme in the deacetylation of the acetyl-CoA synthase enzyme AceCS2 is revealing (Hallows et al., 2006; Schwer et al., 2006). Under appropriate nutritional conditions, AceCS2 is acetylated on a specic lysine residue that inhibits the ability of the enzyme to convert acetate into acetyl-CoA, a regulatory scheme conserved from Salmonella to mammals (Hirschey et al., 2011; Starai et al., 2002). SIRT3-mediated deacetylation of AceCS2 reactivates the enzyme. One potential reason for justifying why AceCS2 is
deacetylated by a sirtuin enzyme is that the product of the reaction is not acetate, which might create a futile cycle, but instead O-acetyl-ADP-ribose and nicotinamide. Alternatively, sirtuinmediated production of the latter metabolite might avail it for biosynthetic or regulatory purposes (Hassa et al., 2006). On a more macroscopic level, one can consider the ability of SIRT1 to deacetylate the PGC1a transcriptional coactivator. PGC1a coordinately regulates many genes whose products conspire to control intermediary metabolism in many tissues of the body. When heavily acetylated, PGC1a is inactive (Lerin et al., 2006; Rodgers et al., 2005). SIRT1-mediated deacetylation reactivates PGC1a (Lerin et al., 2006). In the cases of both AceCS2 and PGC1a, access to ample nutrients can simplistically be understood to inhibit the activities of the two proteins via acetyl-CoA-mediated acetylation. This inhibition, in turn, can be respectively counterbalanced by the mitochondrial SIRT3 and nuclear SIRT1 enzymes. Caloric restriction would logically be expected to demand the activity of the sirtuin family of deacetylase enzymes. For example, SIRT3-mediated deacetylation of AceCS2 would be desired to maximize production of acetyl-CoA from acetate under conditions of caloric restriction, and SIRT1-mediated deacetylation of PGC1a would help activate transcription of the appropriate battery of nuclear genes important for adaptation to starvation or caloric restriction. Evidence has been reported that the levels of expression of sirtuin enzymes can adapt to metabolic state (Hirschey et al., 2011). It has likewise been reported that NAD+ levels may increase upon caloric restriction, thereby offering an alternative means of sirtuin activation. Although it is counterintuitive to consider that cells or tissues would produce higher levels of NAD+ under conditions of caloric restriction, where the need of the cofactor as an electron acceptor for oxidation of hydrocarbons should be diminished, this interpretation has gained widespread acceptance (Canto and Auwerx, 2011; Guarente, 2011b). Such interpretations contradict classical studies showing that NAD+ levels do not increase as a function of starvation. The collective work of Krebs and Veech exhaustively demonstrated that NAD+/NADH levels do not change as a function of starvation, irrespective of whether one measures bound or free fractions of the cofactors (Krebs and Veech, 1969; Veech et al., 1969). It has also been reported that NAD+ levels uctuate as a function of the circadian cycle, thereby instructing nuclear sirtuin enzymes to control the epigenetic state of chromatin in an NAD+-regulated manner. Mouse embryo broblast (MEF) cells decient in the CLOCK transcription factor were reported to contain only 4%5% as much NAD+ as wild-type MEF cells (Nakahata et al., 2009). When NAD+ levels were measured in liver tissue of wild-type mice, two ultradian sets of peaks and troughs of NAD+ abundance were observed per 24 hr cycle (Ramsey et al., 2009). The peak-to-trough uctuation in NAD+ abundance varied by 20%30%, as reported in the latter study. By contrast, when NAD+ levels were measured as a function of the YMC, which is far more robust in amplitude than metabolic uctuation taking place as a function of the circadian cycle, no changes in NAD+ levels were observed (Tu et al., 2007). Likewise, extensive studies of yeast cells exposed to a variety of nutritional states, including caloric restriction, have shown no
alteration in NAD+ or nicotinamide levels that could be interpreted to increase the activity of sirtuin enzymes upon glucose restriction (Evans et al., 2010). Thus it remains unclear whether sirtuin activity is operatively linked to metabolic state via uctuations in the intracellular levels of NAD+. What is clear, however, is that sirtuin enzymes sit in diametric opposition to protein acetylating and that protein acetylation can be inuenced by intracellular levels of acetyl-CoA. In the case of the AceCS2 enzyme that uses acetate to produce mitochondrial acetyl-CoA, SIRT3 serves to induce the catalytic activity of the enzyme by removing an inhibitory acetylation mark. In the case of PGC1a, the nuclear SIRT1 enzyme serves to deacetylate this transcriptional coactivator, thereby liberating its capacity to induce the expression of genes whose products are required in energy-depleted cells. Recent studies have provided evidence that the genes encoding SIRT3 and SIRT6 are tumor suppres n et al., 2012). In this regard, it sors (Kim et al., 2010; Sebastia is particularly intriguing that the battery of SIRT6 target genes n et al., 2012) has been reported to overlap signicantly (Sebastia with genes codependent on the c-Myc oncoprotein and the SAGA/GCN5 histone acetylase complex (Cai et. al., 2011; Ji et al., 2011; McMahon et al., 1998, 2000). NAD+ Independent Histone Deacetylases A recent study has raised the possibility of a different sort of connection between an abundant metabolite and the NAD+independent class of histone deacetylase enzyme. b-hydroxybutyrate, which is one of the three ketone bodies, is produced at mM quantities after prolonged exercise or starvation (Candido et al., 1978). Like sodium butyrate, b-hydroxybutyrate inhibits the activities of many NAD+-independent histone deacetylase enzymes (Shimazu et al., 2013). By administering b-hydroxybutyrate to laboratory mice via an intraperitoneal pump, Shimazu and colleagues were able to demonstrate enhancement of H3K9 and H3K14 acetylation, induced expression of FOXO3Aregulated genes, and resistance to oxidative stress. These data provide evidence that a distinct metabolite, b-hydroxybutyrate, is able not only to fuel metabolic adaptation to starvation but also to help sustain a protective epigenetic state by inhibiting the activities of NAD+-independent histone deacetylase enzymes. Methylation of DNA and Histones: SAM and the Activated Methyl Cycle S-adenosylmethionine (SAM) contains the active methyl donor group utilized by most methyltransferase enzymes (Figure 3). Tetrahydrofolate (THF), when methylated on its N-5 atom (N5MTHF), also acts as a methyl donor. Unlike SAM, the transfer potential of the methyl donor group of N5-MTHF is not sufciently high for most biosynthetic methylation reactions. SAM is produced by the condensation of methionine and ATP during the rst of nine steps required for the conversion of methionine to succinyl-CoA. The methyl group of SAM is chemically activated via the positive charge on the adjacent sulfur atom, which causes the SAM methyl group to be considerably more reactive than the methyl group on N5-MTHF. Enzyme-catalyzed donation of the methyl group of SAM to an acceptor macromolecule yields S-adenosylhomocysteine (SAH), which, in turn, is hydrolyzed to homocysteine and adenosine. The activated methyl
Figure 3. Metabolism and Methyltransferases

DNA and histone methyltransferases use S-adenosylmethionine (SAM), derived from methionine, as a methyl donor, resulting in the generation of S-adenosylhomocysteine (SAH). SAH is converted to homocysteine, which is then converted back to methionine in a vitamin B12dependent reaction that utilizes carbons derived from either choline or folate. DHF, dihydrofolate; THF, tetrahydrofolate; 5,10-MTHF, 5,10-methylene THF; CH3, methyl. Also shown are steps requiring vitamin B6 and B2. For simplicity, enzymes catalyzing the various reactions are not shown.
cycle can then be looped back via the transfer of a methyl group from N5-MTHF to homocysteine, regenerating methionine. SAM and Histone/DNA Methylation Histone methylation represents a covalent modication that is of equal importance to histone acetylation in dening the epigenetic state of chromatin (Kouzarides, 2002; Zhang and Reinberg, 2001). DNA itself is also subject to methylation on the C5 atom of cytosine (Bird, 2002). Intense studies reported over the past decade have led to the identication of a multitude of enzymes that afford the methylation and demethylation of both histone and DNA substrates. Both histone and DNA methylation require SAM as the high-energy methyl donor (Figure 3). Parallel with the aforementioned thinking concerning the possibility that acetylCoA levels might specify epigenetic state, the question can be raised as to whether ambient levels of intracellular or intranuclear SAM might help drive histone methylation. The conversion of methionine to SAM is catalyzed in an ATP-dependent manner by methionine adenosyltransferase (MAT) enzymes (Sakata et al., 1993). A recent study has given evidence that one of the MAT isoforms, MATIIa, associates with a gene-specic transcription factor designated MafK. The latter protein is a small bZip transcription factor that, depending upon its heterodimeric partner, can either repress or activate gene expression (Hintze et al., 2007; Igarashi and Sun, 2006; Muto et al., 1998; Ochiai et al., 2006; Zhang et al., 2006). Afnity chromatography experiments employing a tagged version of the MafK protein led to the discovery of its interaction with MATIIa, raising the possibility that the MafK transcription factor might recruit this enzyme directly to its target genes (Katoh et al., 2011). Cytological experiments revealed the nuclear localization of the MATIIa enzyme, and the results of a variety of molecular biological experiments were interpreted to indicate that the association of the MATIIa methyltransferase enzyme with MafK target genes may be required for transcriptional repression. If correct, these observations offer the possibility that a localized increase in the production of SAM might help establish the epigenetic state of cells. Threonine Dehydrogenase and SAM Related interpretations have evolved from studies of mouse ESCs. As mentioned earlier in this review, mouse ESCs express exceptionally high levels of the TDH enzyme, which catabolizes threonine into glycine and acetyl-CoA (Figure 4). When ESCs are exposed to a selective chemical inhibitor of the TDH enzyme,
intracellular levels of acetyl-CoA drop signicantly (Alexander et al., 2011). Simultaneous increases in the intracellular abundance of threonine and 5-aminoimidazole-4-carboxamide ribonucleotide (AICAR) were observed upon TDH inhibition in ESCs. The increase in threonine can logically be attributed to diminution in the activity of the enzyme that degrades it. Increases in AICAR, a biosynthetic intermediate that awaits N5MTHF-mediated one-carbon donation to continue along the pathway of purine biosynthesis, could likewise be attributed to attenuation in the production of mitochondrial glycine. The latter catabolite of TDH-mediated degradation of threonine is known to feed the mitochondrial glycine cleavage enzyme complex responsible for converting tetrahydrofolate to N5-MTHF (Figure 4). Not surprisingly, chemical inhibition of the TDH enzyme in mouse ESCs causes a precipitous drop in the level of intracellular N5-MTHF (Alexander et al., 2011). Having observed reduced levels of N5-MTHF and increased levels of the AICAR intermediate in purine biosynthesis upon chemical inhibition of the TDH enzyme, it was straightforward to predict that catabolism of threonine is vitally important for ESCs to biosynthesize the required building blocks for DNA synthesis. Given that the cell division cycle for mouse ESCs (45 hr in length) is more rapid than even the fastest growing of cultured cancer cell lines, it may reasonably be concluded that these cells require the specialized features of the TDH catabolic pathway to convert threonine into both acetyl-CoA and the glycine-dependent methylation capacity essential for the biosynthesis of nucleotides. Perplexingly, unique among all primates, mammals, and perhapsall metazoan species, humans do not encode a functional TDH enzyme (Edgar, 2002). As such, the unique metabolic properties engendered by copious expression of the TDH enzyme in mouse ESCs cannot apply to human stem cells. More recent work on mouse ESCs has led to the conclusion that TDH-mediated catabolism of threonine might also be important for maintaining high levels of SAM (Shyh-Chang et al., 2013). Knowing that N5-MTHF is required for the regeneration of methionine from homocysteine, it was logical to predict that intracellular levels of SAM might drop when ESCs are deprived of threonine. Intriguingly, threonine restriction was observed to dramatically impede deposition of the H3K4me2 and H3K4me3 marks on chromatin. These impediments were selective; threonine
Figure 4. TDH-Mediated Catabolism of Threonine

Threonine is catabolized to acetyl-CoA and glycine via a two-step process, rst involving the ratelimiting threonine dehydrogenase (TDH) enzyme yielding the short-lived intermediate 2-amino-3ketobutyrate. This intermediate is subsequently subject to the 2-amino-3-ketobutyrate ligase (KBL) enzyme that, supplemented by coenzyme A (CoA), yields the nal products of the reaction, acetyl-CoA and glycine. Both steps of the catabolic reaction take place in the mitochondria of eukaryotic cells. The former product, acetyl-CoA, can be fed into the TCA cycle or used as an anabolic building block for other metabolites. The latter product, glycine, is used to feed the mitochondrial glycine cleavage system for the conversion of tetrahydrofolate (THF) into N5, N10-methylene-tetrahydrofolate (MTHF). MTHF, in turn, is capable of one-carbon donation in biosynthetic reactions involving purine and pyrimidine synthesis, as well as the regeneration of methionine from homocysteine.
deprivation had no effect on the deposition of H3K4me1, H3K9me3, H3K27me3, H3K36me3, or H3K79me3 marks on histone tails. If correct, these observations offer a logical interpretation of the connection between nutritional and epigenetic states analogous to what has been learned from studies of acetyl-CoA levels driving the epigenetic state of prototrophic yeast cells growing under nutrient-limited conditions (Cai et al., 2011). The H3K4me2 and H3K3me3 marks that disappear when mouse ESCs are deprived of threonine (Shyh-Chang et al., 2013) are part of a bivalent epigenetic state believed to be uniquely important for keeping the chromatin structure of genes encoding developmentally important transcription factors in a properly repressed/poised state for subsequent induction as a function of embryogenesis (Bernstein et al., 2006). It is therefore possible that, by catabolizing threonine along pathways important for the production of both acetyl-CoA necessary for histone acetylation and methylation capacity via the generation of N5-MTHF and SAM, the TDH enzyme may be of fundamental importance in controlling two unique properties of mouse ESCstheir incredibly rapid rate of cell division and their unique retention of developmental pluripotency. FAD and Histone Demethylation Flavin adenine dinucleotide (FAD) is derived from the vitamin riboavin (vitamin B2) and functions as the prosthetic group for certain oxidation-reduction enzymes. Riboavin is phosphorylated by riboavin kinase to generate riboavin 50 -phosphate (sometimes called avin monucleotide or FMN), which is then converted to FAD by FAD synthetase (also called FMN adenyltransferase). The former enzyme appears to be rate limiting for FAD production. In a landmark paper, Yang Shi and coworkers showed that LSD1 (also called KDM1A or AOF2) is a nuclear FAD-dependent enzyme capable of demethylating methylated H3K4 both in vitro and in vivo, thereby establishing that histone methylation is a dynamic, reversible process (Shi et al., 2004). The chemical reaction catalyzed by LSD1 requires a protonated lysine epsilon amino group, thereby limiting its activity to mono62 Cell 153, March 28, 2013 2013 Elsevier Inc.
methylated and dimethylated H3K4 (Figure 5). LSD1 and its paralog LSD2 (also called KDM1B or AOF1) can also inuence H3K9 methylation and DNA methylation, either due to their participation in multiprotein complexes that contain additional chromatin-modifying enzymes, including deacetylases and methyltransferases, or due to indirect effects of H3K4 methylation on recruitment of such enzymes (Ciccone et al., 2009; Lan et al., 2008; Wang et al., 2009b). It has also been argued that the reactive oxygen species produced by the LSD1 histone demethylation reaction can react with neighboring DNA and other macromolecules and thereby affect transcription (Perillo et al., 2008). FAD is produced in mitochondria. It has been suggested that the nuclear location of LSD1 might render it particularly sensitive to changes in FAD availability (and the ratio of FAD to FADH2) arising from the activities of other avin-linked dehydrogenases and oxidases, including those associated with fatty acid b-oxidation and the TCA cycle (Hino et al., 2012). LSD1, in turn, regulates mitochondrial respiration and energy expenditure. Specically, LSD1 binds directly to genes such as PPARg coactivator-1a (PGC1a), PDK4, FATP1, and ATGL and represses their transcription associated with loss of H3K4 methylation (Hino et al., 2012). 2-Oxoglutarate-Dependent Dioxygenases JmjC and TET Demethylases A number of chromatin-modifying enzymes, including the approximately 30 JmjC domain-containing histone demethylases and the three TET (ten-eleven translocation) proteins, are 2-oxoglutarate-dependent dioxygenases (Loenarz and Schoeld, 2011) (Figure 5). The JmjC histone demethylases, in contrast to LSD proteins, are capable of demethylating trimethylated lysines and arginines. Different JmjC histone demethylases exhibit preferences for different histone methylation marks. For example, KDM5A (also called RBP2 or JARID1A) specically recognizes methylated H3K4. TET proteins can oxidize 5-methylcytosine (5mC) to 5-hydroxymethylcytosine (5hmC), 5-formyl cytosine (5fC), and 5-carboxylcytosine (5caC) (He et al., 2011; Ito et al., 2011; Tahiliani et al., 2009). 5hmC in the nervous system
Figure 5. Histone Demethylase Reactions

LSD demethylases oxidize monomethylated and dimethylated histones using FAD (avin adenine dinucleotide) as a cofactor. The oxidized methyl group is unstable and, after attack by water, given off as formaldehyde (HCHO). FAD is derived from FMN (avin mononucleotide), which in turn is derived from riboavin. JmjC demethylases hydroxylate methylated histones in a reaction coupled to decarboxylation of 2-oxoglutarate to succinate. 2-oxoglutarate (also called a-ketoglutarate) can be derived from several sources including isocitrate and glutamic acid. Spontaneous release of the hydroxylated methyl group results in demethylation.
is associated with actively transcribed genes and is recognized by the histone-reader protein MeCP2, which is mutationally inacn tivated in the neurological disorder of Rett syndrome (Melle et al., 2012). TET-catalyzed oxidation of 5mC followed by decarboxylation or base excision is strongly suspected of contributing to DNA demethylation (Cortellino et al., 2011; He et al., 2011; Ito et al., 2011), yet there are currently conicting reports regarding the effect of TET inactivation on global DNA methylation patterns in somatic cells (Figueroa et al., 2010; Ko et al., 2010; Yamazaki et al., 2012). Iron and Oxygen 2-oxoglutarate-dependent dioxygenases require oxygen and Fe (II) in addition to 2-oxoglutarate (also called a-ketoglutarate). The latter metabolite is decarboxylated to succinate during the oxidation reaction (Figure 5). These enzymes are also sensitive to reactive oxygen species and their activity is enhanced in the presence of ascorbic acid. These biochemical attributes render these enzymes potentially susceptible to carcinogenic metals such as nickel, arsenic, and chromium, which displace iron, contribute to oxidative stress, or do both (Chervona and Costa, 2012). Certain 2-oxoglutarate-dependent dioxygenases, such as the EglN prolyl hydroxylases that mark the HIF transcription factor for destruction, have O2 Km values at or near atmospheric oxygen levels and hence their activity is sensitive to changes in oxygen availability within a physiologically relevant range (Hirsila et al., 2003). By contrast, the collagen prolyl hydroxylases have very low O2 Km, presumably so that they can function in environ et al., 2003). ments such as poorly vascularized wounds (Hirsila The O2 Km values for the JmjC histone demethylases and TET proteins are not known, although indirect evidence suggests that at least some of the former could be oxygen sensitive. Specically, the messenger RNAs (mRNAs) for multiple JmjC histone
demethylase genes, including JMJD1A (an H3K9 demethylase), JMJD2B (an H3K9 demethylase), and JARIDC (an H3K4 demethylase), are induced by hypoxia and HIF, conceivably to compensate for their diminished catalytic activity under low-oxygen conditions (Chervona and Costa, 2012; Xia et al., 2009) (Figure 6A). Moreover, increased H3K4 and H3K9 histone methylation has been documented in cells treated with hypoxia in vitro (Johnson n et al., 2011; Zhou et al., 2010). et al., 2008; Tausendscho HIF synthesis is inuenced by the PI3K-AKT-mTOR pathway, which senses nutrients such as glucose and amino acids, while its degradation is under the control of the EglN prolyl hydroxylases, which respond to changes in oxygen and different Krebs cycle intermediates (Figure 6A). Therefore, HIF provides another link between metabolism and epigenetics through its transcriptional regulation of JmjC histone demethylases and other modiers of epigenetic state. 2-Oxoglutarate 2-oxoglutarate is produced from isocitrate in the mitochondria by the action of isocitrate dehydrogenases 2 and 3 (IDH2 and IDH3) in the TCA cycle and can also be generated from several amino acids including arginine, proline, histidine, and glutamine, which can be converted to glutamic acid and transaminated to produce 2-oxoglutarate. 2-oxoglutarate generated in the course of the TCA cycle is converted to succinyl CoA by the action of the 2-oxoglutarate dehydrogenase complex and is also consumed during the conversion of cysteine and lysine to pyruvate and acetyl-CoA, respectively. 2-oxoglutarate generated in the mitochondria can enter the cytosol, either as 2-oxoglutarate or after transamination to produce glutamic acid, via specic transporter proteins such as the malate-2-oxoglutarate antiporter, which participates in the malate-aspartate shuttle. A third IDH paralog, IDH1, resides in the cytosol and peroxisomes and provides an alternative source for nonmitochondrial 2-oxoglutarate. Intracellular 2-oxoglutarate levels are estimated to be in the low millimolar range, well above the 2-oxoglutarate Km values of the JmjC histone demethylases and TET proteins determined to date (Chowdhury et al., 2011; Pritchard, 1995). A caveat, however, is that such Km values are typically determined under idealized in vitro conditions using puried enzymes in the
Figure 6. Mutant Metabolic Enzymes, Epigenetics, and Cancer

(A) The abundance of the HIF transcription factor is regulated by the 2-oxoglutarate-dependent EglN dioxygenases, which are sensitive to changes in oxygen and metabolism, and by the PI3K-AKT-mTOR pathway, which is involved in nutrient sensing and frequently mutationally activated in cancer. HIF induces the transcription of a number of JmjC histone demethylases. (B) Mutational inactivation of the Krebs Cycle Enzymes SDH and FH leads to the accumulation of succinate and fumarate, respectively, whereas tumor-derived IDH1 and IDH2 mutants produce high levels of R-2-hydroxyglutarate (R-2HG). Succinate, fumarate, and R-2HG can inhibit 2-oxoglutarate-dependent dioxygenases, including JmjC histone demethylases and TET DNA hydroxylases.
absence of endogenous inhibitory molecules such as fumarate, succinate, and reactive oxygen species. In this regard, intracellular fumarate and succinate concentrations are estimated to be in the high micromolar and low millimolar range, respectively (Johnson et al., 2008). In model systems, many metabolic enzymes display Km values in vivo that are higher than their corresponding in vitro values and certainly closer to the concentrations of their cosubstrates (Bennett et al., 2009; Yuan et al., 2009). This presumably allows such enzymes to respond to changes in the concentrations of available agonists and antagonists. Moreover, the concentrations of 2-oxoglutarate in specic subcellular compartments such as the nucleus are not known, nor is it known how much 2-oxoglutarate is free versus tightly bound to other proteins. It thus remains possible, but not proven, that JmjC proteins and TET proteins can respond under physiological conditions to changes in agonists such as 2-oxoglutarate or antagonists such as succinate and fumarate. The functions of these proteins do, however, appear to be deregulated as a result of altered metabolism under the pathological conditions found in the specic cancers described below. SDH, FH, and IDH Mutations and Cancer Inactivating mutations affecting the mitochondrial succinate dehydrogenase (SDH) complex subunits and fumarate hydratase (FH) have been identied in cancers, particularly in familial paragangliomas and papillary renal cancers, respectively (Kaelin, 2009). These mutations lead to the marked accumulation of succinate and fumarate, respectively (Figure 6B). Somatic mutations affecting cytosolic IDH1 and mitochondrial IDH2 have been identied in gliomas, acute myelogenous leukemia, chondrosarcomas, and cholangiosarcomas. Somatic mosaicism for such mutations causes Ollier Disease syndrome and Maffucci syndrome, which are characterized by the development of endochondromas and spindle cell hemangiomas (Amary et al., 2011; Pansuriya et al., 2011). Tumor-associated IDH mutations
unmask a latent ability of these enzymes to produce the R enantiomer of 2-hydroxyglutarate, which accumulates to low millimolar levels in IDH mutant tumors (Dang et al., 2009; Gross et al., 2010; Ward et al., 2010). R-2HG, succinate, and fumarate are all capable of inhibiting multiple 2-oxoglutarate-dependent dioxygenases, including the JmjC histone demethylases and TET proteins, when present at sufciently high concentrations (Figure 6B) (Cervera et al., 2009; Chowdhury et al., 2011; Figueroa et al., 2010; Koivunen et al., 2012; Lu et al., 2012; Smith et al., 2007; Turcan et al., 2012; Xiao et al., 2012; Xu et al., 2011). The challenge for the eld is to determine which of these enzymes are actually inhibited by these metabolites in vivo in tumors bearing the appropriate mutations and which of these enzymes are causally linked to the transformed phenotype. For example, the HIF1a transcription factor accumulates in SDH and FH mutant tumors, presumably due to inhibition of the EglN prolyl hydroxylases (Dahia et al., 2005; Isaacs et al., 2005; Pollard et al., 2005, 2007; Selak et al., 2005). Moreover, it is plausible that HIF1a promotes the formation of such tumors based on its repertoire of downstream targets. Deletion of HIF1a, however, worsened the pathological changes observed in mice engineered to lack FH, suggesting that it normally constrains, rather than promotes, the emergence of FH-defective tumors (Adam et al., 2011). A number of 2-oxoglutarate-dependent enzymes that act on chromatin are themselves targets of inactivating mutations in cancer, including the UTX H3K27 demethylase (also called KDM6A), JARID1C H3K4 demethylase (KDM5C), and TET2, making them attractive candidates for pathogenically relevant inhibition by succinate, fumarate, and R-2HG (Abdel-Wahab et al., 2009; Dalgliesh et al., 2010; Tefferi et al., 2009; van Haaften et al., 2009). In this regard, TET2 and IDH mutations are both found in acute myelogenous leukemia and are mutually exclusive, consistent with the idea that R-2HG produced by mutant IDH provides an alterative means of inactivating TET2 (Figueroa
et al., 2010). Moreover, IDH mutant tumors have been reported to display DNA hypermethylation changes consistent with TET2 inactivation. In addition, TET2 inactivation is sufcient to promote hematopoietic stem cell self-renewal and to block differentiation (Figueroa et al., 2010; Li et al., 2011; Moran-Crusio et al., 2011; Noushmehr et al., 2010; Turcan et al., 2012; Losman et al., 2013). Therefore, inactivation of TET2 probably contributes to leukemic transformation at the hands of mutant IDH and the R-2HG oncometabolite it produces. Interestingly, R-2HG is sufcient to promote cytokine independence and block differentiation of hematopoietic cells in vitro and these effects are fairly rapidly reversed upon R-2HG withdrawal (Losman et al., 2013; K. Yen, personal communication). This suggests either that these phenotypes are not due to epigenetic targets of R-2HG or that R-2HG-induced epigenetic changes are surprisingly dynamic. In summary, SDH, FH, and IDH mutations cause the accumulation of succinate, fumarate, and R-2HG, respectively. These metabolites appear to cause cancer by affecting the behavior of various 2-oxoglutarate-dependent dioxygenases, including dioxygenases linked to DNA and histone methylation, and therefore have the potential to alter the epigenome. Nutrition, Epigenetics, and Disease Altered epigenetics is believed to play a part in a variety of diseases in addition to cancer, including diabetes, obesity, dyslipidemia, hypertension, and neurodegeneration. The biochemical considerations outlined herein provide a conceptual framework for understanding how environmental changes, including changes in nutrition, could affect epigenetics and therefore diseases where epigenetic alterations play a role. In this regard, particular attention has been paid to the inuence of diet on one-carbon metabolism, which plays an important role in DNA and histone methylation and in risk of disease. Studies in rodents and sheep have conrmed that the availability of nutrients required for one-carbon metabolism (for example, folate, choline, methionine, and betaine, as well as selected B vitamins) at the time of conception and during pregnancy can induce epigenetically driven phenotypes in their offspring (Dolinoy et al., 2007; Sinclair et al., 2007; Waterland et al., 2006; Wolff et al., 1998). There is also evidence that changes in glucose and glucose metabolism can leave lasting epigenetic marks (Park et al., 2012; Pirola et al., 2010). Observational studies in humans are also consistent with a possible role of nutrition in epigenetics and disease. For example, studies of individuals conceived during famine conditions have revealed decreased DNA methylation of specic loci, such as the IGF2 gene, associated with an increased risk of obesity, dyslipidemia, and insulin resistance later in life (Dominguez-Salas et al., 2012). Genetic variation in a number of genes linked to one-carbon metabolism, including methylenetetrahydrofolate dehydrogenase 1, FTHF dehydrogenase, 510 methyleneTHF reductase, methionine synthase, and glycine-N-methyltransferase, has been linked to a variety of disease phenotypes including cancer and developmental defects (Stover, 2011). Biomarkers or dietary histories indicative of nutritional deciencies associated with defects in one-carbon metabolism (for example, folate deciency) have frequently been associated
with an increased risk of cancer in epidemiological studies. Disappointingly, however, intervention trials have failed to demonstrate a reduction of cancer risk in individuals randomized to receive folic acid supplements compared to controls (Andreeva et al., 2012; Cole et al., 2007; Song et al., 2012; Zhang et al., 2008). This might suggest that folate deciency correlates with, but does not cause, cancer or that folate is important during an early time window, perhaps occurring at an early age and prior to or during tumor initiation, but not thereafter. In this regard, animal studies suggest that folate might actually increase cancer progression if given to nascent tumors in the colon (Kim, 2004). There are many bioactive molecules present in food and herbs, in addition to folate, that are capable of inuencing epigenetics and that have been touted as potential chemopreventative agents. Examples include various B vitamins, retinoic acid, vitamin D3, reservatrol, genistein and daidzein, epigallocatechin-3-gallate (EGCG), and curcumin (Gerhauser, 2013; Stefanska et al., 2012). Moving forward, it will be important to determine whether these agents inuence epigenetic enzymes at concentrations that can be achieved in vivo (as opposed to cell culture studies), to understand their dose-response relationships, and to determine when, during the lifetime of an individual, they can exert their salutary effects. Conclusions and Future Questions Many epigenetic enzymes are potentially susceptible to changes in the levels of cosubstrates and cofactors such as oxygen, ATP, acetyl-CoA, MTHF, S-adenosylmethionine, NAD, FAD, and 2-oxoglutarate and are hence poised to respond to changes in nutrient intake and metabolism. We need to learn more, however, about how much the levels of some of these cosubstrates and cofactors can vary in space (for example, in different cellular subcompartments and in different tissues) and how much they can uctuate over time (for example, in response to changes in nutrition or as a function of age). If they do not vary signicantly, how are they sensed and how are they buffered? If they do vary signicantly, do they actually become limiting for specic epigenetic enzymes in vivo? If so, does this contribute to disease? There is increasing evidence that early exposures, including intrauterine exposures, can lead to lasting epigenetic changes and that epigenetic differences can inuence many phenotypes, including the risk of disease. Among the relevant exposures are exposures related to nutrition. This has led to the hypothesis that we are what we eat but also what our parents ate (DominguezSalas et al., 2012). We need better biomarkers for assessing variability in nutrition and metabolism in the population and its effects on epigenetics. From a public health perspective we need to better understand which alterations in metabolism and epigenetics cause, rather than correlate with, disease and whenand howit might be possible to intervene.
ACKNOWLEDGMENTS The authors apologize to colleagues whose work was not cited due to space limitations or our oversight. Please bring errors and egregious omissions to our attention. S.M. is the founder and chairman of the scientic advisory board of
Peloton Therapeutics. In these roles, S.M. has received compensation in the form of both consulting fees and stock equity in the company. W.K. owns equity in, and consults for, Agios, Fibrogen, Lilly, Nextech, Peloton, and Tracon. REFERENCES Abdel-Wahab, O., Mullally, A., Hedvat, C., Garcia-Manero, G., Patel, J., Wadleigh, M., Malinge, S., Yao, J., Kilpivaara, O., Bhat, R., et al. (2009). Genetic characterization of TET1, TET2, and TET3 alterations in myeloid malignancies. Blood 114, 144147. Adam, J., Hatipoglu, E., OFlaherty, L., Ternette, N., Sahgal, N., Lockstone, H., Baban, D., Nye, E., Stamp, G.W., Wolhuter, K., et al. (2011). Renal cyst formation in Fh1-decient mice is independent of the Hif/Phd pathway: roles for fumarate in KEAP1 succination and Nrf2 signaling. Cancer Cell 20, 524537. Alexander, P.B., Wang, J., and McKnight, S.L. (2011). Targeted killing of a mammalian cell based upon its specialized metabolic state. Proc. Natl. Acad. Sci. USA 108, 1582815833. Amary, M.F., Damato, S., Halai, D., Eskandarpour, M., Berisha, F., Bonar, F., McCarthy, S., Fantin, V.R., Straley, K.S., Lobo, S., et al. (2011). Ollier disease and Maffucci syndrome are caused by somatic mosaic mutations of IDH1 and IDH2. Nat. Genet. 43, 12621265. Andreeva, V.A., Touvier, M., Kesse-Guyot, E., Julia, C., Galan, P., and Hercberg, S. (2012). B vitamin and/or u-3 fatty acid supplementation and cancer: ancillary ndings from the supplementation with folate, vitamins B6 and B12, and/or omega-3 fatty acids (SU.FOL.OM3) randomized trial. Arch. Intern. Med. 172, 540547. Bennett, B.D., Kimball, E.H., Gao, M., Osterhout, R., Van Dien, S.J., and Rabinowitz, J.D. (2009). Absolute metabolite concentrations and implied enzyme active site occupancy in Escherichia coli. Nat. Chem. Biol. 5, 593599. Bernstein, B.E., Mikkelsen, T.S., Xie, X., Kamal, M., Huebert, D.J., Cuff, J., Fry, B., Meissner, A., Wernig, M., Plath, K., et al. (2006). A bivalent chromatin structure marks key developmental genes in embryonic stem cells. Cell 125, 315326. Bird, A. (2002). DNA methylation patterns and epigenetic memory. Genes Dev. 16, 621. Cai, L., and Tu, B.P. (2011). On acetyl-CoA as a gauge of cellular metabolic state. Cold Spring Harb. Symp. Quant. Biol. 76, 195202. Cai, L., and Tu, B.P. (2012). Driving the cell cycle through metabolism. Annu. Rev. Cell Dev. Biol. 28, 5987. Cai, L., Sutter, B.M., Li, B., and Tu, B.P. (2011). Acetyl-CoA induces cell growth and proliferation by promoting the acetylation of histones at growth genes. Mol. Cell 42, 426437. Candido, E.P., Reeves, R., and Davie, J.R. (1978). Sodium butyrate inhibits histone deacetylation in cultured cells. Cell 14, 105113. , C., and Auwerx, J. (2011). NAD+ as a signaling molecule modulating Canto metabolism. Cold Spring Harb. Symp. Quant. Biol. 76, 291298. Cervera, A.M., Bayley, J.P., Devilee, P., and McCreath, K.J. (2009). Inhibition of succinate dehydrogenase dysregulates histone modication in mammalian cells. Mol. Cancer 8, 89. Chervona, Y., and Costa, M. (2012). The control of histone methylation and gene expression by oxidative stress, hypoxia, and metals. Free Radic. Biol. Med. 53, 10411047. Chowdhury, R., Yeoh, K.K., Tian, Y.M., Hillringhaus, L., Bagg, E.A., Rose, N.R., Leung, I.K., Li, X.S., Woon, E.C., Yang, M., et al. (2011). The oncometabolite 2-hydroxyglutarate inhibits histone lysine demethylases. EMBO Rep. 12, 463469. Ciccone, D.N., Su, H., Hevi, S., Gay, F., Lei, H., Bajko, J., Xu, G., Li, E., and Chen, T. (2009). KDM1B is a histone H3K4 demethylase required to establish maternal genomic imprints. Nature 461, 415418. Cole, B.F., Baron, J.A., Sandler, R.S., Haile, R.W., Ahnen, D.J., Bresalier, R.S., McKeown-Eyssen, G., Summers, R.W., Rothstein, R.I., Burke, C.A., et al.;
Polyp Prevention Study Group. (2007). Folic acid for the prevention of colorectal adenomas: a randomized clinical trial. JAMA 297, 23512359. Cortellino, S., Xu, J., Sannai, M., Moore, R., Caretti, E., Cigliano, A., Le Coz, M., Devarajan, K., Wessels, A., Soprano, D., et al. (2011). Thymine DNA glycosylase is essential for active DNA demethylation by linked deamination-base excision repair. Cell 146, 6779. Dahia, P.L., Ross, K.N., Wright, M.E., Hayashida, C.Y., Santagata, S., Barontini, M., Kung, A.L., Sanso, G., Powers, J.F., Tischler, A.S., et al. (2005). A HIF1alpha regulatory loop links hypoxia and mitochondrial signals in pheochromocytomas. PLoS Genet. 1, 7280. Dalgliesh, G.L., Furge, K., Greenman, C., Chen, L., Bignell, G., Butler, A., Davies, H., Edkins, S., Hardy, C., Latimer, C., et al. (2010). Systematic sequencing of renal carcinoma reveals inactivation of histone modifying genes. Nature 463, 360363. Dang, L., White, D.W., Gross, S., Bennett, B.D., Bittinger, M.A., Driggers, E.M., Fantin, V.R., Jang, H.G., Jin, S., Keenan, M.C., et al. (2009). Cancer-associated IDH1 mutations produce 2-hydroxyglutarate. Nature 462, 739744. Dawson, M.A., and Kouzarides, T. (2012). Cancer epigenetics: from mechanism to therapy. Cell 150, 1227. Denu, J.M. (2005). The Sir 2 family of protein deacetylases. Curr. Opin. Chem. Biol. 9, 431440. Dolinoy, D.C., Huang, D., and Jirtle, R.L. (2007). Maternal nutrient supplementation counteracts bisphenol A-induced DNA hypomethylation in early development. Proc. Natl. Acad. Sci. USA 104, 1305613061. Dominguez-Salas, P., Cox, S.E., Prentice, A.M., Hennig, B.J., and Moore, S.E. (2012). Maternal nutritional status, C(1) metabolism and offspring DNA methylation: a review of current evidence in human subjects. Proc. Nutr. Soc. 71, 154165. Edgar, A.J. (2002). The human L-threonine 3-dehydrogenase gene is an expressed pseudogene. BMC Genet. 3, 18. Evans, C., Bogan, K.L., Song, P., Burant, C.F., Kennedy, R.T., and Brenner, C. (2010). NAD+ metabolite levels as a function of vitamins and calorie restriction: evidence for different mechanisms of longevity. BMC Chem. Biol. 10, 2. Feldman, J.L., Dittenhafer-Reed, K.E., and Denu, J.M. (2012). Sirtuin catalysis and regulation. J. Biol. Chem. 287, 4241942427. Figueroa, M.E., Abdel-Wahab, O., Lu, C., Ward, P.S., Patel, J., Shih, A., Li, Y., Bhagwat, N., Vasanthakumar, A., Fernandez, H.F., et al. (2010). Leukemic IDH1 and IDH2 mutations result in a hypermethylation phenotype, disrupt TET2 function, and impair hematopoietic differentiation. Cancer Cell 18, 553567. Finkel, T., Deng, C.X., and Mostoslavsky, R. (2009). Recent progress in the biology and physiology of sirtuins. Nature 460, 587591. Gerhauser, C. (2013). Cancer chemoprevention and nutriepigenetics: state of the art and future challenges. Top. Curr. Chem. 329, 73132. Gross, S., Cairns, R.A., Minden, M.D., Driggers, E.M., Bittinger, M.A., Jang, H.G., Sasaki, M., Jin, S., Schenkein, D.P., Su, S.M., et al. (2010). Cancer-associated metabolite 2-hydroxyglutarate accumulates in acute myelogenous leukemia with isocitrate dehydrogenase 1 and 2 mutations. J. Exp. Med. 207, 339344. Guarente, L. (2011a). Franklin H. Epstein Lecture: Sirtuins, aging, and medicine. N. Engl. J. Med. 364, 22352244. Guarente, L. (2011b). Sirtuins, aging, and metabolism. Cold Spring Harb. Symp. Quant. Biol. 76, 8190. Haberland, M., Montgomery, R.L., and Olson, E.N. (2009). The many roles of histone deacetylases in development and physiology: implications for disease and therapy. Nat. Rev. Genet. 10, 3242. Haigis, M.C., and Sinclair, D.A. (2010). Mammalian sirtuins: biological insights and disease relevance. Annu. Rev. Pathol. 5, 253295. Hallows, W.C., Lee, S., and Denu, J.M. (2006). Sirtuins deacetylate and activate mammalian acetyl-CoA synthetases. Proc. Natl. Acad. Sci. USA 103, 1023010235.
Hardie, D.G. (2011). Adenosine monophosphate-activated protein kinase: a central regulator of metabolism with roles in diabetes, cancer, and viral infection. Cold Spring Harb. Symp. Quant. Biol. 76, 155164. Hassa, P.O., Haenni, S.S., Elser, M., and Hottiger, M.O. (2006). Nuclear ADPribosylation reactions in mammalian cells: where are we today and where are we going? Microbiol. Mol. Biol. Rev. 70, 789829. He, Y.F., Li, B.Z., Li, Z., Liu, P., Wang, Y., Tang, Q., Ding, J., Jia, Y., Chen, Z., Li, L., et al. (2011). Tet-mediated formation of 5-carboxylcytosine and its excision by TDG in mammalian DNA. Science 333, 13031307. Hino, S., Sakamoto, A., Nagaoka, K., Anan, K., Wang, Y., Mimasu, S., Umehara, T., Yokoyama, S., Kosai, K., and Nakao, M. (2012). FAD-dependent lysine-specic demethylase-1 regulates cellular energy expenditure. Nat. Commun. 3, 758. Hintze, K.J., Katoh, Y., Igarashi, K., and Theil, E.C. (2007). Bach1 repression of ferritin and thioredoxin reductase1 is heme-sensitive in cells and in vitro and coordinates expression with heme oxygenase1, beta-globin, and NADP(H) quinone (oxido) reductase1. J. Biol. Chem. 282, 3436534371. Hirschey, M.D., Shimazu, T., Huang, J.Y., Schwer, B., and Verdin, E. (2011). SIRT3 regulates mitochondrial protein acetylation and intermediary metabolism. Cold Spring Harb. Symp. Quant. Biol. 76, 267277. , M., Koivunen, P., Gu nzler, V., Kivirikko, K.I., and Myllyharju, J. (2003). Hirsila Characterization of the human prolyl 4-hydroxylases that modify the hypoxia-inducible factor. J. Biol. Chem. 278, 3077230780. Igarashi, K., and Sun, J. (2006). The heme-Bach1 pathway in the regulation of oxidative stress response and erythroid differentiation. Antioxid. Redox Signal. 8, 107118. Isaacs, J.S., Jung, Y.J., Mole, D.R., Lee, S., Torres-Cabala, C., Chung, Y.L., Merino, M., Trepel, J., Zbar, B., Toro, J., et al. (2005). HIF overexpression correlates with biallelic loss of fumarate hydratase in renal cancer: novel role of fumarate in regulation of HIF stability. Cancer Cell 8, 143153. Ito, S., Shen, L., Dai, Q., Wu, S.C., Collins, L.B., Swenberg, J.A., He, C., and Zhang, Y. (2011). Tet proteins can convert 5-methylcytosine to 5-formylcytosine and 5-carboxylcytosine. Science 333, 13001303. Ji, H., Wu, G., Zhan, X., Nolan, A., Koh, C., De Marzo, A., Doan, H.M., Fan, J., Cheadle, C., Fallahi, M., et al. (2011). Cell-type independent MYC target genes reveal a primordial signature involved in biomass accumulation. PLoS ONE 6, e26057. Johnson, A.B., Denko, N., and Barton, M.C. (2008). Hypoxia induces a novel signature of chromatin modications and global repression of transcription. Mutat. Res. 640, 174179. Jorgensen, P., Nishikawa, J.L., Breitkreutz, B.J., and Tyers, M. (2002). Systematic identication of pathways that couple cell growth and division in yeast. Science 297, 395400. Kaelin, W.G., Jr. (2009). SDH5 mutations and familial paraganglioma: somewhere Warburg is smiling. Cancer Cell 16, 180182. Katoh, Y., Ikura, T., Hoshikawa, Y., Tashiro, S., Ito, T., Ohta, M., Kera, Y., Noda, T., and Igarashi, K. (2011). Methionine adenosyltransferase II serves as a transcriptional corepressor of Maf oncoprotein. Mol. Cell 41, 554566. Kim, Y.I. (2004). Will mandatory folic acid fortication prevent or promote cancer? Am. J. Clin. Nutr. 80, 11231128. Kim, H.S., Patel, K., Muldoon-Jacobs, K., Bisht, K.S., Aykin-Burns, N., Pennington, J.D., van der Meer, R., Nguyen, P., Savage, J., Owens, K.M., et al. (2010). SIRT3 is a mitochondria-localized tumor suppressor required for maintenance of mitochondrial integrity and metabolism during stress. Cancer Cell 17, 4152. Klevecz, R.R., Bolen, J., Forrest, G., and Murray, D.B. (2004). A genomewide oscillation in transcription gates DNA replication and cell cycle. Proc. Natl. Acad. Sci. USA 101, 12001205. Ko, M., Huang, Y., Jankowska, A.M., Pape, U.J., Tahiliani, M., Bandukwala, H.S., An, J., Lamperti, E.D., Koh, K.P., Ganetzky, R., et al. (2010). Impaired hydroxylation of 5-methylcytosine in myeloid cancers with mutant TET2. Nature 468, 839843.
Koivunen, P., Lee, S., Duncan, C.G., Lopez, G., Lu, G., Ramkissoon, S., Losman, J.A., Joensuu, P., Bergmann, U., Gross, S., et al. (2012). Transformation by the (R)-enantiomer of 2-hydroxyglutarate linked to EGLN activation. Nature 483, 484488. Kouzarides, T. (2002). Histone methylation in transcriptional control. Curr. Opin. Genet. Dev. 12, 198209. Krebs, H.A., and Veech, R. (1969). Pyridine nucleotide interrelations in The Energy Level and Metabolic Control in Mitochondria, S. Papa, J. Tager, E. Quagliariello, and E. Slater, eds. (Bari, Italy: Adriatica Editrice), pp. 329382. Lan, F., Nottke, A.C., and Shi, Y. (2008). Mechanisms involved in the regulation of histone lysine demethylases. Curr. Opin. Cell Biol. 20, 316325. Langer, M.R., Fry, C.J., Peterson, C.L., and Denu, J.M. (2002). Modulating acetyl-CoA binding in the GCN5 family of histone acetyltransferases. J. Biol. Chem. 277, 2733727344. Lee, K.K., and Workman, J.L. (2007). Histone acetyltransferase complexes: one size doesnt t all. Nat. Rev. Mol. Cell Biol. 8, 284295. Lerin, C., Rodgers, J.T., Kalume, D.E., Kim, S.H., Pandey, A., and Puigserver, P. (2006). GCN5 acetyltransferase complex controls glucose metabolism through transcriptional repression of PGC-1alpha. Cell Metab. 3, 429438. Li, Z., Cai, X., Cai, C.L., Wang, J., Zhang, W., Petersen, B.E., Yang, F.C., and Xu, M. (2011). Deletion of Tet2 in mice leads to dysregulated hematopoietic stem cells and subsequent development of myeloid malignancies. Blood 118, 45094518. Lipmann, F., and Kaplan, N.O. (1946). A common factor in the enzymatic acetylation of sulfanilamide and of choline. J. Biol. Chem. 162, 743744. Loenarz, C., and Schoeld, C.J. (2011). Physiological and biochemical aspects of hydroxylations and demethylations catalyzed by human 2-oxoglutarate oxygenases. Trends Biochem. Sci. 36, 718. Losman, J.A., Looper, R., Koivunen, P., Lee, S., Schneider-Kramann, R., Cowley, G., Root, D.E., Ebert, B., and Kaelin, W.G. (2013). (R)-2-Hydroxyglutarate is sufcient to reversibly promote leukemogenesis and its effect are reversible. Science. Lu, C., and Thompson, C.B. (2012). Metabolic regulation of epigenetics. Cell Metab. 16, 917. Lu, C., Ward, P.S., Kapoor, G.S., Rohle, D., Turcan, S., Abdel-Wahab, O., Edwards, C.R., Khanin, R., Figueroa, M.E., Melnick, A., et al. (2012). IDH mutation impairs histone demethylation and results in a block to cell differentiation. Nature 483, 474478. McMahon, S.B., Van Buskirk, H.A., Dugan, K.A., Copeland, T.D., and Cole, M.D. (1998). The novel ATM-related protein TRRAP is an essential cofactor for the c-Myc and E2F oncoproteins. Cell 94, 363374. McMahon, S.B., Wood, M.A., and Cole, M.D. (2000). The essential cofactor TRRAP recruits the histone acetyltransferase hGCN5 to c-Myc. Mol. Cell. Biol. 20, 556562. n, M., Ayata, P., Dewell, S., Kriaucionis, S., and Heintz, N. (2012). MeCP2 Melle binds to 5hmC enriched within active genes and accessible chromatin in the nervous system. Cell 151, 14171430. Moran-Crusio, K., Reavie, L., Shih, A., Abdel-Wahab, O., Ndiaye-Lobry, D., Lobry, C., Figueroa, M.E., Vasanthakumar, A., Patel, J., Zhao, X., et al. (2011). Tet2 loss leads to increased hematopoietic stem cell self-renewal and myeloid transformation. Cancer Cell 20, 1124. Muto, A., Hoshino, H., Madisen, L., Yanai, N., Obinata, M., Karasuyama, H., Hayashi, N., Nakauchi, H., Yamamoto, M., Groudine, M., and Igarashi, K. (1998). Identication of Bach2 as a B-cell-specic partner for small maf proteins that negatively regulate the immunoglobulin heavy chain gene 30 enhancer. EMBO J. 17, 57345743. Nakahata, Y., Sahar, S., Astarita, G., Kaluzova, M., and Sassone-Corsi, P. (2009). Circadian control of the NAD+ salvage pathway by CLOCK-SIRT1. Science 324, 654657. Noushmehr, H., Weisenberger, D.J., Diefes, K., Phillips, H.S., Pujara, K., Berman, B.P., Pan, F., Pelloski, C.E., Sulman, E.P., Bhat, K.P., et al.; Cancer Genome Atlas Research Network. (2010). Identication of a CpG island
methylator phenotype that denes a distinct subgroup of glioma. Cancer Cell 17, 510522. Ochiai, K., Katoh, Y., Ikura, T., Hoshikawa, Y., Noda, T., Karasuyama, H., Tashiro, S., Muto, A., and Igarashi, K. (2006). Plasmacytic transcription factor Blimp-1 is repressed by Bach2 in B cells. J. Biol. Chem. 281, 3822638234. Pansuriya, T.C., van Eijk, R., dAdamo, P., van Ruler, M.A., Kuijjer, M.L., Oosting, J., Cleton-Jansen, A.M., van Oosterwijk, J.G., Verbeke, S.L., Meijer, D., et al. (2011). Somatic mosaic IDH1 and IDH2 mutations are associated with enchondroma and spindle cell hemangioma in Ollier disease and Maffucci syndrome. Nat. Genet. 43, 12561261. Park, J., Sarode, V.R., Euhus, D., Kittler, R., and Scherer, P.E. (2012). Neuregulin 1-HER axis as a key mediator of hyperglycemic memory effects in breast cancer. Proc. Natl. Acad. Sci. USA 109, 2105821063. Perillo, B., Ombra, M.N., Bertoni, A., Cuozzo, C., Sacchetti, S., Sasso, A., Chiariotti, L., Malorni, A., Abbondanza, C., and Avvedimento, E.V. (2008). DNA oxidation as triggered by H3K9me2 demethylation drives estrogeninduced gene expression. Science 319, 202206. Pirola, L., Balcerczyk, A., Okabe, J., and El-Osta, A. (2010). Epigenetic phenomena linked to diabetic complications. Nat. Rev. Endocrinol. 6, 665675. ` re, J.J., Alam, N.A., Barwell, J., Barclay, E., Wortham, N.C., Pollard, P.J., Brie Hunt, T., Mitchell, M., Olpin, S., Moat, S.J., et al. (2005). Accumulation of Krebs cycle intermediates and over-expression of HIF1alpha in tumours which result from germline FH and SDH mutations. Hum. Mol. Genet. 14, 22312239. Pollard, P.J., Spencer-Dene, B., Shukla, D., Howarth, K., Nye, E., El-Bahrawy, M., Deheragoda, M., Joannou, M., McDonald, S., Martin, A., et al. (2007). Targeted inactivation of fh1 causes proliferative renal cyst development and activation of the hypoxia pathway. Cancer Cell 11, 311319. Pritchard, J.B. (1995). Intracellular alpha-ketoglutarate controls the efcacy of renal organic anion transport. J. Pharmacol. Exp. Ther. 274, 12781284. Ramsey, K.M., Yoshino, J., Brace, C.S., Abrassart, D., Kobayashi, Y., Marcheva, B., Hong, H.-K., Chong, J.L., Buhr, E.D., Lee, C., et al. (2009). Circadian clock feedback cycle through NAMPT-mediated NAD+ biosynthesis. Science 324, 651654. Rodgers, J.T., Lerin, C., Haas, W., Gygi, S.P., Spiegelman, B.M., and Puigserver, P. (2005). Nutrient control of glucose homeostasis through a complex of PGC-1alpha and SIRT1. Nature 434, 113118. Rowicka, M., Kudlicki, A., Tu, B.P., and Otwinowski, Z. (2007). High-resolution timing of cell cycle-regulated gene expression. Proc. Natl. Acad. Sci. USA 104, 1689216897. Sakata, S.F., Shelly, L.L., Ruppert, S., Schutz, G., and Chou, J.Y. (1993). Cloning and expression of murine S-adenosylmethionine synthetase. J. Biol. Chem. 268, 1397813986. Sauve, A.A., Wolberger, C., Schramm, V.L., and Boeke, J.D. (2006). The biochemistry of sirtuins. Annu. Rev. Biochem. 75, 435465. Schwer, B., Bunkenborg, J., Verdin, R.O., Andersen, J.S., and Verdin, E. (2006). Reversible lysine acetylation controls the activity of the mitochondrial enzyme acetyl-CoA synthetase 2. Proc. Natl. Acad. Sci. USA 103, 10224 10229. n, C., Zwaans, B.M., Silberman, D.M., Gymrek, M., Goren, A., Zhong, Sebastia L., Ram, O., Truelove, J., Guimaraes, A.R., Toiber, D., et al. (2012). The histone deacetylase SIRT6 is a tumor suppressor that controls cancer metabolism. Cell 151, 11851199. Selak, M.A., Armour, S.M., MacKenzie, E.D., Boulahbel, H., Watson, D.G., Manseld, K.D., Pan, Y., Simon, M.C., Thompson, C.B., and Gottlieb, E. (2005). Succinate links TCA cycle dysfunction to oncogenesis by inhibiting HIF-alpha prolyl hydroxylase. Cancer Cell 7, 7785. Shahbazian, M.D., and Grunstein, M. (2007). Functions of site-specic histone acetylation and deacetylation. Annu. Rev. Biochem. 76, 75100. Shi, Y., Lan, F., Matson, C., Mulligan, P., Whetstine, J.R., Cole, P.A., Casero, R.A., and Shi, Y. (2004). Histone demethylation mediated by the nuclear amine oxidase homolog LSD1. Cell 119, 941953. Shimazu, T., Hirschey, M.D., Newman, J., He, W., Shirakawa, K., Le Moan, N., Grueter, C.A., Lim, H., Saunders, L.R., Stevens, R.D., et al. (2013). Suppres-
sion of oxidative stress by b-hydroxybutyrate, an endogenous histone deacetylase inhibitor. Science 339, 211214. Shyh-Chang, N., Locasale, J.W., Lyssiotis, C.A., Zheng, Y., Teo, R.Y., Ratanasirintrawoot, S., Zhang, J., Onder, T., Unternaehrer, J.J., Zhu, H., et al. (2013). Inuence of threonine metabolism on S-adenosylmethionine and histone methylation. Science 339, 222226. Sinclair, K.D., Allegrucci, C., Singh, R., Gardner, D.S., Sebastian, S., Bispham, J., Thurston, A., Huntley, J.F., Rees, W.D., Maloney, C.A., et al. (2007). DNA methylation, insulin resistance, and blood pressure in offspring determined by maternal periconceptional B vitamin and methionine status. Proc. Natl. Acad. Sci. USA 104, 1935119356. Smith, E.H., Janknecht, R., and Maher, L.J., 3rd. (2007). Succinate inhibition of alpha-ketoglutarate-dependent enzymes in a yeast model of paraganglioma. Hum. Mol. Genet. 16, 31363148. Song, Y., Manson, J.E., Lee, I.M., Cook, N.R., Paul, L., Selhub, J., Giovannucci, E., and Zhang, S.M. (2012). Effect of combined folic acid, vitamin B(6), and vitamin B(12) on colorectal adenoma. J. Natl. Cancer Inst. 104, 1562 1575. Starai, V.J., Celic, I., Cole, R.N., Boeke, J.D., and Escalante-Semerena, J.C. (2002). Sir2-dependent activation of acetyl-CoA synthetase by deacetylation of active lysine. Science 298, 23902392. Stefanska, B., Karlic, H., Varga, F., Fabianowska-Majewska, K., and Haslberger, A. (2012). Epigenetic mechanisms in anti-cancer actions of bioactive food componentsthe implications in cancer prevention. Br. J. Pharmacol. 167, 279297. Stover, P.J. (2011). Polymorphisms in 1-carbon metabolism, epigenetics and folate-related pathologies. J Nutrigenet Nutrigenomics 4, 293305. Tahiliani, M., Koh, K.P., Shen, Y., Pastor, W.A., Bandukwala, H., Brudno, Y., Agarwal, S., Iyer, L.M., Liu, D.R., Aravind, L., and Rao, A. (2009). Conversion of 5-methylcytosine to 5-hydroxymethylcytosine in mammalian DNA by MLL partner TET1. Science 324, 930935. Takahashi, H., McCaffery, J.M., Irizarry, R.A., and Boeke, J.D. (2006). Nucleocytosolic acetyl-coenzyme a synthetase is required for histone acetylation and global transcription. Mol. Cell 23, 207217. n, M., Dehne, N., and Bru ne, B. (2011). Hypoxia causes epigeTausendscho netic gene regulation in macrophages by attenuating Jumonji histone demethylase activity. Cytokine 53, 256262. Tefferi, A., Lim, K.H., and Levine, R. (2009). Mutation in TET2 in myeloid cancers. N. Engl. J. Med. 361, 11171118. Teperino, R., Schoonjans, K., and Auwerx, J. (2010). Histone methyl transferases and demethylases; can they link metabolism and transcription? Cell Metab. 12, 321327. Tu, B.P., and McKnight, S.L. (2006). Metabolic cycles as an underlying basis of biological oscillations. Nat. Rev. Mol. Cell Biol. 7, 696701. Tu, B.P., and McKnight, S.L. (2009). Evidence of carbon monoxide-mediated phase advancement of the yeast metabolic cycle. Proc. Natl. Acad. Sci. USA 106, 1429314296. Tu, B.P., Kudlicki, A., Rowicka, M., and McKnight, S.L. (2005). Logic of the yeast metabolic cycle: temporal compartmentalization of cellular processes. Science 310, 11521158. Tu, B.P., Mohler, R.E., Liu, J.C., Dombek, K.M., Young, E.T., Synovec, R.E., and McKnight, S.L. (2007). Cyclic changes in metabolic state during the life of a yeast cell. Proc. Natl. Acad. Sci. USA 104, 1688616891. Turcan, S., Rohle, D., Goenka, A., Walsh, L.A., Fang, F., Yilmaz, E., Campos, C., Fabius, A.W., Lu, C., Ward, P.S., et al. (2012). IDH1 mutation is sufcient to establish the glioma hypermethylator phenotype. Nature 483, 479483. van Haaften, G., Dalgliesh, G.L., Davies, H., Chen, L., Bignell, G., Greenman, C., Edkins, S., Hardy, C., OMeara, S., Teague, J., et al. (2009). Somatic mutations of the histone H3K27 demethylase gene UTX in human cancer. Nat. Genet. 41, 521523. Veech, R.L., Eggleston, L.V., and Krebs, H.A. (1969). The redox state of free nicotinamide-adenine dinucleotide phosphate in the cytoplasm of rat liver. Biochem. J. 115, 609619.
Verdin, E., Hirschey, M.D., Finley, L.W., and Haigis, M.C. (2010). Sirtuin regulation of mitochondria: energy production, apoptosis, and signaling. Trends Biochem. Sci. 35, 669675. Wang, J., Alexander, P., Wu, L., Hammer, R., Cleaver, O., and McKnight, S.L. (2009a). Dependence of mouse embryonic stem cells on threonine catabolism. Science 325, 435439. Wang, J., Hevi, S., Kurash, J.K., Lei, H., Gay, F., Bajko, J., Su, H., Sun, W., Chang, H., Xu, G., et al. (2009b). The lysine demethylase LSD1 (KDM1) is required for maintenance of global DNA methylation. Nat. Genet. 41, 125129. Wang, J., Alexander, P., and McKnight, S.L. (2011). Metabolic specialization of mouse embryonic stem cells. Cold Spring Harb. Symp. Quant. Biol. 76, 183193. Ward, P.S., Patel, J., Wise, D.R., Abdel-Wahab, O., Bennett, B.D., Coller, H.A., Cross, J.R., Fantin, V.R., Hedvat, C.V., Perl, A.E., et al. (2010). The common feature of leukemia-associated IDH1 and IDH2 mutations is a neomorphic enzyme activity converting alpha-ketoglutarate to 2-hydroxyglutarate. Cancer Cell 17, 225234. Waterborg, J.H. (2002). Dynamics of histone acetylation in vivo. A function for acetylation turnover? Biochem Cell Biol. 80, 363378. Waterland, R.A., Dolinoy, D.C., Lin, J.R., Smith, C.A., Shi, X., and Tahiliani, K.G. (2006). Maternal methyl supplements increase offspring DNA methylation at Axin Fused. Genesis 44, 401406. Wellen, K.E., Hatzivassiliou, G., Sachdeva, U.M., Bui, T.V., Cross, J.R., and Thompson, C.B. (2009). ATP-citrate lyase links cellular metabolism to histone acetylation. Science 324, 10761080. Wolff, G.L., Kodell, R.L., Moore, S.R., and Cooney, C.A. (1998). Maternal epigenetics and methyl supplements affect agouti gene expression in Avy/a mice. FASEB J. 12, 949957. Xia, X., Lemieux, M.E., Li, W., Carroll, J.S., Brown, M., Liu, X.S., and Kung, A.L. (2009). Integrative analysis of HIF binding and transactivation reveals its role in
maintaining histone methylation homeostasis. Proc. Natl. Acad. Sci. USA 106, 42604265. Xiao, M., Yang, H., Xu, W., Ma, S., Lin, H., Zhu, H., Liu, L., Liu, Y., Yang, C., Xu, Y., et al. (2012). Inhibition of a-KG-dependent histone and DNA demethylases by fumarate and succinate that are accumulated in mutations of FH and SDH tumor suppressors. Genes Dev. 26, 13261338. Xu, W., Yang, H., Liu, Y., Yang, Y., Wang, P., Kim, S.H., Ito, S., Yang, C., Wang, P., Xiao, M.T., et al. (2011). Oncometabolite 2-hydroxyglutarate is a competitive inhibitor of a-ketoglutarate-dependent dioxygenases. Cancer Cell 19, 1730. Yamazaki, J., Taby, R., Vasanthakumar, A., Macrae, T., Ostler, K.R., Shen, L., Kantarjian, H.M., Estecio, M.R., Jelinek, J., Godley, L.A., and Issa, J.P. (2012). Effects of TET2 mutations on DNA methylation in chronic myelomonocytic leukemia. Epigenetics 7, 201207. Yuan, J., Doucette, C.D., Fowler, W.U., Feng, X.J., Piazza, M., Rabitz, H.A., Wingreen, N.S., and Rabinowitz, J.D. (2009). Metabolomics-driven quantitative analysis of ammonia assimilation in E. coli. Mol. Syst. Biol. 5, 302. Zhang, Y., and Reinberg, D. (2001). Transcription regulation by histone methylation: interplay between different covalent modications of the core histone tails. Genes Dev. 15, 23432360. Zhang, J., Ohta, T., Maruyama, A., Hosoya, T., Nishikawa, K., Maher, J.M., Shibahara, S., Itoh, K., and Yamamoto, M. (2006). BRG1 interacts with Nrf2 to selectively mediate HO-1 induction in response to oxidative stress. Mol. Cell. Biol. 26, 79427952. Zhang, S.M., Cook, N.R., Albert, C.M., Gaziano, J.M., Buring, J.E., and Manson, J.E. (2008). Effect of combined folic acid, vitamin B6, and vitamin B12 on cancer risk in women: a randomized trial. JAMA 300, 20122021. Zhou, X., Sun, H., Chen, H., Zavadil, J., Kluz, T., Arita, A., and Costa, M. (2010). Hypoxia induces trimethylated H3 lysine 4 by inhibition of JARID1A demethylase. Cancer Res. 70, 42144221.
Reversible Disruption of mSWI/SNF (BAF) Complexes by the SS18-SSX Oncogenic Fusion in Synovial Sarcoma
Cigall Kadoch1,2,3,4 and Gerald R. Crabtree2,3,4,*
in Cancer Biology Hughes Medical Institute 3Department of Pathology 4Department of Developmental Biology Stanford University School of Medicine, Stanford, CA 94305, USA *Correspondence: crabtree@stanford.edu http://dx.doi.org/10.1016/j.cell.2013.02.036
2Howard 1Program
SUMMARY
Recent exon sequencing studies have revealed that over 20% of human tumors have mutations in subunits of mSWI/SNF (BAF) complexes. To investigate the underlying mechanism, we studied human synovial sarcoma (SS), in which transformation results from the translocation of exactly 78 amino acids of SSX to the SS18 subunit of BAF complexes. We demonstrate that the SS18-SSX fusion protein competes for assembly with wild-type SS18, forming an altered complex lacking the tumor suppressor BAF47 (hSNF5). The altered complex binds the Sox2 locus and reverses polycomb-mediated repression, resulting in Sox2 activation. Sox2 is uniformly expressed in SS tumors and is essential for proliferation. Increasing the concentration of wild-type SS18 leads to reassembly of wild-type complexes retargeted away from the Sox2 locus, polycomb-mediated repression of Sox2, and cessation of proliferation. This mechanism of transformation depends on only two amino acids of SSX, providing a potential foundation for therapeutic intervention.
INTRODUCTION Exon sequencing in human malignancy has provided paradigmchanging insights into pathogenesis (Lander, 2011) but is often limited by the fact that mutation frequencies are correlative, leaving open the possibility that other primary events are responsible for tumor initiation. This correlative aspect has emerged particularly from recent exon-sequencing studies of human cancers, which have dened frequent mutations in chromatin regulators (Dawson and Kouzarides, 2012). By contrast, precise chromosomal translocations, which dene cancer subsets, provide strong support for an initiating role. Chromatin regulation has often been thought to play supportive roles, and hence
a potential instructive or initiating function for chromatin regulators in human cancer is less clear. Chromatin regulation is essential for appropriate and timely gene expression. This process is achieved by several mechanisms, including DNA methylation, histone modications, and ATP-dependent chromatin remodeling. One of the most wellcharacterized chromatin-remodeling complexes studied to date is the SWI/SNF (BAF) complex, which was discovered in yeast (Peterson and Herskowitz, 1992) and plays a general role in gene activation through nucleosome remodeling, thereby allowing accessibility of transcription factors to their recognition sites. In ies, the Swi2/Snf2 ATPase homolog Brahma was discovered in screens for trithroax genes (Tamkun et al., 1992) and opposes polycomb function. Mammalian complexes have two SWI2-like ATPases (Brg1 and Brm) and a second ATPase, b-actin, and are combinatorially assembled from gene families that encode the 15 subunits. Fewer than half of the subunits are related to yeast SWI/SNF (others are related to RSC and SWR1 subunits [Cairns et al., 1996; Krogan et al., 2003; Mizuguchi et al., 2004]), and hence the name BAF (Brg/Brm-associated complexes) is commonly used. The complexes appear to have undergone evolutionary changes in response to the emergence of multicellularity, polycomb-mediated repression, DNA methylation, and a larger genome size (Wu et al., 2009). The role of combinatorial assembly is seen most clearly in the mammalian nervous system, in which neurons have a family of highly specialized neuron-specic complexes involved in dendritic morphogenesis (Lessard et al., 2007; Wu et al., 2007; Yoo et al., 2009). Recent genetic studies in ies have suggested that the y homolog of the neural-specic BAF (nBAF) subunit BAF53b has an instructive role in targeting dendritic trees to their correct termini (Tea and Luo, 2011). Instructive roles are also suggested from studies demonstrating that forcing the formation of nBAF complexes leads to the conversion of broblasts to neurons (Yoo et al., 2011). Specialized complexes are also found in pluripotent cells (esBAF complexes) (Ho et al., 2009); recreating the esBAF complex subunit composition in broblasts facilitates iPS cell formation (Singhal et al., 2010). These recent studies suggest an instructive role for these ATP-dependent chromatin regulators that was not anticipated from earlier studies.
Cell 153, 7185, March 28, 2013 2013 Elsevier Inc. 71
Recent exome sequencing studies of primary, early human cancers have repeatedly discovered mutations to subunits of polymorphic BAF complexes. Indeed, analysis of the 44 exome sequencing studies published to date indicate that 19.6% of all human cancers have mutations in at least one subunit (C.K., D.C. Hargreaves, C.H. Hodges, L. Elias, L. Ho, J. Ranish, and G.R.C., unpublished data). For example, BAF250a is mutated in 57% of clear-cell ovarian cancers; BAF180 (polybromo) is mutated in 41% of renal cancers (Varela et al., 2011); and medulloblastomas have frequent mutations in Brg, BAF53a, or BAF60b (Jones et al., 2012). The signicance of perturbation to ATP-dependent chromatin-remodeling complexes in tumorigenesis has been most strongly demonstrated in studies focusing on a particular class of tumors, malignant rhabdoid tumors (MRTs), in which the subunit BAF47 (hSNF5) is biallelically inactivated in nearly 100% of cases reported (Versteege et al., 1998). Patients often have inherited a defective SNF5 allele, and the remaining wild-type allele is lost in the tumors, implicating SNF5 as a tumor suppressor. Conditional biallelic inactivation of Snf5 in mouse models results in a fully penetrant phenotype with median onset to tumor development at only 11 weeks (Roberts et al., 2002). The preference for mutation of specic subunits in specic malignancies suggests that different combinatorial assemblies have roles in tissue-specic oncogenic processes, consistent with roles for specialized BAF complexes in neurogenesis and other biologic processes (de la Serna et al., 2001; Lickert et al., 2004; Wu et al., 2009). Because of the possibility that the frequent BAF subunit mutations might be playing a relatively nonspecic role in oncogenesis, we initiated studies on a cancer type, human synovial sarcoma (SS), in which nearly all tumors have a precise translocation involving a specic subunit, indicating that the translocation is the initiating oncogenic event. Human synovial sarcoma accounts for 8%10% of all softtissue malignancies and most commonly arises in the extremities of young adults (Weiss et al., 2001). A recurrent chromosomal translocation, t(X;18)(p11.2;q11.2), fuses the SS18 gene on chromosome 18 to one of three closely related genes on the X chromosome, SSX1, SSX2, and rarely SSX4, resulting in an in-frame fusion protein in which the eight C-terminal amino acids of SS18 are replaced with 78 amino acids (aa) from the SSX C terminus (Clark et al., 1994; de Leeuw et al., 1995; Skytting et al., 1999). This remarkably precise translocation is present in greater than 95% of cases and has been established as pathognomonic for the disease with clinical diagnosis conrmed by karyotyping and RT-PCR for SS18-SSX transcripts (Hiraga et al., 1998; Sandberg and Bridge, 2002). The presence of this translocation is the dening feature of synovial sarcomas and is often the only cytogenetic abnormality (dos Santos et al., 1997; Limon et al., 1991); hence, this is very likely to be the driving oncogenic event in the development of these tumors. However, the mechanism of SS18-SSX transformation has been unclear. Both SS18 and SSX proteins lack known DNA binding motifs, yet they appear to be acting through transcriptional regulatory mechanisms. SS18 is a nuclear protein that has been suggested to interact with chromatin-remodeling factors, such as the Brg/ Brm-containing complexes (Nagai et al., 2001; Thaete et al., 1999), and the transformation potential of the SS18-SSX fusion
72 Cell 153, 7185, March 28, 2013 2013 Elsevier Inc.
has been shown to require Brg/Brm (Nagai et al., 2001). Fusion partners SSX1, 2, and 4 are members of a family of nine human SSX genes that encode highly similar proteins with 73%92% homology (cDNA homology 87%96%) (Smith and McNeel, 2010) and conserved intron/exon junctions. SSX3 and SSX5 have not been found as fusion partners in tumors, although they are highly similar to the oncogenic fusion partners. mRNA expression of SSX genes is restricted to the testes and has been detected at low levels in the thyroid. Here, we demonstrate that SS18 is a dedicated, highly stable subunit of BAF complexes. We nd that the fusion of SS18 with SSX produces a protein that binds to the complex and evicts both the wild-type SS18 and the tumor suppressor BAF47 (SNF5). This altered complex then binds to Sox2, relieving H3K27me3 repression, thereby activating Sox2, which we nd is required for proliferation. Importantly, SS18-SSXdriven complex disruption is determined by a 2 aa hydrophilic region of SSX. Assembly of wild-type complexes and proliferative quiescence can be achieved by increasing the concentration of wild-type SS18, making this region an excellent drug target. RESULTS SS18 Is a Subunit of Mammalian SWI/SNF-like BAF Complexes To better understand the composition of BAF complexes, we used a rapid biochemical/afnity purication approach to isolate endogenous complexes from nontransformed cells. Ammonium sulfate fractionation was followed by rapid afnity purication using a highly specic antibody to a genetically nonessential epitope in the Brg/Brm ATPase subunit (Ho et al., 2009). SS18 (Synovial sarcoma translocation, chromosome 18) peptides were found in highly pure, endogenous BAF complexes in all tissue types examined, with the exception of postmitotic adult neurons. Numbers of peptides and percent coverage for the protein SS18 were comparable to those of established BAF complex subunits, suggesting it is a subunit of BAF complexes (Figure 1A). Immunoprecipitation studies using anti-Brg as well as antibodies specic to other established mSWI/SNF complex components, including BAF250a, BAF155, and BAF47, conrmed the association of SS18 with native BAF complexes; similarly, reciprocal immunoprecipitation using an antibody to SS18 revealed known components of BAF complexes (Figure 1B and Figure S1A available online). Two bands are detected for human SS18 due to alternative splicing (Figure S1B). Purication of complexes using anti-Brg and anti-SS18 antibodies revealed similar banding patterns upon silver stain analyses (Figure S1C). In order to determine whether SS18 was dedicated exclusively to BAF complexes, we performed glycerol gradient sedimentation analyses that demonstrated the presence of SS18 only in those fractions containing Brg and other BAF complex subunits (fractions 1215). SS18 did not associate with polycomb repressive complexes PRC1 or PRC2, as indicated by Bmi1 or Ezh2 immunoblots, respectively, or as a free monomer in earlier fractions of the gradient (Figure 1C). Results were comparable in several cell types assayed, including cell lines ES E14, Raji, 293T, and CCRF-CEM, as well as primary human broblasts.
A
PROTEIN Brg (Smarca4) BAF250a (Arid1a) BAF250b (Arid1b) BAF155 (Smarcc1) BAF57 (Smarce1) BAF53a (Actl6a) BAF47 (Smarcb1) SS18 Bmi1 Ezh2 ESC % coverage # peptides 45.7 101 55.5 34 17.5 27 62.8 130 76.2 23 52.9 26 54 27 39.3 14 n.d. n.d. n.d. n.d. Cell type MEF P1.5 whole brain % coverage # peptides % coverage # peptides 34.3 68 42.4 123 40.8 21 44.2 163 25.3 52 34.3 96 54.4 80 42 75 67.2 32 66.3 111 37.1 18 50.8 31 27 14 43.1 26 13.7 3 13.1 10 n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d.
B
kDa 250 190 55 47
*n.d. (not detected)
C
10%
Fx:
glycerol 1 2 3 4 5
Brg Brg
10 11 12
13
14
15 16
17
18
Brg BAF250 BAF47 SS18 Bmi1 Ezh2
<300kDa
~2MDa
Nuclear extracts
[Urea] (M)
IP: anti-Brg
Density (% of Control)
100 80 60 40 20 0
Ig G An ti An -Brg tiSS 18
BAF250 Brg SS18 BAF47 30% 19 20
>5MDa
Brg SS18
5 1 1.2
0 0.2
0.7
1.7
5 2.5 5
UREA TREATMENT 0-->5 M
Brg BAF250a
BAF170 BAF155
Anti-Brg IP Immunoblot
SS18 BAF47 BAF45d
BAF250a BAF47 BAF155

0 1 2 3 4
BAF170 BAF45d 5
[urea] (M)
Figure 1. SS18 Is a Dedicated, Stable Subunit of mSWI/SNF (BAF) Complexes
(A) Composition of BAF complexes isolated from ES cells, MEFs, and brain as determined by mass spectrometric analysis. See also Figure S1C. (B) Immunoprecipitation (IP) using anti-Brg and anti-SS18 antibodies in 293T nuclear extracts (NEs). See also Figures 1A and 1B. (C) Glycerol gradient (10%30%) analysis on ES cell NEs. (D) Left: Schematic for urea-based denaturation analyses. Right: Anti-Brg IPs on 293T NE preps treated with 05 M urea. (E) Quantitative densitometry on urea denaturation immunoblots.
Using urea-based denaturation studies, we determined that SS18 was remarkably stably bound to the complex, to a greater extent than most other subunits, including BAF47, BAF155, and BAF 170, requiring denaturing conditions of greater than 5 M urea to dissociate (Figures 1D and 1E), similar to ribosomal subunits. The observation that SS18 remains bound when other subunits have dissociated indicates that SS18 binds directly to a stable core complex of Brg, BAF53a, and b-actin (Zhao et al., 1998). These results demonstrate that SS18 is a dedicated subunit of mSWI/SNF or BAF complexes, with binding characteristics similar to those of ribosomal subunits.
SS18-SSX Integrates into BAF Complexes and Alters Complex Composition The invariant molecular feature of human synovial sarcoma is the SS18-SSX fusion protein, in which the C-terminal 78 aa of SSX are fused in frame with aa1379 of the SS18 subunit (Figure 2A). To investigate the oncogenic mechanism, we used two biphasic synovial sarcoma (SS) lines, Aska-SS and Yamato-SS, both of which bear the SS18-SSX1 chromosomal translocation (Naka et al., 2010). Anti-Brg immunoprecipitation studies performed on nuclear extracts isolated from synovial sarcoma cell lines, as compared to control 293T cells (and various other cell types),
387aa
188aa
t(X;18)(p11.2;q11.2) SNH 54aa QPGY 200aa SSXRD 34aa
SS18
379aa
SSX COOH
78aa
8aa
BAF250 Brg BAF170 BAF155 SS18 BAF47
29
COOH
COOH
Ig G Br g Ig G Br g Ig G Br g
SS18
SSX
C
10%
glycerol
30%
Fx: Brg BAF250 BAF170 SS18 BAF47 Ezh2
1 2
3 4 5
6 7 8
9 10 11 12 13 14 15 16 17 18 19 20
SS18-SSX SS18
D
Aska-SS Fx: 3 4 15 16 Yamato-SS 3 4 15 16 Brg SS18 SSX1
E
Un
Immunodepletion
Ab used: Round: BAF155 1 2 SSX1 1 2
Brg BAF155 SS18 BAF47 GAPDH
de pl .
293T
Un
Ab used: Round:
BAF155 1 2
SSX1 1 2
Brg BAF155 SS18 BAF47 TBP GAPDH
Aska-SS
Figure 2. mSWI/SNF (BAF) Complexes Are Disrupted in Synovial Sarcoma Cells Bearing the SS18-SSX1 Fusion Protein
(A) Diagram of the SS18-SSX fusion protein resulting from t(x;18) translocation, hallmark to synovial sarcoma. (B) Anti-Brg IP (left) and total protein (right) in 293T cells as compared to synovial sarcoma (SS) cell lines, Aska-SS, and Yamato-SS. See also Figures S2AS2C. (C) Glycerol gradient (10%30%, fractions 120) analysis on Aska-SS cell NEs. See also Figures S2D and S2E. (D) Side-by-side comparison of fractions 3,4 and 15,16 of Aska-SS glycerol gradient analysis. (E) Immunodepletion studies performed on 293T and Aska-SS cells using anti-BAF155 and anti-SSX1 antibodies. Undepleted, antibody not added.
demonstrated that when SS18 was fused to its translocation partner SSX, the SS18-SSX1 fusion protein was indeed bound to BAF complexes, as reected by an appropriate upshift in molecular weight of SS18 from 55 kDa to 66 kDa upon immunoblot analysis (Figure 2B, left). Remarkably, we observed that both synovial sarcoma lines, as compared to several other cell types assayed, exhibited lower to absent total protein levels of the tumor suppressor subunit BAF47 (hSNF5 or INI1) (Figures 2B, right, and S2A, left), while transcripts were largely comparable (Figure S2A, right). Immunoprecipitated BAF complexes containing the SS18-SSX1 fusion protein showed nearly absent levels of wild-type SS18 on the complex. Input protein levels of the wild-type-sized SS18 protein were also lowered, as were mRNA levels, suggesting reduced transcription (Figure S2B), consistent with previously reported ndings (Brodin et al.,
2001). In addition, a prominent Brg peak is located at the promoter and in an intronic region of the SS18 gene as determined by chromatin immunoprecipitation (ChIP)-seq analysis in murine ES cells (Figure S2C) (Ho et al., 2011), suggesting autoregulation of this locus. Density sedimentation analyses performed on nuclear extracts isolated from Aska-SS and Yamato-SS lines revealed disruption of BAF complex composition, in that wild-type SS18 protein no longer associated with the BAF complex fractions (fractions 15,16) and rather existed in fractions 3 and 4, suggesting its presence as a monomer (Figures 2C and S2D). Quantitative densitometry of an anti-SS18 immunoblot of the glycerol gradient revealed that only a small percentage (2%8%) of BAF complexes contain the wild-type SS18 protein in these cells (Figure S2E). Side-by-side molecular weight comparisons indicated that the SS18-SSX fusion protein,
de
pl .
3T As ka Ya -SS m at oS
BAF250 Brg BAF170 BAF155 SS18 BAF47
SNH 54aa
QPGY 200aa
KRAB 63aa
SSXRD 34aa
293T
Aska-SS Yamato-SS
Input
in both SS lines, was almost entirely associated with the BAF complex (denoted by Brg peaks in fractions 15,16) and the wild-type SS18 protein was present, albeit at lower protein levels, in the monomeric fractions of the gradient (fractions 3,4) (Figure 2D). This was further conrmed by immunoblotting using an anti-SSX1 antibody, which demonstrated the presence of SSX1 only in fractions containing Brg. As shown above, in SS lines containing the SS18-SSX fusion, BAF47 no longer associated with BAF complexes and was nearly absent from nuclear extracts indicative of degradation. This is particularly interesting given that BAF47 is a known tumor suppressor; loss of this subunit from the complex as a result of the integration of SS18-SSX might produce functional consequences similar to those of SNF5 inactivation. In order to further assess the degree of dedication of SS18 and SS18-SSX to the BAF complex, we performed depletion studies using two rounds of immunoprecipitation with polyclonal antibodies specic to a known complex subunit, BAF155, as well as to SS18s fusion partner, SSX1 (Figure 2E). In 293T cells, BAF155 antibodies depleted SS18 protein from the nuclear extracts; SSX1 antibody did not deplete the lysate, as expected, in the wild-type setting. In the Aska-SS cell line, immunodepletion using the SSX1 antibody signicantly depleted complex subunits Brg, BAF155, and SS18-SSX proteins from nuclear extracts to comparable levels as with anti-BAF155 antibody. These results collectively demonstrate that both wild-type SS18 and the SS18-SSX1 fusion protein in synovial sarcoma are dedicated to BAF complexes but that the fusion protein alters subunit composition. To understand how incorporation of SS18-SSX alters the biochemical subunit composition of BAF complexes, we produced N-terminally GFP-tagged constructs of SS18 FL (full length, aa1387), SS18 aa1379 (lacking the last C-terminal 8 aa, which are lost in the fusion), and SS18-SSX using a pEGFP-based expression system (Figure 3A). Previous studies have established that the N-terminal SNH domain of SS18 is responsible for its BAF complex association (Nagai et al., 2001). Anti-GFP immunoprecipitations were performed to isolate those BAF complexes that had incorporated the exogenously introduced SS18 or SS18-SSX variants. Intriguingly, we found that expressing the SS18-SSX fusion protein resulted in the loss of BAF47 from the complex at 72 hr posttransfection (Figure 3B). Wild-type SS18 FL or SS18 1379 both incorporated into BAF complexes but did not alter BAF47 binding to the complex. Input levels of BAF47 at this time point (72 hr), following introduction of SS18-SSX (and all variants tested), were comparable to those of untreated cells. Immunoblot analysis performed on total input protein harvested at 96 hr posttransfection with SS18-SSX indicated a marked decrease in BAF47 levels, while mRNA levels remained stable, suggesting that BAF47 is rst lost from the complex upon integration of SS18-SSX and subsequently degraded (Figures 3C and S3). To understand the means by which BAF47 is degraded under normal conditions, we performed cyclohexamide (CH) chase experiments over 24 hr, plus and minus proteasome inhibitor treatment using MG-132 at the 24 hr time point. The protein half-life of BAF47 was approximately 10 hr after the addition of CH; BAF47 levels could be rescued from CH treatment with MG-132 to >85% of control levels, indicative of proteasome-mediated degradation (Fig-
ure 3D). Treatment of Aska-SS cells with MG-132 resulted in a substantial increase in BAF47 total protein levels (Figure 3E). Upon infection of SS18-SSX1 into 293T broblasts, wild-type SS18-containing complexes were readily replaced by SS18SSX-containing complexes (fractions 14,15,16), and BAF47 levels were reduced as determined by glycerol gradient analyses (Figure 3F). Wild-type SS18 was observed in free, monomeric fractions of the glycerol gradient (fractions 35), as well as Brgassociated fractions 911 (likely indicative of partially formed, lower molecular weight complexes). These studies indicate that the SS18-SSX fusion incorporates into BAF complexes, replacing wild-type SS18 and ejecting and destabilizing BAF47. To understand whether low protein levels of BAF47 result specically from the presence of the SS18-SSX1 fusion in SS cells, we generated shRNA-based knockdown (KD) constructs specic for the 30 UTR of SSX (based on Takenaka et al., 2010) to exclusively target SS18-SSX, but not wild-type SS18. Remarkably, we noted a substantial increase in BAF47 total protein levels upon KD of the SS18-SSX oncogenic fusion (Figure 3G). In addition, wild-type SS18 protein levels increased, suggesting relieved repression of SS18 upon KD of the SS18SSX fusion. We assessed the effect of SS18-SSX KD on proliferation of both synovial sarcoma cell lines. Importantly, KD of the SS18-SSX fusion and of Brg, to which the SS18-SSX fusion was bound, resulted in a profound decrease in proliferation of synovial sarcoma cells (Figure 3H). By contrast, KD of wildtype SS18 and BAF47, subunits not contained in the SS18SSX-containing BAF complexes, had little to no effect on synovial sarcoma cell proliferation (Figure 3H), suggesting that the aberrant residual complex is responsible for driving and maintaining cell proliferation. In human primary broblasts with wild-type complexes, KD of Brg, SS18, and BAF47 reduced proliferation; KD of SS18-SSX1 did not alter proliferation as compared to control hairpin (Figure 3I). These studies indicate that the eviction of BAF47 inactivates it and that it is no longer required for proliferation of the SS cell lines. Hence, the free BAF47 protein does not acquire a new function enabling transformation. Synovial Sarcoma Cell Gene Expression Features Recapitulated: SS18-SSX Induces Sox2 Expression Several studies have demonstrated that SS cells harbor stemcell-like gene expression proles (Garcia et al., 2012; Naka et al., 2010). Moreover, Roberts and colleagues observed that tumors lacking the BAF47 tumor suppressor subunit also express stem-cell-like signatures (Wilson et al., 2010). Naka and colleagues demonstrated that Aska-SS and Yamato-SS lines, as well as 15/15 human tumor specimens of synovial sarcoma tested, express mRNA transcripts of pluripotency factors Sox2, Oct4, and Nanog (Naka et al., 2010). We focused on Sox2 because of its role in oncogenesis (Bass et al., 2009). Introduction of SS18-SSX dramatically induced Sox2 mRNA in primary, untransformed human neonatal foreskin broblasts by 15 days postinfection and selection (Figure 4A). This induction was specic to the full SS18-SSX1 fusion and did not occur when the C-terminal 34 aa of the conserved SSXRD domain were removed from SSX1. To determine if Sox2 mRNA induction was driven by the partially formed complexes, we tested the
Figure 3. SS18-SSX1 Ejects BAF47 and Wild-Type SS18 to Recapitulate BAF Complex Phenotype in Synovial Sarcoma Cells
(A) N-terminal GFP-tagged constructs of SS18 FL, SS18 1379 (8 aa), and SS18-SSX. (B) Anti-GFP IP of BAF complexes 72 hr posttransfection with various pEGFP constructs in 293T broblasts. See also Figure S3. (C) Immunoblot analysis on total protein isolated from transfected 293T cells at time t = 0 to t = 168 hr posttransfection with GFP-SS18-SSX. (D) Top: Cyclohexamide (CH) chase treatment of 293T cells, t = 0 to t = 24 hr, MG-132 proteosome inhibitor. Bottom: Quantitative densitometry of BAF47 protein levels on immunoblot. (E) Immunoblot analysis for BAF47 protein in Aska-SS cells treated with MG-132 proteosome inhibitor for t = 8 and t = 16 hr. (F) Glycerol gradient analyses on 293T cells infected with lentivirus (LV) containing either empty vector (top half) or SS18-SSX (bottom half). (G) shRNA-mediated knockdown (KD) of SS18-SSX and wild-type SS18 in Aska-SS cells. (H) Cell proliferation analyses of Aska-SS cells infected with shScramble control vector or delivery constructs containing shRNA KD to BAF subunits. Cells plated in triplicate at 105 cells/well per condition. Error bars, SD of n = 3 experiments. (I) Cell proliferation analyses of human primary neonatal foreskin broblasts infected with shScramble control vector or shRNA KD to BAF complex subunits. Cells plated in triplicate at 105 cells/well per condition. Error bars, SD of n = 3 experiments.
Figure 4. SS18-SSX1 Induces Sox2 mRNA Expression, which Drives SS Cell Proliferation
(A) Sox2 mRNA levels at day 10 postinfection with LV containing either SS18/SS18-SSX variants or shRNAs to BAF complex subunits. (Normalized to GAPDH; ***p < 0.005, **p < 0.01, error bars reect SD in n = 5 separate experiments.) See also Figure S4A. (B) Time course of Sox2, Oct4, and Nanog mRNA levels postinfection with SS18-SSX-containing LV. (Normalized to GAPDH; error bars reect SD in n = 5 separate experiments.) (C) shRNA-mediated knockdown of Sox2 in Aska-SS cells: top, immunoblot analysis; bottom, Sox2 mRNA levels. (D) Proliferative analysis of Aska-SS cells treated with Sox2 shRNA KD LV. Control, shScramble. (E) Left: Immunoblot analysis on Aska-SS cells treated with shControl or with either shSS18-SSX1 or shSox2-1. Right: Sox2 mRNA relative expression (normalized to GAPDH). (F) Left: Anti-BAF155 ChIP on human primary broblasts treated with either empty vector or SS18-SSX1, with subsequent qPCR for regions at the human Sox2 promoter and two Sox2 transcription factor (TF) binding sites within the exon. Right: Anti-H3K27me3 ChIP. Error bars, SD of n = 3 experiments. See also Figures S4B and S4C.
(legend on next page)
effect of shRNA-mediated KD of SS18 and BAF47 in broblasts on Sox2 mRNA induction. Intriguingly, KD of SS18 and BAF47 both resulted in a statistically signicant increase in Sox2 mRNA to levels nearly comparable to those resulting from overexpression of SS18-SSX (Figure 4A). At the protein level, BAF47 and SS18 appear to reciprocally regulate one anothers stability in broblasts as determined by KD of BAF47 and SS18 and immunoblot analysis for protein levels of each (Figure S4A). KD of Brg alone resulted in >70% reduction in protein levels, but did not induce Sox2. Collectively, these data suggest that the activity of aberrant complexes, which lack BAF47 and wildtype SS18, are responsible for Sox2 mRNA induction. Sox2 mRNA levels increased 23-fold by day 25 postinfection with SS18-SSX1 as compared to control (Figure 4B). Oct4 and Nanog mRNA were not induced signicantly. We sought to determine whether Sox2 was important for synovial sarcoma cell proliferation. To this end, we generated lentivirus containing two different shRNA hairpins to Sox2, both of which effectively reduced Sox2 mRNA and protein in Aska SS cells (Figure 4C), and assessed proliferative capacity in vitro. shRNA-mediated KD of Sox2 profoundly reduced proliferation of Aska-SS cells as compared to scrambled shRNA control (Figure 4D). Intriguingly, upon KD of the SS18-SSX1 fusion, Sox2 mRNA and protein levels were reduced in Aska-SS cells to levels comparable to those of cells treated with Sox2 shRNA itself (Figure 4E), indicating that elevated levels of Sox2 were specically due to the presence of SS18-SSX fusion. To understand the mechanism of Sox2 induction by SS18SSX, we assessed BAF complex occupancy at the Sox2 promoter as well as two clusters of transcription factor (TF) binding sites within the Sox2 exonic region using our afnitypuried BAF155 polyclonal antibody. Intergenic regions were selected as normalization controls. SS18-SSX1-infected primary human broblasts demonstrated a signicant increase in BAF complex occupancy at all three sites within the human Sox2 locus as compared to control broblasts (Figure 4F). In MEFs, there is a prominent H3K27me3 peak over the Sox2 locus as shown by MEF ChIP-seq studies (Mikkelsen et al., 2007), consistent with absent Sox2 expression in these cells (Figure S4B). Lentiviral introduction of SS18-SSX1 into primary human broblasts resulted in a striking decrease in H3K27me3 enrichment at all three sites tested within the Sox2 locus (Figure 4F). To determine if the 78 aa tail of SSX was itself responsible for the targeting of BAF complexes to the Sox2 locus (perhaps by binding a transcription factor), we infected human broblasts with V5-tagged SSX78aa (as well as SS18 FL and SS18-SSX).
However, we did not nd that the 78 aa SSX fragment localized to the Sox2 locus (Figure S4C). These studies indicate that the SS18-SSX fusion functioning within the altered BAF complexes binds to and activates the Sox2 locus in broblasts by disrupting H3K27me3-mediated repression, which is likely directed by the actions of PRC2, the only complex known to place this mark (Chamberlain et al., 2008). Molecular Requirements of SS18-SSX for BAF47 Ejection from BAF Complexes Because expression of SS18-SSX1 resulted in the ejection and subsequent degradation of the BAF47 subunit, we aimed to understand the features of the 78 aa SSX tail that could be responsible for this. We generated a series of truncation mutants: deleting the conserved SSXRD domain of 34 aa, deleting half of the SSXRD domain (17 aa, hydrophobic), and adding amino acids in increments of 10 aa to the SS18 C terminus (+10 through +70). We noted that SS18+10 through SS18+70 did not result in signicant ejection of BAF47 from the complex as determined by immunoblot analysis and quantitative densitometry performed on immunoprecipitated complexes (Figure 5A). This implies that a region in the last 8 aa (SDPEEDDE) is required for BAF47 ejection. Deleting 1/2SSXRD resulted in slightly decreased levels of BAF47. Upon introduction of these variants into human broblasts, Sox2 mRNA induction was only observed with SS18-SSX1 (Figure 5B). Because none of these truncation mutants fully recapitulated the SS18-SSX1-induced BAF47 ejection and Sox2 mRNA induction phenotype, we turned to the fact that the only translocations that have been observed in human synovial sarcoma are SS18-SSX1, SS18-SSX2, and SS18-SSX4. SS18SSX3 has never been observed in a human tumor. This family of nine genes (SSX1-9) located at ch Xp11.2 is highly similar; protein homology among members ranges from 73%93% (Smith and McNeel, 2010). We used Kyte-Doolittle hydrophobicity analysis to compare the 78 C-terminal aa of SSX1,2,4 versus SSX3, which revealed a signicant difference in hydrophobicity between aa4050 (Figure 5C), as highlighted. Upon peptide alignment of the 78 aa of SSX14, it became clear that the amino acid composition at position 43, 44 was most discrepant between the SSX members observed in human SS tumors (1,2 and 4) and the nononcogenic SSX3. SSX1,2,4 contains lysine (K) and arginine (R), glutamic acid (E) and arginine (R), and lysine (K) and threonine (T), respectively, at position 43,44, whereas SSX3 contains a methionine (M) and isoleucine (I) at these positions (Figure 5D, arrows). Given that SSX1 is
Figure 5. Molecular Requirements of the 78 aa SSX Peptide for BAF47 Subunit Ejection from mSWI/SNF-like BAF Complexes
(A) Left: Immunoblot analysis for BAF47 and Brg on anti-Brg IPs (top) of 293T cells transfected with various SS18/SS18-SSX constructs (bottom). Right: Quantitative densitometry depicting BAF47/Brg protein ratios in IP studies. (B) Sox2 mRNA levels in human broblasts day 15 postinfection with LV containing SS18 and SS18-SSX variants. Error bars = SD. (C) Hydrophobicity determination using the Kyte-Doolittle algorithm for 78 C-terminal amino acids (aa) of SSX1-SSX4 proteins. Region of signicant difference highlighted in yellow. (D) Peptide alignment of SSX1- SSX5 C-terminal 78 aa. Pink arrows indicate amino acids of signicant difference between SSX1/2/4 and SSX3 or SSX1/2/4 and SSX5; yellow highlight indicates regions determined to be critical for BAF47 ejection. (E and F) Immunoblot analysis for BAF47 and Brg on anti-GFP IPs of 293T cells transfected (E) with constructs as per above as well as SS18-SSX3 and (F) with SS18-SSX1, SS18-SSX1Daa43,44 (KR/MI), SS18-SSX3, and SS18-SSX3Daa43,44 (MI/KR). See also Figure S5A. (G) Quantitative densitometry depicting BAF47/Brg protein ratios in IP studies. Error bars = SD. (H) Sox2 mRNA levels in human broblasts day 15 postinfection with LV containing various constructs. See also Figure S5B. Error bars = SD.
a common fusion partner of SS18 in synovial sarcoma and SSX3 is not, we then sought to understand if SSX3 fused to SS18 could result in BAF47 ejection and Sox2 induction. To this end, we generated an SS18-SSX3 fusion protein (379 aa of SS18 fused to 78 aa of the SSX3 C terminus). SS18-SSX3 was able to integrate into BAF complexes, as assessed by anti-GFP immunoprecipitation of BAF complexes, but failed to eject BAF47 (Figure 5E) from the complexes. Remarkably, replacement of amino acids 43,44 of SSX1 (KR) with those of SSX3 (MI) in the SS18-SSX1 fusion resulted in substantial loss of the ability to displace BAF47 (Figure 5F). Reciprocal amino acid substitution at position 43,44 in SS18-SSX3 (MI to KR) resulted in the gained ability of SS18-SSX3 to eject BAF47. Comparative densitometry accounting for the BAF47/Brg ratio is shown, representative of n = 3 experiments (Figure 5G). Intriguingly, SS18-SSX1, as well as SS18-SSX3 (D43,44 MI/KR), signicantly induced Sox2 mRNA; no other variant produced this phenotype (Figure 5H), lending further evidence that the loss of BAF47 (hSNF5) from mSWI/SNF complexes is necessary for the induction of Sox2 mRNA expression in synovial sarcoma. All three fusions reported in human synovial sarcomas (SSX1,2,4) produced BAF47 eviction, whereas SS18-SSX3 and SS18-SSX5 (which bears an amino acid change in the last 8 aa of SSX) fusions did not (Figures S5A and S5B). Reversibility of BAF Complex Subunit Composition and Targeting in Human Synovial Sarcoma Our observation that SS18 was displaced or failed to assemble into BAF complexes in the presence of somewhat higher concentrations of the SS18-SSX fusion protein (Figures 2C and 3F) led us to investigate the possibility that the transforming fusion protein and the wild-type protein might exist in a concentration-dependent equilibrium or could be competing for assembly into newly formed complexes. Urea-based denaturation experiments demonstrated that SS18 and SS18-SSX are both stably bound to BAF complexes and dissociate to comparable degrees from 0 to 8 M urea as shown by immunoblot and quantitative densitometry analyses (Figures 6A, 6B, and S6). BAF complex components dissociated at comparable levels across the urea denaturation series from V5-tagged SS18 and SS18-SSX, indicative of equal afnity binding of wild-type SS18 and SS18-SSX (Figure 6C). Moreover, Brg and b-actin remained bound to V5-SS18/SS18-SSX-puried complexes to >5 M urea, suggesting that SS18/SS18-SSX is part of a highly stable core complex of Brg, BAF53a, and b-actin. Given these ndings and having observed that shRNA-mediated KD of the SS18-SSX1 fusion could restore BAF47 total protein levels (Figure 3G), we sought to determine whether overexpression of wild-type SS18 could also be sufcient to allow normal complexes to reform in synovial sarcoma cell lines and whether this could reverse the misassembly of synovial sarcoma BAF complexes and correct the gene expression phenotypes. Intriguingly, introduction of SS18 FL or SS18 1379 resulted in a profound increase in BAF47 total protein levels by day 10 postinfection (Figure 6D, left). Moreover, BAF complexes in Aska-SS cells infected with SS18 regained normal incorporation of wildtype SS18 and BAF47 subunits, suggesting concentrationdriven reintegration of SS18 (Figure 6D, right). Introduction of
SS18-SSX1 into 293T broblasts resulted in reduction of BAF47 total protein to a comparable degree as shRNA-mediated KD of BAF47 (Figure 6E). These studies indicate that the SS18SSX fusion protein and the wild-type SS18 protein compete for assembly into BAF complexes and that the transforming fusion protein can be displaced from BAF complexes to yield wildtype complexes by increasing the concentration of the wildtype SS18 protein. Proliferation of SS cells was inhibited by introduction of wildtype SS18 and SS18 1379, to a similar degree as in cells treated with shRNA-mediated KD of the SS18-SSX1 fusion (Figure 6F). In contrast, introduction of SS18-SSX into SS18-SSX-bearing synovial sarcoma Aska-SS cells had no appreciable effect on proliferation as compared to control. Sox2 mRNA expression levels in Aska-SS cells were reduced by 3- and 4-fold, upon overexpression of SS18 FL and SS18 1379, respectively (Figure 6G). In contrast, overexpression of SS18-SSX1 in these lines already bearing one translocated allele caused Sox2 mRNA levels to increase 1.7-fold above control levels relative to empty vector control, indicating that the levels of Sox2 produced by the SS18-SSX fusion protein were not at maximum. Finally, Aska-SS cells infected with SS18 FL to reverse the BAF complex phenotype exhibited a dramatically decreased occupancy of BAF complexes at the human Sox2 locus with a concomitant increase in H3K27me3 occupancy (Figure 6H). These studies indicate that normal BAF complexes can be reassembled in malignant cells by overexpression of the wild-type SS18 protein, leading to BAF complex removal from the Sox2 gene and resumption of normal repression of Sox2 by H3K27 trimethylation. Finally, we aimed to test the potential for BAF47 overexpression to promote reassembly of wild-type BAF complexes containing BAF47 and SS18 in SS cells and its effect on proliferation. Notably, overexpressed V5-tagged BAF47 was unable to bind SS18-SSX-containing complexes in both SS cell lines tested, as evidenced by low protein levels on complexes detected by anti-Brg and anti-V5 immunoprecipitations as well as by total protein immunoblots, suggestive of rapid degradation (Figure 7A). To test whether shifting aberrant complex assembly back to that of wild-type would allow for integration of the exogenous BAF47-V5 into complexes, we infected SS cells containing BAF47-V5 with either SS18 FL or shSS18-SSX. Indeed, in both lines, overexpression of SS18 FL or KD of the SS18-SSX fusion resulted in increased incorporation and stabilization of BAF47-V5 as indicated by anti-Brg immunoprecipitation (Figure 7B). Intriguingly, BAF47 overexpression had no effect on SS cell proliferation in culture; however, proliferation was dramatically attenuated upon cointroduction of overexpressed SS18 FL or KD of SS18-SSX, suggesting that BAF47 can only assemble into wild-type SS18-containing complexes and not complexes bearing the SS18-SSX fusion (Figure 7C). DISCUSSION Our studies demonstrate that in the two synovial sarcoma cell lines we have used, the fusion of SS18 with SSX, which is diagnostic of this tumor type, leads to assembly of aberrant BAF complexes that become targeted to the Sox2 locus, with loss
Figure 6. Reversible Integration, Gene Expression, and Occupancy by SS18 and SS18-SSX Containing mSWI/SNF (BAF) Complexes
(A) Denaturation studies using 08 M urea with subsequent immunoblot analysis for SS18 in 293T cells and SS18-SSX in Aska-SS cells. See also Figure S6. (B) Quantitative densitometry of SS18 or SS18-SSX1 protein immunoblots from n = 3 experimental replicates of urea denaturation 0 < [urea] < 8 M. y axis: band quantitation/untreated control. Error bars = SD. (C) IP using anti-V5 antibody in urea-treated nuclear extracts isolated from 293T broblasts infected with either V5-SS18 or V5-SS18-SSX with immunoblotting for BAF complex components. (D) Left: Immunoblot analysis on total protein isolated from Aska-SS cells with either SS18 or SS18 1-379 or SS18-SSX1 introduced via LV. Right: Anti-Brg IP of complexes in either empty vector or V5-SS18-FL-treated conditions. (E) Introduction of SS18-SSX1 and shBAF47 into 293T cells with subsequent immunoblot analysis on total protein. (F) Cell proliferation analyses of Aska-SS cells infected with control vector, SS18, SS18 1-379, and SS18-SSX. Error bars = SD. (G) Sox2 mRNA relative expression (normalized to GAPDH) 10 days postinfection with LV containing either control shScramble or overexpression of SS18, SS18 1-379, or SS18-SSX. Error bars = SD. (H) Left: Anti-BAF155 ChIP on Aska-SS cells treated with either empty vector or SS18 FL, with subsequent qPCR for regions at the human Sox2 promoter and two Sox2 transcription factor (TF) binding sites within the exon. Right: Anti-H3K27me3 ChIP at Sox2 locus in Aska-SS control-treated and SS18 FL-treated cells. Error bars = SD.
Figure 7. A Model for Reversible Transformation by the SS18-SSX1 Oncogenic Fusion

(A) Left: Anti-Brg IPs on 293T, B35, Aska-SS, and Yamato-SS cells bearing introduced BAF47-V5. Middle: Anti-V5 IPs. Right: Total protein inputs. (B) Anti-Brg IPs on nuclear extracts of Aska-SS and Yamato-SS cells with stably introduced BAF47-V5 and coinfection with either control vector, SS18 FL, or shSS18-SSX. (C) Proliferation analyses of Aska-SS cells infected with control vector, BAF47-V5, or BAF47V5+coinfected SS18 FL, BAF47-V5+coinfected shSS18-SSX. Error bars = SD. (D) Model for reversible disruption of BAF complex composition and action upon SS18-SSX incorporation.
of repressive H3K27me3 marks, which drives Sox2 expression and proliferation of these cells (Figure 7D). The observation that Sox2 is activated in all SS studied (Naka et al., 2010) suggests this is a general mechanism of oncogenesis in these tumors. We nd that the SS18-SSX fusion incorporates into BAF complexes and activates Sox2 expression, explaining the uniform activation of this gene in SS. But, how do complexes containing the SS18-SSX fusion activate Sox2? BAF complexes containing the SS18-SSX fusion could be targeted by the interaction of SSX with a factor that binds the Sox2 locus. Alternatively, an incorrectly assembled complex could target the Sox 2 locus by changes to bromo-, chromo-, and PHD domain presentation. We nd that the 78 aa of SSX alone are not targeted to the Sox2 locus when expressed in human broblasts (Figure S4B), indicating that it is the aberrantly assembled complex that targets the inactive Sox2 locus, reversing H3K27Me3-mediated repression, and leading to Sox2 activation. Remarkably, the wild-type SS18 protein is capable of replacing the SS18-SSX fusion in BAF complexes when expressed at somewhat higher levels than the fusion protein. The incorporation of wild-type and mutant proteins is unlikely to be due to
direct binding competition. This conclusion arises from the fact that 8 M urea is required to remove either the wildtype SS18 protein from the wild-type complexes or the SS18-SSX fusion from the malignant complexes. Hence, the two proteins most likely compete for assembly into complexes, with the product of the fusion allele winning in SS cells because of increased concentration. The ability of SS18-SSX to disrupt BAF complexes maps to two regions of the SSX protein: the C-terminal 8 aa (SDPEEDDE) and a polar region of 2 aa present in the oncogenic members of the SSX family of proteins. Substitution of KR with MI, found in the nontransforming SSX3, restores normal complex assembly and gene regulation; substitution of MI with KR in SS18-SSX3 results in BAF47 ejection and increased Sox2 mRNA. In this regard, SSX5 is interesting in that it has amino acids lysine (K) and threonine (T) at position 43, 44, combined with an amino acid substitution of P for E in the 8 terminal aa; SS18-SSX5 has not been found in translocations and does not eject BAF47, conrming the importance of both regions for oncogenicity. These two regions could interact to facilitate complex dissolution or form dimers in the malignant complexes. Structural studies will be necessary to dene the precise mechanism. However, the ability of such a small region to lead to complex dissolution and the observation that the wild-type and malignant proteins are in a dynamic equilibrium indicates that the fusion containing the 2 aa essential region in the SSX tail (K43, R44) is an excellent target for developing therapeutics for this disease. A decoy molecule that causes SSX1 to resemble SSX3 would be expected to prevent eviction of BAF47 and thereby reverse the effects of the aberrant SS-BAF complex. This notion is consistent with the precision of the oncogenic translocation in that all translocations discovered to date add exactly 78 aa of SSX1,2 or 4 to the SS18 protein at position 379. In SS cells, the partially assembled complex gains the ability to bind the Sox2 gene, reversing H3K27Me3-mediated
repression. Forcing correct assembly by expressing the wildtype SS18 causes the reassembly of wild-type complexes without the fusion, thus re-establishing normal repression of Sox2 by polycomb. The y Brahma protein was discovered from its ability to oppose polycomb and hence is known as a trithorax gene; however, the underlying biochemical mechanisms have been controversial. In some studies, polycomb was found to prevent Brahma (BAP) complex binding, whereas in others it seemed that BAP or SWI/SNF directly recruited PolII, thereby opposing polycomb. Our studies suggest that somehow BAF complexes evict polycomb; however, our temporal resolution is limited to the infection times (2472 hr), and hence we are unable to determine if the mechanism is direct physical eviction or dilution of H3K27Me3 by nucleosome exchange with cell division because the measured rates of nucleosome turnover (Deal et al., 2010) are sufcient to remove most H3K27Me3 if methylation were prevented by the SS BAF complex. Evidence for BAF-polycomb opposition in malignancy has also been found with inactivation of BAF47 (hSNF5 or Ini1) in human malignant rhabdoid sarcoma (MRTs). In these tumors and in mouse models, polycomb was found to be removed from the INK4a locus upon introduction of BAF47 (hSNF5) (Kia et al., 2008). Understanding the underlying mechanism of polycomb opposition will require techniques that allow rapid recruitment of BAF complexes with a high degree of temporal and spatial control (Hathaway et al., 2012). Synovial sarcoma is largely resistant to conventional, chemotherapy-based forms of treatment, underlining the need for an understanding of its pathogenesis. Disease-specic biologic agents that target SS18-SSX or its interactions have not been developed to date. Here, we have shown that the SS18-SSX1 oncogenic fusion usurps SWI/SNF-like BAF complexes, resulting in activation of Sox2, which drives proliferation. Remarkably, the oncogenic fusion and wild-type SS18 bind to BAF complexes with comparable afnities, allowing directed assembly of oncogenic or wild-type complexes. Moreover, the composition of SS18-SSX-containing BAF complexes (lacking BAF47 and wild-type SS18) can be reversed by reducing the levels of SS18-SSX or by increasing levels of wild-type SS18. The observation that eviction of BAF47 from the complexes is dependent upon only 2 aa in SSX demonstrates an unusual mechanism of oncogenesis and opens a potential therapeutic avenue.
EXPERIMENTAL PROCEDURES Nuclear Extract Preparation and Proteomic Studies Nuclear extract (NE) preparation and immunoprecipitation (IP) studies were performed as described in Ho et al. (2009) and Extended Experimental Procedures. Antibody specications are presented in Table S1. Transfection Studies Briey, cells were plated in 6-well plates to 80% conuence prior to transfection using polyethylenimine (PEI) in a 3:1 PEI:DNA ratio and were harvested at the appropriate time points thereafter. Cell Proliferation Analyses Cells were assessed for >95% viability prior to being plated at 105 cells/well in triplicate/condition in 12-well plates. Cell counts were determined using trypan blue exclusion-based methods.
Urea Denaturation Studies NEs (150 mg) were subjected to partial urea denaturation, ranging from 0.25 to 8 M urea (in IP buffer), for 15 min at room temperature (RT) prior to anti-Brg IP. The coprecipitated proteins were analyzed by immunoblot. Quantitative densitometry analyses were performed with the Li-Cor Oddessy Imaging System (Li-COR Biosciences, Lincoln, NE, USA). Density Sedimentation Analyses NE (800 mg) was resuspended in 300 ml of 0% glycerol HEMG buffer and carefully overlaid onto a 10 ml 10%30% glycerol (in HEMG buffer) gradient prepared in a 14 3 89 mm polyallomer centrifuge tube (331327, Beckman Coulter, Brea, CA, USA). Tubes were centrifuged in an SW40 rotor at 4 C for 16 hr at 40 K rpm. Fractions (0.5 ml) were collected and used in analyses. See Extended Experimental Procedures. Cyclohexamide/MG-132 Studies MG-132 (474790, Calbiochem, San Diego) (10 mg/ml in DMSO) was used at 1:1,000, and cyclohexamide (C4859, Sigma-Aldrich, St. Louis) (100 mg/ml) was used at 1:100 in cell culture media. Briey, cells were plated in 6-well plates and treated with the above agents for 0 to 24 hr and harvested with RIPA lysis buffer. Gene Expression Proling and Analysis Total RNA was isolated using TRIzol reagent (Invitrogen, Carlsbad, CA, USA) and reverse transcribed into cDNA (SuperScript III RT kit, Invitrogen). Realtime PCR was performed using TaqMan Universal Master Mix with Taqman probes and/or SYBR green method with custom-designed primers, normalized to GAPDH and/or 18S rRNA expression. All primers are listed in Table S2. shRNA-Mediated Knockdown and Lentiviral Generation shRNAs specic for human Brg1, BAF47, SS18, and Sox2 were purchased from Open Biosystems (Thermo Scientic, Waltham, MA, USA) (Table S3). shRNA KD constructs for SS18-SSX and shScramble control were generated by annealed oligos (Table S3) and subsequent cloning into the pLK0.1 vector. Lentiviral (LV) was produced as described by Tiscornia et al., 2006. See Extended Experimental Procedures. ChIP Analyses Briey, cells were crosslinked in formaldehyde, washed, and sonicated as described in Extended Experimental Procedures. Antibodies used for ChIP studies include anti-BAF155 (in-house generated), anti-H3K27me3 (07-449, Millipore, Billerica, MA, USA), and V5 (46-0705, Invitrogen). Primers used for real-time PCR are listed in Table S2. SUPPLEMENTAL INFORMATION Supplemental Information includes Extended Experimental Procedures, six gures, and three tables and can be found with this article online at http:// dx.doi.org/10.1016/j.cell.2013.02.036. ACKNOWLEDGMENTS This work was supported in part by grants from the National Institutes of Health (NIH) (HD55391, RO1NS046789, and R01CA163195) (to G.R.C.). G.R.C. is an Investigator of the Howard Hughes Medical Institute. C.K. is supported by the National Science Foundation (Graduate Research Fellowship Program). The authors are grateful to Kazuyuki Itoh, Norifume Naka, and Satoshi Takenaka (Osaka University, Japan) for kindly providing the Aska-SS and Yamato-SS cell lines and cDNA clone of the SS18-SSX fusion. We are also grateful to Clara Lee for technical assistance and N. Hathaway, O. Bell, A. Shalizi, and J. Ronan for helpful discussions and their critical review of the manuscript. Received: September 18, 2012 Revised: December 18, 2012 Accepted: February 13, 2013 Published: March 28, 2013
REFERENCES Bass, A.J., Watanabe, H., Mermel, C.H., Yu, S., Perner, S., Verhaak, R.G., Kim, S.Y., Wardwell, L., Tamayo, P., Gat-Viks, I., et al. (2009). SOX2 is an amplied lineage-survival oncogene in lung and esophageal squamous cell carcinomas. Nat. Genet. 41, 12381242. Brodin, B., Haslam, K., Yang, K., Bartolazzi, A., Xie, Y., Starborg, M., Lundeberg, J., and Larsson, O. (2001). Cloning and characterization of spliced fusion transcript variants of synovial sarcoma: SYT/SSX4, SYT/SSX4v, and SYT/SSX2v. Possible regulatory role of the fusion gene product in wild type SYT expression. Gene 268, 173182. Cairns, B.R., Lorch, Y., Li, Y., Zhang, M., Lacomis, L., Erdjument-Bromage, H., Tempst, P., Du, J., Laurent, B., and Kornberg, R.D. (1996). RSC, an essential, abundant chromatin-remodeling complex. Cell 87, 12491260. Chamberlain, S.J., Yee, D., and Magnuson, T. (2008). Polycomb repressive complex 2 is dispensable for maintenance of embryonic stem cell pluripotency. Stem Cells 26, 14961505. Clark, J., Rocques, P.J., Crew, A.J., Gill, S., Shipley, J., Chan, A.M., Gusterson, B.A., and Cooper, C.S. (1994). Identication of novel genes, SYT and SSX, involved in the t(X;18)(p11.2;q11.2) translocation found in human synovial sarcoma. Nat. Genet. 7, 502508. Dawson, M.A., and Kouzarides, T. (2012). Cancer epigenetics: from mechanism to therapy. Cell 150, 1227. de la Serna, I.L., Carlson, K.A., and Imbalzano, A.N. (2001). Mammalian SWI/ SNF complexes promote MyoD-mediated muscle differentiation. Nat. Genet. 27, 187190. de Leeuw, B., Balemans, M., Olde Weghuis, D., and Geurts van Kessel, A. (1995). Identication of two alternative fusion genes, SYT-SSX1 and SYTSSX2, in t(X;18)(p11.2;q11.2)-positive synovial sarcomas. Hum. Mol. Genet. 4, 10971099. Deal, R.B., Henikoff, J.G., and Henikoff, S. (2010). Genome-wide kinetics of nucleosome turnover determined by metabolic labeling of histones. Science 328, 11611164. rtner, F., dos Santos, N.R., de Bruijn, D.R., Balemans, M., Janssen, B., Ga Lopes, J.M., de Leeuw, B., and Geurts van Kessel, A. (1997). Nuclear localization of SYT, SSX and the synovial sarcoma-associated SYT-SSX fusion proteins. Hum. Mol. Genet. 6, 15491558. Garcia, C.B., Shaffer, C.M., Alfaro, M.P., Smith, A.L., Sun, J., Zhao, Z., Young, P.P., VanSaun, M.N., and Eid, J.E. (2012). Reprogramming of mesenchymal stem cells by the synovial sarcoma-associated oncogene SYT-SSX2. Oncogene 31, 23232334. Hathaway, N.A., Bell, O., Hodges, C., Miller, E.L., Neel, D.S., and Crabtree, G.R. (2012). Dynamics and memory of heterochromatin in living cells. Cell 149, 14471460. Hiraga, H., Nojima, T., Abe, S., Sawa, H., Yamashiro, K., Yamawaki, S., Kaneda, K., and Nagashima, K. (1998). Diagnosis of synovial sarcoma with the reverse transcriptase-polymerase chain reaction: analyses of 84 soft tissue and bone tumors. Diagn. Mol. Pathol. 7, 102110. Ho, L., Ronan, J.L., Wu, J., Staahl, B.T., Chen, L., Kuo, A., Lessard, J., Nesvizhskii, A.I., Ranish, J., and Crabtree, G.R. (2009). An embryonic stem cell chromatin remodeling complex, esBAF, is essential for embryonic stem cell self-renewal and pluripotency. Proc. Natl. Acad. Sci. USA 106, 51815186. Ho, L., Miller, E.L., Ronan, J.L., Ho, W.Q., Jothi, R., and Crabtree, G.R. (2011). esBAF facilitates pluripotency by conditioning the genome for LIF/STAT3 signalling and by regulating polycomb function. Nat. Cell Biol. 13, 903913. ger, N., Kool, M., Zichner, T., Hutter, B., Sultan, M., Cho, Y.J., Jones, D.T., Ja tz, A.M., et al. (2012). Dissecting the genomic Pugh, T.J., Hovestadt, V., Stu complexity underlying medulloblastoma. Nature 488, 100105. Kia, S.K., Gorski, M.M., Giannakopoulos, S., and Verrijzer, C.P. (2008). SWI/ SNF mediates polycomb eviction and epigenetic reprogramming of the INK4b-ARF-INK4a locus. Mol. Cell. Biol. 28, 34573464. Krogan, N.J., Keogh, M.C., Datta, N., Sawa, C., Ryan, O.W., Ding, H., Haw, R.A., Pootoolal, J., Tong, A., Canadien, V., et al. (2003). A Snf2 family
ATPase complex required for recruitment of the histone H2A variant Htz1. Mol. Cell 12, 15651576. Lander, E.S. (2011). Initial impact of the sequencing of the human genome. Nature 470, 187197. Lessard, J., Wu, J.I., Ranish, J.A., Wan, M., Winslow, M.M., Staahl, B.T., Wu, H., Aebersold, R., Graef, I.A., and Crabtree, G.R. (2007). An essential switch in subunit composition of a chromatin remodeling complex during neural development. Neuron 55, 201215. Lickert, H., Takeuchi, J.K., Von Both, I., Walls, J.R., McAuliffe, F., Adamson, S.L., Henkelman, R.M., Wrana, J.L., Rossant, J., and Bruneau, B.G. (2004). Baf60c is essential for function of BAF chromatin remodelling complexes in heart development. Nature 432, 107112. Limon, J., Mrozek, K., Mandahl, N., Nedoszytko, B., Verhest, A., Rys, J., Niezabitowski, A., Babinska, M., Nosek, H., Ochalek, T., et al. (1991). Cytogenetics of synovial sarcoma: presentation of ten new cases and review of the literature. Genes Chromosomes Cancer 3, 338345. Mikkelsen, T.S., Ku, M., Jaffe, D.B., Issac, B., Lieberman, E., Giannoukos, G., Alvarez, P., Brockman, W., Kim, T.K., Koche, R.P., et al. (2007). Genome-wide maps of chromatin state in pluripotent and lineage-committed cells. Nature 448, 553560. Mizuguchi, G., Shen, X., Landry, J., Wu, W.H., Sen, S., and Wu, C. (2004). ATPdriven exchange of histone H2AZ variant catalyzed by SWR1 chromatin remodeling complex. Science 303, 343348. Nagai, M., Tanaka, S., Tsuda, M., Endo, S., Kato, H., Sonobe, H., Minami, A., Hiraga, H., Nishihara, H., Sawa, H., and Nagashima, K. (2001). Analysis of transforming activity of human synovial sarcoma-associated chimeric protein SYT-SSX1 bound to chromatin remodeling factor hBRM/hSNF2 alpha. Proc. Natl. Acad. Sci. USA 98, 38433848. Naka, N., Takenaka, S., Araki, N., Miwa, T., Hashimoto, N., Yoshioka, K., Joyama, S., Hamada, K., Tsukamoto, Y., Tomita, Y., et al. (2010). Synovial sarcoma is a stem cell malignancy. Stem Cells 28, 11191131. Peterson, C.L., and Herskowitz, I. (1992). Characterization of the yeast SWI1, SWI2, and SWI3 genes, which encode a global activator of transcription. Cell 68, 573583. Roberts, C.W., Leroux, M.M., Fleming, M.D., and Orkin, S.H. (2002). Highly penetrant, rapid tumorigenesis through conditional inversion of the tumor suppressor gene Snf5. Cancer Cell 2, 415425. Sandberg, A.A., and Bridge, J.A. (2002). Updates on the cytogenetics and molecular genetics of bone and soft tissue tumors. Synovial sarcoma. Cancer Genet. Cytogenet. 133, 123. zo-Bravo, M.J., Han, D.W., Greber, B., Singhal, N., Graumann, J., Wu, G., Arau ler, H.R. (2010). Chromatin-remodeling Gentile, L., Mann, M., and Scho components of the BAF complex facilitate reprogramming. Cell 141, 943955. n, M., and Skytting, B., Nilsson, G., Brodin, B., Xie, Y., Lundeberg, J., Uhle Larsson, O. (1999). A novel fusion gene, SYT-SSX4, in synovial sarcoma. J. Natl. Cancer Inst. 91, 974975. Smith, H.A., and McNeel, D.G. (2010). The SSX family of cancer-testis antigens as target proteins for tumor therapy. Clin. Dev. Immunol. 2010, 150591. Takenaka, S., Naka, N., Araki, N., Hashimoto, N., Ueda, T., Yoshioka, K., Yoshikawa, H., and Itoh, K. (2010). Downregulation of SS18-SSX1 expression in synovial sarcoma by small interfering RNA enhances the focal adhesion pathway and inhibits anchorage-independent growth in vitro and tumor growth in vivo. Int. J. Oncol. 36, 823831. Tamkun, J.W., Deuring, R., Scott, M.P., Kissinger, M., Pattatucci, A.M., Kaufman, T.C., and Kennison, J.A. (1992). brahma: a regulator of Drosophila homeotic genes structurally related to the yeast transcriptional activator SNF2/SWI2. Cell 68, 561572. Tea, J.S., and Luo, L. (2011). The chromatin remodeling factor Bap55 functions through the TIP60 complex to regulate olfactory projection neuron dendrite targeting. Neural Dev. 6, 5. Thaete, C., Brett, D., Monaghan, P., Whitehouse, S., Rennie, G., Rayner, E., Cooper, C.S., and Goodwin, G. (1999). Functional domains of the SYT and
SYT-SSX synovial sarcoma translocation proteins and co-localization with the SNF protein BRM in the nucleus. Hum. Mol. Genet. 8, 585591. Tiscornia, G., Singer, O., and Verma, I.M. (2006). Production and purication of lentiviral vectors. Nat. Protoc. 1, 241245. Varela, I., Tarpey, P., Raine, K., Huang, D., Ong, C.K., Stephens, P., Davies, H., Jones, D., Lin, M.L., Teague, J., et al. (2011). Exome sequencing identies frequent mutation of the SWI/SNF complex gene PBRM1 in renal carcinoma. Nature 469, 539542. venet, N., Lange, J., Rousseau-Merck, M.F., Ambros, P., Versteege, I., Se Handgretinger, R., Aurias, A., and Delattre, O. (1998). Truncating mutations of hSNF5/INI1 in aggressive paediatric cancer. Nature 394, 203206. Weiss, S.W., Goldblum, J.R., and Enzinger, F.M. (2001). Enzinger and Weisss Soft Tissue Tumors, Fourth Edition (St. Louis: Mosby). Wilson, B.G., Wang, X., Shen, X., McKenna, E.S., Lemieux, M.E., Cho, Y.J., Koellhoffer, E.C., Pomeroy, S.L., Orkin, S.H., and Roberts, C.W. (2010). Epigenetic antagonism between polycomb and SWI/SNF complexes during oncogenic transformation. Cancer Cell 18, 316328.
Wu, J.I., Lessard, J., Olave, I.A., Qiu, Z., Ghosh, A., Graef, I.A., and Crabtree, G.R. (2007). Regulation of dendritic development by neuron-specic chromatin remodeling complexes. Neuron 56, 94108. Wu, J.I., Lessard, J., and Crabtree, G.R. (2009). Understanding the words of chromatin regulation. Cell 136, 200206. Yoo, A.S., Staahl, B.T., Chen, L., and Crabtree, G.R. (2009). MicroRNA-mediated switching of chromatin-remodelling complexes in neural development. Nature 460, 642646. Yoo, A.S., Sun, A.X., Li, L., Shcheglovitov, A., Portmann, T., Li, Y., Lee-Messer, C., Dolmetsch, R.E., Tsien, R.W., and Crabtree, G.R. (2011). MicroRNA-mediated conversion of human broblasts to neurons. Nature 476, 228231. Zhao, K., Wang, W., Rando, O.J., Xue, Y., Swiderek, K., Kuo, A., and Crabtree, G.R. (1998). Rapid and phosphoinositol-dependent binding of the SWI/SNFlike BAF complex to chromatin after T lymphocyte receptor signaling. Cell 95, 625636.
Hijacking the Neuronal NMDAR Signaling Circuit to Promote Tumor Growth and Invasion
Leanne Li1 and Douglas Hanahan1,*
1Swiss Institute for Experimental Cancer Research, School of Life Science, Swiss Federal Institute of Technology Lausanne (EPFL), Lausanne 1015, Switzerland *Correspondence: douglas.hanahan@ep.ch http://dx.doi.org/10.1016/j.cell.2013.02.051
SUMMARY
Glutamate and its receptor N-methyl-D-aspartate receptor (NMDAR) have been associated with cancer, although their functions are not fully understood. Herein, we implicate glutamate-driven NMDAR signaling in a mouse model of pancreatic neuroendocrine tumorigenesis (PNET) and in selected human cancers. NMDAR was upregulated at the periphery of PNET tumors, particularly invasive fronts. Moreover, elevated coexpression of NMDAR and glutamate exporters correlated with poor prognosis in cancer patients. Treatment of a tumorderived cell line with NMDAR antagonists impaired cancer cell proliferation and invasion. Flow conditions mimicking interstitial uid pressure induced autologous glutamate secretion, activating NMDAR and its downstream MEK-MAPK and CaMK effectors, thereby promoting invasiveness. Congruently, pharmacological inhibition of NMDAR in mice with PNET reduced tumor growth and invasiveness. Therefore, beyond its traditional role in neurons, NMDAR may be activated in human tumors by uid ow consequent to higher interstitial pressure, inducing an autocrine glutamate signaling circuit with resultant stimulation of malignancy.
INTRODUCTION Invasion and metastasis is a dening hallmark in the pathogenesis of most forms of human cancer (Hanahan and Weinberg, 2011; Nguyen et al., 2009). Metastasis is a major cause of cancer morbidity, and invasiveness contributes both to metastatic dissemination as well as to locally invasive tumor growth with concomitant tissue damage. The limited efcacy of most conventional and targeted anticancer therapies may relate in part to their largely ineffectual inhibition of cancer progression via invasion and metastasis. Previously our laboratory identied an invasion modier locus on chromosome 17 in the RIP1-Tag2 transgenic mouse model of pancreatic neuroendocrine tumorigenesis (PNET) (Chun et al.,
2010). Within this locus, an N-Methyl-D-aspartate receptor (NMDAR)-associated gene, dlgap1, was one of several candidate proinvasive genes selectively upregulated in invasive carcinomas. We were led, therefore, to consider the potential involvement of the NMDAR in PNET tumorigenesis, especially in invasion. NMDAR is a receptor governing synaptic plasticity in the CNS, where it plays important roles in learning, memory, and neuron maturation. NMDARs have also been detected in various human tumor samples and cell lines, and patch-clamp experiments in several cancer cell lines have demonstrated receptor functionality (Stepulak et al., 2009). However, the functional importance of NMDAR signaling in cancers is unclear. Notably, the mechanistic contributions and pathologic signicance of NMDAR activation in elaborating cancer phenotypes are poorly understood (Prickett and Samuels, 2012). Glutamate, the major physiological agonist of the NMDAR, has long been implicated in cancer (Rzeski et al., 2001). A role in promoting tumor growth and invasion was rst established in glioma (Takano et al., 2001). Subsequently, an increasing number of cancer cells have been found to secrete glutamate (Seidlitz et al., 2009; Sharma et al., 2010) although the effector mechanisms and functional importance of secreted glutamate remain elusive. Glutamate is a ligand for two classes of receptors that are either G protein coupled or ion channels; NMDAR is a member of the ionotropic class. Protumoral effects of glutamate have been attributed to its signaling via G-protein-coupled glutamate receptors (Nicoletti et al., 2007) or via the AMPA receptor (Herner et al., 2011), another ionotropic glutamate receptor; in contrast, there is little evidence implicating glutamate signaling via NMDAR in cancer phenotypes. Motivated by these various considerations, we sought to determine whether the NMDAR and its ligand glutamate might be involved in invasive growth in the mouse model of PNET and, if so, to investigate the regulation and mechanistic effects of NMDAR signaling in such tumors and the possible translational relevance to human cancer. RESULTS NMDAR Is Upregulated in Genetically Engineered Mouse Models of Cancer The RIP1-Tag2 line of transgenic mice presents a genetically engineered mouse model (GEMM) of human PNET (Hanahan,
1985). We found that both subunits of the heterodimeric NMDAR, namely NR1 (Figure S1 available online) and NR2b (Figure 1A), were expressed in PNETs arising in this model. NR2b expression was elevated toward the tumor periphery (Figures 1A and 1B, i), particularly at invasion fronts (Figure 1C). The increased NR2b expression at the tumor periphery was evident in 96.6% of PNETs examined (Figure 1B, i), and the elevation of NR2b expression toward the tumor periphery was more evident as tumor size increased (Figure 1B, ii). Moreover, NR2b phosphorylation at Y1252, which enhances NMDAR activity (Takasu et al., 2002), was more pronounced at the tumor periphery than in the tumor center (Figure 1D). We also examined the expression of NMDAR in mouse models of pancreatic ductal adenocarcinoma (PDAC) (Grippo and Tuveson, 2010) and breast cancer (Fantozzi and Christofori, 2006). NR2b was also variably upregulated at the tumor periphery or in invasive cells in these GEMMs (Figures 1E1G and Table S1). Although the RIP1-Tag2 model is highly synchronized, presenting with discrete tumors as a function of age that renders them easily quantiable, both the breast and PDAC models are temporally and histologically heterogeneous, with merged lesions of varying tumor grades noted in late stage disease. Thus we were not able to quantify the patterns of NMDAR expression in these two GEMMs. NMDAR Pathway Is Evident in Multiple Human Cancers and Is Associated with Poor Cancer Patient Prognosis Having documented the elevated expression of NMDAR in several GEMMs of human cancer, we next audited NMDAR expression using human tissue microarrays (TMA). High NR2b expression was noted in some samples from pancreatic ductal carcinoma (Figure S2A), breast cancer (Figures 2A and 2B), ovarian cancer (Figure 2C), and glioma (Figure 2E), but not in others (Table S2). In the breast cancer TMA, we found that high NR2b expression was associated with the HER2 subtype, whereas negative NR2b expression was observed more in the luminal subtype (Figures 2A and S2B), suggesting that NR2b had different expression patterns among different subtypes. Interestingly, one patient sample in the breast cancer TMA expressed an intermediate to high level of NR2b and showed invasion into adjacent adipose tissues (Figure 2B), whereas a paired sample from the same patient in the same TMA, which was not immediately adjacent to the invasion fronts, expressed only low level of NR2b (Figure 2B). Glutamate is the major agonist for NMDAR. In neurons, vesicular glutamate transporters (vGlut1, -2, and -3) export glutamate to initiate signaling. Thus we assessed vGlut expression in conjunction with NMDAR for possible association with cancer patient survival in the TCGA database. In our survey of human cancer, TMAs we had found overexpression of NR2b in glioma (Figure 2E) and in ovarian cancer (Figure 2C). When we segregated glioblastoma (grade 4 glioma) patients by levels of both NR2b and vGlut2 into higher and lower expressing groups (see Extended Experimental Procedures), the difference was striking: the median survival was 4.4 months longer in the vGlut2/NR2blow group, reaching 15.2 months as compared to 10.8 months for the vGlut2/NR2b-high group (Figures 2F and S2D). Notably, vGlut1 and vGlut3 levels had a similar correlation (Figure S2D).
We also analyzed the ovarian cancer data set in TCGA and found a similar trend: the median survival was one year longer in the NR2b/vGlut2-low expression group than in the high expression group (Figures 2D and S2C). In contrast, in the TCGA lung squamous cell carcinoma data set, low versus high expression of NR2b/vGluts was not associated with differential patient prognosis (Figure S2E), suggesting, quite reasonably, that not all tumor types are affected by variable levels of NMDAR. Inhibiting NMDAR Has Antiproliferative and Anti-Invasive Effects In Vitro Motivated by our observations that the NMDAR was expressed at elevated levels in various human cancer types compared to cognate normal tissues and that patients whose tumors had comparatively higher levels had worse prognosis, we returned to the mouse PNET model to investigate possible roles of NMDAR signaling in this form of cancer. First, we employed the bTC-3 cancer cell line, derived from a PNET tumor in a RIP1-Tag2 mouse (Efrat et al., 1988), to investigate the potential involvement of NMDAR in cancer cell phenotypes in vitro. We used an NMDAR antagonist, MK801, to study the possible contributions of NMDARvia its inhibitionon proliferation and apoptosis. MK801 is a selective, noncompetitive NMDAR inhibitor. MK801 blocks the calcium channel of the NR1 subunit with high afnity; as such it is one of the most potent known NMDAR antagonists. We applied MK801 to cultures of bTC-3 cancer cells and observed a time-dependent decrease in proliferation and increase in apoptosis (Figure 3A). In our previous experience, cultured PNET cancer cells (exemplied by the bTC-3 cell line) are weakly invasive in the traditional transwell invasion assay (Du et al., 2007). Therefore, we employed a modied invasion assay developed by Swartz and colleagues (Shields et al., 2007) (Figures 3B and S3), which uses hydrostatic pressure to create a mimetic of interstitial uid pressure and consequent uid ow. As compared to the traditional static invasion assay, the ow-based invasion assay (Figure 3B) signicantly increased bTC-3 invasiveness, which could be blocked by MK801, indicating a substantive role for NMDAR in cancer cell invasion in this ex vivo assay (Figure 3C). Because AMPAR, another glutamate receptor, has previously been implicated in promoting cancer invasion (Herner et al., 2011), we also examined the effect of an AMPAR antagonist GYKI52466in the ow-based invasion assay. Modest antiinvasive activity was observed, albeit much weaker than that of MK801 (Figure 3C). To verify that these were not off-target effects of MK801, we knocked down the obligatory NMDAR subunit NR1 in bTC-3 cells and found that NR1 siRNA phenocopied the effects of the NMDAR inhibitor MK801 on cancer cell survival and invasiveness (Figure 3D). In our TMA survey, both pancreatic adenocarcinomas (Figure S2A) and breast cancers (Figures 2A and 2B) were found to express NR2b. Therefore, we performed some of the key in vitro experiments on a panel of human breast and pancreatic cancer cell lines that covered a variety of different subtypes (Table S3). Varying responses in terms of reduced cell survival (Figure 3E) and (ow-guided) invasion (Figure 3F) were noted in MK801-treated groups. Notably, the effect of MK801 on
invasiveness was in general more pronounced than on survival. When the tested cell lines were divided into three groups according to their invasiveness in the modied invasion assay, we found that within both highly and moderately invasive groups, the response to MK801 correlated with NR2b mRNA expression (Figure 3G). In contrast, NR2b was barely detectable in the weakly invasive group, and in this group the treatment response was instead associated with NR2a expression (Figure 3G). Interestingly, there was no association between the level of NR2b or other NMDAR subunits and reduced cell survival upon treatment with MK801 (data not shown). Overall, the results suggest that elevated levels of NR2b expression are particularly deterministic for ow-mediated invasion. NMDAR Signaling in PNET Is Autologous and Can Be Activated by Interstitial Flow In light of demonstrating that NMDAR signaling was enhancing the cancer cell phenotypes of proliferation and invasiveness in conditions mimicking interstitial pressure-driven ow, we sought to assess the possible involvement of its ligand glutamate in regulating such phenotypes. By analyzing mRNA isolated from different stages of PNET tumorigenesis using qRT-PCR, we found increased vGlut gene expression in the solid tumor stage, as compared to normal pancreatic islets and premalignant stages (Figures 4A and S4A). This result is consistent with previous studies showing that vGluts are not expressed in normal mouse islet b cells but are detectable and fully functional in bTC-6 (Bai et al., 2003), another PNET cell line derived from the RIP1-Tag2 mouse model (Poitout et al., 1995). Notably, the expression level varied considerably among tumors, as shown by the oating bars covering minimum to maximum values (Figure 4A). This observation is indicative of heterogeneous activation of NMDAR signaling among PNETs, consistent with the considerable range of NR2b staining in different tumors (Figure 1B). To pinpoint the cellular sources of glutamate, we used uorescence-activated cell sorting (FACS) to separate PNETs into their different constituent cell types. In neurons, classical paracrine NMDAR signaling involves presynaptic neurons expressing
vGluts that mediate glutamate secretion to stimulate postsynaptic neurons expressing the glutamate receptor NMDAR, as well as glial cells (and sometimes neurons) expressing EAATs (excitatory amino-acid transporters) that remove extracellular glutamate to modulate NMDAR signaling. By qRT-PCR, we identied cancer cells as the major expressers of all three components of this signaling loop (Figures 4B and S4B), indicative of autocrine NMDAR signaling. We also examined another glutamate transporter, xCT, and found it to be highly expressed by inltrating immune (inammatory) cells (Figure S4B). Therefore, paracrine glutamate signaling from immune inammatory cells to the cancer cells might be operative in PNETs as well. vGlut family proteins were also expressed in the bTC-3 cancer cell line (Figure 4C), and immunostaining showed a typical punctate localization of vGlut3 in the cytoplasm (Figure 4D). The nding that the glutamate transporters and NMDAR were coexpressed in bTC-3 cancer cells led us to investigate whether autocrine glutamate secretion was involved in their capability for invasion. Our initial experiments revealed that bTC-3 invasiveness was enhanced by hydrostatic ow (Figure 3C). Congruently, we found increased levels of glutamate in medium conditioned by bTC-3 cancer cells in ow conditions (Figure 4E), consistent with interstitial ow enhancing glutamate secretion and autocrine signaling via NMDAR in PNET cancer cells. Interstitial Flow Promotes NMDAR Surface Localization Having determined that autologous glutamate secretion was enhanced under the ow conditions that promoted bTC-3 invasion in the modied invasion assay, we added glutamate to the traditional (static) invasion assay. Interestingly, adding glutamate could not fully recapitulate the degree of invasiveness seen in the ow-based invasion assay (data not shown). Therefore, the following question emerged: in addition to increasing the levels of secreted ligand, might interstitial ow be directly affecting the glutamate receptor NMDAR? We used ow cytometry to analyze NMDAR surface expression in regular 2D-cultured (static) bTC-3 cancer cells and found heterogeneous surface expression of both NR1 and NR2b (Figure S5A), which was also observed by immunostaining
Figure 1. Involvement of NMDAR Signaling in Genetically Engineered Mouse Models of CancerDescriptive Evidence
(A) In the RIP1-Tag2 mouse model of pancreatic neuroendocrine tumors (PNET), expression of the NMDAR subunit 2b (NR2b) is selectively elevated at tumor periphery as compared to tumor center (images were taken from the same tumor). (B) Semiquantication of NR2b expression in mouse PNETs. (i) A pair of images was taken from each tumor analyzed, one from the center and the other from the periphery. Image pairs from 59 PNETs excised from nine mice were digitally quantied for NR2b staining intensity (as described in the Extended Experimental Procedures), which revealed that NR2b was in most tumors signicantly overexpressed at the periphery. Only three out of 59 tumors (spots marked in red) fell under the red dashed-line (slope = 1) representing equal level of staining at the periphery and center. (ii) Using the data from (i), a ratio for each tumor was generated by dividing the average staining intensity for NR2B at the tumor periphery with that of the center. This ratio of overexpression was positively associated with tumor diameter measured on semithin tissue sections (nonparametric correlation, two-tailed Spearmans test, with rs = 0.6621). (C) NR2b overexpression was particularly evident at invasion fronts; immunostaining for the oncoprotein expressed by the transgenic RIP1-Tag oncogene reveals the cancer cells. (D) Phsopho-NR2b, indicative of signaling activity, was also preferentially detected at the tumor periphery; a transition from positive to equivocal/negative staining was observed in elds near the tumor center. (E) In the MMTV-PyMT mouse model of breast cancer, increased NR2b expression was observed in cancer cells invading into adipose tissues and muscle layers compared to those at the tumor core. (Blue asterisks, adipocytes; green asterisks, muscle layer; red arrow heads, invading cancer cells). (F) High magnication of breast cancer cells invading into the muscle layer showed elevated NR2b staining compared to cells in the tumor core. (Green asterisks, muscle layer; red arrow heads, invading cancer cells). (G) In a mouse model of pancreatic ductal adenocarcinoma (PDAC), a similar trend of NR2b overexpression at the tumor periphery (i) compared to the tumor center (ii) was also noted. See also Figure S1 and Table S1.
Figure 2. Involvement of NMDAR Signaling in Human CancersDescriptive Evidence

(A) A nonexhaustive survey involving human cancer tissue microarrays (TMA) revealed varying intensity of NR2b staining in the tumor samples. The staining intensity was categorized into four levels: negative, low, intermediate, and high. The percentage of each category from the TMA was documented. Interestingly, NR2b staining was associated with different breast cancer subtypes. p < 0.05, Chi-square test. (B) In one patient sample in the TMA, breast cancer cells with intermediate to high level of NR2b expression were invading into the adjacent adipose tissue, whereas a paired sample that was not immediately adjacent to an invasive front showed negative to low levels of NR2b expression. (Green asterisks, adipocytes; blue arrows, cancer cells). (C) Example from a similar analysis of an ovarian cancer TMA, which also revealed elevated NR2b expression in a subset of cancer cells. (Blue arrow: ovarian cancer cells). (legend continued on next page)
(Figure S5B). However, when we stained for both surface and intracellular NMDARs, uniform levels of NR1 and NR2b were detected (Figure S5A). This result suggested that there was an intracellular pool of NMDARs primed for surface recruitment in response to appropriate signals, similar to the situation in neurons (Lau and Zukin, 2007). This observation suggested that interstitial ow might be involved in regulating the surface expression of NMDAR. Indeed, ow cytometry analysis revealed that the levels of the two NMDAR subunits on the cell surface were increased in ow conditions as compared to the static condition (Figure 5A). Thus interstitial ow modulates surface localization of NMDAR as well as glutamate secretion. Interstitial Flow Activates the CaMK and MEK-MAPK Pathways Downstream of NMDAR Ligand-stimulated NMDAR induces calcium inux in neurons, which activates two major downstream signaling circuits: the Ca2+/calmodulin kinase (CaMK) pathway and the MEK-MAPK pathway (Hardingham and Bading, 2010); both lead to phosphorylation of the transcription factor CREB (cAMP response element-binding protein) at Ser133 (Figure 5B), a prerequisite for recruiting the transcriptional coactivator CREB-bindingprotein (CBP) to the promoter regions of effector genes. To clarify which pathway(s) downstream of NMDAR is involved in cancer cell invasion, we analyzed phosphoprotein expression in bTC-3 cancer cells in the invasion assay. We found that interstitial ow increased NR2b phosphorylation (Figure 5C), which is known to potentiate NMDAR activity (Takasu et al., 2002) and to promote NR2b surface localization (Braithwaite et al., 2006) in neurons. We also observed increased phosphorylation of calmodulin kinase type II (CaMK-II), calmodulin kinase type IV (CaMK-IV), MAPK/ERK kinase (MEK), and p44/p42 mitogen-activated protein kinase (MAPK) in ow conditions as compared to static conditions, leading to a modest increase in CREB phosphorylation at Ser133 (Figure 5C). Consistent with our hypothesis that NMDAR mediates ow-activated signaling, the addition of the NMDAR antagonist MK801 to the ow-stimulated invasion assay markedly decreased the phosphorylation of these effector proteins, with the exception of CAMK-IV (Figure 5C). Strikingly, pretreatment with BAPTAAM, a potent intracellular calcium chelator commonly used to block intracellular calcium signaling, in particular calciumdependent NMDAR signaling (Marsden et al., 2007), was able to abolish ow-induced protein phosphorylation of all these effectors in both MEK-MAPK and CAMK pathways, including CAMK-IV (Figures 5D and S5D). The results establish that the ow-mediated activation of the CaMK and MEK-MAPK pathways is calcium-dependent, principally involving the calciumdependent NMDAR.
In Vivo the NMDAR Antagonists Have Therapeutic Efcacy Having characterized NMDAR signaling in assays involving cultured cancer cells, we proceeded to perform experimental therapeutic trials in the RIP1-Tag2 mouse model in order to assess the importance of NMDAR signaling for tumors in vivo. The synchronized, multistage tumorigenesis pathway to PNET in RIP1-Tag2 mice renders this GEMM a powerful tool for experimental trials of mechanism-targeted drugs. By 1214 weeks of age, 2%4% of the approximately 400 pancreatic islets have progressed through premalignant stages to become solid tumors with varying degrees of invasiveness (Chun et al., 2010), and the mice reach end stage at around 1416 weeks. Several distinctive experimental trial regimens have proved informative about the molecular, histological, and pathologic effects of anticancer drugs: intervention trials start at 1011 weeks of age and last 34 weeks, aiming to determine if a drug can intervene in the expansive growth of nascent solid tumors; regression trials start at 1213 weeks, to assess a drugs effect when substantial solid tumors have developed, and the mice are at a late stage of disease progression (Bergers et al., 1999). Regression trials in this mouse model therefore mimic a common situation in the clinic, when treatment commences in patients with advanced solid tumors. In an intervention trial with MK801, tumor burden (cumulative volume of multiple tumors in the pancreas) and tumor number were both decreased by the treatment (Figure 6A), concomitant with reduced proliferation (Figure 6B). Tumor invasiveness was also attenuated (Figure 6A). In addition, NR2b expression and phosphorylation at the tumor periphery were decreased in MK801-treated tumors (Figure 6B). We then performed a regression trial, and found that MK801 was even more effective in decreasing tumor burden at this late stage of progression (Figure 6C). We also tested a much weaker, clinically approved NMDAR antagonist, memantine, and observed an antitumoral effect in the late stage regression trial, but not in the early stage intervention trial (Figure S6A), perhaps reecting its weaker activity. We then performed a preclinical trial with MK801 in a second mouse model, involving orthotopic transplantation of primary breast cancer cells from MMTV-PyMT transgenic mice. After 3 weeks of treatment, we observed a trend toward decreased tumor burden (Figure S6B); we infer that the lack of statistical signicance reects similar heterogeneity in NMDAR expression to what we described above both in breast tumors in the GEMM and in patient samples (by immunohistochemistry and mRNA proling). Collectively, the preclinical trials support the implications from the analysis of clinical data that NMDAR signaling is functionally important in some tumor types (Figure S2 and Tables S1 and S2) and that individual tumors of particular
(D) Querying the TCGA database of ovarian cancer patients for the combination of NR2b and the glutamate transporter vGlut2 revealed that a group with low expression levels of both NR2b and vGlut2 (below mean) had more than one year of survival advantage compared to a distinctive NR2b/vGlut2 high group (above mean). Median survival, 49.4 months in NR2b/vGlut2 low group; 36.9 months in NR2b/vGlut2 high group. (E) NR2b was also expressed in samples from a glioma TMA, as exemplied by one tumor. (F) Survival analysis using glioblastoma patient data in the TCGA data set revealed that NR2b expression levels alone were signicantly associated with prognosis (median survival, 14.9 months in NR2b low group versus 11.5 months in NR2b high group). Moreover, the combination of low versus high levels of NR2b plus vGlut2 further separated the curves (median survival, 15.2 months in NR2b/vGlut2 low group versus 10.8 months in NR2b/vGlut2 high group). See also Figure S2 and Table S2.
cancer types and subtypes could have heterogeneous expression and consequently variable responses to MK801 treatment (Figure 3E). DISCUSSION The results presented herein provide provocative evidence for a functional role of the NMDAR signaling pathway in tumor progression. In addition, they unveil a mechanism of NMDAR pathway activation in cancer: physical cues in cancer microenvironment, namely interstitial pressure differences and uid ow, serve to activate an autocrine pathway that stimulates proliferation and invasiveness of cancer cells (Figure 7). Glutamate Is Subject to Flow-Regulated Secretion for Autocrine Stimulation of NMDAR We have revealed a mechanism that regulates glutamate bioavailability in the tumor microenvironment, demonstrating that during tumorigenesis, glutamate bioavailability and signaling are evidently elevated by three mechanisms. First, the vGlut genes are transcriptionally upregulated in the cancer cells. Second, increased interstitial pressure and ow in the microenvironment elicit increased glutamate secretion by those cancer cells. Third, ow conditions increase both expression and cellsurface localization of the glutamate receptor NMDAR so as to enable stronger signaling in response to the glutamate ligand. Concordantly, the combination of higher expression of NR2b and vGluts was predictive of poor prognosis for cancer patients. Interstitial Flow as a Microenvironmental Signal Deciphering the complex crosstalk between cancer cells and their microenvironment stands as an important challenge to the cancer eld (Hanahan and Weinberg, 2011). Among the various environmental parameters, cancer-associated stromal cells are prominent, and their importance is increasingly well established (Hanahan and Coussens, 2012). In contrast, invisible cues, such as the physical forces within the tumor microenvironment, have been less well studied and appreciated. Recently, mechanotransduction has begun to emerge as an
instructor for cancer progression (DuFort et al., 2011). Elevated tumor interstitial uid pressure (IFP) and consequently increased interstitial ow have been associated with tumor invasion and lymph node metastasis (Shieh and Swartz, 2011) and with poor patient prognosis (Heldin et al., 2004). Interstitial ow is typically highest at tumor margins due to the differential interstitial uid pressure between tumor and adjacent normal tissue (Dafni et al., 2002; Harrell et al., 2007). Congruently, as demonstrated in Figure 5, ow conditions increased both surface expression of NMDAR and NR2b phosphorylation and enhanced invasion by PNET cancer cells, consistent with the observation that NR2b expression and NR2b phosphorylation preferentially occurred at the periphery of PNET tumors (Figures 1A1D). Moreover, IFP and interstitial ow are known to increase signicantly as tumor size increases (Gutmann et al., 1992), consistent with our observation that peripherally elevated expression of NR2b was positively associated with tumor size (Figure 1B, ii). Notably, the therapeutic benet of the NMDAR inhibitors MK801 and memantine are much more signicant in the late stage regression trial than in the early stage intervention trial (Figures 6A, 6C, and S6A), again implicating the preferential activation of NMDAR in late stage tumors via increased interstitial pressure and consequent ow, leading to heightened malignancy. Proliferation and Invasion May Be Governed by Distinct Branched Pathways Downstream of NMDAR In the in vitro assay, the anti-invasive effect of MK801 (evident within overnight culture) occurred much faster than its antiproliferative and proapoptotic effects (evident after 3 days) on cancer cells in vitro (Figures 3A and 3C). This result suggests that NMDAR-mediated invasion might involve different downstream effectors than those that modulate survival. The analysis of protein phosphorylation in the invasion assay (Figure 5C) provides some insights into the downstream pathways of NMDAR. By comparing cells in ow conditions with those in ow plus MK801 conditions, we could attribute the observed reductions in invasion by MK801 in ow conditions (Figure 5C) to the
Figure 3. Functional Importance of NMDAR Signaling in CancerIn Vitro Evidence

(A) Treating cultured PNET cancer cells with the NMDAR antagonist MK801 decreased proliferation and increased apoptosis. (B) Schematic of the modied invasion assay that mimics interstitial uid pressure-mediated pressure gradients and uid ow. (C) In the modied invasion assay shown in (B), ow signicantly increased cancer cell invasiveness, which could be inhibited both by the NMDAR antagonist MK801 and (to a lesser extent) by the AMPAR antagonist GYKI52466. (D) The invasion- and growth-inhibiting effects of MK801 could be recapitulated with siRNA-mediated knockdown of the obligatory NMDAR subunit 1 (NR1). (E and F) Treatment of a panel of human breast and PDAC cancer cell-lines with MK801 variably reduced cell survival (E), and invasiveness using the modied invasion assay (F). (G) Out of the 14 cell lines shown in (E), 9 cell lines were selected for evaluation of NR2a/b mRNA expression. The 9 cell lines could be divided into three groups according to their invasiveness: high (DanG, BT549, HCC1806), intermediate (SKBR3, SUIT2, 3.27), and weak (MCF7, MDAMB157, BxPC3). Interestingly, NR2b expression was associated with the response to MK801 within high and intermediate invasiveness groups. NR2b was barely detectable in the weakly invasive group; instead, MK801s effects on invasion were associated with differing levels of NR2a expression in that group. Cell lines expressing neither NR2 subunit were not responsive to MK801, indicative of its specic targeting of NMDAR. (H) Among the cell lines, DanG expressed the highest level of NR2b and was the most responsive to MK801 treatment in the modied invasion assay. A representative picture from the MK801-treated DanG cells showed markedly decreased numbers of invading cells compared to the control group. Pictures were originally taken in DAPI channel with monochromic camera, shown with inverted black and white. Data are represented as mean with SEM; a two-tailed Students t test was performed to compare control and treatment groups from (A)(E); in (F) and (G); data were normalized to the control group each time, thus a two-tailed one-sample t test was performed to determine if the ratio was signicantly different from hypothetical value 1, representing the control. *p < 0.05, **p < 0.01, ***p < 0.001. See also Figure S3 and Table S3.
Figure 4. The NMDAR Circuit in PNET: Upstream Activators

(A) Comparative analysis of mRNA levels of the glutamate transporters in the different stages in PNET tumorigenesis revealed that expression of both vGlut1 and -2 was increased in PNETs as compared to normal pancreatic islets and premaligant stages. Normal islet: three independent islet pools; hyperplastic and angiogenic islet: one islet pool each; PNETs: 14 tumors. Floating bars showing minimum to maximum with line showing the mean. (B) Ex vivo, qRT-PCR with cDNA generated from FACS-sorted constituent cell types from PNETs revealed that cancer cells were the major expressers of NR1, vGlut1, and vGlut2, consistent with possible autocrine glutamate to NMDAR signaling. Data shown was from one cell sorting. Sortings were repeated three times, with similar results. Each cell sorting was performed by pooling multiple PNETs from one to two mice. FACS, uorescence-activated cell sorting. (C) In vitro, all three vGlut family proteins were expressed in bTC-3 cancer cells generated from a mouse PNET, as shown by ow cytometric analysis with specic antibodies. (D) Immunocytochemistry conrmed typical punctate cytoplasmic staining of vGlut3 in bTC-3 cells. (Red arrow heads, vGlut3.) (E) Interstitial ow (Figure 3B) increased glutamate concentration in the medium of the transwell invasion assay as compared to the static condition. Unpaired t test, one-tailed. The data are represented as mean with SEM; the data shown was from one of >5 replicate experiments, each with similar trends. See also Figure S4.
decreased phosphorylation of NR2b and its downstream effectors. Notably, however, MK801-treated cancer cells in the modied invasion assay were still more invasive than cancer cells in the static assay (Figure 3C). It is possible that the CaMK-IV phosphorylation was sufcient to confer the remaining 1.5-fold increase in invasiveness when compared to the static group, whereas the other phosphorylated proteins contributed to the
near-6-fold higher levels of the ow group with no inhibition of NMDAR. Concordantly, the knockdown of CaMK-IV expression with siRNA, which largely abrogated invasiveness with minimal effect on proliferation, supports the selective involvement of this kinase in orchestrating an invasive program (Figure S7). Taken together, we infer that the MEK-MAPK pathway is preferentially governing proliferation and survival
Figure 5. The NMDAR Circuit in Cancer: Downstream Effectors

(A) Hydrodynamic pressure and ow through transwells enhanced cell-surface expression of the NMDAR on bTC-3 cells, as revealed by ow cytometry analysis with live cancer cells. Data are represented as mean with SEM. (B) Schematic of NMDAR signaling: receptor activation leads to calcium-dependent stimulation of two major downstream signaling pathways: the CaMK-II/IV pathway and the MEK-MAPK pathway; NR2b phosphorylation at Y1252 and Y1336 are known to potentiate NMDAR activity. CaMKII, calmodulin kinase type II; CaMKIV, calmodulin kinase type IV; MEK, MAPK/ERK kinase; MAPK, p44/p42 mitogen-activated protein kinase; CREB, cAMP response element-binding protein. (C) Flow in the modied transwell invasion assay promoted NR2b phosphorylation in bTC-3 cells, and activated both CaMK and MEK-MAPK pathways downstream of NMDAR, which led to CREB phosphorylation. (*p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001) MFI, mean uorescence intensity; DMFI, difference of MFI between staining and negative control. One-sample t test was performed to determine if the ratio generated from each experiments was signicantly different from 1, representing DMFI of the static group. Data are represented as mean with SEM. (D) These ow-mediated effects could be abolished by an intracellular calcium chelator BAPTA-AM; the red shadow (baseline control, static group) and blue line (ow group) almost totally overlapped in the BAPTA-AM-treated group (right), as compared to the separate lines in the control group (left). (E) A schematic, based on the MFI analysis, suggesting that the MEK-MAPK and CaMK pathways are differentially activated. In static conditions, the MEK-MAPK pathway was already highly activated. In ow conditions, including increased secretion of glutamate, the increased NMDAR phosphorylation only modestly increased MEK-MAPK pathway activity. In contrast, the CaMK pathway activity, which was comparatively low in static condition, is appreciably upregulated by ow conditions. See also Figure S5.
Figure 6. Functional Importance of NMDAR Signaling in CancerIn Vivo Evidence

(A) A preclinical intervention trial targeting early-staged tumors in the RIP1-Tag2 model of PNET using the NMDAR antagonist MK801 decreased tumor burden, tumor number, and the incidence of highly invasive carcinomas in RIP1-Tag2 mice (n = 912 mice per group). Denition: IT, tumor margin < 10% invasive; IC1, margin 10%50% invasive; IC2, margin > 50% invasive. At least 39 tumors per group were graded. (legend continued on next page)
Therapeutic Benets of Inhibiting NMDAR Signaling in Tumors The NMDAR inhibitor MK801 has previously shown antitumoral effects when used to treat various xenograft tumors (North et al., 2010a; North et al., 2010b). Our study broadens the scope, demonstrating therapeutic efcacy of MK801 in an immunocompetent mouse model of endogenous tumor progression. In addition, whereas tissue invasion cannot be thoroughly assessed in traditional subcutaneous xenografts, this GEMM revealed a role for NMDAR in cancer invasion. Notably, the effect of MK801 in vivo was not as striking as that in vitro, and we suggest an explanation: MK801 has a very short half-life, of about an hour (Wegener et al., 2011). Therefore, in tumors in vivo the exposure was only for 1 hr a day, whereas in the ow-modied invasion assay, the drug was continuously present. As mentioned previously, we infer that PNET survival and invasion are governed by different downstream pathways. Thus, the 1 hr daily exposure experienced by tumors in vivo might be sufcient to markedly impair proliferation but insufcient to fully inhibit invasion. Although demonstrably important, NMDAR is not the sole driver of invasion in the RIP1-Tag2 model: various regulatory pathways have been proved instrumental for PNET invasion, including IGF-2/IGF-1R signaling (Christofori et al., 1994; Lopez and Hanahan, 2002), loss of E-cadherin and altered NCAM function (Perl et al., 1999; Perl et al., 1998), loss of desmosomal adhesions (Chun and Hanahan, 2010), and extracellular matrix degrading enzymes supplied by immune cells (Joyce et al., 2004). Therefore, the incomplete inhibition of PNET tumor invasiveness by NMDAR antagonists may also reect parallel, independent capabilities manifested by these other signaling events; the possibility that ow-activated glutamate-to-NMDAR signaling regulates one or another of these various proinvasive signaling pathways deserves future investigation.
Figure 7. A Model: Interstitial Flow Activates Glutamate-to-NMDAR Signaling in the Tumor Microenvironment
At the margins of solid tumors, an interstitial uid pressure (IFP) drop and consequent uid ow into adjacent normal tissue induces membrane localization and phosphorylation of the mechanosensitive NMDAR, and elevates expression of glutamate transporters and consequent secretion of glutamate, constituting an autocrine signaling circuit that stimulates cancer cell proliferation (data not shown) and invasiveness. Downstream of glutamate-activated NMDAR, the CaMK-II/IV and MEK-MAPK signal transducers in turn phosphorylate and activate the transcription factor CREB, which associates with the CREB-binding protein (CBP) to regulate a presumptive transcriptional program that mediates tumor growth and invasion. See also Figure S7.
downstream of NMDAR activation, whereas the CaMK pathway, in particular CaMKIV, is playing the major role in invasion. Still to be claried in future studies is the role of glutamate-stimulated AMPAR signaling, which is implicated in invasion (Figure 3C); AMPAR signaling would also be abrogated by BAPTA but not MK801, thus potentially explaining the differences in degrees of inhibition of downstream effector phosphorylation and of invasiveness.
Implications of NMDAR Pathway Activation in Human Cancer Our experimental design, which focused on the mechanism and effects of NMDAR activation in a GEMM of human cancer, involved only a limited (albeit provocative) survey for evidence of NMDAR signaling in different forms of human cancer. There are, therefore, several considerations. First, because NMDAR is in some cases evidently upregulated at the tumor periphery or in tissue-invading cancer cells, TMAs composed of core needle biopsies may miss the informative margins. Moreover, genome-wide expression-proling data (e.g., from TCGA) may in some cases fail to identify such focal upregulation of mRNA at the margins, obscured by the predominant core of large solid tumors. Second, although we clearly implicated the NR2b subunit in the stimulation of tumor invasion and aggressiveness in the PNET model, we cannot exclude the involvement of alternative NMDAR subunits in other tumor types. In our in vitro survey of human cancer cell lines, some did not express high NR2b, but rather NR2a (such as MDAMB157, SUIT2), and these
(B) MK801 treatment decreased tumor proliferation (BrdU staining), NR2b expression at tumor periphery, and NR2b phosphorylation. BrdU, n = 1425 tumors/ three mice per group; NR2b, n = 1116 tumors/four to ve mice per group; p-NR2b, n = minimum of six mice per group. (C) Similarly, a regression trial targeting late-stage tumors with MK801 also had antitumoral effects. n = 912 mice per group. Data are represented as mean with SEM. See also Figure S6.
were also responsive to MK801; in contrast, cell lines expressing neither subunit were poorly responsive to MK801. Finally, the three vGlut transporters may be variably expressed; whereas all were expressed in the PNET model, this might not be the case in other tumor types. Thus the three vGluts as well as the various NMDAR subunits should be audited in the context of surveying human tumor types, ideally incorporating in situ histological methods that can detect localized upregulation at tumor margins and invasive fronts. Translational Implications of NMDAR Antagonists for Cancer Therapy The results collectively suggest that inhibition of NMDAR signaling could have therapeutic benet in some forms of human cancer. There are, nevertheless, additional considerations. First, patient selection will likely be important. We observed heterogeneous expression of NMDAR in the human cancer TMA survey, as well as varying responses to MK801 treatment in a panel of human cancer cell lines and in a mouse model of breast cancer. Therefore, NMDAR antagonists may be more effective in patients with broader (less focal/peripheral) and higher expression/activation of the NMDAR signaling axis. Additionally, combination therapies with conventional drugs targeting the tumor core along with NMDAR inhibitors targeting the invasive periphery might prove benecial; moreover, because AMPAR is implicated as a second proinvasive glutamate receptor (Figure 3C), it will be of interest to explore combinatorial targeting of NMDAR and AMPAR. A second and important consideration is that rened NMDAR inhibitors are needed. As mentioned previously, the short half-life of MK801 may partially account for its incomplete inhibition of tumor invasiveness; as such, a second-generation drug with a longer half-life and better exposure would likely improve efcacy. Additionally, in light of the well-known (side) effects of NMDAR antagonists on learning, memory, and behavior (Wu et al., 2005), it would be highly desirable to develop new NMDAR antagonists that dont cross the blood-brain-barrier. Although such drugs might not have optimal efcacy in glioma, they would likely prove more tolerable clinically for patients with NMDARexpressing tumors outside of the CNS. Finally, inhibitors of glutamate biosynthesis and secretion are also worth investigating as agents for targeting this proinvasive signaling axis. Perspective In conclusion, this study reveals how cancer cells hijack the glutamate-to-NMDAR signaling pathway operative in neurons to instead promote invasive tumor growth. These ndings potentially link the long-recognized existence of high interstitial uid pressure (IFP) in tumors with the hallmark capability for tumor invasion, whereby comparatively higher IFP in solid tumors creates a pressure drop at the tumor margin, with consequent uid ow into adjacent normal tissue. Via mechanosensory transduction, this pressure drop and uid ow evidently activate autocrine glutamate secretion, concomitant with NMDAR phosphorylation and transport to the cell surface to engage the glutamate ligand, with consequent activation of downstream signaling (Figure 7). The results establish a potentially widespread mechanism for inducing tumor invasiveness. This mech98 Cell 153, 86100, March 28, 2013 2013 Elsevier Inc.
anism offers a potentially important target for future cancer therapeutics, whereby long-lasting, periphery-acting NMDAR antagonists may have promise for treating certain human cancers.
EXPERIMENTAL PROCEDURES Genetically Engineered Mouse Model of Cancer The RIP1-Tag2 mice (Hanahan, 1985) were bred and genotyped as previously described (Herzig et al., 2007). In brief, RIP1-Tag2 mice are transgenic mice, inbred into C57Bl/6, carrying a hybrid oncogene composed of the SV40 early region, encoding the large and small T antigen oncoproteins (Tag), fused to rat insulin gene promoter (RIP). Male mice heterozygous for the RIP1-Tag2 transgene are used for breeding with wild-type C57Bl/6J females. Pups were genotyped for SV40 oncogene using qRT-PCR by Transnetyx (Cordova, TN, http://www.transnetyx.com/). The qRT-PCR probes are designed by the company and procedures are detailed on the companys website. The probes for conventional genotyping using PCR detection are listed as following: forward primer 50 -GCTCTGCTGACATAGAAGAATGG-30 ; reverse primer 50 -GTACTCATTCATGGTGACTATTCCAG-30 (amplicon 454 bp). Male RIP1Tag2 mice are typically larger in size than females, therefore, only male mice with body weight over 23 g at starting point of an experiment were included for trials and analysis (See e.g., Chun et al., 2010). The MMTV-PyMT mice were bred and genotyped as previously described (Malanchi et al., 2012). The genetic modications of the PDAC GEMMs used were listed in the Table S1. The PDAC mice were bred as described previously (Olson et al., 2011). Modied Invasion Assay The hydrostatic-pressure-based modication of the classic Boyden chamber assay was performed as previously described (Shields et al., 2007). In brief, cells were seeded into a mixture of 1.2 mg/ml rat tail collagen type I (BD Biosciences, NJ) and 10%20% of growth factor reduced matrigel (356231, BD biosciences, NJ), then placed onto transwell inserts (Millipore http://www. millipore.com/catalogue/module/C10504, 12 mm PCF, 8.0 mm pore size), and then incubated for one hour at 37 C to solidify the matrix. Additional basal medium was added to the top of the wells to generate 1cm water head (650 ml on the top, 150 ml in the bottom for ow condition; 150 ml on the top, 650 ml on the bottom for the static condition). After overnight incubation, the gels were discarded and the upper side of the membrane was cleaned carefully with cotton tips to ensure that all the cells that didnt cross the membrane were removed. Then, cells on the bottom side of the membrane were xed with ice-cold methanol and stained with DAPI. Five images/well were taken at constant positions (as illustrated in Figure S3) with a 103 objective, which in sum covered most of the area of the membrane. The results were quantied using Fiji Image Analysis software, as described (Schindelin et al., 2012). Each experiment was performed in triplicate and repeated at least three times. Analysis of Protein Phosphorylation in the Invasion Assay The cancer cells were incubated on 6-well hanging inserts (FA-353493, BD Biosciences, NJ) in three different conditions as described in Figure 3B: static, ow and ow plus MK801, following the same gel casting protocol described in Modied Invasion Assay. The gels from the invasion assays mimicking 3D culture were collected from the transwells and digested with collagenase D (11088866001, Roche) according to manufacturers protocol to release the cells. The cells were counted using a cell counter (Countess Cell Counter, Life Technologies), and equal numbers of cells were distributed into individual wells or tubes for antibody staining. For the intracellular staining, the cells were xed with Cytox/Cytoperm buffer (BD Biosciences, NJ), whereas for analysis of NMDAR surface expression, the cells were not xed. Then the cells were blocked with an anti-mouse CD16/32 Fc blocking agent (BioLegend http:// www.biolegend.com/). Primary antibodies were diluted 1:100 in the blocking solution, and secondary antibodies (donkey-anti-rabbit IgG Alexa488 A21206, donkey-anti-rabbit IgG Alexa568 A10042, goat-anti-mouse IgG Alexa488 A11029, donkey-anti-mouse Alexa647 A31571, all from Life Technologies) were diluted at 1:1000 in the staining solution (BD Biosciences, NJ). Staining was performed on ice for 30 min. Samples were then subjected
to ow cytometry in a CyAn ADPS ow cytometry analyzer (Beckman Coulter) in the EPFL ow cytometry core facility. Flow cytometry data were analyzed with FlowJo software, and mean uorescence intensity (MFI) was calculated using geometric mean. SUPPLEMENTAL INFORMATION Supplemental Information includes Extended Experimental Procedures, seven gures, and four tables and can be found with this article online at http://dx.doi. org/10.1016/j.cell.2013.02.051. ACKNOWLEDGMENTS We wish to thank M. Luisa Iruela-Arispe, Igor Allaman, and Anita Wolfer for advice, discussions, and insightful comments on the manuscript; Jessica Sordet-Dessimoz and Alessandra Piersigilli for pathology, Jacques Rougemont and Marion Leleu for statistical consultations; Ehud Drori, Mei-Wen Peng, hl for help with animal experiments; Yunyun Han, Hsinand Jean-Paul Abbu Ying Huang, Sukhvinder Sidhu, and all Hanahan lab members for biological samples and discussions; Melody Swartz, Adrian C. Shieh, and Jacqueline Shields for advice and instruction on the modied invasion assay; Janet Iwasa for the schematic gure; and the EPFL School of Life Sciences technology cores and animal care facility. This research was supported by a core grant from EPFL. Some results are based on data generated by the TCGA pilot project. Received: September 26, 2012 Revised: January 8, 2013 Accepted: February 21, 2013 Published: March 28, 2013 REFERENCES Bai, L., Zhang, X., and Ghishan, F.K. (2003). Characterization of vesicular glutamate transporter in pancreatic alpha - and beta -cells and its regulation by glucose. Am. J. Physiol. Gastrointest. Liver Physiol. 284, G808G814. Bergers, G., Javaherian, K., Lo, K.M., Folkman, J., and Hanahan, D. (1999). Effects of angiogenesis inhibitors on multistage carcinogenesis in mice. Science 284, 808812. Braithwaite, S.P., Adkisson, M., Leung, J., Nava, A., Masterson, B., Urfer, R., Oksenberg, D., and Nikolich, K. (2006). Regulation of NMDA receptor trafcking and function by striatal-enriched tyrosine phosphatase (STEP). Eur. J. Neurosci. 23, 28472856. Christofori, G., Naik, P., and Hanahan, D. (1994). A second signal supplied by insulin-like growth factor II in oncogene-induced tumorigenesis. Nature 369, 414418. Chun, M.G., and Hanahan, D. (2010). Genetic deletion of the desmosomal component desmoplakin promotes tumor microinvasion in a mouse model of pancreatic neuroendocrine carcinogenesis. PLoS Genet. 6, e1001120. Chun, M.G., Mao, J.H., Chiu, C.W., Balmain, A., and Hanahan, D. (2010). Polymorphic genetic control of tumor invasion in a mouse model of pancreatic neuroendocrine carcinogenesis. Proc. Natl. Acad. Sci. USA 107, 17268 17273. Dafni, H., Israely, T., Bhujwalla, Z.M., Benjamin, L.E., and Neeman, M. (2002). Overexpression of vascular endothelial growth factor 165 drives peritumor interstitial convection and induces lymphatic drain: magnetic resonance imaging, confocal microscopy, and histological tracking of triple-labeled albumin. Cancer Res. 62, 67316739. Du, Y.C., Lewis, B.C., Hanahan, D., and Varmus, H. (2007). Assessing tumor progression factors by somatic gene transfer into a mouse model: Bcl-xL promotes islet tumor cell invasion. PLoS Biol. 5, e276. DuFort, C.C., Paszek, M.J., and Weaver, V.M. (2011). Balancing forces: architectural control of mechanotransduction. Nat. Rev. Mol. Cell Biol. 12, 308319. Efrat, S., Linde, S., Kofod, H., Spector, D., Delannoy, M., Grant, S., Hanahan, D., and Baekkeskov, S. (1988). Beta-cell lines derived from transgenic mice
expressing a hybrid insulin gene-oncogene. Proc. Natl. Acad. Sci. USA 85, 90379041. Fantozzi, A., and Christofori, G. (2006). Mouse models of breast cancer metastasis. Breast Cancer Res. 8, 212. Grippo, P.J., and Tuveson, D.A. (2010). Deploying mouse models of pancreatic cancer for chemoprevention studies. Cancer Prev. Res. (Phila.) 3, 13821387. Gutmann, R., Leunig, M., Feyh, J., Goetz, A.E., Messmer, K., Kastenbauer, E., and Jain, R.K. (1992). Interstitial hypertension in head and neck tumors in patients: correlation with tumor size. Cancer Res. 52, 19931995. Hanahan, D. (1985). Heritable formation of pancreatic beta-cell tumours in transgenic mice expressing recombinant insulin/simian virus 40 oncogenes. Nature 315, 115122. Hanahan, D., and Coussens, L.M. (2012). Accessories to the crime: functions of cells recruited to the tumor microenvironment. Cancer Cell 21, 309322. Hanahan, D., and Weinberg, R.A. (2011). Hallmarks of cancer: the next generation. Cell 144, 646674. Hardingham, G.E., and Bading, H. (2010). Synaptic versus extrasynaptic NMDA receptor signalling: implications for neurodegenerative disorders. Nat. Rev. Neurosci. 11, 682696. Harrell, M.I., Iritani, B.M., and Ruddell, A. (2007). Tumor-induced sentinel lymph node lymphangiogenesis and increased lymph ow precede melanoma metastasis. Am. J. Pathol. 170, 774786. Heldin, C.H., Rubin, K., Pietras, K., and Ostman, A. (2004). High interstitial uid pressure - an obstacle in cancer therapy. Nat. Rev. Cancer 4, 806813. Herner, A., Sauliunaite, D., Michalski, C.W., Erkan, M., De Oliveira, T., Abiatari, I., Kong, B., Esposito, I., Friess, H., and Kleeff, J. (2011). Glutamate increases pancreatic cancer cell invasion and migration via AMPA receptor activation and Kras-MAPK signaling. Int. J. Cancer 129, 23492359. Herzig, M., Savarese, F., Novatchkova, M., Semb, H., and Christofori, G. (2007). Tumor progression induced by the loss of E-cadherin independent of beta-catenin/Tcf-mediated Wnt signaling. Oncogene 26, 22902298. Joyce, J.A., Baruch, A., Chehade, K., Meyer-Morse, N., Giraudo, E., Tsai, F.Y., Greenbaum, D.C., Hager, J.H., Bogyo, M., and Hanahan, D. (2004). Cathepsin cysteine proteases are effectors of invasive growth and angiogenesis during multistage tumorigenesis. Cancer Cell 5, 443453. Lau, C.G., and Zukin, R.S. (2007). NMDA receptor trafcking in synaptic plasticity and neuropsychiatric disorders. Nat. Rev. Neurosci. 8, 413426. Lopez, T., and Hanahan, D. (2002). Elevated levels of IGF-1 receptor convey invasive and metastatic capability in a mouse model of pancreatic islet tumorigenesis. Cancer Cell 1, 339353. nez, A., Susanto, E., Peng, H., Lehr, H.A., Malanchi, I., Santamaria-Mart Delaloye, J.F., and Huelsken, J. (2012). Interactions between cancer stem cells and their niche govern metastatic colonization. Nature 481, 8589. Marsden, K.C., Beattie, J.B., Friedenthal, J., and Carroll, R.C. (2007). NMDA receptor activation potentiates inhibitory transmission through GABA receptor-associated protein-dependent exocytosis of GABA(A) receptors. J. Neurosci. 27, 1432614337. , J. (2009). Metastasis: from dissemiNguyen, D.X., Bos, P.D., and Massague nation to organ-specic colonization. Nat. Rev. Cancer 9, 274284. Nicoletti, F., Arcella, A., Iacovelli, L., Battaglia, G., Giangaspero, F., and Melchiorri, D. (2007). Metabotropic glutamate receptors: new targets for the control of tumor growth? Trends Pharmacol. Sci. 28, 206213. North, W.G., Gao, G., Jensen, A., Memoli, V.A., and Du, J. (2010a). NMDA receptors are expressed by small-cell lung cancer and are potential targets for effective treatment. Clin Pharmacol 2, 3140. North, W.G., Gao, G., Memoli, V.A., Pang, R.H., and Lynch, L. (2010b). Breast cancer expresses functional NMDA receptors. Breast Cancer Res. Treat. 122, 307314. Olson, P., Chu, G.C., Perry, S.R., Nolan-Stevaux, O., and Hanahan, D. (2011). Imaging guided trials of the angiogenesis inhibitor sunitinib in mouse models predict efcacy in pancreatic neuroendocrine but not ductal carcinoma. Proc. Natl. Acad. Sci. USA 108, E1275E1284.
Perl, A.K., Wilgenbus, P., Dahl, U., Semb, H., and Christofori, G. (1998). A causal role for E-cadherin in the transition from adenoma to carcinoma. Nature 392, 190193. Perl, A.K., Dahl, U., Wilgenbus, P., Cremer, H., Semb, H., and Christofori, G. (1999). Reduced expression of neural cell adhesion molecule induces metastatic dissemination of pancreatic beta tumor cells. Nat. Med. 5, 286291. Poitout, V., Stout, L.E., Armstrong, M.B., Walseth, T.F., Sorenson, R.L., and Robertson, R.P. (1995). Morphological and functional characterization of beta TC-6 cellsan insulin-secreting cell line derived from transgenic mice. Diabetes 44, 306313. Prickett, T.D., and Samuels, Y. (2012). Molecular pathways: dysregulated glutamatergic signaling pathways in cancer. Clin. Cancer Res. 18, 42404246. Rzeski, W., Turski, L., and Ikonomidou, C. (2001). Glutamate antagonists limit tumor growth. Proc. Natl. Acad. Sci. USA 98, 63726377. Schindelin, J., Arganda-Carreras, I., Frise, E., Kaynig, V., Longair, M., Pietzsch, T., Preibisch, S., Rueden, C., Saalfeld, S., Schmid, B., et al. (2012). Fiji: an open-source platform for biological-image analysis. Nat. Methods 9, 676682. Seidlitz, E.P., Sharma, M.K., Saikali, Z., Ghert, M., and Singh, G. (2009). Cancer cell lines release glutamate into the extracellular environment. Clin. Exp. Metastasis 26, 781787. Sharma, M.K., Seidlitz, E.P., and Singh, G. (2010). Cancer cells release glutamate via the cystine/glutamate antiporter. Biochem. Biophys. Res. Commun. 391, 9195.
Shieh, A.C., and Swartz, M.A. (2011). Regulation of tumor invasion by interstitial uid ow. Phys. Biol. 8, 015012. Shields, J.D., Fleury, M.E., Yong, C., Tomei, A.A., Randolph, G.J., and Swartz, M.A. (2007). Autologous chemotaxis as a mechanism of tumor cell homing to lymphatics via interstitial ow and autocrine CCR7 signaling. Cancer Cell 11, 526538. Stepulak, A., Luksch, H., Gebhardt, C., Uckermann, O., Marzahn, J., Sifringer, M., Rzeski, W., Staufner, C., Brocke, K.S., Turski, L., and Ikonomidou, C. (2009). Expression of glutamate receptor subunits in human cancers. Histochem. Cell Biol. 132, 435445. Takano, T., Lin, J.H., Arcuino, G., Gao, Q., Yang, J., and Nedergaard, M. (2001). Glutamate release promotes growth of malignant gliomas. Nat. Med. 7, 10101015. Takasu, M.A., Dalva, M.B., Zigmond, R.E., and Greenberg, M.E. (2002). Modulation of NMDA receptor-dependent calcium inux and gene expression through EphB receptors. Science 295, 491495. Wegener, N., Nagel, J., Gross, R., Chambon, C., Greco, S., Pietraszek, M., Gravius, A., and Danysz, W. (2011). Evaluation of brain pharmacokinetics of (+)MK-801 in relation to behaviour. Neurosci. Lett. 503, 6872. Wu, J., Zou, H., Strong, J.A., Yu, J., Zhou, X., Xie, Q., Zhao, G., Jin, M., and Yu, L. (2005). Bimodal effects of MK-801 on locomotion and stereotypy in C57BL/6 mice. Psychopharmacology (Berl.) 177, 256263.
Endogenous Retrotransposition Activates Oncogenic Pathways in Hepatocellular Carcinoma

oz-Lopez,3 Daniel J. Gerhardt,2 Malcolm E. Fisher,1 Thu Nguyen,2 Ruchi Shukla,1,11 Kyle R. Upton,2,11 Martin Mun Paul M. Brennan,4 J. Kenneth Baillie,1 Agnese Collino,5 Serena Ghisletti,5 Shruti Sinha,5 Fabio Iannelli,5 Enrico Radaelli,6 Alexandre Dos Santos,7,8 Delphine Rapoud,7,8 Catherine Guettier,7,8 Didier Samuel,7,8 Gioacchino Natoli,5 Piero Carninci,9 Francesca D. Ciccarelli,5 Jose Luis Garcia-Perez,3 Jamila Faivre,7,8 and Geoffrey J. Faulkner1,2,10,*
1Division of Genetics and Genomics, The Roslin Institute and Royal (Dick) School of Veterinary Studies, University of Edinburgh, Easter Bush EH25 9RG, UK 2Cancer Biology Program, Mater Medical Research Institute, South Brisbane QLD 4101, Australia 3Department of Human DNA Variability, Pzer-University of Granada and Andalusian Government Center for Genomics and Oncology (GENYO), 18007 Granada, Spain 4Edinburgh Cancer Research Centre, The University of Edinburgh, Western General Hospital, Crewe Road South, Edinburgh EH4 2XR, UK 5Department of Experimental Oncology, European Institute of Oncology (IEO), Via Adamello 16, 20139 Milan, Italy 6DIVET, School of Veterinary Medicine, University of Milan, Via Celoria, 20133 Milan, Italy 7INSERM U785, Centre He patobiliaire, Villejuif 94800, France 8Universite Paris-Sud, Faculte de Me decine, Villejuif 94800, France 9RIKEN Yokohama Institute, Omics Science Center, 1-7-22 Suehiro-cho , Tsurumi-ku, Yokohama, Kanagawa 230-0045, Japan 10School of Biomedical Sciences, University of Queensland, Brisbane QLD 4072, Australia 11These authors contributed equally to this work *Correspondence: faulknergj@gmail.com http://dx.doi.org/10.1016/j.cell.2013.02.032
SUMMARY
INTRODUCTION Liver cancer accounts for 9% of all cancer deaths worldwide and 12% in developing countries (Jemal et al., 2011). Pathological inspection indicates hepatocellular carcinoma (HCC) in 80% of liver tumors, with infection by hepatitis B virus (HBV) and hepatitis C virus (HCV) being the most prevalent risk factors, followed by chronic alcoholism (Jemal et al., 2011; Perz et al., 2006; Tateishi and Omata, 2012). Although early detection and monitoring of patients with liver cirrhosis can substantially improve 5 year survival rates, progression to advanced HCC reduces average life expectancy to less than 8 months (Llovet et al., 2008). As for other cancers, genome and exome resequencing have elucidated molecular pathways frequently perturbed in HCC (Guichard et al., 2012; Tateishi and Omata, 2012; Totoki et al., 2011), potentially enabling therapeutic intervention informed by the mutational signature of a given tumor. The capacity to catalog the full spectrum of genetic aberrations occurring in HCC is therefore of critical importance. LINE-1 (L1) retrotransposons are a major source of endogenous mutagenesis in humans (Burns and Boeke, 2012; Levin and Moran, 2011). These mobile genetic elements utilize a copy-and-paste mechanism to retrotranspose to new genomic loci, with such success in germ cells that 500,000 L1 copies comprise 17% of the genome (Lander et al., 2001). Of these copies, only 80100 are transposition competent, with distinct subsets of frequently activeor hotL1s driving insertional mutagenesis in each individual genome (Beck et al.,
LINE-1 (L1) retrotransposons are mobile genetic elements comprising 17% of the human genome. New L1 insertions can profoundly alter gene function and cause disease, though their signicance in cancer remains unclear. Here, we applied enhanced retrotransposon capture sequencing (RC-seq) to 19 hepatocellular carcinoma (HCC) genomes and elucidated two archetypal L1-mediated mechanisms enabling tumorigenesis. In the rst example, 4/19 (21.1%) donors presented germline retrotransposition events in the tumor suppressor mutated in colorectal cancers (MCC). MCC expression was ablated in each case, enabling oncogenic b-catenin/Wnt signaling. In the second example, suppression of tumorigenicity 18 (ST18) was activated by a tumor-specic L1 insertion. Experimental assays conrmed that the L1 interrupted a negative feedback loop by blocking ST18 repression of its enhancer. ST18 was also frequently amplied in HCC nodules from Mdr2/ mice, supporting its assignment as a candidate liver oncogene. These proof-of-principle results substantiate L1-mediated retrotransposition as an important etiological factor in HCC.
2010; Brouha et al., 2003). Retrotransposon insertions can profoundly alter gene structure and expression (Cordaux and Batzer, 2009; Faulkner et al., 2009; Han et al., 2004; Levin and Moran, 2011) and have been found in nearly 100 cases of disease (Faulkner, 2011; Hancks and Kazazian, 2012). L1 activity is consequently suppressed in most somatic cells by methylation of a CpG island in the internal L1 promoter (Coufal et al., 2009; Swergold, 1990). By contrast, L1 is often hypomethylated in tumor cells, removing a key obstacle to retrotransposition (Levin and Moran, 2011). Despite this failure to repress L1 transcription, only a handful of L1 insertions had been found in human tumors until very recently (Liu et al., 1997; Miki et al., 1992). High-throughput L1 integration site sequencing has since revealed 9 and 69 de novo L1 insertions, respectively, in lung and colorectal tumors (Iskow et al., 2010; Solyom et al., 2012), whereas cancer genome resequencing elucidated a further 183 tumor-specic L1 insertions in colorectal, ovarian, and prostate cancer (Lee et al., 2012). In this latter study, more than half of all insertions were found in a single colorectal tumor; the other individuals presented fewer than ve tumor-specic L1 insertions on average. These data suggest L1 mobilization may be common in epithelial tumors, though the reasons for possible cell-of-origin restriction are currently unknown. Tumor-specic L1 retrotransposition has not previously been observed in HCC. For several reasons it is, however, a logical cancer in which to expect L1 mobilization. First, HCC is epithelial in origin. Second, HBV and HCV infection are common in HCC; viruses can suppress host defense factors, such as APOBEC proteins, that control retrotransposon activation. APOBEC3G has been shown, for instance, to inhibit both HBV replication and endogenous retrotransposition (Esnault et al., 2005; Turelli et al., 2004). Third, liver inammation precedes HCC and may, via cellular stress, stimulate retrotransposition (Fornace and Mitchell, 1986). Given these facts, we aimed to map L1 integration sites in HCC using retrotransposon capture sequencing (RC-seq) and assess their impact upon oncogenic and tumor suppressor pathways. RESULTS Enhanced Retrotransposon Capture Sequencing To test the hypothesis that L1 mobilizes in HCC, we applied an updated RC-seq protocol to 19 HCC tumors and matched adjacent liver tissue that were conrmed positive for HBV or HCV infection (Table 1). An earlier RC-seq design (Baillie et al., 2011) was modied to incorporate multiplex liquid-phase sequence capture (Figure 1A) using a rened probe pool (Table S1 available online) and a reduced insert size of 220 nt, which enabled high-condence assembly of overlapping paired-end 150 nt reads (Figure 1B). This change simplied genomic alignment and, more importantly, enabled single-nucleotide resolution of retrotransposon integration sites (Figure 1C). After stringent ltering and mapping, an average of 2 million reads were retained per library with >95% identity to active L1, Alu, and SVA families, as well as the most recently active human LTR endogenous retroviruses (Table S2). Optimized sequence capture led to a 4-fold increase in reads aligned to nonreference
Table 1. Nonreference Genome Insertions Detected by RC-Seq Private Validated Germline Germline Tumor-Specic Donor Gender Virus Age Insertions Insertions Insertions 12 15 21 29 32 33 35 42 47 48 49 60 62 70 86 89 95 106 116 M M M M M F F F M M M M M M F M M M M HCV HBV HCV HCV HBV HCV HCV HCV HBV HBV HCV HCV HBV HCV HBV HBV HBV HBV HBV 65 53 51 52 73 57 78 67 61 35 68 48 33 55 56 60 54 60 62 2,082 1,845 2,019 1,602 1,681 1,982 1,786 1,594 1,581 1,744 1,644 1,570 1,750 1,673 1,701 1,739 1,773 2,141 1,532 202 216 271 44 100 234 96 43 77 212 58 33 153 82 50 163 88 48 71 3 1 0 0 0 2 0 0 2 0 0 0 0 0 0 4 0 0 0
F, female; M, male. Please see Tables S2 and S3 for supporting data and details.
genome L1s per library compared to previous RC-seq based on solid-phase arrays and similar sequencing depth (Baillie et al., 2011). The improved resolution of RC-seq also allowed us to discriminate a required minimum of two unique amplicons in support of any nonreference genome insertion (see Extended Experimental Procedures). Frequent Retrotransposition in the Human Germline A total of 7,689 nonreference genome insertions were detected in 19 tumor (T) samples and 19 matched nontumor (NT) liver samples. Of these, we annotated 7,644 as putatively germline (Table S3) because of their presence in (1) databases of retrotransposon-induced polymorphisms (Beck et al., 2010; Ewing and Kazazian, 2010; Iskow et al., 2010; Wang et al., 2006), (2) pre-existing insertions annotated by pooled blood RC-seq (Baillie et al., 2011), (3) multiple individuals, or (4) nontumor liver. L1, Alu, SVA, and LTR-anked retrotransposons comprised 13.5%, 81.8%, 4.3%, and 0.4% of germline insertions, respectively. As expected, L1-Ta and L1-pre-Ta (99.3%) and AluY (99.7%) were the main L1 and Alu subfamilies active in germ cells (Mills et al., 2007). A total of 2,241 germline insertions were found in only one individual each (Table 1 and Table S3) and were not annotated by the aforementioned retrotransposon polymorphism databases, suggesting that these were private or rare mutations or, alternatively, had occurred in early development (Garcia-Perez et al., 2007; Kano et al., 2009). RC-seq detected 1,489 (66.4%) insertions at both their 50 and 30 ends, enabling us to model the characteristic sequence features of L1-mediated retrotransposition.
Each individual genome contained on average 244 nonreference genome L1 insertions, a gure 60% and 80% higher, respectively, than recent L1 insertion site sequencing on cell lines (Ewing and Kazazian, 2010) and single cells (Evrony et al., 2012). Therefore, to assess the RC-seq false-positive rate, we randomly selected 200 germline insertions (173 Alu, 14 L1, 11 SVA, and 2 LTR) for site-specic PCR validation (Table S5). Of these, we conrmed 197 (98.5%). The remaining three insertions (2 SVA and 1 Alu) occurred in repetitive genomic regions and were detected by multiple unique reads in at least ten different samples each, indicating that these may have represented PCR false negatives. These comparisons and experiments together demonstrate the sensitive and accurate mapping of bona de retrotransposition events by RC-seq and further highlight ongoing L1 retrotransposition in the global human population (Beck et al., 2010; Ewing and Kazazian, 2010; Huang et al., 2010; Iskow et al., 2010). Activation of b-Catenin/Wnt Signaling via L1-Mediated Ablation of MCC To assess the potential tumorigenic consequences of the identied nonreference genome insertions, we selected and validated, by insertion site PCR, 31 L1, Alu, and SVA insertions in genes generally implicated to play a causal role in cancer (Futreal et al., 2004) or specically in HCC (Guichard et al., 2012), including L1 insertions in the proto-oncogene ALK and the tumor suppressor FHIT (Table S5). Quantitative RT-PCR indicated, however, that 28/31 of these germline insertions did not signicantly perturb host gene expression in tumor or nontumor liver versus control liver from ve unaffected individuals (data not shown). Strikingly, the three remaining insertions all coincided with strong inhibition of the tumor suppressor mutated in colorectal cancers (MCC) (Higgins et al., 2007). MCC is expressed in liver (Senda et al., 1999) and regulates the oncogenic b-catenin/Wnt signaling pathway frequently activated in HCC (Fukuyama et al., 2008; Guichard et al., 2012; Totoki et al., 2011). In vitro experiments have established that siRNA knockdown of MCC mRNA dramatically increases b-catenin (CTNNB1) expression, whereas MCC overexpression inhibits cellular proliferation (Fukuyama et al., 2008; Matsumine et al., 1996). MCC is also an intriguing HCC candidate gene because of its genomic proximity to APC, a major tumor suppressor mutated in familial adenomatous polyposis preceding colorectal cancer (Groden et al., 1991; Kinzler et al., 1991). It is important to note that mutated APC occurs in <2% of HCC cases versus >60% of colorectal carcinomas (Guichard et al., 2012; Powell et al., 1992). We therefore hypothesized that germline retrotransposition events specically inhibited MCC tumor suppressor function in liver. To test this prediction, we assessed the impact of each MCC mutation upon MCC, APC, and CTNNB1 expression. Three germline retrotransposon insertions were found in MCC. The rst of these, labeled MCC-L1-a, comprised a 5.3 kb L1-Ta oriented in sense to MCC in donors 70 and 95 (Figure 3A). Another L1-Ta, labeled MCC-L1-b, was full-length (6 kb), occurred at a different genomic position in donor 116, and was oriented antisense to MCC (Figure 3B). Finally, in donor 33, we found an AluY (MCC-Alu; Figure 3C) inserted in an
Figure 1. Enhanced RC-Seq

(A) Multiplexed Illumina libraries are hybridized to liquid-phase sequence capture probes targeting the 50 and 30 ends of recently active human retrotransposons (Table S1). (B) Paired-end 150-mer sequencing of 220 nt inserts enables contig assembly of each read pair into a single read. (C) Assembled reads with a 50 or 30 section of an active retrotransposon at one end (highlighted in red) are retained. The opposite end is then aligned to the reference genome, indicating the position of known and novel insertions.
Without any additional sequencing, we were able to analyze insertions for the presence of target site duplications (TSDs), an L1-endonuclease recognition motif (Jurka, 1997), and a polyA tail (Figures 2A and 2B). These features consistently resembled target-primed reverse transcription (TPRT) for L1, Alu, and SVA, again illustrating the primary retrotransposition mechanism in germ cells (Cost et al., 2002; Jurka, 1997). We also identied 160 previously undetected full-length (>99.9%) L1 copies, including 115 with paired 50 /30 detection (Figure 2C; Table S4) and 82 each found in a single donor only. All were annotated as L1-Ta or pre-Ta. These potentially hot L1s added to a recent cohort of full-length L1 insertions found in six geographically diverse individuals via fosmid screening and sequencing (Beck et al., 2010). Of 68 L1 insertions reported by Beck et al. (2010), we detected 49 (72.1%), including 15/18 (83.3%) with an allelic frequency >5%. Of the 49 insertions common to both studies, 46 (93.9%) were base-pair identical in genomic position. These results conrm strong agreement between RC-seq and the conservative fosmid-based approach of Beck et al. (2010).
Figure 2. Characteristics of Recent Germline Retrotransposition in Humans

(A) Distributions of target site duplication length and poly-A tail length for L1, Alu, and SVA. (B) Consensus sequence motifs (Crooks et al., 2004) at integration sites closely resembled the canonical L1 endonuclease recognition sequence. (C) Genomic positions (indicated by red lines) of 115 previously unobserved full-length L1 insertions detected at both termini by RC-seq. Please see Tables S4 and S5 for further supporting data.
ENCODE-delineated enhancer (Thurman et al., 2012). Insertion site PCR revealed that MCC-L1-a was heterozygous in donor 70 and homozygous (or possibly hemizygous) in donor 95, whereas MCC-L1-b and MCC-Alu were heterozygous in donor 116 and donor 33, respectively (Figure 3D). An immunoblot indicated that MCC was dramatically less abundant in tumor and nontumor samples from all four donors compared with control liver tissue (Figure 4A). By contrast, CTNNB1 was expressed much more strongly in the affected donors than in controls (Figure 4A). This inverse relationship was consistent with MCC suppression of CTNNB1 through protein-protein interactions, as reported elsewhere (Fukuyama et al., 2008). As a corroborating example, immunohistochemistry performed on tumor and nontumor tissue from donor 116 conrmed cytoplasmic CTNNB1 accumulation (Figure S1), a strong indicator that the factors controlling CTNNB1 expression outside of the plasma membrane were absent and that many cells had entered a proliferative state (Nhieu et al., 1999). Quantitative RT-PCR indicated that MCC transcription was severely reduced (p < 0.02p < 0.002, t test, degrees of freedom [df] = 19) in all four tumors compared to normal liver (Figures 4B). MCC-L1-a and MCC-L1-b strongly suppressed MCC expression in donor 95 and donor 116s nontumor liver, respectively (Figure 4B). MCC was also signicantly downregulated in tumor versus nontumor in all four individuals (p < 0.0001, t test, df = 4). Capillary sequencing of each MCC exon revealed only one missense mutation, a homozygous SNP (570 A > G) in donor 33 MCC exon 5, producing an Arg > Lys substitution in the putative CTNNB1 binding domain of MCC (Fukuyama et al., 2008).
Therefore, MCC-L1-a, MCC-L1-b, and MCC-Alu were the primary enactors of MCC transcriptional inhibition, potentially assisted by other modications to MCC or its upstream regulatory pathway. Finally, we performed qRT-PCR to evaluate APC transcription coincident with mutated MCC. We found no signicant differential transcription of APC in tumor or nontumor liver from the four affected donors versus normal liver controls (Figure S2). In donor 95, APC was downregulated signicantly in tumor versus nontumor (p < 0.003, t test, df = 4) but only by 30% versus normal liver controls. By contrast, MCC-L1-a, the homozygous L1 insertion in donor 95, severely reduced MCC transcription in both tumor (83%) and nontumor (63%) samples compared with normal liver controls (Figure 4B), demonstrating that the primary effect of MCC-L1-a was on MCC rather than APC. These data in sum conrmed that (1) L1-mediated retrotransposition in MCC specically repressed MCC and not APC and (2) CTNNB1 was strongly induced in all four affected individuals, indicating activation of a major HCC oncogenic pathway. Somatic L1 Mobilization in HCC Forty-ve nonreference genome insertions were annotated as tumor specic. These consisted of 17 L1, 27 Alu, and 1 SVA. We rst validated each L1 insertion with insertion site PCR, including capillary sequencing of their 50 and 30 ends (Table S6). All 17 L1s successfully amplied; 12 conrmed as tumor-specic, and 5 were found in both tumor and nontumor liver. Further examination of the tumor-specic set revealed uniform usage of the degenerate L1 endonuclease motif highlighted in Figure 2B. In
Figure 3. Structure and Validation of Germline L1 and Alu Insertions in MCC

(A) MCC mutant allele MCC-L1-a: a 5.3 kb L1-Ta detected by RC-seq at its 50 and 30 ends in 70T, 70NT, 95T, and 95NT. The L1 was anked by a 13 nt TSD. Primers used for PCR validation (1,2,3) and RC-seq reads (red/white bars) are indicated above the gene structure. Note: L1 not drawn to scale. (B) MCC mutant allele MCC-L1-b: a full-length (6 kb) L1-Ta detected by RC-seq at its 50 and 30 ends in 116T and 116NT. The L1 was antisense to MCC and was anked by a 14 nt TSD. Primers are indicated as for (A). (C) MCC mutant allele MCC-Alu: an AluY detected by RC-seq at its 30 end in 33T and 33NT. The AluY was antisense to MCC, had a 15 nt TSD, and bisected an annotated enhancer (Thurman et al., 2012). Primers (1,2) are indicated below the gene structure. (D) Insertion-site PCR validation conrmed that MCC-L1-a, MCC-L1-b, and MCC-Alu were present in the corresponding tumor and nontumor samples (Table S5). The wild-type allele was absent for donor 95, indicating a homozygous L1 insertion.
two examples, PCR amplication of the 50 junction was repeatedly unsuccessful, preventing TSD characterization, an outcome possibly due to gross genomic abnormality at the L1 insertion site (Gilbert et al., 2002). Eight of the other integration sites incorporated TSDs, whereas the remaining two examples involved small genomic deletions 30 of the insertion site and no TSD. Somatic L1 mobilization occurred in donors 12, 15, 33, 47, and 89 (Table 1), with the latter individual presenting four insertions. Two L1 copies (chr11:60136439 and chrX:99180431) were greater than 5.3 kb in length, but no insertions were full-length. All 12 somatic L1 insertions were from the L1-Ta subfamily. We next evaluated 13 Alu insertions and the single SVA insertion found only in tumor, using insertion site PCR. In all cases, amplication occurred in both tumor and adjacent liver DNA, indicating germline insertions. Our primary explanation for this result is that there are several thousand potentially active AluY copies in the genome, compared to fewer than 100 active L1s (Bennett et al., 2008; Brouha et al., 2003). As seen previously, the RC-seq read count per Alu is consequently 75% lower than for L1 (Baillie et al., 2011), making false-negatives in the nontumor control more likely for Alu than for L1. A secondary explanation is that chromosomal gain is very common in HCC (Guichard et al., 2012), increasing the probability that some germline insertions are detected in tumor but not in adjacent nontumor liver. A nal possibility is that mutations in individual precancerous cells are clonally amplied in tumors and are called as tumor-specic by RC-seq and germline by insertion
site PCR. However, this was unlikely, as we consistently observed strong PCR amplication in both tumor and nontumor liver in these cases. Consequently, RCseq reliably identies new L1, Alu, and SVA mobilization events but requires insertion site PCR to annotate tumor-specic insertions. In recent work, we reported somatic L1 mobilization in the normal brain but did not evaluate other organs (Baillie et al., 2011). For the current study, somatic L1 insertions in nontumor liver were considered difcult to evaluate because of the frequent occurrence of chromosomal loss in tumors. In this scenario, germline L1 insertions may be deleted in tumor but retained in nontumor liver and called somatic events. Nonetheless, we identied 21 L1 insertions restricted to nontumor liver in the set putatively annotated as germline and as a proof-of-principle experiment selected an example (chr13:27423763) for insertion site PCR and capillary sequencing (Table S6). This 2.5 kb L1Ta insertion was detected only in liver and, interestingly, had a long (127 nt) TSD (Figure S3). A germline L1 insertion deleted in tumor cells would reasonably be expected to be detected in the nontransformed cells (e.g., lymphocytes) inltrating a tumor (Unitt et al., 2005). Therefore, this very likely represented a bona de liver-specic somatic L1 insertion in the preneoplastic liver of donor 47. Consequently, hepatocytes, or their progenitor cells, may support limited somatic L1 mobilization, though the contribution of this activity to malignancy remains unclear. L1 Hypomethylation Enables Tumor-Specic Mobilization To assess whether L1 activity and L1 methylation state were correlated in HCC samples, we performed bisulphite conversion
Figure 4. Downregulation of MCC

(A) Relative expression of MCC and CTNNB1 in control liver tissue compared to tumor and nontumor liver tissue from donors 33, 70, 95, and 116. An immunoblot performed against anti-MCC, anti-CTNNB1, and anti-GAPDH (loading control) antibodies detected strong MCC expression only in controls and strong CTNNB1 expression only in MCC mutant donors. MCC was also detected weakly in donor 70NT, in line with qRT-PCR results shown in (B). Expected protein molecular weights are marked on right. Note: anti-MCC and anti-CTNNB1 antibodies produce double bands. See Figure S1 for donor 116 CTNNB1 immunohistochemistry. (B) Downregulation of MCC transcription in MCC mutant donors: qRT-PCR revealed that MCC mRNA was signicantly reduced compared to control liver tissue in donors 33 (tumor only), 70 (tumor only), 95 (tumor and nontumor), and 116 (tumor and nontumor). **p < 0.002 and *p < 0.02, two-tailed t test, df = 19. In all four donors, MCC was also strongly downregulated in tumor versus nontumor samples (p < 0.0001, two-tailed t test, df = 10). Data are presented as mean SD. See Figure S2 for APC qRT-PCR.
Figure 5. L1 Promoter Activation in HCC

(A) Bisulphite analyses in HCC patients versus controls revealed a signicant decrease in L1 promoter methylation in tumor samples. Each column represents the methylation of 20 CpG residues found within the internal L1-Ta promoter. Values are presented as the mean percent of CpG methylation SEM (***p < 0.0005, ****p < 2.5 3 1018, chi-square test). Please see Figure S4 for detailed analysis. (B) TaqMan qRT-PCR measurement of L1 ORF2 indicated signicantly increased L1 transcription in tumor and adjacent matched liver tissue versus controls. Data for each group (tumor, nontumor, and control) were pooled and presented as mean SEM (**p < 0.003, two-tailed t test, df = 22, Bonferroni correction). (C) As for (B), except observed at the L1 50 UTR (*p < 0.006). Please see Figure S3 for an example of a somatic L1 insertion in nontumor liver.
of gDNA and capillary sequenced the CpG island present in the canonical L1 promoter. Eight tumors (15T, 47T, 48T, 62T, 89T, 95T, 106T, and 116T) matched adjacent liver samples, and control liver samples were analyzed. In the tumor group, 54.8% of L1-promoter CpG dinucleotides were methylated, compared with 69.2% in nontumor liver, a strongly signicant difference (p < 2.5 3 1018, chi-square test, n = 8) (Figure 5A). On average, all but one CpG was hypomethylated in tumor, with the remaining CpG being equally methylated in tumor and nontumor liver (Figure S4A). Hypomethylation was not observed in grouped adjacent nontumor liver tissue versus controls. As shown in Figure S4B, a subset of four individuals (donors 47, 89, 106, and 116) presented much stronger L1-promoter hypomethylation in their tumor (40.5%) versus nontumor liver (72.3%) samples compared with the remaining individuals (69.2% versus 66.1%). The three individuals with tumor-specic L1 insertions and L1 methylation data (donors 15, 47, and 89) yielded a strong correlation between L1 hypomethylation percentage and tumor-specic L1 insertion count (r = 0.97; n = 3). Donor 89 exhibited the strongest tumor-specic L1 hypomethy106 Cell 153, 101111, March 28, 2013 2013 Elsevier Inc.
lation and also had the most tumor-specic L1 insertions (Figure S4C). Donor 15 showed only tumor-specic hypomethylation distal to the L1 50 end, whereas donors 47 and 89 were hypomethylated across the L1 promoter (Figure S4C). Hypomethylation of the L1 promoter enables transcription of full-length L1 mRNAs that are translated to form the L1 mobilization machinery (Ostertag and Kazazian, 2001a). We therefore used cDNA synthesized with L1-specic primers (Wissing et al., 2012) to quantify L1 expression levels by TaqMan qRT-PCR. In this analysis, we measured L1 mRNA levels using primers targeting L1 ORF2 (Figure 5B) and the L1 50 UTR (Figure 5C). In both
Figure 6. A Tumor-Specic L1 Insertion Causes Induction of ST18

(A) ST18 mutant allele: a 0.4 kb L1-Ta insertion antisense to ST18. Primers used for PCR validation (1,2) are indicated above the gene. (B) L1 insertion, magnied view: RC-seq detected the L1 50 and 30 termini, indicating a 17 nt TSD and a 50 inversion. (C) Insertion-site PCR validation: the L1 was detected only in 47T, whereas the empty site was found in both 47T and 47NT. (D) qRT-PCR: ST18 was upregulated 4-fold in 47T versus 47NT (*p < 0.005, two-tailed t test, df = 4). Data are presented as mean SD. (E) ST18 immunoblot: ST18 (115 kDa) was enriched in 47T versus 47NT and normal liver controls. (F) ST18 immunohistochemistry: accumulation of ST18 (brown) was observed in tumor nodules compared to surrounding nontumor regions. Nuclei were stained with hematoxylin (blue). (G) A palindromic sequence motif was bisected by the L1. Each 8 nt unit (a and b, light green) contained a subsequence 1 nt different to a PIT1enhancer motif known to bind MYT1 (Rhodes et al., 1993). A second motif 58 bp from the L1 integration site matched the consensus CEBPA binding motif (orange). (H) ChIP followed by quantitative real-time PCR in Huh7 cells conrmed enrichment for ST18 bound to the putative ST18-enhancer element illustrated in (G), compared to GAPDH. Data from antibodies targeting both the N termini and C termini of ST18 are shown. Signicance values were calculated using two-tailed t tests (df = 4). Data are presented as mean SD. Please see Tables S6 and S7 and Figures S5 and S6 for further information regarding tumorspecic L1 insertions and additional ST18 characterization.
cases, signicant enrichment was observed in tumor and nontumor versus normal controls (p < 0.003 for ORF2, p < 0.006 for 50 UTR, t tests, df = 22). Together, these data showed that L1 was activated and transcribed in HCC, coincident with hypomethylation of the L1 promoter. ST18 Activated by a Tumor-Specic L1 Insertion Tumor-specic L1 insertions were observed in six protein-coding genes (Table S6). Quantitative RT-PCR indicated that two of these genes (STXBP5L and SLC5A8) were not expressed in liver. The expression of three other genes was reduced 2-fold to 6-fold in tumor versus adjacent liver (p < 0.05, t test, df = 4), including a 30 UTR insertion in SLC2A1 and intronic insertions in PHGDH and EFHD1 (Figure S5). These examples resemble those seen in other cancers in which intragenic L1 insertions in tumors coincided with reduced host gene expression (Lee et al., 2012). To our knowledge, downregulation of SLC2A1, PHGDH, or EFHD1 has not previously been associated with cancer. The remaining tumor-specic L1 insertion occurred in donor 47 and was associated with activation of the transcriptional
repressor suppression of tumorigenicity 18 (ST18), a member of the MYT1 zinc-nger transcription factor family (Yee and Yu, 1998). Contrasting reports depict ST18 as a tumor suppressor and as an oncogene in different cancers (Jandrig et al., 2004; Steinbach et al., 2006). ST18 is, however, very poorly expressed in liver (Jandrig et al., 2004), making it unlikely to act as a tumor suppressor in this context. Ectopic host gene expression was an unusual consequence of an L1 insertion given that these events are usually repressive (Han et al., 2004). As such, we hypothesized that ST18 was a candidate liver oncogene activated via an unknown mechanism triggered by an intronic L1 insertion. Initial data from RC-seq indicated a heavily 50 truncated, 410 bp L1-Ta arranged antisense to ST18 (Figure 6A). The integration site incorporated a 17 nt TSD, a degenerate L1 endonuclease motif (GC/AAAA), and a 112 bp 50 inversion of the L1 (Figure 6B), consistent with twin priming (Ostertag and Kazazian, 2001b). We then conrmed these features by PCR amplication and capillary sequencing of the L1 50 and 30 junctions, indicating a tumor-specic L1 insertion (Figure 6C). PCR on DNA extracted from three distinct biopsies taken from the same tumor detected
the L1 in all three regions, suggesting clonal amplication of tumor cells with the L1 mutant ST18. As noted above, qRT-PCR indicated that ST18 expression was signicantly increased in tumor versus adjacent nontumor liver (p < 0.005, t test, df = 4) (Figure 6D). To corroborate this result, we performed an immunoblot and immunohistochemistry with an anti-ST18 antibody and found ST18 was indeed ectopically expressed in donor 47 tumor (Figures 6E and 6F). Chromosomal gain and regional copy number variation (CNV) have previously been reported for chromosome 8q, the genomic region containing ST18 (Guichard et al., 2012). However, quantitative real-time PCR on gDNA indicated no ST18 CNV in donor 47 tumor. Thus, tumor cells containing the ST18 L1 mutation were clonally amplied without CNV of the ST18 locus, followed by ST18 transcriptional activation. In response, we predicted that ST18 was activated by insertional mutagenesis of a cis-regulatory element proximal to the L1. In silico analysis of the L1 integration site indicated that it bisected a palindromic motif containing two 8 bp units differing by one nucleotide and separated by 3 bp (Figure 6G). The probability of a random insertion in this motif, even allowing for a mismatch in the palindrome and a generous gap of %11bp, was less than 1/1,000 (permutation test). Intriguingly, each unit was only one nucleotide different to a strong MYT1 binding motif found in the enhancer of PIT1 (Rhodes et al., 1993). Previous experiments predicted that these units would bind MYT1 with reduced efciency (Jiang et al., 1996), though transcription factors incorporating two zinc-nger domains, as for MYT1, are known to greatly gain efciency through binding tandem DNA motifs (Yee and Yu, 1998). The putative MYT1 binding site was proximal to a strong binding site for CEBPA, a transcription factor enriched in liver and known to bind active enhancers (Johnson et al., 1987). Based on this computational analysis, we predicted that the L1 bisected an enhancer normally bound to the zinc ngers of the ST18 MYT1 domain. To test this experimentally, we performed chromatin immunoprecipitation (ChIP) of DNA bound to the ST18 protein in Huh7 cells, followed by PCR amplication of the putative ST18 enhancer. This assay conrmed that, absent an L1 insertion, ST18 was preferentially bound to its own enhancer (p < 0.0004, t test, df = 4) (Figure 6H). An L1 insertion in the ST18 binding site would reasonably be expected to displace this repressive mark from the enhancer. Thus, we experimentally validated a model of ST18 activation in which a negative feedback loop was interrupted by a tumor-specic L1 insertion. Finally, in view of the clonal amplication of tumor cells containing ectopically expressed ST18, we engaged complementary in vitro and in vivo experimental models to assess ST18 oncogenic function in HCC. Although ST18 is poorly expressed in liver, we found it to be abundant in several liver cancer cell lines (Figure S6A). We then determined the frequency of ST18 CNV in an Mdr2/ mouse model of inammation-driven HCC. TaqMan quantitative real-time PCR detected ST18 amplication in 4/23 Mdr2/ HCC nodules and no deletions (Table S7). A disproportionately high percentage of advanced tumors (75%) presented ST18 amplication. ST18 expression was also significantly higher in nodules with amplied ST18 compared with
wild-type mouse liver (p < 0.0001, t test, df = 19) (Figure S6B). These experiments demonstrate concordance of frequent ST18 amplication and upregulation in human and mouse models of HCC, results consistent with ST18 functioning as a candidate liver oncogene. DISCUSSION The present study highlights endogenous L1-mediated retrotransposition in the germline and somatic cells of HCC patients. We report two archetypal mechanisms revealing MCC and ST18 as HCC candidate genes. MCC is, for the many reasons highlighted above, a highly plausible liver tumor suppressor. Four out of 19 individuals studied here, including two cases each of HBV and HCV infection, presented distinct germline L1 or Alu insertions contributing to MCC suppression in tumor and nontumor liver tissue. Strong upregulation of CTNNB1 in all four donors was consistent with prior observations that CTNNB1 is inhibited by MCC (Fukuyama et al., 2008). It is also interesting that MCC-L1-a was homozygous in donor 95, and therefore, MCC was almost certainly downregulated in the liver of this patient prior to HBV infection, i.e., preceding viral challenge, cirrhosis, and tumorigenesis. We also demonstrate that MCC transcriptional repression in all four affected donors was exclusive of APC. Mutated APC is common in colorectal cancer but rare in HCC (Guichard et al., 2012; Powell et al., 1992). Even in colon, MCC presents numerous properties of a tumor suppressor (Bouwmeester et al., 2004; Fukuyama et al., 2008; Kohonen-Corish et al., 2007; Matsumine et al., 1996). Indeed, a Sleeping Beauty transposon mutagenesis screen using a mouse model of colorectal cancer found specic mutations in MCC and APC at a 1:9 ratio (Starr et al., 2009). Very recently, exome resequencing identied sporadic MCC point mutations in HCC (Guichard et al., 2012). Thus, MCC has potential to act as a liver tumor suppressor independent of APC, and our results support this potentially pivotal line of enquiry. Tumorigenic retrotransposition in somatic cells was rst observed 20 years ago, coincidentally in the APC gene of an individual with colorectal cancer (Miki et al., 1992). High-throughput sequencing has since provided the means to test whether tumor-specic retrotransposition is a common feature of cancer. Our results indicate that L1 mobilization occurs in a minority of HCC tumors, adding to the list of epithelial cancers (lung, colon, ovarian, and prostate) known to support the phenomenon (Iskow et al., 2010; Lee et al., 2012; Miki et al., 1992; Solyom et al., 2012). Although transformed tumor cells, including liver cancer cell lines, support frequent transgenic L1 mobilization (Moran et al., 1996), it is unknown whether endogenous L1 activation precedes neoplastic transformation in vivo. For this reason, it was interesting that L1 transcription was found in liver tissue adjacent to tumors, in addition to an example of somatic L1 mobilization. Finally, in a small cohort of tumor-specic L1 insertions, we identied mobilization via TPRT, twin priming, and a third mechanism resulting in a small deletion and no TSD, as reported elsewhere (Gilbert et al., 2002). These observations highlight the multiple routes by which L1 mobilization alters the tumor cell genome.
The results presented here corroborate recent data generated via whole-genome sequencing of other cancers. As in our study, Lee et al. (2012) described tumor-specic L1 insertions bearing the hallmark features of TPRT and also found intragenic L1 insertions in differentially expressed genes (Lee et al., 2012). One distinct feature of the current study is our discovery that germline L1 and Alu insertions signicantly perturb expression of genes relevant to HCC. Another advance is our explanation for the occasional activation of host genes by tumor-specic L1 insertions, based on an example of an interrupted negative feedback loop. The method presented by Lee et al. (2012) is convenient inasmuch as existing whole-genome sequencing data can be reanalyzed to identify novel retrotransposon insertions. However, we generated similar results with per sample sequencing depth 1/12 that of Lee et al. (2012), suggesting RC-seq is more efcient for new studies specically focused on retrotransposons. L1-mediated insertional mutagenesis revealed ST18 as a candidate oncogene in HCC. Numerous corroborating observations support this possibility, including (1) clonal amplication of tumor cells containing the L1 mutant ST18, (2) ectopic ST18 transcription and translation in tumor not seen in adjacent nontumor liver or control liver, (3) consistent ST18 expression in transformed liver cancer cell lines, (4) frequent amplication of ST18 in HCC nodules taken from Mdr2/ mice, and (5) induction of ST18 transcription in those animals. However, we do not make any conclusion regarding the function of ST18 as a tumor suppressor or oncogene outside of the liver and draw attention in this matter to KLF4, a transcriptional repressor known to function as a tumor suppressor and as an oncogene, depending on context (Rowland et al., 2005). Overall, our results illustrate the conuence of multiple genetic aberrations in HCC, where inherited and de novo retrotransposition events form part of a wider mutational landscape. The experiments presented here and elsewhere suggest L1 activity varies substantially between individuals and cancer types (Iskow et al., 2010; Lee et al., 2012; Solyom et al., 2012). It remains to be proven whether this phenomenon correlates with prognosis, is useful in a diagnostic capacity, or can be subjected to exogenous interference in vivo. Nonetheless, we can conclude that L1-mediated retrotransposition is a potentially crucial source of mutations that can reduce the tumor suppressive capacity of somatic cells in HCC.
EXPERIMENTAL PROCEDURES Full protocols can be found in the Extended Experimental Procedures. Samples Tumor and nontumor liver tissues from 19 HCC patients with a conrmed HBV patobiliaire, Paul-Brousse or HCV infection were provided by the Centre He Hospital. DNA and RNA were extracted with a DNeasy Blood and Tissue Kit (QIAGEN, Hilden, Germany) and a mirVana miRNA Isolation Kit (Life Technologies, Carlsbad, CA, USA), respectively. Control liver samples from ve donors were provided by the Edinburgh Sudden Death Brain and Tissue Bank. DNA and RNA were isolated through standard phenol-chloroform extraction and RNA-Bee RNA isolation reagent (Tel-Test), respectively. Samples were analyzed with approval from the French Institute of Medical Research and Health (Ref: 11-047), the East of Scotland Research Ethics Service (Ref: LR/11/ES/0022), and the Mater Health Services Human Research Ethics Committee (Ref: 1915A).
RC-Seq Library Preparation, Sequencing, and Analysis Multiplexed DNA sequencing libraries were constructed for HCC tumor and nontumor samples using a paired-end Illumina TruSeq Kit with substantial modications. Briey, 1 mg of sonicated DNA size selected for an insert size of 200250 bp was used for each library and amplied by six cycles of ligation-mediated PCR (LM-PCR). Libraries were then pooled in groups of 4 to 6 and hybridized to an updated custom Roche NimbleGen sequence capture array comprising oligos tiling the 50 and 30 termini of active human retrotransposon consensus sequences (Figure 1; Table S1). Libraries were again amplied by six cycles of LM-PCR and sequenced on an Illumina HiSeq2000. After quality ltering, each read pair was assembled into a contig, aided by 2 3 150mer sequencing and a 220 nt insert size. Read contigs were then aligned to retrotransposon consensus sequences to determine their retrotransposon donor family, aligned to the human reference genome (hg19) to determine their genomic position, and nally formed into clusters. PCR Validation Germline retrotransposon insertions detected by RC-seq were rst validated by a standard empty site/lled site PCR assay and then, if unsuccessful, with PCR targeting an insertion site 50 or 30 end. Tumor-specic insertions were characterized with a similar strategy but also incorporated 50 and 30 end capillary sequencing. All validation was performed on nonamplied DNA stored and handled separately from postamplication RC-seq products. Primers were designed using custom Python scripts and Primer3. qRT-PCR Complementary DNA was synthesized from total RNA using random hexamers, except for L1 analyses, where a specic sense L1 primer was used. qRT-PCR was performed using a LightCycler 480 (Roche, Indianapolis, IN, USA), and values were normalized to TATA-binding protein (TBP). For primer sequences, see Table S8. ACCESSION NUMBERS RC-seq FASTQ les were deposited in the European Nucleotide Archive (ERP001476). SUPPLEMENTAL INFORMATION Supplemental Information includes Extended Experimental Procedures, six gures, and eight tables and can be found with this article online at http:// dx.doi.org/10.1016/j.cell.2013.02.032. ACKNOWLEDGMENTS We thank Professor Haig Kazazian, Professor Bert Vogelstein, Dr. Bruno Amati, and Dr. Alister Funnell for helpful discussion. We thank Professor Daniel Adam and Dr. Eric Vibert Azoulay, Professor Denis Castaing, Professor Rene patobiliaire, Villejuif), and the Tissue Biobank Group (AP-HP and (Centre He Paris-Sud) for providing HCC specimens. We thank Dr. Christine Universite Beck and Professor John Moran for providing the genomic coordinates of a recently published full-length L1 insertion cohort (Beck et al., 2010). Research performed in the laboratories of D.S., G.N., P.C., F.D.C., J.F., and G.J.F. was funded by the European Unions Seventh Framework Programme (FP7/2007-2013) under grant agreement No. 259743 underpinning the MODHEP consortium. J.K.B. and P.M.B. were supported by Wellcome Trust Clinical Fellowships (090385/Z/09/Z and 090386/Z/09/Z, respectively) through the Edinburgh Clinical Academic Track (ECAT). J.L.G.P was supported by FP7-PEOPLE-2007-4-3-IRG, CICE-FEDER-P09-CTS-4980, PeS-FEDER-PI002, FIS-FEDER-PI11/01489, and the Howard Hughes Medical Institute (IECS-55007420). J.F. acknowledges the support of the Institut National de et de la Recherche Me dicale (INSERM), the Association pour la la Sante Recherche contre le Cancer (ARC 4866), and the Institut National du Cancer (INCa 2009-PAIR-CHC). G.J.F. acknowledges the support of a New Investigator Award from the British BBSRC (BB/H005935/1) and a C.J. Martin Overseas Based Biomedical Fellowship from the Australian NHMRC (575585).
Received: July 27, 2012 Revised: December 21, 2012 Accepted: February 19, 2013 Published: March 28, 2013 REFERENCES Baillie, J.K., Barnett, M.W., Upton, K.R., Gerhardt, D.J., Richmond, T.A., De Sapio, F., Brennan, P.M., Rizzu, P., Smith, S., Fell, M., et al. (2011). Somatic retrotransposition alters the genetic landscape of the human brain. Nature 479, 534537. Beck, C.R., Collier, P., Macfarlane, C., Malig, M., Kidd, J.M., Eichler, E.E., Badge, R.M., and Moran, J.V. (2010). LINE-1 retrotransposition activity in human genomes. Cell 141, 11591170. Bennett, E.A., Keller, H., Mills, R.E., Schmidt, S., Moran, J.V., Weichenrieder, O., and Devine, S.E. (2008). Active Alu retrotransposons in the human genome. Genome Res. 18, 18751883. Bouwmeester, T., Bauch, A., Ruffner, H., Angrand, P.O., Bergamini, G., Croughton, K., Cruciat, C., Eberhard, D., Gagneur, J., Ghidelli, S., et al. (2004). A physical and functional map of the human TNF-alpha/NF-kappa B signal transduction pathway. Nat. Cell Biol. 6, 97105. Brouha, B., Schustak, J., Badge, R.M., Lutz-Prigge, S., Farley, A.H., Moran, J.V., and Kazazian, H.H., Jr. (2003). Hot L1s account for the bulk of retrotransposition in the human population. Proc. Natl. Acad. Sci. USA 100, 52805285. Burns, K.H., and Boeke, J.D. (2012). Human transposon tectonics. Cell 149, 740752. Cordaux, R., and Batzer, M.A. (2009). The impact of retrotransposons on human genome evolution. Nat. Rev. Genet. 10, 691703. Cost, G.J., Feng, Q., Jacquier, A., and Boeke, J.D. (2002). Human L1 element target-primed reverse transcription in vitro. EMBO J. 21, 58995910. Coufal, N.G., Garcia-Perez, J.L., Peng, G.E., Yeo, G.W., Mu, Y., Lovci, M.T., Morell, M., OShea, K.S., Moran, J.V., and Gage, F.H. (2009). L1 retrotransposition in human neural progenitor cells. Nature 460, 11271131. Crooks, G.E., Hon, G., Chandonia, J.M., and Brenner, S.E. (2004). WebLogo: a sequence logo generator. Genome Res. 14, 11881190. Esnault, C., Heidmann, O., Delebecque, F., Dewannieux, M., Ribet, D., Hance, A.J., Heidmann, T., and Schwartz, O. (2005). APOBEC3G cytidine deaminase inhibits retrotransposition of endogenous retroviruses. Nature 433, 430433. Evrony, G.D., Cai, X., Lee, E., Hills, L.B., Elhosary, P.C., Lehmann, H.S., Parker, J.J., Atabay, K.D., Gilmore, E.C., Poduri, A., et al. (2012). Single-neuron sequencing analysis of L1 retrotransposition and somatic mutation in the human brain. Cell 151, 483496. Ewing, A.D., and Kazazian, H.H., Jr. (2010). High-throughput sequencing reveals extensive variation in human-specic L1 content in individual human genomes. Genome Res. 20, 12621270. Faulkner, G.J. (2011). Retrotransposons: mobile and mutagenic from conception to death. FEBS Lett. 585, 15891594. Faulkner, G.J., Kimura, Y., Daub, C.O., Wani, S., Plessy, C., Irvine, K.M., Schroder, K., Cloonan, N., Steptoe, A.L., Lassmann, T., et al. (2009). The regulated retrotransposon transcriptome of mammalian cells. Nat. Genet. 41, 563571. Fornace, A.J., Jr., and Mitchell, J.B. (1986). Induction of B2 RNA polymerase III transcription by heat shock: enrichment for heat shock induced sequences in rodent cells by hybridization subtraction. Nucleic Acids Res. 14, 5793 5811. Fukuyama, R., Niculaita, R., Ng, K.P., Obusez, E., Sanchez, J., Kalady, M., Aung, P.P., Casey, G., and Sizemore, N. (2008). Mutated in colorectal cancer, a putative tumor suppressor for serrated colorectal cancer, selectively represses beta-catenin-dependent transcription. Oncogene 27, 60446055. Futreal, P.A., Coin, L., Marshall, M., Down, T., Hubbard, T., Wooster, R., Rahman, N., and Stratton, M.R. (2004). A census of human cancer genes. Nat. Rev. Cancer 4, 177183.
Garcia-Perez, J.L., Marchetto, M.C., Muotri, A.R., Coufal, N.G., Gage, F.H., OShea, K.S., and Moran, J.V. (2007). LINE-1 retrotransposition in human embryonic stem cells. Hum. Mol. Genet. 16, 15691577. Gilbert, N., Lutz-Prigge, S., and Moran, J.V. (2002). Genomic deletions created upon LINE-1 retrotransposition. Cell 110, 315325. Groden, J., Thliveris, A., Samowitz, W., Carlson, M., Gelbert, L., Albertsen, H., Joslyn, G., Stevens, J., Spirio, L., Robertson, M., et al. (1991). Identication and characterization of the familial adenomatous polyposis coli gene. Cell 66, 589600. Guichard, C., Amaddeo, G., Imbeaud, S., Ladeiro, Y., Pelletier, L., Maad, I.B., Calderaro, J., Bioulac-Sage, P., Letexier, M., Degos, F., et al. (2012). Integrated analysis of somatic mutations and focal copy-number changes identies key genes and pathways in hepatocellular carcinoma. Nat. Genet. 44, 694698. Han, J.S., Szak, S.T., and Boeke, J.D. (2004). Transcriptional disruption by the L1 retrotransposon and implications for mammalian transcriptomes. Nature 429, 268274. Hancks, D.C., and Kazazian, H.H., Jr. (2012). Active human retrotransposons: variation and disease. Curr. Opin. Genet. Dev. 22, 191203. Higgins, M.E., Claremont, M., Major, J.E., Sander, C., and Lash, A.E. (2007). CancerGenes: a gene selection resource for cancer genome projects. Nucleic Acids Res. 35(Database issue), D721D726. Huang, C.R., Schneider, A.M., Lu, Y., Niranjan, T., Shen, P., Robinson, M.A., Steranka, J.P., Valle, D., Civin, C.I., Wang, T., et al. (2010). Mobile interspersed repeats are major structural variants in the human genome. Cell 141, 1171 1182. Iskow, R.C., McCabe, M.T., Mills, R.E., Torene, S., Pittard, W.S., Neuwald, A.F., Van Meir, E.G., Vertino, P.M., and Devine, S.E. (2010). Natural mutagenesis of human genomes by endogenous retrotransposons. Cell 141, 1253 1261. Jandrig, B., Seitz, S., Hinzmann, B., Arnold, W., Micheel, B., Koelble, K., Siebert, R., Schwartz, A., Ruecker, K., Schlag, P.M., et al. (2004). ST18 is a breast cancer tumor suppressor gene at human chromosome 8q11.2. Oncogene 23, 92959302. Jemal, A., Bray, F., Center, M.M., Ferlay, J., Ward, E., and Forman, D. (2011). Global cancer statistics. CA Cancer J. Clin. 61, 6990. Jiang, Y., Yu, V.C., Buchholz, F., OConnell, S., Rhodes, S.J., Candeloro, C., Xia, Y.R., Lusis, A.J., and Rosenfeld, M.G. (1996). A novel family of Cys-Cys, His-Cys zinc nger transcription factors expressed in developing nervous system and pituitary gland. J. Biol. Chem. 271, 1072310730. Johnson, P.F., Landschulz, W.H., Graves, B.J., and McKnight, S.L. (1987). Identication of a rat liver nuclear protein that binds to the enhancer core element of three animal viruses. Genes Dev. 1, 133146. Jurka, J. (1997). Sequence patterns indicate an enzymatic involvement in integration of mammalian retroposons. Proc. Natl. Acad. Sci. USA 94, 18721877. Kano, H., Godoy, I., Courtney, C., Vetter, M.R., Gerton, G.L., Ostertag, E.M., and Kazazian, H.H., Jr. (2009). L1 retrotransposition occurs mainly in embryogenesis and creates somatic mosaicism. Genes Dev. 23, 13031312. Kinzler, K.W., Nilbert, M.C., Vogelstein, B., Bryan, T.M., Levy, D.B., Smith, K.J., Preisinger, A.C., Hamilton, S.R., Hedge, P., Markham, A., et al. (1991). Identication of a gene located at chromosome 5q21 that is mutated in colorectal cancers. Science 251, 13661370. Kohonen-Corish, M.R., Sigglekow, N.D., Susanto, J., Chapuis, P.H., Bokey, E.L., Dent, O.F., Chan, C., Lin, B.P., Seng, T.J., Laird, P.W., et al. (2007). Promoter methylation of the mutated in colorectal cancer gene is a frequent early event in colorectal cancer. Oncogene 26, 44354441. Lander, E.S., Linton, L.M., Birren, B., Nusbaum, C., Zody, M.C., Baldwin, J., Devon, K., Dewar, K., Doyle, M., FitzHugh, W., et al.; International Human Genome Sequencing Consortium. (2001). Initial sequencing and analysis of the human genome. Nature 409, 860921. Lee, E., Iskow, R., Yang, L., Gokcumen, O., Haseley, P., Luquette, L.J., 3rd, Lohr, J.G., Harris, C.C., Ding, L., Wilson, R.K., et al.; Cancer Genome Atlas
Research Network. (2012). Landscape of somatic retrotransposition in human cancers. Science 337, 967971. Levin, H.L., and Moran, J.V. (2011). Dynamic interactions between transposable elements and their hosts. Nat. Rev. Genet. 12, 615627. Liu, J., Nau, M.M., Zucman-Rossi, J., Powell, J.I., Allegra, C.J., and Wright, J.J. (1997). LINE-I element insertion at the t(11;22) translocation breakpoint of a desmoplastic small round cell tumor. Genes Chromosomes Cancer 18, 232239. Llovet, J.M., Ricci, S., Mazzaferro, V., Hilgard, P., Gane, E., Blanc, J.F., de Oliveira, A.C., Santoro, A., Raoul, J.L., Forner, A., et al.; SHARP Investigators Study Group. (2008). Sorafenib in advanced hepatocellular carcinoma. N. Engl. J. Med. 359, 378390. Matsumine, A., Senda, T., Baeg, G.H., Roy, B.C., Nakamura, Y., Noda, M., Toyoshima, K., and Akiyama, T. (1996). MCC, a cytoplasmic protein that blocks cell cycle progression from the G0/G1 to S phase. J. Biol. Chem. 271, 1034110346. Miki, Y., Nishisho, I., Horii, A., Miyoshi, Y., Utsunomiya, J., Kinzler, K.W., Vogelstein, B., and Nakamura, Y. (1992). Disruption of the APC gene by a retrotransposal insertion of L1 sequence in a colon cancer. Cancer Res. 52, 643645. Mills, R.E., Bennett, E.A., Iskow, R.C., and Devine, S.E. (2007). Which transposable elements are active in the human genome? Trends Genet. 23, 183191. Moran, J.V., Holmes, S.E., Naas, T.P., DeBerardinis, R.J., Boeke, J.D., and Kazazian, H.H., Jr. (1996). High frequency retrotransposition in cultured mammalian cells. Cell 87, 917927. Nhieu, J.T., Renard, C.A., Wei, Y., Cherqui, D., Zafrani, E.S., and Buendia, M.A. (1999). Nuclear accumulation of mutated beta-catenin in hepatocellular carcinoma is associated with increased cell proliferation. Am. J. Pathol. 155, 703710. Ostertag, E.M., and Kazazian, H.H., Jr. (2001a). Biology of mammalian L1 retrotransposons. Annu. Rev. Genet. 35, 501538. Ostertag, E.M., and Kazazian, H.H., Jr. (2001b). Twin priming: a proposed mechanism for the creation of inversions in L1 retrotransposition. Genome Res. 11, 20592065. Perz, J.F., Armstrong, G.L., Farrington, L.A., Hutin, Y.J., and Bell, B.P. (2006). The contributions of hepatitis B virus and hepatitis C virus infections to cirrhosis and primary liver cancer worldwide. J. Hepatol. 45, 529538. Powell, S.M., Zilz, N., Beazer-Barclay, Y., Bryan, T.M., Hamilton, S.R., Thibodeau, S.N., Vogelstein, B., and Kinzler, K.W. (1992). APC mutations occur early during colorectal tumorigenesis. Nature 359, 235237. Rhodes, S.J., Chen, R., DiMattia, G.E., Scully, K.M., Kalla, K.A., Lin, S.C., Yu, V.C., and Rosenfeld, M.G. (1993). A tissue-specic enhancer confers Pit-1dependent morphogen inducibility and autoregulation on the pit-1 gene. Genes Dev. 7, 913932.
Rowland, B.D., Bernards, R., and Peeper, D.S. (2005). The KLF4 tumour suppressor is a transcriptional repressor of p53 that acts as a context-dependent oncogene. Nat. Cell Biol. 7, 10741082. Senda, T., Matsumine, A., Yanai, H., and Akiyama, T. (1999). Localization of MCC (mutated in colorectal cancer) in various tissues of mice and its involvement in cell differentiation. J. Histochem. Cytochem. 47, 11491158. Solyom, S., Ewing, A.D., Rahrmann, E.P., Doucet, T., Nelson, H.H., Burns, M.B., Harris, R.S., Sigmon, D.F., Casella, A., Erlanger, B., et al. (2012). Extensive somatic L1 retrotransposition in colorectal tumors. Genome Res. 22, 23282338. Starr, T.K., Allaei, R., Silverstein, K.A., Staggs, R.A., Sarver, A.L., Bergemann, T.L., Gupta, M., OSullivan, M.G., Matise, I., Dupuy, A.J., et al. (2009). A transposon-based genetic screen in mice identies genes altered in colorectal cancer. Science 323, 17471750. Steinbach, D., Schramm, A., Eggert, A., Onda, M., Dawczynski, K., Rump, A., Pastan, I., Wittig, S., Pfaffendorf, N., Voigt, A., et al. (2006). Identication of a set of seven genes for the monitoring of minimal residual disease in pediatric acute myeloid leukemia. Clin. Cancer Res. 12, 24342441. Swergold, G.D. (1990). Identication, characterization, and cell specicity of a human LINE-1 promoter. Mol. Cell. Biol. 10, 67186729. Tateishi, R., and Omata, M. (2012). Hepatocellular carcinoma in 2011: genomics in hepatocellular carcinomaa big step forward. Nat. Rev. Gastroenterol. Hepatol. 9, 6970. Thurman, R.E., Rynes, E., Humbert, R., Vierstra, J., Maurano, M.T., Haugen, E., Shefeld, N.C., Stergachis, A.B., Wang, H., Vernot, B., et al. (2012). The accessible chromatin landscape of the human genome. Nature 489, 7582. Totoki, Y., Tatsuno, K., Yamamoto, S., Arai, Y., Hosoda, F., Ishikawa, S., Tsutsumi, S., Sonoda, K., Totsuka, H., Shirakihara, T., et al. (2011). High-resolution characterization of a hepatocellular carcinoma genome. Nat. Genet. 43, 464469. Turelli, P., Mangeat, B., Jost, S., Vianin, S., and Trono, D. (2004). Inhibition of hepatitis B virus replication by APOBEC3G. Science 303, 1829. Unitt, E., Rushbrook, S.M., Marshall, A., Davies, S., Gibbs, P., Morris, L.S., Coleman, N., and Alexander, G.J. (2005). Compromised lymphocytes inltrate hepatocellular carcinoma: the role of T-regulatory cells. Hepatology 41, 722730. Wang, J., Song, L., Grover, D., Azrak, S., Batzer, M.A., and Liang, P. (2006). dbRIP: a highly integrated database of retrotransposon insertion polymorphisms in humans. Hum. Mutat. 27, 323329. oz-Lopez, M., Macia, A., Yang, Z., Montano, M., Collins, W., Wissing, S., Mun Garcia-Perez, J.L., Moran, J.V., and Greene, W.C. (2012). Reprogramming somatic cells into iPS cells activates LINE-1 retroelement mobility. Hum. Mol. Genet. 21, 208218. Yee, K.S., and Yu, V.C. (1998). Isolation and characterization of a novel member of the neural zinc nger factor/myelin transcription factor family with transcriptional repression activity. J. Biol. Chem. 273, 53665374.
The Lipid Mediator Protectin D1 Inhibits Inuenza Virus Replication and Improves Severe Inuenza
Masayuki Morita,1,14 Keiji Kuba,1,14 Akihiko Ichikawa,1 Mizuho Nakayama,1 Jun Katahira,2 Ryo Iwamoto,4,5 Tokiko Watanebe,6,8,10 Saori Sakabe,6,10 Tomo Daidoji,11 Shota Nakamura,3 Ayumi Kadowaki,1 Takayo Ohto,1 Hiroki Nakanishi,7 Ryo Taguchi,7 Takaaki Nakaya,11 Makoto Murakami,12 Yoshihiro Yoneda,2 Hiroyuki Arai,4 Yoshihiro Kawaoka,6,8,10 Josef M. Penninger,13 Makoto Arita,4,9 and Yumiko Imai1,*
of Biological Informatics and Experimental Therapeutics, Graduate School of Medicine, Akita University, Akita 010-8543, Japan Networks Laboratories, Biomolecular Dynamics Laboratory, Graduate School of Frontier Biosciences 3Department of Infection Metagenomics, Research Institute for Microbial Diseases Osaka University, Osaka 565-0871, Japan 4Department of Health Chemistry 5Business-Academia-Collaborative Laboratory, Graduate School of Pharmaceutical Science 6Division of Virology, Department of Microbiology and Immunology and International Research Center for Infectious Diseases, Institute of Medical Science 7Department of Metabolome, Graduate School of Medicine University of Tokyo, Tokyo 113-0033, Japan 8ERATO Infection-induced Host Responses Project 9PRESTO Japan Science and Technology Agency, Saitama 332-0012, Japan 10Inuenza Research Institute, University of Wisconsin-Madison, Madison, WI 53711, USA 11Department of Infectious Diseases, Kyoto Prefectural University of Medicine, Kyoto 602-8566, Japan 12Biomembrane Signaling Project, Tokyo Metropolitan Institute of Medical Science, Tokyo 156-8506, Japan 13IMBA (Institute of Molecular Biotechnology in the Austrian Academy of Sciences), Vienna 1030, Austria 14These authors contributed equally to this work and are co-rst authors *Correspondence: imai@med.akita-u.ac.jp http://dx.doi.org/10.1016/j.cell.2013.02.027
2Biomolecular 1Department
SUMMARY
INTRODUCTION Despite immunization programs, inuenza A viruses are a major cause of morbidity and mortality throughout the world, especially among individuals with high risk factors, such as diabetes, asthma, or pregnancy (Clark and Lynch, 2011). Highly pathogenic avian H5N1 inuenza viruses continue to circulate with mortality rates of up to 60% of infected patients (Beigel et al., 2005). Inuenza strains have also recently emerged that exhibit resistance to antiviral drugs such as oseltamivir (Shankaran and Bearman, 2012). Recently, a virus with its hemagglutinin (HA) derived from H5N1 was identied that is transmissible in ferrets, further supporting that H5N1 viruses have pandemic potential (Imai and Kawaoka, 2012). Moreover, currently used antiviral drugs, though benecial if administered early in the disease (Kumar, 2011), are not effective in critically ill patients after inuenza virus infections (Beigel et al., 2005; Clark and Lynch, 2011; Kumar et al., 2010). Inuenza A viruses are negative-sense RNA viruses that exploit the host cellular machinery for multiple phases of their life cycle (Neumann et al., 2004). Recent genome-wide RNAi screens have identied several host genes and molecular networks crucial for inuenza virus replication (Karlas et al., nig et al., 2010; Shapira et al., 2009). 2010; Hao et al., 2008; Ko
Inuenza A viruses are a major cause of mortality. Given the potential for future lethal pandemics, effective drugs are needed for the treatment of severe inuenza such as that caused by H5N1 viruses. Using mediator lipidomics and bioactive lipid screen, we report that the omega-3 polyunsaturated fatty acid (PUFA)-derived lipid mediator protectin D1 (PD1) markedly attenuated inuenza virus replication via RNA export machinery. Production of PD1 was suppressed during severe inuenza and PD1 levels inversely correlated with the pathogenicity of H5N1 viruses. Suppression of PD1 was genetically mapped to 12/15-lipoxygenase activity. Importantly, PD1 treatment improved the survival and pathology of severe inuenza in mice, even under conditions where known antiviral drugs fail to protect from death. These results identify the endogenous lipid mediator PD1 as an innate suppressor of inuenza virus replication that protects against lethal inuenza virus infection.
However, whether bioactive lipid mediators and lipid metabolic pathways might be involved in inuenza virus infections is unknown. Omega-3 polyunsaturated fatty acids (PUFA), such as docosahexaenoic acid (DHA) and eicosapentaenoic acid (EPA), both of which are enriched in sh oils, serve as substrates for the production of potent bioactive anti-inammatory lipid mediators, such as resolvins, protectins, maresins, and their biosynthetic intermediates (Serhan, 2007; Serhan et al., 2012). Here, we report that the DHA-derived protectin D1 isomer (PD1; 10S, 17S-dihydroxydocosahexaenoic acid) markedly attenuates inuenza virus replication via interference with the virus RNA nuclear export machinery. PD1 was identied in self-limited resolving inammatory exudates in vivo (Serhan et al., 2002), where it regulates the innate local response and stimulates resolution. Production of PD1 was downregulated during severe inuenza virus infections, e.g., with H5N1 avian inuenza viruses, and inversely correlated with the pathogenicity of different H5N1 virus isolates. Importantly, PD1 treatment improved the survival of inuenza-virus-infected mice even at later stages of the disease, when current antiviral therapies are not effective (Arya et al., 2010), suggesting that PD1 could be a novel therapeutic target for the treatment of severe inuenza virus infections. RESULTS The PUFA-Derived Mediator Protectin D1 Reduces Inuenza Virus Replication To investigate the potential role of lipid metabolites in inuenza infections, we performed a screen of PUFA-derived lipids in human lung epithelial (A549) cells infected with the inuenza A virus stain A/Puerto Rico/8/34 (H1N1) (PR8 virus) by using bioactive lipid libraries that include prostaglandins (PGs), resolvins, protectins, and other PUFA-derived metabolites (Figure 1A). PR8-infected cells (multiplicity of infection [MOI]: 0.2) were treated with PUFA-derived metabolites (1 mM each) and inuenza virus nucleoprotein (NP) mRNA expression was quantied at 8 hr after infection (Figure 1A). Compared with vehicle treatment, the arachidonic acid (AA)-derived products 12-HETE and 15-HETE, as well as the DHA-derived 17HDoHE and PD1, inhibited the expression of NP mRNA by more than 30%. Of note, some compounds, e.g., 8,9-DHET, resulted in increased NP mRNA levels that required further analysis; here, we focus on compounds that can inhibit virus replication because they may represent potential therapeutic targets for inuenza virus infection. Immunohistochemistry demonstrated that PD1, 17-HDoHE, 12-HETE, and 15-HETE administration markedly decreased the numbers of cells that were positive for the inuenza virus NP protein and viral antigen R309 (Figure 1B). In addition, PR8 virus titers were markedly decreased by the treatment with 12-HETE, 15-HETE, 17HDoHE, and PD1 (Figure 1C). Consistent with the results of our screen, the DHA-derived resolvins RvD1 and RvD2 showed no apparent inhibition (Figures 1B and 1C). Because PD1 treatment among all bioactive mediators tested led to the most potent inhibition of PR8 virus replication, we focused on PD1 for our subsequent studies. We conrmed that PD1 treatment attenuated inuenza virus replication, as assessed by NP protein production in PR8-virus-infected A549
cells (Figure S1A available online). NP mRNA expression was inhibited by PD1 treatment in a dose-dependent fashion (Figure S1B). Importantly, PD1 treatment also attenuated the replication of highly pathogenic H5N1 inuenza viruses, as assessed by markedly reduced viral M protein mRNA expression (Figures 1D and S1C), and decreased virus titers (Figures 1E and S1D) in A549 and MDCK cells. Thus, our screen for PUFAderived metabolites identied PD1 as a suppressor of inuenza virus replication in cultured cells. Reduction of Endogenous PD1 Production in Lungs of Mice in Severe Inuenza Virus Infection To examine the metabolic proles of PUFA-derived metabolites in severe inuenza infection in vivo, we conducted a mediator lipidomics for AA-, DHA- and EPA-derived metabolites in lung tissues obtained from wild-type mice subjected to intratracheal PR8-virus infections (Ichikawa et al., 2013). All PR8-infected animals died within 8 days (Figure S2A). NP mRNA expression (Figure S2B) and virus titers (Figure S2C) rapidly increased after infection peaking on day 2. PR8 infections also resulted in markedly impaired pulmonary functions as assessed by tissue resistance and tissue elastance (Figure S2D) (Sly et al., 2003). Histologically, numbers of inammatory cells were slightly increased in the PR8-infected lungs on day 2, and on day 5 we observed pulmonary edema, hemorrhage, and hyaline membrane formation (Figure S2E). Thus, in line with previous reports (Smith et al., 2011; Ichikawa et al., 2013), intratracheal infection with PR8 inuenza viruses induces lung pathologies that closely mimic the symptoms of severe inuenza in humans. To identify the endogenous lipid mediators and pathway metabolites during PR8 infections, lung tissues were obtained from PR8- and mock-infected mice at 6, 12, 24, and 48 hr after infection and subjected to liquid chromatography-tandem mass spectrometry (LC-MS/MS)-based lipidomics analysis. This technology has been recently developed and allows researchers to identify and quantify more than 250 PUFAderived endogenous lipid mediators and pathway metabolites including prostaglandins (PGs), leukotrienes (LTs), lipoxins (LXs), resolvins, protectins, and other AA-, EPA-, DHA-derived products (Arita, 2012). As shown in Tables S1, S2, and S3, several lipid mediators were detectable in noninfected lungs under baseline control conditions. In particular, cyclooxygenase (COX) pathway products such as PGE2 and 6-keto PGF1a and the 12/15-lipoxygenase (12/15-LOX) pathway metabolites 12HETE, 15-HETE, and 14-HDoHE were present in the lung tissues of untreated control mice (Tables S1 and S2). Next, we examined the relative ratios of AA-derived (Figure 2A), DHA-derived (Figure 2B), and EPA-derived (Figure 2C) metabolites in PR8-infected versus control lungs. Intriguingly, among all the identied mediators and metabolites, endogenous production of AAderived 12-HETE and 15-HETE, as well as DHA-derived 17HDoHE and PD1, showed the greatest level of reduction during the course of the PR8 inuenza virus infection (Figures 2A2C); these are the same series of lipid mediators that showed the largest impact on inuenza replication in A549 cells (Figure 1). Akin to its effects on viral replication in cell culture, PD1 was one of the most reduced lipid mediator in the lungs of PR8-infected mice (Figure 2B).
Figure 1. Identication of PUFA-Derived Products Involved in Inuenza Virus Replication

(A) Lipid library screening for inuenza replication. A549 cells were infected with inuenza PR8 virus (MOI: 0.2) and treated with PUFA-derived compound (1 mM) or vehicle immediately after infection. Virus NP mRNA expression was analyzed by calculating the relative ratio of the indicated AA-, DHA-, and EPA-derived compounds to vehicle treatment. Averages from three independent screens shown. (B) Immunocytochemistry of inuenza virus nucleoprotein (NP) and virus antigen R309. PR8-virus-infected A549 cells (MOI: 0.2) were treated with the indicated compounds (1 mM each) or vehicle. Infected cells were xed after a 24 hr postinfection and stained for NP (green) and virus antigen R309 (red). Nuclei (blue) were counterstained with DAPI.
To extend our results to other inuenza virus strains, we intranasally (i.n.) infected wild-type mice with 2009 H1N1 virus, avian H5N1 (H5N1) virus, avian H5N1 virus carrying an attenuating Glu mutation at position 627 of PB2 (H5N1 PB2-627E), or mock infected them. Virus titers were higher in the lungs of mice infected with the avian H5N1 virus as compared to 2009 H1N1 or H5N1PB2-627E-virus-infected mice (Figure S2F), conrming previous results (Shinya et al., 2004). Lung tissues were obtained at 24 and 48 hr after infection and subjected to LC-MS/MS-based lipidomics analysis. Again, endogenous production of AA-derived 12-HETE and 15-HETE, and of DHA-derived 17-HDoHE and PD1, was markedly downregulated in lungs infected with the highly pathogenic avian H5N1 virus at both time points analyzed (Figure 3). PD1 production was the most suppressed in H5N1virus-infected lungs, whereas it was even elevated in lungs infected with the less pathogenic 2009 H1N1 virus and H5N1 PB2-627E virus strains (Figure 3), suggesting that endogenous PD1 levels inversely correlate with the pathogenicity of inuenza virus strains. Taken together, our data show that endogenous production of PD1 is markedly suppressed in the lungs of mice intratracheally infected with PR8 virus or nasally infected with the highly pathogenic avian H5N1 inuenza virus. PD1 Protects against Severe Inuenza In Vivo We next tested whether the PUFA-derived 12-HETE, 15-HETE, 17-HDoHE, and PD1, all of which attenuated virus replication in cultured cells (Figure 1), could also protect against severe inuenza in mice. Mice were infected intratracheally with PR8 virus (1,500 TCID50); lipids mediators (1 mg/mouse) or vehicle were given intravenously (i.v.) 12 hr before and immediately after the PR8 virus infection. The lipids metabolites were given intravenously because they are autacoids and, as such, are rapidly metabolized and inactivated in vivo. Consistent with our in vitro data, PD1 treatment improved the survival of PR8-infected mice (Figure 4A). To extend our results to a human pathogenic virus isolate, we tested whether PD1 treatment could also affect the survival of mice infected with the 2009 H1N1 virus. When PD1 was given i.v. 12 hr before and immediately after 2009 H1N1 virus infections, the survival of the virus-infected mice improved (Figure 4B). We then tested whether the protective actions of PD1 resulted from its anti-inammatory functions in macrophages and neutrophils (Schwab et al., 2007). However, macrophage and neutrophil numbers in bronchioalveolar lavage (BAL) uid were seemingly unaffected by PD1 treatment in the lungs of PR8 inuenza-virus-infected mice (Figure S3A), although the PD1-treated lungs did show slightly improved histological changes at 2 days after infection (Figure S3B). Furthermore, PD1 treatment did not signicantly decrease the levels of the proinammatory cytokines IL-6, IP-10, and CXCL2 nor mRNA expression (IL-6, IFN-b, CXCL1, and CXCL2) in the lungs on days 2 and 5 after PR8 infection (Figures 4C and S3C). In contrast, virus NP mRNA expression (Figure 4D), virus titers (Fig-
ure 4E), and the number of virus antigen-positive cells (Figure 4F) were all signicantly reduced in the lungs of PD1-treated mice, indicating that PD1 acts on viral replication. Similar to our in vitro data (Figure 1), intravenous treatment (1 mg/mouse) with RvD1 or RvD2 at 12 hr before and immediately after PR8 virus infection had no apparent effect on the survival of PR8-virus-infected mice (Figure 4G). Because LXA4 showed a mild inhibition of virus replication (Figure 1) and LXA4 production was suppressed during the course of severe inuenza (Figure 2A), we also tested the effect of LXA4 on the survival of mice infected with PR8 inuenza virus in vivo. However, treatment (1 mg/mouse, i.v.) with LXA4 also had no apparent effect on the survival of PR8-virus-infected mice (Figure 4G). Given that RvD1, RvD2, and LXA4 are known to have anti-inammatory properties (Serhan and Levy, 2003), we next asked whether they could decrease the inammatory cytokine production initiated in response to virus infection. However, RvD1, RvD2, or LXA4 treatment did not decrease the proinammatory cytokine levels (IL-6, IP-10, CXCL2) in the lungs on days 2 and 5 after PR8 infection (Figure S3D). In addition, virus replication, as assessed by NP mRNA, was apparently not affected following treatment with these lipid metabolites (Figure 4H). We nally examined whether PD1 treatment has a therapeutic benet after the infection was established. Mice were infected i.t. with PR8 virus and treated with PD1 (100 ng/mouse; i.t.) or vehicle 2 days after the initial infection; infected mice were then mechanically ventilated, mimicking the therapeutic setting for critically ill inuenza patients in intensive care units. We chose the 48 hr time point because currently available anti-inuenza drugs are thought to be benecial if administered early after infection but ineffective if administered around 48 hr after infection (Kumar, 2011). Intriguingly, such therapeutic PD1 treatment signicantly improved the pulmonary functions of PR8-infected mice as assessed by tissue resistance (Figure 4I) and tissue elastance (Figure 4J). A similar administration regiment using RvD1 or RvD2 did not improve pulmonary functions (Figures S3E and S3F). Importantly, when mice were treated with the antiviral drug peramivir (Arya et al., 2010) or PD1 alone at 48 hr postinfection, 70%75% of the animals still died within 16 days (Figure 4K). However, PD1 plus peramivir treatment beginning 48 hr after infection completely rescued the mice from death due to the inuenza infection (Figure 4K). Of note, if treated with peramivir immediately after infection, all animals survived (Figure S3G). These results show that PD1 has a marked benecial effect on severe inuenza virus infections both prophylactically and, most importantly, therapeutically in vivo. PD1 Inhibits Nuclear Export of Viral Transcripts Next, we explore the mechanisms by which PD1 inhibits inuenza virus replication. Consistent with our in vivo data (Figures 4C, and S3C), expression of antiviral (IFNb) or antiinammatory (IL-8) cytokines was unaffected by PD1 treatment
(C) Virus titers. PR8-virus-infected A549 cells (MOI: 5) were treated with the indicated compounds (8 mM each) or vehicle. Infectious virus particles were quantied at 24 hr after infection by focus forming unit (FFU) replication assay. **p < 0.01, *p < 0.05 compared with vehicle. (D and E) Effects of PD1 on H5N1 virus replication. H5N1 inuenza-virus-infected A549 cells (MOI: 0.02, 0.2) were treated with PD1 (8 mM) or vehicle immediately after infection. Viral M protein mRNA expression (D) and virus titers (E) were quantied at 24 hr after infection by qRT-PCR and FFU assay. **p < 0.01 compared with vehicle. Data in (C)(E) are presented as means S.E.M. See also Figure S1.
A
1.5
TXB2
Flu/PBS Flu/PBS
8-HETE
9-HETE
1.5
11-HETE
1.5
19-HETE
1.5
Flu/PBS
1.0 0.5
TXA2
Flu/PBS
1.5 1.0 0.5
Flu/PBS
1.0 0.5
1.0 0.5
1.0 0.5
Flu/PBS Flu/PBS
1.5
AA
1.5
1.0 0.5
15-deoxy PGJ2
1.5
PGD2
1.5
PGH2
COX
16-HETE
Flu/PBS
Flu/PBS
Flu/PBS
1.0 0.5
1.0 0.5
Flu/PBS
AA
5-LOX 12/15-LOX
2.5 2.0 1.5 1.0 0.5
17-HETE n.d.
18-HETE
1.5 1.0 0.5
20-HETE
1.5 1.0 0.5
15-keto PGE2
1.5
1.5
PGE2
Flu/PBS
Flu/PBS
5-HETE
2.5 2.0 1.5 1.0 0.5
12-HETE
Flu/PBS Flu/PBS
1.5 1.0 0.5
15-HETE
1.5 1.0 0.5
15-oxo-ETE
3.0 2.5 2.0 1.5 1.0
1.0 0.5
1.0 0.5
Flu/PBS
PGF2a
1.5
5-LOX
LXA4
1.5
Flu/PBS
1.0 0.5
5-oxo-ETE
12-oxo-ETE
2.5 2.0 1.5 1.0 0.5
PGI2
6-keto-PGF1a
1.5
2.5 2.0 1.5 1.0 0.5 0.0
Flu/PBS
Flu/PBS
Flu/PBS
Flu/PBS
1.0 0.5
Flu/PBS
1.0 0.5
5-HEPE
Flu/PBS
C B
1.5 2.5
1.0 0.5
TXB3
1.5
TXA3
EPA
1.5
LTB5 n.d. LTA5
Flu/PBS
DHA
Flu/PBS
Flu/PBS
4-HDoHE
Flu/PBS
2.0 1.5 1.0 0.5
11-HDoHE
1.0
1.5 1.0 0.5
13-HDoHE
1.5
1.0
Flu/PBS
1.0 0.5
Flu/PBS
1.0
COX
0.5
0.5
PGD3
1.5
PGH3 COX
5-LOX
0.5
DHA
Flu/PBS
7-HDoHE
2.5
Flu/PBS
16-HDoHE
1.5
EPA
18-HEPE
2.0
1.0 0.5
2.0 1.5 1.0 0.5
Flu/PBS
12/15-LOX
1.0
12/15-LOX
Flu/PBS
0.5
PGE3
1.5
8-HEPE
Flu/PBS
1.5 1.0 0.5
14-HDoHE
Flu/PBS
1.5
1.5
Flu/PBS
17-HDoHE
Flu/PBS
1.0
Flu/PBS
1.5 1.0 0.5
Flu/PBS
Flu/PBS
8-HDoHE n.d.
10-HDoHE
1.5
20-HDoHE
2.0
12-HEPE
1.5 1.0 0.5
15-HEPE
1.5 1.0 0.5
1.5 1.0 0.5
1.0 0.5
1.0
Flu/PBS
0.5 0.5
PGF3a
1.5 1.0
9-HEPE
1.5
RvE1 n.d. RvE2 n.d.
RvD5 Maresin 1 (7,17-diHDoHE) (7,14-diHDoHE) n.d. n.d. 4,14-diHDoHE n.d.

PD1
1.5
Flu/PBS
RvD1 n.d. RvD2 n.d.
0.5
1.5
PGI3
0.5
1.0
LXA5 n.d.
Flu/PBS Flu/PBS
Flu/PBS
21-HDoHE
Flu/PBS
1.0
1.0 0.5
0.5
11-HEPE 6-keto-PGF1a-17delta
1.5 1.5 1.0 0.5
Flu/PBS
Flu/PBS
1.0
1.0 0.5
0.5
19-HEPE n.d.
20-HEPE
2.5 2.0 1.5 1.0 0.5
Figure 2. Mediator Lipidomics in Severe Inuenza in Mice

(AC) Lung tissues were obtained from intratracheally PR8 virus (1 3 104 TCID50)- or mock (PBS)-infected mice at 0, 6, 12, 24, and 48 hr after infection and subjected to LC-MS/MS lipidomics analysis. Changes in the production of PR8-virus-infected lungs (Flu) relative to PBS-treated controls (PBS) are shown for (A) AA-, (B) DHA-, and (C) EPA-derived metabolites at 0, 6, 12, 24, and 48 hr after infection. n = 5 per each group. See also Figure S2, Tables S1, S2, and S3.
of inuenza-virus-infected A549 cells (Figure S4A), suggesting that PD1 is unlikely to attenuate inuenza virus replication through its possible antiviral or anti-inammatory responses.
Thus, we wondered whether PD1 could affect inuenza virus replication via controlling virus life cycle. The inuenza viral genome consists of negative-sense RNA (vRNA) packaged in
2009 H1N1
24 h 48 h
H5N1
24 h 48 h
H5N1 PB2-627E
24 h 48 h PGE2 LXA4 LXB4 PGD2 15-deoxy PGJ2 PGF2a 6-keto PGF1a TX B2 LTB4 5-HETE 8-HETE 12-HETE 15-HETE 4-HDoHE 7-HDoHE 10-HDoHE 13-HDoHE 17-HDoHE 20-HDoHE RvD5 (7,17-diHDoHE) PD1 RvD1 RvD2 14-HDoHE Maresin 1 (7,14-diHDoHE) 5-HEPE 8-HEPE 12-HEPE 15-HEPE 18-HEPE LTB5 PGE3 RvE1
Figure 3. Mediator Lipidomics in H5N1 Inuenza Virus Infection in Mice

WT mice were intranasally infected with 2009 H1N1 virus (2009 H1N1), avian H5N1 virus (H5N1), H5N1 virus with an attenuating mutation of Glu at position 627 of PB2 (H5N1 PB2-627E), or mock infected with PBS as a control. Lung tissues were obtained from mice at 24 and 48 hr after infection and subjected to lipidomics analysis. Red columns indicate percent increases and blue columns percent decreases of the indicated lipid mediators in inuenza virus (2009 H1N1, H5N1, H5N1 PB2-627E)-infected lungs as compared to PBS-treated controls. Grey columns indicate that the products were not detectable. n = 3 per group.
AA derived-
viral ribonucleoprotein (vRNP) complexes. vRNA is replicated into cRNA, which serves as the template for vRNA neosynthesis. After the formation of vRNPs in the nucleus, vRNAs are exported to the cytoplasm (Neumann et al., 2004). In addition, an initial round of transcription produces 50 capped and 30 poly(A) viral mRNA that is also exported to the cytoplasm. To examine whether PD1 affects the nuclear export of inuenza virus RNA, we visualized the localization of PR8 virus segment 7 mRNA and vRNA in infected A549 cells using uorescence in situ hybridization (FISH). Upon PR8 virus infection, virus mRNA and vRNA were primarily localized in the nucleus at 5 hr after infection but shifted to the cytoplasm at 8 hr in the vehicle-treated cells (Figure 5A). PD1 treatment markedly attenuated cytoplasmic
translocation of virus mRNA and vRNA (Figure 5A). Of note, RvD1 treatment did not affect the export of viral RNAs (Figure 5A). As a consequence, virus mRNA and vRNA expression was signicantly reduced in the PD1-treated but not RvD1-treated cytoplasmic fraction 8 hr after PR8 infections (Figure 5B). Thus, our data indicate that PD1 inhibits nuclear export of inuenza virus RNAs. Genome-wide RNAi screens for host factors implicated in inuenza virus replication identied a critical role for the mRNA transporter NXF1 (also known as Tap) (Karlas et al., 2010; Hao et al., 2008), although the precise mechanisms remained unknown. We therefore assessed a potential involvement of NXF1 in the PD1-regulated nuclear export of inuenza virus RNA. Indeed, knockdown of NXF1 in A549 cells (Figure S4B), but not of the related transporters p15/NXT1 (Herold et al., 2001), NXT2, or CRM1/ XPO1 (Stade et al., 1997), signicantly attenuated PR8 inuenza virus replication as assessed by NP mRNA expression (Figure 5C), NP protein levels (Figure S4C), and virus titers (Figure S4D). Consistent with a previous report (Read and Digard, 2010), FISH analyses further conrmed that knockdown of NXF1 inhibited the nuclear export of virus mRNA and vRNA (Figure S4E). Of note, although knockdown of p15/NXT1 or its homolog NXT2 (Herold et al., 2000) alone did not attenuate virus replication, knockdown of both p15/NXT1 and NXT2 attenuated the inuenza virus replication (Figure 5C). Knockdown of CRM1/XPO1 did not affect the nuclear export of inuenza virus transcripts (Figure S4F). NXF1 forms a functional heterodimer ter et al., 1998; Herold et al., with the p15/NXT1 protein (Gru guez-Navarro and Hurt, 2011). 2000; Katahira et al., 1999; Rodr NXF1 proteins, via their middle (M) and C-terminal domains (Figure S4G), bind to phenylalanine-glycine (FG) repeat-containing nucleoporins (Nups) that line the nuclear pore channel (Ho
EPA derived-
DHA derived-
A
Percent survival
100 80 60 40 20 0 0 2 4 6 8 10 PR8 virus 1500 TCID50
Vehicle 12-HETE 15-HETE 17-HDoHE PD1
B
100
Percent survival
80 60 40 20 0
PD1 Vehicle
**
2009 H1N1 virus 1000 TCID50

0 2 4 6 8 10 12 14
12
14
16
18
Days after i.t. infection
Days after infection
CXCL2 (pg/mg protein)
1000
IL-6 (pg/mg protein)
IP-10 (pg/mg protein)
n.s. n.s.
1500
n.s.
300
n.s.
n.s.
Virus titers (FFU / g tissue)
C
800 600 400 200 0 veh PD1 veh PD1
D
3.5 3.0
10 7 10 6 10 5 10 4 10 3 10 2 10 1 10 0
Vehicle PD1
NP/18S (AU)
1000
200
2.5 2.0 1.5 1.0 0.5
**
n.s.
500
100
**
Vehicle PD1
0 veh PD1 veh PD1
0 veh PD1 veh PD1
0.0
Day 2
Day 5
Day 2 Day 5
Day 2
Day 5
F
Vehicle PD1
G
Percent survival
100 80 60 40 20 0 0 2 4 6 8 PR8 virus 1500 TCID50 Vehicle RvD1 RvD2 LXA4
H
1000 800 600 400 200 0 Vehicle RvD1 RvD2 LXA4
NP/18S(AU)
10
Day 2
Day 5
Days after infection
I
Tissue resistance (cmH 2O/mL)
40 30 20 10
J
Tissue elastance (cmH 2O/mL)
i.t. PD1 or vehicle at 2 days post infection
200 150 100 50
K
100
i.t. PD1 or vehicle at 2 days post infection
Percent survival
80 60 40 Pe ramiv ir i.v . 20 0 0 2 4 6 8
PD1 i.v.
**
Vehicle
**
Peramivir + PD1 Peramivir PD1 Vehicle

PR8 v irus 500 TCID50
Vehicle
0
PD1
0.0 0.5 1.0 1.5 2.0 2.5
PD1 0.0 0.5 1.0 1.5 2.0 2.5
Time (hrs)
Time (hrs)
10 12 14 16 18 20
Days after i.t. infection

Figure 4. Protective Role of PD1 against Inuenza Virus Infection In Vivo
(A) Survival of inuenza-virus-infected WT mice upon treatment with PUFA-derived metabolies. n = 7 for vehicle, n = 7 for 12-HETE, n = 8 for 15-HETE, n = 8 for 17-HDoHE, and n = 8 for PD1 group. **p < 0.01 compared with vehicle. (B) Survival of WT mice intratracheally infected with 2009 H1N1 virus upon treatment with PD1 (1 mg/mouse) (n = 6) or vehicle (n = 6). PD1 treatment improved the survival of virus-infected mice (p = 0.03). (CF) Effects of PD1 on host cytokine levels and viral replication. Cytokine (IL-6, IP-10, and CXCL2) levels in lungs at 2 and 5 days postinfection are shown (C). NP mRNA expression (D) and virus titers (E) were measured in lung 24 hr postinfection. n = 5 per each group. n.s., not signicant. **p < 0.01 compared with vehicle. (F) Lung tissues at 2 days after infection were stained with the R309 antibody to detect inuenza virus antigen.
et al., 2002; Liker et al., 2000). FG-Nups facilitate RNA transport as their repeats provide docking sites for carriers during their movement across the nuclear pore complex (Reed and Hurt, 2002; Stewart, 2010). Among FG-Nups, knockdown of Nup62, but not Nup98 or Nup214, signicantly attenuated virus replication assessed by NP mRNA expression (Figure 5D) and virus titers (Figure S5A) in A549 cells, as previously reported for immunodeciency virus type 1 (HIV-1) infections (Monette et al., 2011). FISH analyses further showed that Nup62 knockdown inhibited the nuclear export of virus mRNA and vRNA (Figure S5B), whereas knockdown of Nup98 nor Nup214 did not affect the nuclear export of inuenza virus transcripts (Figure S5C). Whether inuenza virus RNAs can directly bind NXF1 has been elusive. By using a cell-free in vitro RNA gel shift assay, we found that recombinant protein expressing full-length NXF1 1-619 and p15/NXT1 (NXF1-p15) generated band shifts of inuenza virus mRNA and vRNA (Figure S6A), indicating that inuenza virus RNAs directly binds to NXF1-p15. In contrast, we failed to detect a direct binding of Nup62-FG protein to virus RNAs (Figure S6B). Interestingly, coincubation of the NXF1-p15 and Nup62-FG proteins generated a supershift of inuenza virus mRNA and vRNA (Figure 5E), supporting the notion that NXF1-p15 and Nup62-FG form a protein complex in vitro, which can bind virus RNAs. Of note, immunoprecipitation experiments showed that Nup62 indeed associated with NXF1 in noninfected and infected cells (Figure S6C). These data indicate that inuenza virus RNAs directly bind to NXF1 and that NXF1, in cooperating with a key docking protein Nup62, facilitates inuenza virus RNAs export and virus replication. We next tested whether PD1 could interfere with the recruitment of virus RNAs to NXF1 that can bind Nup62 using an RNA immunoprecipitation (RIP) assays. PD1, but not RvD1, treatment signicantly reduced the amount of virus mRNA and vRNA that immunoprecipitated with the anti-NXF1 antibody (Ab) (Figure 5F). The Ab used was immunoprecipitated with NXF1 (Figure S6D). To further conrm this nding, we conducted a RIP sequencing (RIPseq) analysis. The RNAs that immunoprecipitated with the anti-NXF1 Ab indeed mapped to inuenza virus RNAs. PD1 treatment signicantly reduced the copy number of all inuenza virus RNAs that immunoprecipitated with the antiNXF1 Ab (Figure 5G). Furthermore, we detected the presence of Nup62 microclusters around the nuclear rim and a distinct speckled cytoplasmic staining at 8 hr after infection (Figure 5H). Importantly, upon PD1, but not RvD1, treatment, Nup62 remained at the nuclear envelop in the virus-infected cells (Figure 5H). Taken together, our data indicate that PD1 inhibits the recruitment of virus RNAs to NXF1 that can cooperate with
Nup62 for RNA export, thereby attenuating virus RNA export and replication. Finally, we asked whether PD1 also affects nuclear export of host poly(A) mRNA. FISH analysis showed that PD1 treatment had no apparent impact on the nuclear export of host bulk poly(A) mRNA in the inuenza-virus-infected cells (Figure 5I). To conrm this nding, we isolated nuclear and cytoplasmic RNA fractions from the cells at 8 hr after mock or PR8 virus infection and subjected to RNA sequencing (RNAseq) analysis. Consistent with our FISH data (Figure 5I), the distributions of nuclear and cytoplasmic RNA sequences were largely unaffected by PD1 treatment (Figure 5J), although some RNAs appeared changed following PD1 treatment (Table S4); the signicance of these changes needs to be determined, though none of these changed RNA have been yet implicated in virus replication. PD1 Production Is 12/15-LOX Dependent 12/15-LOX (the murine ortholog of human 15-LOX1) (Chanez et al., 2002) is known to be a key enzyme in the biosynthesis of PD1 (Yamada et al., 2011). To test whether the generation of PD1 in the inuenza-virus-infected lung was indeed dependent on 12/15-LOX, we conducted lipidomics for PUFA-derived metabolites in 12/15-LOX knockout (KO) (Sun and Funk, 1996) versus control wild-type (WT) mice. Compared with WT mice, levels of PD1 were reduced in 12/15 LOX KO mice at baseline (Figure 6A). Upon inuenza virus infection for 24 hr, production of PD1 was markedly reduced in 12/15 LOX KO mice (Figure 6A). In addition, the production of 17-HDoHE, the upstream product of PD1, was also markedly reduced in 12/15 LOX KO mice at baseline and at 24 hr postinfection (Figure 6A). Of note, production of LXA4 was apparently not affected by the genetic inactivation of 12/15-LOX (Figure S7A). 5-lipoxygenase followed by 12-lipoxygenase action (Papayianni et al., 1995) might be involved in the production of lipoxin A4. These data indicate that the endogenous production of PD1 is critically dependent on 12/15-LOX expression. In lungs, 12/15-LOX is predominantly expressed in epithelial cells and leukocytes (Chanez et al., 2002). In PR8-infected mice, mRNA (Figure 6B) and protein levels (Figure 6C) of 12/ 15-LOX were decreased in lungs. NP mRNA levels (Figure 6D) and virus titers (Figure 6E) were signicantly higher in the lung tissues of PR8-virus-infected 12/15-LOX decient mice at 2 days postinfection. 12/15-LOX KO mice also exhibited greater body weight loss after PR8 virus infections (Figure 6F). Furthermore, we isolated mouse embryonic broblasts (MEFs) from WT and 12/15-LOX KO mice and infected them with the PR8
(G and H) RvD1, RvD2, LXA4, or vehicle was given i.v. 12 hr before and immediately after PR8 infection of WT mice (n = 6 for each group). Percent survival is shown (G). NP mRNA expression was measured in lungs at 2 and 5 days postinfection (H). (I and J) Therapeutic effects of PD1 on lung function of PR8-infected mice. Two days postinfection mice were intratracheally treated with PD1 or vehicle (veh), and then mechanically ventilated for 2.5 hr. Changes in pulmonary tissue resistance (I) and tissue elastance (J) were analyzed. n = 6 per group. **p < 0.01 for the entire time course compared with vehicle treatment. (K) Survival of inuenza-virus-infected mice upon delayed treatment with the antiviral drug peramivir (n = 5), PD1 (n = 7), peramivir plus PD1 (n = 7), or vehicle (n = 7). WT mice were infected intratracheally with PR8 virus. In the peramivir group, 2 days postinfection, peramivir (10 mg/kg) was given i.v. In the PD1 group, from 2 days postinfection PD1 (1 mg/mouse/day) was given i.v. for 3 consecutive days. In the peramivir plus PD1 group, peramivir (10 mg/kg) was given i.v. at 2 days postinfection and PD1 (1 mg/mouse/day) was given i.v. for three consecutive days as indicated. Data in (C), (D), (E), and (H)(J) are presented as means S.E.M. See also Figure S3.
A
Virus mRNA 3h 5h 8h
Vehicle
B
Virus mRNA/18S (AU)
Virus vRNA/18S (AU)
Virus vRNA 3h 5h 8h
2.0 1.5 1.0 0.5 0.0

ve h PD R 1 vD 1 ve h PD R 1 vD 1
2.0
* *
1.5 1.0 0.5 0.0
* *
RvD1
PD1
Nuclei Cytoplasm
C
0.05 0.04
NP/18s (AU)
D
0.10
E
Virus mRNA/18S (AU)
F
0.8
0.03 0.02 0.01 0.00

on
NP/18S (AU)
0.08
Virus vRNA/18S (AU)
N
mRNA
0.6 0.4 0.2 0.0
0.06 0.04 0.02 0.00
**
**
**
N
**
ve h PD R 1 vD 1 ve PDh R 1 vD 1
Nuclei Cytoplasm
1.0 0.8 0.6 0.4 0.2 0.0
**
t C R N rol M X 1/ F X p1 P 1 p1 5/ O N 1 5/ X N XT N T1 1+ XT N 2 XT 2
tr o up l 6 N 2 up N 98 up 21 4
on
vRNA
ve PDh R 1 vD 1 ve PDh R 1 vD 1
siRNA
Input
IP: NXF1 Ab
siRNA
H
Nup62
Vehicle
NXF1
I
3h
Vehicle
Host Poly(A) mRNA

5h 8h
J
Nucleus
PD1 [Log 2(FPKM)]
RvD1
PD1
Cytoplasm
PD1 [Log2(FPKM)]
15 10 5 0
15 10 5 0
Flu (-)
Flu (+)
Flu (-)
PD1
10
15 0
10
15
10
ve P h R D1 vD 1 ve h P R D1 vD 1
Input IP: NXF1 Ab
Merge
Flu (+)
15 0
10
Vehicle [Log2 (FPKM)] Vehicle [Log2 (FPKM)]
Vehicle [Log2 (FPKM)] Vehicle [Log2 (FPKM)]
15
Figure 5. PD1 Inhibits Nuclear Export of Viral Transcripts

(A and B) A549 cells were infected with inuenza PR8 virus (MOI: 2) and treated with PD1 (2.7 mM) or vehicle immediately after infection. Infected cells were xed at the indicated time points, and PR8 virus mRNA and virus vRNA (A) were detected by uorescence in situ hybridization. Nuclei were counterstained with DAPI. Cytoplasmic and nuclear fractions were separately isolated from the cells at 8 hr after infection, and virus mRNA and vRNA (B) were quantied by qRT-PCR. *p < 0.05 compared with vehicle. (C and D) siRNA knockdown of RNA export genes (C) or FG-nucleoporins (D) in PR8 inuenza-infected A549 cells (MOI: 0.2). NP mRNA expression was quantied at 24 hr after infection by qRT-PCR. **p < 0.01 compared with control siRNA.
viruses. NP mRNA levels (Figure 6G) and virus titers (Figure S7C) were again markedly higher in 12/15-LOX KO MEFs. Importantly, the elevated NP mRNA expression (Figure 6G) and virus titers (Figure S7C) in the 12/15-LOX KO MEFs were signicantly attenuated by the treatment with PD1, 17-HDoHE, or a combination of both. These results show that the 12/15-LOX is the key enzyme required for the biosynthesis of PD1 and 17-HDoHE in inuenzavirus-infected mouse lung. DISCUSSION Novel families of biologically active lipid mediators derived from PUFA precursors have recently been uncovered and their complete stereochemical assignments established (Schwab et al., 2007; Serhan, 2007; Serhan et al., 2011; Isobe et al., 2012). These lipid mediators include AA-derived lipoxin A4 (LXA4), DHA-derived PD1 (also known neuroprotectin D1 when generated in neuronal tissues), D series resolvins (RvD1, RvD2), and the EPA-derived E series resolvins (Serhan, 2007). These mediators are known to have anti-inammatory properties and display potent protective actions in experimental inammatory diseases such as peritonitis (Schwab et al., 2007; Yamada et al., 2011) and even Alzheimers disease (Lukiw et al., 2005). Here, we used mediator lipidomics to show the metabolic proles and functions of these mediators in the pathology of inuenza virus infection. To examine whether such bioactive mediators and biosynthetic intermediates are involved in inuenza virus replication, we screened PUFA-derived products for their effects on inuenza virus replication. Among them, PD1 markedly attenuated inuenza virus replication and, most importantly, promoted survival of infected mice. Similar to PD1, RvD1 and RvD2 are DHA-derived products of the 12/15-LOX pathway (Serhan, 2007). However, RvD1 and RvD2 were under the detection limit after PR8 or H5N1 virus infections, possibly because of scarce inltration with neutrophils. The modes of interaction of NXF1 with cellular and viral RNAs are likely to be different. Constitutive transport elements (CTE) of ter retroviral RNAs can directly bind NXF1 (Braun et al., 1999; Gru et al., 1998), and NXF1 has been shown to control ICP27-mediated export of Herpes simplex virus (HSV)-1 mRNA (Yatherajam et al., 2011). Although inuenza virus infections have been shown to downregulate the host mRNA export machinary by forming an inhibitory complex consisting of viral NS1 and the host NXF1, p15, and other mRNA export factors (Satterly et al., 2007), it was unknown whether NXF1 is in fact involved in the virus RNA export. We show that inuenza virus RNAs directly bind to NXF1 and that NXF1, in cooperating with Nup62, facilitates inu-
enza virus RNAs export from the nucleus to the cytoplasm. However, to date it remains largely unknown the involvements of viral proteins and/or adaptor proteins in the NXF1-mediated inuenza virus RNAs export (Schneider and Wolff, 2009). Importantly, we found that PD1 attenuates the recruitment of virus RNAs to NXF1 that can bind Nup62 and suppresses the altered localization of Nup62 in the virus-infected cells. The precise mechanisms involved in this process including the PD1s interaction to viral proteins such as NS1 should be claried in the future study. Furthermore, a yet unidentied specic receptor for PD1 would further raise the important question of how endogenous PD1 controls the downstream pathways connecting to NXF1-Nup62, although PD1 could also have many yet unexplored effects beyond regulation of the nuclear pores. A key approach to treat severe inuenza has been pharmacological targeting of both the virus life cycle and the subsequent extreme pulmonary inammation. Current anti-inuenza virus drugs are benecial only if administered early after infection; they are much less effective if administered 48 hr after infection (Kumar, 2011). In this context, a key nding of our study is that treatment of PD1 beginning around 48 hr after the initial infection, combined with peramivir, completely rescued mice from u mortality, indicating that PD1 has a marked benecial effect on severe PR8 inuenza virus infections both prophylactically and, most importantly, therapeutically. Deep RNA sequencing showed that PD1 exhibits only minor effects on host poly(A) mRNA nuclear export under both noninfected as well as infected conditions, although it has not yet identied how PD1 inhibits the export specic for virus RNA rather than host RNA. Antiviral cytokine IFNb expression levels were also not affected by PD1 treatment in vivo and in vitro. These results may imply that PD1 treatment is not likely to induce a harmful inhibition of host RNA export that could possibly produce detrimental side effects in host cells. Given its demonstrated potency, its potent antiinammatory properties in sterile and bacterial inammation (Serhan, 2007), and its production in healthy human airways and reduction in disease (Levy et al., 2007), PD1 could serve as a novel antiviral drug as well as a biomarker for severe and lethal inuenza virus infections.
EXPERIMENTAL PROCEDURES Viruses A/Puerto Rico/8/34 (H1N1; PR8), A/California/04/2009 H1N1 (2009 H1N1), A/Vietnam/1203/04 H5N1 (H5N1), and a mutant H5N1 virus possessing a single amino-acid substitution, from lysine to glutamic acid, at position 627 in PB2 (H5N1 PB2-627E) were used. BSL3 conditions were used for experiments with H5N1 and H5N1 PB2-627E viruses.
(E) Coincubation of NXF1-p15 plus Nup62-FG proteins generated a supershift of inuenza virus mRNA and vRNA. Virus RNA alone (N) served as a negative control. (F and G) RNA immunoprecipitaion and viral RNA sequencing. Input and immunoprecipitaed RNAs were analyzed by RT-qPCR for virus mRNA and vRNA (F). The immunoprecipitated RNAs were subjected to RNA sequencing. Individual graphs represent the read depth Log10 (RPM) for each segment of the virus (G). (H) Immunocytochemistry for Nup62 (green) and NXF1 (red) expression in the PD1-, RvD1-, or vehicle treated cells at 8 hr after infection. (I) The infected cells were xed at the indicated time points, and Poly(A) mRNA was detected by uorescence in situ hybridization. Nuclei were counterstained with DAPI. (J) Host mRNA sequencing. Nuclear and cytoplasmic fractions were isolated from the cells at 8 hr after mock or PR8 virus infection and the extracted RNAs subjected to RNA sequencing. Individual graphs represent the pairwise relationships of the expression level (FPKM) comparing PD1 (y axis) and vehicle (x axis) treatment in nuclear (left) and cytoplasmic (right) RNAs under noninfected and infected conditions. Data in (B), (C), (D), and (F) are presented as means S.E.M. See also Figures S4, S5, S6, and Table S4.
G F
Figure 6. 12/15 LOX-Dependent PD1 Production in Inuenza Virus-Infected Lungs

(A) Lipidomics for 12/15-LOX knockout (KO) mice. WT and 12/15-LOX KO mice were infected intratracheally with PR8 virus (1 3 104 TCID50). Lung tissues were sampled at time point 0 (baseline) and 24 hr postinfection. The concentrations (pg/mg tissue) of the DHA-derived metabolites are shown; n = 5 per group. (B) WT mice were intratracheally infected with mock (PBS) or PR8 virus. Lung tissues were sampled at the indicated time points postinfection and 12/15-LOX mRNA levels measured by qRT-PCR. *p < 0.05 for the whole time course compared to PBS controls. (C) Lung tissues were sampled from WT mice at the indicated time points postinfection, and used for western blotting using Abs to 12/15-LOX and b-actin as a loading control. Lung tissues obtained from 12/15 LOX KO mice served as a negative control.
Animals C57BL/6 mice were bred at our animal facility. 12/15-LOX decient mice on the C57BL/6 background were purchased from the Jackson Laboratory. Animal experiments and housing of mice were performed in accordance with institutional guidelines. Intratracheal Inuenza Virus Infection Model and Measurements of Respiratory Function Mice were intratracheally infected with inuenza viruses as described elsewhere (Ichikawa et al., 2013). For respiratory function, airway elastance and resistance were measure as described elsewhere (Imai et al., 2005, 2008) using the exiVent system (SCIREQ). Mediator Lipidomics LC-MS/MS-based lipidomics analyses were performed using an highperformance liquid chromatography (HPLC) system (Waters UPLC) with a linear ion trap quadrupole mass spectrometer (QTRAP5500; AB SCIEX) equipped with an Acquity UPLC BEH C18 column (Waters) as described (Arita, 2012). MS/MS analyses were conducted in negative ion mode, and fatty acid metabolites were identied and quantied by multiple reaction monitoring (MRM). Lipid Metabolites The Protectin D1 isomer (PD1, 10S,17S-diHDoHE), RvD1, RvD2, and 15HETE-d8, LTB4-d4, and PGE2-d4 were purchased from Cayman Chemical. To screen for PUFA-derived compounds, we also used the Bio-active screen Lipid II Library (Cayman Chemical). Virus Infection of Cultured Cells The A549 human lung epithelial cells, mouse embryonic broblasts and Madin Darby canine kidney cells (MDCK) were washed with PBS and then infected with virus at the indicated MOIs in DMEM 0.1% BSA for 1 hr at 37 C. Cells were then washed and incubated for the indicated time periods at 37 C in Dulbeccos modied Eagles medium (DMEM) supplemented with 10% FCS. Immunocytochemistry A549 cells on glass coverslips were xed with 4% paraformaldehyde and incubated with anti-NP Abs (ATCC, Hb65), anti-R309 Abs (Itoh et al., 2009), antiNXF1-C Abs (Katahira et al., 1999), or anti-Nup62 MAbs (BD Transduction Laboratories), and then incubated with appropriate secondary antibodies. Coverslips were mounted with mounting medium containing 4,6-diamidino2-phenylindole (DAPI). Images were analyzed with a multiphoton laser microscope (A1R MP, Nikon). Fluorescent In Situ Hybridization FISH for inuenza virus RNAs and poly(A) mRNA was conducted as described previously (Amorim et al., 2007). To generate a digoxigenin labeled positive- or negative-sense RNA probe, we linearized pSPT19-PR8-M plasmid and transcribed it in vitro with SP6 or T7 RNA polymerase, respectively, by using DIG-RNA labeling kit (Roche). A Cy3-labeled oligo (dT) probe was synthesized for poly(A) mRNA FISH. A549 cells were xed and permeabilized. After 1 hr prehybridization, the cells were hybridized with the RNA probe for 16 hr at 37 C. A digoxigenin labeled probe was detected by indirect immunouorescence with an antidigoxigenin uorescein Fab fragment (Roche). Images were analyzed using multiphoton laser microscopy (A1R MP, Nikon).
Cellular Fractionation Cytoplasmic and nuclear fractions of the cells were segregated by use of a subcellular protein fractionation kit (Thermo Scientic) in the presence of protease inhibitor cocktail (Roche) and RNase inhibitor (Toyobo). Quantitative Real-Time PCR RNA was extracted from cells using the RNeasy Mini Kit (QIAGEN, Valencia, CA). First-strand cDNA was synthesized from DNA-free RNA using a Primescript RT reagent kit (Takara). Samples of rst strand cDNA were subjected to real-time PCR quantication (Takara) using specic primers for the indicated RNAs with GAPDH or 18S as an internal control. Relative amounts of RNAs were calculated by using the comparative CT method. RNA-Binding Protein Immunoprecipitation Assay and RIP Sequencing RIP was carried out using an immunoprecipitation kit (RIP-assay kit; MBL). Briey, cells were homogenized and incubated overnight with protein G agarose beads (Thermo) preincubated with an anti-NXF1 Ab (BD Biosciences) or control Ab. RNAs bound to the beads were puried, and qPCR analysis was carried out. For RIP sequencing, immunoprecipitated RNAs were subjected to RNA sequencing analysis using HiSeq2000 (Illumina). The sequencing run yielded 30 million reads on average for the PD1 and vehicle treatment samples. Low-quality bases in sequencing reads were trimmed, and the trimmed reads were mapped to the inuenza genome of the strain A/Puerto Rico/8/34(H1N1) by using Bowtie2 (Langmead and Salzberg, 2012). Read counts were normalized to reads per million reads (RPM). RNAseq Nuclear and cytoplasmic fractionated RNAs were subjected to RNA sequencing analysis using HiSeq2000 (Illumina). The sequencing run yielded 12 million reads on average. The quality-ltered reads were aligned to the human genome hg19 by using TopHat (Trapnell et al., 2009). The relative expression levels (fragments per kilobase per million mapped fragments, FPKM) were calculated using cuffdiff within the software package Cufinks version 2.0.2 (Trapnell et al., 2010). The software package CummeRbund version 2.0.0 was used to analyze the output data of cuffdiff (Trapnell et al., 2012). In Vitro RNA-Binding Assay To generate PR8 M positive-sense mRNA and negative-sense vRNA, linearized pSPT19-PR8-M was in vitro transcribed with SP6 RNA polymerase and T7 RNA polymerase, respectively, by using MEGAscript kit (Ambion), followed by biotinylation using an RNA30 end biotinylation kit (Pierce). In vitro RNA gel shift assays were carried out using a chemiluminescent RNA electrophoretic mobility gel shift assay (EMSA) kit (Thermo Scientic) following the manufacturers instructions. RNA-protein complexes were separated by 4% native polyacrylamide gel, electroblotted to hybond N+ membrane, and the bands visualized with streptavidin-horseradish peroxidase conjugate and chemiluminescence. Statistical Analyses Measurements at single time points were analyzed by using an unpaired t test or analysis of variance (ANOVA). Time courses were analyzed by repeated measurements (mixed model) of ANOVA. Log-rank tests were performed on Kaplan-Meier survival curves. All statistical tests were calculated using the GraphPad Prism 5.00 program. p < 0.05 was considered to indicate statistical signicance.
(D and E) WT and 12/15-LOX KO mice were infected intratracheally with PR8 virus (1 3 104 TCID50) Virus NP mRNA (D) and virus titers (E) were quantied in lung tissues at 24 hr postinfection by qRT-PCR and focus forming unit (FFU) replication assay. n = 5 per each group. *p < 0.05 compared with WT. (F) Body weight changes in inuenza-infected 12/15-LOX KO mice. WT (n = 10) and 12/15-LOX KO (n = 11) mice were infected intratracheally with PR8 virus (10 TCID50). (G) Mouse embryonic broblasts (MEFs), obtained from WT or 12/15-LOX KO mice, were infected with PR8 virus (MOI: 2) and treated with PD1 (8 mM), 17-HDoHE (8 mM), or a combination of both (4 mM each) immediately after infection. Virus NP mRNA expression was quantied in the infected cells at 24 hr postinfection. **p < 0.01. *p < 0.05. Data in (A), (B), (D), (E), (F),and (G) are presented as means S.E.M. See also Figure S7.
SUPPLEMENTAL INFORMATION Supplemental Information includes Extended Experimental Procedures, seven gures, and ve tables and can be found with this article online at http://dx.doi. org/10.1016/j.cell.2013.02.027. ACKNOWLEDGMENTS We thank all members of our laboratories for helpful discussions. We gratefully thank Ms. Michiko Kamio for technical assistance on LC-MS/MS analyses. Y.I. and K.K. are supported by the Funding Program for Next Generation WorldLeading Researchers (NEXT Program, JSPS). Y.K. is supported by the Center of Education and Research for the Advanced Genome-Based Medicine-For Personalized Medicine and the Control of Worldwide Infectious Diseases, by a grant-in-aid for Specially Promoted Research and by the Japan Initiative for Global Research Network on Infectious Diseases from the Ministry of Education, Culture, Sports, Science and Technology, by ERATO (Japan Science and Technology Agency), and by Public Health Service research grants from the National Institute of Allergy and Infectious Diseases. M.A. is supported by a grant-in-aid from the Japan Science and Technology Agency Precursory Research for Embryonic Science and Technology (PRESTO), and a grant-in-aid from the Ministry of Education, Culture, Sports, Science, and Technology of Japan. J.M.P. is supported by IMBA and an Advanced ERC grant by the European Union. H.A. and M.A. are supported in part by the Program for Promotion of Basic and Applied Research for Innovations in Bio-Oriented Industry. Received: June 7, 2012 Revised: December 8, 2012 Accepted: February 13, 2013 Published: March 7, 2013 REFERENCES Amorim, M.J., Read, E.K., Dalton, R.M., Medcalf, L., and Digard, P. (2007). Nuclear export of inuenza A virus mRNAs requires ongoing RNA polymerase II activity. Trafc 8, 111. Arita, M. (2012). Mediator lipidomics in acute inammation and resolution. J. Biochem. 152, 313319. Arya, V., Carter, W.W., and Robertson, S.M. (2010). The role of clinical pharmacology in supporting the emergency use authorization of an unapproved antiinuenza drug, peramivir. Clin. Pharmacol. Ther. 88, 587589. Beigel, J.H., Farrar, J., Han, A.M., Hayden, F.G., Hyer, R., de Jong, M.D., Lochindarat, S., Nguyen, T.K., Nguyen, T.H., Tran, T.H., et al.; Writing Committee of the World Health Organization (WHO) Consultation on Human Inuenza A/H5. (2005). Avian inuenza A (H5N1) infection in humans. N. Engl. J. Med. 353, 13741385. Braun, I.C., Rohrbach, E., Schmitt, C., and Izaurralde, E. (1999). TAP binds to the constitutive transport element (CTE) through a novel RNA-binding motif that is sufcient to promote CTE-dependent RNA export from the nucleus. EMBO J. 18, 19531965. Chanez, P., Bonnans, C., Chavis, C., and Vachier, I. (2002). 15-lipoxygenase: a Janus enzyme? Am. J. Respir. Cell Mol. Biol. 27, 655658. Clark, N.M., and Lynch, J.P., 3rd. (2011). Inuenza: epidemiology, clinical features, therapy, and prevention. Semin. Respir. Crit. Care Med. 32, 373392. ter, P., Tabernero, C., von Kobbe, C., Schmitt, C., Saavedra, C., Bachi, A., Gru Wilm, M., Felber, B.K., and Izaurralde, E. (1998). TAP, the human homolog of Mex67p, mediates CTE-dependent RNA export from the nucleus. Mol. Cell 1, 649659. Hao, L., Sakurai, A., Watanabe, T., Sorensen, E., Nidom, C.A., Newton, M.A., Ahlquist, P., and Kawaoka, Y. (2008). Drosophila RNAi screen identies host genes important for inuenza virus replication. Nature 454, 890893. Herold, A., Suyama, M., Rodrigues, J.P., Braun, I.C., Kutay, U., Carmo-Fonseca, M., Bork, P., and Izaurralde, E. (2000). TAP (NXF1) belongs to a multigene
family of putative RNA export factors with a conserved modular architecture. Mol. Cell. Biol. 20, 89969008. Herold, A., Klymenko, T., and Izaurralde, E. (2001). NXF1/p15 heterodimers are essential for mRNA nuclear export in Drosophila. RNA 7, 17681780. Ho, D.N., Coburn, G.A., Kang, Y., Cullen, B.R., and Georgiadis, M.M. (2002). The crystal structure and mutational analysis of a novel RNA-binding domain found in the human Tap nuclear mRNA export factor. Proc. Natl. Acad. Sci. USA 99, 18881893. Ichikawa, A., Kuba, K., Morita, M., Chida, S., Tezuka, H., Hara, H., Sasaki, T., Ohteki, T., Ranieri, V.M., dos Santos, C.C., et al. (2013). CXCL10-CXCR3 enhances the development of neutrophil-mediated fulminant lung injury of viral and nonviral origin. Am. J. Respir. Crit. Care Med. 187, 6577. Imai, M., and Kawaoka, Y. (2012). The role of receptor binding specicity in interspecies transmission of inuenza viruses. Curr. Opin. Virol. 2, 160167. Imai, Y., Kuba, K., Neely, G.G., Yaghubian-Malhami, R., Perkmann, T., van Loo, G., Ermolaeva, M., Veldhuizen, R., Leung, Y.H., Wang, H., et al. (2008). Identication of oxidative stress and Toll-like receptor 4 signaling as a key pathway of acute lung injury. Cell 133, 235249. Imai, Y., Kuba, K., Rao, S., Huan, Y., Guo, F., Guan, B., Yang, P., Sarao, R., Wada, T., Leong-Poi, H., et al. (2005). Angiotensin-converting enzyme 2 protects from severe acute lung failure. Nature 436, 112116. Isobe, Y., Kato, T., and Arita, M. (2012). Emerging roles of eosinophils and eosinophil-derived lipid mediators in the resolution of inammation. Front Immunol. 3, 270. Itoh, Y., Shinya, K., Kiso, M., Watanabe, T., Sakoda, Y., Hatta, M., Muramoto, Y., Tamura, D., Sakai-Tagawa, Y., Noda, T., et al. (2009). In vitro and in vivo characterization of new swine-origin H1N1 inuenza viruses. Nature 460, 10211025. Karlas, A., Machuy, N., Shin, Y., Pleissner, K.P., Artarini, A., Heuer, D., Becker, D., Khalil, H., Ogilvie, L.A., Hess, S., et al. (2010). Genome-wide RNAi screen identies human host factors crucial for inuenza virus replication. Nature 463, 818822. sser, K., Podtelejnikov, A., Mann, M., Jung, J.U., and Hurt, E. Katahira, J., Stra (1999). The Mex67p-mediated nuclear mRNA export pathway is conserved from yeast to human. EMBO J. 18, 25932609. nig, R., Stertz, S., Zhou, Y., Inoue, A., Hoffmann, H.H., Bhattacharyya, S., Ko Alamares, J.G., Tscherne, D.M., Ortigoza, M.B., Liang, Y., et al. (2010). Human host factors required for inuenza virus replication. Nature 463, 813817. Kumar, A. (2011). Early versus late oseltamivir treatment in severely ill patients with 2009 pandemic inuenza A (H1N1): speed is life. J. Antimicrob. Chemother. 66, 959963. Kumar, S., Havens, P.L., Chusid, M.J., Willoughby, R.E., Jr., Simpson, P., and Henrickson, K.J. (2010). Clinical and epidemiologic characteristics of children hospitalized with 2009 pandemic H1N1 inuenza A infection. Pediatr. Infect. Dis. J. 29, 591594. Langmead, B., and Salzberg, S.L. (2012). Fast gapped-read alignment with Bowtie 2. Nat. Methods 9, 357359. Levy, B.D., Kohli, P., Gotlinger, K., Haworth, O., Hong, S., Kazani, S., Israel, E., Haley, K.J., and Serhan, C.N. (2007). Protectin D1 is generated in asthma and dampens airway inammation and hyperresponsiveness. J. Immunol. 178, 496502. Liker, E., Fernandez, E., Izaurralde, E., and Conti, E. (2000). The structure of the mRNA export factor TAP reveals a cis arrangement of a non-canonical RNP domain and an LRR domain. EMBO J. 19, 55875598. Lukiw, W.J., Cui, J.G., Marcheselli, V.L., Bodker, M., Botkjaer, A., Gotlinger, K., Serhan, C.N., and Bazan, N.G. (2005). A role for docosahexaenoic acidderived neuroprotectin D1 in neural cell survival and Alzheimer disease. J. Clin. Invest. 115, 27742783. , N., and Mouland, A.J. (2011). HIV-1 remodels the nuclear Monette, A., Pante pore complex. J. Cell Biol. 193, 619631. Neumann, G., Brownlee, G.G., Fodor, E., and Kawaoka, Y. (2004). Orthomyxovirus replication, transcription, and polyadenylation. Curr. Top. Microbiol. Immunol. 283, 121143.
Papayianni, A., Serhan, C.N., Phillips, M.L., Rennke, H.G., and Brady, H.R. (1995). Transcellular biosynthesis of lipoxin A4 during adhesion of platelets and neutrophils in experimental immune complex glomerulonephritis. Kidney Int. 47, 12951302. Read, E.K., and Digard, P. (2010). Individual inuenza A virus mRNAs show differential dependence on cellular NXF1/TAP for their nuclear export. J. Gen. Virol. 91, 12901301. Reed, R., and Hurt, E. (2002). A conserved mRNA export machinery coupled to pre-mRNA splicing. Cell 108, 523531. guez-Navarro, S., and Hurt, E. (2011). Linking gene regulation to mRNA Rodr production and export. Curr. Opin. Cell Biol. 23, 302309. Satterly, N., Tsai, P.L., van Deursen, J., Nussenzveig, D.R., Wang, Y., Faria, P.A., Levay, A., Levy, D.E., and Fontoura, B.M. (2007). Inuenza virus targets the mRNA export machinery and the nuclear pore complex. Proc. Natl. Acad. Sci. USA 104, 18531858. Schneider, J., and Wolff, T. (2009). Nuclear functions of the inuenza A and B viruses NS1 proteins: do they play a role in viral mRNA export? Vaccine 27, 63126316. Schwab, J.M., Chiang, N., Arita, M., and Serhan, C.N. (2007). Resolvin E1 and protectin D1 activate inammation-resolution programmes. Nature 447, 869874. Serhan, C.N. (2007). Resolution phase of inammation: novel endogenous anti-inammatory and proresolving lipid mediators and pathways. Annu. Rev. Immunol. 25, 101137. Serhan, C.N., and Levy, B. (2003). Novel pathways and endogenous mediators in anti-inammation and resolution. Chem. Immunol. Allergy 83, 115145. Serhan, C.N., Hong, S., Gronert, K., Colgan, S.P., Devchand, P.R., Mirick, G., and Moussignac, R.L. (2002). Resolvins: a family of bioactive products of omega-3 fatty acid transformation circuits initiated by aspirin treatment that counter proinammation signals. J. Exp. Med. 196, 10251037. Serhan, C.N., Krishnamoorthy, S., Recchiuti, A., and Chiang, N. (2011). Novel anti-inammatorypro-resolving mediators and their receptors. Curr. Top. Med. Chem. 11, 629647. Serhan, C.N., Dalli, J., Karamnov, S., Choi, A., Park, C.K., Xu, Z.Z., Ji, R.R., Zhu, M., and Petasis, N.A. (2012). Macrophage proresolving mediator maresin 1 stimulates tissue regeneration and controls pain. FASEB J. 26, 17551765. Shankaran, S., and Bearman, G.M. (2012). Inuenza virus resistance to neuraminidase inhibitors: implications for treatment. Curr. Infect. Dis. Rep. 14, 155160.
Shapira, S.D., Gat-Viks, I., Shum, B.O., Dricot, A., de Grace, M.M., Wu, L., Gupta, P.B., Hao, T., Silver, S.J., Root, D.E., et al. (2009). A physical and regulatory map of host-inuenza interactions reveals pathways in H1N1 infection. Cell 139, 12551267. Shinya, K., Hamm, S., Hatta, M., Ito, H., Ito, T., and Kawaoka, Y. (2004). PB2 amino acid at position 627 affects replicative efciency, but not cell tropism, of Hong Kong H5N1 inuenza A viruses in mice. Virology 320, 258266. Sly, P.D., Collins, R.A., Thamrin, C., Turner, D.J., and Hantos, Z. (2003). Volume dependence of airway and tissue impedances in mice. J. Appl. Physiol. 94, 14601466. Smith, J.H., Nagy, T., Barber, J., Brooks, P., Tompkins, S.M., and Tripp, R.A. (2011). Aerosol inoculation with a sub-lethal inuenza virus leads to exacerbated morbidity and pulmonary disease pathogenesis. Viral Immunol. 24, 131142. Stade, K., Ford, C.S., Guthrie, C., and Weis, K. (1997). Exportin 1 (Crm1p) is an essential nuclear export factor. Cell 90, 10411050. Stewart, M. (2010). Nuclear export of mRNA. Trends Biochem. Sci. 35, 609617. Sun, D., and Funk, C.D. (1996). Disruption of 12/15-lipoxygenase expression in peritoneal macrophages. Enhanced utilization of the 5-lipoxygenase pathway and diminished oxidation of low density lipoprotein. J. Biol. Chem. 271, 24055 24062. Trapnell, C., Pachter, L., and Salzberg, S.L. (2009). TopHat: discovering splice junctions with RNA-Seq. Bioinformatics 25, 11051111. Trapnell, C., Williams, B.A., Pertea, G., Mortazavi, A., Kwan, G., van Baren, M.J., Salzberg, S.L., Wold, B.J., and Pachter, L. (2010). Transcript assembly and quantication by RNA-Seq reveals unannotated transcripts and isoform switching during cell differentiation. Nat. Biotechnol. 28, 511515. Trapnell, C., Roberts, A., Goff, L., Pertea, G., Kim, D., Kelley, D.R., Pimentel, H., Salzberg, S.L., Rinn, J.L., and Pachter, L. (2012). Differential gene and transcript expression analysis of RNA-seq experiments with TopHat and Cufinks. Nat. Protoc. 7, 562578. Yamada, T., Tani, Y., Nakanishi, H., Taguchi, R., Arita, M., and Arai, H. (2011). Eosinophils promote resolution of acute peritonitis by producing proresolving mediators in mice. FASEB J. 25, 561568. Yatherajam, G., Huang, W., and Flint, S.J. (2011). Export of adenoviral late mRNA from the nucleus requires the Nxf1/Tap export receptor. J. Virol. 85, 14291438.
Somatic Mutations of the Immunoglobulin Framework Are Generally Required for Broad and Potent HIV-1 Neutralization
Florian Klein,1,12 Ron Diskin,3,5,12 Johannes F. Scheid,1,6 Christian Gaebler,1,7 Hugo Mouquet,1 Ivelin S. Georgiev,8 Marie Pancera,8 Tongqing Zhou,8 Reha-Baris Incesu,1,9 Brooks Zhongzheng Fu,3 Priyanthi N.P. Gnanapragasam,3 Thiago Y. Oliveira,1,10 Michael S. Seaman,11 Peter D. Kwong,8 Pamela J. Bjorkman,3,4,13 and Michel C. Nussenzweig1,2,13,*
of Molecular Immunology Hughes Medical Institute The Rockefeller University, New York, NY 10065, USA 3Division of Biology 4Howard Hughes Medical Institute California Institute of Technology, 1200 E. California Boulevard, Pasadena, CA 91125, USA 5Department of Structural Biology, Weizmann Institute of Science, Rehovot 76100, Israel 6Charite Universita tsmedizin, D-10117 Berlin, Germany 7Faculty of Medicine Carl Gustav Carus, Technische Universita t Dresden, D-01307 Dresden, Germany 8Vaccine Research Center, National Institute of Allergy and Infectious Diseases, National Institutes of Health (NIH), Bethesda, MD 20892, USA 9Universita tsklinikum Hamburg-Eppendorf, D-20246 Hamburg, Germany 10Department of Genetics, Medical School of Ribeirao Preto/USP, National Institute of Science and Technology for Stem Cells and Cell Therapy and Center for Cell-based Therapy, Ribeirao Preto, SP 14051140, Brazil 11Beth Israel Deaconess Medical Center, Boston, MA 02215, USA 12These authors contributed equally to this work 13These authors contributed equally to this work *Correspondence: nussen@rockefeller.edu http://dx.doi.org/10.1016/j.cell.2013.03.018
2Howard 1Laboratory
SUMMARY
INTRODUCTION A fraction of HIV-1-infected individuals mount a broadly neutralizing serologic response (Doria-Rose et al., 2010; Simek et al., 2009) 23 years after infection (Mikell et al., 2011). Antibodies generated by these individuals are of great interest for vaccine design because they can protect macaques from infection (Mascola et al., 2000; Moldt et al., 2012; Shibata et al., 1999). Moreover, combinations of broadly neutralizing antibodies can control an established HIV-1 infection in humanized mice (Klein et al., 2012b). Despite their potential importance to vaccine development and HIV-1 therapy, little was known about the molecular composition of the human anti-HIV-1 antibody response until single-cell antibody cloning techniques were developed and used for characterizing IgGs from the sera of HIV-1-infected individuals with broadly neutralizing activity (Scheid et al., 2009a; Scheid et al., 2009b). This analysis revealed highly potent bNAbs, all of which might eventually be used in vaccine development (Corti et al., 2010; Huang et al., 2012; Morris et al., 2011; Mouquet et al., 2012; Scheid et al., 2011; Walker et al., 2009, 2011b; Wu et al., 2010). A surprising observation was that anti-HIV-1 antibodies are highly somatically mutated when compared to other immunoglobulins (IgGs) cloned from the same patients (Scheid et al., 2009a; Xiao et al., 2009a, 2009b). Whereas most human
Broadly neutralizing antibodies (bNAbs) to HIV-1 can prevent infection and are therefore of great importance for HIV-1 vaccine design. Notably, bNAbs are highly somatically mutated and generated by a fraction of HIV-1-infected individuals several years after infection. Antibodies typically accumulate mutations in the complementarity determining region (CDR) loops, which usually contact the antigen. The CDR loops are scaffolded by canonical framework regions (FWRs) that are both resistant to and less tolerant of mutations. Here, we report that in contrast to most antibodies, including those with limited HIV-1 neutralizing activity, most bNAbs require somatic mutations in their FWRs. Structural and functional analyses reveal that somatic mutations in FWR residues enhance breadth and potency by providing increased exibility and/or direct antigen contact. Thus, in bNAbs, FWRs play an essential role beyond scaffolding the CDR loops and their unusual contribution to potency and breadth should be considered in HIV-1 vaccine design.
antibodies that have undergone afnity maturation carry 1520 VH-gene somatic mutations (Tiller et al., 2007), potent broadly neutralizing antibodies carry 40100 VH-gene mutations (Corti et al., 2010; Scheid et al., 2011; Walker et al., 2009, 2011b; Wu et al., 2010; Xiao et al., 2009a, 2009b). These mutations are essential because reversion to the antibody germline sequence drastically reduces neutralizing potency and breadth (Mouquet et al., 2010; Scheid et al., 2011; Wu et al., 2011; Xiao et al., 2009b; Zhou et al., 2010). However, why so many mutations appear to be required is not known. Wu and Kabat rst divided antibody variable regions into complementarity determining regions (CDRs) and framework regions (FWRs) based on the number of somatic hypermutations in these regions (Wu and Kabat, 1970) (Figures 1A and 1B). The CDRs consist primarily of loops that form the sites of contact between the antibody and antigen (Amzel and Poljak, 1979) and account for the specicities of most antibody molecules as demonstrated by CDR grafting experiments (Jones et al., 1986). The structural integrity of the variable domains is maintained by the FWRs, which encode nine antiparallel b strands arranged into two b sheets (one sheet containing strands A, B, E, and D and the other containing strands C, C, C, F and G; Figures 1A and 1B). The relatively invariant b strands of the FWRs serve as a scaffold for three CDR loops, which connect strands B and C, C and C, and F and G (Figures 1A and 1B) (Amzel and Poljak, 1979). Somatic mutations are preferentially found in the CDR loops where they can alter the antibody combining site without affecting the overall structure of the variable domain (Wu and Kabat, 1970). Mutations in the FWR are usually poorly tolerated and generally biased to neutral substitutions to avoid changes that would destroy the structural underpinnings of the variable domain (Reynaud et al., 1995; Wagner et al., 1995). Here we examine the role of somatic mutations in the development of broadly neutralizing anti-HIV-1 antibodies. In contrast to most other antibodies, including anti-HIV-1 antibodies with limited neutralization activity, we found that FWR mutations, including noncontact residues, are essential for the neutralizing activity of most potent bNAbs. We propose that the requirement to alter the FWR, without destroying its essential structural elements, accounts for the high mutation load found in broadly neutralizing anti-HIV-1 antibodies and possibly for the difculty and prolonged latency with which such antibodies develop. RESULTS Somatic Hypermutation in HIV-1-Neutralizing Antibodies To examine the role of somatic hypermutations in anti-HIV-1 antibody neutralization breadth and potency, we selected a group of 9 HIV-1-reactive antibodies with activity limited to easy to neutralize (Tier 1) HIV-1 strains (Seaman et al., 2010), and 17 antibodies with broad neutralization activity (Figures 1C and S1 and Table S1 available online). The antibodies with limited neutralizing activity included antibodies recognizing the CD4-binding site (CD4bs; 6-187, 9913, and 11989) (Mouquet et al., 2011; Scheid et al., 2009a), the core epitope (1479, 2 491, and 11591) (Mouquet et al., 2011; Pietzsch et al., 2010;
Scheid et al., 2009a), the V3-loop (447-52D and 10188) (Gorny et al., 1993; Mouquet et al., 2011), and the CD4-induced site (17b) (Thali et al., 1993) (Table S1). Eight of the 17 bNAbs also recognize the CD4bs (VRC01, NIH45-46, 3BNC60, 12A12, 1NC9, 8ANC131, 12A21, and 3BNC117) (Scheid et al., 2011; Wu et al., 2010), whereas others recognized the V1/V2 loop (PG16) (Walker et al., 2009), carbohydrates (2G12) (Calarese et al., 2003; Trkola et al., 1996), the core epitope (HJ16) (Corti et al., 2010), the base of the V3-loop (10-1074 and PGT128) (Mouquet et al., 2012; Walker et al., 2011a), the membrane proximal external region (MPER; 4E10 and 2F5) (Buchacher et al., 1994; Muster et al., 1993) and two antibodies (3BC176 and 8ANC195) (Klein et al., 2012a; Scheid et al., 2011) for which the precise epitopes are not yet determined (Figure 1C and Table S1). Antibodies with limited neutralizing activity differ from bNAbs in that they generally carry fewer somatic mutations (Figures 1C and S1 and Table S1). We used the well-accepted Kabat system (Wu and Kabat, 1970) that utilizes sequence comparisons for FWR/CDR assignments. However, direct comparisons between Kabat and the IMGT numbering system (Giudicelli et al., 2006) (Figure 1B), which includes antibody structural data, were also performed for a subset of antibodies. Complete reversion of somatic mutations in the heavy and light chain V genes (FWR1-3 and CDR1/2) drastically reduces anti-HIV-1 antibody binding and neutralization activity (Buchacher et al., 1994; Mouquet et al., 2010; Scheid et al., 2011; Xiao et al., 2009b; Zhou et al., 2010). Moreover, reverting only the CDR1 and CDR2 in 3BNC60 and NIH45-46 strongly diminished binding and neutralization (Figure S2A and Data S1A). To determine the functional consequences of FWR mutations, we reverted the framework residues to their germline counterparts (FWR-GL) in each of the 26 selected antibodies (Data S1B1D) and evaluated binding to the HIV-1 envelope protein as well as their neutralization activities. HIV-1-Reactive Antibodies with Limited Activity As expected, reversion of somatic mutations in the FWR residues (FWR-GL) of the HIV-1 antibodies with limited breadth had only minimal effects on binding of most of these antibodies to gp140YU2 (Figure S2B) as measured by ELISA and conrmed by surface plasmon resonance (SPR) (data not shown). Only two of these FWR-GL antibodies (9-913 and 10-188) showed a decrease in binding (Figure S2B). In agreement with the ELISA and SPR experiments, we found little or no change in neutralizing activity in most of the FWR-GL antibodies with limited neutralization activity on a panel of up to six Tier 1 viruses representing clades A, B, and C (Figure S2B and Table S2). Only antibodies 9-913 and 10-188, which displayed decreased binding to gp140YU2, showed a decrease (9-913) or complete loss (10188) in neutralizing activity (Figure S2B and Table S2). We conclude that with two exceptions out of nine antibodies tested, FWR mutations do not alter the binding or neutralizing activity of anti-HIV-1 antibodies with limited neutralizing activity. Thus, despite their signicantly higher levels of somatic mutation, HIV-1-neutralizing antibodies with limited breadth resemble previously characterized antibodies to other antigens in that FWR mutations seem not to be essential.
Figure 1. Somatic Mutations in the Framework Regions of HIV-1-Reactive Antibodies

(A) Ribbon representation of the variable domains of 3BNC60 (Scheid et al., 2011), illustrating the CDRs (magenta) and the FWRs of the immunoglobulin heavy (blue) and light (cyan) chain. (B) Illustration of Kabat and IMGT CDR (magenta) and FWR (IgH, blue; IgL, cyan) assignments for the variable heavy and light chain domains of 3BNC60. Gray arrows indicate b strands dened by the crystal structure of the 3BNC60 Fab (Scheid et al., 2011). (C) Position of FWR mutations in heavy and light chain of the 17 investigated antibodies with broad neutralizing activity (see also Data S1). Indicated are silent (black) and replacement (red) mutations. Insertions are illustrated in blue. For mAbs 2G12, 2F5, and 4E10, only amino acid replacements are shown (red). Number of replacement mutations within CDR1/2 and FWR1-4 are listed in the two columns at the very right (see also Table S1). HIV-1-reactive antibodies with limited neutralization are displayed in Figure S1.
Potent Broadly Neutralizing Antibodies In contrast, reversion of the FWR mutations in most of the 17 broadly neutralizing antibodies decreased their binding to gp140YU2 (Figures 2 and 3). Three of the antibodies, PG16 (Walker
et al., 2009), 8ANC195 (Scheid et al., 2011), and 3BC176 (Klein et al., 2012a) bound poorly to gp140YU2 when measured up to 8 mg/ml and differences in binding between mutated and reverted antibodies could not be evaluated (Figure 2).
ELISA
Neutralizing activity
300 100 10
ELISA
300 100 10
3BNC60
3
mature
FWR-GL
3
PGT128
g/ml
OD405nm 2 1
mature
FWR-GL
g/ml
OD405nm
2 1
1 0.1 0.01
1 0.1 0.01
0.01
0.1
48
0.001
ab c de f h i j k
ab c de f h i j k
0.01
0.1
48
0.001
ab c d f gh i j k
ab c d f gh i j k
12A12
3
300 100 10
10-1074
3
300 100 10
g/ml
g/ml
0.01 0.1 1 48
OD405nm
OD405nm
2 1
1 0.1 0.01
2 1
1 0.1 0.01
0.01
0.1
48
0.001
ab c de f h i j k
ab c de f h i j k
0.001
abcde f gh i j k
abcde f gh i j k
1NC9
3
300 100 10
4E10
3
300 100 10
g/ml
OD405nm
2 1
OD405nm
1 0.1 0.01
g/ml
0.01 0.1 1 48
2 1
1 0.1 0.01
0.01
0.1
48
0.001
a b c d h i j k
a b c d h i j k
0.001
abcde f gh i j k
abcde f gh i j k
8ANC131
3
300 100 10
2F5
3
300 100 10
g/ml
g/ml
0.01 0.1 1 48
OD405nm
2 1
OD405nm
1 0.1 0.01
2 1
1 0.1 0.01
0.01
0.1
48
0.001
b c d e f h j k
b c d e f h j k
0.001
a b c d f g h j
a b c d f g h j
HJ16
3
300 100 10
PG16
3
300 100 10
g/ml
OD405nm
OD405nm
2 1
1 0.1 0.01
g/ml
0.01 0.1 1 48
2 1
1 0.1 0.01
0.01
0.1
48
0.001
0.001
a c d e f g h i j k
a c d e f g h i j k
8ANC195
3
300 100 10
3BC176
3
300 100 10
g/ml
OD405nm
2 1
OD405nm
1 0.1 0.01
g/ml
0.01 0.1 1 48
2 1
1 0.1 0.01
0.01
0.1
48
0.001
b d f h i
b d f h i
0.001
a b c e g h i
a b c e g h i
g/mlmAb
HIV strains
2G12
3
300 100 10
Virus strains (Name, Tier, Clade)

a - MW965.26 (1, C) b - SF162.LS (1, B) c - BaL.26 (1, B) d - SS1196.1 (1, B) e - DJ263.8 (1, A) f - 6535.3 (1, B)
b c d e f h i k b c d e f h i k
OD405nm
2 1
1 0.1 0.01
g - RHPA4259.7 (2, B) h - SC422661.8 (2, B) i - TRO.11 (2, B) j - YU2.DG (2, B) k - PVO.4 (3, B)
10-14 10-12 10-10 10-8 10-6 10-4 10-2 100
IC50 (< 0.001 g/ml) IC50 (0.001 - < 0.01 g/ml) IC50 (0.01 - < 0.1 g/ml) IC50 (0.1 - < 1 g/ml) IC50 (1 - 10 g/ml) IC50 (> 10 g/ml) IC50 > measured conc.
g/ml
0.01
0.1
48
0.001
mature mAbs
FWR-reverted mAbs
g/mlmAb
HIVstrains
The neutralizing activity of the reverted antibodies was tested on a panel of 11 viruses (Tier 1 to Tier 3) representing HIV-1 clades A, B, and C (Figures 2 and 3 and Table S2). The majority of the FWR-GL antibodies lost nearly all their neutralizing activity against the tested strains. Similar results were also obtained when we produced FWR-GL antibody versions of VRC01, 3BNC60, and 8ANC131 according to IMGT alignment (Data S1F and Figure S2C). 4E10, which is among the least potent antibodies of the bNAbs tested, was the exception to the rule in retaining binding to gp140YU2 as well as potency and breadth. PG16 retained most of its neutralizing breadth, but like the other bANbs, it lost potency by at least 10-fold after FWR reversion (geometric mean IC50 values increased from 0.95 to 9.69, Figure 2 and Table S2). 2G12 represents a special case for interpreting the effects of FWR residue reversion because the two Fabs of 2G12 IgG form a domain-swapped (Fab)2 unit that creates a single antigen-binding site for recognizing a constellation of host-derived high mannose carbohydrates on gp120 (Calarese et al., 2003) (Figure S3). FWR residues are critical for the Fab dimerization via domain swapping (Huber et al., 2010) as well as for formation of a novel carbohydrate-binding site at the VHVH interface (Calarese et al., 2003). To provide a general understanding of the relationship between direct FWR contacts and CDRH3 length on the degree of FWR somatic mutation, we analyzed these parameters for bNAbs for which antibody-antigen structures have been determined. No correlation was observed with CDRH3 length and degree of heavy chain, light chain, or total amount of FWR somatic mutation (Figure S4). Notably, however, the FWR contact surface area with antigen did correlate with the total number of amino acid changes in FWRs (Figure S4; p value = 0.0088; rho = 0.533). To determine whether the effects of FWR reversion on binding and neutralization were simply due to alterations in contact residues, we selectively reverted all mutated FWR residues except for the contact residues (FWR-GLCR+; Figure 3) in four bNAbs whose contact residues were known (VRC01, NIH45-46, 12A21, and 3BNC117) (Scheid et al., 2011; Wu et al., 2011; Zhou et al., 2010; unpublished data). Despite retaining their FWR contact residues, all of the partially reverted antibodies showed lower levels of binding by ELISA and SPR. Interestingly, the SPR-binding curves showed that loss of afnity appeared to be primarily due to an increased dissociation rate (off rate; kd) (Figure 3). We conclude that FWR mutations enhance the afnity of broadly neutralizing antibodies primarily by decreasing the dissociation rate. Most importantly, FWR-GLCR+ antibodies that retained somatically mutated FWR contact residues lost both neutralizing breadth and potency (Figure 3 and Table S2).
We conclude that FWR mutations in noncontact residues are essential for the binding, breadth, and potency of most broadly neutralizing anti-HIV-1 antibodies. Importance of a FWR Insertion in 3BNC60 Insertions and deletions are infrequent byproducts of somatic hypermutation that occur as a result of double-strand DNA breaks induced by deamination of neighboring cytidine residues by activation-induced cytidine deaminase (AID) (Pavri and Nussenzweig, 2011; Wilson et al., 1998). Nevertheless, several potent anti-CD4bs antibodies, including 3BNC60, contain insertions within FWRs that are acquired during somatic mutation (Scheid et al., 2011). Sequence analysis of the clonal relatives of 3BNC60 (clone RU01) revealed a correlation between the presence of this insertion and neutralizing activity (Figure 4). Superimposition of the 3BNC60 Fab structure onto the Fab portion of the NIH45-46/gp120 complex structure (Diskin et al., 2011) suggested that this insertion might enhance binding by interacting with the V1/V2 loop region that was truncated in the gp120 construct that was crystallized (Figure 5A) (similar results were found when superimposing the 3BNC60 Fab onto the 3BNC117 Fab, two nearly identical antibodies, in the 3BNC117/gp120 cocrystal structure [unpublished data]). To assess the effects of the insertion within FWR3 of the 3BNC60 heavy chain, we constructed a 3BNC60 mutant (3BNC60DI) in which the Trp-Asp-Phe-Asp insertion was removed (Figure 5B) and evaluated its neutralization activity against a panel of seven viruses chosen to include strains that were resistant to VRC01 but sensitive to 3BNC60 (Figure 5C). 3BNC60DI lost neutralization potency against all seven viruses. Adding the insertion to 3BNC55, a weaker variant of 3BNC60 isolated from the same donor (Scheid et al., 2011), increased the neutralization potency of 3BNC55+I compared to 3BNC55 against one viral strain (TRO.11) (Figure 5C). Addition of the insertion did not, however, restore the ability of a VRC01 plus insertion mutant to neutralize VRC01-resistant viruses (Data S1E and Table S2). Taken together, these results demonstrate that the FWR insertion is critical for the neutralization activity of 3BNC60, but that its ability to improve potency requires a precise recognition geometry that is not always found in other potent anti-CD4bs antibodies. Crystal Structure of a Partially Reverted Fab We previously noted a disruption in the canonical variable domain fold of the VH domain of 3BNC60 (Scheid et al., 2011); namely, the main chain hydrogen bonding pattern between strands C and C was disrupted by the presence of Pro61 (Kabat numbering position 60; IMGT position 68) located at the
Figure 2. Binding and Neutralization Activity of Mature and FWR-Reverted (FWR-GL) Broadly Neutralizing Antibodies
Evaluation of binding to gp140YU2 ELISA (left) of mature antibodies (green) and antibodies with FWRs reverted to germline (FWR-GL; blue, Data S1B). Panels on the right compare IC50 values for neutralization of Tier 1 (MW965.26, SF162.LS, Bal.26, SS1196.1, DJ263.8, and 6535.3) Tier 2 (RHPA4259.7, SC422661.8, TRO.11, and YU2.DG), and Tier 3 (PVO.4) viruses (Table S2). Only viruses are shown that were neutralized by the mature version of the antibody. Neutralization activity is color coded (blue arrow, <0.001 mg/ml; dark red, IC50 between 0.001 and <0.01 mg/ml; red, 0.01<0.1 mg/ml; orange, 0.1<1 mg/ml; light orange, 110 mg/ml; yellow, >10 mg/ml; white, IC50 was not achieved up to the tested concentration). Antibodies (NIH45-46 and 3BN60) with reverted CDR1/2 are shown in Figure S2A and HIV-1-reactive antibodies with limited neutralization are displayed in Figure S2B. The FWRs of VRC01, 3BNC60, and 3BNC131 were also reverted according to IMGT (Giudicelli et al., 2006) and results are shown in Figure S2C. A detailed illustration of the FWR mutations in 2G12 is shown in Figure S3.
ELISA
3 OD405nm 2 1
SPR
on-rate
1.0 normalized RU 0.8 0.6 0.4 0.2 0.0 0 100 200 300
off-rate
300 100 10
VRC01
mature
FWR-GL
FWR-GLCR+
g/ml
1 0.1 0.01
0.01
0.1
48
0.001
abcde f h i j k
abcde f h i j k
abcde f h i j k
3 OD405nm 2 1
NIH45-46
normalized RU
1.0 0.8
300 100 10
g/ml
0 100 200 300
0.6 0.4 0.2 0.0
1 0.1 0.01
0.01
0.1
48
0.001
abcde f h i j k
abcde f h i j k
abcde f h i j k
3 OD405nm 2 1
3BNC117
normalized RU
1.0
300 100 10
g/ml
0 100 200 300
0.8 0.6 0.4 0.2 0.0
1 0.1 0.01
0.01
0.1
48
0.001
abcde f g hi j k
abcde f g hi j k
abcde f g hi j k
3 OD405nm 2 1
12A21
normalized RU
1.0 0.8
BD
g/ml
0 100 200 300
300 100 10 1 0.1 0.01 0.001
0.6 0.4 0.2 0.0
0.01
0.1
g/mlmAb
mature mAbs
48
a b c e g h i j k
a b c e g h i j k
a b c e g h i j k
Time (sec)
Virus strains (Name, Tier, Clade)
HIVstrains
IC50 (< 0.001 g/ml) IC50 (0.001 - < 0.01 g/ml) IC50 (0.01 - < 0.1 g/ml) IC50 (0.1 - < 1 g/ml) IC50 (1 - 10 g/ml) IC50 (> 10 g/ml) IC50 > measured concentration
FWR-reverted mAbs FWR-reverted (except contact residues) mAbs
a - MW965.26 (1, C) b - SF162.LS (1, B) c - BaL.26 (1, B) d - SS1196.1 (1, B) e - DJ263.8 (1, A) f - 6535.3 (1, B)
g - RHPA4259.7 (2, B) h - SC422661.8 (2, B) i - TRO.11 (2, B) j - YU2.DG (2, B) k - PVO.4 (3, B)
Figure 3. Binding and Neutralization Activity of Mature, FWR-GL, and FWR-GLCR+ Broadly Neutralizing Antibodies
Evaluation of binding to gp140YU2 ELISA (left) and SPR (middle) of mature antibodies (green), antibodies with FWRs reverted to germline (FWR-GL, blue), and antibodies with germline-reverted FWRs except gp120-contacting residues (FWR-GLCR+; light blue; Data S1D). SPR results are shown for starting concentrations of 1 mM. BD (below detection; no binding of the antibody was observed). Panels on the right compare IC50 values for neutralization of Tier 1 to Tier 3 viruses (Table S2) as in Figure 2. Neutralization activity is color coded as indicated.
C-terminal end of strand C (Figure 6A). In contrast, the germline VH sequence and other potent anti-CD4bs antibodies contain an alanine at this position (Scheid et al., 2011) (Data S1C and S1D). A proline within a b strand cannot form a main chain hydrogen bond with a carbonyl oxygen in an adjacent b strand because it lacks a hydrogen atom attached to its mainchain nitrogen (Figure 6A). Nevertheless, this proline mutation is associated
with increased antibody potency among clonal members of the 3BNC60 family (Figure 4) (Scheid et al., 2011). Although classied by Kabat (Kabat et al., 1991) as part of CDRH2, residue 61 is within the C b strand of the Ig V domain fold and is classied as a FWR residue by IMGT (Lefranc et al., 1999) (Figure 1B, S5). Thus we were interested in its potential role in antigen recognition. In the structure of the free 3BNC60
74.5 38.5 40.3
WDFD Insertion + Pro61 Pro61

99.8
82.3 99.7
97.2 99.4
46.8 33.3 61.3 95.9 64.9 95.0
3BNC95HC 3BNC117HC 3BNC62HC 3BNC176HC 3BNC60HC 3BNC91HC 3BNC55HC 3BNC89HC 3BNC108HC 3BNC153HC 3BNC156HC 3BNC72HC 3BNC158HC 3BNC66HC 3BNC53HC 3BNC142HC 3BNC42HC 0
Potent Neutralization Intermediate Neutralization
Figure 4. Effects of a FWR Insertion and a C b Strand Proline in Clone RU01

Phylogenetic tree of the antibodies (Ig heavy chain) derived from the RU01 clone that members include 3BNC117 and 3BNC60 (Scheid et al., 2011). Antibodies that carry both the four amino acid insertion in FWR3 and the A61P somatic mutation are shown in red, antibodies with only the A61P mutation are shown in orange, and antibodies without either feature are shown in black. Bootstrap values (1,000 trials, seed = 111) of the phylogenetic tree are indicated. Structure of 3BNC117 IGVH in its gp120-bound conformation is shown in Figure S5.
Poor Neutralization
20
15
10
Amino Acid Substitutions
Fab, the region surrounding Pro61 is stabilized by a crystal contact (Figure S6) into a position that would clash with the CD4-binding loop of gp120 (Diskin et al., 2011; Scheid et al., 2011; Wu et al., 2011; Zhou et al., 2010). The potential clash with gp120 suggests that the region surrounding 3BNC60 Pro61 rearranges upon gp120 binding. Indeed, in the structure of a gp120 complex with the nearly identical antibody 3BNC117, the Fab exhibits a canonical VH domain structure in its gp120-bound conformation (unpublished data) (Figure 6A). In addition, a murine Fab structure including a proline at this position shows minimal disruption of the b sheet including strand C (Staneld et al., 1990) (Figure 6A). To address the unusual properties of the C strand in 3BNC60, crystal structure of 3BNC60P61A (pdb code we solved the 2.65 A 4GW4), a single amino acid revertant mutant form of 3BNC60 (Figure 6A and Table S3). Like other Fabs containing alanine at position 61, including VRC01 (Zhou et al., 2010), VRC03, VRCPG04 (Wu et al., 2011) and NIH45-46 (Diskin et al., 2011; Scheid et al., 2011), strand C of 3BNC60P61A Fab exhibited normal hydrogen bonding to the neighboring C strand (Figure 6A). Although the displaced region of strand C in the 3BNC60 Fab was involved in interactions with a crystallographic neighbor, perhaps stabilizing it in the conformation observed in the structure (Scheid et al., 2011), the corresponding region of strand C in the 3BNC60P61A Fab did not contact a crystallographic neighbor despite isomorphous packing interactions in the 3BNC60 and 3BNC60P61A crystals (Figure S6). To further evaluate the effects of the Pro to Ala substitution in strand C, we compared the thermal stability of the 3BNC60 and 3BNC60P61A Fabs (Figure 6B). Both proteins exhibited denaturation proles characteristic of two-state (native to denatured) unfolding; however, the 3BNC60P61A Fab showed increased thermal stability compared with the wild-type 3BNC60 Fab, as demonstrated by a higher transition midpoint (Tm) (Figure 6B). The functional consequences of the Pro to Ala substitution were evaluated by comparing the neutralization potencies of 3BNC60 and 3BNC60P61A IgGs against 19 representative strains sensitive to 3BNC60 (Figure 6C). The substitution affected the neutralizing activity of the antibody against seven of the strains (highlighted red; Figure 6C). We conclude that somatic mutation in the b sheet framework involving strand C is essential for the potency and breadth of 3BN60.
DISCUSSION The b sandwich structure of the immunoglobulin fold lends itself to a natural division into relatively structurally invariant b strand FWRs and the more structurally diverse loops connecting the b strands, three of which form the hypervariable CDRs (Figures 1A and 1B). Thus, mutations in FWRs are usually poorly tolerated and selected against, whereas mutations in the structurally diverse CDRs are well tolerated. Starting with the rst crystal structure of a Fab bound to a protein antigen (Amit et al., 1986), antibody-antigen complex structures have conrmed that residues within the CDR loops usually form the majority of contacts with the antigen. Therefore the primary role of FWR residues is to provide a scaffold for the antigen-contacting CDRs, as evidenced by the common practice of CDR grafting (Jones et al., 1986). However, we nd that in contrast to most antibodies, including HIV-1-reactive antibodies with limited neutralizing activity (Figures S1 and S2B), somatically mutated FWR residues are critical for the breadth and potency of broadly neutralizing anti-HIV-1 antibodies (Figure 2 are 3). Thus, the FWRs in these unusual antibodies serve an essential function beyond that of a scaffold for antigen-binding CDRs. Understanding the excessive somatic hypermutation found in bNAbs requires consideration of the process by which antibodies are mutated. B cells undergo somatic hypermutation in germinal centers during T-cell-dependent immune responses (Victora and Nussenzweig, 2012). The mutations are introduced by AID, which preferentially targets cytosines embedded in RGYW nucleotide sequences (in which R can be A or G, Y can be C or T, and W can be A or T) in antibody variable regions (Pavri and Nussenzweig, 2011). However, mutations are not limited to cytidine residues because error prone repair mechanisms also contribute to the repair of the initial lesions (Pavri and Nussenzweig, 2011; Peled et al., 2008). Thus, the mutation process is far more random than would occur if only RGYW cytidines were targeted. Mutations that enhance antibody afnity are rare, but they are positively selected in the germinal center as a result of increased antigen uptake and MHC-peptide presentation, which results in increased T-cell-mediated help (Victora and Nussenzweig, 2012). Mutations that increase antibody afnity can do so by increasing the on-rate or by decreasing the off rate. The on-rate
Figure 5. Analysis of 3BNC60 Insertion

(A) Superimposition of the structure of the VH domain of 3BNC60 (cyan) (Scheid et al., 2011) onto the NIH45-46 VH domain from cocrystal structure of NIH45-46 (gray) bound to gp120 (gold) (Diskin et al., 2011) highlighting the four residue insertion in FWR3 of 3BNC60 (cyan arrow) and a potential interaction between the insertion and the gp120 V1/V2 loop (note that the V1/V2 loop was truncated in the gp120 construct used for cocrystallization with NIH45-46 and VRC01) (Diskin et al., 2011; Zhou et al., 2010). (B) Alignment of the heavy chain FWR3 sequences of 3BNC60 (Scheid et al., 2011), 3BNC60 without the 3BNC60 insertion (3BNC60DI), 3BNC55 (Scheid et al., 2011), and 3BNC55 containing the 3BNC60 insertion (3BNC55+I; Data S1E). The four amino acid insertion in FWR3 of 3BNC60 is shown in cyan and the region grafted into 3BNC60 from 3BNC55 in order to delete the insertion without disrupting the structure is shown in gray. The amino acid changes to introduce the insertion into 3BNC55 are shown in blue. (C) In vitro neutralization data (IC50 values in mg/ml) comparing the potencies of the antibodies. The upper panel compares neutralization of selected viral strains by 3BNC60 and 3BNC60DI. The lower panel compares the neutralization by a less potent member of the RU01 clone (3BNC55), in which the FWR3 insertion of 3BNC60 was introduced (3BNC55+I). Reduced and increased neutralization activity of the engineered antibodies (3BNC60DI, 3BNC55+I) is highlighted red and green, respectively. Neutralization data on VRC01 with and without the 3BNC60 insertion is displayed in Table S2.
is diffusion limited, and the off rate is limited by speed of antigen internalization because once the antigen is in a degradative endosome it will be digested irrespective of its rate of release from the antibody (Batista and Neuberger, 1998; Foote and Milstein, 1991). Thus, naturally developing antibody afnities do not normally exceed an afnity of 1011 M. In humans, this degree of afnity maturation is usually achieved with an average of 1015 nucleotide mutations focused in the antigen contact residues in the CDR loops. Although somatic mutations occur throughout the variable region, mutations that alter the amino acid coding sequence of the antibody accumulate preferentially in CDRs in part because of higher level of degeneracy of the codons used in the FWRs (Jolly et al., 1996; Reynaud et al., 1995; Wagner et al., 1995). Any alterations in the FWRs are constrained by the fact that they must conserve the overall structure of the antibody because B cells that fail to express surface Ig are destined to die by apoptosis (Rajewsky, 1996). Indeed the FWRs appear to have evolved a nucleotide coding sequence that resists mutation in order to prevent changes in relatively invariant b strands that are required to scaffold the CDR loops (Jolly et al., 1996; Reynaud et al., 1995; Wagner et al., 1995). Thus, a large number of random nucleotide mutations would be required to alter the FWRs in a manner that optimizes the broadly neutral-
izing anti-HIV-1 antibodies while conserving essential structural elements. The mutated FWR residues in the broadly neutralizing antiHIV-1 antibodies contribute to binding and enhance breadth and potency in several different ways. In some cases, the substituted FWR residues directly contact the antigen; for example, crystal structures of the bNAbs VRC01 and NIH45-46 complexed with gp120 (Diskin et al., 2011; Zhou et al., 2010) reveal that FWR residue Arg71 in both VRC01 and NIH45-46 forms a salt bridge with gp120 residue Asp368. Thus, FWR mutation can serve to directly increase the binding interface by recruiting portions of the antibody that are not normally involved in antigen recognition. Moreover the ligand-bound structure of PGT128 (Pejchal et al., 2011) shows that the carbohydrate attached to Asn121gp120 interacts with residues within the C b strand of the antibody heavy chain (Trp56, Thr57, His59 and Lys64). Two of these residues represent somatic mutations. In other cases, FWR residues do not appear to directly contact the antigen, yet are required for potency. For example, the FWRs of PG16 do not contribute substantially to direct antigen contacts and interaction with the antigen is mainly mediated through a long CDR3 loop (unpublished data) (Figure S4). Somatic mutation of residues within the b sheet framework of the V domain can also indirectly affect the V domain structure so
Figure 6. Comparison of 3BNC60 and 3BNC60P61A Structures, Thermal Denaturation Proles, and Neutralizing Activities
(A) Left: 3BNC60P61A (Table S3; magenta, heavy chain; orange, light chain) was superimposed on the structure of 3BNC60 (cyan, heavy chain; red, light chain). The C b strand within FWR3 of the VH domain differs in conformation between the two structures. Middle/right: Close-up of the C and C b strands of 3BNC60P61A (magenta), 3BNC117 bound to go120 (yellow), 3BNC60 (cyan), and a murine Fab (green). The main chain atoms of the VH domain C and C b strands of 3BNC60P61A exhibit a typical hydrogen-bonding pattern for an antiparallel b sheet. Three of the ve inter-b strand hydrogen bonds in 3BNC60P61A are found in all three structures (yellow dashed lines), whereas 3BNC60 lacks two and the murine Fab/3BNC117 lacks one of the hydrogen bonds (green dashed lines in 3BNC60P61A). Cocrystallization of the 3BNC60-relative 3BNC117 with gp120 shows that Proline61 is accommodated without disrupting the C-C b sheet when 3BNC117 is bound to gp120. Overview of the packing in crystals of 3BNC60 and 3BNC60P61A is shown in Figure S6. (B) Thermal denaturation proles of the 3BNC60 and 3BNC60P61A Fabs monitored by the CD signal at 218 nm. Tms (indicated with arrows) were derived by estimating the half-point of the ellipticity change between the beginning and end of each transition. Errors bars indicate SD. (C) In vitro neutralization data (IC50 values in mg/ml) comparing the 3BNC60 and 3BNC60P61A IgGs for a panel of 19 viruses. Reduced neutralization activity is highlighted in red. An additional nine viral strains (T250-4, T278-50, 620345.c01, X2088_c9, 89-F1_2_25, 6540.v4.c1, CAP45.2.00.G3, 6545.v4.c1, and Du422.1) were resistant to both 3BNC60 and 3BNC60P61A. The 3BNC60P61A mutant was not signicantly more potent than 3BNC60 against any of the 28 strains tested.
as to facilitate HIV-1 antigen recognition, as exemplied by the Pro61 residue in the broadly neutralizing antibodies 3BNC60/ 3BNC117 (Figure 6). Residue 61 falls within the b sheet framework of the Ig V domain. Substitution of this proline to the germline alanine in 3BNC60 resulted in increased thermal stability, but
a loss of neutralization activity against some HIV-1 strains. The structural effects of the proline to alanine substitution were revealed by comparison of the crystal structures of the 3BNC60P61A and 3BNC60 Fabs: the VH domain b strand that was disrupted by the proline in the 3BNC60 structure was
restored to its canonical position in the mutant Fab structure (Figure 6A). In the 3BNC117/gp120 structure (unpublished data), the Pro61-containing b strand is shifted from its position in the free 3BNC60 structure (where it was likely stabilized into the observed conformation by crystal packing forces) in order to avoid clashing with the CD4-binding loop in gp120, as predicted previously for 3BNC60/gp120 complexes (Scheid et al., 2011) (Figure 6A). The lower thermal stability of 3BNC60 Fab compared with 3BNC60P61A Fab (Figure 6B) is consistent with exibility that would allow this sort of displacement. Superimposition of the 3BNC60P61A Fab into the 3BNC117/ gp120 cocrystal structure (unpublished data) suggests that an alanine would be accommodated equally as well as a proline at position 61, and the fact that the P61A mutant of 3BNC60 neutralizes some HIV-1 strains equally as well as wild-type 3BNC60 suggests that Pro61 does not make any direct contacts to HIV-1 gp120 that are critical for binding to gp120 and neutralization. Instead we speculate that in the case of 3BNC60 and 3BNC117, the ability to disrupt the CC portion of the CCCFG b sheet, which provides exibility for VH residues 60 66, is necessary in order to accommodate antigenic sequence heterogeneity in or near the gp120 V5 loop. Taken together, these results provide a counterintuitive example of a neutral or functionally favorable somatic mutation that decreases the Fab stability by disrupting the canonical VH domain fold, allowing exibility of the b sheet framework that is needed for optimal antigen binding and neutralization (Figure 6C). The high level of mutation found in broadly neutralizing antibodies would be difcult to explain in the context of an immune response to a conventional antigen and a single round of germinal center selection. However, HIV-1 differs from conventional antigens in that it presents the host with a continuously evolving target that is selected on the basis of its ability to evade the antibody response (Wei et al., 2003). We speculate that HIV-1 variants selected for their ability to evade antibodies due to lowered afnity will recall memory B cells to the germinal center for additional rounds of mutation and selection. Thus, iterative rounds of antibody mutation, selection, and viral escape would facilitate the accumulation of essential mutations in the FWR that conserve some key aspects of antibody structure while altering others to enhance anti-HIV-1 breadth and potency. In conclusion, our experiments suggest a molecular and a structural rationale for the requirement of high levels of somatic mutation found in broadly neutralizing antibodies and possibly for the observation that it takes several years for infected individuals to develop such antibodies. The high relevance of FWR mutations should be considered in the approach to designing an HIV-1 vaccine.
EXPERIMENTAL PROCEDURES Sequence Analysis of HIV-1-Reactive Antibodies Analysis of heavy and light chain gene segment usage, number of somatic mutations, and the presence of deletions or insertions was carried out using the NCBI IgBLAST software (http://www.ncbi.nlm.nih.gov/igblast/). CDRs and FWRs were designated according to the Kabat (Wu and Kabat, 1970) (Figures 1, Data S1AS1E, and Table S1) or IMGT numbering system (Figure 1B and Data S1F) using IgBLAST software. VRC01- and NIH45-46-FWR-GLCR+ antibodies were designed to carry reverted FWRs with unreverted contact
residues based on the crystal structure of a gp120-VRC01 complex (Zhou et al., 2010). 3BNC117- and 12A21-FWR-GLCR+-antibodies were based on the cocrystal structures as described by Peter Kwong and colleagues (PDB codes 4JPV and 4JPW, unpublished data). Any residue with a buried surface 2 was considered signicant. area (BSA) greater than 5 A Sequences of the RU01 clone (Scheid et al., 2011) were aligned using Clustal V from the DNAstar package using PAM 250 matrix. The phylogenetic tree was constructed using the DNAstar package, which employs a neighbor joining method. Bootstrap values (1,000 trials, seed = 111) are as indicated (Figure 4). Classication of neutralization potency (RU01 clone) was based on the neutralization activity. Antibodies were grouped into potent, intermediate, and poor neutralizers by taking into account the potency (IC50) against BaL.26, DJ263.8, 6535.3, RHPA4259.7, Tro.11, PVO.4, and YU2.DG as well as breadth (number of the tested strains neutralized) (Scheid et al., 2011). Cloning, Expression, and Purication of Immunoglobulins The variable regions of all mature, FWR-GL, FWR-GLCR+, and CDR1/2-GL antibodies were cloned into the same IgH, Igk, or Igl backbones encoding the constant domains of the antibodies. Reversions of the FWRs were introduced by overlapping PCR or by DNA synthesis (Integrated DNA Technologies). The sequence of 17b was obtained from the structure with CD4 (pdb code 1G9M). The P61A mutation was introduced into the 3BNC60 heavy chain gene by site-directed mutagenesis using QuikChange Site-Directed Mutagenesis Kit (Agilent Technologies). Because the region of the FWR3 insertion is highly mutated and the exact location of inserted amino acids is difcult to determine, surrounding regions from 3BNC55 and 3BNC60 were included in each construct (Figure 5B). Likewise, the residues surrounding the engrafted insertion in VRC01 were chosen based on the crystal structures of VRC01 and 3BNC60 (Data S1E) (Scheid et al., 2011; Zhou et al., 2010). Antibodies were expressed and puried as previously described (Mouquet et al., 2011) (Diskin et al., 2011). ELISA ELISAs for analyzing antibody binding to gp140YU2 (Figures 2, 3, and S2) was performed as previously described (Mouquet et al., 2011). Neutralization Assays Neutralizing activities of mature, FWR-GL, FWR-GLCR+, and CDR1/2-GL antibodies were determined using a TZM.bl assay as previously described (Diskin et al., 2011; Li et al., 2005; Monteori, 2005; Seaman et al., 2010). Briey, TZM.bl cells were infected with different Tier 1 to Tier 3 HIV-1-Env-pseudoviruses in the presence of serial dilutions of the antibodies tested. Antibodies with neutralizing activity inhibit the infection and a reduction of luciferase reporter gene expression can be measured 48 hr after infection. IC50 values were calculated based on the antibody concentration that resulted in a 50% reduction of relative luminescence units (RLU). Surface Plasmon Resonance Experiments were performed with a Biacore T200 (Biacore) as described previously (Diskin et al., 2011). Briey, YU-2 gp140 was primary aminecoupled on CM5 chips (Biacore) at an immobilization level of 1,000 RUs. IgG antibodies were injected over ow cells at 1 mM, at ow rates of 90 ml/min. The sensor surface was regenerated by a 50 s injection of 10 mM glycineHCl pH 2.5 at a ow rate of 90 ml/min. Crystallization and Structure of 3BNC60P61A Fab 3BNC60P61A Fab (pdb code 4GW4) was concentrated to 13.7 mg/ml in 20 mM Tris (pH 8.0), 150 mM sodium chloride, 0.02% sodium azide (TBS) buffer. Crystals of Fab 3BNC60P61A were obtained by mixing a protein solution at 13.7 mg/ml with 16% polyethylene glycol 6000, 0.1 M citric acid (pH 3.9), 0.15 M lithium sulfate monohydrate at 20 C. For cryoprotection, crystals were briey soaked in mother liquor solutions supplemented with 15% and subsequently with 30% ethylene glycol before ash cooling in liquid nitrogen. 3BNC60P61A Fab crystals grew in space group P21 (a = 64.6, b = 154.9, c = ; b = 109.7 ) and were isomorphous to 3BNC60 Fab crystals (Scheid 74.2 A
et al., 2011). Data were indexed, integrated, and scaled using XDS (Kabsch, 2010). We used Phaser (McCoy et al., 2007) to nd a molecular replacement solution for two Fabs per asymmetric unit (chains A and H and chains B and L for the heavy and light chains, respectively) using the 3BNC60 Fab structure (PDB code 3RPI) as a search model after omitting residues 5966 of the heavy chain. Residues 5966 were built into Fo-Fc difference electron density maps after a few rounds of iterative renement including noncrystallographic symmetry restraints using Phenix (Adams et al., 2010) and Coot (Emsley resolution atomic model for two 3BNC60P61A et al., 2010). The nal 2.65 A Fabs (Rwork = 21.5%; Rfree = 25.6%) includes 12,598 protein atoms (of which 6,198 are hydrogen atoms), 199 water molecules, and 28 ligand atoms (Nacetylglucosamine attached to Asn70 of the light chain) (Table S3). Using the numbering established for the 3BNC60 Fab structure (Scheid et al., 2011), the residues included in the nal model are 1132 and 141217 of chain H, 2132 and 141217 of chain A, 4198 of chain L, and 4199 of chain B. The rst glutamine of the 3BNC60P61A heavy chain was modeled as 5-pyrrolidone-2-carboxylic acid. A total of 94.5%, 5.3%, and 0.3% of the residues were in the favored, allowed and disallowed regions of the Ramachandran plot, respectively. Thermal Stability Comparisons Puried 3BNC60 and 3BNC60P61A Fabs were concentrated to 10 mM in 1 mM dithiothreitol and 5 mM sodium chloride for CD studies. Far UV wavelength scans (200 nm to 250 nm) were recorded in 1 nm increments using an averaging time of 5 s on an Aviv 62A DS spectropolarimeter. Both spectra showed a distinct negative signal at 218 nm, thus this wavelength was chosen for the thermal stability comparisons. Thermostability proles were obtained by monitoring the CD signal at 218 nm as the temperature was raised from 40 C to 95 C in 2 C increments, allowing 1 min of equilibration time after each temperature step and averaging the CD signal over 30 s of measurement. Transition midpoints (Tms) were obtained by estimating the half-point of the ellipticity change between the native and denatured states. SUPPLEMENTAL INFORMATION Supplemental Information includes Extended Experimental Procedures, six gures, three tables, and one data le, and can be found with this article online at http://dx.doi.org/10.1016/j.cell.2013.03.018. ACKNOWLEDGMENTS We thank Louise Scharf for her help on generating structural images and KaiHui Yao for technical assistance. We thank Paola Marchovecchio and Han Gan for preparation of 3BNC60P61A as well as the members of the Nussenzweig and Bjorkman laboratory for helpful discussions. We also thank R. Wyatt for the YU2-gp140 plasmid; Davide Corti and Antonio Lanzavecchia for the HJ16 plasmids, and Francine McCutchan, Beatrice Hahn, David Monteori, Michael Thomson, Ronald Swanstrom, Lynn Morris, Jerome Kim, Linqi Zhang, Dennis Ellenberger, and Carolyn Williamson for contributing HIV-1 envelope plasmids used in neutralization assays. F.K. was supported by the German Research Foundation (DFG, KL 2389/1-1), the Stavros Niarchos Foundation and the Robert Mapplethorpe Foundation. C.G. and R.-B.I. were supported by The German National Academic Foundation. I.G., M.P., T.Z., and P.D.K. were supported by intramural funding to the Vaccine Research Center, NIAID, NIH. M.S.S., P.J.B., and M.C.N. were supported by the Bill and Melinda Gates Foundation (M.S.S., Comprehensive Antibody Vaccine Immune Monitoring Consortium grant 1032144; P.J.B. and M.C.N., Collaboration for AIDS Vaccine Discovery grants 38660 [P.J.B.] and OPP1033115 [M.C.N.]). M.C.N is supported by the CHAVI-ID Award UM1AI100663 and The NIH grant AI081677. M.C.N. and P.J.B. are HHMI investigators. Received: April 6, 2012 Revised: January 7, 2013 Accepted: March 11, 2013 Published: March 28, 2013
REFERENCES czi, G., Chen, V.B., Davis, I.W., Echols, N., Adams, P.D., Afonine, P.V., Bunko Headd, J.J., Hung, L.W., Kapral, G.J., Grosse-Kunstleve, R.W., et al. (2010). PHENIX: a comprehensive Python-based system for macromolecular structure solution. Acta Crystallogr. D Biol. Crystallogr. 66, 213221. Amit, A.G., Mariuzza, R.A., Phillips, S.E., and Poljak, R.J. (1986). Three-dimensional structure of an antigen-antibody complex at 2.8 A resolution. Science 233, 747753. Amzel, L.M., and Poljak, R.J. (1979). Three-dimensional structure of immunoglobulins. Annu. Rev. Biochem. 48, 961997. Batista, F.D., and Neuberger, M.S. (1998). Afnity dependence of the B cell response to antigen: a threshold, a ceiling, and the importance of off-rate. Immunity 8, 751759. Buchacher, A., Predl, R., Strutzenberger, K., Steinfellner, W., Trkola, A., Purtscher, M., Gruber, G., Tauer, C., Steindl, F., Jungbauer, A., et al. (1994). Generation of human monoclonal antibodies against HIV-1 proteins; electrofusion and Epstein-Barr virus transformation for peripheral blood lymphocyte immortalization. AIDS Res. Hum. Retroviruses 10, 359369. Calarese, D.A., Scanlan, C.N., Zwick, M.B., Deechongkit, S., Mimura, Y., Kunert, R., Zhu, P., Wormald, M.R., Staneld, R.L., Roux, K.H., et al. (2003). Antibody domain exchange is an immunological solution to carbohydrate cluster recognition. Science 300, 20652071. Corti, D., Langedijk, J.P., Hinz, A., Seaman, M.S., Vanzetta, F., FernandezRodriguez, B.M., Silacci, C., Pinna, D., Jarrossay, D., Balla-Jhagjhoorsingh, S., et al. (2010). Analysis of memory B cell responses and isolation of novel monoclonal antibodies with neutralizing breadth from HIV-1-infected individuals. PLoS ONE 5, e8805. Diskin, R., Scheid, J.F., Marcovecchio, P.M., West, A.P., Jr., Klein, F., Gao, H., Gnanapragasam, P.N., Abadir, A., Seaman, M.S., Nussenzweig, M.C., and Bjorkman, P.J. (2011). Increasing the potency and breadth of an HIV antibody by using structure-based rational design. Science 334, 12891293. Doria-Rose, N.A., Klein, R.M., Daniels, M.G., ODell, S., Nason, M., Lapedes, A., Bhattacharya, T., Migueles, S.A., Wyatt, R.T., Korber, B.T., et al. (2010). Breadth of human immunodeciency virus-specic neutralizing activity in sera: clustering analysis and association with clinical variables. J. Virol. 84, 16311636. Emsley, P., Lohkamp, B., Scott, W.G., and Cowtan, K. (2010). Features and development of Coot. Acta Crystallogr. D Biol. Crystallogr. 66, 486501. Foote, J., and Milstein, C. (1991). Kinetic maturation of an immune response. Nature 352, 530532. Giudicelli, V., Duroux, P., Ginestoux, C., Folch, G., Jabado-Michaloud, J., Chaume, D., and Lefranc, M.P. (2006). IMGT/LIGM-DB, the IMGT comprehensive database of immunoglobulin and T cell receptor nucleotide sequences. Nucleic Acids Res. 34(Database issue), D781D784. Gorny, M.K., Xu, J.Y., Karwowska, S., Buchbinder, A., and Zolla-Pazner, S. (1993). Repertoire of neutralizing human monoclonal antibodies specic for the V3 domain of HIV-1 gp120. J. Immunol. 150, 635643. Huang, J., Ofek, G., Laub, L., Louder, M.K., Doria-Rose, N.A., Longo, N.S., Imamichi, H., Bailer, R.T., Chakrabarti, B., Sharma, S.K., et al. (2012). Broad and potent neutralization of HIV-1 by a gp41-specic human antibody. Nature 491, 406412. Huber, M., Le, K.M., Doores, K.J., Fulton, Z., Staneld, R.L., Wilson, I.A., and Burton, D.R. (2010). Very few substitutions in a germ line antibody are required to initiate signicant domain exchange. J. Virol. 84, 1070010707. Jolly, C.J., Wagner, S.D., Rada, C., Klix, N., Milstein, C., and Neuberger, M.S. (1996). The targeting of somatic hypermutation. Semin. Immunol. 8, 159168. Jones, P.T., Dear, P.H., Foote, J., Neuberger, M.S., and Winter, G. (1986). Replacing the complementarity-determining regions in a human antibody with those from a mouse. Nature 321, 522525. Kabat, E., Wu, T.T., Reid-Miller, M., Perry, H.M., Gottesman, K.S., and Foeller, C. (1991). Sequences of Proteins of Immunological Interest (Bethesda, MD: US Department of Health and Human Services, National Institutes of Health).
Kabsch, W. (2010). Integration, scaling, space-group assignment and postrenement. Acta Crystallogr. D Biol. Crystallogr. 66, 133144. Klein, F., Gaebler, C., Mouquet, H., Sather, D.N., Lehmann, C., Scheid, J.F., Kraft, Z., Liu, Y., Pietzsch, J., Hurley, A., et al. (2012a). Broad neutralization by a combination of antibodies recognizing the CD4 binding site and a new conformational epitope on the HIV-1 envelope protein. J. Exp. Med. 209, 14691479. Klein, F., Halper-Stromberg, A., Horwitz, J.A., Gruell, H., Scheid, J.F., Bournazos, S., Mouquet, H., Spatz, L.A., Diskin, R., Abadir, A., et al. (2012b). HIV therapy by a combination of broadly neutralizing antibodies in humanized mice. Nature 492, 118122. ller, W., Bontrop, R., Lefranc, M.P., Giudicelli, V., Ginestoux, C., Bodmer, J., Mu , V., and Chaume, D. (1999). IMGT, the internaLemaitre, M., Malik, A., Barbie tional ImMunoGeneTics database. Nucleic Acids Res. 27, 209212. Li, M., Gao, F., Mascola, J.R., Stamatatos, L., Polonis, V.R., Koutsoukos, M., Voss, G., Goepfert, P., Gilbert, P., Greene, K.M., et al. (2005). Human immunodeciency virus type 1 env clones from acute and early subtype B infections for standardized assessments of vaccine-elicited neutralizing antibodies. J. Virol. 79, 1010810125. Mascola, J.R., Stiegler, G., VanCott, T.C., Katinger, H., Carpenter, C.B., Hanson, C.E., Beary, H., Hayes, D., Frankel, S.S., Birx, D.L., and Lewis, M.G. (2000). Protection of macaques against vaginal transmission of a pathogenic HIV-1/SIV chimeric virus by passive infusion of neutralizing antibodies. Nat. Med. 6, 207210. McCoy, A.J., Grosse-Kunstleve, R.W., Adams, P.D., Winn, M.D., Storoni, L.C., and Read, R.J. (2007). Phaser crystallographic software. J. Appl. Cryst. 40, 658674. Mikell, I., Sather, D.N., Kalams, S.A., Altfeld, M., Alter, G., and Stamatatos, L. (2011). Characteristics of the earliest cross-neutralizing antibody response to HIV-1. PLoS Pathog. 7, e1001251. Moldt, B., Rakasz, E.G., Schultz, N., Chan-Hui, P.Y., Swiderek, K., Weisgrau, K.L., Piaskowski, S.M., Bergman, Z., Watkins, D.I., Poignard, P., and Burton, D.R. (2012). Highly potent HIV-specic antibody neutralization in vitro translates into effective protection against mucosal SHIV challenge in vivo. Proc. Natl. Acad. Sci. USA 109, 1892118925. Monteori, D.C. (2005). Evaluating neutralizing antibodies against HIV, SIV, and SHIV in luciferase reporter gene assays. Curr. Protoc. Immunol., Chapter 12, Unit 12 11. Morris, L., Chen, X., Alam, M., Tomaras, G., Zhang, R., Marshall, D.J., Chen, B., Parks, R., Foulger, A., Jaeger, F., et al. (2011). Isolation of a human antiHIV gp41 membrane proximal region neutralizing antibody by antigen-specic single B cell sorting. PLoS ONE 6, e23532. Mouquet, H., Scheid, J.F., Zoller, M.J., Krogsgaard, M., Ott, R.G., Shukair, S., Artyomov, M.N., Pietzsch, J., Connors, M., Pereyra, F., et al. (2010). Polyreactivity increases the apparent afnity of anti-HIV antibodies by heteroligation. Nature 467, 591595. Mouquet, H., Klein, F., Scheid, J.F., Warncke, M., Pietzsch, J., Oliveira, T.Y., Velinzon, K., Seaman, M.S., and Nussenzweig, M.C. (2011). Memory B cell antibodies to HIV-1 gp140 cloned from individuals infected with clade A and B viruses. PLoS ONE 6, e24078. Mouquet, H., Scharf, L., Euler, Z., Liu, Y., Eden, C., Scheid, J.F., Halper-Stromberg, A., Gnanapragasam, P.N., Spencer, D.I., Seaman, M.S., et al. (2012). Complex-type N-glycan recognition by potent broadly neutralizing HIV antibodies. Proc. Natl. Acad. Sci. USA 109, E3268E3277. ker, Muster, T., Steindl, F., Purtscher, M., Trkola, A., Klima, A., Himmler, G., Ru F., and Katinger, H. (1993). A conserved neutralizing epitope on gp41 of human immunodeciency virus type 1. J. Virol. 67, 66426647. Pavri, R., and Nussenzweig, M.C. (2011). AID targeting in antibody diversity. Adv. Immunol. 110, 126. Pejchal, R., Doores, K.J., Walker, L.M., Khayat, R., Huang, P.S., Wang, S.K., Staneld, R.L., Julien, J.P., Ramos, A., Crispin, M., et al. (2011). A potent and broad neutralizing antibody recognizes and penetrates the HIV glycan shield. Science 334, 10971103.
Peled, J.U., Kuang, F.L., Iglesias-Ussel, M.D., Roa, S., Kalis, S.L., Goodman, M.F., and Scharff, M.D. (2008). The biochemistry of somatic hypermutation. Annu. Rev. Immunol. 26, 481511. Pietzsch, J., Scheid, J.F., Mouquet, H., Klein, F., Seaman, M.S., Jankovic, M., Corti, D., Lanzavecchia, A., and Nussenzweig, M.C. (2010). Human anti-HIVneutralizing antibodies frequently target a conserved epitope essential for viral tness. J. Exp. Med. 207, 19952002. Rajewsky, K. (1996). Clonal selection and learning in the antibody system. Nature 381, 751758. Reynaud, C.A., Garcia, C., Hein, W.R., and Weill, J.C. (1995). Hypermutation generating the sheep immunoglobulin repertoire is an antigen-independent process. Cell 80, 115125. Scheid, J.F., Mouquet, H., Feldhahn, N., Seaman, M.S., Velinzon, K., Pietzsch, J., Ott, R.G., Anthony, R.M., Zebroski, H., Hurley, A., et al. (2009a). Broad diversity of neutralizing antibodies isolated from memory B cells in HIV-infected individuals. Nature 458, 636640. Scheid, J.F., Mouquet, H., Feldhahn, N., Walker, B.D., Pereyra, F., Cutrell, E., Seaman, M.S., Mascola, J.R., Wyatt, R.T., Wardemann, H., and Nussenzweig, M.C. (2009b). A method for identication of HIV gp140 binding memory B cells in human blood. J. Immunol. Methods 343, 6567. Scheid, J.F., Mouquet, H., Ueberheide, B., Diskin, R., Klein, F., Oliveira, T.Y., Pietzsch, J., Fenyo, D., Abadir, A., Velinzon, K., et al. (2011). Sequence and structural convergence of broad and potent HIV antibodies that mimic CD4 binding. Science 333, 16331637. Seaman, M.S., Janes, H., Hawkins, N., Grandpre, L.E., Devoy, C., Giri, A., Coffey, R.T., Harris, L., Wood, B., Daniels, M.G., et al. (2010). Tiered categorization of a diverse panel of HIV-1 Env pseudoviruses for assessment of neutralizing antibodies. J. Virol. 84, 14391452. Shibata, R., Igarashi, T., Haigwood, N., Buckler-White, A., Ogert, R., Ross, W., Willey, R., Cho, M.W., and Martin, M.A. (1999). Neutralizing antibody directed against the HIV-1 envelope glycoprotein can completely block HIV-1/SIV chimeric virus infections of macaque monkeys. Nat. Med. 5, 204210. Simek, M.D., Rida, W., Priddy, F.H., Pung, P., Carrow, E., Laufer, D.S., Lehrman, J.K., Boaz, M., Tarragona-Fiol, T., Miiro, G., et al. (2009). Human immunodeciency virus type 1 elite neutralizers: individuals with broad and potent neutralizing activity identied by using a high-throughput neutralization assay together with an analytical selection algorithm. J. Virol. 83, 73377348. Staneld, R.L., Fieser, T.M., Lerner, R.A., and Wilson, I.A. (1990). Crystal structures of an antibody to a peptide and its complex with peptide antigen at 2.8 A. Science 248, 712719. Thali, M., Moore, J.P., Furman, C., Charles, M., Ho, D.D., Robinson, J., and Sodroski, J. (1993). Characterization of conserved human immunodeciency virus type 1 gp120 neutralization epitopes exposed upon gp120-CD4 binding. J. Virol. 67, 39783988. Tiller, T., Tsuiji, M., Yurasov, S., Velinzon, K., Nussenzweig, M.C., and Wardemann, H. (2007). Autoreactivity in human IgG+ memory B cells. Immunity 26, 205213. Trkola, A., Purtscher, M., Muster, T., Ballaun, C., Buchacher, A., Sullivan, N., Srinivasan, K., Sodroski, J., Moore, J.P., and Katinger, H. (1996). Human monoclonal antibody 2G12 denes a distinctive neutralization epitope on the gp120 glycoprotein of human immunodeciency virus type 1. J. Virol. 70, 11001108. Victora, G.D., and Nussenzweig, M.C. (2012). Germinal centers. Annu. Rev. Immunol. 30, 429457. Wagner, S.D., Milstein, C., and Neuberger, M.S. (1995). Codon bias targets mutation. Nature 376, 732. Walker, L.M., Phogat, S.K., Chan-Hui, P.Y., Wagner, D., Phung, P., Goss, J.L., Wrin, T., Simek, M.D., Fling, S., Mitcham, J.L., et al.; Protocol G Principal Investigators. (2009). Broad and potent neutralizing antibodies from an African donor reveal a new HIV-1 vaccine target. Science 326, 285289. Walker, L.M., Huber, M., Doores, K.J., Falkowska, E., Pejchal, R., Julien, J.P., Wang, S.K., Ramos, A., Chan-Hui, P.Y., Moyle, M., et al.; Protocol G Principal
Investigators. (2011a). Broad neutralization coverage of HIV by multiple highly potent antibodies. Nature 477, 466470. Walker, L.M., Sok, D., Nishimura, Y., Donau, O., Sadjadpour, R., Gautam, R., Shingai, M., Pejchal, R., Ramos, A., Simek, M.D., et al. (2011b). Rapid development of glycan-specic, broad, and potent anti-HIV-1 gp120 neutralizing antibodies in an R5 SIV/HIV chimeric virus infected macaque. Proc. Natl. Acad. Sci. USA 108, 2012520129. Wei, X., Decker, J.M., Wang, S., Hui, H., Kappes, J.C., Wu, X., Salazar-Gonzalez, J.F., Salazar, M.G., Kilby, J.M., Saag, M.S., et al. (2003). Antibody neutralization and escape by HIV-1. Nature 422, 307312. Wilson, P.C., de Bouteiller, O., Liu, Y.J., Potter, K., Banchereau, J., Capra, J.D., and Pascual, V. (1998). Somatic hypermutation introduces insertions and deletions into immunoglobulin V genes. J. Exp. Med. 187, 5970. Wu, T.T., and Kabat, E.A. (1970). An analysis of the sequences of the variable regions of Bence Jones proteins and myeloma light chains and their implications for antibody complementarity. J. Exp. Med. 132, 211250. Wu, X., Yang, Z.Y., Li, Y., Hogerkorp, C.M., Schief, W.R., Seaman, M.S., Zhou, T., Schmidt, S.D., Wu, L., Xu, L., et al. (2010). Rational design of envelope
identies broadly neutralizing human monoclonal antibodies to HIV-1. Science 329, 856861. Wu, X., Zhou, T., Zhu, J., Zhang, B., Georgiev, I., Wang, C., Chen, X., Longo, N.S., Louder, M., McKee, K., et al.; NISC Comparative Sequencing Program. (2011). Focused evolution of HIV-1 neutralizing antibodies revealed by structures and deep sequencing. Science 333, 15931602. Xiao, X., Chen, W., Feng, Y., and Dimitrov, D.S. (2009a). Maturation Pathways of Cross-Reactive HIV-1 Neutralizing Antibodies. Viruses 1, 802817. Xiao, X., Chen, W., Feng, Y., Zhu, Z., Prabakaran, P., Wang, Y., Zhang, M.Y., Longo, N.S., and Dimitrov, D.S. (2009b). Germline-like predecessors of broadly neutralizing antibodies lack measurable binding to HIV-1 envelope glycoproteins: implications for evasion of immune responses and design of vaccine immunogens. Biochem. Biophys. Res. Commun. 390, 404409. Zhou, T., Georgiev, I., Wu, X., Yang, Z.Y., Dai, K., Finzi, A., Kwon, Y.D., Scheid, J.F., Shi, W., Xu, L., et al. (2010). Structural basis for broad and potent neutralization of HIV-1 by antibody VRC01. Science 329, 811817.
Glioblastoma Stem Cells Generate Vascular Pericytes to Support Vessel Function and Tumor Growth
Lin Cheng,1,7 Zhi Huang,1,7 Wenchao Zhou,1 Qiulian Wu,1 Shannon Donnola,1 James K. Liu,2 Xiaoguang Fang,1 Andrew E. Sloan,3 Yubin Mao,4 Justin D. Lathia,1 Wang Min,5 Roger E. McLendon,6 Jeremy N. Rich,1 and Shideng Bao1,*
of Stem Cell Biology and Regenerative Medicine, Lerner Research Institute of Neurosurgery Cleveland Clinic, Cleveland, OH 44195, USA 3Brain Tumor and Neuro-Oncology Center, University Hospitals, Case Western Reserve University, Cleveland, OH 44106, USA 4Department of Pathology, Medical College of Xiamen University, Xiamen, Fujian 361005, China 5Department of Pathology, Yale University School of Medicine, New Haven, CT 06520, USA 6Department of Pathology, Duke University Medical Center, Durham, NC 27710, USA 7These authors contributed equally to this work *Correspondence: baos@ccf.org http://dx.doi.org/10.1016/j.cell.2013.02.021
SUMMARY
Glioblastomas (GBMs) are highly vascular and lethal brain tumors that display cellular hierarchies containing self-renewing tumorigenic glioma stem cells (GSCs). Because GSCs often reside in perivascular niches and may undergo mesenchymal differentiation, we interrogated GSC potential to generate vascular pericytes. Here, we show that GSCs give rise to pericytes to support vessel function and tumor growth. In vivo cell lineage tracing with constitutive and lineage-specic uorescent reporters demonstrated that GSCs generate the majority of vascular pericytes. Selective elimination of GSC-derived pericytes disrupts the neovasculature and potently inhibits tumor growth. Analysis of human GBM specimens showed that most pericytes are derived from neoplastic cells. GSCs are recruited toward endothelial cells via the SDF-1/CXCR4 axis and are induced to become pericytes predominantly by transforming growth factor b. Thus, GSCs contribute to vascular pericytes that may actively remodel perivascular niches. Therapeutic targeting of GSC-derived pericytes may effectively block tumor progression and improve antiangiogenic therapy.
INTRODUCTION Glioblastomas (GBMs) are fatal tumors with orid vascularization that correlates with tumor malignancy and clinical prognosis (Norden et al., 2009). Targeting endothelial cells (ECs) has been a major focus of antiangiogenic therapeutics, although tumor vessels consist of two distinct but interdependent cellular compartments: ECs and pericytes (Bergers and Song, 2005;
Carmeliet and Jain, 2011). However, most current therapies targeting ECs are not curative and may transform tumor growth ` ezpattern toward a more invasive phenotype in GBMs (Pa Ribes et al., 2009), suggesting that targeting ECs alone is not sufcient for effective tumor control. Therefore, further insights into tumor vascular development and maintenance have direct translational implications. Vascular pericytes play critical roles in various physiological contexts, including support of vascular structure and function, maintenance of blood-brain barrier, facilitation of vessel maturation, and initiation of vessel sprouting (Armulik et al., 2010; Bell et al., 2010; Bergers and Song, 2005; Winkler et al., 2011). Pericytes and ECs communicate with each other by direct physical contact and reciprocal paracrine signaling to maintain vessel integrity and function (Franco et al., 2011; Carmeliet and Jain, 2011; Song et al., 2005). Altered association between pericytes and ECs has been shown in tumor vessels (Carmeliet and Jain, 2011; Winkler et al., 2011). Tumor vessels with less pericyte coverage appear more vulnerable to radiation and chemotherapy, suggesting that pericytes are critical to protect ECs and may promote therapeutic resistance (Bergers et al., 2003; Franco et al., 2011). When therapies target ECs in tumors, the pericyte network often maintains a functional core of pre-existing blood vessels (Carmeliet and Jain, 2011). The tumor vasculature frequently exhibits structural and functional abnormality with irregular pericytes on endothelial tubules. The pericyte-EC interaction also differs substantially between tumors and normal tissues (Morikawa et al., 2002; Winkler et al., 2011). However, the mechanisms underlying the abnormality and difference are poorly understood. To better understand the vascular development and maintenance in tumors and lay the foundation for improved targeting therapy, it is essential to determine the interplay between cancer cells and vascular compartments. GBMs display remarkable cellular hierarchies with tumorigenic glioma stem cells (GSCs) at the apex (Bao et al., 2006a; Calabrese et al., 2007; Zhou et al., 2009), although the cancer
stem cell (CSC) model remains controversial for some tumor types (Magee et al., 2012). We previously demonstrated that GSCs promote tumor angiogenesis through elevated expression of vascular endothelial growth factor (VEGF) (Bao et al., 2006b). This study has been extended by others (Ehtesham et al., 2009; Folkins et al., 2009). GSCs are often located in perivascular niches and interact with ECs in a bidirectional manner (Bao et al., 2006b; Calabrese et al., 2007). Within this context, there was an excitement generated by reports suggesting that GSCs may transdifferentiate into ECs (Ricci-Vitiani et al., 2010; Soda et al., 2011; Wang et al., 2010). These reports have been controversial because the frequency of GSC-EC conversion was not dened, and ECs do not contain cancer genetic alterations in human GBMs (Kulla et al., 2003; Rodriguez et al., 2012). Because pericytes are physically proximal to ECs on vessels, distinguishing ECs and pericytes by location alone poses a challenge. A complementary or competing hypothesis would be a lineage commitment of GSCs to vascular pericytes. There are important reasons to consider GSCs as potential pericyte progenitors. GSCs have the ability to undergo mesenchymal differentiation (deCarvalho et al., 2010; Ricci-Vitiani et al., 2008). GSCs share properties with neural stem cells (NSCs) that display the potential to transdifferentiate into pericytes (Ii et al., 2009; Morishita et al., 2007). Further, pericytes are similar to mesenchymal stem cells (MSCs) (Crisan et al., 2008). Thus, we investigated the potential of GSCs to generate vascular pericytes and contribute to the remodeling of perivascular niches and determined the signicance of GSC-derived pericytes (G-pericytes) in maintaining functional vessels to support GBM tumor growth. RESULTS GSCs Are Able to Assume a Pericyte Lineage In Vitro To investigate a potential lineage link between GSCs and pericytes, we initially examined the capacity of GSCs to differentiate into pericytes in vitro. GSCs were isolated from GBM tumors and validated through functional assays (self-renewal, multipotency, and tumor formation) as previously described (Bao et al., 2006a; Guryanova et al., 2011). Immunouorescent (IF) staining of freshly sorted GSCs from primary GBMs and the GSC-generated tumorspheres demonstrated SOX2 expression but complete absence of the pericyte markers a smooth muscle actin (a-SMA) and NG2 (Figures S1A and S1B available online), supporting a lack of contamination of GSC populations by pericytes. After GSCs or tumorspheres were induced for differentiation, the differentiated cells contained a fraction (4%11%) of cells expressing multiple pericyte markers (a-SMA, NG2, CD248, and CD146) (Figures S1CS1E). To further determine GSC ability to assume a pericyte lineage, we examined the cellular fate of
single GSC-derived tumorsphere that did not contain any cell expressing pericyte markers (Figure 1A). Upon differentiation, cells derived from the single GSC-derived tumorsphere contained a fraction of cells expressing pericyte markers (Figure 1B). To rule out potential contamination of host-derived pericyte progenitors in xenograft-derived GSCs, we performed secondary sorting of enriched GSCs with positive selection for the human cell-specic surface antigen TRA-1-85 and negative selection for the pericyte marker CD146. We conrmed that the single GSC-generated spheres derived from the resorted GSCs (SOX2+) did not contain any cell expressing pericyte markers (Figure S1F), whereas differentiated cells derived from the single GSC-derived sphere contained pericyte-marker-expressing cells (Figure S1G). These pericyte-marker-positive cells also expressed the human-cell-specic nuclear antigen NuMA (Figure S1G), conrming that these pericytes were derived from human GSCs, but not from murine pericytes or their progenitors. Collectively, these data demonstrate that GSCs have the capacity to assume a pericyte lineage in vitro. GSCs Give Rise to Vascular Pericytes in GBM Xenografts In Vivo To extend the lineage analysis of GSCs in vivo, we examined the origin of pericytes in GBM xenografts and found that pericytes (CD146+CD248+; 2.63%6.14% of total cells) sorted from the xenografts were largely positive for human NuMA and TRA-185 (Figures S1H, S1I, and S1L). In contrast, puried ECs (CD31+CD105+) from GBM xenografts were completely negative for human NuMA and TRA-1-85 (Figures S1JS1L). We then performed a lineage tracing study by transducing GSCs with GFP constitutive expression and implanted the GSCs orthotopically to establish xenografts. Tumor sections of the xenografts derived from the green uorescent protein (GFP)-labeled GSCs were immunostained for an EC marker (CD31) and several pericyte markers (a-SMA, Desmin, NG2, CD146, CD248, Ang1, CD13, and platelet-derived growth factor receptor b [PDGFRb]) because these pericyte markers are expressed in normal brain and primary GBMs (Figures S2AS2C). No tumor showed GFP-positive ECs, but most tumor vessels were adorned with GFP-positive cells with typical pericytic location and morphology on the vascular external surface (Figures 1C and 1D). IF analyses of pericyte markers further conrmed that expression of pericyte markers (Desmin, a-SMA, NG2, PDGFRb, CD248, and CD146) overlapped with GFP in the majority (mean 78%, range 57%89%) of pericytes (Figures 2A and 2B; Figures S2D and S2E), indicating that the majority of vascular pericytes were derived from GSCs. We validated this result in 21 GBM xenografts using GFP-labeled GSCs isolated from 12 primary GBMs and 9 GBM xenografts, suggesting that the contribution of GSCs to pericytes is a common event during
Figure 1. GSCs Have the Potential to Assume a Pericyte Lineage

(A) IF staining of SOX2 (a GSC marker) and a-SMA (a pericyte marker) in the single GSC-derived tumorsphere. Nuclei were stained with DAPI. (B) IF staining of a-SMA and NuMA (a human cell-specic nuclear antigen) in differentiated cells derived from the single GSC-derived tumorsphere. (C and D) In vivo lineage tracing of GSCs with GFP constitutive expression. Sections of GBM tumors derived from the GFP-labeled GSCs (D456 or CCF2170) were immunostained for CD31 to mark ECs and counterstained with DAPI. Arrows indicate GFP+ cells with pericytic location. (E) IF staining of CD31 in peritumoral brain adjacent to the GBM tumor derived from GFP-labeled GSCs (CCF2170). All scale bars represent 25 mm. See also Figure S1.
Figure 2. GSCs Generate Pericytes Expressing Specic Markers In Vivo

(A and B) In vivo lineage tracing of GSCs and IF staining of pericyte marker Desmin (A) or a-SMA (B) in GBM tumors derived from GFP-labeled GSCs (D456). Quantications show fractions of G-pericytes (GFP+ and Desmin+/a-SMA+). (C) IF staining of Desmin in peritumoral brain adjacent to GBM tumor derived from GFP-labeled GSCs (CCF1468). A vessel containing G-pericytes (Desmin+ and GFP+) in peritumoral brain was marked (#) and enlarged. All scale bars represent 20 mm. The error bars represent SD. See also Figure S2.
GBM growth. Notably, a minor fraction (mean 22%, range 11% 43%) of vascular pericytes in the GBM xenografts did not overlap with GFP expression, indicating that these pericytes were host derived. Most tumor vessels had a mixture of GSC- and hostderived pericytes (Figure 2A). Taken together, these data demonstrate that GSCs have the capacity to generate the majority of vascular pericytes in GBM xenografts. Peritumoral Brain Vessels Contain G-Pericytes Because GBMs commonly invade into normal brain, we examined whether GSCs contribute to vascular pericytes in peritumoral brain. We found that a subset of vessels in peritumoral brain adjacent to the GFP-labeled GSC xenograft also contain GFP-positive pericytes (Figure 1E). IF analyses validated a fraction of vessels coexpressing pericyte markers and GFP and in brain tissue near the GFP tumor (Figure 2C). These data indicate that GSCs can also give rise to pericytes in the peritumoral brain. Notably, G-pericytes (GFP+) were detectable not only in peritumoral brain but also in tumor-free brain up to 0.86 mm distant from the tumor edge, suggesting that GSCs were recruited by ECs in the peritumoral brain to generate pericytes. Thus, GSCs also generate vascular pericytes in the peritumoral brain. Validation of G-Pericytes by Lineage-Specic Fluorescent Reporters To provide direct evidence validating GSC capacity to generate pericytes in vivo, we performed in vivo cell lineage tracing of GSCs with a pericyte marker (Desmin or a-SMA) promoterdriven expression of GFP or mCherry, which served as uorescent reporters of pericyte lineage. We cloned the human Desmin promoter (Li and Paulin, 1991) and a-SMA core promoter (Keogh et al., 1999; Nakano et al., 1991) and then generated lentiviral constructs for the Desmin promoter-driven GFP expression (DesPro-GFP) or a-SMA promoter-driven mCherry expression (aSMAPro-mCherry). We conrmed that the cloned Desmin and a-SMA promoters were functional and pericyte specic because GFP or mCherry expression occurred specically in human brain vascular pericytes (HBVPs) (Figure S3A, left). We then implanted DesPro-GFPtransduced GSCs into mouse brains and examined tumor vessels by IF analysis. DesPro-driven GFP expression specically marked perivascular cells that expressed pericyte markers, including Desmin, NG2, PDGFRb, CD248, Ang1, and CD13 (Figures 3A3C), validating that GSCs generated vascular pericytes in the GBM xenografts. The G-pericytes also expressed the gap junction protein connexin45 (Cx45) that is often localized at pericyte-EC contacts (Figures 3D and 3E). Notably, GFP-positive cells were mainly located in perivascular regions close to vessels but rarely detected in regions distant from vessels in tumors (Figure 3F). We further performed an additional pericyte lineage tracing of GSCs cotransduced with DesPro-GFP and aSMAPro-mCherry and detected coexpression of mCherry and GFP in perivascular cells (Figures 3G and 3H). GFP+ perivascular cells were abundant around vessels and the majority of pericyte-marker-positive cells (>83%) expressed GFP (Figures 3A3F), conrming that GSCs generated the majority of pericytes in these tumors.
Tumor pericytes often exhibit abnormal morphologies, sometimes extending their processes away from the endothelium (Morikawa et al., 2002). The G-pericytes often displayed such irregular morphology (Figures 3A and 3F). Recent appreciation of intertumoral heterogeneity of GBMs has informed a mesenchymal subtype in contrast to proneural and classical subtypes (Verhaak et al., 2010). Interestingly, in vivo lineage tracing showed that mesenchymal GSCs have signicantly greater ability to generate pericytes than classic and proneural GSCs in xenografts (Figures S3B and S3C; Table S1). Collectively, these data provide direct evidence demonstrating that GSCs have the capacity to generate pericytes in vivo. Because our in vivo cell fate tracing of GSCs with GFP constitutive expression failed to detect GSC-derived ECs (Figures 1C and 1D), we performed the cell lineage tracing of GSCs with an EC marker (CD31 or CD105) promoter-driven GFP expression to directly address whether GSCs generate ECs. We cloned us et al., 1998) and the human CD105 (endoglin) promoter (R the CD31 (PECAM-1) promoter restricted to ECs (Almendro et al., 1996; Gumina et al., 1997) and then generated lentiviral constructs for conditional GFP expression driven by CD31 or CD105 promoter (CD31Pro-GFP or CD105Pro-GFP). We validated that the cloned CD31 and CD105 promoters were functional and EC specic because CD31Pro- or CD105Pro-driven GFP expression specically occurred in ECs (human brain microvessel endothelial cells [HBMECs]) (Figure S3A, right). To perform EC lineage tracing of GSCs, GSCs with CD31Pro-GFP or CD105Pro-GFP were orthotopically implanted into mouse brains. In conrmation with our earlier studies, no GFP expression was detectable in tumor ECs marked by CD31 and Glut1 staining (Figures S3D and S3E), further ruling out the possibility of GSC-derived ECs in GBM xenografts. To further characterize the G-pericytes, we examined pericyte marker expression in G-pericytes and HBVP pericytes. We isolated G-pericytes by sorting GFP+CD146+ cells from GBM xenografts derived from the DesPro-GFP-GSCs. Comparative RT-PCR analyses of key pericyte markers (a-SMA, Desmin, CD248, NG2, CD146, and PDGFRb) in the sorted G-pericytes and HBVPs conrmed similar marker expression in the GBM xenografts (Figure S4A). To address whether G-pericytes still express GSC markers after lineage switching, we examined expression of several putative GSC markers (SOX2, OLIG2, CD133, and Nestin) and pericyte markers in sorted GSCs (CD15+L1CAM+) and G-pericytes (GFP+CD146+). RT-PCR analyses showed that G-pericytes no longer express the GSC markers (Figure S4B). This result was conrmed by IF staining of SOX2, OLIG2, or Nestin on frozen sections of the DesProGFP-GSC xenografts. Consistently, GFP expression was turned on specically in perivascular cells that rarely (<0.8%) expressed SOX2, OLIG2, or Nestin (Figure S4C, S4D, and S4F). In contrast, the SOX2, OLIG2, or Nestin-expressing cells (GSCs) are localized near perivascular niches (Figure S4C and S4D). The mutually exclusive expression of GSC and pericyte markers suggests that GSCs undergo differentiation to generate G-pericytes rather than being a GSC subpopulation adjacent to ECs in GBM tumors. In addition, G-pericytes do not express astrocyte markers such as glial brillary acidic protein (GFAP) and S100b (Figures S4A, S4E, and S4F), indicating that G-pericytes are
not a subpopulation of astrocytes. Consistently, pericytes and astrocytes are distinct cell populations without overlapping expression of specic markers in primary GBMs (Figure S4G). These data demonstrate that G-pericytes are unique cells expressing specic pericyte markers. Pericytes in Primary GBMs Are Commonly Derived from Neoplastic Cells To examine whether pericytes are lineage related to cancer cells in human primary GBMs, we performed uorescence in situ hybridization (FISH) analyses of common GBM genetic changes (Cancer Genome Atlas Research Network, 2008) in combination with IF staining of a pericyte marker (a-SMA) to determine if pericytes carry cancer genetic alterations in GBMs. Because gains of chromosome 7 (EGFR amplication) or losses of chromosome 10 (PTEN loss) are frequent in GBM cells, permitting a lineage tracing to the neoplastic cells, we employed DNA probes for centromeres of chromosome 7 (CEP-7) and 10 (CEP-10), EGFR, and PTEN to detect cancer genetic alterations in pericytes, ECs, and tumor cells in GBM tissue microarrays. FISH analyses showed the majority of tumor pericytes (mean 76%, range 58%83%) carried the same genetic alterations (CEP-7 polysomy, EGFR trisomy or amplication, CEP-10 loss, or PTEN loss) as cancer cells in 49 GBMs (Figures 4A and 4B), indicating that tumor pericytes are commonly derived from cancer cells. In contrast, we rarely detected relevant genetic changes in ECs in these GBMs (Figures S5A and S5B). To further conrm these results, we isolated pericytes (CD146+CD248+; 2.18% 5.26% of total cells) and ECs (CD31+CD105+) from primary GBMs and performed similar FISH analyses. The majority (>72%) of sorted tumor pericytes (a-SMA+) carried the same genetic alteration (CEP-7 polysomy) as matched GSCs (Figures 4C and 4D). In contrast, sorted ECs (CD31+CD105+) expressed Glut1 but did not share the GSC genetic alterations (Figure S5C and S5D). These results support a tumor source for pericytes, but not for ECs in human primary GBMs. To further address whether pericytes in endogenous GBMs are derived from cancer cells, we examined pericytes in the genetically engineered mouse GBMs (Nestin-tva/Ink4a/Arf// HA-PDGFB models; Hambardzumyan et al., 2009). IF staining of hemagglutinin-tagged platelet-derived growth factor B (HAPDGFB) and pericyte markers (Desmin, NG2, CD248, or a-SMA) showed that a signicant fraction (mean 63%) of tumor pericytes expressed HA-PDGFB, supporting a tumor origin (Figure 4E, 4F; data not shown). In contrast, staining of EC markers (CD31 or Glut1) and HA-PDGFB showed no tumor-cell-derived ECs in these mouse GBMs (Figure S5E). These data demonstrate that pericytes in the genetically engineered mouse GBMs are also largely derived from neoplastic cells.
Selective Elimination of G-Pericytes Disrupts Tumor Vessels and Inhibits Tumor Growth To determine the functional signicance of G-pericytes, we examined effects of selective elimination of G-pericytes on vessels and tumor growth. GSCs were transduced with Desmin-promoter-driven expression of herpes simplex virus thymidine kinase (HsvTK) (Figure 5A) to achieve conditional HsvTK expression in G-pericytes. Because HsvTK metabolizes ganciclovir (GCV) into a toxic agent specically in cells expressing HsvTK (Culver et al., 1992), G-pericytes expressing HsvTK should be sensitive to GCV and thus eliminated by GCV treatment. To conrm selective killing of G-pericytes expressing Desmin-promoter-driven HsvTK by GCV treatment, we generated a construct for coexpression of HsvTK and GFP under the same promoter (DesPro-TK-GFP) (Figures 5A and S6A). As expected, after the DesPro-TK-GFP-transduced GSCs were induced to differentiate, GFP was expressed in a fraction of differentiated cells (G-pericytes) (Figure S6B). Apoptotic detection showed that GCV treatment selectively induced apoptosis in cells coexpressing GFP and HsvTK (Figure S6C). These data indicate that selective elimination of G-pericytes is achievable by using Desmin-promoter-driven HsvTK conditional expression with GCV treatment. To examine the impact of selective targeting of G-pericytes on tumor vessels, we implanted DesPro-TK-GFP-GSCs into mouse brains. Mice bearing the tumors were treated with vehicle control or GCV daily to induce HsvTK-mediated toxicity to G-pericytes. Apoptotic detection by TUNEL staining demonstrated that GCV treatment for 3 days selectively induced cell death in G-pericytes (GFP+) in vivo (Figure 5B). Further, GCV treatment for 1 week caused almost a complete depletion of G-pericytes, collapse of vessel lumens, and disruption of endothelial walls in GBM tumors (Figures 5C, 5D, S6D and S6E). Moreover, measurement of vascular function by uorescein isothiocyanate (FITC)-conjugated mega-dextran showed that GCV treatment for 1 week to deplete G-pericytes severely attenuated vascular function in the DesPro-TK-GSC xenografts, because perfusion of FITCmega-dextran into the tumors was dramatically reduced (Figures 5E, 5F, S6F and S6G). Collectively, these data demonstrate that selective elimination of G-pericytes potently disrupts vascular structure and function in GBM tumors. To evaluate the impact of selective targeting of G-pericytes on tumor growth, we initially used subcutaneous tumor experiments to track sequential tumor volumes. The established subcutaneous tumors derived from the DesPro-TK-GFP-GSCs were treated with GCV or vehicle control for 3 weeks. GCV treatment caused signicant regression of the tumors (Figures 5G and 5H), indicating that selective elimination of G-pericytes by HsvTKinduced GCV toxicity inhibited tumor growth. To further validate
Figure 3. In Vivo Lineage Tracing of GSCs with Pericyte-Specic Promoter-Driven Fluorescent Reporters
(AF) In vivo lineage tracing of GSCs with Desmin promoter-driven GFP (DesPro-GFP). Sections of GBM tumors derived from DesPro-GFP-GSCs were immunostained for an EC marker (CD31), a pericyte marker (Desmin, NG2, PDGFRb, CD248, Ang1, or CD13) (A), or the pericyte-EC junction marker Cx45 (D). Quantications show fractions of GFP+ pericytes (B), coverage by GFP+ pericytes (C), or the fraction of GFP+ pericytes expressing Cx45 (E). An arrow indicates rare GFP+ pericytes away from vessels (F). (G and H) In vivo lineage tracing of GSCs with coexpression of Desmin promoter-driven GFP and a-SMA promoter-driven mCherry in GBMs. Quantication shows the fraction of GFP+ cells with mCherry. All scale bars represent 25 mm. The error bars represent SD. See also Figures S3, S4, and Table S1.
C D
Figure 4. Pericytes Are Commonly Derived from Neoplastic Cells in Primary GBMs
(A and B) FISH analyses of genetic alterations with the CEP-7, CEP-10, EGFR, or PTEN probe (in red) in pericytes (a-SMA+) in primary GBMs. Quantication shows average fractions of pericytes carrying the cancer genetic alterations (CEP-7 polysomy, EGFR amplication or trisomy, CEP-10 loss, or PTEN loss) in GBM tissue arrays (B). (C and D) FISH analyses with CEP-7 probe in sorted pericytes (a-SMA+) and GSCs from primary GBMs. Quantication shows the fraction (mean 72%) of pericytes carrying the GSC genetic alterations (D). (E and F) IF staining of a pericyte marker (Desmin or a-SMA) and HA-PDGFB in the genetically engineered mouse GBMs (Nestin-tva/Ink4a/Arf//HA-PDGFB model). Quantications show fractions (mean 63%) of HA-PDGFB+ pericytes (F) The scale bars represent 10 mm (A and C) and 25 mm (E). The error bars represent SD. See also Figure S5.
this result in orthotopic tumors, we transduced GSCs either with DesPro-GFP (control) or DesPro-TK and implanted these GSCs into mouse brains to establish GBM xenografts. Both groups of mice bearing the tumors were administered with GCV to eliminate the G-pericytes expressing HsvTK. GCV treatment for 2 weeks caused extensive vessel regression in GBM tumors derived from DesPro-TK-GSCs, but not from DesPro-GFPGSCs (Figures S6H and S6I). Moreover, GCV treatment for 3 weeks markedly inhibited intracranial tumor growth in GBM xenografts derived from DesPro-TK-GSCs, but not in control tumors from DesPro-GFP-GSCs (Figures 5I and 5J). Alternatively, treatment by GCV, but not vehicle control suppressed intracranial tumor growth in the GBM xenografts derived from DesPro-TK-GSCs (Figure S6J). As a consequence, GCV treatment signicantly increased survival of animals implanted with the DesPro-TK-GSCs (Figure 5K). These data demonstrate that selective elimination of G-pericytes suppressed GBM tumor growth and malignant progression. GSCs Are Recruited toward ECs via the SDF-1/CXCR4 Axis To understand the mechanisms underlying GSC recruitment toward ECs, we examined whether GSCs can be recruited by HBMECs to support the maintenance of EC complexes in vitro. GSCs labeled with the green uorescent tracer CFSE were mixed with HBMEC complexes labeled with the red uorescent tracer CMTRX. Integration of GSCs-derived cells into EC complexes was detected on day 2 after cell mixing, and the integration stabilized the EC complexes for extended periods (2.6-fold) relative to EC complex alone (Figure 6A). To address whether pericyte lineage specication of GSCs can be induced by EC complexes, we cocultured GSCs with HBMECs and detected integration of G-pericytes by a-SMA staining (Figure 6B). Because the attachment of pericytes to ECs can be mediated through adherens junctions containing N-cadherin (Gerhardt et al., 2000), we examined N-cadherin expression and found that N-cadherin was localized to the contact sites between G-pericytes (a-SMA+) and EC complexes (Figure 6B). To dene the molecular mechanisms underlying GSC recruitment by ECs, we analyzed the effect of several chemotactic factors (SDF-1a, PDGFB, and transforming growth factor b [TGF-b]) secreted by HBMECs on GSC migration. We found that SDF-1 potently stimulated GSC migration (Figures S7A and S7B), whereas PDGFB only modestly attracted GSCs. To further address whether HBMECs attract GSCs via SDF-1, we cocultured GSCs and HBMECs in separate chambers of transwells and detected that HBMECs potently attracted GSCs, an effect dependent on SDF-1 because an anti-SDF-1 antibody attenuated the effects (Figures S7C and S7D). Because ECs in brain and GBMs constitutively express SDF-1 (Kokovay et al., 2010; Komatani et al., 2009), we conrmed that abundant SDF-1 formed a gradient around vessels with greater SDF-1 proximal to vessels in brain and GBMs (Figure S7E). Because SDF-1 is secreted by ECs and GSCs express the SDF-1 receptor CXCR4 (Ehtesham et al., 2009; Folkins et al., 2009), we hypothesized that brain ECs may recruit GSCs at least in part through the SDF-1/CXCR4 axis. CXCR4 knockdown in GSCs reduced the recruitment of GFP-labeled GSCs to EC
complexes (Figures 6C and 6D). In addition, an SDF-1 blocking antibody signicantly reduced the integration of GFP-labeled GSCs into HBMEC complexes (Figures S7F and S7G). As a further conrmation, we examined the effect of a CXCR4 inhibitor (AMD3100) on GSC recruitment to EC complexes. AMD3100 treatment signicantly reduced the integration of GSCs (CMTRX labeled, in red) into CFSE-labeled HBMECs (in green) (Figures S7H and S7I). To further determine whether GSC recruitment to ECs depends on the SDF-1/CXCR4 axis during tumor vascularization, GFP-labeled GSCs were transduced with shCXCR4 or nontargeting small hairpin RNA (shNT) and implanted into mouse brains. In shNT xenografts, tumor vessels were covered with abundant G-pericytes (GFP+ and Desmin+), whereas G-pericytes and total pericyte coverage on vessels was signicantly reduced in shCXCR4 xenografts (Figures 6E6G). Immunohistochemical (IHC) staining conrmed that CXCR4 knockdown signicantly decreased vessel density in the tumors (Figures S7J and S7K). Collectively, these data suggest that ECs recruit GSCs via the SDF-1/CXCR4 axis and that targeting this pathway reduces G-pericytes in GBMs. TGF-b Induces Differentiation of GSCs into Pericytes We next sought to understand the molecular mechanisms underlying the pericyte lineage specication of GSCs. To identify the potential factors inducing GSC differentiation into pericytes, we examined the effect of several EC-secreted cytokines (SDF-1, PDGFB, and TGF-b) on GSC differentiation into pericytes. Immunoblot analysis showed that TGF-b dominantly induced expression of a-SMA when GSCs were cultured in differentiation media (Figures 7A and 7B). IF staining of multiple pericyte markers (NG2, a-SMA, CD146 and CD248) conrmed that TGF-b treatment increased the fraction of cells expressing pericyte markers in the differentiated cells (Figures 7C and 7D; data not shown). Further, TGF-b treatment induced GFPexpressing cells in differentiated cells derived from DesProGFP-GSCs (Figures 7E and 7F). To address whether ECs induce GSC differentiation into pericytes through TGF-b, we cocultured DesPro-GFP-GSCs and HBMEC complexes and monitored GFP-expressing cells (G-pericytes) over time. GFP+ cells were induced and integrated into EC complexes, an effect that was attenuated by incubation of the EC complexes with an antiTGF-b antibody (Figure 7G). Immunoblot analysis validated that coculture of GSCs with HBMECs or their conditioned media induced expression of pericyte marker a-SMA in differentiated cells, an effect that was reduced by a TGF-b neutralizing antibody (Figure 7H). Collectively, these data demonstrate that HBMECs induce pericyte lineage specication of GSCs at least in part through TGF-b. Thus, the recruitment of GSCs toward ECs via the SDF-1/CXCR4 axis and the induction of GSC differentiation into pericytes by TGF-b are two events controlled by different molecular mechanisms (Figure 7I). DISCUSSION Pericytes play essential roles to maintain functional vessels to support tumor growth. Tumor pericytes are thought to be derived from their progenitors from the surrounding normal
I G
A
HBMECs only
Figure 6. GSCs Are Recruited toward ECs through the SDF-1/CXCR4 Axis to Support Endothelial Complex
Day 0 Day 2 Day 6 Day 8
(A) In vitro endothelial complex formation of HBMECs (labeled with red uorescent tracer CMTRX) with or without GSCs (labeled with CFSE, in green). (B) IF staining of a-SMA and N-cadherin in complexes of HBMECs and GSC-derived cells. Nuclei were stained with DAPI. (C and D) Endothelial complexes of HBMECs with GFP-labeled GSCs expressing shCXCR4 or shNT. Quantication shows fractions of GSC-derived cells (GFP+) on HBMEC complexes (D). *p < 0.001. (EG) In vivo lineage tracing of GSCs with GFP constitutive expression and IF staining of CD31 or Desmin in tumors derived from GSCs expressing shCXCR4 or shNT. Quantications show fractions of G-pericytes (GFP+) (F) and total pericyte coverage (G) on vessels. *p < 0.001. The scale bars represent 100 mm (AC) and 25 mm (E). The error bars represent SD. See also Figure S7.
HBMECs + GSCs
Day 0
Day 2
Day 6
Day 8
HBMECs + GSC
-SMA DAPI
N-Cadherin DAPI
Overlay
Phase
Fraction (%) of GSCs integrated into EC complex
C
HBMECs
GSCs + shNT
GSCs + shCXCR4
100 80 60 40 20
*
shNT shCXCR4-1 shCXCR4-2
E
GFPDesminDAPI GFPCD31DAPI
shNT
shCXCR4
0 GSCs: CCF2170 shCXCR4
shNT 100 80 60 40 20 0
tissue or from the bone-marrow-derived cells homing in tumors after treatments (De Palma et al., 2005; Du et al., 2008). In this study, we demonstrate that the majority of vascular pericytes
Figure 5. Selective Elimination of G-Pericytes Disrupts Tumor Vessels and Inhibits Tumor Growth
(A) Schematic illustrations of Desmin-promoter-driven expression of HsvTK and coexpression of HsvTK and GFP. (B) TUNEL assay detecting selective apoptosis (in red) of G-pericytes (GFP+) induced by ganciclovir (GCV) in GBM tumors derived from DesPro-TK-GFP-GSCs (CCF2170). (C and D) IF staining of CD31 (in red) shows effects of selective elimination of G-pericytes (GFP+) by GCV on vessels in GBM tumors derived from DesPro-TKGFP-GSCs. Quantication shows the reduced G-pericyte coverage by GCV treatment (D). *p < 0.001. (E and F) Assessment of vascular function using the FITC-conjugated mega-dextran after selective elimination of G-pericytes in GBM tumors derived from DesPro-TK-GSCs. Quantication shows intensity of perfused FITC-mega-dextran into the control or GCV-treated tumors (F). *p < 0.001. (G and H) The effect of targeting G-pericytes by GCV treatment on growth of subcutaneous tumors derived from DesPro-TK-GFP-GSCs. Quantication shows mean tumor weights in the control and GCV-treated mice (H). *p < 0.001 (n = 12). (I and J) The effect of selective elimination of G-pericytes on GBM growth in mouse brains. Mice bearing tumors derived from DesPro-TK-GSCs or DesPro-GFPGSCs (control) were treated with GCV for 3 weeks. Images of whole brains (I) and histological analysis (hematoxylin and eosin [H&E] staining) on brain sections (J) are shown. (K) Kaplan-Meier survival curves of mice bearing GBM tumors derived from DesPro-TK-GSCs or DesPro-GFP-GSCs (control) after GCV treatment. *p < 0.001 (n = 6). The scale bars represent 25 mm (B) and 100 mm (C and E). The error bars represent SD. See also Figure S6.
Fraction (%) of pericytes from GSCs (GFP+)
Pericyte coverage (%) on vessels
in GBMs are derived from GSCs. Because G-pericytes express similar pericyte markers as normal brain vascular pericytes, GSCs function as pericyte progenitors and contribute to vasculature formation in GBMs. The ability of GSCs to D456 generate vascular pericytes in vivo G shNT shCXCR4 suggests that GSCs may actively remodel * * 100 their microenvironment and create a supportive niche, permitting functional 80 vessels to augment tumor growth without 60 depending on the limited source of 40 normal pericyte progenitors from 20 surrounding tissues. 0 Because NSCs can transdifferentiate into pericytes (Ii et al., 2009; Morishita et al., 2007), a lineage link between NSCs and pericytes is present in normal tissues. Because GSCs share regulatory programs with NSCs, the plasticity of GSCs toward a pericyte lineage may be a product of aberrant
Figure 7. TGF-b Induces Differentiation of GSCs into Pericytes

(A) Immunoblot (IB) analysis of pericyte marker (a-SMA) expression in differentiated cells from GSCs (CCF1992) in the presence of indicated cytokines (1 ng/ml) in culture media. (B) IB analysis of pericyte markers (a-SMA, CD248, and NG2) and a GSC marker (SOX2) in GSCs and differentiated cells with or without treatment of TGF-b (2 ng/ml). (C and D) IF staining of pericyte markers (NG2 and a-SMA) in GSCs (CW1217) and differentiated cells induced by serum or TGF-b. Quantication shows pericyte fractions (D). *p < 0.001. (E and F) In vitro pericyte lineage tracing of GSCs with Desmin promoter-driven GFP induced by serum or TGF-b (2 ng/ml). Quantication shows fractions of GFP+ cells in the differentiated cells (F). *p < 0.001. (G) In vitro HBMEC complex formation with DesPro-GFP-GSCs in the presence of anti-TGFb antibody (monoclonal antibody [mAb]) or immunoglobulin G (IgG). (H) IB analysis of a-SMA expression after coculture of GSCs with HBMECs or their conditioned media in the presence of anti-TGF-b antibody or IgG. (I) A schematic illustration showing the recruitment of GSCs toward ECs and the differentiation of GSCs into pericytes in GBMs. GSCs expressing CXCR4 are recruited toward ECs by SDF-1 and induced predominantly by TGF-b to become pericytes to support vessel function and tumor growth. The scale bars represent 25 mm (C and E) and 100 mm (G). The error bars represent SD.
F E
developmental biology. Although previous reports suggest that GSCs may give rise to ECs in GBMs (Ricci-Vitiani et al., 2010; Soda et al., 2011; Wang et al., 2010), such an event may be very rare because ECs in GBMs rarely carry the cancer genetic mutations as demonstrated in our study and others (Kulla et al., 2003; Rodriguez et al., 2012). Moreover, our complementary lineage tracing studies failed to demonstrate GSC-derived ECs in vivo, although in the culture condition we occasionally observed rare EC-marker-expressing cells (<0.6%) in differentiated cells from GSCs. Because vascular pericytes closely attach to ECs and both cells appear very thin, prior studies may have missed the true identity of tumor-derived cells on vessels. Because both ECs and pericytes express Tie2 (De Palma et al., 2005), the use of Tie2 promoter-driven HsvTK expression for targeting GSC-derived ECs (Ricci-Vitiani et al., 2010) might actually eliminate the G-pericytes. Our in vivo lineage tracing with pericyte- or EC-specic promoter-driven uorescent reporters
directly demonstrated that GSCs give rise to pericytes rather than ECs in vivo. The contribution of GSCs to vascular pericytes requires GSC recruitment toward ECs. Because ECs in brain and GBMs express abundant SDF-1 forming chemoattractant gradient, the expression of CXCR4 (the receptor for SDF-1) in GSCs (Ehtesham et al., 2009; Folkins et al., 2009) may provide a paracrine loop for recruitment of GSCs toward ECs. A recent study showed that NSCs can be recruited to perivascular niches in normal brain through the CXCR4/SDF-1 axis (Kokovay et al., 2010). The recruitment of pericyte progenitors to ECs in normal tissues also depends on SDF-1/CXCR4 signaling (Song et al., 2009). SDF-1 expression has been proposed as one of the mechanisms underlying the resistance to antiangiogenic therapy in GBM trials (Batchelor et al., 2007). Elevated SDF-1 signaling may enhance GSC recruitment toward ECs and increase G-pericyte coverage to protect tumor vessels, leading to resistance to antiangiogenic therapy. The potent capacity of GSCs to generate vascular pericytes allows active vascularization in GBMs to support tumor growth. Because GSCs contribute to the majority of vascular pericytes in GBMs, G-pericytes may have a crucial role in mediating therapeutic resistance in GBMs. Because pericytes juxtacrine to
ECs express signicant levels of VEGF and other factors to support EC survival (Franco et al., 2011; Song et al., 2005; Winkler et al., 2011), G-pericytes may protect ECs and render ECs less responsive to antiangiogenic agents in GBMs. Thus, targeting G-pericytes may synergize with current therapies targeting ECs to achieve more effective outcome. Because CSCs are present in other solid cancers (Magee et al., 2012), it is important to determine whether CSCs can generate vascular pericytes in other malignant tumors with orid angiogenesis. Our studies demonstrate that GSCs not only interact with perivascular niches but also have the capacity to remodulate their microenvironment by contributing pericyte compartments of the neovasculature. Because selective elimination of G-pericytes potently disrupted vessels and inhibited tumor growth, therapeutic targeting of G-pericytes may have a signicant impact on improving GBM treatment efcacy.
EXPERIMENTAL PROCEDURES Isolation of GSCs and Non-Stem Tumor Cells from GBMs GBM surgical specimens were collected in accordance with a Cleveland Clinic Institutional Review Board-approved protocol. GSCs and non-stem tumor cells were derived from GBM tumors and functionally validated as described previously (Bao et al., 2006a; Guryanova et al., 2011). For the detailed procedure, please see Extended Experimental Procedures. Pericyte or EC-Specic Promoter-Driven Expression of GFP or mCherry Human Desmin promoter (312 bp) with an enhancer (284 bp) (Li and Paulin, 1991), a-SMA promoter (262 bp) with an enhancer (123 bp) (Keogh et al., us 1999; Nakano et al., 1991), CD105 promoter plus enhancer (955 bp) (R et al., 1998), and CD31 promoter plus enhancer (887 bp) restricted to ECs (Almendro et al., 1996; Gumina et al., 1997) were cloned by PCR and conrmed by sequencing. The specic promoter with enhancer was inserted into pCDH-CMV-EF1-Puro lentiviral vector (System Biosciences) to replace the original CMV promoter. The ORF of GFP or mCherry was then inserted into the vector to generate lentiviral constructs. Lentiviruses were produced and tittered as described elsewhere (Guryanova et al., 2011). Cell Lineage Tracing of GSCs To perform cell lineage tracing, GSCs were transduced with GFP or mCherry constitutive expression or conditional expression driven by the pericyte or EC-specic promoter through lentiviral infection and then transplanted into brains of athymic BALB/c nu/nu mice to establish xenografts as described elsewhere (Guryanova et al., 2011). To trace cell lineage of GSCs in vivo, sections of mouse brains bearing the xenografts were immunostained for pericyte or EC markers and analyzed for GFP or mCherry expression. IF and IHC stainings were performed as described (Guryanova et al., 2011). Tumor sections of the genetically engineered mouse GBMs were provided by Dr. Dolores Hambardzumyan. For detailed methods and the antibody information, please see Extended Experimental Procedures. Selective Targeting of G-Pericytes in GBM Xenografts GSCs were transduced with Desmin or CD31-promoter-driven expression of HsvTK, GFP, or HsvTK plus GFP through lentiviral infection and then transplanted into brains of athymic mice. Mice bearing the xenografts received GCV (Sigma-Aldrich) at 75 mg/kg/day or vehicle control daily through intraperitoneal injection. The xenografts were collected for IF and IHC staining and uorescent analysis. To evaluate the targeting effect on animal survival, mice were maintained until the development of neurological signs. HBVPs, HBMECs, and EC Complex Formation HBVPs and HBMECs were obtained from ScienCell. HBMECs with low passage were used for coculture and endothelial complex formation assays
as described (Bao et al., 2006b). For the detailed procedure and the labeling of GSCs and HBMECs, please see Extended Experimental Procedures. Statistical Analysis All quantied data were statistically analyzed. Grouped data are presented as mean SD. The difference between experimental groups was assessed by one-way ANOVA or one-way ANOVA on ranks testing. For the animal survival experiments, log-rank survival analysis was performed. For further details, please see Extended Experimental Procedures. SUPPLEMENTAL INFORMATION Supplemental Information includes Extended Experimental Procedures, seven gures, and one table and can be found with this article online at http://dx.doi. org/10.1016/j.cell.2013.02.021. ACKNOWLEDGMENTS We thank Brain Tumor and Neuro-Oncology Centers at Cleveland Clinic, University Hospitals, and Duke Medical Center for providing GBM specimens. We also thank Dr. Dolores Hambardzumyan for providing the genetically engineered mouse GBMs. We are grateful to Cathy Shemo at the Flow Cytometry Core and Judith Drazba at the Imaging Core for their help. This work was supported by the Cleveland Clinic Foundation and National Institutes of Health R01 grant NS070315 (to S.B.). Received: August 19, 2011 Revised: May 12, 2012 Accepted: February 11, 2013 Published: March 28, 2013 REFERENCES n, T., Rius, C., Lastres, P., Langa, C., Corb , A., and BernaAlmendro, N., Bello u, C. (1996). Cloning of the human platelet endothelial cell adhesion molebe cule-1 promoter and its tissue-specic expression. Structural and functional characterization. J. Immunol. 157, 54115421. , G., Ma e, M., Nisancioglu, M.H., Wallgard, E., Niaudet, C., Armulik, A., Genove He, L., Norlin, J., Lindblom, P., Strittmatter, K., et al. (2010). Pericytes regulate the blood-brain barrier. Nature 468, 557561. Bao, S., Wu, Q., McLendon, R.E., Hao, Y., Shi, Q., Hjelmeland, A.B., Dewhirst, M.W., Bigner, D.D., and Rich, J.N. (2006a). Glioma stem cells promote radioresistance by preferential activation of the DNA damage response. Nature 444, 756760. Bao, S., Wu, Q., Sathornsumetee, S., Hao, Y., Li, Z., Hjelmeland, A.B., Shi, Q., McLendon, R.E., Bigner, D.D., and Rich, J.N. (2006b). Stem cell-like glioma cells promote tumor angiogenesis through vascular endothelial growth factor. Cancer Res. 66, 78437848. Batchelor, T.T., Sorensen, A.G., di Tomaso, E., Zhang, W.T., Duda, D.G., Cohen, K.S., Kozak, K.R., Cahill, D.P., Chen, P.J., Zhu, M., et al. (2007). AZD2171, a pan-VEGF receptor tyrosine kinase inhibitor, normalizes tumor vasculature and alleviates edema in glioblastoma patients. Cancer Cell 11, 8395. Bell, R.D., Winkler, E.A., Sagare, A.P., Singh, I., LaRue, B., Deane, R., and Zlokovic, B.V. (2010). Pericytes control key neurovascular functions and neuronal phenotype in the adult brain and during brain aging. Neuron 68, 409427. Bergers, G., and Song, S. (2005). The role of pericytes in blood-vessel formation and maintenance. Neuro-oncol. 7, 452464. Bergers, G., Song, S., Meyer-Morse, N., Bergsland, E., and Hanahan, D. (2003). Benets of targeting both pericytes and endothelial cells in the tumor vasculature with kinase inhibitors. J. Clin. Invest. 111, 12871295. Calabrese, C., Poppleton, H., Kocak, M., Hogg, T.L., Fuller, C., Hamner, B., Oh, E.Y., Gaber, M.W., Finklestein, D., Allen, M., et al. (2007). A perivascular niche for brain tumor stem cells. Cancer Cell 11, 6982.
Cancer Genome Atlas Research Network. (2008). Comprehensive genomic characterization denes human glioblastoma genes and core pathways. Nature 455, 10611068. Carmeliet, P., and Jain, R.K. (2011). Principles and mechanisms of vessel normalization for cancer and other angiogenic diseases. Nat. Rev. Drug Discov. 10, 417427. Crisan, M., Yap, S., Casteilla, L., Chen, C.W., Corselli, M., Park, T.S., Andriolo, G., Sun, B., Zheng, B., Zhang, L., et al. (2008). A perivascular origin for mesenchymal stem cells in multiple human organs. Cell Stem Cell 3, 301313. Culver, K.W., Ram, Z., Wallbridge, S., Ishii, H., Oldeld, E.H., and Blaese, R.M. (1992). In vivo gene transfer with retroviral vector-producer cells for treatment of experimental brain tumors. Science 256, 15501552. deCarvalho, A.C., Nelson, K., Lemke, N., Lehman, N.L., Arbab, A.S., Kalkanis, S., and Mikkelsen, T. (2010). Gliosarcoma stem cells undergo glial and mesenchymal differentiation in vivo. Stem Cells 28, 181190. De Palma, M., Venneri, M.A., Galli, R., Sergi Sergi, L., Politi, L.S., Sampaolesi, M., and Naldini, L. (2005). Tie2 identies a hematopoietic lineage of proangiogenic monocytes required for tumor vessel formation and a mesenchymal population of pericyte progenitors. Cancer Cell 8, 211226. , E., Song, H., Du, R., Lu, K.V., Petritsch, C., Liu, P., Ganss, R., Passegue Vandenberg, S., Johnson, R.S., Werb, Z., and Bergers, G. (2008). HIF1alpha induces the recruitment of bone marrow-derived vascular modulatory cells to regulate tumor angiogenesis and invasion. Cancer Cell 13, 206220. Ehtesham, M., Mapara, K.Y., Stevenson, C.B., and Thompson, R.C. (2009). CXCR4 mediates the proliferation of glioblastoma progenitor cells. Cancer Lett. 274, 305312. Folkins, C., Shaked, Y., Man, S., Tang, T., Lee, C.R., Zhu, Z., Hoffman, R.M., and Kerbel, R.S. (2009). Glioma tumor stem-like cells promote tumor angiogenesis and vasculogenesis via vascular endothelial growth factor and stromal-derived factor 1. Cancer Res. 69, 72437251. Franco, M., Roswall, P., Cortez, E., Hanahan, D., and Pietras, K. (2011). Pericytes promote endothelial cell survival through induction of autocrine VEGF-A signaling and Bcl-w expression. Blood 118, 29062917. Gerhardt, H., Wolburg, H., and Redies, C. (2000). N-cadherin mediates pericytic-endothelial interaction during brain angiogenesis in the chicken. Dev. Dyn. 218, 472479. Gumina, R.J., Kirschbaum, N.E., Piotrowski, K., and Newman, P.J. (1997). Characterization of the human platelet/endothelial cell adhesion molecule-1 promoter: identication of a GATA-2 binding element required for optimal transcriptional activity. Blood 89, 12601269. Guryanova, O.A., Wu, Q., Cheng, L., Lathia, J.D., Huang, Z., Yang, J., MacSwords, J., Eyler, C.E., McLendon, R.E., Heddleston, J.M., et al. (2011). Nonreceptor tyrosine kinase BMX maintains self-renewal and tumorigenic potential of glioblastoma stem cells by activating STAT3. Cancer Cell 19, 498511. Hambardzumyan, D., Amankulor, N.M., Helmy, K.Y., Becher, O.J., and Holland, E.C. (2009). Modeling adult gliomas using RCAS/t-va technology. Transl. Oncol. 2, 8995. Ii, M., Nishimura, H., Sekiguchi, H., Kamei, N., Yokoyama, A., Horii, M., and Asahara, T. (2009). Concurrent vasculogenesis and neurogenesis from adult neural stem cells. Circ. Res. 105, 860868. Keogh, M.C., Chen, D., Schmitt, J.F., Dennehy, U., Kakkar, V.V., and Lemoine, N.R. (1999). Design of a muscle cell-specic expression vector utilising human vascular smooth muscle alpha-actin regulatory elements. Gene Ther. 6, 616628. Kokovay, E., Goderie, S., Wang, Y., Lotz, S., Lin, G., Sun, Y., Roysam, B., Shen, Q., and Temple, S. (2010). Adult SVZ lineage cells home to and leave the vascular niche via differential responses to SDF1/CXCR4 signaling. Cell Stem Cell 7, 163173. Komatani, H., Sugita, Y., Arakawa, F., Ohshima, K., and Shigemori, M. (2009). Expression of CXCL12 on pseudopalisading cells and proliferating microvessels in glioblastomas: an accelerated growth factor in glioblastomas. Int. J. Oncol. 34, 665672.
Kulla, A., Burkhardt, K., Meyer-Puttlitz, B., Teesalu, T., Asser, T., Wiestler, O.D., and Becker, A.J. (2003). Analysis of the TP53 gene in laser-microdissected glioblastoma vasculature. Acta Neuropathol. 105, 328332. Li, Z.L., and Paulin, D. (1991). High level desmin expression depends on a muscle-specic enhancer. J. Biol. Chem. 266, 65626570. Magee, J.A., Piskounova, E., and Morrison, S.J. (2012). Cancer stem cells: impact, heterogeneity, and uncertainty. Cancer Cell 21, 283296. Morikawa, S., Baluk, P., Kaidoh, T., Haskell, A., Jain, R.K., and McDonald, D.M. (2002). Abnormalities in pericytes on blood vessels and endothelial sprouts in tumors. Am. J. Pathol. 160, 9851000. Morishita, R., Nagata, K., Ito, H., Ueda, H., Asano, M., Shinohara, H., Kato, K., and Asano, T. (2007). Expression of smooth muscle cell-specic proteins in neural progenitor cells induced by agonists of G protein-coupled receptors and transforming growth factor-beta. J. Neurochem. 101, 10311040. Nakano, Y., Nishihara, T., Sasayama, S., Miwa, T., Kamada, S., and Kakunaga, T. (1991). Transcriptional regulatory elements in the 50 upstream and rst intron regions of the human smooth muscle (aortic type) alpha-actin-encoding gene. Gene 99, 285289. Norden, A.D., Drappatz, J., and Wen, P.Y. (2009). Antiangiogenic therapies for high-grade glioma. Nat. Rev. Neurol. 5, 610620. ` ez-Ribes, M., Allen, E., Hudock, J., Takeda, T., Okuyama, H., Vin als, F., Pa Inoue, M., Bergers, G., Hanahan, D., and Casanovas, O. (2009). Antiangiogenic therapy elicits malignant progression of tumors to increased local invasion and distant metastasis. Cancer Cell 15, 220231. Ricci-Vitiani, L., Pallini, R., Larocca, L.M., Lombardi, D.G., Signore, M., Pierconti, F., Petrucci, G., Montano, N., Maira, G., and De Maria, R. (2008). Mesenchymal differentiation of glioblastoma stem cells. Cell Death Differ. 15, 14911498. Ricci-Vitiani, L., Pallini, R., Biffoni, M., Todaro, M., Invernici, G., Cenci, T., Maira, G., Parati, E.A., Stassi, G., Larocca, L.M., and De Maria, R. (2010). Tumour vascularization via endothelial differentiation of glioblastoma stemlike cells. Nature 468, 824828. us, C., Smith, J.D., Almendro, N., Langa, C., Botella, L.M., Marchuk, D.A., R u, C. (1998). Cloning of the promoter region of human Vary, C.P., and Bernabe endoglin, the target gene for hereditary hemorrhagic telangiectasia type 1. Blood 92, 46774690. Rodriguez, F.J., Orr, B.A., Ligon, K.L., and Eberhart, C.G. (2012). Neoplastic cells are a rare component in human glioblastoma microvasculature. Oncotarget 3, 98106. Soda, Y., Marumoto, T., Friedmann-Morvinski, D., Soda, M., Liu, F., Michiue, H., Pastorino, S., Yang, M., Hoffman, R.M., Kesari, S., and Verma, I.M. (2011). Transdifferentiation of glioblastoma cells into vascular endothelial cells. Proc. Natl. Acad. Sci. USA 108, 42744280. Song, S., Ewald, A.J., Stallcup, W., Werb, Z., and Bergers, G. (2005). PDGFRbeta+ perivascular progenitor cells in tumours regulate pericyte differentiation and vascular survival. Nat. Cell Biol. 7, 870879. Song, N., Huang, Y., Shi, H., Yuan, S., Ding, Y., Song, X., Fu, Y., and Luo, Y. (2009). Overexpression of platelet-derived growth factor-BB increases tumor pericyte content via stromal-derived factor-1alpha/CXCR4 axis. Cancer Res. 69, 60576064. Verhaak, R.G., Hoadley, K.A., Purdom, E., Wang, V., Qi, Y., Wilkerson, M.D., Miller, C.R., Ding, L., Golub, T., Mesirov, J.P., et al.; Cancer Genome Atlas Research Network. (2010). Integrated genomic analysis identies clinically relevant subtypes of glioblastoma characterized by abnormalities in PDGFRA, IDH1, EGFR, and NF1. Cancer Cell 17, 98110. Wang, R., Chadalavada, K., Wilshire, J., Kowalik, U., Hovinga, K.E., Geber, A., Fligelman, B., Leversha, M., Brennan, C., and Tabar, V. (2010). Glioblastoma stem-like cells give rise to tumour endothelium. Nature 468, 829833. Winkler, E.A., Bell, R.D., and Zlokovic, B.V. (2011). Central nervous system pericytes in health and disease. Nat. Neurosci. 14, 13981405. Zhou, B.B., Zhang, H., Damelin, M., Geles, K.G., Grindley, J.C., and Dirks, P.B. (2009). Tumour-initiating cells: challenges and opportunities for anticancer drug discovery. Nat. Rev. Drug Discov. 8, 806823.
Heritable Remodeling of Yeast Multicellularity by an Environmentally Responsive Prion

Daniel L. Holmes,1 Alex K. Lancaster,2 Susan Lindquist,2,3,4 and Randal Halfmann1,*
of Biochemistry, University of Texas Southwestern Medical Center, 5323 Harry Hines Boulevard, Dallas, TX 75390-9038, USA Institute for Biomedical Research, 9 Cambridge Center, Cambridge, MA 02142, USA 3Department of Biology 4Howard Hughes Medical Institute Massachusetts Institute of Technology, 77 Massachusetts Avenue, Cambridge, MA 02139, USA *Correspondence: randal.halfmann@utsouthwestern.edu http://dx.doi.org/10.1016/j.cell.2013.02.026
2Whitehead 1Department
SUMMARY
Prion proteins undergo self-sustaining conformational conversions that heritably alter their activities. Many of these proteins operate at pivotal positions in determining how genotype is translated into phenotype. But the breadth of prion inuences on biology and their evolutionary signicance are just beginning to be explored. We report that a prion formed by the Mot3 transcription factor, [MOT3+], governs the acquisition of facultative multicellularity in the budding yeast Saccharomyces cerevisiae. The traits governed by [MOT3+] involved both gains and losses of Mot3 regulatory activity. [MOT3+]-dependent expression of FLO11, a major determinant of cellcell adhesion, produced diverse lineage-specic multicellular phenotypes in response to nutrient deprivation. The prions themselves were induced by ethanol and eliminated by hypoxiaconditions that occur sequentially in the natural respiro-fermentative cycles of yeast populations. These data demonstrate that prions can act as environmentally responsive molecular determinants of multicellularity and contribute to the natural morphological diversity of budding yeast.
INTRODUCTION The evolution of multicellularity is among the most notable transitions in the history of life (Grosberg and Strathmann, 2007). Despite its reputation as a simple, unicellular eukaryote, the budding yeast Saccharomyces cerevisiae has proven to be a powerful model for this transition (Koschwanez et al., 2011; Ratcliff et al., 2012). It frequently and reversibly abandons a solitary lifestyle in favor of diverse multicellular growth forms (re sch, 2012). In doing so, individual ckner and Mo viewed in Bru yeast cells cooperate to protect themselves from the environment, from other organisms, and from starvation. The cell surface adhesins that drive multicellularity are encoded by some of the most heavily regulated and yet rapidly
ckner evolving genes in the yeast genome (Hahn et al., 2005; Bru sch, 2012). Repeated selection for new adhesin phenoand Mo types might support mechanisms that expedite their genetic diversication, including intragenic tandem repeats, gene multiplicity, and a subtelomeric location, all of which produce high recombination rates that drive the frequent appearance of new sch, 2012). ckner and Mo functional variants (Bru In opposition to the environmental pressures for diversication, multicellularity itself necessitates conformity: individual cells must act concertedly and cooperatively to maintain the integrity and therefore adaptive benets of multicellular structures. This dichotomy may have favored the evolution of switch-like regulation in adhesin expression, resulting in binary or dimorphic transitions. Such switches are both genetic and epigenetic in nature. Mutations occur frequently in trans-acting regulators of adhesins (Halme et al., 2004), whereas positiondependent genomic silencing (Halme et al., 2004), cis-interfering noncoding RNAs (Bumgarner et al., 2012), and stochastic associations with low-abundance transcription factors (Octavio et al., 2009) can each act as epigenetic toggle switches of adhesin expression. Self-perpetuating switches in the folding of certain proteins, known as prions, might also play a role (Patel et al., 2009; Halfmann et al., 2012). Among these very different molecular mechanisms, the fact that prions are based on protein folding might give them an intrinsic capacity to transduce environmental changes into dominant and highly stable phenotypic changes. Prion-forming proteins can exist in profoundly different structural states, at least one of which is a self-templating (prion) state. Proteins in the prion conformation interact with nonprion conformers of the same protein, converting them to prion conformers. The biochemically characterized prions are insoluble amyloid-like brils that cause heritable changes in the cellular distribution of the protein. Prions create phenotypes by diverse meansin many cases by sequestering proteins away from their normal cellular activities, but in others, endowing them with new activities (Derkatch et al., 2001; Rogoza et al., 2010; Suzuki et al., 2012). Notably, the self-templating nature of prion conformations allows such phenotypes to be immediately and robustly heritable (Serio et al., 2000; Satpute-Krishnan and Serio, 2005). This has generated a vigorous discussion about the potential signicance of prions in yeast evolution (Pigliucci, 2008; Wickner et al., 2011; Koonin, 2012).
Figure 1. Prions Converge on the Principal Determinant of Yeast Multicellularity, FLO11

(A) Venn diagram demonstrating the overlap in the regulons of known prion-forming yeast transcription factors. Only two genes, FLO11 and HXT2, are regulated by all four. (B) A genetic reporter for [MOT3+]. Mot3 represses transcription of DAN1 under normal growth conditions. When the DAN1 ORF is replaced with URA3, laboratory yeast strains can become prototrophic for uracil by prion-driven inactivation of Mot3. (C) Cells harboring the reporter for [MOT3+] were transformed with either empty vector (E.V.) or a galactose-inducible vector encoding Mot3. Following 24 hr of growth in galactose, 5-fold serial dilutions of cells were plated to media lacking uracil ( Uracil) to select for those that contained stably inactivated Mot3. Cells plated to uracilcontaining media (+ Uracil), at the highest dilution, are also shown. Transient overexpression of Mot3 strongly induces conversion to a Ura+ phenotype. (D) qRT-PCR using mRNA isolated from isogenic [MOT3+] and [mot3] colonies reveals that [MOT3+] cells have increased FLO11 expression. Values are normalized to ACT1. Error bars represent SD from triplicate colonies. See also Figure S1 and Tables S1 and S2.
Yeast cells normally switch between prion and nonprion states at low frequencies, but protein folding is extraordinarily sensitive to environmental stress (Chiti and Dobson, 2006). Therefore, when cells are not well suited to their environment, i.e., when they are stressed, prion switching may accelerate (Tyedmers et al., 2008). The net result is that a small fraction of stressed cells explores alternative phenotypes. Often, newly arising prions are detrimental (McGlinchey et al., 2011). On occasion, however, the phenotypes revealed by prions are adaptive, enabling the prion-containing lineages to survive at times when they otherwise might perish (True and Lindquist, 2000; Halfmann and Lindquist, 2010). Theoretical work supports the concept that prion switching may constitute a sophisticated form of bet-hedging (Griswold and Masel, 2009; Lancaster et al., 2010). Despite the plausible functions of prion-based switches in gene regulation, the topic remains intensely controversial (Liebman and Chernoff, 2012). Denitive evidence for prion functionality, including mechanisms that would link specic prion-protein conformational switches to environmental changes that are explicitly relevant to their functions, is wanting. Here, we report that prion formation by the Mot3 transcriptional repressor regulates the acquisition of multicellular growth forms in budding yeast. The formation, elimination, and phenotypic manifestation of Mot3 prions each respond to specic environmental conditions. The effects of Mot3 prions are further determined by heritable variation between different yeast isolates. Mot3 prion switching is thus a molecular mechanism that couples natural environmental changes to heritable changes in
gene expression. It might also provide a new route to the evolution of cooperative multicellular behaviors. RESULTS A Convergence of Prions at the Multicellularity Determinant, FLO11 The functions of known yeast prions are heavily biased for gene regulation (Halfmann and Lindquist, 2010). To identify regulatory functions that might be associated with prion switching, we compared the published regulons of transcription factors that we and others have shown to be capable of forming prions (Alberti et al., 2009; Patel et al., 2009; Rogoza et al., 2010). The regulons of all three experimentally veried prion-forming transcription factors (Cyc8, Mot3, and Sfp1), as well as Gln3, a prion-like transcription factor that is itself the major regulatory target of the Ure2 prion (Wickner, 1994; Kulkarni et al., 2001), overlapped at only two genes: FLO11, encoding a cell wall-anchored glycoprotein; and HXT2, encoding a high-afnity glucose transporter (Figure 1A). Both genes are naturally induced by stress and by nutrient depri rkel, 1999; Bru ckner and vation (Ozcan and Johnston, 1999; Tu sch, 2012). The probability that any given ORF would be reguMo lated by all four of these transcription factors, by chance, is 1.21 3 105 (Extended Experimental Procedures available online). Notably, FLO11 is also regulated by the chromatin-remodeling factor Swi1 (Barrales et al., 2012), yet another type of transcriptional regulator that is capable of forming a prion (Du et al., 2008). Other transcription factors are likely to be prions. Therefore, to reduce bias, we also undertook a complementary computational
analysis to determine if the 32 transcription factors annotated to regulate FLO11 are signicantly enriched for the presence of prion-like domains, including those that have not yet been discovered to act as prions. We used a modied hidden Markov model algorithm (Alberti et al., 2009) to predict the prion propensities for all yeast transcription factors. We then asked whether the observed distribution of this propensity score for the regulators of FLO11 was signicantly different from random samples of 32 transcription factors using a bootstrap approach (Extended Experimental Procedures). We found that regulators of FLO11 had a higher enrichment for prion propensity than expected by chance (p = 0.017). Why might the FLO11 regulatory interactome be inundated with prion regulators? Three aspects of its biology suggest an answer. First, as the principal determinant of multicellularity; Flo11 enables single cells to differentiate into multicellular struc ckner and tures, including cell clumps, chains, and biolms (Bru sch, 2012). The success of any multigenerational developMo mental program of this nature necessitates a stable molecular commitment (which prions provide) rather than the shortterm responses of conventional regulatory networks. Second, FLO11 expression switches between expression states at multiple frequencies, ranging from approximately once every three cell divisions to once in a thousand cell divisions (Kuthan et al., 2003; Halme et al., 2004; Octavio et al., 2009). This is relevant because prion-based switches have also been found to occur at multiple frequencies, from 107 to 102 per cell division (Liebman and Chernoff, 2012). Finally, FLO11 is induced by, and ckner is thought to protect against, environmental stresses (Bru sch, 2012). Prions are intrinsically sensitive to changes and Mo in protein homeostasis (Tyedmers et al., 2008) and may, therefore, be able to act as sensors of environmental stress. These considerations led us to test whether the gain and loss of prion states by prion-forming transcription factors might induce facultative multicellular transitions in response to environmental adversities. Isolation of [MOT3+] Colonies Given the well-known instability of FLO11 expression, and the multitude of factors responsible for regulating it, we reasoned that simply correlating FLO11 expression dynamics with individual prions would not provide a sufciently rigorous test of our hypothesis. Instead, we employed a reverse approach, in which we isolated stable prion states on the basis of a completely orthogonal phenotype and then queried their effects on FLO11. To this end, and to facilitate follow-up experiments, we focused on Mot3, a transcription factor with a relatively discrete set of gene targets and well-characterized biology. The consensus binding motif for Mot3 (nucleotide sequence HAGGYA) occurs 16 times upstream of FLO11 (Table S1), far more frequently than expected by chance (cumulative binomial probability of 0.007). Mot3 normally represses FLO11 (Carter et al., 2007) and other genes involved in remodeling the yeast cell surface (Table S2; annotation clusters 1, 2, and 6 designate cell walland cell membrane-localized gene products; cluster 4 designates gene products involved with membrane biosynthesis). Oxygen depletion alleviates this repression (Sertil et al., 2003; Lai et al., 2006).
To provide a facile orthogonal measure of Mot3 prion switching, we placed a URA3 gene under the control of a Mot3-regulated promoter (DAN1), in a strain that also carries a ura3 deletion (Alberti et al., 2009). Mot3 normally represses this promoter. When Mot3 switches to the prion state, however, it should sequester the protein away from the genome, derepressing URA3 and allowing cells to grow without uracil (Figure 1B). In keeping with prion nomenclature, this state is designated [MOT3+] (Alberti et al., 2009), with capital letters denoting dominance in genetic crosses and brackets denoting cytoplasmic inheritance. Because most laboratory yeast have recessive null mutations in FLO8, a master regulator of FLO11 (Liu et al., 1996), we mated the Mot3-reporter strain with a ura3 FLO8 strain (S1278b) to produce a hybrid diploid competent for FLO11 expression. This strain spontaneously acquired a Ura+ phenotype at a frequency of 104 to 103, depending on the stringency of selection (Figure S1). One dening feature of prion biology is that overexpression of the prion protein increases the frequency at which the prion appears in a population of prion minus cells (Wickner, 1994; Alberti et al., 2009). A second is that once the protein has acquired the prion conformation, overexpression is no longer necessary to maintain that state. To induce [MOT3+], we therefore transformed cells with a galactose-inducible plasmid encoding Mot3, or with an empty vector, and grew the cells in galactose media for 24 hr. We then plated the cells to glucose media lacking uracil to restore endogenous Mot3 expression levels. This selected for cells in which Mot3 was inactivated in a heritable way. Indeed, the transient overexpression greatly increased the appearance of Ura+ colonies (Figure 1C). Such colonies retained their phenotype after repeated passaging. Another characteristic common to prions formed by amyloidogenic proteins is a requirement for Hsp104. This proteinremodeling factor fragments prion amyloid bers and enables prion templates to be inherited by daughter cells. To test if the Ura+ phenotype was due to a prion switch, we passaged cells on media containing 3 mM guanidine hydrochloride (GdHCl), a selective inhibitor of Hsp104 (Ferreira et al., 2001). We then passaged them to media lacking GdHCl before testing for the continued inheritance of the Ura+ phenotype. As was previously demonstrated in a different genetic background by Alberti et al. (2009), this treatment restored Ura+ cells to the original, ura phenotype (Figure S1B). To ensure that this was not due to an off-target effect of GdHCl, we used a genetic approachtransiently expressing a dominantnegative variant of Hsp104 (Chernoff et al., 1995), Hsp104DN. This treatment also restored the ura phenotype. Overexpression of WT Hsp104 did not (Figures S1C and S1D). Thus, the inheritance of the Ura+ phenotype requires the continuous activity of Hsp104. As noted above, the self-templating structure for most prions is an amyloid ber. Most amyloids are extraordinarily resistant to solubilization by detergents, and this property can be used to distinguish amyloids from other noncovalent protein complexes (Alberti et al., 2009). To verify that the prion responsible for the Ura+ phenotype is indeed formed by Mot3, we used semidenaturing detergent-agarose gel electrophoresis (SDD-AGE).
SDD-AGE resolves amyloids from nonamyloid aggregates and soluble proteins based on size and insolubility in SDS. By probing SDD-AGE blots for the naturally occurring hexa-histidine motif of Mot3, we found that SDS-resistant aggregates of Mot3 occurred in Ura+ cells, but not in ura cells (Figure S1B). We conclude that the Ura+ cells are [MOT3+]. Having isolated [MOT3+] by virtue of the transcriptional derepression of the DAN1 locus, we next asked if FLO11 was simultaneously affected. We prepared mRNA from isogenic [MOT3+] and [mot3] cells and used quantitative RT-PCR (qRT-PCR) to evaluate FLO11 expression. Indeed, [MOT3+] increased FLO11 transcripts by 10-fold (Figure 1D). [MOT3+] Governs the Acquisition of Multicellular Phenotypes FLO11 expression mediates the development of diverse multicellular phenotypes in response to specic environmental signals. Under standard nutrient-rich laboratory growth conditions, [MOT3+] conferred only modest FLO11-related phenotypes (data not shown). We therefore explored the synergistic effects of [MOT3+] with environmental conditions that naturally induce FLO11. The best characterized of these is nitrogen starvation. When challenged with limiting or poor nitrogen sources, cells elongate, bud in a unipolar fashion, and remain attached after cell division (Gimeno et al., 1992). The resulting chains of cells extend well beyond the original colony boundaries and can even invade the underlying growth substratum. To explore the effects of [MOT3+] on invasive growth, we plated cells on media in which the only source of nitrogen is a poor one, proline. After 5 days of growth, noninvasive cells were dislodged by rinsing the plates vigorously with running water. Prion minus cells, [mot3], were easily washed away. In contrast, [MOT3+] colonies acquired invasive laments that could not be dislodged (Figure 2A). In addition to invasive growth in response to nitrogen starvation, Flo11 can induce complex colony architectures in response to starvation for fermentable carbon sources (Granek and Magwene, 2010). Emerging evidence suggests that such colonies are, in fact, rudimentary biolms that protect cells from stress chova et al., and enable cells to cooperate metabolically (Va 2011). To investigate, we plated [MOT3+] and [mot3] cells to media containing a variety of different carbon sources. [mot3] cells formed smooth, simple colonies regardless of carbon source. [MOT3+] colonies were indistinguishable from those of [mot3] on glucose and galactose, which are fermentable carbon sources. But on glycerol and ethanol, which are nonfermentable, [MOT3+] cells formed elaborate colonies with prominent ridges and invaginations (Figure 2A; data not shown). Another stimulus of Flo11-dependent colony differentiation is growth on a semisolid substratum (Reynolds and Fink, 2001). When grown on semisolid media, with ethanol as a carbon source, [MOT3+] cells differentiated into an elaborate compound structure consisting of well-developed microcolonies embedded in a transparent gelatinous matrix (Figure 2B). Notably, the matrix was impenetrable to isogenic yeast cells applied to the exterior of the colony (Figure S2). In striking contrast, [mot3] colonies again failed to differentiate, instead producing a simple smooth colony lacking a prominent gelatinous exterior.
Figure 2. [MOT3+] Enables Facultative Multicellular Growth

(A) Schematics on the left depict multicellular phenotypes (dark cells) that are induced by each of the common growth conditions indicated. On plates containing only proline as a nitrogen source, [MOT3+] cells formed invasive laments that penetrated the agar surface (top; invasive growth remains after dislodging surface cells under running water). On plates containing only ethanol as a carbon source, [MOT3+] cells formed complex colony morphol chova ogies (middle). The schematic for colony morphology is based on Va et al. (2011). When grown to saturation in liquid media, [MOT3+] cells exhibited a decreased tendency to occulate (bottom). Each of these behaviors of [MOT3+] cells was eliminated by treating the cells transiently with GdHCl, an inhibitor of the prion-partitioning factor Hsp104. (B) [MOT3+] cells form elaborate, biolm-like structures when grown on semisolid ethanol media. Scale bars, 1 cm. See also Figure S2.
Finally, we examined a quite different multicellular behavior. Some yeast strains clump together, or occulate, toward the end of fermentation in liquid media. Unlike phenotypes typically driven by Flo11, which result from cell-cell interactions between mother cells and their daughters, occulation results from horizontal association between cells. This behavior is industrially desirable but notoriously unstable; occulation competence is frequently and stochastically gained and lost (Verstrepen et al., 2003). We asked if [MOT3+] can also contribute to occulation phenotypes by allowing cells to grow to saturation in rich liquid media. [MOT3+] cultures became moderately occulent 1 day before [mot3] cultures. The nal extent of occulation, however, was greater in [mot3] cultures (Figure 2A). The occulation phenotype, in both [MOT3+] and [mot3] cells, was driven by Flo1, a cell-wall-anchored adhesin similar to Flo11 (Supplemental Information). Like FLO11, the FLO1 promoter contains numerous binding motifs for Mot3.
Figure 3. The N-Terminal Region Is Required for Prion-Mediated Inactivation of Mot3

(A) A schematic of full-length Mot3 and two functionally distinct truncation mutants. Regions of high predicted prion-forming propensity are indicated in red. The C-terminal nonprion-like region of Mot3 contains two C2H2 zinc ngers, indicated in blue, which are involved in DNA binding. The prion-determining region, PrD, as initially dened by Alberti et al. (2009) encompasses the N-terminal 295 residues of the protein. A poly-asparagine (poly-N) tract stretches from residues 143157. In the present study, we construct a variant, DPrD, that lacks much of the PrD including the poly-N tract. An endogenous hexa-histidine motif is indicated in green. (B) Ectopic expression of DPrD ([DPrD) from a constitutive promoter (TEF1) suppresses the Ura+ phenotype of [MOT3+]. In contrast, [MOT3+] cells ectopically expressing full-length Mot3 ([WT) remain Ura+. This is because WT Mot3, but not DPrD, accumulates as SDS-resistant aggregates, as visualized by SDD-AGE and immunoblotting against the endogenous hexa-histidine motif. Amyloids from endogenous Mot3 are too low abundance to be detected in this exposure. An SDS-PAGE blot of lysates boiled in 2% LDS demonstrates comparable expression of the Mot3 variants.
[MOT3+] Prion Phenotypes Result from Both Losses and Gains of Mot3 Function Most prion phenotypes mimic loss-of-function mutations in the genes that encode them. To determine whether [MOT3+] phenotypes are due to simple inactivation of Mot3, we employed two genetic approaches. First, we asked whether the prions phenotypes could be complemented by supplying Mot3s normal cellular activity in trans. To this end, we constructed a variant of Mot3 with a deletion in the prion-forming region of the protein, known as the PrD (Alberti et al., 2009). This variant, DPrD, was immune to prion-mediated inactivation. When expressed from a plasmid in [MOT3+] cells, it maintained solubility and fully reversed the Ura+ phenotype (Figure 3). It also reversed the complex colony morphology and the hypoocculent phenotypes (Figure 4A). Thus, a loss of Mot3s normal cellular activity is necessary for the prion phenotypes. Importantly, when the plasmid expressing DPrD was lost, [MOT3+] phenotypes returned, demonstrating that DPrD simply masked the prion phenotypes while allowing the full-length protein to maintain the prion state. Next, we asked if genetic deletion of MOT3 conferred the same spectrum of phenotypes as [MOT3+]. To a limited extent, it did: Dmot3 cells formed rufed colonies on ethanol media and invasive laments on nitrogen-limited media (Figure 4B). However, other phenotypes were not fully recapitulated. Whereas [MOT3+] prions altered both the kinetics and magnitude of occulation, Dmot3 cells were entirely nonocculent. Most surprisingly, deletion of MOT3 did not fully derepress the URA3 reporter for Mot3 activity. Instead, derepression was only partially penetrant: Dmot3 cells were predominantly ura but produced abundant papilla with unstable Ura+ phenotypes (Figure 4C). What might explain the difference between [MOT3+] and Dmot3 phenotypes? Taking advantage of the facile readout of the URA3 reporter, we investigated potential gains of function by the prion. We rst tested if ectopically expressed prion particles could stabilize a Ura+ phenotype in Dmot3 cells. To do so,
we transformed cells with a plasmid that overexpressed fulllength Mot3 from the strong TEF1 promoter. These cells became Ura+ at an increased frequency (Figure 4D). The phenotype was not maintained when the plasmid was lost, indicating that it required continuous expression of Mot3 (data not shown). To verify that the phenotype was due to PrD-mediated aggregation, and not some other effect of overexpressed Mot3, we repeated the experiment with plasmids encoding either the DPrD variant of Mot3 or the highly aggregation-prone PrD alone (Alberti et al., 2009). As expected, expression of DPrD suppressed the formation of Ura+ colonies. To our surprise, however, expression of the PrD had no effect (Figure 4D). These results suggest that the Mot3 prion templates are themselves insufcient to stabilize the Ura+ phenotype of Dmot3 cells. Instead, the stable phenotype requires the prion-mediated sequestration of a region of the Mot3 protein that is outside the prion-forming region. It seems likely that the gain of function of [MOT3+] results from the cosequestration of one or more transcriptional corepressors that interact with this region of Mot3, as illustrated in Figure 4E. A Typical Environmental Condition, Ethanol Stress, Induces [MOT3+] Multicellular behaviors are commonly induced by environmental stresses. At least one prion, formed by the translation termination factor, Sup35, also shares this property (Tyedmers et al., 2008). To determine if [MOT3+] can be induced by stress, we exposed [mot3] cells to a variety of chemical stressors. Following 6 hr of treatment, cells were plated to uracil-decient media to select for those that had acquired [MOT3+]. Most of the treatments did not change the frequency of [MOT3+] (Figure S3A; data not shown). Only one stressor, ethanol, had an effect. Treating cells for 6 hr with a concentration of ethanol that is commonly reached in wine fermentations (12%) increased the frequency of Ura+ colonies by 10-fold (Figures 5A and S3). Cells bearing a DPrD MOT3 gene were unable to form Ura+ colonies in response to ethanol. Importantly, the treatment was
Figure 4. Prion-Mediated Inactivation of Mot3 Is Required but Is Not Sufcient to Convey the Full Spectrum of [MOT3+] Phenotypes
(A) Ectopic expression of DPrD ([DPrD) from a constitutive promoter (TEF1) suppresses the complex colony morphology, invasion, and hypoocculation phenotypes of [MOT3+] cells. In contrast, [MOT3+] cells that ectopically express full-length Mot3 ([WT) retain [MOT3+] phenotypes. (B) Genetic deletion of MOT3 confers some phenotypes that mimic those of [MOT3+], including biolm formation and invasion. Dmot3 cells do not occulate when grown to saturation in liquid media. (C) Genetic deletion of MOT3 is not sufcient to fully derepress the DAN1 promoter (PDAN1-URA3). Unlike [MOT3+] cells (which are Ura+), Dmot3 cells partition dynamically between fully repressed (ura) and derepressed (Ura+) states. (D) Ectopic expression of WT Mot3 stabilizes the derepressed (Ura+) state, whereas the nonaggregating variant of Mot3, DPrD, restores the fully repressed (ura) state. Expression of PrD or empty vector has no effect. Duplicate transformants are shown, and serial dilutions were made onto the same plate. Bottom view is a western blot comparing expression levels of the Mot3 variants. (E) In a hypothetical model for the gain of function of Mot3 prion conformers, Mot3 interacts with another protein that corepresses transcription from the DAN1 promoter. In the [MOT3+] state, Mot3 prion particles sequester this protein, resulting in the full derepression of the DAN1 promoter. In cells altogether lacking Mot3, this protein can still bind to and partially repress the DAN1 promoter. Scale bars, 1 cm.
not overtly toxic to cells, indicating that ethanol did not merely select for pre-existing [MOT3+] cells. How does ethanol stimulate Mot3 prion conversion? It did not simply increase Mot3 protein levels (Figure S3B). Ethanol stress causes protein misfolding and strongly induces the proteinremodeling factor Hsp104 (Figure S3B; Sanchez et al., 1992). Hsp104 has previously been found to stimulate prion conversion by other prion proteins (Shorter and Lindquist, 2006; Kryndushkin et al., 2011). To determine if Hsp104 also stimulates [MOT3+], we transformed [mot3] cells with a plasmid expressing Hsp104 from a strong constitutive promoter. This produced a greater than 10-fold increase in the frequency of Ura+ colonies (Figure 5B). Overexpressed Hsp104 did not induce Ura+ colonies in cells bearing a DPrD MOT3 gene. These experiments suggest that ethanol stress accelerates the formation of [MOT3+] by perturbing protein homeostasis.
Another Typical Environmental Condition, Hypoxia, Eliminates [MOT3+] A strategy used by yeast cells to inhibit competition from other organisms is to rapidly ferment glucose to ethanol. The subsequent respiration of that ethanol can produce an anoxic environment by the end of the respiro-fermentative cycle. Hypoxia represses MOT3 transcription (Figure S3C; Sertil et al., 2003). To test if this is sufcient to eliminate [MOT3+] prions and reset cells back to the [mot3] state, we passaged [MOT3+] cells under hypoxic (<1% O2) conditions. After 5 days, cells were repassaged on fresh nonselective media and incubated under normal atmospheric conditions. Colonies were then replicaplated to uracil-decient media to determine the presence or absence of [MOT3+]. The hypoxic treatment proved remarkably effective: [MOT3+] cultures reverted uniformly back to [mot3] (Figure 5C). Cells that were passaged in parallel under normoxic conditions remained [MOT3+]. Notably, all [MOT3+]-associated phenotypes, including the ethanol- and proline-dependent multicellular growth forms, were also eliminated. Thus, by way of a [MOT3+] to [mot3] prion switch, transient oxygen depletion
Figure 5. Environmental Conditions Govern Mot3 Prion Switching

(A) The spontaneous switching of Mot3 to its prion state is increased by ethanol stress. Diploid [mot3] cells, or as a negative control, WT/DPrD heterozygotes, were incubated for 6 hr in media alone (mock), or media containing 12% ethanol (EtOH), prior to plating to media lacking uracil. Ura+ colonies were counted after 4 days and normalized by the number of total colony-forming units. Data represent mean SD from three independent experiments. (B) [mot3] cells were transformed with either empty vector or a plasmid expressing Hsp104 from a strong constitutive promoter (GPD). Transformants were inoculated to rich media overnight, and then the fraction of [MOT3+] cells was determined as in (A). (C) Passaging [MOT3+] cells under hypoxic conditions reverts them to [mot3]. (D) The effect of hypoxia is specic to [MOT3+]. A strain harboring prion states of three different proteins ([MOT3+], [PSI+], and [RNQ+]) was treated transiently with either GdHCl or hypoxia. Amyloids representing each prion state were detected by SDD-AGE. GdHCl, which targets the prion-partitioning factor Hsp104, eliminated all three prions, whereas hypoxia specically eliminated [MOT3+]. See also Figure S3.
induced specic and heritable changes to respiration-associated phenotypes. To determine if the effect of hypoxia is specic to [MOT3+], we created a single yeast strain harboring the prion states of three different proteins: Mot3, Sup35, and Rnq1. We then subjected cells of this strain to hypoxia followed by SDD-AGE to detect the continued presence of amyloids of each prion protein. Mot3 amyloids were eliminated. Sup35 and Rnq1 amyloids remained (Figure 5D). Hypoxia did not simply counterselect [MOT3+] cells (Figures S3D and S3E). To verify that the unique response of [MOT3+] to hypoxia results from the transcriptional repression of MOT3, we supplemented [MOT3+] cells with Mot3 expressed from a heterologous promoter (SUP35) that is not regulated by hypoxia. When these [MOT3+] cells were passaged under hypoxia, they retained the prion state (Figure S3F). Cells that contained empty vector did not, demonstrating that repression of MOT3 under hypoxia is required for prion elimination. This natural transcriptional regulation of MOT3 ensures that [MOT3+] prions will be eliminated from cells as they reach the end of the respirofermentative cycle. [MOT3+] Produces Different Multicellular Behaviors in Different Genetic Backgrounds Natural yeast isolates exhibit a rich diversity of multicellular behaviors and fermentative characteristics (Casalone et al.,
2005; Liti et al., 2009; Granek and Magwene, 2010). The profound effects that [MOT3+] exerted on FLO11-dependent phenotypes in the laboratory strain suggested that it might also contribute to the acquisition of new phenotypes in the wild. To analyze Mot3 prion formation in nonlaboratory yeast, we integrated a dominant drug-resistance reporter for Mot3 activity (Figure 6A) into genetically and ecologically diverse, yet experimentally tractable, natural isolates of S. cerevisiae (Liti et al., 2009). We then overexpressed the Mot3 prion-determining region and selected [MOT3+] derivatives on drug-containing media. Of 32 strains tested (Table S3), 28 (Figure S4A) readily acquired the resistance phenotype indicative of the prion. We further focused on ten divergent strains (Figure 6B) that retained strong resistance phenotypes even after loss of the inducing plasmid (as expected for cells harboring the prion). When examined by SDD-AGE, they contained Mot3 amyloids (Figure S4C). Hypoxia eliminated the phenotype and eliminated the amyloids (Figures S4B and S4C). These results demonstrate that diverse genetic backgrounds have the capacity to acquire [MOT3+] and to naturally regulate its inheritance through a hypoxic growth phase. We then compared the Flo11-dependent phenotypes of [MOT3+] isolates and their [mot3] counterparts. These differed widely between genetic backgrounds (Figures 6C, 6D, and S4DS4H; Table S4). Two strains, the grape/wine isolates L-1374 and Y-55, exhibited strong increases in multiple multicellular behaviors. In the absence of the prion, they were noninvasive and formed smooth colonies under all growth conditions.
Figure 6. [MOT3+] Phenotypes Are Dictated by Genetic Variation

(A) A genetic reporter for [MOT3+] in prototrophic strains. A gene encoding resistance to the antibiotic nourseothricin (NAT) was placed under control of the DAN1 promoter, such that yeast carrying the reporter are NAT resistant only when they are [MOT3+]. (B) The reporter was integrated into genetically and ecologically diverse strains of S. cerevisiae. Each node on the phylogenetic tree (adapted from Liti et al., 2009) represents a different yeast strain that could be induced to form a NAT-resistant phenotype by overexpressed Mot3 PrD. Labels denote strains that were evaluated for [MOT3+]-dependent multicellular phenotypes. Green labels indicate that [MOT3+] produced multiple robust multicellular phenotypes, yellow indicates more limited effects of [MOT3+], and red indicates no observable effect. The two strains labeled in blue exhibited [MOT3+]-dependent changes in occulation. (C) [MOT3+] produces a different array of multicellular phenotypes in each strain, as demonstrated by invasive behaviors and colony morphologies in four divergent strains. ferm., fermenting.
But when they contained [MOT3+], they became invasive on proline media and acquired complex colony morphologies on ethanol media. The colony morphologies were also manifest, to a lesser extent, on glucose media. [MOT3+] had specic multicellular effects in other strains. In the fermenting fruit juice isolate DBVPG6040, [MOT3+] conferred a strong invasive phenotype but had no effect on colony morphology. In contrast, the tecc (honey wine) isolate DBVPG1853 had altered colony morphologies in the [MOT3+] state but acquired invasive laments independently of prion status. Only one of the ten strains investigated, the soil isolate DBVPG1373, had no apparent multicellular effect of [MOT3+]. There are two general mechanisms by which [MOT3+] might produce different phenotypes in different genetic backgrounds. First, the physical properties of the prion (including the nature of the prion fold and its efciency of propagation) may themselves be inuenced by the genetic background, which in turn would inuence how it interacts with Mot3-regulated loci. Second, [MOT3+] may regulate (directly or indirectly) genes that contain polymorphisms that are phenotypically silent in the absence of the prion. To distinguish between these possibilities, we deleted MOT3 from two genetically divergent strains, L-1374 and UWOPS83-787.3, that had very different phenotypic manifestations of [MOT3+]. If the causative genetic differences acted exclusively by inuencing the prion itself, we would expect the deletion to have the same phenotypic effect in both backgrounds. If, conversely, the causative genetic differences are themselves subject to prion regulation, the genetic ablation of Mot3 activity would be expected to produce different effects in each background. As shown in Figure 6D, the latter scenario proved true: MOT3 deletions produced different effects in the two strains and, in both, phenocopied [MOT3+]. Altogether, these data demonstrate that [MOT3+] alters the expression of standing genetic variation in S. cerevisiae, resulting in combinations of multicellular phenotypes that differ between lineages. To verify the role of FLO11 in [MOT3+]-dependent multicellularity, we deleted FLO11 from a [MOT3+] isolate of L-1374. The resulting cells lost the multicellular phenotypes of [MOT3+], despite retaining the prion itself (Figure 6D; data not shown). The number and arrangement of Mot3 binding motifs in the FLO11 promoter did not differ signicantly between strains and, therefore, are unlikely to account for [MOT3+]-dependent phenotypic differences (Table S1). Instead, we discovered that the FLO11-coding region is itself highly polymorphic, with different strains harboring different repeat length variants (Figure S4I). The number of tandem repeats within cell surface adhesins correlates with their activity (Verstrepen et al., 2005), suggesting that prion formation by Mot3 could enable the expression of functional differences between FLO11 alleles. Differences in the activities of the other numerous regulators of
FLO11 are quite likely to further synergize with [MOT3+] in diversifying multicellular behaviors. To gain insight into the extent of [MOT3+]-induced transcriptional changes, we analyzed the expression of other Mot3regulated genes in both L-1374 and UWOPS83-787.3. Using qRT-PCR, we found that [MOT3+] derepressed loci throughout the genomes of both strains (Figure 6E; Table S5). Two important trends emerged. First, the effects of genetic background were quite specic. Although multiple transcripts exhibited quantitatively different responses between strains, only FLO11 differed qualitatively: it was derepressed by [MOT3+] in L-1374 but was unaffected in UWOPS83-787.3. Second, the relative effect of [MOT3+] and Dmot3 differed among individual genes. DAN1, for example, was derepressed much more strongly by the prion than by the deletion, whereas the reverse was true for ANB1. Therefore, the gain of function of [MOT3+] appears to impact a specic subset of the Mot3 regulon. These ndings illustrate that, despite genetic divergence between the strains leading to distinct multicellular effects of [MOT3+], the unique transcriptional response of the prion is nevertheless shared between them. DISCUSSION The evolutionary constraints on multicellularity are stringent (Grosberg and Strathmann, 2007). Multicellularity involves mechanisms for cell differentiation, cooperation, and the control of defector cells that might promote their own growth at the expense of the integrity of the multicellular structure. We have demonstrated that a self-perpetuating conformational switch in a protein provides a mechanism to drive multicellular growth in a unicellular organism. The Mot3 transcription factor lies at the helm of a Flo11dependent developmental program. The successful deployment of that program necessitates an epigenetic commitment that spans multiple generations. The self-perpetuating nature of prion propagation ensures that commitment. Prion conversion by Mot3 activates the multicellularity program and stably perpetuates that state to subsequent generations. By virtue of their common descent and shared prion inheritance, all cells in the lineage remain physically and metabolically connected. The complex heritable phenotypes that emerge from this cooperation would be difcult to achieve by a group of cells each acting independently. Yeast multicellularity is deeply intertwined with carbon metabolism. The multicellularity determinants FLO11 and FLO1 are coregulated with glucose, sucrose, and starch metabolic programs (Verstrepen and Klis, 2006). Mot3 appears to be an important regulator of carbon metabolism: its targets are most heavily enriched for sugar transporters, sterol biosynthetic enzymes, and alcohol dehydrogenases (Table S2). Mathematical
(D) [MOT3+]-dependent morphological differences between L-1374 and UWOPS83-787.3 are reproduced by deleting MOT3. [MOT3+]-dependent multicellularity in L-1374 requires FLO11. (E) Demonstrative qRT-PCR analyses of the expression of Mot3-regulated genes for isogenic [MOT3+], [mot3], and Dmot3 cells in L-1374 and UWOPS83-787.3. The direct effect of Dmot3 on FLO11 was not determined. Error bars represent SD from separate experiments. Scale bars, 1 cm. See also Figure S4 and Tables S4 and S5.
models suggest that multicellularity may have evolved in response to competition between carbon metabolism strategies (Pfeiffer et al., 2001), a hypothesis supported by correlations between dimorphic transitions and glucose depletion (Verstrepen and Klis, 2006). Thus, Mot3 establishes a paradigm that links prions with carbon metabolism and facultative multicellularity. Importantly, there may be many such prions. Two major regulators of carbon metabolism and multicellularity, Cyc8 and Swi1, were recently found to form prions (Du et al., 2008; Patel et al., 2009). Another prion, which involves the transmembrane proton pump, Pma1, also governs carbon source utilization (Brown and Lindquist, 2009). Prion Formation as a Potential Function of Mot3 Mot3 is a key player in the cellular response to oxygen availability. When oxygen is limiting, Mot3 levels rapidly decrease, resulting in the derepression of its genetic targets (Sertil et al., 2003; Lai et al., 2006). Paradoxically, however, hypoxia does not produce the same multicellular responses as [MOT3+]. In fact, oxygen is strictly required for Flo11-dependent colony morphologies and lamentation (Wright et al., 1993; Zupan and Raspor, 2010). It seems that prion conversion, therefore, enables phenotypic manifestations of the Mot3 regulon that may otherwise be inaccessible. How might Mot3 regulation, via prion formation, be advantageous during aerobic growth? [MOT3+] effectively primed cells to respond to limitations in either glucose or nitrogen. Under those conditions, yeast cells fulll most of their carbon and nitrogen needs by oxidative metabolism. Ethanol is the predominant source of carbon in the postdiauxic phase of wine fermentations, and proline is by far the most abundant source of nitrogen in grapes (Ough and Stashak, 1974). Each of these nutrients can only be metabolized in the presence of oxygen (Ingledew et al., 1987). Indeed, the strongest [MOT3+]-dependent responses occurred when cells were forced to grow with either ethanol as a carbon source or proline as a nitrogen source. Physiological responses to these nutrients would be counterproductive during anaerobic growth. The natural regulation of MOT3 by hypoxia parsimoniously overrides the carbon- and nitrogenlimitation signals by eliminating [MOT3+]. Why should cells employ a priona semistochastic switch to regulate fundamental metabolic pathways? Prion switching divides large populations into mosaics of distinct subpopulations, and recent work reveals that cell-to-cell heterogeneity in metabolic processes can be maintained through cooperative interactions that benet the population as a whole (Beardmore et al., 2011; MacLean, 2008). For a population of [mot3] and [MOT3+] cells, the whole may be greater than the sum of its parts. Environmentally Regulated Inheritance Our data are consistent with a function for [MOT3+] during postdiauxic growth. Two environmental signals that delimit the beginning and end of that period, ethanol and hypoxia, inversely regulated prion switching by Mot3 (Figure 7). These signals naturally happen in sequence: as ethanol concentrations reach a peak, yeast switch from fermentative to respirative metabolism. The oxidative respiration of ethanol and remaining sugars
Figure 7. Model for a Function for Mot3 Prion Switching in the Respiro-Fermentative Cycle of Wine Yeasts
Yeast cells begin the cycle upon inoculation to glucose-rich grape must. Glucose is fermented to ethanol during the exponential growth phase. The ensuing ethanol stress triggers prion conversion to [MOT3+], which, in conjunction with glucose exhaustion, drives some of the cells into a multicellular growth program that protects them from stress and/or increases metabolic efciency. As the population respires ethanol and remaining sugars, oxygen levels decline resulting in the accelerated reappearance of [mot3] cells within [MOT3+] subpopulations.
then depletes molecular oxygen from the local environment (Bauer and Pretorius, 2000). The ensuing hypoxia reverts cells back to the [mot3] state. Accordingly, Mot3 prion switching may constitute an explicit mechanism that contributes to the recently discovered (Mitchell et al., 2009) capacity of yeast to adaptively anticipate the environmental changes that occur sequentially in the course of wine production. With the exception of phenomena that involve the uptake of foreign nucleic acids (Koonin, 2012), the elimination of [MOT3+] by hypoxia is, to our knowledge, the only example of a stably inherited phenotypic change that is induced en masse by a specic environmental change. Complexity of Multicellular Phenotypes [MOT3+] differs in two key respects from other mechanisms that underlie rapid morphological switching in yeast. First, unlike other genetic and epigenetic elements, [MOT3+] is structurally autonomous from nucleic acids, enabling it to act in a dominant and promoter-extrinsic manner to alter gene expression throughout the genome. Second, the prion is not simply an on/off switch for Mot3. Instead, it represses the proteins normal activity while endowing it with new activities. Both effects are required for the full spectrum of [MOT3+] phenotypes. One simple explanation for the gain of function of [MOT3+] is that the prion particles sequester other transcriptional regulators that interact with Mot3. Indeed, Mot3 functions within a regulatory framework rife with other transcription factors whose sequences are particularly well suited for prion-like interactions. To what extent these proteins inuence each others aggregation and how this, in turn, contributes to the complexity and
plasticity of [MOT3+] phenotypes are exciting subjects for exploration. In sum, we have demonstrated that prions formed by a yeast transcription factor enable the differentiation of yeast cells into facultative multicellular structures. The formation, inheritance, and phenotypic manifestation of the prions were dictated by natural environmental signals and by an abundance of heritable genetic variation between yeast isolates. These relationships explicate some of the natural diversity and plasticity of multicellular phenotypes and promise to further our understanding of the molecular mechanisms that enable the evolution of social behaviors.
EXPERIMENTAL PROCEDURES Computational Analyses Genetic targets of transcription factors were obtained from YEASTRACT (Teixeira et al., 2006) and from Wong and Struhl for the Cyc8-Tup1 complex (Wong and Struhl, 2011). Prion propensities were predicted using a previously developed algorithm (Alberti et al., 2009) with parameters updated according to experimentally conrmed prion hits. An independent algorithm (Toombs et al., 2012) was also used, with comparable results. Functional annotations were determined using the functional clustering analysis in DAVID release 6.7 (Huang et al., 2009) with the highest classication stringency. Mot3 binding motifs in the 3 kb regulatory region 50 upstream of the FLO11 gene (Rupp et al., 1999) were identied in all strains according to a frequency matrix for the Mot3 motif (Bryne et al., 2008). Yeast Genetics and Molecular Biology Standard cloning procedures were used (Alberti et al., 2009). Oligos and plasmid sequences are available upon request. Chromosomal integrations were achieved using PCR-based mutagenesis and the delitto perfetto method (Goldstein and McCusker, 1999; Storici and Resnick, 2006). Yeast were transformed with a standard lithium-acetate protocol as described by Gietz et al. (1992). Yeast strains are listed in Table S3. Yeast Media and Phenotypic Analyses Standard yeast media and growth conditions were used. Where appropriate, glucose was replaced with 2% ethanol or glycerol, and ammonium sulfate was replaced with 0.05% proline. Semisolid media contained 0.3% agar. For agar invasion analyses, colonies were allowed to grow for 5 days at 25 C. Plates were photographed before and after the removal of noninvasive cells by dislodging surface cells under running water. Flocculation was assayed after growth to saturation in YPD containing 20% glucose. Hypoxia Treatments Cells were point inoculated to YPD plates supplemented with Tween 80 and ergosterol (Hongay et al., 2002) and placed in a sealed bag containing a BD GasPak EZ Anaerobe Pouch System sachet. After 5 days at 30 C, cells were retrieved from the outer perimeter of colonies to a fresh YPD plate and allowed to form colonies under standard laboratory conditions prior to subsequent analyses. SDD-AGE SDD-AGE was performed as described (Halfmann et al., 2012). qRT-PCR Total RNA was extracted using MasterPure Yeast RNA Purication Kit (Epicenter) as instructed by the manufacturer. cDNAs were generated using oligo(dT)20 and the Superscript III First-Strand Synthesis System (Invitrogen). qPCR was carried out using the DyNAmo HS SYBR Green qPCR Kit (Thermo Fisher Scientic) on the HT 7900 Real-Time PCR System (Applied Biosystems). RQ values were normalized against those for ACT1 or TFC1. Primers are listed in Table S6.
SUPPLEMENTAL INFORMATION Supplemental Information includes Extended Experimental Procedures, four gures, and six tables and can be found with this article online at http://dx. doi.org/10.1016/j.cell.2013.02.026. ACKNOWLEDGMENTS This work was supported primarily by the NIH Directors Early Independence Award Program, grant number DP5-OD009152-01 (to R.H.) and by the Sara and Frank McKnight Fellowship (R.H.). S.L. is a Howard Hughes Medical Institute (HHMI) investigator. This work was supported by grants from the G. Harold and Leila Y. Mathers Foundation and HHMI (to S.L.). We thank Cintia Hongay for fruitful early discussions about Mot3, Oliver King for assistance with the prion prediction algorithm, and Sunil Laxman and members of the S.L. lab for critical reading of the manuscript. We thank Gerry Fink and Fred Winston for yeast strains S1278b and FY2609, respectively, and Mikko Taipale for unpublished plasmids. Received: September 29, 2012 Revised: December 22, 2012 Accepted: February 12, 2013 Published: March 28, 2013 REFERENCES Alberti, S., Halfmann, R., King, O., Kapila, A., and Lindquist, S. (2009). A systematic survey identies prions and illuminates sequence features of prionogenic proteins. Cell 137, 146158. Barrales, R.R., Korber, P., Jimenez, J., and Ibeas, J.I. (2012). Chromatin modulation at the FLO11 promoter of Saccharomyces cerevisiae by HDAC and Swi/ Snf complexes. Genetics 191, 791803. Bauer, F.F., and Pretorius, I.S. (2000). Yeast stress response and fermentation efciency: how to survive the making of winea review. S. Afr. J. Enol. 21, 2751. Beardmore, R.E., Gudelj, I., Lipson, D.A., and Hurst, L.D. (2011). Metabolic trade-offs and the maintenance of the ttest and the attest. Nature 472, 342346. Brown, J.C., and Lindquist, S. (2009). A heritable switch in carbon source utilization driven by an unusual yeast prion. Genes Dev. 23, 23202332. sch, H.U. (2012). Choosing the right lifestyle: adhesion ckner, S., and Mo Bru and development in Saccharomyces cerevisiae. FEMS Microbiol. Rev. 36, 2558. Bryne, J.C., Valen, E., Tang, M.H., Marstrand, T., Winther, O., da Piedade, I., Krogh, A., Lenhard, B., and Sandelin, A. (2008). JASPAR, the open access database of transcription factor-binding proles: new content and tools in the 2008 update. Nucleic Acids Res. 36(Database issue), D102D106. Bumgarner, S.L., Neuert, G., Voight, B.F., Symbor-Nagrabska, A., Grisa, P., van Oudenaarden, A., and Fink, G.R. (2012). Single-cell analysis reveals that noncoding RNAs contribute to clonal heterogeneity by modulating transcription factor recruitment. Mol. Cell 45, 470482. Carter, G.W., Prinz, S., Neou, C., Shelby, J.P., Marzolf, B., Thorsson, V., and Galitski, T. (2007). Prediction of phenotype and gene expression for combinations of mutations. Mol. Syst. Biol. 3, 96. Casalone, E., Barberio, C., Cappellini, L., and Polsinelli, M. (2005). Characterization of Saccharomyces cerevisiae natural populations for pseudohyphal growth and colony morphology. Res. Microbiol. 156, 191200. Chernoff, Y.O., Lindquist, S.L., Ono, B., Inge-Vechtomov, S.G., and Liebman, S.W. (1995). Role of the chaperone protein Hsp104 in propagation of the yeast prion-like factor [psi+]. Science 268, 880884. Chiti, F., and Dobson, C.M. (2006). Protein misfolding, functional amyloid, and human disease. Annu. Rev. Biochem. 75, 333366. Derkatch, I.L., Bradley, M.E., Hong, J.Y., and Liebman, S.W. (2001). Prions affect the appearance of other prions: the story of [PIN(+)]. Cell 106, 171182.
Du, Z., Park, K.W., Yu, H., Fan, Q., and Li, L. (2008). Newly identied prion linked to the chromatin-remodeling factor Swi1 in Saccharomyces cerevisiae. Nat. Genet. 40, 460465. Ferreira, P.C., Ness, F., Edwards, S.R., Cox, B.S., and Tuite, M.F. (2001). The elimination of the yeast [PSI+] prion by guanidine hydrochloride is the result of Hsp104 inactivation. Mol. Microbiol. 40, 13571369. Gietz, D., St Jean, A., Woods, R.A., and Schiestl, R.H. (1992). Improved method for high efciency transformation of intact yeast cells. Nucleic Acids Res. 20, 1425. Gimeno, C.J., Ljungdahl, P.O., Styles, C.A., and Fink, G.R. (1992). Unipolar cell divisions in the yeast S. cerevisiae lead to lamentous growth: regulation by starvation and RAS. Cell 68, 10771090. Goldstein, A.L., and McCusker, J.H. (1999). Three new dominant drug resistance cassettes for gene disruption in Saccharomyces cerevisiae. Yeast 15, 15411553. Granek, J.A., and Magwene, P.M. (2010). Environmental and genetic determinants of colony morphology in yeast. PLoS Genet. 6, e1000823. Griswold, C.K., and Masel, J. (2009). Complex adaptations can drive the evolution of the capacitor [PSI], even with realistic rates of yeast sex. PLoS Genet. 5, e1000517. Grosberg, R.K., and Strathmann, R.R. (2007). The evolution of multicellularity: a minor major transition? Annu. Rev. Ecol. Evol. Syst. 38, 621654. Hahn, M.W., De Bie, T., Stajich, J.E., Nguyen, C., and Cristianini, N. (2005). Estimating the tempo and mode of gene family evolution from comparative genomic data. Genome Res. 15, 11531160. Halfmann, R., and Lindquist, S. (2010). Epigenetics in the extreme: prions and the inheritance of environmentally acquired traits. Science 330, 629632. Halfmann, R., Jarosz, D.F., Jones, S.K., Chang, A., Lancaster, A.K., and Lindquist, S. (2012). Prions are a common mechanism for phenotypic inheritance in wild yeasts. Nature 482, 363368. Halme, A., Bumgarner, S., Styles, C., and Fink, G.R. (2004). Genetic and epigenetic regulation of the FLO gene family generates cell-surface variation in yeast. Cell 116, 405415. Hongay, C., Jia, N., Bard, M., and Winston, F. (2002). Mot3 is a transcriptional repressor of ergosterol biosynthetic genes and is required for normal vacuolar function in Saccharomyces cerevisiae. EMBO J. 21, 41144124. Huang, W., Sherman, B.T., and Lempicki, R.A. (2009). Systematic and integrative analysis of large gene lists using DAVID bioinformatics resources. Nat. Protoc. 4, 4457. Ingledew, W.M., Magnus, C.A., and Sosulski, F.W. (1987). Inuence of oxygen on proline utilization during the wine fermentation. Am. J. Enol. Vitic. 38, 246248. Koonin, E.V. (2012). Does the central dogma still stand? Biol. Direct 7, 27. Koschwanez, J.H., Foster, K.R., and Murray, A.W. (2011). Sucrose utilization in budding yeast as a model for the origin of undifferentiated multicellularity. PLoS Biol. 9, e1001122. Kryndushkin, D.S., Engel, A., Edskes, H., and Wickner, R.B. (2011). Molecular chaperone Hsp104 can promote yeast prion generation. Genetics 188, 339348. Kulkarni, A.A., Abul-Hamd, A.T., Rai, R., El Berry, H., and Cooper, T.G. (2001). Gln3p nuclear localization and interaction with Ure2p in Saccharomyces cerevisiae. J. Biol. Chem. 276, 3213632144. , B., Slaninova , I., Jacq, C., and Palkova , Z. Kuthan, M., Devaux, F., Janderova (2003). Domestication of wild Saccharomyces cerevisiae is accompanied by changes in gene expression and colony morphology. Mol. Microbiol. 47, 745754. Lai, L.C., Kosorukoff, A.L., Burke, P.V., and Kwast, K.E. (2006). Metabolicstate-dependent remodeling of the transcriptome in response to anoxia and subsequent reoxygenation in Saccharomyces cerevisiae. Eukaryot. Cell 5, 14681489. Lancaster, A.K., Bardill, J.P., True, H.L., and Masel, J. (2010). The spontaneous appearance rate of the yeast prion [PSI+] and its implications for the
evolution of the evolvability properties of the [PSI+] system. Genetics 184, 393400. Liebman, S.W., and Chernoff, Y.O. (2012). Prions in yeast. Genetics 191, 1041 1072. Liti, G., Carter, D.M., Moses, A.M., Warringer, J., Parts, L., James, S.A., Davey, R.P., Roberts, I.N., Burt, A., Koufopanou, V., et al. (2009). Population genomics of domestic and wild yeasts. Nature 458, 337341. Liu, H., Styles, C.A., and Fink, G.R. (1996). Saccharomyces cerevisiae S288C has a mutation in FLO8, a gene required for lamentous growth. Genetics 144, 967978. MacLean, R.C. (2008). The tragedy of the commons in microbial populations: insights from theoretical, comparative and experimental studies. Heredity (Edinb) 100, 471477. McGlinchey, R.P., Kryndushkin, D., and Wickner, R.B. (2011). Suicidal [PSI+] is a lethal yeast prion. Proc. Natl. Acad. Sci. USA 108, 53375341. Mitchell, A., Romano, G.H., Groisman, B., Yona, A., Dekel, E., Kupiec, M., Dahan, O., and Pilpel, Y. (2009). Adaptive prediction of environmental changes by microorganisms. Nature 460, 220224. Octavio, L.M., Gedeon, K., and Maheshri, N. (2009). Epigenetic and conventional regulation is distributed among activators of FLO11 allowing tuning of population-level heterogeneity in its expression. PLoS Genet. 5, e1000673. Ough, C.S., and Stashak, R.M. (1974). Further studies on proline concentration in grapes and wines. Am. J. Enol. Vitic. 25, 712. Ozcan, S., and Johnston, M. (1999). Function and regulation of yeast hexose transporters. Microbiol. Mol. Biol. Rev. 63, 554569. Patel, B.K., Gavin-Smyth, J., and Liebman, S.W. (2009). The yeast global transcriptional co-repressor protein Cyc8 can propagate as a prion. Nat. Cell Biol. 11, 344349. Pfeiffer, T., Schuster, S., and Bonhoeffer, S. (2001). Cooperation and competition in the evolution of ATP-producing pathways. Science 292, 504507. Pigliucci, M. (2008). Is evolvability evolvable? Nat. Rev. Genet. 9, 7582. Ratcliff, W.C., Denison, R.F., Borrello, M., and Travisano, M. (2012). Experimental evolution of multicellularity. Proc. Natl. Acad. Sci. USA 109, 1595 1600. Reynolds, T.B., and Fink, G.R. (2001). Bakers yeast, a model for fungal biolm formation. Science 291, 878881. Rogoza, T., Goginashvili, A., Rodionova, S., Ivanov, M., Viktorovskaya, O., Rubel, A., Volkov, K., and Mironova, L. (2010). Non-Mendelian determinant [ISP+] in yeast is a nuclear-residing prion form of the global transcriptional regulator Sfp1. Proc. Natl. Acad. Sci. USA 107, 1057310577. Rupp, S., Summers, E., Lo, H.J., Madhani, H., and Fink, G. (1999). MAP kinase and cAMP lamentation signaling pathways converge on the unusually large promoter of the yeast FLO11 gene. EMBO J. 18, 12571269. Sanchez, Y., Taulien, J., Borkovich, K.A., and Lindquist, S. (1992). Hsp104 is required for tolerance to many forms of stress. EMBO J. 11, 23572364. Satpute-Krishnan, P., and Serio, T.R. (2005). Prion protein remodelling confers an immediate phenotypic switch. Nature 437, 262265. Serio, T.R., Cashikar, A.G., Kowal, A.S., Sawicki, G.J., Moslehi, J.J., Serpell, L., Arnsdorf, M.F., and Lindquist, S.L. (2000). Nucleated conformational conversion and the replication of conformational information by a prion determinant. Science 289, 13171321. Sertil, O., Kapoor, R., Cohen, B.D., Abramova, N., and Lowry, C.V. (2003). Synergistic repression of anaerobic genes by Mot3 and Rox1 in Saccharomyces cerevisiae. Nucleic Acids Res. 31, 58315837. Shorter, J., and Lindquist, S. (2006). Destruction or potentiation of different prions catalyzed by similar Hsp104 remodeling activities. Mol. Cell 23, 425438. Storici, F., and Resnick, M.A. (2006). The delitto perfetto approach to in vivo site-directed mutagenesis and chromosome rearrangements with synthetic oligonucleotides in yeast. Methods Enzymol. 409, 329345.
Suzuki, G., Shimazu, N., and Tanaka, M. (2012). A yeast prion, Mod5, promotes acquired drug resistance and cell survival under environmental stress. Science 336, 355359. Teixeira, M.C., Monteiro, P., Jain, P., Tenreiro, S., Fernandes, A.R., Mira, N.P., -Correia, I. (2006). The Alenquer, M., Freitas, A.T., Oliveira, A.L., and Sa YEASTRACT database: a tool for the analysis of transcription regulatory associations in Saccharomyces cerevisiae. Nucleic Acids Res. 34(Database issue), D446D451. Toombs, J.A., Petri, M., Paul, K.R., Kan, G.Y., Ben-Hur, A., and Ross, E.D. (2012). De novo design of synthetic prion domains. Proc. Natl. Acad. Sci. USA 109, 65196524. True, H.L., and Lindquist, S.L. (2000). A yeast prion provides a mechanism for genetic variation and phenotypic diversity. Nature 407, 477483. rkel, S. (1999). Hyperosmotic stress represses the transcription of HXT2 Tu and HXT4 genes in Saccharomyces cerevisiae. Folia Microbiol. (Praha) 44, 372376. Tyedmers, J., Madariaga, M.L., and Lindquist, S. (2008). Prion switching in response to environmental stress. PLoS Biol. 6, e294. chova , L., Stov cek, O., Chernyavskiy, O., Stpa nek, L., Kub cek, V., Hlava Va , L., and Palkova , Z. (2011). Flo11p, drug efux pumps, and the extracelnova lular matrix cooperate to form biolm yeast colonies. J. Cell Biol. 194, 679687.
Verstrepen, K.J., and Klis, F.M. (2006). Flocculation, adhesion and biolm formation yeasts. Mol. Microbiol. 60, 515. Verstrepen, K.J., Derdelinckx, G., Verachtert, H., and Delvaux, F.R. (2003). Yeast occulation: what brewers should know. Appl. Microbiol. Biotechnol. 61, 197205. Verstrepen, K.J., Jansen, A., Lewitter, F., and Fink, G.R. (2005). Intragenic tandem repeats generate functional variability. Nat. Genet. 37, 986990. Wickner, R.B. (1994). [URE3] as an altered URE2 protein: evidence for a prion analog in Saccharomyces cerevisiae. Science 264, 566569. Wickner, R.B., Edskes, H.K., Bateman, D., Kelly, A.C., and Gorkovskiy, A. (2011). The yeast prions [PSI+] and [URE3] are molecular degenerative diseases. Prion 5, 258262. Wong, K.H., and Struhl, K. (2011). The Cyc8-Tup1 complex inhibits transcription primarily by masking the activation domain of the recruiting protein. Genes Dev. 25, 25252539. Wright, R.M., Repine, T., and Repine, J.E. (1993). Reversible pseudohyphal growth in haploid Saccharomyces cerevisiae is an aerobic process. Curr. Genet. 23, 388391. Zupan, J., and Raspor, P. (2010). Invasive growth of Saccharomyces cerevisiae depends on environmental triggers: a quantitative model. Yeast 27, 217228.
An RNA Degradation Machine Sculpted by Ro Autoantigen and Noncoding RNA

Xinguo Chen,1 David W. Taylor,2 Casey C. Fowler,3 Jorge E. Galan,1,3 Hong-Wei Wang,2,4 and Sandra L. Wolin1,2,*
of Cell Biology of Molecular Biophysics and Biochemistry 3Department of Microbial Pathogenesis Yale School of Medicine, New Haven, CT 06510, USA 4Present address: Tsinghua-Peking Center for Life Sciences, School of Life Sciences, Tsinghua University, Beijing 100084, China *Correspondence: sandra.wolin@yale.edu http://dx.doi.org/10.1016/j.cell.2013.02.037
SUMMARY
Many bacteria contain an ortholog of the Ro autoantigen, a ring-shaped protein that binds noncoding RNAs (ncRNAs) called Y RNAs. In the only studied bacterium, Deinococcus radiodurans, the Ro ortholog Rsr functions in heat-stress-induced ribosomal RNA (rRNA) maturation and starvation-induced rRNA decay. However, the mechanism by which this conserved protein and its associated ncRNAs act has been obscure. We report that Rsr and the exoribonuclease polynucleotide phosphorylase (PNPase) form an RNA degradation machine that is scaffolded by Y RNA. Single-particle electron microscopy, followed by docking of atomic models into the reconstruction, suggests that Rsr channels single-stranded RNA into the PNPase cavity. Biochemical assays reveal that Rsr and Y RNA adapt PNPase for effective degradation of structured RNAs. A Ro ortholog and ncRNA also associate with PNPase in Salmonella Typhimurium. Our studies identify another ribonucleoprotein machine and demonstrate that ncRNA, by tethering a protein cofactor, can alter the substrate specicity of an enzyme.
INTRODUCTION Noncoding RNAs (ncRNAs) are involved in an enormous variety of cellular processes. Many ncRNAs assemble with proteins and function as ribonucleoprotein (RNP) complexes. These RNPs include small nuclear RNPs that function in pre-mRNA splicing, small nucleolar RNPs that modify pre-rRNAs, the telomerase RNP that maintains chromosome ends, and the microRNA/Argonaute complexes that modulate mRNA translation and stability (Hannon et al., 2006). For each of these RNPs, the ncRNA moiety base pairs with nucleic acid targets to direct enzymatic activity to specic RNA and DNA sequences. For other RNPs, the ncRNA inuences the function of bound proteins. For example, in the signal recognition particle (SRP), which mediates targeting of
secretory and membrane proteins to plasma membranes, the ncRNA increases the interaction of SRP with its receptor and stimulates the GTPase activities of the SRP-receptor complex (Ataide et al., 2011). In contrast to the many well-characterized RNPs, the function of Ro RNPs has been mysterious since their discovery (Lerner et al., 1981). The major protein component, the ring-shaped Ro 60 kDa autoantigen, is a clinically important target of the immune response in patients with the rheumatic diseases grens syndrome (Sim systemic lupus erythematosus and Sjo and Wolin, 2011). Ro orthologs are present in most metazoans and approximately 5% of sequenced bacterial genomes (Perreault et al., 2007; Sim and Wolin, 2011). Bacteria containing likely Ro orthologs are present in most phyla and include the radiation-resistant Deinococcus radiodurans and the human pathogens Bacteroides thetaiotaomicron and Salmonella enterica serovar Typhimurium (S. Typhimurium) (Sim and Wolin, 2011). In all studied organisms, Ro binds ncRNAs called Y RNAs. These ncRNAs, which are 100 nt in length, fold into secondary structures consisting of one or more large internal loops and a long stem that contains the Ro- binding site (Teunissen et al., 2000; Sim and Wolin, 2011). Many species contain between two and four distinct Y RNAs that differ largely in the sequences and sizes of the internal loops (Sim and Wolin, 2011). Ro binding stabilizes Y RNAs from degradation et al., 1999). Conversely, Y RNAs block a nuclear accu(Labbe mulation signal on Ro, retaining Ro in the cytoplasm (Sim et al., 2009). Because the Ro protein also binds misfolded ncRNAs in some animal cell nuclei, it is proposed to function in ncRNA quality control (OBrien and Wolin, 1994; Shi et al., 1996; Chen et al., 2003; Hogg and Collins, 2007). Structural and biochemical studies revealed that Ro binds misfolded RNAs that contain both a 30 single-stranded end and adjacent protein-free helices (Fuchs et al., 2006). The 30 ends of these RNAs insert through the Ro ring, while helices contact the Ro outer surface. Because the binding of Ro to misfolded ncRNAs is largely sequence nonspecic, Ro may scavenge aberrant RNAs that fail to assemble with their correct RNA-binding proteins (Fuchs et al., 2006). However, both the mechanism by which Ro affects misfolded RNA metabolism and whether Y RNAs contribute to this function are unknown.
Figure 1. Rsr, Y RNA, and PNPase Form an RNP

(A) After purication from D. radiodurans, the complex was fractionated in glycerol gradients and proteins detected by silver staining (top). Protein identities were conrmed by immunoblotting. By subjecting RNA from the fractions to northern analysis, we detected a 30 nt Y RNA fragment (bottom). Positions of Rsr, PNPase, and thyroglobulin run in parallel gradients are shown. Asterisks denote fractions containing the complex. Extra bands in lanes 2 and 18 are keratins. (B) After E. coli expression, the puried complex was fractionated in glycerol gradients. Proteins were visualized with Coomassie blue (top). Extracted RNA was visualized with ethidium bromide (bottom). Asterisks, complex. In both (A) and (B), samples were analyzed in multiple gels that were joined at the lines.
RESULTS An Rsr/Y RNA/PNPase RNP To identify components of the Rsr/ PNPase complex, we subjected a D. radiodurans strain in which Rsr and PNPase each carried a distinct epitope tag to two rounds of afnity purication. Rsr was fused to the IgG-binding domains of Staphylococcus aureus Protein A, while PNPase was fused to three copies of FLAG. After extracts were applied to IgG Sepharose, Rsr-containing complexes were eluted with TEV protease and applied to anti-FLAG agarose to isolate the Rsr/ PNPase complex. Fractionation of the nal eluate on glycerol gradients, followed by silver staining, revealed that the complex migrated at 300450 kD and that the major components were Rsr and PNPase (Figure 1A). Northern analyses revealed that a Y RNA degradation fragment comigrated with the two proteins (Figure 1A). Because the apparent molecular size suggested that the complex consisted of Rsr (57 kD), one PNPase trimer (262 kD), and a Y RNA (42 kD), we attempted a reconstitution by coexpressing these components in E. coli, a bacterium that lacks a Ro ortholog. Epitope-tagged forms of Rsr and PNPase were expressed in E. coli in the presence or absence of Y RNA. Complex formation was only detected when Y RNA was coexpressed. After purication, the complex formed in E. coli migrated identically in glycerol gradients to the complex puried from D. radiodurans (Figure 1B). Thus, Rsr and PNPase form a stable complex that also contains Y RNA. Y RNA Tethers Rsr to PNPase Since Rsr and PNPase only associated when Y RNA was present, we hypothesized that Y RNA scaffolds the complex. To test this idea, we established an assay to evaluate how Rsr and PNPase interact via Y RNA. Rsr and PNPase were mixed with 32P-labeled Y RNA and the RNPs were separated from unbound RNA in nondenaturing gels. Rsr and PNPase each formed discrete RNPs with Y RNA (Figure 2A). When both Rsr
To have a genetically tractable system, we characterized Ro and a Y RNA in the rst sequenced bacterium with a Ro ortholog, D. radiodurans (Chen et al., 2000). These studies revealed that the ortholog ro sixty-related (Rsr) functions with 30 to 50 exoribonucleases during some types of environmental stress. Rsr and two exoribonucleases, RNase II and RNase PH, are required for efcient 23S rRNA maturation during heat stress (Chen et al., 2007). In stationary phase, Rsr and the ringshaped exoribonuclease polynucleotide phosphorylase (PNPase) are important for rRNA degradation (Wurtmann and Wolin, 2010). Rsr and PNPase are found together in immunoprecipitates, and the sedimentation of PNPase with ribosomal subunits in stationary phase requires Rsr (Chen et al., 2007; Wurtmann and Wolin, 2010). However, the components of the putative Rsr/PNPase complex, the way in which Rsr inuences PNPase activity, and whether Y RNA is involved have not been addressed. To understand how a Ro protein can inuence the function of an exoribonuclease, we puried the Rsr/PNPase complex from D. radiodurans and examined its composition, molecular architecture, and activity. We report that Y RNA tethers Rsr to PNPase to form an RNA degradation machine. Single-particle electron microscopy (EM), followed by docking Ro and PNPase atomic structures into the three-dimensional reconstruction, is consistent with a model in which single-stranded RNA threads from the Rsr ring into the PNPase cavity. Biochemical experiments demonstrate that Rsr and Y RNA specialize PNPase for effective degradation of structured RNAs. Notably, a Ro ortholog and a ncRNA also associate with PNPase in an evolutionarily distant bacterium, S. Typhimurium. We discuss the similarities between the Rsr/Y RNA/PNPase RNP and other RNA degradation machines.
Figure 2. Y RNA Is a Scaffold

(A) 32P-labeled Y RNA was mixed with no protein (lane 1), Rsr (lane 2), PNPase (lane 3), or Rsr and PNPase (lane 4) and RNPs were separated in native gels. RsrH189S was assayed in lanes 5 and 6. (B) Proposed secondary structures for wild-type and mutant Y RNAs. A fraction of the RNA forms a conformer in which the nucleotides in the conserved helix (box) are base paired (Green et al., 1998). In the helix swap mutant, the base pairs are reversed. To reduce alternative structures, we converted the Us at positions 24 and 25 to Cs (asterisks). These changes do not affect Rsr binding. In the truncated RNA, nts are numbered according to positions in the full-length RNA. (C and D) Wild-type and mutant Y RNAs were incubated with Rsr and PNPase as indicated. (E) PNPase and mutants lacking the KH or S1 domain were incubated with Y RNA in the absence or presence of Rsr. See also Figure S1.
and PNPase were present, a new RNP appeared that migrated more slowly than the two binary complexes (Figure 2A). As expected if Y RNA tethers Rsr to PNPase to form a ternary complex, this slowest migrating RNP failed to form in the presence of a mutant Rsr (Rsr-H189S) that does not bind Y RNA (Chen et al., 2007) (Figure 2A). By excising the bands, fractionating the proteins in SDS-PAGE, and performing western blotting, we conrmed that the slowest migrating RNP contained both Rsr and PNPase (Figures S1A and S1B, available online). If Y RNA is a scaffold, then Rsr and PNPase should bind separate sites on the RNA. In vertebrates, Ro makes specic contacts with the 50 strand of a conserved helix in a stem formed by base pairing the 50 and 30 Y RNA ends (Stein et al., 2005) (Figure 2B). As expected if Rsr interacts similarly, a mutant Y RNA in which the conserved base pairs were reversed (Figure 2B, helix
swap) did not bind Rsr or form the Rsr/Y RNA/PNPase RNP, although PNPase binding to the mutant RNA was unaffected (Figure 2C). A second feature of Y RNAs is the presence of one or more large internal loops (Teunissen et al., 2000; Sim and Wolin, 2011) (Figure 2B). A Y RNA lacking these loops exhibited reduced PNPase binding and did not form the Rsr/Y RNA/ PNPase RNP, although Rsr bound the truncated RNA indistinguishably from the full-length RNA (Figure 2D). The Y RNA loops probably interact with the K-homology (KH) and S1 domains of PNPase, since PNPases lacking either of these single-stranded RNA-binding domains (Lunde et al., 2007) failed to bind Y RNA (Figure 2E), although the mutant proteins formed trimers similar to full-length PNPase as judged by fractionation in native gels and electron microscopy (Figure S1).
Figure 3. Architecture of the Rsr/Y RNA/ PNPase RNP

(A) Representative reference-free two-dimensional class averages of native PNPase (top row) compared with those of the GraFix-prepared complex (middle row). Class averages of PNPase and Rsr/Y RNA/PNPase in the same column are aligned in or to the same orientation. The difference maps between each pair of images are shown in the bottom row. Scale bar represents 10 nm. (B) Rsr/Y RNA/PNPase single-particle reconstruction from negatively stained complexes. (CF) The atomic models of S. antibioticus PNPase (PDB 1E3P) and X. laevis Ro complexed with Y RNA (PDB 1YVP) and misfolded RNA (PDB 2I91) fragments were docked into the EM density to obtain a model of the RNP. PNPase, blue; Ro, purple; Y RNA, red; duplex portion of substrate, yellow; single-stranded portion of substrate, light green. The PNPase KH/S1 domains, which were only partly visualized in S. antibioticus PNPase, are boxed in (D). The Y RNA fragment is oriented such that in the full-length RNA, the 50 and 30 ends point toward PNPase (C and D), and the part of the Y RNA that is absent from the crystal structure is depicted as a dashed red line (C-E). Front and back sections of the reconstruction were removed from the view to show the single-stranded end of the substrate threading through Rsr toward the PNPase cavity in (F). Scale bar denotes 5 nm. See also Figure S2.
Since mutations that reduce binding of either Rsr or PNPase to Y RNA impair complex formation, we conclude that Y RNA scaffolds the complex. Molecular Architecture of the Rsr/Y RNA/PNPase RNP We examined the structure of the Rsr/Y RNA/PNPase RNP by negative-stain EM. PNPase appeared as a homogeneous mono diamdisperse globular particle with a central cavity and 100 A eter in the longest direction in raw micrographs (Figure S1F). Two-dimensional class averages showed a large ring-shaped particle containing three major domains of density with 3-fold symmetry (Figure 3A). Since the Rsr/Y RNA/PNPase RNP dissociated under the conditions used for negative-stain EM, with particles corresponding to PNPase and smaller ring-like particles in the size and shape of Rsr (Ramesh et al., 2007) easily recognizable in raw micrographs, we used the GraFix method (Kastner et al., 2008) to improve recovery. The use of GraFix, which involves a nal purication step of sedimenting the
complex through a double gradient of glutaraldehyde and glycerol, improved the homogeneity of the RNP dramatically. Large globular particles similar to PNPase but containing a marked additional density and a total longest diameter of appeared (Figure S2A). The 140150 A majority of the two-dimensional class averages of the Rsr/Y RNA/PNPase RNP appeared as a small ring connected to a larger ring (Figures 3A and S2B). No differences were observed between native PNPase and PNPase in the GraFixprepared RNP (Figure 3A), indicating that the GraFix method does not affect PNPase structure. Using single-particle reconstruction methods, we obtained a three-dimensional reconstruction of the Rsr/Y RNA/PNPase resolution based on the 0.5 Fourier shell correlaRNP at 25 A tion (FSC) criterion from 9,000 particle images (Figures 3B and S2). In agreement with the reference-free two-dimensional class averages, the reconstruction showed a small elongated ring-shaped density positioned at an angle almost directly above the central channel of the larger ring-shaped structure. The smaller and larger densities have the sizes and shapes of Rsr (Ramesh et al., 2007) and PNPase (Symmons et al., 2000), respectively, and reference-free two-dimensional class averages of Rsr/Y RNA/PNPase complexes labeled with an antibody against the His-tag of Rsr conrm the identity of the smaller ring as Rsr (Figures S2D and S2E). There is an additional rod-shaped
density that forms the major connection between the two rings along one side, which may represent Y RNA. By docking atomic models of Rsr and PNPase into the EM map, we obtained a model for the complex (Figure 3C). The smaller ring accommodates the crystal structure of Rsr (Ramesh et al., 2007) or X. laevis Ro (Stein et al., 2005). The Rsr structure (PDB 2NVO) more closely resembles the size and shape of the smaller ring-shaped density and could be docked into the map with the highest cross-correlation coefcient (0.939). Because a X. laevis Ro structure included a Y RNA fragment (Stein et al., 2005), we used this structure (PDB 1YVP) for nal modeling by superimposing its coordinates onto the Rsr structure already deposited into the EM density. This orientation places one end of the Y RNA fragment, corresponding to the 50 and 30 termini of the intact RNA, in close proximity to one of the KH and S1 domains of the PNPase trimer (Figure 3C). The Y RNA fragment used for crystallization (Stein et al., 2005) contains only the part of the stem encompassing the conserved helix, and we propose that the remainder of the stem and the distal loops continue along the Ro surface and across the rod-shaped density to allow the loops to contact a second KH/S1 domain of PNPase, thus anchoring the two protein components (Figures 3D and 3E). To examine how the Rsr/Y RNA/PNPase RNP could interact with substrates, we superimposed the coordinates of X. laevis Ro bound to a misfolded RNA fragment (PDB 2I91; Fuchs et al., 2006) onto the structure of Ro deposited into the EM density. In this structure, the single-stranded 30 end of the misfolded RNA passes through the Ro cavity, while a duplex binds on the outer surface (Fuchs et al., 2006). Docking of this substrate (Figures 3D3F) places the duplex on the Rsr outer face (colored yellow) with the single-stranded end (colored green) passing through the Rsr cavity directly above the PNPase channel. The Streptomyces antibioticus PNPase structure (PDB 1E3P; Symmons et al., 2000) can t into the larger ring of the reconstruction in two opposite orientations with similar crosscorrelation coefcients (0.835 and 0.841); however, the docking result that places the KH and S1 domains facing Rsr and the protruding single-stranded RNA is most consistent with the biochemical data (Figures 2 and S1). In this orientation, RNA could thread from Rsr into the PNPase cavity (Figure 3F). Rsr and Y RNA Specialize PNPase for Degrading Structured RNAs Our model predicts that Rsr and Y RNA should inuence the activity of PNPase in the RNP complex. By testing a substrate containing a 7GC base pair stemloop and a 35 nt 30 extension (SL7-N35 RNA; Figure S3A), we determined that both PNPase and the puried RNP exhibit processive phosphate-dependent degradation (Figure 4A). As reported for E. coli PNPase (Spickler and Mackie, 2000), both PNPase and the RNP stalled 78 nt 30 to the stemloop (Figures 4A, bracket; and S3B). Importantly, the RNP degraded the substrate more effectively than PNPase alone, as measured by both increased degradation of the input RNA and the accumulation of limit oligonucleotides (Figures 4A, 4D, and S3B). Consistent with a role in unwinding RNA, the RNP complex also stalled slightly closer to the stemloop than PNPase (Figures 4A and S3B, arrows). Enhanced RNA degradation was not detected when puried Rsr, Y RNA, and PNPase
were added to the reaction under conditions that result in inefcient complex assembly, indicating that RNP formation is required (Figure S3B). We next examined the activity of the Rsr/Y RNA/PNPase RNP on substrates that resemble most cellular RNAs in consisting of short helices separated by bulges and loops. One substrate consisted of a misfolded pre-5S rRNA that is bound by X. laevis Ro (Fuchs et al., 2006) (Figure S3A). Since Rsr and PNPase function in rRNA degradation (Wurtmann and Wolin, 2010), we also tested a 160 nt substrate resembling the D. radiodurans 16S rRNA 30 end (Figure S3A). The RNP was more active than PNPase in degrading both these RNAs (Figures 4B, 4C, 4E, and 4F). To identify the specic RNA features that result in more effective degradation by the RNP, we assayed a series of duplex-containing substrates similar to those used to characterize other exoribonucleases (Cheng and Deutscher, 2005; Lorentzen et al., 2008). In addition to a GC-rich 17 bp duplex, the substrates contained poly(U) 30 extensions of varying length (Figure S4A). Both PNPase and the Rsr/Y RNA/PNPase RNP degraded substrates with 30 overhangs of at least 20 nt, with more efcient decay occurring when the overhang contained 30 nt (Figure S4B). As reported for E. coli PNPase (Cheng and Deutscher, 2005), neither D. radiodurans PNPase nor the RNP could degrade the 17 bp duplex; however, both digested through a 13 bp duplex (Figure S4E). Importantly, when we tested substrates containing shorter duplexes, the RNP degraded the RNAs more effectively than PNPase alone. The difference was evident with a substrate containing an 11 bp duplex and a 30 nt overhang, as measured by increased degradation of the input RNA and decreased stalling 30 to the duplex (Figures 5A, 5C, and 5E). Although the enhancement of PNPase activity was less than that observed on other structured RNAs (Figure 4), this may be because disruption of the rst few bps of the short duplex by either PNPase or the RNP results in dissociation of the remaining helix. Notably, the RNP was less active than PNPase in degrading a singlestranded RNA corresponding to the overhang strand of the duplex (Figures 5B, 5D, and 5F), most likely because the S1 and KH domains that contribute to single-stranded RNA binding by PNPase are less accessible in the RNP (Figure 3D). Similar effects were seen with duplex and single-stranded substrates containing poly(A) tails (data not shown). Together with the increased activity of the Rsr/Y RNA/PNPase RNP that we observed on stemloop-containing substrates (Figure 4), we conclude that Rsr and Y RNA increase the activity of PNPase in degrading structured RNAs. In addition to its role as an exoribonuclease, PNPase adds A-rich tails to RNA using nucleoside diphosphates as precursors (Mohanty and Kushner, 2011). No differences were detected between the RNP and PNPase when assayed for polyadenylation activity using a structured RNA that is a substrate for other PNPases (Sohlberg et al., 2003) (Figures S3CS3E). To determine whether the RNP contributes to tail synthesis in vivo, we compared the A-rich tails in RNA isolated from wild-type D. radiodurans and strains lacking PNPase, Rsr, or the Y RNA. This revealed that PNPase, but not Rsr or the Y RNA, is required for tail formation (Figure S3F). As we do not detect effects of Rsr and the Y RNA on A-tail addition,
Figure 4. Enhanced Degradation of Structured RNAs by the RNP

(AC) The activity of PNPase and the RNP was compared on 50 -labeled SL7-N35 RNA (A), misfolded X. laevis pre-5S rRNA (B), and the D. radiodurans 16S rRNA 30 end substrate (C). At intervals, aliquots were removed and RNAs fractionated in denaturing gels (top panels). Brackets denote sites of enzyme stalling. Compared to PNPase, the RNP stalls closer to the stemloop of the SL7-N35 RNA (A, arrow). The gel in (B) was run further than those in (A) and (C). Aliquots were subjected to immunoblotting to conrm that equal amounts of PNPase were present (bottom panels). (DF) Data from the degradation assays were plotted. Data are represented as mean values from three replicates SEM. See also Figure S3.
Figure 5. The RNP Degrades a Duplex More Effectively than PNPase but is Less Active on Single-Stranded RNA
(A and B) Sequences of ds11-U30 and ss17-U30 substrates. The ss17-U30 RNA corresponds to the overhang strand of the ds11-U30 duplex. In both substrates the 50 end of the ss17-U30 RNA was labeled with [g-32P]ATP. (C and D) The activity of PNPase and the RNP was compared on the ds11-U30 (C) and ss17-U30 (D) substrates. Aliquots were removed at intervals and fractionated in denaturing gels to visualize the labeled RNA (top panels). Aliquots were also assayed by immunoblotting to conrm that equal amounts of PNPase were present (bottom panels). Notably, although more of the RNP may have been used in the ss17-U30 assay (D, bottom panel), the RNP is less active than PNPase on this substrate (top panel). (E and F) Degradation reactions using the ds11-U30 and ss17-U30 substrates were performed in triplicate. Data are represented as mean values from three replicates SEM. In (F), error bars for PNPase are too small to be visible. See also Figure S4.
we conclude that Rsr and Y RNA sculpt PNPase for effective degradation of structured RNAs. A Ro Ortholog and a ncRNA Associate with PNPase in S. Typhimurium To determine whether the role we identied for Rsr in D. radiodurans could be conserved in an evolutionarily distant species, we examined S. Typhimurium, an enteric bacterium closely related to E. coli. S. Typhimurium Rsr is encoded within the rtcBA operon, whose E. coli orthologs encode the RNA ligase (rtcB) and RNA cyclase (rtcA) components of a s54-regulated RNA repair operon (Figure 6A) (Genschik et al., 1998; Tanaka
and Shuman, 2011). In E. coli, this locus is tightly regulated by the adjacent rtcR transcriptional activator, as rtcBA transcription is not detected during standard growth but becomes detectable when an N-terminally truncated, constitutively active form of RtcR is overexpressed (Genschik et al., 1998). Consistent with similar regulation, we could only detect S. Typhimurium Rsr when a parallel version of the truncated RtcR was expressed from a plasmid (Figure 6B). Immunoprecipitation with an antibody against S. Typhimurium Rsr, followed by end-labeling of associated RNAs, revealed two ncRNAs encoded 30 to rsr (Figures 6A, 6C, and 6D). Both ncRNAs are only detected when the truncated RtcR is
D E
Figure 6. Rsr and ncRNA Associate with PNPase in S. Typhimurium

(A) Maps of the S. Typhimurium rsr locus and the corresponding E. coli region. (B) S. Typhimurium HA-rsr cells carrying empty vector or the truncated RtcR under control of the arabinose-inducible promoter (pRtcRDN) were grown in the presence or absence of arabinose. Lysates were subjected to immunoblotting with anti-HA (top) and anti-PNPase antibodies (bottom panel). Expression in the absence of arabinose (lane 3) is due to promoter leakiness. (C) Lysates from cells carrying the empty vector or pRtcRDN were subjected to immunoprecipitation with anti-S. Typhimurium Rsr antibodies. Total RNA (lanes 1 and 2) and RNAs within immunoprecipitates (lanes 3 and 4) were visualized by end labeling. (D) Possible secondary structures of YrlA and YrlB. (E) RNA from cells carrying the indicated plasmids was subjected to northern blotting to detect YrlA and YrlB. (F) FLAG3-pnp HA-rsr lysates were subjected to afnity purication using anti-FLAG. After fractionating eluates in glycerol gradients, HA-Rsr and PNPase were detected by immunoblotting. YrlA was detected by northern blotting. Samples were analyzed in multiple gels that were joined at the lines. Positions of ovalbumin (44 kD), bovine g-globulin (158 kD), and thyroglobulin (670 kD) run in parallel gradients are shown. See also Figure S5.
overexpressed (Figures 6C and 6E). Thus, similar to D. radiodurans Rsr and Y RNA, which are upregulated following several forms of environmental stress (Sim and Wolin, 2011),
transcription of S. Typhimurium Rsr and its associated ncRNAs may be regulated in response to an external stimulus. We call the S. Typhimurium ncRNAs YrlA (Y RNA-like) and YrlB.
PNPase, as both components were greatly reduced when the afnity purication was performed from an untagged strain (Figure S5). Thus, while the larger size of the S. Typhimurium Rsr/ YrlA/PNPase complex suggests that it may contain additional components or differ in stoichiometry from the minimal D. radiodurans RNP, the association of PNPase with Rsr and a ncRNA is conserved. DISCUSSION The function of the Ro protein and its associated ncRNAs have been enigmatic for over 30 years. By studying a bacterial Ro protein, we identied a ribonucleoprotein machine in which the ortholog Rsr is tethered via Y RNA to the ring-shaped exoribonuclease PNPase. Single-particle EM of the RNP revealed a double-ring architecture, suggesting that Rsr channels RNA into the PNPase cavity for degradation. Biochemical studies demonstrated that Rsr and Y RNA specialize PNPase for degrading structured RNAs. This work has identied another role for ncRNA and revealed a bacterial RNA degradation machine that resembles the archaeal and eukaryotic exosomes. A ncRNA Functions as a Tether We showed that Y RNA tethers D. radiodurans Rsr to PNPase to form an RNP machine. Since proteins with RNA-binding domains that could interact with Y RNA loops are widespread, Y RNAs could potentially tether Ro orthologs to additional RNA remodeling proteins, such as helicases and RNA chaperones. In this case, Y RNAs and their associated Ro rings would function as trans-acting modules that can be attached to multiple distinct proteins to augment their handling of structured RNAs. Conversely, Y RNAs could tether proteins that affect the activity or subcellular location of Ro. The use of loops to tether interacting proteins ensures that Y RNAs will not be substrates for enzymes that require an RNA end for activity, such as exoribonucleases and some helicases. The role of Y RNAs as tethers is supported by reports of proteins that interact with mammalian Ro proteins through binding Y RNAs (Bouffard et al., 2000; Fabini et al., 2001; Fouraux et al., 2002; Hogg and Collins, 2007; Sim et al., 2012). Although in most cases the signicance is unknown, binding of the zipcodebinding protein ZBP1 to a mouse Y RNA allows nuclear export of the RNP (Sim et al., 2012). Because the binding sites of Y RNAs and misfolded ncRNAs overlap on the Ro surface (Fuchs et al., 2006), it has been unclear how Y RNAs could be repositioned to allow substrates to access the Ro cavity. We propose that binding of proteins such as PNPase to Y RNA loops removes this portion of Y RNA from the Rsr ring, rendering the cavity accessible (Figure 7A). An appealing feature of this model is that in addition to its role as a scaffold, Y RNA could function as a gate to ensure that RNA substrates enter the Rsr cavity only when PNPase or another remodeling enzyme is available to act on them. Although Y RNA-mediated tethering of Rsr is needed for the enhanced activity of PNPase on the tested RNAs, Rsr may be able to assist some enzymes as a Y RNA-free protein. This is suggested by the requirement for Rsr in heat-stress-induced 23S rRNA maturation, which involves RNase II and RNase PH
Figure 7. Model for Rsr/Y RNA/PNPase RNP Function

(A) Formation of the Rsr/Y RNA/PNPase RNP. Residues on the Rsr outer surface contact conserved bases in the Y RNA stem, while the rest of the RNA interacts with basic residues on the Rsr surface, preventing binding of other RNAs (Stein et al., 2005). When PNPase is present, Y RNA loops bind the PNPase KH and S1 domains, repositioning the RNA so that substrates can access the Rsr cavity. (B) The Rsr/Y RNA/PNPase RNP resembles archaeal and eukaryotic exosomes. All three complexes contain rings with six RNase PH domains. PNPase is a trimer in which each monomer contains two RNase PH domains. The archaeal ring contains three copies of each of two proteins containing single RNase PH domains, while in eukaryotes six distinct proteins form the ring. Both exosomes contain RNA-binding caps formed by KH and S1 domain proteins. We propose that, similar to the archaeal exosome, RNA passes through an RNA-binding ring (Rsr) to undergo phosphorolytic degradation in the catalytic ring (PNPase). As yeast and human exosomes have a catalytically inactive RNase PH domain ring, RNA passes through the nine-subunit Exo1-9 ring to reach a hydrolytic exonuclease.
To examine whether S. Typhimurium Rsr associates with PNPase, we fused PNPase to three copies of FLAG and carried out afnity purication using anti-FLAG agarose. Fractionation of the eluate in glycerol gradients, followed by immunoblotting, revealed that PNPase sedimented both at the expected size for the homotrimer (Figure 6F, fractions 8 and 9) and in heavier fractions. Rsr was also present in the eluate. However, while some Rsr and PNPase migrated similarly in the gradient to the D. radiodurans complex (Figure 6F, fractions 11 and 12), Rsr was present predominantly in heaver fractions (fractions 1622). Using northern blotting, we detected YrlA, but not YrlB, in the gradient fractions. Although some YrlA was degraded during sedimentation, making it difcult to assess where it peaked, we detected YrlA in all of the gradient fractions that contained Rsr (Figure 6F). The presence of Rsr and YrlA in the eluate was dependent on
(Chen et al., 2007). In this case, either deleting the Y RNA or increasing the levels of Y RNA-free Rsr results in constitutive 23S rRNA maturation, implying that Rsr may assist 30 end trimming by some exonucleases without a Y RNA tether. If so, the ability of Rsr and other Ro proteins to act both as free proteins and as Ro/Y RNA modules would increase their functional repertoires. The Rsr/Y RNA/PNPase RNP: A Bacterial Exosome? Our characterization of the Rsr/Y RNA/PNPase RNP expands the inventory of RNA degradation machines and reveals another way in which exonucleases such as PNPase can be adapted to degrade structured RNAs. Previously, the only described bacterial machine was the degradosome, which in E. coli contains PNPase, the scaffolding endonuclease RNase E, an RNA helicase, and the metabolic enzyme enolase (Gorna et al., 2012). Although the Rsr/Y RNA/PNPase RNP may function with helicases or endonucleases in vivo, these enzymes are not required to enhance the activity of PNPase in degrading at least some RNAs. Moreover, as some bacteria could have both an Rsr/Y RNA/PNPase RNP and a degradosome, there may be selective advantages associated with each assembly. We note that in many bacteria the degradosome contains different exonucleases, endonucleases, helicases, and metabolic enzymes than those present in the E. coli complex (Gorna et al., 2012). If the Rsr/Y RNA/PNPase RNP diverges similarly in composition (as suggested by the apparent size differences between the D.radiodurans and S. Typhimurium complexes), Rsr and Y RNA could associate with additional nucleases or remodeling proteins in some bacteria. Although the use of ncRNA to scaffold an RNA degradation machine is unique, the D. radiodurans Rsr/Y RNA/PNPase RNP exhibits structural and functional similarities to the multiprotein nuclease complexes known as exosomes. Like the archaeal exosome (Evguenieva-Hackenberg, 2011), the Rsr/Y RNA/ PNPase RNP contains an RNA-binding ring atop an RNase PH domain-containing ring that catalyzes both phosphorolytic degradation of RNA and 30 tail addition (Figure 7B). Similar to the trimeric RNA-binding ring of the archaeal exosome (Evguenieva-Hackenberg et al., 2008), the Rsr ring is important for efcient RNA degradation but is not needed for synthesis of A-rich tails. Also, the conformational exibility present in both the archaeal RNA-binding ring and Rsr should allow both rings to interact with diverse substrates (Ramesh et al., 2007; Evguenieva-Hackenberg, 2011). However, while the exosome RNA-binding ring is coaxial to the degradation ring, Rsr sits at an angle above the PNPase cavity, due to the asymmetric bulk of the ncRNA tether. Notably, while the RNA-binding ring of the archaeal exosome enhances the activity of the RNase PH ring in degrading both single-stranded and structured RNAs (Evguenieva-Hackenberg et al., 2008), the Rsr/Y RNA/PNPase RNP is less active than PNPase on single-stranded RNAs (Figure 5). A likely explanation for the decreased activity of the RNP on single-stranded RNA is that Y RNA-mediated tethering of Rsr to the S1/KH domains of PNPase obstructs one or more of these single-stranded RNAbinding surfaces. In this model, the RNA-binding surface of PNPase would be largely replaced with that of Rsr (Figure 7B).
If the RNA-binding platform of Rsr, which is composed of HEAT repeats, resembles X. laevis Ro in preferring substrates that contain both a single-stranded tail and adjacent helices (Fuchs et al., 2006), the presence of Rsr would specialize PNPase for binding structured RNAs. As described for the yeast exosome, in which RNA threads through the catalytically inactive Exo19 ring to reach the Exo11 RNase (Bonneau et al., 2009) (Figure 7B), Rsr may contribute to ATP-independent unwinding of structured RNAs. Such a role is consistent with our nding that Rsr did not contribute to polyadenylation by PNPase, since addition of nucleotides to an already accessible end may be less dependent on RNA unwinding than exonucleolytic decay. We note that our experiments do not allow us to determine whether Rsr and/or Y RNA dissociate after PNPase engages a substrate or whether they remain on the substrate throughout degradation. Moreover, as crystal structures of Ro and Rsr have revealed considerable exibility in the organization of the HEAT repeat ring and in the size of the central hole (Stein et al., 2005; Ramesh et al., 2007), the Rsr ring could open to allow RNA to enter the channel and/or Rsr to disengage from the substrate. Atomic resolution structures of the Rsr/Y RNA/ PNPase RNP complexed with substrates will be required to determine the actual paths of RNA substrates. Nonetheless, the crystal structures showing single-stranded RNA traversing the X. laevis Ro cavity (Stein et al., 2005; Fuchs et al., 2006), the juxtaposition of the Rsr and PNPase rings observed with EM, and the increased activity of the RNP in degrading structured RNA are all consistent with a mechanism in which RNA passes from the Rsr ring into the PNPase cavity. Finally, the homology between bacterial and vertebrate Ro proteins and Y RNAs, together with the nding that vertebrate Ro proteins bind misfolded ncRNAs (OBrien and Wolin, 1994; Chen et al., 2003), suggests that the RNA remodeling function that we identied for Rsr could be conserved in eukaryotes.
EXPERIMENTAL PROCEDURES Purication of the Rsr/Y RNA/PNPase RNP from D. radiodurans Strain PTR17 expressing Protein A-TEV-Rsr and PNPase-FLAG3 was grown to OD600 = 0.8; resuspended in buffer A (40 mM Tris-HCl [pH 7.5], 100 mM NaCl, 10% glycerol, 0.1% NP-40, 3 mM MgCl2, 3 mM MnCl2, and 1 mM DTT) with 0.5 mM PMSF, 1 mM EGTA, 0.1% diethyl pyrocarbonate, 2 mM vanadyl ribonuclease complex (VRC), and 1 3 protease inhibitor cocktail (Roche Applied Science); and lysed with a French press. After clearing in a Beckman Type 50.2Ti rotor at 20,000 rpm for 30 min, the supernatant was mixed with IgGSepharose (GE Healthcare) for 2 hr at 4 C. After washing with 40 ml buffer A containing 0.5 mM PMSF and 20 ml TEV buffer (10 mM Tris-HCl [pH 7.4], 150 mM NaCl, 0.1% NP-40, 1 mM MgCl2, 1 mM MnCl2, 1 3 protease inhibitor cocktail, and 0.5 mM PMSF), the beads were incubated with 100 units TEV protease for 16 hr at 4 C. The eluate was mixed with anti-FLAG agarose (Sigma-Aldrich) for 2 hr at 4 C, washed with buffer B (40 mM Tris-HCl [pH 7.4], 100 mM NaCl, 0.1% NP-40, 1 mM MgCl2, and 1 mM MnCl2), and eluted with 1 mg/ml 3XFLAG peptide in buffer B containing 0.02% NP-40. This eluate was layered on a 10%40% glycerol gradient in 20 mM Tris-HCl (pH 7.5), 50 mM NaCl, 1 mM MgCl2, 1 mM MnCl2, and 2 mM b-mercaptoethanol and sedimented at 36,000 rpm in a Beckman SW41 rotor for 20 hr at 4 C. Fractions were subjected to SDS-PAGE and silver staining. For detection of Y RNA, extracted RNA was separated in 6% polyacrylamide, 8.3 M urea gels, transferred to Zeta-Probe GT (Bio-Rad) and probed with a [a-32P]rCTPlabeled RNA complementary to the Y RNA.
Electrophoretic Mobility Shift Assays For the visualization of RNP formation, 9.1 pmol of unlabeled Y RNA was mixed with 86 fmol of Y RNA that was transcribed in the presence of [a-32P]rATP, heated to 95 C in binding buffer (20 mM Tris-HCl [pH 7.5], 50 mM NaCl, and 2 mM MnCl2), frozen on dry ice, and thawed on ice. The refolded RNA was incubated with 17.4 pmol Rsr for 30 min at 4 C and 30 min at 30 C. This mixture was incubated with 6.8 pmol PNPase in a nal volume of 5 ml (20 mM Tris-HCl [pH 7.5], 50 mM NaCl, 1.25 mM MnCl2, 2 mM DTT, and 5% glycerol) for 30 min at 4 C and 30 min at 30 C. Next, 1 ml of binding buffer containing 10% glycerol was added and the reaction fractionated in 4% polyacrylamide gels (80:1, acrylamide:bisacrylamide) in 25 mM Tris, 25 mM boric acid, and 1 mM EDTA. Gels were run at 4 C for 1.5 hr at 250 V. Purication of PNPase and the Rsr/Y RNA/PNPase RNP after E. coli Expression Because attempts to separate in vitro-assembled RNPs from unassembled PNPase were unsuccessful, we used coexpression of tagged components in E. coli, followed by purication, to visualize the complex by EM and assay its activity. For PNPase, E. coli expressing D. radiodurans PNPase carrying an N-terminal Strep-tag (Schmidt and Skerra, 2007) were lysed with a French press in 20 mM HEPES (pH 7.5), 200 mM NaCl, 5% glycerol, 2 mM b-mercaptoethanol, 0.5 mM PMSF, 1 mM Pefabloc, and 1 3 protease inhibitor cocktail. Strep-PNPase was puried on Strep-Tactin Superow Plus (QIAGEN), eluted with 5 mM d-desthiobiotin, 20 mM HEPES (pH 7.5), 200 mM NaCl, 2 mM b-mercaptoethanol, and 5% glycerol and concentrated with Amicon Ultra centrifugal lter units to 0.8 ml. Next, 0.3 ml of Strep-PNPase was applied to 11 ml 10%40% glycerol gradients in 20 mM HEPES, (pH 7.5), 50 mM NaCl, 1 mM MgCl2, 1 mM MnCl2, 1 mM Pefabloc, and 2 mM b-mercaptoethanol and sedimented at 36,000 rpm in a Beckman SW41 rotor for 20 hr at 4 C. To express the Rsr/Y RNA/PNPase RNP, we cloned Rsr containing an N-terminal His6-tag and Strep-PNPase into pRSFDuet-1 (Novagen). In addition, Y RNA, followed by a hammerhead ribozyme and a T7 terminator, was cloned into pETDuet-1. To purify the Rsr/Y RNA/PNPase RNP, we grew E. coli carrying both plasmids in LB with ampicillin and kanamycin (100 mg/ml each), induced with IPTG, and lysed the cells by passing through a French press in buffer C (20 mM HEPES [pH 7.5], 50 mM NaCl, 5% glycerol, 1 mM MgCl2, 1 mM MnCl2, 2 mM b-mercaptoethanol, 0.5 mM PMSF, 1.25 mM VRC, 1 mM Pefabloc, and 1 3 protease inhibitor cocktail). After clearing at 20,000 rpm in a Beckman 50.2Ti rotor for 30 min, lysate was incubated with Ni-NTA agarose (QIAGEN) for 2 hr, and washed with buffer C and Rsr-containing complexes eluted with 160 mM imidazole in buffer C. The eluate was incubated with Strep-Tactin resin (QIAGEN) for 2 hr, washed with buffer C lacking VRC, and eluted from the resin with 0.3 ml of 5 mM d-desthiobiotin, 20 mM HEPES (pH 7.5), 50 mM NaCl, 2 mM b-mercaptoethanol, 5% glycerol, 1 mM MgCl2, 1 mM MnCl2, and 0.5% PMSF. This eluate was fractionated in 11 ml 10%40% glycerol gradients in 20 mM HEPES (pH 7.5), 50 mM NaCl, 2 mM b-mercaptoethanol, and 1 mM MgCl2, 1 mM MnCl2, 1 mM Petabloc as described for PNPase. One liter of E. coli yielded 44106 pmols of RNP. Variations in which we used forms of Rsr and PNPase with cleavable tags, followed by tag cleavage, resulted in lower yields, probably due to the extra manipulations needed to remove the tags. Because the RNP dissociated upon freezing, fractions containing the complex were stored at 4 C and used within 3 weeks. To allow exact comparisons, we stored PNPase under identical conditions. For EM, 0.1% glutaraldehyde (EM grade, Sigma) was added to the 40% glycerol solution prior to gradient preparation (Kastner et al., 2008). Negative-Stain EM Sample Preparation and Data Collection PNPase and the Rsr/Y RNA/PNPase RNP were diluted to 2550 nM and immediately applied to glow-discharged holey carbon grids with a thin layer of carbon over the holes. To form the RNP-antibody complex, we incubated 100 nM of the GraFix-prepared RNP and 50 nM of monoclonal anti-His (A00186, GenScript) at 0 C for 30 min in 10 ml prior to grid application. After 1 min, grids were stained consecutively in 3 droplets of 2% (w/v) uranyl formate solution and excess stain removed by blotting with lter paper. For the two-dimensional analysis, samples were examined with an FEI Technai-12 electron microscope equipped with a LaB6 lament operated at 120 keV acceleration voltage using a nominal magnication of 42,000. Images
were recorded on a 4k 3 4k Ultrascan4000 CCD camera (Gatan) using low 2. The defocus used to dose mode with an exposure dose of 2030 eA collect the raw image was between 0.8 mm and 1.1 mm. Electron micrographs recorded using the Technai-12 electron microscope had a pixel size and were directly used for image processing. of 2.6 A For three-dimensional reconstruction of the Rsr/Y RNA/PNPase RNP, the sample was examined using a Technai 12 Bio-TWIN operated at 120 keV. per pixel The RNP was imaged at a nominal magnication of 49,000 (2.18 A at specimen level) and data were acquired on an F416CMOS 4k 3 4k CCD 2. Tilt pairs of images were recorded camera (TVIPS) using a dose of 20 eA automatically using the RCT application (Yoshioka et al., 2007) within the LEGINON data collection software (Suloway et al., 2005). Briey, the tilted (+60 ) image was collected rst using a defocus of approximately 1.5 mm, and then the stage was tilted back to 0 and a second image was collected using a defocus range of 0.7 mm to 1.4 mm. Image processing and threedimensional reconstruction are described in the Extended Experimental Procedures. Exonuclease Assays Degradation assays contained 20 fmol of PNPase or Rsr/Y RNA/PNPase RNP and 94 fmol RNA in 20 mM Tris-HCl [pH 7.5], 50 mM KCl, 1 mM MgCl2, 2 mM DTT, and 10 mM phosphate at 30 C. At intervals, aliquots were added to equal volumes 90% formamide, 1 mM EDTA, 0.25% SDS, 0.03% bromophenol blue, 0.03% xylene cyanol and fractionated in 8% or 15% polyacrylamide/ 8.3 M urea gels. A PhosphorImager (Molecular Dynamics), followed by ImageQuant analysis, was used to quantify degradation. Aliquots were subjected to immunoblotting with anti-PNPase (Chen et al., 2007) to conrm equal amounts of PNPase were added. For details, see the Extended Experimental Procedures. ACCESSION NUMBERS The EM map of the Rsr/Y RNA/PNPase RNP has been deposited into the Electron Microscopy Data Bank with accession number EMDB-5389. The GenBank accession numbers for YrlA and YrlB are KC584024 and KC584025, respectively. SUPPLEMENTAL INFORMATION Supplemental Information includes ve gures and Extended Experimental Procedures and can be found with this article online at http://dx.doi.org/10. 1016/j.cell.2013.02.037. ACKNOWLEDGMENTS X.C. identied the Rsr/Y RNA/PNPase RNP, reconstituted it in E. coli, and performed all biochemical experiments. D.W.T. performed the single-particle electron microscopy and three-dimensional reconstruction. C.C.F. prepared S. Typhimurium strains and extracts. All authors contributed to planning experiments and analyzing data. We thank J. Button for assistance in preparing S. Typhimurium strains; A. Carpousis, N. Kucera, A. Ramesh, J. Sacchettini, S. Sim, and E. Wurtmann for reagents; H. Zheng and K. Reinisch for assistance with baculovirus; G. Lander, M. Cianfrocco, and P. Grob for electron microscopy assistance; E. Conti for advice; and S. Baserga, E. De La Cruz, A. Horwich, J. Steitz, E. Ullu, and Y. Zuo for comments on the manuscript. We thank the Yale Center for Cellular and Molecular Imaging and E. Nogales for use of the EM facility at UC-Berkeley. D.W.T is an NSF Graduate Research Fellow. This work was supported by the Smith Family Awards Program for Excellence in Biomedical Research (to H.W.), NIH grants R01 AI079022 (to J.E.G.) and R01 GM073863 (to S.L.W.), and Molecular Biophysics training grant 5 T32 GM008283 (to D.W.T.). Received: April 10, 2012 Revised: January 10, 2013 Accepted: February 19, 2013 Published: March 28, 2013
REFERENCES Ataide, S.F., Schmitz, N., Shen, K., Ke, A., Shan, S.O., Doudna, J.A., and Ban, N. (2011). The crystal structure of the signal recognition particle in complex with its receptor. Science 331, 881886. Bonneau, F., Basquin, J., Ebert, J., Lorentzen, E., and Conti, E. (2009). The yeast exosome functions as a macromolecular cage to channel RNA substrates for degradation. Cell 139, 547559. ` re, F., and Boire, G. (2000). Interaction cloning and Bouffard, P., Barbar, E., Brie characterization of RoBPI, a novel protein binding to human Ro ribonucleoproteins. RNA 6, 6678. Chen, X., Quinn, A.M., and Wolin, S.L. (2000). Ro ribonucleoproteins contribute to the resistance of Deinococcus radiodurans to ultraviolet irradiation. Genes Dev. 14, 777782. Chen, X., Smith, J.D., Shi, H., Yang, D.D., Flavell, R.A., and Wolin, S.L. (2003). The Ro autoantigen binds misfolded U2 small nuclear RNAs and assists mammalian cell survival after UV irradiation. Curr. Biol. 13, 22062211. Chen, X., Wurtmann, E.J., Van Batavia, J., Zybailov, B., Washburn, M.P., and Wolin, S.L. (2007). An ortholog of the Ro autoantigen functions in 23S rRNA maturation in D. radiodurans. Genes Dev. 21, 13281339. Cheng, Z.F., and Deutscher, M.P. (2005). An important role for RNase R in mRNA decay. Mol. Cell 17, 313318. Evguenieva-Hackenberg, E. (2011). The archaeal exosome. Adv. Exp. Med. Biol. 702, 2938. Evguenieva-Hackenberg, E., Roppelt, V., Finsterseifer, P., and Klug, G. (2008). Rrp4 and Csl4 are needed for efcient degradation but not for polyadenylation of synthetic and natural RNA by the archaeal exosome. Biochemistry 47, 1315813168. Fabini, G., Raijmakers, R., Hayer, S., Fouraux, M.A., Pruijn, G.J., and Steiner, G. (2001). The heterogeneous nuclear ribonucleoproteins I and K interact with a subset of the Ro ribonucleoprotein-associated Y RNAs in vitro and in vivo. J. Biol. Chem. 276, 2071120718. Fouraux, M.A., Bouvet, P., Verkaart, S., van Venrooij, W.J., and Pruijn, G.J. (2002). Nucleolin associates with a subset of the human Ro ribonucleoprotein complexes. J. Mol. Biol. 320, 475488. Fuchs, G., Stein, A.J., Fu, C., Reinisch, K.M., and Wolin, S.L. (2006). Structural and biochemical basis for misfolded RNA recognition by the Ro autoantigen. Nat. Struct. Mol. Biol. 13, 10021009. Genschik, P., Drabikowski, K., and Filipowicz, W. (1998). Characterization of the Escherichia coli RNA 30 -terminal phosphate cyclase and its sigma54-regulated operon. J. Biol. Chem. 273, 2551625526. Gorna, M.W., Carpousis, A.J., and Luisi, B.F. (2012). From conformational chaos to robust regulation: the structure and function of the multi-enzyme RNA degradosome. Q. Rev. Biophys. 45, 105145. Green, C.D., Long, K.S., Shi, H., and Wolin, S.L. (1998). Binding of the 60-kDa Ro autoantigen to Y RNAs: evidence for recognition in the major groove of a conserved helix. RNA 4, 750765. Hannon, G.J., Rivas, F.V., Murchison, E.P., and Steitz, J.A. (2006). The expanding universe of noncoding RNAs. Cold Spring Harb. Symp. Quant. Biol. 71, 551564. Hogg, J.R., and Collins, K. (2007). Human Y5 RNA specializes a Ro ribonucleoprotein for 5S ribosomal RNA quality control. Genes Dev. 21, 30673072. Kastner, B., Fischer, N., Golas, M.M., Sander, B., Dube, P., Boehringer, D., Hartmuth, K., Deckert, J., Hauer, F., Wolf, E., et al. (2008). GraFix: sample preparation for single-particle electron cryomicroscopy. Nat. Methods 5, 5355. , J.C., Hekimi, S., and Rokeach, L.A. (1999). The levels of the RoRNPLabbe associated Y RNA are dependent upon the presence of ROP-1, the Caenorhabditis elegans Ro60 protein. Genetics 151, 143150.
Lerner, M.R., Boyle, J.A., Hardin, J.A., and Steitz, J.A. (1981). Two novel classes of small ribonucleoproteins detected by antibodies associated with lupus erythematosus. Science 211, 400402. Lorentzen, E., Basquin, J., Tomecki, R., Dziembowski, A., and Conti, E. (2008). Structure of the active subunit of the yeast exosome core, Rrp44: diverse modes of substrate recruitment in the RNase II nuclease family. Mol. Cell 29, 717728. Lunde, B.M., Moore, C., and Varani, G. (2007). RNA-binding proteins: modular design for efcient function. Nat. Rev. Mol. Cell Biol. 8, 479490. Mohanty, B.K., and Kushner, S.R. (2011). Bacterial/archaeal/organellar polyadenylation. Wiley Interdiscip Rev RNA 2, 256276. OBrien, C.A., and Wolin, S.L. (1994). A possible role for the 60-kD Ro autoantigen in a discard pathway for defective 5S rRNA precursors. Genes Dev. 8, 28912903. Perreault, J., Perreault, J.-P., and Boire, G. (2007). Ro-associated Y RNAs in metazoans: evolution and diversication. Mol. Biol. Evol. 24, 16781689. Ramesh, A., Savva, C.G., Holzenburg, A., and Sacchettini, J.C. (2007). Crystal structure of Rsr, an ortholog of the antigenic Ro protein, links conformational exibility to RNA binding activity. J. Biol. Chem. 282, 1496014967. Schmidt, T.G.M., and Skerra, A. (2007). The Strep-tag system for one-step purication and high-afnity detection or capturing of proteins. Nat. Protoc. 2, 15281535. Shi, H., OBrien, C.A., Van Horn, D.J., and Wolin, S.L. (1996). A misfolded form of 5S rRNA is complexed with the Ro and La autoantigens. RNA 2, 769784. Sim, S., and Wolin, S.L. (2011). Emerging roles for the Ro 60-kDa autoantigen in noncoding RNA metabolism. Wiley Interdiscip Rev RNA 2, 686699. Sim, S., Weinberg, D.E., Fuchs, G., Choi, K., Chung, J., and Wolin, S.L. (2009). The subcellular distribution of an RNA quality control protein, the Ro autoantigen, is regulated by noncoding Y RNA binding. Mol. Biol. Cell 20, 15551564. Sim, S., Yao, J., Weinberg, D.E., Niessen, S., Yates, J.R., 3rd, and Wolin, S.L. (2012). The zipcode-binding protein ZBP1 inuences the subcellular location of the Ro 60-kDa autoantigen and the noncoding Y3 RNA. RNA 18, 100110. Sohlberg, B., Huang, J., and Cohen, S.N. (2003). The Streptomyces coelicolor polynucleotide phosphorylase homologue, and not the putative poly(A) polymerase, can polyadenylate RNA. J. Bacteriol. 185, 72737278. Spickler, C., and Mackie, G.A. (2000). Action of RNase II and polynucleotide phosphorylase against RNAs containing stem-loops of dened structure. J. Bacteriol. 182, 24222427. Stein, A.J., Fuchs, G., Fu, C., Wolin, S.L., and Reinisch, K.M. (2005). Structural insights into RNA quality control: the Ro autoantigen binds misfolded RNAs via its central cavity. Cell 121, 529539. Suloway, C., Pulokas, J., Fellmann, D., Cheng, A., Guerra, F., Quispe, J., Stagg, S., Potter, C.S., and Carragher, B. (2005). Automated molecular microscopy: the new Leginon system. J. Struct. Biol. 151, 4160. Symmons, M.F., Jones, G.H., and Luisi, B.F. (2000). A duplicated fold is the structural basis for polynucleotide phosphorylase catalytic activity, processivity, and regulation. Structure 8, 12151226. Tanaka, N., and Shuman, S. (2011). RtcB is the RNA ligase component of an Escherichia coli RNA repair operon. J. Biol. Chem. 286, 77277731. Teunissen, S.W., Kruithof, M.J., Farris, A.D., Harley, J.B., Venrooij, W.J., and Pruijn, G.J. (2000). Conserved features of Y RNAs: a comparison of experimentally derived secondary structures. Nucleic Acids Res. 28, 610619. Wurtmann, E.J., and Wolin, S.L. (2010). A role for a bacterial ortholog of the Ro autoantigen in starvation-induced rRNA degradation. Proc. Natl. Acad. Sci. USA 107, 40224027. Yoshioka, C., Pulokas, J., Fellmann, D., Potter, C.S., Milligan, R.A., and Carragher, B. (2007). Automation of random conical tilt and orthogonal tilt data collection using feature-based correlation. J. Struct. Biol. 159, 335346.
Single-Cell Dynamics of Genome-Nuclear Lamina Interactions

Jop Kind,1,* Ludo Pagie,1 Havva Ortabozkoyun,1 Shelagh Boyle,3 Sandra S. de Vries,1 Hans Janssen,2 Mario Amendola,1 Leisha D. Nolen,3 Wendy A. Bickmore,3 and Bas van Steensel1,*
of Gene Regulation Microscopy Facility Netherlands Cancer Institute, Amsterdam, 1066 CX, the Netherlands 3MRC Human Genetics Unit, MRC Institute of Genetics and Molecular Medicine, University of Edinburgh, Edinburgh, EH4 2XU, United Kingdom *Correspondence: j.kind@nki.nl (J.K.), b.v.steensel@nki.nl (B.v.S.) http://dx.doi.org/10.1016/j.cell.2013.02.028
2Electron 1Division
SUMMARY
The nuclear lamina (NL) interacts with hundreds of large genomic regions termed lamina associated domains (LADs). The dynamics of these interactions and the relation to epigenetic modications are poorly understood. We visualized the fate of LADs in single cells using a molecular contact memory approach. In each nucleus, only 30% of LADs are positioned at the periphery; these LADs are in intermittent molecular contact with the NL but remain constrained to the periphery. Upon mitosis, LAD positioning is not detectably inherited but instead is stochastically reshufed. Contact of individual LADs with the NL is linked to transcriptional repression and H3K9 dimethylation in single cells. Furthermore, we identify the H3K9 methyltransferase G9a as a regulator of NL contacts. Collectively, these results highlight principles of the dynamic spatial architecture of chromosomes in relation to gene regulation.
INTRODUCTION It is well established that the positioning of interphase chromosomes within the three-dimensional space of the nucleus is nonrandom, yet it remains unclear how this spatial organization is brought about and what players are involved. One proposed mechanism involves the nuclear lamina (NL), a lamentous protein layer composed of A- and B-type lamins that coats the nucleoplasmic side of the inner nuclear membrane (Prokocimer et al., 2009; Shevelyov and Nurminsky, 2012). The NL provides a large surface area that appears to be an anchoring platform for chromosomes (Chubb et al., 2002; van Steensel and Dekker, 2010). Mapping by means of the DamID technology has revealed that mammalian genomes contain about 1,1001,400 lamina associated domains (LADs), which are regions of 0.110 Mb that specically associate with the NL. LADs are relatively gene poor, have a repressive chromatin signature, and are often
demarcated by specic sequence elements (Kind and van Steensel, 2010). LADs are found on all chromosomes and collectively cover 35%40% of the mammalian genome (Guelen et al., 2008; Peric-Hupkes et al., 2010). This remarkably large proportion suggests that LAD-NL interactions impose major constraints on the shape and positioning of chromosomes. Hence, much may be learned about chromosome architecture by studying LAD-NL interactions in more detail. For example, knowledge of the dynamics of these interactions throughout the cell cycle may shed light on the plasticity of chromosome folding, and comparison of LAD-NL interactions between mother and daughter cells could address the intriguing question of whether chromosome folding is heritable. In addition, an understanding of the molecular cues that target LADs to the NL will provide insight into some of the mechanisms that determine the overall architecture of chromosomes. Current knowledge of the dynamics of LAD-NL interactions in individual cells is mostly based on indirect evidence. Fluorescence in situ hybridization (FISH) microscopy of a handful of LADs has indicated that they are indeed preferentially, but not always, located near the NL (Guelen et al., 2008; Peric-Hupkes et al., 2010; Zullo et al., 2012), suggesting that the LAD-NL interactions are dynamic to some degree. However, FISH can only be performed in xed cells, and due to the limited resolution of light microscopy it is not possible to know whether a LAD makes direct molecular contact with the NL, or is merely nearby. Other approaches, such as tagging of single loci with Lac operon (LacO) arrays, as well as photobleaching and photoactivation experiments, have demonstrated that interphase chromatin is locally mobile but rarely moves over long distances, except during the rst 2 hr of G1 phase (Strickfaden et al., 2010; Thomson et al., 2004,; Vazquez et al., 2001; Walter et al., 2003). Yet it is not known whether this principle also applies to LADs, which could be rmly anchored to the NL and thereby provide chromosomes with a stable backbone. Furthermore, a few reports have addressed whether the nuclear position of single loci and entire chromosomes is heritable (Gerlich et al., 2003; Strickfaden et al., 2010; Thomson et al., 2004), but it is not clear whether the overall folding state of chromosomes is transmitted from mother to daughter cells.
Likewise, little is known about the mechanisms that drive LADNL interactions. Recently it was reported that long (GA)n repeats account for the peripheral positioning of certain LADs (Zullo et al., 2012). Besides specic DNA sequences, local chromatin characteristics of LADs may play a role, because a wide range of molecular interactions have been reported between chromatin components and NL proteins (Prokocimer et al., 2009) and methylated histone H3 lysine 9 was found to be important for the NL anchoring of certain genes in C. elegans (Towbin et al., 2012). Here, we report the use of DNA adenine methylation as an articial epigenetic tag to visualize, track and manipulate LADs in single human cells. A key feature of this approach is that regions of genomic DNA are permanently labeled in living cells once they contact the NL, enabling us to follow their fate over time. Using this method, we demonstrate that LAD-NL contacts are highly dynamic and that positioning of LADs at the NL is determined in an apparently stochastic manner early after mitosis. Furthermore, we nd that the stochastic contacts of LADs with the NL are directly linked to gene repression and their level of the histone modication H3K9me2 and that the H3K9 methyltransferase G9a controls the contact frequency of LADs with the NL. RESULTS A-Tracer Technology Our method of tracking genome-NL interactions in single cells builds upon the DamID technology, in which genomic regions that are in molecular contact with a nuclear protein of interest are tagged in vivo with adenine-6-methylation (m6A), a modication unknown to higher eukaryotes (van Steensel and Henikoff, 2000). This method was previously used to map genome-NL contacts (Guelen et al., 2008; Pickersgill et al., 2006). Specically, by expressing a fusion protein consisting of E. coli DNA adenine methyltransferase (Dam) and Lamin B1, any DNA in molecular contact with the NL is adenine methylated. Because m6 A is a stable covalent modication, we reasoned that visualization of this mark inside cells would highlight not only DNA in direct contact with the NL but also any DNA that was adenine methylated in a previous contact event. Hence, this strategy could provide a microscopic history tracking of genome-NL interactions. In order to visualize m6A in the genome of intact cells, we required a protein module that specically recognizes this DNA modication. We reasoned that the restriction endonuclease DpnI, which cuts the sequence Gm6ATC but not GATC (Hermann and Jeltsch, 2003), might contain a domain that selectively binds to Gm6ATC. Because DpnI has been poorly characterized, we systematically tested eight different truncations of this protein for the desired specicity (Figure 1A). We fused each truncation to enhanced green uorescent protein (eGFP) and coexpressed it with either Dam-Lamin B1 or Dam-only in the human brosarcoma cell line HT1080 (Figure 1B). Four overlapping DpnI truncation fragments yielded a conspicuous nuclear rim staining when coexpressed with Dam-Lamin B1 but not with Dam alone, indicating that they harbor a Gm6ATC-binding domain. This is in agreement with the recently reported crystal structure of DpnI (Siwek et al., 2012). We proceeded with truncation #7, a C-terminal fragment of 109 amino acids that illuminated DNA at the
m6
nuclear periphery most consistently when coexpressed with Dam-Lamin B1 (data not shown). We will refer to this fragment fused to eGFP as m6A-Tracer. Biochemical characterization indicated that the afnity of m6ATracer is much higher for fully adenine-methylated GATC motifs compared to unmethylated or hemimethylated DNA. Even a 1,000-fold excess of unmethylated competitor does not detectably affect binding to fully methylated GATC (Figure 1C). None of the DpnI truncations, including #7, harbor any endonuclease activity (Figure 1D), which would be undesirable for its application in living cells. To further verify that m6A-Tracer, when coexpressed with Dam-Lamin B1, indeed binds LADs, we conducted chromatin immunoprecipitation (ChIP) experiments with an antibody against eGFP. Quantitative PCR (qPCR) of immunoprecipitated material conrmed that m6A-Tracer in the presence of DamLamin B1 is enriched in LADs (Figure 1E). Immunouorescence microscopy showed that Lamin B1 is essentially absent from the nuclear interior in HT1080 cells (Figures S1A and S1B, available online); consistent with this we observe virtually no m6ATracer signal in the nuclear interior 24 hr after cotransfection of Dam-Lamin B1 and m6A-Tracer (Figures S1C and S1D). In contrast, Lamin A is clearly present in the nuclear interior, and Dam-Lamin A yields substantial internal m6A-Tracer signals (Figures S1AS1D). Collectively, these data demonstrate that m6 A-Tracer, when coexpressed with Dam-Lamin B1, can be used to specically visualize genomic regions that engage in molecular contact with the NL. Interphase LADs Are Mobile within a Narrow Zone underneath the NL To gain insight into the dynamics of LADs, we generated an HT1080-derived clonal cell line in which both m6A-Tracer and Dam-Lamin B1 can be induced independently (Figure 2A). The induction of m6A-Tracer is based on the Tet-Off system, where the removal of Doxycycline (Dox) results in the activation of transcription (Gossen and Bujard, 1992). Inducible Dam-Lamin B1 expression was established by the fusion of a destabilization domain (DD) (Banaszynski et al., 2006), which causes DamLamin B1 to be rapidly targeted for proteosomal degradation unless the protein is shielded by the synthetic small molecule Shield1. We refer to this cell line as clone3. Neither the fusion of the DD nor the relatively high expression levels of Dam-Lamin B1as compared to the standard DamID protocol (Meuleman et al., 2013)affected the genome-NL interactions (Figure S2A). Furthermore, m6A-Tracer and Dam-Lamin B1 expression did not detectably affect cell health, viability, or cell-cycle progression (Figure S2B and Extended Experimental Procedures), nor did it alter local chromatin compaction or the histone modication patterns of LADs (Figures S2CS2E). In order to study the dynamics of genome-NL contacts in interphase, we arrested clone3 cells in mitosis by the addition of nocodazole to the medium; after 2 hr, we collected the mitotic cells by mitotic shake-off, released them from the nocodazole block, and induced Dam-Lamin B1 expression. We then monitored the m6A-Tracer patterns throughout interphase over 25 hr. Within 5 hr after Dam-Lamin B1 induction, the m6A-Tracer signal became visible at the periphery of the nucleus (Figure 2B).
A
eGFP eGFP eGFP eGFP
50 1
100
150
200
250
2 3 4
eGFP eGFP
B
Dam Dam-Lamin B1
eGFP Lamin B1
eGFP eGFP
5 6 7 8
C
competitor competitor 0 10 100 1000 competitor 0 10 100 1000 0 10 100 1000
2
Fold binding over eGFP-H2B
70 60
Exp1
50 40 30 20 10 0
Exp2
5
200bp 100bp
FL NEB
6
E
0.8
% recovery over Input
m6A-Tracer m6A-Tracer+Dam m6A-Tracer+Dam-Lamin B1
0.7 0.6 0.5 0.4 0.3 0.2 0.1 0

LAD 2 LAD 8 CFH R3 CYP 2C1 9 UBE 2B iLAD 1 LAD 2 LAD 8 CFH R3 CYP 2C1 9 UBE 2B iLAD 1 LAD 2 LAD 8 CFH R3 CYP 2C1 9 UBE 2B iLAD 1
Exp1 Exp2
Figure 1. The m6A-Tracer System to Study the Dynamics of NL-Genome Interactions

(A) Schematic representation of eight DpnI truncations fused to eGFP. Scale bar is in amino acids. (B) Cellular localization of each eGFP-DpnI truncation upon coexpression with either Dam (left column) or Dam-Lamin B1 (right column). (C) In vitro m6A-binding of eGFP-DpnI truncation #7 to a DNA fragment containing an unmethylated, hemimethylated or fully methylated GATC sequence, in the presence of various amounts (fold excess) of unmethylated competitor DNA. Data from two independent experiments (Exp1 and Exp2) are shown.
m6
nnnGATCnnn nnnCTAGnnn
m6
m6
Digest
This early signal appeared in patches, but over time most of the NL became lined with a layer of m6A-Tracer signal. Interestingly, already at 5 hr, we observed occasional signals that were up to 1mm away from the NL, which is evidently too far to be in molecular contact. Such detached signals became progressively more frequent over time (Figure 2B). Because the m6A tags are stable, these regions must have contacted the NL at an earlier time point and then moved away from the NL. However, even after 25 hr the vast majority of m6A-Tracer signals remained within 1 mm from the NL (Figures 2B and 2C). From these cumulative labeling and tracking experiments we infer that genome-NL contacts during interphase are generally dynamic, because regions that have contacted the NL may move up to 1 mm from the NL within a few hours. Migration beyond this distance is very rare, suggesting that the mobility of LADs during interphase is conned to a narrow zone near the NL. It has been suggested that the transcription machinery may play an important role in nuclear organization (Cook, 2010). For example, active genes may be anchored to transcription factories in the nuclear interior, and as a consequence, inactive parts of the genome may be passively driven to the nuclear periphery. We therefore tested whether the peripheral positioning of LADs could be perturbed by inhibition of global transcription. After 24 hr exposure of cells to alpha-amanitin, which causes nearcomplete shutdown of RNA polymerase II activity, we observed a signicant dispersal of m6A-Tracer signal (Figures S2FS2I). However, the overall peripheral distribution was only partially disrupted, suggesting that other mechanisms contribute to the NL connement of LADs. Dynamic Genome-NL Contacts Involve Classical H3K9me2-Marked Heterochromatin In order to study the nature of the narrow dynamic LAD zone in more detail, we conducted electron microscopy after immunogold labeling of eGFP in U2OS cells cotransfected with DamLamin B1 and the m6A-Tracer. Twenty-four hours after transfection, we found m6A-Tracer signals predominantly in a layer less than 1 mm from the NL consistent with light microscopy observations (Figures 2D and S2J). The immunogold particles are preferentially found throughout the peripheral electrondense chromatin material that is classically dened as heterochromatin. Thus, the dynamic contacts of the genome with the NL primarily involve this dense type of heterochromatin. The condensed appearance is not caused by m6A-Tracer binding, because coexpression of m6A-Tracer with unfused Dam yields immunogold labeling that is mostly located in less electrondense regions of the nucleus (Figures S2K and S2L). The histone modication H3K9me2 was previously found at the nuclear periphery of mammalian cell nuclei (Wu et al., 2005; Yokochi et al., 2009) and found to be enriched in LADs (Guelen et al., 2008; Peric-Hupkes et al., 2010; Wen et al., 2009). Staining of clone3 and parental HT1080 cells with an anti-
body against H3K9me2 conrmed that this mark is enriched at the nuclear periphery and generally overlaps with the m6A-Tracer signals (Figures 2E and S2M). ChIP of H3K9me2 conrms this enrichment in LADs (Figure S2E). In contrast, the related modication H3K9me3 is less enriched at the periphery, consistent with previous data (Wu et al., 2005; Yokochi et al., 2009), and typically does not overlap with m6A-Tracer (Figures 2E and S2M). By ChIP we also did not nd H3K9me3 enriched in LADs (Figure S2E). We note, however, that H3K9me3 has been predominantly found at centromeric regions (Peters et al., 2001; Wu et al., 2005), which cannot be detected by the m6A-Tracer technology because centromeric repeats lack GATC sequence motifs. Together, these results indicate that LADs at the nuclear periphery correspond primarily to heterochromatin that appears condensed in transmission electron microscopy and is marked by H3K9me2. Disruption of LAD-NL Contacts by Targeting of an Activator We wondered whether the heterochromatic state of LADs is required for their molecular contacts with the NL. To address this question we made use of the acidic-activating domain (AAD) of the viral protein VP16, which has been reported to disrupt the heterochromatic state and peripheral positioning of an articial repeat array (Chuang et al., 2006; Tumbar et al., 1999). We targeted the AAD specically to LADs by fusing it to the m6A-Tracer and coexpressing it with Dam-Lamin B1. Strikingly, targeting of a tandem array of three copies of this AAD to LADs caused substantial broadening and a more dispersed appearance of the sub-NL zone of m6A-Tracer signals (Figures 3A and 3B), even though the integrity of the NL was not visibly affected (Figure 3A). Thus, the VP16 AAD, when targeted to LADs, causes destabilization of the peripheral positioning of LADs. However, not all LADs appear equally sensitive to VP16, because in all nuclei the most peripheral shell retained substantial m6A-Tracer signals. This destabilization of LAD positioning by VP16 is not due to activation of genes within LADs, because messenger RNA sequencing (mRNA-seq) showed very few signicant increases in gene expression within LADs after targeting of the AAD for 24 hr (Figure 3C). Because targeted VP16 is known to cause changes in histone modications (Carpenter et al., 2005), we asked whether increased acetylation of histones might account for the relocation of LADs. By quantitative immunouorescence microscopy we observed that histone H3 acetylation (H3Ac) increased in LADs when VP16 was fused to m6A-Tracer compared to m6A-Tracer alone (Figure 3D, rst and third column). However, this increase was similar in LADs that remained at the NL, compared to LADs that had moved to the nuclear interior (Figure 3E). It is therefore unlikely that elevated H3ac alone causes detachment of LADs from the NL.
(D) Restriction endonuclease activity of the DpnI truncations and full-length DpnI (fused to eGFP) and commercial pure DpnI (NEB), using a dsDNA fragment containing a single fully methylated GATC sequence as substrate. Arrow points to the digested product. (E) ChIP of m6A-Tracer, without or with coexpression of Dam alone or Dam-Lamin B1, using an antibody against eGFP. Recovery relative to input was determined for four LAD (red) and two inter-LAD (black) sequences (Table S3). Data are from two independent experiments (Exp1 and Exp2). Note that Dam methylates all loci, whereas Dam-Lamin B1 specically methylates LADs. See also Figure S1 and Table S3.
A
Unstable +Shield1 Stable LAD tTA -Dox tetO LAD 150 5hr 10hr Harvest
Intensity
m6A-Tracer m6A-Tracer Lamin B1
DD-Dam-LaminB1
m6A-Tracer m6A
B
5 hour 10 hour 15 hour 20 hour 25 hour shake off +shield/-Dox
7.8 +/- 0.8 8.4 +/- 1.7 8.7 +/- 1.3 8.5 +/- 1.1 8.2 +/- 1.2
100
15hr 20hr 25hr
5hr
50
1 pixel= 87.4nm
0
12.8 m
10 20 30 Pixel distance from the NL
40
10hr
10 m
NL Cytoplasm
15hr
Nucleus
10 m
500nM
20hr
m6A-Tracer
H3K9me3
m6A-Tracer K9me3
10 m
1
H3K9me2
3
m6A-Tracer K9me2
25hr
10 m
K9me2
K9me3
Figure 2. Conned Dynamics of LADs during Interphase

(A) Design of the inducible m6A-Tracer/Dam-Lamin B1 system. (B) Representative confocal sections showing accumulation of m6A-Tracer signal at the nuclear periphery at indicated times after induction of Dam-Lamin B1 expression, which was started immediately upon the release from a nocodazole block. Note that LADs marked by m6A-Tracer (green) can move over short distances from the NL (arrow) but are not typically found in the nuclear interior. Lamin B1 staining is shown in blue.
m6A-Tracer m6A-Tracer Lamin B1
VP16-m6A-Tracer
m6A-Tracer Lamin B1
m6A-Tracer-VP16 m6A-Tracer Lamin B1
B
eGFP VP16 Intensity DpnI 7th truncation 70 60 50 40 30 20 10 0 0 10 20 30 40 50 60 Pixel distance from the NL
m6A-Tracer m6A-Tracer-VP16
9.5 +/- 2.2 12.7 +/- 2.2
p=4.50e-05 VP16- m6A-Tracer 15.4 +/- 4.4 p=1.08e-07 1 pixel= 87.4nM
C
domain genes up/down up 3 2 LAD 5716 down 6 3 up 5 2 iLAD 24776 down 11 8
m6 A-Tracer m6A-Tracer Dapi H3ac
m6 A-Tracer m6A-Tracer Dapi
m6 A-Tracer-VP16 m6A-Tracer Dapi H3ac
m6 A-Tracer-VP16 m6A-Tracer Dapi
-Dox -Shield1 -Dox +Shield1
H3K9me2
H3K9me2
Nucleoplasm Periphery
p=0.1698
Mean H3K9me2 in LADs
p=1.49e-08
Mean H3ac in LADs
200 150 100 50 0
200 150 100 50 0
Figure 3. VP16 Targeting Dislodges LADs from the NL

(A) Representative images showing the displacement of the m6A-Tracer signal away from the NL after targeting of the VP16 AAD to LADs (center and right columns) compared to m6A-Tracer without the AAD (left column). Note that the displacement is similar for N- and C-terminal AAD fusions. Bottom panels show separate channels at higher magnications. (B) Average radial distribution of m6A-Tracer signals after induction of the m6A-Tracer alone (black line) or fused to the AAD (blue lines). Numbers in the upper-right corner represent average ( SD) distance (in pixels) from the NL at which m6A-Tracer signals are at half-maximum intensity (n = 25). p value according to an unpaired Wilcoxon test. (C) Gene expression analysis 24 hr after induction of either m6A-Tracer-VP16 alone (top row in gray) or in combination with Dam-Lamin B1 (bottom row in gray) compared to control samples without induction. Numbers in gray boxes indicate genes with signicantly altered expression levels. (D) Immunouorescence images showing H3ac (red) inltration into the NL compartment upon AAD targeting (third column, arrow), an environment normally devoid of H3Ac (rst column, arrow). Second and fourth column, H3K9me2 (red) is reduced on LADs that moved to the nuclear interior after AAD targeting. m6ATracer is shown in green and DNA in blue. (E and F) Per-pixel quantication of the H3Ac (E) and H3K9me2 (F) intensity levels on LADs located in the nucleoplasm (left bar) and at the NL (right bar). LADs are dened here as m6A-Tracer signals with pixel values R 150. Data are represented as mean SD (n = 28). p value according to a paired Wilcoxon test.
We then investigated the state of H3K9me2. Remarkably, after targeting of VP16, LADs that remained at the nuclear periphery retained H3K9me2, while LADs that relocated to the nuclear interior showed signicantly less H3K9me2 (Figure 3D,
second and fourth column; Figure 3F). This correlation is suggestive of a role for H3K9me2 in the retention of LADs near the NL. We therefore propose that alterations in the chromatin state can dislodge LADs from the NL in the absence of gene
(C) Average radial distribution of m6A-Tracer signals at indicated times after Dam-Lamin B1 induction (n = 50, except for t = 5 hr with n = 32). Numbers in the upperright corner represent average ( SD) distance (in pixels) from the NL at which m6A-Tracer signals are at half-maximum intensity. (D) Electron micrograph of a U2OS nucleus after immunogold labeling of m6A-Tracer coexpressed with Dam-Lamin B1. Note that gold particles are present throughout heterochromatin, but much less in euchromatin. (E) Immuouorescent labeling showing that m6A-Tracer (green) after induction of Dam-Lamin B1 colocalizes substantially with H3K9me2 (images 2, 4, and 6) but much less with H3K9me3 (images 3, 5, and 7). See also Figure S2.
A
prophase pro-metaphase metaphase
anaphase
telophase
cytokinesis
m6A-Tracer Lamin B1 Alpha-Tubulin
H3K27ac CREST
H3K4me2 CREST
m6A-Tracer Lamin A
m6A-Tracer Lamin A
m6A-Tracer CREST
m6A-Tracer CREST
m6A-Tracer Lamin A Dapi
m6A-Tracer Lamin A Dapi
Figure 4. Fate of LADs during Mitosis

(A) LAD distribution at different stages of mitosis. Confocal sections show m6A-Tracer (green), Lamin B1 (blue), and alpha-tubulin (red).
activation. Additional evidence for a causal role of H3K9me2 will be presented below. LADs Form Discrete Metaphase Chromosome Bands but Do Not Nucleate NL Reassembly in Telophase To investigate the fate of LADs during mitosis, we activated DamLamin B1 in unsynchronized clone3 cells with Shield1 for 20 hr, after which we xed the cells and counterstained with antibodies against alpha-tubulin and Lamin B1, which enabled us to identify cells in specic stages of mitosis (Figure 4A). In prophase, LADs become condensed and rounded, presumably reecting the ongoing chromosomal compaction. Strikingly, in prometaphase and metaphase (when the NL is disassembled), LADs form a clear banded pattern along the condensed chromosomes (Figure 4A). This pattern alternates with active histone marks such as H3K27ac and H3K4me2 (Figure 4B), indicating that the spatial segregation of LADs and transcriptionally active regions during interphase is largely preserved in mitotic chromosomes. The LAD bands typically span the entire width of metaphase chromosomes, with no apparent preferential radial positioning. At the anaphase-telophase transition, the NL is reassembled at the surface of the chromosomes (Moir et al., 2000). Recently, an ectopically integrated LAD-derived element was reported to act as a nucleation site for NL assembly (Zullo et al., 2012). In order to test whether endogenous LADs generally perform this function, we analyzed the location of Lamin A relative to the m6 A-Tracer signals in early and late telophase cells. Confocal microscopy imaging showed deposition of Lamin A along most of chromosomal surface in late telophase but without any detectable preference for LADs (Figure 4C, see arrows in bottom panels). Thus, in general, chromosomal regions that were LADs in the mother cell do not act as preferred nucleation points for initial NL assembly in the daughter cells. Stochastic Reshufing of LADs after Mitosis Even though the newly formed NL shows no preference for LADs in telophase, we hypothesized that LADs would return to the NL early in G1 phase, when chromatin is highly mobile (Dimitrova and Gilbert, 1999; Thomson et al., 2004). To monitor this, we conducted time-lapse microscopy of the m6A-Tracer, using Red Fluorescent Protein-tagged histone H2B (H2B-mRFP) to simultaneously visualize bulk chromatin. Dam-Lamin B1 expression was induced for 20 hr prior to imaging of telophase cells. We thus could specically track the LADs that had been in contact with the NL in the mother cells (mother LADs) and study their fate in early G1 of the daughter cells. The results revealed that in telophase most mother LADs are not yet positioned at the periphery of the newly formed nuclei (Figures 5A and S3A). Only after 40 min into G1 phase, a subset of LADs appears to reassociate with the NL (Figure 5A). However, many of the mother LADs do not return to the NL; even 5 hr after mitosis, we found the majority of m6A-Tracer signal to be more than 1 mm from the nuclear rim (Figure 5B). These internal signals
show a modest preference to be located close to nucleoli (Figures 5B and 5C). To ensure that the m6A-Tracer itself does not hamper the reassociation of mother LADs with the NL after mitosis, we also activated Dam-Lamin B1 prior to mitosis and turned on m6A-Tracer expression only early in G1 of the next cell cycle, i.e., after LAD positioning had stabilized. Again, in the daughter cells many mother LADs were located in the nuclear interior (Figure S3B), indicating that the failure of individual LADs to localize to the NL is not caused by m6A-Tracer binding. We therefore conclude that the peripheral positioning of LADs after mitosis is a rather erratic process. The apparent stochastic association of LADs with the NL raised the question of what proportion of all LADs are NL associated on average. In order to quantify this, we performed DNA FISH with a pool of 140 3 105 oligonucleotides designed to simultaneously detect 25 LADs distributed over most chromosomes (Table S1). This revealed that on average 32.0% 11.6% (mean SD) of these LADs is located in close proximity (less than 1 mm) to the NL (Figures 5D and 5E). If the LAD-NL contacts after mitosis are established in a purely stochastic manner, then after every cell division each LAD should have a renewed chance to associate with the NL. Because the m6A tags deposited by Dam-Lamin B1 are stable, this predicts that the pool of LADs should acquire additional m6 A with each new mitosis (provided that Dam-Lamin B1 is continuously expressed). We tested this by quantitative measurements of m6A in several LADs over time in synchronized cell populations that express Shield1-inducible Dam-Lamin B1. The m6A levels were determined using the restriction endonuclease DpnII (van Steensel and Henikoff, 2000), which specically cuts unmethylated but not hemimethylated or fully methylated GATC sequences (Hermann and Jeltsch, 2003) (see Table S2 for an overview of the m6A assays used). We initially monitored accumulation of m6A during the rst interphase when Dam-Lamin B1 expression was switched on. In agreement with the m6A-Tracer microscopy time series (Figure 2B), the m6A levels of individual LADs (but not of inter-LAD regions) increased over time; however, after 15 hr a plateau was reached (Figures 5F and S3C), indicating that all LAD sequences in the peripheral dynamic zone had touched the NL and were fully methylated. When cells were subsequently allowed to go through another mitosis, the cumulative m6A levels of each tested LAD nearly doubled in the next interphase (Figure 5G). This doubling indicates that NL association in the daughter cells involves a LAD subset that is largely different from the LAD subset that had contacted the NL in the mother cells. Lastly, we asked whether the stochastic decision to position a LAD at the periphery is made independently for each of the two daughter cells or instead is somehow coordinated. By systematic scoring of LAD FISH signals in pairs of daughter cells, we found no signicant correlation between the radial positions in the two daughter cells (Figures S3D and S3E). This corroborates
(B) Distribution of LADs (green) in relation to H3K27ac (red, left panel) or H3K4me2 (red, right panel). CREST antibody (blue) marks centromeric regions. Insets show line scans along the chromosomal axis. (C) Lamin A (red) during early (left panel) and late (right panel) telophase reassembles onto chromosomes (blue) but shows no preference for LADs (green). Arrows mark some Lamin A-rich regions devoid of LADs.
A
0min 6min 12min
18min
24min
30min
36min
42min
48min
m6 A-Tracer RFP-H2B
B
m6 A-Tracer m6 A-Tracer Dapi m6 A-Tracer NPM1
C
Dapi Intensity 80 60 40 20 0 0 20 40 60 80 100 Pixel distance from the Nucleolus
1 pixel= 43.7nm
D
LAD probe pool Dapi LAD probe pool Dapi
E
30 Percentage 25 20 15 10 5 0 0-8 9-16 17-24 25-32 33-40 41-48 Pixel distance from the nuclear rim
**
LAD probe pool Nuclear volume

1 pixel= 87.4nm
F
25 20 % m6 A 15 10 5 0
LADs
iLADs 0hr 5hr 10hr 15hr 20hr
G
50 40 % m6 A 30 20 10 0
LADs
iLADs 18hr 36hr
61
2B
B3
AD
B3
AD iL
A6
A6
BE
LA
LA
LA
S4
S4
LA
iL
Figure 5. Reorganization of LADs after Mitosis

(A) Time-lapse imaging of LADs (green and grayscale) and Histone H2B (red) starting from late telophase. (B) Organization of LADs (m6A deposited by Dam-Lamin B1 expression in the mother cell) in a representative daughter cell about 5 hr after mitosis. Nucleoli are labeled using an antibody against nucleophosmin (NPM1; red).
BE
2B
61
63
63
earlier observations for laco-tagged loci (Thomson et al., 2004) and underscores that the peripheral position of LADs after mitosis is largely determined by a stochastic mechanism. H3K9me2 State of LADs Is Directly Linked to Stochastic Positioning Given the enrichment of H3K9me2 at the nuclear periphery (Figure 2E) (Wu et al., 2005; Yokochi et al., 2009) and the conspicuous loss of this mark on the dislodged LADs after AAD targeting (Figures 3D and 3F), we asked whether the naturally occurring stochastic contacts with the NL are related to the H3K9me2 status of the LADs. We developed an approach that enabled us to infer a direct relationshipat the level of LAD copies in individual cells from population-based measurements. As before, we isolated mitotic cells, activated DamLamin B1 expression by treatment with Shield1, and harvested the cells 18 hr later, i.e., before the next mitosis. We collected chromatin from these cells and performed ChIP with antibodies directed against H3K9me2 (or total H3 as a control). On this immunopuried material we then conducted the DpnII-based assay to estimate NL contact frequencies, enabling us to test whether the stochastic NL contacts are linked to variation in H3K9me2 levels (Table S2). Remarkably, seven out of eight LADs show a signicantly higher propensity to be associated with the NL when enriched for H3K9me2, than when not (Figure 6A). A striking example is the promoter region of CDH12, which is found at the NL in 58% of the cases when modied by H3K9me2, as compared to a general frequency of 25% (Figure 6A, rst panel). The same experiment conducted in parallel without Shield1 shows that the protection of DNA against digestion by DpnII is due to the presence of m6A and not due to the inability of the enzyme to digest formerly formaldehyde crosslinked material (Figure S4A). These ndings imply that the stochastic positioning of a LAD is directly linked to cell-to-cell or allele-to-allele variation in the H3K9me2 level of this LAD (see Discussion). The H3K9 Methyltransferase G9a Promotes Lad-NL Contacts We next tested whether H3K9me2 is required for LAD-NL associations. In mammals, a heteromeric complex of the histone lysine methyltransferases (HKMTs) G9a and GLP1 is primarily responsible for H3K9 dimethylation (Shinkai and Tachibana, 2011). We used RNA interference (RNAi) to knock down G9a and conrmed that this lowered the overall H3K9me2 levels (Figures S4B and S4D), as was reported previously (Tachibana et al., 2005; Yokochi et al., 2009). Strikingly, this caused all eight tested LADs to be less frequently associated with the NL
compared to control-treated cells (Figure 6B). A similar effect was observed after treatment with BIX01294, an inhibitor of the HKMT activity of both G9a and GLP1 (Kubicek et al., 2007; Quinn et al., 2010) (Figure 6B). We found no additional effect when the G9a knockdown was combined with BIX01294 treatment (Figure 6B). In contrast, overexpression of mCherrytagged G9a caused increased overall H3K9me2 levels and enhanced NL interactions (Figures 6B, S4C, and S4E). The magnitude of this effect may be limited by the availability of the G9a dimerization partner GLP1. In summary, we conclude that G9a modulates LAD-NL contacts, presumably through dimethylation of H3K9. Leaky LAD Genes Are More Active when Located Away from the NL While most genes in LADs are expressed at undetectable or very low levels, a small proportion produce substantial amounts of mRNA (Figure 7A). We asked whether these leaky LAD genes might become derepressed in particular when stochastically located in the nuclear interior. To address this question we rst visualized mother LADs in daughter cells together with H3K4me3. We found that this marker of transcriptional activity is generally low in mother LADs, both at the periphery and nuclear interior (Figure 7B). However, some overlap of internal LADs and H3K4me3 could be detected, and quantitative image analysis suggested slightly higher overall levels of H3K4me3 on internal mother LADs compared to peripheral mother LADs (Figure 7C). Although we note that this overlap may be overestimated due to the resolution of confocal microscopy, this result suggests that some genes within LADs may become active when located in the nuclear interior. To investigate this at molecular resolution, we repeated the ChIP-DamID experiments as in Figure 6, but this time using an antibody against H3K36me3, which is a marker of transcriptional elongation (Bannister et al., 2005). We focused on ve LAD genes with detectable expression according to mRNA-seq (Figure 7A). Strikingly, m6A levels deposited on these genes by Dam-Lamin B1, again measured in mother cells prior to mitosis, were consistently lower in the H3K36me3-positive chromatin fraction compared to the total H3-containing fraction (Figure 7D). Note that this result is exactly opposite to that obtained for H3K9me2 (Figure 6A). We conclude that the H3K36me3 state (and thus presumably the transcription activity) of these genes is higher in cells in which the respective LADs are not in contact with the NL. This implies that contact of these genes with the NL is inversely linked to stochastic variation of their expression in the population of cells.
(C) Average distribution of m6A-Tracer signals in daughter cells (n = 39) as a function of distance to the nearest nucleolus. Positions less than 1.4 mm from the NL were excluded from this analysis. (D) FISH of 25 LADs (Table S1) in metaphase (left panel) and interphase (right panel). (E) Radial distribution of FISH signals (n = 29 nuclei). **p < 0.001 according to a paired Wilcoxon test. (F) Quantitative analysis of m6A accumulation within one interphase at indicated time points after Dam-Lamin B1 induction, which was initiated simultaneously with the release from a nocodazole block. Data were generated in clone11 cells that do not express m6A-Tracer; a similar result was obtained in clone3 cells that do express m6A-Tracer (Figure S3C). (G) Same as in (F) except that Dam-Lamin B1 was induced for either 18 hr (gray bars), when m6A deposition is due to NL interactions in mother cells, or 36 hr (black bars), when m6A levels are the cumulative result of NL interactions in mother and daughter cells. Data are represented as mean SD (n = 3). See also Figure S3 and Tables S1, S2, and S3.
Histone H3 H3K9me2 CDH12 LAD2
B
CDH12 LAD2
0.8 Fraction m6A 0.6 0.4 0.2

0
**
0.8 0.6 0.4 0.2

0
Relative NL association
2.5 2.0 1.5 1.0 0.5 0 2.5 2.0 1.5 1.0 0.5 0 2.5 2.0 1.5 1.0 0.5 0 2.5 2.0 1.5 1.0 0.5 0
2.5 2.0
G9a RNAi BIX G9a RNAi/BIX mCherry-G9a
** ** * *
1.5 1.0 0.5 0
* * * *
0.8 Fraction m6A 0.6 0.4 0.2

0
CYP2C19
0.8 0.6
LAD6
CYP2C19
2.5 2.0
LAD6
* ** * *
1.5 1.0 0.5 0
* ** * *
**
0.4 0.2
0
**
0.8 Fraction m6A 0.6 0.4 0.2

0
CFHR3
0.8 0.6
LAD8
CFHR3
2.5 2.0
LAD8
* *
* ** ** *
1.5 1.0 0.5 0
**
0.4 0.2
0
**
0.8 Fraction m6A 0.6 0.4 0.2

0
LAD1
0.8 0.6 0.4 0.2

0
LAD63
LAD1
2.5
LAD63
2.0 1.5
**
**
1.0
0.5 0
Figure 6. H3K9me2 by G9a Mediates LAD-NL Associations

(A) Stochastic NL interactions are linked to H3K9me2 status. Chromatin from Dam-Lamin B1-expressing cells was subjected to ChIP with antibodies directed against Histone H3 (gray) or H3K9me2 (black), followed by quantication of m6A levels (representing NL contact frequencies), in eight LADs (n = 3). (B) G9a controls LAD-NL interactions. Bars show fold change in NL contact frequencies of eight LADs upon siRNA-mediated knockdown of G9a (G9a RNAi), G9a inhibition by BIX01294 (BIX), a combination of G9a siRNA and BIX01294 (G9a RNAi/BIX), or overexpression of Cherry-G9a (Cherry-G9a). Data are based on four biological replicates and represented as mean SD. *p < 0.05, **p < 0.01 according to a paired t test. See also Figure S4 and Tables S2 and S3.
DISCUSSION Here we demonstrated the use of a molecular contact memory approach to follow the fate of NL-genome contacts over time and to identify a role for H3K9me2 in the stochastic positioning of LADs at the periphery. Tracking and Targeting of Native Genomic Loci by m6 A-Tracer We have harnessed DNA adenine methylation as an articial epigenetic mark to track and manipulate genome-NL contacts in single human cells. Because m6A is a stable modication, we could use the combination of inducible Dam-Lamin B1 and m6 A-Tracer constructs as a history tracking system to follow the fate of genomic regions that contact Lamin B1 at a certain point in time. In addition, we demonstrated that these regions
can be collectively manipulated by the fusion of a transcription activator to the m6A-Tracer. We expect that various other chromatin regulators and cofactors may be directed in a similar fashion. Moreover, because m6A can be specically deposited at target loci of a wide range of chromatin proteins (Filion et al., 2010), the m6A-Tracer system may be applicable to study many aspects of chromatin biology. Constrained Mobility of LADs during Interphase Our results show that within interphase, contacts of the genome with the NL are dynamic. Nevertheless, during interphase loci generally do not migrate far from the NL after a contact event. This constrained mobility of chromatin is in agreement with previous studies that employed photobleaching of nuclear regions or uorescent tagging of single loci (Abney et al., 1997; Chubb et al., 2002; Vazquez et al., 2001), although these studies
1.2
B
LAD iLAD
m6 A-Tracer H3K4me3
m6A-Tracer
Figure 7. Dissociation of LADs from the NL Can Be Associated with Gene Activation
(A) Distribution of gene expression levels in LADs and inter-LADs (iLADs) based on mRNA-seq of uninduced (+Dox/-Shield1) cells as described in Figure 3C. Squares indicate the expression levels of ve active LAD (red) and two active iLAD (blue) genes. (B) Partial overlap of H3K4me3 with internal LADs. Dam-Lamin B1 was expressed in mother cells; daughter cells were xed 5 hr after mitosis and stained for H3K4me3 (red). Right-hand panels show higher magnication of the left panels. Arrow marks a LAD region that overlaps with H3K4me3 signal. (C) Quantitative analysis of mean H3K4me3 pixel value distributions in daughter cells as in (B) (n = 25) for regions covered by m6A-Tracer signals (LADs) or not (No LAD). Light gray: less than or equal to 1.4 mm from the NL; dark gray: more than 1.4 mm from the NL. p value is according to a paired Wilcoxon test. (D) Stochastic NL interactions of active LAD genes are inversely linked to their H3K36me3 status. Chromatin from Dam-Lamin B1-expressing mother cells was subjected to ChIP with antibodies directed against Histone H3 (gray) or H3K36me3 (black), followed by quantication of m6A levels (representing NL contact frequencies) in promoter (P), exon (E), and 30 UTR (U) regions of ve active genes located in LADs (red) and two active genes in iLADs (blue) (Table S3). Note the generally lower NL contact frequencies of LAD gene copies that carry H3K36me3 as compared to bulk (H3 control ChIP) (n = 3). Data are represented as mean SD. *p < 0.05, **p < 0.01 according to a paired t test. See also Tables S2 and S3.
Density
0.2
0.4
0.6
0.8
1.0
m6 A-Tracer Dapi 0.0
H3K4me3
Gene Expression log10 (count + 1)
C
Nucleoplasm Periphery P < 1e-7 P < 1e-7 Mean H3K4me3 120
m6 A-Tracer H3K4me3 Dapi
merge
40
80
LAD
LAD
No LAD
0.6
Active LAD genes
Active iLAD genes Histone H3 H3K36me3
0.5
Fraction m6A
0.4
0.3
**
cells (Tumbar and Belmont, 2001) and the reported erasure of H3K9me3 at a Lacotagged locus by VP16 targeting (Hathaway * ** 0.1 * * et al., 2012). Hundreds of LAD regions ** were previously found to relocate relative 0 to the NL upon cellular differentiation; E U E U E U P E E U E U P E U ALCAM PLOD2 LDHB STAG2 PEX13 UBE2B SERINC1 in many instances this occurred in the absence of changes in gene expression (Peric-Hupkes et al., 2010). It is possible were not able to detect direct molecular contacts with the NL. that the underlying mechanism is similar to the one responsible Our electron microscopy data indicate that the mobility of for the repositioning effect that we observed here. peripheral LADs is largely conned to the layer of densely staining heterochromatin. Thus, although the mobility of chromatin Reshufing of LADs after Mitosis within this layer is high, it rarely mixes with the neighboring After mitosis, LADs are stochastically reshufed, indicating that euchromatin. positioning of genomic loci relative to the NL is generally not heritable. Mother-to-daughter cell transmission of radial positioning Activator-Induced Repositioning of LADs is also not supported by a study of two LacO-tagged loci (ThomWe demonstrated that targeting of the VP16 AAD can drive the son et al., 2004) or by anecdotal observations based on the escape of a subset of LADs from the nuclear periphery, most labeling of peripheral chromatin by photoactivation (Strickfaden likely through alterations in the chromatin state rather than through et al., 2010). activation of gene expression. This is consistent with the AADWe observed that some of the mother LADs that end up in the driven relocation of a single Laco-tagged locus in chinese hamster nuclear interior after mitosis become closely associated with
0.2
nucleoli. Genome-wide mapping studies have found that nucleolus-associated domains (NADs) overlap substantially with LADs meth et al., 2010; van Koningsbruggen et al., 2010). It is (Ne possible that each LAD/NAD has a distinct distribution probability between nucleoli and the NL. NL Interactions of LADs Are Linked to Stochastic Histone Modications Our results indicate that the stochastic NL contacts of individual LAD copies are closely linked to their H3K9me2 state. Furthermore, for ve active genes embedded in LADs we found that the H3K36me3 state (and thus presumably the transcription state [Bannister et al., 2005]) is inversely linked to stochastic NL contacts. These results imply that the histone modication state of individual LADs is also stochastic, i.e., it must exhibit considerable cell-to-cell variation or allele-to-allele variation in single cells or both. No technique is presently available to directly detect histone modications on individual loci in single cells. However, di- and trimethylated H3K9 have previously been linked to the variegated expression of reporter genes embedded in heterochromatin, which could point to stochastic variation of these histone marks themselves (Fodor et al., 2010). Stochastic expression of genes is well documented but the underlying mechanisms are still poorly understood (Lionnet and Singer, 2012; Raj and van Oudenaarden, 2009). Contact of a LAD with the NL may help to repress the genes within this LAD, which is consistent with studies showing that articial anchoring of genes to the NL can lead to reduced transcriptional activity (Dialynas et al., 2010; Finlan et al., 2008; Reddy et al., 2008). The mechanism of NL-mediated repression still remains to be elucidated. G9a Promotes Nl Contacts Our study indicates a role of G9a in promoting LAD-NL contacts, presumably via H3K9me2. Importantly, a recent study in C. elegans also found evidence for a role of H3K9 methylation in the targeting of genomic regions to the NL (Towbin et al., 2012). In contrast, a FISH study in mouse embryonic stem cells found no signicant change in the peripheral localization of a number of late-replicating G9a target genes after knockout of G9a (Yokochi et al., 2009). This difference may be explained by speciesor cell-type specic effects, or by technological differences, because our approach detects molecular contacts whereas DNA-FISH is limited by the resolution of light microscopy. We cannot rule out the possibility that positioning of LADs at the NL also enforces H3K9me2, i.e., that the causal relationship is circular. If this is the case, it would establish a positive feedback loop that would enforce the H3K9me2-positive state and keep peripherally located LADs at the NL during interphase (Towbin et al., 2012). Although G9a was recently found to interact with the NL-interacting protein BAF (Montes de Oca et al., 2011), we are not aware of direct evidence that G9a or other H3K9 methyltransferases are preferentially active at the NL. Together, the results reported here provide a foundation for further studies of the relationships between dynamic chromosome organization and transcription regulation in single cells.
EXPERIMENTAL PROCEDURES Detailed methods are described in the Extended Experimental Procedures. Cell Lines Stable clonal cell lines were derived from HTC75, which is an HT1080 line that stably expresses the Tet-off transactivator (van Steensel and de Lange, 1997). The Shield1-inducible Dam-Lamin B1 clonal line clone11 was derived by transfecting HTC75 cells with pTuner-DD-Dam-Lamin B1; the m6A-Tracer line clone3 was derived from clone11 by transfection with an m6A-Tracer expression vector under a Tet-off responsive promoter. Stable clonal cell lines expressing the m6A-Tracer-VP16 fusion constructs were created as described for the clone3 line.
m6 A-Tracer Microscopy and FISH Images of xed cells were taken on Leica TCS SP2 or SP5 microscopes. For time-lapse imaging, clone3 cells were transfected with H2B-mRFP and imaged on a Leica TCS SP5 confocal laser- scanning microscope with a 37 C climate chamber. 3D DNA-FISH was performed as described (Boyle et al., 2011), using a custom-made Texas Red-labeled probe pool consisting of 140,000 oligonucleotides of an average TM of 76 C that tile each of the 25 LADs. Genomic positions of the probed LADs are listed in Table S1. To visualize a single LADs in both daughter cells we generated a probe with a fosmid (W12-1931L04) that corresponds to the region chr3:136427835-136466953 (GRCh37/hg19).
ChIP-m6A Assay Clone11 cells were arrested for 4 hr and collected by mitotic shake-off. Then Dam-Lamin B1 expression was induced by addition of Shield1 for 18 hr. Cells were xed with formaldehyde, chromatin was prepared and subjected to ChIP using antibodies as indicated, precipitated DNA was puried and treated with or without DpnII, after which qPCR across individual GATC sites was done with primers listed in Table S3. ACCESSION NUMBERS Genome-wide DamID and RNA-seq data are available from the Gene Expression Omnibus (http://www.ncbi.nlm.nih.gov/geo/), accession number GSE40112. SUPPLEMENTAL INFORMATION Supplemental Information includes four gures, three tables, and Extended Experimental Procedures and can be found with this article online at http:// dx.doi.org/10.1016/j.cell.2013.02.028. ACKNOWLEDGMENTS We thank Wouter Meuleman for help with FISH probe pool design; Federica Morettini for the scoring of FISH images; the NKI Genomics Core Facility for microarray hybridizations and RNA-seq; and the Wolthuis and Neefjes laboratories for reagents. This work was supported by EMBO LTF and NWO-ALW VENI (J.K.), NWO-ALW VICI and ERC Advanced grant 293662 (B.v.S.), and MRC and ERC Advanced grant 249956 (W.A.B.). Received: August 30, 2012 Revised: December 17, 2012 Accepted: February 5, 2013 Published: March 21, 2013 REFERENCES Abney, J.R., Cutler, B., Fillbach, M.L., Axelrod, D., and Scalettar, B.A. (1997). Chromatin dynamics in interphase nuclei and its implications for nuclear structure. J. Cell Biol. 137, 14591468.
Banaszynski, L.A., Chen, L.C., Maynard-Smith, L.A., Ooi, A.G., and Wandless, T.J. (2006). A rapid, reversible, and tunable method to regulate protein function in living cells using synthetic small molecules. Cell 126, 9951004. Bannister, A.J., Schneider, R., Myers, F.A., Thorne, A.W., Crane-Robinson, C., and Kouzarides, T. (2005). Spatial distribution of di- and tri-methyl lysine 36 of histone H3 at active genes. J. Biol. Chem. 280, 1773217736. Boyle, S., Rodesch, M.J., Halvensleben, H.A., Jeddeloh, J.A., and Bickmore, W.A. (2011). Fluorescence in situ hybridization with high-complexity repeatfree oligonucleotide probes generated by massively parallel synthesis. Chromosome Res. 19, 901909. Carpenter, A.E., Memedula, S., Plutz, M.J., and Belmont, A.S. (2005). Common effects of acidic activators on large-scale chromatin structure and transcription. Mol. Cell. Biol. 25, 958968. Chuang, C.H., Carpenter, A.E., Fuchsova, B., Johnson, T., de Lanerolle, P., and Belmont, A.S. (2006). Long-range directional movement of an interphase chromosome site. Curr. Biol. 16, 825831. Chubb, J.R., Boyle, S., Perry, P., and Bickmore, W.A. (2002). Chromatin motion is constrained by association with nuclear compartments in human cells. Curr. Biol. 12, 439445. Cook, P.R. (2010). A model for all genomes: the role of transcription factories. J. Mol. Biol. 395, 110. Dialynas, G., Speese, S., Budnik, V., Geyer, P.K., and Wallrath, L.L. (2010). The role of Drosophila Lamin C in muscle function and gene expression. Development 137, 30673077. Dimitrova, D.S., and Gilbert, D.M. (1999). The spatial position and replication timing of chromosomal domains are both established in early G1 phase. Mol. Cell 4, 983993. Filion, G.J., van Bemmel, J.G., Braunschweig, U., Talhout, W., Kind, J., Ward, L.D., Brugman, W., de Castro, I.J., Kerkhoven, R.M., Bussemaker, H.J., and van Steensel, B. (2010). Systematic protein location mapping reveals ve principal chromatin types in Drosophila cells. Cell 143, 212224. Finlan, L.E., Sproul, D., Thomson, I., Boyle, S., Kerr, E., Perry, P., Ylstra, B., Chubb, J.R., and Bickmore, W.A. (2008). Recruitment to the nuclear periphery can alter expression of genes in human cells. PLoS Genet. 4, e1000039. Fodor, B.D., Shukeir, N., Reuter, G., and Jenuwein, T. (2010). Mammalian Su(var) genes in chromatin control. Annu. Rev. Cell Dev. Biol. 26, 471501. Gerlich, D., Beaudouin, J., Kalbfuss, B., Daigle, N., Eils, R., and Ellenberg, J. (2003). Global chromosome positions are transmitted through mitosis in mammalian cells. Cell 112, 751764. Gossen, M., and Bujard, H. (1992). Tight control of gene expression in mammalian cells by tetracycline-responsive promoters. Proc. Natl. Acad. Sci. USA 89, 55475551. Guelen, L., Pagie, L., Brasset, E., Meuleman, W., Faza, M.B., Talhout, W., Eussen, B.H., de Klein, A., Wessels, L., de Laat, W., and van Steensel, B. (2008). Domain organization of human chromosomes revealed by mapping of nuclear lamina interactions. Nature 453, 948951. Hathaway, N.A., Bell, O., Hodges, C., Miller, E.L., Neel, D.S., and Crabtree, G.R. (2012). Dynamics and memory of heterochromatin in living cells. Cell 149, 14471460. Hermann, A., and Jeltsch, A. (2003). Methylation sensitivity of restriction enzymes interacting with GATC sites. Biotechniques 34, 924926, 928, 930. Kind, J., and van Steensel, B. (2010). Genome-nuclear lamina interactions and gene regulation. Curr. Opin. Cell Biol. 22, 320325. Kubicek, S., OSullivan, R.J., August, E.M., Hickey, E.R., Zhang, Q., Teodoro, M.L., Rea, S., Mechtler, K., Kowalski, J.A., Homon, C.A., et al. (2007). Reversal of H3K9me2 by a small-molecule inhibitor for the G9a histone methyltransferase. Mol. Cell 25, 473481. Lionnet, T., and Singer, R.H. (2012). Transcription goes digital. EMBO Rep. 13, 313321. Meuleman, W., Peric-Hupkes, D., Kind, J., Beaudry, J.B., Pagie, L., Kellis, M., Reinders, M., Wessels, L., and van Steensel, B. (2013). Constitutive nuclear
lamina-genome interactions are highly conserved and associated with A/Trich sequence. Genome Res. 23, 270280. Moir, R.D., Yoon, M., Khuon, S., and Goldman, R.D. (2000). Nuclear lamins A and B1: different pathways of assembly during nuclear envelope formation in living cells. J. Cell Biol. 151, 11551168. Montes de Oca, R., Andreassen, P.R., and Wilson, K.L. (2011). Barrier-toAutointegration Factor inuences specic histone modications. Nucleus 2, 580590. meth, A., Conesa, A., Santoyo-Lopez, J., Medina, I., Montaner, D., Pe tera, Ne ngst, G. (2010). Initial genomics of B., Solovei, I., Cremer, T., Dopazo, J., and La the human nucleolus. PLoS Genet. 6, e1000889. Peric-Hupkes, D., Meuleman, W., Pagie, L., Bruggeman, S.W., Solovei, I., f, S., Flicek, P., Kerkhoven, R.M., van Lohuizen, M., et al. Brugman, W., Gra (2010). Molecular maps of the reorganization of genome-nuclear lamina interactions during differentiation. Mol. Cell 38, 603613. fer, C., Peters, A.H., OCarroll, D., Scherthan, H., Mechtler, K., Sauer, S., Scho Weipoltshammer, K., Pagani, M., Lachner, M., Kohlmaier, A., et al. (2001). Loss of the Suv39h histone methyltransferases impairs mammalian heterochromatin and genome stability. Cell 107, 323337. Pickersgill, H., Kalverda, B., de Wit, E., Talhout, W., Fornerod, M., and van Steensel, B. (2006). Characterization of the Drosophila melanogaster genome at the nuclear lamina. Nat. Genet. 38, 10051014. Prokocimer, M., Davidovich, M., Nissim-Rania, M., Wiesel-Motiuk, N., Bar, D.Z., Barkan, R., Meshorer, E., and Gruenbaum, Y. (2009). Nuclear lamins: key regulators of nuclear structure and activities. J. Cell. Mol. Med. 13, 10591085. Quinn, A.M., Allali-Hassani, A., Vedadi, M., and Simeonov, A. (2010). A chemiluminescence-based method for identication of histone lysine methyltransferase inhibitors. Mol. Biosyst. 6, 782788. Raj, A., and van Oudenaarden, A. (2009). Single-molecule approaches to stochastic gene expression. Annu Rev Biophys 38, 255270. Reddy, K.L., Zullo, J.M., Bertolino, E., and Singh, H. (2008). Transcriptional repression mediated by repositioning of genes to the nuclear lamina. Nature 452, 243247. Shevelyov, Y.Y., and Nurminsky, D.I. (2012). The nuclear lamina as a genesilencing hub. Curr. Issues Mol. Biol. 14, 2738. Shinkai, Y., and Tachibana, M. (2011). H3K9 methyltransferase G9a and the related molecule GLP. Genes Dev. 25, 781788. Siwek, W., Czapinska, H., Bochtler, M., Bujnicki, J.M., and Skowronek, K. (2012). Crystal structure and mechanism of action of the N6-methyladeninedependent type IIM restriction endonuclease R.DpnI. Nucleic Acids Res. 40, 75637572. hler, D., and Strickfaden, H., Zunhammer, A., van Koningsbruggen, S., Ko Cremer, T. (2010). 4D chromatin dynamics in cycling cells: Theodor Boveris hypotheses revisited. Nucleus 1, 284297. Tachibana, M., Ueda, J., Fukuda, M., Takeda, N., Ohta, T., Iwanari, H., Sakihama, T., Kodama, T., Hamakubo, T., and Shinkai, Y. (2005). Histone methyltransferases G9a and GLP form heteromeric complexes and are both crucial for methylation of euchromatin at H3-K9. Genes Dev. 19, 815826. Thomson, I., Gilchrist, S., Bickmore, W.A., and Chubb, J.R. (2004). The radial positioning of chromatin is not inherited through mitosis but is established de novo in early G1. Curr. Biol. 14, 166172. lez-Aguilera, C., Sack, R., Gaidatzis, D., Kalck, V., Towbin, B.D., Gonza Meister, P., Askjaer, P., and Gasser, S.M. (2012). Step-wise methylation of histone H3K9 positions heterochromatin at the nuclear periphery. Cell 150, 934947. Tumbar, T., and Belmont, A.S. (2001). Interphase movements of a DNA chromosome region modulated by VP16 transcriptional activator. Nat. Cell Biol. 3, 134139. Tumbar, T., Sudlow, G., and Belmont, A.S. (1999). Large-scale chromatin unfolding and remodeling induced by VP16 acidic activation domain. J. Cell Biol. 145, 13411354.
van Koningsbruggen, S., Gierlinski, M., Schoeld, P., Martin, D., Barton, G.J., Ariyurek, Y., den Dunnen, J.T., and Lamond, A.I. (2010). High-resolution whole-genome sequencing reveals that specic chromatin domains from most human chromosomes associate with nucleoli. Mol. Biol. Cell 21, 3735 3748. van Steensel, B., and de Lange, T. (1997). Control of telomere length by the human telomeric protein TRF1. Nature 385, 740743. van Steensel, B., and Dekker, J. (2010). Genomics tools for unraveling chromosome architecture. Nat. Biotechnol. 28, 10891095. van Steensel, B., and Henikoff, S. (2000). Identication of in vivo DNA targets of chromatin proteins using tethered dam methyltransferase. Nat. Biotechnol. 18, 424428. Vazquez, J., Belmont, A.S., and Sedat, J.W. (2001). Multiple regimes of constrained chromosome motion are regulated in the interphase Drosophila nucleus. Curr. Biol. 11, 12271239. Walter, J., Schermelleh, L., Cremer, M., Tashiro, S., and Cremer, T. (2003). Chromosome order in HeLa cells changes during mitosis and early G1, but
is stably maintained during subsequent interphase stages. J. Cell Biol. 160, 685697. Wen, B., Wu, H., Shinkai, Y., Irizarry, R.A., and Feinberg, A.P. (2009). Large histone H3 lysine 9 dimethylated chromatin blocks distinguish differentiated from embryonic stem cells. Nat. Genet. 41, 246250. Wu, R., Terry, A.V., Singh, P.B., and Gilbert, D.M. (2005). Differential subnuclear localization and replication timing of histone H3 lysine 9 methylation states. Mol. Biol. Cell 16, 28722881. Yokochi, T., Poduch, K., Ryba, T., Lu, J., Hiratani, I., Tachibana, M., Shinkai, Y., and Gilbert, D.M. (2009). G9a selectively represses a class of late-replicating genes at the nuclear periphery. Proc. Natl. Acad. Sci. USA 106, 1936319368. -Regi, R., Gaffney, D.J., Epstein, C.B., SpooZullo, J.M., Demarco, I.A., Pique ner, C.J., Luperchio, T.R., Bernstein, B.E., Pritchard, J.K., Reddy, K.L., and Singh, H. (2012). DNA sequence-dependent compartmentalization and silencing of chromatin at the nuclear lamina. Cell 149, 14741487.
The Arabidopsis Nucleosome Remodeler DDM1 Allows DNA Methyltransferases to Access H1-Containing Heterochromatin
Assaf Zemach,1,2 M. Yvonne Kim,1,2 Ping-Hung Hsieh,1,2 Devin Coleman-Derr,1 Leor Eshed-Williams,1,3 Ka Thao,1 Stacey L. Harmer,1,4 and Daniel Zilberman1,*
of Plant and Microbial Biology, University of California, Berkeley, Berkeley, CA 94720, USA authors contributed equally to this work 3Present address: Faculty of Agriculture, Food and Environment, Hebrew University of Jerusalem, Rehovot 76100, Israel 4Present address: Department of Plant Biology, University of California, Davis, Davis, CA 95616, USA *Correspondence: danielz@berkeley.edu http://dx.doi.org/10.1016/j.cell.2013.02.033
2These 1Department
SUMMARY
Nucleosome remodelers of the DDM1/Lsh family are required for DNA methylation of transposable elements, but the reason for this is unknown. How DDM1 interacts with other methylation pathways, such as small-RNA-directed DNA methylation (RdDM), which is thought to mediate plant asymmetric methylation through DRM enzymes, is also unclear. Here, we show that most asymmetric methylation is facilitated by DDM1 and mediated by the methyltransferase CMT2 separately from RdDM. We nd that heterochromatic sequences preferentially require DDM1 for DNA methylation and that this preference depends on linker histone H1. RdDM is instead inhibited by heterochromatin and absolutely requires the nucleosome remodeler DRD1. Together, DDM1 and RdDM mediate nearly all transposon methylation and collaborate to repress transposition and regulate the methylation and expression of genes. Our results indicate that DDM1 provides DNA methyltransferases access to H1-containing heterochromatin to allow stable silencing of transposable elements in cooperation with the RdDM pathway.
INTRODUCTION DNA methylation in owering plants occurs in three sequence contexts: CG, CHG, and CHH (asymmetric), where H is any nucleotide except G. Methylation in each context is believed to be primarily catalyzed by a specic family of DNA methyltransferases: MET1 (homologous to animal Dnmt1) for CG, chromomethylases (CMT) for CHG, and DRM (homologous to animal Dnmt3) for CHH (Law and Jacobsen, 2010). The majority of plant methylation is found in transposable elements (TEs), where methylation occurs in all sequence contexts and is crucial for
the repression of TE expression and transposition (Law and Jacobsen, 2010). Substantial methylation is also found in the bodies of active genes, where methylation is generally restricted to the CG context (Law and Jacobsen, 2010; Zemach et al., 2010b). The establishment of plant DNA methylation in all sequence contexts and the maintenance of CHH methylation are mediated by a specialized branch of the RNA interference pathway referred to as RNA-directed DNA methylation (RdDM) (Law and Jacobsen, 2010). RdDM relies on two plant-specic homologs of RNA polymerase II: Pol IV and Pol V. The Pol IV branch of RdDM is thought to synthesize the long RNA molecules that are made double stranded by RNA-dependent RNA polymerase 2 (RDR2) and processed by Dicer-like nucleases into small interfering RNA (sRNA). RNA Pol V and associated factors are believed to produce nascent transcripts from target loci that are recognized by sRNA-containing AGO4 complexes that target DNA methylation via DRM enzymes (Haag and Pikaard, 2011). DNA methylation is also inuenced by chromatin factors: CHG methylation by CMT3 requires dimethylation of histone H3 at lysine 9 (H3K9me2), to which CMT3 binds via its chromo- and bromo-adjacent homology domains (Du et al., 2012; Law and Jacobsen, 2010), and CHG methylation is kept out of genes by a histone demethylase, IBM1, which removes H3K9me2 from gene bodies (Saze and Kakutani, 2011). A more enigmatic chromatin factor that is essential for normal DNA methylation is the Snf2 family nucleosome remodeler DDM1 (Jeddeloh et al., 1999; Lippman et al., 2004). Snf2 remodelers hydrolyze ATP to move along DNA, altering nucleosome composition and placement and allowing other proteins to access the DNA (Ryan and Owen-Hughes, 2011). DDM1 can shift nucleosomes in vitro (Brzeski and Jerzmanowski, 2003), and its mutation has been reported to cause a profound loss of methylation from some TEs and repeats (Jeddeloh et al., 1999), but not from genes (Lippman et al., 2004). Mutation of Lsh, the mouse homolog of DDM1, causes a similar methylation phenotype (Tao et al., 2011), indicating that DDM1 remodelers are ancient components of the DNA methylation pathway. The loss of DDM1 leads to strong transcriptional activation of TEs (Lippman et al., 2004), and inbred ddm1 mutant lines have
increased rates of transposition (Tsukahara et al., 2009). sRNAs correspond to the TEs hypomethylated in ddm1 mutant plants, leading to the suggestion that DDM1 participates in RdDM (Lippman et al., 2004). However, DDM1 and RdDM synergize to silence rDNA loci (Blevins et al., 2009), indicating that RdDM can function without DDM1, and DDM1 can mediate CHH methylation independently of RdDM at some TEs (Teixeira et al., 2009). Thus, the related questions of how DDM1 interacts with the methyltransferase pathways, including RdDM, and why some sequences require DDM1 for methylation and others do not, remain largely unanswered. Here, we report genome-wide analyses of DNA methylation, RNA expression, and TE movement in plants lacking DDM1 and RdDM. We nd that DDM1 facilitates methylation in all sequence contexts separately from RdDM by countering the inuence of linker histone H1. We also nd that DDM1-mediated, RdDM-independent CHH methylation is catalyzed by the chromomethylase CMT2. DDM1 and RdDM synergistically mediate nearly all Arabidopsis TE methylation, prevent transposition, and maintain proper patterns of gene expression. DDM1 is important for DNA methylation in genes as well as TEs, and the strength of the DDM1 requirement is positively correlated with heterochromatin, which we nd inhibits RdDM. Our data suggest that DDM1 is specialized for remodeling heterochromatic, H1bound nucleosomes to allow DNA methyltransferases, and likely other proteins, access to the DNA. While the genome can be naturally subdivided into genes and TEs, our results suggest that all genomic sequences are part of a chromatin continuum that cuts across TE and gene annotations and better explains how DNA binding and modifying proteins reach their target sites. RESULTS DDM1 and RdDM Separately Mediate Nearly All DNA Methylation in TEs To understand how DDM1 mediates DNA methylation, we quantied genomic methylation of ddm1 mutant Arabidopsis thaliana plants (Table S1 available online). Lack of DDM1 caused a 58%, 57%, and 32% overall reduction of CG, CHG, and CHH methylation, respectively (Figure S1A), reected in much lower TE methylation in all sequence contexts (Figure 1A). The strong loss of CHH methylation may be caused by a dependence of RdDM on methylation in other contexts, as proposed to explain the loss of non-CG methylation in met1 mutant plants (Figure S1A) (Lister et al., 2008), or it may indicate that a large fraction of plant CHH methylation is mediated by DDM1 independently of RdDM (Teixeira et al., 2009). To answer this question, we determined DNA methylation in plants with a mutation in RdDM pathway gene RDR2, which is required for the production of all endogenous sRNA molecules (Xie et al., 2004). We also analyzed methylation in plants lacking DRD1, a Snf2 remodeler that positively regulates RdDM at a number of loci (Kanno et al., 2004; Law et al., 2010) and forms a complex with RdDM pathway proteins that is thought to facilitate Pol V activity (Law et al., 2010; Zhong et al., 2012). Lack of RDR2 and DRD1 caused a relatively modest loss of CHH methylation (Figures 1A and S1A), demonstrating that the majority of CHH methylation does not require RdDM (Wierzbicki et al., 2012). Combining the ddm1
mutation with either rdr2 or drd1 caused a nearly complete loss of CHH methylation (Figures 1A and S1A), as well as of CG and CHG methylation in TEs (Figure 1A), demonstrating that a great deal of Arabidopsis CHH methylation is mediated by DDM1 separately from RdDM and that the two pathways together are responsible for almost all DNA methylation of TEs. The methylation phenotypes of drd1 and ddm1drd1 mutants are virtually indistinguishable from those of rdr2 and ddm1rdr2, respectively (Figure 1A), indicating that DRD1 is absolutely required for RdDM. DDM1-Dependent CHH Methylation Is Catalyzed by CMT2 The presence of extensive RdDM-independent CHH methylation raises the question of which DNA methyltransferase is responsible. Previous studies have demonstrated that MET1 and CMT3 mediate virtually all Arabidopsis CG and CHG methylation, respectively (Cokus et al., 2008; Lister et al., 2008), and our data conrm these results (Figure 1B and S1A). However, even plants lacking DRM1 (which appears to be specically active during early seed development) (Jullien et al., 2012), DRM2, and CMT3 have substantial residual CHH methylation (Cokus et al., 2008; Lister et al., 2008), indicating that another methyltransferase must be involved. Mutation of DRM2 causes CHH methylation loss that closely resembles that in RdDM mutants (Figures 1A1C), consistent with the established link between DRM2 and RdDM (Law and Jacobsen, 2010). Unexpectedly, lack of CMT2, a homolog of CMT3 (Figure 1D), has little impact on CHG methylation but causes a major loss of CHH methylation (Figures 1B, S1A, and S1B). The observation that DRM2 accounts for RdDM (Figures 1A1C) indicates that CMT2 is responsible for the DDM1-mediated, sRNA-independent CHH methylation (Figure 1A). In support of this conclusion, residual methylation in cmt2 plants is correlated with that in ddm1 and anticorrelated with that in RdDM mutants (Table S2), whereas residual methylation in drm2 plants (mediated by CMT2) is uncorrelated with sRNA abundance (Table S2). CMT2 appears to have evolved prior to the radiation of angiosperms (Figure 1D) to methylate a different sequence context than canonical chromomethylases (Cokus et al., 2008; Du et al., 2012; Zemach et al., 2010b). DDM1 and RdDM Mediate Methylation of Distinct TE Sizes and Domains The ddm1 and drd1 mutant lines exhibit a similar absolute level but very different patterns of CHH methylation loss (Figures 1A and S1A). The drd1 mutation strongly reduces TE CHH methylation near the points of alignment, whereas ddm1 has a larger effect away from the points of alignment (Figure 1A). The same distinction is evident for CG and CHG methylation (Figure 1A) and for the CHH methylation phenotypes of drm2 and cmt2 (Figure 1B). Lack of DDM1 and RdDM thus affects TEs very differently. More than 80% of annotated Arabidopsis TEs are shorter than 1,000 base pairs (bp) (Buisine et al., 2008), and such TEs would only contribute to the patterns of TE methylation shown in Figures 1A and 1B close to the points of alignment, suggesting that the differences between ddm1 and drd1, as well as between
A
0.8
CG methylation
drd1
WT
B
0.9
CG methylation
WT
E
1
CG methylation
WT drd1 ddm1
drm2
cmt2
F
0.8
CG methylation
WT drd1 ddm1
rdr2
ddm1
met1
ddm1drd1
0
0 1 2 3 4 5 6kb
ddm1rdr2 ddm1drd1 0 -4 -3 -2 -1 0 1 2 3 4 3 2 1 0 -1 -2 -3 -4 kb
0 -4 -3 -2 -1 0 1 2 3 4 3 2 1 0 -1 -2 -3 -4 kb
0 -4
ddm1drd1
-3 -2 -1 0 1 2 3 4 3 2 1 0 -1 -2 -3 -4 kb
0.5
CHG methylation
WT drd1 rdr2 ddm1
0.6
CHG methylation
WT
0.6
CHG methylation
WT cmt2 drm2 drd1 ddm1drd1

0 1 2 3 4 5 6kb
0.5
CHG methylation
WT
cmt2 cmt3
drd1 ddm1
ddm1
ddm1drd1 ddm1rdr2 0 -4 -3 -2 -1 0 1 2 3 4 3 2 1 0 -1 -2 -3 -4 kb
0 -4 -3 -2 -1 0 1 2 3 4 3 2 1 0 -1 -2 -3 -4 kb
ddm1drd1
-4 -3 -2 -1 0 1 2 3 4 3 2 1 0 -1 -2 -3 -4 kb
0.14
CHH methylation
WT drd1 ddm1 rdr2 ddm1drd1 ddm1rdr2
0.14
CHH methylation
WT drm2
0.16
CHH methylation
WT ddm1 cmt2 drd1 drm2
0.3
CHH methylation
WT
cmt2
ddm1 drd1 ddm1drd1
0 -4 -3 -2 -1 0 1 2 3 4 3 2 1 0 -1 -2 -3 -4 kb
0 -4 -3 -2 -1 0 1 2 3 4 3 2 1 0 -1 -2 -3 -4 kb
ddm1drd1
0 1 2 3 TE size 4 5 6kb
0 -4 -3 -2 -1 0 1 2 3 4 3 2 1 0 -1 -2 -3 -4 kb
C
wild-type drd1
1 0 1 0 1
Short TEs
Long TE
CHH methylation
D
PO
Aquca_13 3_00030_ Acoerulea Aquca _078_0 0023_A coerul ea
CMT1/3
PT R_
g1 25 XP 70 .1_ Si _00 Os 03 35 ata 38 71 iva 90 678 .1_ m _S Bd ita ist ac lic hy a on
drm2 0
1
cle
LO
cmt2 0
1
C_
Bra
XP_
_0
ddm1drd1 0 Genes TEs

40 0 0.4 -0.4 250 0 2.7 -2.7 8 0 8 0 8
024
02
Os
e0
.9
03
ddm1 0
en
tin
0028
CMT3_Athaliana
340 _B rap 8734 a .1_A lyrata
12
4m
_C
1 7. 16 05 olor 11 Sbic 00 0.1_ 3 _ 4 8 NP 06g02 sb
m _Z
et
2_
Zm
ay
XP_
XP_00351
sRNA GC content nucleosomes H3K9me2 RNA expression
8 CM 8927 T1 6.1 _A _A th lyr ali at an a a
6.1
5_0
_10
0035
drd1
4871
000
8.1_
wild-type
Gma x ra XP_002267685.2_Vvinife unis m m a co 77_R tin m0026 en 29827. lem Cc m_ 2 2 35 00 .9_ e0 tin n me cle
communis 000332_R 28582.m era vinif .1_V 2 3 9 na 75 enti 0022 XP_ lem ax c C Gm m_ 204 34.1_ 002 65 .9_ 351 ne0 _00 enti a P m p X a Cle Br 7_ 85 19 a0 r B
_A
coe
rule
00 01 s0 47 00 .1
002 283 355 .2_V vinif

ch oc ar
m en tin a cle
7640.1_Gm
_P tri
XP
78
_00
297
.1 335
_Sm
oell
end
orff
ii
era pa
ax
10
a alic _Sit 01m 210 Si0 1_Sbicolor Sb09g007390. LOC XP_003561758.1_Bdistachyon _Os 05g 137 80.1 _Os ativ a
Br X MT2 a0 P_0 _ 13 02 Atha lian 37 870 0_ 0 a Br 05.1 _A ap lyr a ata
ddm1drd1
Chr1 5890741-5920460
Figure 1. DDM1 and CMT2 Mediate RdDM-Independent CHH Methylation of Long TEs
(A) Patterns of TE DNA methylation (CG, CHG, and CHH) in wild-type and indicated mutants. Arabidopsis TEs were aligned at the 50 end or the 30 end, and average methylation for all cytosines within each 100 bp interval is plotted. The dashed lines represent the points of alignment. (B) Patterns of TE methylation in met1, cmt3, cmt2, and drm2 plants. (C) CHH methylation, sRNA, GC content, nucleosomes, H3K9me2, and RNA levels of a representative region. Genes and TEs oriented 50 to 30 and 30 to 50 are shown above and below the line, respectively. (D) Phylogenetic tree of angiosperm chromomethylases, with Selaginella moellendorfi (black) as an outgroup. Dicots are shown in green and monocots in red. (E) Locally weighted scatterplot smoothing (LOWESS) t of CG, CHG, and CHH methylation levels in wild-type and indicated mutants calculated in 50 bp windows and plotted against TE size. (F) DNA methylation in wild-type and indicated mutants was averaged specically in long TEs (>4 kb), as described in (A). See also Figure S1 and Table S1.
cmt2 and drm2, may be caused by differential hypomethylation of short and long TEs. Indeed, the ddm1 and cmt2 hypomethylation effects are positively correlated with TE size, whereas drd1 and drm2 hypomethylation is negatively correlated with
TE size (Figures 1C and 1E). Pericentric heterochromatin and chromosome arms are enriched for long and short TEs, respectively (Figure S1C) (Ahmed et al., 2011); consequently, DDM1 preferentially mediates DNA methylation near the centromeres,
XP
Aq
_0
02
XP_
uca
ddm1 0
CMT2
whereas RdDM predominantly functions along the chromosome arms (Figures S1D and S1E). The methylation patterns in Figure 1A are also consistent with DDM1 and DRD1 mediating methylation differently at the edges and inside the bodies of TEs. To examine this issue, we averaged methylation across TEs longer than 4 kb so that short TEs would not inuence the methylation level near the points of alignment. The hypomethylation induced by drd1 is strongest at TE edges, whereas ddm1 hypomethylation is greater within TE bodies (Figures 1C and 1F). The little remaining non-CG methylation in the ddm1drd1 double mutant is evenly distributed across the entire TE sequence (Figure 1F). Similarly, cmt2 nearly eliminates CHH methylation of TEs longer than 4 kb except at the edges, where CHH methylation is mediated by DRM2 (Figures 1C and S1F). Taken together, our results indicate that DDM1 is preferentially required for DNA methylation within the bodies of long TEs, which is catalyzed by MET1 (CG), CMT3 (CHG), and CMT2 (CHH), whereas RdDM mostly targets short TEs and TE edges through DRM2 (Figures 1C, 1E, 1F, and S1F), which is consistent with the observed enrichment of Pol V at such sequences (Zhong et al., 2012). Heterochromatin Requires DDM1 for DNA Methylation and Inhibits RdDM Because DDM1 can remodel nucleosomes, we asked whether chromatin features are responsible for the differential requirement of DDM1 for DNA methylation. Sequence composition is thought to be a major determinant of the nucleosome landscape (Iyer, 2012); indeed, ddm1 TE demethylation in all sequence contexts is strongly correlated with nucleosome occupancy and GC content (Figure 2A). Short TEs and TE edges are relatively AT rich and nucleosome depleted (Figures 2B, 2C, and S2A), consistent with the preferential requirement of DDM1 to maintain DNA methylation in the bodies of long TEs (Figures 1E, 1F, and 2A). However, nucleosome occupancy alone is unlikely to determine the dependence of DNA methylation on DDM1 because genic GC content is similar to that of long TEs and nucleosome occupancy in genes is higher than in most TEs (Figure 2B), presumably because genes and long TEs are composed largely of protein-coding sequences. The properties of nucleosomes in euchromatic genes do differ from those of heterochromatic TEs, as exemplied by different sets of posttranslational histone modications (Roudier et al., 2011). Somewhat unexpectedly, we nd that H3K9me2, the best studied histone modication associated with plant heterochromatin, is enriched in the bodies of long TEs compared to short TEs (Figures 2B and 2C). More generally, DDM1mediated TE DNA methylationthe residual methylation in the drd1 mutantis correlated with heterochromatic histone modications, anticorrelated with euchromatic ones, and not correlated with sRNA abundance (Figures 2A and S2B). Furthermore, the dependence of short TE methylation on DDM1 also correlates with GC content, nucleosome occupancy, and histone modications (Figure 2D), just like methylation of all TEs (Figure 2A), which indicates that the chromatin environment, rather than TE size, ultimately determines the extent to which DDM1 is required for maintenance of DNA methylation.
RdDM would be expected to occur at sRNA-associated loci. Indeed, DRD1-mediated methylationthe residual methylation in the ddm1 mutantis positively correlated with sRNA abundance (Figure S2C), and TE edges, which are hypermethylated at CHH sites in comparison to internal sequences in wild-type plants (Figure 1F), are preferentially targeted by sRNA (Figure 2C) (Lee et al., 2012). However, sRNA molecules are also abundantly derived from TE bodies (Figure 2C). To understand how chromatin structure affects RdDM, we analyzed DRD1-mediated DNA methylation at sequences with similar levels of sRNA. With sRNA levels held constant, DRD1-mediated methylation is negatively correlated with GC content, nucleosome occupancy, and heterochromatic histone modications (Figures 2E, S2B, and S2D). Thus, RdDM is inhibited by heterochromatin, as has been suggested by (Schoft et al., 2009). Histone H1 Mediates the Dependence of Heterochromatic DNA Methylation on DDM1 Our results indicate that DDM1 is preferentially required for methylation of heterochromatic sequences in all contexts, most likely by allowing methyltransferases access to the DNA. To determine which component of heterochromatin blocks enzyme access, we examined histone H1, which binds to the nucleosome core and the linker DNA that separates nucleosomes (Thomas, 1999); condenses chromatin and inhibits nucleosome mobility and transcription in vitro (Pennings et al., 1994; Robinson and Rhodes, 2006); and is associated with more compact, less accessible, and transcriptionally silent chromatin in vivo (Ascenzi and Gantt, 1999; Barra et al., 2000; Fan et al., 2005; Raghuram et al., 2009). Loss of H1 has been reported to cause disparate changes in genomic DNA methylation. Mice with reduced H1 specically lose DNA methylation at the regulatory regions of several imprinted genes (Fan et al., 2005), whereas loss of H1 leads to extensive hypermethylation in the fungus Ascobolus immersus (Barra et al., 2000) and apparently stochastic methylation gains and losses in Arabidopsis (Wierzbicki and Jerzmanowski, 2005). Plants with mutant alleles in the two canonical Arabidopsis H1 genes (Rea et al., 2012; Wierzbicki and Jerzmanowski, 2005) (h1 plants; Figures S3AS3C) exhibit a complex DNA methylation phenotype. Euchromatic TEs (those with low H3K9me2) lose DNA methylation in h1 (Figure 3A), whereas H3K9me2-rich heterochromatic TEs exhibit a global increase of DNA methylation (Figures 3A and 3B), supporting our hypothesis that H1 impedes access of DNA methyltransferases to heterochromatin. Loss of H1 almost completely suppresses the reduction of TE CHH methylation in ddm1 and greatly ameliorates the reduction of TE CG and CHG methylation (Figures 3B, 3C, S3D, and S3E). Most strikingly, H3K9me2-rich heterochromatic TEs are not preferentially hypomethylated in h1ddm1 plants, as they are in ddm1 (Figures 3C and S3D). Instead, h1ddm1 causes heterochromatic TEs to lose less DNA methylation than more euchromatic TEs (black traces in Figure 3C), similar to h1 (green traces in Figure 3C), indicating that the loss of DDM1 affects euchromatic and heterochromatic TEs similarly when H1 is not present (Figure 3D). Lack of H1 still destabilizes the methylation of euchromatic TEs when combined with ddm1, but heterochromatic TEs are methylated rather efciently when both DDM1 and H1
A
0.6
DDM1-mediated DNA methylation in TEs

Sequence context Heterochromatin Nucleosome occupancy
B
GC content
.6
TEs
.4
Genes
Spearman correlation coefficient
0.4
0.2
Euchromatin
0 -0.2
Nucleosomes H3K9me2
CG methylation CHG methylation CHH methylation
-0.4
ge
es
e1
e3
e2
e2
e3
si ze
e3
en
ub
ed
9m
4m
4m
nt
27
2B
so
TE
co
27
36
y/
sR
3K
3K
od
eo
3K
3K
3K
-B
cl
TE
nu
3K
7/0.46 sRNA/GC sRNA nucleosomes GC H3K9me2
90/4.2 nuc / H3K9me2
-2
20
-0.6
40
60
80
.2
2 11 50. .5 <0
24
>4
2 11 50. .5 <0
24
>4
0/0.31 -4 -3 -2 -1
4 kb
10/-1 -1 -2 -3 -4
TE size (kb)
Gene size (kb)
D
0.6
DDM1-mediated DNA methylation in short TEs

Heterochromatin Nucleosomes
E
0.3
DRD1-mediated CHH methylation

Euchromatin
0.4
0.2 0.1 0 -0.1 -0.2 -0.3 -0.4 -0.5
Nucleosomes
Heterochromatin
0.2
Euchromatin
0 -0.2
CG methylation
-0.4
sRNA=2 sRNA=6 sRNA=10
CHG methylation CHH methylation
-0.6
Figure 2. Heterochromatin Requires DDM1 for DNA Methylation and Inhibits RdDM
(A) Spearman correlation coefcients between DDM1-mediated methylation (drd1 methylation minus ddm1drd1 methylation) and DNA sequence and chromatin features of TEs in 50 bp windows. (B) Box plots showing GC content, nucleosome occupancy, and H3K9me2 levels in 50 bp windows within TEs and genes of the indicated size. (C) sRNA, GC content, nucleosome occupancy and H3K9me2 levels were averaged in long TEs (>4 kb) as described in Figure 1. (D) Spearman correlation coefcients between DDM1-mediated methylation in short TEs (<500 bp) and chromatin features. (E) Spearman correlation coefcients between DRD1-mediated CHH methylation (ddm1 methylation minus ddm1drd1 methylation) and chromatin features, calculated for 50 bp windows with three different levels of sRNA. See also Figure S2 and Table S2.
are absent (Figures 3C, S3D, and S3E). Our results indicate that the differential importance of DDM1 for the maintenance of DNA methylation in heterochromatic versus euchromatic TEs is governed by H1. Methylation of TE Families Depends on sRNA Abundance and Chromatin Features The Arabidopsis genome contains a variety of TE families that have different mechanisms of transposition, internal structure, and chromosomal localization (Ahmed et al., 2011; Buisine
et al., 2008). We chose four such families to examine DNA methylation mediated by DDM1 and RdDM in more detail: Gypsy, Copia, MuDR, and LINE elements. Gypsy elements are long terminal repeat (LTR)-anked retrotransposons that are concentrated in pericentric heterochromatin (Figure S1C). Copia LTR elements are more evenly dispersed, as are LINE non-LTR retrotransposons and the terminal inverted-repeat-anked MuDR DNA transposons (Figure S1C). The four TE families have distinct sRNA distributions: Gypsy elements have high levels of sRNA across the entire sequence;
3K
Nu
Nu
3K
co
2B
cl
K H3
K H3
K H3
cl
K H3
K H3
K H3
36
2B
36
nt
ub
ub
s eo
2 3K
2 3K
co
s eo
2 3K
2 3K
en
e3
om
9m
7m
7m
4m
4m
nt
om
9m
7m
7m
4m
4m
e3
es
H3
e2
e1
e3
e2
e3
en
es
H3
e2
e1
e3
e2
e3
A
5
hypomethylation
hypermethylation
hypomethylation
hypermethylation
hypomethylation
hypermethylation
CG Heterochromatic TEs
CHG Euchromatic TEs
Density
Density
Euchromatic TEs
0 -0.8
-0.4
0.4
0.8
-0.4
0.4
Density
Heterochromatic TEs
CHH Euchromatic TEs Heterochromatic TEs
-0.4
0.4
Methylation difference (h1-WT)

h1

h1
0.7
h1ddm1 rep2
WT
h1ddm1 rep1
0.5
WT h1ddm1 rep2 h1ddm1 rep1
0.12
h1 h1ddm1 rep2
WT h1ddm1 rep1
mCHG
mCHH
mCG
ddm1
0 -4
ddm1
-2 0 2 4 2 0 -2 -4
0 -4
ddm1
-2 0 2 4 2 0 -2 -4
0 -4
-2
-2
-4 kb
C
log2(methylation mut/WT)
0
CG
0 -1
CHG
0 -0.4 -0.8 -1.2 -1.6 0 -0.6 0 -0.6 -3 -2 euchromatic -1 0 1 2 heterochromatic
CHH
-2
h1 h1ddm1_rep1 ddm1 h1ddm1/h1

-3 -2 euchromatic -1 0 1 2 heterochromatic
-1 -2
0 -0.6
-3 -2 euchromatic
-1
1 2 heterochromatic
H3K9me2
H3K9me2
H3K9me2
Figure 3. Lack of H1 Ameliorates the Loss of Methylation in ddm1 Plants

(A) Kernel density plots of methylation differences between h1 and wild-type (WT) (positive numbers indicate greater methylation in h1). TEs with H3K9me2 log2 scores lower than 1 and higher than 1 are considered euchromatic and heterochromatic, respectively. The colored arrows emphasize global differences (a shifted peak) or extensive local differences (a broad shoulder). (B) Average methylation of TEs in sibling wild-type (WT), h1, ddm1, and h1ddm1 (two biological replicates) seedlings is plotted as in Figure 1. (C and D) M-spline curve ts of log2 DNA methylation ratios in 50 bp windows plotted against H3K9me2 level. See also Figure S3.
Copia and MuDR elements preferentially accumulate sRNA at their 50 and 30 terminal repeats; and LINEs, which lack 50 repeats, have a spike in sRNA abundance at the 30 end (Figure 4A). DDM1-mediated DNA methylationthe residual methylation in the drd1 mutantdoes not appear to be inuenced by sRNA, as exemplied by efcient methylation of sRNApoor Copia and LINE TE bodies as well as sRNA-rich Gypsy elements in drd1 (compare the black traces in Figures 4B and S4A with that in Figure 4A); this supports our family-independent TE analysis (Figure 2A). CHH RdDMthe residual methylation in the ddm1 mutantresembles the distribution of sRNA in Copia, MuDR, and LINE elements (compare the blue trace in Figure 4B with the trace in Figure 4A). Although Gypsy elements
are evenly covered by sRNA (Figure 4A), RdDM is still preferentially localized at their edges (Figure 4B), supporting our conclusion that the heterochromatic environment of internal Gypsy sequences inhibits RdDM (Figure 2E). Lack of H1 ameliorates the CG, CHG, and CHH methylation losses caused by ddm1 in all TE families (Figures 4C and S4B). In particular, CHH methylation in h1ddm1 is similar to that in h1 at Gypsy and MuDR elements and, to a lesser extent, at Copia and LINE TEs (Figure 4C). CHH methylation in drm2 plants closely resembles that in drd1 mutants at all four TE families (compare the black traces in Figures 4B and 4D), further substantiating the link between RdDM and DRM2. The cmt2 CHH methylation phenotype is
A
7
Gypsy
7
MuDR
7
Copia
7
LINE
sRNA
sRNA
sRNA
0 -4
-2
-2
-4
0 -4 0.2
sRNA
0 -4 0.2
-2
-2
-4
-2
-2
-4
0 -4 0.2
-2
-2
-4kb
B
0.2
WT mCHH ddm1
drd1 mCHH mCHH mCHH

0 -4
ddm1drd1
0 -4 -2 0 2 4 2 0 -2 -4 0 -4 -2 0 2 4 2 0 -2 -4 0 -4 -2 0 2 4 2 0 -2 -4 -2 0 2 4 2 0 -2 -4kb
C
0.2
h1 mCHH
mCHH
mCHH
WT ddm1
-2 0 2 4 2 0 -2 -4
0 -4
0 -4
-2
-2
-4
0 -4
-2
-2
-4
mCHH
0 -4 -2
h1ddm1 rep2 h1ddm1 rep1
0.2
0.2
0.2
-2
-4kb
0.2
0.2
0.2
WT mCHH mCHH mCHH
0.2
drm2
0 -4 -2 0 2 4
cmt2
2 0 -2 -4 0 -4 -2 0 2 4 2 0 -2 -4 0 -4 -2 0 2 4 2 0 -2 -4 0 -4 -2 0 2 4 2 0 -2 -4kb
Figure 4. Methylation of TE Families Depends on sRNA Abundance and Chromatin Features

(AD) Averaged sRNA abundance (A) and CHH methylation levels (BD) are plotted as in Figure 1 for TEs belonging to four distinct families. The ddm1 trace in (C) represents siblings of the wild-type (WT), h1, and h1ddm1 seedlings analyzed in this panel and is independent of the ddm1 roots analyzed in (B). See also Figure S4.
similar to but stronger than that of ddm1 at MuDR, Copia, and LINE elements (compare the blue traces in Figures 4B and 4D)virtually all CHH methylation is lost in cmt2 plants except at sRNA-targeted terminal repeat sequences (Figure 4D), supporting our conclusion that CMT2 mediates CHH methylation independently of sRNA. It is surprising that lack of CMT2 essentially eliminates CHH methylation at Gypsy elements (Figure 4D), even though Gypsy CHH methylation can be maintained by RdDM (Figure 4B). This result suggests that CHH methylation feeds back on the levels of the sRNA molecules that mediate RdDM and is consistent with the recent nding that sRNA production from short TEs requires Pol V but sRNA production from Gypsy elements does not (Lee et al., 2012). Lack of the Pol V pathway, as exemplied by drd1, leads to major CHH hypomethylation of short TEs and TE edges (Figures 1E and 1F) and would thus be expected to reduce CHH methylation-dependent sRNA production from such sequences, whereas sRNA production from Gypsy elements that maintain RdDM-independent CHH methylation at sRNA-corresponding sequences (Figure 4B) would not require RdDM but would require CMT2. DDM1 and RdDM Collaborate to Repress TE Transcription and Transposition Consistent with the importance of DDM1 for the maintenance of DNA methylation in the bodies of long TEs, where TEencoded genes required for transposition are located, our RNA sequencing (RNA-seq) analysis revealed that many TEs (2,294) are reactivated in ddm1 mutant plants (Figure 5A). In contrast,
lack of DRD1, which primarily affects RdDM of noncoding short TEs and TE edges, caused the reactivation of just 44 TEs (Figure 5A). In agreement with our methylation analyses, TEs reactivated in ddm1 are longer and more heterochromatic than those reactivated in drd1 (Figure 5B). In both mutants, TE reactivation is associated with DNA hypomethylation (Figures 5C and S5A). The ddm1drd1 double mutant, which showed additive to synergistic hypomethylation (Figures 5C and S5A), led to stronger TE transcriptional reactivation than either of the single mutants (Figure 5D). This is particularly exemplied by Gypsy elements that require both DDM1 and RdDM for full methylation (Figures 4B and S4A), which are synergistically hyperactivated in ddm1drd1 (Figure 5E). Mutations in DDM1 and MET1 have been shown to cause transposition of a few TEs, including CACTA and EVADE, but only after several generations of inbreeding (Mirouze et al., 2009; Tsukahara et al., 2009). We found that CACTA and EVADE transpose within the rst homozygous generation of ddm1drd1 and ddm1rdr2 mutants (Figures 5F and S5B). This result emphasizes that the DDM1 and RdDM pathways collaborate to prevent TE expression and mobilization. DDM1 Mediates Gene Body DNA Methylation Plant genes, including those with presumed or demonstrated biological functions, can have methylation patterns that resemble TEs (You et al., 2012; Zemach et al., 2010a) and may therefore be regulated more like TEs than conventional genes. TEs are generally highly methylated at CG sites (Figure 6A, left), whereas most Arabidopsis genes have lower methylation
mCHH
(mut-WT) CHH methylation difference
ddm1 30 drd1 5 39
2225
drd1
ddm1
ddm1drd1
(mut-WT) CHH methylation difference
C
0 .1
Highly upregualted TEs in drd1
Highly upregualted TEs in ddm1
-.1
-.3
-.2
-.3
-.2
-.1
.1
drd1
ddm1
ddm1drd1
116 ddm1drd1 drd1 ddm1 ddm1drd1
D
15
Highly upregualted TEs in drd1 TE expression Log FC TE expression Log FC

15
Highly upregualted TEs in ddm1
10
10
12
drd1 TE size (kb) 4 6 8 H3K9me2 1 2
ddm1
ddm1drd1
Highly upregulated TEs in:
10
drd1
ddm1
ddm1drd1
F
ddm1rdr2 -6.1 1.8 ddm1drd1 -1.8 ddm1 -1.8 drd1 -1.8
drd1 ddm1
1.8 1.8 6.1
DNA copy
drd1
ddm1
EVADE
Genes TEs
CG methylation
15
E
TE expression Log FC
drd1
ddm1
ddm1drd1
wild-type drd1 ddm1 ddm1drd1 wild-type drd1 ddm1 ddm1drd1 wild-type drd1 ddm1 ddm1drd1
1 0 1 0 1 0 1 0 1
10
CHH methylation
0 1 0 1 0 1 0 9
RNA expression
0 9 0 9 0 9 0
-5 COPIA
LINE
MuDR
GYPSY
Figure 5. DDM1 and RdDM Collaborate to Repress TE Expression and Transposition

(A) Venn diagram of signicantly upregulated TEs in drd1, ddm1, and ddm1drd1 mutants. (BD) Box plots of the sizes and H3K9me2 levels (B), absolute fractional CHH demethylation of 50 bp windows (C), and expression compared to wild-type (D) of TEs that are at least 32-fold overexpressed either in drd1 or in ddm1. (E) Box plots of TE family expression in the indicated mutants with respect to wild-type. (F) DNA sequencing coverage (log2[reads in mutant/reads in wild-type]), DNA methylation, and RNA levels near the LTR retrotransposon EVADE (AT5TE20395). The sequence coverage is indicative of EVADE copy number relative to wild-type (Tsukahara et al., 2009). See also Figure S5.
levels, except for a distinct group that resembles TEs (Figure 6A). A cutoff of 60% overall CG methylation separates the two genic groups rather cleanly (Figure 6A). The 1,284 highly methylated genes resemble TEs in many aspects. For example, they are concentrated in pericentric heterochromatin (Figure S6A) and enriched for non-CG methylation (Figures 6B and S6B) and H3K9me2 (Figure 6C). CG methylation of such genes is lost in ddm1 but not drd1 mutants (Figure 6A), and non-CG methylation is lost partially in both mutants and synergistically in ddm1drd1 (Figures 6B and S6B). Because of these characteristics, we refer to the highly methylated TE-like genes as heterochromatic genes and to the rest as euchromatic genes. Despite the reported requirement of DDM1 and its mouse homolog Lsh for maintenance of DNA methylation in TEs but not genes (Lippman et al., 2004; Tao et al., 2011), we nd that CG methylation of at least 5,348 euchromatic gene bodies
(50% of all methylated euchromatic genes) is signicantly reduced in ddm1 plants (Fishers exact test p value < 0.0005), whereas only 85 genes are signicantly hypermethylated, leading to an overall methylation loss of 20% (Figures 6D and S6C). Lack of H1 also destabilizes genic methylation (Figures S3E and S6C), with at least 3,712 genes signicantly hypomethylated and 2,003 genes hypermethylated (p value < 0.0005). Nevertheless, h1 suppresses the genic hypomethylation caused by ddm1 (Figures 6D, 6E, and S3E), with numbers of signicantly hypomethylated (3,878) and hypermethylated (1,488) genes in h1ddm1 resembling h1. These results demonstrate that DDM1 is important for DNA methyltransferase access to all types of sequences and that the DDM1 requirement depends on the extent of heterochromatin, from the most heterochromatic long TEs to the less heterochromatic short TEs, to the least heterochromatic genes (Figure 6F).
Loss of DDM1 Causes Extensive Genic CHG Hypermethylation Loss of DDM1 has been reported to cause CHG hypermethylation of a few genes (Lippman et al., 2004; Saze and Kakutani, 2007), and genic hypermethylation was also reported in mouse cells lacking Lsh (Tao et al., 2011), but whether ddm1 causes extensive genic CHG hypermethylation, as has been reported for Arabidopsis met1 and ibm1 mutants (Lister et al., 2008; Saze and Kakutani, 2011), is unknown. We nd that lack of DDM1 leads to substantial CHG hypermethylation in euchromatic gene bodies (Figure 6D), but this hypermethylation differs from that caused by met1. Hypermethylation in ddm1 plants is higher toward the 30 end, whereas met1-mediated CHG hypermethylation is more prevalent near the 50 end (Figures 6D and S6D). CHG hypermethylation in ddm1 is restricted to genes that exhibit CG methylation and is excluded from 50 and 30 genic sequences like wild-type CG methylation (Figure 6G). In contrast, met1 causes CHG methylation of genes that lack CG methylation in wild-type, and the hypermethylation extends to the 30 and, to a lesser extent, the 50 regions of genes that are normally not CG methylated (Figure 6G). Like the ddm1 phenotype, CHG hypermethylation in plants lacking the H3K9me2 demethylase IBM1 (Figure 6D) is conned almost exclusively to genic regions that bear CG methylation (Figure 6G). Such genes also exhibit some CHG methylation in wild-type plants (Figure 6G), which may be due to imperfect activity of IBM1. The ibm1 mutation also causes substantial CHH hypermethylation of genes (Figure 6D) (Coleman-Derr and Zilberman, 2012) that is insensitive to drd1 (Figure S6E) and is thus likely mediated by CMT2. However, hypermethylation in ddm1 plants is unlikely to be primarily caused by IBM1 malfunction (Miura et al., 2009) because genic CHG methylation is 50 biased in ibm1, unlike the 30 -biased methylation in ddm1 (Figures 6D and S6D). This interpretation is also supported by the normal expression of the IBM1 transcript in ddm1, which is disrupted in met1 (Figure S6F) (Rigal et al., 2012). Thus, lack of MET1, IBM1, and DDM1 leads to distinct genic hypermethylation phenotypes likely mediated by different although potentially overlapping mechanisms. DDM1 and RdDM Synergistically Regulate Gene Expression The importance of TE silencing by DNA methylation is not restricted to preventing transpositiondemethylation and activation of TEs can also alter the expression of nearby genes (Henderson and Jacobsen, 2008; Hsieh et al., 2011). We found 179 genes upregulated in ddm1 plants (Figures 7A and S7A), all but one of which are heterochromatic (Figure 7B), consistent with the importance of DDM1 for the methylation of heterochromatin generally and heterochromatic genes specically (Figures 6A, 6B, 7C, and S6B). Only 10 genes are upregulated in drd1 plants (Figure 7A), six of which are euchromatic (Figure 7B) and only two of which overlap with genes upregulated by ddm1. One of the heterochromatic genes upregulated in drd1 is AT1G59930 (Figure 7C), a maternally expressed imprinted gene that is activated by DNA demethylation (Hsieh et al., 2011). All six drd1upregulated euchromatic genes (AT1G21940, AT1G35730, AT4G01985, AT4G09350, AT4G16460, and AT5G41830) have
short TEs or TE edges that are hypomethylated in drd1 but not in ddm1 in close proximity to the transcriptional start sites, such as the 150 bp Copia fragment (META1) upstream of AT1G35730 (Figure S7B). More genes (214) are overexpressed in the ddm1drd1 double mutant, including 23 euchromatic genes and ve of the six euchromatic genes upregulated in drd1 (Figures 7A and 7B). One such gene is SDC (Figure 7D), the overexpression of which confers the characteristic phenotype of the drm1drm2cmt3 methyltransferase mutant line (Henderson and Jacobsen, 2008) that is also exhibited by ddm1drd1 plants (Figure 7E). Similarly to Gypsy elements, the repetitive SDC promoter is targeted by sRNA and H3K9me2 (Figure 7D), and its methylation and silencing is mediated by DDM1 and RdDM (Figure 7D). SDC is also a maternally expressed imprinted gene regulated by DNA demethylation (Hsieh et al., 2011), as is FWA, which is modestly upregulated in ddm1drd1 plants due to demethylation of its SINE-related promoter (Figure S7C). Overall, our results demonstrate that DDM1 and RdDM collaborate to methylate geneadjacent repetitive elements, thereby maintaining appropriate patterns of gene expression. DISCUSSION The targeting of plant DNA methylation has been carefully dissected at individual loci, and the methyltransferases that catalyze CG and CHG methylation throughout the genome are known (Law and Jacobsen, 2010), but the identity of the pathways that mediate the bulk of genomic methylation has remained a mystery. Here, we nd that DDM1 and RdDM separately mediate nearly all DNA methylation in Arabidopsis TEs. DDM1 is required for methylation in all sequence contexts of highly heterochromatic TEs (Figures 1A and 2A). This requirement is reduced at less heterochromatic elements, is least in euchromatic genes (Figure 6F), and depends on histone H1 (Figures 3B3D). Together with the preferential demethylation of short euchromatic TEs during plant sexual development (Ibarra et al., 2012; Zemach et al., 2010a), our results demonstrate that a division of the genome into genes and TEs can only explain the biology of DNA methylation if chromatin conguration is also considered. Lack of access to DNA is postulated to be a core property of heterochromatin that enforces gene silencing by preventing binding of transcription factors and RNA polymerase (Grewal and Jia, 2007). At the same time, stable maintenance of inaccessible heterochromatin requires DNA methylation and histone modications like H3K9me2 that are catalyzed by enzymes that need to access chromatin. This conundrum is exemplied by the RdDM pathway, which silences TE expression yet requires TE transcripts to function (Haag and Pikaard, 2011). Our results indicate that H1 restricts access to nucleosomal DNA and that DDM1 overcomes this restriction to enable the maintenance of DNA methylation and silencing of diverse TEs. Without DDM1, DNA methyltransferases cannot efciently methylate inaccessible heterochromatic TEs (Figures 1A and 2A), leading to derepression and transposition (Figures 5 and S5). Without H1, less heterochromatic sequences lose methylation (Figures 3A and S3E), presumably because they become more
A
CG methylation
wild-type
1 0.8 0.6 0.4 0.2 0 0 2 4 6 8 1 0.8 0.6 0.4 0.2 10kb 0 0 2
wild-type
1 0.8 0.6 0.4 0.2 4 6 8 10kb 0 0 2
drd1
1 0.8 0.6 0.4 0.2 4 6 8 10kb 0 0 2
ddm1
1 0.8 0.6 0.4 0.2 4 6 8 10kb 0 0 2
ddm1drd1
TE size
Gene size
Gene size
Gene size
Gene size
10kb
B
1
TEs
CHG methylation
CG methylation
mCG<0.6 Genes
mCG>0.6 Genes
C
2 3
D
0.6 ibm1
WT
h1
drd1
0.8
H3K9me2
0.6
ddm1
h1ddm1 rep2
h1ddm1 rep1
0.4
met1 -1 0 1 2 3 4 3 2 1 ibm1 0 -1 -2 kb
0.2
0 -2 0.55 mCG <0.6 mCG >0.6
-1
WT
WT
WT
drd1
ddm1 ddm1 drd1
CHG methylation
met1
ddm1
E
wild type CG methylation
0.02
1 0 1
WT
-1 0 1 2 3
drd1 4 3 2 1 0 -1 -2 kb
h1
0 1
0 -2 0.05
0 1
CHH methylation
ddm1 h1ddm1
0
ibm1 met1
WT
Genes TEs
ddm1 0 1 2 3 4 3 2 1 0 -1 -2 kb
0 drd1 -2 -1
F
1 1 4
G
CG methylation
wild-type
ddm1
met1
ibm1
WT - ddm1 CG methylation difference
WT CG methylation
.8
.5
.6
H3K9me2
>4 <0 G kb .5k ene b s
TEs
TEs
TEs
CHG methylation
-3 0
.4
-.5
-2
3 kb
-3
3 kb
-3
3 kb
-3
3 kb
-3
Figure 6. DDM1 Mediates Genic DNA Methylation

(A) CG methylation was averaged in TEs (leftmost panel) and genes for the indicated genotypes (four right panels) and plotted against TE or gene size. Note the group of relatively short genes above the red line that are highly methylated in wild-type and signicantly hypomethylated in ddm1 and ddm1drd1 mutants, similarly to TEs. (B) Box plots of averaged CHG methylation in TEs, euchromatic genes (mCG < 0.6), and heterochromatic genes (mCG > 0.6). WT, wild-type. (C) Box plots of H3K9me2 in euchromatic and heterochromatic genes. (D) Genes with average CG methylation between 20% and 60% (euchromatic genes) were aligned as described in Figure 1A. The y axis of the CHG plot was broken at 0.02 to improve visualization. WT, wild-type. (E) Distribution of CG methylation in representative genes AT1G04700, AT1G04750, and AT1G67220 (emphasized by horizontal black bars) that lose methylation in ddm1 but not in h1ddm1. (F) Box plots of wild-type CG methylation (left), absolute fractional CG demethylation in ddm1 (middle), and H3K9me2 (right) of 50 bp windows within long TEs (>4 kb), short TEs (<500 bp), and euchromatic genes.
A
All genes
Only euchromatic genes (mCG<0.6) drd1 1 ddm1 5 1
wild-type drd1 ddm1 ddm1drd1 wild-type
1 0 1 0 1 0 1 0 1 0 1
mCG mCHG
ddm1 drd1 17 2 169 8
17 ddm1drd1 drd1
drd1 0 1 ddm1 0 1 ddm1drd1 0 wild-type 0.5 0 drd1 0 0.5 ddm1 0 0.5 ddm1drd1 0
0.5
mCHH
36
ddm1drd1
ddm1 ddm1drd1
AT1G59930
wild-type drd1 ddm1 ddm1drd1 sRNA H3K9me2
6 0 6 0 6 0 6 0
Genes TEs mRNA
D
wild-type ddm1 drd1 ddm1drd1 wild-type ddm1 drd1 ddm1drd1
promoter
6.5 0 0.6 -0.8 0.2 0 0.2 0 0 0 0.2 0
E
mCHH Genes
wild-type
ddm1drd1
6.6 0 6.6 0 6.6 0 6.6 0
SDC
Figure 7. DDM1 and RdDM Synergistically Regulate Gene Expression

(A and B) Venn diagram of signicantly (p < 0.05) upregulated genes in drd1, ddm1, and ddm1drd1 mutants. (C) DNA methylation and RNA levels near AT1G46696, and the linked genes AT1G59920 and AT1G59930. (D) sRNA, H3K9me2, CHH methylation, and RNA levels near SDC (AT2G17690). (E) Phenotypes of wild-type (at leaves) and ddm1drd1 (leaves curled downward) plants. See also Figure S7.
accessible to enzymes that catalyze euchromatic histone modications and antagonize DNA methylation (Ibarra et al., 2012; Probst et al., 2004; Qian et al., 2012). The balance between exclusion and access is thus essential for the stable propagation of chromatin states. The inuence of the chromatin environment on DNA methylation has important implications for how different classes of TEs are silenced. Short TEs, which are preferentially found near active genes (Ibarra et al., 2012; Zemach et al., 2010a), are generally relatively euchromatic (Figure 2B) and rely on RdDM for silencing (Figure 5B), whereas silencing of the more heterochromatic longer TEs (Figure 2B) that are usually found away from genes (Ibarra et al., 2012; Zemach et al., 2010a) relies primarily on DDM1 (Figure 5B). Despite these differences, DDM1 and RdDM contribute to the methylation and silencing of most TEs, leading to a synergistic loss of methylation (Figures 1A, 4B, and 5C) and repression (Figures 5D and 5E), as well as to enhanced transposition (Figures 5F and S5B), when both pathways are mutated. Our data suggest that RdDM operates primarily through DRM2 and is thus responsible for a relatively minor fraction of genomic methylation (Figures 1A and 1B) (Wierzbicki et al.,
(G) Heat maps of CG (red) and CHG (yellow) DNA methylation in genes aligned at the 50 end (left half of each panel) and the 30 end (right half of each panel). More intense color indicates greater methylation. Genes without wild-type CG methylation (shown in the top half of each panel) were stacked from the top of chromosome 1 to the bottom of chromosome 5; genes containing CG methylation islands (shown at the bottom of each panel) were sorted based on the starting position (for 50 panels) or ending position (for 30 panels) of the wild-type CG methylation island. See also Figure S6.
mRNA
2012). We show that the majority of CHH methylation is mediated by CMT2 (Figure S1A) independently of RdDM (Figures 1A1C, 4D; Table S2), presumably by binding to H3K9me2 like its CMT3 homolog (Du et al., 2012). CMT2 forms a distinct family in monocots and dicots (Figure 1D), indicating that this enzyme catalyzes CHH methylation throughout owering plants, including important crops such as rice. Curiously, we have not been able to identify a CMT2 homolog in maize, suggesting that CHH methylation may be entirely dependent on RdDM in this species. RdDM, by denition, only targets loci that generate sRNA and therefore affects TEs, which are at least somewhat heterochromatic compared to genes, almost exclusively (Figures 1, 4B, and 6D) (Zhong et al., 2012). Production of sRNA can be inuenced by many factors, including TE structure and copy number (Martienssen, 2003). Thus, the abundant Gypsy LTR retrotransposons that make up the bulk of pericentric heterochromatin generate sRNA differently from the more dispersed Copia, LINE, and MuDR elements (Figure 4A), and RdDM is crucial for Gypsy silencing (Figure 5E) despite the highly heterochromatic nature of these elements. Nonetheless, RdDM is more efcient at less heterochromatic sRNA-producing loci (Figure 2E) (Schoft
et al., 2009). This is a curious observation because RdDM, which functions to methylate and silence TEs, might be expected to work well in heterochromatin. Furthermore, why would DDM1, a protein apparently adapted to remodel heterochromatic nucleosomes, not facilitate RdDM? Our observation that DRD1 is required for RdDM offers a potential explanation. DRD1 is associated with RNA polymerase V, a derivative of RNA polymerase II that shares most Pol II subunits (Haag and Pikaard, 2011). Pol II has evolved to transcribe genes; thus, the machinery associated with Pol V is likely adapted to function in euchromatin. Unlike DDM1, DRD1 may remodel heterochromatic nucleosomes inefciently. Because RdDM is a branch of the RNA interference pathway that functions to cleave aberrant and viral mRNA, it likely evolved independently of heterochromatic DDM1-associated pathways. Thus, plants possess two separate mechanisms for methylating and silencing TEs that rely on distinct remodelers with differing nucleosome preferences.
EXPERIMENTAL PROCEDURES Genomic Data Acquisition Bisulte conversion, Illumina library construction, sequencing, and data processing were performed exactly as described by Ibarra et al. (2012). DNA for the MNase sequencing library was prepared essentially as described by Zilberman et al. (2008). RNA-seq analysis was performed as described by Zemach et al. (2010b). Phylogenetic Analysis Phylogenetic analysis was performed as described by Zemach et al. (2010b). Gene and TE DNA Methylation and Chromatin Pattern Analyses Gene and TE meta-analysis was performed as described by Coleman-Derr and Zilberman (2012). Genes with CG methylation over 60% were excluded from the analysis because they behave like transposons. Genes with low CG methylation (<20%) were also excluded from methylation analyses because they decrease the dynamic range without substantively contributing to the analysis. To avoid complications in calculating TE size caused by serial TE insertions near the centromeres, only TEs located on the chromosome arms were included in analyses where TEs were ltered by size. GC content was calculated by averaging in 50 bp windows. Kernel Density Plots Density plots in Figure 3A were generated with the difference between h1 and wild-type root methylation in 50 bp windows with at least 10 informative sequenced cytosines and fractional methylation of at least 0.3 for CG and 0.1 for CHG and CHH in h1 or wild-type. Genes with over 60% and under 20% CG methylation were excluded from the analysis. Expression Analysis RNA-seq data sets were mapped to the TAIR cDNA and TE annotations and analyzed using the Bioconductor package edgeR (Robinson et al., 2010). ACCESSION NUMBERS Sequencing data have been deposited in Gene Expression Omnibus under accession number GSE41302. SUPPLEMENTAL INFORMATION Supplemental Information includes Extended Experimental Procedures, seven gures, and two tables and can be found with this article online at http://dx.doi. org/10.1016/j.cell.2013.02.033.
ACKNOWLEDGMENTS We thank Toshiro Nishimura for computational support, Minyong Chung for Illumina sequencing, and Robert Fischer for critical reading of the manuscript. A.Z. was supported by a fellowship from the Jane Cofn Childs Memorial Fund for Medical Research. P-H.H. is supported by a Taiwan Ministry of Education Scholarship. D.Z. is a Young Investigator of the Arnold and Mabel Beckman Foundation. Received: September 15, 2012 Revised: December 1, 2012 Accepted: February 11, 2013 Published: March 28, 2013 REFERENCES Ahmed, I., Sarazin, A., Bowler, C., Colot, V., and Quesneville, H. (2011). Genome-wide evidence for local DNA methylation spreading from small RNA-targeted sequences in Arabidopsis. Nucleic Acids Res. 39, 69196931. Ascenzi, R., and Gantt, J.S. (1999). Subnuclear distribution of the entire complement of linker histone variants in Arabidopsis thaliana. Chromosoma 108, 345355. Barra, J.L., Rhounim, L., Rossignol, J.L., and Faugeron, G. (2000). Histone H1 is dispensable for methylation-associated gene silencing in Ascobolus immersus and essential for long life span. Mol. Cell. Biol. 20, 6169. Blevins, T., Pontes, O., Pikaard, C.S., and Meins, F., Jr. (2009). Heterochromatic siRNAs and DDM1 independently silence aberrant 5S rDNA transcripts in Arabidopsis. PLoS ONE 4, e5932. Brzeski, J., and Jerzmanowski, A. (2003). Decient in DNA methylation 1 (DDM1) denes a novel family of chromatin-remodeling factors. J. Biol. Chem. 278, 823828. Buisine, N., Quesneville, H., and Colot, V. (2008). Improved detection and annotation of transposable elements in sequenced genomes using multiple reference sequence sets. Genomics 91, 467475. Cokus, S.J., Feng, S., Zhang, X., Chen, Z., Merriman, B., Haudenschild, C.D., Pradhan, S., Nelson, S.F., Pellegrini, M., and Jacobsen, S.E. (2008). Shotgun bisulphite sequencing of the Arabidopsis genome reveals DNA methylation patterning. Nature 452, 215219. Coleman-Derr, D., and Zilberman, D. (2012). Deposition of histone variant H2A.Z within gene bodies regulates responsive genes. PLoS Genet. 8, e1002988. Du, J., Zhong, X., Bernatavichute, Y.V., Stroud, H., Feng, S., Caro, E., Vashisht, A.A., Terragni, J., Chin, H.G., Tu, A., et al. (2012). Dual binding of chromomethylase domains to H3K9me2-containing nucleosomes directs DNA methylation in plants. Cell 151, 167180. Fan, Y., Nikitina, T., Zhao, J., Fleury, T.J., Bhattacharyya, R., Bouhassira, E.E., Stein, A., Woodcock, C.L., and Skoultchi, A.I. (2005). Histone H1 depletion in mammals alters global chromatin structure but causes specic changes in gene regulation. Cell 123, 11991212. Grewal, S.I., and Jia, S. (2007). Heterochromatin revisited. Nat. Rev. Genet. 8, 3546. Haag, J.R., and Pikaard, C.S. (2011). Multisubunit RNA polymerases IV and V: purveyors of non-coding RNA for plant gene silencing. Nat. Rev. Mol. Cell Biol. 12, 483492. Henderson, I.R., and Jacobsen, S.E. (2008). Tandem repeats upstream of the Arabidopsis endogene SDC recruit non-CG DNA methylation and initiate siRNA spreading. Genes Dev. 22, 15971606. Hsieh, T.F., Shin, J., Uzawa, R., Silva, P., Cohen, S., Bauer, M.J., Hashimoto, M., Kirkbride, R.C., Harada, J.J., Zilberman, D., and Fischer, R.L. (2011). Regulation of imprinted gene expression in Arabidopsis endosperm. Proc. Natl. Acad. Sci. USA 108, 17551762. Ibarra, C.A., Feng, X., Schoft, V.K., Hsieh, T.F., Uzawa, R., Rodrigues, J.A., Zemach, A., Chumak, N., Machlicova, A., Nishimura, T., et al. (2012). Active
DNA demethylation in plant companion cells reinforces transposon methylation in gametes. Science 337, 13601364. Iyer, V.R. (2012). Nucleosome positioning: bringing order to the eukaryotic genome. Trends Cell Biol. 22, 250256. Jeddeloh, J.A., Stokes, T.L., and Richards, E.J. (1999). Maintenance of genomic methylation requires a SWI2/SNF2-like protein. Nat. Genet. 22, 9497. Jullien, P.E., Susaki, D., Yelagandula, R., Higashiyama, T., and Berger, F. (2012). DNA methylation dynamics during sexual reproduction in Arabidopsis thaliana. Curr. Biol. 22, 18251830. Kanno, T., Mette, M.F., Kreil, D.P., Aufsatz, W., Matzke, M., and Matzke, A.J. (2004). Involvement of putative SNF2 chromatin remodeling protein DRD1 in RNA-directed DNA methylation. Curr. Biol. 14, 801805. Law, J.A., and Jacobsen, S.E. (2010). Establishing, maintaining and modifying DNA methylation patterns in plants and animals. Nat. Rev. Genet. 11, 204220. Law, J.A., Ausin, I., Johnson, L.M., Vashisht, A.A., Zhu, J.K., Wohlschlegel, J.A., and Jacobsen, S.E. (2010). A protein complex required for polymerase V transcripts and RNA- directed DNA methylation in Arabidopsis. Curr. Biol. 20, 951956. Lee, T.F., Gurazada, S.G., Zhai, J., Li, S., Simon, S.A., Matzke, M.A., Chen, X., and Meyers, B.C. (2012). RNA polymerase V-dependent small RNAs in Arabidopsis originate from small, intergenic loci including most SINE repeats. Epigenetics 7, 781795. Lippman, Z., Gendrel, A.V., Black, M., Vaughn, M.W., Dedhia, N., McCombie, W.R., Lavine, K., Mittal, V., May, B., Kasschau, K.D., et al. (2004). Role of transposable elements in heterochromatin and epigenetic control. Nature 430, 471476. Lister, R., OMalley, R.C., Tonti-Filippini, J., Gregory, B.D., Berry, C.C., Millar, A.H., and Ecker, J.R. (2008). Highly integrated single-base resolution maps of the epigenome in Arabidopsis. Cell 133, 523536. Martienssen, R.A. (2003). Maintenance of heterochromatin by RNA interference of tandem repeats. Nat. Genet. 35, 213214. Mirouze, M., Reinders, J., Bucher, E., Nishimura, T., Schneeberger, K., Ossowski, S., Cao, J., Weigel, D., Paszkowski, J., and Mathieu, O. (2009). Selective epigenetic control of retrotransposition in Arabidopsis. Nature 461, 427430. Miura, A., Nakamura, M., Inagaki, S., Kobayashi, A., Saze, H., and Kakutani, T. (2009). An Arabidopsis jmjC domain protein protects transcribed genes from DNA methylation at CHG sites. EMBO J. 28, 10781086. Pennings, S., Meersseman, G., and Bradbury, E.M. (1994). Linker histones H1 and H5 prevent the mobility of positioned nucleosomes. Proc. Natl. Acad. Sci. USA 91, 1027510279. Probst, A.V., Fagard, M., Proux, F., Mourrain, P., Boutet, S., Earley, K., Lawrence, R.J., Pikaard, C.S., Murfett, J., Furner, I., et al. (2004). Arabidopsis histone deacetylase HDA6 is required for maintenance of transcriptional gene silencing and determines nuclear organization of rDNA repeats. Plant Cell 16, 10211034. Qian, W., Miki, D., Zhang, H., Liu, Y., Zhang, X., Tang, K., Kan, Y., La, H., Li, X., Li, S., et al. (2012). A histone acetyltransferase regulates active DNA demethylation in Arabidopsis. Science 336, 14451448. Raghuram, N., Carrero, G., Thng, J., and Hendzel, M.J. (2009). Molecular dynamics of histone H1. Biochem. Cell Biol. 87, 189206. Rea, M., Zheng, W., Chen, M., Braud, C., Bhangu, D., Rognan, T.N., and Xiao, W. (2012). Histone H1 affects gene imprinting and DNA methylation in Arabidopsis. Plant J. 71, 776786. lissier, T., and Mathieu, O. (2012). DNA methylation in an Rigal, M., Kevei, Z., Pe intron of the IBM1 histone demethylase gene stabilizes chromatin modication patterns. EMBO J. 31, 29812993.
Robinson, M.D., McCarthy, D.J., and Smyth, G.K. (2010). edgeR: a Bioconductor package for differential expression analysis of digital gene expression data. Bioinformatics 26, 139140. Robinson, P.J., and Rhodes, D. (2006). Structure of the 30 nm chromatin bre: a key role for the linker histone. Curr. Opin. Struct. Biol. 16, 336343. rard, C., Sarazin, A., Mary-Huard, T., Cortijo, S., Roudier, F., Ahmed, I., Be Bouyer, D., Caillieux, E., Duvernois-Berthet, E., Al-Shikhley, L., et al. (2011). Integrative epigenomic mapping denes four main chromatin states in Arabidopsis. EMBO J. 30, 19281938. Ryan, D.P., and Owen-Hughes, T. (2011). Snf2-family proteins: chromatin remodellers for any occasion. Curr. Opin. Chem. Biol. 15, 649656. Saze, H., and Kakutani, T. (2007). Heritable epigenetic mutation of a transposon-anked Arabidopsis gene due to lack of the chromatin-remodeling factor DDM1. EMBO J. 26, 36413652. Saze, H., and Kakutani, T. (2011). Differentiation of epigenetic modications between transposons and genes. Curr. Opin. Plant Biol. 14, 8187. Schoft, V.K., Chumak, N., Mosiolek, M., Slusarz, L., Komnenovic, V., Browneld, L., Twell, D., Kakutani, T., and Tamaru, H. (2009). Induction of RNA-directed DNA methylation upon decondensation of constitutive heterochromatin. EMBO Rep. 10, 10151021. Tao, Y., Xi, S., Shan, J., Maunakea, A., Che, A., Briones, V., Lee, E.Y., Geiman, T., Huang, J., Stephens, R., et al. (2011). Lsh, chromatin remodeling family member, modulates genome-wide cytosine methylation patterns at nonrepeat sequences. Proc. Natl. Acad. Sci. USA 108, 56265631. Teixeira, F.K., Heredia, F., Sarazin, A., Roudier, F., Boccara, M., Ciaudo, C., Cruaud, C., Poulain, J., Berdasco, M., Fraga, M.F., et al. (2009). A role for RNAi in the selective correction of DNA methylation defects. Science 323, 16001604. Thomas, J.O. (1999). Histone H1: location and role. Curr. Opin. Cell Biol. 11, 312317. Tsukahara, S., Kobayashi, A., Kawabe, A., Mathieu, O., Miura, A., and Kakutani, T. (2009). Bursts of retrotransposition reproduced in Arabidopsis. Nature 461, 423426. Wierzbicki, A.T., and Jerzmanowski, A. (2005). Suppression of histone H1 genes in Arabidopsis results in heritable developmental defects and stochastic changes in DNA methylation. Genetics 169, 9971008. Wierzbicki, A.T., Cocklin, R., Mayampurath, A., Lister, R., Rowley, M.J., Gregory, B.D., Ecker, J.R., Tang, H., and Pikaard, C.S. (2012). Spatial and functional relationships among Pol V-associated loci, Pol IV-dependent siRNAs, and cytosine methylation in the Arabidopsis epigenome. Genes Dev. 26, 1825 1836. Xie, Z., Johansen, L.K., Gustafson, A.M., Kasschau, K.D., Lellis, A.D., Zilberman, D., Jacobsen, S.E., and Carrington, J.C. (2004). Genetic and functional diversication of small RNA pathways in plants. PLoS Biol. 2, E104. You, W., Tyczewska, A., Spencer, M., Daxinger, L., Schmid, M.W., Grossniklaus, U., Simon, S.A., Meyers, B.C., Matzke, A.J., and Matzke, M. (2012). Atypical DNA methylation of genes encoding cysteine-rich peptides in Arabidopsis thaliana. BMC Plant Biol. 12, 51. Zemach, A., Kim, M.Y., Silva, P., Rodrigues, J.A., Dotson, B., Brooks, M.D., and Zilberman, D. (2010a). Local DNA hypomethylation activates genes in rice endosperm. Proc. Natl. Acad. Sci. USA 107, 1872918734. Zemach, A., McDaniel, I.E., Silva, P., and Zilberman, D. (2010b). Genome-wide evolutionary analysis of eukaryotic DNA methylation. Science 328, 916919. Zhong, X., Hale, C.J., Law, J.A., Johnson, L.M., Feng, S., Tu, A., and Jacobsen, S.E. (2012). DDR complex facilitates global association of RNA polymerase V to promoters and evolutionarily young transposons. Nat. Struct. Mol. Biol. 19, 870875. Zilberman, D., Coleman-Derr, D., Ballinger, T., and Henikoff, S. (2008). Histone H2A.Z and DNA methylation are mutually antagonistic chromatin marks. Nature 456, 125129.
Cand1 Promotes Assembly of New SCF Complexes through Dynamic Exchange of F Box Proteins
Nathan W. Pierce,1,9 J. Eugene Lee,1,9,11 Xing Liu,1 Michael J. Sweredoski,2 Robert L.J. Graham,2 Elizabeth A. Larimore,5,8 Michael Rome,1,3 Ning Zheng,6 Bruce E. Clurman,7,8 Sonja Hess,2 Shu-ou Shan,3,10 and Raymond J. Deshaies1,4,10,*
of Biology, MC 156-29 Exploration Laboratory of the Beckman Institute, MC 139-74 3Division of Chemistry and Chemical Engineering, MC 147-75 4Howard Hughes Medical Institute California Institute of Technology, 1200 East California Boulevard, Pasadena, CA 91125, USA 5Molecular and Cellular Biology Program 6Howard Hughes Medical Institute, Department of Pharmacology University of Washington, P.O. Box 357280, Seattle, WA 98195, USA 7Division of Clinical Research 8Human Biology Fred Hutchinson Cancer Research Center, Seattle, WA 98109, USA 9These authors contributed equally to this work 10These authors contributed equally to this work 11Present address: Center for Bioanalysis, Division of Metrology for Quality of Life, Korea Research Institute of Standards and Science, Daejeon 305-340, Korea *Correspondence: deshaies@caltech.edu http://dx.doi.org/10.1016/j.cell.2013.02.024
2Proteome 1Division
SUMMARY
The modular SCF (Skp1, cullin, and F box) ubiquitin ligases feature a large family of F box protein substrate receptors that enable recognition of diverse targets. However, how the repertoire of SCF complexes is sustained remains unclear. Real-time measurements of formation and disassembly indicate that SCFFbxw7 is extraordinarily stable, but, in the Nedd8-deconjugated state, the cullin-binding protein Cand1 augments its dissociation by onemillion-fold. Binding and ubiquitylation assays show that Cand1 is a protein exchange factor that accelerates the rate at which Cul1-Rbx1 equilibrates with multiple F box protein-Skp1 modules. Depletion of Cand1 from cells impedes recruitment of new F box proteins to pre-existing Cul1 and profoundly alters the cellular landscape of SCF complexes. We suggest that catalyzed protein exchange may be a general feature of dynamic macromolecular machines and propose a hypothesis for how substrates, Nedd8, and Cand1 collaborate to regulate the cellular repertoire of SCF complexes.
INTRODUCTION Three enzymes work in succession to covalently attach ubiquitin and ubiquitin chains to target proteins: an ubiquitin activating
enzyme (E1), an ubiquitin conjugating enzyme (E2), and an ubiquitin ligase (E3) (Dye and Schulman, 2007). The proteasome, a massive multisubunit protease, recognizes and degrades proteins conjugated to lysine 48- or lysine 11-linked polyubiquitin chains containing at least four ubiquitins (Thrower et al., 2000; Wickliffe et al., 2011). Cullin-RING ubiquitin ligases (CRLs) are the largest family of E3s and are typied by the SCF (Skp1, cullin, and F box) complexes, which comprise four subunits: the scaffold Cul1, the RING domain protein Rbx1, the adaptor Skp1, and a substrate-binding F box protein (Petroski and Deshaies, 2005; Dye and Schulman, 2007). Sixty-nine proteins in the human genome have F box motifs, at least 42 of which form SCF complexes (Lee et al., 2011). Although this modularity of SCF complexes allows for recognition of diverse substrates, how SCF complex formation is regulated remains unclear. Cullin-associated and neddylation-dissociated protein 1 (Cand1) is a Cul1-associated protein whose binding is mutually exclusive with the F box protein-Skp1 subcomplex and is also blocked by attachment of the ubiquitin-like protein Nedd8 to lysine 720 of Cul1 (Liu et al., 2002; Zheng et al., 2002a; Goldenberg et al., 2004; Bornstein et al., 2006; Duda et al., 2008; Siergiejuk et al., 2009). Neddylation of Cul1 induces a major conformational rearrangement in Cul1-Rbx1 that eliminates a Cand1-binding site and stimulates ubiquitin transfer from associated E2s to substrates (Duda et al., 2008; Saha and Deshaies, 2008). In vitro, Cand1 acts as a stoichiometric inhibitor of cullin neddylation, SCF assembly, and SCF ubiquitin ligase activity (Liu et al., 2002; Zheng et al., 2002a; Goldenberg et al., 2004; Bornstein et al., 2006; Duda et al., 2008; Siergiejuk et al., 2009). However, genetic evidence indicates that Cand1
Figure 1. FRET Reveals Properties of SCF Assembly

(A) Fluorescence emission spectra from excitation at 430 nm of 70 nM CFPCul1-Rbx1, 70 nM Fbxw7TAMRA-Skp1, a mixture of the two, or buffer alone revealed FRET with 30% efciency upon complex formation. Signals were normalized to peak donor emission at 478 nm. (B) The change in donor uorescence versus time in a stopped ow apparatus with 5 nM CFPCul1Rbx1 and varying concentrations of Fbxw7TAMRASkp1. Signal changes were t to single exponential curves. (C) The rate of signal change in (B) versus the concentration of Fbxw7TAMRA-Skp1. Fitting the data to (kobs = kon*[Fbxw7] + koff) gave kon of 4 3 106 M1 s1 regardless of Cul1s neddylation status. Error bars, SD, n R 3. (D) 700 nM Skp2-Skp1 (chase) competed FRET away if preincubated with 70 nM Fbxw7TAMRASkp1 before, but not after addition of 70 nM CFP Cul1 for 5 min. (E) Fluorescence emission at 478 nm versus time after addition of chase to preincubated CFPCul1Rbx1 and Fbxw7TAMRA-Skp1 normalized to peak donor emission in (D). Single exponential t with a xed end point of 1 gave koff of 8.5 3 107 s1. KD is thus 2 3 1013 M. Error bars, SD, n = 3. See also Figure S1.
is a positive regulator of SCF and other CRLs in vivo (Zheng et al., 2002a; Chuang et al., 2004; Feng et al., 2004; Lo and Hannink, 2006; Zhang et al., 2008; Bosu et al., 2010; Kim et al., 2010), suggesting that Cand1-mediated recycling of substrate receptor modules is important for proper CRL function (Liu et al., 2002; Cope and Deshaies, 2003; Dye and Schulman, 2007; Zhang et al., 2008; Schmidt et al., 2009). Here, we demonstrate a protein exchange factor activity for Cand1 that resolves the conicting biochemical and genetic data and together with other data on the Nedd8 conjugation cycle enables us to propose a specic model for how the cellular repertoire of CRL complexes is controlled. RESULTS Dynamics of SCFFbxw7 Assembly and Disassembly To reconcile the conicting observations on Cand1, we sought to characterize the assembly properties of SCF complexes by developing a real-time assay based on uorescence resonance energy transfer (FRET) that monitors the binding dynamics between F box protein-Skp1 and Cul1-Rbx1. Fbxw7 coexpressed with Skp1 was tagged at its C terminus, via the Sortase reaction (Popp et al., 2009; Proft, 2010), with the peptide GGGGK conjugated to the uorescent dye TAMRA, producing covalently labeled Fbxw7TAMRA. The transpeptidation reaction
was efcient (data not shown) and the tag did not compromise ubiquitylation activity (Figure S1A available online). We observed FRET when Fbxw7TAMRA-Skp1 was mixed with Cul1-Rbx1 in which the N terminus of Cul1 is fused to cyan uorescent protein (CFPCul1) (Figure 1A). The association rate constant (kon) for complex assembly, 4 3 106 M1 s1, was determined by monitoring donor CFPCul1-Rbx1 uorescence at varying concentrations of Fbxw7TAMRA-Skp1 in a stopped ow apparatus (Figures 1B and 1C). The FRET observed in our assay could be competed away by excess nonuorescent Skp2-Skp1 (Figure 1D). Using this chase assay, we measured a dissociation rate constant (koff) for SCFFbxw7 of 9 3 107 s1, or 0.5 week1 (Figure 1E). These measurements revealed an extraordinarily tight complex with a KD of 200 fM. Neddylation of Cul1 did not affect the maximum FRET efciency in our assay nor the rate of association or dissociation of SCFFbxw7, conrming that the neddylation-induced major conformational change in Rbx1 and Cul1s C-terminal domain does not affect the binding interface of F box protein-Skp1 to Cul1s N-terminal domain (Figure 1C and S1BS1D) (Zheng et al., 2002b; Petroski and Deshaies, 2005; Duda et al., 2008). Cand1 Accelerates SCFFbxw7 Disassembly F box protein-Skp1 and Cand1 antagonize each others binding to Cul1-Rbx1 (Liu et al., 2002; Zheng et al., 2002a; Goldenberg et al., 2004; Bornstein et al., 2006; Duda et al., 2008; Siergiejuk et al., 2009; Lee et al., 2011). To test whether this is a general property, we prepared cell lysate from 293 cells that expressed
Figure 2. Cand1 Actively Removes Fbxw7Skp1 from Cul1 by Altering Off Rate
(A) As in Figures 1A and 1D except with the addition of 100 nM Cand1. (B) As in (A), except using neddylated CFPCul1. (C) The change in donor uorescence versus time in a stopped ow apparatus upon addition of 150 nM Cand1 to 50 nM CFPCul1-Rbx1 preincubated with 50 nM Fbxw7TAMRA-Skp1. (D) The single exponential observed rates of SCF disassembly for various Cand1 concentrations mixed with 5 nM CFPCul1-Rbx1 or 5 nM neddylated CFPCul1-Rbx1 preincubated with 5 nM Fbxw7TAMRA-Skp1. Chase indicates 700 nM Skp2Skp1. Error bars, SD, n R 3. (E) As in Figure 1E except with 150 nM or 300 nM Cand1 and 700 nM Skp2-Skp1 chase mixed with 70 nM neddylated CFPCul1 preincubated with 70 nM Fbxw7TAMRA-Skp1. Error bars: range of values, n = 2. (F) As in Figure 1C, except with 150 nM Cand1 preincubated with 5 nM CFPCul1-Rbx1. Error bars, SD, n R 3. See also Figure S2.
F
actively promoted the disassembly of the SCFFbxw7 complex. To directly test this hypothesis, we measured the loss of FRET in real time upon addition of Cand1 to a preformed complex in a stopped ow apparatus (Figure 2C). Titration of Cand1 revealed increasingly rapid rates of SCFFbxw7 dissociation that followed saturation kinetics with a maximum rate of 1.3 s1 and a half maximal concentration (KM) of 26 nM (Figure 2D). To eliminate interference from reassociation of Fbxw7TAMRA-Skp1 with CFPCul1-Rbx1, we repeated our measurements with unlabeled Skp2-Skp1 competitor in the reaction (Figure 2D). The maximal rate of Cand1-dependent dissociation remained unchanged while the KM increased to 53 nM (Figure 2D). In agreement with previous results, SCFFbxw7 formed with neddylated Cul1 showed no Cand1-induced dissociation over 30 s (Figure 2D). However, when the reactions were allowed to proceed for hours, we observed that saturating Cand1 accelerated disassembly of neddylated SCFFbxw7 by 45-fold (Figure 2E). Four main points arise from this analysis. First, the KM of 53 nM sets an upper limit on the KD between Cand1 and CFPCul1Rbx1. Second, the saturation kinetics reveal the existence of a transient complex that contained Cand1, CFPCul1-Rbx1, and Fbxw7TAMRA-Skp1. Third, the maximal observed rate of 1.3 s1 represents the rate of Fbxw7TAMRA-Skp1 dissociation from the transient complex, which is accelerated by greater than one-million-fold compared to the spontaneous dissociation of SCFFbxw7. Finally, neddylation of Cul1 attenuated the effect of Cand1 by 30,000-fold. These observations are reminiscent of a recent analysis that shows that binding of IkBa to NF-kB both slows the association of NF-kB with DNA and causes a signicant increase in the
tetracycline-inducible, FLAG-tagged Cul1 and were substantially depleted of Cand1 by stable expression of a lentiviral shRNA construct (Tet-FLAGCul1 Cand1kd cells). SCF complexes should be stable in this lysate due to the near absence of Cand1. This lysate was either mock treated or supplemented with a large excess of pure Cand1. Addition of Cand1 resulted in reduced recovery of 20 of 21 F box proteins observed in FLAG Cul1 immunoprecipitates, whereas the level of Rbx1 association was not signicantly affected (Figure S2). To measure directly Cand1s effect on SCF assembly in the absence of confounding factors that might be present in cell lysates, we added two-fold molar excess Cand1 to preformed pure SCFFbxw7 assembled from Fbxw7TAMRA-Skp1 and CFPCul1-Rbx1. We observed a signicant reduction in FRET after 5 min, indicating that Cand1 interfered with this assemblage (Figure 2A). This observation was not an artifact of our FRET assay because when we repeated these experiments with neddylated CFP Cul1-Rbx1, Cand1s effect was eliminated (Figure 2B). The short time span used for Figure 2A is in direct contrast to the slow dissociation of Fbxw7TAMRA in the presence of excess unlabeled Skp2-Skp1 observed in Figure 1E, suggesting that Cand1 was not a conventional competitive inhibitor that trapped Cul1-Rbx1 as it dissociated from Fbxw7-Skp1 but instead
Figure 3. F Box Proteins Rapidly Remove Cand1 from Cul1

(A) GST-Rbx1-Cul1-Cand1TAMRA (100 nM) was supplemented with 1 mM Cand1. At indicated times, aliquots were removed and incubated with glutathione resin for 15 min. Resin-associated proteins were fractionated by SDS-PAGE and detected by uorography. (B) The ratio of released Cand1TAMRA to total Cand1 over time was t to a single exponential giving koff of 1.2 3 105 s1. Error bars, SD, n = 3. (C) GST-Rbx1-Cul1-Cand1TAMRA (100 nM) preincubated with glutathione resin was supplemented with buffer or 1 mM of indicated proteins. Bound and released proteins were collected at indicated times and distribution of Cand1TAMRA was evaluated as in (A). (D) Summary of the rates measured here. Transient complexes are in brackets. See also Figure S3.
rate of dissociation of NF-kB from DNA (Bergqvist et al., 2009). Thus, it seemed likely that Cand1s dramatic effect on the dissociation rate of SCFFbxw7 would also be accompanied by a decrease in the rate of assembly of SCFFbxw7. However, in the presence of saturating Cand1, we measured an association rate of 2.2 3 106 M1 s1 for Fbxw7TAMRA-Skp1 and CFPCul1Rbx1 (Figure 2F). This is equivalent (within the error of our measurements) to the association rate without Cand1 (Figure 1C). As expected, when CFPCul1 was modied with Nedd8, Cand1 also had no effect on the association rate (Figure 2F). The above result has profound implications for building a model of Cand1s function. First, Cand1 is not a competitive inhibitor. Whereas competitive inhibitors decrease kon and have no effect on koff, Cand1 had no effect on kon but increased koff. Second, unlike allosteric inhibitors, such as IkBa, that decrease the association rate of their targets in addition to increasing the dissociation rate, Cand1 only affected the dissociation rate of SCFFbxw7. This specic type of behavior is found in the mechanism of guanine nucleotide exchange factors (GEFs) that actively promote the exchange of guanosine triphosphate (GTP) for GDP from target GTPases (Klebe et al., 1995; Goody and Hofmann-Goody, 2002; Guo et al., 2005). In these cases, GEF binding to its target GTPase increases the dissociation rate of guanine nucleotides yet does not inhibit their association rate. This is accomplished by toggling between stable GEFGTPase and GTPase-nucleotide complexes through a transient GEF-GTPase-nucleotide ternary complex in which the dissociation rates of both GEF and nucleotide are dramatically increased relative to their stable binary complexes (Klebe et al., 1995; Goody and Hofmann-Goody, 2002; Guo et al., 2005). Our results thus far conform to this model. Both Fbxw7-Skp1 and Cand1 formed tight complexes with Cul1-Rbx1 and there exists a transient ternary intermediate that exhibited a greatly increased rate of Fbxw7-Skp1 dissociation from Cul1-Rbx1.
F Box Proteins Remove Cand1 from Cul1 If Cand1 conforms to the behavior of GEFs, Cand1s dissociation rate from Cul1-Rbx1 should be very slow but should increase dramatically in the presence of F box protein-Skp1 complexes. To measure this, we rst developed a dissociation assay in which we monitored competitive displacement of Cand1TAMRA from GSTRbx1-Cul1 in the presence of 10-fold excess of unlabeled Cand1. Displacement was monitored either by native gel electrophoresis (Figures S3A and S3B) or rapid pull-down of GST on glutathione resin (Figure 3A). This protocol revealed that spontaneous dissociation of Cand1TAMRA from Cul1 was extremely slow, with a rate of 1.2 3 105 s1 (Figure 3B). This explains why Ubc12 is unable to conjugate Nedd8 to Cul1 that is bound to Cand1 (Goldenberg et al., 2004; Bornstein et al., 2006; Duda et al., 2008; Siergiejuk et al., 2009). Strikingly, this rate was greatly accelerated in the presence of Fbxw7-Skp1, such that displacement of Cand1TAMRA was largely completed within 5 min (Figure 3C). This key result indicates that Fbxw7-Skp1 is sufcient to displace Cand1 from Cul1, without requiring assistance from the Nedd8 conjugation pathway. This is consistent with reports that loss of Nedd8 conjugation activity has very little effect on the steady-state repertoire of SCF ubiquitin ligases (Bennett et al., 2010; Lee et al., 2011). A similar enhancement of Cand1 displacement by three different F box protein-Skp1 complexes was observed using an indirect assay wherein removal of Cand1 was evidenced by the ability to conjugate Nedd8 to the liberated Cul1 (Figures S3C and S3D). Displacement of Cand1 by Fbxw7-Skp1 was very specic, because Cand1 was not displaced by either Skp1 alone nor Fbxw7 bound to a mutant of Skp1 (Skp1DD) used for crystallography that lacks two loops that would be expected to clash with Cand1 (Figures 3C and S3B). Importantly, Skp1DD was fully able to sustain ubiquitylation, indicating that it was competent to bind Cul1 (Figure S3E). Combining the results above with the data gathered so far, we constructed a kinetic framework similar to that of GEFs (Klebe
SCF complexes given that: (1) the main interaction of F box proteins with Cul1 occurs through Skp1; (2) three different F box proteins rapidly evicted Cand1 from Cul1-Rbx1 (Figure S3D); and (3) Cand1 was able to dislodge 20 of 21 F box proteins from FLAGCul1 in cell lysates (Figure S2). Cand1 Functions as an F Box Protein Exchange Factor Given that Cand1 can disrupt multiple SCF complexes in a cell lysate, we wondered if Cand1 can promote exchange of one Cul1-bound F box protein for another. To test this directly, we measured the effect of Cand1 on the ability of Fbxw7TAMRASkp1 to gain access to Cul1 sequestered into complexes with b-TrCP-Skp1. As expected, preincubation of b-TrCPSkp1 and CFPCul1-Rbx1 diminished the observed association rate of 210 nM Fbxw7TAMRA-Skp1 with CFPCul1-Rbx1 from 0.9 s1 to 5 3 105 s1, a reduction of 18,000-fold (Figure 4A). Remarkably, addition of 150 nM Cand1 to this assay increased the observed association rate to 0.07 s1, a 1,400-fold rescue (Figure 4B). Thus, Cand1 greatly reduced the timescale with which Cul1-Rbx1 equilibrates with the total population of b-TrCP-Skp1 and Fbxw7-Skp1. To determine whether SCFFbxw7 assembled in the presence of Cand1 was active, we measured ubiquitylation of a cyclin E peptide (CycE) that serves as an SCFFbxw7 substrate (Hao et al., 2007; Pierce et al., 2009). Addition of b-TrCP-Skp1 to a preformed SCFFbxw7 did not affect the rate of ubiquitylation of CycE (Figure 4C). However, switching the order in which b-TrCP-Skp1 and Fbxw7-Skp1 were incubated with Cul1-Rbx1 effectively inhibited the ubiquitylation of CycE because b-TrCP-Skp1 formed stable complexes with Cul1, thereby denying access to CycE-Fbxw7-Skp1. Under these conditions, addition of Cand1 potently stimulated ubiquitylation of CycE (Figure 4C). The same was true if Cul1-Skp1 was preassembled with Skp2-Skp1 (Figure S3F). To our knowledge, all prior in vitro experiments implicated Cand1 as an inhibitor of SCF ubiquitin ligase activity. However, the design of our assaywhich more faithfully mimics what happens in vivo as new F box proteins are being synthesizedhighlights the ability of Cand1 to activate an SCF complex by enabling the new F box protein to gain access to Cul1 assembled into pre-existing SCF complexes. Cand1 Modulates the SCF Repertoire in Cells Consistent with the exchange activity observed with puried components, addition of excess puried b-TrCP-Skp1 to lysates of Tet-FLAGCul1 cells substantially reduced the recovery of 13 out of 15 endogenous F box proteins detected in anti-FLAG immunoprecipitates (Figure S4). However, this effect was significantly attenuated for 11 of these F box proteins by shRNAmediated depletion of Cand1. In contrast, excess b-TrCP-Skp1 had no effect on recovery of Rbx1 regardless of the presence or absence of Cand1. Together with our biochemical data on puried proteins, these results point toward a general ability of Cand1 to act as a protein exchange factor that equilibrates Cul1-Rbx1 with the available pool of F box protein-Skp1 complexes. To address whether Cand1s exchange activity inuences SCF complexes in cells, we quantied the steady state and dynamic nature of the Cul1-associated proteome in Tet-FLAGCul1 cells with or without Cand1 depletion. To evaluate
Figure 4. Cand1 Functions as an F Box Protein Exchange Factor

(A) Fluorescence emission at 478 nm versus time after addition of 210 nM Fbxw7TAMRA-Skp1 to 70 nM CFPCul1 preincubated with 70 nM b-TrCP-Skp1. A single exponential t gave koff of 5 3 105 s1. Error bars: range of values, n = 2. (B) The change in donor uorescence versus time in a stopped ow apparatus upon addition of 150 nM Cand1 to 70 nM CFPCul1-Rbx1 preincubated rst with 70 nM b-TrCP-Skp1 and second with 210 nM Fbxw7TAMRA-Skp1. (C) Cul1-Rbx1 (150 nM) was preincubated with 500 nM Fbxw7-Skp1 (lanes 16) or 660 nM b-TrCP-Skp1 (lanes 718) for 5 min, followed by addition of 600 nM radiolabeled cycE peptide substrate and either 660 nM b-TrCP-Skp1 (lanes 16) or 500 nM Fbxw7-Skp1 (lanes 718). Either buffer (lanes 112) or 200 nM Cand1 (lanes 1318) were then added, and reactions were incubated an additional 5 min prior to initiation of an ubiquitylation assay (all lanes) by supplementation of all lanes with 60 mM ubiquitin, 1 mM ubiquitin E1, and 10 mM Cdc34b. See also Figure S4.
et al., 1995; Goody and Hofmann-Goody, 2002; Guo et al., 2005) in which a transient ternary species comprising Fbxw7-Skp1, Cul1-Rbx1, and Cand1 rapidly collapses into subcomplexes that contain Cul1-Rbx1 bound by either Cand1 or Fbxw7-Skp1 (Figure 3D). We can build a similar cycle for neddylated Cul1Rbx1 in which Cand1 rapidly dissociates from the ternary species, yet Fbxw7-Skp1 slowly dissociates form the ternary species (Figure 3D). Our observations on Skp1DD, coupled with the report that deletion of a short b-hairpin from Cand1 allows formation of a stable Cand1-Skp1DD-Cul1-Rbx1 complex (Goldenberg et al., 2004) leads us to propose that the transient ternary complex we observed in Figure 2D is a high-energy intermediate created by clashes involving the exible acidic loops in Skp1 and the b hairpin in Cand1, enabling rapid dissociation of either Cand1 or F box protein-Skp1 from Cul1-Rbx1. In addition, we suggest that the interaction dynamics between Fbxw7-Skp1 and Cand1 should apply to most or all F box proteins that form
Figure 5. Cellular Cand1 Shapes SteadyState and Dynamic SCF Landscape
(A) Tet-FLAGCul controlkd and Tet-FLAGCul1 Cand1kd cells (kd refers to knockdown) grown in medium with isotopically light or heavy lysine plus arginine, respectively, were induced with 1 mg/ml tetracycline for 1 hr and lysed in 1 mM MLN4924 and 2 mM o-phenanthroline 24 hr later. Two experiments were performed according to this protocol. In the rst experiment, we used pseudoMRM mass spectrometry to measure the relative amounts of 14 observable F box proteins in total cell lysate from Cand1-depleted and control cells (white bars). In the second experiment, we retrieved FLAGCul1 and measured the relative amounts of 34 F box proteins in the immunoprecipitates (black bars). All isotopic ratios were normalized to FLAGCul1s (0.94), which was set to 1.0. For both experiments, results represent the ratio Cand1kd:controlkd of each protein in anti-FLAG IP measured by mass spectrometry. Each protein had R two peptides. Error bars B C represent standard errors of overall protein group ratios, calculated from bootstrap analysis of two biological replicates (the second replicate was performed as a label swap). Abundance changes in Cul1 IP for all the proteins listed in Figure 5A except for Fbxo8, Fbxw9, and Fbxw4 achieved p values < 0.05. Fbxo44a and b correspond to IPI00647771 and IPI00414844, respectively. Statistical analysis is provided in Table S1. (B) Immunoblot validation with indicated antibodies of results in (A). (C) The same cells used in (A) were transfected D with a plasmid that encodes FLAGCry1. Forty-eight hours later, a chase was initiated by addition of 40 mg/ml cycloheximide. Cells were harvested at the indicated times and their content of FLAGCry1 and GAPDH was evaluated by SDS-PAGE and immunoblotting (left), and quantied (right). (D) The same cells used in (A) and grown in isotopically light lysine plus arginine were induced with 1 mg/ml tetracycline for 1 hr at t = 0 hr, treated with 5 mM epoxomicin at t = 48 hr, shifted to isotopically heavy lysine plus arginine at t = 49 hr, and lysed at t = 61 hr in 1 mM MLN4924 and 2 mM o-phenanthroline. Two experiments were performed according to this protocol. In the rst experiment shown here, we used data-dependent mass spectrometry to discover and measure the fraction of F box proteins in FLAGCul1 IPs that was heavy (i.e. made in the 12 hr prior to lysis). In the second experiment (Figure S5B), we used pseudoMRM to target nine F box proteins (italicized) and measure the fraction of heavy-labeled species in total cell lysate from Cand1-depleted and control cells. Each protein had R2 peptides. Error bars represent standard errors of overall protein group ratios, calculated from bootstrap analysis of two biological replicates. The F box proteins shown to the left of the dotted line are those for which the different association observed in control and Cand1kd cells achieved a p value < 0.05. See also Figure S5 and Table S2.
the steady-state architecture of the SCF proteome, we grew control cells in medium formulated with isotopically light lysine and arginine (light medium) and Cand1-depleted cells in medium formulated with isotopically heavy lysine and arginine (heavy medium). Both cultures were pulsed with tetracycline for 1 hr at t = 0 hr to induce transient synthesis of FLAGCul1. After t = 25 hr FLAGCul1 immunoprecipitates were prepared, mixed,
and analyzed by quantitative mass spectrometry. Analysis of isotopic ratios revealed that Cand1 depletion had no effect on Cul1 levels but exerted far-reaching effects on the SCF repertoire, with 14 of 34 SCF complexes increasing by R2-fold, and 2 decreasing by R1.5-fold (Figure 5A, black bars). Meanwhile, Cul1-bound Skp1 was increased 1.7-fold in Cand1-decient cells. The net assembly of SCF complexes in Cand1-depleted
cells is in agreement with the relative abundances of Cul1 (302 nM) and Cand1 (390 nM) in 293 cells (Bennett et al., 2010) and our biophysical data that excess Cand1 causes a net disassembly of SCFFbxw7 (Figure 2A). In addition, the levels of Cul1associated Nedd8 and COP9 signalosome (CSN) were increased by 2.4 and 1.5-fold, respectively. These changes are consistent with the increased overall assembly of SCF complexes in Cand1-depleted cells, and the ability of F box proteins to stabilize binding of CSN (Enchev et al., 2012) while suppressing deneddylation of Cul1 (Emberley et al., 2012; Enchev et al., 2012). The effect of Cand1 on F box proteins was specic because the tightly bound Cul1 partner Rbx1 exhibited very little change. These and other changes observed in Figure 5A were validated qualitatively by western blotting (Figure 5B). Importantly, mass-spectrometry-based quantication of peptides diagnostic for a select group of F box proteins revealed that the change observed in the steady-state repertoire of SCF complexes in Cand1-depleted cells was not a simple reection of altered abundance of F box proteins in total cell lysate (Figure 5A, white bars). To evaluate the relationship between altered abundance of SCF complexes and their role in protein degradation, we examined the degradation of an Fbxl3 substrate, cryptochrome Cry1 (Busino et al., 2007; Siepka et al., 2007). Whereas SCFFbxl3 was more abundant in Cand1depleted cells (Figure 5A), the rate of degradation of Cry1 was reduced (Figure 5C) as was the association between Cry1 and Cul1 (Figure S5A). To connect our in vitro observations with what happens in cells, we sought to evaluate SCF dynamics in vivo. Control and Cand1-depleted cells growing in light medium were pulsed with tetracycline for 1 hr at t = 0 hr to induce transient synthesis of FLAGCul1. At t = 49 hr, the cells were pulsed with heavy medium in the presence of the proteasome inhibitor epoxomicin to suppress apparent F box protein exchange due to degradation. At t = 61 hr, FLAGCul1 immunoprecipitates were prepared and analyzed by quantitative mass spectrometry to determine the fraction of each bound protein that was produced during the 12 hr pulse-label. In a parallel experiment, we monitored the heavy:light ratios in total cell lysate of peptides diagnostic for 12 different F box proteins to evaluate the rate at which newly synthesized forms of these proteins accumulated. As shown in Figure 5D, several newly synthesized F box proteins including Fbxw11, Fbxo11, and Fbxo21 exhibited reduced incorporation into the pool of pre-existing Cul1 in the absence of Cand1, consistent with Cand1 being an F box protein exchange factor. Importantly, the isotopic ratios for F box proteins in total cell lysate (including the three mentioned above) were not affected by the presence or absence of Cand1 (Figure S5B), indicating that low penetration of these pulse-labeled F box proteins into the Cul1-bound pool of Cand1-depleted cells was not due to a reduced rate of synthesis. DISCUSSION Cand1 Is an F Box Protein Exchange Factor and Modulates the Cellular Repertoire of SCF Complexes Here we establish the rst kinetic framework for the dynamic assembly of SCF complexes. Prior biochemical studies sug212 Cell 153, 206215, March 28, 2013 2013 Elsevier Inc.
gested that Cand1 is an inhibitor of SCF complexes, whereas genetic studies indicated that it promotes SCF function. Our results resolve this apparent paradox. We show that Cand1 can unambiguously stimulate SCF activity in vitro by enabling an F box protein-Skp1 complex to access Cul1 that was previously occupied by a different F box protein-Skp1 complex, and that Cand1 promotes assembly in vivo of new F box proteins with pre-existing Cul1 molecules. We conclude that Cand1 serves as an exchange factor for F box protein-Skp1 complexes and conforms to the scheme originally proposed for GEFs (Klebe et al., 1995; Goody and Hofmann-Goody, 2002; Guo et al., 2005). However, unlike GEFs that exchange GDP for GTP, Cand1 promotes exchange of multiple F box proteins on the Cul1 scaffold. Because Cand1 can bind other cullins (Lo and Hannink, 2006; Bosu et al., 2010; Chua et al., 2011) and inuences ubiquitylation or degradation of multiple CRL substrates in vivo (Zheng et al., 2002a; Chuang et al., 2004; Feng et al., 2004; Lo and Hannink, 2006; Zhang et al., 2008; Bosu et al., 2010; Kim et al., 2010), we suggest that it serves as an exchange factor for CRL adaptors in general. In anticipation of a widespread role for Cand1 exchange activity in CRL biology and by analogy to guanine nucleotide exchange factors (GEFs), we suggest that it be referred to as a substrate receptor exchange factor (SREF). A perplexing feature of prior genetic studies is that Cand1 deciency causes a partial reduction of function for some SCF complexes but not others (Chuang et al., 2004; Feng et al., 2004; Zhang et al., 2008; Bosu et al., 2010; Kim et al., 2010). We suggest that the impact of Cand1 on the steady-state distribution of SCF complexeswith some complexes showing large changes in level whereas other complexes exhibit minimal perturbation (Figure 5A)may partially account for this heretofore unexplained behavior. Another puzzling observation made in both human (Lo and Hannink, 2006) and Arabidopsis (Chuang et al., 2004) Cand1-decient cells is that accumulation of a given CRL complex paradoxically correlates with stabilization of its substratea phenomenon also observed here. Our results suggest the interesting possibility that Cand1 biases the assembly of SCF complexes to favor F box proteins for which substrates are available. Substrate Sculpting of the CRL Repertoire: An Hypothesis Based on the recent observations that substrates and other ligands can greatly reduce binding and deneddylation of CRLs by CSN (Fischer et al., 2011; Emberley et al., 2012; Enchev et al., 2012), we propose the hypothesis shown in Figure 6, which provides a potential mechanism for how the cellular repertoire of CRL complexes could be optimized to match substrate demand. In our model, which builds upon previous proposals (Cope and Deshaies, 2003; Schmidt et al., 2009), the exchange activity of Cand1 coupled with cycles of Nedd8 conjugation and deconjugation enables CRL complexes to toggle between two radically different statesthe Cand1 exchange regime and the Nedd8 stable state. When a cullin is conjugated with Nedd8, it can have extraordinary (subpicomolar) afnity for its adaptorbound substrate receptor (SR), which binds in a manner that is essentially irreversible (t1/2 9 days for Fbxw7-Skp1). Therefore, we envision that a CRL exists in a stable, active state when it is
the signicant phenotypic consequences of Cand1 mutation (Chuang et al., 2004; Feng et al., 2004; Zhang et al., 2008; Bosu et al., 2010; Kim et al., 2010; Helmstaedt et al., 2011), it is apparent that ongoing synthesis and degradation of adaptor-SR modules (Bennett et al., 2010) or other mechanisms of SCF disassembly (Yen et al., 2012) by themselves do not sufce to sustain a fully functional CRL network. Although other examples of factors that actively promote complex disassembly have been described (Klebe et al., 1995; Goody and Hofmann-Goody, 2002; Guo et al., 2005; Bergqvist et al., 2009; Chan et al., 2009), to our knowledge Cand1 is the rst example of a factor that has the potential to promote equilibration of a protein scaffold with a large number of interacting partners. We suggest that protein exchange factors that work analogously to Cand1 may play important roles in processes such as DNA replication, transcription, mRNA splicing, and vesicle trafcking that rely on protein machines that engage in transactions notable for their speed and afnity.
Figure 6. Hypothesis for Control of CRL Assembly by Substrate, Cand1, and Nedd8
Rapid exchange of multiple CRL adaptor-bound substrate receptors occurs in the Cand1 exchange regime through the formation and decay of transient ternary complexes shown in brackets. Cand1 and adaptor are drawn as deformed in these complexes, to emphasize the proposal that they clash sterically, yielding an unstable state. In the presence of substrates, CRLs that pass through an intermediate state become neddylated and enter a stable state where ubiquitylation of substrates occurs. Loss of substrates facilitates recruitment of CSN, removal of Nedd8, and a return to the exchange regime effected by Cand1. EXPERIMENTAL PROCEDURES FRET Assay Fluorimeter scans were performed on FluoroLog-3 (Jobin Yvon) in a buffer containing 30 mM Tris (pH 7.6), 100 mM NaCl, 0.5 mM DTT, and 1 mg/ml Ovalbumin (Sigma) in a volume of 250 ml. Mixtures were excited at 430 nm and the emissions were scanned from 450 nm to 650 nm. Stopped ow reactions were performed on a Kintek stopped ow machine in the same buffer as the uorimeter scans. Ubiquitylation Assay CycE was incubated with [g-32P] ATP (132 nM) and Protein Kinase A for 45 min at 30 C to make radiolabeled CycE. Ubiquitylation reactions contained ATP (2 mM), ubiquitin (60 mM), ubiquitin E1 (1 mM), Cdc34b (10 mM), Cul1-Rbx1 (150 nM), and Fbxw7-Skp1 (varying concentrations). Additional proteins were included as mentioned in the text. Reactions were performed and quenched in buffers previously described for ubiquitylation assays (Pierce et al., 2009). Reactions were analyzed by running on 16% gels, drying, and quantifying with a phosphor screen (Molecular Devices). Cul1-Cand1 Dissociation Assay Cul1 with N-terminal tagged GST-Rbx1 (0.1 mM) was preincubated with Cand1TAMRA (0.1 mM) for 10 min and captured on glutathione sepharose 4B resin. Aliquots of resin were transferred to Micro Bio-Spin columns (BioRad), resuspended in 1 mM of the indicated proteins, and incubated for 15 s or 5 min. Reactions were terminated by separation of beads and supernatant by centrifugation, and equivalent portions of each were fractionated by SDSPAGE. Gels were scanned by a Typhoon scanner to quantify Cand1TAMRA. To analyze the dissociation rate of the Cul1-Cand1 complex, at various time points following the addition of 1 mM Cand1 to 0.1 mM GST-Rbx1-Cul1Cand1TAMRA, an aliquot was withdrawn and incubated with glutathione sepharose 4B resin for 15 min and processed for SDS-PAGE analysis and uorography. The gels were then stained by SilverQuest staining kit (Invitrogen) to detect the total Cand1 bands. The intensity of Cand1TAMRA and total Cand1 band was measured with ImageJ (NIH), and the intensity of Cand1TAMRA was normalized by the intensity of total Cand1 at each time point. Kinetic Analysis Regressions were performed in Matlab with the exception of Figure 3B, which was generated in Prism. Mass Spectrometric Analyses All data-dependent liquid chromatography-mass spectrometry and data analyses for Cul1-bound proteins were performed as described previously (Lee et al., 2011) with the following modications. In experiments where we
saturated with substrate, which occludes CSN binding. The Nedd8-conjugated CRL ubiquitylates substrate and can recruit downstream factors involved in substrate degradation (Bandau et al., 2012; den Besten et al., 2012). Once substrate is depleted, the ability of CSN to bind the CRL and remove Nedd8 is enhanced. In this metastable intermediate state, the complex can either bind a new substrate and become reactivated by Nedd8 conjugation to return to the stable state, or it can bind Cand1 and enter the exchange regime, resulting in up to a one-million-fold increase in dissociation rate of adaptorSR. The resulting Cand1-cullin-Rbx complex rapidly assembles with any of the available adaptor-SR complexes to form an unstable ternary intermediate that promptly decays to regenerate Cand1-cullin-Rbx or yield a new CRL complex. Neddylation of the latter species, which is stabilized by substrate, completes the cycle. We envision that substrates, by shielding CRL complexes from the actions of CSN and Cand1, help to sculpt the cellular repertoire of CRL complexes. Other factors are likely to contribute as well, including the rate of synthesis and degradation of substrate receptor proteins, as well as posttranslational modications and localization controls that modulate the access of CRL complexes to the Nedd8-Cand1 cycle. An implication of our model is that in Cand1-decient cells, the assembly state of CRLs would become uncoupled from substrate demand, such that the cullin could potentially become tied up in superuous and nonproductive CRL complexes. This situation may be tolerated in cells with constitutively high turnover of F box and other substrate receptor proteins, but given
tested the effect of excess b-TrCP or Cand1 on SCF complex composition, puried b-TrCP or Cand1 was added to the lysate for 2 hr at 23 C, followed by immunoprecipitation for 15 min at the same temperature. To calculate the overall protein group ratio for an experiment, median evidence ratios were rst calculated for each biological replicate, and overall ratio was calculated from the mean of the biological replicates. Standard errors of the overall protein group ratios were calculated using bootstrap analysis. PseudoMRM (Greco et al., 2010), also referred to as targeted peptide monitoring (Sandhu et al., 2008; Hewel et al., 2010) or peptide ion monitoring (Kulasingam et al., 2008) was used for the quantication of the selected peptides of F box proteins in global lysates. SUPPLEMENTAL INFORMATION Supplemental Information includes Extended Experimental Procedures, ve gures, and ve tables and can be found with this article online at http://dx. doi.org/10.1016/j.cell.2013.02.024. ACKNOWLEDGMENTS We thank J. Vielmetter of the Caltech Protein Expression Facility for providing Fbxw7-Skp1, b-TrCP-Skp1, and ubiquitin E1. We thank B. Schulman, L. Busino, M. Pagano, F. Bassermann, O. Schneewind, A. Saha, and W. den Besten for gifts of reagents. We thank G. Smith for assistance with mass spectrometry analyses. We thank all the members of the Deshaies and Shan lab for support and helpful discussions. We thank D. Wolf and T. Kurz for communicating unpublished results. N.W.P. was supported by the Gordon Ross Fellowship, and an NIH Training Grant. J.E.L. was supported by the Ruth L. Kirschstein NRSA Fellowship (CA138126) from the NIH. R.J.D. is an Investigator of the HHMI. This work was supported in part by NIH GM065997 to R.J.D. The Proteome Exploration lab is supported in part by grants from the Gordon and Betty Moore Foundation and the Beckman Institute and an instrumentation grant from NIH (10565784). Received: December 10, 2012 Revised: January 24, 2013 Accepted: February 12, 2013 Published: February 28, 2013 REFERENCES Bandau, S., Knebel, A., Gage, Z.O., Wood, N.T., and Alexandru, G. (2012). UBXN7 docks on neddylated cullin complexes using its UIM motif and causes HIF1a accumulation. BMC Biol. 10, 36. Bennett, E.J., Rush, J., Gygi, S.P., and Harper, J.W. (2010). Dynamics of cullinRING ubiquitin ligase network revealed by systematic quantitative proteomics. Cell 143, 951965. Bergqvist, S., Alverdi, V., Mengel, B., Hoffmann, A., Ghosh, G., and Komives, E.A. (2009). Kinetic enhancement of NF-kappaBxDNA dissociation by IkappaBalpha. Proc. Natl. Acad. Sci. USA 106, 1932819333. Bornstein, G., Ganoth, D., and Hershko, A. (2006). Regulation of neddylation and deneddylation of cullin1 in SCFSkp2 ubiquitin ligase by F-box protein and substrate. Proc. Natl. Acad. Sci. USA 103, 1151511520. Bosu, D.R., Feng, H., Min, K., Kim, Y., Wallenfang, M.R., and Kipreos, E.T. (2010). C. elegans CAND-1 regulates cullin neddylation, cell proliferation and morphogenesis in specic tissues. Dev. Biol. 346, 113126. Busino, L., Bassermann, F., Maiolica, A., Lee, C., Nolan, P.M., Godinho, S.I.H., Draetta, G.F., and Pagano, M. (2007). SCFFbxl3 controls the oscillation of the circadian clock by directing the degradation of cryptochrome proteins. Science 316, 900904. Chan, C., Beltzner, C.C., and Pollard, T.D. (2009). Colin dissociates Arp2/3 complex and branches from actin laments. Curr. Biol. 19, 537545. Chua, Y.S., Boh, B.K., Ponyeam, W., and Hagen, T. (2011). Regulation of cullin RING E3 ubiquitin ligases by CAND1 in vivo. PLoS ONE 6, e16071.
Chuang, H.-W., Zhang, W., and Gray, W.M. (2004). Arabidopsis ETA2, an apparent ortholog of the human cullin-interacting protein CAND1, is required for auxin responses mediated by the SCF(TIR1) ubiquitin ligase. Plant Cell 16, 18831897. Cope, G.A., and Deshaies, R.J. (2003). COP9 signalosome: a multifunctional regulator of SCF and other cullin-based ubiquitin ligases. Cell 114, 663671. den Besten, W., Verma, R., Kleiger, G., Oania, R.S., and Deshaies, R.J. (2012). NEDD8 links cullin-RING ubiquitin ligase function to the p97 pathway. Nat. Struct. Mol. Biol. 19, 511516, S1. Duda, D.M., Borg, L.A., Scott, D.C., Hunt, H.W., Hammel, M., and Schulman, B.A. (2008). Structural insights into NEDD8 activation of cullin-RING ligases: conformational control of conjugation. Cell 134, 9951006. Dye, B.T., and Schulman, B.A. (2007). Structural mechanisms underlying posttranslational modication by ubiquitin-like proteins. Annu. Rev. Biophys. Biomol. Struct. 36, 131150. Emberley, E.D., Mosadeghi, R., and Deshaies, R.J. (2012). Deconjugation of Nedd8 from Cul1 is directly regulated by Skp1-F-box and substrate, and the COP9 signalosome inhibits deneddylated SCF by a noncatalytic mechanism. J. Biol. Chem. 287, 2967929689. Enchev, R.I., Scott, D.C., da Fonseca, P.C.A., Schreiber, A., Monda, J.K., Schulman, B.A., Peter, M., and Morris, E.P. (2012). Structural basis for a reciprocal regulation between SCF and CSN. Cell Rep 2, 616627. Feng, S., Shen, Y., Sullivan, J.A., Rubio, V., Xiong, Y., Sun, T.-P., and Deng, X.W. (2004). Arabidopsis CAND1, an unmodied CUL1-interacting protein, is involved in multiple developmental pathways controlled by ubiquitin/proteasome-mediated protein Degradation. Plant Cell 16, 18701882. hm, K., Matsumoto, S., Lingaraju, G.M., Faty, M., Fischer, E.S., Scrima, A., Bo Yasuda, T., Cavadini, S., Wakasugi, M., Hanaoka, F., et al. (2011). The molecular basis of CRL4DDB2/CSA ubiquitin ligase architecture, targeting, and activation. Cell 147, 10241039. Goldenberg, S.J., Cascio, T.C., Shumway, S.D., Garbutt, K.C., Liu, J., Xiong, Y., and Zheng, N. (2004). Structure of the Cand1-Cul1-Roc1 complex reveals regulatory mechanisms for the assembly of the multisubunit cullin-dependent ubiquitin ligases. Cell 119, 517528. Goody, R.S., and Hofmann-Goody, W. (2002). Exchange factors, effectors, GAPs and motor proteins: common thermodynamic and kinetic principles for different functions. Eur. Biophys. J. 31, 268274. Greco, T.M., Seeholzer, S.H., Mak, A., Spruce, L., and Ischiropoulos, H. (2010). Quantitative mass spectrometry-based proteomics reveals the dynamic range of primary mouse astrocyte protein secretion. J. Proteome Res. 9, 27642774. Guo, Z., Ahmadian, M.R., and Goody, R.S. (2005). Guanine nucleotide exchange factors operate by a simple allosteric competitive mechanism. Biochemistry 44, 1542315429. Hao, B., Oehlmann, S., Sowa, M.E., Harper, J.W., and Pavletich, N.P. (2007). Structure of a Fbw7-Skp1-cyclin E complex: multisite-phosphorylated substrate recognition by SCF ubiquitin ligases. Mol. Cell 26, 131143. Helmstaedt, K., Schwier, E.U., Christmann, M., Nahlik, K., Westermann, M., Harting, R., Grond, S., Busch, S., and Braus, G.H. (2011). Recruitment of the inhibitor Cand1 to the cullin substrate adaptor site mediates interaction to the neddylation site. Mol. Biol. Cell 22, 153164. Hewel, J.A., Liu, J., Onishi, K., Fong, V., Chandran, S., Olsen, J.B., Pogoutse, O., Schutkowski, M., Wenschuh, H., Winkler, D.F.H., et al. (2010). Synthetic peptide arrays for pathway-level protein monitoring by liquid chromatography-tandem mass spectrometry. Mol. Cell. Proteomics 9, 24602473. Kim, S.-H., Kim, H.-J., Kim, S., and Yim, J. (2010). Drosophila Cand1 regulates Cullin3-dependent E3 ligases by affecting the neddylation of Cullin3 and by controlling the stability of Cullin3 and adaptor protein. Dev. Biol. 346, 247257. Klebe, C., Prinz, H., Wittinghofer, A., and Goody, R.S. (1995). The kinetic mechanism of Rannucleotide exchange catalyzed by RCC1. Biochemistry 34, 1254312552. Kulasingam, V., Smith, C.R., Batruch, I., Buckler, A., Jeffery, D.A., and Diamandis, E.P. (2008). Product ion monitoring assay for prostate-specic antigen in serum using a linear ion-trap. J. Proteome Res. 7, 640647.
Lee, J.E., Sweredoski, M.J., Graham, R.L.J., Kolawa, N.J., Smith, G.T., Hess, S., and Deshaies, R.J. (2011). The steady-state repertoire of human SCF ubiquitin ligase complexes does not require ongoing Nedd8 conjugation. Mol. Cell. Proteomics 10, M110, 006460. Liu, J., Furukawa, M., Matsumoto, T., and Xiong, Y. (2002). NEDD8 modication of CUL1 dissociates p120(CAND1), an inhibitor of CUL1-SKP1 binding and SCF ligases. Mol. Cell 10, 15111518. Lo, S.-C., and Hannink, M. (2006). CAND1-mediated substrate adaptor recycling is required for efcient repression of Nrf2 by Keap1. Mol. Cell. Biol. 26, 12351244. Petroski, M.D., and Deshaies, R.J. (2005). Function and regulation of cullinRING ubiquitin ligases. Nat. Rev. Mol. Cell Biol. 6, 920. Pierce, N.W., Kleiger, G., Shan, S.-O., and Deshaies, R.J. (2009). Detection of sequential polyubiquitylation on a millisecond timescale. Nature 462, 615619. Popp, M.W.-L., Antos, J.M., and Ploegh, H.L. (2009). Site-specic protein labeling via sortase-mediated transpeptidation. Curr. Protoc. Protein Sci. 56, 15.3.115.3.9. Proft, T. (2010). Sortase-mediated protein ligation: an emerging biotechnology tool for protein modication and immobilisation. Biotechnol. Lett. 32, 110. Saha, A., and Deshaies, R.J. (2008). Multimodal activation of the ubiquitin ligase SCF by Nedd8 conjugation. Mol. Cell 32, 2131. Sandhu, C., Hewel, J.A., Badis, G., Talukder, S., Liu, J., Hughes, T.R., and Emili, A. (2008). Evaluation of data-dependent versus targeted shotgun proteomic approaches for monitoring transcription factor expression in breast cancer. J. Proteome Res. 7, 15291541. Schmidt, M.W., McQuary, P.R., Wee, S., Hofmann, K., and Wolf, D.A. (2009). F-box-directed CRL complex assembly and regulation by the CSN and CAND1. Mol. Cell 35, 586597.
Siepka, S.M., Yoo, S.-H., Park, J., Song, W., Kumar, V., Hu, Y., Lee, C., and Takahashi, J.S. (2007). Circadian mutant Overtime reveals F-box protein FBXL3 regulation of cryptochrome and period gene expression. Cell 129, 10111023. Siergiejuk, E., Scott, D.C., Schulman, B.A., Hofmann, K., Kurz, T., and Peter, M. (2009). Cullin neddylation and substrate-adaptors counteract SCF inhibition by the CAND1-like protein Lag2 in Saccharomyces cerevisiae. EMBO J. 28, 38453856. Thrower, J.S., Hoffman, L., Rechsteiner, M., and Pickart, C.M. (2000). Recognition of the polyubiquitin proteolytic signal. EMBO J. 19, 94102. Wickliffe, K.E., Williamson, A., Meyer, H.-J., Kelly, A., and Rape, M. (2011). K11-linked ubiquitin chains as novel regulators of cell division. Trends Cell Biol. 21, 656663. Yen, J.L., Flick, K., Papagiannis, C.V., Mathur, R., Tyrrell, A., Ouni, I., Kaake, R.M., Huang, L., and Kaiser, P. (2012). Signal-induced disassembly of the SCF ubiquitin ligase complex by Cdc48/p97. Mol. Cell 48, 288297. l, L.D., and Gray, W.M. (2008). Zhang, W., Ito, H., Quint, M., Huang, H., Noe Genetic analysis of CAND1-CUL1 interactions in Arabidopsis supports a role for CAND1-mediated cycling of the SCFTIR1 complex. Proc. Natl. Acad. Sci. USA 105, 84708475. Zheng, J., Yang, X., Harrell, J.M., Ryzhikov, S., Shim, E.H., Lykke-Andersen, K., Wei, N., Sun, H., Kobayashi, R., and Zhang, H. (2002a). CAND1 binds to unneddylated CUL1 and regulates the formation of SCF ubiquitin E3 ligase complex. Mol. Cell 10, 15191526. Zheng, N., Schulman, B.A., Song, L., Miller, J.J., Jeffrey, P.D., Wang, P., Chu, C., Koepp, D.M., Elledge, S.J., Pagano, M., et al. (2002b). Structure of the Cul1-Rbx1-Skp1-F boxSkp2 SCF ubiquitin ligase complex. Nature 416, 703709.
Phospholipase C Hydrolyzes Perinuclear Phosphatidylinositol 4-Phosphate to Regulate Cardiac Hypertrophy

Lianghui Zhang,1,4 Sundeep Malik,1,4 Jinjiang Pang,3 Huan Wang,2,5 Keigan M. Park,1 David I. Yule,1 Burns C. Blaxall,1,3,6 and Alan V. Smrcka1,2,3,*
of Pharmacology and Physiology of Biochemistry and Biophysics 3Aab Institute of Cardiovascular Research University of Rochester School of Medicine and Dentistry, 601 Elmwood Avenue, Rochester, NY 14642, USA 4These authors contributed equally to this work 5Present address: Wistar Institute, Philadelphia, PA 19104, USA 6Present address: Cincinnati Childrens Hospital Medical Center, Cincinnati, OH 45229-3039, USA *Correspondence: alan_smrcka@urmc.rochester.edu http://dx.doi.org/10.1016/j.cell.2013.02.047
SUMMARY
Phospholipase C (PLC) is a multifunctional enzyme implicated in cardiovascular, pancreatic, and inammatory functions. Here we show that conditional deletion of PLC in mouse cardiac myocytes protects from stress-induced pathological hypertrophy. PLC small interfering RNA (siRNA) in ventricular myocytes decreases endothelin-1 (ET-1)-dependent elevation of nuclear calcium and activation of nuclear protein kinase D (PKD). PLC scaffolded to muscle-specic A kinase-anchoring protein (mAKAP), along with PKC and PKD, localizes these components at or near the nuclear envelope, and this complex is required for nuclear PKD activation. Phosphatidylinositol 4-phosphate (PI4P) is identied as a perinuclear substrate in the Golgi apparatus for mAKAPscaffolded PLC. We conclude that perinuclear PLC, scaffolded to mAKAP in cardiac myocytes, responds to hypertrophic stimuli to generate diacylglycerol (DAG) from PI4P in the Golgi apparatus, in close proximity to the nuclear envelope, to regulate activation of nuclear PKD and hypertrophic signaling pathways.
INTRODUCTION Heart failure is a major global health problem and a leading cause of mortality. In response to high blood pressure or other cardiac stress, the adult heart remodels and enlarges in what is thought to be an initial compensatory mechanism but that is maladaptive in the long term and leads to cardiac decompensation and heart failure. One major feature of cardiac remodeling is hypertrophic growth of cardiac myocytes, driven in part by increased levels of neurohumoral agonists such as norepinephrine and endothelin-1
(ET-1), acting on G protein-coupled receptors (GPCRs) in cardiac myocytes (Rockman et al., 2002). Many hypertrophic agonists couple to activation of phosphoinositide-specic phospholipase C (PLC) to stimulate phosphatidylinositol 4,5-bisphosphate (PIP2) hydrolysis and produce inositol 1,4,5 trisphosphate (IP3) and diacylglycerol (DAG), a key reaction involved in mediating cardiomyocyte hypertrophy. Gaq, a GTP-binding protein that directly activates PLC, is a major participant in the hypertrophic process in mice (DAngelo et al., 1997; Dorn and Brown, 1999), and a splice variant of PLCb, PLCb1b, is required for cardiomyocyte hypertrophy in vitro driven by a-adrenergic receptor activation (Filtz et al., 2009; Grubb et al., 2011). A novel isoform of PLC, phospholipase C (PLC), is downstream of both GPCRs and receptor tyrosine kinases by virtue of its ability to be regulated by small GTPases, including Ras, Rho, and Rap, and by heterotrimeric G protein bg subunits (Kelley et al., 2001; Smrcka et al., 2012). We recently demonstrated that hypertrophy of neonatal rat ventricular myocytes (NRVMs) driven by ET-1, norepinephrine, or isoproterenol (Iso) was inhibited by small interfering RNA (siRNA)-mediated knockdown of PLC (Zhang et al., 2011). Furthermore, it was shown that PLC scaffolds to muscle-specic A kinase-anchoring protein (mAKAP) at the nuclear envelope (NE), and disruption of this scaffolding interaction prevents development of agonist-induced hypertrophy (Zhang et al., 2011). This suggests that PLC generates second messengers at the NE that are required for hypertrophy. Canonical GPCR-mediated PIP2 hydrolysis by PLC occurs at the plasma membrane (PM), where PIP2 is preferentially enriched. PIP2 is not readily detectable in intracellular organelle membranes, and PI-dependent signaling processes in intracellular membranes have not been well characterized. A phosphoinositide cycle involving PIP2 hydrolysis is present in the nuclear matrix (Divecha et al., 1993; Keune et al., 2011; Ramazzotti et al., 2011). Turnover of nuclear PIP2 is not subject to regulation by GPCRs, despite the presence of PLCb in the nucleus of some cells, but IGF-1 has been shown to stimulate nuclear
PIP2 hydrolysis (Cocco et al., 1989; Divecha et al., 1991). Surprisingly, nuclear PIP2 is not associated with NE membranes but rather is in a poorly dened detergent-resistant structure in the matrix (Keune et al., 2011). With no apparent PIP2 substrate at the NE, a role for a PLC-hydrolytic activity associated with the NE is not obvious, although it is theoretically possible that an NE-scaffolded PLC could gain access to nuclear PIP2. Phosphatidylinositol 4-phosphate (PI4P) is an alternate substrate for puried PLC in vitro that is found in intracellular membranes, but PI4P has not been shown to be a native physiological substrate for PLC in cells. PI4P has been thought of primarily as a precursor for replenishment of PI4,5P2 during active receptor-stimulated PIP2 hydrolysis, but recent studies indicate that PI4P itself plays other important roles in cell function (Hammond et al., 2012). One of the key signaling events that leads to expression of hypertrophic genes is phosphorylation of histone deacetylase (HDAC5) and subsequent binding of phosphorylated HDAC to 14-3-3 proteins in the cytoplasm (Frey and Olson, 2003). Two key kinases that phosphorylate HDAC at the nucleus are calcium calmodulin-dependent protein kinase II (CamKII) and protein kinase D (PKD) (Frey and Olson, 2003; McKinsey, 2007). CamKII is regulated by calcium, and PKD is regulated by DAG and phosphorylation by protein kinase C (PKC) (Rozengurt, 2011; Steinberg, 2012), suggesting that Ca2+ and DAG levels need to be elevated locally at the nucleus to maintain their activities. How plasma-membrane GPCR-generated signals regulate these nuclear proteins has not been carefully examined. It has been generally assumed that IP3 generated as a result of PM PLCb activiation can diffuse through the cytoplasm to NE, where IP3 receptors are enriched in cardiac myocytes to release a local pool of Ca2+ (Higazi et al., 2009; Wu et al., 2006). DAG, on the other hand, is diffusible within membranes but not between membranes, so how PKD is activated at the nucleus is unclear. There is evidence that PKD can be activated at the PM and subsequently diffuse to the nucleus (Bossuyt et al., 2011), but whether PM signals are sufcient for this process has not been examined. To determine the mechanistic role of PLC in cardiac function and failure, we deleted PLC specically in cardiac myocytes in mice after development and examined the subcellular scaffolded signals generated in cardiac myocytes by PLC. We show that mice with cardiac myocyte-specic deletion of PLC are protected from pressure-overload-induced hypertrophy, clearly identifying a gene critical for the development of heart failure. Further, we dene a role of nuclear-scaffolded PLC in regulating signaling proteins that drive hypertrophy. Notably, we identify PI4P in the perinuclear Golgi apparatus as the key substrate for PLC at the NE for generation of local DAG and subsequent nuclear PKD activation. RESULTS Cardiac-Specic Deletion of PLC in PLCox/ox a-MHC-MerCreMer Mice Our recent results in NRVMs show that siRNA-based depletion of PLC prevents agonist-induced hypertrophy (Zhang et al., 2011). This contrasts with our previous work demonstrating that global knockout of PLC exacerbates hypertrophy in
response to isoproterenol infusion (Wang et al., 2005). To clarify the role of PLC in cardiac function and failure, we deleted PLC specically in cardiac myocytes in mice after development (Figure 1A and Experimental Procedures). PLCox/ox mice (PLCox/ox Cre) were bred with a-MHC-MerCreMer (a-MHC-MCM) mice to generate PLCox/ox Tg(a-MHC-MCM), (PLCox/ox Cre+) mice (Figure 1A) (Sohal et al., 2001). After 1 month of postnatal development, mice from both genotypes were given three daily injections of 40 mg/kg body weight tamoxifen intraperitoneally (i.p.). PLC messenger RNA (mRNA) was decreased by 80% in myocytes isolated from PLCox/ox Cre+ mice compared to either PLCox/ox Cre or PLC wild-type (WT) mice (Figure 1B). PLC protein was signicantly reduced (>70%) in myocytes isolated from PLCox/ox Cre+ mice relative to PLCox/ox Cre mice, whereas in lung, the level of PLC protein was unchanged (Figure 1C). PLC deletion in PLCox/ox Cre+ mice had no signicant effect on ejection fraction, fractional shortening, ventricular-wall dimensions, or left ventricular (LV) mass in the absence of stress (Figure 1E). Previous data from global PLC/ mice showed no changes in levels of other PLCb1, PLCb3, PLCg1, or PLCd3 compared to PLC+/+ mice (Wang et al., 2005). Cardiac-Specic Deletion of PLC Inhibits Development of Hypertrophy after Transverse Aortic Constriction After 4 weeks of transverse aortic constriction (TAC), PLCox/ox Cre mice had increased heart size, LV and right ventricular (RV) wall thickness, and interstitial brosis (Figure 1D, cre). These morphological changes were signicantly inhibited in the PLCox/ox Cre+ animals (Figure 1D, cre+). PLCox/ox Cre mice developed a severe decrement in heart function with a drastic decrease in ejection fraction and fractional shortening, had a signicant increase in heart weight (HW) to tibia length (TL), and showed large increases in hypertrophic gene expression (Figure 1E and Figures S1A and S1B available online). In contrast, PLCox/ox Cre+ mice were signicantly protected from these hypertrophy-related changes (Figures 1E, S1A, and S1B). These data indicate that cardiomyocyte PLC plays a role in mediating cardiac hypertrophy, supporting the mechanistic data obtained in NRVMs. Characterization of Signaling Pathways Associated with PLC Deletion in Cardiomyocytes In the global PLC knockout (KO) animal, CamKII phosphorylation was basally increased (see Figure S5 in Zhang et al., 2011), but here in the cardiac-specic KO, there was no detectable increase in CamKII phosphorylation nor PKD phosphorylation under basal conditions (Figures 1F and S1C). TAC increased phosphorylation of PKD at S916 and S744/748 in PLCox/ox Cre relative to control (Ctl) animals. This phosphorylation is signicantly blunted in the PLCox/ox Cre+ animals (Figures 1F and S1C). Similarly, CamKII phosphorylation is increased by TAC in the PLCox/ox Cre mice and is signicantly blunted in the PLCox/ox Cre+ mice (Figures 1F and S1C). These two kinases phosphorylate HDAC to cause translocation of HDAC from the nucleus to the cytoplasm, leading to MEF-dependent transcription of hypertrophic genes (Frey and Olson, 2003). HDAC phosphorylation was also blunted in the PLCox/ox
exon6, 106 amino acids
B
X X Y C2 RA1 RA2 Y C2 RA1 RA2
CDC25 CDC25
PH PH
6 PLC /GAPDH 4 2 0 Cre- Cre+ WT KO
control
TAC
Figure 1. Conditional Deletion of PLC in Cardiac Myocytes Prevents Development of Cardiac Hypertrophy
(A) Domain Structure of PLC and Strategy for Conditional Deletion of PLC. Exon 6 encodes the rst common exon of two PLC splice variants at the amino terminus of the CDC25 domain. Exon 6 was anked by two LoxP sites as shown. Small arrows indicate location of primers for genotyping, and small bars indicate the location of Southern blot probes. (B) One-month-old PLC/Cre+ and PLC/Cre mice were injected with 40 mg/kg of tamoxifen once/day for 3 consecutive days. PLC mRNA was measured by real-time quantitative PCR and normalized to GAPDH levels. (C) PLC protein was immunoprecipitated and analyzed by western blotting. GAPDH from the lysates was immunoblotted as a loading control. KO controls are from globally deleted PLC/ mice shown for comparison. (D) Anatomical view, histological HE-stained sections, and trichome-stained (for brosis, blue) sections from hearts from tamoxifen-treated PLC/ mice with or without 4 weeks of TAC. (E) Quantitation of heart weight to tibia length (HW/TL), ANF/GAPDH mRNA levels by RT-PCR, ejection fraction (by echocardiography), and fractional shortening (by echocardiography) ( standard error of the mean [SEM]). Analyzed by one-way ANOVA: 7 mice Cre, 11 mice Cre+, *p < 0.05, **p < 0.01, ***p < 0.005. (F) Western blots of PKD, phosphoPKD, CamKII, phosphoCamKII, HDAC, phosphoHDAC, and GAPDH from tamoxifen-injected mice with and without TAC. See also Figure S1.
wild type PLC exon6

EcoRV EcoRV EcoRV EcoRI BamHI EcoRI XbaI
10.7kb
EcoRV EcoRV EcoRI
EcoRV
lox p targeted PLC exon6

EXON6
EXON6
LoxP
EcoRI XbaI
CrePLC
fl/fl
Neo
8.5kb
MHC-MerCreMer + Tamoxifen
EcoRV EcoRI XbaI
EcoRV
EcoRI
Myocytes
Cre+ Cre+ Cre+ CreCre-
Lung
Cre+ CreCreWT KO Lysis buffer
Cre+
C
PLC GAPDH
1 mm
HW/TL (mg/cm)
ANF/GAPDH
200 150 100 50 0
**
25 20 15 10 5 0
***
Cre-
Cre- Cre+ Cre- Cre+
Cre- Cre+ Cre- Cre+
Fractional shortening (%)
Ejection Fraction (%)
Ctl
100 75 50 25 0
TAC
Ctl
60 40 20 0
TAC
**
Cre+
Base 4wk Base 4wk
Cre-Cre
Base 4wk Base 4wk
Cre+
-TAC
Cre+TAC
Cre+
+Cre
+TAC
100 um
-TAC
CrepPKD(S744/748) pPKD(S916) PKD pCamKII CamKII pHDAC HDAC GAPDH
Cre+
Cre+ compared to PLCox/ox Cre mice (Figures 1F and S1C). Based on these data, we hypothesize that PLC may be involved in regulation of PKD- and CAMKII-dependent phosphorylation of HDAC in cardiac myocytes. Gaq-Dependent Hypertrophy Requires mAKAP-Bound PLC Gaq signaling has been shown to be a primary driver of hypertrophy both in NRVMs and in mice (DAngelo et al., 1997; Knowlton et al., 1993). To directly test the involvement of PLC in Gaqdependent hypertrophy, adenovirus expressing Gaq was used to infect NRVMs along with an adenovirus expressing either PLC siRNA or a random siRNA Ctl sequence. As previously reported, expression of Gaq caused an increase in NRVM cell area (Figure 2A) and atrial natriuretic factor (ANF) mRNA (Figure 2B) expression, relative to infection with Ctl lacZ- or yellow uorescent
protein (YFP)-expressing viruses. Depletion of PLC by siRNA completely inhibited the development of hypertrophy by Gaq assessed by both of these measures (Figures 2A and 2B). To test whether mAKAP-PLC scaffolding was important for Gaq-dependent hypertrophy, we coexpressed Gaq with the mAKAPspectrin repeat homology domain 1 (SR1) domain, documented previously to disrupt mAKAP-PLC interactions at the NE (Zhang et al., 2011). Expression of the SR-1 domain almost completely inhibited hypertrophy driven by Gaq (Figure 2C), indicating that NE scaffolding of PLC via mAKAP is important for Gaq-dependent NRVM hypertrophy. Scaffolding of PLC, PKC, and PKD to mAKAP mAKAP is a large scaffolding protein localized to the NE in cardiac myocytes, and it scaffolds to many partners, including PLC (Dodge-Kafka et al., 2005; Kapiloff et al., 2001; Zhang et al., 2011). We have shown by biochemical fractionation that PLC is localized to the nuclear fraction of cardiac myocyte lysates along with mAKAP (Zhang et al., 2011). Expression of PLC fused to mCherry was enriched at the periphery of the nucleus in NRVMs
A
Ctl siRNA
LacZ
Gq
Figure 2. Gaq-Stimulated Hypertrophy in NRVMs Is Blocked by PLC siRNA or Disruption of NE Scaffolding via mAKAP
(A) NRVMs were transduced with 50 MOI of adenovirus expressing WT Gaq or LacZ. Cells were cotransduced with either PLC siRNA or control scrambled siRNA (Ctl) adenovirus. After 48 hr, cells were xed, permeabilized, and stained for a-actinin. Cell area was calculated from 200 cells each treatment with NIH ImageJ and pooled from three separate experiments in the right panel ( SEM). (B) Same as (A), except ANF/GAPDH ratio was determined by quantitative real-time PCR ( SEM). (C) NRVMs were cotransduced with adenoviruses expressing Gaq and mAKAP-SR domain. ANF/ GAPDH ratio was measured after 48 hr ( SEM). All experiments were repeated three times and were analyzed by one-way ANOVA, *p < 0.05; ***p < 0.005.
PLC siRNA
Cell Area (m2)
Ctl siRNA
PLC siRNA
B
ANF/GAPDH (fold change)
3
C
5 4
2
ANF/GAPDH
3 2 1
*
1 0 YFP Gq YFP Gq
0 LacZ Gq
Ctl siRNA
PLC siRNA
YFP
(Figure 3A), and some PLC is localized to endoplasmic reticulum/ sarcoplasmic reticulum (ER/SR)-like structures. To identify other components in this complex that are potentially involved in PLC signaling, we immunoprecipitated PLC from mouse heart lysates and immunoblotted for key potential upstream regulators and downstream targets. First we looked for Epac1 association because Epac is upstream of PLC-dependent regulation of cardiac contractile function and has been previously identied in mAKAP immunoprecipitates (Dodge-Kafka et al., 2005; Oestreich et al., 2007, 2009). Both PLC and mAKAP immunoprecipitates contained Epac1 (Figure 3B). When PLC was cotransfected into HEK293 cells with Epac1, Epac1 was detected in PLC immunoprecipitates (Figure S2A). Coexpression of mAKAP did not alter the efciency of the immunoprecipitation, suggesting that PLC directly binds to Epac1. To further characterize the interaction, puried GSTEpac1 was tested for binding to puried PLC. GST-Epac1 specically bound to PLC, conrming a direct interaction (Figure S2B). To determine whether interaction of Epac1 with PLC was required for association with AKAP in the heart, we compared AKAP immunoprecipitates from PLC/ and PLC+/+ mouse heart lysates (Figure S2C). Epac1 was detectable in the lysates from both genotypes, but the level of Epac1 was signicantly lower in immunoprecipitates from PLC/ mouse hearts, suggesting that binding of Epac1 to PLC contributes to scaffolding of Epac1 to mAKAP. We had also previously shown that PLC regulates phosphorylation of Ryr2 through a PKC- and CamKII-dependent mechanism during acute b-adrenergic receptor (bAR) activation (Oestreich et al., 2009). We examined PLC immunoprecipitates
for the presence of Ryr2 and found Ryr2 immunoreactivity only in lysates from hearts isolated from PLC+/+ mice (Figure 3C). Ryr2 has been shown to bind to mAKAP, so it is possible that this interacLacZ Gq tion is indirect. When coexpressed in HEK293 cells, Ryr2 weakly interacted mAKAP-SR with PLC in a manner that was not affected by coexpression of mAKAP (data not shown). This suggests that mAKAP does not mediate Ryr2 scaffolding to PLC, but because the PLC-Ryr2 interaction is weaker in cotransfected cells than in heart lysates, another protein is likely to participate in PLC-Ryr2 scaffolding. Immunodepletion of heart lysates with mAKAP antiserum reduces, but does not eliminate, Ryr2 immunoprecipitation with PLC (Figure S2D). This indicates that a fraction of PLC is scaffolded to Ryr2 but is not associated with mAKAP, supporting the existence of multiple pools of PLC in cardiac myocytes. Because PKC and PKD are critical hypertrophic kinases that we propose are downstream of PLC, we examined their scaffolding to mAKAP and PLC. We had previously shown that PLC mediates PKC activation in adult cardiac myocytes during acute bAR stimulation (Oestreich et al., 2009). PLC and mAKAP immunoprecipitates from heart lysates contained both PKC and PKD immunoreactivity (Figures 3D and 3E). PLC immunoprecipitates from hearts isolated from PLC/ mice did not contain PKC immunoreactivity conrming the specicity of the immunoprecipitation. Together, these data indicate that many of the signaling components required for PLC action in cardiac myocytes involved in either regulation of contraction or hypertrophy are assembled into a signaling complex with mAKAP. More importantly, these data show that the key hypertrophic regulators, PKC and PKD, are scaffolded with PLC and mAKAP at the NE, where they play key functional roles analyzed below. mAKAP-Bound PLC Is Involved in Nuclear PKD Activation in NRVMs To test the involvement of PLC in ET-1-dependent PKD activation, PLC was depleted in NRVMs using siRNA, and PKD S916
PLC-mCherry
PLC-mCherry +DAPI
IP: No Ab EpacI
PLC mAKAP
C
IP: No Ab PLC Ryr2 IB: Ryr2
IP-PLC EpacI
IP-NoAb
PLC+/+ PLC-/- PLC+/+
IP: Ryr2 PLC PLC PLC+/+ PLC+/+ PLC-/-
IB: Ryr2
IP: No Ab
PLC mAKAP
E
IP:PLC mAKAP No Ab PKD
the nuclear pool of PKD, NRVMs were transduced with an adenovirus expressing a nuclear-targeted uorescence resonance energy transfer (FRET)-based reporter of PKD activity, nuclear D kinase activity reporter (nDKAR) (Bossuyt et al., 2011; Kunkel et al., 2007). Figure 4C is a YFP/ cyan uorescent protein (CFP) ratiometric image from NRVMs expressing nDKAR, with CFP excitation showing basal FRET in the nucleus as previously reported (Bossuyt et al., 2011). NRVMs were treated with either vehicle or ET-1, and nuclear FRET was monitored over time. ET-1 treatment causes a signicant decrease in nuclear FRET over the course of 20 min compared to vehicle Ctl (Figures 4D and 4E). In NRVMs treated with PLC siRNA, the ET-1induced decrease in FRET was completely eliminated (Figures 4D and 4E). Expression of mAKAP-SR1 in NRVMs also completely inhibited the ET-1-dependent decrease in nDKAR FRET (Figure 4F). This indicates that ET-1 activates nuclear PKD, and that this nuclear PKD activation requires mAKAP-scaffolded perinuclear PLC. PLC Is Important for Localized Nuclear Calcium Signals ET-1 induces calcium increases in the nuclei of cardiac ventricular myocytes; this process is dependent on IP3 presumably binding to type II IP3 receptors localized at the NE (Higazi et al., 2009; Wu et al., 2006). Several reports indicate that the myocyte nuclear Ca2+ release via IP3 receptors can mediate cardiomyocyte hypertrophy (Arantes et al., 2012; Higazi et al., 2009; Nakayama et al., 2010). To monitor changes in nuclear calcium levels, we utilized Fura-2-loaded NRVMs and specically monitored the nuclear regions of the cells by ratio imaging epiuorescence microscopy (Figure 5A). NRVMs were pretreated with nifedipine (L-type Ca2+ channel blocker) and mibefradil (L- and T-type Ca2+ channel blocker) to block Ca2+ entry and Ca2+ release associated with myocyte excitation contraction (Higazi et al., 2009). ET-1 stimulated an increase in nuclear [Ca2+] that was inhibited by PLC siRNA treatment (Figure 5A). Expression of the RA domain of PLC to disrupt AKAP-dependent nuclear scaffolding also inhibited nuclear Ca2+ release but only by 10%20% (Figure 5A, middle right panel). This nuclear Ca2+ release is likely IP3 dependent, as previously reported (Higazi et al., 2009), because it was not affected by blocking the other major Ca2+-release channel, Ryr2, with ryanodine (Figure 5A, right panel). To conrm that the ET-1-dependent Ca2+ increases are localized to the nucleus, we used two-photon microscopy to examine ET-1-dependent increases in nuclear Ca2+ signals in NRVMs. Treatment with ET-1 results in a robust Ca2+ release that is primarily restricted to the nucleus (Figure 5B; Movie S1), as previously reported. These data indicate that IP3 generated by PLC-dependent PIP2 hydrolysis contributes to nuclear Ca2+ signals that may be important for regulation of hypertrophy, but that PLC scaffolded at the NE is not a major participant in this release. This suggests that PLC localized elsewhere, possibly at the PM, contributes to diffusible IP3 generation necessary for release of Ca2+ at the NE. Perinuclear PI4P Localization To identify potential perinuclear substrates for PLC, we utilized specic uorescent probes containing GFP fused to domains that recognize specic phosphoinositides (Balla et al., 2009).
PKC IP-PLC PKC IP-NoAb
PLC+/+ PLC-/- PLC+/+
Figure 3. PLC Is in a Multicomponent Signaling Complex with mAKAP, Epac, PKC, PKD, and Ryr2 in the Heart
(A) perinuclear localization of PLC in NRVMs. mCherry-tagged PLC was expressed in NRVMs (left panel) and costained with DAPI (right panel). The boundaries of the cell are outlined with dashed lines. (B) Epac coimmunoprecipitates with PLC and mAKAP from heart lysates. (C) Ryr2 immunoprecipitates with PLC from heart lysates. (D) PKC immunoprecipitates with PLC and mAKAP from heart lysates. (E) PKD immunoprecipitates with PLC and mAKAP from heart lysates. PLC/ mice were tested as a specicity control in (B), (C), and (D). See also Figure S2.
phosphorylation was measured. PKD phosphorylation increases after 10 min of ET-1 stimulation and declines thereafter but remains active for at least 4 hr (Figure S3). ET-1 stimulation of PKD phosphorylation in PLC siRNA-treated NRVMs was blunted compared to Ctl siRNA-treated cells at all times measured. We more carefully examined inhibition by PLC depletion after stimulation with ET-1 and norepinephrine for 1 hr and ET-1 for 24 hr (Figure 4A). At both times, ET-1-dependent PKD phosphorylation was inhibited in PLC siRNA-treated NRVMs by approximately 50%. This indicates that PLC is important for activation of PKD, but because inhibition of total cellular PKD activation by PLC depletion is partial, PLC may be regulating a specic subcellular pool of PKD. To determine whether mAKAP-associated PLC was involved in PKD activation, PLC-mAKAP interactions were disrupted by expression of mAKAP-SR1 or PLC-ras-association 1 (RA1) domains (Zhang et al., 2011), and ET-1-stimulated phosphorylation of PKD at 1 hr and 24 hr was measured. Expression of either PLC-RA1 or mAKAP-SR1 domains inhibited ET-1-dependent global PKD phosphorylation at 1 hr (30%) and 24 hr (70%) (Figure 4B). To determine whether PLC is required for activation of
Ctl siRNA
A
130 130
PLC siRNA
Ctl ET-1 NE pPKD(S916) PKD
normalized p-PKD/PKD
1.5 1.0 0.5 0.0
Ctl
ET-1 NE
1h
CON ET-1 NE
Figure 4. PLC Regulates Nuclear PKD Activation

(A) NRVMs were infected with PLC or control (Ctl) siRNA, followed by treatment with 100 nM ET-1 or 10 mM norepinephrine (NE) for 1 hr and ET-1 for 24 hr and assessment of PKD phosphorylation by western blotting, each repeated 34 times. Statistically different from Ctl siRNA, **p < 0.01, ***p < 0.005. (B) NRVMs were infected with either control YFP, mAKAP-SR, or PLC-RA expressing adenoviruses to disrupt PLC-mAKAP scaffolding, followed by treatment with 100 nM ET-1 for 1 hr or 24 hr and assessment of total PKD activation. Statistically different from Ctl siRNA, ***p < 0.005. (C) NRVMs were transduced with adenovirus-expressing nDKAR. The YFP to CFP ratio is shown as a pseudocolor image to emphasize the high level of FRET in the nucleus. (D) A region in the nucleus from NRVMs expressing nDKAR FRET was selected, and the YFP/CFP ratio was followed for the indicated times after 100 nM ET-1 addition in PLC siRNA or Ctl siRNA adenovirus-treated NRVMs. (E) Pooled data for the YFP/CFP ratio of nDKAR analyzed 20 min after addition of ET-1 (+) or vehicle () from ve independent experiments in PLC siRNA- or Ctl siRNA-expressing NRVMs (50100 cells each condition); ***p < 0.005. (F) NRVMs were cotransfected with plasmids expressing nDKAR and mAKAP-SR1 or control LacZ. ET-1 (100 nM) was added at the indicated time. Data are pooled from four independent cells for each treatment from two separate NRVM preparations. All quantitative data are SEM. See also Figure S3.
*** **
Ctl siRNA
PLC siRNA
Ctl siRNA
PLC siRNA
p-PKD/PKD intensity
ET-1
ET-1
ET-1
ET-1
ET-1
ET-1
0.25 0.20 0.15 0.10 0.05 0.00
Ctl
Ctl
130 130
24h
***
p - PKD S916 PKD
Ctl siRNA PLC siRNA
normalized p-PKD/PKD
1.5 1.0 0.5 0.0
B
Ctl
130 130
YFP ET-1
PLC - RA Ctl ET-1
mAKAP-SR Ctl ET-1 pPKD (S916) PKD
1h
***
CON ET-1
***
YFP PLC-RA
mAKAP-SR
p-PKD/PKD intensity
YFP
PLC - RA mAKAP-SR
0.4 0.3 0.2 0.1 0.0
ET-1
ET-1
ET-1
Ctl
Ctl
Ctl
24h
** **
130 130
pPKD (S916) PKD
YFP
PLC-RA mAKAP-SR
Normalized YFP/CFP
1.02 1.00 0.98 0.96 0.94 0.92 0.90 0

PLC siRNA Ctl siRNA
ET-1
5 10 15 20 Time (min)
Normalized YFP/CFP
Normalized YFP/CFP
1.00 0.95 0.90
***
***
1.04 1.00 0.96

ET-1
0.92 0.88 0
ET-1 ET-1+SR1
ET-1
- +
- +
5 10 15 20 25 30
Ctl siRNA PLC siRNA
Time (min)
A
0.1 340/380 ratio unit
Nif/mib ET-1
PBSS
340/380 ratio
***
60 sec
Ctl siRNA
PLC siRNA
YFP
PLC-RA
Ryan:
+
ET-1
Caffeine
+block
Nifedipine and mibefradil 1000 Fluorescence A.U. ET-1
MP ex 810 nm em >500 nm
+endothelin
30 s
31 s
33 s
35 s
60 sec
37 s
39 s
41 s
43 s
45 s
47 s
Figure 5. PLC Is Involved in Regulation of Nuclear Ca2+ Elevation

(A) NRVMs were loaded with Fura-2, and the Fura-2 ratio in the nucleus was measured with time after 200 nM ET-1 addition in the presence of a 10 mM nifedepine/ 2 mM mibefradil. Left panel is a representative trace from two individual cells. Middle left panel is the combined data from multiple cells comparing Ctl siRNA- and PLC siRNA-treated NRVMs, ***p < 0.005. Middle right panel is the combined data from cells expressing either YFP or the YFP-RA domain to disrupt mAKAP scaffolding. Both sets of data are from four independent experiments from cells isolated from four different NRVM preparations with >10 cells for each coverslip and 3 coverslip averages for each NRVM preparation; *p < 0.05, calculated based on an average of 12 coverslip averages for each condition. Right panel shows that the ET-1-dependent nuclear Ca2+ response is not blocked by Ryanodine, but the caffeine response is (this experiment was repeated with two separate sets of NRVMs; data are representative of one experiment with 30 cells each condition). All data are SEM. (B) Two-photon microscopy was used to measure nuclear Ca2+ levels in Fluo-4-loaded NRVMs. Cells were perfused with imaging buffer containing 10 mM nifedepine and 1.8 mM mibefradil to block Ca2+ transients associated with voltage-dependent Ca2+ release followed by addition of 100 nM ET-1. Individual panels show the level of Fluo-4 uorescence at the indicated times (see Movie S1). The boxes in the rst panel correspond to the areas shown in the traces shown on the right.
To localize PIP2, we transfected cells with reporters containing GFP fused to the pleckstrin homology (PH) domain of PLCd or the PIP2-binding domain of Tubby and examined the cells by confocal uorescence microscopy. Both of these constructs bind to PIP2 specically, but PLCd-PH also binds IP3. Transfection of GFP-Tubby (Figure 6A) or GFP-PLCd-PH (not shown) into HEK293 cells revealed a prominent PM distribution of PIP2 with some intracellular uorescence but no obvious localization near the NE (Figure 6A). Transfection of NRVMs with the same constructs revealed a similar PM distribution of PIP2 (Figure 6A). Thus PIP2 is not detectable at the NE in NRVMs and is likely not available as a substrate for NE-scaffolded PLC activity. PI4P is also a substrate for PLC and other mammalian PLC isoforms in vitro, but its utilization as a substrate in cells has not been demonstrated (Rhee et al., 1989; Seifert et al., 2004; Smrcka et al., 1991). To determine the localization of PI4P in NRVMs, cells were transfected with reporters containing GFP fused to the PH domains of either four-phosphate adaptor protein (FAPP) or oxysterol-binding protein (OSBP). These reporters selectively monitor PI4P subcellular localization in living cells, but this binding is codependent on interaction with the small GTPase ARF (Balla et al., 2005). Thus they readily detect PI4P at intracellular membranes containing ARF. As previ222 Cell 153, 216227, March 28, 2013 2013 Elsevier Inc.
ously reported in other cultured cells, transfection of GFPFAPP-PH into HEK293 cells leads to prominent uorescence at asymmetrically localized structures that correspond to the Golgi apparatus (Figure 6B, left panel). Surprisingly, and in striking contrast to other cultured cells, both FAPP-PH-GFP and OSBP-PH-GFP strongly show a ring of perinuclear uorescence surrounding the nucleus in NRVMs (Figure 6B, middle and right panels and Figure S4A). In some cells, other intracellular structures resembling either Golgi or SR were labeled (Figure 6C, for example), but this was generally less prominent. Treatment with Brefeldin A to inhibit ARF resulted in rapid depletion of perinuclear FAPP-PH-GFP uorescence (Figure 6C). Perinuclear FAPP-PH-GFP uorescence was also completely disrupted by treatment with PI4-kinase inhibitor phenylarsine oxide (PAO) (Figure 6D). Brefeldin A is an agent used to disrupt the Golgi apparatus through its ability to block Golgi-localized ARF. ARF and PI4P are enriched in the Golgi in most cells and play important roles in Golgi-membrane vesicle and lipid trafcking. This suggested that the perinuclear PI4P may be in Golgi closely associated with the outer surface of the nucleus. Treating cells with a BODIPY-TR ceramide to uorescently label the Golgi revealed a perinuclear ring similar to what is observed with the PI4P
HEK293 Tubby
NRVM Tubby
NRVM PLC-PH
reporters (Figure S4B). Indeed, previous studies have shown that the Golgi surrounds the nucleus in myocytes and is intimately associated with the NE at a constant distance of 100200 nm (Kronebusch and Singer, 1987; Tassin et al., 1985). Together these data indicate that PI4P is localized to the Golgi apparatus, in close proximity to the NE, where it could be accessed by mAKAP-scaffolded PLC. PLC Activation at the NE Leads to PI4P Depletion and DAG Generation To assess whether perinuclear PI4P is a substrate for PLC, we treated cells with a specic activator of Epac, 8-(4-chlorophenylthio)-20 -O-methyl-cAMP (cpTOME). Epac-dependent Rap activation stimulates PLC but no other PLC isoforms, and Epac is scaffolded at the NE with PLC and mAKAP. Thus, PLC in this NE-signaling complex can be activated selectively with cpTOME. Treatment of NRVMs with cpTOME for 1 hr caused a signicant depletion of perinuclear Golgi FAPP-PHGFP uorescence (Figure 7A). Selected regions of perinuclear FAPP uorescence were monitored in single live cells over time after addition of vehicle Ctl, cpTOME, or ET-1 (Figure 7B; Movie S2). Both cpTOME and ET-1 treatment enhanced the rate of depletion of FAPP-PH-GFP uorescence compared to vehicle Ctl. ET-1-dependent decreases in PI4P were blocked by preincubation with the ET-1a receptor antagonist BQ-123 (Figure 7B). We used a PI4P strip mass assay to measure total PI4P levels in NRVMs after vehicle or cpTOME treatment (Dowler et al., 2002). Treatment with cpTOME led to a signicant decrease in cellular PI4P mass compared to vehicle-treated NRVMs (Figure 7C). Similar studies examining PIP2 at the PM indicate that cpTOME does not stimulate PM-dependent PIP2 depletion (Figure S5A). To show that ET-1-dependent perinuclear PI4P depletion was dependent on mAKAP-scaffolded PLC, cells were cotransfected with GFP-FAPP-PH and either mAKAP-SR1 or Ctl vector (Figure 7D). In cells transfected with mAKAP-SR1, ET-1-dependent depletion of PI4P was blocked. Additionally, as shown in Figure 7E, depletion of PLC with PLC-siRNA completely eliminated cpTOME-dependent depletion of perinuclear PI4P. These data strongly indicate that mAKAP-scaffolded PLC is directly involved in perinuclear PI4P depletion consistent with its activity as a perinuclear enzyme that can hydrolyze PI4P (Figure S5B) to produce DAG. One consideration is that although PIP2 is undetectable in the perinuclear Golgi in cardiac myocytes (Figure 6A), there could be a very low level of PIP2 that is the direct substrate for perinuclear PLC. In this scenario, PI4P might be indirectly depleted to replace the hydrolyzed PIP2 pool. To test this, a PI5 phosphatase was targeted to the Golgi or the PM with a rapamycinrecruitable type IV PI5P ptase domain fused to FKBP12 (Varnai et al., 2006). This PI5 ptase cleaves the 5-phosphate from PI4,5P2, effectively depleting PIP2. Figure S6A shows that targeting of the PI5 ptase to the PM eliminates GFP-Tubby labeling of the PM, indicating that the targeted PI5 ptase depletes PM PIP2. Targeting of the PI5 ptase to the Golgi with rapamycin does not alter PI4P levels (Figures S6B and S6C, black squares). Treatment with cpTOME depletes PI4P at the perinuclear Golgi at the same rate in the presence or absence of rapamycin-dependent targeting of the PI5P ptase to the Golgi
HEK-293 FAPP-PH
NRVM FAPP-PH
NRVM OSBP-PH
0 min
20 min
30 min
40 min
Vehicle
Brefeldin A
0 min
PAO
30 min
Figure 6. PI4P Localizes to Perinuclear Golgi Surrounding the NE in Cardiac Myocytes

(A) Detection of PI4,5P2 localization in NRVMs. Left panel: HEK293 cells transfected with Tubby-GFP and stained with DAPI; middle: NRVMs transfected with Tubby-GFP; right: NRVMs transfected with PLCd-PH-GFP and analyzed by confocal microscopy. (B) Detection of PI4P at the NE. The indicated cell types were transfected with either OSBP-PH-GFP or FAPP-PH-GFP and analyzed by confocal microscopy. (C) Inhibition of ARF eliminates perinuclear staining with FAPP-PH-GFP. NRVMs transfected with FAPP-PH-GFP were treated with 1 mg/ml Brefeldin A, and GFP uorescence monitored with time. Cells treated with vehicle are shown in the bottom panels and indicate a lack of photobleaching in these experiments. (D) Inhibition of PI4kinase with PAO depletes perinuclear uorescence associated with FAPP-PH-GFP. NRVMs transfected with FAPP-PH-GFP were treated with 10 mM PAO, and uorescence analyzed by confocal microscopy. All experiments were repeated a minimum of three times. See also Figure S4.
Relative Fluorescence
pmol PI4P stand.

20 15 10 5 4
PI4P
Before cpTOME
cpTOME (1 hr)
C
100
Veh
100 90 80 70 60 50 0
cpTOME cpTOME Veh
P Is
PI PI3P PI5P
2
90 80 70 60 50 0
ET-1 BQ+ET-1 ET-1
PI3,4P2 PI3,5P2 PI4,5P

3
cpTOME
2 1 0.5
PI3,4,5P
Relative PI4P
E
100 90 80 70 60 50 0
ET-1 SR1+ET-1 ET-1
100 75 50 25 0
***
10 20 30 Time (min)
40
10
20
30
40
Time (min)
100 75 50 25 0
0 Min ET-1 ET-1+BQ
***
***
125
***
100 75 50 25 0 veh
100 90 80 70 60 50 0
PLC siRNA Ctl siRNA
cpTOME
**
0 Min cpTOME
Veh
cpTOME
10
20
30
40
10
20
30
1.15 1.10 1.05 1.00 0.95 0 2 4 6 8 10 12
50 min
50 min
1.05 1.00 0.95 0.90 0.85
Time (min)
100 75 50 25 0
0 Min ET-1 ET-1+SR1
Time (min)
Normalized YFP/CFP
Normalized Fpn /Fc
***
***
***
ET-1
ET-1 ET-1+Sac1
10 15 20 25 30
50 min
Time (min)
Time (min)
Figure 7. Perinuclear PI4P Is a Substrate for mAKAP-Scaffolded PLC, Stimulated by either Epac or ET-1 Receptors
(A) NRVMs transfected with FAPP-PH-GFP were analyzed by live-cell confocal microscopy before and after treatment with 10 mM cpTOME for 1 hr (see also Movie S2). (B) Individual regions of GFP uorescence in NRVMs transfected with FAPP-PH-GFP were monitored with confocal microscopy and followed with time after treatment with vehicle, 10 mM cpTOME, 50 nM ET-1, or 50 nM ET-1+ 100 nM BQ-123 (top panels are representative traces). Data were pooled from four experiments at 0 and 50 min for quantitation and statistics (bottom panels). (C) NRVMs were treated with vehicle or 10 mM cpTOME for 50 min followed by extraction of PI4P and assay using a PI4P protein-lipid overlay assay according to the manufacturers instructions. Data from three separate experiments are quantitated in the bottom panel and analyzed by a Students t test. (D) NRVMs were cotransfected with FAPP-PH-GFP and either mAKAP-SR1 or control plasmid. Perinuclear GFP uorescence was monitored as in (B). ET-1 experiments in (B) and (D) were done in parallel so the ET-1 alone representative traces are the same in both panels. Top panel is a representative trace, and bottom panel is pooled data from three experiments analyzed by one-way ANOVA. (E) NRVMs were transfected with FAPP-PH-GFP and transduced with viruses expressing PLC siRNA or random control siRNA, and perinuclear GFP uorescence was monitored as in (B) and (D). Data are pooled from ve independent experiments each. (F) NRVMs were transfected with YFP-C1b-Y123W to detect DAG localization. Left panel shows localization of YFP-C1b-Y123W by confocal microscopy. Right panel: NRVMs transfected with YFP-C1b-Y123W were stimulated with 10 mM cpTOME, and perinuclear regions and cytoplasm were imaged over time. The ratio of perinuclear uorescence (Fpn) to the cytoplasmic uorescence (Fc) was calculated and normalized to the starting Fpn/Fc before cpTOME addition. Data are pooled from four independent experiments; Fpn/Fc at 912 min were individually compared with Fpn/Fc at 0 min with a one-way ANOVA. (G) Golgi-specic depletion of PI4P blocks ET-1-dependent nuclear PKD activation. Left panel: Cells were transduced with GFP-tagged, Golgi-targeted Sac-1 and imaged by confocal microscopy. Right panel: Cells were cotransfected with plasmids expressing nDKAR and Golgi-targeted Flag-tagged Sac-1 plasmids, and nDKAR FRET was monitored as in Figures 4D and 4F. Fifty nanomolar ET-1 was added at the indicated time. Data were pooled from four independent cells for each treatment from two separate NRVM preparations. All data are SEM. See also Figures S5, S6, and S7 and Movie S2.
(Figure S6C). This indicates that perinuclear Golgi PI4P depletion is not dependent on PIP2 hydrolysis in the Golgi, further supporting the idea that the direct substrate for perinuclear PLC is PI4P. To test for production of perinuclear DAG, cells were transfected with a DAG-binding domain (C1b domain of PKCbII with a Y123W mutation) that is fused to YFP (YFP-C1b-Y123W), binds with high afnity to DAG, and translocates to membranes where DAG is produced (Kunkel and Newton, 2010). In resting cells, there was basal association of the reporter with perinuclear
and other cellular membranes (Figure 7F). This could be due to basal turnover of lipids in the Golgi in the presence of serum, or the very high afnity of this reporter for DAG could stabilize a turning-over DAG pool. Nevertheless, treatment with cpTOME increased the association of a DAG reporter with the perinuclear region of the cell. The increase was small (10% increase) but reproducible and consistent with what has been previously observed for this DAG reporter in the Golgi in other cell types (Kunkel and Newton, 2010). Thus DAG is generated in close proximity to the NE by stimulation of PLC.
To test whether PI4P is required for nuclear PKD activation, PI4P was depleted in NRVMs using PAO (as shown in Figure 6D), and ET-1-dependent nuclear PKD activity was monitored using nDKAR (Figure S7). PAO depletes cellular PI4P in cells without depleting the PIP2 pools necessary for conventional PLC-mediated PIP2 hydrolysis (Hammond et al., 2012). By examining the effect of PAO treatment on GFP-Tubby uorescence, we also found that PAO does not deplete NRVM PIP2 (data not shown). PAO treatment completely eliminated ET-1-dependent nuclear PKD activation, indicating that PI4P is a required substrate for nuclear PKD activation, which supports the idea that Golgi PI4P is required for nuclear PKD activation. To more directly show that Golgi-localized PI4P is required for nuclear PKD activation, we transfected cells with a variant of the PI4P-specic phosphatase, Sac1, with a mutation leading to specic localization in the Golgi (Sac1-K2A) (Blagoveshchenskaya et al., 2008; Rohde et al., 2003). In the left panel of Figure 7G is an NRVM transfected with GFP-Sac1-K2a, showing perinuclear Golgi localization of Sac1-K2A. Sac1-K2A transfection completely eliminated the ET-1-dependent decrease in nDKAR FRET, demonstrating that PI4P in the Golgi is required for nuclear PKD activation (Figure 7G, right panel). We conclude that mAKAP-scaffolded PLC generates DAG from PI4P in close proximity to the NE required for nuclear PKD activation. DISCUSSION We previously demonstrated that PLC integrates multiple hypertrophic stimuli in neonatal rat ventricular myocytes (Zhang et al., 2011). Here we demonstrate that cardiac-specic deletion of PLC after 1 month of development signicantly inhibits cardiac hypertrophy development. This strongly suggests that PLC signaling in the cardiac myocyte is important for hypertrophy development, and that the neonatal myocyte analysis of PLC function is largely relevant to whole-heart function in an animal. Our previously published data established that both phosphoinositide hydrolysis by PLC and NE scaffolding of PLC via mAKAP are required for PLC-dependent hypertrophy in NRVMs (Zhang et al., 2011). This suggested that PLC must be generating IP3, DAG, or both at the NE to drive hypertrophic signaling cascades at the nucleus. IP3 generated at the PM is diffusible, so agonist-dependent nuclear Ca2+ release could be controlled by IP3 generated from PM PIP2 pools. DAG, on the other hand, is a membrane-bound lipid, not freely diffusible between membranes, and therefore must be produced in the vicinity of the nucleus to regulate nuclear signaling proteins. Here we show that PLC is important for agonist-dependent regulation of both nuclear Ca2+ release and PKD activation. PLC scaffolded at the NE is required to generate perinuclear DAG important for nuclear PKD activation, but surprisingly, the substrate for this reaction is in the uniquely localized perinuclear Golgi apparatus. PKD is recruited to the Golgi apparatus in a DAG-dependent manner in other cell types where it is involved in lipid and vesicle trafcking (Campelo and Malhotra, 2012). PLC, PKD, and ARF are critical regulators of Golgi ssion involved in the transport of cargo to the PM. Here, with the unique architecture of the Golgi apparatus in cardiac myocytes
and the perinuclear scaffolding of PLC by mAKAP, the process of PI hydrolysis and DAG generation appears to have been coopted to recruit and activate PKD in the vicinity of the nucleus, where it can regulate nuclear gene expression. Although we and others have shown that PI4P is a substrate for mammalian PLC isoforms, including PLC, in vitro (Figure S5B), PI4P has not been shown to be a native physiological substrate for any mammalian PLC in cells. Thus, we demonstrate here that PI4P hydrolysis is a physiologically relevant PLC reaction that likely performs a widespread function in cell biology. PI4P has generally been thought to function primarily as a precursor to replace PIP2 as it is depleted by receptor-stimulated PIP2 hydrolysis. Recently however, it has been shown that depletion of PI4P does not signicantly alter the level of PIP2 in cells, nor does it affect acute receptor-stimulated PIP2 hydrolysis, indicating that PI4P has functional roles other than simply serving as a PIP2 precursor (Hammond et al., 2012). We also show that PLC is involved in regulation of ET-1dependent nuclear Ca2+ elevation. Previous studies have indicated that local Ca2+ signaling at the nucleus is IP3 receptor dependent and regulates HDAC nuclear export via activation of CamKII (Wu et al., 2006). The source of IP3 was undened in these experiments but has been suggested to diffuse from the PM. Hydrolysis of PI4P by PLC activity would generate DAG and inositol 1,4-bisphosphate (IP2) and thus would not be a relevant reaction for local IP3-dependent Ca2+ release because IP2 does not regulate Ca2+ release through IP3 receptors. Thus the role of PI4P hydrolysis by mAKAP-scaffolded PLC at the Golgi appears to be DAG generation for PKD activation. We propose that a different pool of PLC, perhaps at the PM in cooperation with PLCb, generates IP3 from PIP2, which can diffuse to the nucleus and release Ca2+ via IP3 receptors in the nucleus. Overall, our model is that PLC is located in different subcellular compartments, allowing for regulation of hypertrophy and CICR via different mechanisms. In conclusion, we have discovered a novel integrator of hypertrophic signals in the heart, PLC, whose scaffolding at the NE to generate DAG from the novel PLC substrate PI4P in the perinuclear Golgi apparatus is critical for this function. This suggests that PLC catalytic activity could be a novel target for heart failure. On the other hand, PLC is found in many cell types and has multiple functions (Smrcka et al., 2012). Global deletion of PLC increases the propensity for heart failure (Wang et al., 2005). mAKAP has a more restricted distribution and thus so does the PLC-mAKAP complex (Kapiloff et al., 1999). A more targeted strategy for treatment of heart failure could involve developing reagents that interfere with PLC scaffolding to mAKAP.
EXPERIMENTAL PROCEDURES See the Extended Experimental Procedures for detailed methodology. Cardiomyocyte-Specic Deletion of PLC and Induction and Analysis of Hypertrophy PLCox/ox Cre+ and PLCox/ox Cre mice were generated as described in the Extended Experimental Procedures. At 30 days after birth, mice were injected intraperitoneally (i.p.) three times on consecutive days with 40 mg/kg tamoxifen to excise exon 6 of the PLCe1 gene.
TAC Two- to three-month-old tamoxifen (3060 days after tamoxifen injection) injected male PLC WT (PLCox/ox Cre) and cardiac KO (PLCox/ox Cre+) mice were subjected to TAC or sham operation (opening and closing chest without aortic banding). Mice were anesthetized with isourane anesthesia (2% inhaled) and intubated. Essentially as described by ourselves and others, a transaortic constriction was produced (Ram et al., 2011; Rockman et al., 1991). Echocardiography Transthoracic 2D and M-mode echocardiography analysis was used to assess heart function after 4 weeks of TAC in mice under minimal isoorane anesthesia with similar heart rates across animals using a Visualsonics Vevo 2100 echocardiography system (Visualsonics, Toronto, ON), using methods as we have recently described (Ram et al., 2011). Morphological Assessment of Hypertrophy After 6 weeks of TAC, mice were euthanized, and hearts were harvested. Heart and tibias were dissected. Heart weight and TL were measured. For histological analysis, hearts were perfused and xed with 10% formalin. Isolation, Culture, Adenoviral Infection, siRNA Treatment, and Hypertrophy Analysis of NRVMs Isolation of NRVMs, adenoviral-mediated siRNA of PLC, hypertrophy induction, and analysis were as previously described (Zhang et al., 2011). Imaging of Phosphoinositide and DAG Reporters Cells were imaged 2448 hr after transfection with the appropriate plasmid at 12 mg. Transfection efciency was approximately 5%. Fluorescent cells were identied and imaged by confocal microscopy. During imaging and cpTOME, ET-1, BFA, and PAO treatments, cells were in culture medium containing serum. EGFP: excitation 488 nm, emission 510 nm; YFP: excitation 515 nm, emission 527 nm; mCherry and mRFP: excitation 559 nm, emission 618 nm; CFP: excitation 440, emission 476; DAPI and Hoeschst stains: excitation 405 nm, emission 461 nm. Ca2+ Imaging NRVMs were cultured on laminin (25 mg/ml; Invitrogen) coated coverslips. For epiuorescent Ca2+ imaging, myocytes were loaded with Fura-2AM (2 mM) in imaging buffer (HBSS containing in mM: 5.5 glucose, 0.56 MgCl2, 4.7 KCL, 1 Na2HPO4, 10 HEPES, 1.2 CaCl2, pH 7.4). One hundred micromolar ryanodine was loaded in conjunction with Fura-2 dye for the ryanodine receptor block experiments. For measurement of nuclear Ca2+ by two-photon microscopy, myocytes were loaded with Fluo-4AM (2 mM) in imaging buffer. Fluo-4: excitation 820 nm, emission 520 nm. nDKAR FRET NRVMs were transduced with an nDKAR-expressing adenovirus. Transduced cells were identied with CFP excitation and emission. FRET was determined as the ratio of YFP emission at 535 nm to CFP emission at 480 nm after CFP excitation at 440 nm. Measurement of PI4P Mass PI4P was extracted from either vehicle- or cpTOME-treated myocytes (1 hr) using a MeOH, CHCl3, HCl extraction protocol (Gray et al., 2003). The extract was spotted onto a nitrocellulose lter, and PI4P was detected using a PI4P protein lipid overlay assay according to the manufacturers instructions (Echelon Biosciences Inc.) (Dowler et al., 2002). SUPPLEMENTAL INFORMATION Supplemental Information includes Extended Experimental Procedures, seven gures, and two movies and can be found with this article online at http://dx. doi.org/10.1016/j.cell.2013.02.047.
ACKNOWLEDGMENTS We would like to thank Tamas Balla for uorescent phosphoinositide reporter fusion and PI5 ptase constructs and for valuable critical reading of this manuscript. We would also like to thank Peter Mayinger for providing the Sac1-K2A construct and Alexandra Newton for the nDKAR and DAG reporter constructs. This work was supported by National Institutes of Health grants 5R01GM053536 (A.V.S.), R01DE041756 (D.I.Y.), and R01 HL089885 and R01 HL091475 (B.C.B.). Received: October 26, 2012 Revised: January 22, 2013 Accepted: February 14, 2013 Published: March 28, 2013
REFERENCES , N.C.G., Andrade, L.M., Arantes, L.A.M., Aguiar, C.J., Amaya, M.J., Figueiro Rocha-Resende, C., Resende, R.R., Franchini, K.G., Guatimosim, S., and Leite, M.F. (2012). Nuclear inositol 1,4,5-trisphosphate is a necessary and conserved signal for the induction of both pathological and physiological cardiomyocyte hypertrophy. J. Mol. Cell. Cardiol. 53, 475486. rnai, P., and Balla, T. (2005). Balla, A., Tuymetova, G., Tsiomenko, A., Va A plasma membrane pool of phosphatidylinositol 4-phosphate is generated by phosphatidylinositol 4-kinase type-III alpha: studies with the PH domains of the oxysterol binding protein and FAPP1. Mol. Biol. Cell 16, 12821295. Balla, T., Szentpetery, Z., and Kim, Y.J. (2009). Phosphoinositide signaling: new tools and insights. Physiology (Bethesda) 24, 231244. dler, A., Blagoveshchenskaya, A., Cheong, F.Y., Rohde, H.M., Glover, G., Kno Nicolson, T., Boehmelt, G., and Mayinger, P. (2008). Integration of Golgi trafcking and growth factor signaling by the lipid phosphatase SAC1. J. Cell Biol. 180, 803812. Bossuyt, J., Chang, C.-W., Helmstadter, K., Kunkel, M.T., Newton, A.C., Campbell, K.S., Martin, J.L., Bossuyt, S., Robia, S.L., and Bers, D.M. (2011). Spatiotemporally distinct protein kinase D activation in adult cardiomyocytes in response to phenylephrine and endothelin. J. Biol. Chem. 286, 33390 33400. Campelo, F., and Malhotra, V. (2012). Membrane ssion: the biogenesis of transport carriers. Ann. Rev. Biochem. 81, 407427. Cocco, L., Martelli, A.M., Gilmour, R.S., Ognibene, A., Manzoli, F.A., and Irvine, R.F. (1989). Changes in nuclear inositol phospholipids induced in intact cells by insulin-like growth factor I. Biochem. Biophys. Res. Commun. 159, 720725. DAngelo, D.D., Sakata, Y., Lorenz, J.N., Boivin, G.P., Walsh, R.A., Liggett, S.B., and Dorn, G.W., 2nd. (1997). Transgenic Galphaq overexpression induces cardiac contractile failure in mice. Proc. Natl. Acad. Sci. USA 94, 81218126. , H., and Irvine, R.F. (1991). The polyphosphoinositide cycle Divecha, N., Banc exists in the nuclei of Swiss 3T3 cells under the control of a receptor (for IGF-I) in the plasma membrane, and stimulation of the cycle increases nuclear diacylglycerol and apparently induces translocation of protein kinase C to the nucleus. EMBO J. 10, 32073214. , H., and Irvine, R.F. (1993). Inositides and the nucleus and Divecha, N., Banc inositides in the nucleus. Cell 74, 405407. Dodge-Kafka, K.L., Soughayer, J., Pare, G.C., Carlisle Michel, J.J., Langeberg, L.K., Kapiloff, M.S., and Scott, J.D. (2005). The protein kinase A anchoring protein mAKAP coordinates two integrated cAMP effector pathways. Nature 437, 574578. Dorn, G.W., 2nd, and Brown, J.H. (1999). Gq signaling in cardiac adaptation and maladaptation. Trends Cardiovasc. Med. 9, 2634. Dowler, S., Kular, G., and Alessi, D.R. (2002). Protein lipid overlay assay. Sci. STKE 2002, pl6.
Filtz, T.M., Grubb, D.R., McLeod-Dryden, T.J., Luo, J., and Woodcock, E.A. (2009). Gq-initiated cardiomyocyte hypertrophy is mediated by phospholipase Cbeta1b. FASEB J. 23, 35643570. Frey, N., and Olson, E.N. (2003). Cardiac hypertrophy: the good, the bad, and the ugly. Annu. Rev. Physiol. 65, 4579. Gray, A., Olsson, H., Batty, I.H., Priganica, L., and Peter Downes, C. (2003). Nonradioactive methods for the assay of phosphoinositide 3-kinases and phosphoinositide phosphatases and selective detection of signaling lipids in cell and tissue extracts. Anal. Biochem. 313, 234245. Grubb, D.R., Iliades, P., Cooley, N., Yu, Y.L., Luo, J., Filtz, T.M., and Woodcock, E.A. (2011). Phospholipase Cbeta1b associates with a Shank3 complex at the cardiac sarcolemma. FASEB J. 25, 10401047. Hammond, G.R.V., Fischer, M.J., Anderson, K.E., Holdich, J., Koteci, A., Balla, T., and Irvine, R.F. (2012). PI4P and PI(4,5)P2 are essential but independent lipid determinants of membrane identity. Science 337, 727730. Higazi, D.R., Fearnley, C.J., Drawnel, F.M., Talasila, A., Corps, E.M., Ritter, O., McDonald, F., Mikoshiba, K., Bootman, M.D., and Roderick, H.L. (2009). Endothelin-1-stimulated InsP3-induced Ca2+ release is a nexus for hypertrophic signaling in cardiac myocytes. Mol. Cell 33, 472482. Kapiloff, M.S., Schillace, R.V., Westphal, A.M., and Scott, J.D. (1999). mAKAP: an A-kinase anchoring protein targeted to the nuclear membrane of differentiated myocytes. J. Cell Sci. 112, 27252736. Kapiloff, M.S., Jackson, N., and Airhart, N. (2001). mAKAP and the ryanodine receptor are part of a multi-component signaling complex on the cardiomyocyte nuclear envelope. J. Cell Sci. 114, 31673176. Kelley, G.G., Reks, S.E., Ondrako, J.M., and Smrcka, A.V. (2001). Phospholipase C(): a novel Ras effector. EMBO J. 20, 743754. Keune, Wj., Bultsma, Y., Sommer, L., Jones, D., and Divecha, N. (2011). Phosphoinositide signalling in the nucleus. Adv. Enzyme Regul. 51, 9199. Knowlton, K.U., Michel, M.C., Itani, M., Shubeita, H.E., Ishihara, K., Brown, J.H., and Chien, K.R. (1993). The a 1A-adrenergic receptor subtype mediates biochemical, molecular, and morphologic features of cultured myocardial cell hypertrophy. J. Biol. Chem. 268, 1537415380. Kronebusch, P.J., and Singer, S.J. (1987). The microtubule-organizing complex and the Golgi apparatus are co-localized around the entire nuclear envelope of interphase cardiac myocytes. J. Cell Sci. 88, 2534. Kunkel, M.T., and Newton, A.C. (2010). Calcium transduces plasma membrane receptor signals to produce diacylglycerol at Golgi membranes. J. Biol. Chem. 285, 2274822752. Kunkel, M.T., Toker, A., Tsien, R.Y., and Newton, A.C. (2007). Calcium-dependent regulation of protein kinase D revealed by a genetically encoded kinase activity reporter. J. Biol. Chem. 282, 67336742. McKinsey, T.A. (2007). Derepression of pathological cardiac genes by members of the CaM kinase superfamily. Cardiovasc. Res. 73, 667677. Nakayama, H., Bodi, I., Maillet, M., DeSantiago, J., Domeier, T.L., Mikoshiba, K., Lorenz, J.N., Blatter, L.A., Bers, D.M., and Molkentin, J.D. (2010). The IP3 receptor regulates cardiac hypertrophy in response to select stimuli. Circ. Res. 107, 659666. Oestreich, E.A., Malik, S., Goonasekera, S.A., Blaxall, B.C., Kelley, G.G., Dirksen, R.T., and Smrcka, A.V. (2009). Epac and phospholipase Cepsilon regulate Ca2+ release in the heart by activation of protein kinase Cepsilon and calciumcalmodulin kinase II. J. Biol. Chem. 284, 15141522. Oestreich, E.A., Wang, H., Malik, S., Kaproth-Joslin, K.A., Blaxall, B.C., Kelley, G.G., Dirksen, R.T., and Smrcka, A.V. (2007). Epac-mediated activation of phospholipase C() plays a critical role in b-adrenergic receptor-dependent
enhancement of Ca2+ mobilization in cardiac myocytes. J. Biol. Chem. 282, 54885495. Ram, R., Mickelsen, D.M., Theodoropoulos, C., and Blaxall, B.C. (2011). New approaches in small animal echocardiography: imaging the sounds of silence. Am. J. Physiol. Heart Circ. Physiol. 301, H1765H1780. Ramazzotti, G., Faenza, I., Fiume, R., Matteucci, A., Piazzi, M., Follo, M.Y., and Cocco, L. (2011). The physiology and pathology of inositide signaling in the nucleus. J. Cell. Physiol. 226, 1420. Rhee, S.G., Suh, P.G., Ryu, S.H., and Lee, S.Y. (1989). Studies of inositol phospholipid-specic phospholipase C. Science 244, 546550. Rockman, H.A., Ross, R.S., Harris, A.N., Knowlton, K.U., Steinhelper, M.E., Field, L.J., Ross, J., Jr., and Chien, K.R. (1991). Segregation of atrial-specic and inducible expression of an atrial natriuretic factor transgene in an in vivo murine model of cardiac hypertrophy. Proc. Natl. Acad. Sci. USA 88, 8277 8281. Rockman, H.A., Koch, W.J., and Lefkowitz, R.J. (2002). Seven-transmembrane-spanning receptors and heart function. Nature 415, 206212. Rohde, H.M., Cheong, F.Y., Konrad, G., Paiha, K., Mayinger, P., and Boehmelt, G. (2003). The human phosphatidylinositol phosphatase SAC1 interacts with the coatomer I complex. J. Biol. Chem. 278, 5268952699. Rozengurt, E. (2011). Protein kinase D signaling: multiple biological functions in health and disease. Physiology (Bethesda) 26, 2333. Seifert, J.P., Wing, M.R., Snyder, J.T., Gershburg, S., Sondek, J., and Harden, T.K. (2004). RhoA activates puried phospholipase C- by a guanine nucleotide-dependent mechanism. J. Biol. Chem. 279, 4799247997. Smrcka, A.V., Hepler, J.R., Brown, K.O., and Sternweis, P.C. (1991). Regulation of polyphosphoinositide-specic phospholipase C activity by puried Gq. Science 251, 804807. Smrcka, A.V., Brown, J.H., and Holz, G.G. (2012). Role of phospholipase C in physiological phosphoinositide signaling networks. Cell. Signal. 24, 1333 1343. Sohal, D.S., Nghiem, M., Crackower, M.A., Witt, S.A., Kimball, T.R., Tymitz, K.M., Penninger, J.M., and Molkentin, J.D. (2001). Temporally regulated and tissue-specic gene manipulations in the adult and embryonic heart using a tamoxifen-inducible Cre protein. Circ. Res. 89, 2025. Steinberg, S.F. (2012). Regulation of protein kinase D1 activity. Mol. Pharmacol. 81, 284291. Tassin, A.M., Paintrand, M., Berger, E.G., and Bornens, M. (1985). The Golgi apparatus remains associated with microtubule organizing centers during myogenesis. J. Cell Biol. 101, 630638. Varnai, P., Thyagarajan, B., Rohacs, T., and Balla, T. (2006). Rapidly inducible changes in phosphatidylinositol 4,5-bisphosphate levels inuence multiple regulatory functions of the lipid in intact living cells. J. Cell Biol. 175, 377382. Wang, H., Oestreich, E.A., Maekawa, N., Bullard, T.A., Vikstrom, K.L., Dirksen, R.T., Kelley, G.G., Blaxall, B.C., and Smrcka, A.V. (2005). Phospholipase C modulates b-adrenergic receptor-dependent cardiac contraction and inhibits cardiac hypertrophy. Circ. Res. 97, 13051313. Wu, X., Zhang, T., Bossuyt, J., Li, X., McKinsey, T.A., Dedman, J.R., Olson, E.N., Chen, J., Brown, J.H., and Bers, D.M. (2006). Local InsP3-dependent perinuclear Ca2+ signaling in cardiac myocyte excitation-transcription coupling. J. Clin. Invest. 116, 675682. Zhang, L., Malik, S., Kelley, G.G., Kapiloff, M.S., and Smrcka, A.V. (2011). Phospholipase C scaffolds to muscle-specic A kinase anchoring protein (mAKAPbeta) and integrates multiple hypertrophic stimuli in cardiac myocytes. J. Biol. Chem. 286, 2301223021.
Metformin Retards Aging in C. elegans by Altering Microbial Folate and Methionine Metabolism
,1 Tahereh Noori,1 Filipe Cabreiro,1 Catherine Au,1,4 Kit-Yi Leung,2,4 Nuria Vergara-Irigaray,1 Helena M. Cocheme David Weinkove,3 Eugene Schuster,1 Nicholas D.E. Greene,2 and David Gems1,*
of Healthy Ageing, and G.E.E., University College London, London WC1E 6BT, UK Development Unit, Institute of Child Health, University College London, London WC1N 1EH, UK 3School of Biological and Biomedical Sciences, Durham University, Durham DH1 3LE, UK 4These authors contributed equally to this study *Correspondence: david.gems@ucl.ac.uk http://dx.doi.org/10.1016/j.cell.2013.02.035
2Neural 1Institute
SUMMARY
The biguanide drug metformin is widely prescribed to treat type 2 diabetes and metabolic syndrome, but its mode of action remains uncertain. Metformin also increases lifespan in Caenorhabditis elegans cocultured with Escherichia coli. This bacterium exerts complex nutritional and pathogenic effects on its nematode predator/host that impact health and aging. We report that metformin increases lifespan by altering microbial folate and methionine metabolism. Alterations in metformin-induced longevity by mutation of worm methionine synthase (metr-1) and S-adenosylmethionine synthase (sams-1) imply metformin-induced methionine restriction in the host, consistent with action of this drug as a dietary restriction mimetic. Metformin increases or decreases worm lifespan, depending on E. coli strain metformin sensitivity and glucose concentration. In mammals, the intestinal microbiome inuences host metabolism, including development of metabolic disease. Thus, metformin-induced alteration of microbial metabolism could contribute to therapeutic efcacyand also to its side effects, which include folate deciency and gastrointestinal upset.
INTRODUCTION Metformin is the worlds most widely prescribed drug, as an oral antihyperglycemic agent for type 2 diabetes (T2D) and in the treatment of metabolic syndrome. However, the real and potential benets of metformin therapy go beyond its prescribed usage, including reduced risk of cancer (Dowling et al., 2011) and, in animal models, delayed aging, an effect seen in rodents (Anisimov et al., 2011) and in the nematode Caenorhabditis elegans (Onken and Driscoll, 2010). The mechanisms underlying these positive effects remain unclear. One possibility is that met228 Cell 153, 228239, March 28, 2013 2013 Elsevier Inc.
formin recapitulates the effects of dietary restriction (DR), the controlled reduction of food intake that can improve late-life health and increases lifespan in organisms ranging from nematodes and fruit ies to rodents and rhesus monkeys (Mair and Dillin, 2008). In mammals, the action of metformin is partly mediated by AMPK activation, which results in downregulation of TOR and the IGF-1/AKT pathways to reduce energy-consuming processes (Pierotti et al., 2012). An unexplored possibility is that metformin alters mammalian physiology via its effects on gut microbiota (Bytzer et al., 2001). The gut microbiome (or microbiota) plays a major role in the effects of nutrition on host metabolic status (Nicholson et al., 2012), as well as contributing to metabolic disorders such as obesity, diabetes, metabolic syndrome, autoimmune disorders, inammatory bowel disease, liver disease, and cancer (Delzenne and Cani, 2011; Kau et al., 2011; Nicholson et al., 2012). It may also inuence the aging process (Ottaviani et al., 2011). It has been argued that the host and its symbiotic microbiome acting in association (holobiont) should be considered as a unit of selection in evolution (Zilber-Rosenberg and Rosenberg, 2008). Coevolution of microbiota facilitates host adaptation by enabling e.g., nutrient acquisition, vitamin synthesis, xenobiotic detoxication, immunomodulation, and gastrointestinal maturation. In return, the host provides a sheltered incubator with nutrients ckhed et al., 2005). Thus, the two components of the holo(Ba biont are symbiotic, but microbiota can also be commensal or pathogenic. Dening interactions between drug therapy, microbiome and host physiology is experimentally challenging given the complex and heterogeneous nature of mammalian gut microbiota. Here simple animal models amenable to genetic manipulation can be helpful. For example, in the fruit y Drosophila, microbiota modulates host development and metabolic homeostasis via the TOR pathway (Storelli et al., 2011). C. elegans is particularly convenient for such studies because under standard culture conditions only a single microbe is present (as a food source): the human gut bacterium Escherichia coli (Brenner, 1974). Active bacterial metabolism is a critical nutritional requirement for C. elegans, the absence of which retards development and extends lifespan (Lenaerts et al., 2008).
A 1.0
0.8 Fraction alive 0.6 0.4 0.2 0 0 Wild type Metformin 10 20 30 40 50 60 0 mM 25 mM 50 mM 100 mM
B 1.0
0.8 Fraction alive 0.6 0.4 0.2 0 0 Wild type Phenformin 0 mM 1.5 mM 3 mM 4.5 mM
Figure 1. The Biguanide Drugs Phenformin and Metformin Decelerate Aging in C. elegans
(A and B) Metformin (A) and phenformin (B) extend lifespan in a dose-dependent manner. Phenformin alters fecundity and reduces body size (see Figures S1AS1D). (C) Phenformin (4.5 mM) does not increase lifespan in the presence of 50 mM metformin, consistent with similar mechanism of drug action. (D) Metformin decreases the exponential increase in age-related mortality (for survival curve see Figure S1E). (E) Later-life administration (day 8) of metformin increases lifespan at lower concentrations (25, but not 50 or 100 mM). See also Figure S1. For statistics, see Table S1.
10
20
30
40
50
C 1.0
0.8 Fraction alive 0.6 0.4 0.2 0 0 Wild type Metformin 10 20 0 mM 50 mM Metformin 4.5 mM Phenf. + 50 mM Metf. 4.5 mM Phenformin
D
Ln mortality
0 -2 -4 -6 -8 0 mM 50 mM 0 10 20 Days 30 40 50 Wild type Metformin
30 0 mM
40
50
E
Fraction alive
1.0 0.8 0.6 0.4 0.2 0 0 Wild type Metformin 10 20 30 Day 8
25 mM 50 mM 100 mM
Days
40
50
Moreover, worms are sometimes long-lived on mutant E. coli with metabolic defects (Saiki et al., 2008; Virk et al., 2012) and on microbial species thought to enhance human health, e.g., from the genera Lactobacillus and Bidobacterium (Ikeda et al., 2007). These observations suggest that E. coli plays a more active role in C. elegans nutrition and metabolism than as a mere food source, and in some respects acts as microbiota (Lenaerts et al., 2008). C. elegans has also been used extensively to identify genes that specify endocrine, metabolic, and dietary regulation of aging (Kenyon, 2010). In this study, we examine the mechanism by which metformin extends lifespan in C. elegans. We report that its effects are mediated by the cocultured E. coli, where metformin inhibits bacterial folate and methionine metabolism. This, in turn, leads to altered methionine metabolism in the worm, and increased lifespan. These ndings reveal how drug action on host-microbiome interactions can impact health and longevity. RESULTS Extension of C. elegans Lifespan by Metformin Is Mediated by Live E. coli We rst veried the effects on worm lifespan of metformin, and also the more potent biguanide drug phenformin. Metformin at 25, 50, and 100 mM increased mean lifespan by 18%, 36%, and 3% (Figure 1A; Table S1 available online). Phenformin at 1.5, 3, and 4.5 mM also increased lifespan, by 5%, 21%, and 26% (Figure 1B; Table S1). As expected, maximal effects on lifespan of these pharmacologically similar drugs were nonadditive
(Figure 1C; Table S1). Metformin reduced the exponential age increase in mortality rate (Figure 1D), demonstrating that it slows aging (at least until day 18) rather than reducing risk of death. Metformin also modestly increased mean lifespan when administered from middle age onward, but only at 25 mM (+8%, p < 0.001; Figure 1E; Table S1). In most trials, the DNA replication inhibitor FUdR was used to prevent progeny production, but effects of metformin on lifespan are not FUdR-dependent (Figures S1F and S1G; Table S1) (Onken and Driscoll, 2010). These results conrm the robust effects of biguanide drugs on aging in C. elegans. Interventions altering E. coli can affect C. elegans lifespan (Garigan et al., 2002; Gems and Riddle, 2000; Saiki et al., 2008). To test the possibility that metformin increases worm lifespan by altering the E. coli, we assessed its effects in the absence of bacteria (axenic culture). As expected, culture on axenic medium (Lenaerts et al., 2008) and bacterial deprivation (Kaeberlein et al., 2006) caused an increase in worm lifespan, typical of DR. Under these conditions, metformin did not increase worm lifespan, but instead markedly reduced it (Figures 2A, S2A, and S2B; Table S2). UV-irradiation of E. coli impairs bacterial viability and extends worm lifespan without reducing fertility, suggesting a mechanism distinct from DR (Gems and Riddle, 2000). Under these conditions, metformin still shortened lifespan (16%, p < 0.001; Figure 2B; Table S2). Next, we raised E. coli in the presence of metformin and then transferred it to drug-free agar plates. Drug pretreatment of E. coli robustly extended worm lifespan (+33%, p < 0.001; Figure 2C; Table S2). We conclude that the life-extending effect of metformin is mediated by live E. coli. Moreover, in the absence of E. coli, metformin shortens C. elegans lifespan, likely reecting drug toxicity. One possibility is that metformin extends worm lifespan by reducing E. coli pathogenicity. Proliferating E. coli block the alimentary canal in older worms, and antibiotic treatment can both prevent this proliferation and increase worm lifespan
A
Fraction alive
1.0 0.8 0.6 0.4 Wild type 0.2 Metformin 0 0 10 20 30 40 50 60
Fraction alive
Live OP50 0 mM 50 mM Axenic (no bacteria) 0 mM 50 mM
1.0 0.8 0.6 0.4 0.2 0 0 Wild type Metformin 10 20 30
Live OP50 0 mM 50 mM UV OP50 0 mM 50 mM
Figure 2. Metformin Extending Effects on C. elegans Lifespan Require Live Bacteria

(A) Metformin shortens lifespan of C. elegans cultured axenically (i.e., in the absence of E. coli). (B) Metformin shortens lifespan of C. elegans cultured on UV-irradiated E. coli (OP50). (C) Metformin pretreatment of bacteria is sufcient to extend lifespan. (D) Metformin extends lifespan in the absence of E. coli proliferation (blocked by carbenicillin). (E) Metformin extends lifespan in the presence of the less pathogenic bacterium Bacillus subtilis. (F) Retardation of bacterial growth by metformin, monitored over an 18 hr period. (G) Biguanide drugs cause altered bacterial lawn morphology. (H) Bacterial viability is reduced by carbenicillin and UV treatment, but not metformin. (I) Metformin extends lifespan in the presence of multi-antibiotic resistant E. coli OP50-R26. See also Figure S2. For statistics, see Table S2.
70
80
40 Live OP50 0 mM 50 mM Carb OP50 0 mM 50 mM
50
C
Fraction alive
1.0 0.8 0.6 0.4 0.2 0 0 Wild type Metformin 10 20 30
D
Fraction alive Pre-treated OP50 0 mM 50 mM
1.0 0.8 0.6 0.4 0.2 0 0 0.5 0.4 0.3 0.2 Wild type Metformin 10 20 30
40
50
Days
40
50
1.0 0.8
Fraction alive
B. subtilis 0 mM 50 mM
0.6 0.4 Wild type Metformin
an antibiotic. Notably, the drug concentration thresholds for bacterial and worm 0.1 0.2 lifespan effects were similar, and also 0 pH-dependent (Figures S2G and S2H 0.0 0 200 400 600 800 1000 0 10 20 30 40 50 and Table S2). Time (min) Days G H 16 I We then asked if the antibiotic effects 1.0 of metformin were bacteriocidal or bacteOP50 Control 12 0 mM (no drug) 0.8 riostatic. When subcultured from metfor50 mM min plates, E. coli showed no reduction 0.6 OP50-R26 4.5 mM 8 0 mM in colony forming units (Figure 2H), phenformin 0.4 50 mM implying that metformin has bacterio4 Wild type Metformin 100 mM 0.2 static rather than bacteriocidal effects. metformin 0 To probe whether metformin acts via 0 50 0 50 0 mM 0 0 10 20 30 40 50 metformin Control Carbenicillin UV one of the major, known antibiotic mechDays anisms, we employed the R26 P-group plasmid that confers resistance to carbe(Garigan et al., 2002). To determine whether metformin extends nicillin, neomycin, kanamycin, tetracycline, streptomycin, gentaworm lifespan by preventing E. coli proliferation, we tested its micin, mercuric ions, and sulfonamides. However, metformin still effects in the presence of carbenicillin. This antibiotic is bacterio- extended lifespan in worms on R26-transformed E. coli (39%, static, blocking bacterial proliferation without greatly reducing its p < 0.001; Figure 2I and Table S2). What is the property of E. coli whose alteration by metformin viability. Metformin increased lifespan to a similar degree in the absence (+25%) or presence (+24%) of carbenicillin (p < 0.001; increases worm lifespan? Coenzyme Q (ubiquinone) deciency Figure 2D; Table S2). Thus, metformin does not increase lifespan in E. coli increases C. elegans lifespan due to impairment of by preventing bacterial proliferation. Culture of C. elegans with bacterial respiration (Saiki et al., 2008). We therefore tested Bacillus subtilis increases lifespan (Garsin et al., 2003), suggest- whether metformin can increase lifespan of worms on Q-deing that this microbe is less pathogenic to C. elegans than E. coli. cient ubiG mutant E. coli and found that it does (+20%, p < Metformin increased lifespan of worms cultured on B. subtilis 0.001; Figure 3A). We then tested whether metformin reduces (+9%, p < 0.001; Figure 2E; Table S2). These ndings suggest respiration rate in E. coli OP50. Although metformin transiently that reduced bacterial pathogenicity is not the cause of metfor- reduced respiration rate, long-term exposure increased it (Figure 3B). Taken together, these ndings suggest that metformins min-induced longevity. effect on worm lifespan is not caused by inhibition of bacterial respiration. Biguanides Have Bacteriostatic Effects at Lipopolysaccharides (LPS) are the major component of the Concentrations that Increase Lifespan Biguanides induced a dose-dependent inhibition of E. coli prolif- outer wall of Gram-negative bacteria. The structure of E. coli eration (Figures 2F and S2C) and an alteration in bacterial lawn LPS can affect C. elegans lifespan (Maier et al., 2010). To test morphology (Figure 2G). Similar results were obtained with whether metformin action is dependent upon E. coli LPS type, B. subtilis (Figures S2DS2F). Thus, metformin can also act as we looked at worm lifespan on seven E. coli strains with a variety
C.F.U. per ml (x103)
Bacterial lawn
Fraction alive
Bacterial Growth O.D (595nm)
OP50 0 mM 25 mM 50 mM 100 mM Metformin
A
1.0 Fraction alive 0.8 0.6 0.4 Wild type 0.2 Metformin E. coli K-12 0 0 10 HT115 0 mM 50 mM GD1 0 mM 50 mM
B
Bacterial respiration (nmol O/min/OD)
350 300 250 200 150 100 50 n.s n.s Transient Long-term
** n.s ***
Figure 3. Metformin Effects on C. elegans Lifespan Correlate with Effects of Metformin on Bacterial Growth
(A) Metformin extends lifespan in the presence of respiratory-decient E. coli strain GD1. (B) Growth in the presence of metformin does not impair respiration in E. coli OP50. (CE) Effects on lifespan are independent of bacterial subgroup (B or K-12) and lipopolysaccharide (LPS) structure. K-12 strains possess longer LPS structures than B strains and CS2429. CS2429 is an LPS truncated mutant derived from isogenic parent strain CS180. HB101 is a B/K-12 hybrid. (F) Relationship between bacterial growth inhibition by metformin (50 mM) and effects on lifespan among different E. coli strains. (G) OP50-MR E. coli is resistant to growth inhibition by metformin. This strain also shows crossresistance to phenformin (Figures S3CS3E). (H) Metformin shortens lifespan in the presence of OP50-MR. Error bars represent SEM. *p < 0.05; **p < 0.01; ***p < 0.001. See also Figure S3. For statistics, see Table S3.
20
30
40
50 OP50 0 mM 50 mM BL21G 0 mM 50 mM
60
0 50 mM Metformin
- + - + +
OP50
- + - + +
GD1
C 1.0
0.8 Fraction alive 0.6 0.4 Wild type 0.2 Metformin E. coli B 0 0 10 1.0 0.8 Fraction alive 0.6 0.4 Wild type 0.2 Metformin E. coli B/K-12 0 0 10
D 1.0
0.8 Fraction alive 0.6 0.4 Wild type 0.2 Metformin E. coli K-12 0 0 10 80 70 60 50 40 -60 CS180 0 mM 50 mM CS2429 0 mM 50 mM
20
30
40 HB101 0 mM 50 mM
50
20
Bacterial growth ratio (% of growth 50mM Metf./ control)
Days
30
40
50
MG1655 BL21-G OP50-MR HT115 CS2429 HB101 GD1
y=-0.199x+63.99 R2=0.82; p=0.0007 -40
OP50 CS180
20
Days
30
40
50
Bacterial growth ratio (% of non-treated control)
100 80 60 40 20 0 0
**
**
***
Fraction alive
**
OP50 OP50-MR
-20 0 20 40 60 Lifespan ratio (% of the mean 50mM Metf./ control) OP50-MR 0 mM 50 mM
1.0 0.8 0.6 0.4 E.coli B 0.2 Wild type Metformin 0 0 10
predominate. However, inhibition of bacterial proliferation per se is not the cause of worm life extension, as already shown (Figure 2D; Table S2).
25 100 50 mM Metformin
150
of LPS structures. Although effects of metformin on worm lifespan differed between E. coli strains (Figures 3A3E and S3A and Table S3), this variation did not correlate with the E. coli LPS type. Interestingly, among E. coli strains there was a strong positive correlation between the capacity of metformin to increase worm lifespan and to inhibit bacterial growth (R2 = 0.82, p < 0.0007; Figure 3F). There was no correlation between bacterial metformin sensitivity and effect on worm lifespan in the absence of metformin (R2 = 9.6 3 105, p = 0.98; Figure S3B). This suggests that the capacity of the drug to extend worm lifespan is a function of the microbial sensitivity to growth inhibition by metformin. To test this directly, we isolated a metformin-resistant OP50 derivative (OP50-MR) (Figures 3G and S3CS3E) that proved to contain eight mutations (see Extended Experimental Procedures). As predicted, on this strain 50 mM metformin shortened worm lifespan (37%, p < 0.001; Figure 3H). We conclude that in metformin-resistant E. coli strains, life-shortening toxic effects
Metformin Disrupts Folate Metabolism in E. coli It was recently discovered that C. elegans live longer on an E. coli mutant with reduced folate levels (aroD) (Virk et al., 2012). Moreover, metformin can de20 30 40 crease folate levels in patients (Sahin Days et al., 2007). We therefore asked whether metformin increases worm lifespan by altering bacterial folate metabolism. Folates are B-group vitamins whose structure incorporates a pteridine ring, p-aminobenzoic acid (pABA), and glutamic acid(s). Folates are typically present as the reduced forms, dihydrofolate (DHF) and tetrahydrofolate (THF). THF can be substituted with a variety of one-carbon units (including formyl and methyl groups) that function as a coenzyme in metabolic reactions involving transfer of one-carbon moieties (Figure 4A). These are involved in the biosynthesis of purines and pyrimidines, in amino acid interconversions, and for the provision of methyl groups in methylation reactions (Kwon et al., 2008). Metformin markedly changed the folate composition in OP50 (Figure 4B), as detected by LC-MS/MS. It increased levels of 5-methyl-THF (+116%, p = 2.5 3 106), 5,10-methylene-THF (+99%, p = 5.9 3 106), and DHF (+38%, p = 7.1 3 104), whereas levels of the remaining folates were decreased. It also increased folate polyglutamylation, particularly n = 6 and 7 glutamates (Figures 4C, S4A, and S4B; Table S4). Folate
A
5-Methyl THF SAH SAMe Hcy MS
SAMS Met
MTHFR
dUMP Thymidylate synthesis 5,10-Methylene THF dTMP 5,10-Methenyl THF 10-Formyl THF Glu pABA GTP DHF pABGlu pteridines Folic acid
Figure 4. Metformin Folate Metabolism
Inhibits
Bacterial
THF Proteins
Purine synthesis DHFR
Trimethoprim (TRI)
B
% of total folate
40 30 n.s
***
n.s
*** *** *** * *

Ctrl Res 5-Methyl THF
**
20 10 0
**
OP50 (Ctrl) 0 mM 50 mM OP50-MR (Res) 0 mM 50 mM
***
n.s
Ctrl Res DHF
Ctrl Res THF
Ctrl Res Ctrl Res Ctrl Res Formyl Methenyl Methylene THF THF THF Folate metabolites (Glun=1-7) in E. coli
C
% of total 5-Methyl-THF
50 40
30 OP50-MR 0 mM 50 mM 20 10 0
Fraction alive
OP50 0 mM 50 mM
***
1.0 0.8 0.6 0.4 0.2
Trimethoprim (g/ml) 0 0.1 0.2 0.5 1.0
(A) The folate and methionine cycles. Metabolites analyzed, red; enzymes, blue; supplements, purple. DHF, dihydrofolate; DHFR, dihydrofolate reductase; Glu, glutamate; Hcy, homocysteine; Met, methionine; MS, methionine synthetase; MTHFR, methylenetetrahydrofolate reductase; pABA, p-aminobenzoic acid; SAH, S-adenosylhomocysteine; SAMe, S-adenosylmethionine; SAMS, S-adenosylmethionine synthase; THF, tetrahydrofolate; TRI, trimethoprim. Dotted lines represent feedback loops. (B) Metformin alters folate homeostasis in E. coli OP50 but not OP50-MR. The values for each metabolite are the sum of the values for the different glutamate side chains (17) divided by sum of all folate metabolites measured. (C) Metformin alters 5-methyl-THF polyglutamylation in OP50 but not OP50-MR. (D) The DHFR inhibitor TRI increases C. elegans lifespan in a dose-dependent manner. See Figure S4D for E. coli growth retardation by TRI. (E) Effects of metformin and TRI on lifespan are nonadditive, consistent with similar modes of action. (F) Principal component analysis (Metaboanalyst) of OP50 metabolites with TRI and metformin. Note that TRI abolishes effects of metformin. Error bars represent SEM. *p < 0.05; **p < 0.01; ***p < 0.001. See also Figure S4. For statistics, see Table S4.
*** ***
1
OP50 0 10 20 Days 30 40
(0.2, 0.5, and 1 mg/ml) increased lifespan by 16%, 30%, and 38%, respectively E F (p < 0.001) (Figure 4D). By contrast, in *** 27.5 n.s the presence of 50 mM metformin, 0.2 mg/ml TRI caused only a slight n.s 25 *** TRI (1 g/ml) increase in lifespan (+8%, p < 0.001), Control TRI+Metf (no treatment) whereas at higher concentrations it either 22.5 *** had no effect (0.5 mg/ml TRI, 2%, p = 0.17) or reduced lifespan (1 mg/ml TRI, 20 Metformin *** Metformin *** 19%, p < 0.001; Figure 4E). Such 0 mM (50 mM) nonadditivity was recapitulated in the n.s 50 mM 17.5 lack of effect of metformin on metabolic 1000 -1000 0 0 0.1 0.2 0.5 1.0 proles of OP50 when cotreated with PC 1 (48.2 %) g/ml Trimethoprim 1 mg/ml TRI (Figures 4F and S4E). These nonadditive effects of metformin and TRI imply a shared mechanism of action, sugpolyglutamylation increases their retention in the cell, and gesting that altered bacterial folate metabolism by metformin bioavailability for reactions involving folate-dependent enzymes increases worm lifespan. (Kwon et al., 2008). By contrast, in the resistant strain OP50-MR, metformin did not affect polyglutamylation (Figures 4C and S4C), Metformin Disrupts C. elegans Methionine Metabolism or DHF levels. 5-Methyl-THF and 5,10-methylene-THF were To explore whether metformin-induced alterations in microbial still increased (Figure 4B), but by only 29% (p = 0.018) and folate metabolism increase host lifespan by altering worm folate 17% (p = 0.003). Genome sequencing of OP50-MR revealed metabolism, we rst examined worm folate proles under standard culture conditions (agar plates with E. coli OP50). In a mutation in glyA, which encodes a folate cycle enzyme. To explore whether metformin effects on bacterial folate worms, as in humans, 5-methyl-THF was the predominant folate metabolism affect worm lifespan, we used the antibiotic trimeth- (59%) and treatment with metformin did not alter the ratio of oprim (TRI) that inhibits dihydrofolate reductase (DHFR). TRI different folate forms (Figure 5A). However, it did decrease
2 3 4 5 6 7 5-Methyl THF (Glun=1-7) in E. coli PC 2 ( 19.3 %)
Mean lifespan (Days)
-1000
1000
***
***
***
***
***
% (Fol-Glun=1-3/Fol-Glun=1-7)
A
% of total folate
70 60 50 40 30 20 10 0
0 mM 50 mM
60 50 40 30 20 10 0
0 mM 50 mM
p=0.057
Figure 5. Effect of Metformin on the Methionine Cycle but Not the Folate Cycle in C elegans
(A) Effect of metformin on C. elegans/E. coli system: little effect on nematode folate homeostasis. (B) Metformin induces a shift toward shorter-chain (n = 13) glutamate folate forms in C. elegans. (C) Metformin increases S-adenosylmethionine (SAMe) levels in E. coli (OP50). (D) Mutation of metr-1(ok521) (methionine synthetase, MS) increases lifespan only in the presence of metformin. (E) In C. elegans, metformin greatly reduces SAMe levels and increases S-adenosylhomocysteine (SAH) levels. (F) Metformin shortens lifespan in S-adenosylmethionine synthase-decient sams-1 (ok3033) mutants. Error bars represent SEM. *p < 0.05; **p < 0.01; ***p < 0.001. See also Figure S5. For statistics, see Table S5.
n.s
**
DHF
Methylene 5-Methyl THF THF n=1-7 Fol-Glu in N2 worms 0.0008 0.0006 0.0004 0.0002 0 0 50 Metformin
p=0.055
THF
DHF
THF
Methylene 5-Methyl THF THF Wild type 0 mM 50 mM metr-1(ok521) 0 mM 50 mM
C
SAMe (nmol/mg) in E. coli SAMe/SAH in E. coli
D
1000 800 600 400 200 0 0 50 mM Metformin
p=0.065
**
SAH (nmol/mg) in E. coli
0.8 0.6 0.4 0.2 0
1.0 0.8 0.6 0.4 0.2 0 0 OP50 Metformin 10 20 30
0 50 Metformin
Fraction alive
40
50
E
SAMe (nmol/mg) in N2 worms SAH (nmol/mg) in N2 worms SAMe/SAH in N2 worms
F
5 4 3 2 1 0 0 50 Metformin
**
0.25 0.2 0.15 0.1 0.05 0
**
50 40 30 20 10 0 0 50 mM Metformin
Fraction alive
70 60
1.0
**
0.8 0.6 0.4 0.2 0 0 OP50 Metformin 10
0 50 Metformin
glutamate chain length (n = 13) (Figures 5B and S5; Table S5), suggesting a possible change in the activity of folate-dependent enzymes. Thus, disruption of microbial folate metabolism increases host lifespan but with little effect on host folate levels. One possibility is that products of other E. coli folate-associated pathways inuence C. elegans lifespan. Inhibition of bacterial methionine synthase (MS) causes 5-methyl-THF accumulation via the methyl trap mechanism, so-called because of the irreversible conversion of 5,10-methylene-THF to 5-methyl-THF (Mato et al., 2008) (Figure 4A). Consistent with MS inhibition, metformin not only strongly increased 5-methyl-THF levels but also reduced levels of THF (14%, p = 0.003) (Figure 4B). Metformin also impaired the bacterial methionine cycle, causing an 86% increase in S-adenosylmethionine (SAMe) levels (p = 0.0032) and a 33% increase of S-adenosylhomocysteine (SAH) (p = 0.055; Figures 4A and 5C), consistent with the lack of homocysteine (Hcy) remethylation if MS is inhibited. SAMe, the major corepressor of genes encoding enzymes of methionine biosynthesis, also inhibits the folate cycle and reduces methionine production by blocking methylene-THF reductase (MTHFR) (Banerjee and Matthews, 1990) (Figure 4A). Thus, the accumulation of the substrates SAMe, SAH, 5-methyl-THF, and 5,10-
methylene-THF, and reduction of the product THF imply that metformin also reduces microbial methionine availability. This suggests that metformin might sams-1(ok3033) increase lifespan by reducing levels of 0 mM 50 mM bacterial-derived methionine in the host. To explore this, we employed a C. elegans MS mutant, metr-1(ok521), which cannot synthesize methionine and is therefore 20 30 40 50 wholly dependent upon exogenous Days methionine (Hannich et al., 2009). In the absence of metformin, metr-1 did not increase worm lifespan (p = 0.85; Figure 5D). Interestingly however, metr-1 did increase lifespan in the presence of 50 mM metformin (+67%, p < 0.001; Figure 5D). Thus, metr-1 sensitizes C. elegans to the life-extending effects of metformin. This suggests that microbes are the main source of dietary methionine, but the worms also synthesize some methionine of their own using METR-1. Thus, effects of metr-1 on lifespan are only detected when dietary methionine levels are reduced. Supporting this scenario, metformin treatment lowered SAMe levels in C. elegans (72%, p = 0.005) and increased SAH levels (+181%, p = 0.002; Figure 5E). In summary, in E. coli metformin increases SAMe and 5-methyl-THF. By contrast, in C. elegans it decreases SAMe and the SAMe/SAH ratio without affecting 5-methyl-THF levels. In C. elegans, SAMe is synthesized by the SAMe synthase SAMS-1, RNAi knockdown of which extends lifespan (Hansen et al., 2005). Notably, sams-1 RNAi does not increase eat-2 mutant lifespan, suggesting a shared mechanism with eat-2induced DR (Ching et al., 2010; Hansen et al., 2005). If metformin increases lifespan by the same mechanism as loss of sams-1, then metformin should not increase lifespan in the absence of sams-1. To test this, we employed a sams-1(ok3033) null mutant that, as expected, extended lifespan (+35%, p < 0.001;
Wild type 0 mM 50 mM
p-AAK-2/actin expression (a.u.)
2.5 2.0 1.5 1.0 0.5 0
B
* * *
Fraction alive 1.0 0.8 0.6 0.4 0.2 Phenformin 0 0 1.5 3 4.5 mM Phenformin 0 10 20 Days 30 40 50 Wild type 0 mM 4.5 mM aak-2(ok524) 0 mM 4.5 mM
Figure 6. AMP kinase and SKN-1 Protect Against Biguanide Toxicity

(A) Phenformin increases pAAK-2 levels, suggesting AMPK activation (2-day-old adults). (B) Phenformin shortens lifespan in aak-2(ok524) AMPK loss-of-function mutants. (C) Phenformin does not extend lifespan in skn1(zu135) mutants. (D) AMPK-dependent induction of expression by phenformin of SKN-1-activated reporter gst-4::gfp in L4 animals. (E) aak-2(ok524) but not daf-16(mgDf50) mutants are hypersensitive to growth inhibition by metformin, as measured by the food clearance assay. (F) skn-1(zu135) increases sensitivity to growth inhibition by metformin. (G) Metformin increases expression of gst-4::gfp under conditions that do not increase lifespan (maintenance on E. coli HT115). (H) Life extension by metformin pretreatment of E. coli is partially AMPK-dependent. (I) Life extension by metformin pretreatment of E. coli is not SKN-1-dependent. Error bars represent SEM of at least three independent biological replicates. *p < 0.05; **p < 0.01; ***p < 0.001. See also Figure S6. For statistics, see Table S6.
p-AAK Actin
C
1.0 0.8 Fraction alive 0.6 0.4 0.2 0 0 Phenformin 10 20 Days 30 40 50 Wild type 0 mM 4.5 mM skn-1(zu135) 0 mM 4.5 mM
D
gst-4::gfp expression (a.u.)/ area of worm (m2) N2 - 0 mM 200 N2 - 4.5 mM
***
150 100 50 0 0 4.5 WT 0
n.s
mM 4.5 Phenformin
aak-2(ok524)
E
Time to drop to 50% of food available (% of non-treated) 250 200 150 100 50 0 0 25 50 mM Metformin 100 Wild type aak-2(ok524) daf-16(mgDf50)
F
100
48 h 60 h
72 h 84 h
G
gst-4::gfp expression (a.u)/ worm area (m2) 500 400 300 200 100
* ***
% of L4 worms
80 60 40 20 0 mM Metformin 0 50 0 50
***
***
WT skn-1(zu135)
H
1.0 0.8 Fraction alive 0.6 0.4 0.2 Metformin Pre-treated E. coli 0 0 10 20 Wild type 0 mM 50 mM aak-2(ok524) 0 mM 50 mM
I
1.0 0.8 Fraction alive 0.6 0.4 0.2 Metformin Pre-treated E. coli 0 0 10 20
Days
30
40
50
Figure 5F). Strikingly, in a sams-1 mutant, metformin reduced lifespan (38%, p < 0.001), reminiscent of the effect of metformin on eat-2 mutants (Onken and Driscoll, 2010). These results suggest the possibility that metformin and eat-2-induced DR act by similar disruptions of methionine-associated functions. AMP Kinase and SKN-1 Protect C. elegans Against Metformin Toxicity Metformin-induced longevity requires the worm AMP-dependent protein kinase (AMPK) (Onken and Driscoll, 2010). This is consistent with the fact that biguanide drugs activate AMPK (Hawley et al., 2003). However, if extension of C. elegans lifespan by biguanide drugs is mediated by E. coli, why should this effect require the worm AMPK? To explore this, we rst tested whether
biguanides activate worm AMPK, by measuring phosphorylation of Thr-172 in the worm AMPKa subunit AAK-2. Phen0 mM 0 50 formin, but not metformin, detectably Metformin WT HT115 increased pAMPK levels (Figures 6A and S6A), perhaps reecting the greater Wild type membrane permeability of phenformin. 0 mM We then veried the AMPK-dependence 50 mM of the effect of biguanides on worm lifeskn-1(zu135) 0 mM span in the presence of E. coli. Lifespan 50 mM in aak-2 mutants was not increased by either metformin (Figure S6B; Table S6), as previously noted (Onken and Driscoll, 2010), or phenformin (Figure 6B). In fact, 30 40 50 Days phenformin reduced lifespan (15%, p < 0.001; Table S6). Notably, the metformin-induced deceleration of the age increase in mortality rate was still present in aak-2 mutants, but initial mortality rates were markedly greater (Figure S6C), consistent with increased sensitivity to metformin toxicity. The life-extending effects of both biguanides also required the SKN-1 Nrf2 transcription factor (Figures 6C and S6H), and induced expression of the SKN-1 target gst-4 (glutathione S-transferase 4) in an AMPK-dependent fashion (Figures 6D and S6IS6K), consistent with previous ndings (Onken and Driscoll, 2010). Thus, both biguanides cause AMPK-dependent activation of SKN-1, and induce detoxication gene expression. Our ndings imply that the impact of metformin on worm lifespan reects the sum of indirect, E. coli-mediated life-extending effects and direct life-shortening effects. A possible interpretation of the AMPK and SKN-1 dependence of biguanide effects
A
1.0 0.8 Fraction alive 0.6 0.4 0.2 0 0 Wild type Metformin 10 20 Days 30 40 50 Control glucose 0 mM 50 mM
D-Glucose
B
Bacterial growth ratio (% of non-treated control) 100 80 60 40 20 0
***
Figure 7. High Glucose Diet Suppresses Metformin-Induced Life Extension

(A) Metformin decreases lifespan on 0.25% D-glucose. See Figure S7A for 1% D-glucose. (B) Metformin does not inhibit bacterial growth in the presence of 0.25% D-glucose. (C) Scheme summarizing direct and indirect effects of metformin on the C. elegans/E. coli system. Dotted lines indicate hypothetical feedback loops. Error bars represent SEM of at least three independent biological replicates. *p < 0.05; **p < 0.01; ***p < 0.001. See also Figure S7. For statistics, see Table S7.
(0.25%) 0 mM 50 mM
Metformin (mM) Ctrl 50 0 D-Glucose (%)
Ctrl 50 0.25%
Metformin
NH NH N NH 2
E. coli
NH N N NH
C. elegans
How might SAMe levels regulate AMPK? Increased levels of SAMe can nez-Chaninhibit AMPK activation (Mart tar et al., 2006). To probe this we tested whether longevity induced by sams-1 RNAi is AMPK-dependent, and this proved to be the case (Figure S6G). This suggests that metformin increases lifespan at least in part via the AMPK-activating effects of reduced SAMe levels.
Folate metabolism 5-MeTHF NH Methionine

2
Met 5-Methyl THF 5,10-Methylene THF Hcy NH NH MS N N NH Met

2
METR
5-MeTHF Hcy
SAMS-1
SAH SAMe
+ SAMS
SAMe/SAH
SAM
N
NH N
NH NH2
Metabolic DR
AAK-2 SKN-1
THF
LIFESPAN
on lifespan is that these proteins protect worms against drug toxicity. To test this, we compared growth inhibition by metformin in wild-type and mutant C. elegans using a food clearance assay. aak-2 and skn-1 but not daf-16 mutants showed increased sensitivity to growth inhibition by biguanides (Figures 6E, 6F, and S6D). Note that metformin-induced life extension is not daf-16-dependent (Onken and Driscoll, 2010). We also observed that metformin induced a similar level of gst-4 expression in worms on E. coli OP50 and HT115 (+29 and +30%, respectively) even though the drug increases lifespan only with the former strain (Figure 6G). These ndings further suggest that aak-2 and skn-1 protect worms against biguanide toxicity. To test this further, we raised E. coli with or without metformin, and then transferred it to metformin-free plates with carbenicillin to prevent further growth. Carbenicillin does not affect E. colimediated effects of metformin (Figure 2D). Notably, metforminpretreated E. coli caused a larger increase in mean lifespan in wild-type worms than aak-2 worms (+48 and +29%, respectively, p < 0.001, Figure 6H) but not skn-1 worms (+17 and +21%, respectively, p < 0.0001; Figure 6I). Moreover, extension of lifespan by blocking folate metabolism with 1 mg/ml TRI (Figure S6E) or by a folate- decient mutant E. coli aroD also appeared to be partially aak-2-dependent (Figure S6F). These results suggest that AMPK-dependence of life extension by metformin is partly due to resistance against drug toxicity, but also partly to AMPK mediation of microbial effects on the worm. By contrast, skn-1 activation appears to act solely by protecting against the life shortening effect of metformin.
Metformin Does Not Extend Lifespan on a High Glucose Diet Metformin is a treatment for hyperglycemia caused by diabetes. We wondered whether metformin is able to provide protection against high glucose levels, which can shorten worm lifespan (Lee et al., 2009). In fact, metformin proved unable to extend the lifespan of worms supplemented with 0.25% or 1% glucose (Figures 7A and S7A; Table S7), but instead shortened lifespan. Next we tested whether high glucose affected inhibition of bacterial growth by metformin. Strikingly, glucose supplementation suppressed metformin-induced inhibition of bacterial growth (Figures S7BS7D). This may reect a switch from amino acid-based to glucose-based metabolism for growth, relieving the need of glucogenic amino acids (e.g., methionine) as a source of carbon. Thus, a diet high in glucose can abrogate the benecial effects of metformin on lifespan, a nding of potential relevance to mammals. DISCUSSION In this study we have shown how metformin slows aging in C. elegans by metabolic alteration of the E. coli with which it is cultured. Metformin disrupts the bacterial folate cycle, leading to reduced levels of SAMe and decelerated aging in the worm. Two Mechanisms of Action of Metformin on C. elegans The effect of metformin on worm lifespan was strongly dependent upon the accompanying microbes. In the presence of some E. coli strains, metformin increased lifespan, whereas with other strains or in the absence of microbes it shortened lifespan. This study demonstrates that metformin has both direct
Stress resistance Detoxification
and indirect effects on C. elegans. Metformin (50 mM) acts directly to shorten worm lifespan, likely reecting drug toxicity, and indirectly to increase lifespan by impairing microbial folate metabolism. The actual effect of metformin on lifespan depends on whether direct or indirect effects predominate. Given metformin-sensitive E. coli strains (e.g., OP50), drug treatment impairs folate metabolism and slows aging. But given metformin-resistant strains (e.g., OP50-MR), folate metabolism is less affected, the toxic effect predominates, and lifespan is shortened. It is possible that in other host organisms the capacity for metformin to slow aging is also microbiome-dependent. For example, the recent observation that metformin activates AMPK but does not increase lifespan in Drosophila (Slack et al., 2012) might reect the presence of metformin-resistant microbiota. Our ndings imply that life-extending effects of metformin are not due to rescue from proliferation-mediated bacterial pathogenicity. Instead, the drug alters bacterial metabolism, leading to a state of nutritional restriction in the worm, which increases lifespan. Consistent with this, as under DR, concentrations of biguanides that increase lifespan also reduce egg laying rate (Onken and Driscoll, 2010) (Figures S1A and S1B) and reduce the rate of increase in age-specic mortality (Figures 1D and S6C) (Wu et al., 2009). It was previously demonstrated that AMPK-dependent activation of SKN-1 is essential for metformin benets on health span and lifespan (Onken and Driscoll, 2010). Our ndings show that AMPK and SKN-1 promote resistance to biguanide toxicity, and imply it is for this reason that in their absence drug-induced life extension is not seen. However, AMPK (but not SKN-1) is also required for the full microbe-mediated life extension (Figure 6H). Metformin Effects on Methionine Metabolism in E. coli and C. elegans We investigated the likely bacterial target of metformin, rst ruling out DHF reductase as a target (Figures S4ES4G). Instead, metformin induction of a methyl trap, in which 5-methyl-THF accumulates, is consistent with lowered MS activity (Nijhout et al., 2004) and therefore attenuated methionine biosynthesis. Moreover, metformin also increases bacterial levels of SAMe, which is known to inhibit transcription of genes involved in methionine biosynthesis (Banerjee and Matthews, 1990). Studies in mammalian liver cells show that SAMe can act both as an allosteric activator of SAMS and a feedback inhibitor of MTHFR leading to reduced levels of methionine. In addition, increased levels of 5-methyl-THF block methyltransferases (e.g., glycine N-methyltransferase) (Mato et al., 2008). This provides a potential explanation for the observed rise of SAMe in addition to MS inhibition by metformin, and strongly suggest that it reduces bacterial methionine levels (Figure 7C). Consistent with this, treating the C. elegans/E. coli system with metformin caused a 5-fold decrease in SAMe levels and a drop in the SAMe/SAH ratio in the worm. Moreover, mutation of the worm MS gene metr-1 enhanced metformin-induced life extension, again consistent with MS inhibition in metformintreated E. coli, and also with methionine restriction as a mechanism of worm life extension. The latter is further supported by the inability of metformin to extend the lifespan of sams-1
mutant worms, which have a 65% decrease in SAMe levels (Walker et al., 2011). Both sams-1 RNAi and metformin increase lifespan in wildtype but not eat-2 (DR) mutants worms, and both treatments are thought to recapitulate the effects of DR (Hansen et al., 2005; Onken and Driscoll, 2010). Indeed, metformin induces a DR-like state that, similarly to decreased levels of sams-1 by RNAi, reduces brood size, delays reproductive timing, and increases lifespan independently of the transcription factor DAF-16/FoxO but not in eat-2 DR mutants (Onken and Driscoll, 2010). Also, sams-1 mRNA levels are reduced 3-fold in eat-2 mutants (Hansen et al., 2005). Similar DR-like phenotypes, including reduced body size, were observed in our study when using phenformin (Figures S1AS1D). Moreover, restriction of dietary methionine can extend lifespan in fruit ies and rodents (Grandison et al., 2009; Orentreich et al., 1993). Taken with these observations, our ndings suggest a potential common mechanism underlying the action of metformin, knockdown of sams-1 and DR, which will be interesting to investigate in future studies. Potential mechanisms by which reduced SAMe might increase lifespan include reduced protein synthesis and altered fat metabolism (Ching et al., 2010; Hansen et al., 2005; Walker et al., 2011). Additionally, reduced SAMe/SAH ratio, as a measure of reduced methylation potential, could modulate lifespan via histone methylation (i.e., epigenetic effects). One possibility is that the relative abundance of metabolites such as SAMe allows the cell to assess its energy state and respond accordingly, creating a link between diet, metabolism and gene expression to modulate physiology and consequently lifespan. Metformin and Gut Microbiota in Humans Our ndings are of potential relevance to mammalian biology and human health. Bacteria in the human gut play a central role in nutrition and host biology, and affect the risk of obesity and associated metabolic disorders such as diabetes, inammation, and liver diseases (Cani and Delzenne, 2007). Our nding that metformin inuences C. elegans aging by altering microbial metabolism raises the possibility that this drug might similarly inuence mammalian biology by affecting microbial metabolism or composition. Metformin is the most prescribed drug to treat T2D, with doses ranging from 5002,500 mg/day (Scarpello and Howlett, 2008). Drug concentration in the jejunum is 30- to 300-fold higher than in the plasma in metformin recipients (Bailey et al., 2008) and concentrations above 20 mM have been detected in the intestinal lumen after administration of 850 mg metformin (Proctor et al., 2008). Interestingly, common side effects include gastrointestinal disorders (e.g., bloating and diarrhea) (Bytzer et al., 2001), reduced folate, and increased homocysteine levels (Sahin et al., 2007). Similarly, we nd that metformin impairs bacterial folate metabolism and reduces host SAMe/SAH ratio. Factors causing perturbation of the microbiome (dysbiosis), e.g., obesity, a high-fat diet, and antibiotics, often lead to metabolic dyshomeostasis in the host (Delzenne et al., 2011; Nicholson et al., 2012) e.g., due to release of proinammatory microbial LPS into the bloodstream. Our data show that the effects of metformin are bacterial strain-dependent but independent of
LPS. One possibility is that metformin might promote a better balance of gut microbiota species. We were able to develop a metformin-resistant bacterial strain that confers benets to the host (Figure S3F) suggesting that long-term administration of metformin could benet the host even after treatment is ceased. Indeed, metformin administration to rats causes a change in the composition of the microbiome (Pyra et al., 2012), although it remains unclear what effect this has upon the host. Moreover, the antibiotic noroxacin can induce alteration of mouse gut microbiome that has benecial effects, e.g., enhanced glucose tolerance (Membrez et al., 2008). Lowering dietary glucose can benet humans with metabolic syndrome or T2D (Venn and Green, 2007). Diet strongly inuences the metabolism of the human microbiota (Turnbaugh et al., 2009). We have found that elevated dietary glucose suppresses the effects of metformin on bacterial growth and worm lifespan. This suggests that a high-sugar diet might impair microbe-mediated benets of metformin. Overall, our ndings point to the potential therapeutic efcacy of drugs that alter gut microbiota, particularly to prevent or treat metabolic disease (Delzenne et al., 2011). In addition, it underscores the value of C. elegans as a model to study host-microbe interactions. E. coli as Food Source and Microbiome for C. elegans Mammals, including humans, coexist with intestinal microbes in a relationship that includes elements of commensalism, symbiosis, and pathogenesis, and microbiota strongly inuences host metabolism (Delzenne et al., 2011; Nicholson et al., 2012). Several observations suggest that in at least some respects E. coli could act as microbiome for C. elegans. Although worms can be cultured on semidened media in the absence of E. coli (axenically), such media do not support normal growth and fertility. C. elegans seems to require live microbes for normal growth, reproduction, and aging (Lenaerts et al., 2008; Smith et al., 2008). However, unlike microbiota and their mammalian hosts, E. coli is the principal food source for C. elegans. Studies of GFPlabeled E. coli imply that in late stage larvae (L4), bacterial cells are largely broken down by the pharynx prior to entering the intestine (Kurz et al., 2003), although by day 2 of adulthood intact E. coli are visible in the intestine (Labrousse et al., 2000). In senescent worms, E. coli contribute to the demise of their host, clogging the lumen of the alimentary canal and invading the intestine (Garigan et al., 2002; Labrousse et al., 2000; McGee et al., 2011). Thus, it appears that in early life C. elegans and E. coli exist in a predator-prey relationship, whereas in late life the tables are turned. But it remains possible that metabolic activity in intact or lysed E. coli within the worm contributes to intestinal function and host metabolism throughout life. Presumably, C. elegans has evolved in the constant presence of metabolically active intestinal microbes. We postulate that, consequently, intestinal function requires their presence. Thus, it may only be possible to fully understand C. elegans metabolism as it operates within the C. elegans/E. coli holobiont (Zilber-Rosenberg and Rosenberg, 2008). Our account of how metformin impacts on the two organisms is consistent with this view.
EXPERIMENTAL PROCEDURES Strains and Culture Conditions Nematode and bacterial strains used and generated in this study are described in the Extended Experimental Procedures. Where indicated, molten NGM agar was supplemented with drugs. Axenic plates were prepared as previously described (Lenaerts et al., 2008). Lifespan Analysis This was performed as follows, unless otherwise indicated. Briey, trials were initiated by transfer of L4-stage worms (day 0) on plates supplemented with 15 mM FUdR. Statistical signicance of effects on lifespan was estimated using the log rank test, performed using JMP, Version 7 (SAS Institute). GST-4::GFP Fluorescence Quantitation Animals were raised from L1 stage on control or drug-treated plates. Quantication of GFP expression at the L4 stage was carried out using a Leica DMRXA2 epiuorescence microscope, an Orca C10600 digital camera (Hamamatsu, Hertfordshire, UK), and Volocity image analysis software (Improvision, UK). GFP intensity was measured as the pixel density in the entire cross-sectional area of each worm from which the background pixel density was subtracted (90 worms per condition). Bacterial Growth Assay Liquid bacterial growth was performed in microtiter plates containing the respective bacterial strain (previously grown overnight in LB and diluted 1,000-fold) and drugs in 200 ml of LB at pH 7.0. Absorbance (OD 600 nm) was measured every 5 min over an 18 hr period with shaking at 37 C using a Tecan Innite M2000 microplate reader and Magellan V6.5 software. For colony forming unit counts, see Extended Experimental Procedures. Bacterial Respiration This was measured in a Clark-type oxygen electrode (Rank Brothers, Cambridge, UK) in a 1 ml stirred chamber at 37 C (Lenaerts et al., 2008). Metabolite Analysis by LC-MS/MS Bacterial and nematode metabolite analysis was performed as described in Extended Experimental Procedures. Metabolomic Principal Component Analysis Raw LC-MS/MS spectral data were uploaded into MetaboAnalyst. To avoid propensity to data overtting, PCA analysis was used to create the 2D analysis plot. Western Blotting Briey, phosphorylation of AAK-2 subunit (pAMPKa) was detected using pAMPKa (Cell Signaling) at a 1:1,000 dilution. Films were scanned and the density of each band or the entire lane was quantied by densitometry using ImageQuant TL (GE Healthcare Europe Gmb, UK). Food Clearance Assay The effect of biguanide compounds on C. elegans physiology was monitored from the rate at which 50% of the E. coli food suspension was consumed, as a read out for C. elegans growth, survival, or fecundity. SUPPLEMENTAL INFORMATION Supplemental Information includes Extended Experimental Procedures, seven gures, and seven tables and can be found with this article online at http://dx. doi.org/10.1016/j.cell.2013.02.035. ACKNOWLEDGMENTS We thank Dan Ackerman, Joy Alcedo, Nazif Alic, Caroline Araiz, Alex Benedetto, Steve Clarke, Gonc alo Correia, Monica Driscoll, Michael Murphy, Brian Onken, Matthew Piper, Bhupinder Virk, and Matthias Ziehm for useful
discussion and other help. Strains were provided by the CGC and CGSC (DBI0742708). We acknowledge funding from the Wellcome Trust (Strategic Award), the European Union (LifeSpan) and the MRC (J003794). Received: August 6, 2012 Revised: November 7, 2012 Accepted: February 11, 2013 Published: March 28, 2013 REFERENCES Anisimov, V.N., Berstein, L.M., Popovich, I.G., Zabezhinski, M.A., Egormin, P.A., Piskunova, T.S., Semenchenko, A.V., Tyndyk, M.L., Yurova, M.N., Kovalenko, I.G., and Poroshina, T.E. (2011). If started early in life, metformin treatment increases lifespan and postpones tumors in female SHR mice. Aging (Albany NY) 3, 148157. ckhed, F., Ley, R.E., Sonnenburg, J.L., Peterson, D.A., and Gordon, J.I. Ba (2005). Host-bacterial mutualism in the human intestine. Science 307, 1915 1920. Bailey, C.J., Wilcock, C., and Scarpello, J.H. (2008). Metformin and the intestine. Diabetologia 51, 15521553. Banerjee, R.V., and Matthews, R.G. (1990). Cobalamin-dependent methionine synthase. FASEB J. 4, 14501459. Brenner, S. (1974). The genetics of Caenorhabditis elegans. Genetics 77, 7194. Bytzer, P., Talley, N.J., Jones, M.P., and Horowitz, M. (2001). Oral hypoglycaemic drugs and gastrointestinal symptoms in diabetes mellitus. Aliment. Pharmacol. Ther. 15, 137142. Cani, P.D., and Delzenne, N.M. (2007). Gut microora as a target for energy and metabolic homeostasis. Curr. Opin. Clin. Nutr. Metab. Care 10, 729734. Ching, T.T., Paal, A.B., Mehta, A., Zhong, L., and Hsu, A.L. (2010). drr-2 encodes an eIF4H that acts downstream of TOR in diet-restriction-induced longevity of C. elegans. Aging Cell 9, 545557. Delzenne, N.M., and Cani, P.D. (2011). Gut microbiota and the pathogenesis of insulin resistance. Curr. Diab. Rep. 11, 154159. Delzenne, N.M., Neyrinck, A.M., Backhed, F., and Cani, P.D. (2011). Targeting gut microbiota in obesity: effects of prebiotics and probiotics. Nat. Rev. Endocrinol. 7, 639646. Dowling, R.J., Goodwin, P.J., and Stambolic, V. (2011). Understanding the benet of metformin use in cancer treatment. BMC Med. 9, 33. Garigan, D., Hsu, A.L., Fraser, A.G., Kamath, R.S., Ahringer, J., and Kenyon, C. (2002). Genetic analysis of tissue aging in Caenorhabditis elegans: a role for heat-shock factor and bacterial proliferation. Genetics 161, 11011112. Garsin, D.A., Villanueva, J.M., Begun, J., Kim, D.H., Sifri, C.D., Calderwood, S.B., Ruvkun, G., and Ausubel, F.M. (2003). Long-lived C. elegans daf-2 mutants are resistant to bacterial pathogens. Science 300, 1921. Gems, D., and Riddle, D.L. (2000). Genetic, behavioral and environmental determinants of male longevity in Caenorhabditis elegans. Genetics 154, 15971610. Grandison, R.C., Piper, M.D., and Partridge, L. (2009). Amino-acid imbalance explains extension of lifespan by dietary restriction in Drosophila. Nature 462, 10611064. Hannich, J.T., Entchev, E.V., Mende, F., Boytchev, H., Martin, R., Zagoriy, V., lker, H.J., and Kurzchalia, T.V. Theumer, G., Riezman, I., Riezman, H., Kno (2009). Methylation of the sterol nucleus by STRM-1 regulates Dauer larva formation in Caenorhabditis elegans. Dev. Cell 16, 833843. Hansen, M., Hsu, A.L., Dillin, A., and Kenyon, C. (2005). New genes tied to endocrine, metabolic, and dietary regulation of lifespan from a Caenorhabditis elegans genomic RNAi screen. PLoS Genet. 1, 119128. kela , T.P., Hawley, S.A., Boudeau, J., Reid, J.L., Mustard, K.J., Udd, L., Ma Alessi, D.R., and Hardie, D.G. (2003). Complexes between the LKB1 tumor suppressor, STRAD alpha/beta and MO25 alpha/beta are upstream kinases in the AMP-activated protein kinase cascade. J. Biol. 2, 28.
Ikeda, T., Yasui, C., Hoshino, K., Arikawa, K., and Nishikawa, Y. (2007). Inuence of lactic acid bacteria on longevity of Caenorhabditis elegans and host defense against salmonella enterica serovar enteritidis. Appl. Environ. Microbiol. 73, 64046409. Kaeberlein, T.L., Smith, E.D., Tsuchiya, M., Welton, K.L., Thomas, J.H., Fields, S., Kennedy, B.K., and Kaeberlein, M. (2006). Lifespan extension in Caenorhabditis elegans by complete removal of food. Aging Cell 5, 487494. Kau, A.L., Ahern, P.P., Grifn, N.W., Goodman, A.L., and Gordon, J.I. (2011). Human nutrition, the gut microbiome and the immune system. Nature 474, 327336. Kenyon, C.J. (2010). The genetics of ageing. Nature 464, 504512. ` s, E., Aurouze, M., Vallet, I., Michel, G.P., Uh, M., Kurz, C.L., Chauvet, S., Andre Celli, J., Filloux, A., De Bentzmann, S., et al. (2003). Virulence factors of the human opportunistic pathogen Serratia marcescens identied by in vivo screening. EMBO J. 22, 14511460. Kwon, Y.K., Lu, W., Melamud, E., Khanam, N., Bognar, A., and Rabinowitz, J.D. (2008). A domino effect in antifolate drug action in Escherichia coli. Nat. Chem. Biol. 4, 602608. Labrousse, A., Chauvet, S., Couillault, C., Kurz, C.L., and Ewbank, J.J. (2000). Caenorhabditis elegans is a model host for Salmonella typhimurium. Curr. Biol. 10, 15431545. Lee, S.J., Murphy, C.T., and Kenyon, C. (2009). Glucose shortens the lifespan of C. elegans by downregulating DAF-16/FOXO activity and aquaporin gene expression. Cell Metab. 10, 379391. Lenaerts, I., Walker, G.A., Van Hoorebeke, L., Gems, D., and Vaneteren, J.R. (2008). Dietary restriction of Caenorhabditis elegans by axenic culture reects nutritional requirement for constituents provided by metabolically active microbes. J. Gerontol. A Biol. Sci. Med. Sci. 63, 242252. Maier, W., Adilov, B., Regenass, M., and Alcedo, J. (2010). A neuromedin U receptor acts with the sensory system to modulate food type-dependent effects on C. elegans lifespan. PLoS Biol. 8, e1000376. Mair, W., and Dillin, A. (2008). Aging and survival: the genetics of lifespan extension by dietary restriction. Annu. Rev. Biochem. 77, 727754. zquez-Chantada, M., Garnacho, M., Latasa, M.U., nez-Chantar, M.L., Va Mart nez-Cruz, L.A., Parada, L.A., Varela-Rey, M., Dotor, J., Santamaria, M., Mart Lu, S.C., and Mato, J.M. (2006). S-adenosylmethionine regulates cytoplasmic HuR via AMP-activated kinase. Gastroenterology 131, 223232. nez-Chantar, M.L., and Lu, S.C. (2008). Methionine metaboMato, J.M., Mart lism and liver disease. Annu. Rev. Nutr. 28, 273293. McGee, M.D., Weber, D., Day, N., Vitelli, C., Crippen, D., Herndon, L.A., Hall, D.H., and Melov, S. (2011). Loss of intestinal nuclei and intestinal integrity in aging C. elegans. Aging Cell 10, 699710. Membrez, M., Blancher, F., Jaquet, M., Bibiloni, R., Cani, P.D., Burcelin, R.G., , K., and Chou, C.J. (2008). Gut microbiota modulation with Corthesy, I., Mace noroxacin and ampicillin enhances glucose tolerance in mice. FASEB J. 22, 24162426. Nicholson, J.K., Holmes, E., Kinross, J., Burcelin, R., Gibson, G., Jia, W., and Pettersson, S. (2012). Host-gut microbiota metabolic interactions. Science 336, 12621267. Nijhout, H.F., Reed, M.C., Budu, P., and Ulrich, C.M. (2004). A mathematical model of the folate cycle: new insights into folate homeostasis. J. Biol. Chem. 279, 5500855016. Onken, B., and Driscoll, M. (2010). Metformin induces a dietary restriction-like state and the oxidative stress response to extend C. elegans Healthspan via AMPK, LKB1, and SKN-1. PLoS ONE 5, e8758. Orentreich, N., Matias, J.R., DeFelice, A., and Zimmerman, J.A. (1993). Low methionine ingestion by rats extends lifespan. J. Nutr. 123, 269274. Ottaviani, E., Ventura, N., Mandrioli, M., Candela, M., Franchini, A., and Franceschi, C. (2011). Gut microbiota as a candidate for lifespan extension: an ecological/evolutionary perspective targeted on living organisms as metaorganisms. Biogerontology 12, 599609.
Pierotti, M.A., Berrino, F., Gariboldi, M., Melani, C., Mogavero, A., Negri, T., Pasanisi, P., and Pilotti, S. (2012). Targeting metabolism for cancer treatment and prevention: metformin, an old drug with multi-faceted effects. Oncogene. http://dx.doi.org/10.1038/onc.2012.181. Proctor, W.R., Bourdet, D.L., and Thakker, D.R. (2008). Mechanisms underlying saturable intestinal absorption of metformin. Drug Metab. Dispos. 36, 16501658. Pyra, K.A., Saha, D.C., and Reimer, R.A. (2012). Prebiotic ber increases hepatic acetyl CoA carboxylase phosphorylation and suppresses glucosedependent insulinotropic polypeptide secretion more effectively when used with metformin in obese rats. J. Nutr. 142, 213220. Sahin, M., Tutuncu, N.B., Ertugrul, D., Tanaci, N., and Guvener, N.D. (2007). Effects of metformin or rosiglitazone on serum concentrations of homocysteine, folate, and vitamin B12 in patients with type 2 diabetes mellitus. J. Diabetes Complications 21, 118123. Saiki, R., Lunceford, A.L., Bixler, T., Dang, P., Lee, W., Furukawa, S., Larsen, P.L., and Clarke, C.F. (2008). Altered bacterial metabolism, not coenzyme Q content, is responsible for the lifespan extension in Caenorhabditis elegans fed an Escherichia coli diet lacking coenzyme Q. Aging Cell 7, 291304. Scarpello, J.H., and Howlett, H.C. (2008). Metformin therapy and clinical uses. Diab. Vasc. Dis. Res. 5, 157167. Slack, C., Foley, A., and Partridge, L. (2012). Activation of AMPK by the putative dietary restriction mimetic metformin is insufcient to extend lifespan in Drosophila. PLoS ONE 7, e47699. Smith, E.D., Kaeberlein, T.L., Lydum, B.T., Sager, J., Welton, K.L., Kennedy, B.K., and Kaeberlein, M. (2008). Age- and calorie-independent lifespan exten-
sion from dietary restriction by bacterial deprivation in Caenorhabditis elegans. BMC Dev. Biol. 8, 49. Storelli, G., Defaye, A., Erkosar, B., Hols, P., Royet, J., and Leulier, F. (2011). Lactobacillus plantarum promotes Drosophila systemic growth by modulating hormonal signals through TOR-dependent nutrient sensing. Cell Metab. 14, 403414. Turnbaugh, P.J., Ridaura, V.K., Faith, J.J., Rey, F.E., Knight, R., and Gordon, J.I. (2009). The effect of diet on the human gut microbiome: a metagenomic analysis in humanized gnotobiotic mice. Sci. Transl. Med. 1, 6ra14. Venn, B.J., and Green, T.J. (2007). Glycemic index and glycemic load: measurement issues and their effect on diet-disease relationships. Eur. J. Clin. Nutr. 61(Suppl 1), S122S131. Virk, B., Correia, G., Dixon, D.P., Feyst, I., Jia, J., Oberleitner, N., Briggs, Z., Hodge, E., Edwards, R., Ward, J., et al. (2012). Excessive folate synthesis limits lifespan in the C. elegans: E. coli aging model. BMC Biol. 10, 67. Walker, A.K., Jacobs, R.L., Watts, J.L., Rottiers, V., Jiang, K., Finnegan, D.M., Shioda, T., Hansen, M., Yang, F., Niebergall, L.J., et al. (2011). A conserved SREBP-1/phosphatidylcholine feedback circuit regulates lipogenesis in metazoans. Cell 147, 840852. Wu, D., Rea, S.L., Cypser, J.R., and Johnson, T.E. (2009). Mortality shifts in Caenorhabditis elegans: remembrance of conditions past. Aging Cell 8, 666675. Zilber-Rosenberg, I., and Rosenberg, E. (2008). Role of microorganisms in the evolution of animals and plants: the hologenome theory of evolution. FEMS Microbiol. Rev. 32, 723735.
Diet-Induced Developmental Acceleration Independent of TOR and Insulin in C. elegans

Lesley T. MacNeil,1,2 Emma Watson,1,2 H. Efsun Arda,1,2,4 Lihua Julie Zhu,3 and Albertha J.M. Walhout1,2,*
in Systems Biology in Molecular Medicine 3Program in Gene Function and Expression University of Massachusetts Medical School, Worcester, MA 01605, USA 4Present address: Department of Developmental Biology, Stanford University School of Medicine, Stanford, CA 93405, USA *Correspondence: marian.walhout@umassmed.edu http://dx.doi.org/10.1016/j.cell.2013.02.049
2Program 1Program
SUMMARY
Dietary composition has major effects on physiology. Here, we show that developmental rate, reproduction, and lifespan are altered in C. elegans fed Comamonas DA1877 relative to those fed a standard E. coli OP50 diet. We identify a set of genes that change in expression in response to this diet and use the promoter of one of these (acdh-1) as a dietary sensor. Remarkably, the effects on transcription and development occur even when Comamonas DA1877 is diluted with another diet, suggesting that Comamonas DA1877 generates a signal that is sensed by the nematode. Surprisingly, the developmental effect is independent from TOR and insulin signaling. Rather, Comamonas DA1877 affects cyclic gene expression during molting, likely through the nuclear hormone receptor NHR-23. Altogether, our ndings indicate that different bacteria elicit various responses via distinct mechanisms, which has implications for diseases such as obesity and the interactions between the human microbiome and intestinal cells.
INTRODUCTION The amount and nutritional content of food are important determinants of organismal health and inuence life-history traits such as developmental rate and fecundity. Thus, cells and organisms must sense and interpret dietary state and alter their physiology accordingly. Although the inuence of caloric intake on organismal health and life-history traits is well appreciated, little is known about the physiological effects of different diets and the mechanisms involved in coordinating diet and physiology. In response to differences in nutrient availability from different diets, metabolic networks are modulated to meet cellular and organismal needs. Further, as it is not benecial to turn on cata240 Cell 153, 240252, March 28, 2013 2013 Elsevier Inc.
bolic and anabolic uxes simultaneously, metabolites generated by a specic catabolic pathway often act as inhibitors of the opposing anabolic pathway and vice versa. Metabolic network modulation also occurs at the level of transcription; complex information processing mechanisms relay nutritional input to signal transduction pathways that impinge upon transcription factors to regulate metabolic gene expression. The nematode C. elegans is a powerful genetic model to studythe impact of diet on gene expression and life-history traits such as developmental rate, reproduction, and lifespan. C. elegans is a soil-dwelling bacterivore with a simple anatomy of fewer than 1,000 somatic cells, 20 of which form the intestine, a single organ that functions as both gut and liver with digestive as well as endocrine functions. In the laboratory, C. elegans can be fed a variety of bacterial species and strains (Avery and You, 2012; Coolon et al., 2009). The standard laboratory diet for C. elegans is E. coli OP50. In the wild, however, C. elegans is not likely to encounter E. coli; rather, its diet constitutes a variety of bacteria that grow on rotting vegetation. A number of bacteria have been isolated from soil samples containing C. elegans, including Comamonas (Avery and Shtonda, 2003). Altering the C. elegans diet can affect a number of traits, including roaming time, lifespan, and fecundity and pharyngeal pumping rate (Coolon et al., 2009; Shtonda and Avery, 2006; Soukas et al., 2009). In response to starvation, animals can enter into larval stage 1 (L1) diapause, which is a short-term developmental delay. If food does not become available, C. elegans also employs a long-term survival strategy by entering into dauer, an alternate L3 phase (Sommer and Ogawa, 2011). Caloric restriction has been shown to decrease fecundity and developmental rate and to increase lifespan (Lakowski and Hekimi, 1998). Starvation and caloric restriction versus ample food availability exemplify extreme conditions in the dietary spectrum. However, challenges are not simply the presence or absence of food, but instead, diverse food sources with different content and quality may be encountered. Several nutrient-response systems have been studied in numerous model systems. The target of rapamycin (TOR) pathway, for instance, detects a variety of conditions, including amino acid availability, energy levels, and stress and affects
numerous physiological processes, including growth, metabolism, and lifespan (Laplante and Sabatini, 2012). In C. elegans, the TOR ortholog let-363 is essential for development (Long et al., 2002). Other phenotypes associated with perturbation of the TOR pathway include decreased brood size and developmental rate and increased lifespan (Honjoh et al., 2009; Korta et al., 2012; Pan et al., 2007; Soukas et al., 2009). The insulin/ IGF signaling pathway is another major pathway that regulates lifespan, diapauses, and stress response, for instance, in response to a lack of nutrients (Narasimhan et al., 2009). Insulin signaling occurs through the DAF-2 receptor and impinges on the FoxO transcription factor DAF-16. Finally, nuclear hormone receptors (NHRs) sense a variety of signals produced under specic metabolic and environmental conditions. NHRs are ligand-regulated transcription factors that affect numerous physiological processes, including development, growth, and metabolism (Pardee et al., 2011; Sonoda et al., 2008). Remarkably, the C. elegans genome encodes 271 NHRs (Reece-Hoyes et al., 2005), whereas the human genome encodes fewer than 50 (Reece-Hoyes et al., 2011; Sonoda et al., 2008). Most C. elegans NHRs are homologs of HNF4, and only a few have been characterized experimentally. Here, we use C. elegans to investigate the relationships between diet and life-history traits and the mechanisms involved. We nd that, when fed the soil bacteria Comamonas, C. elegans develop faster, lay fewer eggs, and live shorter than when fed E. coli OP50. By expression proling, we identify a core set of C. elegans genes that differ in expression on the different diets. We establish a transgenic dietary sensor strain that harbors the promoter of one of these genes, acdh-1, to drive expression of the green uorescent protein (GFP). On the standard laboratory diet of E. coli OP50, GFP expression levels are high. In contrast, GFP expression is barely detectable when the animals are fed Comamonas DA1877. This demonstrates that the dietary response occurs at the level of transcription. Remarkably, when Comamonas DA1877 is dramatically diluted with E. coli OP50, GFP expression is low and developmental rate is accelerated. This shows that the Comamonas DA1877 effect is dominant over that of E. coli OP50 and that the response does not simply reect differences in caloric intake, for instance, by mimicking starvation. We show that the developmental acceleration caused by a Comamonas DA1877 diet is independent of TOR and insulin signaling. Instead, we nd that Comamonas DA1877 affects cycling gene expression during larval molts, likely through NHR-23. RESULTS Dietary Modication of Life-History Traits We measured developmental rate, fecundity, and lifespan in C. elegans fed three different diets: E. coli OP50; E. coli HT115, a strain used in RNA interference (RNAi) by feeding experiments (Timmons et al., 2001); and Comamonas DA1877. The latter has been proposed to be a healthier diet because it alleviates developmental delays in Eat mutants that have defects in pharyngeal pumping and a compromised ability to eat (Avery and Shtonda, 2003; Shtonda and Avery, 2006). To assess developmental rate, we synchronized animals fed each of the diets in the L1
stage and selected one time point to monitor the developmental age of a population of animals (Figure S1 available online). We found an increased number of older animals in the Comamonas DA1877-fed population than in that fed E. coli OP50 or HT115, suggesting that Comamonas DA1877-fed animals develop faster (Figure 1A). In order to separate the effects of diet on developmental rate from recovery from L1 starvation, we switched animals from an E. coli OP50 to a Comamonas DA1877 diet midway through development. Even following recovery from starvation on E. coli OP50, animals fed a Comamonas DA1877 diet displayed accelerated development (Figure 1B). This demonstrates that effects on development are not the result of differences in recovery from starvation and are not limited to early developmental time points. To investigate the timing of the changes in developmental rate, we used a transgenic strain harboring a molting-dependent reporter Pmlt-10::GFP-pest. This transgenic strain expresses a destabilized GFP protein under the control of the mlt-10 promoter, a gene whose expression oscillates with molting (Frand et al., 2005). We examined GFP expression during development in animals fed E. coli OP50 or Comamonas DA1877. As reported previously, animals fed E. coli OP50 displayed oscillatory GFP expression (Frand et al., 2005). Surprisingly, the amplitude of GFP oscillations was dramatically reduced in all four molting cycles in animals fed Comamonas DA1877 (Figure 1C). Thus, Comamonas DA1877 may alter developmental rate by modulating the molting program. We measured two additional life-history traits in C. elegans fed the different diets and found that animals fed Comamonas DA1877 have a decreased average brood size of 158 eggs compared to 234 and 232 for animals fed E. coli OP50 or HT115, respectively (Figure 1D). Finally, animals fed Comamonas DA1877 exhibit a much-reduced lifespan (Figure 1E). The Dietary Effect on Development and Lifespan Is Not Due to Pathogenic Infection In addition to serving as food, some bacteria can also be pathogenic to C. elegans (Gravato-Nobre and Hodgkin, 2005; Tan et al., 1999). To test whether the life-history trait changes result from a pathogenic response, we used killed bacteria. Comamonas DA1877 proved difcult to kill using standard protocols (Sutphin and Kaeberlein, 2009) (data not shown). However, a combination of UV irradiation, peptone-free media, and antibiotics effectively killed both E. coli OP50 and Comamonas DA1877. On killed bacteria, both diets result in similar brood sizes (Figure 1F). In contrast, there was still a diet-induced difference in developmental rate and lifespan (Figures 1G and 1H). This demonstrates that the short lifespan and fast development are not the result of a pathogenic infection but rather are more likely to be caused by a dietary effect. Thus, we focused on these phenotypes in the rest of our study. Diet-Induced Changes in Gene Expression Next, we examined the gene expression changes elicited by the three different bacterial diets in young adult animals using microarray expression proling. In animals fed Comamonas DA1877, 389 genes changed signicantly (R2-fold; p < 0.001) when compared to E. coli OP50 (Figures 2A and 2B).
A
100
B
OP50
100
24hrs
OP50 DA1877
100 90
Percent total animals
Percent GFP positive
80 70 60 50 40 30 20 10 OP50 DA1877
O P O 50 D P5 A1 0 8 7 to D 7 A1 87 7
56
O
60
32
20
24 28
44
48
32
0 T1 15
P5
87
hours after L1 arrest

F
300
A1
300 250
E
100 90 80 70 60 50 40 30 20 10 DA1877 OP50 HT115
36 40
16
Total number of offspring
Total number of offspring
Percent survival
200 150 100 50
T1 D 15 A1 87 7
2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32
52
250 200 150 100 50
0 P5
22 24 28 30
P5
100
H
Adult Late L4 Mid-late L4 Mid L4 Early L4 L3
100 90 80 70 60 50 40 30 20 10
Percent survival
Killed OP50 Killed DA1877
7 P5 0 D A1 87 7
8 10 12
14
16
18
20
Days Adult
Killed
P5
87
Live
Days Adult
A1
Killed
Figure 1. A Comamonas DA1877 Diet Affects C. elegans Life-History Traits

(A) Developmental progression on three different diets. Synchronized N2 wild-type animals (L1 stage) were grown on three different diets as indicated on the x axis and scored after 43 hr. Larval stage was visually determined based on the stage of vulval development (see Figure S1).
34
A1
36
26
87
64
Gene ontology (GO) analysis (Ashburner et al., 2000) of the genes that change on a Comamonas DA1877 diet revealed an enrichment of terms related to molting (Figure 2C). This is consistent with the ability of Comamonas DA1877 to affect the cycling expression of mlt-10 (Figure 1C). Thus, it is likely that Comamonas DA1877 affects the C. elegans molting program and, hence, the expression of associated genes. However, in contrast to what we observed with the mlt-10 reporter, the majority of the moltingassociated genes that change in response to diet decrease in expression in response to Comamonas DA1877. Because animals fed E. coli OP50 and Comamonas DA1877 grow at different rates, animals were collected for RNA isolation based on visual examination of developmental age in the population. We therefore wondered whether some changes in the dietinduced expression proles could result from differences in staging of the population. Therefore, we performed an additional microarray expression proling experiment at a later time point that is well separated from the oscillations that occur during molting. At this stage, even more genes changed in expression in response to Comamonas DA1877 (Figure 2D). This is likely due, at least in part, to the onset of reproduction. A comparison between the two expression proling experiments revealed a set of 87 core genes that are affected by a Comamonas DA1877 diet at both stages (Figure 2E and Table S2). The core includes upregulated and downregulated genes that encode metabolic enzymes, as well as numerous worm-specic genes. A Dietary Sensor in Living Animals The expression proles do not reveal whether these changes occur at the level of transcription or messenger RNA (mRNA) stability. Among the core genes, acdh-1 exhibits the most dramatic change in expression (Figure 2A and Table S1). We previously generated a transgenic strain that expresses GFP in the intestine and hypodermis under the control of the acdh-1 promoter and wondered whether we could use this as a dietary sensor in living animals (Arda et al., 2010). We integrated the Pacdh-1::GFP transgene into the C. elegans genome and fed the animals different diets. GFP expression was high on E. coli OP50, intermediate on E. coli HT115, and barely detectable on Comamonas DA1877, which recapitulates the microarray data (Figure 3A). The effect on GFP expression was transmitted to the progeny and maintained prior to the start of feeding (Figure 3B). To test whether the sensor can respond to individual nutrients, we exposed the Pacdh-1::GFP transgenic animals to 5 mM glucose on each of the three diets and found that this results in an increase in GFP expression, most notably on an E. coli HT115 diet (Figure 3C). Animals fed Comamonas DA1877 required higher glucose concentrations to induce
acdh-1 promoter activity (Figure 3D). Previously, it has been shown that acdh-1 expression is reduced upon starvation (Van Gilst et al., 2005a). When deprived of food, GFP expression was indeed reduced in Pacdh-1::GFP transgenic animals (Figure 3E). Altogether, these results demonstrate that acdh-1 expression is modulated in response to dietary conditions at least in part through the activity of its promoter and thus at the level of transcription. We obtained similar results with a strain that expresses GFP under the control of acdh-2, which is homologous to acdh-1 and also changes in response to diet (Figure S2). The Dietary Effect Is Distinct from the Starvation Response Both a diet of Comamonas DA1877 and starvation repress the dietary sensor. Previous studies have demonstrated that caloric restriction affects life-history traits (Lakowski and Hekimi, 1998), as does a Comamonas DA1877 diet. Althoughsupercially this may suggest that caloric intake differs between the different bacterial diets, caloric restriction would be expected to reduce developmental rate and increase lifespan, which is opposite to the effects of Comamonas DA1877. We performed two experiments to test whether Comamonas DA1877 confers a (partial) starvation response. In the rst experiment, we used nCounter technology (Geiss et al., 2008) to compare the expression changes of endogenous acdh-1 and acdh-2 to three starvation-induced genes, acs-11, gst-4, and cpt-3, as well pqm-1, a stress-responsive gene (Tawe et al., 1998; Van Gilst et al., 2005a). We did not observe signicant changes in the expression of these genes on any diet (Figure 4A). In the second experiment, we reasoned that, if Comamonas DA1877 lacks specic nutrients and thereby induces a starvation response, then combining it with E. coli OP50 would alleviate this and increase acdh-1 expression. Interestingly, however, a mixed diet of E. coli OP50 and Comamonas DA1877 resulted in barely detectable GFP expression, similar to Comamonas DA1877 alone (Figure 4B). Remarkably, even when diluted dramatically with E. coli OP50, Comamonas DA1877 still exerted a repressive effect (Figure 4B). Together, these observations demonstrate that Comamonas DA1877 is not simply nutrient poor. Indeed, we did not observe major differences in bulk protein, carbohydrate, or lipid between the different bacteria (Figure S3). A second implication of these observations is that Comamonas DA1877 generates a signal to which the animal responds. Diluting Comamonas DA1877 with E. coli OP50 was sufcient to induce an increase in developmental rate (Figure 4C) but did not affect lifespan (Figure 4D). This suggests that developmental rate and lifespan are regulated by different Comamonas-derived
(B) Developmental progression of animals at 48 hr post-L1 synchronization, OP50 to DA1877 indicates animals switched from E. coli OP50 food to Comamonas DA1877 at 24 hr. Development on Comamonas DA1877 is shown for comparison. (C) Expression of GFP in a synchronized population of Pmlt-10::GFP-pest animals fed E. coli OP50 or Comamonas DA1877 throughout development. (D) Brood size on three different diets. Wild-type N2 animals were grown on three different diets indicated on the x axis. Bars represent the average total number of progeny per animal, with the SD indicated for all animals combined. (E) Lifespan of adult animals fed each of the indicated diets. OP50: E. coli OP50; HT115: E. coli HT115; DA1877: Comamonas DA1877. (F) Brood size of animals on killed E. coli OP50 or killed Comamonas DA1877 (as in D). Bars represent the average total number of progeny per animal, with the SD indicated for all animals combined. (G) Developmental progression of animals grown on live or killed E. coli OP50 or Comamonas 1877 at 45 hr. (H) Survival of adult animals fed each of the indicated diets.
DA1877
10
Figure 2. Diet-Induced Changes in Gene Expression

(A) Scatterplot indicating changes in gene expression in animals fed E. coli HT115 or Comamonas DA1877 relative to animals fed E. coli OP50 detected by microarray expression proling. Each square indicates a gene. See also Table S1. (B) Venn diagram indicating the total number of genes changing in response to diet on E. coli HT115 and Comamonas DA1877 diets and the overlap between these, relative to standard laboratory diet of E. coli OP50. (C) GO analysis of genes that decrease in expression on a Comamonas DA1877 diet. Numbers in parentheses indicate enrichment score. See also Table S1. (D) Venn diagrams indicating a comparison of Comamonas DA1877-responsive genes at two different stages of development. (E) Categorization of core dietary response genes. See also Table S2.
5 log2 fold change -5 -5 acdh-1 -10 HT115 5 10
-10
Increased
1147
41 30
Core upregulated
Core downregulated
signals. Alternatively, the effects may be caused by the same signal, but lifespan may be more dose dependent. Lifespan experiments require animals to be maintained on bacterial plates for longer periods of time during which the Comamonas DA1877 signal may diminish. We performed microarray expression proling on animals fed diluted Comamonas DA1877 (0.1%, Figure 4E). The expression prole of animals grown on diluted bacteria was very similar to that of animals grown on undiluted Comamonas DA1877. In fact, of 87 core genes, 67 also changed on the diluted dietary condition, although in general, the magnitude of these changes was lower (Figure 4E). These observations indicate that a small amount Comamonas DA1877 is sufcient to change gene expression as well as accelerate development. Amphid Sensing Is Not Required for the Dietary Response C. elegans senses many environmental cues neuronally and relays the information to the rest of the animal through the use
of secreted ligands and neuropeptides (Ezcurra et al., 2011). Therefore, we tested whether amphid neurons that are exposed to the environment and that sense external signals are involved in the response to Comamonas DA1877. We crossed the dietary sensor into daf-6(e1377) animals in which socket cells are abnormal, preventing the T interaction of amphid neurons with the exterior environment (Albert et al., 1981). GS We found that daf-6(e1377) mutants still T respond to Comamonas DA1877 (FigW ure 5A). The dietary response of endogenous core gene acdh-1, acdh-2, and ech-6 expression is also unaffected by a mutation in daf-6 (Figure 5B). Together, these observations demonstrate that the Comamonas DA1877 signal is not sensed by amphid neurons but may rather be interpreted as the bacteria traverse the digestive tract upon ingestion. Effects of TOR and Insulin Pathways on the Dietary Response Insulin and TOR signaling act as nutrient response pathways and affect life-history traits in many organisms (Laplante and Sabatini, 2012; Narasimhan et al., 2009). To test their possible roles in dietary response, we knocked down known pathway components and examined the effects on GFP expression in the dietary sensor using RNAi by feeding with E. coli HT115 bacteria on which the sensor displays intermediate levels of GFP expression (Figure 3A). Although acdh-1 was previously reported as a target of daf-16 (Murphy et al., 2003), knockdown of neither daf-2 nor daf-16 markedly affected GFP expression (Figure S4). We crossed the dietary sensor into daf-2(e1370) mutants and observed a slight increase in GFP on an E. coli OP50 diet (Figure 5C). When these animals were fed Comamonas DA1877, however, GFP expression decreased as in wild-type sensor
A
OP50
5 mM glucose
Figure 3. A Dietary Sensor in Living Animals

(A) Pacdh-1::GFP transgenic animals respond transcriptionally to the different bacterial diets and can be used as a dietary sensor in living animals. The Comamonas DA1877 uorescence image was acquired with a 25-fold longer exposure time. See also Figure S2. (B) Eggs display maternal GFP expression. (C) The addition of 5 mM glucose activates the dietary sensor. (D) High levels of glucose can activate the dietary sensor on Comamonas DA1877. (E) The dietary sensor is repressed upon starvation.
DA1877
HT115
exp x25
exp x25
200 M
DA1877 (long exposure)
200 M 20 M
E
Fed-OP50
of Comamonas DA1877 to that of E. coli HB101, we measured development in wild-type and rict-1(ft7) mutant animals. As previously reported, E. coli HB101 accelerated growth in wild-type animals, but not rict-1(ft7) mutant animals. In contrast, Comamonas DA1877 accelerated growth in both wild-type and rict1(ft7) mutant animals (Figure 5G). We next tested whether rsks-1, a TORC1 pathway component, was required for accelerated growth on Comamonas DA1877. As previously reported (Pan et al., 2007), rsks-1 mutants develop more slowly than wild-type animals; however, they do develop faster on Comamomas DA1877 than on E. coli OP50 (Figure 5H). Together, these results demonstrate that, although the dietary sensor is affected by TOR pathway inhibition, the accelerated developmental rate of animals fed Comamonas DA1877 is independent of TOR.
OP50
Starved
DA1877
100 mM glucose
NGM
200 M
animals. Similarly, in the absence of daf-16, neither the dietary response nor the developmental acceleration by Comamonas DA1877 was affected (Figures 5D and 5E). Altogether, these ndings demonstrate that the insulin signaling pathway is not required for the dietary sensor and developmental timing aspects of the response to Comamonas DA1877. Knockdown of several components of the TOR pathway reduced GFP expression in the dietary sensor, including rict-1 and ruvb-1 (Figures 5F and S4). There are two TOR complexes, TORC1 and TORC2, that induce different signaling pathways, and inhibition of either pathway decelerates development (Pan et al., 2007; Soukas et al., 2009). Wild-type animals fed E. coli HB101 develop faster than those grown on E. coli OP50, and this acceleration is suppressed by a mutation in rict-1, a TORC2 component (Soukas et al., 2009). To compare the effect
Nuclear Hormone Receptors Affect the Dietary Sensor NHRs are ligand-regulated transcription factors that act as sensors for hormones, vitamins, and lipids and play broad roles in development and physiology (Sonoda et al., 2008). Therefore, they are excellent candidates to mediate the response to dietary signals. By RNAi, we found that nhr-49, a known nutrientresponse mediator (Van Gilst et al., 2005b), does not affect the dietary sensor (Figure S4). We previously identied NHR-10 as a direct regulator of acdh-1 (Arda et al., 2010). We crossed the dietary sensor into nhr-10(tm4695) mutant animals that carry a deletion in the nhr-10 gene. As expected, we observed a strong decrease in GFP expression in nhr-10 mutants on E. coli OP50 (Figure 6A) (Arda et al., 2010). However, GFP expression was still readily detectable, most notably in the posterior intestine. Interestingly, in nhr-10 mutants, GFP expression was further reduced on Comamonas DA1877 (Figure 6A). In addition, although nhr-10 mutants developed more slowly than wild-type animals, their
40000
1400
30000
mRNA counts
1000 DA1877 1/1000 DA1877
20000 600 10x 10000 200

0
200 M
1/10000 DA1877

5
OP50 HT115 DA1877
B
OP50 1/100 DA1877
100
dh ec 1 h6
D
100 80
OP50 DA1877 1/200 DA1877 1/1000 DA1877
10
Percent survival
60 40
log 2 fold change

-10 -5 0 10
-5
20 0
10 12 14 16 18 20 22 24 26 28
-10
Days past Adult

Figure 4. Comamonas DA1877 Dietary Effect Can Be Diluted
(A) nCounter analysis of gene expression of C. elegans fed different bacterial diets. gst-4 and pqm-1 are induced by oxidative stress. cpt-3 and acs-11 are starvation responsive genes. (B) Mixing diets: numbers indicate proportion of Comamonas DA1877 to E. coli OP50. (C) Animals grown on diluted Comamonas DA1877 develop faster relative to animals grown on OP50. (D) Postdevelopmental survival of adult animals fed each of the indicated diets. OP50: E. coli OP50; DA1877: Comamonas DA1877; 1/200 and 1/1,000 refers to the dilution of Comamonas DA1877 in E. coli OP50. All bacteria were seeded onto peptone-free plates to prevent bacterial growth. (E) Changes in gene expression of animals fed a diet of Comamonas DA1877 diluted in E. coli OP50 compared to Comamonas DA1877 diet alone. See also Figure S3.
development was still accelerated on a Comamonas DA1877 diet (Figure 6B). This demonstrates that NHR-10 is involved in acdh-1 regulation but is not solely responsible for the dietary response. The effect on the molting program by Comamonas DA1877 is either a cause or a consequence of accelerated development. In order to discriminate between these possibilities, we compared the oscillation of the mlt-10 molting reporter between animals fed Comamonas DA1877 and E. coli HB101. Remarkably, although animals fed E. coli HB101 develop more rapidly, the oscillatory pattern of GFP expression was not affected (Figure 6C). This demonstrates that the effect is specic to Comamonas DA1877 and not solely a result of accelerated growth. NHR-23 is a regulator of molting, and mlt-10 is a known transcriptional target of NHR-23. Therefore, we asked whether nhr-23 may be involved in the dietary response. In our microarray experiment of young adult animals, we found that the
30
32
GO term molting cycle is enriched in the genes that change in expression on a Comamonas DA1877 diet. Many of these genes also change in expression in a published study where nhr-23 knockdown was compared to control L1 larvae (Kouns et al., 2011). nhr-23 is an essential gene, and its knockdown causes larval arrest due to molting defects (Kostrouchova et al., 1998). To circumvent this, we diluted bacteria producing nhr-23 double-stranded RNA (dsRNA) 20-fold with bacteria containing vector alone. We examined larval development in both animals with decreased and increased nhr-23 expression (from transgene gaIs269 that expresses an NHR-23::GFP fusion protein). Surprisingly, both knockdown and overexpression of nhr-23 slowed development (Figure 6D). Knockdown of the exogenous copy of nhr-23 by GFP knockdown partially suppressed the latter effect, demonstrating that it is a result of an increase in nhr-23. Thus, both a reduction and an increase in nhr-23 expression decelerate development. The
DA1877 (0.1%)
DA1877 (100%)
O 1 / D P5 2 1/ 00 A18 0 10 D 7 00 A1 7 D 87 A1 7 87 7
ac
ac
1/1000
dh gs -2 pq t-4 m ac -1 scp 11 t-3
OP50 exposure/20
oscillations in mlt-10 expression are part of a tightly regulated system, controlled, at least in part, by nhr-23. Changing nhr-23 expression levels may perturb these oscillations, and adjusting to these perturbations may result in a developmental delay. Diluted nhr-23 RNAi results in a dramatic reduction in GFP expression in adults (Figure 6E). NHR-10 is a direct regulator of acdh-1 (Arda et al., 2010). NHR-23 could therefore act indirectly, for instance, by activating NHR-10. Alternatively, NHR-23 may act in parallel to NHR-10 to affect acdh-1, either directly or indirectly. To test whether NHR-23 mediates its effect on acdh-1 expression through NHR-10, we performed nhr-23 RNAi in the nhr-10(tm4695) mutant animals carrying Pacdh-1::GFP. If NHR-23 affects acdh-1 expression by modulating NHR-10, we would expect GFP levels following RNAi of nhr-23 in an nhr-10 null animal the same as in the nhr-10 mutant alone. However, knocking down nhr-23 further reduced GFP expression (Figure 6F), demonstrating that the effect of nhr-23 on acdh-1 expression is not mediated through nhr-10. DISCUSSION Diet has major effects on the development and health of complex multicellular organisms. We and others have used the nematode C. elegans together with a variety of bacterial diets to unravel the phenotypic adjustments of the animal in response to diet, as well as the mechanisms involved (Coolon et al., 2009; Shtonda and Avery, 2006; Soukas et al., 2009). Different bacteria likely provide different degrees and types of nutrition to the nematode. We focused on the effects of the soil bacterium Comamonas that was isolated together with C. elegans and therefore likely represents a natural food source (Avery and Shtonda, 2003). Relative to the standard laboratory diet of E. coli OP50, a diet of Comamonas DA1877 accelerates development, reduces fecundity, and shortens lifespan. In contrast, unc-119(ed3) mutant animals produce more offspring on Comamonas DA1877 than on E. coli OP50 (data not shown). Thus, our ndings indicate that bacterial diets are not simply healthy or unhealthy but that specic qualities of these diets may be optimal under different conditions and for different life-history traits. In addition, the response to Comamonas DA1877 is likely not a trait that was acquired when C. elegans was established as a genetic model system in the laboratory because another nematode species, C. briggsae, exhibits similar changes in life-history traits when fed the different diets (data not shown). A complication in understanding the inuence of diet on C. elegans physiology is the fact that bacteria are not only a source of food; they can also be pathogenic (Tan and Shapira, 2011). The observation that Comamonas DA1877 can exert its effects on development and lifespan when killed demonstrates that these effects are not due to pathogenicity. However, we noted that several genes previously reported to change in response to specic pathogens were also changed in response to the diets we tested (data not shown). In studies of the C. elegans pathogen response, pathogenic bacteria also serve as food, and therefore, it will be important to disentangle dietary from pathogenic effects in the future.
In the wild, it is likely that C. elegans encounters complex mixtures of bacterial species. Our observation that the effect of Comamonas DA1877 on development and gene expression occurs even when mixed with E. coli OP50 illustrates that the animal is capable of responding to even low amounts of particular types of bacteria. Further, this indicates that it is not the caloric content that is responsible for the effects on physiology and gene expression but rather that the bacteria generate a signal that is interpreted by the nematode. We demonstrate that this signal is not sensed by amphid neurons. Rather, it is likely that this response occurs after ingestion of the bacteria, potentially directly by cells in the intestine. Relative to E. coli OP50, diluted Comamonas DA1877 accelerated C. elegans development but did not affect lifespan. This may suggest that different effects of Comamonas DA1877 may be elicited by different signals. Future biochemical fractionation experiments with Comamonas DA1877 extracts may shed light on the nature of these signals and thereby further illuminate the mechanism by which it is interpreted by the nematode. TOR and insulin represent major nutrient-sensing pathways in a number of organisms (Laplante and Sabatini, 2012; Narasimhan et al., 2009). TOR signaling is regulated by amino acid availability and stress and affects downstream processes, including protein synthesis, energy metabolism, and proliferation. Similarly, insulin signaling regulates the starvation response and lifespan. TOR and insulin signaling pathways regulate developmental rate and likely adjust this rate in accordance with cellular and environmental states. Remarkably, the developmental acceleration of C. elegans in response to Comamonas DA1877 is independent of both of these pathways. We did, however, observe changes in the expression of GFP in the dietary sensor in response to perturbation of TOR. This is likely related to the response of the dietary sensor to food deprivation, as loss of TOR signaling may mimic starvation. NHR-10 directly binds and activates the acdh-1 promoter (Arda et al., 2010) (this study). However, NHR-10 is not solely responsible for the dietary response. In the accompanying paper in this issue of Cell, we nd that NHR-10 is partly responsible for the response to endogenous metabolic network perturbations and identify additional NHRs that affect the dietary sensor (Watson et al., 2013). Here, we identify NHR-23 as a candidate mediator of the response to Comamonas DA1877. Together, these results indicate that a variety of NHRs function coordinately to ensure appropriate transcriptional and physiological responses to different exogenous and endogenous cues. Comamonas DA1877 affects the C. elegans molting program, which is also regulated by nhr-23. Comamonas DA1877 dampens the oscillatory activity of the mlt-10 promoter. Further, in young adult animals, many molting genes are repressed in response to this dietof 266 genes downregulated by nhr-23 knockdown, 39 are also repressed by Comamonas DA1877 (Kouns et al., 2011) (this study). These changes could be either the cause or consequence of the effect that Comamonas DA1877 exerts on cyclic gene expression during molting. Our observation of dampened mlt-10 oscillations demonstrates that this diet does in fact affect molting. mlt-10 is a target of NHR-23, suggesting that this NHR may be activated by
A
Wild type
OP50
DA1877 (exp x100)
C
Wild type
OP50
DA1877
daf-2(e1370)
daf-6 (e1377)
B
60000 50000 acdh-2 acdh-1 ech-6
N2
daf-16(mgDf50)
OP50
100 Adult Late L4 Mid-Late L4 Mid L4 Early-Mid L4 Early L4
40000 30000 20000 10000

0
DA1877
O P H 50 D T11 A1 5 8 O 77 P5 H 0 D T11 A1 5 87 7
N2
daf-6(e1377)
Vector control
100
GFP
100X
rict-1
rheb-1
O P H 50 B1 D 0 A1 1 8 O 77 P5 H 0 B D 10 A1 1 87 7
100X
rsks-1
N2
rict-1(ft7)
200M
Figure 5. Analysis of Known Genes and Pathways Indicates that the Comamonas DA1877 Effect on Development Is Independent of TOR and Insulin Signaling
(A) The dietary sensor is not affected by a loss-of-function mutation in daf-6. (B) nCounter analysis of diet-responsive genes in N2 wild-type and daf-6(e1377) mutant animals. (C) daf-2(e1370) mutant animals respond to Comamonas DA1877 like wild-type animals. (D) daf-16(mgDf50) mutant animals respond to Comamonas DA1877 like wild-type animals.
O D P50 A1 8 O 77 P D 50 A1 87 7
N2 rsks-1(ok1255)
A1 0 8 O 77 P D 50 A1 87 7
200M
mRNA counts
0
O
P5
100
da
f-1
6(
200M
gD
f5
0)
nhr-10(tm4695) OP50
100 Percent total animals
nhr-10 N2 (tm4695) Adult Late L4 Mid-Late L4 Mid L4 Early L4 L3
Figure 6. Two Nuclear Hormone Receptors Affect the Dietary Sensor

(A) nhr-10(tm4695) mutant animals have reduced acdh-1 promoter activity but still respond to Comamonas DA1877. (B) Developmental progression of animals at 45 hr post-L1 synchronization of nhr-10(tm4695) mutant animals fed the indicated diets. (C) Expression of GFP in a synchronized population of Pmlt-10::GFP-pest animals over time in animals fed E. coli OP50, E. coli HB101, or Comamonas DA1877. (D) Developmental rate in N2 and Pnhr-23::NHR23::GFP animals that overexpress NHR-23 (OE). Stages are as in (B). (E) nhr-23 RNAi (diluted 1 in 20) greatly reduces acdh-1 promoter activity. (F) nhr-23 RNAi (diluted 1 in 20) reduces acdh-1 promoter activity in nhr-10(tm4695) mutant animals.
DA1877
C
100 80
O D P5 A1 0 8 O 77 D P5 A1 0 87 7
200M
D
100
N2
OE
Percent GFP positive
mRNA expression does not change in response to diet (data not shown), 60 suggesting a modulation of its activity at the protein level. The C. elegans genome 40 encodes six NHR-23 variants that differ in OP50 their N-terminal domains. One possibility 20 DA1877 is that diet affects different NHR-23 variHB101 ants that have distinct sets of target 0 0 genes at different stages in the animals lifetime. Hours after L1 arrest Oscillating gene expression is pivotal in numerous biological processes, including the cell cycle, molting cycles, and circaE F dian rhythms. Our data show that diet Pacdh-1::GFP Pacdh-1::GFP; nhr-10(tm4695) can affect oscillatory gene expression, which likely affects developmental rate. Comamonas DA1877 both dampens mlt-10 oscillation and shortens its period. On the other hand, E. coli HB101 shortens 3x exp the period without dampening the amplitude. Thus, both diets likely accelerate development by impinging on the molting program. Whereas Comamonas DA1877 3x exp affects developmental rate in a TOR200M independent manner, E. coli HB101 accelerates growth in a TOR-dependent manner. Nutrition also affects developComamonas DA1877. In adults fed this diet, however, the mental rate in Drosophila (Layalle et al., 2008). In ies, the expression of other NHR-23 targets is decreased, suggesting nutritional effect can occur via the TOR pathway, which impinges that NHR-23 may be repressed by this diet. NHRs can both action the molting hormone Ecdysone (Layalle et al., 2008). Ecdyvate and repress transcription (Pardee et al., 2011). nhr-23 sone initiates a transcriptional cascade that activates the
nhr-23 RNAi
nh ve r-2 cto 3 rR R N ve NA Ai ct i 1 o / G r R 20 FP N R Ai N Ai
15 17 19 21 23 25 27 29 31
HT115 vector
(E) Developmental progression of animals at 43 hr post-L1 synchronization of daf-16(mgDf50) and wild-type (N2) animals grown on E. coli OP50 or Comamonas DA1877. (F) RNAi analysis of TOR pathway components. Animals harboring the dietary sensor were fed E. coli HT115 bacteria that express dsRNA for the indicated genes. See also Figure S4. (G) Developmental progression of animals at 48 hr post-L1 synchronization of rict-1(ft7) mutant animals fed the indicated diets. (H) Developmental progression of animals at 50 hr post-L1 synchronization of rsks-1(ok1255) mutant animals fed the indicated diets.
response in the presence of another. This indicates that unhealthy foods can illicit physiological responses in the presence of an otherwise healthy diet or vice versa. This has major implications for treatments for diseases affected by diet such as diabetes, obesity, and cancer. Another implication of our study for human health relates to the observation that bacteria can generate a signal that is interpreted by gene regulatory networks in nematode cells, most likely in the intestine. Numerous bacterial species, known as the microbiome, colonize the human intestine. These commensal bacteria provide numerous benets to our healththey are important in immunity to ward off harmful bacteria, produce essential nutrients and vitamins, and regulate gut development. However, under adverse conditions, the gut ora can inict infections or affect disease progression. It is likely that the microbiota generates a cacophony of signals that can affect the metabolic network of the surrounding intestinal cells. We propose that the nematode C. elegans can be used as a model to provide further insights into the communication between microbes and mammalian cells.
EXPERIMENTAL PROCEDURES Strains C. elegans strains were cultured and maintained by standard protocols (Brenner, 1974). Construction Pacdh-1::GFP was previously described (Arda et al., 2010). The extrachromosomal array was integrated by UV irradiation using standard methods (Evans, 2006) to generate VL749 wwIs24 [Pacdh1::GFP + unc-119(+)]. Integrated lines were outcrossed three times to N2 wild-type animals. The nhr-10(tm4695) mutant was kindly provided by the National Bioresource Project, Japan. Additional strains were obtained from the C. elegans Genetics Center (CGC). The daf-6(e1377) mutant was crossed into VL749 to generate VL840. Animals were genotyped, and dye lling was performed to verify the identity of the daf-6(e1377) mutant, which is dye-lling defective. Bacterial strains E. coli OP50, E. coli HT115(DE3), and Comamonas DA1877 were obtained from the CGC. GR1395 mgIs49 [mlt-10::GFP-pest; ttx-1::GFP] was used to monitor molting (Frand et al., 2005). The RW10429 strain [gaIs269 [nhr-23::TY1::EGFP::3xFLAG + unc-119(+)]was obtained from the CGC. Phenotypic Analysis of Life-History Traits Animals were grown at 20 C. For all diet-specic assays, animals were grown on the appropriate diet for at least one generation prior to the assay. We measured development by rst synchronizing animals by L1 arrest. Briey, animals were grown on the relevant diet, and eggs were collected by bleaching, washed three times in M9 buffer, and allowed to hatch in M9 buffer for 18 hr. Following synchronization, animals were transferred to nematode growth media (NGM) plates and incubated at 20 C. At the indicated times in Figures 1, 4, 5, and 6, animals were washed off the plates, mounted on agarose pads, and examined on a compound microscope. Animals were visually categorized into age groups based on the development of the vulva (Figure S1). At least 40 animals were scored for each diet. Brood sizes were determined by picking individual L4 animals onto plates containing different diets. Animals were transferred daily, and number of offspring on the plates was counted. For lifespan analysis, L4 animals were transferred onto NGM plates seeded with either of the three diets. The following day, the animals were transferred to NGM plates seeded with the appropriate bacteria. Every two days, animals were checked for pharyngeal pumping. If pumping was not observed, animals were lightly prodded with a platinum wire. If animals did not respond, they were considered dead and were scored and removed. For lifespan analysis using diluted bacteria, animals were cultured on peptone-free NGM. Killed bacterial lawns were prepared using an overnight culture of bacteria (E. coli OP50 or Comamonas DA1877). Bacterial cultures were concentrated 2-fold and
Figure 7. Model for Dietary Regulation of Developmental Rate

Specic dietary components impinge on different nutrient-sensing pathways to regulate developmental rate.
NHR-23 ortholog DHR3. This suggests an evolutionarily conserved mechanism that links nutrition, molting, and developmental transitions (Figure 7). In humans, circadian oscillations in gene expression involve the NHR-23 ortholog RORa. Like NHR23, there are multiple isoforms of RORa. These vary in their N termini and differ in their DNA binding specicity, suggesting ` re et al., that they may regulate different target genes (Gigue 1994). Circadian rhythms are greatly affected by diet, and timing of food ingestion can have major effects on physiology (Froy, 2007). Further studies of dietary effects on oscillatory gene expression and the relationships to development, physiology, and lifespan will shed more light on common as well as distinct mechanisms that have evolved in different species. Our study has several implications for human health and nutrition. The rst tantalizing implication relates to our observation that even small amounts of one diet can elicit a physiological
seeded on peptone-free NGM containing tetracycline and spectinomycin. When lawns were dry, plates were UV irradiated using a Stratagene crosslinker as described (Sutphin and Kaeberlein, 2009). Expression Proling Analysis N2 wild-type animals were grown on each diet for one generation prior to egg collection. Eggs were collected and synchronized in L1. All animals were grown on standard NGM plates. Animals fed Comamonas DA1877 developed faster, and thus, all samples could not be collected simultaneously but were instead collected when most animals on the plates reached the young adult stage as judged by the presence of a fully developed vulva or as gravid adults as judged by the presence of eggs in the gonad. Animals were washed twice in M9 buffer, pelleted by centrifugation, and frozen at 80 C in Trizol. RNA was collected using Trizol extraction followed by DNase I treatment and cleanup using RNeasy prep kit (QIAGEN). Three biological replicates were prepared for each condition. Microarray expression proling was performed by the Genomics Core facility at University of Massachusetts Medical School using C. elegans genome arrays that contain probes for predicted coding sequences (Affymetrix). The RMA method in the Affy package from Bioconductor was used in R to summarize the probe level data and to normalize the data set to remove across-array variation. Log-transformed data were used in subsequent analyses. Moderated T statistics in Limma (Smyth, 2004) was used to calculate signicance. Signicance was determined using an adjusted p value (Benjamini and Hochberg, 1995). Changes in gene expression that were 2-fold (p < 0.001) or greater were considered signicant. Microarray data have been submitted to GEO (GSE43959). nCounter mRNA Quantication nCounter assays were performed as per manufacturers instructions (NanoString Technologies). Probes were hybridized with 300 ng of total RNA. Resulting counts were normalized to ama-1 mRNA levels. Probe sequences are provided in Table S3. Mixing Bacterial Diets Liquid cultures were grown overnight at 37 C in Luria-Bertani (LB) broth. E. coli OP50 and Comamonas DA1877 bacterial cultures were diluted to the same OD600 and mixed in ratios indicated in Figures 3F and 3G. Bacterial suspensions were spread onto peptone-free NGM to minimize bacterial growth. Eggs were immediately added to plates, and animals were allowed to develop to adults. For growth assays, eggs were prepared by hypochlorite bleaching of animals grown on E. coli OP50. Eggs were allowed to hatch in M9 to generate a synchronized L1 population. Starvation Assay To assay the response to starvation, animals were rst grown on E. coli OP50 bacteria. L4 animals were washed ve times in M9 and then transferred to unseeded peptone-free NGM plates. After 24 hr, animals were collected and examined for GFP expression. Gene Ontology Analysis Gorilla software (Eden et al., 2009) was used to identify GO term enrichment. Up- or downregulated genes were compared to all genes present on the array. GO terms with a p value < 104 are provided in Figure 2C. RNAi Gene Knockdown Plates were prepared by adding IPTG to a nal concentration of 5 mM to NGM agar. RNAi clones were obtained either from the ORFeome RNAi library (Rual et al., 2004), the Ahringer RNAi library (Kamath et al., 2003), or cloned into the RNAi Gateway Destination vector from ORFeome clones or using genomic DNA. E. coli HT115 RNAi cultures were grown in LB broth containing 100 mg/ml ampicillin to log phase and were concentrated by centrifugation, resuspended in 1/10 volume of M9, and added to RNAi feeding plates. VL749 or N2 animals were synchronized by hypochlorite bleaching, washed in M9 buffer, and added to the E. coli HT115 seeded plates. For nhr-23 RNAi, cultures were grown as described above. Following growth of the cultures, HT115 bacteria containing the nhr-23 RNAi construct were diluted 1/20 with HT115 bacteria containing the empty RNAi vector alone.
Generation of Molting Curves In order to generate molting curves, embryos were collected from E. coli OP50 grown adults by hypochlorite bleaching. Animals were allowed to hatch and were synchronized in L1 by incubation in M9 for 18 hr. Animals were then added to the specied foods and incubated at 20 C. For the longer time course (Figure 1), two experiments were started 12 hr apart, and animals were scored every hour for 12 hr. Animals were scored on a uorescenceequipped dissecting microscope. Animals with appreciable GFP expression were scored positive. At least 50 animals were scored for each time point. ACCESSION NUMBERS The GEO accession number for the microarray data reported in this paper is GSE43959. SUPPLEMENTAL INFORMATION Supplemental Information includes Extended Experimental Procedures, four gures, and three tables and can be found with this article online at http:// dx.doi.org/10.1016/j.cell.2013.02.049. ACKNOWLEDGMENTS We thank members of the Walhout laboratory and Amy Walker for discussion and critical reading of the manuscript. We thank Heidi Tissenbaum for RNAi clones and discussions. We thank Victor Ambros for discussions. Some nematode strains used in this work were provided by the CGC, which is funded by the NIH Ofce of Research Infrastructure Programs (P40 OD010440). This work was supported by NIH grant DK068429 to A.J.M.W., a CIHR postdoctoral fellowship to L.T.M., and an American Heart Association fellowship to E.W. Received: August 15, 2012 Revised: December 18, 2012 Accepted: February 5, 2013 Published: March 28, 2013 REFERENCES Albert, P.S., Brown, S.J., and Riddle, D.L. (1981). Sensory control of dauer larva formation in Caenorhabditis elegans. J. Comp. Neurol. 198, 435451. Arda, H.E., Taubert, S., Conine, C., Tsuda, B., Van Gilst, M.R., Sequerra, R., Doucette-Stam, L., Yamamoto, K.R., and Walhout, A.J.M. (2010). Functional modularity of nuclear hormone receptors in a Caenorhabditis elegans metabolic gene regulatory network. Mol. Syst. Biol. 6, 367. Ashburner, M., Ball, C.A., Blake, J.A., Botstein, D., Butler, H., Cherry, J.M., Davis, A.P., Dolinski, K., Dwight, S.S., Eppig, J.T., et al.; The Gene Ontology Consortium. (2000). Gene ontology: tool for the unication of biology. Nat. Genet. 25, 2529. Avery, L., and Shtonda, B.B. (2003). Food transport in the C. elegans pharynx. J. Exp. Biol. 206, 24412457. Avery, L., and You, Y.J. (2012). C. elegans feeding. In WormBook, The C. elegans Research Community, ed. http://dx.doi.org/10.1895/wormbook. 1.150.1, http://www.wormbook.org. Benjamini, Y., and Hochberg, Y. (1995). Controlling the false discovery rate: a practical and powerful approach to multiple testing. J. R. Stat. Soc., B 57, 289300. Brenner, S. (1974). The genetics of Caenorhabditis elegans. Genetics 77, 7194. Coolon, J.D., Jones, K.L., Todd, T.C., Carr, B.C., and Herman, M.A. (2009). Caenorhabditis elegans genomic response to soil bacteria predicts environment-specic genetic effects on life history traits. PLoS Genet. 5, e1000503. Eden, E., Navon, R., Steinfeld, I., Lipson, D., and Yakhini, Z. (2009). GOrilla: a tool for discovery and visualization of enriched GO terms in ranked gene lists. BMC Bioinformatics 10, 48.
Evans, T.C. (2006). Transformation and microinjection. In WormBook, The C. elegans Research Community, ed. http://www.ncbi.nlm.nih.gov/books/ NBK19648/. Ezcurra, M., Tanizawa, Y., Swoboda, P., and Schafer, W.R. (2011). Food sensitizes C. elegans avoidance behaviours through acute dopamine signalling. EMBO J. 30, 11101122. Frand, A.R., Russel, S., and Ruvkun, G. (2005). Functional genomic analysis of C. elegans molting. PLoS Biol. 3, e312. Froy, O. (2007). The relationship between nutrition and circadian rhythms in mammals. Front. Neuroendocrinol. 28, 6171. Geiss, G.K., Bumgarner, R.E., Birditt, B., Dahl, T., Dowidar, N., Dunaway, D.L., Fell, H.P., Ferree, S., George, R.D., Grogan, T., et al. (2008). Direct multiplexed measurement of gene expression with color-coded probe pairs. Nat. Biotechnol. 26, 317325. ` re, V., Tini, M., Flock, G., Ong, E., Evans, R.M., and Otulakowski, G. Gigue (1994). Isoform-specic amino-terminal domains dictate DNA-binding properties of ROR alpha, a novel family of orphan hormone nuclear receptors. Genes Dev. 8, 538553. Gravato-Nobre, M.J., and Hodgkin, J. (2005). Caenorhabditis elegans as a model for innate immunity to pathogens. Cell. Microbiol. 7, 741751. Honjoh, S., Yamamoto, T., Uno, M., and Nishida, E. (2009). Signalling through RHEB-1 mediates intermittent fasting-induced longevity in C. elegans. Nature 457, 726730. Kamath, R.S., Fraser, A.G., Dong, Y., Poulin, G., Durbin, R., Gotta, M., Kanapin, A., Le Bot, N., Moreno, S., Sohrmann, M., et al. (2003). Systematic functional analysis of the Caenorhabditis elegans genome using RNAi. Nature 421, 231237. Korta, D.Z., Tuck, S., and Hubbard, E.J. (2012). S6K links cell fate, cell cycle and nutrient response in C. elegans germline stem/progenitor cells. Development 139, 859870. Kostrouchova, M., Krause, M., Kostrouch, Z., and Rall, J.E. (1998). CHR3: a Caenorhabditis elegans orphan nuclear hormone receptor required for proper epidermal development and molting. Development 125, 16171626. Kouns, N.A., Nakielna, J., Behensky, F., Krause, M.W., Kostrouch, Z., and Kostrouchova, M. (2011). NHR-23 dependent collagen and hedgehog-related genes required for molting. Biochem. Biophys. Res. Commun. 413, 515520. Lakowski, B., and Hekimi, S. (1998). The genetics of caloric restriction in Caenorhabditis elegans. Proc. Natl. Acad. Sci. USA 95, 1309113096. Laplante, M., and Sabatini, D.M. (2012). mTOR signaling in growth control and disease. Cell 149, 274293. opold, P. (2008). The TOR pathway couples Layalle, S., Arquier, N., and Le nutrition and developmental timing in Drosophila. Dev. Cell 15, 568577. Long, X., Spycher, C., Han, Z.S., Rose, A.M., Muller, F., and Avruch, J. (2002). TOR deciency in C. elegans causes developmental arrest and intestinal atrophy by inhibition of mRNA translation. Curr. Biol. 12, 14481461. Murphy, C.T., McCarroll, S.A., Bargmann, C.I., Fraser, A., Kamath, R.S., Ahringer, J., Li, H., and Kenyon, C. (2003). Genes that act downstream of DAF-16 to inuence the lifespan of Caenorhabditis elegans. Nature 424, 277283. Narasimhan, S.D., Yen, K., and Tissenbaum, H.A. (2009). Converging pathways in lifespan regulation. Curr. Biol. 19, R657R666.
Pan, K.Z., Palter, J.E., Rogers, A.N., Olsen, A., Chen, D., Lithgow, G.J., and Kapahi, P. (2007). Inhibition of mRNA translation extends lifespan in Caenorhabditis elegans. Aging Cell 6, 111119. Pardee, K., Necakov, A.S., and Krause, H. (2011). Nuclear receptors: small molecule sensors that coordinate growth, metabolism and reproduction. Subcell. Biochem. 52, 123153. Reece-Hoyes, J.S., Deplancke, B., Shingles, J., Grove, C.A., Hope, I.A., and Walhout, A.J.M. (2005). A compendium of Caenorhabditis elegans regulatory transcription factors: a resource for mapping transcription regulatory networks. Genome Biol. 6, R110. Reece-Hoyes, J.S., Barutcu, A.R., McCord, R.P., Jeong, J.S., Jiang, L., MacWilliams, A., Yang, X., Salehi-Ashtiani, K., Hill, D.E., Blackshaw, S., et al. (2011). Yeast one-hybrid assays for gene-centered human gene regulatory network mapping. Nat. Methods 8, 10501052. Rual, J.-F., Ceron, J., Koreth, J., Hao, T., Nicot, A.-S., Hirozane-Kishikawa, T., Vandenhaute, J., Orkin, S.H., Hill, D.E., van den Heuvel, S., and Vidal, M. (2004). Toward improving Caenorhabditis elegans phenome mapping with an ORFeome-based RNAi library. Genome Res. 14(10B), 21622168. Shtonda, B.B., and Avery, L. (2006). Dietary choice behavior in Caenorhabditis elegans. J. Exp. Biol. 209, 89102. Smyth, G.K. (2004). Linear models and empirical bayes methods for assessing differential expression in microarray experiments. Stat. Appl. Genet. Mol. Biol. 3, Article3. Sommer, R.J., and Ogawa, A. (2011). Hormone signaling and phenotypic plasticity in nematode development and evolution. Curr. Biol. 21, R758R766. Sonoda, J., Pei, L., and Evans, R.M. (2008). Nuclear receptors: decoding metabolic disease. FEBS Lett. 582, 29. Soukas, A.A., Kane, E.A., Carr, C.E., Melo, J.A., and Ruvkun, G. (2009). Rictor/ TORC2 regulates fat metabolism, feeding, growth, and life span in Caenorhabditis elegans. Genes Dev. 23, 496511. Sutphin, G.L., and Kaeberlein, M. (2009). Measuring Caenorhabditis elegans life span on solid media. J. Vis. Exp. 27, 1152. Tan, M.W., and Shapira, M. (2011). Genetic and molecular analysis of nematode-microbe interactions. Cell. Microbiol. 13, 497507. Tan, M.W., Mahajan-Miklos, S., and Ausubel, F.M. (1999). Killing of Caenorhabditis elegans by Pseudomonas aeruginosa used to model mammalian bacterial pathogenesis. Proc. Natl. Acad. Sci. USA 96, 715720. hrsen, K. (1998). Tawe, W.N., Eschbach, M.L., Walter, R.D., and Henkle-Du Identication of stress-responsive genes in Caenorhabditis elegans using RT-PCR differential display. Nucleic Acids Res. 26, 16211627. Timmons, L., Court, D.L., and Fire, A. (2001). Ingestion of bacterially expressed dsRNAs can produce specic and potent genetic interference in Caenorhabditis elegans. Gene 263, 103112. Van Gilst, M.R., Hadjivassiliou, H., and Yamamoto, K.R. (2005a). A Caenorhabditis elegans nutrient response system partially dependent on nuclear receptor NHR-49. Proc. Natl. Acad. Sci. USA 102, 1349613501. Van Gilst, M.R., Hadjivassiliou, H., Jolly, A., and Yamamoto, K.R. (2005b). Nuclear hormone receptor NHR-49 controls fat consumption and fatty acid composition in C. elegans. PLoS Biol. 3, e53. Watson, E., MacNeil, L.T., Arda, H.E., Zhu, L.J., and Walhout, A.J.M. (2013). Integration of metabolic and gene regulatory networks modulates the C. elegans dietary response. Cell 153, this issue, 253266.
Integration of Metabolic and Gene Regulatory Networks Modulates the C. elegans Dietary Response
Emma Watson,1,2,4 Lesley T. MacNeil,1,2,4 H. Efsun Arda,1,2,5 Lihua Julie Zhu,3 and Albertha J.M. Walhout1,2,*
in Systems Biology in Molecular Medicine 3Program in Gene Function and Expression University of Massachusetts Medical School, Worcester, MA 01605, USA 4These authors contributed equally to this work 5Present address: Department of Developmental Biology, Stanford University School of Medicine, Stanford, CA 93405, USA *Correspondence: marian.walhout@umassmed.edu http://dx.doi.org/10.1016/j.cell.2013.02.050
2Program 1Program
SUMMARY
Expression proles are tailored according to dietary input. However, the networks that control dietary responses remain largely uncharacterized. Here, we combine forward and reverse genetic screens to delineate a network of 184 genes that affect the C. elegans dietary response to Comamonas DA1877 bacteria. We nd that perturbation of a mitochondrial network composed of enzymes involved in amino acid metabolism and the TCA cycle affects the dietary response. In humans, mutations in the corresponding genes cause inborn diseases of amino acid metabolism, most of which are treated by dietary intervention. We identify several transcription factors (TFs) that mediate the changes in gene expression upon metabolic network perturbations. Altogether, our ndings unveil a transcriptional response system that is poised to sense dietary cues and metabolic imbalances, illustrating extensive communication between metabolic networks in the mitochondria and gene regulatory networks in the nucleus.
INTRODUCTION To maintain homeostasis, a cell must be able to sense its own energy state, assess nutrient availability, and modulate metabolic pathways in a coordinated fashion. Organisms utilize and integrate endocrine, allosteric, and transcriptional mechanisms to respond to nutrients and activate or repress appropriate metabolic pathways accordingly. Different diets provide nutrients in different proportions and affect the transcription of metabolic genes in different ways. Dramatic changes in gene expression occur following dietary shifts in C. elegans, D. melanogaster, and M. musculus (Carsten et al., 2005; Du et al., 2010; MacNeil et al., 2013 [this issue of Cell]).
Ensuring optimal metabolic tuning to complex dietary signals requires integration of nutrient sensors and their downstream targets to produce a coherent response. Studies in yeast have revealed coexpression of metabolic enzymes as a central mechanism of nutrient response. For example, the yeast transcription factor (TF) Gcn4 activates enzymes involved in amino acid synthesis upon amino acid starvation (Hinnebusch, 2005). This type of coordinated, need-based metabolic transcriptional response to nutrients is an important mechanism in yeast, but the extent to which it occurs in more complex multicellular organisms remains largely unknown. Inborn errors of metabolism are relatively rare recessive disorders that are characterized by a buildup of metabolites that cannot be processed and/or a lack of metabolites required for basic cellular processes (Saudubray et al., 2006). Patients with such diseases can display a variety of symptoms such as failure to thrive, developmental delay, and seizures (Acosta, 2010). These diseases are often clinically managed by altering the patients diet to avoid buildup of toxic metabolites and to supplement limiting nutrients. For example, maple syrup urine disease, a disorder in which patients fail to catabolize the branched chain amino acids (BCAAs) leucine, isoleucine, or valine, is managed by limiting the intake of protein-rich foods. The nematode C. elegans is a free-living bacterivore that is a genetically tractable model instrumental in understanding mechanisms relating to human disease. For instance, genomescale RNA interference (RNAi) resources have been used to nnichsen et al., identify essential genes (Kamath et al., 2003; So 2005), genes affecting fat storage (Ashra et al., 2003), and other processes. Further, the animal is transparent, which has enabled the use of green uorescent protein (GFP) to monitor spatiotemporal gene expression in living animals (Chale et al., 1994; Grove et al., 2009; Hunt-Newbury et al., 2007; Martinez et al., 2008). The core metabolic networks and major nutrient-sensing pathways, including insulin and TOR, are highly conserved between C. elegans and humans. However, the TF family of nuclear hormone receptors (NHRs) that regulate metabolic gene expression has greatly expanded, with a total of 271 members, compared to 48 in humans (Reece-Hoyes et al.,
2011a, 2011b). NHRs function as ligand-gated TFs and are postulated to provide the animal with a repertoire of responses to different environmental conditions (Arda et al., 2010; Taubert et al., 2011). C. elegans exhibits an altered gene expression prole when fed a bacterial diet of Comamonas DA1877 compared to the standard laboratory diet of E. coli OP50, and these changes are accompanied by altered life-history traits such as developmental rate, number of offspring, and lifespan (MacNeil et al., 2013). The metabolic gene acdh-1 exhibits the most dramatic change in expression between these diets. We developed a dietary sensor strain comprising the acdh-1 promoter driving the expression of GFP when fed the standard laboratory diet of E. coli OP50, GFP expression is high, but on a Comamonas DA1877 diet, GFP expression is dramatically reduced (MacNeil et al., 2013). Here, we perform forward and reverse genetic screens to identify genes that can alter the transcriptional response to diet. We identify 184 genes, including 146 activators and 38 repressors. Most of the 38 repressors encode enzymes that function within four mitochondrial metabolic pathways: BCAA breakdown, methionine metabolism, glycine cleavage system, and the tricarboxylic acid cycle (TCA). Human orthologs of most of the dietary response repressors confer inborn diseases of amino acid metabolism when mutated. In addition to acdh-1, the transcription of several other metabolic genes from these four pathways is modulated in response to both diet and metabolic network perturbations. These genes represent two modules that exhibit reciprocal behaviorone module, which includes acdh-1, is repressed by Comamonas DA1877 but is activated by specic metabolic network perturbations, whereas the other is activated by Comamonas DA1877 but is repressed by these metabolic network perturbations. We identify the TFs SBP-1, NHR-10, and NHR-68 as candidate mediators of the response to metabolic perturbations. Taken together, we uncover extensive communication between mitochondrial metabolic networks and nuclear gene regulatory networks, which may serve to optimize metabolic ux under different dietary and/or metabolic conditions. RESULTS A Forward Genetic Screen for Dietary Sensor Repressors Animals harboring the Pacdh-1::GFP dietary sensor display much lower levels of GFP expression when fed Comamonas DA1877 than when fed the standard laboratory diet of E. coli OP50 (MacNeil et al., 2013). ACDH-1 has been annotated as an acyl-CoA dehydrogenase involved in b-oxidation of short chain fatty acids (Van Gilst et al., 2005) and as a short branched chain acyl-CoA dehydrogenase involved in BCAA breakdown (Murphy et al., 2003). Sequence comparison of the ACDH-1 protein with the human ACADSB enzyme (which functions in BCAA breakdown) and the human SCAD enzyme (which functions in b-oxidation) indicates that ACDH-1 more closely resembles ACADSB (Figure S1 available online). We performed a forward genetic screen to identify genes that, when mutated, cause activation of the acdh-1 promoter on a Comamonas DA1877 diet. We mutagenized the dietary sensor
strain, screened 10,000 genomes, and isolated 45 fertile F2 mutants that produced GFP-positive offspring (Figure 1A). These mutants varied in both level and pattern of GFP expression. For example, some mutants displayed relatively broad expression in the intestine, hypodermis, and muscle, whereas others express GFP only in the intestine (Figure 1B and Table S1). We mapped 20 mutations to four chromosomes. Five mutants were selected for whole-genome sequencing, one from each linkage group and one additional mutant from linkage group IV. We identied ve mutations in four genes: mmcm-1, pccb-1, F32B6.2, and F25B4.1, all of which encode mitochondrial metabolic enzymesmmcm-1 encodes methylmalonyl-CoA mutase, pccb-1 encodes the b subunit of propionyl-CoA carboxylase, and F32B6.2 encodes an ortholog of a methylcrotonoyl-CoA carboxylase. These three proteins are all involved in BCAA breakdown. The fourth gene, F25B4.1, encodes the T protein of the glycine cleavage system. We have named F32B6.2 mccc-1 and F25B4.1 gcst-1 to reect these annotated biochemical functions. By complementation tests and sequencing, we identied additional alleles of three of the four genes (Figure 1C). To verify that the mutations truly affect the expression of acdh-1, we examined endogenous messenger RNA (mRNA) levels in four mutants. When grown on Comamonas DA1877, all mutants showed an increase in acdh-1 mRNA levels relative to wild-type animals, which conrms the results obtained with the dietary sensor (Figure 1D, orange bars). We found similar changes in expression of two other diet-responsive genes, acdh-2 and ech-6, in these mutants (Figure 1D and data not shown). When fed E. coli OP50, the expression of these genes was similar in wild-type and mutant animals (Figure 1D, purple bars), demonstrating that the mutations do not cause a general increase in their expression. In three of the four mutants, the expression of acdh-1 was lower on Comamonas DA1877 than on E. coli OP50, indicating that they are still somewhat able to respond to dietary cues (Figure 1D, compare orange and purple bars). The dietary response was completely impaired only in the mmcm-1(ww5) mutant (Figure 1D). The acdh-1 promoter not only responds to different bacterial diets but is also repressed upon starvation (MacNeil et al., 2013; Van Gilst et al., 2005). We tested the response to starvation in our mutants. When starved for 24 hr, mccc-1(ww4) mutant animals retain GFP expression, which indicates that the starvation response is impaired in these mutants as well. In contrast, in mmcm-1, pccb-1, and gcst-1 mutants, GFP expression was reduced, indicating that the response to starvation is retained (Figure 1E). mmcm-1(ww5) mutants are completely impaired in the dietary response (Figure 1D) but retain a starvation response, conrming that these two responses are distinct (MacNeil et al., 2013). Altogether, these observations show that the dietary sensor can respond not only to diet but also to endogenous metabolic network perturbations. A Genome-Scale RNAi Screen To identify additional genes that can affect the dietary sensor, we explored the use of RNAi by feeding, which is carried out in E. coli HT115 bacteria (Timmons et al., 2001). When fed E. coli HT115, the Pacdh-1::GFP animals express intermediate levels of GFP,
Figure 1. Forward Genetic Screen for Dietary Sensor Regulators

(A) Flow chart of EMS mutagenesis screen. (B) Examples of GFP expression in wild-type animals and two mutant strains. (C) Cartoon illustrating amino acid changes in four proteins identied in the forward genetic screen. BD, biotin binding domain; B12B, vitamin B12 binding domain. (D) Endogenous acdh-1 and acdh-2 levels in mutants as measured by nCounter assays and normalized to ama-1. (E) Mutants exhibit different responses to starvation. See also Tables S1 and S2.
Figure 2. Reverse Genetic Screen for Dietary Sensor Regulators

(A) RNAi knockdown works efciently in the presence of Comamonas DA1877. Pacdh-1::GFP animals were fed E. coli HT115, producing either double-stranded RNA (dsRNA) directed against mccc-1 (which was found in the forward genetic screen) or no dsRNA (vector control) in the presence and absence of Comamonas DA1877. Comamonas DA1877 had no effect on Pmir-63::GFP. (B) Flow chart of genome-scale RNAi screen. (C) qRT-PCR of endogenous acdh-1 mRNA in wild-type animals subjected to RNAi of 11 genes found in the RNAi screen. Values were normalized to expression in a vector only control and plotted as log2 fold change. Yellow bars indicate genes whose knockdown caused increased GFP expression, and blue bars indicate genes whose knockdown caused decreased GFP expression. Error bars indicated the SE in three technical repeats. (D) Validation of RNAi results using mutant animals. Bright eld and uorescent images of Pacdh-1::GFP animals with wild-type or mutant backgrounds, fed either E. coli OP50 or Comamonas DA1877. Yellow indicates increases in GFP expression, and blue indicates decreased GFP expression compared to wild-type Pacdh-1:GFP dietary sensor animals. Exposure times that are different from Pacdh-1:GFP dietary sensor animals fed E. coli OP50 are indicated in the top right corner of the respective panel. (E) qRT-PCR of endogenous acdh-1 mRNA in Dsams-5 and Dhlh-11 mutants compared to wild-type animals fed E. coli OP50, E. coli HT115, or Comamonas DA1877. Changes in acdh-1 expression were plotted as log2 fold change compared to acdh-1 levels in wild-type animals fed E. coli OP50. Error bars indicated the SE in three technical repeats. See also Figures S3 and S4 and Table S2, S3, and S5.
enabling us to identify both dietary sensor repressors and activators. In addition, we took advantage of the fact that Comamonas DA1877 can be diluted and still repress the dietary sensor (MacNeil et al., 2013). Adding a small amount of Comamonas DA1877 to the E. coli HT115 RNAi feeding lawn repressed the dietary sensor while maintaining RNAi knockdown efciency (Figure 2A). This enabled us to perform a genome-scale RNAi screen in the presence of Comamonas DA1877 (Figure 2B). We screened the ORFeome RNAi library, which covers more than half of all predicted protein-coding genes (Rual et al.,
2004). After screening the library once in each dietary condition, we obtained 836 hits (8%, Figure 2B). These hits were rearrayed and retested four times to remove false positives. The 554 genes that rescored in at least two of four retests were considered further. This list of hits contained many genes involved in general protein expression that may not specically affect the dietary sensor. To identify such potential false positives, we screened the 554 hits against another transgene, Pmir-63::GFP, that also expresses GFP in the intestine but is not sensitive to Comamonas DA1877 (Figures 2A and 2B). We retained 179
high-condence hits (2% of the genes tested). Of these, 146 caused a decrease in GFP expression (activators), whereas 33 caused an increase (repressors) (Table S2). We retrieved nhr-10 and mdt-15, known activators of acdh-1 (Arda et al., 2010), as well as two of the genes (mccc-1 and gcst-1) found in the forward genetic screen. The other two genes (mmcm-1 and pccb-1) were not retrieved, although they were present in the RNAi library. This could be due to the inherent variability and the relatively high false negative rate of RNAi experiments (Kamath et al., 2003). Indeed, when we retested these clones in a small-scale, directed RNAi experiment, we found that their knockdown does increase GFP expression (data not shown). RNAi Screen Validation To verify that the genes found by RNAi affect endogenous acdh-1 expression, we used quantitative RT-PCR (qRT-PCR) in wild-type (N2) animals subjected to RNAi of 11 representative genes found in the screen. As shown in Figure 2C, the qRTPCR recapitulated what was observed in the Pacdh-1::GFP dietary sensor strain for all 11 knockdowns. We obtained C. elegans mutant strains for several genes found in the RNAi screen, including the TFs nhr-68 and hlh-11, and the enzymes ZK669.4, sams-5, sams-1, and metr-1 (Figure S2). We introduced each mutation into the Pacdh-1::GFP dietary sensor strain and examined GFP levels on E. coli OP50 and Comamonas DA1877 (Figure 2D). These mutants exhibited GFP expression levels that recapitulate the effects observed by RNAimutations in the enzymes and hlh-11 all activated GFP expression, whereas a deletion in nhr-68 decreased GFP expression. Further, we conrmed the increased expression of endogenous acdh-1 in animals fed E. coli OP50 and Comamonas DA1877 by qRT-PCR in Dhlh-11 and Dsams-5 mutant animals (Figure 2E). To probe potential functional relationships among the genes discovered in the genetic screens, we assessed their connectedness in WormNet, a probabilistic functional gene network constructed through integration of different data types (Lee et al., 2008). The combined 181 genes found by both the forward and reverse genetic screens are signicantly connected in WormNet and form a closely linked functional network (p = 1.97 3 1078) (Figure S3). Thus, these 181 genes are more functionally interrelated than would be expected for a random set of genes, supporting the overall quality of the RNAi screen. Dietary Sensor Activators The majority (74%) of the 146 activators are expressed in the intestine (Figure S3), suggesting that they may act cell autonomously. Gene ontology (GO) analysis revealed an enrichment for several terms, including growth, sex differentiation, and ribosome biogenesis (Figure S3). Several of these genes are predicted to function in a number of biological processes, including general gene expression, splicing and translation, metabolism, and the regulation of transcription (Figure S3). Because acdh-1 is also repressed in response to starvation, the dietary sensor strain cannot discriminate between genes involved in, or inducing, a starvation response and those involved in the dietary response. To identify which activators may mediate a starvation response and which may be more
specic to the dietary response, we used two additional transgenic strains in which GFP expression is repressed by starvation, but not affected by bacterial diet (Pgst-4::GFP and PC53A3.2::GFP, Figure S4). We found 37 activators that affect at least one of these transgenes, 12 of which affected both, indicating that they may induce or mediate a starvation rather than dietary response (Table S3). The remaining 109 activators were specic to Pacdh-1::GFP. Dietary Sensor Repressors Function in Four Metabolic Pathways The combined 35 dietary sensor repressors obtained in both screens are enriched for GO terms relating to metabolism, including mitochondrion, TCA cycle, and acetyl-CoA catabolic process (Figure S3). Overall, 83% of these are annotated metabolic genes, compared to 15% of the activators, and the vast majority are expressed in the intestine (Figure S3). Interestingly, most (24) of the repressors encode enzymes specically involved in four metabolic pathways: BCAA breakdown (n = 8), methionine metabolism (n = 5), the glycine cleavage system (n = 2), and the TCA cycle (n = 9) (Figure 3). We wondered whether other genes in these pathways that were not retrieved in the genetic screens would affect the dietary sensor as well. We obtained deletion mutants for two genes involved in BCAA breakdown (mce-1 and pcca-1) and one from the methionine metabolism pathway (cbl-1) (Figure 3). These deletions were introduced into the Pacdh-1::GFP dietary sensor strain. Deletions in mce-1 and pcca-1 both caused increased GFP expression on all three diets, and a deletion in cbl-1 caused increased GFP on Comamonas DA1877, but not on E. coli OP50 (Figure S5). Altogether, these ndings indicate that disrupting these four metabolic pathways affects the dietary response to Comamonas DA1877. The observation that loss of several individual genes within a common pathway can induce similar changes in the expression of acdh-1 suggests that these perturbations may converge onto a common regulatory signal. Interestingly, acdh-1 itself was found as a repressor in the RNAi screen (Table S2). We obtained an acdh-1 deletion mutant and introduced it into the Pacdh-1::GFP dietary sensor strain. We observed higher GFP levels in the acdh-1 mutant relative to the wild-type dietary sensor strain when the animals were fed E. Coli OP50 or HT115 diets but observed less of an increase on a Comamonas DA1877 diet. On the latter diet, acdh-1 expression is normally low (Figure 4A). Thus, feedback on the acdh-1 promoter occurs mainly on diets in which endogenous acdh-1 levels are normally high, suggesting that acdh-1 expression is regulated according to need. Cellular metabolic needs are unlikely to be met by modulating the expression of a single enzyme; rather, enzymes within a metabolic pathway or network may need to be coordinately ning et al., 2010). In addition to acdh-1, several regulated (Gru other genes predicted to function within BCAA breakdown and methionine metabolism pathways are also transcriptionally regulated in response to diet (Figure 3) (MacNeil et al., 2013). This suggests that these genes comprise a coordinately regulated module. We wondered whether these genes were also transcriptionally sensitive to metabolic network perturbation. We examined the expression of seven Comamonas
Figure 3. Dietary Repressors Are Enriched in Four Metabolic Pathways

Cartoon of a metabolic map showing C. elegans BCAA breakdown, methionine metabolism, glycine cleavage system, and TCA cycle. Rectangles indicate genes; circles are metabolites. See also Figures S1 and S5.
DA1877-responsive genes (four from BCAA breakdown, two from methionine metabolism, and one NHR) by qRT-PCR in several mutant strains that exhibit altered acdh-1 expression (Dmetr-1, Dsams-1, Dmce-1, Dacdh-1, and Dhlh-11). We found that several BCAA breakdown genes, as well as nhr-68, are coinduced with acdh-1 in response to metabolic network perturbation (Figure 4B). This further supports the modular regulation of these functionally related genes and indicates that this may perhaps be mediated by nhr-68 (see below). The expression of acdh-2, a close homolog of acdh-1, is also increased in an acdh-1 deletion mutant, which may indicate a compensation mechanism of acdh-2 for the loss of acdh-1 function. Interestingly, the Comamonas DA1877-induced gene cbs-1 (MacNeil et al., 2013) exhibited opposite behavior compared to the Comamonas DA1877-repressed genes because it is repressed in the metabolic gene mutants. To characterize the transcriptomic response to metabolic gene perturbations in more detail, we performed a microarray expression proling experiment with two metabolic gene mutants on a Comamonas DA1877 diet. One of these genes,
pcca-1, functions in BCAA breakdown, whereas the other, metr-1, is involved in methionine metabolism. The two mutations elicit highly similar changes in gene expression (Figure 4C), further suggesting that these two different metabolic mutations may cause similar metabolomic changes that may impinge on a common regulatory signal. Strikingly, most of the core genes that change in response to a Comamonas DA1877 diet (MacNeil et al., 2013) also exhibited expression changes in one or both metabolic mutants (Figure 4D). In fact, there are two modules of coexpressed genes that exhibit reciprocal behaviorone that is repressed by a Comamonas DA1877 diet but is activated in the metabolic gene mutants and one that is activated by a Comamonas DA1877 diet but is repressed in the metabolic gene mutants (Figure 4E and Table S4). Human Orthologs of Dietary Repressors Are Involved in Inborn Metabolic Disease Many metabolic reactions are evolutionarily conserved, and most of the C. elegans enzymes found in our genetic screens have human orthologs (Table S2). Most of these are associated
Figure 4. Metabolic Feedback and Transcriptional Compensation

(A) Increased acdh-1 promoter activity in Dacdh-1 mutants identies feedback control. (B) qRT-PCR of seven genes (rows) in different strains (columns). (C) Overlap in gene expression changes between two metabolic gene mutants. (D) Overlap between genes that change in expression in the two metabolic gene mutants and those that change in response to a Comamonas DA1877 diet. (E) Opposite changes in gene expression in response to a Comamonas DA1877 diet versus metabolic network perturbations. See also Figure S2 and Table S4.
with inborn disorders of amino acid metabolism (Figure 5). For instance, we identied multiple orthologs of genes that, when mutated in humans, cause maple syrup urine disease, methylmalonic acidemia, homocystinuria, or propionic acidemia. These diseases are characterized by toxic buildups of amino acids and intermediate metabolites and are treated by limiting amino acid
(protein) intake and supplementing vitamins and other cofactors (Figure 5). We also found orthologs of TCA cycle genes that cause diseases such as lactic acidosis and pyruvate dehydrogenase deciency. These may also result in the accumulation of intermediate metabolites or in the inability to generate sufcient amounts of energy (ATP).
Figure 5. Orthologs of Dietary Sensor Repressors Confer Human Inborn Metabolic Disorders Relating to Amino Acid Metabolism
Network connecting dietary sensor repressors, C. elegans phenotypes, human orthologs, human inborn metabolic diseases, and dietary treatments. Orange nodes, C. elegans phenotypes; colored squares, C. elegans genes found in the screen; diamonds, human orthologs/homologs; gray circles, human diseases; hexagons, nutrients. Red edges indicate dietary avoidance; green edges indicate dietary supplementation.
Mutations in two C. elegans metabolic genes (Dmetr-1 and Dpcca-1) cause highly similar changes in gene expression (Figure 4C). metr-1 is the ortholog of human methionine synthase (MTR), which, when mutated, causes methylcobalamin (vitamin B12) deciency. pcca-1 is the ortholog of human PCCA, which, when mutated, causes propionic acidemia. Our observations suggest that mutations in these genes may result in similar metabolic effects. The human disorders caused by mutations in these genes are treated in part by supplementation with vitamin B12, and both disorders can be revealed by elevated propionylcarnitine levels in newborn screening (Weisfeld-Adams et al., 2010). This suggests that, as in C. elegans, these two human diseases have at least partially overlapping molecular phenotypes. Limited Involvement of TOR and Insulin Signaling Pathways in Response to Metabolic Network Perturbations To determine whether known nutrient sensing pathways such as TOR or the insulin signaling pathway are involved in upregulating acdh-1 in response to metabolic network perturbations, we performed RNAi on a panel of genes from these pathways in metabolic mutants harboring the Pacdh-1::GFP transgene. Knockdown of daf-2 or daf-16 had no effect on GFP expression in any of the mutants (Figure 6A). We also crossed the Dmetr-1 and Dacdh-1 metabolic mutations into a transgenic strain expressing a DAF-16::GFP fusion protein to determine whether these metabolic mutants affected DAF-16 nuclear localization and found that neither had any effect (data not shown). Thus, insulin signaling does not mediate the response to metabolic network perturbations. The TOR pathway is an attractive candidate for mediating the effects of metabolic perturbations because it affects acdh-1 in wild-type animals and is regulated by amino acid levels, specifically leucine (Laplante and Sabatini, 2012). We tested whether the TOR pathway is involved in mediating the response to endogenous metabolic network perturbations and found limited involvement (Figures 6A and 6B). Knockdown of rheb-1 or ruvb-1 reduced GFP levels in most metabolic gene mutants. However, in almost every case, GFP was still higher in these mutant strains than in the wild-type strain (Figures 6A and 6B). Therefore, it is likely that TOR signaling may be partially involved in, but not solely responsible for, mediating the induction of acdh-1 in metabolic mutants. Multiple TFs Affect the Response to Metabolic Network Perturbations In wild-type animals, both nhr-10 and nhr-23 activate the dietary sensor (Arda et al., 2010; MacNeil et al., 2013). In the RNAi screen, we identied additional TFs that affect acdh-1 expression, most of which are NHRs (Table S2). We knocked down nhr-10, nhr-23, nhr-68, nhr-173, and nhr-74, as well as sbp-1, by RNAi in multiple metabolic gene mutants and examined changes in GFP expression. We observed complex gene regulatory effects of different TFs (Figures 6A and 6B). First, sbp-1 RNAi resulted in a dramatic decrease in GFP expression in all animals tested, indicating that SBP-1 might function at the top of the metabolic gene regulatory hierarchy. However, sbp-1 RNAi also reduced GFP expression in the two starvation-responsive
transgenic strains (Table S3). Previously, SBP-1 has been shown to regulate genes involved methionine metabolism (Walker et al., 2011). Altogether, these observations suggest that SBP-1 may be a general regulator of metabolic gene expression in the C. elegans intestine. Second, nhr-23 RNAi has only mild effects on GFP expression in most mutants, whereas it has stronger effects in the wild-type sensor strain, suggesting that it primarily controls acdh-1 expression in response to diet. However, it is important to note that nhr-23 RNAi was performed under dilute conditions to avoid larval arrest. Third, nhr-10 and nhr-68 knockdown dramatically reduces GFP expression in almost all mutants. For example, GFP levels in mce-1 and pcca-1 deletion animals are equivalent to those in wild-type animals subjected to nhr-10 or nhr-68 knockdown (Figure 6B). Therefore, nhr-10 and nhr-68 may mediate the increase in acdh-1 expression in response to specic metabolic network perturbations. Knockdown of nhr-10 in an nhr-68 deletion mutant or vice versa had little additional effect on GFP expression, indicating that these two NHRs may function together (Figure 6A). Knockdown of the other NHRs tested, nhr-173 and nhr-74, had mild effects on GFP expression in wild-type animals and had similarly mild effects in the context of metabolic network perturbations. However, knockdown of nhr-173 further reduces GFP expression in an nhr-10 deletion mutant, indicating that these NHRs may function in parallel. DISCUSSION We have unraveled a complex network of genes that affect dietary gene expression in C. elegans. This network is likely not yet complete for several reasons. First, forward genetics identied mutations in four metabolic genes, and although we have multiple alleles for three of these genes, the screen is not yet saturated. Second, the reverse genetic RNAi screen has not revealed all genes involved because the ORFeome RNAi library contains only half of all C. elegans protein-coding genes (Rual et al., 2004) and because RNAi screens are prone to a high false negative rate (Kamath et al., 2003). This is supported by the initial retrieval of only two of the four genes found by forward genetics and the identication of additional genes through targeted analysis of mutants in the metabolic pathways involved. We primarily focused on dietary sensor repressors, most of which encode mitochondrial enzymes. We found that these repressors function in an intricate feedback and compensation system that results in regulation of metabolic gene expression. This is perhaps best exemplied by the feedback in which deletion of acdh-1 results in an increase in its own promoter activity under dietary conditions when acdh-1 levels are normally high. We know that this feedback extends from acdh-1 to other genes involved in BCAA breakdown because additional genes in this pathway either change in expression in response to a Comamonas DA1877 diet, confer an increase in acdh-1 promoter activity when perturbed, or both. Thus, as has been observed in simpler ning et al., unicellular organisms such as bacteria and yeast (Gru 2010), metabolic changes induced genetically or by diet likely result in disruption of metabolic ux that, in turn, elicit a compensatory transcriptional response of relevant metabolic genes.
Figure 6. Mediators of the Response to Metabolic Network Perturbation

(A) Summary of RNAi knockdown analysis in different mutant strains. INS, insulin signaling pathway components. The colors indicate visually examined changes in GFP expression in Pacdh-1::GFP dietary sensor animals. (B) DIC and uorescent images of indicated strains (columns) exposed to different RNAi knockdowns (rows). Yellow arrowheads indicate an increase in GFP expression in the head hypodermis in acdh-1 RNAi knockdown animals.
Altogether, our study reveals extensive communication between mitochondrial metabolic networks and nuclear gene regulatory networks that function to dial metabolic gene expression according to cellular or organismal need (Figure 7). The mechanisms by which information is relayed from the mitochondria to the nucleus remain poorly understood. There are several shared metabolites among the four metabolic pathways found in the screens, including acetyl-CoA, succinylCoA, pyruvate, and the cofactor cobalamin (vitamin B12) (Figure 3). These shared metabolites represent points of con262 Cell 153, 253266, March 28, 2013 2013 Elsevier Inc.
vergence for metabolic ux through each of the individual pathways. Perturbing ux through methionine metabolism by mutating metr-1, for instance, may affect ux through BCAA breakdown due to buildup or depletion of shared metabolites. This could explain the observation that several genes from these two pathways are coregulated in response to diet and metabolic perturbation in either pathway. These metabolites may mediate the communication from the mitochondria to the nucleus. It is well known that these metabolites have broad ranging effects on the cell. For instance, epigenetic effects can be mediated
Figure 7. Model for Integration of Metabolic Network Perturbation and Dietary Response
Perturbing metabolic networks in mitochondria may lead to imbalances that are sensed by gene regulatory networks in the nucleus and may interfere with the dietary response.
by histone acetylation, methylation, and phosphorylation, which require acetyl-CoA, SAM, and ATP, respectively. It is possible that overall or specic protein modication levels are affected by metabolic network perturbations and result in changes in gene expression. Further, these metabolites may directly or indirectly target TFs such as NHRs (see below). Interestingly, there was a high degree of overlap in both upregulated and downregulated genes in the gene expression proles of two metabolic gene mutants, Dmetr-1 and Dpcca-1. This suggests that the functional defects caused by perturbing either the BCAA breakdown pathway or methionine metabolism may elicit similar metabolomic shifts in the animal, which may impinge upon a common regulatory signal. Both metabolic gene mutations reverse many of the gene expression changes conferred by a Comamonas DA1877 diet in wild-type animals. Two modules of coexpressed genes exhibit reciprocal seesaw behaviorone module is repressed by a Comamonas DA1877 diet but is activated in the metabolic mutants, whereas
another is activated by a Comamonas DA1877 diet but is repressed in the metabolic gene mutants. This implies that both metabolic network perturbations and a Comamonas DA1877 diet converge onto the same regulatory network to elicit converse effects on the two modules, perhaps via two opposing regulatory signals (Figure 7). NHR-10 directly binds the acdh-1 promoter (Arda et al., 2010) and mediates the induction of acdh-1 in response to metabolic network perturbations in several mutants, particularly those that function within BCAA breakdown. In addition, NHR-68 and several other NHRs either function together or in parallel with NHR-10. Epistasis experiments with additional NHRs, as well as other TFs, will further illuminate their function in metabolic gene regulation. NHRs are ligand-regulated TFs that may be able to directly sense metabolite accumulation and regulate genes accordingly. For instance, NHR ligands may be produced as a result of specic metabolic network perturbations. Alternatively, NHRs may be inactivated by a metabolite, and depletion
of that metabolite upon network perturbation might result in NHR activation. Perturbation of different metabolic pathways may result in the accumulation or depletion of metabolites that may be sensed by different NHRs with overlapping downstream targets. Previously, we found functional modularity of NHRs in a C. elegans gene regulatory network (Arda et al., 2010). Here, we nd a modular organization at the level of metabolic gene expression. Together, these ndings support a model in which gene regulatory network modules facilitate responses to physiological and environmental cues. Most dietary sensor repressors encode enzymes involved in inborn disorders of amino acid metabolism in humans. These diseases are usually treated by dietary interventions that are designed to (1) avoid buildup of toxic metabolites, (2) supplement patients with depleted metabolites, and/or (3) supplement cofactors that may improve residual enzyme activity (Figure 5) (Acosta, 2010). In C. elegans, mutations in genes from different pathways within the metabolic network (BCAA and methionine metabolism) can cause highly similar changes in gene expression. Thus, it may be that a dietary regimen used to mitigate one type of metabolic disease may also be benecial for another. C. elegans presents an attractive model with which transcriptomic changes in response to different diets and metabolic network perturbations can be compared and potentially can be used in screens for drugs or other small molecules that affect these changes. Our study establishes C. elegans as a powerful model to dissect the mechanisms of dietary responses, inborn metabolic diseases, and the connections between them. It is likely that other parts of the metabolic network respond to different cues. We provide a framework to identify candidate sensors to these cues and to combine these sensors with genetic screens and epistasis to dissect the mechanisms involved.
EXPERIMENTAL PROCEDURES Strains C. elegans strains were grown as described (Brenner, 1974). N2 (Bristol) was used as the wild-type strain. The following mutant strains were provided by the C. elegans Gene Knockout Consortium (CGC): nhr-68(gk708), sams1(ok2946), cbl-1(ok2954), mce-1(ok243), pcca-1(ok2282), metr-1(ok521), ZK669.4(ok3001), sams-5(gk147), and hlh-11(ok2944). The nhr-10(tm4695) mutant strain was provided by the National Bioresource Project, Japan. The Pacdh-1::GFP strain (VL749) is described in MacNeil et al. (2013). VL405 [Pmir-63::GFP + unc-119(+)] is described in Martinez et al. (2008). BC12350 dpy-5(e907); sIs12119 [rCesC53A3.2::GFP + pCeh361] is described in McKay et al. (2003). CL2166 dvIs19[pAF15(Pgst-4::GFP::NLS)] was obtained from the CGC (Tawe et al., 1998). EMS Screen VL749 animals were treated with 0.5% ethyl methyl sulfoxide (EMS) for 4 hr. Mutagenized animals were transferred to nematode growth media (NGM) plates seeded with Comamonas DA1877 and allowed to lay eggs. Eggs were collected, and 100 10 cm Comamonas DA1877-NGM plates were seeded with 50 F1 animals each. F2 animals were screened for visible GFP. A single GFP-positive F2 animal was picked from each F1 plate and transferred to individual plates. Of the 100 mutants, 45 were fertile and produced GFPpositive offspring when fed Comamonas DA1877. Mutant Mapping Chromosomal assignment was done by crossing mutants carrying the Pacdh1::GFP reporter into the CB4856 (Hawaiian) strain followed by SNP mapping of
GFP-positive pools of F2 animals as described (Wicks et al., 2001). Mutants that produced GFP-positive F1 males were assigned to linkage group (LG) X. Primers used for chromosomal assignment are as described (Davis et al., 2005). Following chromosomal assignment, ne mapping was undertaken for ve mutants using chromosomal primers as described (Davis et al., 2005; Wicks et al., 2001). We created him-5(e1467) lines carrying the Pacdh1::GFP transgene and the mutation of interest to perform complementation tests. These him-5 lines were used to generate males that were crossed into mutant hermaphrodites carrying Pacdh-1::GFP, and GFP expression was assessed in F1 animals grown on Comamonas DA1877. Whole-Genome Sequencing Before sequencing, mutants were outcrossed three times to N2 wild-type animals. Total genomic DNA was prepared for mutants ww2, ww4, ww10, ww11, and ww17 as described (Sarin et al., 2010). Libraries were made and barcoded using the NextFlex barcoding system and library construction kits (Bioo Scientic). Samples were sequenced by the IIGB Genomic Core facility, UC Riverside, using Illuminas HiSeq2000 platform (51 bp reads). After ltering out low-quality reads using the default Illumina pipeline quality lter, we recovered 184.8 million reads representing an 183 average coverage. Reads were mapped onto the C. elegans genome sequence (WS201), and variants were identied using MAQgene with default parameter setting except that minimum fraction of non-wild-type reads was set to 0.8, minimum span of uncovered bases to report as uncovered region was set to 1, and maximum sum of error qualities was set to 150 (Bigelow et al., 2009). Mismatches were compared between all ve mutants. Base pair changes common to all mutants were ignored, as these were likely present in our starting strain. Variants were validated by PCR and sequencing. Additional alleles were identied by PCR amplication, followed by sequencing of targeted genes. RNAi Screen RNAi screening was performed on 96-well plates using NGM agar containing 5 mM IPTG. E. coli HT115 RNAi cultures were grown in LB containing 100 mg/ml ampicillin to log phase in 96-well deep well dishes. Bacteria were centrifuged and resuspended in 1/20 volume of M9. 10 ml of the resuspended cultures was added to the RNAi screening plates. For the Comamonas DA1877 condition screen, Comamonas DA1877 was grown overnight to saturation in Luria-Bertani (LB), then sonicated and diluted 1/10 with LB containing 100 mg/ml ampicillin. 10 ml drops were added to each well of the E. coli HT115-RNAi seeded plates. VL749 animals were synchronized by hypochlorite bleaching and washed in M9 media, and 25 eggs were added to each well. After 60 hr, animals reached the early adult stage and were visually screened for GFP expression. RNAi clones found to affect GFP in the rst pass of the screen (856 clones) were rearrayed and retested four times. 554 genes retested at least two out of four times and were rearrayed again. These clones were tested three times for specicity using strain VL405, which contains an integrated Pmir-63::GFP construct (Martinez et al., 2008). 192 genes did not affect Pmir-63::GFP in any of the three tests and were considered specic for Pacdh-1::GFP. All nal hits were sequence veried. Mutant Validation nhr-68(gk708), sams-1(ok2946), cbl-1(ok2954), mce-1(ok243), pcca1(ok2282), metr-1 (ok521), ZK669.4 (ok3001), sams-5 (gk147), nhr-10 (tm4695), and hlh-11 (ok2944) mutants were outcrossed three times to the wild-type N2 strain and homozygosed with the exception of the ok3001 allele, a deletion in the gene ZK669.4, which produced sterile homozygotes. Pictures of sterile ZK669.4(ok3001) homozygotes from heterozygous parents were taken, and animals were retrospectively genotyped. RNA was isolated from the hlh-11 and sams-5 deletion mutant strains grown on E. coli OP50, E. coli HT115, or Comamonas DA1877 for qRT-PCR analysis. All outcrossed mutant strains were also crossed to VL749 and homozygosed. qRT-PCR Animals were synchronized and grown on E. coli OP50, E. coli HT115, or Comamonas DA1837, and 1,000 adult animals were harvested for each condition. Animals were thoroughly washed in M9 buffer, and total RNA was isolated using Trizol (Invitrogen), followed by DNaseI treatment
and cleanup using QIAGEN RNeasy columns. Complementary DNA (cDNA) was prepared from 1 mg of RNA using oligo-dT and Mu-MLV enzyme (NEB). Primer sequences for qRT-PCR were generated using the GETprime database (Gubelmann et al., 2011) and are listed in Table S5. qRT-PCR was performed in triplicate using the Applied Biosystems StepOnePlus Real-Time PCR System and Fast Sybr Green Master Mix (Invitrogen). Relative transcript abundance was determined using the DDCt method and normalized to averaged ama-1 and act-1 mRNA expression levels (Livak and Schmittgen, 2001). nCounter Analysis RNA was extracted as described above. nCounter analysis was performed as per manufacturers instructions (NanoString Technologies) using 300 ng of total RNA, and expression between samples was normalized to ama-1 mRNA levels. Probes used are listed in Table S5. Functional Annotation GO analyses were performed using David (Huang et al., 2009). WormNet v.2 was used as described (Lee et al., 2008). Genes found to affect Pacdh1::GFP in the screens were subjected to functional categorization based on manual curation. Gene descriptions in WormBase were used to assign genes to categories when available, or the described function for the closest human homolog was used. If no functional annotation could be assigned to a gene, it was placed in the unknown category. If a genes described function was not within one of the major categories, it was placed in the other category. Tissue expression was assigned based on WormBase WS232 annotations and was categorized into intestinal, nonintestinal, and unknown groups. If the subset of tissues in which a gene is expressed included the intestine, it was placed in the intestinal group. If the tissue expression for a gene was described but did not include the intestine, it was placed in the nonintestinal group. If no expression patterns were described, it was placed in the unknown group. Metabolic Map Annotations for C. elegans BCAA breakdown, methionine metabolism, and the TCA pathway were obtained from the KEGG database (Kanehisa et al., 2010) and from WormBase. Microarray Expression Proling Mutant animals were grown on Comamonas DA1877 on standard NGM plates for two generations prior to egg collection. Eggs were collected and synchronized in L1 stage and grown until they reached the gravid adult stage. Animals were washed twice in M9 buffer, pelleted by centrifugation, and frozen at 80 C in Trizol. RNA was collected using Trizol extraction followed by DNase I treatment and cleanup using RNeasy prep kit (QIAGEN). Three biological replicates were prepared for each condition. Microarray expression proling was performed by the Genomics Core facility at University of Massachusetts Medical School using C. elegans genome arrays (Affymetrix). The RMA method in the Affy package from Bioconductor was used in R to summarize the probe level data and to normalize the data set to remove across-array variation. Log-transformed data were used in subsequent analyses. Moderated T statistics in Limma (Smyth, 2004) was used to calculate signicance. Signicance was determined using an adjusted p value (Benjamini and Hochberg, 1995). Changes in gene expression that were 2-fold (p < 0.001) or greater were considered signicantly changed. Starvation Experiments Animals were rst grown to L4 on E. coli OP50 bacteria. L4 animals were washed ve times in M9 and then transferred to unseeded peptone-free NGM plates. After 24 hr, animals were collected and examined for GFP expression. ACCESSION NUMBERS The GEO accession number for the microarray data reported in this paper is GSE43952.
SUPPLEMENTAL INFORMATION Supplemental Information includes ve gures and ve tables and can be found with this article online at http://dx.doi.org/10.1016/j.cell.2013.02.050. ACKNOWLEDGMENTS We thank members of the Walhout laboratory, Amy Walker, and Victor Ambros for discussions and critical reading of the manuscript. We thank Jean Francois Rual and Marc Vidal (CCSB at the Dana-Farber Cancer Institute) for a copy of the ORFeome RNAi library and L. Safak Yilmaz for help with the creation of Figure 3. This work was supported by the National Institutes of Health (NIH) grant DK068429 to A.J.M.W. L.T.M. was supported by a CIHR postdoctoral fellowship. E.W. was supported by an AHA Founders Afliate Predoctoral Fellowship (11PRE6590003). Some nematode strains used in this work were provided by the CGC, which is funded by the NIH National Center for Research Resources (NCRR). Received: August 16, 2012 Revised: December 15, 2012 Accepted: February 5, 2013 Published: March 28, 2013 REFERENCES Acosta, P.B. (2010). Nutrition Management of Patients with Inherited Metabolic Disorders (Sudbury, MA: Jones and Bartlett Publishers). Arda, H.E., Taubert, S., Conine, C., Tsuda, B., Van Gilst, M.R., Sequerra, R., Doucette-Stam, L., Yamamoto, K.R., and Walhout, A.J.M. (2010). Functional modularity of nuclear hormone receptors in a Caenorhabditis elegans gene regulatory network. Mol. Syst. Biol. 6, 367. Ashra, K., Chang, F.Y., Watts, J.L., Fraser, A.G., Kamath, R.S., Ahringer, J., and Ruvkun, G. (2003). Genome-wide RNAi analysis of Caenorhabditis elegans fat regulatory genes. Nature 421, 268272. Benjamini, Y., and Hochberg, Y. (1995). Controlling the false discovery rate: a practical and powerful approach to multiple testing. J. R. Stat. Soc., B 57, 289300. Bigelow, H., Doitsidou, M., Sarin, S., and Hobert, O. (2009). MAQGene: software to facilitate C. elegans mutant genome sequence analysis. Nat. Methods 6, 549. Brenner, S. (1974). The genetics of Caenorhabditis elegans. Genetics 77, 7194. Carsten, L.D., Watts, T., and Markow, T.A. (2005). Gene expression patterns accompanying a dietary shift in Drosophila melanogaster. Mol. Ecol. 14, 32033208. Chale, M., Tu, Y., Euskirchen, G., Ward, W.W., and Prasher, D.C. (1994). Green uorescent protein as a marker for gene expression. Science 263, 802805. Davis, M.W., Hammarlund, M., Harrach, T., Hullett, P., Olsen, S., and Jorgensen, E.M. (2005). Rapid single nucleotide polymorphism mapping in C. elegans. BMC Genomics 6, 118. Du, D., Shi, Y.H., and Le, G.W. (2010). Microarray analysis of high-glucose diet-induced changes in mRNA expression in jejunums of C57BL/6J mice reveals impairment in digestion, absorption. Mol. Biol. Rep. 37, 18671874. Grove, C.A., De Masi, F., Barrasa, M.I., Newburger, D.E., Alkema, M.J., Bulyk, M.L., and Walhout, A.J. (2009). A multiparameter network reveals extensive divergence between C. elegans bHLH transcription factors. Cell 138, 314327. ning, N.-M., Lehrach, H., and Ralser, M. (2010). Regulatory crosstalk of the Gru metabolic network. Trends Biochem. Sci. 35, 220227. Gubelmann, C., Gattiker, A., Massouras, A., Hens, K., David, F., Decouttere, F., Rougemont, J., and Deplancke, B. (2011). GETPrime: a gene- or transcript-specic primer database for quantitative real-time PCR. Database (Oxford) 2011, bar040.
Hinnebusch, A.G. (2005). Translational regulation of GCN4 and the general amino acid control of yeast. Annu. Rev. Microbiol. 59, 407450. Huang, W., Sherman, B.T., and Lempicki, R.A. (2009). Systematic and integrative analysis of large gene lists using DAVID bioinformatics resources. Nat. Protoc. 4, 4457. Hunt-Newbury, R., Viveiros, R., Johnsen, R., Mah, A., Anastas, D., Fang, L., Halfnight, E., Lee, D., Lin, J., Lorch, A., et al. (2007). High-throughput in vivo analysis of gene expression in Caenorhabditis elegans. PLoS Biol. 5, e237. Kamath, R.S., Fraser, A.G., Dong, Y., Poulin, G., Durbin, R., Gotta, M., Kanapin, A., Le Bot, N., Moreno, S., Sohrmann, M., et al. (2003). Systematic functional analysis of the Caenorhabditis elegans genome using RNAi. Nature 421, 231237. Kanehisa, M., Goto, S., Furumichi, M., Tanabe, M., and Hirakawa, M. (2010). KEGG for representation and analysis of molecular networks involving diseases and drugs. Nucleic Acids Res. 38(Database issue), D355D360. Laplante, M., and Sabatini, D.M. (2012). mTOR signaling in growth control and disease. Cell 149, 274293. Lee, I., Lehner, B., Crombie, C., Wong, W., Fraser, A.G., and Marcotte, E.M. (2008). A single gene network accurately predicts phenotypic effects of gene perturbation in Caenorhabditis elegans. Nat. Genet. 40, 181188. Livak, K.J., and Schmittgen, T.D. (2001). Analysis of relative gene expression data using real-time quantitative PCR and the 2(-Delta Delta C(T)) Method. Methods 25, 402408. MacNeil, L.T., Watson, E., Arda, H.E., Zhu, L.J., and Walhout, A.J.M. (2013). Diet-induced developmental acceleration independent of TOR and insulin in C. elegans. Cell 153, this issue, 240252. Martinez, N.J., Ow, M.C., Reece-Hoyes, J.S., Barrasa, M.I., Ambros, V.R., and Walhout, A.J. (2008). Genome-scale spatiotemporal analysis of Caenorhabditis elegans microRNA promoter activity. Genome Res. 18, 20052015. McKay, S.J., Johnsen, R., Khattra, J., Asano, J., Baillie, D.L., Chan, S., Dube, N., Fang, L., Goszczynski, B., Ha, E., et al. (2003). Gene expression proling of cells, tissues, and developmental stages of the nematode C. elegans. Cold Spring Harb. Symp. Quant. Biol. 68, 159169. Murphy, C.T., McCarroll, S.A., Bargmann, C.I., Fraser, A., Kamath, R.S., Ahringer, J., Li, H., and Kenyon, C. (2003). Genes that act downstream of DAF-16 to inuence the lifespan of Caenorhabditis elegans. Nature 424, 277283. Reece-Hoyes, J.S., Barutcu, A.R., McCord, R.P., Jeong, J.S., Jiang, L., MacWilliams, A., Yang, X., Salehi-Ashtiani, K., Hill, D.E., Blackshaw, S., et al. (2011a). Yeast one-hybrid assays for gene-centered human gene regulatory network mapping. Nat. Methods 8, 10501052. Reece-Hoyes, J.S., Diallo, A., Lajoie, B., Kent, A., Shrestha, S., Kadreppa, S., Pesyna, C., Dekker, J., Myers, C.L., and Walhout, A.J.M. (2011b). Enhanced
yeast one-hybrid assays for high-throughput gene-centered regulatory network mapping. Nat. Methods 8, 10591064. Rual, J.-F., Ceron, J., Koreth, J., Hao, T., Nicot, A.-S., Hirozane-Kishikawa, T., Vandenhaute, J., Orkin, S.H., Hill, D.E., van den Heuvel, S., and Vidal, M. (2004). Toward improving Caenorhabditis elegans phenome mapping with an ORFeome-based RNAi library. Genome Res. 14(10B), 21622168. Sarin, S., Bertrand, V., Bigelow, H., Boyanov, A., Doitsidou, M., Poole, R.J., Narula, S., and Hobert, O. (2010). Analysis of multiple ethyl methanesulfonate-mutagenized Caenorhabditis elegans strains by whole-genome sequencing. Genetics 185, 417430. Saudubray, J.-M., Sedel, F., and Walter, J.H. (2006). Clinical approach to treatable inborn metabolic diseases: an introduction. J. Inherit. Metab. Dis. 29, 261274. Smyth, G.K. (2004). Linear models and empirical bayes methods for assessing differential expression in microarray experiments. Stat. Appl. in Genet. Mol. Biol. 3, Article3. nnichsen, B., Koski, L.B., Walsh, A., Marschall, P., Neumann, B., Brehm, M., So Alleaume, A.M., Artelt, J., Bettencourt, P., Cassin, E., et al. (2005). Full-genome RNAi proling of early embryogenesis in Caenorhabditis elegans. Nature 434, 462469. Taubert, S., Ward, J.D., and Yamamoto, K.R. (2011). Nuclear hormone receptors in nematodes: evolution and function. Mol. Cell. Endocrinol. 334, 4955. hrsen, K. (1998). Tawe, W.N., Eschbach, M.L., Walter, R.D., and Henkle-Du Identication of stress-responsive genes in Caenorhabditis elegans using RT-PCR differential display. Nucleic Acids Res. 26, 16211627. Timmons, L., Court, D.L., and Fire, A. (2001). Ingestion of bacterially expressed dsRNAs can produce specic and potent genetic interference in Caenorhabditis elegans. Gene 263, 103112. Van Gilst, M.R., Hadjivassiliou, H., and Yamamoto, K.R. (2005). A Caenorhabditis elegans nutrient response system partially dependent on nuclear receptor NHR-49. Proc. Natl. Acad. Sci. USA 102, 1349613501. Walker, A.K., Jacobs, R.L., Watts, J.L., Rottiers, V., Jiang, K., Finnegan, D.M., Shioda, T., Hansen, M., Yang, F., Niebergall, L.J., et al. (2011). A conserved SREBP-1/phosphatidylcholine feedback circuit regulates lipogenesis in metazoans. Cell 147, 840852. Weisfeld-Adams, J.D., Morrissey, M.A., Kirmse, B.M., Salveson, B.R., Wasserstein, M.P., McGuire, P.J., Sunny, S., Cohen-Pfeffer, J.L., Yu, C., Caggana, M., and Diaz, G.A. (2010). Newborn screening and early biochemical follow-up in combined methylmalonic aciduria and homocystinuria, cblC type, and utility of methionine as a secondary screening analyte. Mol. Genet. Metab. 99, 116123. Wicks, S.R., Yeh, R.T., Gish, W.R., Waterston, R.H., and Plasterk, R.H. (2001). Rapid gene mapping in Caenorhabditis elegans using a high density polymorphism map. Nat. Genet. 28, 160164.
Erratum
b-Catenin-Driven Cancers Require a YAP1 Transcriptional Complex for Survival and Tumorigenesis
Joseph Rosenbluh, Deepak Nijhawan, Andrew G. Cox, Xingnan Li, James T. Neal, Eric J. Schafer, Travis I. Zack, Xiaoxing Wang, Aviad Tsherniak, Anna C. Schinzel, Diane D. Shao, Steven E. Schumacher, Barbara A. Weir, Francisca Vazquez, Glenn S. Cowley, David E. Root, Jill P. Mesirov, Rameen Beroukhim, Calvin J. Kuo, Wolfram Goessling, and William C. Hahn*
*Correspondence: william_hahn@dfci.harvard.edu http://dx.doi.org/10.1016/j.cell.2013.03.007
(Cell 151, 14571473; December 13, 2012) During the preparation of gures for the above article, we inadvertently inserted incorrect images in Figures 1E, 2E, and 6C. The correct and incorrect images were all derived from the same experiments and are quite similar in appearance. We mislabeled them when we retrieved them for gure preparation. In Figure 1E, we inserted an incorrect image in the loading control (actin) for SW480. Similarly, for Figure 2E, we inserted an incorrect image for the loading controls (actin) for the + YAP1 samples. In Figure 6C, we inserted an incorrect panel for immunoblot for YAP1 in the + YES1 (T348I) samples. These errors do not affect the results or interpretation of any of these experiments. The corrected images have been inserted into the gures below, and the gures have been corrected online. We regret these errors and apologize for the inconvenience that they may have caused.
Figure 1. Identication of Genes Essential for b-Catenin-Active Cell Lines
Figure 2. YAP1 Is Essential for Tumorigenicity of b-Catenin-Dependent Cancer Cell Lines
Figure 6. Expression of YES1 Is Essential for Formation of the YAP1-b-Catenin-TBX5 Complex
272
Ca
+
SnapShot: Inammasomes
CASR
ATP
Maninjay K. Atianand,1 Vijay A. Rathinam,1 and Katherine A. Fitzgerald1 Division of Infectious Diseases and Immunology, University of Massachusetts Medical School, Worcester, MA 01605, USA
Basic inammasome structure

P2X7 receptor Gram-negative Bacteria
Signal 2
PLC
Potassium efflux Lysosomal destabilization GBP5 Bacterial mRNA
TLR4
Bacterium containing type III secretion system Salmonella typhimurium Pseuodomonas aeruginosa
AC
Signal 3
Flagellin
RECEPTOR
NLR or ALR
Inhibition
ADAPTOR
ASC
PrgJ
Inammasome complex
InsP3 of cAMP
TRIF
Ca
+
Cell 153, March 28, 2013 2013 Elsevier Inc.

ROS
Procaspase-1
EFFECTOR
Active caspase-1
ROS ER stress
Ca
+
NLRP3
Type I interferons Oxidized mitochondrial DNA ASC
Mitochondrion
NAIPs5/6 NAIP2
PKC-
pro-IL-1b pro-IL-18
IL-1b IL-18
NLRC4/IPAF
P P
Signal 1
H+
M2 channel Endoplasmic reticulum (ER) Trans-golgi
Procaspase-1
H+
Active caspase -11

Procaspase-1
NUCLEUS
Cell death IL-1a & HMGB1 release
DOI http://dx.doi.org/10.1016/j.cell.2013.03.009 Active caspase-1

Cell death IL-1a & HMGB1 release pro-IL-1b pro-IL-18
PA
IL-1b IL-18
Cytosolic bacteria Francisella tularensis Listeria monocyotgenes
Anthrax lethal toxin

Yersinia pestis
Intestinal inammation and homeostasis

Microbes ?
Diacylated lipopeptides
AIM2
ASC
LF
NLRP1b
NLRP6
NLRP12
NLRP7
Procaspase-1
(In human monocytes & macrophages)
DNA viruses mCMV Vaccinia Herpes viruses
IFI16
KSHV DNA
NUCLEUS
ASC
See online version for legend and references.
Procaspase-1
SnapShot: Inammasomes
1
Maninjay K. Atianand,1 Vijay A. Rathinam,1 and Katherine A. Fitzgerald1 Division of Infectious Diseases and Immunology, University of Massachusetts Medical School, Worcester, MA 01605, USA
Large multiprotein complexes referred to as inflammasomes form in the cytosol during infection or in response to endogenous danger signals released from damaged or dying cells and regulate the processing and secretion of IL-1 and IL-18 as well as an inflammatory form of cell death called pyroptosis. Inflammasomes have emerged as central mediators of the host response to pathogens and are also important contributors to sterile inflammatory diseases. Inammasome Assembly and Functions The proinflammatory cytokines IL-1 and IL-18 are secreted by many cell types and play crucial roles in both the innate and adaptive arms of host immunity. In contrast to the majority of other proinflammatory cytokines such as TNF, IL-1 and IL-18 are regulated at both the transcriptional and posttranslational levels. Upon transcriptional induction by PRRs (signal 1), IL-1 and IL-18 are synthesized as biologically inactive proteins, which are subsequently processed by the cysteine protease caspase-1 (ICE). Conversion of procaspase-1 itself into an enzymatically active form, caspase-1, is mediated by the inflammasome (signal 2). There are distinct types of inflammasomes, differentiated by their protein constituents, activators, and effectors. A basic inflammasome complex consists of a cytosolic sensor (which can be either an NLR or a member of the ALR family), the adaptor ASC, and the effector molecule procaspase-1. The inflammasome adaptor ASC has a bipartite structure containing an N-terminal PYD and the C-terminal CARD, and therefore, it acts as a bridge between the sensor (NLRs or ALRs) and the effector procaspase-1 by utilizing homotypic PYD:PYD and CARD:CARD interactions, respectively. In addition to IL-1 and IL-18 processing, active caspase-1 also regulates a form of inflammatory cell death referred to as pyroptosis and leads to unconventional protein secretion. Several inflammasome complexes, which are illustrated here, have been identified in recent years. The NLR-Containing Inammasomes The NLRP3 inflammasome is the best-studied inflammasome, and its activation is triggered by bacterial, viral, parasitic, and fungal infections as well as by endogenous danger signals such as ATP and uric acid crystals. Although much progress has been made in discovering activators for the NLRP3 inflammasome, what has not been clear is how the assembly of this large multiprotein structure is controlled. Several NLRP3-inflammasome-activating mechanisms have been linked to NLRP3 inflammasome complex assembly and include K+ efflux, lysosomal destabilization, generation of ROS, involvement of Gbp5, presence of cytosolic bacterial RNA during infection, and the release of oxidized mitochondrial DNA. More recently, the murine CASR has also been shown to act upstream of the Nlrp3 inflammasome. The CASR acts via PLC that catalyzes the generation of InsP3, which leads to the release of Ca 2+ from the ER leading to the assembly of the Nlrp3 inflammasome. Given the diverse nature of these mechanisms, it is likely that the NLRP3 inflammasome is an indirect sensor of a common cellular event associated with these proposed mechanisms. Recently, a TLR4-TRIF-type I IFN-dependent activation of caspase-11 pathway (signal 3) has been implicated upstream of caspase-1 activation for the NLRP3-inflammasome-mediated IL-1 secretion in response to infection with Gramnegative bacteria, adding yet another layer of complexity to the mechanisms underlying NLRP3-inflammasome-dependent effector functions. The NLRC4 (IPAF) inflammasome is another well-studied inflammasome that assembles via interactions with specific NAIPs in response to the type III secretion apparatus of S. typhimurium. Notably, NLRC4 contains an N-terminal CARD (and lacks the PYD), and therefore, it does not require ASC for inflammasome assembly. The phosphorylation of NLRC4 by protein kinase C- is also essential for NLRC4 activation. The NLRP1b and NLRP12 inflammasomes sense anthrax LT and Y. pestis infection, respectively. The NLRP6 inflammasome plays a role in regulating intestinal homeostasis, and the NLRP7 inflammasome has been shown to recognize diacylated lipopeptides in human macrophages. The AIM2 and IFI16 Inammasome Several groups independently identified AIM2 as a cytosolic receptor for DNA. AIM2 is a member of a larger HIN200- and Pyrin-domain-containing protein family (PYHIN or ALR family). AIM2 binds DNA of self and nonself origin in a sequence-independent manner via its C-terminal HIN200 domain. Notably, the AIM2 inflammasome is the first inflammasome where a direct receptor:ligand interaction has been formally demonstrated. Subsequent studies in AIM2-deficient mice have further demonstrated an essential role for the AIM2 inflammasome in protective immunity to the intracellular bacterium F. tularensis or following infection with mouse cytomegaovirus. Recently, the related ALR protein, IFI16, has also been linked to inflammasome activation. The IFI16 inflammasome has been implicated in detection of both KSHV and herpes simplex virus infection, where IFI16 is thought to recognize the herpesviral DNA in the nucleus. Abbreviations AIM2, absent in melanoma 2; ALR, AIM2-like receptor; ASC, apoptotic speck-like protein containing CARD; ATP, adenosine triphosphate; CARD, caspase-1 activation and recruitment domain; CASR, calcium-sensing receptor; ER, endoplasmic reticulum; GBP5, guanylate-binding protein 5; HMGB1, high mobility group B1; ICE, IL-1 -converting enzyme; IFI16, interferon inducible protein 16; IL-1, interleukin 1; InsP3, inositol-1,3,5 triphosphate; KSHV, Kaposi-sarcoma-associated herpes virus; LF, lethal factor; LT, lethal toxin; MAMP, microbe-associated molecular pattern; NLR, NOD-like receptors; PA, protective antigen; PLC, phospholipase C; PRR, pattern recognition receptor; PYD, pyrin domain; PYHIN, pyrin and HIN200 domain containing; ROS, reactive oxygen species; TLR, Toll-like receptor; TRIF, Toll/interleukin-1 receptor (TIR)-domain-containing adaptor inducing IFN-. ACKNOWLEDGMENTS We apologize to the colleagues whose work could not be cited here because of space limitations. REFErENCES
Elinav, E., Strowig, T., Kau, A.L., Henao-Mejia, J., Thaiss, C.A., Booth, C.J., Peaper, D.R., Bertin, J., Eisenbarth, S.C., Gordon, J.I., and Flavell, R.A. (2011). NLRP6 inflammasome regulates colonic microbial ecology and risk for colitis. Cell 145, 745757. Fernandes-Alnemri, T., Yu, J.W., Juliana, C., Solorzano, L., Kang, S., Wu, J., Datta, P., McCormick, M., Huang, L., McDermott, E., et al. (2010). The AIM2 inflammasome is critical for innate immunity to Francisella tularensis. Nat. Immunol. 11, 385393. Kerur, N., Veettil, M.V., Sharma-Walia, N., Bottero, V., Sadagopan, S., Otageri, P., and Chandran, B. (2011). IFI16 acts as a nuclear pathogen sensor to induce the inflammasome in response to Kaposi Sarcoma-associated herpesvirus infection. Cell Host Microbe 9, 363375. Khare, S., Dorfleutner, A., Bryan, N.B., Yun, C., Radian, A.D., de Almeida, L., Rojanasakul, Y., and Stehlik, C. (2012). An NLRP7-containing inflammasome mediates recognition of microbial lipopeptides in human macrophages. Immunity 36, 464476. Kofoed, E.M., and Vance, R.E. (2011). Innate immune recognition of bacterial ligands by NAIPs determines inflammasome specificity. Nature 477, 592595. Qu, Y., Misaghi, S., Izrael-Tomasevic, A., Newton, K., Gilmour, L.L., Lamkanfi, M., Louie, S., Kayagaki, N., Liu, J., Kmves, L., et al. (2012). Phosphorylation of NLRC4 is critical for inflammasome activation. Nature 490, 539542. Rathinam, V.A., Jiang, Z., Waggoner, S.N., Sharma, S., Cole, L.E., Waggoner, L., Vanaja, S.K., Monks, B.G., Ganesan, S., Latz, E., et al. (2010). The AIM2 inflammasome is essential for host defense against cytosolic bacteria and DNA viruses. Nat. Immunol. 11, 395402. Rathinam, V.A., Vanaja, S.K., Waggoner, L., Sokolovska, A., Becker, C., Stuart, L.M., Leong, J.M., and Fitzgerald, K.A. (2012). TRIF licenses caspase-11-dependent NLRP3 inflammasome activation by gram-negative bacteria. Cell 150, 606619. Shenoy, A.R., Wellington, D.A., Kumar, P., Kassa, H., Booth, C.J., Cresswell, P., and MacMicking, J.D. (2012). GBP5 promotes NLRP3 inflammasome assembly and immunity in mammals. Science 336, 481485. Vladimer, G.I., Weng, D., Paquette, S.W., Vanaja, S.K., Rathinam, V.A., Aune, M.H., Conlon, J.E., Burbage, J.J., Proulx, M.K., Liu, Q., et al. (2012). The NLRP12 inflammasome recognizes Y. pestis. Immunity 37, 96107.
272.e1Cell 153, March 28, 2013 2013 Elsevier Inc. DOI http://dx.doi.org/10.1016/j.cell.2013.03.009

Cell - 28 March 2013

Încărcat de

Informații document

Titlu original

Drepturi de autor

Formate disponibile

Partajați acest document

Partajați sau inserați document

Opțiuni de partajare

Vi se pare util acest document?

Este necorespunzător acest conținut?

Drepturi de autor:

Formate disponibile

Cell - 28 March 2013

Încărcat de

Drepturi de autor:

Formate disponibile

Leading Edge

Neurotransmitter Signaling Drives Cancer Progression

Lipid Protector Battles Inuenza

Antibody Mutations that Are Worth the Wait

Achilles Heel for Glioblastoma in GSC-Derived Vasculature

Y an RNA Reshapes Enzyme Function

Mitosis Reshufes the DNA Deck at the Nuclear Lamina

Prions Foster Budding Friendships

Horses for Courses in DNA Methylation

Speedy Ligase ExchangeYes, We Cand1!

Golgi Lipids Drive Hypertrophy

Life-Extending Drug Acts through Microbes

Lifestyle Choice Affected by Diet

Dietary Response Networks

Where Do Phenotypes eXIST?

Transcription versus Transcript

Lncing Yin and Yang

Cell 153, March 28, 2013 2013 Elsevier Inc. 7

Pier Paolo Pandol

Cell 153, March 28, 2013 2013 Elsevier Inc. 9

lncRNAs Complicate Cancer

Assessing RNA Therapies

Revealing ncRNA Disease Roles

10 Cell 153, March 28, 2013 2013 Elsevier Inc.

Ring around the Ro-sie: RNA-Mediated Alterations of PNPase Activity

12 Cell 153, March 28, 2013 2013 Elsevier Inc.

Ring around the Ro-sie: RNA-Mediated Alterations of PNPase Activity

12 Cell 153, March 28, 2013 2013 Elsevier Inc.

Cullins Getting Undressed by the Protein Exchange Factor Cand1

14 Cell 153, March 28, 2013 2013 Elsevier Inc.

Cullins Getting Undressed by the Protein Exchange Factor Cand1

14 Cell 153, March 28, 2013 2013 Elsevier Inc.

Figure 1. Substrate Receptor Exchange Mediated by Cand1

Cell 153, March 28, 2013 2013 Elsevier Inc. 15

16 Cell 153, March 28, 2013 2013 Elsevier Inc.

Figure 1. Somatic IDH1/2 Mutations Produce the Oncometabolite 2HG

Figure 3. Cancer-Associated Mutations in the RNA-Splicing Machinery

Figure 5. Genetic Alterations Disrupt Multiple Cellular Processes

Cell 153, March 28, 2013 2013 Elsevier Inc. 31

32 Cell 153, March 28, 2013 2013 Elsevier Inc.

Cell 153, March 28, 2013 2013 Elsevier Inc. 33

34 Cell 153, March 28, 2013 2013 Elsevier Inc.

Cell 153, March 28, 2013 2013 Elsevier Inc. 35

36 Cell 153, March 28, 2013 2013 Elsevier Inc.

Cell 153, March 28, 2013 2013 Elsevier Inc. 37

Figure 1. Representative Epigenetic States

Figure 2. Histone H3 Lysine Writers, Erasers, and Readers

Figure 3. Genetic Alterations in Epigenetic Regulators

(legend continued on next page)

44 Cell 153, March 28, 2013 2013 Elsevier Inc.

Cell 153, March 28, 2013 2013 Elsevier Inc. 45

Figure 4. Genetic Disruption of Epigenetic Control at H3K27 in Cancer

Figure 5. Interplay between the Cancer Genome and Epigenome

50 Cell 153, March 28, 2013 2013 Elsevier Inc.

Cell 153, March 28, 2013 2013 Elsevier Inc. 51

52 Cell 153, March 28, 2013 2013 Elsevier Inc.

Cell 153, March 28, 2013 2013 Elsevier Inc. 53

54 Cell 153, March 28, 2013 2013 Elsevier Inc.

Cell 153, March 28, 2013 2013 Elsevier Inc. 55

Figure 1. Metabolism and Acetylation/Deacetylation

Figure 3. Metabolism and Methyltransferases

Figure 4. TDH-Mediated Catabolism of Threonine

Figure 5. Histone Demethylase Reactions