Sunteți pe pagina 1din 147

Cancer Cell

Previews
Methylome Alterations Mark New Therapeutic Opportunities in Glioblastoma
Eric H. Raabe1,2,* and Charles G. Eberhart2
of Pediatric Oncology of Pathology Johns Hopkins University School of Medicine, Baltimore, MD 21287, USA *Correspondence: eraabe2@jhmi.edu http://dx.doi.org/10.1016/j.ccr.2012.10.001
2Division 1Division

In this issue of Cancer Cell, Sturm et al. report that global DNA methylation patterns in glioblastoma multiforme divide adult and pediatric tumors into subgroups that have characteristic DNA mutations, mRNA proles, and most importantly, different clinical behaviors. These ndings suggest novel opportunities for therapeutics for this dreaded disease.
Glioblastoma multiforme (GBM) is the most aggressive brain tumor and is associated with very poor overall survival. GBM occurs in adults much more frequently than in children or adolescents, and pediatric GBM has genetic abnormalities that make it distinct from adult tumors, suggesting that although the microscopic appearance and grim prognosis are shared (Figure 1), pediatric and adult GBM have different underlying biologies (Paugh et al., 2010). In this issue of Cancer Cell, Sturm et al. (2012) show that pediatric GBM contains two epigenetic subgroups that are distinct from those found in adult tumors. The epigenetic proles of these groups correlate tightly with mutations in H3F3A. This gene encodes the replication-independent histone H3.3, which predominantly binds transcriptionally active loci and telomeres. H3.3 is frequently methylated at or near the residues that are mutated. Alterations in histone methylation affect the accessibility of the associated DNA and may promote changes in DNA methylation (Turcan et al., 2012). H3F3A mutations therefore provide a potential mechanism underlying the global methylation changes observed in these pediatric GBM subgroups. These ndings build on earlier studies that showed that adult GBM could be subdivided into three epigenetic subgroups, one of which correlated with mutations in the metabolic gene IDH1 (Noushmehr et al., 2010; Parsons et al., 2008; Verhaak et al., 2010). Patients without IDH1 mutation are older and have more rapidly progressive disease, whereas those with IDH1 mutation frequently have an antecedent low grade glioma, are younger, have more frequent TP53 mutations, and are less likely to have receptor tyrosine kinase amplication (Parsons et al., 2008). The IDH1 mutation is a gain of function alteration that creates a novel onco-metabolite, 2hydroxyglutarate (2HG), which interferes with the normal cellular methylation machinery. This in turn leads to widespread increases in global methylation known as the CpG-island methylator phenotype (CIMP) (Turcan et al., 2012). mRNA expression proling identied four subgroups of adult GBM (proneural, neural, classical, and mesenchymal) and determined that each subgroup contains distinct pathway alterations (Verhaak et al., 2010). Tumors with IDH1 mutations fall primarily into the proneural expression prole, and in epigenetic analyses, CIMP+ tumors segregate into that same group (Noushmehr et al., 2010). IDH1 is rarely altered in pediatric GBMs. Indeed, pediatric GBM contain signicantly different genomic alterations from adult tumors (Paugh et al., 2010). The recent discovery that mutations in the histone gene H3F3A occur predominantly in pediatric high grade gliomas (Schwartzentruber et al., 2012) further emphasizes the differences between adult and pediatric GBM. Sturm et al. (2012) now show that an expanded analysis of the methylome of GBM, which includes pediatric GBM cases, can divide this tumor into six subgroups, two of which are primarily pediatric, and the remainder of which are primarily adult. Methylation groups largely correlate with the pattern of expression proling that was previously reported, emphasizing the importance of epigenetic regulation in GBM (Verhaak et al., 2010). Sturm et al. (2012) show that three of the epigenetic subgroups (two pediatric and one adult) correlate with mutations in H3F3A and IDH1. H3F3A and IDH1 mutations are non-overlapping, suggesting that they represent different paths to achieve widespread alterations in genomic methylation. The two mutations in H3F3A are seven amino acid residues apart and are associated with dramatically different methylation proles. The H3F3A G34 mutation is associated with a hypomethylation phenotype, which is most prominent at the ends of chromosomes. The H3F3A K27 mutation is associated with a methylation pattern that is distinct from both that of H3F3A G34 and the CIMP+ phenotype associated with IDH1 mutation. There is also increasing evidence that alterations in H3F3A and IDH1 interfere with the normal differentiation of neural progenitors. The abnormal metabolite 2HG produced by mutant IDH1 leads to increased repressive methylation at H3K27 and inhibits immortalized neural cell differentiation, suggesting a potential common mechanism for tumorigenesis in IDH1 and H3F3A K27M mutant groups (Lu et al., 2012). IDH1 mutation also leads to increased neural stem cell marker expression and decreased expression of mature differentiation genes in response to differentiation signals, which are hallmarks of pre-cancerous cells (Turcan et al., 2012). Sturm et al. (2012) show in H3F3A G34 mutated glioblastoma a similar decrease in expression of OLIG2, an important neuro-developmental gene.

Cancer Cell 22, October 16, 2012 2012 Elsevier Inc. 417

Previews

Cancer Cell

REFERENCES Loss of OLIG2 is associated with increased methylation Heaphy, C.M., Subhawong, A.P., at the OLIG2 locus, which Hong, S.M., Goggins, M.G., Montoccurs despite the global hygomery, E.A., Gabrielson, E., Netto, G.J., Epstein, J.I., Lotan, T.L., pomethylation of the genome Westra, W.H., et al. (2011). Am. J. in H3F3A G34 mutated glioPathol. 179, 16081615. blastoma. Although there Louis, D.N., Ohgaki, H., Wiestler, are intriguing similarities in O.D., and Cavenee, W.K. (2007). the pathways between IDH1 WHO Classication of Tumours of and H3F3A mutated glioblasthe Central Nervous System (Lyon, France: IARC Press). toma, the disparate methylation signatures and clinical Lu, C., Ward, P.S., Kapoor, G.S., course of these three groups Rohle, D., Turcan, S., Abdel-Wahab, O., Edwards, C.R., Khanin, R., Figinvolving children and young ueroa, M.E., Melnick, A., et al. adults suggest that the (2012). Nature 483, 474478. mechanism by which these Noushmehr, H., Weisenberger, D.J., alterations lead to transforDiefes, K., Phillips, H.S., Pujara, K., Figure 1. Similar Appearances, Different Methylomes, and mation of normal cells will Berman, B.P., Pan, F., Pelloski, Divergent Outcomes C.E., Sulman, E.P., Bhat, K.P., require extensive study. A pediatric midline GBM of the pons (a diffuse intrinsic pontine glioma, et al.; Cancer Genome Atlas left) and a hemispheric glioblastoma multiforme arising in a young adult Although this study shows Research Network. (2010). Cancer (right) have a similar histologic appearance on high power hematoxylin and that different GBM methCell 17, 510522. eosin photomicrographs. Yet, as demonstrated by Sturm et al. (2012) in this ylation subgroups have issue of Cancer Cell, these two tumors affect different patient populations, Parsons, D.W., Jones, S., Zhang, X., arise in anatomically distinct regions, have different methylation proles, somewhat variable clinical Lin, J.C., Leary, R.J., Angenendt, P., and behave differently. For both images, magnication = 2003 and scale courses, the overall outMankoo, P., Carter, H., Siu, I.M., bar = 100 mM. Gallia, G.L., et al. (2008). Science comes of children and adults 321, 18071812. with GBM is extremely poor, with few long-term survivors reported hypomethylated preferentially at the Paugh, B.S., Qu, C., Jones, C., Liu, Z., Adamoin the literature (Louis et al., 2007). It chromosome ends and features the wicz-Brice, M., Zhang, J., Bax, D.A., Coyle, B., Barrow, J., Hargrave, D., et al. (2010). J. Clin. Onremains to be seen how identication of alternative lengthening of telomeres col. 28, 30613068. methylation alterations will translate into (ALT) phenotype. Although the precise improved therapeutic options for these mechanism of ALT remains unknown, Schwartzentruber, J., Korshunov, A., Liu, X.Y., Jones, D.T., Pfaff, E., Jacob, K., Sturm, D., Fontepatients. The signicant changes in the presence of promyelocytic leukemia basso, A.M., Quang, D.A., To njes, M., et al. (2012). global methylation patterns identied by bodies and evidence of heterologous Nature 482, 226231. Sturm et al. (2012) make epigenetic recombination between telomeres proSturm, D., Witt, H., Hovestadt, V., Khuong-Quang, altering agents an attractive therapeutic vide enticing targets for therapy (Heaphy D.-A., Jones, D.T.W., Konerman, C., Pfaff, E., modality. The Childrens Oncology Group et al., 2011). The recent increases in our Tonjes, M., Sill, M., Bender, S., et al. (2012). Cancer is currently investigating altering chro- understanding of the genetics of GBM, Cell 22, this issue, 425437. matin structure with histone deacetylase and in particular, the integration of Turcan, S., Rohle, D., Goenka, A., Walsh, L.A., Fang, inhibitors in GBM (NCT01236560). How- both pediatric and adult tumors into an F., Yilmaz, E., Campos, C., Fabius, A.W., Lu, C., ever demethylating agents have not overall epigenomic classication scheme Ward, P.S., et al. (2012). Nature 483, 479483. been widely investigated in pediatric by Sturm et al. (2012), set the stage for Verhaak, R.G., Hoadley, K.A., Purdom, E., Wang, GBM. translation of molecular advances into V., Qi, Y., Wilkerson, M.D., Miller, C.R., Ding, L., Golub, T., Mesirov, J.P., et al.; Cancer Genome Sturm et al. (2012) show that the improved care for patients with this Atlas Research Network. (2010). Cancer Cell 17, 98110. G34 subgroup of pediatric GBM is disease.

418 Cancer Cell 22, October 16, 2012 2012 Elsevier Inc.

Cancer Cell

Previews
Cell Cycle-Based Therapies Move Forward
Marcos Malumbres1,*
1Cell Division and Cancer Group, Spanish National Cancer Research Centre (CNIO), Madrid E-28029, Spain *Correspondence: malumbres@cnio.es http://dx.doi.org/10.1016/j.ccr.2012.09.024

Targeted therapies directed against cell cycle regulators have been difcult to translate into the clinic. In this issue of Cancer Cell, Choi et al. and Sawai et al. rekindle the therapeutic value of inhibiting specic cyclin-dependent kinase complexes by demonstrating their requirements in the maintenance of breast tumors and leukemias.
The real problem is to understand why cancer cells grow when their normal counterparts would not. (Hunt, 2008) Tumor cells invariably display defects in the machinery that controls the cell division cycle. Yet, whether the cell cycle is a useful therapeutic target is still being intensely debated mostly due to the essential role of this process in tissue homeostasis and the difculties of nding a therapeutic window (Malumbres and Barbacid, 2009). Progression through the cell cycle is driven by several protein kinases, including cyclin-dependent kinases (CDKs). Molecular analysis of tumor cells has provided strong evidence that the activity of these kinases is deregulated in human tumors suggesting their potential therapeutic use. However, the clinical benet of the rst generation of CDK inhibitors has been limited due to toxicity and lack of specicity, possibly derived from the critical function of CDKs in DNA replication and chromosome segregation. The pioneering work by P. Sicinskis group in 2001 (Yu et al., 2001) showed that a specic interphase cyclin, cyclin D1, was required for RAS- and HER2induced mammary tumors, but was dispensable for WNT- or MYC-induced mammary tumors. This work showed for the rst time that the therapeutic value of a cell cycle regulator may be cell-type specic and dependent on specic oncogenic alterations. These studies were performed using germline cyclin D1 knockout mice, which raised some important questions at that time. Since cyclin D1 knockout mice displayed minor defects in mammary gland development, it was argued that the effect observed could be due to special requirements for cyclin D1 in the cell of origin of these tumors. Two studies in 2006 suggested that this was not the case, as a similar therapeutic benet was observed using CDK4-decient mice or knockin mice expressing a mutant cyclin D1, which does not bind CDKs but maintains other CDK-independent functions (Landis et al., 2006; Yu et al., 2006). None of these models displayed defects in mammary gland development; yet, they were resistant to HER2-induced mammary gland carcinomas. In addition, these studies suggested that CDK4, one of the kinase partners of cyclin D1, was the critical enzymatic activity to be targeted in HER2-positive mammary gland carcinomas. Subsequent studies by Barbacid and colleagues (Puyol et al., 2010) showed that CDK4, but not CDK2 or CDK6, was critical for the development of K-RAS-induced lung tumors, whereas lack of this interphase kinase did not affect the development of lungs in germline knockout mice. Because the genetic modication was present since conception, the remaining question was whether the acute inhibition of cyclin D1-CDK complexes would be effective in already developed breast tumors. A new study in this issue of Cancer Cell by Choi et al. (2012) now shows that conditional genetic ablation of cyclin D1 in adult mice that bear tumors inhibits the growth of HER2-positive mammary carcinomas. A similar effect is obtained using a small-molecule inhibitor of CDK4/6 (PD 0332991) currently being studied in clinical trials. Importantly, this effect is not accompanied by toxicities associated with the acute inhibition of cyclin D1-CDK complexes in adult individuals (Puyol et al., 2010; Choi et al., 2012; Sawai et al., 2012), suggesting that sideeffects of such treatments will be minimal. Based on the previous results, cyclin D1-CDK4 complexes seem to be critical targets in the proliferation of epithelial tumors, at least breast and lung carcinomas (Figure 1). What about other interphase CDK complexes? Previous studies established a critical role for cyclin D3 and, at least partially, CDK6 in lymphocytes and in the development of T cell leukemias (Hu et al., 2009; Sicinska et al., 2003). The relevance of these ndings in cancer therapy has now been addressed using conditional knockouts or acute treatments with kinase inhibitors. Genetic ablation of cyclin D3 or inhibition of CDK4/6 complexes in Notch1-induced T cell acute lymphoblastic leukemia (T-ALL) results in tumor regression without causing major abnormalities in other tissues (Choi et al., 2012; Sawai et al., 2012; in this issue of Cancer Cell). Importantly, the response of leukemic cells to these treatments is dramatically different from that of epithelial cells. Both lung and breast cancer cells display characteristics of cell cycle arrest and senescence after genetic ablation or chemical inhibition of cyclin D1-CDK4 complexes (Choi et al., 2012; Puyol et al., 2010). Similar treatments, however, result in apoptotic cell death in mouse and human leukemic cells (Choi et al., 2012; Sawai et al., 2012). This response seems to be associated with the nding that Notch1-induced tumors display high levels of cyclin D2/D3 and cyclin D3-CDK6 complexes in agreement with the modulation of cyclin D3 expression by the Notch1 pathway (Choi et al., 2012; Joshi et al., 2009). In fact, Notch1-negative T-ALLs or other types of leukemic cells did not undergo apoptosis following CDK4/6 inhibition (Choi et al., 2012). Whether the

Cancer Cell 22, October 16, 2012 2012 Elsevier Inc. 419

Previews

Cancer Cell

REFERENCES susceptibility to apoptosis is provided by the activation of Choi, Y.J., Li, X., Hydbring, P., Notch1 or the specic Sanda, T., Stefano, J., Christie, inhibition of cyclin D3-(and A.L., Signoretti, S., Look, A.T., Kung, A.L., von Boehmer, H., and possibly CDK6) complexes Sicinski, P. (2012). Cancer Cell 22, is not clear at present. Interthis issue, 438451. estingly, cyclin D2 cannot Hu, M.G., Deshpande, A., Enos, M., compensate for the lack of Mao, D., Hinds, E.A., Hu, G.F., cyclin D3 when expressed Chang, R., Guo, Z., Dose, M., Mao, from the Ccnd3 locus sugC., et al. (2009). Cancer Res. 69, 810818. gesting intrinsic differences in the function of these two Hunt, T. (2008). Cell Cycle 7, 3789 3790. proteins (Sawai et al., 2012). The molecular basis for the Joshi, I., Minter, L.M., Telfer, J., differential response to Demarest, R.M., Capobianco, A.J., Aster, J.C., Sicinski, P., Fauq, A., CDK4/6 inhibitionsenesGolde, T.E., and Osborne, B.A. cence versus apoptosis (2009). Blood 113, 16891698. may be considered a new Landis, M.W., Pawlyk, B.S., Li, T., avenue of high interest in Figure 1. An Initial Road Map for the Treatment of Human Sicinski, P., and Hinds, P.W. Malignancies with Specic CDK4/6 Inhibitors the therapeutic evaluation of (2006). Cancer Cell 9, 1322. Since the small-molecule inhibitor PD 0332991 inhibits both CDK4 and these cell cycle regulators. CDK6 complexes, the participation of CDK4 in human leukemias cannot Malumbres, M., and Barbacid, M. It is been a long road20 be discarded at this time. Similarly, it is not clear which specic cyclins (2009). Nat. Rev. Cancer 9, 153166. activate CDK4 in lung tumors, although this information is likely to be disyearssince the initial genepensable for the selection of small-molecule kinase inhibitors. Further ration of genetically-engin, A., Dubus, P., Puyol, M., Mart research will be necessary to evaluate the relevance of specic cyclin-CDK Mulero, F., Pizcueta, P., Khan, G., neered mice with cell cycle complexes in similar pathologies induced by other oncogenes (e.g., mammary a, D., and Guerra, C., Santamar gland or lung tumors without activation of the RAS pathway) or in other mutations. The discovery Barbacid, M. (2010). Cancer Cell tumor types. that individual interphase 18, 6373. cyclins or CDKs were disSawai, C., Freund, J., Oh, P., pensable for the developNdiaye-Lobry, D., Bretz, J.C., ment and homeostasis of most tissues human tumors may be sensitive to Strikoudis, A., Genesca, L., Trimarchi, T., Kelliher, was considered a demonstration of specic cyclin-CDK complexes, as long M.A., Clark, M., et al. (2012). Cancer Cell 22, this issue, 452465. the developmental plasticity of mamma- as we identify the specic complexes lian tissues and raised doubts on the that mediate the response to specic Sicinska, E., Aifantis, I., Le Cam, L., Swat, W., Borusefulness of inhibiting these activities oncogenic pathways in each specic owski, C., Yu, Q., Ferrando, A.A., Levin, S.D., in tumors (Malumbres and Barbacid, cell type. A long road in which difcult Geng, Y., von Boehmer, H., and Sicinski, P. (2003). Cancer Cell 4, 451461. 2009). The recent studies in breast, questions such as why cancer cells lung, and T-ALL make a strong case grow when their normal counterparts Yu, Q., Geng, Y., and Sicinski, P. (2001). Nature for the identication of cellular contexts would not (Hunt, 2008) need to be 411, 10171021. in which the inhibition of specic cy- addressed in a cell-type- and onco m, M., Zaclin-CDK complexes may have thera- gene-specic manner. Cancer patients Yu, Q., Sicinska, E., Geng, Y., Ahnstro gozdzon, A., Kong, Y., Gardner, H., Kiyokawa, H., peutic value (Figure 1). Based on these will undoubtedly benet from these Harris, L.N., Sta l, O., and Sicinski, P. (2006). Cancer Cell 9, 2332. results, one could propose that many studies.

420 Cancer Cell 22, October 16, 2012 2012 Elsevier Inc.

Cancer Cell

Previews
The Not-so-Skinny on Muscle Cancer
Ken Kikuchi1 and Charles Keller1,*
1Pediatric

Family Pediatric Research Institute, Department of Pediatrics, Oregon Health & Science University, Cancer Biology Program, Pape Portland, OR 97239 USA *Correspondence: keller@ohsu.edu http://dx.doi.org/10.1016/j.ccr.2012.09.018

The childhood cancer embryonal rhabdomyosarcoma can arise in tissue without skeletal muscle elements. In this issue of Cancer Cell, Hatley and colleagues report that non-skeletal muscle progenitors can be a cell of origin for Sonic Hedgehog-driven embryonal rhabdomyosarcoma in an adipocyte-restricted conditional mouse model.
One of the most intriguing clinical and scientic questions in pediatric oncology is how rhabdomyosarcoma, a tumor with a myogenic phenotype, can arise in tissue without hypaxial-like skeletal muscle elements. This conundrum is especially evident for the embryonal subtype of rhabdomyosarcoma (eRMS), which can arise from the salivary glands, skull base (parameninges), biliary tree, and genitourinary tract (bladder/prostate) (Gurney et al., 1999; Shapiro and Bhattacharyya, 2006). The work by Hatley et al. (2012) in this issue of Cancer Cell begins to address this conundrum by highlighting the plasticity of cells traditionally thought to be in the adipogenic lineage. In this paper, Hatley et al. (2012) conditionally activate a mutant Smoothened (SmoM2) allele to drive Hedgehog signaling in the adipogenic lineage using a 5.4 kb promoter/enhancer fragment from the adipocyte fatty acid binding protein 4 gene (Fabp4, also called aP2) to drive Cre. Eighty percent of aP2-Cre; SmoM2/+ mice developed eRMS in the head and ventral neck by 2 months of age. aP2-Cre Cdkn2aFlox/Flox mice did not develop eRMS, whereas deletion of Cdkn2a in aP2-Cre;SmoM2/+ mice decreased latency and increased tumor penetrance, with all mice having tumors by 55 days. This nding implicates Cdkn2a locus loss as a secondary factor (disease modier) of eRMS progression. The authors validated tumors as true eRMS by histology and immunohistochemistry (Desmin, MyoD, and Myogenin). Similarly, an embryonal muscle gene signature (MyoD1, Myogenin, Pax7, Myf5, Myh3, and Myh8) was evident by RT-PCR assay. To further afrm the diagnosis, Hatley et al. (2012) performed gene expression proling of aP2-Cre;SmoM2/+ tumors in comparison to human eRMS and tumors of previously reported mouse eRMS models. Overall, 67% of the probe pairs and 58% of the ortholog gene pairs showed agreement in gene expression between their mouse tumors and previously published eRMS models (Rubin et al., 2011) or human tumors, respectively. Hatley et al. (2012) concluded that, despite being of an adipogenic lineage of origin, aP2-Cre;SmoM2/+ mouse tumors are an accurate preclinical eRMS model based on these histological and transcriptome parameters. The most interesting aspect of this paper is that these tumors originated from aP2 expressing cells. Previously, activity of the aP2 promoter in transgenic mice has been documented as specic for the adipocyte lineage and to be distinct from any skeletal muscle lineage (Urs et al., 2006). Moreover, the specic aP2-Cre transgenic line used by Hatley et al. (2012) was shown to be active in adipose tissue but not in skeletal muscle, at least as evidenced from recombination in the sternocleidomastoid muscle (SCM) of aP2-Cre;R26-LacZ reporter mice and aP2-Cre;R26-YFP mice and whole-tissue examination. Interestingly, it was found that eRMS tumors in situ were completely surrounded by non-neoplastic adipose tissue adjacent to and clearly separated from the SCM at P14. To compare the tumorigenic potential of SmoM in myogenic lineages, Hatley et al. (2012) activated the SmoM2 allele in early muscle development, employing Pax3-, Myf5-, or MyoD1-Cre. All of these crosses resulted in embryonic lethality without tumor formation. However, activation of SmoM2 with Myogenin-Cre, which is specic for myoblast-stage muscle differentiation, resulted in tongue tumors in 100% of the mice. Mice with activation of SmoM2 in terminally differentiated skeletal muscle (MCK-Cre) were viable with no evidence of tumorigenesis. This result is the best by far in asking whether differentiated versus differentiating myobers can transform into rhabdomyosarcoma. To address the tumorigenic potential of SmoM2 in postnatal muscle stem cells (satellite cells), a Pax7CreERT2 was employed, and mice were aged until 150 dayswithout development of tumors. It would have been intriguing to see if prior reports of nonmyogenic sarcomas arising from a satellite cell lineage (Rubin et al., 2011) would have been observed if the mice had been aged longer. Another caveat of these ndings is that the promoter of SmoM2 in this system was Rosa26, not the native Smo promoter, and it is therefore difcult to say that this is a molecularly physiological eRMS model. Furthermore, given that most Hedgehog pathway inhibitors currently in clinical trials are Smoothened antogonists, this model may have highest value as a genetic model of eRMS for the time being and as a preclinical therapeutic model only when GLI inhibitors emerge as clinical candidate agents. Taking a look onto mesenchymal progenitor cell biology, the plasticity of these stem cells may not be so surprising. It is known that Myf5 expressing cell can develop into both muscle and brown fat tissue via PRDM16 and PPARg signaling (Seale et al., 2008). Furthermore, satellite cells can undergo adipogenesis under certain experimental conditions (Asakura et al., 2001; Shefer et al., 2004) or may apparently undergo transdifferentiation into broblasts (Brack et al., 2007). Hedgehog activation can also reduce fat

Cancer Cell 22, October 16, 2012 2012 Elsevier Inc. 421

Previews

Cancer Cell

formation via PPARg (Suh therapeutic targets for select et al., 2006). Such reports patients. Applying knowledge lend one to speculate that from these specialized prethe origin of aP2-Cre; clinical models to clinical use SmoM2/+ mice tumors might will no doubt require novel be cells that have undergone statistical designs for clinical a transdifferentiation event trials, given that the overall from the adipocyte to the annual incidence of pediatric myogenic lineage (Figure 1). rhabdomyosarcoma is only in Hatley et al. (2012) recapitthe hundreds, not thousands. ulate another important ndPersonalized rhabdomyosaring by gene expression procoma science inches ever ling that aP2-Cre;SmoM2/+ closer. tumors show an activated satellite cell phenotype reREFERENCES ective of the eRMS tumor Asakura, A., Komaki, M., and Rudpathology rather than the nicki, M. (2001). Differentiation 68, lineage of origin for the 245253. tumor. Certainly, activation Brack, A.S., Conboy, M.J., Roy, S., of Hedgehog signaling was Lee, M., Kuo, C.J., Keller, C., and evident by RT-PCR studies Rando, T.A. (2007). Science 317, M2/+ of aP2-Cre;Smo tumors, 807810. yet, by contrast, less than Gurney, J.G., Young, J.L., Jr., 30% of human eRMS exhibit Roffers, S.D., Smith, M.A., and Bunin, G.R. (1999). Soft Tissue a gene expression signature Sarcomas. In Cancer Incidence and consistent with a Hedgehog Survival among Children and pathway on overdrive state Adolescents: United States SEER Program 19751995, National Figure 1. Model of Skeletal Myogenesis and Possible Cellular (Rubin et al., 2011), and only Cancer Institute, SEER Program. Origins of eRMS rarely is Hedgehog overdrive NIH Pub. No. 994649. Ries LAG, Postnatal muscle maintenance and regeneration are regulated by Pax3, Pax7, SM, J.G. Gurney, M. Linet, T. Tamra, a solitary signaling abnorand muscle regulatory factors (MyoD, Myf5, Myogenin, and Myf6). Studies J.L. Young, and G.R. Bunin, eds. mality. Instead, p53 loss of described in the text suggest that the cells of origin for eRMS include differen(Bethesda, MD: National Cancer tiating myoblasts and adipocytes. function was most frequent Institute). (Rubin et al., 2011)implying Hatley, M.E., Tang, W., Garcia, M.R., that p53 loss precedes Finkelstein, D., Millay, D.P., Liu, N., Hedgehog overdrive temporally in rhab- eRMS patients it models, and closer Graff, J., Galindo, R.L., and Olson, E.N. (2012). domyosarcomagenesis. The p53 status examination of the preneoplastic and Cancer Cell 22, this issue, 536546. of aP2-Cre;SmoM2/+ tumors was unex- early neoplastic lesions will invariably Rubin, B.P., Nishijo, K., Chen, H.I., Yi, X., Schuetze, plored, but it would have been of interest help us understand the microenvironment D.P., Pal, R., Prajapati, S.I., Abraham, J., Arenkiel, B.R., Chen, Q.R., et al. (2011). Cancer Cell 19, to have known the p53 status of these in which these tumor initiating cells 177191. tumors, particularly given the 1969 histor- transform. P., Bjork, B., Yang, W., Kajimura, S., Chin, Overall, the heterogeneity in human Seale, ical precedent of Li and Fraumeni ` , A., Devarakonda, S., Conroe, S., Kuang, S., Scime describing a familial eRMS syndrome rhabdomyosarcoma phenotypes may H.M., Erdjument-Bromage, H., et al. (2008). Nature now known to be attributable to germline result from the balance between genetic 454, 961967. p53 mutation. factors (mutational prole of the tumor Shapiro, N.L., and Bhattacharyya, N. (2006). In the Hatley model, eRMS tumors including the initial mutation[s] and modi- Otolaryngol. Head Neck Surg. 134, 631634. occurred from only the head and neck, ers) and epigenetic factors (the cell of Shefer, G., Wleklinski-Lee, M., and Yablonkayet human eRMS occur also in the origin). In an era of precision medicine, Reuveni, Z. (2004). J. Cell Sci. 117, 53935404. urogenital tract, extremities, and retro- this transgenic model representing the Suh, J.M., Gao, X., McKay, J., McKay, R., Salo, Z., peritoneum. Nevertheless, this cranial- subset of human head and neck ERMS and Graff, J.M. (2006). Cell Metab. 3, 2534. oriented eRMS model may be of specic not derived from myogenic precursors Urs, S., Harrington, A., Liaw, L., and Small, D. interest for the subset of head and neck may be particularly valuable in dening (2006). Transgenic Res. 15, 647653.

422 Cancer Cell 22, October 16, 2012 2012 Elsevier Inc.

Cancer Cell

Previews
A Mitochondrial Power Play in Lymphoma
Ralph J. DeBerardinis1,*
1Childrens Medical Center Research Institute, University of Texas Southwestern Medical Center, 5323 Harry Hines Boulevard, Dallas, TX 75390-8502, USA *Correspondence: ralph.deberardinis@utsouthwestern.edu http://dx.doi.org/10.1016/j.ccr.2012.09.023

Deregulated energetics is a hallmark of malignancy, but metabolic heterogeneity among individual tumors is unknown. A study by Caro et al. in this issue of Cancer Cell demonstrates that a subset of lymphomas is dened by reliance on mitochondrial energy generation and is selectively killed when this pathway is impaired.
Altered metabolism was among the rst cancer biomarkers (Koppenol et al., 2011). By 1925, Otto Warburg had observed that tumors rapidly took up glucose and converted it to lactic acid, which was released into the milieu. This was a departure from the other organs he studied, which imported less glucose, oxidized a higher fraction of it to carbon dioxide, and secreted much less lactate. This remarkable phenotype, called the Warburg effect, provides the rationale for two modern imaging technologies in cancer diagnosis: 18uoro-2-deoxyglucose positron emission tomography (FDG-PET) to identify tumors by rapid glucose import and 1H magnetic resonance spectroscopy to identify areas of elevated lactate. Over the last two decades, the Warburg effect has been mechanistically tied to the molecular basis of transformation, as tumor-promoting mutations in many different oncogenes and tumor suppressors have been demonstrated to stimulate glycolysis (Ward and Thompson, 2012). Cells have two ways to produce adenosine triphosphate (ATP) for energy: glycolysis and oxidative phosphorylation (OxPhos) (Figure 1). In glycolysis, glucose is converted to pyruvate, generating NADH from NAD+ and ATP from ADP. If the pyruvate is reduced to lactate, NAD+ is regenerated and glycolysis continues. Although glycolysis is rapid, it is deemed inefcient because most of the energy that could be generated from glucose is lost when the cell secretes lactate. In contrast, OxPhos is highly efcient. When substrates like pyruvate are oxidized in the mitochondria, reducing equivalents are delivered to the electron transport chain, creating a proton gradient coupled to ATP synthesis (Figure 1). This system yields nearly 20 times as many ATP molecules per glucose as the Warburg effect. Other oxidizable substrates, like fatty acids, produce an even higher energy yield. Warburg considered tumor glycolysis to be a metabolic anomaly, famously postulating that it was the consequence of irreversible defects in respiration (i.e., OxPhos) which were the root cause of cancer (Warburg, 1956). This model has been disproven as a general principle, because most cancer cells do not have static defects in their respiratory machinery, and more pointedly, because nonmalignant cells also display the Warburg effect if stimulated to grow (Wang et al., 1978). Nevertheless, the concept that glycolysis is a universal feature of aggressive tumor growth, and that OxPhos is counter-productive to tumorigenesis still pervades the literature. It was therefore surprising when a bioinformatics study on diffuse large B cell lymphoma (DLBCL) revealed that some 30% of these tumors belonged to a subset dened by high expression of genes involved in OxPhos (Monti et al., 2005). DLBCL is the most common nonHodgkins lymphoma in Western populations and is characterized by rapid growth. Several attempts have been made to classify DLBCL by molecular typing, particularly through shared patterns of gene expression. One such approach revealed three major DLBCL subtypes: the OxPhos subset; the B cell receptor (BCR) subset, characterized by expression of genes involved in B cell receptor signaling; and the host response (HR) subset, expressing markers of inltrating inammatory cells (Monti et al., 2005). Now, a study in this issue of Cancer Cell (Caro et al., 2012) examines the metabolic phenotypes of these subtypes and found that OxPhos cell lines and primary biopsies contained elevated expression of many mitochondrial proteins. OxPhos cells displayed enhanced glucose oxidation relative to lactate formation, better defenses against oxidative stress, and a robust ability to oxidize fatty acids in the mitochondria. Providing exogenous fatty acids stimulated survival and growth in OxPhos cells, but not in cells from other subtypes. Overall, cells from OxPhos tumors generated a higher fraction of their ATP in the mitochondria, and this activity was required for cellular tness. Importantly, inhibition of mitochondrial metabolism selectively killed OxPhos cells. The nuclear receptor peroxisome proliferator-activated receptor gamma (PPARg), which regulates the expression of many genes encoding mitochondrial proteins, was overexpressed in OxPhos DLBCLs. PPARg antagonists suppressed fatty acid oxidation and killed OxPhos cells in culture. Inhibitors of fatty acid oxidation had the same effect, as did silencing of an enzyme required to synthesize the antioxidant glutathione. None of these treatments affected nonOxPhos cells. The work illustrates several key principles in cancer metabolism. First, it demonstrates signicant metabolic heterogeneity among tumors of the same type. This point should be emphasized, because it tempers the paradigm that tumors are monolithic in their preference for glycolytic energy production. Second, metabolic heterogeneity is programmed, and it persists when cells are adapted to culture. This argues strongly against the concept that the glycolytic phenotype of most cancer cell lines is

Cancer Cell 22, October 16, 2012 2012 Elsevier Inc. 423

Previews

Cancer Cell

an unavoidable consequence of new work forties Warburgs historGlucose culture conditions. Even the ical claim that metabolism is an suppression of OxPhos in other actionable biomarker (even if DLBCL subtypes was programmed, he would have been surprised by because these cells could be stimuthese particular results). PPARg + ADP, P NAD i lated to activate fatty acid oxidation antagonism may provide a starting NADH ATP simply by silencing Syk, a kinase in point for targeting OxPhos tumors, Fatty BCR signaling. Third, high rates of and it would also be interesting Pyruvate Lactate Acids OxPhos can support growth in to determine whether OxPhos precancer cell lines. This may also be dicts sensitivity to therapies true in vivo. Although a complete currently in use for cancer. New Acetyl-CoA clinical description of the OxPhos approaches to image tumor metaDLBCL is lacking, patients with bolism noninvasively (Kurhanewicz Oxaloacetate Citrate these tumors do not appear to et al., 2011) may soon make it Reducing TCA survive any longer than those with possible to stratify patients to equivalents Cycle other types (Monti et al., 2005). The therapeutic regimens solely by aggressive course of DLBCL as a probing metabolism in vivo. ATP ADP, Pi whole suggests that a preference ATP ATP Electron ATP for OxPhos does not prevent rapid ATP Transport Chain cell growth in the large subset REFERENCES ATP ATP mitochondrion ATP of tumors using that metabolic platCaro, P., Kishan, A.U., Norberg, E., Stanley, form. Thus, metabolic programming I.A., Chapuy, B., Ficarro, S.B., Polak, K., to use the mitochondria rather than Tondera, D., Gounarides, J., Yin, H., glycolysis as the major site of energy Glycolysis OxPhos et al. (2012). Cancer Cell 22, this issue, 547560. formation is an effective strategy for fraction of ATP generation cancer cell growth. Koppenol, W.H., Bounds, P.L., and Dang, The ndings also emphasize that C.V. (2011). Nat. Rev. Cancer 11, Non-OxPhos OxPhos 325337. cancer cells, like normal cells, use cells cells multiple pathways concurrently to Kurhanewicz, J., Vigneron, D.B., Brindle, K., produce energy (Figure 1). The Chekmenev, E.Y., Comment, A., Cunningham, C.H., Deberardinis, R.J., Green, G.G., Figure 1. Variable Reliance on Oxidative vast majority of cancer cells use Leach, M.O., Rajan, S.S., et al. (2011). Phosphorylation in DLBCL both glycolysis and OxPhos Neoplasia 13, 8197. Cells produce ATP from glycolysis (blue pathway) and oxidatogether, although the balance tive phosphorylation (OxPhos, green). Cancer cell lines vary Maher, E.A., Marin-Valencia, I., Bachoo, between the two varies widely (Zu according to what fraction of their total ATP comes from R.M., Mashimo, T., Raisanen, J., Hatanpaa, each pathway, as shown on the spectrum at the bottom. A and Guppy, 2004). DLBCL cells K.J., Jindal, A., Jeffrey, F.M., Choi, C., subset of DLBCL tumors was dened by high expression of are no exception. Despite their Madden, C., et al. (2012). NMR Biomed. genes related to OxPhos. These cells produced most of their Published online March 15, 2012. http://dx. ability to oxidize pyruvate and ATP from OxPhos and required this pathway for survival and doi.org/10.1002/nbm.2794. growth. Non-OxPhos cells, which did not share this gene fatty acids in the mitochondria, expression signature, produced a higher fraction of their ATP OxPhos cells still consumed gluMonti, S., Savage, K.J., Kutok, J.L., from glycolysis, and were resistant to OxPhos inhibition. Feuerhake, F., Kurtin, P., Mihm, M., Wu, cose and secreted a modest Abbreviation: TCA, tricarboxylic. B., Pasqualucci, L., Neuberg, D., Aguiar, amount of lactate. Conversely, the R.C., et al. (2005). Blood 105, 18511861. non-OxPhos cells still produced half of their ATP in the mitochondria (Fig- factors that stimulate cell growth and Wang, T., Sheppard, J.R., and Foker, J.E. (1978). ure 1). Along these lines, a large majority may contribute to tumorigenesis (Wein- Science 201, 155157. of DLBCL tumors are readily detected berg et al., 2010). Warburg, O. (1956). Science 123, 309314. It remains to be seen whether the by FDG-PET, and this likely includes P.S., and Thompson, C.B. (2012). Cancer OxPhos tumors. Thus, avid uptake of OxPhos phenotype is driven by tumori- Ward, Cell 21, 297308. glucose analogs may occur even when genic mutations or other features of the mitochondria are fully active. Another DLBCL biology, and whether it bestows Weinberg, F., Hamanaka, R., Wheaton, W.W., Weinberg, S., Joseph, J., Lopez, M., Kalyanararecent study on metabolic ux in glio- some context-specic advantage to this man, B., Mutlu, G.M., Budinger, G.R., and Chanblastoma reached this same conclusion subset. Regardless of the mechanism, it del, N.S. (2010). Proc. Natl. Acad. Sci. USA 107, (Maher et al., 2012). Besides ATP, mito- is encouraging that molecular phenotyp- 87888793. chondria produce biosynthetic precur- ing led to the discovery of metabolic Zu, X.L., and Guppy, M. (2004). Biochem. Biophys. sors, reactive oxygen species, and other vulnerabilities in DLBCL. In a sense, the Res. Commun. 313, 459465.

424 Cancer Cell 22, October 16, 2012 2012 Elsevier Inc.

Cancer Cell

Article
Hotspot Mutations in H3F3A and IDH1 Dene Distinct Epigenetic and Biological Subgroups of Glioblastoma
Dominik Sturm,1,42 Hendrik Witt,1,7,42 Volker Hovestadt,2,42 Dong-Anh Khuong-Quang,11,42 David T.W. Jones,1 njes,2 Martin Sill,4 Sebastian Bender,1 Marcel Kool,1 Marc Zapatka,2 Carolin Konermann,3 Elke Pfaff,1 Martje To Natalia Becker,4 Manuela Zucknick,4 Thomas Hielscher,4 Xiao-Yang Liu,11 Adam M. Fontebasso,12 Marina Ryzhova,13 Steffen Albrecht,14 Karine Jacob,11 Marietta Wolter,15 Martin Ebinger,16 Martin U. Schuhmann,17 Timothy van Meter,18 Michael C. Fru hwald,19 Holger Hauch,20 Arnulf Pekrun,21 Bernhard Radlwimmer,2 Tim Niehues,22 Gregor von Komorowski,23 Matthias Du rken,23 Andreas E. Kulozik,7 Jenny Madden,24 Andrew Donson,24 Nicholas K. Foreman,24 Rachid Drissi,25 Maryam Fouladi,25 Wolfram Scheurlen,26 Andreas von Deimling,5,9 Camelia Monoranu,27 Wolfgang Roggendorf,27 Christel Herold-Mende,8 Andreas Unterberg,8 Christof M. Kramm,28 rg Felsberg,15 Christian Hartmann,29 Benedikt Wiestler,10 Wolfgang Wick,10 Till Milde,6,7 Olaf Witt,6,7 Jo Anders M. Lindroth,3 Jeremy Schwartzentruber,30 Damien Faury,11 Adam Fleming,11 Magdalena Zakrzewska,31 Pawel P. Liberski,31 Krzysztof Zakrzewski,32 Peter Hauser,33 Miklos Garami,33 Almos Klekner,34 Laszlo Bognar,34 Sorana Morrissy,35 Florence Cavalli,35 Michael D. Taylor,35 Peter van Sluis,36 Jan Koster,36 Rogier Versteeg,36 Richard Volckmann,36 Tom Mikkelsen,37 Kenneth Aldape,38 Guido Reifenberger,15 V. Peter Collins,39 Jacek Majewski,40 Andrey Korshunov,5 Peter Lichter,2 Christoph Plass,3,41,* Nada Jabado,11,41,* and Stefan M. Pster1,7,41,*
of Pediatric Neurooncology of Molecular Genetics 3Division of Epigenetics and Cancer Risk Factors 4Division of Biostatistics 5Clinical Cooperation Unit Neuropathology 6Clinical Cooperation Unit Pediatric Oncology German Cancer Research Center (DKFZ) Heidelberg, 69120 Heidelberg, Germany 7Department of Pediatric Oncology, Hematology, and Immunology 8Department of Neurosurgery 9Department of Neuropathology 10Department of Neurooncology Heidelberg University Hospital, 69120 Heidelberg, Germany 11Departments of Pediatrics and Human Genetics, McGill University 12Division of Experimental Medicine, Montreal Childrens Hospital McGill University Health Center Research Institute, Montreal, QC H3Z 2Z3, Canada 13NN Burdenko Neurosurgical Institute, Moscow, 125047, Russia 14Department of Pathology, Montreal Childrens Hospital, McGill University Health Center Research Institute, Montreal, QC H3H 1P3, Canada 15Department of Neuropathology, Heinrich-Heine-University, 40225 Du sseldorf, Germany 16Department of Hematology and Oncology, Childrens University Hospital Tu bingen, 72076 Tu bingen, Germany 17Department of Neurosurgery, University Hospital Tu bingen, 76076 Tu bingen, Germany 18Virginia Commonwealth University, Richmond, VA 23298, USA 19Pediatric Hospital, Klinikum Augsburg, 86156 Augsburg, Germany 20Pediatric Hospital, Klinikum Heilbronn, 74078 Heilbronn, Germany 21Prof.-Hess-Kinderklinik, Klinikum Bremen-Mitte, 28177 Bremen, Germany 22Childrens Hospital, Helios Clinics, 47805 Krefeld, Germany 23Childrens University Hospital, 68135 Mannheim, Germany 24Department of Pediatrics, University of Colorado Denver, Aurora, CO 80045, USA 25Division of Oncology, Cincinnati Childrens Hospital Medical Center, Cincinnati, OH 45229, USA 26Cnopfsche Kinderklinik, Nu rnberg Childrens Hospital, 90419 Nu rnberg, Germany 27Department of Neuropathology, Institute of Pathology, University Wu rzburg, 97080 Wu rzburg, Germany 28University Childrens Hospital, Martin Luther University Halle-Wittenberg, 06120 Halle, Germany 29Institute of Pathology, Department of Neuropathology, Hannover Medical School, 30175 Hannover, Germany 30McGill University and Genome Quebec Innovation Centre, Montreal, QC H3A 1A4, Canada 31Department of Molecular Pathology and Neuropathology, Medical University of Lodz 92-216 Poland 32Department of Neurosurgery, Polish Mothers Memorial Hospital Research Institute, Lodz 93-338 Poland 332nd Department of Paediatrics, Semmelweis University, Budapest H-1094 Hungary 34Department of Neurosurgery, Medical and Health Science Center, University of Debrecen, H-4032 Debrecen, Hungary 35Program in Developmental and Stem Cell Biology, Division of Neurosurgery, Arthur and Sonia Labatt Brain Tumour Research Centre, Hospital for Sick Children, University of Toronto, Toronto, ON M4N 1X8, Canada 36Department of Oncogenomics, AMC, University of Amsterdam, Amsterdam 1105 AZ, The Netherlands 37Departments of Neurology and Neurosurgery, Henry Ford Hospital, Detroit, MI 48202, USA 38Department of Neuro-Oncology, University of Texas MD Anderson Cancer Center, Houston, TX 77030, USA 39Division of Molecular Histopathology, Department of Pathology, University of Cambridge, Cambridge, CB2 0QQ, United Kingdom
2Division 1Division

Cancer Cell 22, 425437, October 16, 2012 2012 Elsevier Inc. 425

Cancer Cell
Epigenetic and Biological Subgroups of Glioblastoma

40Department 41These

of Human Genetics, McGill University, Montreal, QC H3Z 2Z3, Canada authors contributed equally to this work 42These authors contributed equally to this work *Correspondence: c.plass@dkfz.de (C.P.), nada.jabado@mcgill.ca (N.J.), s.pster@dkfz.de (S.M.P.) http://dx.doi.org/10.1016/j.ccr.2012.08.024

SUMMARY

Glioblastoma (GBM) is a brain tumor that carries a dismal prognosis and displays considerable heterogeneity. We have recently identied recurrent H3F3A mutations affecting two critical amino acids (K27 and G34) of histone H3.3 in one-third of pediatric GBM. Here, we show that each H3F3A mutation denes an epigenetic subgroup of GBM with a distinct global methylation pattern, and that they are mutually exclusive with IDH1 mutations, which characterize a third mutation-dened subgroup. Three further epigenetic subgroups were enriched for hallmark genetic events of adult GBM and/or established transcriptomic signatures. We also demonstrate that the two H3F3A mutations give rise to GBMs in separate anatomic compartments, with differential regulation of transcription factors OLIG1, OLIG2, and FOXG1, possibly reecting different cellular origins.

INTRODUCTION Glioblastoma (GBM; World Health Organization [WHO] grade IV), the most common primary brain tumor, carries a universally dismal prognosis in children and adults (Louis et al., 2007). With evidence emerging recently of age-specic molecular and genetic differences, it is now becoming apparent that pediatric GBM is largely biologically distinct from adult GBM. Based on similarities in recurrent genomic aberrations (Bax et al., 2010; McLendon et al., 2008; Paugh et al., 2010; Qu et al., 2010; Schiffman et al., 2010; Zarghooni et al., 2010), it was long thought that pediatric GBMs more closely resembled adult secondary GBM, which arise from a preceding lower-grade lesion. However, stepwise transformation from less-malignant gliomas into GBM rarely occurs in children (Broniscer et al., 2007). Furthermore, IDH1 or IDH2 mutations, which are found in up to 98% of adult secondary GBM, are very rare in childhood GBM (<10%) (Antonelli et al., 2010; Balss et al., 2008; De Carli et al., 2009; Paugh et al., 2010; Pollack et al., 2011; Schiffman et al., 2010; Setty et al., 2010; Yan et al., 2009). We recently identied two recurrent somatic mutations in the H3F3A gene, affecting highly conserved residues of its encoded protein, the replication-independent histone 3 variant H3.3, in one-third of pediatric GBMs (Schwartzentruber et al., 2012). Mutations in a protein complex comprised of H3.3 and ATRX/

DAXX were detected in 45% of cases, and were shown to be associated with TP53 mutations and alternative lengthening of telomeres (ALT). The H3.3 mutations result in amino acid substitutions at K27 or G34at or near residues targeted by key post-translational modications that regulate H3.3s activity in governing gene expression (Hyland et al., 2011)and were shown to be linked to distinct transcriptional proles (Schwartzentruber et al., 2012). Methylation of K27 and K36 is also disrupted by elevated levels of the onco-metabolite 2-hydroxyglutarate (2-HG) resulting from gain-of-function mutations in IDH1 (Chowdhury et al., 2011; Xu et al., 2011), which was previously shown to be associated with a distinct Glioma-CpG-Island Methylator Phenotype (G-CIMP) (Noushmehr et al., 2010). In the present study, we further investigate the heterogeneity of glioblastoma across the entire age spectrum, and elucidate the impact of H3F3A mutations on the GBM epigenome. RESULTS Integrated Molecular Classication of Glioblastoma We used an integrative approach based on epigenetic, copynumber, expression, and genetic analyses to investigate the heterogeneity of glioblastoma across all age groups. An overview of all GBM samples subjected to various analyses is given in Figure S1A available online.

Signicance GBM is the most common and also the most devastating brain tumor, with a 5-year survival rate below 10%. We present strong evidence that GBM can be subclassied into multiple groups, indistinguishable by histological appearance, but correlating with molecular-genetic factors as well as key clinical variables such as patient age and tumor location. We identied six epigenetic GBM subgroups displaying characteristic global DNA methylation patterns, harboring distinct hotspot mutations, DNA copy-number alterations, and transcriptomic patterns. These ndings may guide the identication of innovative subgroup-specic treatments, e.g., targeted epigenetic therapies for H3.3-mutated variants, and improve the design of future clinical trials. Our study enables classication of GBM across the entire age continuum into biologically meaningful subgroups carrying clinical implications.
426 Cancer Cell 22, 425437, October 16, 2012 2012 Elsevier Inc.

Cancer Cell
Epigenetic and Biological Subgroups of Glioblastoma

GBM samples (n = 210)


Methylated

Controls

0.5

Expression Methylation

TCGA subgroups
Fetal normal brain (n=4) Adult normal brain (n=2) M.SssI-DNA (n=2) WGA-DNA (n=2)

G-CIMP+ Cluster #2 Cluster #3 Proneural Neural Classical Mesenchymal

90

DNA methylation probes (n = 8,000)

Color scale (beta-values)

Unmethylated

60

Age
21 0

Methylation cluster TCGA Methylation TCGA Expression MUT IDH1 K27M G34R/V H3F3A WT NA TP53 Chr. 7 gain Chr. 10 loss CDKN2A del. EGFR ampl. PDGFRA ampl.

IDH

K27

G34

RTK I 'PDGFRA'

Mesenchymal

RTK II 'Classic'

p < 0.001 p < 0.001 p < 0.001 p < 0.001 p < 0.001 p < 0.001 p < 0.001 p < 0.001 p < 0.001 p < 0.001

Figure 1. Methylation Proling Reveals the Existence of Six Epigenetic GBM Subgroups
Heatmap of methylation levels in six GBM subgroups identied by unsupervised k-means consensus clustering, and control samples as indicated. Each row represents a probe; each column represents a sample. The level of DNA methylation (beta-value) is represented with a color scale as depicted. For each sample (n = 210), patient age, subgroup association, mutational status, and cytogenetic aberrations are indicated. See also Figure S1 and Tables S1 and S2.

We investigated a cohort of GBMs from children (n = 59) and adult patients (n = 77) for genome-wide DNA methylation patterns using the Illumina 450k methylation array, and complemented our data with unpublished proles of 74 adult GBM samples generated by The Cancer Genome Atlas (TCGA) Consortium (McLendon et al., 2008) (Table S1). Consensus clustering using the 8,000 most variant probes across the data set robustly identied six distinct DNA methylation clusters (Figures 1 and S1B). Based on correlations with mutational status, DNA copy-number aberrations, and gene expression signatures, as outlined below, we have labeled these subgroups IDH, K27, G34, RTK I (PDGFRA), Mesenchymal, and RTK II (Classic). A striking nding of this integrated analysis is that H3F3A K27 and G34 mutations were exclusively distributed to the K27 (18/18) and G34 (18/18) clusters, respectively (p < 0.001; Fishers exact test) (Figure 1). The IDH group contained 88% of IDH1mutated tumors (23/26) (p < 0.001) and displayed concerted,

Cytogenetics

Mutations

global hypermethylation (Figures 1, 2A, and 2B), thereby expanding the previously described link between IDH1 mutation and G-CIMP+ tumors to a pediatric setting (Noushmehr et al., 2010). In contrast, tumors in the G34 cluster specically showed widespread hypomethylation across the whole genome, and especially in nonpromoter regions, when compared with all other subgroups (Figures 2A and 2B). This suggests the existence of a more global version of a CpG hypomethylator phenotype (CHOP), as proposed for a small number of genes in gastric cancer (Kaneda et al., 2002). More detailed mapping of differentially methylated regions revealed that the hypomethylation observed in H3F3A G34-mutated tumors was particularly prominent at chromosome ends (Figures 2C and 2D), potentially linking subtelomeric demethylation to alternative lengthening of telomeres, which is most frequently observed in this subgroup (Schwartzentruber et al., 2012). Of note, all mutations in H3F3A and IDH1 were mutually exclusive (p < 0.001) (Figure 1). To further test this observation, we
Cancer Cell 22, 425437, October 16, 2012 2012 Elsevier Inc. 427

Cancer Cell
Epigenetic and Biological Subgroups of Glioblastoma

A
1.0

Overall
1.0

Promoter
1.0

Non-promoter
IDH

0.8

Hypomethylation

0.8

0.8

K27 G34 RTK I MES RTK II Control

Probabilty Fn(probes)

0.6

0.6

IDH K27 G34 RTK I MES

0.6

0.4

0.4

0.4

0.2

Hypermethylation

0.2

RTK II Control

0.2

0.0 0.0 0.2 0.4 0.6 0.8 1.0

0.0 0.0 0.2 0.4 0.6 0.8 1.0

0.0 0.0 0.2 0.4 0.6 0.8 1.0

DNA methylation (beta-value)

DNA methylation (beta-value)


0.75

DNA methylation (beta-value)

B
0.60

Mean methylation (mean betavalue)

0.55

***

***

***

**

0.35

***
0.50 0.30

***

***

***

0.70

***

***

***

0.65

0.60 0.45 0.25

0.55

*
0.40

***

* ***
0.20 RTK II RTK II Control Control 0.50

***

***

***

***
0.45 RTK II MES Control RTK II

0.35 K27 G34 MES IDH RTK I

K27

G34

K27

MES

G34

RTK I

C
15

D
0.02 10

Chromosome end (4Mb)

Number of probes (x 1,000)

Mean methylation (mean betavalue)

Fraction CpG Island

0.4
5

0.01

0.2 0

0.00

Normalized DNA Methylation

0 0.04

-0.01

0.00

-0.04 0.0e+00 5.0e+06 1.0e+07 1.5e+07 2.0e+07 2.5e+07 3.0e+07

IDH G34 K27 RTK I MES RTK II

-0.02

-0.03

***
-0.04
K27 IDH

***
G34

RTK I

***

***
RTK I

Distance to chromosome end [bp]

Figure 2. Global DNA Methylation Patterns in GBM Subgroups


(A) Distinct patterns of global DNA methylation in GBM subgroups as identied by consensus clustering. The empirical cumulative distribution function for DNA methylation levels (beta-values) is plotted individually for each subgroup. (B) Overall DNA methylation levels (mean beta-values) of individual GBM methylation subgroups. Signicant differences (***p < 0.001; **p < 0.01; *p < 0.05) to IDH and G34 subgroups are indicated. (C) Upper panel: Probe density in respect of distance to chromosome end. The fraction of probes located within CpG-Islands (red line) remains stable. Lower panel: Mean methylation value per subgroup within windows of 500kb, normalized to control samples. Individual samples are normalized by the mean overall methylation value. (D) Mean methylation value within 4 Mb to the chromosome end normalized to the mean overall methylation value and to control samples. Signicant differences (***p < 0.001) between subgroups compared to G34 tumors are indicated. MES, Mesenchymal. See also Figure S2.

428 Cancer Cell 22, 425437, October 16, 2012 2012 Elsevier Inc.

MES

IDH

IDH

***

Cancer Cell
Epigenetic and Biological Subgroups of Glioblastoma

extended the targeted sequencing analysis of H3F3A and IDH1 to include 460 GBM samples from patients covering a broad age range (Figure S1C; Table S2). Even in this expanded series, no co-occurring mutations in H3F3A and IDH1 were detected (p < 0.001), and the age distribution conrmed reported associations of certain mutations with GBM in children (H3F3A K27), adolescent patients (H3F3A G34), and young adult patients (IDH1) (Khuong-Quang et al., 2012; Schwartzentruber et al., 2012; Yan et al., 2009) (Figure S1C; Table S2). As we have shown, TP53 mutations largely overlap with H3F3A mutations in pediatric GBM (Schwartzentruber et al., 2012), similar to the association of TP53 and IDH1 mutations in adults (Yan et al., 2009). This observation also holds true in our larger cohort, with a high enrichment of TP53 mutations in the G34 (18/18), IDH (22/24), and K27 (13/18) clusters (p < 0.001) (Figure 1). Since pediatric GBMs have been shown to display a distinct spectrum of focal copy-number aberrations (CNAs) compared with their adult counterparts (Bax et al., 2010; Paugh et al., 2010; Qu et al., 2010), we integrated DNA methylation clusters with copy-number data derived from the methylation arrays (Figures 1 and S1D). Interestingly, PDGFRA amplication was signicantly more common in the RTK I PDGFRA cluster than any other subgroup (11/33; p < 0.001), hence our proposed name for this group. The RTK II Classic cluster demonstrated a very high frequency of whole chromosome 7 gain (50/56; p < 0.001) and whole chromosome 10 loss (56/56; p < 0.001), as well as frequent homozygous deletion of CDKN2A (35/56; p < 0.001) and amplication of EGFR (39/56; p < 0.001) (Figures 1 and S1D)all hallmark CNAs of adult GBM (Louis et al., 2007), as reected by the complete absence of pediatric patients in this cluster. Overall, tumors from the IDH, K27, and G34 clusters were mostly devoid of the detected CNAs associated with the other GBM subgroups (amplications of PDGFRA and EGFR, deletion of CDKN2A, chromosome 7 gain, and chromosome 10 loss) (Figure 1; Table S1), in keeping with a previously reported nding in G-CIMP+ tumors (Noushmehr et al., 2010). To additionally place the methylation subgroups proposed here into the context of previous GBM classication systems, we used the gene expression signature described by the TCGA to classify 122 of the above tumors with available transcriptome data into one of four gene expression subtypes: Proneural, Neural, Mesenchymal, and Classical (Verhaak et al., 2010) (Figure 1; Table S1). This further conrmed the prototypic nature of tumors in the RTK II Classic cluster, which was clearly enriched for Classical expression proles (p < 0.001). The RTK I PDGFRA cluster was highly enriched for Proneural expression (p = 0.01), further substantiating the previously reported association of PDGFRA amplication with this expression subtype (Verhaak et al., 2010). As expected, all tumors in the IDH cluster displayed Proneural expression patterns. Interestingly, the K27 cluster also showed a clear enrichment of tumors with a Proneural signature (p < 0.01), indicating that this expression subtype can be divided into subgroups harboring distinct genomic aberrations based on methylation proling and targeted gene sequencing. Mesenchymal gene expression was mostly restricted to one methylation subgroup (p < 0.001), that showed a much lower incidence of typical GBM-related CNAs, generally fewer copy-number changes, and no character-

istic point mutations. We therefore termed this methylation cluster, which displayed the largest similarity with normal brain methylation patterns, Mesenchymal. Copy-number aberrations in these samples were, however, observed at a similar amplitude as in other cases, indicating an absence of excess stromal contamination. Our nding of six GBM methylation clusters is different from a TCGA study using Illumina 27k arrays, which identied three methylation clusters in an adult GBM cohort (Noushmehr et al., 2010). Applying their signature to our data set, however, showed that two clusters (G-CIMP+ and Cluster #3) mapped almost exactly to two of our subgroups (IDH and RTK II Classic,, respectively, p < 0.001) (Figure 1). By adding pediatric cases to the study cohort, we demonstrate that TCGA methylation Cluster #2 can be further divided into four biologically distinct subgroups, dened by a clear enrichment for mutations (K27, G34), CNAs (PDGFRA), and/or gene expression signatures (Mesenchymal). The same DNA methylation clusters were apparent when restricting our analyses to the pediatric population, with the exception of the RTK II Classic cluster, which is not represented in the pediatric population (Figure S1E). Notably, by analyzing tumors from patients spanning a broad age spectrum, we further observed a clear age-dependent increase in overall DNA methylation levels (Figure S2A), even after adjusting our analysis to exclude tumors with age-related mutations in IDH1 or H3F3A (Figure S2B). GBM Subgroups Show Correlations with Clinicopathological Variables The DNA methylation clusters described here were closely associated with specic age groups, pointing toward the biological diversity of epigenetic GBM subgroups (Figure 1). While the K27 cluster predominantly consisted of childhood patients (median age 10.5 years, range 523 years), patients in the G34 cluster were found mostly around the threshold between the adolescent and adult populations (median age 18 years, range 942 years), as previously suggested (Schwartzentruber et al., 2012). The RTK I PDGFRA cluster also harbored a proportion of pediatric patients (median age 36 years, range 874 years), in line with reports of PDGFRA CNAs being more prevalent in childhood high-grade gliomas (Bax et al., 2010; Paugh et al., 2010; Qu et al., 2010). The Mesenchymal cluster displayed a widespread age distribution (median age 47, range 285 years). The IDH and RTK II Classic clusters were mostly comprised of younger adult (median age 40 years, range 1371 years) and older adult (median age 58, range 3681 years) patients, respectively, reecting the established differences in patient age between IDH1-mutated/G-CIMP+ and IDH1 WT adult GBM (Noushmehr et al., 2010; Yan et al., 2009). The epigenetic GBM subgroups identied here also showed mutation-specic patterns of tumor location in the central nervous system (Figure 3A). While K27-mutated tumors were predominantly seen in midline locations, e.g., thalamus, pons, and the spinal cord (21/25 cases with available data), tumors from all other subgroups almost exclusively arose in the cerebral hemispheres (86/92, p < 0.001). To further investigate this association of mutation type and location, we investigated the transcriptomic proles of H3F3A-mutated samples (n = 13). Gene signatures characteristic for K27 and G34 mutant GBMs were
Cancer Cell 22, 425437, October 16, 2012 2012 Elsevier Inc. 429

Cancer Cell
Epigenetic and Biological Subgroups of Glioblastoma

A
Frontal lobe

Basal ganglia

Parietal lobe

Frontal lobe
3
2

Parietal lobe

2 2

5
2

14
2

6
2

3
2

11
Thalamus

3
2

7 10 7
Occipital lobe Temporal lobe
3
2

n = 119 Pons IDH (36) K27 (27) G34 (22) RTK I 'PDGFRA' (15) Mesenchymal (10) RTK II 'Classic' (9)

Cerebellum Brainstem

Occipital lobe

= 1 Case
2

5
Spinal cord

= 5 Cases

B
1 0.8 0.6 0.4 0.2 0 censored IDH (10) K27 (14) G34 (12) RTK I 'PDGFRA' (23) Mesenchymal (50) RTK II 'Classic' (43) n = 152

C
1 0.8 0.6 0.4 p<0.01 0.2 0 p=0.05 p<0.001 censored IDH1 MUT (47) H3F3A K27 (24) H3F3A G34 (13) H3F3A/IDH1 WT (199) n = 283

Survival probabiliy

12

24

36

48

60

72

84

96

108

120

12

24

36

48

60

72

84

96

108

120

Overall survival (months)

Overall survival (months)

Figure 3. Epigenetic Subgroups of GBM Correlate with Distinct Clinical Characteristics


(A) Location of 119 GBMs in the human central nervous system grouped by methylation clusters. The number of cases in each group is indicated within the circles. Circles without numbers represent single cases. Different colors indicate methylation cluster afliation. Tumors occurring in midline locations are depicted in the sagittal view (left panel), tumors occurring in the cerebral and cerebellar hemispheres are depicted in the exterior view (right panel). (B and C) Kaplan-Meier survival curves for GBM subgroups dened by methylation proling (B), and mutation analysis (C). The p values were computed by log rank tests between subgroups. See also Figure S3.

applied to a published series of 1,340 transcriptomic proles representing multiple regions of the developing and adult human brain (Kang et al., 2011; Figure S3). The G34 mutant signature appeared to be most strongly expressed in early embryonic regions and early- to mid-fetal stages of neocortex and striatum development. In contrast, the K27 signature most closely matched with mid- to late-fetal stages of striatum and thalamus development. Thus, G34 and K27 mutant GBMs seem to show expression patterns of early developmental stages correlating with their subsequent tumor location, possibly indicating different cellular origins and/or time of tumor initiation for these two subgroups.
430 Cancer Cell 22, 425437, October 16, 2012 2012 Elsevier Inc.

Correlating our proposed methylation clusters with patient survival indicated differences between mutation-dened subgroups, but this was somewhat restricted by the low number of patients with available survival data in each subgroup (Figure 3B). We therefore enlarged our survival analysis to include all tumors with known H3F3A and IDH1 mutation status (Figure 3C). As expected, patients with IDH1 mutant tumors had a signicantly longer overall survival (OS) than patients with H3F3A and IDH1 WT tumors (p < 0.001) (Noushmehr et al., 2010; Parsons et al., 2008; Yan et al., 2009). Notably, G34 mutant GBM patients also showed a trend toward a better OS than WT tumor patients, with marginal statistical signicance (p = 0.05). In

Cancer Cell
Epigenetic and Biological Subgroups of Glioblastoma

Fold change >=0 Log10 (pvalue, BH), fold change <0 log10 (pvalue, BH)

A
Significance (-Log10 p-value)

OLIG1 OLIG2

B
Significance (-Log10 p-value)

C
FOXG1 OLIG1

OLIG1

53
OLIG2

39

OLIG2

4 475 2

FOXG1

Gene expression

87 2

254

710

-4 14 -5 0
FOXG1 10

0 0.6 0.4 0.2 0.0 0.2 0.4 0.6

0 -10 -5 0 5 10

DNA methylation difference (beta-values)

Gene expression difference (Log2 fold change)

DNA methylation
Difference >=0 Log10 (pvalue, BH), Difference <0 Log10 (pvalue, BH)

D
Chromosome 21 RTK I G34 K27 IDH 1.0

E
OLIG1 expression (Log2) 12 10 8 6 4

F
12

***

***

***

**
OLIG2 expression (Log2)
10

IDH K27 G34 RTK I MES RTK II

r= 0.599
8

Mesenchymal

0.5

OLIG2 expression (Log2)

12 10 8

***

***

***

**

6 4 RTK I RTK II IDH K27 MES G34


0.0 0.2 0.4 0.6 0.8 1.0

RTK II

0 OLIG1 OLIG2

DNA methylation (beta-value)

Figure 4. Identication of Marker Genes Affected by Differential Methylation and Expression in GBM Subgroups
(A and B) Volcano plots illustrating differences in DNA methylation (A) and gene expression (B) between tumors from the K27 and G34 subgroups. Difference in betavalues (A) and Log2 fold change in gene expression values (B) are plotted on the x axis, adjusted p values calculated using the SAM method are plotted on the y axis. (C) Starburst plot integrating DNA methylation (x axis) and gene expression (y axis) data. (D) Methylation levels at the OLIG1 and OLIG2 loci across all 210 GBM samples investigated. Each row represents one sample; each vertical bar represents one CpG-site. Light blue bars indicate promoter regions. Methylation levels are represented by a color scale as indicated. (E) Mean gene expression levels of OLIG1 (upper panel) and OLIG2 (lower panel) across GBM subgroups (n = 48). Signicant differences (***p < 0.001; **p < 0.01; *p < 0.05) between subgroups compared to G34 tumors are indicated. (F) Inverse correlation of promoter methylation (x axis) and gene expression (y axis) of OLIG2 across GBM methylation subgroups (n = 48; Pearsons correlation coefcient, r). MES, Mesenchymal. See also Figure S4 and Table S3.

contrast, patients with K27 mutations tended toward an even shorter OS than patients with WT tumors, although this did not reach statistical signicance (p = 0.12). Comparing the two H3F3A mutations, patients harboring G34-mutated tumors clearly had a longer OS than patients with tumors carrying the K27 mutation (p < 0.01). While this association may be partly linked to G34-mutated tumors being more accessible to surgery than the midline K27-mutated tumors, the better prognosis of G34 versus K27 was independent of location for those cases where both mutation type and tumor site information were available (p = 0.02; HR = 0.20, 95% CI = 0.050.77; Cox proportional hazards model). Integrating Methylome and Transcriptome Data Identies Marker Genes of GBM Subgroups A combined analysis of DNA methylation and gene expression data was used to identify subgroup-specic differentially regu-

lated genes (Figures 4A4C and S4A4C; Table S3). This analysis revealed Oligodendrocyte Lineage Genes 1 and 2 (OLIG1 and OLIG2) and the neural development gene FOXG1 as top candidates for further analysis in H3F3A-mutated GBMs (Figure 4A4C). DNA hypermethylation across the OLIG1 and OLIG2 loci occurred exclusively in G34-mutated tumors, which concurrently displayed signicantly lower OLIG1 and OLIG2 gene expression (Figure 4D4F). Interestingly, this pattern closely mimics that of embryonic stem cells, where epigenetic inactivation of OLIG1 and OLIG2 has been proposed as a mechanism to prevent neural lineage commitment (Meissner et al., 2008). Expression of FOXG1 was signicantly lower in K27mutated tumors than in all other subgroups, accompanied by higher levels of promoter methylation (Figures S4A, S4D, and S4E). This comparative analysis also further supported our suggestion of a CHOP-like phenotype in G34 tumors, as most of the differentially methylated genes were found to be
Cancer Cell 22, 425437, October 16, 2012 2012 Elsevier Inc. 431

Cancer Cell
Epigenetic and Biological Subgroups of Glioblastoma

7%

4%

IDH1 (2/2)

B
IDH1R132H OLIG2 / FOXG1
+ + -

OLIG2

FOXG1

15%
G34 (6/6)

48%
WT (9/9)

OLIG2 / FOXG1
26%
K27 (8/8)

OLIG2 / FOXG1
+

OLIG2 / FOXG1 n=143

C 100%
80% 60% 40% 20% 6 26

3 14

2 8 52 17

100% 80% 60% 40% 20% 4 OLIG2- / FOXG1+ OLIG2+ / FOXG10% 3 OLIG2- / FOXG1OLIG2+ / FOXG1+ 9 23 13 7 18 51

ATRX

ALT

OLIG2- / FOXG1+

OLIG2+ / FOXG1-

OLIG2- / FOXG1-

OLIG2+ / FOXG1+

0%

ATRX+

ATRX

ALT+

ALT

5 m

10 m

Figure 5. Identication of H3F3A-Mutated GBMs by Differential Protein Expression Patterns


(A) Classication of 143 pediatric GBMs according to protein expression of OLIG2, FOXG1, and mutated IDH1 (IDH1R132H). Numbers in brackets indicate samples with known H3F3A and IDH1 mutation status as predicted by immunohistochemistry and veried by targeted gene sequencing, respectively. (B) Typical pattern of OLIG2/FOXG1+ cells with concomitant loss of ATRX protein expression and ALT as observed in G34-mutated GBMs. Insets show contrasting staining results for comparison. Scale bars represent 100 mm unless indicated differently. (C and D) Correlation of GBMs as classied in (A) with ATRX loss (C), and ALT (D). See also Figure S5 and Table S4.

hypomethylated (1653/1946, 85%, Figure S4B) in this subgroup, in contrast to the hypermethylator G-CIMP pattern observed in the IDH subgroup (Figure S4C). Hypermethylation and concurrent downregulation of TP73 antisense RNA 1 (TP73-AS1) was identied as a unique characteristic of this IDH/G-CIMP+ cluster (Figures S4D and S4F). Interestingly, inactivation of this gene by promoter methylation has been reported as a common mechanism in a high proportion of oligodendrogliomas, 80% of which are also known to harbor IDH1 mutations (Pang et al., 2010). Immunohistochemical Analysis Correctly Subclassies Mutation-Dened GBM Subgroups In an attempt to subgroup GBM samples based on differential protein expressiona method which is likely to be more suitable for possible clinical applicationwe used commercially available antibodies against OLIG2, FOXG1, and mutated IDH1 (R132H) to stain a tissue-microarray (TMA) with cores from 143 pediatric GBMs, and classied tumors according to their protein expression patterns (Figures 5A and 5B; Table S4). The resulting fractions of tumors with predicted mutations in IDH1 (IDH1R132H, n = 6) and H3F3A (OLIG2+/FOXG1 for K27, n = 37, and OLIG2/ FOXG1+ for G34, n = 21) were consistent with the frequency of each mutation in the pediatric population as detected by targeted gene sequencing (Figure S1C). Our approach correctly classied GBMs with known H3F3A and IDH1 mutation status, and revealed a frequent association between OLIG2/FOXG1+ tumors (assumed to be G34-mutated), loss of ATRX protein expression, and an ALT phenotype (Figures 5BD), as previously reported for H3F3A G34-mutated tumors (Schwartzentruber et al., 2012). The putative H3F3A mutant groups also did not overlap with tumors harboring IDH1 (R132H) mutations, and only one case with EGFR amplication and homozygous
432 Cancer Cell 22, 425437, October 16, 2012 2012 Elsevier Inc.

CDKN2A deletion was detected therein (Figure S5A). The correlation with clinicopathological variables, such as tumor location and patient survival, also reected our ndings from the arraybased analysis (Figures S5B and S5C). Of note, rare tumors represented on the TMA occurring in the basal ganglia and the spinal cord were almost always found in the OLIG2+/FOXG1 subgroup (and therefore predicted to harbor the H3F3A K27 mutation), further strengthening our hypothesis of the H3F3A K27 mutation as a unifying characteristic of midline GBM. DISCUSSION We have identied six biological subgroups of GBM based on global DNA methylation patterns, which correlate with specic molecular-genetic alterations and key clinical parameters. Our ndings suggest that at least 30%40% of pediatric/young adult GBMs are likely characterized by disrupted epigenetic regulatory mechanisms, associated with recurrent and mutually exclusive mutations in either H3F3A or IDH1, and aberrant DNA methylation patterns. Placing these subgroups into the context of previous molecular GBM classication schemes described by the TCGA (Noushmehr et al., 2010; Verhaak et al., 2010) revealed a clear correlation with DNA methylation clusters and a corresponding enrichment for previously established expression signatures in different epigenetic subgroups. We have also demonstrated that our proposed classication can rene that described by the TCGA for adult GBM, to give a stratication system that is applicable across all ages, and denes additional biologically meaningful subgroups. A simplied graphical summary of the key molecular and biological characteristics of the GBM subgroups as identied by our integrated classication strategy is given in Figure 6.

Cancer Cell
Epigenetic and Biological Subgroups of Glioblastoma

Epigenetic and Biological Subgroups of Glioblastoma


IDH
IDH1mut

K27
H3F3Amut K27

G34
H3F3Amut G34

RTK I

MESENCHYMAL

RTK II
EGFR ampl. 7+ 10CDKN2A del.

PDGFRA ampl.

Mutations / Cytogenetics
TP53mut
CIMP+

TP53mut

TP53mut
CHOP+

CDKN2A del.

CNVlow

DNA Methylation
-0.6 0 0.6 -0.6 0 0.6 -0.6 0 0.6 -0.6 0 0.6 -0.6 0 0.6 -0.6 0 0.6

Gene Expression IHC Protein Marker

Proneural

Proneural

Mixed

Proneural

Mesenchymal

Classical

IDH1R132H

OLIG2+/FOXG1-

OLIG2-/FOXG1+ OLIG2+/FOXG1+

OLIG2+/FOXG1+ OLIG2+/FOXG1+

Age Distribution (years)


0 30 60 90 0 30 60 90 0 30 60 90 0 30 60 90 0 30 60 90 0 30 60 90

Tumor Location

Patient Survival (months)


0 60 120 0 60 120 0 60 120 0 60 120 0 60 120 0 60 120

Figure 6. Graphical Summary of Key Molecular and Biological Characteristics of GBM Subgroups
Simplied schematic representation of key genetic and epigenetic ndings in six GBM subgroups as identied by methylation proling and correlations with clinical patient data.

We and others have recently described a high frequency of H3F3A K27 mutations in thalamic GBMs and in diffuse intrinsic pontine gliomas (DIPGs), suggesting that the latter likely represent an anatomically-dened subset of K27 mutant GBM (Khuong-Quang et al., 2012; Schwartzentruber et al., 2012; Wu et al., 2012). We now extend this observation to a larger subgroup of GBM, characterized by the K27M mutation, which almost exclusively occurs in midline locations, including rare tumors in the basal ganglia and the spinal cord. This is in line with a recent study by Puget et al. (2012) in which gene expression patterns of brainstem gliomas were found to resemble midline/thalamic tumors, indicating a closely related origin. The K27 subgroup also displays markedly lower expression of the ventral telencephalic marker FOXG1 than other subgroups. Conversely, non-K27 tumors were restricted to hemispheric

locations, further underlining the biological divergence of epigenetic GBM subgroups. While recurrent focal amplication of PDGFRA has been suggested as a key oncogenic event in pediatric DIPGs in some studies (Paugh et al., 2011; Puget et al., 2012; Zarghooni et al., 2010), midline-associated tumors in the K27 or OLIG2+/FOXG1 subgroups (including ten brainstem gliomas with known PDGFRA copy-number status) lacked this common feature in our series. PDGFRA amplication was, however, enriched in a subgroup of supratentorial hemispheric GBMs. In part, this discrepancy may be explained by the use of autopsy (and therefore post radio/chemotherapy) material in previous study cohorts of DIPGs, which might have been confounded by the higher incidence of PDGFRA amplications observed in radiation-induced gliomas (Paugh et al., 2010). Nevertheless, amplications of PDGFRA have also been
Cancer Cell 22, 425437, October 16, 2012 2012 Elsevier Inc. 433

Cancer Cell
Epigenetic and Biological Subgroups of Glioblastoma

detected in small numbers of pretreatment samples (Paugh et al., 2011; Puget et al., 2012; Zarghooni et al., 2010), and post-treatment samples were not found to show increased widespread genomic instability (Paugh et al., 2011). This particularly clinically challenging subset of tumors clearly warrants further investigation, underlining the importance of routine stereotactic biopsy of DIPGs at the time of primary diagnosis. OLIG2 has previously been reported as a universal marker for diffuse gliomas (Ligon et al., 2004), and OLIG2-positive progenitor-like cells of the subventricular zone have been suggested as potential glioma-initiating cells (Wang et al., 2009). There is also evidence that OLIG2-mediated modication of p53 function is required for complete inactivation of the latter in malignant gliomas, which typically show indirect loss of p53 activity through MDM2 amplication or p14ARF deletion (Mehta et al., 2011). Here, we describe a distinct subgroup of GBM, harboring the H3F3A G34 mutation, in which OLIG1 and OLIG2 protein expression is absent. Given the 100% mutation frequency of TP53 in this subgroup, this may indicate a different pathogenesis of G34-mutated GBM, in which direct inactivation of p53 is required rather than via an OLIG2dependent mechanism. The previously reported association of H3F3A mutations, particular the G34 mutation, with loss of ATRX and ALT (Schwartzentruber et al., 2012) is further expanded upon here. Interestingly, the global CHOP that we observed in G34 mutants was particularly pronounced in subtelomeric regions, suggesting a possible mechanistic link with ALT in these tumors (Gonzalo et al., 2006). Whether this is a more general phenotype that can be observed in clinically and etiologically distinct subgroups of other human cancers, remains to be investigated. The close link between H3F3A mutation type, tumor location, and differential expression of key neuronal lineage markers leads us to speculate that there may be differences in the cell of origin and/or the time of tumor development between these GBM subgroups. Although supported by the differential expression of mutant-specic gene signatures at different stages of human brain development, this remains to be formally shown. Also requiring further validation in larger, prospective cohorts is the association of the G34 mutation with better overall survival compared with H3F3A and IDH1 WT tumors, and that of K27 mutation with potentially poorer prognosis, as observed in our series and a recent cohort of pediatric DIPGs (Khuong-Quang et al., 2012). Given the location of the H3F3A mutations at or near critical regulatory histone residues, and their distinct methylation proles, we consider it likely that the H3.3 mutations are directly involved in producing widespread aberrant DNA methylation and deregulation of gene expression. This has recently been shown for IDH1 mutations, which alone are sufcient to induce the global epigenetic reprogramming of the G-CIMP phenotype in normal astrocytes (Turcan et al., 2012). Overproduction of the oncometabolite 2-hydroxyglutarate in IDH1-mutated cells inhibits the TET family of 5-methlycytosine hydroxylases and H3K27-specic demethylases. This is thought to lead to decreased 5-hydroxylmethycytosine and increased H3K27 methylation (Xu et al., 2011), resulting in aberrant DNA and histone methylation, and a block to differentiation (Christensen et al., 2011; Dang et al., 2009; Lu et al., 2012).
434 Cancer Cell 22, 425437, October 16, 2012 2012 Elsevier Inc.

Seminal studies have shown that DNA methylation patterns are tightly linked to histone 3 lysine K27 trimethylation (H3K27me3) patterns (Brinkman et al., 2012; Statham et al., 2012), and in high-CpG-density promoters, loss of H3K4me3 and retention of H3K4me2 or H3K27me3 is correlated with an increase in DNA methylation (Meissner et al., 2008). Therefore, mutations affecting H3K27 methylation are likely to affect DNA methylation. In addition, mutations in ATRX have been shown to give rise to changes in the patterns of DNA methylation of several highly repeated sequences, which further supports the link between chromatin remodeling machinery and DNA methylation (Gibbons et al., 2000). Based on our current knowledge, the incorporation of H3.3 variants into the genome, and the subsequent effects on gene regulation, involve the H3.3 chaperone complex (including ATRX and DAXX) in a replication independent et al., 2010; Goldberg et al., 2010) but other manner (Drane factors also likely play a role. The exact mechanism by which the H3F3A mutations might be inducing epigenetic reprogramming requires further elucidation. In conclusion, this study describes a number of ndings that enhance our understanding of the heterogeneity of GBM, as well as shedding light on potential cellular origins and oncogenic pathways leading to gliomagenesis. We have identied potential prognostic biomarkers, which may be further exploited for molecular diagnostic purposes, and also provided a focus for future work at a basic and translational/targeted therapeutic level, particularly in a pediatric and young adult setting.
EXPERIMENTAL PROCEDURES Patients and Tumor Samples Primary tumor samples for methylation (n = 136; Table S1), mutation (n = 460; Table S2), and gene expression (n = 69) analysis and all clinical data were collected at the DKFZ (Heidelberg, Germany) and at McGill University (Montreal, Canada). Parafn-embedded samples (n = 143; Table S4) for TMA analysis were collected from the Burdenko Neurosurgical Institute (Moscow, rzburg Russia) and from the Department of Neuropathology, University of Wu (Germany). Patient clinical details can be found in Table S1 for the methylation analysis data set and in Table S4 for the TMA cohort. All of the tumors were banked at the time of primary diagnosis between 1994 and 2011 in accordance with research ethics board approval from the respective institutes. Informed consent was obtained from all patients included in this study. An overview of all samples included in different data collections is given in Figure S1A. All of the samples were independently reviewed by senior pediatric neuropathologists (S.A. and A.K.) according to the WHO guidelines. Detailed information about samples provided by TCGA can be found elsewhere (http:// cancergenome.nih.gov). DNA Methylation Proling For genome-wide assessment of DNA methylation GBM samples (n = 136) and controls (n = 10; four fetal and two adult samples of non-neoplastic cerebellum; two samples of Whole-Genome Amplied DNA (unmethylated control; two samples of M.SssI-treated DNA [100% methylated control]) were arrayed using the Illumina HumanMethylation450 BeadChip according to the manufacturers instructions at the DKFZ. Methylation data of additional adult glioblastoma samples (n = 74) were obtained from the TCGA website (https:// tcga-data.nci.nih.gov; available data from TCGA batches 79 and 111). The following ltering criteria were applied: Removal of probes targeting the X and Y chromosomes (n = 11,551), removal of probes containing a singlenucleotide polymorphism (dbSNP132 Common) within ve base pairs of and including the targeted CpG-site (n = 24,536), and probes not mapping uniquely to the human reference genome (hg19) allowing for one mismatch (n = 9,993). In total, 438,370 probes were kept for analysis.

Cancer Cell
Epigenetic and Biological Subgroups of Glioblastoma

For a subset of differentially methylated genes from the 450k array, MassARRAY technology (Sequenom, San Diego, CA, USA) was used to validate our results, allowing us to compare DNA methylation levels at 29 individual CpG-sites investigated by both techniques. DNA methylation measurements of those 29 CpG dinucleotides were highly correlated (median Pearsons correlation: r = 0.96; range: 0.711.00). Gene Expression Proling Glioblastoma samples for which RNA of sufcient quantity and quality was available (n = 69) were analyzed on the Affymetrix GeneChip Human Genome U133 Plus 2.0 Array at the Microarray Department of the University of Amsterdam, the Netherlands. Sample library preparation, hybridization, and quality control were performed according to the manufacturers protocols. Expression data were normalized using the MAS5.0 algorithm of the GCOS program (Affymetrix Inc). Gene expression data of additional adult glioblastoma samples (n = 74) were obtained from the TCGA website (https://tcga-data. nci.nih.gov; available data from TCGA batches 79 and 111). Predictive analysis of microarrays was used to assign TCGA methylation and gene expression subgroups to each of the samples in the present study. Detection of CNAs Copy-number aberrations were detected from the 450k Innium methylation array in a custom approach using the sum of both methylated and unmethylated signals (Figure S1D). For the detection of EGFR and PDGFRA high-level amplications, homozygous CDKN2A deletions, and CNAs affecting chromosomes 7 and 10 (as depicted in Figure 1), automatic scoring was veried by manual curation of the respective loci for each individual prole, and compared with results obtained from SNP proling and uorescence in situ hybridization (FISH) analysis where available. The three methodologies showed very high concordance. Statistical Analysis and Measurement of Differential DNA Methylation and Gene Expression For unsupervised consensus clustering we used the 8,000 most variable methylated probes (by standard deviation) across the data set (R package: clusterCons) (Monti et al., 2003; Wilkerson and Hayes, 2010). The consensus matrix was calculated using the k-means algorithm (10 random starting sets, maximum of 1,000 iterations) on a fraction of probes (0.8) in 1,000 iterations. The signicance analysis of microarrays (SAM) method was used to identify genes that are differentially methylated or differentially expressed between subgroups. Correction for multiple testing was performed using the Benjamini-Hochberg method. Genes were considered signicantly differentially methylated/expressed between two subgroups when displaying an adjusted p value < 0.01 and a methylation difference of 0.2 or a 2-fold change in expression. Statistical Analysis of Clinical and Molecular Data Kaplan-Meier analysis was performed to estimate the survival time of different GBM subgroups and a log rank test was used to test for differences of more than one survival curve. Comparisons of binary and categorical patient characteristics between subgroups were performed by the use of a two-sided Fishers exact test. An unpaired t test was used to test for differences between the mean values for continuous variables in GBM subgroups. Immunohistochemistry and FISH Hematoxylin and eosin stained sections from all 143 parafn blocks were prepared to dene representative tumor regions for inclusion in the TMA. Antibodies against the following antigens were applied: OLIG2 (Millipore, Billerica, MA, USA; AB9610; dilution 1:250), FOXG1 (Abcam, Cambridge, UK; ab18259; dilution 1:50), ATRX (Sigma, HPA001906; dilution 1:750), and mutated IDH1 (R132H; (Capper et al., 2010; dianova, DIA H09). Multicolor interphase FISH analysis for PDGFRA, EGFR, and CDKN2A was performed as described (Pster et al., 2009). Telomere-specic FISH was done using a standard formalin-xed parafn-embedded FISH protocol (Heaphy et al., 2011) using a FITC peptide nucleic acid telomere probe from Dako (Glostrup, Denmark). Genomic Sequencing Targeted sequencing of H3F3A (rst coding exon), IDH1 (exon 4), and TP53 (all exons) was performed by QIAGEN (Hilden, Germany) in both forward and

reverse directions using puried PCR products. PCR procedures were as previously described (Pfaff et al., 2010). Primer sequences are available upon request. ACCESSION NUMBERS The complete CpG methylation values have been deposited in NCBIs Gene Expression Omnibus (GEO; http://www.ncbi.nlm.nih.gov/geo) and are accessible through GEO Series accession number GSE36278. The complete gene expression values are accessible through GEO Series accession numbers GSE36245 and, as part of a previously reported series, GSE34824 (Schwartzentruber et al., 2012). SUPPLEMENTAL INFORMATION Supplemental Information includes ve gures, four tables, and Supplemental Experimental Procedures and can be found with this article online at http://dx. doi.org/10.1016/j.ccr.2012.08.024. ACKNOWLEDGMENTS We thank Andrea Wittmann and Laura Sieber from the Division of Pediatric Neurooncology at the DKFZ for excellent technical support and Matthias Schick and Roger Fischer from the DKFZ Genomics and Proteomics Core Facility for performing the microarray analyses to a very high standard. We also acknowledge the outstanding technical assistance of Hannelore Schraut rzburg) and Leonore Senf (Nu rnberg Childrens Hospital). The (University of Wu project was supported by grants from the German Cancer Aid (109252 and 108456); the BMBF (to A.K., B.R., O.W., P.L., S.M.P., G.R., and J.F.; ICGC PedBrain, NGFNPlus #01GS0883); Koningin Wilhelmina Fonds (UvA-20104713) and KIKA (to M.K.); the Cole Foundation; the Canadian Institute of Health nome Canada and Ge nome Research; the Institute of Cancer Research, Ge ller Award for Neuro-Oncology to S.M.P. Quebec (to N.J.); and the Alfred-Mu Received: February 22, 2012 Revised: June 3, 2012 Accepted: August 24, 2012 Published: October 15, 2012 REFERENCES Antonelli, M., Buttarelli, F.R., Arcella, A., Nobusawa, S., Donofrio, V., Oghaki, H., and Giangaspero, F. (2010). Prognostic signicance of histological grading, p53 status, YKL-40 expression, and IDH1 mutations in pediatric high-grade gliomas. J. Neurooncol. 99, 209215. Balss, J., Meyer, J., Mueller, W., Korshunov, A., Hartmann, C., and von Deimling, A. (2008). Analysis of the IDH1 codon 132 mutation in brain tumors. Acta Neuropathol. 116, 597602. Bax, D.A., Mackay, A., Little, S.E., Carvalho, D., Viana-Pereira, M., Tamber, N., Grigoriadis, A.E., Ashworth, A., Reis, R.M., Ellison, D.W., et al. (2010). A distinct spectrum of copy number aberrations in pediatric high-grade gliomas. Clin. Cancer Res. 16, 33683377. Brinkman, A.B., Gu, H., Bartels, S.J., Zhang, Y., Matarese, F., Simmer, F., Marks, H., Bock, C., Gnirke, A., Meissner, A., and Stunnenberg, H.G. (2012). Sequential ChIP-bisulte sequencing enables direct genome-scale investigation of chromatin and DNA methylation cross-talk. Genome Res. 22, 11281138. Broniscer, A., Baker, S.J., West, A.N., Fraser, M.M., Proko, E., Kocak, M., Dalton, J., Zambetti, G.P., Ellison, D.W., Kun, L.E., et al. (2007). Clinical and molecular characteristics of malignant transformation of low-grade glioma in children. J. Clin. Oncol. 25, 682689. ger, D., Ackermann, Capper, D., Weissert, S., Balss, J., Habel, A., Meyer, J., Ja U., Tessmer, C., Korshunov, A., Zentgraf, H., et al. (2010). Characterization of R132H mutation-specic IDH1 antibody binding in brain tumors. Brain Pathol. 20, 245254.

Cancer Cell 22, 425437, October 16, 2012 2012 Elsevier Inc. 435

Cancer Cell
Epigenetic and Biological Subgroups of Glioblastoma

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. Christensen, B.C., Smith, A.A., Zheng, S., Koestler, D.C., Houseman, E.A., Marsit, C.J., Wiemels, J.L., Nelson, H.H., Karagas, M.R., Wrensch, M.R., et al. (2011). DNA methylation, isocitrate dehydrogenase mutation, and survival in glioma. J. Natl. Cancer Inst. 103, 143153. 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. De Carli, E., Wang, X., and Puget, S. (2009). IDH1 and IDH2 mutations in gliomas. N. Engl. J. Med. 360, 2248. , P., Ouararhni, K., Depaux, A., Shuaib, M., and Hamiche, A. (2010). The Drane death-associated protein DAXX is a novel histone chaperone involved in the replication-independent deposition of H3.3. Genes Dev. 24, 12531265. Gibbons, R.J., McDowell, T.L., Raman, S., ORourke, D.M., Garrick, D., Ayyub, H., and Higgs, D.R. (2000). Mutations in ATRX, encoding a SWI/SNF-like protein, cause diverse changes in the pattern of DNA methylation. Nat. Genet. 24, 368371. Goldberg, A.D., Banaszynski, L.A., Noh, K.M., Lewis, P.W., Elsaesser, S.J., Stadler, S., Dewell, S., Law, M., Guo, X., Li, X., et al. (2010). Distinct factors control histone variant H3.3 localization at specic genomic regions. Cell 140, 678691. 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. 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. 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. Kaneda, A., Kaminishi, M., Yanagihara, K., Sugimura, T., and Ushijima, T. (2002). Identication of silencing of nine genes in human gastric cancers. Cancer Res. 62, 66456650. Kang, H.J., Kawasawa, Y.I., Cheng, F., Zhu, Y., Xu, X., Li, M., Sousa, A.M., Pletikos, M., Meyer, K.A., Sedmak, G., et al. (2011). Spatio-temporal transcriptome of the human brain. Nature 478, 483489. Khuong-Quang, D.A., Buczkowicz, P., Rakopoulos, P., Liu, X.Y., Fontebasso, A.M., Bouffet, E., Bartels, U., Albrecht, S., Schwartzentruber, J., Letourneau, L., et al. (2012). K27M mutation in histone H3.3 denes clinically and biologically distinct subgroups of pediatric diffuse intrinsic pontine gliomas. Acta Neuropathol. 124, 439447. Ligon, K.L., Alberta, J.A., Kho, A.T., Weiss, J., Kwaan, M.R., Nutt, C.L., Louis, D.N., Stiles, C.D., and Rowitch, D.H. (2004). The oligodendroglial lineage marker OLIG2 is universally expressed in diffuse gliomas. J. Neuropathol. Exp. Neurol. 63, 499509. Louis, D.N., Ohgaki, H., Wiestler, O.D., and Cavenee, W.K. (2007). WHO Classication of Tumors of the Central Nervous System (Lyon: IARC). 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. McLendon, R., Friedman, A., Bigner, D., Van Meir, E.G., Brat, D.J., Mastrogianakis, M., Olson, J.J., Mikkelsen, T., Lehman, N., Aldape, K., et al; Cancer Genome Atlas Research Network. (2008). Comprehensive genomic characterization denes human glioblastoma genes and core pathways. Nature 455, 10611068. Mehta, S., Huillard, E., Kesari, S., Maire, C.L., Golebiowski, D., Harrington, E.P., Alberta, J.A., Kane, M.F., Theisen, M., Ligon, K.L., et al. (2011). The central nervous system-restricted transcription factor Olig2 opposes p53

responses to genotoxic damage in neural progenitors and malignant glioma. Cancer Cell 19, 359371. Meissner, A., Mikkelsen, T.S., Gu, H., Wernig, M., Hanna, J., Sivachenko, A., Zhang, X., Bernstein, B.E., Nusbaum, C., Jaffe, D.B., et al. (2008). Genomescale DNA methylation maps of pluripotent and differentiated cells. Nature 454, 766770. Monti, S., Tamayo, P., Mesirov, J., and Golub, T. (2003). Consensus clustering: A resampling-based method for class discovery and visualization of gene expression microarray data. Mach. Learn. 52, 91118. 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. Pang, J.C., Li, K.K., Lau, K.M., Ng, Y.L., Wong, J., Chung, N.Y., Li, H.M., Chui, Y.L., Lui, V.W., Chen, Z.P., et al. (2010). KIAA0495/PDAM is frequently downregulated in oligodendroglial tumors and its knockdown by siRNA induces cisplatin resistance in glioma cells. Brain Pathol. 20, 10211032. 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. Paugh, B.S., Qu, C., Jones, C., Liu, Z., Adamowicz-Brice, M., Zhang, J., Bax, D.A., Coyle, B., Barrow, J., Hargrave, D., et al. (2010). Integrated molecular genetic proling of pediatric high-grade gliomas reveals key differences with the adult disease. J. Clin. Oncol. 28, 30613068. Paugh, B.S., Broniscer, A., Qu, C., Miller, C.P., Zhang, J., Tatevossian, R.G., Olson, J.M., Geyer, J.R., Chi, S.N., da Silva, N.S., et al. (2011). Genomewide analyses identify recurrent amplications of receptor tyrosine kinases and cell-cycle regulatory genes in diffuse intrinsic pontine glioma. J. Clin. Oncol. 29, 39994006. Pfaff, E., Remke, M., Sturm, D., Benner, A., Witt, H., Milde, T., von Bueren, ttler, A., Jorch, N., et al. (2010). TP53 mutation is A.O., Wittmann, A., Scho frequently associated with CTNNB1 mutation or MYCN amplication and is compatible with long-term survival in medulloblastoma. J. Clin. Oncol. 28, 51885196. Pster, S., Remke, M., Benner, A., Mendrzyk, F., Toedt, G., Felsberg, J., Wittmann, A., Devens, F., Gerber, N.U., Joos, S., et al. (2009). Outcome prediction in pediatric medulloblastoma based on DNA copy-number aberrations of chromosomes 6q and 17q and the MYC and MYCN loci. J. Clin. Oncol. 27, 16271636. Pollack, I.F., Hamilton, R.L., Sobol, R.W., Nikiforova, M.N., Lyons-Weiler, M.A., Laframboise, W.A., Burger, P.C., Brat, D.J., Rosenblum, M.K., Holmes, E.J., et al. (2011). IDH1 mutations are common in malignant gliomas arising in adolescents: A report from the Childrens Oncology Group. Childs Nerv. Syst. 27, 8794. Published online August 20, 2010. http://dx.doi.org/10. 1007/s00381-010-1264-1. Puget, S., Philippe, C., Bax, D.A., Job, B., Varlet, P., Junier, M.P., Andreiuolo, F., Carvalho, D., Reis, R., Guerrini-Rousseau, L., et al. (2012). Mesenchymal transition and PDGFRA amplication/mutation are key distinct oncogenic events in pediatric diffuse intrinsic pontine gliomas. PLoS ONE 7, e30313. Qu, H.Q., Jacob, K., Fatet, S., Ge, B., Barnett, D., Delattre, O., Faury, D., Montpetit, A., Solomon, L., Hauser, P., et al. (2010). Genome-wide proling using single-nucleotide polymorphism arrays identies novel chromosomal imbalances in pediatric glioblastomas. Neuro-oncol. 12, 153163. Schiffman, J.D., Hodgson, J.G., VandenBerg, S.R., Flaherty, P., Polley, M.Y., Yu, M., Fisher, P.G., Rowitch, D.H., Ford, J.M., Berger, M.S., et al. (2010). Oncogenic BRAF mutation with CDKN2A inactivation is characteristic of a subset of pediatric malignant astrocytomas. Cancer Res. 70, 512519. Schwartzentruber, J., Korshunov, A., Liu, X.Y., Jones, D.T.W., Pfaff, E., Jacob, njes, M., et al. (2012). Driver K., Sturm, D., Fontebasso, A.M., Quang, D.A., To mutations in histone H3.3 and chromatin remodelling genes in paediatric glioblastoma. Nature 482, 226231.

436 Cancer Cell 22, 425437, October 16, 2012 2012 Elsevier Inc.

Cancer Cell
Epigenetic and Biological Subgroups of Glioblastoma

mel, T., Vladimirova, V., Kramm, C.M., Pietsch, T., Setty, P., Hammes, J., Rotha and Waha, A. (2010). A pyrosequencing-based assay for the rapid detection of IDH1 mutations in clinical samples. J. Mol. Diagn. 12, 750756. Statham, A.L., Robinson, M.D., Song, J.Z., Coolen, M.W., Stirzaker, C., and Clark, S.J. (2012). Bisulte sequencing of chromatin immunoprecipitated DNA (BisChIP-seq) directly informs methylation status of histone-modied DNA. Genome Res. 22, 11201127. 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. 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, Y., Yang, J., Zheng, H., Tomasek, G.J., Zhang, P., McKeever, P.E., Lee, E.Y., and Zhu, Y. (2009). Expression of mutant p53 proteins implicates a lineage relationship between neural stem cells and malignant astrocytic glioma in a murine model. Cancer Cell 15, 514526.

Wilkerson, M.D., and Hayes, D.N. (2010). ConsensusClusterPlus: a class discovery tool with condence assessments and item tracking. Bioinformatics 26, 15721573. 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. 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. Zarghooni, M., Bartels, U., Lee, E., Buczkowicz, P., Morrison, A., Huang, A., Bouffet, E., and Hawkins, C. (2010). Whole-genome proling of pediatric diffuse intrinsic pontine gliomas highlights platelet-derived growth factor receptor alpha and poly (ADP-ribose) polymerase as potential therapeutic targets. J. Clin. Oncol. 28, 13371344.

Cancer Cell 22, 425437, October 16, 2012 2012 Elsevier Inc. 437

Article
The Requirement for Cyclin D Function in Tumor Maintenance
Yoon Jong Choi,1,8,11 Xiaoyu Li,2,9,11 Per Hydbring,1,8 Takaomi Sanda,3,7,10 Joanna Stefano,1 Amanda L. Christie,4 Sabina Signoretti,5,6 A. Thomas Look,3,7,10 Andrew L. Kung,4,10,12 Harald von Boehmer,2,9 and Piotr Sicinski1,8,*
of Cancer Biology of Cancer Immunology and AIDS 3Department of Pediatric Oncology 4Lurie Family Imaging Center 5Department of Medical Oncology Dana-Farber Cancer Institute, Boston, MA 02215, USA 6Department of Pathology, Brigham and Womens Hospital, Boston, MA 02115, USA 7Division of Hematology/Oncology, Childrens Hospital, Boston, MA 02115, USA 8Department of Genetics 9Department of Microbiology and Immunology 10Department of Pediatrics Harvard Medical School, Boston, MA 02115, USA 11These authors contributed equally to this work 12Present address: Division of Pediatric Hematology/Oncology/Stem Cell Transplantation, Columbia University Medical Center, New York, NY 10032, USA *Correspondence: peter_sicinski@dfci.harvard.edu http://dx.doi.org/10.1016/j.ccr.2012.09.015
2Department 1Department

Cancer Cell

SUMMARY

D-cyclins represent components of cell cycle machinery. To test the efcacy of targeting D-cyclins in cancer treatment, we engineered mouse strains that allow acute and global ablation of individual D-cyclins in a living animal. Ubiquitous shutdown of cyclin D1 or inhibition of cyclin D-associated kinase activity in mice bearing ErbB2-driven mammary carcinomas triggered tumor cell senescence, without compromising the animals health. Ablation of cyclin D3 in mice bearing Notch1-driven T cell acute lymphoblastic leukemias (T-ALL) triggered tumor cell apoptosis. Such selective killing of leukemic cells can also be achieved by inhibiting cyclin D associated kinase activity in mouse and human T-ALL models. Inhibition of cyclin D-kinase activity represents a highly-selective anticancer strategy that specically targets cancer cells without signicantly affecting normal tissues.
INTRODUCTION The proliferation of mammalian cells is driven by the core cell cycle machinery operating in the cell nucleus. The key components of this machinery are proteins called cyclins, which bind, activate, and provide substrate specicity to cyclin-dependent kinases (CDKs). Cyclin-CDK complexes phosphorylate cellular proteins, thereby driving cell cycle progression (Malumbres and Barbacid, 2009). Among all cyclin classes, D-type cyclins are of particular importance to the cancer eld, as they represent the ultimate recipients of many oncogenic pathways (Deshpande et al., 2005; Fu et al., 2004; Musgrove et al., 2011). This family is composed of cyclins D1, D2, and D3, which show substantial amino acid sequence similarity, and are expressed in an overlapping, redundant fashion in all proliferating cell types (Musgrove et al., 2011). D-cyclins bind and activate CDK4 and CDK6; cyclin D-CDK4 and D-CDK6 complexes phosphorylate

Signicance D-cyclins are recipients of many oncogenic signals, and are often overexpressed in human cancers. This study tests the consequences of an acute and global inhibition of D-cyclins in vivo, in a living mouse. We found that cyclins D1 or D3 are largely dispensable for normal physiology of adult animals, but they are essential for tumor maintenance. We demonstrated that acute cyclin D1- or D3-inhibition causes tumor cell senescence or apoptosis. Hence, therapeutic targeting of D-cyclins is expected to be highly specic in shutting off cancer progression, without signicantly affecting normal tissues. Our demonstration that inhibition of cyclin D-CDK kinase selectively kills mouse and human leukemic cells in vivo suggests a potential therapeutic strategy in patients bearing these malignancies.
438 Cancer Cell 22, 438451, October 16, 2012 2012 Elsevier Inc.

Cancer Cell
D-Cyclins in Tumor Maintenance

the retinoblastoma tumor suppressor protein, pRB, and pRB-like p107 and p130 proteins, leading to activation or derepression of E2F transcription factors. E2Fs then induce several target genes that are required for entry of cells into the DNA synthesis (S) phase (Trimarchi and Lees, 2002). In addition, cyclin D-CDK4 and D-CDK6 complexes play a second, noncatalytic function in G1 phase progression through sequestration of cell cycle inhibitors p27Kip1 and p21Cip1, which leads to activation of CDK2-containing complexes (Sherr and Roberts, 2004). Amplication of individual cyclin D genes and overexpression of their encoded proteins were documented in a large proportion of human cancers. For example, cyclin D1 is overexpressed in the majority of breast cancer cases, whereas overexpression of cyclin D3 is observed in many lymphoid malignancies (Deshpande et al., 2005; Fu et al., 2004; Musgrove et al., 2011). A comprehensive analysis of many human cancer types revealed that the gene encoding cyclin D1 represents the second most frequently amplied locus in the human cancer genome (Beroukhim et al., 2010). Moreover, in many human cancers overexpression of D-type cyclins takes place in the absence of any detectable genomic alterations (Deshpande et al., 2005; Fu et al., 2004; Musgrove et al., 2011). It was initially assumed that individual D-type cyclins are required for proliferation of normal, nontransformed cells. However, we and others found that knockout mice lacking individual D-cyclins are viable, and display only minor phenotypes, revealing that these proteins are dispensable for development of the overwhelming majority of organs (Sherr and Roberts, 2004). In contrast, particular D-cyclins were shown to be essential for tumor initiation in vivo in the specic compartments. For example, mice lacking cyclin D1 are resistant to ErbB2-driven mammary adenocarcinomas, while cyclin D3 null animals are refractory to Notch1-driven T-ALL (Bowe et al., 2002; Landis et al., 2006; Sicinska et al., 2003; Yu et al., 2001). These analyses established an essential requirement for D-cyclins in tumor initiation. An important unresolved question is whether these proteins are also required for tumor maintenance, and whether their ablation in mice that already developed tumors would have an effect on tumor progression. Another critical question for therapeutic targeting of D-cyclins is what would be the consequence of an acute shutdown of individual D-cyclins in the whole animal. Knockout mice lacking particular D-cyclins displayed only minor phenotypes, but these mice developed from the very beginning in the absence of a cyclin D-protein. It is well-established that constitutive, germline knockout animals often activate compensatory mechanisms (Lin et al., 2008; Sage et al., 2003), whereas an acute shutdown of a protein in an adult animal may have much more profound consequences. To address these questions, we developed mouse models that allowed us to inducibly shut off cyclin D function in the whole animal. Using these models, we acutely and ubiquitously ablated expression of cyclin D1 or D3 in adult mice that developed different types of tumors. RESULTS Acute Ablation of Cyclin D1 in Adult Mice In order to determine the consequence of an acute and global ablation of cyclin D1 in adult mice, we generated conditional

cyclin D1 knockout (D1F/F) animals (Figures S1AS1C available online). We determined that cyclin D1F/F mice developed normally and displayed no phenotypic abnormalities, consistent with the expectation that the oxed cyclin D1 allele is functionally wildtype. We interbred cyclin D1F/F and cyclin D1/ mice and generated heterozygous cyclin D1F/ animals that were used in the analyses described below. These cyclin D1F/ mice were phenotypically normal, as expected from the normal appearance of cyclin D1+/ heterozygotes (Fantl et al., 1995; Sicinski et al., 1995). In order inducibly ablate cyclin D1 expression in adult mice, we crossed cyclin D1F/ mice with Esr1-Cre animals. The latter strain ubiquitously expresses tamoxifen-inducible Cre recombinase. Administration of tamoxifen to Esr1-Cre mice activates Cre, leading to global deletion of the oxed sequences in mouse organs (Hayashi and McMahon, 2002). Adult cyclin D1F//Esr1-Cre mice were injected with tamoxifen, and efcient deletion of cyclin D1 in several organs was veried by semiquantitative PCR (Figure S1D). We then observed the animals for 1 year, without noting any obvious abnormalities or premature lethality. The mice displayed normal biochemical parameters in the peripheral blood, which were periodically monitored (Figures S1E and S1F and data not shown). Hence, acute ablation of cyclin D1 in adult mice had no detectable impact on the animals health, suggesting that cyclin D1 is largely dispensable for normal physiology of adult animals. Acute Ablation of Cyclin D1 in Adult Females Bearing ErbB2-Driven Mammary Carcinomas We next asked what would be the consequence of an acute and global ablation of cyclin D1 in mice bearing ErbB2-driven mammary carcinomas. We used MMTV-ErbB2 mice that overexpress ErbB2 (HER2) oncogene in their mammary epithelium (Muller et al., 1988). Female MMTV-ErbB2 mice develop mammary adenocarcinomas with a median latency of 3040 weeks (Yu et al., 2001). We interbred cyclin D1F/, Esr1-Cre, and MMTV-ErbB2 mice and generated cyclin D1F//Esr1-Cre/ MMTV-ErbB2 females. Once these animals developed palpable breast tumors, we injected them with tamoxifen, thereby triggering cyclin D1 deletion in the whole animal, including in their breast cancers (Figure 1A). We found that ablation of cyclin D1 halted breast cancer progression. For controls, we used cyclin D1F/+/Esr1-Cre/MMTV-ErbB2 females, whose tumors continued to grow following tamoxifen challenge (Figure 1B). Consequently, at the time of sacrice, the animals overall tumor burden was signicantly lower in cyclin D1F/-/Esr1-Cre/MMTVErbB2 females (Figures 1B, S1G, and S1H). Staining of tumor sections for Ki-67, a marker of proliferation, revealed a strongly reduced fraction of cells expressing Ki-67 in cyclin D1-deleted tumors, indicating that cyclin D1 shutdown crippled proliferation of breast cancer cells in vivo (Figures S1I and S1J). Strikingly, we observed that cyclin D1 ablation also triggered senescence of breast cancer cells, as evidenced by wide-spread staining of tumor cells for senescence-associated (SA)-b-galactosidase (Figures 1C and 1D) and trimethylated lysine 9 of histone H3 (H3K9-Me3, Figure S1K). These results indicate that the continued presence of cyclin D1 is required to maintain growth of ErbB2-overexpressing breast cancers, by driving tumor cell proliferation and by protecting tumor cells from senescence.
Cancer Cell 22, 438451, October 16, 2012 2012 Elsevier Inc. 439

Cancer Cell
D-Cyclins in Tumor Maintenance

MMTV-ErbB2
Mammary-specific expression of ErbB2

tamoxifen D1
/-

sacrifice

Figure 1. Acute Ablation of Cyclin D1 Blocks Tumor Progression and Triggers Senescence in ErbB2-Driven Tumors
(A) Schematic representation of the experimental design. (B) Total tumor burden per mouse in cyclin D1F/+/ Esr1-Cre/MMTV-ErbB2 (D1D /+) and in D1F//Esr1Cre/MMTV-ErbB2 (D1D /) mice. Deletion of cyclin D1 was induced (by tamoxifen administration) upon detection of palpable tumors; mice were sacriced after 6 weeks. In the top panel, each dot represents the total tumor burden (g) for a given mouse upon sacrice, horizontal lines represent mean values. Lower panel: mean total tumor burden (g) per mouse. Error bars represent SD, n = 10. (C) Representative SA-b-galactosidase staining of sections of ErbB2-driven mammary tumors following acute cyclin D1 shutdown (cyclin D1D /), or in control animals (cyclin D1D/+). Scale bar represents 50 mm. (D) Mean percentage of SA-b-galactosidasepositive area in tumor sections. For each section, three independent areas were scanned and quantied using software from Aperio Technologies. Error bars represent SEM. (E) Total tumor burden per animal in MMTV-ErbB2 mice treated daily with vehicle only (VO) or PD 0332991 (PD, 150 mg/kg body weight) for 6 weeks after detection of palpable tumors. Upper panel shows tumor burden for each mouse, lower panel shows mean values SD, n = 7. (F) Representative SA-b-galactosidase staining of sections of ErbB2-driven mammary carcinomas from mice treated with vehicle only or PD 0332991 for 6 weeks after detection of palpable tumors. Scale bar represents 50 mm. (G) Mean percentage of area staining positive for SA-b-galactosidase, analyzed as in (D). Error bars represent SEM. See also Figure S1.

Group: D1F/-/Esr1-Cre+ or D1F/+Esr1-Cre+

0 2 4 6 Weeks post tumor detection

Total tumor burden/ mouse (grams)

*
3 2 1 0
1.4

C
SA- -gal p<0.05

Cyclin D1

/+

Cyclin D1

/-

D1

/+

D1

/-

D
70 60

***
p<0.0001

Total tumor burden/ mouse (grams)

% SA- -gal+

1.2 1.0 0.8 0.6 0.4 0.2 0.0

50 40 30 20 10

D1

/+

D1 *

/-

D1

/+

D1

/-

E
Total tumor burden/ mouse (grams)
5 4 3 2 1 0

F
SA- -gal Vehicle only p<0.05

PD 0332991

VO

PD

Total tumor burden/ mouse (grams)

2.5 2.0 1.5 1.0 0.5 0.0

G
% SA- -gal+

50 40 30 20 10

***
p<0.0005

VO

PD

VO

Inhibition of Cyclin D-Cdk Kinase Activity in Adult Females Bearing ErbB2-Driven Mammary Tumors We next asked whether inhibition of cyclin D-associated kinase activity would have the same impact as an acute cyclin D1 ablation. To address this issue, we monitored a group of MMTV-ErbB2 females for breast cancer occurrence. As soon as breast tumors were detected, we started treating mice with PD 0332991, a specic and potent inhibitor of cyclin D-CDK4 and D-CDK6 kinases (Fry et al., 2004). We found that inhibition of cyclin D-associated kinase activity essentially phenocopied an acute cyclin D1 ablation, namely it halted progression of breast cancers, and it triggered tumor cell senescence (Figures 1E1G, S1L, and S1M), without having any overt effect on the animals health (data not shown). Neither cyclin D1 ablation nor
440 Cancer Cell 22, 438451, October 16, 2012 2012 Elsevier Inc.

PD 0332991 treatment had any effect on the apoptotic rate of breast cancer cells (Figure S1N). Collectively, these analyses revealed PD that cyclin D1 and cyclin D1-associated kinase are largely dispensable for the physiology of adult animals, but they are essential for maintenance of ErbB2-driven breast cancers. Molecular Analyses of ErbB2-Driven Breast Cancers In order to gain some understanding of the observed antisenescence function of cyclin D1-associated kinase in breast cancers in vivo, we rst focused on the pRB / E2F pathway. Hypophosphorylated (active) pRB protein was shown to enforce the senescent state by permanently silencing expression of E2F target genes, such as cyclin E1, cyclin A2, or Dhfr (Chicas et al., 2010; Narita et al., 2003; Dean et al., 2010). We found that administration of PD 0332991 to breast cancer-bearing animals resulted in a strong inhibition of pRB phosphorylation in breast cancer cells in vivo (Figure S1O). This led to decreased

Cancer Cell
D-Cyclins in Tumor Maintenance

expression of E2F target genes, including the genes whose silencing was implicated to play a role in the senescence program (Figure S1P). We conclude that inhibition of E2F activity likely contributes to cancer cell senescence in vivo. We recently showed that cyclin D1-CDK4 kinase phosphorylates, stabilizes, and activates FOXM1 transcription factor (Anders et al., 2011), which plays an important role in protecting cancer cells from senescence (Li et al., 2008a; Park et al., 2009). Moreover, treatment of in vitro cultured human cancer cells with PD 0332991 destabilized FOXM1 and led to pRB-independent cancer cell senescence (Anders et al., 2011). Analyses of breast tumors from PD 0332991-treated mice revealed significantly decreased levels of FOXM1 protein in cancer cells in vivo (Figure S1Q). We also established that this led to signicantly reduced levels of FOXM1 transcriptional targets (Figure S1R). It is likely that this inhibition of FOXM1 activity contributes to senescence of breast cancer cells in vivo, upon PD 0332991administration to cancer-bearing mice. It should be noted, however, that because we ablated cyclin D1, or inhibited cyclin D1-associate kinase in the entire animal (not only in tumor cells) we cannot exclude that noncell autonomous mechanisms contributed to inhibition of tumor progression. Ablation of Cyclin D3 in Adult Mice Another instance of a specic requirement for a particular D-cyclin in tumor initiation is the reliance of T cell acute lymphoblastic leukemia (T-ALL) formation on cyclin D3. We previously showed that cyclin D3/ mice are refractory to Notch1-driven T-ALL (Sicinska et al., 2003). To determine whether T-ALL cells also require this cyclin for tumor maintenance, we generated conditional cyclin D3 knockout (D3F/F) mice (Figures S1SS1U). We crossed cyclin D3F/F and Esr1-Cre mice and obtained cyclin D3F/F/Esr1-Cre animals. We challenged adult cyclin D3F/F/ Esr1-Cre mice with tamoxifen, veried efcient deletion of the oxed cyclin D3 alleles in multiple organs (Figure S1V), and observed mice for 1 year. We found no obvious abnormalities upon cyclin D3 ablation throughout the entire observation period, including normal biochemical parameters in the peripheral blood, which were periodically monitored (Figure S1E). Because cyclin D3/ mice displayed hematopoietic abnormalities (Sicinska et al., 2003, 2006; Cooper et al., 2006), we also crossed cyclin D3F/F mice with Mx1-Cre animals. The latter strain allows a very efcient inducible deletion of the oxed sequences in the hematopoietic cells, following administration hn et al., 1995) (Figure S2A). We found of polyI-polyC (pI-pC) (Ku that an acute shutdown of cyclin D3 in the hematopoietic lineages of cyclin D3F/F/Mx1-Cre animals essentially phenocopied the abnormalities observed in constitutive cyclin D3 null mice. Specically, cyclin D3D/D/Mx1-Cre animals displayed very small thymi containing signicantly reduced numbers of doublepositive CD4+CD8+ thymocytes (Figures S2B and S2C). Analyses of peripheral blood revealed modestly decreased red and white blood cell counts (Figures S2DS2F), again resembling the abnormalities found in cyclin D3/ animals (Sicinska et al., 2006). Altogether, these observations revealed that an acute shutdown of cyclin D3 recapitulated the phenotype of cyclin D3 null mice, but it did not cause any additional major abnormalities, indicating that cyclin D3 is largely dispensable for the health of adult animals.

Ablation of Cyclin D3 in T-ALL We next asked what would be the impact of an acute shutdown of cyclin D3 for development and maintenance of T cell acute lymphoblastic leukemia. As before, we utilized a mouse model of Notch1-driven T-ALL, as the majority of human T-ALL cases contain activating lesions in the Notch1 pathway (Weng et al., 2004). We collected bone marrow cells from cyclin D3F/F/ Mx1-Cre+ or D3F/F/Mx1-Cre mice, ow-sorted hematopoietic progenitor cells (HPC), transduced HPC with a retrovirus encoding activated Notch1 (Notch1 intracellular domain, Notch1-ICD) and GFP, and injected the transduced HPC into the bloodstream of sublethally irradiated C57BL/6 wild-type recipient mice (Figure 2A). It is well established that the recipient animals develop T-ALL following this procedure. The rst abnormal, GFP-positive CD4+CD8+ cells appear in the peripheral blood after 2 weeks, mice develop multiple T cell tumors after 6 weeks, leading to the animals death within 1012 weeks (Li et al., 2008b; Chiang et al., 2006). Two weeks after bone marrow transplantation (BMT), when abnormal GFP+ cells appeared in the peripheral blood, we challenged the recipient mice with pI-pC to ablate cyclin D3 expression in the transplanted cells. Unexpectedly, we found that the GFP+ cells essentially disappeared from the peripheral blood following cyclin D3 shutdown (Figures 2B and 2C). Further observation of animals revealed that whereas control mice (transduced with cyclin D3F/F/Mx1-Cre-negative HPC) developed T cell tumors and died within 11 weeks, six out of seven mice that sustained acute cyclin D3 deletion remained leukemia-free and healthy throughout the observation period (Figure 2D). We found that in the single experimental mouse that succumbed to T-ALL, tumors arose from the residual cells containing undeleted cyclin D3 alleles (Figure S3A). The remaining six mice were eventually sacriced after 30 weeks, and the absence of inltrating leukemic cells in their organs was veried by detailed histopathological analysis (Figure S3B). Hence, an acute ablation of cyclin D3 blocked development of Notch1-driven T-ALL in vivo. In the analyses described above, we deleted cyclin D3 at an early time-point of T-ALL development. To investigate the requirement for cyclin D3 in the maintenance of T-ALL tumors, we next switched off cyclin D3 expression (by pI-pC administration) at 6.5 weeks post-BMT, after the animals had developed tumors (Figure 3A). We then collected major organs and enumerated the presence of inltrating GFP+ tumor cells, by FACS. Strikingly, we found that an acute ablation of cyclin D3 led to a strong reduction in the number of tumor cells (Figures 3B and 3C; data not shown). Long-term observation of animals revealed that ablation of cyclin D3 led to a signicant extension of the animals survival (Figure 3D). We veried that tumors which eventually arose in these animals contained GFP+ cells with undeleted cyclin D3F alleles (Figure S3C). Collectively, these analyses established that cyclin D3 is required for maintenance of Notch1-driven T-ALL in vivo, and that an acute ablation of cyclin D3 leads to disappearance of malignant cells. Cell Cycle Arrest and Tumor Cell Apoptosis Following Ablation of Cyclin D3 To examine the cellular mechanisms by which cyclin D3 ablation inhibits Notch1-ICD-induced T-ALL progression, we rst
Cancer Cell 22, 438451, October 16, 2012 2012 Elsevier Inc. 441

Cancer Cell
D-Cyclins in Tumor Maintenance

Notch1-ICD

pI-pC D3
/

A
Harvest Lin- BM Retroviral infection BMT 2

Figure 2. Acute Ablation of Cyclin D3 in a Mouse Model of Notch1-ICD-Driven T Cell Acute Lymphoblastic Leukemia
10

4 6 8 Weeks post-BMT

Donor: D3F/F/Mx-Cre+ or D3F/F/Mx-Cre-

Recipient: C57BL/6 WT

T-ALL formation in transplanted mice

Notch1-ICD-GFP+

D3F/F Blood 36%

D3

Blood 3%

C
%Notch1-ICD-GFP+
35 30 25 20 15 10 5 0

BLOOD

***
p<0.0002

SSC

SSC

D
% Survival

100 80 60 40 20 0 0 10
F/F D3 D3 WT n=7 D3 D3 / n=7

p<0.0002

(A) Schematic representation of the experimental design. After T-ALL formation in the recipient mice, cyclin D3 ablation was induced at 2 weeks post-BMT. (B) Representative FACS proles showing percentage of Notch1-ICD-overexpressing GFP+ cells in peripheral blood right after completion of pI-pC administration (started 2 weeks post-BMT) to cyclin D3F/F/Mx1-Cre+(D3D/D) or D3F/F/Mx1Cre (D3F/F) mice. SSC, side scatter. (C) Mean percentages of GFP+ cells in peripheral blood of mice treated as in (B). Error bars represent SD, n = 7. (D) Kaplan-Meier analysis of survival time in each group, n = 7. See also Figure S2.

D3F/F

D3

20

30

Time (weeks post-BMT)

examined cell cycle distribution of splenic tumor cells following an acute shutdown of cyclin D3. We observed a reduction of cells in the S phase and an increase of G0/G1 cells (Figures S4AS4C). We next asked whether the rapid disappearance of T-ALL cells upon cyclin D3 ablation might be caused by tumor cell apoptosis. To test this, we stained the GFP+ leukemic cells from the peripheral blood, as well as tumor cells that inltrated spleens, with Annexin V/7-AAD, and analyzed by FACS. We detected wide-spread apoptosis of peripheral blood leukemic cells and of tumor cells following cyclin D3 ablation (Figures 4A4D). TUNEL staining of tumor sections conrmed tumor cell apoptosis (Figure 4E). These analyses revealed that cyclin D3 is required to maintain survival of T-ALL cells, and that ablation of cyclin D3 triggers tumor cell apoptosis. Tumor Cell Apoptosis following Inhibition of Cyclin D-Cdk Kinase We wished to determine whether inhibition of cyclin D3-associated kinase activity would have the same impact as an acute cyclin D3 ablation. To address this, mice were injected with wild-type hematopoietic progenitor cells that were transduced with a virus expressing Notch1-ICD and GFP, leading to T-ALL development. Daily treatment with PD 0332991 was started when GFP+ cells were detected in the peripheral blood (2 weeks posttransplantation), or after mice had developed T cell tumors (6.5 weeks posttransplantation). Once the treatment was initiated, we observed a strong reduction in the number of leukemic cells in the peripheral blood and of tumor cells inltrating the peripheral organs (Figures 5A and 5B; data not shown). Continuous long-term treatment resulted in a dramatic improvement of survival rates in both groups, with the majority of animals remaining healthy at 40 weeks post-BMT (Figures 5C and 5D). Importantly, we veried that treatment of tumor cells with PD 0332991 inhibited cell cycle progression and triggered tumor
442 Cancer Cell 22, 438451, October 16, 2012 2012 Elsevier Inc.

cell apoptosis, thereby mimicking the effect of an acute cyclin D3 ablation (Figures 5E, 5F, S5A, and S5B data not shown). Moreover, combined administration of pI-pC (to induce deletion of cyclin D3) plus PD 0332991 to cyclin D3F/F/Mx1-Cre mice did not increase the tumor apoptotic rate above that seen upon cyclin D3 ablation alone, conrming that the effects of PD 0332991 are mediated by inhibiting cyclin D3 activity (Figure S5C). No senescence of T-ALL cells was observed following PD 0332991 treatment (Figure S5D). Collectively, these ndings indicate that cyclin D3-Cdk kinase is required for T-ALL maintenance, by driving cell proliferation and by protecting tumor cells from the apoptotic death. Analyses of Human Leukemic Cells In Vitro and In Vivo We next asked whether inhibition of cyclin D-associated kinase activity in human leukemic cells would also trigger apoptosis. We chose human T-ALL cell lines KOPTK1, DND41, MOLT-16, and RPMI-8402, which are known to harbor activating mutations in the NOTCH1 gene (NOTCH1-ICD-positive) (ONeil et al., 2007). We treated cells in vitro with PD 0332991, stained them with Annexin V/7-AAD, and analyzed by FACS. We observed that inhibition of cyclin D-CDK kinase triggered cell cycle arrest and apoptosis of human T-ALL cells (Figures 6A, 6B, S6A, and S6B; data not shown). No senescence of T-ALL cells was observed after PD 0332991 treatment (Figure S6C). Similar analysis of an additional 20 human leukemic cell lines (including NOTCH1-ICD-negative T-ALL, B cell ALL, and others) revealed that PD 0332991 treatment did not cause substantial increase in the apoptotic rate in these cell lines (Figure S6D; data not shown). This, together with our observations that inhibition of cyclin D3 activity killed Notch1-ICD-positive mouse T-ALL tumor cells (Figures 4 and 5), suggests that cyclin D3-CDK activity is specically required for survival of T-ALL cells with activated Notch1 pathway. Intriguingly, comparison of the levels of D-cyclins and CDKs between four NOTCH1-ICD-positive and negative human

Cancer Cell
D-Cyclins in Tumor Maintenance

Notch1-ICD

pI-pC D3
/

A
Harvest Lin- BM Retroviral infection BMT 2

Figure 3. Cyclin D3 Is Required for Maintenance of Notch1-ICD-Driven T-ALL In Vivo


10

4 6 8 Weeks post-BMT

Donor: D3F/F/Mx-Cre+ or D3F/F/Mx-Cre-

Recipient: C57BL/6 WT

T-ALL formation in transplanted mice

Notch1-ICD-GFP+

D3F/F Spleen 70%

D3

Spleen 7%

C
90 80 70 60 50 40 30 20 10 0

SPLEEN

%Notch1-ICD-GFP+

***
p<0.0001

SSC
100

SSC

D
% Survival

80 60 40 20 0 0

p<0.0008

(A) Schematic representation of the experimental design. After T-ALL formation in the recipient mice, cyclin D3 ablation was induced at 6.5 weeks postBMT. (B) Representative FACS analysis showing percentage of Notch1-ICD-overexpressing GFP+ cells in spleens right after completion of pI-pC administration (started 6.5 weeks post-BMT) in cyclin D3F/F/Mx1-Cre+ (D3D/D) or D3F/F/Mx1-Cre (D3F/F) mice. SSC, side scatter. (C) Mean percentages of GFP+ cells in spleens of mice treated as in (B). Error bars represent SD, n = 4. (D) Kaplan-Meier analysis of survival time in each group, n = 6. See also Figure S3.

D3F/F

D3

F/F n=6 D3 D3WT / D3 D3 / n=6

10

15

Time (weeks post-BMT)

T-ALL lines indicated that the former express signicantly higher levels of cyclin D3 (and cyclin D2) and cyclin D3-CDK6 complexes (Figures S6E and S6F). It is possible that NOTCH1 activation increases the levels of D-cyclins in T-ALL cells, thereby making these cells addicted to cyclin D-CDK kinase for tumor cell survival. To test the response of human NOTCH1-ICD-positive T-ALL cells to cyclin D-CDK inhibition in an in vivo setting, we engineered the human T-ALL cell lines KOPTK1 and DND41 to stably express rey luciferase. Engineered human leukemic cells were injected intravenously into immunocompromised mice, and bioluminescence imaging was used to assess the distribution and burden of leukemic disease. Recipient mice died within 57 weeks due to widely disseminated T cell leukemia. To determine the acute effects of CDK4/6 inhibition, mice with disseminated KOPTK1 and DND41 leukemia were treated for 57 days with PD 0332991. As expected, leukemic burden increased in vehicle-treated mice (Figures 7A and 7C). In contrast, treatment with PD 0332991 either resulted in a nearly 5-fold reduction in tumor burden (KOPTK1 cells, Figure 7A) or blunted this increase (DND41, Figure 7C). We collected peripheral blood and bone marrow cells from PD 0332991 treated animals, and analyzed the proportion of apoptotic human T-ALL cells by Annexin V/7-AAD staining followed by FACS. Treatment of tumor-bearing mice with PD 0332991 induced signicant apoptosis in both KOPTK1 (Figure 7B) and DND41 (Figure 7D) xenografts in vivo without having any overt effect on the animals health (Figure S7; data not shown). To determine the therapeutic efcacy of CDK4/6 inhibition, we treated mice bearing established human T-ALL xenografts with PD 0332991. Long-term daily treatment with PD 0332991 resulted in suppression of leukemia progression as determined by bioluminescence imaging (Figures 7E, 7F, 7H, and 7I), and resulted in a highly signicant prolongation in

survival for both KOPTK1 (Figure 7G) and DND41 (Figure 7J) xenografted animals. Collectively, these ndings suggest that inhibition of cyclin D-associated kinase activity in patients with NOTCH1positive T-ALL might represent an attractive therapeutic approach, as it may lead to selective killing of leukemic cells. Molecular Analyses of T-ALL Cells and Breast Cancer Cells The studies described above revealed that T-ALL cells respond to inhibition of cyclin D-CDK4/6 kinase by undergoing cell cycle arrest and cell death. In order to understand the molecular basis of these phenomena, we rst focused on the pRB / E2F pathway. We observed that treatment of T-ALL cells with PD 0332991 resulted in the appearance of hypophosphorylated (active) pRB species (Figures S5A, S5B, S6A, and S6B). Inhibition of cyclin D3 function led to diminished E2F transcriptional activity, as evidenced by decreased levels of several E2F transcriptional targets, including genes involved in G1 and S phase progression (Figure S4C). We conclude that inhibition of the pRB / E2F pathway is likely responsible for cell cycle arrest of T-ALL cells following PD 0332991 treatment. The pRB / E2F pathway was shown to be also involved in controlling cell survival (Ianari et al., 2009; Young and Longmore, 2004). E2Fs can play both pro-survival and pro-apoptotic roles, depending on cellular context (Chen et al., 2009; Chong et al., 2009). In the developing mouse retinas, E2Fs were shown to play a pro-survival function by inducing expression of p53deacetylating enzymes. Ablation of E2Fs in retinal cells was demonstrated to cause strong hyperacetylation of p53, leading to p53 activation and p53-dependent cell death (Chen et al., 2009). However, we found that cyclin D3-CDK kinase likely does not operate through this mechanism in T-ALL cells. First, we established that PD 0332991 treatment did not overtly change p53 acetylation levels in T-ALL cells (Figure S5E). Moreover, cyclin D-CDK4/6 inhibition also triggered apoptosis of T-ALL cell lines expressing mutant p53: MOLT-16, DND-41, and RPMI-8402 (http://www.sanger.ac.uk/genetics/CGP/CellLines/)
Cancer Cell 22, 438451, October 16, 2012 2012 Elsevier Inc. 443

Cancer Cell
D-Cyclins in Tumor Maintenance

Mice received Notch1-ICD-GFP-transduced cyclin D3F/F/Mx1-Cre+ (D3D/D) or D3F/F/Mx1-Cre (D3F/F) bone marrow cells, and were treated ve times with pI-pC, starting at 6.5 weeks following BMT. (A) Representative FACS proles of peripheral blood stained for Annexin V/7-AAD and gated on GFP+ cells. (B) Mean percentages of apoptotic (Annexin V+) cells in peripheral blood (gated on GFP+ cells) and analyzed as in (A). Error bars represent SD, n = 4. (C) Representative FACS proles of cells from the spleen stained for Annexin V/7-AAD and gated on GFP+. (D) Mean percentages of apoptotic (Annexin V+) cells from spleens (gated on GFP+ cells) and analyzed as in (C). Error bars represent SD, n = 4. (E) TUNEL staining of spleen sections. Scale bar represents 50 mm. See also Figure S4.

Figure 4. Cyclin D3 Ablation Triggers Apoptosis of Notch1-ICD-Driven T-ALL Cells

(Figure 6), indicating that the apoptosis following cyclin D-CDK4/6 inhibition can be p53-independent. Given a differential response to PD 0332991 treatment of T-ALL cells (cell cycle arrest and apoptosis) versus breast cancer cells (cell cycle arrest and senescence, but no apoptosis), we decided to compare the response of these two tumor types to cyclin D-CDK4/6 inhibition at a molecular level. To this end, we treated in vitro cultured human T-ALL cell line KOPTK1 and mouse breast cancer cells derived from MMTV-ErbB2 tumor (V720 cells) (Yu et al., 2006) with PD 0332991. We then analyzed gene expression at a genome-wide level using microarrays. Comparison of T-ALL and breast cancer cell expression proles revealed three distinct gene expression groups (Fig444 Cancer Cell 22, 438451, October 16, 2012 2012 Elsevier Inc.

ure 8A; Tables S1S4). The rst group comprised of genes that change their expression upon PD 0332991 treatment both in T-ALL and in breast cancer cells. We observed a strong enrichment for cell cycle genes by gene ontology analysis in this group (Table S2). Analysis of all cell cycle genes that were significantly regulated by PD 0332991 treatment indicated that the majority of them were downregulated (Figure 8B), which likely explains the observed cell cycle arrest. The second group comprised of genes that, upon PD 0332991 treatment, change their expression in T-ALL, but not in breast cancer cells (Figure 8A). This group showed strong enrichment for apoptosis genes by gene ontology analysis (Table S3). Analysis of all apoptosis-related genes that were signicantly

Cancer Cell
D-Cyclins in Tumor Maintenance

A
Notch1-ICD-GFP+ VO 34% PD 4%

B
%Notch1-ICD-GFP+

35 30 25 20 15 10 5 0

***
p<0.0006

Figure 5. Pharmacological Inhibition of Cdk4/6 Activity in Notch1-ICD-Driven T-ALL


Mice received Notch1-ICD-GFP-transduced C57BL/6 wild-type bone marrow cells, and were fed daily with either vehicle only (VO) or Cdk4/6 inhibitor, PD 0332991 (PD, 100150 mg/kg body weight). (A) PD 0332991 treatment was started at 6.5 weeks post-BMT. Shown are representative FACS proles depicting percentage of GFP+ cells in peripheral blood after 7 days of PD 0332991 administration. SSC, side scatter. (B) Mean percentages of GFP+ cells in peripheral blood, in mice treated as in (A). Error bars represent SD, n = 7. (C and D) Kaplan-Meyer survival curves. (C) Treatment was started at 2 weeks post-BMT (following detection of GFP+ cells in peripheral blood), n = 10 mice per group. (D) Treatment started at 6.5 weeks post-BMT, after animals had developed tumors, n = 10 mice per group. (E and F) Splenocytes from vehicle treated moribund mice were cultured in vitro and treated with vehicle only (VO) or 1 mM of PD 0332991 (PD) for 4 days. Cells were then stained for Annexin V/7AAD, and analyzed by FACS. (E) Shown is a representative FACS prole of Annexin V/7-AAD staining gated on GFP+ splenic tumor cells. (F) Mean percentage of apoptotic (Annexin V+) tumor cells following treatment with vehicle only or 1 mM of PD. Error bars represent SD, n = 4. See also Figure S5.

SSC

SSC

C
% Survival

D
% Survival

VO

PD

VO PD
PD or VO

VO PD
PD or VO

p<0.001

p<0.001

Time (weeks post-BMT)

Time (weeks post-BMT) PD

VO 30% 7-AAD

F
% Apoptosis 84%

Annexin V

100 90 p<0.0008 80 70 60 50 40 30 20 10 0

***

VO

PD

regulated in KOPTK1 cells by PD 0332991 treatment revealed that those genes displayed both up- or downregulation after treatment (Figure 8C). Visual inspection of these genes revealed that cyclin D-CDK4/6 inhibition led to a strong downregulation of BIRC5/Survivin, and resulted in an upregulation of components of the extrinsic apoptotic pathway (TNF, CASP8, CASP10), upregulation of pro-apoptotic BIM, as well an upregulation of GZMA and GZMB (Figure 8D). The third group (Figure 8A) contained genes, which change their expression in breast cancer cells, but not in T-ALL cells. This group showed enrichment for cell division genes as well for TGF-b, consistent with the well-established role of TGF-b in cellular senescence (Kuilman and Peeper, 2009) (Table S4). Collectively these analyses suggest that inhibition of cyclin D-CDK4/6 kinase in NOTCH1-positive T-ALL cells triggers a transcriptional apoptotic program, thereby causing death of tumor cells. DISCUSSION The causative role of cyclin D overexpression in the pathogenesis of human cancers has been very well documented. Amplication and rearrangements of the cyclin D1 gene were observed in nearly 100% of mantle cell lymphomas, a substantial fraction of multiple myelomas, and in several other cancer types including breast, squamous cell, pancreatic, endometrial, and lung carcinomas. Also genes encoding cyclin D2 or D3 are frequently amplied in human lymphoid malignancies and in

testicular cancers (Deshpande et al., 2005; Fu et al., 2004; Musgrove et al., 2011). Overexpression of cyclin D proteins, which results from these genomic abnormalities, is believed to represent an early, causative event in tumor formation (Chaganti and Houldsworth, 2000; Deshpande et al., 2005; Musgrove et al., 2011). For example, in breast cancers, cyclin D1 overexpression represents one of the earliest oncogenic lesions, seen already in ductal carcinoma in situ (Weinstat-Saslow et al., 1995). Consistent with the role of overexpressed D-cyclins as drivers of the oncogenic process, transgenic mice engineered to overexpress cyclin D1 in mammary glands are prone to mammary carcinomas (Wang et al., 1994). Also, targeted overexpression of cyclin D1 in B lymphocytes was shown to cooperate with the Myc oncogene in triggering B cell lymphomas (Bodrug et al., 1994; Lovec et al., 1994). Analyses of knockout mice lacking individual D-cyclins supported an essential role for these proteins in tumor initiation. Thus, mice lacking cyclin D1, or its catalytic partner, Cdk4, are either resistant, or show signicantly reduced sensitivity to Ras- and ErbB2-driven breast tumors, depending on the genetic background and type of transgene used (Bowe et al., 2002; Reddy et al., 2005; Yu et al., 2001, 2006; Zhang et al., 2011). Loss of cyclin D1 also renders mice resistant to rhabdoid tumors arising in Ini1 heterozygotes, decreases intestinal tumorigenesis in ApcMin background, and reduces incidence of Ras-driven skin papillomas (Hulit et al., 2004; Robles et al., 1998; Tsikitis et al., 2005). Cyclin D2 null mice show decreased incidence of ovarian and testicular tumors (Burns et al., 2003), whereas mice lacking cyclin D3 or Cdk6, are resistant to Notch1-driven T cell acute
Cancer Cell 22, 438451, October 16, 2012 2012 Elsevier Inc. 445

Cancer Cell
D-Cyclins in Tumor Maintenance

Figure 6. Pharmacological Inhibition of CDK4/6 Kinase Triggers Apoptosis of Human T-ALL Cell Lines
Human T-ALL cell lines were cultured and treated with vehicle only (VO) or 1 mM of PD 0332991 (PD) for 4 days, stained for Annexin V/7-AAD and analyzed by FACS. (A) Shown are representative FACS proles of KOPTK1, DND41, MOLT-16, and RPMI-8402 cells stained with Annexin V/7-AAD. (B) Quantication of the apoptotic (Annexin V+) cells, analyzed as above. Shown are mean values. Error bars represent SD, n = 4. See also Figure S6.

lymphoblastic leukemias, or to lymphomagenesis triggered by activated Akt (Hu et al., 2009; Sicinska et al., 2003). All these observations point to the requirement for a D-cyclin in oncogenic transformation. However, in order to validate therapeutic targeting of D-cyclins in cancer treatment one must establish whether the continued presence of these proteins is required to maintain tumor progression. Another critically important question is what would be the consequence of an acute and global ablation of a D-type cyclin in a living animal. To address these issues, we developed mouse strains that allowed us to globally shut down expression of individual D-cyclins in mice that developed different types of tumors. We report here that an acute ablation of cyclin D1 in adult mice bearing
446 Cancer Cell 22, 438451, October 16, 2012 2012 Elsevier Inc.

ErbB2-driven breast cancers halted tumor progression, and triggered senescence of cancer cells, without compromising the animals health. We also found that an acute ablation of cyclin D3 blocked T-ALL progression by triggering tumor cell apoptosis. Similar effects were seen upon inhibition of cyclin D-Cdk kinase in vivo, by PD 0332991 administration to tumor-bearing animals. Collectively, these results reveal that individual D-cyclins are not required for normal mouse physiology, but they are absolutely essential for tumor maintenance. Hence, by targeting particular D-cyclins, or by inhibiting their associated kinases, one can selectively kill tumor cells, or irreversibly arrest their proliferation by triggering tumor cell senescence, without affecting normal tissues. It remains to be seen whether ablating all three D-cyclins would have any consequences in normal adult animals. Our results are consistent with recent ndings of Puyol et al. (2010) demonstrating that an acute shutdown of Cdk4 in a mouse model of K-Ras-driven nonsmall cell lung cancers triggered senescence of tumor cells in vivo. Together with our results, these ndings indicate that cyclin D1-CDK4 antisenescence function likely operates in several cancer types. On the other hand, Zhang et al. (2011) recently reported that MMTV-ErbB2 breast tumor cells, after two rounds of in vitro and in vivo passaging, no longer require cyclin D1 for proliferation, due to upregulation of cyclin D3. These results indicate that analyses of serially passaged cell lines need to be treated with caution, and should be veried in vivo using intact tumors, as in the work of Puyol et al. (2010) and in this study. It remains to be determined what makes cancer cells so distinctly dependent on individual D-cyclins. One possibility is that the molecular mechanism of cell cycle progression operates differently in cancer cells versus in normal cells. Moreover, cyclin D-CDK4/6 complexes may play additional, cell

Cancer Cell
D-Cyclins in Tumor Maintenance

Figure 7. Effects of Pharmacological Inhibition of CDK4/6 on Human T-ALL In Vivo


Human KOPTK1-Luc+ or DND41-Luc+ cells were engrafted by tail vein injection into NSG mice and treated daily with vehicle only (VO) or PD 0332991 (PD, 150 mg/kg body weight). (A) Animals with established KOPTK1-Luc+ leukemia were treated daily, for 5 days, with vehicle or PD 0332991. Tumor burden is expressed as relative bioluminescence, normalized to the baseline bioluminescence on treatment day 1. Shown are mean values, error bars represent SEM.

Cancer Cell 22, 438451, October 16, 2012 2012 Elsevier Inc. 447

Cancer Cell
D-Cyclins in Tumor Maintenance

Figure 8. Response of KOPTK1 T-ALL Cells and V720 Breast Cancer Cells to Inhibition of Cyclin D-CDK Kinase
(A) KOPTK1 and V720 cells were treated with PD 0332991 (PD) or vehicle (VO) for 48 hr and 24 hr, respectively. Experiment was done in triplicate. Genes that show changes in expression levels after PD treatment were selected based on an absolute fold-change >1.5, absolute mean difference >50, and p value <0.05, and were classied into three groups: genes with altered expression levels in both cell types (Group 1), in KOPTK1 cells only (Group 2), or in V720 cells only (Group 3). Each row corresponds to one gene and is normalized across the row. (B) Heatmap images representing relative expression levels of cell cycle-related genes that showed signicant expression change after PD treatment in KOPTK1 cells (top) or V720 cells (bottom). (C) A heatmap image representing relative expression levels of apoptosis-related genes that showed signicant expression change after PD treatment in KOPTK1 cells. (D) Expression levels of seven apoptosis-related genes in KOPTK1 cells. Shown are mean values SD, n = 3 per group. See also Tables S1S4.

cycle-independent functions in cancer cells, by protecting these cells from oncogene-induced senescence, or by maintaining tumor cell survival. Indeed, cyclin D-CDK kinase was shown to

protect cells against senescence, by acting through the retinoblastoma protein (Chicas et al., 2010; Dean et al., 2010; Ruas et al., 2007) and through FOXM1 (Anders et al., 2011). It is likely

(B) After 5 days of treatment, peripheral blood and bone marrow were collected from KOPTK1-Luc+ xenografted animals, and apoptosis of human leukemic cells determined by FACS analysis of Annexin V/7-AAD staining, gated on human CD45+ cells. Graphs represent mean percentages of apoptotic (Annexin V+) cells. Error bars represent SD. (C) Animals with established DND41-Luc+ leukemia were treated daily, for 7 days, with vehicle or PD 0332991. Tumor burden is expressed as relative bioluminescence, normalized to the baseline bioluminescence on treatment day 1. Shown are mean values, error bars represent SEM. (D) After 7 days of treatment, peripheral blood and bone marrow were collected from DND41-Luc+ xenografted animals, and apoptosis of human leukemic cells determined by FACS analysis of Annexin V/7-AAD staining, gated on human CD45+ cells. Graphs represent mean percentages of apoptotic (Annexin V+) cells. Error bars represent SD. (E) Bioluminescence imaging of representative vehicle and PD 0332991 treated KOPTK1-Luc+ xenografted animals. (F) Total body bioluminescence (photons s1 ROI1) in vehicle and PD 0332991 treated recipients of KOPTK1-Luc+ cells was monitored over time to determine tumor burden. Shown are mean values, error bars represent SEM. (G) Kaplan-Meier survival curves of KOPTK1-Luc+ recipients treated with vehicle or PD 0332991. (H) Bioluminescence imaging of representative vehicle and PD 0332991 treated DND41-Luc+ xenografted animals. (I) Total body bioluminescence (photons s1 ROI1) in vehicle and PD 0332991 treated recipients of DND41-Luc+ cells was monitored over time to determine tumor burden. Shown are mean values, error bars represent SEM. (J) Kaplan-Meier survival curves of DND41-Luc+ recipients treated with vehicle only or PD 0332991. n = 8 for all groups. See also Figure S7.

448 Cancer Cell 22, 438451, October 16, 2012 2012 Elsevier Inc.

Cancer Cell
D-Cyclins in Tumor Maintenance

that these functions are essential only in certain tumor types, but not in normal tissues, which renders these tumors uniquely sensitive to cyclin D-CDK inhibition. In contrast to the impact of cyclin D1 inhibition in breast tumors, where it caused tumor cell senescence, we found that the shutdown of cyclin D3 in lymphoid tumors triggered apoptosis. Importantly, we found that cyclin D3-CDK inhibition also resulted in killing of human T-ALL cells in vitro and in vivo. Intriguingly, this effect seems to be specic to NOTCH1-ICD-positive T-ALL, suggesting a synthetic lethal interaction between activation of the Notch1 pathway and cyclin D-CDK kinase inhibition in leukemic cells. Although the exact molecular mechanism of the observed cell death remains to be determined, our analyses suggest that inhibition of cyclin D-CDK4/6 kinase activity in NOTCH1-ICDpositive T-ALL cells triggers a transcriptional apoptotic program, which likely contributes to killing of tumor cells. Altogether, our study demonstrates that in addition to the wellestablished roles of overexpressed D-cyclins in tumor initiation, the presence of D-cyclins is essential for tumor maintenance. Our results show that inhibition of cyclin D-kinase activity represents a highly selective anticancer strategy that specically targets cancer cells without signicantly affecting normal tissues.
EXPERIMENTAL PROCEDURES Full experimental procedures and any associated references are available in the Supplemental Experimental Procedures. Generation of Conditional Cyclin D1F/F and D3F/F Mice Please see Supplemental Experimental Procedures for details. All animal procedures conformed to the relevant regulatory standards, guidelines, and regulations, and were approved by the Dana Farber Cancer Institute Animal Care and Users Committee. Genotyping and deletion analysis of oxed or deleted cyclin D1 and D3 alleles were performed with primers: D1-50 (50 -GA GTTTTCCGGGTGCGTT-30 ), D1-30 (50 -CTGTGGTGTCGCTGACA-30 ), D1-D (50 -GGCAGTAGCAAGATCTGTTA-30 ) and D3-50 (50 -CTGCGTTCTGTCCCTTT CCTT-30 ), D3-30 (50 -CGCGATAGACACAGGAACCA-30 ), and D3-D (50 -CCA GACTGGAGCCAGAGATAA-30 ) by predenaturation of genomic DNA at 94 C for 5 min, followed by 33 cycles of amplication: 94 C for 30 s, 60 C for 30 s, 72 C for 30 s, and a nal extension step at 72 C for 10 min. The sizes of PCR products are: for cyclin D1wt = 180 base pairs (bp), D1F = 420 bp, D1D = 476 bp, D3wt = 210 bp, D3F = 500 bp, and D3D = 450 bp. Cyclin D1F/F mice were bred with cyclin D1/ (Sicinski et al., 1995), Esr1-Cre, and MMTV-ErbB2 animals. Cyclin D3F/F mice were bred with Esr1-Cre and Mx1Cre animals. Mx1-Cre, MMTV-ErbB2, and Esr1-Cre mice were from Jackson Laboratory. Acute Ablation of Cyclins D1 or D3 and PD 0332991 Treatment Esr1-Cre expressing mice were injected intraperitoneally with 4-hydroxytamoxifen (Sigma) 0.225 mg/g of body weight on alternate days. To inhibit cyclin D-Cdk kinase, mice were fed daily with PD 0332991, 150 mg/kg of body weight by gastric gavage; every 2 weeks the daily dose was lowered to 100 mg/kg for 23 days. Control mice were treated with vehicle (10% 0.1N HCl, 10% Cremaphor EL, 20% PEG300, 60% 50 mM citrate buffer pH 4.5) 10 ml/kg by gastric gavage. Mx1-Cre expressing mice were injected intraperitoneally with 2.5 mg/g body weight of polyinosinic-polycytidylic acid on alternate days (pI-pC, Amersham). Flow Cytometry Analysis and Sorting Cells were stained with antibodies against CD4 and CD8 (BD Biosciences). For cell cycle analyses, mice were intraperitoneally injected with 12 mg BrdU and sacriced after 2hr. Tissues were harvested and analyzed with BrdU-Flow kit using either anti-BrdU antibody conjugated to phycoerythrin (PE) or uorescein isothiocyanate (FITC) (BD Biosciences). Annexin V/7-AAD staining was

performed on fresh cells according to the manufacturers specications (ApoScreen, Southern Biotech). Cells from human T-ALL xenotransplants were stained with anti-human CD45 antibody (Biolegend). Alternatively cells were xed, permeabilized, and stained with phycoerythrin (PE)-mouse antiphospho pRbS780 antibody (BD Biosciences) or isotype IgG1 control and analyzed by FACS. All data acquisition was performed on LSRII, Fortessa and Facscan and analyzed using Cell Quest and FACSDiva software (BD Biosciences). Sorting was conducted on Moow or FACSaria sorters (BD Biosciences). In Vitro Analyses of Leukemic Cells Spleens were collected from moribund vehicle-treated recipients of Notch1ICD- and GFP-expressing tumor cells and cultured for 4 days in RPMI supplemented with 10% fetal calf serum (Sigma) and 10 mM glutamine in the presence of 1mM PD 0332991 or vehicle. Human T-ALL cell lines were cultured as described in Supplemental Experimental Procedures. Cells were then stained as described above and analyzed by ow cytometry. Human T-ALL Xenografts KOPTK1 and DND41 cells were stably transduced with the FUW-LucmCherry-puro lentivirus (Kimbrel et al., 2009) to generate KOPTK1-Luc+ and DND41-Luc+ cells. To establish orthotopic xenografts, 2 3 106 engineered cells were injected intravenously via the lateral tail vein into NOD-SCIDIL2Rgnull (NSG) mice. Bioluminescence imaging was performed by injecting mice with 75 mg/kg of D-Luciferin (Promega, Madison, WI), followed by imaging on an IVIS Spectrum (Caliper Life Sciences, Hopkinton, MA). Total body bioluminescence was determined by quantifying photon ux through standardized regions of interest using the Living Images software package (Caliper Life Sciences). Mice with established disease, dened by increasing bioluminescence, were entered onto treatment with either PD 0332991 150 mg/kg by gastric gavage, or vehicle 10 ml/kg by gastric gavage. Mice were imaged at the indicated time points, and were sacriced when they showed signs of distress. Statistical Calculations Statistical signicance was calculated by log rank (Mantel-Cox) test for the survival curves and two-way ANOVA test for the bioluminescence data. All other p values were calculated by Students unpaired t test. ACCESSION NUMBERS Expression data can be found at http://www.ncbi.nlm.nih.gov/geo/ under superseries accession number GSE40514. SUPPLEMENTAL INFORMATION Supplemental Information includes seven gures, four tables, and Supplemental Experimental Procedures and can be found with this article online at http://dx.doi.org/10.1016/j.ccr.2012.09.015. ACKNOWLEDGMENTS We thank Drs. K. Kozar for initial help with gene-targeting constructs; E. Sicinska, Y. Geng, L. Anders, and Q. Yu for reagents, advice, and expertise. P.H. was partly supported by a fellowship from the Swedish Wennergren Foundations. This work was supported by NIH Grants R01 CA083688 and P01 CA080111 to P.S., P01 CA109901 to A.T.L., H.v.B., and P.S., and Claudia Adams Barr grant to X.L. P.S. is a consultant and a recipient of a research grant from Novartis. T.S. is supported by grants from the National Cancer Institute (1K99CA157951), the William Lawrence and Blanche Hughes Foundation, the Childrens Leukemia Research Association, and the Japan Society for the Promotion of Science. Received: January 7, 2012 Revised: June 24, 2012 Accepted: September 18, 2012 Published: October 15, 2012

Cancer Cell 22, 438451, October 16, 2012 2012 Elsevier Inc. 449

Cancer Cell
D-Cyclins in Tumor Maintenance

REFERENCES Anders, L., Ke, N., Hydbring, P., Choi, Y.J., Widlund, H.R., Chick, J.M., Zhai, H., Vidal, M., Gygi, S.P., Braun, P., and Sicinski, P. (2011). A systematic screen for CDK4/6 substrates links FOXM1 phosphorylation to senescence suppression in cancer cells. Cancer Cell 20, 620634. 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. Bodrug, S.E., Warner, B.J., Bath, M.L., Lindeman, G.J., Harris, A.W., and Adams, J.M. (1994). Cyclin D1 transgene impedes lymphocyte maturation and collaborates in lymphomagenesis with the myc gene. EMBO J. 13, 21242130. Bowe, D.B., Kenney, N.J., Adereth, Y., and Maroulakou, I.G. (2002). Suppression of Neu-induced mammary tumor growth in cyclin D1 decient mice is compensated for by cyclin E. Oncogene 21, 291298. Burns, K.H., Agno, J.E., Sicinski, P., and Matzuk, M.M. (2003). Cyclin D2 and p27 are tissue-specic regulators of tumorigenesis in inhibin alpha knockout mice. Mol. Endocrinol. 17, 20532069. Chaganti, R.S., and Houldsworth, J. (2000). Genetics and biology of adult human male germ cell tumors. Cancer Res. 60, 14751482. Chen, D., Pacal, M., Wenzel, P., Knoeper, P.S., Leone, G., and Bremner, R. (2009). Division and apoptosis of E2f-decient retinal progenitors. Nature 462, 925929. Chiang, M.Y., Xu, M.L., Histen, G., Shestova, O., Roy, M., Nam, Y., Blacklow, S.C., Sacks, D.B., Pear, W.S., and Aster, J.C. (2006). Identication of a conserved negative regulatory sequence that inuences the leukemogenic activity of NOTCH1. Mol. Cell. Biol. 26, 62616271. Chicas, A., Wang, X., Zhang, C., McCurrach, M., Zhao, Z., Mert, O., Dickins, R.A., Narita, M., Zhang, M., and Lowe, S.W. (2010). Dissecting the unique role of the retinoblastoma tumor suppressor during cellular senescence. Cancer Cell 17, 376387. enz-Robles, M.T., Nair, V., Ferrey, A., Hagan, Chong, J.L., Wenzel, P.L., Sa J.P., Gomez, Y.M., Sharma, N., Chen, H.Z., Ouseph, M., et al. (2009). E2f1-3 switch from activators in progenitor cells to repressors in differentiating cells. Nature 462, 930934. Cooper, A.B., Sawai, C.M., Sicinska, E., Powers, S.E., Sicinski, P., Clark, M.R., and Aifantis, I. (2006). A unique function for cyclin D3 in early B cell development. Nat. Immunol. 7, 489497. Dean, J.L., Thangavel, C., McClendon, A.K., Reed, C.A., and Knudsen, E.S. (2010). Therapeutic CDK4/6 inhibition in breast cancer: key mechanisms of response and failure. Oncogene 29, 40184032. Deshpande, A., Sicinski, P., and Hinds, P.W. (2005). Cyclins and CDKs in development and cancer: a perspective. Oncogene 24, 29092915. Fantl, V., Stamp, G., Andrews, A., Rosewell, I., and Dickson, C. (1995). Mice lacking cyclin D1 are small and show defects in eye and mammary gland development. Genes Dev. 9, 23642372. Fry, D.W., Harvey, P.J., Keller, P.R., Elliott, W.L., Meade, M., Trachet, E., Albassam, M., Zheng, X., Leopold, W.R., Pryer, N.K., and Toogood, P.L. (2004). Specic inhibition of cyclin-dependent kinase 4/6 by PD 0332991 and associated antitumor activity in human tumor xenografts. Mol. Cancer Ther. 3, 14271438. Fu, M., Wang, C., Li, Z., Sakamaki, T., and Pestell, R.G. (2004). Minireview: cyclin D1: normal and abnormal functions. Endocrinology 145, 54395447. Hayashi, S., and McMahon, A.P. (2002). Efcient recombination in diverse tissues by a tamoxifen-inducible form of Cre: a tool for temporally regulated gene activation/inactivation in the mouse. Dev. Biol. 244, 305318. Hu, M.G., Deshpande, A., Enos, M., Mao, D., Hinds, E.A., Hu, G.F., Chang, R., Guo, Z., Dose, M., Mao, C., et al. (2009). A requirement for cyclin-dependent kinase 6 in thymocyte development and tumorigenesis. Cancer Res. 69, 810818. Hulit, J., Wang, C., Li, Z., Albanese, C., Rao, M., Di Vizio, D., Shah, S., Byers, S.W., Mahmood, R., Augenlicht, L.H., et al. (2004). Cyclin D1 genetic heterozy-

gosity regulates colonic epithelial cell differentiation and tumor number in ApcMin mice. Mol. Cell. Biol. 24, 75987611. Ianari, A., Natale, T., Calo, E., Ferretti, E., Alesse, E., Screpanti, I., Haigis, K., Gulino, A., and Lees, J.A. (2009). Proapoptotic function of the retinoblastoma tumor suppressor protein. Cancer Cell 15, 184194. Kimbrel, E.A., Davis, T.N., Bradner, J.E., and Kung, A.L. (2009). In vivo pharmacodynamic imaging of proteasome inhibition. Mol. Imaging 8, 140147. hn, R., Schwenk, F., Aguet, M., and Rajewsky, K. (1995). Inducible gene Ku targeting in mice. Science 269, 14271429. Kuilman, T., and Peeper, D.S. (2009). Senescence-messaging secretome: SMS-ing cellular stress. Nat. Rev. Cancer 9, 8194. Landis, M.W., Pawlyk, B.S., Li, T., Sicinski, P., and Hinds, P.W. (2006). Cyclin D1-dependent kinase activity in murine development and mammary tumorigenesis. Cancer Cell 9, 1322. Li, S.K., Smith, D.K., Leung, W.Y., Cheung, A.M., Lam, E.W., Dimri, G.P., and Yao, K.-M. (2008a). FoxM1c counteracts oxidative stress-induced senescence and stimulates Bmi-1 expression. J. Biol. Chem. 283, 1654516553. Li, X., Gounari, F., Protopopov, A., Khazaie, K., and von Boehmer, H. (2008b). Oncogenesis of T-ALL and nonmalignant consequences of overexpressing intracellular NOTCH1. J. Exp. Med. 205, 28512861. Lin, Y., Bloodgood, B.L., Hauser, J.L., Lapan, A.D., Koon, A.C., Kim, T.K., Hu, L.S., Malik, A.N., and Greenberg, M.E. (2008). Activity-dependent regulation of inhibitory synapse development by Npas4. Nature 455, 11981204. ro y, T. (1994). Cyclin Lovec, H., Grzeschiczek, A., Kowalski, M.B., and Mo D1/bcl-1 cooperates with myc genes in the generation of B-cell lymphoma in transgenic mice. EMBO J. 13, 34873495. Malumbres, M., and Barbacid, M. (2009). Cell cycle, CDKs and cancer: a changing paradigm. Nat. Rev. Cancer 9, 153166. Muller, W.J., Sinn, E., Pattengale, P.K., Wallace, R., and Leder, P. (1988). Single-step induction of mammary adenocarcinoma in transgenic mice bearing the activated c-Neu oncogene. Cell 54, 105115. Musgrove, E.A., Caldon, C.E., Barraclough, J., Stone, A., and Sutherland, R.L. (2011). Cyclin D as a therapeutic target in cancer. Nat. Rev. Cancer 11, 558572. ~nez, S., Heard, E., Narita, M., Lin, A.W., Hearn, S.A., Spector, Narita, M., Nu D.L., Hannon, G.J., and Lowe, S.W. (2003). Rb-mediated heterochromatin formation and silencing of E2F target genes during cellular senescence. Cell 113, 703716. ONeil, J., Grim, J., Strack, P., Rao, S., Tibbitts, D., Winter, C., Hardwick, J., Welcker, M., Meijerink, J.P., Pieters, R., et al. (2007). FBW7 mutations in leukemic cells mediate NOTCH pathway activation and resistance to gamma-secretase inhibitors. J. Exp. Med. 204, 18131824. Park, H.J., Carr, J.R., Wang, Z., Nogueira, V., Hay, N., Tyner, A.L., Lau, L.F., Costa, R.H., and Raychaudhuri, P. (2009). FoxM1, a critical regulator of oxidative stress during oncogenesis. EMBO J. 28, 29082918. n, A., Dubus, P., Mulero, F., Pizcueta, P., Khan, G., Guerra, C., Puyol, M., Mart a, D., and Barbacid, M. (2010). A synthetic lethal interaction between Santamar K-Ras oncogenes and Cdk4 unveils a therapeutic strategy for non-small cell lung carcinoma. Cancer Cell 18, 6373. a, X., Litvin, J., and Reddy, E.P. Reddy, H.K., Mettus, R.V., Rane, S.G., Gran (2005). Cyclin-dependent kinase 4 expression is essential for Neu-induced breast tumorigenesis. Cancer Res. 65, 1017410178. Robles, A.I., Rodriguez-Puebla, M.L., Glick, A.B., Trempus, C., Hansen, L., Sicinski, P., Tennant, R.W., Weinberg, R.A., Yuspa, S.H., and Conti, C.J. (1998). Reduced skin tumor development in cyclin D1-decient mice highlights the oncogenic ras pathway in vivo. Genes Dev. 12, 24692474. Ruas, M., Gregory, F., Jones, R., Poolman, R., Starborg, M., Rowe, J., Brookes, S., and Peters, G. (2007). CDK4 and CDK6 delay senescence by kinase-dependent and p16INK4a-independent mechanisms. Mol. Cell. Biol. 27, 42734282. rez-Mancera, P.A., Wysocki, J.M., and Jacks, T. Sage, J., Miller, A.L., Pe (2003). Acute mutation of retinoblastoma gene function is sufcient for cell cycle re-entry. Nature 424, 223228.

450 Cancer Cell 22, 438451, October 16, 2012 2012 Elsevier Inc.

Cancer Cell
D-Cyclins in Tumor Maintenance

Sherr, C.J., and Roberts, J.M. (2004). Living with or without cyclins and cyclindependent kinases. Genes Dev. 18, 26992711. Sicinska, E., Aifantis, I., Le Cam, L., Swat, W., Borowski, C., Yu, Q., Ferrando, A.A., Levin, S.D., Geng, Y., von Boehmer, H., and Sicinski, P. (2003). Requirement for cyclin D3 in lymphocyte development and T cell leukemias. Cancer Cell 4, 451461. Sicinska, E., Lee, Y.M., Gits, J., Shigematsu, H., Yu, Q., Rebel, V.I., Geng, Y., Marshall, C.J., Akashi, K., Dorfman, D.M., et al. (2006). Essential role for cyclin D3 in granulocyte colony-stimulating factor-driven expansion of neutrophil granulocytes. Mol. Cell. Biol. 26, 80528060. Sicinski, P., Donaher, J.L., Parker, S.B., Li, T., Fazeli, A., Gardner, H., Haslam, S.Z., Bronson, R.T., Elledge, S.J., and Weinberg, R.A. (1995). Cyclin D1 provides a link between development and oncogenesis in the retina and breast. Cell 82, 621630. Trimarchi, J.M., and Lees, J.A. (2002). Sibling rivalry in the E2F family. Nat. Rev. Mol. Cell Biol. 3, 1120. Tsikitis, M., Zhang, Z., Edelman, W., Zagzag, D., and Kalpana, G.V. (2005). Genetic ablation of Cyclin D1 abrogates genesis of rhabdoid tumors resulting from Ini1 loss. Proc. Natl. Acad. Sci. USA 102, 1212912134. Wang, T.C., Cardiff, R.D., Zukerberg, L., Lees, E., Arnold, A., and Schmidt, E.V. (1994). Mammary hyperplasia and carcinoma in MMTV-cyclin D1 transgenic mice. Nature 369, 669671.

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). Activating mutations of NOTCH1 in human T cell acute lymphoblastic leukemia. Science 306, 269271. Weinstat-Saslow, D., Merino, M.J., Manrow, R.E., Lawrence, J.A., Bluth, R.F., Wittenbel, K.D., Simpson, J.F., Page, D.L., and Steeg, P.S. (1995). Overexpression of cyclin D mRNA distinguishes invasive and in situ breast carcinomas from non-malignant lesions. Nat. Med. 1, 12571260. Young, A.P., and Longmore, G.D. (2004). Differential regulation of apoptotic genes by Rb in human versus mouse cells. Oncogene 23, 25872599. Yu, Q., Geng, Y., and Sicinski, P. (2001). Specic protection against breast cancers by cyclin D1 ablation. Nature 411, 10171021. m, M., Zagozdzon, A., Kong, Y., Yu, Q., Sicinska, E., Geng, Y., Ahnstro l, O., and Sicinski, P. (2006). Gardner, H., Kiyokawa, H., Harris, L.N., Sta Requirement for CDK4 kinase function in breast cancer. Cancer Cell 9, 2332. Zhang, Q., Sakamoto, K., Liu, C., Triplett, A.A., Lin, W.C., Rui, H., and Wagner, K.U. (2011). Cyclin D3 compensates for the loss of cyclin D1 during ErbB2induced mammary tumor initiation and progression. Cancer Res. 71, 7513 7524.

Cancer Cell 22, 438451, October 16, 2012 2012 Elsevier Inc. 451

Article
Therapeutic Targeting of the Cyclin D3:CDK4/6 Complex in T Cell Leukemia
Catherine M. Sawai,1,6 Jacquelyn Freund,1 Philmo Oh,1 Delphine Ndiaye-Lobry,1 Jamieson C. Bretz,2 Alexandros Strikoudis,1 Lali Genesca,3 Thomas Trimarchi,1 Michelle A. Kelliher,4 Marcus Clark,5 Jean Soulier,3 Selina Chen-Kiang,2 and Iannis Aifantis1,*
of Pathology and Howard Hughes Medical Institute, New York University School of Medicine, New York, NY 10016, USA of Pathology and Graduate Program in Immunology and Microbial Pathogenesis, Weill Medical College of Cornell University, New York, NY 10065, USA 3INSERM U944 and University Paris Diderot, Saint-Louis Hospital, Paris, 75010 France 4Department of Cancer Biology, University of Massachusetts Medical School, Worcester, MA 01605, USA 5Department of Medicine, Section of Rheumatology and Gwen Knapp Center for Lupus Research, University of Chicago, Chicago, IL 60637, USA 6Present address: Department of Microbiology and Immunology, Columbia University, New York, NY 10032, USA *Correspondence: iannis.aifantis@nyumc.org http://dx.doi.org/10.1016/j.ccr.2012.09.016
2Department 1Department

Cancer Cell

SUMMARY

D-type cyclins form complexes with cyclin-dependent kinases (CDK4/6) and promote cell cycle progression. Although cyclin D functions appear largely tissue specic, we demonstrate that cyclin D3 has unique functions in lymphocyte development and cannot be replaced by cyclin D2, which is also expressed during blood differentiation. We show that only combined deletion of p27Kip1 and retinoblastoma tumor suppressor (Rb) is sufcient to rescue the development of Ccnd3/ thymocytes. Furthermore, we show that a small molecule targeting the kinase function of cyclin D3:CDK4/6 inhibits both cell cycle entry in human T cell acute lymphoblastic leukemia (T-ALL) and disease progression in animal models of T-ALL. These studies identify unique functions for cyclin D3:CDK4/6 complexes and suggest potential therapeutic protocols for this devastating blood tumor.
INTRODUCTION D-type cyclins (D1, D2, and D3) bind cyclin-dependent kinases 4 and 6 (CDK4/6), and the activity of cyclin D:CDK4/6 complexes promotes entry into the cell cycle (Sherr, 1995; Sherr and Roberts, 2004). Cyclin D:CDK4/6 complexes are believed to promote cell cycle progression through at least two functions: by interacting with cell cycle inhibitors, such as p21Cip1 and p27Kip1, and by the phosphorylation of the retinoblastoma tumor suppressor (Rb). Cyclin D:CDK4/6 are thought to form ternary complexes that bind cyclin-dependent kinase inhibitors (CDKIs) of the p21Cip/p27Kip1 family (Sherr and Roberts, 2004). This facilitates downstream cyclin E:CDK2 complex activity that, along with cyclin D:CDK4/6, inactivates Rb and allows activation of E2F transcription factors and progression through the cell cycle. The functions of D-type cyclins have been studied using germline gene deletion. Each knockout mouse was viable, but displayed distinct tissue-specic defects (Ciemerych et al., 2002; Kozar et al., 2004; Sicinski et al., 1995, 1996; Sicinska et al., 2003, 2006). When these deciencies were combined, complete hematopoietic failure was observed, demonstrating the absolute requirement for D-type cyclins within the hematopoietic system (Kozar et al., 2004). Cyclin D2-decient (Ccnd2/) mice display reduced proliferation of mature splenic B cells and lack CD5+ peritoneal B cells (Solvason et al., 2000). Cyclin D3 knockout (Ccnd3/) animals show defects in early B and T cell differentiation, as well as impaired proliferation of

Signicance While dispensable for proliferation of many tissues, D-type cyclins are absolutely required in the hematopoietic system, specically for early lymphocyte development. Aberrant expression of D-type cyclins is associated with hematopoietic neoplasms; thus deciphering the functions of cyclin D is critical for the development of strategies to prevent an oncogenic cell cycle. We present genetic evidence that combined deletion of p27Kip1 and Rb, which regulate progression through the G1 phase of the cell cycle, rescue the defect in early T cell development from deletion of cyclin D3. Furthermore, we show that inhibition of cyclin D:CDK4/6 activity abrogates proliferation in T-ALL cell lines and primary human cells as well as progression of disease in animal models of T-ALL.
452 Cancer Cell 22, 452465, October 16, 2012 2012 Elsevier Inc.

Cancer Cell
Cyclin D3:CDK4/6 Role in the Progression of T-ALL

granulocytes (Cooper et al., 2006; Peled et al., 2010; Sicinska et al., 2003, 2006). Cyclin D1 was recently suggested to play a key role in hematopoietic stem cell quiescence and self-renewal (Zou et al., 2011); however, Ccnd1/ mice do not display striking hematopoietic effects, most likely due to redundancy with D2 and D3 (Sicinski et al., 1995). Previous work has suggested that defects associated with individual cyclin D deciency stem from their tissue-specic expression and that D-type cyclins are largely functionally redundant. For example, high expression of cyclin D1 protein, but not D2 or D3, is observed in both the retina and mammary tissue, and Ccnd1/animals correspondingly have reduced proliferation of both the cells that contribute to the retina and breast epithelium compartment (Sicinski et al., 1995). Genetic studies in which endogenous Ccnd1 was substituted with Ccnd2 complementary DNA (cDNA) have demonstrated that cyclin D2 can largely replace cyclin D1 function in mammary and retina tissue development (Carthon et al., 2005). However, these tissues typically express a single D-type cyclin, so whether D-type cyclins can functionally replace one another in cells that express more than one cyclin, such as developing lymphocytes, remains unclear. Aberrant cell cycle regulation is a common thread to all forms of cancer (Hunter and Pines, 1994). Deregulated expression of all D-type cyclins is frequently observed in hematopoietic malignancies (Bergsagel et al., 2005; Motokura and Arnold, 1993). We have previously shown that induction of T cell acute lymphoblastic leukemia (T-ALL), a disease caused by transformation of lymphocyte progenitors, requires cyclin D3, as expression of the oncogenic intracellular domain of Notch1 (ICN1) in Ccnd3/ bone marrow progenitors fails to initiate disease. Consistent with these animal studies, cyclin D overexpression is commonly seen in human T-ALL, with specic cyclin D expression associated with distinct T-ALL subsets (Li et al., 2008; Sicinska et al., 2003). Early thymocyte progenitor-ALL is characterized by cyclin D2 overexpression (Coustan-Smith et al., 2009), whereas more mature forms of T-ALL are associated with D3 overexpression (Joshi et al., 2009; Li et al., 2008). Finally, previous data have suggested that Notch signaling directly regulates cyclin D3 expression, and blocking cyclin D3 expression by g-secretase inhibition of Notch signaling prevents cell cycle progression in human T-ALL cell lines in vitro (Joshi et al., 2009). These data suggested that D-type cyclins and/or their downstream interacting partners could be attractive therapeutic targets in this type of disease. RESULTS Unique Roles for Cyclin D3 in Lymphocyte Development We have previously shown that cyclins D2 and D3 are both expressed during early stages of lymphocytic differentiation; however, only loss of cyclin D3 leads to signicant effects on cell differentiation (Cooper et al., 2006; Sicinska et al., 2003). To genetically test the ability of cyclin D2 to replace cyclin D3 function, we generated mice in which Ccnd2 cDNA was targeted to the Ccnd3 locus, such that Ccnd2 was regulated by the Ccnd3 50 and 30 untranslated region (Figure S1 available online). The unique Ccnd3/2 transcript generated from the knock-in allele was not detected in wild-type,

Ccnd2/, or Ccnd3/ cells using quantitative PCR (qPCR) analysis (Figure 1A). This unique transcript was specically produced in Ccnd3+/Ccnd2-Neo lymphocytes at low levels, but deletion of the neomycin resistance cassette resulted in a signicant increase in messenger RNA (mRNA) expression in Ccnd3+/Ccnd2 cells. Analysis of total Ccnd2 mRNA showed comparable expression in Ccnd3+/Ccnd2 cells to that of wildtype cells. As expected, Ccnd2 transcripts were decreased in Ccnd2+/ thymocytes and diminished in Ccnd2/ cells (Figure 1B). To further conrm knock-in allele expression, we generated Ccnd2/Ccnd3+/Ccnd2 animals and analyzed cyclin D2 expression in lymphocytes. Although the protein was not detected in Ccnd2/Ccnd3+/Ccnd2-Neo cells that retained the neomycin resistance cassette, cyclin D2 was readily detected in cells from knock-in animals (Ccnd2/ Ccnd3+/Ccnd2) that had deleted the selection cassette (Figure 1C). Furthermore, cyclin D2 expression was increased in Ccnd3Ccnd2/Ccnd2 (hereafter, Ccnd3D2/D2) cells compared to both wild-type and Ccnd3/ cells, demonstrating that cyclin D2 protein was specically generated from the knock-in allele. Given the role of cyclin D3 in early thymopoiesis (Sicinska et al., 2003), we investigated T cell development in Ccnd3D2/D2 animals (Figure 1D). We found that thymus size and total thymus cellularity of Ccnd3D2/D2 animals was reduced similarly to that of Ccnd3/ mice, and both were signicantly decreased from those of controls. Loss of cyclin D3 is associated with an increase in the percentage of CD4CD8 double negative (DN) thymocytes and a corresponding decrease in CD4+CD8+ double positive (DP) cells (Sicinska et al., 2003); although the overall number of DN cells was not altered, the number of Ccnd3/ DP was signicantly reduced. We found that the total number of Ccnd3D2/D2 DN and DP cells was similar to that observed in Ccnd3/ animals (Figure 1E). Although loss of cyclin D3 did not affect the absolute number of DN cells generated, the percentage of DN cells was increased, while the percentage of DP cells was decreased compared to controls (Ccnd3/ DN 10.2% and DP 62.4%; control DN 2.3% and DP 83.8%). As observed in Ccnd3/ thymocytes, the percentage of Ccnd3D2/D2 DN cells was also increased and the percentage of DP cells was reduced compared to controls (Figure 1F). Finally, we quantied proliferation of Ccnd3D2/D2 pro- and pre-T cells, which was comparable to that observed in Ccnd3/ pro- and pre-T cells (Figure 1G); however, both of these populations showed a decrease in rates of proliferation in comparison to wild-type cells. Even with high cyclin D2 expression, cyclin D:CDK4/6 specic phosphorylation of Rb (S807/811) was reduced in Ccnd3D2/D2 thymocytes compared to wild-type and was similar to that observed in Ccnd3/ cells (Figure 1H). These data demonstrated that, in early Ccnd3D2/D2 thymocytes, despite the abundance of cyclin D2, Rb remained hypophosphorylated and the cells could not efciently pass into S-phase. In addition to its role in T cell development, cyclin D3 is also required for early B lymphopoiesis (Cooper et al., 2006). We found that the overall bone marrow cellularity in Ccnd3D2/D2 animals was reduced to that found in Ccnd3/ mice (Figure S1), while control bone marrow contained a signicantly higher number of cells. Furthermore, knock-in animals had a decreased number of pre-B cells compared to controls, and knock-in pre-B cells had an impaired ability to proliferate compared to
Cancer Cell 22, 452465, October 16, 2012 2012 Elsevier Inc. 453

Cancer Cell
Cyclin D3:CDK4/6 Role in the Progression of T-ALL

Figure 1. Cyclin D3 Has Unique Functions in Normal Lymphopoiesis


(AC) Total thymocytes were isolated from wild-type control, Ccnd2+/, Ccnd2/, Ccnd2/Ccnd3+/Ccnd2-Neo, Ccnd2/Ccnd3+/Ccnd2, or Ccnd3Ccnd2/Ccnd2 animals and analyzed for expression of (A) the unique transcript generated from the Ccnd3Ccnd2 knock-in allele and (B) total Ccnd2 transcript normalized to Gapdh by qPCR or (C) for cyclin D2, D3, and actin (loading control) expression by Western blot. Bands for cyclin D2 and actin were quantied and the relative amount of protein indicated with the wild-type control set to 1. (DF) Images of thymuses, left panel (D). Thymus cellularity of control, Ccnd3/, and Ccnd3D2/D2 mice was measured, and (E) the fraction of DN (CD4CD8) and DP (CD4+CD8+) cells were calculated using percentages from (F) ow cytometry of CD4 and CD8 expression. (G) Cell cycle status of pro- (CD4CD8CD44CD25+) and pre-T cells (CD4CD8CD44CD25lo) was measured by incorporation of DAPI using FACS analysis. (H) Wild-type control, Ccnd3/ and Ccnd3D2/D2 thymocytes were analyzed for Rb (S807/811), total Rb, cyclin D3, and actin by Western blot. Data represent mean SD (***p < 0.001) and are representative of at least three independent experiments. See also Figure S1.

454 Cancer Cell 22, 452465, October 16, 2012 2012 Elsevier Inc.

Cancer Cell
Cyclin D3:CDK4/6 Role in the Progression of T-ALL

Figure 2. Cyclin D3 Has Nonredundant Functions in the Induction of T-ALL


(A) Lethally irradiated B6.SJL animals were transplanted with 3 3 105 pMIGR1 ICN1-IRES-EGFP wild-type or Ccnd3D2/D2 bone marrow progenitors, and peripheral blood was collected and analyzed for DP cells by FACS. Peripheral blood analysis shown is 31 days upon transplant of transduced cells. (B) Kaplan-Meier curve shows the survival of transplanted mice for the period of observation (**p < 0.005). Data are representative of two independent experiments with ve mice per condition.

controls. Collectively, these results indicated that cyclin D2 expression could not rescue the defects in lymphocyte development caused by the lack of cyclin D3. Unique Roles for Cyclin D3 in Acute Lymphocytic Leukemia Although Ccnd2 expression from the Ccnd3 locus did not rescue the defect during the normal development of Ccnd3/ early lymphocytes, we wanted to test its ability to replace the requirement for cyclin D3 in oncogenic transformation, specically in T-ALL (Sicinska et al., 2003). We selected a Notch-driven model of T-ALL, as more than 90% of human T-ALL shows signs of constitutive Notch pathway activation and the majority of human T-ALL lines are addicted to Notch1 function (Palomero et al., 2007; Weng et al., 2004). We transduced lineagec-Kit+ bone marrow progenitors from wild-type and Ccnd3D2/D2 mice with retrovirus encoding the constitutively active intracellular domain of Notch1-IRES-enhanced green uorescent protein (EGFP) (ICN1-EGFP), as previously described (Espinosa et al., 2010; Vilimas et al., 2007; Walkley et al., 2005). These cells were transplanted into lethally irradiated congenic wild-type mice, and animals were monitored for the presence of CD4+CD8+ DP leukemic cells in the peripheral blood. At 2 weeks after transplant, ICN1-EGFP+ leukemic cells were detected in the blood of control animals. Furthermore, 1 month posttransplant, the peripheral blood of control animals contained a signicant number of ICN1-EGFP+ cells, of which 71% were DP cells, while only a minute number of ICN1-EGFP+ cells were observed in recipients of Ccnd3D2/D2 cells (Figure 2A). The development of disease in hosts that received wild-type cells transduced with ICN1 retrovirus was rapid, with all control animals succumbing to the disease 46 weeks after transplant (Figure 2B). In contrast,

animals that received transduced Ccnd3D2/D2 cells displayed signicant protection from disease, with all animals remaining disease-free for the entire 6-month period of observation. These results demonstrated a specic requirement for cyclin D3, but not cyclin D2, in the induction of T-ALL. Endogenous Ccnd2 Protein Induction Fails to Rescue Ccnd3/ Phenotypes Considering that Ccnd2 knock-in into the Ccnd3 locus did not rescue the specic requirements for cyclin D3, we further investigated the regulation of expression of these molecules in early lymphocytes. We readily detected cyclin D2 protein in puried wild-type pro-T (CD4825+44); however, this expression decreased in wild-type pre-T (CD4825low/neg44) cells (Figure S1). However, in Ccnd3/ cells, cyclin D2 protein was signicantly increased in pro- and pre-T cells as well as total thymocytes. This nding suggested that physiological cyclin D2 overexpression was not able to rescue the developmental block caused by cyclin D3 deciency, in agreement with our observations from the knock-in described in Figure 1. Cyclin D2 overexpression appears to be posttranscriptional, as no differences in Ccnd2 mRNA levels were detected from wild-type and Ccnd3/ total thymocytes (Figure S1). Furthermore, immunouorescence analysis of cyclin D2 and D3 protein from puried DN wild-type thymocytes conrmed previous ndings, as it showed preferential induction of cyclin D2 protein in Ccnd3/ thymocyte progenitors (Figure S1). Cyclin D3 expression remained unchanged in response to Ccnd2 deletion (data not shown; Figure S1). To further investigate regulation of cyclin D2 protein, we treated thymocytes with cycloheximide, which inhibits de novo protein synthesis, and analyzed expression over time. Cyclin D2 half-life was
Cancer Cell 22, 452465, October 16, 2012 2012 Elsevier Inc. 455

Cancer Cell
Cyclin D3:CDK4/6 Role in the Progression of T-ALL

signicantly increased in Ccnd3/ cells (2.5-4-fold) when compared to wild-type thymocytes (Figure S3D). We also observed that cyclin D2 protein was stabilized upon treatment with the proteasome inhibitor MG-132, suggesting regulation by the ubiquitin-proteasome system (Figure S1). These studies further support our previous hypothesis, as they demonstrate that physiological upregulation of cyclin D2 protein is unable to rescue normal lymphocytic differentiation and induction of leukemia. Partially Restores T Cell Development Loss of p27 in Ccnd3/ Mice We next sought to further dene the functions of cyclin D3 in regulating cell cycle during early lymphocyte development. Of the downstream cell cycle regulators, we focused on the cell cycle inhibitor p27Kip1, as mice decient for p27Kip1 show increased thymic cellularity (Fero et al., 1996; Nakayama et al., 1996), and p27Kip1 is the only CDKI dynamically regulated at early stages of T cell development (I.A., unpublished data). To genetically test the interaction between cyclin D3 and p27Kip1, we crossed Ccnd3/ and Cdkn1b/ mice. Analysis of cyclin D3 and p27 protein from total thymocytes of Ccnd3/ Cdkn1b/ mice conrmed the genotypes of the mice (Figure 3A). We measured total thymic cellularity and found that the average number of Ccnd3/Cdkn1b/ cells was signicantly increased compared to that of Ccnd3/ thymocytes, but remained signicantly reduced compared to wild-type controls (Figure 3B). Differences in total thymocyte numbers were associated with differences in percentages and average number of DP cells (Figures 3C and 3D). While Ccnd3/ Cdkn1b/ animals displayed signicantly higher percentages and numbers of DP thymocytes than Ccnd3/ animals, they had signicantly fewer DP cells than control mice. Taken together, these observations suggested that ablation of p27Kip1 only partially restores development of Ccnd3/ thymocytes. We have previously shown that the Ccnd3/ T cell defect stems from a reduction in S-phase entry and cell cycle progression (Sicinska et al., 2003); thus, we next assessed the cell cycle status of Ccnd3/Cdkn1b/ early lymphocytes. We evaluated cell cycle using DAPI and observed an increase from 11.6% to 17.5% of pre-T cells in S-G2-M phases of the cell cycle from Ccnd3/ and Ccnd3/Cdkn1b/ animals, respectively (Figure 3E). In contrast, the percentage of Ccnd3/Cdkn1b/ proliferating pre-T cells was still reduced compared to controls. To further investigate cell cycle, we measured CDK2-associated activity using an in vitro kinase assay. CDK2-containing complexes immunoprecipitated from Ccnd3/ total thymocytes showed little kinase activity toward exogenous Rb (Figure 3F). In contrast, Ccnd3/Cdkn1b/ thymocytes showed an increase in CDK2-associated activity, but this activity did not reach the levels of control CDK2 activity. These combined studies indicated that loss of p27Kip1 only partially restored the cell cycle and developmental defects of early Ccnd3/ T cells. Notch1-Induced Transformation of Ccnd3/Cdkn1b/ Cells We also tested the ability of Ccnd3/Cdkn1b/ cells to be transformed by Notch1 activation. Peripheral blood analysis 2 weeks after transplant revealed a small population of ICN1456 Cancer Cell 22, 452465, October 16, 2012 2012 Elsevier Inc.
Kip1

EGFP+ DP cells in hosts that received Ccnd3/Cdkn1b/ transduced cells (Figure S2A). Similar to control mice that received wild-type ICN1-transduced cells, all recipients of Ccnd3/Cdkn1b/ transduced cells developed disease, albeit with slightly delayed kinetics, and succumbed to disease by 10 weeks posttransplant (Figure S2B). These data show that additional loss of p27Kip1 was sufcient to rescue transformation and progression to disease of Ccnd3/ cells. Loss of Rb Partially Restores Ccnd3/ T Cell Development As loss of p27 did not completely rescue Ccnd3/ lymphopoiesis at the steady state, we hypothesized that remaining activities of additional key cell cycle regulators could prevent cells from properly entering the cell cycle in the absence of cyclin D3. We observed that Rb was expressed and phosphorylated in wild-type pre-T cells and that its expression, specically the cyclin D:CDK4/6-phosphorylated pRb (S807/811) species, was reduced in Ccnd3/ cells (Figure 4A). To genetically test the importance of Rb regulation, we generated cyclin D3/Rb doubly decient animals by crossing Ccnd3/ mice to Rb1F/F mice carrying the Mx1-Cre transgene. Administration of the doublestrand RNA mimic poly(I:C) to these animals induced Cre recombinase expression, mediating deletion of Rb1, and these animals hereafter are referred to as Ccnd3/Rb1/. Rb was not detectable in total thymocytes from Ccnd3/Rb1/ mice, demonstrating efcient deletion of the oxed Rb1 alleles (Figure 4B). As with Ccnd3/Cdkn1b/ mice, the thymic cellularity of Ccnd3/Rb1/ animals was signicantly increased compared to Ccnd3/ animals, yet signicantly decreased compared to controls (Figure 4C). Ccnd3/Rb1/ animals displayed trends in the percentages and average number of DP cells that were similar to those seen in Ccnd3/Cdkn1b/ mice (Figures 4D and 4E), indicating that deletion of Rb1 only partially restored Ccnd3/ early T cell development. We next analyzed the cell cycle of Ccnd3/Rb1/ pre-T cells and detected approximately 21% of cells in S-G2-M phases, whereas only 11.5% of Ccnd3/ pre-T cells had transitioned beyond G1 (Figure 4F). However, the percentage of Ccnd3/Rb1/ proliferating preT cells was signicantly lower than that of controls (approximately 30%). We also tested CDK2-associated kinase activity in Ccnd3/Rb1/ total thymocytes by in vitro kinase assay. CDK2-containing complexes isolated from Ccnd3/Rb1/ T cells showed partially restored ability to phosphorylate exogenous Rb compared to Ccnd3/ thymocytes (Figure 4G). However, the CDK2 kinase activity in control thymocytes was signicantly higher, providing an explanation for the only partial restoration of progenitor cell proliferation. Combined Loss of p27Kip1 and Rb Restores Ccnd3/ T Cell Development Our genetic analyses showed that neither p27Kip1 nor Rb loss was sufcient to completely rescue Ccnd3/ developmental defects. Although there could be several alternative explanations for this incomplete rescue (including redundancy with other CDKI or Rb pocket proteins), we hypothesized that concomitant inactivation of p27Kip1 and Rb could provide efcient rescue, suggesting that cyclin D3 acts by simultaneously altering pRb

Cancer Cell
Cyclin D3:CDK4/6 Role in the Progression of T-ALL

Figure 3. Loss of p27 Partially Rescues the Development of Ccnd3/ T Cell Progenitors
(A) Total thymocytes analyzed for cyclin D3, p27, and actin expression by immunoblot. (BD) Total thymus cellularity of littermate control, Ccnd3/Cdkn1b/, and Ccnd3/ mice was quantied (B), and the relative number of DN and DP cells was calculated using total cell numbers and percentages from ow cytometry (C and D). (E) The proliferation of pro- and pre-T cells was analyzed by measuring DAPI incorporation by ow cytometric analysis. (F) Protein lysates were subjected to immunoprecipitation with antibody to CDK2 or normal rabbit immunoglobulin G (IgG) as control, and the ability of precipitated complexes to phosphorylate recombinant Rb was measured by in vitro kinase assay. Results represent mean SD (*p < 0.05, **p < 0.01, ***p < 0.0005, ****p < 0.00005) and are representative of three independent experiments of three mice of each genotype. See also Figure S2.

activity and binding p27Kip1. We thus generated Ccnd3/ Cdkn1b/Rb1/ compound mutant animals and analyzed early T cell development. We found that ablation of both p27Kip1 and Rb in Ccnd3/ animals resulted in a signicant increase in total thymocyte numbers, comparable to that of controls (Figure 5A). The number of DN cells in compound mutant animals was also similar to wild-type thymuses (Figure 5B). Moreover, Ccnd3/Cdkn1b/Rb1/ and control thymuses contained an almost identical number of DP cells

(Figures 5B and 5C). Finally, we measured pre-T cell proliferation. Only 13.0% of Ccnd3/ cells were in S-G2-M phases (Figure 5D). In contrast, approximately 29% of Ccnd3/ Cdkn1b/Rb1/ pre-T cells were in S-G2-M phases, a percentage similar to that found in control littermate animals. These results indicated that only simultaneous genetic deletion of p27Kip1 and Rb was sufcient to completely rescue cell cycle progression and the development of Ccnd3/ early T cells.
Cancer Cell 22, 452465, October 16, 2012 2012 Elsevier Inc. 457

Cancer Cell
Cyclin D3:CDK4/6 Role in the Progression of T-ALL

(A and B) Analysis of total Rb, pRb (S807/811), cyclin D3, and actin by immunoblot from pro- or pre-T puried from wild-type, Ccnd3/Rb1/, and Ccnd3/ total thymocytes. (CE) Total thymus numbers of littermate control, Ccnd3/Rb1/, and Ccnd3/ mice were quantied (C), and the relative number of DN and DP cells was calculated using total cell numbers and percentages from ow cytometry (D and E). (F) Cell cycle analysis of pro- and pre-T cells by DAPI incorporation measured by FACS. (G) Immune complexes were pulled-down with anti-Cdk2 antibody or normal rabbit IgG, and phosphorylation of exogenous Rb was measured by kinase assay. Results represent mean SD (*p < 0.05, ***p < 0.0005) and are representative of three independent experiments of three mice of each genotype.

Figure 4. Deletion of Rb Partially Rescues the Development of Ccnd3/ T Cell Progenitors

458 Cancer Cell 22, 452465, October 16, 2012 2012 Elsevier Inc.

Cancer Cell
Cyclin D3:CDK4/6 Role in the Progression of T-ALL

Figure 5. Combined Loss of p27 and Rb Restores Ccnd3/ T Lymphopoiesis

(AC) Total thymus cellularity of littermate control, Ccnd3/Cdkn1b/Rb1/, and Ccnd3/ mice was quantied (A), and the relative number of DN and DP cells was calculated using total cell numbers and percentages from ow cytometry (B and C). (D) Cell cycle status of pro- and pre-T cells was measured by DAPI incorporation by ow cytometric analysis. Results represent mean SD (**p < 0.005) and are representative of two independent experiments of four mice of each genotype.

Inhibition of Cyclin D3:CDK4/6 Complex Activity Suppresses Human T-ALL Cell Growth Our studies suggested that cyclin D3 has unique functions in lymphocyte development and transformation, likely in conjunction with CDK4/6, to titrate CDKIs and inactivate Rb. We have previously shown that cyclin D3 expression is essential for the induction of Notch-driven T-ALL (Sicinska et al., 2003). However, these studies did not address the potential of cyclin D3:CDK4/6 targeting during disease progression, a question of signicant clinical relevance. To address this question, we used PD0332991, a CDK4/6 specic small molecule inhibitor currently in clinical trials for multiple myeloma treatment (Baughn et al., 2006; Marzec et al., 2006; Menu et al., 2008). Initially, to conrm interaction between cyclin D3 and CDK4/6 in T cells, we performed immunoprecipitation of endogenous CDK6 in wild-type

thymocytes and observed specic interaction with cyclin D3 (Figure 6A). Similar cyclin D3:CDK4/6 interaction was also observed in human T-ALL cell lines. Having conrmed this interaction, we initiated in vitro treatments of mouse and human T-ALL cell lines with PD-0332991. All human T-ALL lines utilized carried NOTCH1 mutations, and the majority was absolutely dependent on Notch activity. Although the mouse T-ALL lines were driven by overexpression of TAL1, they also contained Notch1-truncating PEST mutations (ONeil et al., 2006). We tested PD-0332991 at both 0.5 and 1 mM and found similar effects in vitro. PD-0332991 treatment efciently inhibited S-phase entry of all cell lines within 15 hr, leading to accumulation of cells in G0/G1 phases (Figure 6B). The effects of drug treatment were reversible, as removal of PD-0332991 led to efcient re-entry in the cell cycle (Figure S3). To further expand
Cancer Cell 22, 452465, October 16, 2012 2012 Elsevier Inc. 459

Cancer Cell
Cyclin D3:CDK4/6 Role in the Progression of T-ALL

Figure 6. Inhibition of Cyclin D3:CDK4/6 Activity Inhibits T-ALL Cell Growth In Vitro
(A) Protein lysates from wild-type and Ccnd3/ thymocytes were prepared and subjected to pull-down with antibody to CDK6 or normal rabbit IgG and then analyzed by immunoblot cyclin D3 antibody.

460 Cancer Cell 22, 452465, October 16, 2012 2012 Elsevier Inc.

Cancer Cell
Cyclin D3:CDK4/6 Role in the Progression of T-ALL

these studies using primary leukemia samples, we in vitro treated cells isolated from two T-ALL patients. PD-0332991 treatment led to inhibition of cell cycle progression and accumulation in the G0/G1 cell cycle phase in the two T-ALL primary samples (Figure 6C). We also compared primary T-ALL leukemia cells at diagnosis to cells at relapse or after engraftment in immune-decient mice. We found a strong inhibition in all conditions, despite higher cycling rates in the relapse and xenograft cells, suggesting that PD-0332991 could be efcient in the treatment of relapsed T-ALL (Figure 6C). To gain a better molecular and biochemical understanding of PD-0332991 function on human T-ALL cells, we have used immunoblotting to dene expression and activation of known cell cycle regulators. PD-0332991 treatment efciently suppressed pRb (S807/811) phosphorylation and increased the expression of the p27Kip1 CDKI, both hallmarks of a G0/G1 arrest (Figure 6D). Whole-transcriptome analysis using four human T-ALL lines led to similar results (Figures 6E and 6F). PD0332991 treatment suppressed the expression of key mitosis regulators, including E2f2, Ccna2, Skp2, Cdc25a, Ccne2, and Cdt1. PD-0332991 treatment did not affect expression of Ccnd3 or Cdk4/Cdk6 mRNA. Gene set enrichment analysis (GSEA) showed that PD-0332991 treatment led to signicant gene expression correlation with gene sets related to cell cycle progression, including DNA replication, S-phase entry, and entry into mitosis (Figure 6F). Furthermore, after 4 days exposure to PD-0332991, we observed a signicant increase in annexin V expression compared to controls, indicating progression to cell death after treatment (Figure 6G). These combined studies suggested that, by inducing cell cycle arrest and apoptosis of leukemic cells, PD-0332991-mediated inhibition of cyclin D3:CDK4/6 activity could be an attractive therapy for T-ALL. Inhibition of Cyclin D3:CDK4/6 Complex Activity Inhibits T-ALL Progression In Vivo To test the ability of PD-0332991 to suppress disease progression in preclinical models of T-ALL, we treated animals that received cells expressing potent oncogenic forms of NOTCH1. Upon establishment of disease at week 3, we initiated treatment of leukemic mice by oral administration of PD-0332991 for ten consecutive days. The effects of the treatment were rapid and signicant, as all untreated control mice died by week 12, while the majority of PD-0332991-treated mice survived during the period of observation (Figure 7A). Peripheral white blood counts signicantly decreased, and ICN1-EGFP+ DP leukemic cells disappeared from the peripheral blood of the vast majority of

the treated animals (Figures 7B and 7C). In contrast, control animals displayed splenomegaly, and analysis of control splenocytes showed that the majority of cells were ICN1-EGFP+ DP leukemic cells (Figures 7D and 7E). Histologic examination demonstrated a signicant reduction of leukemic cell inltrations in all tissues studied from PD-0332991-treated animals (Figure 7F). Furthermore, ICN1-EGFP+ cells showed increased annexin V expression in PD-0332991-treated animals compared to controls (Figure 7G). T-ALL cells were addicted to CDK4/6 function, as interruption of drug administration led to relapse of the disease (data not shown), suggesting that the disease can be reinitiated by a small number of noncycling cells with leukemia-initiating abilities. To extend these studies using a xenograft model, we transplanted lethally irradiated Rag2/Il2rg/ mice with the human T-ALL CEM cell line and administered PD-0332991 3 weeks posttransplant for 14 consecutive days. Peripheral blood analysis at 4 weeks posttransplant showed 55% huCD45+ cells from untreated control mice compared to 6% huCD45+ cells from PD-0332991-treated mice. All untreated control mice died within 30 days of transplant, while the majority of treated mice survived during the observation period (Figure S4). Collectively, these studies demonstrated that cyclin D3:CDK4/6 inhibition was able to efciently suppress T-ALL progression leading to disease regression in vivo. DISCUSSION We demonstrate here that the cyclin D3:CDK4/6 complex has unique functions in the expansion of normally developing T cell progenitors and induction of T cell leukemia. We show that cyclin D2, a D-type cyclin also expressed in developing T cell progenitors, cannot replace cyclin D3. Although we detected cyclin D2 expression specically generated by the knock-in allele, Ccnd2 is unable to rescue the developmental defects caused by Ccnd3 deciency. Two additional ndings support the notion that cyclin D2 is unable to sustain expansion of thymocyte progenitors. Initially, we found high expression of cyclin D2 protein in pro-T cells; however, the cells are largely quiescent, suggesting that elevated levels of D-type cyclins do not always correlate with high rates of proliferation. Moreover, we were able to demonstrate that, in the absence of cyclin D3, D2 expression was signicantly elevated. This nding suggested that differentiating progenitors respond to the loss of cyclin D3 by upregulating cyclin D2; however, as with Ccnd3D2/D2 genetic replacement, this endogenous upregulation of cyclin D2 fails to

(B) Human and mouse (Ms.) T-ALL cell lines were treated with 1 mM PD-0332991 for 15 hr, followed by pulse with 30 mM BrdU. The percentage of cells in the S-phase was measured by FACS analysis of BrdU incorporation and propidium iodide (PI). (C) Primary human T-ALL cells (puried from the same patient) at diagnosis, relapse, and as a xenograft were grown on MS5-DL1 stromal cells for 2 days, then treated with 0.5 mM PD-0332991 for 15 hr and pulsed with 30 mM BrdU. The percentages of cells in the S-phase were calculated using percentages from FACS analysis of BrdU incorporation and PI. (D) Alternatively, cell lines were treated for 12 hr with 0.5 mM PD-0332991. Protein extracts were analyzed by SDS-PAGE followed by immunoblot with antibodies against p27, Rb, and pRb (Ser807/811), or actin as a loading control. (E) Heat map showing downregulation of genes associated with the S-phase of cell cycle and DNA replication upon treatment with 1 mM PD-0332991. (F) GSEA showing signicant downregulation of genes associated with DNA replication, mitotic cell cycle, the S-phase of the cell cycle, and the cell cycle checkpoints. (G) Human T-ALL cell lines were treated with 1 mM PD-0332991 for 4 days and analyzed for apoptosis by annexin V and 7-amino-actinomycin D. Data represent mean SD and are representative of at least three independent experiments. See also Figure S3.

Cancer Cell 22, 452465, October 16, 2012 2012 Elsevier Inc. 461

Cancer Cell
Cyclin D3:CDK4/6 Role in the Progression of T-ALL

Figure 7. PD-0332991 Inhibits Cyclin D3:CDK4/6 Activity and T-ALL Progression In Vivo
(A and B) Kaplan Meier survival curve of lethally irradiated CD45.1+ B6.SJL mice transplanted with ICN1-EGFP-transduced bone marrow. Mice were treated with 150 mg/kg PD-0332991 or vehicle for ten consecutive days beginning on day 21 (***p < 0.0001). Peripheral blood was analyzed for expression of GFP, CD4, and CD8 at both 21 and 28 days posttransplant. (C) White blood cell counts control and PD-0332991-treated animals were measured at the endpoint analysis, 7 weeks posttransplantation. PD-0332991-treated animals showed signicantly reduced blood counts compared to controls (**p < 0.01). (D and E) Spleens were harvested and control mice showed increased splenic mass compared to treated animals. Splenocytes were analyzed for GFP, CD4, and CD8 expression by FACS analysis. (F) Liver and spleen were parafn embedded and tissue sections were stained by hematoxylin and eosin staining (H&E). Scale bars correspond to 100 mm at 10X magnication. (G) Apoptosis in peripheral blood, liver, and lung was measured by expression of annexin V. Data are representative of two independent experiments. See also Figure S4.

rescue the Ccnd3/ phenotypes. These experiments suggest that the two cyclins have different functions during early hematopoiesis, in agreement with previous reports using distinct tissue systems. Indeed it was demonstrated that D-type cyclins could have distinct cellular localizations and different abilities to bind CDK inhibitors, such as p27 and p21 (Tamamori-Adachi et al., 2008). Moreover, biochemical studies showed differential substrate specicity between cyclin D1:CDK4 and cyclin D3:CDK4 complexes (Sarcevic et al., 1997). Finally, other studies suggested differential utilization of the LxCxE Rb-binding motif between cyclin D1 and D3 (Baker et al., 2005). Although we were able to demonstrate that cyclin D2 is unable to efciently phosphorylate Rb in the absence of cyclin D3, further studies
462 Cancer Cell 22, 452465, October 16, 2012 2012 Elsevier Inc.

are required to elucidate the mechanistic differences among the three kinases in hematopoiesis. In addition to its requirement in normal T cell development, cyclin D3 is required for the induction of T-ALL, a disease with rapid kinetics and characterized by increased rates of cell division. The inhibition of T-ALL induction is not due to a complete inhibition of T cell progenitor differentiation, as Ccnd3D2/D2 thymuses generate a signicant number of mature T cell receptor ab+ cells. Finally, the inability of cyclin D2 to rescue the Ccnd3/ phenotypes is unlike previous models of cyclin rescue experiments (Carthon et al., 2005; Geng et al., 1999). This phenotypic disparity may reect the requirement for precise regulation of cell cycle in lymphocytes, a cell type expressing

Cancer Cell
Cyclin D3:CDK4/6 Role in the Progression of T-ALL

more than one D-type cyclin, compared to cell types where a single D-type cyclin is expressed. Since the Ccnd3/ defects stemmed from the failure of lymphocyte progenitors to proliferate optimally, we focused on putative functions of cyclin D3 that directly regulate the cell cycle. Using genetic animal crosses, we demonstrated that the inhibitor p27Kip1 is only one of the critical regulators of cell cycle during T cell development. The p27Kip1 inhibitor appears to be a key regulator in early T cell development, as its protein expression is immediately decreased upon expression of the pre-T cell receptor, a prerequisite for both progression of T cell differentiation and induction of T-ALL (Aifantis et al., 2008). It would be interesting to determine the specic functions of p27Kip1 in developing T cells using specic mutants that disrupt the interaction between p27Kip1 and cyclin D:CDK4/6. While elimination of p27Kip1 is sufcient to allow normal proliferation of Ccnd1/ mammary and retina cells (Geng et al., 2001), loss of p27Kip1 does not restore optimal proliferation of Ccnd3/ T cells. Indeed, phosphorylation and inactivation of Rb function by cyclin D3:CDK4/6 complexes are another critical function, as demonstrated using the Ccnd3/Rb1/ animals. p27Kip1 and Rb are thought to have both cell-autonomous and cell-nonautonomous effects (Chien et al., 2006; Walkley et al., 2007). Although we cannot exclude cell-nonautonomous effects from contributing to the rescue of Ccnd3/ thymocyte development, we believe there is a cell-intrinsic component as well, since ICN1-EGFPtransduced Ccnd3/Cdkn1b/ cells were able to be transformed and produce disease after transplant into wild-type hosts (Figure S4). Our data provide genetic evidence suggesting that only simultaneous elimination of both layers of regulation sufciently lowers the threshold for developing progenitors to enter the cell cycle. These data suggested that targeting cyclin D3:CDK4/6 complex function can directly target induction and progression of T cell leukemia and, at the same time, cause minimal hematologic side effects, based on the phenotypes of animals lacking expression of cyclin D3 or CDK4/6 (Cooper et al., 2006; Malumbres et al., 2004; Rane et al., 1999; Sicinska et al., 2003, 2006; Tsutsui et al., 1999). To pharmacologically target this kinase complex, we have used PD-0332991, a small molecule selectively inhibiting CDK4/6 function (Fry et al., 2004). PD-0332991 is a pyridopyrimidine that exhibits an IC50 value less than 0.01 mmol/L against cyclin D3:CDK4/6 complexes. PD0332991 is currently in clinical trials for the treatment of multiple myeloma (Baughn et al., 2006; Menu et al., 2008), advanced adult solid tumors, and refractory non-Hodgkins lymphoma (Schwartz et al., 2011). The drug has a relatively long half-life and is generally well tolerated with minimal side effects. Our results and studies now demonstrate that PD-0332991 could be an efcient treatment for pediatric and adult T cell leukemia. Indeed, we were able to show rapid induction of cell cycle arrest in both mouse and human T-ALL cell lines. Furthermore, we show that the treatment results in cell cycle arrest of primary human leukemia cells both at diagnosis and relapse, as well as loss of leukemic cells by apoptosis, and inhibition of disease progression. Previous data have shown that PD-0332991 effectively inhibits the cell cycle in a breast cancer cell line, and this is dependent on Rb, as chronic loss of Rb eventually leads to resistance to PD-0332991 (Dean et al., 2010). Further studies

are required to determine whether Rb is required for PD0332991-induced cell cycle arrest in T-ALL. PD-0332991 efciency against relapsed disease is particularly exciting, as such leukemia cells are hyperproliferating, particularly aggressive, and usually nonresponsive to current therapeutic protocols. Although we have observed that PD-0332991 administration can efciently suppress disease progression, it most likely does not target putative leukemia-initiating cells, as the disease slowly relapses upon discontinuation of treatment. To further potentiate its activity and achieve sustained remission, PD0332991 could be used in combinatorial treatment protocols together with either conventional chemotherapy or next generation T-ALL-targeted therapeutic compounds, including g-secretase inhibitors (Real et al., 2009) or Velcade (Menu et al., 2008; Vilimas et al., 2007).
EXPERIMENTAL PROCEDURES Mice For generation of Ccnd3D2/D2 mice, see Supplemental Experimental Procedures. Ccnd2/, Ccnd3/, and Rb1F/F mice were kindly provided by Piotr Sicinski and Kay Macleod, respectively, and genotyped following published procedures (Ciemerych et al., 2002). Rb1F/F mice were subsequently crossed to Mx1-Cre+ mice (Jackson Laboratory). Deletion of oxed alleles was induced by intraperitoneal injection of 20 mg/kg polyI:C. Mice were injected ve times, once every other day, and analyzed 2 weeks after the last injection. C57BL/6 and Cdkn1b/ mice were purchased from the Jackson Laboratory. All animals were used between 410 weeks of age and housed in the sterile Smilow Animal Facility at NYU Medical Center (New York, NY). All experiments were performed according to the guidelines of the Institutional Animal Care and Use Committee of NYU Medical Center. Notch1 Transduction and T-ALL Induction Retroviral supernatant was generated by transfection of HEK293T cells with pMIGR1-Notch1-IC-IRES-EGFP retroviral construct by the calcium phosphate method. c-Kit+ bone marrow progenitors from wild-type control, Ccnd3Ccnd2/Ccnd2, Ccnd3/, or Ccnd3/Cdkn1b/ animals were puried by magnetic selection and cultured in complete OPTI-MEM supplemented with stem cell factor and Flt3 at 50 ng/ml and interleukin (IL)-6 and IL-7 at 10 ng/ml. Cells were incubated with retroviral supernatant plus 8 mg/ml polybrene and subjected to three rounds of spinoculation prior to transplant. Prior to transplantation, green uorescent protein (GFP) expression was analyzed by ow cytometry, and 3 3 105 ICN1-EGFP+ cells were injected retro-orbitally into lethally irradiated C57B6.SJL (CD45.1+) hosts. Mice were monitored for symptoms of disease by peripheral blood analysis for the presence of CD4+CD8+ DP cells. RNA and Protein Analyses RNA and protein expression analyses were performed using standard molecular biology procedures. Please refer to the Supplemental Experimental Procedures for further details. Flow Cytometry Flow cytometry was performed essentially, as previously described (Cooper et al., 2006). Additional details are found in Supplemental Experimental Procedures. PD-0332991 Treatment and Microarray Analysis The CDK4/6-specic inhibitor PD-0332991 was generously provided by Pzer Global Research and Development. Human T-ALL lines (CEM, DND41, HPB-ALL, Jurkat, and TAL1) and mouse (#130, #720, and #5146) were treated with 0.5, 1, or 5 mM of PD-0332991, with concentrations of 0.5 and 1 mM yielding similar results in vitro. Total RNA was extracted from PD-0332991-treated cells using the RNEasy plus mini kit (QIAGEN). RNA was amplied with the Ovation RNA Amplication System V2 (Nugen) for

Cancer Cell 22, 452465, October 16, 2012 2012 Elsevier Inc. 463

Cancer Cell
Cyclin D3:CDK4/6 Role in the Progression of T-ALL

complementary RNA (cRNA) amplication and labeling. Labeled cRNA was then hybridized to Human Genome U133 Plus 2.0 GeneChips (Affymetrix) for microarray analysis. Affymetrix gene expression proling data were normalized with the robust multiarray average algorithm using GeneSpring GX software (Agilent). The gene expression intensity presentation was generated with Multi Experiment Viewer software (http://www.tm4/org/mev/). Geneset enrichment analysis was performed using GSEA software (http://www. broadinstitute.org/gsea/) using phenotype as permutation type, 1,000 permutations, and signal-to-noise ratio as metric for ranking genes (Mootha et al., 2003; Subramanian et al., 2005). Gene sets used in the analysis were taken from the MSig database of the Broad Institute (http://www.broadinstitute. org/gsea/msigdb/cards/). For in vivo studies of PD-0332991 treatment, 3 weeks after transplant of Notch1-expressing cells, transplanted mice received either 150 mg/kg PD-0332991 in 50 mM sodium lactate or vehicle control by gavage daily for ten consecutive days. Four weeks after transplant, peripheral blood was analyzed for GFP, CD4, and CD8 expression. Animals were sacriced 7 to 8 weeks posttransplant as control mice became moribund. Upon autopsy, tissues were harvested and prepared for uorescence-activated cell sorting (FACS) and histological analysis. Tissues were xed for 24 hr in 10% buffered formalin, dehydrated, and then embedded in parafn. Parafn blocks were sectioned at 5 mm and stained with hematoxylin and eosin. For xenograft studies, sublethally irradiated Rag2/Il2rg/ mice were transplanted with 5 3 105 human T-ALL CEM cells. Three weeks posttransplant, mice were given either vehicle control or 150 mg/kg PD-0332991 by gavage for 14 days. Peripheral blood was analyzed for huCD45+ cells 4 weeks posttransplant, and survival of mice was monitored. Primary Human T-ALL Proliferation Assays The study was approved by the Institut Universitaire dHematologie Institutional Review Board (Saint-Louis Hospital, Paris, France), and informed consent was obtained from the patients. Human and xenografted T-ALL cell samples were thawed in T-ALL medium and seeded onto MS5-DLH1 stromal cells, as previously described (Clappier et al., 2011). After 48 hr, blasts were reseeded onto new MS5-DLH1 (6 mm multiwell at 106 cells/ml) and treated overnight with 0.5 mM of PD-0332991 inhibitor or solvent (DMSO), followed by 10 mM bromodeoxyuridine (BrdU) pulse for 30 min. Bulk cultures were harvested, and stromal cells were removed from the suspension by two rounds of ltration through a 70 mm strainer and panning for 30 min. The cell suspension was then incubated with an anti-CD45 antibody, xed and labeled with BrdU antibody. Statistical Analysis All data for cell numbers, expression analysis by qPCR, and FACS analysis of cell cycle and apoptosis represent mean SD. Statistical signicance was calculated using Students t test, with p < 0.05 considered signicant. Significance of survival differences was determined by log rank test. ACCESSION NUMBERS The gene expression data from treated T-ALL cell lines can be found at the GEO database (http://www.ncbi.nlm.nih.gov/geo/) using the accession number GSE40635. SUPPLEMENTAL INFORMATION Supplemental Information includes four gures and Supplemental Experimental Procedures and can be found with this article online at http://dx.doi. org/10.1016/j.ccr.2012.09.016. ACKNOWLEDGMENTS We would like to thank members of the Aifantis Lab and C.W. Brains for discussions and comments, Dr. J. Zavadil and the NYU Genome Technology Center (supported in part by NIH/NCI P30 CA016087-30 Grant) for expert assistance with microarray experiments, the NYU Flow Cytometry facility (supported in part by NIH/NCI 5 P30CA16087-31) for expert cell sorting, the NYU

Histology Core (5P30CA16087-31), and the Transgenic Mouse Core (NYU Cancer Institute Center Grant 5P30CA16087-31). I.A. was supported by the National Institutes of Health (RO1CA133379, RO1CA105129, R21CA141399, RO1CA149655, and RO1GM088847). I.A. was also supported by the William Lawrence and Blanche Hughes Foundation, the Leukemia & Lymphoma Society, and The V Foundation for Cancer Research. We are grateful for the support by a Feinberg Lymphoma Grant. L.G. is supported by a grant from the Institute National du Cancer (INCa). T.T. is supported by the NYU Molecular and Cellular Biology Training Program. A.S. is supported by NYSTEM institutional NYU Stem Cell Training Grant (C026880). I.A. is a Howard Hughes Medical Institute Early Career Scientist. Received: January 6, 2012 Revised: June 18, 2012 Accepted: September 13, 2012 Published: October 15, 2012 REFERENCES Aifantis, I., Raetz, E., and Buonamici, S. (2008). Molecular pathogenesis of T-cell leukaemia and lymphoma. Nat. Rev. Immunol. 8, 380390. Baker, G.L., Landis, M.W., and Hinds, P.W. (2005). Multiple functions of D-type cyclins can antagonize pRb-mediated suppression of proliferation. Cell Cycle 4, 330338. Baughn, L.B., Di Liberto, M., Wu, K., Toogood, P.L., Louie, T., Gottschalk, R., Niesvizky, R., Cho, H., Ely, S., Moore, M.A., and Chen-Kiang, S. (2006). A novel orally active small molecule potently induces G1 arrest in primary myeloma cells and prevents tumor growth by specic inhibition of cyclin-dependent kinase 4/6. Cancer Res. 66, 76617667. Bergsagel, P.L., Kuehl, W.M., Zhan, F., Sawyer, J., Barlogie, B., and Shaughnessy, J., Jr. (2005). Cyclin D dysregulation: an early and unifying pathogenic event in multiple myeloma. Blood 106, 296303. Carthon, B.C., Neumann, C.A., Das, M., Pawlyk, B., Li, T., Geng, Y., and Sicinski, P. (2005). Genetic replacement of cyclin D1 function in mouse development by cyclin D2. Mol. Cell. Biol. 25, 10811088. Chien, W.M., Rabin, S., Macias, E., Miliani de Marval, P.L., Garrison, K., Orthel, J., Rodriguez-Puebla, M., and Fero, M.L. (2006). Genetic mosaics reveal both cell-autonomous and cell-nonautonomous function of murine p27Kip1. Proc. Natl. Acad. Sci. USA 103, 41224127. Ciemerych, M.A., Kenney, A.M., Sicinska, E., Kalaszczynska, I., Bronson, R.T., Rowitch, D.H., Gardner, H., and Sicinski, P. (2002). Development of mice expressing a single D-type cyclin. Genes Dev. 16, 32773289. Clappier, E., Gerby, B., Sigaux, F., Delord, M., Touzri, F., Hernandez, L., Ballerini, P., Baruchel, A., Pumio, F., and Soulier, J. (2011). Clonal selection in xenografted human T cell acute lymphoblastic leukemia recapitulates gain of malignancy at relapse. J. Exp. Med. 208, 653661. Cooper, A.B., Sawai, C.M., Sicinska, E., Powers, S.E., Sicinski, P., Clark, M.R., and Aifantis, I. (2006). A unique function for cyclin D3 in early B cell development. Nat. Immunol. 7, 489497. Coustan-Smith, E., Mullighan, C.G., Onciu, M., Behm, F.G., Raimondi, S.C., Pei, D., Cheng, C., Su, X., Rubnitz, J.E., Basso, G., et al. (2009). Early T-cell precursor leukaemia: a subtype of very high-risk acute lymphoblastic leukaemia. Lancet Oncol. 10, 147156. Dean, J.L., Thangavel, C., McClendon, A.K., Reed, C.A., and Knudsen, E.S. (2010). Therapeutic CDK4/6 inhibition in breast cancer: key mechanisms of response and failure. Oncogene 29, 40184032. Espinosa, L., Cathelin, S., DAltri, T., Trimarchi, T., Statnikov, A., Guiu, J., s-Esteve, J., Nomdedeu, J., Bellosillo, B., et al. (2010). The Rodilla, V., Ingle Notch/Hes1 pathway sustains NF-kB activation through CYLD repression in T cell leukemia. Cancer Cell 18, 268281. Fero, M.L., Rivkin, M., Tasch, M., Porter, P., Carow, C.E., Firpo, E., Polyak, K., Tsai, L.H., Broudy, V., Perlmutter, R.M., et al. (1996). A syndrome of multiorgan hyperplasia with features of gigantism, tumorigenesis, and female sterility in p27(Kip1)-decient mice. Cell 85, 733744.

464 Cancer Cell 22, 452465, October 16, 2012 2012 Elsevier Inc.

Cancer Cell
Cyclin D3:CDK4/6 Role in the Progression of T-ALL

Fry, D.W., Harvey, P.J., Keller, P.R., Elliott, W.L., Meade, M., Trachet, E., Albassam, M., Zheng, X., Leopold, W.R., Pryer, N.K., and Toogood, P.L. (2004). Specic inhibition of cyclin-dependent kinase 4/6 by PD 0332991 and associated antitumor activity in human tumor xenografts. Mol. Cancer Ther. 3, 14271438. Geng, Y., Whoriskey, W., Park, M.Y., Bronson, R.T., Medema, R.H., Li, T., Weinberg, R.A., and Sicinski, P. (1999). Rescue of cyclin D1 deciency by knockin cyclin E. Cell 97, 767777. Geng, Y., Yu, Q., Sicinska, E., Das, M., Bronson, R.T., and Sicinski, P. (2001). Deletion of the p27Kip1 gene restores normal development in cyclin D1decient mice. Proc. Natl. Acad. Sci. USA 98, 194199. Hunter, T., and Pines, J. (1994). Cyclins and cancer. II: Cyclin D and CDK inhibitors come of age. Cell 79, 573582. Joshi, I., Minter, L.M., Telfer, J., Demarest, R.M., Capobianco, A.J., Aster, J.C., Sicinski, P., Fauq, A., Golde, T.E., and Osborne, B.A. (2009). Notch signaling mediates G1/S cell cycle progression in T cells via cyclin D3 and its dependent kinases. Blood 113, 16891698. Kozar, K., Ciemerych, M.A., Rebel, V.I., Shigematsu, H., Zagozdzon, A., Sicinska, E., Geng, Y., Yu, Q., Bhattacharya, S., Bronson, R.T., et al. (2004). Mouse development and cell proliferation in the absence of D-cyclins. Cell 118, 477491. Li, X., Gounari, F., Protopopov, A., Khazaie, K., and von Boehmer, H. (2008). Oncogenesis of T-ALL and nonmalignant consequences of overexpressing intracellular NOTCH1. J. Exp. Med. 205, 28512861. n, J., Cerezo, A., Ortega, S., a, D., Gala Malumbres, M., Sotillo, R., Santamar Dubus, P., and Barbacid, M. (2004). Mammalian cells cycle without the D-type cyclin-dependent kinases Cdk4 and Cdk6. Cell 118, 493504. Marzec, M., Kasprzycka, M., Lai, R., Gladden, A.B., Wlodarski, P., Tomczak, E., Nowell, P., Deprimo, S.E., Sadis, S., Eck, S., et al. (2006). Mantle cell lymphoma cells express predominantly cyclin D1a isoform and are highly sensitive to selective inhibition of CDK4 kinase activity. Blood 108, 17441750. Menu, E., Garcia, J., Huang, X., Di Liberto, M., Toogood, P.L., Chen, I., Vanderkerken, K., and Chen-Kiang, S. (2008). A novel therapeutic combination using PD 0332991 and bortezomib: study in the 5T33MM myeloma model. Cancer Res. 68, 55195523. Mootha, V.K., Lindgren, C.M., Eriksson, K.F., Subramanian, A., Sihag, S., le, M., Laurila, E., et al. Lehar, J., Puigserver, P., Carlsson, E., Ridderstra (2003). PGC-1alpha-responsive genes involved in oxidative phosphorylation are coordinately downregulated in human diabetes. Nat. Genet. 34, 267273. Motokura, T., and Arnold, A. (1993). Cyclin D and oncogenesis. Curr. Opin. Genet. Dev. 3, 510. Nakayama, K., Ishida, N., Shirane, M., Inomata, A., Inoue, T., Shishido, N., Horii, I., Loh, D.Y., and Nakayama, K. (1996). Mice lacking p27(Kip1) display increased body size, multiple organ hyperplasia, retinal dysplasia, and pituitary tumors. Cell 85, 707720. ONeil, J., Calvo, J., McKenna, K., Krishnamoorthy, V., Aster, J.C., Bassing, C.H., Alt, F.W., Kelliher, M., and Look, A.T. (2006). Activating Notch1 mutations in mouse models of T-ALL. Blood 107, 781785. Palomero, T., Sulis, M.L., Cortina, M., Real, P.J., Barnes, K., Ciofani, M., Caparros, E., Buteau, J., Brown, K., Perkins, S.L., et al. (2007). Mutational loss of PTEN induces resistance to NOTCH1 inhibition in T-cell leukemia. Nat. Med. 13, 12031210. Peled, J.U., Yu, J.J., Venkatesh, J., Bi, E., Ding, B.B., Krupski-Downs, M., Shaknovich, R., Sicinski, P., Diamond, B., Scharff, M.D., and Ye, B.H. (2010). Requirement for cyclin D3 in germinal center formation and function. Cell Res. 20, 631646. Rane, S.G., Dubus, P., Mettus, R.V., Galbreath, E.J., Boden, G., Reddy, E.P., and Barbacid, M. (1999). Loss of Cdk4 expression causes insulin-decient diabetes and Cdk4 activation results in beta-islet cell hyperplasia. Nat. Genet. 22, 4452. Real, P.J., Tosello, V., Palomero, T., Castillo, M., Hernando, E., de Stanchina, E., Sulis, M.L., Barnes, K., Sawai, C., Homminga, I., et al. (2009). Gamma-

secretase inhibitors reverse glucocorticoid resistance in T cell acute lymphoblastic leukemia. Nat. Med. 15, 5058. Sarcevic, B., Lilischkis, R., and Sutherland, R.L. (1997). Differential phosphorylation of T-47D human breast cancer cell substrates by D1-, D3-, E-, and A-type cyclin-CDK complexes. J. Biol. Chem. 272, 3332733337. Schwartz, G.K., LoRusso, P.M., Dickson, M.A., Randolph, S.S., Shaik, M.N., Wilner, K.D., Courtney, R., and ODwyer, P.J. (2011). Phase I study of PD 0332991, a cyclin-dependent kinase inhibitor, administered in 3-week cycles (Schedule 2/1). Br. J. Cancer 104, 18621868. Sherr, C.J. (1995). D-type cyclins. Trends Biochem. Sci. 20, 187190. Sherr, C.J., and Roberts, J.M. (2004). Living with or without cyclins and cyclindependent kinases. Genes Dev. 18, 26992711. Sicinska, E., Aifantis, I., Le Cam, L., Swat, W., Borowski, C., Yu, Q., Ferrando, A.A., Levin, S.D., Geng, Y., von Boehmer, H., and Sicinski, P. (2003). Requirement for cyclin D3 in lymphocyte development and T cell leukemias. Cancer Cell 4, 451461. Sicinska, E., Lee, Y.M., Gits, J., Shigematsu, H., Yu, Q., Rebel, V.I., Geng, Y., Marshall, C.J., Akashi, K., Dorfman, D.M., et al. (2006). Essential role for cyclin D3 in granulocyte colony-stimulating factor-driven expansion of neutrophil granulocytes. Mol. Cell. Biol. 26, 80528060. Sicinski, P., Donaher, J.L., Parker, S.B., Li, T., Fazeli, A., Gardner, H., Haslam, S.Z., Bronson, R.T., Elledge, S.J., and Weinberg, R.A. (1995). Cyclin D1 provides a link between development and oncogenesis in the retina and breast. Cell 82, 621630. Sicinski, P., Donaher, J.L., Geng, Y., Parker, S.B., Gardner, H., Park, M.Y., Robker, R.L., Richards, J.S., McGinnis, L.K., Biggers, J.D., et al. (1996). Cyclin D2 is an FSH-responsive gene involved in gonadal cell proliferation and oncogenesis. Nature 384, 470474. Solvason, N., Wu, W.W., Parry, D., Mahony, D., Lam, E.W., Glassford, J., Klaus, G.G., Sicinski, P., Weinberg, R., Liu, Y.J., et al. (2000). Cyclin D2 is essential for BCR-mediated proliferation and CD5 B cell development. Int. Immunol. 12, 631638. Subramanian, A., Tamayo, P., Mootha, V.K., Mukherjee, S., Ebert, B.L., Gillette, M.A., Paulovich, A., Pomeroy, S.L., Golub, T.R., Lander, E.S., and Mesirov, J.P. (2005). Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression proles. Proc. Natl. Acad. Sci. USA 102, 1554515550. Tamamori-Adachi, M., Goto, I., Yamada, K., and Kitajima, S. (2008). Differential regulation of cyclin D1 and D2 in protecting against cardiomyocyte proliferation. Cell Cycle 7, 37683774. Tsutsui, T., Hesabi, B., Moons, D.S., Pandol, P.P., Hansel, K.S., Koff, A., and Kiyokawa, H. (1999). Targeted disruption of CDK4 delays cell cycle entry with enhanced p27(Kip1) activity. Mol. Cell. Biol. 19, 70117019. Vilimas, T., Mascarenhas, J., Palomero, T., Mandal, M., Buonamici, S., Meng, F., Thompson, B., Spaulding, C., Macaroun, S., Alegre, M.L., et al. (2007). Targeting the NF-kappaB signaling pathway in Notch1-induced T-cell leukemia. Nat. Med. 13, 7077. Walkley, C.R., Fero, M.L., Chien, W.M., Purton, L.E., and McArthur, G.A. (2005). Negative cell-cycle regulators cooperatively control self-renewal and differentiation of haematopoietic stem cells. Nat. Cell Biol. 7, 172178. Walkley, C.R., Shea, J.M., Sims, N.A., Purton, L.E., and Orkin, S.H. (2007). Rb regulates interactions between hematopoietic stem cells and their bone marrow microenvironment. Cell 129, 10811095. 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). Activating mutations of NOTCH1 in human T cell acute lymphoblastic leukemia. Science 306, 269271. Zou, P., Yoshihara, H., Hosokawa, K., Tai, I., Shinmyozu, K., Tsukahara, F., Maru, Y., Nakayama, K., Nakayama, K.I., and Suda, T. (2011). p57(Kip2) and p27(Kip1) cooperate to maintain hematopoietic stem cell quiescence through interactions with Hsc70. Cell Stem Cell 9, 247261.

Cancer Cell 22, 452465, October 16, 2012 2012 Elsevier Inc. 465

Article
STAT3-Driven Upregulation of TLR2 Promotes Gastric Tumorigenesis Independent of Tumor Inammation

Cancer Cell

Hazel Tye,1 Catherine L. Kennedy,1 Meri Najdovska,1 Louise McLeod,1 William McCormack,3 Norman Hughes,4 Anouk Dev,2 William Sievert,2 Chia Huey Ooi,5 Tomo-o Ishikawa,6 Hiroko Oshima,6 Prithi S. Bhathal,4 Andrew E. Parker,3 Masanobu Oshima,6 Patrick Tan,5,7,8 and Brendan J. Jenkins1,*
for Innate Immunity and Infectious Diseases, Monash Institute of Medical Research of Medicine, Monash Medical Centre Monash University, Clayton, Victoria 3168, Australia 3OPSONA Therapeutics, Ltd., Institute of Molecular Medicine, Trinity Centre for Health Sciences, St. James Hospital, Dublin 8, Ireland 4Department of Pathology, University of Melbourne, Footscray, Victoria 3011, Australia 5Cancer and Stem Cell Biology, Duke-NUS Graduate Medical School, 169857, Singapore 6Division of Genetics, Cancer Research Institute, Kanazawa University, Kanazawa 920-1192, Japan 7Genome Institute of Singapore, 138672, Singapore 8Cancer Sciences Institute of Singapore, National University of Singapore, 117456, Singapore *Correspondence: brendan.jenkins@monash.edu http://dx.doi.org/10.1016/j.ccr.2012.08.010
2Department 1Centre

SUMMARY

Gastric cancer (GC) is associated with chronic inammation; however, the molecular mechanisms promoting tumorigenesis remain ill dened. Using a GC mouse model driven by hyperactivation of the signal transducer and activator of transcription (STAT)3 oncogene, we show that STAT3 directly upregulates the epithelial expression of the inammatory mediator Toll-like receptor (TLR)2 in gastric tumors. Genetic and therapeutic targeting of TLR2 inhibited gastric tumorigenesis, but not inammation, characterized by reduced proliferation and increased apoptosis of the gastric epithelium. Increased STAT3 pathway activation and TLR2 expression were also associated with poor GC patient survival. Collectively, our data reveal an unexpected role for TLR2 in the oncogenic function of STAT3 that may represent a therapeutic target in GC.
INTRODUCTION Gastric cancer (GC) is the second most lethal cancer worldwide (Parkin et al., 2005) and represents a growing number of cancers, such as colon, lung, liver, and prostate, that are associated with inammation (Gao et al., 2005; Kawada et al., 2006; Ogata et al., 2006; Walser et al., 2008). Although the molecular mechanisms underlying the pathogenesis of these cancers remain unclear, a causal correlation has been established between inammation triggered by microbes and gastrointestinal cancers, as evidenced by chronic gastritis caused by infection with the Gram-negative pathogen Helicobacter pylori being a major risk factor for human GC (Uemura et al., 2001). The involvement of pathogenic microbes in gastrointestinal inammation and carcinogenesis has implicated Toll-like receptors (TLRs), a key family of microbial sensors of the host innate and adaptive immune systems (Kawai and Akira, 2010), in mediating chronic inammatory responses that promote tumorigenesis. For instance, genetic ablation of the common TLR and interleukin (IL)-1 receptor family signaling adaptor MyD88 in mice alleviates intestinal tumorigenesis induced in ApcMin/+ mice (Rakoff-Nahoum and Medzhitov, 2007). With respect to human gastric disease, TLR2 and TLR4 gene expression is elevated in H. pylori-positive gastritis patients (Uno et al., 2007), and TLR2 and TLR4 gene polymorphisms are associated with an increased GC risk (Hold et al., 2007; Tahara et al., 2007). However, the

Signicance Hyperactivation of the STAT3 oncogene is a hallmark of many epithelial tumors, including gastric, however, the full spectrum of actions by which STAT3 elicits its oncogenicity remains ill dened. We have discovered that in gastric epithelial cells, STAT3 directly upregulates the expression of the innate immune pathogen recognition receptor TLR2. While gastric tumorigenesis is associated with inammation, we unexpectedly revealed that TLR2 promoted gastric epithelial cell survival and proliferation rather than tumor inammation. Given the emergence of TLR signaling cascades as key oncogenic drivers in various tumors also characterized by STAT3 hyperactivation (e.g., liver, colon), our studies suggest that selective blockade of TLR2 may provide a therapeutic approach to suppress tumorigenesis.
466 Cancer Cell 22, 466478, October 16, 2012 2012 Elsevier Inc.

Cancer Cell
TLR2 Promotes STAT3-Driven Gastric Tumorigenesis

expression status of these TLRs in human GC is unclear. In addition to the link between TLRs and the inammatory response, more recent studies have revealed that epithelial-expressed TLRs can promote noninammatory-related epithelial responses including cell survival, proliferation and migration (Shaykhiev et al., 2008), and angiogenesis (West et al., 2010). While these ndings constitute compelling circumstantial evidence for the involvement of TLRs in GC and other inammatory-related cancers, a causal role for specic TLRs in these cancers remains to be dened. We have previously established the gp130F/F mouse, which by 6 weeks of age spontaneously develops gastric inammation and intestinal-type gastric tumors (Tebbutt et al., 2002). These mice are homozygous for a knockin phenylalanine substitution at tyrosine 757 in the cytoplasmic domain of the IL-6 cytokine family receptor signaling subunit gp130, which disrupts the negative feedback mechanism on gp130 signaling mediated by suppressor of cytokine signaling (SOCS)3 (Tebbutt et al., 2002). The gastric phenotype of gp130F/F mice is driven by hyperactivation of the oncogenic latent transcription factor STAT3 via the IL-6 cytokine family member, IL-11 (Ernst et al., 2008; Jenkins et al., 2005), the clinical relevance of which is underscored by upregulated IL-11 production and STAT3 hyperactivation being common traits of human GC (Ellmark et al., 2006; Gong et al., 2005; Yakata et al., 2007). Although the full spectrum of molecular mechanisms underlying the potent oncogenic properties of STAT3 in cancers remains undetermined, the oncogenicity of STAT3 has been attributed to, at least in part, its direct transcriptional upregulation of genes, which promote angiogenesis, cell cycle progression, and cell survival (Bromberg et al., 1999). Moreover, the complex role of STAT3 in cancer is evidenced by the fact STAT3 can promote both a protumorigenic chronic inammatory microenvironment (Bollrath and Greten, 2009) and conversely suppress antitumor innate and adaptive immune responses (Kortylewski et al., 2005). Regarding the former, emerging evidence invokes functional overlap between STAT3 and nuclear factor-kB (NF-kB) oncogenic and inammatory signaling pathways in driving inammation-associated cancers (Bollrath and Greten, 2009; Yang et al., 2007). Here, we investigate whether crosstalk between STAT3 and TLR signaling contributes to the molecular pathogenesis of GC. RESULTS Augmented TLR2 Expression in Gastric Tumors Correlates with STAT3 Hyperactivation We initially identied changes in the expression prole of genes associated with the TLR network during gastric tumorigenesis in gp130F/F mice by using RT2 Proler PCR Arrays specic for 84 key TLR-associated genes. The majority of genes upregulated in gastric antrum tumor tissue of 24-week-old gp130F/F mice compared to tumor-free antrum tissue from gp130+/+ mice were NF-kB-regulated, including Tnfa and Il1b proinammatory cytokine genes, as well as Tlr2, previously implicated in human GC (El-Omar et al., 2003; Tahara et al., 2007) (Table S1 available online). Independent quantitative real-time PCR (qPCR) assays conrmed that among TLRs and other pattern recognition receptors (PRRs) implicated in gastric disease,

including Nod1 and Nod2 (Rosenstiel et al., 2006), only Tlr2 messenger RNA levels were signicantly upregulated in gp130F/F gastric tumor tissue compared to gastric tumor and tumor-free tissue from gp130F/F:Stat3-/+ mice and gastric tumorfree tissue from gp130F/F:Il11r/ mice displaying reduced STAT3 activation (Ernst et al., 2008) and expression of STAT3target genes, Stat3 and Il11 (Figures 1A and S1A). In support of the notion that TLR-induced responses are associated with gastric tumorigenesis, the gastric expression of TLR-regulated proinammatory genes Il1b, Cxcl2, and Tnfa, as well as the Cd14 gene encoding the coreceptor for TLR2 and TLR4, was signicantly increased in gp130F/F mice compared to compound mutant mice (Figure S1A). The coexistent hyperactivation of STAT3, augmented IL-11 expression, and specic upregulation of TLR2 in gastric tumorigenesis was also independently observed in the transgenic K19-Wnt/C2mE (Gan) mouse model of GC that simultaneously overexpresses the cyclooxygenase2/prostaglandin E2 and Wnt signaling pathways in the gastric mucosa (Itadani et al., 2009; Oshima et al., 2006) (Figures S1BS1E). The augmented gastric expression prole of specic STAT3target genes and inammatory mediators in gp130F/F mice translated to human disease, since TLR2, IL1B, CD14, IL8 (human homolog of Cxcl2), STAT3, and IL11 were signicantly upregulated in tumor tissue from GC patients compared to adjacent nontumor tissue and antrum tissue from healthy individuals (Figure 1B). Furthermore, analysis of 78 primary human gastric tumors (Ooi et al., 2009) revealed an overall positive correlation between increased activation of the STAT3 pathway and elevated TLR2 expression (Figure 1C). Among these 78 gastric tumors, we also compared 25 tumors in which STAT3 activation-TLR2 gene expression was high against 21 tumors in which STAT3 activation-TLR2 expression was low. These 46 GC patients segregated into two distinct groups whereby those displaying a combined high STAT3-TLR2 status had a signicantly impaired overall survival compared to patients with a combined low STAT3-TLR2 status, thus suggesting that STAT3 hyperactivation and TLR2 upregulation are poor prognostic factors in GC (Figure 1D). In further support of a STAT3TLR2 axis in human GC, analysis of 18 human GC cell lines previously characterized by gene expression signatures representing the activity status of various oncogenic pathways, including STAT3 (Ooi et al., 2009), revealed a signicant positive correlation between the levels of STAT3 pathway activation and TLR2 gene expression (Figure 1E). STAT3-Mediated Upregulation of TLR2 in Gastric Epithelial Cells Since the gastric epithelium plays a key role in modulating the immune response to microbes, we next evaluated whether TLR2 was transcriptionally regulated by STAT3 in gastric epithelial cells. Overexpression of the hyperactive STAT3-C mutant (Bromberg et al., 1999) in mouse gastric epithelial IMGE-5 cells induced gene expression for Socs3, a known STAT3-target gene (Table S2), and Tlr2, but not other TLR family members (Figures 2A and 2B), and this result was also conrmed in human GC epithelial cell lines MKN-28 and AGS (Figures 2C and 2D). The transcriptional induction of TLR2 by STAT3-C was further conrmed by the STAT3-C-driven transactivation
Cancer Cell 22, 466478, October 16, 2012 2012 Elsevier Inc. 467

Cancer Cell
TLR2 Promotes STAT3-Driven Gastric Tumorigenesis

Figure 1. Augmented STAT3 Activation Correlates with Increased Expression of TLR2 in Gastric Tumors of gp130F/F Mice and Human GC
(A) qPCR expression analyses on antral gastric tumor (T) or nontumor (NT) tissue of 24-week-old mice of the indicated genotypes. Expression data from at least ve samples per genotype are shown following normalization for 18S expression. *p < 0.05, **p < 0.01, and ***p < 0.001. (B) qPCR of the indicated genes in human gastric antral biopsies corresponding to normal (N) tissue from healthy individuals (n = 6), as well as tumor (T) and matched adjacent nontumor (NT) tissue from GC patients (n = 12). Expression data are shown following normalization for 18S. *p < 0.05, **p < 0.01, and ***p < 0.001. (C) Top panel, STAT3 activation levels in 78 human gastric tumors and tumors with positive STAT3 activation scores (STAT3 high), and those with negative STAT3 activation scores (STAT3 low) are indicated by different color schemes. Bottom panel, TLR2 gene expression levels in 78 tumors in the same order as the top panel. The y axis values are presented on a log scale, with positive and negative values representing high and low TLR2 gene expression, respectively. Spearman rank correlation coefcient (r) = 0.3395. p = 0.0025. (D) Kaplan-Meier survival analysis of a subset of 46 human GC patients with nonzero STAT3 activation scores stratied into two groups based on either low (n = 21) or high (n = 25) combined STAT3-TLR2 status. (E) Correlation between STAT3 activation and TLR2 expression in human GC cell lines. STAT3 activation levels in 18 human GC cell lines (-) were determined based on the average STAT3 pathway activation scores (x axis). The y axis represents the log2-transformed TLR2 gene expression value. r = 0.6633, p = 3.86 3 105. In (A and B), data are presented from replicate analysis as the mean SEM. See also Figure S1 and Table S1.

of the full-length TLR2 promoter-luciferase reporter TLR2-2391 (Figure 2E). Moreover, bioinformatic analyses identied ve putative STAT3-binding sites in the 50 promoter region of TLR2 (Figure 2F), and chromatin immunoprecipitation on either IL-11-stimulated gastric epithelial cells or gp130F/F tumor tissue revealed that tyrosine-phosphorylated (pY) STAT3 was recruited to the TLR2 50 region spanning the putative STAT3 site at bases 1241/1249 (Figures 2F2H), thus suggesting that TLR2 is a direct STAT3 target gene. In support of these data, STAT3-C did not transactivate either the truncated (TLR21248) or site-directed mutant (TLR2-TT > AA) TLR2-specic
468 Cancer Cell 22, 466478, October 16, 2012 2012 Elsevier Inc.

to gp130+/+ (Figure S2A).

promoter reporters that disrupted the 1241/1249 STAT3 site (Figure 2E). The functional relevance of these ndings was validated by the augmented production of IL-6 and IL-1b in primary gastric epithelial cells from gp130F/F compared mice stimulated with the TLR2 agonist Pam3Cys

TLR2 Promotes Gastric Tumorigenesis in gp130F/F Mice To genetically dene a causal role for TLR2 in STAT3-driven gastric tumorigenesis, we generated gp130F/F mice lacking TLR2 (gp130F/F:Tlr2/). The stomachs of 24-week-old gp130F/F: Tlr2/ mice were noticeably smaller in size compared to those of age-matched gp130F/F littermates (Figure 3A), analogous to the reduced stomach size observed previously in gp130F/F: Stat3-/+ mice due to suppression of STAT3-driven hyperplasia

Cancer Cell
TLR2 Promotes STAT3-Driven Gastric Tumorigenesis

Figure 2. STAT3 Directly Regulates TLR2 Gene Expression in Gastric Epithelial Cells
(A) qPCR of Socs3 and Tlr genes in gastric epithelial IMGE-5 cells stably-transfected with STAT3-C/green uorescent protein (GFP) (+) or GFP alone (). Data from three independent experiments are shown following normalization for 18S. *p < 0.05. Cell lysates were immunoblotted with STAT3 and actin antibodies with exogenous STAT3-C/GFP fusion protein (120kDa) detected in transfected cells only. (B) qPCR of endogenous Tlr2 and Tlr4 gene expression in IMGE-5 cells transiently-transfected with increasing amounts of STAT3-C/GFP. Data from three independent experiments are shown following normalization for 18S. *p < 0.05 versus cells transfected with GFP alone. (C and D) qPCR of endogenous gene expression for TLR family members and STAT3 in human MKN-28 (C) and AGS (D) cells transiently-transfected with 500ng STAT3-C/GFP or GFP alone. Data from four independent experiments are shown for each cell line following normalization for ACTIN. *p < 0.05. (E) Human epithelial 293 cells were cotransfected with luciferase reporter plasmids for the fulllength TLR2 promoter (TLR2-2391), the truncated TLR2 promoter (TLR2-1248), and the sitedirected mutant TLR2 promoter (TLR2-TT > AA) together with the STAT3-C/GFP construct or an empty GFP vector (GFP). Data from four independent experiments are expressed as the mean SEM-fold change relative to TK-renilla normalized reporter activity in GFP-transfected cells. **p < 0.01. (F) Schematic diagram of the TLR2 gene promoter. The vertical arrow indicates where the TLR2 promoter fragment was truncated in the TLR2-1248 construct and the underlined TT bases at the 1241/1248 STAT3 site (bold type) were mutated to AA in the TLR2-TT > AA construct. Arrowheads depict primers (used in ChIP assays) spanning ve putative STAT3 binding sites in the TLR2 promoter. (G) Cells were incubated without or with IL-11 for 1 hr and subjected to ChIP assays using pY-STAT3 or IgG isotype control antibodies. The graph represents quantication by qPCR of the immunoprecipitated TLR2 promoter fragment containing the 1241/1248 STAT3 site bound to each antibody. Data from four independent experiments are expressed as the mean SEM % of input DNA. Gels represent semiquantitative PCR of specic promoter fragments that were amplied from coprecipitated DNA from IL-11-treated cells using primers spanning the 1241 STAT3 site in the TLR2 promoter (F) and, as a control, the STAT3 binding site at position 70 in the SOCS3 promoter. (H) Gel representing semiquantitative PCR of ChIP assay for pY-STAT3 interaction with the TLR2 gene promoter as per (G), except on gastric tumor tissue from gp130F/F mice. In (A)(D), all data are presented from replicate analysis as the mean SEM-fold induction relative to cells transfected with GFP alone. See also Figure S2 and Table S2.

throughout the gastric epithelium (Jenkins et al., 2005). The reduction in stomach size and mass of gp130F/F:Tlr2/ mice was more pronounced in males (50% reduction in mass) than females (30%) when compared to their gp130F/F counterparts (Figures 3A and 3B). The gastric tumor mass of gp130F/F:Tlr2/ mice was also signicantly reduced (70% in male and 40% in F/F female) compared to gp130 mice of the corresponding sex (Figure 3C). The total incidence of gastric lesions in gp130F/F: Tlr2/ male (2.76 0.28) and gp130F/F:Tlr2/ female (4.22 0.76) mice was also lower than in gp130F/F males (8.30 0.49) and gp130F/F females (7.50 1.22), with a noticeable reduction in the numbers of large (>4 mm diameter) tumors (Figure 3D). We also note that STAT3 tyrosine phosphorylation and IL-11 expression levels in gp130F/F:Tlr2/ tumors remained elevated and were comparable to those of gp130F/F gastric tumors (Ernst et al., 2008) (Figures 3E and 3F), consistent with the IL-11/STAT3 signaling axis being an upstream regulator of TLR2. The pathological requirement for TLR2 in gastric tumorigenesis was specic, since the gross appearance of stomachs and extent

of gastric tumorigenesis in gp130F/F:Tlr4/ male and female mice were comparable to those of gp130F/F littermates (Figures S2B2F). TLR2 Promotes Gastric Epithelial Cell Survival and Proliferation, but Not Inammation We have previously reported that gastric tumor formation in gp130F/F mice is characterized by chronic inammatory cell (lymphoplasmacytoid) aggregates in the submucosa (Ernst et al., 2008). However, despite reduced gastric tumorigenesis in gp130F/F:Tlr2/ mice, histological examination of hematoxylin and eosin (H&E)-stained gastric sections from 24-weekold mice indicated that the extent of these submucosal inammatory aggregates was comparable to that of gp130F/F mice (Figure 3G). Further histological and immunohistochemical analyses revealed a similar composition of both innate and adaptive inammatory cell inltrates between the two genotypes (Figures S3AS3C). These observations were further veried by ow cytometry, which indicated comparable
Cancer Cell 22, 466478, October 16, 2012 2012 Elsevier Inc. 469

Cancer Cell
TLR2 Promotes STAT3-Driven Gastric Tumorigenesis

Figure 3. Alleviation of Gastric Tumor Formation, but Not Inammation or STAT3 Hyperactivation, in gp130F/F Mice Lacking TLR2
(A) Representative appearance of stomachs from 24-week-old male (M) and female (F) mice of the indicated genotypes. Arrows indicate macroscopically visible tumors >4 mm in diameter. Fundic (f) and antral (a) stomach regions are labeled. (BD) Scatter plots depicting the total mass (grams) of stomachs (B) and gastric tumors (C), as well as the incidence of tumors (D), from either male or female gp130F/F and gp130F/F:Tlr2/ 24week-old mice. Data for each genotype are expressed as the mean SEM. *p < 0.05, **p < 0.01, ***p < 0.001. (E) Western blots of gastric tumor lysates from 24week-old gp130F/F and gp130F/F:Tlr2/ mice with the indicated antibodies. (F) Top panels: Representative photomicrographs showing immunohistochemistry for tyrosinephosphorylated STAT3 in the surface epithelium of gastric tumors of 24-week-old gp130F/F and gp130F/F:Tlr2/ mice. Scale bars = 200 mm. Bottom panels: Representative photomicrographs showing pY-STAT3+ cells in the mucosal epithelium (m) and submucosa (sm) containing inammatory cells in gp130F/F and gp130F/F:Tlr2/ mice. Scale bars = 200 mm. (G) Representative photomicrographs showing H&E-stained cross-sections through the antral region of stomachs from 24-week-old mice of the indicated genotypes. Arrows point to patches of inammatory cell accumulates in the gastric submucosa of gp130F/F and gp130F/F:Tlr2/ mice under low power magnication. Scale bars = 200 mm. See also Figure S3.

frequencies of Gr-1+CD11b+ (myeloid-derived suppressor cells [MDSC]), CD11b+F4/80+ (monocytes/macrophages), B220+ (B), and CD3+ (T) cells in the stomachs and perigastric lymph nodes of gp130F/F and gp130F/F:Tlr2/ mice (Figures S3D and S3E). In addition, the activation status of B (B220+CD86+) and T (CD4+CD69+ and CD8+CD69+) cells, which are the predominant gastric tumor-inltrating immune cells in gp130F/F and gp130F/F:Tlr2/ mice, in the stomach and perigastric lymph node tissue of the two genotypes was comparable (Figure S3F), as were the percentages of TNFa- or IFNg-producing gastric or perigastric lymph node lymphocytes (Figure S3F). Despite these observations supporting the notion that TLR2 does not promote the activation of these inammatory cells, the expression of a subset of TLR2-regulated cytokines and chemokines
470 Cancer Cell 22, 466478, October 16, 2012 2012 Elsevier Inc.

was reduced in gp130F/F:Tlr2/ gastric tumors (Figure S3G), as well as in gastric epithelial cells from gp130F/F:Tlr2/ mice stimulated with microbial-containing gastric gp130F/F homogenates (Figure S3H), thus conrming the reduced TLR2 responsiveness of the gastric epithelium in gp130F/F:Tlr2/ mice. These data therefore suggest that gastric epithelial stimulation by TLR2 does not promote the inltration or activation of inammatory cells. We next investigated whether the reduced gastric tumorigenesis in gp130F/F:Tlr2/ mice correlated with any proliferative, apoptotic, and/or angiogenic changes within the gastric mucosa. By contrast to the expansion of proliferating gastric cells observed throughout the mucosal epithelium of gp130F/F mice, the proliferation zone in the gastric epithelium of gp130F/F:Tlr2/ mice was more dened and resembled that of gp130+/+ mice with no apparent proliferating-cell nuclear antigen (PCNA)-positive staining in the surface epithelium (Figures 4A and 4B). Furthermore, apoptotic TUNEL-positive cells were found predominantly within the surface gastric

Cancer Cell
TLR2 Promotes STAT3-Driven Gastric Tumorigenesis

epithelial region of gp130F/F:Tlr2/ mice, whereas little to no TUNEL-positive gastric cells were detected within the mucosal epithelium of gp130F/F mice (Figure 4C). The extent of cellular proliferation and apoptosis within the gastric submucosal inammatory cell inltrates, however, was similar between gp130F/F and gp130F/F:Tlr2/ mice (Figures S4A and S4B). With respect to angiogenesis, the expression of the CD31 angiogenic marker in the gastric mucosa and submucosa of gp130F/F and gp130F/F:Tlr2/ tumors was comparable (Figures S4C and S4D). Collectively, these results unexpectedly suggest that TLR2 promotes gastric epithelial cell growth rather than gastric tumor inammation or angiogenesis. The notion that TLR2 signaling can promote cell proliferation was also conrmed in two human GC epithelial cell lines, whereby TLR2 ligand stimulation of MKN-28 and NUGC4 cells resulted in a dose-dependent signicant increase in cell proliferation (Figure 4D). TLR2 ligand stimulation of these human GC cell lines also signicantly upregulated the expression of numerous cell cycle progression and cell survival/antiapoptosis genes (Figures 4E and 4F), among which CCND2, BCL2A1, IER3, and BIRC3 are highly expressed in human GC (Li et al., 2011; Park et al., 1997; Sasada et al., 2004; Takano et al., 2000). These observations were also supported in vivo in the gp130F/F tumor model, where the expression of these cell survival and cell cycle progression genes was signicantly reduced in gp130F/F:Tlr2/ tumors compared to gp130F/F tumors (Figure 4G). To identify signaling pathways that facilitated the TLR2 ligandinduced proliferation of human GC cells, cells were pretreated with a series of well-documented signaling pathway inhibitors at doses shown to suppress the TLR2-driven activation of NF-kB; phosphatidylinositol 3-kinase (PI3K)/Akt; stress-activated protein kinase/c-Jun NH2-terminal kinase (SAPK/JNK); p38; and extracellular signal-related kinase (ERK)1/2 mitogenactivated protein kinases (MAPKs) (Figure 5A). Suppression of the PI3K/Akt, ERK1/2, and JNK MAPK, or NF-kB pathways resulted in a signicant inhibition of TLR2-stimulated cell growth, whereas blocking the activity of p38 MAPK did not impair TLR2-induced cell proliferation (Figure 5B). In support of these ndings, blockade of at least one of these known TLR2 signaling pathways in human GC cells signicantly suppressed the TLR2-induced expression of 6/8 cell cycle regulation and survival genes (Figure 6A). These in vitro ndings were also validated in vivo, where treatment of gp130F/F mice with the U0126 mitogen-activated protein/extracellular signal-regulated kinase kinase (MEK) inhibitor suppressed the TLR2-induced expression of Bcl2a1, Ccnd1, and c-Myc (Figures 6B and 6C). Taken together, these results suggest that TLR2 utilizes multiple signaling cascades to promote the growth of GC cells. TLR2-Expressing Bone Marrow-Derived Inammatory Cells Do Not Promote Gastric Tumors We have previously employed bone marrow chimeras to demonstrate that bone marrow-derived inammatory cell inltrates displaying STAT3 hyperactivation do not contribute to gastric inammation-associated tumorigenesis (Ernst et al., 2008). However, since TLR2 was expressed on both epithelial cells (surface mucosa) and inammatory cells (submucosa) in the stomachs of gp130F/F mice (Figure 7A), we investigated whether TLR2-expressing inammatory cells played a role in gastric

tumorigenesis. Reconstitution of the hematopoietic compartment of irradiated gp130F/F recipient mice with bone marrow from donor gp130F/F:Tlr2/ mice (Figure S5A) did not alleviate the size and number of gastric tumors, nor the overall size of hyperplastic stomachs of the recipient mice (Figures 7B7E). Furthermore, the reciprocal reconstitution of gp130F/F:Tlr2/ recipient mice with gp130F/F donor bone marrow did not exacerbate gastric tumorigenesis (Figures S5B and S5C). As predicted, gastric tumorigenesis and the overall appearance of stomachs of control gp130F/F and gp130F/F:Tlr2/ recipient mice reconstituted with autologous gp130F/F and gp130F/F:Tlr2/ donor bone marrow, respectively, were indistinguishable from their naive gp130F/F and gp130F/F:Tlr2/ counterparts (Figures 7B7E). Collectively, these data further support a role for gastric epithelial TLR2-mediated responses in promoting tumorigenesis. Therapeutic Targeting of TLR2 in gp130F/F Mice Alleviates Gastric Tumorigenesis To determine whether TLR2 may serve as a bona de therapeutic target for GC, we next explored whether therapeutic suppression of TLR2 activity would alleviate the growth of established gastric tumors in gp130F/F mice. We initially examined the efcacy of the OPN-301 TLR2 blocking antibody (Meng et al., 2004) in suppressing acute TLR2-driven gastric responses by pretreating gp130F/F mice by intraperitoneal (i.p.) injection with OPN-301 antibody or isotype control for 30 min prior to administration with Pam3Cys over 6 hr. qPCR conrmed that OPN-301 signicantly suppressed the TLR2-dependent induction of Cxcl2 and Tnfa genes in gastric antrum tissue (Figure 8A). Since the onset of gastric hyperplasia and tumors in gp130F/F mice occurs from 46 weeks of age and progresses in severity up to 24 weeks of age (Jenkins et al., 2005; Tebbutt et al., 2002), we undertook a therapeutic approach by injecting 12week-old gp130F/F mice displaying established gastric disease with OPN-301 or isotype control twice weekly over 10 weeks. Notably, OPN-301 treatment substantially reduced the stomach size and overall tumor burden, including the number of gastric tumors, of gp130F/F mice (Figures 8B8D). Collectively, these data demonstrate that therapeutic targeting of TLR2 suppresses both the initiation and growth of gastric tumors in gp130F/F mice. DISCUSSION Here, we reveal that uncontrolled activation of the oncogenic transcription factor STAT3 promotes gastric tumorigenesis through the specic upregulation of TLR2 in the gastric epithelium. Surprisingly, the composition and activation status of gastric inammatory cell inltrates in tumors of gp130F/F mice was comparable in the presence or absence of TLR2, suggesting that TLR2 signaling does not contribute to tumor-associated gastric inammation. Rather, we observed that TLR2 promoted cell survival and proliferation in the mouse gastric tumor epithelium. Furthermore, our bone marrow reconstitution studies imply a protumorigenic role for TLR2 in a tumor-cell intrinsic rather than extrinsic (i.e., inammatory cell) manner, consistent with our previous data for STAT3 (Ernst et al., 2008). A role for the STAT3-TLR2 axis in gastric tumorigenesis is also supported by the comparable reductions observed in both stomach size and tumor mass upon either ablation of TLR2 or genetically reducing
Cancer Cell 22, 466478, October 16, 2012 2012 Elsevier Inc. 471

Cancer Cell
TLR2 Promotes STAT3-Driven Gastric Tumorigenesis

Figure 4. TLR2 Promotes Gastric Epithelial Tumor Cell Survival and Proliferation
(A and B) Representative photomicrographs showing PCNA-stained cross-sections under low power (A) and high power (B) magnication through the antral region of stomachs from 24-week-old mice of the indicated genotypes. In (A), arrows point to the dened zone of proliferating mucous neck cells of the gastric epithelium. Scale bars = 100 mm. (C) Representative photomicrographs of TUNEL-stained cross-sections through the antral stomach region from mice of the indicated genotypes. Scale bars = 100 mm.

472 Cancer Cell 22, 466478, October 16, 2012 2012 Elsevier Inc.

Cancer Cell
TLR2 Promotes STAT3-Driven Gastric Tumorigenesis

Figure 5. TLR2-Driven Cell Proliferation in Human GC Cells Occurs via Activation of Multiple Signaling Pathways
(A) Representative western blots with the indicated antibodies of lysates from MKN-28 cells pretreated with DMSO vehicle or specic pathway inhibitors at the indicated concentrations for 30 min, and then stimulated without or with 10 mg/ml Pam3Cys for 30 min. Densitometry quantication from two independent experiments for the indicated phosphorylated signaling molecules and relative expression was determined against the corresponding amount of total protein and is presented as the mean SD. (B) Proliferative response of MKN-28 cells incubated over 24 hr without () or with (+) Pam3Cys (10 mg/ml) in the presence of pathway inhibitors for PI3K (Wortmannin), ERK MAPK (U0126), JNK MAPK (SP600125), NF-kB (MG132), and p38 MAPK (SB203580). Data are from four independent experiments in which cells were pretreated for 30 min with either DMSO vehicle () or the indicated concentrations of specic inhibitors (Inhib) and are expressed as the mean SEM % increase in proliferation relative to DMSO pretreated cells in the absence of Pam3Cys. *p < 0.05, **p < 0.01, and **p < 0.001 versus Pam3Cysincubated cells pretreated with DMSO.

the level of STAT3 activation in gp130F/F mice, the latter of which diminishes the STAT3-driven hyperplastic response throughout the entire gastric epithelium (Jenkins et al., 2005). Collectively, our ndings not only support emerging evidence for a broader role of TLRs in regulating noninammatory processes, including epithelial cell proliferation and survival (Shaykhiev et al., 2008), but also demonstrate in a STAT3-driven cancer model that TLR2 is required for tumorigenesis in a cell autonomous manner independently of inammation. Another key nding of our current study was that TLR2 stimulation promoted the proliferation of human gastric epithelial (cancer) cells, as well as the transcriptional induction of numerous genes associated with cell cycle regulation and antiapoptosis, via multiple downstream signaling pathways, namely PI3K/Akt, ERK1/2 and JNK MAPKs, and NF-kB. The notion that these signaling pathways represent an important mecha-

nism promoting gastric epithelial cell growth is supported by previous observations that blockade of either the PI3K/Akt, ERK1/2 MAPK, or NF-kB pathways, which are frequently overexpressed and/or overactivated in human GC (Cinti et al., 2008; Liang et al., 2005; Sasaki et al., 2001), in human GC cell lines suppresses cell proliferation (Asciutti et al., 2011; Lim et al., 2001; Yokota et al., 2010). We also note that a role for the PI3K/Akt pathway in the TLR2-induced survival of epithelial cells has been previously reported (Cario et al., 2007). Furthermore, several of these antiapoptotic and cell proliferation genes (e.g., BCL2A1, BIRC3) have previously been reported to be highly expressed in human GC (Li et al., 2011; Park et al., 1997; Sasada et al., 2004; Takano et al., 2000). We note that despite a subset of these genes, such as CCND1 and C-MYC, being STAT3-regulated genes (Bromberg et al., 1999; Ernst et al., 2008), their reduced expression in tumors of gp130F/F:Tlr2/ versus gp130F/F mice occurs despite comparable STAT3 hyperactivation in tumors from both genotypes. Such a discrepancy can be reconciled by the complex transcriptional regulation of these genes by both

(D) Proliferative response of human MKN-28 and NUGC4 cells to the TLR2 ligand Pam3Cys over 24 hr. Data are from at least three independent experiments in which cells were either untreated or treated with the indicated concentrations of Pam3Cys, and are expressed as the mean SEM % increase in proliferation relative to untreated cells. *p < 0.05, **p < 0.01, ***p < 0.001. (E and F) qPCR gene expression analyses in MKN-28 (E) and NUGC4 (F) cells stimulated (+) with TLR2 ligand (100 ng/ml) for 6 hr. Expression data from at least three independent experiments are shown following normalization for ACTIN. *p < 0.05, **p < 0.01, ***p < 0.001 versus untreated () cells. (G) qPCR gene expression analyses in antral gastric tumor tissue of 24-week-old gp130F/F (n = 5) and gp130F/F:Tlr2/ (n = 5) mice. Expression data are shown following normalization for 18S. *p < 0.05, **p < 0.01. In (E)(G), data are presented from replicate analysis as the mean SEM. See also Figure S4.

Cancer Cell 22, 466478, October 16, 2012 2012 Elsevier Inc. 473

Cancer Cell
TLR2 Promotes STAT3-Driven Gastric Tumorigenesis

Figure 6. TLR2 Upregulates the Expression of Cell Cycle Regulation and Survival Genes in Human GC Cells via Multiple Signaling Pathways
(A) qPCR gene expression analyses in MKN-28 cells pretreated for 30 min with either DMSO vehicle (black bars; ) or specic inhibitors (gray bars) and stimulated with TLR2 ligand (100 ng/ml) for 6 hr. Expression data from three independent experiments are shown following normalization for ACTIN and are presented as the mean SEM % relative to DMSO pretreated cells stimulated with TLR2 ligands. *p < 0.05, **p < 0.01, and ***p < 0.001 versus TLR2 ligand-stimulated cells pretreated with DMSO. Wn, Wortmannin; U0, U0126; SP, SP600125; MG, MG132. (B and C) qPCR gene expression analyses in gastric tumor tissue from gp130F/F (F/F) mice that were either untreated (, white bars) or treated with MEK inhibitor U0126 for 18 hr (U0, gray bars) (B) and either untreated (, white bars), administered with Pam3Cys over 6 hr (P3C, black bars), or pretreated with U0126 for 18 hr prior to administration with Pam3Cys for 6 hr (U0 + P3C, gray bars) (C). Expression data from at least three samples per treatment group are shown following normalization for 18S and are presented from replicate analysis as the mean SEM fold induction. *p < 0.05, **p < 0.01, and ***p < 0.001.

STAT3-dependent and STAT3-independent mechanisms, the latter of which have been widely reported for CCND1 and C-MYC to involve transcriptional regulatory factors such as NF-kB, activating transcription factor 2, activator protein 1, and cAMP response element-binding (Beier et al., 1999; Boulon et al., 2002; Joyce et al., 1999; Kim et al., 2000; Vartanian et al., 2011). Since these transcription factors are also activated by TLR2 (Kawai and Akira, 2006; Mellett et al., 2011; Wang et al., 2011), the reduced gene expression in gp130F/F:Tlr2/ gastric tumors may be a consequence of, at least in part, the loss of gastric TLR2-mediated activation of such transcription factors. Several recent studies support a role for TLR2 in promoting tumorigenesis, albeit by divergent mechanisms, which most likely reect the complex role of TLR2 in innate and adaptive immunity, as well as nonimmune (e.g., epithelial) cellular responses. For instance, TLR2-dependent NF-kB signaling has been associated with WEHI-3B tumor cell survival and disease progression in an in vivo mouse model of myelomonocytic leukemia (Shcheblyakov et al., 2011). In addition, TLR2 can promote the proliferation of hepatocarcinoma tumor cells in a bacterial (Listera monocytogenes) infection-associated ectopic tumor model independent of inuencing immunosuppressive immune cells (e.g., MDSC) (Huang et al., 2007). By contrast, in another ectopic tumor model involving metastatic melanoma cells, it was proposed that TLR2 promoted tumor metastasis via STAT3 activation leading to suppression of antitumor immunity (Yang et al., 2009), an oncogenic mechanism previously assigned to STAT3 (Kortylewski et al., 2005). However, the role of TLR2 in STAT3-mediated gastric tumorigenesis is unlikely to involve such a mechanism; since we have previously
474 Cancer Cell 22, 466478, October 16, 2012 2012 Elsevier Inc.

demonstrated that gastric tumorigenesis in gp130F/F mice is independent of antitumor immunity (Ernst et al., 2008). Furthermore, we demonstrate here that gastric STAT3 hyperactivation in gp130F/F mice is unaffected by targeting TLR2, and bone marrow reconstitution studies conrm that gastric tumor growth is independent of TLR2 expression on hematopoietic-derived tumor-inltrating immune cells. While our current study demonstrates a causal role for TLR2 in gastric tumorigenesis, the role of TLRs in other inammationassociated cancers, in particular colon, is controversial. For instance, our work presented here builds upon recent reports that the MyD88 adaptor protein, which is essential for TLR signaling, promotes intestinal tumorigenesis (Rakoff-Nahoum and Medzhitov, 2007; Schiechl et al., 2011). Similar to our current ndings for gastric tumorigenesis, reduced intestinal tumor formation in MyD88-decient ApcMin/+ mice was associated with higher numbers of apoptotic tumor cells without any change in the frequency of inltrating leukocytes (RakoffNahoum and Medzhitov, 2007). In addition, MyD88-deciency protected mice from colonic tumor development but not inammation in a colitis-associated colorectal cancer (CAC) model induced by oxazolone and azoxymethane (AOM) (Schiechl et al., 2011). On the other hand, others have reported that Myd88/ mice are more susceptible to AOM/dextran sodium sulfate (DSS) CAC (Salcedo et al., 2010). The contrasting roles for TLR2 in CAC induced by AOM/DSS have also been reported, with one study suggesting that TLR2 is dispensable for colonic tumorigenesis (Salcedo et al., 2010), whereas another that tumorigenesis was exacerbated in Tlr2/ mice (Lowe et al., 2010). Collectively, these discrepancies most likely reect inherent differences in the experimental approaches

Cancer Cell
TLR2 Promotes STAT3-Driven Gastric Tumorigenesis

and TLR2 expression in human GC correlates with a poor prognosis, together with the suppression of gastric tumorigenesis in a STAT3-driven preclinical GC mouse model following genetic and therapeutic targeting of TLR2, suggest that TLR2 represents a compelling biomarker and therapeutic target for GC.
EXPERIMENTAL PROCEDURES Human Gastric Biopsy Tissue Collection Antral gastric biopsies were collected at Monash Medical Centre (Melbourne, Australia) from 18 patients undergoing upper gastrointestinal endoscopy or surgical resection for clinical indications. Patients with a history of taking nonsteroidal anti-inammatory drugs, proton pump inhibitors, or antibiotics were excluded from the study. Biopsies were either snap-frozen in liquid nitrogen or stored in 10% formalin, the latter for histopathological assessment on H&E-stained tissue sections using the revised version of the Sydney System (Dixon et al., 1996). Biopsies free of any lesions were considered normal. Full and informed consent was obtained from all patients, and biopsy collection was approved by the Southern Health Human Research Ethics Committee. Mice and Treatments The generation of gp130F/F, gp130F/F:Stat3+/, and gp130F/F:IL-11r/ mice has been previously described (Ernst et al., 2008; Jenkins et al., 2005, Tebbutt et al., 2002). Mice homozygous null for either Tlr2 (Takeuchi et al., 1999) or Tlr4 (Hoshino et al., 1999) were used to generate gp130F/F:Tlr2/ and gp130F/F:Tlr4/ mice. The specicity and pharmacokinetics of the OPN-301 monoclonal TLR2 blocking antibody (Opsona Therapeutics), analogous to clone T2.5 (Meng et al., 2004), have been previously described (Arslan et al., 2010). For acute studies, 12-week-old male gp130F/F mice were i.p. injected with 10mg/kg of OPN-301 antibody or IgG1 isotype control for 30 min prior to i.p. administration with 2mg/kg of Pam3Cys (EMC Microcollections) for 3 or 6 hr. For chronic studies, 12-week-old male gp130F/F mice were i.p. injected with 10mg/kg of OPN-301 or IgG1 isotype control twice weekly over 10 weeks. For treatment with the MEK 1/2 inhibitor U0126 (Cell Signaling Technology), gp130F/F mice were i.p. injected with U0126 (10mg/kg) or vehicle (Dimethyl sulfoxide; DMSO) for 18 hr prior to i.p. injection of Pam3Cys (2 mg/kg) for 6 hr. All experiments were endorsed by the Monash University Animal Ethics Committee and, where appropriate, included genetically-matched (mixed 129Sv 3 C57BL/6 background) gp130+/+ (wild-type) littermate controls. Mice were housed under specic pathogen-free conditions and were agematched for each experiment. Gastric Epithelial Cells The generation of primary mouse gastric epithelial cells (Viala et al., 2004) and maintenance of gastric IMGE-5, AGS, and MKN-28 cells (Jenkins et al., 2005) have been described previously. Transfections of mouse IMGE-5 and human MKN-28 and AGS cells were performed with the Nucleofector II System (Amaxa) and FuGene-6 (Roche), respectively. Chromatin immunoprecipitations were performed using the Imprint ChIP kit (Sigma-Aldrich) with pTyr705-STAT3 and IgG antibodies (Santa Cruz Biotechnology). For experiments involving the use of specic pathway inhibitors, cells were pretreated for 30 min with either DMSO vehicle, U0126, the PI3K inhibitor Wortmannin (Sigma), the NF-kB inhibitor MG132 (Calbiochem), the p38 MAPK inhibitor SB203580 (Calbiochem), or the JNK MAPK inhibitor SP600125 (Calbiochem) at the indicated concentrations prior to stimulation with Pam3Cys. Statistical Analyses All statistical analyses were performed using GraphPad Prism V5.0 software. The normality of data was assessed using the DAgostino and Pearson omnibus K2 normality test. Statistical signicance (p < 0.05) between the means of two groups was determined using Student t tests (normal distribution) or Mann-Whitney U tests (abnormal distribution). Statistical comparisons of the means of multiple (R3) groups were determined using repeatedmeasures one-way ANOVA (normal distribution) or Kruskal Wallis nonparametric analyses (abnormal distribution). Additional procedures are described in Supplemental Experimental Procedures.

Figure 7. TLR2-Mediated Gastric Tumorigenesis Is Independent of Bone Marrow-Derived Inammatory Cells


(A) Representative TLR2-stained cross-sections through the antral stomach tumor region of a 24-week-old gp130F/F mouse showing staining around the surface area of the glandular gastric epithelium (left panel), as well as in submucosal inammatory cell inltrates (right panel). Scale bars = 100 mm. (B) Representative appearance of stomachs from 22-week-old gp130F/F naive mice, or recipient gp130F/F mice reconstituted with autologous gp130F/F (F/FF/F), or heterologous gp130F/F:Tlr2/ (F/FF/F:T2) donor bonemarrow. a, antrum; f, fundus. (CE) Bar graphs depict the total mass (grams) of stomachs (C) and gastric tumors (D) as well as the incidence of tumors (E) from the indicated mice. Data for each group are expressed as the mean SEM. gp130F/F, n = 3; F/FF/F, n = 3; and F/FF/F:T2, n = 10. See also Figure S5.

used by investigators to study chemical-induced intestinal carcinogenesis. In summary, our current study denes that an endogenously expressed PRR promotes gastric tumorigenesis by regulating the expression of genes involved in cell survival and proliferation, and importantly highlights the need to further investigate the broader role of TLR2 in other inammatory-related cancers, for instance colon, prostate, and liver. In this regard, it is noteworthy that deregulated STAT3 activation, including by gp130-acting cytokines, has been linked to the pathogenesis of tumor formation in these organs (Bollrath et al., 2009; Gao et al., 2005; Kawada et al., 2006; Ogata et al., 2006; Rebouissou et al., 2009). Collectively, our ndings that increased STAT3 activation

Cancer Cell 22, 466478, October 16, 2012 2012 Elsevier Inc. 475

Cancer Cell
TLR2 Promotes STAT3-Driven Gastric Tumorigenesis

Figure 8. Therapeutic Suppression of Gastric Tumorigenesis in gp130F/F Mice Treated with a TLR2 Blocking Antibody
(A) qPCR of Tnfa and Cxcl2 gene expression was performed on antral gastric tissue of 12-week-old gp130F/F mice that were untreated (gray bars) or pretreated with either OPN-301 (white bars) or isotype (black bars) control antibodies for 30 min prior to administration with Pam3Cys over 6 hr. Expression data from at least two samples per time point and treatment group are shown following normalization for 18S and are presented from replicate analysis as the mean SEM fold induction. *p < 0.05. (B) Representative appearance of stomachs from 22-week-old gp130F/F mice treated twice weekly over 10 weeks with either OPN-301 or isotype control antibodies. Arrows indicate macroscopically visible tumors >4 mm in diameter. Fundic (f) and antral (a) stomach regions are labeled. (C and D) Scatter plots depicting the total mass (grams) of gastric tumors (C), as well as the incidence of tumors (D), from 22-week-old gp130F/F mice treated twice weekly over 10 weeks with either OPN-301 or isotype control antibodies. Data for each group are expressed as the mean SEM. *p < 0.05.

SUPPLEMENTAL INFORMATION Supplemental Information includes two tables, ve gures, and Supplemental Experimental Procedures and can be found with this article online at http://dx. doi.org/10.1016/j.ccr.2012.08.010. ACKNOWLEDGMENTS We thank E. Vidacs and M. Meilak for technical assistance and N. Watkins and P. Hertzog for critical reading of the manuscript. We are grateful to S. Akira and S. Uematsu (University of Osaka, Japan) for kindly providing the Tlr2/ and Tlr4/ mice. This work was supported by grants from the Association for International Cancer Research (U.K.), Cancer Council of Victoria (Australia), and the National Health and Medical Research Council (Australia), as well as the Operational Infrastructure Support Program by the Victorian Government of Australia. B.J.J. is supported by both a Senior Medical Research Fellowship awarded by the Sylvia and Charles Viertel Foundation and a Monash University Fellowship. We also acknowledge the generosity of the family and friends of the late L. Hender for supporting cancer research. W.M. is a current employee and A.E.P. is a former employee and shareholder of Opsona Therapeutics, a drug discovery and development company focused on the role of TLR signaling in human immunology. Received: July 28, 2011 Revised: April 12, 2012 Accepted: August 14, 2012 Published: October 15, 2012 REFERENCES Arslan, F., Smeets, M.B., ONeill, L.A., Keogh, B., McGuirk, P., Timmers, L., Tersteeg, C., Hoefer, I.E., Doevendans, P.A., Pasterkamp, G., and de Kleijn, D.P. (2010). Myocardial ischemia/reperfusion injury is mediated by leukocytic toll-like receptor-2 and reduced by systemic administration of a novel antitoll-like receptor-2 antibody. Circulation 121, 8090.

Asciutti, S., Akiri, G., Grumolato, L., Vijayakumar, S., and Aaronson, S.A. (2011). Diverse mechanisms of Wnt activation and effects of pathway inhibition on proliferation of human gastric carcinoma cells. Oncogene 30, 956966. Beier, F., Lee, R.J., Taylor, A.C., Pestell, R.G., and LuValle, P. (1999). Identication of the cyclin D1 gene as a target of activating transcription factor 2 in chondrocytes. Proc. Natl. Acad. Sci. USA 96, 14331438. Bollrath, J., and Greten, F.R. (2009). IKK/NF-kappaB and STAT3 pathways: central signalling hubs in inammation-mediated tumour promotion and metastasis. EMBO Rep. 10, 13141319. Bollrath, J., Phesse, T.J., von Burstin, V.A., Putoczki, T., Bennecke, M., Bateman, T., Nebelsiek, T., Lundgren-May, T., Canli, O., Schwitalla, S., et al. (2009). gp130-mediated Stat3 activation in enterocytes regulates cell survival and cell-cycle progression during colitis-associated tumorigenesis. Cancer Cell 15, 91102. , A., Blanchard, J.M., Hipskind, R.A., Boulon, S., Dantonel, J.C., Binet, V., Vie and Philips, A. (2002). Oct-1 potentiates CREB-driven cyclin D1 promoter activation via a phospho-CREB- and CREB binding protein-independent mechanism. Mol. Cell. Biol. 22, 77697779. Bromberg, J.F., Wrzeszczynska, M.H., Devgan, G., Zhao, Y., Pestell, R.G., Albanese, C., and Darnell, J.E., Jr. (1999). Stat3 as an oncogene. Cell 98, 295303. Cario, E., Gerken, G., and Podolsky, D.K. (2007). Toll-like receptor 2 controls mucosal inammation by regulating epithelial barrier function. Gastroenterology 132, 13591374. Cinti, C., Vindigni, C., Zamparelli, A., La Sala, D., Epistolato, M.C., Marrelli, D., Cevenini, G., and Tosi, P. (2008). Activated Akt as an indicator of prognosis in gastric cancer. Virchows Arch. 453, 449455. Dixon, M.F., Genta, R.M., Yardley, J.H., and Correa, P. (1996). Classication and grading of gastritis. The updated Sydney System. International Workshop on the Histopathology of Gastritis, Houston 1994. Am. J. Surg. Pathol. 20, 11611181. El-Omar, E.M., Rabkin, C.S., Gammon, M.D., Vaughan, T.L., Risch, H.A., Schoenberg, J.B., Stanford, J.L., Mayne, S.T., Goedert, J., Blot, W.J., et al.

476 Cancer Cell 22, 466478, October 16, 2012 2012 Elsevier Inc.

Cancer Cell
TLR2 Promotes STAT3-Driven Gastric Tumorigenesis

(2003). Increased risk of noncardia gastric cancer associated with proinammatory cytokine gene polymorphisms. Gastroenterology 124, 11931201. Ellmark, P., Ingvarsson, J., Carlsson, A., Lundin, B.S., Wingren, C., and Borrebaeck, C.A. (2006). Identication of protein expression signatures associated with Helicobacter pylori infection and gastric adenocarcinoma using recombinant antibody microarrays. Mol. Cell. Proteomics 5, 16381646. Ernst, M., Najdovska, M., Grail, D., Lundgren-May, T., Buchert, M., Tye, H., Matthews, V.B., Armes, J., Bhathal, P.S., Hughes, N.R., et al. (2008). STAT3 and STAT1 mediate IL-11-dependent and inammation-associated gastric tumorigenesis in gp130 receptor mutant mice. J. Clin. Invest. 118, 17271738. Gao, L., Zhang, L., Hu, J., Li, F., Shao, Y., Zhao, D., Kalvakolanu, D.V., Kopecko, D.J., Zhao, X., and Xu, D.Q. (2005). Down-regulation of signal transducer and activator of transcription 3 expression using vector-based small interfering RNAs suppresses growth of human prostate tumor in vivo. Clin. Cancer Res. 11, 63336341. Gong, W., Wang, L., Yao, J.C., Ajani, J.A., Wei, D., Aldape, K.D., Xie, K., Sawaya, R., and Huang, S. (2005). Expression of activated signal transducer and activator of transcription 3 predicts expression of vascular endothelial growth factor in and angiogenic phenotype of human gastric cancer. Clin. Cancer Res. 11, 13861393. Hold, G.L., Rabkin, C.S., Chow, W.H., Smith, M.G., Gammon, M.D., Risch, H.A., Vaughan, T.L., McColl, K.E., Lissowska, J., Zatonski, W., et al. (2007). A functional polymorphism of toll-like receptor 4 gene increases risk of gastric carcinoma and its precursors. Gastroenterology 132, 905912. Hoshino, K., Takeuchi, O., Kawai, T., Sanjo, H., Ogawa, T., Takeda, Y., Takeda, K., and Akira, S. (1999). Cutting edge: Toll-like receptor 4 (TLR4)-decient mice are hyporesponsive to lipopolysaccharide: evidence for TLR4 as the Lps gene product. J. Immunol. 162, 37493752. Huang, B., Zhao, J., Shen, S., Li, H., He, K.L., Shen, G.X., Mayer, L., Unkeless, J., Li, D., Yuan, Y., et al. (2007). Listeria monocytogenes promotes tumor growth via tumor cell toll-like receptor 2 signaling. Cancer Res. 67, 43464352. Itadani, H., Oshima, H., Oshima, M., and Kotani, H. (2009). Mouse gastric tumor models with prostaglandin E2 pathway activation show similar gene expression proles to intestinal-type human gastric cancer. BMC Genomics 10, 615622. Jenkins, B.J., Grail, D., Nheu, T., Najdovska, M., Wang, B., Waring, P., Inglese, M., McLoughlin, R.M., Jones, S.A., Topley, N., et al. (2005). Hyperactivation of Stat3 in gp130 mutant mice promotes gastric hyperproliferation and desensitizes TGF-beta signaling. Nat. Med. 11, 845852. Joyce, D., Bouzahzah, B., Fu, M., Albanese, C., DAmico, M., Steer, J., Klein, J.U., Lee, R.J., Segall, J.E., Westwick, J.K., et al. (1999). Integration of Racdependent regulation of cyclin D1 transcription through a nuclear factorkappaB-dependent pathway. J. Biol. Chem. 274, 2524525249. Kawada, M., Seno, H., Uenoyama, Y., Sawabu, T., Kanda, N., Fukui, H., Shimahara, Y., and Chiba, T. (2006). Signal transducers and activators of transcription 3 activation is involved in nuclear accumulation of beta-catenin in colorectal cancer. Cancer Res. 66, 29132917. Kawai, T., and Akira, S. (2006). TLR signaling. Cell Death Differ. 13, 816825. Kawai, T., and Akira, S. (2010). The role of pattern-recognition receptors in innate immunity: update on Toll-like receptors. Nat. Immunol. 11, 373384. Kim, D.W., Gazourian, L., Quadri, S.A., Romieu-Mourez, R., Sherr, D.H., and Sonenshein, G.E. (2000). The RelA NF-kappaB subunit and the aryl hydrocarbon receptor (AhR) cooperate to transactivate the c-myc promoter in mammary cells. Oncogene 19, 54985506. Kortylewski, M., Kujawski, M., Wang, T., Wei, S., Zhang, S., Pilon-Thomas, S., , J., Kerr, W.G., et al. (2005). Inhibiting Stat3 signaling in Niu, G., Kay, H., Mule the hematopoietic system elicits multicomponent antitumor immunity. Nat. Med. 11, 13141321. Li, Z., Chen, J., Chan, K.W., Qiao, L., and Wong, B.C.Y. (2011). A possible role of cIAP2 in Helicobacter pylori-associated gastric cancer. Cancer Lett. 313, 192200. Liang, B., Wang, S., Zhu, X.G., Yu, Y.X., Cui, Z.R., and Yu, Y.Z. (2005). Increased expression of mitogen-activated protein kinase and its upstream

regulating signal in human gastric cancer. World J. Gastroenterol. 11, 623628. Lim, J.W., Kim, H., and Kim, K.H. (2001). Nuclear factor-kappaB regulates cyclooxygenase-2 expression and cell proliferation in human gastric cancer cells. Lab. Invest. 81, 349360. Lowe, E.L., Crother, T.R., Rabizadeh, S., Hu, B., Wang, H., Chen, S., Shimada, K., Wong, M.H., Michelsen, K.S., and Arditi, M. (2010). Toll-like receptor 2 signaling protects mice from tumor development in a mouse model of colitis-induced cancer. PLoS ONE 5, e13027. Mellett, M., Atzei, P., Jackson, R., ONeill, L.A., and Moynagh, P.N. (2011). Mal mediates TLR-induced activation of CREB and expression of IL-10. J. Immunol. 186, 49254935. Meng, G., Rutz, M., Schiemann, M., Metzger, J., Grabiec, A., Schwandner, R., Luppa, P.B., Ebel, F., Busch, D.H., Bauer, S., et al. (2004). Antagonistic antibody prevents toll-like receptor 2-driven lethal shock-like syndromes. J. Clin. Invest. 113, 14731481. Ogata, H., Kobayashi, T., Chinen, T., Takaki, H., Sanada, T., Minoda, Y., Koga, K., Takaesu, G., Maehara, Y., Iida, M., and Yoshimura, A. (2006). Deletion of the SOCS3 gene in liver parenchymal cells promotes hepatitis-induced hepatocarcinogenesis. Gastroenterology 131, 179193. Ooi, C.H., Ivanova, T., Wu, J., Lee, M., Tan, I.B., Tao, J., Ward, L., Koo, J.H., Gopalakrishnan, V., Zhu, Y., et al. (2009). Oncogenic pathway combinations predict clinical prognosis in gastric cancer. PLoS Genet. 5, e1000676. Oshima, H., Matsunaga, A., Fujimura, T., Tsukamoto, T., Taketo, M.M., and Oshima, M. (2006). Carcinogenesis in mouse stomach by simultaneous activation of the Wnt signaling and prostaglandin E2 pathway. Gastroenterology 131, 10861095. Park, I.C., Lee, S.H., Whang, D.Y., Hong, W.S., Choi, S.S., Shin, H.S., Choe, T.B., and Hong, S.I. (1997). Expression of a novel Bcl-2 related gene, B-1, in various human cancers and cancer cell lines. Anticancer Res. 17 (6D), 46194622. Parkin, D.M., Bray, F., Ferlay, J., and Pisani, P. (2005). Global cancer statistics, 2002. CA Cancer J. Clin. 55, 74108. Rakoff-Nahoum, S., and Medzhitov, R. (2007). Regulation of spontaneous intestinal tumorigenesis through the adaptor protein MyD88. Science 317, 124127. Rebouissou, S., Amessou, M., Couchy, G., Poussin, K., Imbeaud, S., Pilati, C., Izard, T., Balabaud, C., Bioulac-Sage, P., and Zucman-Rossi, J. (2009). Frequent in-frame somatic deletions activate gp130 in inammatory hepatocellular tumours. Nature 457, 200204. Rosenstiel, P., Hellmig, S., Hampe, J., Ott, S., Till, A., Fischbach, W., Sahly, H., lsch, U.R., Philpott, D., and Schreiber, S. (2006). Inuence of Lucius, R., Fo polymorphisms in the NOD1/CARD4 and NOD2/CARD15 genes on the clinical outcome of Helicobacter pylori infection. Cell. Microbiol. 8, 11881198. Salcedo, R., Worschech, A., Cardone, M., Jones, Y., Gyulai, Z., Dai, R.M., Wang, E., Ma, W., Haines, D., OhUigin, C., et al. (2010). MyD88-mediated signaling prevents development of adenocarcinomas of the colon: role of interleukin 18. J. Exp. Med. 207, 16251636. Sasada, T., Takedatsu, H., Azuma, K., Koga, M., Maeda, Y., Shichijo, S., Shoumura, H., Hirai, T., Takabayashi, A., and Itoh, K. (2004). Immediate early response gene X-1, a stress-inducible antiapoptotic gene, encodes cytotoxic T-lymphocyte (CTL) epitopes capable of inducing human leukocyte antigenA33-restricted and tumor-reactive CTLs in gastric cancer patients. Cancer Res. 64, 28822888. Sasaki, N., Morisaki, T., Hashizume, K., Yao, T., Tsuneyoshi, M., Noshiro, H., Nakamura, K., Yamanaka, T., Uchiyama, A., Tanaka, M., and Katano, M. (2001). Nuclear factor-kappaB p65 (RelA) transcription factor is constitutively activated in human gastric carcinoma tissue. Clin. Cancer Res. 7, 41364142. Schiechl, G., Bauer, B., Fuss, I., Lang, S.A., Moser, C., Ruemmele, P., RoseJohn, S., Neurath, M.F., Geissler, E.K., Schlitt, H.J., et al. (2011). Tumor development in murine ulcerative colitis depends on MyD88 signaling of colonic F4/80+CD11b(high)Gr1(low) macrophages. J. Clin. Invest. 121, 16921708.

Cancer Cell 22, 466478, October 16, 2012 2012 Elsevier Inc. 477

Cancer Cell
TLR2 Promotes STAT3-Driven Gastric Tumorigenesis

Shaykhiev, R., Behr, J., and Bals, R. (2008). Microbial patterns signaling via Toll-like receptors 2 and 5 contribute to epithelial repair, growth and survival. PLoS ONE 3, e1393. Shcheblyakov, D.V., Logunov, D.Y., Rakovskaya, I.V., Shmarov, M.M., Naroditsky, B.S., and Ginzburg, A.L. (2011). Triggering of toll-like receptor-2 in mouse myelomonocytic leukaemia cells WEHI-3B leads to the suppression of apoptosis and promotes tumor progression in vivo. Acta Naturae 3, 8393. Tahara, T., Arisawa, T., Wang, F., Shibata, T., Nakamura, M., Sakata, M., Hirata, I., and Nakano, H. (2007). Toll-like receptor 2 -196 to 174del polymorphism inuences the susceptibility of Japanese people to gastric cancer. Cancer Sci. 98, 17901794. Takano, Y., Kato, Y., van Diest, P.J., Masuda, M., Mitomi, H., and Okayasu, I. (2000). Cyclin D2 overexpression and lack of p27 correlate positively and cyclin E inversely with a poor prognosis in gastric cancer cases. Am. J. Pathol. 156, 585594. Takeuchi, O., Hoshino, K., Kawai, T., Sanjo, H., Takada, H., Ogawa, T., Takeda, K., and Akira, S. (1999). Differential roles of TLR2 and TLR4 in recognition of gram-negative and gram-positive bacterial cell wall components. Immunity 11, 443451. Tebbutt, N.C., Giraud, A.S., Inglese, M., Jenkins, B.J., Waring, P., Clay, F.J., Malki, S., Alderman, B.M., Grail, D., Hollande, F., et al. (2002). Reciprocal regulation of gastrointestinal homeostasis by SHP2 and STAT-mediated trefoil gene activation in gp130 mutant mice. Nat. Med. 8, 10891097. Uemura, N., Okamoto, S., Yamamoto, S., Matsumura, N., Yamaguchi, S., Yamakido, M., Taniyama, K., Sasaki, N., and Schlemper, R.J. (2001). Helicobacter pylori infection and the development of gastric cancer. N. Engl. J. Med. 345, 784789. Uno, K., Kato, K., Atsumi, T., Suzuki, T., Yoshitake, J., Morita, H., Ohara, S., Kotake, Y., Shimosegawa, T., and Yoshimura, T. (2007). Toll-like receptor (TLR) 2 induced through TLR4 signaling initiated by Helicobacter pylori cooperatively amplies iNOS induction in gastric epithelial cells. Am. J. Physiol. Gastrointest. Liver Physiol. 293, G1004G1012.

Vartanian, R., Masri, J., Martin, J., Cloninger, C., Holmes, B., Artinian, N., Funk, A., Ruegg, T., and Gera, J. (2011). AP-1 regulates cyclin D1 and c-MYC transcription in an AKT-dependent manner in response to mTOR inhibition: role of AIP4/Itch-mediated JUNB degradation. Mol. Cancer Res. 9, 115130. Viala, J., Chaput, C., Boneca, I.G., Cardona, A., Girardin, S.E., Moran, A.P., met, S., Huerre, M.R., Coyle, A.J., et al. (2004). Nod1 responds Athman, R., Me to peptidoglycan delivered by the Helicobacter pylori cag pathogenicity island. Nat. Immunol. 5, 11661174. Walser, T., Cui, X., Yanagawa, J., Lee, J.M., Heinrich, E., Lee, G., Sharma, S., and Dubinett, S.M. (2008). Smoking and lung cancer: the role of inammation. Proc. Am. Thorac. Soc. 5, 811815. Wang, J., Ma, J., Charboneau, R., Barke, R., and Roy, S. (2011). Morphine inhibits murine dendritic cell IL-23 production by modulating Toll-like receptor 2 and Nod2 signaling. J. Biol. Chem. 286, 1022510232. West, X.Z., Malinin, N.L., Merkulova, A.A., Tischenko, M., Kerr, B.A., Borden, E.C., Podrez, E.A., Salomon, R.G., and Byzova, T.V. (2010). Oxidative stress induces angiogenesis by activating TLR2 with novel endogenous ligands. Nature 467, 972976. Yakata, Y., Nakayama, T., Yoshizaki, A., Kusaba, T., Inoue, K., and Sekine, I. (2007). Expression of p-STAT3 in human gastric carcinoma: signicant correlation in tumour invasion and prognosis. Int. J. Oncol. 30, 437442. Yang, H.Z., Cui, B., Liu, H.Z., Mi, S., Yan, J., Yan, H.M., Hua, F., Lin, H., Cai, W.F., Xie, W.J., et al. (2009). Blocking TLR2 activity attenuates pulmonary metastases of tumor. PLoS ONE 4, e6520. Yang, J., Liao, X., Agarwal, M.K., Barnes, L., Auron, P.E., and Stark, G.R. (2007). Unphosphorylated STAT3 accumulates in response to IL-6 and activates transcription by binding to NFkappaB. Genes Dev. 21, 13961408. Yokota, S., Okabayashi, T., Rehli, M., Fujii, N., and Amano, K. (2010). Helicobacter pylori lipopolysaccharides upregulate toll-like receptor 4 expression and proliferation of gastric epithelial cells via the MEK1/2-ERK1/2 mitogen-activated protein kinase pathway. Infect. Immun. 78, 468476.

478 Cancer Cell 22, 466478, October 16, 2012 2012 Elsevier Inc.

Cancer Cell

Article
Loss of Cutaneous TSLP-Dependent Immune Responses Skews the Balance of Inammation from Tumor Protective to Tumor Promoting
-Dante Durham,2 and Freddy Radtke1,* Matteo Di Piazza,1,3 Craig S. Nowell,1,3 Ute Koch,1 Andre
de rale de Lausanne, School of Life Sciences, Swiss Institute for Experimental Cancer Research, Lausanne, Polytechnique Fe Vaud 1015, Switzerland 2Department of Radio-oncology, Centre Hospitalier Universitaire Vaudois, Lausanne, Vaud 1011, Switzerland 3These authors contributed equally to this work *Correspondence: freddy.radtke@ep.ch http://dx.doi.org/10.1016/j.ccr.2012.08.016
1Ecole

SUMMARY

Inammation can promote or inhibit cancer progression. In this study we have addressed the role of the proinammatory cytokine thymic stromal lymphopoietin (TSLP) during skin carcinogenesis. Using conditional loss- and gain-of-function mouse models for Notch and Wnt signaling, respectively, we demonstrate that TSLP-mediated inammation protects against cutaneous carcinogenesis by acting directly on CD4 and CD8 T cells. Genetic ablation of TSLP receptor (TSLPR) perturbs T-cell-mediated protection and results in the accumulation of CD11b+Gr1+ myeloid cells. These promote tumor growth by secreting Wnt ligands and augmenting b-catenin signaling in the neighboring epithelium. Epithelial specic ablation of b-catenin prevents both carcinogenesis and the accumulation of CD11b+Gr1+ myeloid cells, suggesting tumor cells initiate a feed-forward loop that induces protumorigenic inammation.
INTRODUCTION For many years cancer has been regarded predominantly as a cell autonomous process in which genetically transformed cells propagate the development of malignant neoplasms. However, today it is well established that tumor progression requires complex interactions between neoplastic cells and the host-derived stroma, including tumor-associated broblasts, niche-dening cells, and vasculature. Of critical importance is also the relationship between the tumor and the host immune system. Originally, tumor-inltrating immune cells were thought to reect the attempts of the immune system to eliminate cancer cells. However, today a large body of evidence indicates that malignant neoplasms can induce qualitatively distinct types of inammation that are protumorigenic (Coussens and Werb, 2002; Hanahan and Weinberg, 2011). Cancer-associated inammation frequently exhibits proles that promote both immune evasion and the growth of malignant cells. Tumors can evade T-cell-mediated clearance by skewing the acquired immune response from cytotoxic Th1 to permissive Th2 proles (Aspord et al., 2007) or by inducing immunosuppressive regulatory T cells (Quezada et al., 2011). Tumors can also co-opt cells of the innate immune system to become constituents of the protumorigenic stroma. Notably, cells of the myeloid lineage, such as macrophages (Allavena et al., 2008) and neutrophils (Fridlender et al., 2009) can promote tumor growth by a variety of mechanisms, depending on their differentiation status. Furthermore, poorly dened immature myeloid cells can foster tumor growth directly (Kowanetz et al., 2010; Qian et al., 2011; Yang et al., 2011) or by acting as myeloid-derived suppressor cells (Gabrilovich and Nagaraj, 2009). These data demonstrate that tumor cells modulate both acquired and innate arms of the immune system to form an integral component of the tumor stroma. The molecular basis for such immunomodulatory roles, however, is poorly dened,

Signicance Inammatory responses that promote or inhibit cancer development are qualitatively distinct at a cellular and molecular level. In this study we show that TSLP can induce antitumorigenic inammation by acting directly on T cells, which prevent the growth of b-catenin-dependent skin tumors. Ablation of TSLP-mediated immune responses dramatically alters the inammatory prole, resulting in the accumulation of CD11b+Gr1+ myeloid cells that promote tumor growth by secreting Wnts. These ndings provide a mechanistic explanation regarding the reduced risk of cancer in patients who suffer from allergic disorders and demonstrate a tumor protective role for TSLP. Furthermore, we reveal an additional mechanism by which tumor-associated myeloid cells can promote tumor growth.
Cancer Cell 22, 479493, October 16, 2012 2012 Elsevier Inc. 479

Cancer Cell
TSLP Promotes Antitumor Inammation in the Skin

and it is unclear to what extent oncogenes/tumor suppressors inuence tumor development by such indirect mechanisms. Mutations in the Notch signaling pathway are associated with a variety of human cancers and indicate that Notch can function as either a tumor suppressor or an oncogene (Weng and Aster, 2004). In the skin, conditional inactivation of Notch1 in the mouse epidermis results in the development of basal/ squamous cell carcinoma after chemically induced carcinogenesis (Demehri et al., 2009; Nicolas et al., 2003; Proweller et al., 2006), suggesting that Notch functions as a tumor suppressor in this tissue. In accordance, loss-of-function mutations in human Notch receptors correlate with cutaneous and lung squamous cell carcinoma (Agrawal et al., 2011; Stransky et al., 2011; Wang et al., 2011). Factors acting downstream of Notch with respect to epithelial malignancies remain to be fully elucidated. However, deletion of Notch1 in the murine epidermis leads to elevated levels of activated b-catenin (Nicolas et al., 2003), which has been shown to induce skin tumor formation and regulate the maintenance of CD34+ cancer stem cells (Gat et al., 1998; Malanchi et al., 2008). The tumor suppressor activity of Notch signaling in the skin may therefore at least in part be due to negative regulation of the Wnt/b-catenin pathway. Whether b-catenin accumulation is due to cell autonomous deregulation or is mediated by crosstalk between different cell populations is unclear, although there is evidence for both mechanisms (Devgan et al., 2005; Kwon et al., 2011; Sanders et al., 2009). Conditional inactivation of both Notch1 and Notch2 in the murine epidermis results in a chronic inammatory condition resembling a severe form of atopic dermatitis (AD) that occurs as a consequence of elevated TSLP expression by keratinocytes (Demehri et al., 2008; Dumortier et al., 2010). This cytokine has been shown to affect a variety of cell types in the immune system and as such is believed to play a role in allergic disorders, including AD and asthma (Ziegler, 2010). Interestingly, elevated TSLP has also been associated with progression of various epithelial malignancies, including breast and pancreatic cancer (De Monte et al., 2011; Olkhanud et al., 2011; Pedroza-Gonzalez et al., 2011). In light of this, it is intriguing that loss of Notch in the skin results in increased tumor susceptibility and elevated expression of TSLP. It is presently unclear to what extent these processes are linked, although elucidation of how they interact may provide insight into how inammation can exert either anti- or protumorigenic effects. In this study, we have delineated the role of TSLP-mediated inammation during skin tumorigenesis by using loss- and gain-of-function approaches for Notch and Wnt signaling, respectively. RESULTS Loss of TSLP Receptor Signaling in Notch Mutant Mice Results in Cutaneous Malignancy To determine the role of TSLP in cutaneous cancer associated with loss of Notch signaling, mice carrying a K5CreERT transgene and oxed alleles of both Notch1 and Notch2 receptors (N1N2K5) were crossed with Tslpr/ mice, thus generating Notch1lox/lox;Notch2lox/lox;K5CreERT;Tslpr/ mice (herein referred to as N1N2K5 TSLPR). This allowed us to conditionally
480 Cancer Cell 22, 479493, October 16, 2012 2012 Elsevier Inc.

ablate Notch signaling in the epidermis in the absence of TSLPR signaling and therefore to dissect the role of TSLP during chronic inammation in response to Notch ablation. Consistent with our previous study, N1N2K5 mice developed an AD-like disease within 5 weeks of skin-specic inactivation of Notch1 and Notch2, characterized by the degeneration of hair follicles into large cysts. N1N2K5 TSLPR mice also developed a severe skin disorder characterized by hair follicle degeneration (Figures 1A1B; Figures S1AS1C available online). However, in these mutants the cystic lesions could be broadly classied into two subtypes: those lined by nonproliferative epithelial cells consisting of 13 cell layers or larger lesions surrounded by hyperproliferative Ki67+ epithelial cells (Figures 1B and 1C; Figure S1A). The hyperproliferative cystic structures in N1N2K5 TSLPR mice frequently extended deep into the underlying connective tissue of the dermis and were often visible on the surface of the skin (Figures 1A and S1B). Pathological analysis classied these hyperproliferative cysts as dysplastic lesions resembling keratoacanthomas, relatively low-grade hair follicle malignancies that can progress to invasive squamous cell carcinomas. The severity of the phenotype in N1N2K5 TSLPR animals was reected in survival rates, as these mutants died by day 60 post-gene-inactivation (Figure 1D), most likely because of complications resulting from barrier dysfunction and tumor burden (data not shown). These data indicate that TSLPR signaling prevents tumor formation in Notch-decient skin. In support of this, injection of N1N2K5 mice with TSLP-neutralizing antibodies resulted in tumor development in 60% of animals (Figure S1D), thus conrming the tumor protective function of TSLP-TSLPR interactions. The Tumor Protective Effect of TSLP Is Mediated by Hematopoietic Cells The TSLPR is expressed by a variety of hematopoietic cells, including T, B, and dendritic cells. We therefore speculated that elevated levels of TSLP may elicit tumor protective immune responses. To conrm this, we generated bone marrow (BM) chimeras in which CD45.2 N1N2K5 hosts were reconstituted with CD45.1 Tslpr/ BM prior to Notch inactivation. The reciprocal experiment was also performed, in which CD45.1 Tslpr+/+ BM was used to reconstitute CD45.2 N1N2K5 TSLPR mice (Figure 2A). In the former, TSLPR/ / N1N2K5 chimeras developed hyperproliferative dermal cysts, resembling the tumors observed in N1N2K5 TSLPR mice upon ablation of Notch signaling (Figure 2B). Conversely, N1N2K5 TSLPR mice reconstituted with Tslpr+/+ BM were protected from tumor formation and instead developed the AD-like disease that occurs in N1N2K5 animals (Figure 2B). These data conrm that TSLPR signaling elicits antitumor responses in hematopoietic cells. Ablation of TSLPR Signaling Alters the Inammatory Microenvironment in Notch-Decient Skin The impaired antitumor immune response elicited in N1N2K5 TSLPR mice could result from either a decreased dermal inammatory response and/or an altered inammatory prole. To discriminate between these possibilities, we analyzed the inammatory cells present in the dermis of both N1N2K5 and N1N2K5 TSLPR mice by ow cytometry.

Cancer Cell
TSLP Promotes Antitumor Inammation in the Skin

Figure 1. Loss of TSLPR Signaling in Notch-Decient Epidermis Leads to Tumor Formation


(A) Representative images of N1N2 TSLPR control, N1N2K5, and N1N2K5 TSLPR mice 50 days after neonatal deletion of Notch. (B and C) Representative hematoxylin and eosin (H&E) (B) and Ki67 (C) staining on dorsal skin sections of mice shown in (A). Photomicrographs in (B) are shown at different magnications to demonstrate the hyperproliferative layers surrounding cysts (asterisks) in N1N2K5 TSLPR mice compared to N1N2K5 animals. Dotted lines mark lumen/cell border. Scale bars: 100 mm. (D) Survival curve of N1N2 TSLPR (n = 15) and N1N2K5 TSLPR (n = 27) mice. See also Figure S1.

We noted that during the earlier stages subsequent to Notch inactivation, the inammatory response in N1N2K5 TSLPR mice was reduced compared to N1N2K5 mutants (data not shown). However, once the macroscopic phenotype was highly overt at 45 weeks after Notch ablation, the proportion of dermal CD45+ cells in both N1N2K5 and N1N2K5 TSLPR mice was on average 5-fold higher compared to normal littermate controls (Figure 2C). Thus, the inammatory response in N1N2K5 and N1N2K5 TSLPR mice was quantitatively equivalent at the phenotypic endpoint. Detailed phenotypic analysis of the dermal inammatory inltrates revealed several changes between N1N2K5 and N1N2K5 TSLPR mice, including alterations in the proportions of CD117+ mast cells and B220+ B cells (Figure 2D). However, the most striking differences were apparent in the T and myeloid cell compartments. In N1N2K5 animals, the proportion of CD4+ and CD8+ T cells was 12.7% and 10.4%, respectively, whereas in N1N2K5 TSLPR mice, the proportions were reduced on average to 1.8% and 1.1% (Figure 2D). In addition, CD11b+Gr1+ myeloid cells constituted just 7.2% of dermal CD45+ cells in N1N2K5 mice, whereas this population increased dramatically to 36.7% upon ablation of TSLPR (Figure 2D). Collectively, these data demonstrate that loss of TSLPR signaling in Notch-decient epidermis results in reduced proportions of CD4+ and CD8+ T cells and increased inltration of CD11b+Gr1+ myeloid cells.

TSLPR Signaling in T Cells Mediates Protection against Tumor Formation The observation that specic populations of dermal inammatory cells are altered quantitatively upon ablation of TSLPR signaling suggests that the cells responsible for mediating tumor protection in response to TSLP are likely to be within these compartments. Therefore, to functionally assess the role that specic cell types play in TSLP-mediated tumor protection, lethally irradiated N1N2K5 mice were reconstituted with Tslpr+/+ BM derived from different gene targeted mice lacking specic hematopoietic cell types (Figure 2E). In this system, the development of tumors after Notch inactivation is a direct consequence of loss of a specic hematopoietic population. We initially determined the functional contribution of lymphoid populations by reconstituting N1N2K5 mice with Rag2/;gc/ and Rag2/ BM. In both cases, more than 90% of reconstituted N1N2K5 mice developed tumors, indicating that the protective cell type resided within the T or B cell compartment, not the NK cell population (Figures 2F and S2A). Consistent with this, 78% of the mice in which NK cells were depleted by injection of PK1.36 monoclonal antibody remained tumor free, and those mice that did develop tumors presented only mild lesions (Figure S2B). To determine if B cells were required for tumor protection, we reconstituted N1N2K5 mice with Jh/ BM. In this experiment, 83% of mice were resistant to tumor formation, indicating that B cells do not mediate TSLPR-dependent tumor protection and therefore suggesting that the critical cell type resides within the T cell compartment (Figures 2F and S2A). In support of this, antibody-mediated depletion of CD4+ and CD8+ T cells resulted in tumor development in 83% of N1N2K5 mice (Figure S2C). To establish the role of T cells conclusively, we reconstituted N1N2K5 mice with Cd4/ and Cd8a/ BM. In this setting, the majority of mice reconstituted with Cd4/ BM were resistant to tumor formation (73%), whereas 74% of Cd8a/ / N1N2K5 chimeras developed severe tumors (Figures 2F and S2A). Taken together, these data demonstrate
Cancer Cell 22, 479493, October 16, 2012 2012 Elsevier Inc. 481

Cancer Cell
TSLP Promotes Antitumor Inammation in the Skin

Figure 2. CD8+ T Cells Are Required for TSLPR-Mediated Tumor Suppression


(A) Schematic depiction of experimental layouts. Lethally irradiated CD45.2 N1N2K5 (n = 14) and N1N2K5 TSLPR mice (n = 10) were transplanted with CD45.1 Tslpr/ BM and CD45.1 Tslpr+/+ BM, respectively, prior to ablation of Notch. (B) Bar diagram summarizing the results of the experiments described in (A). Data represent the results of three independent experiments. N1N2 (n = 15) and N1N2 TSLPR (n = 12) littermate controls were included. (C) Bar diagram showing the proportion of CD45+ cells in the dermis of N1N2 TSLPR control, N1N2K5, and N1N2K5 TSLPR mice 46 weeks after deletion of Notch. (D) Bar diagram showing the proportions of CD45+ subpopulations present in the dermis of N1N2K5 and N1N2K5 TSLPR mice 46 weeks after deletion of Notch and development of overt skin phenotypes. (C) and (D) Dermal cell inltrates were quantied by ow cytometry; data represent the mean of ve independent experiments (mean SD; *p < 0.01; **p < 0.001). (E) Schematic depiction of experimental layout. N1N2K5 and littermate mice were transplanted with BM from Rag2/;gc/ (n = 12/9), Rag2/ (n = 12/11), Jh/ (n = 12/7), Cd4/ (n = 16/11), or Cd8a/ (n = 15/10) donors. (F) Bar diagram summarizing the combined results of six independent BM transfer experiments as depicted in (E). (G) PCR analysis of Db1-Jb1 TCR rearrangement performed on genomic DNA from CD8+ T cells isolated from lymph nodes (LN) or dermis (Der). Tail and thymus were used as germline and rearranged controls, respectively. Bands corresponding to rearrangements containing Jb1.1-Jb1.5 are indicated. Additional bands are nonspecic. See also Figure S2.

482 Cancer Cell 22, 479493, October 16, 2012 2012 Elsevier Inc.

Cancer Cell
TSLP Promotes Antitumor Inammation in the Skin

that CD8+ T cells are required to mediate the tumor-protective response elicited by TSLPR signaling. Consistent with this, analysis of TCR rearrangement in dermal CD8+ T cells isolated from N1N2K5 mice revealed the predominance of Db1-Jb1.4 rearrangements, suggesting that the cytotoxic T cell response is antigen specic (Figure 2G). Previous studies have demonstrated that TSLP acts directly on CD4+ (Al-Shami et al., 2005; Rochman et al., 2007) and CD8+ (Rochman and Leonard, 2008) T cells during allergic immune responses. Thus, to investigate whether CD8+ T cells alone or T cells in general are sufcient for tumor protection, adoptive transfer experiments were performed in which N1N2K5 TSLPR mice were injected with different combinations of WT CD4+ and/or CD8+ T cells (Figure 3A; Figure S3A). When CD8+ or CD4+ T cells were transferred alone, the proportion of recipients exhibiting tumor resistance was 65% and 90%, respectively (Figure 3B). Chimeric mice that did develop tumors upon transfer of CD4+ or CD8+ T cells showed reduced tumor numbers and severity compared to control mice receiving either PBS, CD11c+ dendritic cells, or B220+ B cells (Figures 3B and S3B; data not shown). The level of protection conferred by CD4+ T cells was surprising in light of the BM reconstitution experiments described above, which indicated that CD4+ T cells are dispensable for tumor protection. However, ow cytometric analysis indicated that adoptively transferred CD4+ T cells induced the recruitment of host-derived CD8+ T cells, thus restoring the protective cytotoxic response (Figure 3C). The highest level of protection was observed when CD4+ and CD8+ T cells were administered together, as recipients in these cases exhibited complete protection from tumor development (Figure 3B). Importantly, tumor protection did not occur as a consequence of inefcient gene inactivation or counterselection of Notch-decient epidermal cells (Figures S3C and S3D). The nding that the adoptive transfer of WT T cells can prevent tumor formation in N1N2K5 TSLPR mice indicates that T cells respond directly to TSLP and do not act in response to another TSLP responsive cell type. Consistent with this, phosphoow cytometric analysis of pSTAT5 expression, which acts downstream of TSLPR ligation, revealed that only dermal CD4+ and CD8+ T cells from N1N2K5 mice are pSTAT5+ (Figures 3D and S3E). Interestingly, TSLPR signaling does not appear to be required for T cell migration in vitro (Figure S3F), and thus it seems likely that its tumor-protective effect is mediated via increased T cell survival/proliferation and/or activation. Finally, to determine if TSLP-responsive T cells could elicit protection against preformed tumors, we adoptively transferred CD4+ and CD8+ T cells into N1N2K5 TSLPR mice 2 weeks post-gene-inactivation and assayed tumor development (Figure 3E). Strikingly, compared to control animals receiving either CD11c+ dendritic cells or B220+ B cells, mice receiving both CD4+ and CD8+ T cells exhibited complete tumor protection (Figure 3F). Thus, TSLP-responsive T cells have the capacity to control preformed malignancies, at least during the early stages of tumor progression. CD11b+Gr1+ Myeloid Cells Promote Tumor Development As described above, a signicant inammatory component in N1N2K5 TSLPR dermis is the CD11b+Gr1+ myeloid population,

which correlates with tumor progression. To assess if CD11b+ Gr1+ myeloid cells promote tumor growth in N1N2K5 TSLPR mice, we depleted Gr1+ cells by intraperitoneal injection of the RB6.8C5 monoclonal antibody (Figure 4A). Depletion of Gr1+ cells was conrmed by ow cytometric analysis of both peripheral blood and dermal inltrates (Figure 4B and data not shown), and tumor progression was assessed by histology. As expected, mice receiving isotype control antibody developed large cystic tumors lined by hyperproliferative epithelial cells (Figures 4C and 4D). In contrast, mice in which Gr1+ cells were depleted developed smaller cysts that were predominantly lined by epithelial cells 12 cell layers thick, and thus resembled the phenotype observed in N1N2K5 mice (Figures 4C and 4D). The above data indicated that CD11b+Gr1+ myeloid cells perform a functional role in promoting tumor development. This observation may reect myeloid-derived suppressor cell activity, in which antitumor T cell responses are suppressed, or may be due to the promotion of tumor growth in a more direct manner. To distinguish between these possibilities, we isolated CD11b+Gr1+ myeloid cells from the tumor-associated microenvironment of N1N2K5 TSLPR mice and assessed their ability to suppress T cell activation in vitro. We could not detect any effect of CD11b+Gr1+ myeloid cells on the proliferation of either CD4+ or CD8+ T cells (Figure 4E). Accordingly, depletion of CD11b+Gr1+ cells did not result in an increase of dermal T cells in N1N2K5 TSLPR mice (Table S1). These data therefore suggest that the CD11b+Gr1+ population does not perturb T cell responses but rather promotes tumor growth directly. Tumor Formation in the Absence of TSLPR Signaling Is b-Catenin Dependent The macroscopic phenotype presented in N1N2K5 TSLPR mice is reminiscent of extracolonic manifestations presented in patients with familial adenomatous polyposis (FAP), in which loss-of-function mutations in the APC gene result in constitutive activation of the Wnt/b-catenin pathway (Jagelman, 1991). Elevated b-catenin-mediated signaling is also implicated in a variety of skin cancers (Gat et al., 1998; Malanchi et al., 2008). The link between Wnt/b-catenin signaling and cancer therefore prompted us to investigate the status of b-catenin in N1N2K5 TSLPR mice. Immunohistochemistry (IHC) for active b-catenin was performed on skin sections from N1N2K5, N1N2K5 TSLPR, and N1N2 TSLPR control mice. In N1N2K5 TSLPR mutants, active b-catenin was present in the majority of hyperproliferative cysts, in contrast to N1N2K5 mice, where it was largely absent (Figure 5A). These data suggest a role for b-catenin in the development and/or maintenance of skin tumors in these animals. To investigate whether b-catenin is functionally required for tumor development, we generated Notch1lox/lox;Notch2lox/lox; Ctnnb1lox/lox;K5CreERT Tslpr/ mice (hereafter N1N2bcatK5 TSLPR), in which Notch1, Notch2, and the b-catenin gene (Ctnnb1) can simultaneously be deleted in the skin (Figures S4A and S4B). Strikingly, ablation of b-catenin prevented the formation of cystic tumors in N1N2bcatK5 TSLPR mice, establishing that tumor development in these animals is b-catenin dependent (Figure 5B). Flow cytometric analysis of dermal inltrates in N1N2bcatK5 TSLPR mice revealed increased inltration of CD45+ cells compared to littermate control animals,
Cancer Cell 22, 479493, October 16, 2012 2012 Elsevier Inc. 483

Cancer Cell
TSLP Promotes Antitumor Inammation in the Skin

Figure 3. T Cells Are Sufcient to Mediate TSLP-Dependent Tumor Protection


(A) Schematic depiction of experimental layout. N1N2K5 TSLPR mice were transplanted with 107 CD45.1+ immune cells in the following combination: CD11c+ cells (n = 3), B220+ B cells (n = 8), CD4+ T cells (n = 14), CD8+ T cells (n = 16), and CD4+ and CD8+ T cells (n = 8). One N1N2K5 TSLPR mouse injected with PBS was included as a positive control in each experiment (n = 5). (B) Bar diagram showing the results of the experiment described in (A). Data represent the results of ve independent experiments. (C) Representative ow cytometric analysis of dermal T cells in a N1N2K5 TSLPR mouse adoptively transferred with WT CD45.1 CD4+ T cells. (Left) Shows proportions of CD4+ and CD8+ T cells in dermal inammatory inltrates. (Middle and Right) Shows contribution to CD4+ and CD8+ T cells by donor-derived (CD45.1) cells respectively. (D) Flow cytometric analysis of pSTAT5 expression in CD4+ and CD8+ T cells in the dermis of N1N2K5 and N1N2K5 TSLPR mice. Analysis was performed 5 weeks after Notch inactivation, when animals displayed an overt skin phenotype. (E) Schematic depiction of experimental layout. Notch signaling was ablated in N1N2K5 TSLPR mice. Two weeks post-gene-inactivation, animals were transplanted with 107 CD11c+ cells (n = 2), B220+ B cells (n = 3), or CD4+ and CD8+ T cells (n = 5) from CD45.1 donors. (F) Bar diagram showing the combined results of two independent experiments (described in E). See also Figure S3.

484 Cancer Cell 22, 479493, October 16, 2012 2012 Elsevier Inc.

Cancer Cell
TSLP Promotes Antitumor Inammation in the Skin

Figure 4. CD11b+Gr1+ Myeloid Cells Promote Tumor Development in N1N2K5 TSLPR Mice
(A) Schematic depiction of experimental layout. N1N2K5 TSLPR mice (n = 11) and N1N2 TSLPR littermate controls (n = 8) were injected with aGr1 antibody (RB6.8C5) to deplete Gr1+ myeloid cells. Administration of aGr1 was initiated 2 days prior to deletion of Notch (induced by ve consecutive tamoxifen injections starting at P30) and subsequently every 5 days after the nal tamoxifen injection. IgG2b isotype control antibody was also administered to N1N2K5 TSLPR mice (n = 4) and N1N2 TSLPR littermates (n = 2; data not shown). Analysis was performed 7 weeks after Notch inactivation. (B) Representative ow cytometric analysis of CD11b+Gr1+ myeloid cells in the dermis of N1N2K5 TSLPR mice injected with IgG2b or aGr1 antibody. (C) Representative H&E staining performed on dorsal skin sections of N1N2K5 TSLPR mice receiving IgG2b isotype control antibody (top) or aGr1 (bottom). Scale bars: 400 mm. (D) Bar diagram summarizing the results of the experiment described in (A). Data represent the results of three independent experiments. (E) Bar diagram showing the results of in vitro T cell suppression assay. Immunosuppressive activity of myeloid cells was assayed by coculturing splenic C57BL/6 T cells with CD45+CD11b+Gr1+ cells isolated from tumor-bearing N1N2K5 TSLPR skin in the ratios indicated. Myeloid cell numbers were kept constant (3.6 3 104 cells/well), and T cell numbers were reduced appropriately to adjust the T cell:myeloid cell ratio. BrdU incorporation was determined by measuring absorbance after 72 hr of coculture. Data represents the results of three independent experiments (mean SD). See also Table S1.

Cancer Cell 22, 479493, October 16, 2012 2012 Elsevier Inc. 485

Cancer Cell
TSLP Promotes Antitumor Inammation in the Skin

Figure 5. Tumor Formation in N1N2K5 TSLPR Mice Is b-Catenin Dependent


(A) Representative IHC staining for active b-catenin on skin sections from N1N2K5 TSLPR and N1N2K5 mice. Genotypes are indicated beside each image. (Left) Shows low magnication image of N1N2K5 TSLPR skin. (Right) Show high magnication of epithelial cells lining degenerative cysts. Asterisks demarcate single cysts; dotted lines mark lumen/cell border. (B) Representative photographs and H&E staining of N1N2bcatK5 TSLPR mice and littermate controls 65 days after gene inactivation. In all experimental settings gene inactivation was performed by ve consecutive daily tamoxifen injections starting at day P6. (C and D) Representative ow cytometric analysis of dermal CD4+ and CD8+ T cells (C) and CD11b+Gr1+ myeloid cells (D) in N1N2bcat TSLPR, N1N2K5, N1N2bcatK5 TSLPR, and N1N2K5 TSLPR mice. Scale bars: 200 mm. See also Figure S4.

conrming that in this context epidermal deletion of Notch continued to elicit a chronic inammatory response (Figure S4C). Consistent with the absence of TSLPR, CD4+ and CD8+ T cells represented less than 3% of dermal CD45+ cells (Figure 5C). Importantly, serum levels of TSLP remained elevated despite the loss of b-catenin (Figure S4D). Interestingly, the accumulation of dermal CD11b+Gr1+ myeloid cells was impaired in N1N2bcatK5 TSLPR mice (Figure 5D),
486 Cancer Cell 22, 479493, October 16, 2012 2012 Elsevier Inc.

indicating that the recruitment and/or development of this population is independent of TSLP but dependent on epithelial-specic Wnt/b-catenin signaling. CD11b+Gr1+ Myeloid Cells Promote Tumor Development by Provision of Wnt Ligands Having established that b-catenin signaling is required for cutaneous tumor development in N1N2K5 TSLPR mice, we next

Cancer Cell
TSLP Promotes Antitumor Inammation in the Skin

addressed how increased Wnt signaling is mediated. Initially, we determined if Wnt ligand expression was upregulated in N1N2K5 TSLPR mice by performing real-time PCR on whole skin. This analysis revealed that compared to normal skin isolated from N1N2 TSLPR mice, skin from N1N2K5 TSLPR mutants displayed increased expression of a variety of Wnt ligands, including Wnt3a, Wnt4, Wnt10a, and Wnt10b (Figure 6A). We then addressed the cellular source of Wnts by performing realtime PCR on uorescence-activated cell sorting (FACS)-puried cell populations isolated from tumor-bearing N1N2K5 TSLPR dermis. For this analysis, we focused our attention on CD11b+ Gr1+ myeloid cells, as these cells are required for tumor development in N1N2K5 TSLPR mice and accumulate in the tumor stroma. Wnt ligand expression was also assessed in tumorassociated CD11b+Gr1- cells and broblasts isolated from N1N2K5 TSLPR mice, as well as normal dermal broblasts isolated from N1N2 TSLPR controls. Interestingly, all cell types isolated from N1N2K5 TSLPR tumors exhibited increased expression of Wnt ligands compared to normal dermal broblasts (data not shown). However, comparison of the different tumor-associated populations demonstrated that the highest levels of Wnt ligand expression were detected in CD11b+Gr1+ myeloid cells, particularly with respect to Wnt3a, which exhibited a 40-fold increase in expression compared to tumor-associated broblasts (Figure 6B). To conrm that CD11b+Gr1+ myeloid cells produced WNT3A protein, we performed immunouorescent (IF) staining on skin sections isolated from N1N2K5 TSLPR and N1N2K5 mice. Consistent with the gene expression analysis, WNT3A colocalized with Gr1+ myeloid cells, which were frequently located in close proximity to epithelial tumor cells (Figure 6C). In N1N2K5 dermis, Gr1+ cells were only rarely detected, although those that were present were also WNT3A+ (Figure 6C and data not shown). We also analyzed WNT3a expression by western blot in tumor-bearing and non-tumor-bearing regions of N1N2K5 TSLPR skin and correlated this data with the proportions of CD11b+Gr1+ myeloid cells present. Consistent with our previous observations, the highest levels of WNT3a expression were detected in tumor-bearing regions of skin containing high proportions of CD11b+Gr1+ myeloid cells (Figure 6D; Figures S5AS5C). We also observed WNT3A expression in cystic epithelial cells in both N1N2K5 and N1N2K5 TSLPR skin (Figure 6C; Figure S5B), indicating that CD11b+Gr1+ myeloid cells are not the exclusive source of this protein. These data suggest that the elevated Wnt/b-catenin signaling observed in N1N2K5 TSLPR skin tumors is at least in part mediated by CD11b+Gr1+ myeloid cells. To explore this possibility further, we performed a coculture assay using a Tcf-luciferase reporter cell line (Tcf-Luc) in conjunction with myeloid cells and/or broblasts isolated from tumor-bearing N1N2K5 TSLPR skin. Coculture with CD11b+Gr1+ cells or broblasts alone did not induce luciferase activity in the Tcf-Luc cells. However, when used in combination, CD11b+Gr1+ cells and tumor broblasts induced a signicant 2.5-fold increase in luciferase activity compared to Tcf-Luc cells cultured alone (Figure 6E), indicating that these cells have the functional capacity to induce Wnt/b-catenin signaling in neighboring cells. The requirement of tumor-associated broblasts likely reects the expression of potentiating factors by these cells, as we observed markedly increased expression of the Wnt-potentiator Periostin (POSTN)

(Malanchi et al., 2012) in the stroma of N1N2K5 TSLPR dermis (Figure S5D). Collectively, these data indicate that the accumulation of CD11b+Gr1+ myeloid cells in N1N2K5 TSLPR mice results in promotion of tumor growth by increased provision of Wnt ligands. In support of this conclusion, depletion of CD11b+Gr1+ cells from the dermis of N1N2K5 TSLPR mice using the RB6.8C5 antibody results in reduced nuclear b-catenin staining in epidermal cysts (Figure S5E). TSLPR Signaling Mediates Tumor Protection in Notch-Independent Skin Cancers We next sought to determine if TSLP signaling could mediate tumor protection in other models of skin cancer in which Notch expression/signaling is maintained in the epidermis. To this end, we utilized Ctnnb1lox(ex3);K5CreERT mice (hereafter bcatD3K5), in which constitutively active b-catenin can be induced in the epidermis by conditionally deleting exon3 of the b-catenin gene (Harada et al., 1999). In this model, activation of b-catenin results in the formation of hair-follicle-derived tumors over a period of several months (Gat et al., 1998). We subsequently generated bcatD3K5 mice lacking TSLPR (bcatD3K5 TSLPR) and induced expression of the dominant active mutant of b-catenin. In agreement with previous reports, bcatD3K5 mice developed mild skin phenotypes at 75 days post-gene-induction (Gat et al., 1998), presenting as cystic hair-follicle-derived tumors (Figure 7A). In contrast, bcatD3K5 TSLPR mutants developed overtly more severe phenotypes within 25 days of b-catenin induction, exhibiting extensive hair loss and thickening of the skin (Figure 7A). When analyzed histologically, cystic tumors were prevalent and were signicantly larger than those observed in bcatD3K5 mice (Figures 7A and 7B), which at this time point had not developed a severe skin phenotype. The difference in phenotypic severity between the TSLPR-decient and competent mice was reected in survival rates, with 80% of bcatD3K5 TSLPR animals succumbing within 30 days of b-catenin induction. By contrast, all bcatD3K5 mice survived beyond 60 days (Figure S6A). Importantly, TSLP serum levels were increased in both bcatD3K5 and bcatD3K5 TSLPR animals compared to littermate controls (Figure 7C). Moreover, analysis of dermal inammatory cells revealed that CD4+ and CD8+ T cells were signicantly reduced in Tslpr/ compared to Tslpr+/+ control mutants (Figure 7D), supporting the hypothesis that TSLP mediates protection from cutaneous tumors by promoting T-cell-mediated immunity. Interestingly, we also observed accumulating CD11b+Gr1+ populations in the cutaneous tumors arising in bcatD3K5 TSLPR mice (Figure 7D). In this system b-catenin signaling is activated in the epithelium autonomously and thus the effects of Wntdriven tumorigenesis should be independent of stromal derived ligands. To conrm this, we depleted CD11b+Gr1+ cells in bcatD3K5 TSLPR mice by intraperitoneal injection of RB6.8C5 and assessed tumor development upon gene activation. Consistent with our hypothesis, tumor development in these animals was not adversely affected by the depletion of Gr1+ myeloid cells (Figures S6BS6D). Taken together, these results suggest that TSLPRmediated inammation protects against different forms of cutaneous tumors. In support of this conclusion, the accompanying
Cancer Cell 22, 479493, October 16, 2012 2012 Elsevier Inc. 487

Cancer Cell
TSLP Promotes Antitumor Inammation in the Skin

Figure 6. Tumor-Associated CD11b+Gr1+ Myeloid Cells Promote Tumor Development by Secreting Wnt Ligands
(A and B) Real-time PCR analysis for Wnt ligand expression performed on RNA isolated from whole-tumor-bearing N1N2K5 TSLPR skin (A) or specic cell populations isolated from the tumor microenvironment (B). In (A) gene expression is expressed relative to whole skin isolated from N1N2 TSLPR controls. In (B) gene expression is expressed relative to tumor-associated broblasts. Data represent the mean SD of three independent experiments. (C) IF staining for Gr1, CD45, and WNT3a expression in the dermis of N1N2K5 and N1N2K5 TSLPR mice. Asterisks demarcate cyst lumen; dotted lines mark the border between the cystic epithelium and the surrounding dermis. The image shown for N1N2K5 skin displays a region with extensive CD45 staining to demonstrate the low proportions of Gr1+ cells. Scale bars: 100 mm. (D) Western blot analysis of WNT3a expression in whole-skin extracts from nontumor and tumor-bearing regions of N1N2K5 TSLPR mice. Tubulin was used as loading control; lanes shown are cropped from the same blot. The original blot is shown in Figure S5C. (E) CD11b+Gr1+ and CD11b+Gr1- myeloid cells were isolated from tumor-bearing skin of N1N2K5 TSLPR mice and cocultured with 293T cells expressing a Tcf-luciferase reporter (Tcf-Luc), with or without broblasts isolated from N1N2K5 TSLPR tumor-bearing skin (Tum Fibr) or WT dermis (WT Fibr). Luciferase activity was measured 36 hr after coculture (mean SD). See also Figure S5.

488 Cancer Cell 22, 479493, October 16, 2012 2012 Elsevier Inc.

Cancer Cell
TSLP Promotes Antitumor Inammation in the Skin

Figure 7. TSLP Mediates Protection against b-Catenin-Dependent Skin Carcinogenesis


(A) Representative images of bcatD3K5 (n = 5) and bcatD3K5 TSLPR (n = 10) mice after induction of activated b-catenin by ve consecutive daily injections of tamoxifen starting at day P30. bcatD3 TSLPR mice were used as normal littermate controls (n = 8). A bcatD3K5 mouse is shown 75 days posttamoxifen to illustrate accelerated tumor development in the absence of TSLPR signaling. H&E staining performed on dorsal skin sections from each mouse are shown below each image. Arrows point to regions with overt skin phenotype. Scale bars: 100 mm. (B) Bar diagram displaying the proportion of mice with the indicated skin phenotype 25 days after induction of activated B-catenin. Data represent the results of three independent experiments. (C) TSLP serum levels were measured in bcatD3K5 TSLPR (n = 5), bcatD3 TSLPR (n = 3), bcatD3K5 (n = 4), and N1N2K5 (n = 3) mice 3 weeks after b-catenin activation or Notch inactivation at day P30. The bar diagram is the result of a single experiment run in triplicate (mean SD; *p < 0.05; **p < 0.001). (D) Flow cytometric analysis of T cells and myeloid cells in the dermis of bcatD3 TSLPR control, bcatD3K5 TSLPR, and bcatD3K5 mice 30 days after induction of activated bcatenin. All data are gated on CD45+ cells. See also Figure S6.

Cancer Cell 22, 479493, October 16, 2012 2012 Elsevier Inc. 489

Cancer Cell
TSLP Promotes Antitumor Inammation in the Skin

manuscript by Demehri et al. (2012) in this issue of Cancer Cell also demonstrates that TSLP-dependent immune responses can protect against DMBA/TPA-induced tumors in a wild-type background. Importantly, this model is independent of any pre-existing genetic aberrations. DISCUSSION In this study, we have revealed a tumor protective role for the cytokine TSLP during cutaneous skin cancer development. This effect is primarily mediated by direct signaling on dermal CD4+ and CD8+ T cells via the TSLPR, which prevents the growth of b-catenin-dependent hair follicle tumors in Notch lossof-function and b-catenin gain-of-function mouse models. Loss of TSLPR responsive T cells results in tumor development and is accompanied by the induction of protumorigenic inammation, which augments Wnt/b-catenin signaling. These ndings reveal a tumor-protective role for TSLP and demonstrate how acquired and innate arms of the immune system can mediate pro- and antitumor effects. This in turn may have broader implications with respect to the role of inammation in other epithelial malignancies. TSLP and T-Cell-Mediated Tumor Protection A key nding in this study is that direct activation of the TSLPR on dermal TCRab+ T cells elicits antitumor immune responses to cutaneous malignancies. The analysis performed in Notch mutant mice indicates that both CD4+ and CD8+ T cells mediate this protective effect, although CD8+ T cells are the essential effector cells. That TSLP can mediate tumor protection by promoting CD8+ T cell responses is surprising, as TSLP is considered to promote Th2-mediated immunity and thus is not normally associated with cytotoxic responses (Liu et al., 2007). However, it has been shown that TSLP can promote the development of cytotoxic CD8+ T cells by inducing the differentiation of cytolytic effectors that express IL-5 and IL-13 (Soumelis and Liu, 2004). We also nd that expression levels of IL-5 and IL-13 are elevated in N1N2K5 dermis (Dumortier et al., 2010), suggesting that TSLP induces a similar cytotoxic state in this model. In addition, it appears likely that TSLP acts to maintain dermal T cells, as their proportions are signicantly reduced upon TSLPR ablation. This is unlikely to be a consequence of impaired homing as Transwell migration studies indicate that Tslpr/ T cells show normal migration (Figure S3F), consistent with a previous report (He et al., 2008). Activation of dermal T cells is also unaffected by loss of TSLPR (data not shown), suggesting that activation-associated proliferation is not perturbed. It is therefore likely that TSLP signaling promotes T cell survival by acting directly on CD4+ and CD8+ T lymphocytes, as demonstrated by earlier studies (Al-Shami et al., 2005; Rochman et al., 2007; Rochman and Leonard, 2008). Regardless of the precise mechanism, the data presented here indicate that the tumor-protective role of TSLP is primarily mediated by a direct effect on dermal T cells. In support of this conclusion, the accompanying manuscript by Demehri et al. (2012) also demonstrates that TSLP elicits antitumor T cell immunity upon ablation of cutaneous Notch signaling. Interestingly, in their model CD4+ T cells are sufcient for tumor protection and CD8+ T cells are dispensable. This is likely due to the use
490 Cancer Cell 22, 479493, October 16, 2012 2012 Elsevier Inc.

of the Msx2-Cre deletor strain, which inactivates the Notch pathway during development and thus primes the developing immune system from an early time point. In our system, Notch deletion is induced only after birth, meaning that any T cell activity directed against epithelial tumor cells is a naive immunological response. Induction of Protumorigenic Inammation upon Loss of TSLPR An important consequence of TSLPR ablation in the mouse models used in this study is the induction of protumorigenic inammation, an important functional component of which are CD11b+Gr1+ myeloid cells. Why these cells accumulate in the dermis of N1N2K5 TSLPR mice but only represent a relatively minor population of myeloid cells in N1N2K5 animals is unclear and needs further investigation. One possible explanation is that the allergic response induced by TSLP is antagonistic to protumorigenic components of the inammatory milieu. Such a mechanism has been proposed by a recent study indicating that histamine produced during allergic immune reactions prevents the accumulation of tumor-promoting immature CD11b+ Gr1+ myeloid cells (Yang et al., 2011). In addition, we cannot formally exclude that the role of TSLP-responsive T cells is to act directly against CD11b+Gr1+ myeloid cells. However, we think this is unlikely based on the tumor-protective effect of TSLP demonstrated in bcatD3K5 mice, where tumor development occurs independently of myeloid cells. An alternative explanation is that the CD11b+Gr1+ myeloid cells are recruited by putative cancer stem cells (CSCs) that secrete promyeloid factors. A recent study has demonstrated that CD34+ cutaneous CSCs can induce a protumorigenic stroma via paracrine effects (Beck et al., 2011). We also observe expanded populations of immature CD34+ epithelial cells in the tumors arising in N1N2K5 TSLPR mice (data not shown), suggesting similar mechanisms may be operating in the models used in this study. Finally, it is possible that loss of TSLP-mediated immune responses results in increased loads of pathogenic organisms that may induce protumorigenic inammation, as has been suggested by the link between specic infections and carcinogenesis (Parkin, 2006). Wnt-Dependent Tumor Promotion by CD11b+Gr1+ Myeloid Cells The tumor-promoting role of CD11b+Gr1+ myeloid cells has been reported previously by a variety of studies (Kowanetz et al., 2010; Qian et al., 2011; Rodriguez and Ochoa, 2008). Here, we demonstrate that these cells can exert this effect by augmenting Wnt/b-catenin signaling in neighboring epithelial cells via the secretion of Wnt ligands. The ability of CD11b+Gr1+ myeloid cells to induce Wnt/ b-catenin signaling is dependent on tumor-associated broblasts, which probably reects increased expression of the Wnt potentiator POSTN. This suggests that the dermal stroma in these mice is primed to facilitate the protumorigenic function of CD11b+Gr1+ myeloid cells. Because the recruitment of CD11b+Gr1+ cells is itself dependent on elevated b-catenin expression in the epidermis, it is likely that Wnt pathway activation is an early event during tumor

Cancer Cell
TSLP Promotes Antitumor Inammation in the Skin

progression in the models described here. Indeed, loss of Notch receptors can increase levels of nuclear b-catenin in a cell autonomous manner (Devgan et al., 2005; Kwon et al., 2011). Therefore, we suggest that once recruited to the tumor stroma, myeloid populations can augment the Wnt signaling cascade and drive tumor growth. In this way, the epithelial population that initiates cancer development establishes a feed-forward loop that favors tumor progression. Context-Dependent Role of TSLP during Cancer Development The central theme of this study is the tumor-protective effect of TSLP during cutaneous cancer development. This is in contrast to previous reports that nd TSLP can promote tumor growth and metastasis in mouse models of breast and pancreatic cancer (De Monte et al., 2011; Olkhanud et al., 2011; PedrozaGonzalez et al., 2011). The relevance of TSLP in this respect is indeed highlighted by the fact that pharmaceutical companies are seeking to develop drugs directed against TSLP (Edwards, 2008). However, within the human population, individuals who suffer from atopic allergic disorders have a reduced risk of developing certain types of cancer (Gandini et al., 2005; Prizment et al., 2007; Vajdic et al., 2009; Wang and Diepgen, 2005). It is tempting to speculate that the context in which TSLP functions as a tumor suppressor or tumor promoter is related to how allergic immune responses affect barrier versus glandular epithelial cells. It will be interesting to determine if TSLP-responsive T cells that induce allergic inammation in other barrier epithelia, such as lung (Al-Shami et al., 2005), are also tumor protective. Thus, whether TSLP is inducing a pro- or antitumorigenic immune response appears to be context- and tissue-dependent, and therefore blocking TSLP-mediated responses as a potential cancer treatment is a hypothesis that must be analyzed with caution.
EXPERIMENTAL PROCEDURES Ethics Statement All animal work was conducted in accordance with Swiss national guidelines. All mice were kept in the animal facility under EPFL animal care regulations. They were housed in individual cages at 23 C 1 C with a 12 hr light/dark cycle. All animals were supplied with food and water ad libitum. This study te rinaire Cantonal of Etat has been reviewed and approved by the Service Ve de Vaud. Mice Notch1lox/lox;Notch2lox/lox;Tslpr/; Notch1lox/lox;Notch2lox/lox;K5CreERT, ERT / mice were previously described (Dumortier et al., K5Cre , and Tslpr 2010). Notch1lox/lox;Notch2lox/lox;Tslpr/;K5CreERT mice were crossed to Ctnnb1lox/lox (Malanchi et al., 2008) mice to generate Notch1lox/lox; Notch2lox/lox; Ctnnb1lox/lox;Tslpr/;K5CreERT mice. Ctnnb1lox(ex3) mice (Harada et al., 1999) were crossed with Tslpr/ mice and K5CreERT transgenic mice (Indra et al., 1999) to generate Ctnnb1lox(ex3);Tslpr/;K5CreERT mice. C57BL/6 CD45.1 (B6.SJL-Ptprca/BoyAiTac) and Rag2/;gc/ (B6.Rag2tm1FwaIl2rgtmWjl) mice were purchased from Taconic Europe (Germany). Tslpr/ and Rag2/ mice were crossed to C57BL/6 CD45.1 mice to obtain Tslpr/ CD45.1 and Rag2/ CD45.1 mice, respectively. Jh/ (B6.129P2-Igh-Jtm1Cgn/J), Cd4/ (B6.129S2-Cd4tm1Mak/J), and Cd8a/ (B6.129S2-Cd8atm1Mak/J) mice were purchased from the Jackson Laboratory (Bar Harbor, ME, USA). All mice used as donors in the bone marrow chimera and adoptive transfer experiments were on a pure C57BL/6

background. All Notch mutants were backcrossed >10 generations with C57BL/6 mice and thus can be considered to be on a pure C57BL/6 background. Depending on the experiment, gene inactivation was achieved by intraperitoneal injection at P6, P30, or P60 of control and oxed mutant mice with 1 mg/20 g body weight of tamoxifen (Sigma-Aldrich, St. Gallen, Switzerland) for ve consecutive days. Dermal Cell Isolation Back skin was aseptically dissected from mice and incubated dermis-side down in a 1 mg/ml solution of collagenase/dispase (Roche, Indianapolis, IN, USA) for 1 hr at 37 C. The dermis was separated from the epidermis, dissociated mechanically, and incubated in a 2 mg/ml solution of collagenase (Sigma-Aldrich) for 45 min at 37 C with gentle agitation. Cell suspensions were ltered through a 70 mm cell strainer (BD Falcon, Franklin Lakes, NJ, USA) and washed twice in staining medium (13 HBSS/25 mM HEPES/2% newborn bovine calf serum). Flow Cytometry Single-cell suspensions were stained with monoclonal antibody conjugates in accordance with standard protocols and analyzed using a CyAn ow cytometer (Dako, Glostrup, Denmark). Data analysis and processing was performed using FlowJo software (Tree Star, Ashland, OR, USA). Immunohistochemistry and Immunouorescence For immunohistochemistry, tissues were xed in 4% PFA and embedded in parafn. After sectioning, 4 mm sections were rehydrated, blocked with 3% H2O2, and incubated in antigen retrieval buffer (10 mM Tris/1 mM EDTA/0.05% Tween 20 [pH 9.0]) at 95 C for 20 min. Sections were then stained using unconjugated primary antibodies and the appropriate HRPconjugated secondary antibodies. Staining was revealed by DAB (SigmaAldrich) revelation. For immunouorescence, tissues were embedded in OCT compound and frozen. Cryosections (8 mm) were then xed in acetone at 4 C for 2 min and air-dried for 20 min. After rinsing in PBS, sections were stained with unconjugated primary antibodies and the appropriate uorophore-conjugated secondary antibodies. Fluorescent images were acquired using an LSM700 confocal microscope (Zeiss, Oberkochen, Germany). See the Supplemental Experimental Procedures for further details. SUPPLEMENTAL INFORMATION Supplemental Information includes six gures, one table, and Supplemental Experimental Procedures and can be found with this article online at http:// dx.doi.org/10.1016/j.ccr.2012.08.016. ACKNOWLEDGMENTS This work was supported in part by the Swiss National Science Foundation, the Swiss Cancer League, the Marie Curie Foundation, EuroSyStem, and OptiStem. We thank Rolf Kemler for providing the conditional b-catenin mice, Makoto Taketo for the conditional b-cateninD3 mice, Pierre Chambon and Daniel Metzger for the K5CreERT mice, Ursula Zimber-Strobl and Lothar Strobl for the conditional Notch2 mice, and Warren Leonard for the Tslpr/ mice. We thank Joerg Huelsken for providing the Wnt-reporter cells and reagents, Nicola Harris and Ilaria Mosconi for providing the TSLP-neutralizing antibody, and Daniel Hohl for the TEWL measurement apparatus. We would like to acknowledge Fabio Aloisio and Alessandra Piersigilli for pathological ` Artacho, Jessica Sordetanalysis of the tumors, Olivier Randin, Jose ` le Ferrand for technical assistance, and Miguel Garcia, Dessimoz, and Gise Gonzalo Tapia, and Sintia Winkler for cell sorting. We thank Shadmehr Demehri and Raphael Kopan for discussion, sharing experimental data, and critical reading of the manuscript. Received: February 8, 2012 Revised: June 8, 2012 Accepted: August 17, 2012 Published: October 15, 2012

Cancer Cell 22, 479493, October 16, 2012 2012 Elsevier Inc. 491

Cancer Cell
TSLP Promotes Antitumor Inammation in the Skin

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. Al-Shami, A., Spolski, R., Kelly, J., Keane-Myers, A., and Leonard, W.J. (2005). A role for TSLP in the development of inammation in an asthma model. J. Exp. Med. 202, 829839. Allavena, P., Sica, A., Garlanda, C., and Mantovani, A. (2008). The Yin-Yang of tumor-associated macrophages in neoplastic progression and immune surveillance. Immunol. Rev. 222, 155161. Aspord, C., Pedroza-Gonzalez, A., Gallegos, M., Tindle, S., Burton, E.C., Su, D., Marches, F., Banchereau, J., and Palucka, A.K. (2007). Breast cancer instructs dendritic cells to prime interleukin 13-secreting CD4+ T cells that facilitate tumor development. J. Exp. Med. 204, 10371047. Beck, B., Driessens, G., Goossens, S., Youssef, K.K., Kuchnio, A., Caauwe, A., Sotiropoulou, P.A., Loges, S., Lapouge, G., Candi, A., et al. (2011). A vascular niche and a VEGF-Nrp1 loop regulate the initiation and stemness of skin tumours. Nature 478, 399403. Coussens, L.M., and Werb, Z. (2002). Inammation and cancer. Nature 420, 860867. De Monte, L., Reni, M., Tassi, E., Clavenna, D., Papa, I., Recalde, H., Braga, M., Di Carlo, V., Doglioni, C., and Protti, M.P. (2011). Intratumor T helper type 2 cell inltrate correlates with cancer-associated broblast thymic stromal lymphopoietin production and reduced survival in pancreatic cancer. J. Exp. Med. 208, 469478. Demehri, S., Liu, Z., Lee, J., Lin, M.H., Crosby, S.D., Roberts, C.J., Grigsby, P.W., Miner, J.H., Farr, A.G., and Kopan, R. (2008). Notch-decient skin induces a lethal systemic B-lymphoproliferative disorder by secreting TSLP, a sentinel for epidermal integrity. PLoS Biol. 6, e123. Demehri, S., Turkoz, A., and Kopan, R. (2009). Epidermal Notch1 loss promotes skin tumorigenesis by impacting the stromal microenvironment. Cancer Cell 16, 5566. Demehri, S., Turkoz, A., Manivasagam, S., Yockey, L.J., Turkoz, M., and Kopan, R. (2012). Elevated epidermal thymic stromal lymphopoietin levels prevent skin tumorigenesis. Cancer Cell 22, this issue, 494505. Devgan, V., Mammucari, C., Millar, S.E., Brisken, C., and Dotto, G.P. (2005). p21WAF1/Cip1 is a negative transcriptional regulator of Wnt4 expression downstream of Notch1 activation. Genes Dev. 19, 14851495. Dumortier, A., Durham, A.D., Di Piazza, M., Vauclair, S., Koch, U., Ferrand, G., Ferrero, I., Demehri, S., Song, L.L., Farr, A.G., et al. (2010). Atopic dermatitislike disease and associated lethal myeloproliferative disorder arise from loss of Notch signaling in the murine skin. PLoS ONE 5, e9258. Edwards, M.J. (2008). Therapy directed against thymic stromal lymphopoietin. Drug News Perspect. 21, 312316. Fridlender, Z.G., Sun, J., Kim, S., Kapoor, V., Cheng, G., Ling, L., Worthen, G.S., and Albelda, S.M. (2009). Polarization of tumor-associated neutrophil phenotype by TGF-beta: N1 versus N2 TAN. Cancer Cell 16, 183194. Gabrilovich, D.I., and Nagaraj, S. (2009). Myeloid-derived suppressor cells as regulators of the immune system. Nat. Rev. Immunol. 9, 162174. Gandini, S., Lowenfels, A.B., Jaffee, E.M., Armstrong, T.D., and Maisonneuve, P. (2005). Allergies and the risk of pancreatic cancer: a meta-analysis with review of epidemiology and biological mechanisms. Cancer Epidemiol. Biomarkers Prev. 14, 19081916. Gat, U., DasGupta, R., Degenstein, L., and Fuchs, E. (1998). De Novo hair follicle morphogenesis and hair tumors in mice expressing a truncated betacatenin in skin. Cell 95, 605614. Hanahan, D., and Weinberg, R.A. (2011). Hallmarks of cancer: the next generation. Cell 144, 646674. Harada, N., Tamai, Y., Ishikawa, T., Sauer, B., Takaku, K., Oshima, M., and Taketo, M.M. (1999). Intestinal polyposis in mice with a dominant stable mutation of the beta-catenin gene. EMBO J. 18, 59315942.

He, R., Oyoshi, M.K., Garibyan, L., Kumar, L., Ziegler, S.F., and Geha, R.S. (2008). TSLP acts on inltrating effector T cells to drive allergic skin inammation. Proc. Natl. Acad. Sci. USA 105, 1187511880. Indra, A.K., Warot, X., Brocard, J., Bornert, J.M., Xiao, J.H., Chambon, P., and Metzger, D. (1999). Temporally-controlled site-specic mutagenesis in the basal layer of the epidermis: comparison of the recombinase activity of the tamoxifen-inducible Cre-ER(T) and Cre-ER(T2) recombinases. Nucleic Acids Res. 27, 43244327. Jagelman, D.G. (1991). Extra-colonic manifestations of familial adenomatous polyposis. Oncology (Williston Park) 5, 2327, discussion 316. Kowanetz, M., Wu, X., Lee, J., Tan, M., Hagenbeek, T., Qu, X., Yu, L., Ross, J., Korsisaari, N., Cao, T., et al. (2010). Granulocyte-colony stimulating factor promotes lung metastasis through mobilization of Ly6G+Ly6C+ granulocytes. Proc. Natl. Acad. Sci. USA 107, 2124821255. Kwon, C., Cheng, P., King, I.N., Andersen, P., Shenje, L., Nigam, V., and Srivastava, D. (2011). Notch post-translationally regulates b-catenin protein in stem and progenitor cells. Nat. Cell Biol. 13, 12441251. Liu, Y.J., Soumelis, V., Watanabe, N., Ito, T., Wang, Y.H., Malefyt, Rde.W., Omori, M., Zhou, B., and Ziegler, S.F. (2007). TSLP: an epithelial cell cytokine that regulates T cell differentiation by conditioning dendritic cell maturation. Annu. Rev. Immunol. 25, 193219. Malanchi, I., Peinado, H., Kassen, D., Hussenet, T., Metzger, D., Chambon, P., Huber, M., Hohl, D., Cano, A., Birchmeier, W., and Huelsken, J. (2008). Cutaneous cancer stem cell maintenance is dependent on beta-catenin signalling. Nature 452, 650653. 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. Nicolas, M., Wolfer, A., Raj, K., Kummer, J.A., Mill, P., van Noort, M., Hui, C.C., Clevers, H., Dotto, G.P., and Radtke, F. (2003). Notch1 functions as a tumor suppressor in mouse skin. Nat. Genet. 33, 416421. Olkhanud, P.B., Rochman, Y., Bodogai, M., Malchinkhuu, E., Wejksza, K., Xu, M., Gress, R.E., Hesdorffer, C., Leonard, W.J., and Biragyn, A. (2011). Thymic stromal lymphopoietin is a key mediator of breast cancer progression. J. Immunol. 186, 56565662. Parkin, D.M. (2006). The global health burden of infection-associated cancers in the year 2002. Int. J. Cancer 118, 30303044. Pedroza-Gonzalez, A., Xu, K., Wu, T.C., Aspord, C., Tindle, S., Marches, F., Gallegos, M., Burton, E.C., Savino, D., Hori, T., et al. (2011). Thymic stromal lymphopoietin fosters human breast tumor growth by promoting type 2 inammation. J. Exp. Med. 208, 479490. Prizment, A.E., Folsom, A.R., Cerhan, J.R., Flood, A., Ross, J.A., and Anderson, K.E. (2007). History of allergy and reduced incidence of colorectal cancer, Iowa Womens Health Study. Cancer Epidemiol. Biomarkers Prev. 16, 23572362. Proweller, A., Tu, L., Lepore, J.J., Cheng, L., Lu, M.M., Seykora, J., Millar, S.E., Pear, W.S., and Parmacek, M.S. (2006). Impaired notch signaling promotes de novo squamous cell carcinoma formation. Cancer Res. 66, 74387444. Qian, B.Z., Li, J., Zhang, H., Kitamura, T., Zhang, J., Campion, L.R., Kaiser, E.A., Snyder, L.A., and Pollard, J.W. (2011). CCL2 recruits inammatory monocytes to facilitate breast-tumour metastasis. Nature 475, 222225. Quezada, S.A., Peggs, K.S., Simpson, T.R., and Allison, J.P. (2011). Shifting the equilibrium in cancer immunoediting: from tumor tolerance to eradication. Immunol. Rev. 241, 104118. Rochman, I., Watanabe, N., Arima, K., Liu, Y.J., and Leonard, W.J. (2007). Cutting edge: direct action of thymic stromal lymphopoietin on activated human CD4+ T cells. J. Immunol. 178, 67206724. Rochman, Y., and Leonard, W.J. (2008). The role of thymic stromal lymphopoietin in CD8+ T cell homeostasis. J. Immunol. 181, 76997705. Rodriguez, P.C., and Ochoa, A.C. (2008). Arginine regulation by myeloid derived suppressor cells and tolerance in cancer: mechanisms and therapeutic perspectives. Immunol. Rev. 222, 180191.

492 Cancer Cell 22, 479493, October 16, 2012 2012 Elsevier Inc.

Cancer Cell
TSLP Promotes Antitumor Inammation in the Skin

oz-Descalzo, S., Balayo, T., Wirtz-Peitz, F., Hayward, P., Sanders, P.G., Mun and Arias, A.M. (2009). Ligand-independent trafc of Notch buffers activated Armadillo in Drosophila. PLoS Biol. 7, e1000169. Soumelis, V., and Liu, Y.J. (2004). Human thymic stromal lymphopoietin: a novel epithelial cell-derived cytokine and a potential key player in the induction of allergic inammation. Springer Semin. Immunopathol. 25, 325333. 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. nez-Maza, O., Becker, N., Vajdic, C.M., Falster, M.O., de Sanjose, S., Mart Bracci, P.M., Melbye, M., Smedby, K.E., Engels, E.A., Turner, J., et al. (2009). Atopic disease and risk of non-Hodgkin lymphoma: an InterLymph pooled analysis. Cancer Res. 69, 64826489.

Wang, H., and Diepgen, T.L. (2005). Is atopy a protective or a risk factor for cancer? A review of epidemiological studies. Allergy 60, 10981111. 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. (2011). Loss-offunction mutations in Notch receptors in cutaneous and lung squamous cell carcinoma. Proc. Natl. Acad. Sci. USA 108, 1776117766. Weng, A.P., and Aster, J.C. (2004). Multiple niches for Notch in cancer: context is everything. Curr. Opin. Genet. Dev. 14, 4854. Yang, X.D., Ai, W., Asfaha, S., Bhagat, G., Friedman, R.A., Jin, G., Park, H., Shykind, B., Diacovo, T.G., Falus, A., and Wang, T.C. (2011). Histamine deciency promotes inammation-associated carcinogenesis through reduced myeloid maturation and accumulation of CD11b+Ly6G+ immature myeloid cells. Nat. Med. 17, 8795. Ziegler, S.F. (2010). The role of thymic stromal lymphopoietin (TSLP) in allergic disorders. Curr. Opin. Immunol. 22, 795799.

Cancer Cell 22, 479493, October 16, 2012 2012 Elsevier Inc. 493

Article
Elevated Epidermal Thymic Stromal Lymphopoietin Levels Establish an Antitumor Environment in the Skin

Cancer Cell

Shadmehr Demehri,1,2,* Ahu Turkoz,1 Sindhu Manivasagam,1 Laura J. Yockey,1 Mustafa Turkoz,1 and Raphael Kopan1,*
1Department of Developmental Biology and Division of Dermatology, Washington University School of Medicine, Box 8103, 660 South Euclid Avenue, St. Louis, MO 63110-1095, USA 2Department of Internal Medicine, St. Lukes Hospital, 232 South Woods Mill Road, Chestereld, MO 63017, USA *Correspondence: demehris@wustl.edu (S.D.), kopan@wustl.edu (R.K.) http://dx.doi.org/10.1016/j.ccr.2012.08.017

SUMMARY

Thymic Stromal Lymphopoietin (TSLP), a cytokine implicated in induction of T helper 2 (Th2)-mediated allergic inammation, has recently been shown to stimulate solid tumor growth and metastasis. Conversely, studying mice with clonal loss of Notch signaling in their skin revealed that high levels of TSLP released by barrier-defective skin caused a severe inammation, resulting in gradual elimination of Notch-decient epidermal clones and resistance to skin tumorigenesis. We found CD4+ T cells to be both required and sufcient to mediate these effects of TSLP. Importantly, TSLP overexpression in wild-type skin also caused resistance to tumorigenesis, conrming that TSLP functions as a tumor suppressor in the skin.

INTRODUCTION The mammalian skin is a model organ used for decades in chemical carcinogenesis studies and contributed to the recognition that carcinogenesis is a stepwise process (Hanahan and Weinberg, 2000). As in other models of cancer, skin tumors arise as a consequence of intrinsic changes in the initiated cells that are amplied through interactions with the tumor microenvironment (Campisi, 2005; Kessenbrock et al., 2010; Sneddon and Werb, 2007). Although immune cells can suppress tumor growth in the tumor microenvironment, their carcinogenesis-promoting role is becoming increasingly appreciated (Coussens and Werb, 2002; Hanahan and Weinberg, 2000, 2011). Because skin is the largest barrier organ in the body, it is under tight surveillance by the immune system, and even subtle changes in its cellular differentiation program can alter the overall susceptibility to cancer by eliciting a persistent inammatory response (Quigley et al., 2009). This is partly due to a direct epidermal contribution to inammation via secretion of multiple cytokines, such as interleukin (IL)-1, IL-6, and TGF-b (Morasso and Tomic-Canic, 2005).

Studies in mice with clonal deletion of Notch pathway components uncovered a role for Notch in tumor promotion via induction of a noncell autonomous feed-forward loop between the epidermis, dermal broblasts, and the immune system (Demehri et al., 2008, 2009b). Notch is a transmembrane receptor that mediates short-range communication between adjacent cells (Kopan and Ilagan, 2009). Upon binding to the ligand presented by a neighboring cell, Notch undergoes proteolysis by g-secretase enzyme to release its intracellular domain. Subsequently, the Notch intracellular domain translocates into the nucleus and binds to its DNA-binding partner, RBPj, and regulates its downstream targets in a context-dependent manner (Kopan and Ilagan, 2009). Notch signaling plays multiple roles in skin development (Mascia et al., 2012), but in the context of carcinogenesis, the most relevant role is in promoting suprabasal differentiation (Blanpain et al., 2006; Demehri et al., 2009b; Nguyen et al., 2006; Nicolas et al., 2003; Pan et al., 2004; Rangarajan et al., 2001). Reduction in Notch signaling leads to aberrant epidermal differentiation and defective barrier formation, which creates a chronic wound-like environment

Signicance We demonstrate unequivocally that TSLP triggers a dominant antitumor response in a Th2-polarized inammatory microenvironment in the skin. Importantly, the antitumor microenvironment created by TSLP-inducers like low-calcemic vitamin D agonists (e.g., Calcipotriol) can prevent and eliminate skin tumors in wild-type mice. Although our ndings may reect a skin-specic effect, it is intriguing to postulate that TSLP plays a common tumor suppressor role during the early stages of solid tumor development. Considering the emergence of TSLP as a potential therapeutic target in treatment of solid cancers, this report points to an alternative utility for TSLP as an antitumor immune factor that can be utilized to optimally combat and ultimately prevent solid cancers.
494 Cancer Cell 22, 494505, October 16, 2012 2012 Elsevier Inc.

Cancer Cell
TSLP Overexpression Protects from Skin Cancer

Figure 1. Mice Lacking RBPj in Portions of Their Epidermis Are Resistant to Skin Tumorigenesis
(A) The calico pattern of EYFP expression (green) induced by Msx2-Cre-mediated gene deletion in a Msx2-Cretg/+; Rosa (LoxP_Stop_LoxP)-Eyfp (Msx2-Cre/+; RosaEYFP) newborn is shown. Image taken under tungsten illumination is shown in magenta. After birth, mutant clones become evident due to hair phenotypes. (B) Reduction in Notch signaling dosage in the skin correlates with shortening of life span and time to spontaneous tumor formation. This trend, however, does not extend to mice lacking Rbpj (green arrow), which live 100 days yet do not develop any skin tumors. n > 20 in each group; % indicates the percentage of mice that developed skin tumors; **p < 0.01, Students t test; error bars represent SD. This gure is modied from Demehri et al., 2009b. (C and D) Time-to-tumor onset (C; p < 0.0001, log rank test) and average tumor number (D) of RBPjCKO and wild-type littermates treated with the standard DMBA/TPA protocol from 3 to 18 weeks of age are shown. n = 7 for each group; * p < 0.05, Students t test; error bars represent SEM. Genotypes: Msx2-Cretg/+; Notch1ox/ox (N1CKO), Msx2-Cretg/+; Notch1ox/ox; Notch2ox/+ (N1N2hCKO), Msx2-Cretg/+; Notch1ox/ox; Notch2ox/+; Notch3/ (N1N2hN3CKO), Msx2Cretg/+; Notch1ox/ox; Notch2ox/ox (N1N2CKO), Msx2-Cretg/+; Ps1ox/ox; Ps2/ (PSDCKO), Msx2Cretg/+; Rbpjox/ox (RBPjCKO).

prone to spontaneous skin tumors. Consistent with these ndings, mice and humans lacking components of the g-secretase complex (Presenilin 1 and 2 [Ps1/2], presenilin enhancer 2 [PEN2], APH1 or Nicastrin) have elevated rates of spontaneous tumor development (Lapins et al., 2001; Li et al., 2007; Wang et al., 2010; Xia et al., 2001). Additionally, the g-secretase inhibitor Semagacestat (LY450139) led to elevated incidence of skin cancer among the participants in a phase III clinical trial (Extance, 2010). Importantly, the latency for spontaneous tumor formation in the epidermis is determined by the degree of disruption in its differentiation program caused by Notch signaling loss (Demehri et al., 2009b). An epidermal-derived cytokine whose level rises as more Notch signaling is lost in the skin is Thymic Stromal Lymphopoietin (TSLP), which could contribute to the susceptibility of Notchdecient skin to tumorigenesis (De Monte et al., 2011; Demehri et al., 2008, 2009b; Olkhanud et al., 2011; Pedroza-Gonzalez et al., 2011). TSLP is an IL-7-like cytokine studied mainly in the context of T helper 2 (Th2)-mediated allergic inammation in the skin and lung (Leonard, 2002; Rochman et al., 2009; Ziegler, 2010); overexpression of TSLP is sufcient to promote the development of atopic dermatitis and asthma, respectively (Ziegler, 2010). Importantly, transient exposure to an allergen in the presence of TSLP is sufcient to prime the skin and lung immune cells, creating long-lasting T cells that can trigger allergic inammation later (Han et al., 2012; Zhang et al., 2009; Demehri et al., 2009a). Although TSLP is not expressed in the skin under physiological conditions, keratinocytes are powerful secretors of TSLP in both humans (Lee et al., 2010) and mice; chronic and

severe barrier disruption can result in TSLP release into the serum up to 5,000-fold over its baseline levels (Demehri et al., 2008; Dumortier et al., 2010; Zhang et al., 2009). Interestingly, systemic TSLP drives a leukemia-like B cell lymphoproliferative disease in newborn mice (Demehri et al., 2008), and constitutively active TSLPR causes acute B-lymphoblastic leukemia in children (Cario et al., 2010; Hertzberg et al., 2010; Shochat et al., 2011). TSLP has also emerged recently as a progrowth cytokine in breast and pancreatic cancers (De Monte et al., 2011; Olkhanud et al., 2011; Pedroza-Gonzalez et al., 2011). These ndings suggest a therapeutic opportunity for TSLP-blocking agents that are in development for the treatment of allergic diseases (Schmitt, 2011) as cancer immunotherapeutic agents. Considering the emerging role of TSLP as a potential therapeutic target in cancer therapy, we set out to investigate the role of epidermal-derived TSLP in skin carcinogenesis. RESULTS Mice Lacking All Canonical Notch Signaling in the Epidermis Are Resistant to Skin Carcinogenesis The Msx2-Cretg/+ line expresses Cre recombinase early, transiently, and only in the dorsal and ventral midline regions of the skin, generating a calico pattern of gene deletion (Figure 1A). Stepwise removal of Notch alleles in epidermal keratinocytes by Msx2-Cretg/+ is associated with a corresponding decline in differentiation, creation of a wound-like environment and increased susceptibility to carcinogenesis (Demehri et al., 2009b; Figures 1A and 1B). Animals lacking all Notch signaling
Cancer Cell 22, 494505, October 16, 2012 2012 Elsevier Inc. 495

Cancer Cell
TSLP Overexpression Protects from Skin Cancer

Demehri et al., 2009b). Resistance to DMBA/TPA might reect a reduced growth potential of RBPj-decient keratinocytes (Blanpain et al., 2006). However, even if this were the case, we would expect initiated wild-type cells in this environment to form tumors (Demehri et al., 2009b). Alternatively, resistance to tumorigenesis in RBPjCKO skin may be due to a switch from a tumor-promoting environment in N1CKO (Demehri et al., 2009b) to a tumor-suppressing environment in RBPjCKO skin. Wild-type Keratinocytes Replace Their Notch Signaling-Decient Neighbors over Time The calico pattern of gene deletion in RBPjCKO animals allowed us to notice a second, potentially related phenotype. As these mice aged, the mutant epidermal clones on their dorsal and ventral surfaces shrank, and RBPj-decient epidermis eventually disappeared in the oldest individuals (Figure 2A and data not shown). Although Msx2-Cretg/+; Notch1ox/ox; Notch2ox/ox (N1N2CKO); and PSDCKO mice die post weaning due to a lethal B-lymphoproliferative disorder (Demehri et al., 2008), they did survive longer if we controlled their B-LPD with a sublethal dose of irradiation (Demehri et al., 2008). When lethality was rescued in this manner, we observed a similar regression of Notch1/2- or Ps1/2-deleted epidermis as N1N2CKO and PSDCKO animals aged (Figure 2B). This too could reect a proliferative disadvantage of Notch signaling-decient keratinocytes or the active process of rejection. H-Ras-Infected Notch-Decient Keratinocytes Are Highly Tumorigenic in Immune-Compromised Mice To ask if the resistance of RBPjCKO mice to carcinogenesis and the loss of mutant keratinocytes are due to a low proliferative capacity, we evaluated their tumor-forming potential in response to the activated H-Ras oncogene in the nude mouse environment (Nicolas et al., 2003). First, we isolated keratinocytes from newborn Rbpjox/ox and Rbpj+/ox littermates. Cells were then infected with activated H-Ras-expressing retrovirus, allowed to recover for 24 hr and then infected with Cre-expressing adenovirus. H-Ras-infected, Cre expressing, Rbpj/ (RBPjKO) or Rbpj+/ (wild-type) cells (1.5 3 106) were injected subcutaneously into nude mice and tumor development was monitored over time. Importantly, RBPjKO keratinocytes formed large tumors with histological features of moderately differentiated squamous cell carcinomas in 30 days, while wild-type keratinocytes did not form a signicant tumor mass (Figure 3). Similar results were obtained using g-secretase-decient (PSDKO) keratinocytes (Figure 3). These results demonstrate that Notch signaling-decient cells are highly proliferative and competent to form tumors in a T cell-decient environment. Therefore, we hypothesized that the apparent resistance of Notch signalingdecient animals to skin tumorigenesis is most likely due to the activation of a tumor-suppressing environment in their skin. Notch-Decient Animals Develop Severe Allergic Skin Inammation Caused Specically by Epidermal TSLP Overexpression Inammation can either promote or suppress tumorigenesis depending on its magnitude and the immune cells involved (Coussens and Werb, 2002; Schreiber et al., 2011). In sharp contrast to the mild inammation, dermal broplasia and

Figure 2. Notch Signaling-Decient Epidermal Clones Regress with Age


(A) The red dotted line and arrowheads delineate the boundaries of the RBPjdecient dorsal epidermis, as determined by hair/epidermal phenotype. Bottom panel shows a-RBPj antibody staining of RBPj-depleted midline skin (left) and wild-type skin in the periphery (right) of an 18-week-old RBPjCKO animal. (B) The red dotted line and arrowheads delineate the boundaries of the mutant dorsal epidermis in N1N2CKO and PSDCKO animals. Asterisk marks the recipients of a sublethal dose of irradiation in the second week of life; representative pictures are shown in all panels; scale bars: 1 cm (mice pictures); 50 mm (histology).

in their skin (e.g., Msx2-Cretg/+; Ps1ox/ox; Ps2/ [PSDCKO; loss of g-secretase enzyme function]) die shortly after birth, not allowing enough time for spontaneous tumor development (Figure 1B). RBPj-decient animals (Msx2-Cretg/+; Rbpjox/ox; or RBPjCKO), however, live for 100 days on average, which is comparable to animals lacking all but one Notch2 allele in their skin (Msx2-Cretg/+; Notch1ox/ox; Notch2ox/+; Notch3/; or N1N2hN3CKO). Surprisingly, while 20% of N1N2hN3CKO mice developed spontaneous skin tumors, none (0/40) of the RBPjCKO mice that have been examined developed spontaneous tumors (Figure 1B). To further rule out the possibility that the short life span of RBPjCKO mice masked incipient tumors, we subjected RBPjCKO mice, maintained through >6 generations of intercrossing in a mixed genetic background (C57BL/6 with FVB, CD1 contribution), to a multistage chemical skin carcinogenesis model (Yuspa et al., 1994). Three-week-old RBPjCKO and their wild-type littermates (dened as all mice not inheriting the Cre transgene) were treated with 25 mg 7,12-dimethylbenz[a] anthracene (DMBA), followed by a twice-weekly dose of 12-Otetradecanoylphorbol-13-acetate (TPA) for 14 weeks. Surprisingly, no tumors were detected in the DMBA/TPA-treated RBPjCKO animals after 15 weeks of follow-up, whereas the majority of wild-type littermates developed more than 20 papillomas/mouse (n = 7 in each group, p < 0.0001; Figures 1C and 1D). These results contrast starkly to the enhanced tumor susceptibility seen in other Notch-decient animals (Figure 1B;
496 Cancer Cell 22, 494505, October 16, 2012 2012 Elsevier Inc.

Cancer Cell
TSLP Overexpression Protects from Skin Cancer

were indistinguishable from RBPjCKO littermates (Figure S1C), indicating that TNFa was not necessary for theses phenotypes. RBPjCKO keratinocytes produced signicantly higher levels of TSLP than those seen in the tumor-prone N1CKO and N1N2hN3CKO animals (Figure 4B), driving the severe Th2 inammation seen in RBPjCKO skin (Demehri et al., 2009a; Dumortier et al., 2010; He et al., 2008). Therefore, we next examined the effect of loss of TSLP signaling on inammation by generating RBPjCKO mice that lack the TSLP receptor by deleting genes encoding its subunits (Il7ra or Crlf2 [Tslpr]). Il7ra/ (IL7raKO) mice were used in place of Tslpr/ (TSLPRKO) mice because Rbpj and Tslpr are linked on the same arm of chromosome 5, and thus we were unable to generate RBPjCKO;TSLPRKO animals. Deleting Il7ra in RBPjCKO or N1N2CKO (Msx2-Cretg/+; Rbpjox/ox; Il7ra/ or RBPjCKO; IL7raKO, Msx2-Cretg/+; Notch1ox/ox; Notch2ox/ox; Il7ra/ or N1N2CKO; IL7raKO) mice led to a marked reduction in skin inammation (Figure 4C; Figure S1D). To conrm that this effect was specic to TSLP and not an indirect consequence of reduced lymphocyte numbers in IL7raKO mice (Peschon et al., 1994), we deleted Tslpr in PSDCKO animals (Msx2-Cretg/+; Ps1ox/ox; Ps2/; Tslpr/ or PSDCKO; TSLPRKO), which signicantly prolonged their lifespan (Dumortier et al., 2010). As with N1N2CKO; IL7raKO and RBPjCKO; IL7raKO, inammation was greatly reduced in PSDCKO;TSLPRKO animals (Figure S1D). Taken together, these results demonstrate a central role for TSLP in regulating the level of inammation in Notch-decient skin. Blocking TSLP Signaling in Notch-Decient Animals Results in the Expansion of the Mutant Skin and Tumorigenesis RBPjCKO;IL7raKO animals showed a clear reversal of the two phenotypes unique to RBPjCKO mice. First, RBPj-decient epidermal clones expanded dramatically in RBPjCKO;IL7raKO mice, forming numerous hypertrophic cysts (Figure 4D). Together with the data in Figure 3, this result excludes a proliferative defect in RBPj-decient keratinocytes. Second, RBPjCKO; IL7raKO mice developed spontaneous, invasive dermal and exophytic tumors over time (Figures 4D and S1E). N1N2CKO; IL7raKO and PSDCKO;TSLPRKO animals also showed expansion of their mutant skin territories (Figure 4D; Dumortier et al., 2010 and the accompanying paper by Di Piazza et al.). All RBPjCKO;IL7raKO and PSDCKO;TSLPRKO animals eventually developed cancerous lesions that invaded through the subcutaneous muscle layer at 10 to 15 weeks of age (Figures 4D and S1E). Treating the skin of these animals with a single dose of DMBA further revealed their susceptibility to tumorigenesis (Figure S1E). Therefore, eliminating TSLP reception reduced inammation, restored a tumor-promoting environment reminiscent of the N1CKO mice, and uncovered the underlying cancer-prone phenotype in mice lacking canonical Notch signaling in their skin. Whether viruses contributed to tumor formation in these animals remains to be determined. The Adaptive Immune System Mediates the Effects of TSLP on Skin Rejection and Tumor Resistance In order to determine if bone marrow (BM)-derived immune cells mediated the effects of TSLP on skin rejection and resistance
Cancer Cell 22, 494505, October 16, 2012 2012 Elsevier Inc. 497

Figure 3. H-Ras-Infected RBPjKO or PSDKO Keratinocytes Are Highly Tumorigenic in Nude Mice
(A) Schema showing the experimental procedure. Cells are infected with retrovirus containing oncogenic H-Ras and then with Adeno-Cre to delete the oxed alleles. H-Ras-infected RBPjKO or PSDKO and wild-type cells injected subcutaneously into the right and left anks of the nude mice, respectively. (B) Nude mice with palpable tumors (red circles) are harvested 30 days after the injection. Hematoxylin and eosin (H&E) and a-RBPj antibody stainings on a tumor formed by RBPjKO keratinocytes are shown. Representative pictures are shown; scale bars: 1 cm (mice pictures), 50 mm (histology).

angiogenesis, which generated a tumor-promoting environment in N1CKO skin (Demehri et al., 2009b), RBPjCKO skin exhibited a signicant accumulation of leukocytes (CD45+ cells) beneath the RBPj-decient epidermal clones (Figure 4A). To test if this high level of dermal inammation could have a suppressing effect on tumor growth (Coussens and Werb, 2002), we rst examined the effect of immunosuppressant drugs on DMBAtreated RBPjCKO mice. Treatment with the maximum tolerable doses of Dexamethasone or Methotrexate did not signicantly reduce inammation in RBPjCKO mice, nor did it affect the rejection of mutant skin cells or allow tumor formation (Figures S1A and S1B available online). Next, we examined if TNFa, a prominent proinammatory cytokine also overexpressed in Notch signaling-decient skin (Dumortier et al., 2010), contributes to skin inammation in RBPjCKO animals. The levels of skin inammation, tumor resistance, and mutant skin patch rejection in Msx2-Cretg/+; Rbpjox/ox; TnfrI/; TnfrII/ mice

Cancer Cell
TSLP Overexpression Protects from Skin Cancer

Figure 4. Blunting the Skin Inammation by Blocking TSLP Signaling in Notch Signaling-Decient Animals Results in Mutant Skin Expansion and Tumorigenesis
(A) H&E and CD45 antibody stainings show the level of skin inammation in RBPjCKO animals compared to the tumor-prone N1CKO and N1N2hN3CKO mice. Scale bars: 50 mm. (B) The circulating TSLP levels during the second week of life are compared among the allelic series of Notch-decient animals. *p < 0.05, Students t test; error bars represent SD. This gure is modied from Demehri et al., 2008. (C) H&E and CD45 antibody stainings highlight the level of dermal inammation in RBPjCKO; IL7raKO animals that lack TSLP signaling. Scale bars: 50 mm. (D) Gross and microscopic features of RBPjCKO; IL7raKO, N1N2CKO;IL7raKO and PSDCKO; TSLPRKO skin are compared to those of RBPjCKO, N1N2CKO, and PSDCKO littermates, respectively. Asterisks mark the recipients of a sublethal dose of irradiation in the second week of life; insets: tumors penetrating the muscle layer; scale bars: 1 cm (mice pictures), 200 mm (histology); representative pictures are shown in all panels. See also Figure S1.

to tumorigenesis, we used littermates with identical major MHC class I haplotypes (Figures S2AS2C) and reconstituted the immune system of lethally irradiated RBPjCKO;IL7raKO and PSDCKO;TSLPRKO mice with BM from their wild-type littermates. We monitored their response to DMBA treatment following BM transplantation (BMT). Interestingly, wild-type BM restored resistance to DMBA-induced tumors in RBPjCKO; IL7raKO and PSDCKO;TSLPRKO mice; concomitantly, the mice gained the ability to eliminate their mutant skin cells (Figures 5A, 5B, and S2D). In a complementary set of experiments, we reconstituted the immune system of lethally irradiated RBPjCKO and PSDCKO mice with BM from their IL7raKO and TSLPRKO littermates, respectively, to determine if a protumor environment would be re-established. This immune reconstitution, however, failed to establish tolerance, and no DMBAinduced tumors formed (Figure S2E). This could be explained by the persistence of irradiation-resistant activated and memory T cells in the mutant animals at the time of transplantation (Figure S2F). Together, these ndings demonstrate that BMderived immune cells act downstream of TSLP, and suggest that the effector cell type(s) formed in RBPjCKO and PSDCKO are resistant to lethal doses of irradiation. Based on the ndings above, we focused on the immune cell repertoire in RBPjCKO (Figure S2G; Demehri et al., 2009b) and targeted cell types that are known to be resistant to irradiation
498 Cancer Cell 22, 494505, October 16, 2012 2012 Elsevier Inc.

(Murphy et al., 1987), express TSLP receptors, and mediate tumor resistance and skin rejection (Vesely et al., 2011). Based on these criteria, we chose to delete CD8+ cytotoxic T lymphocytes (CTLs) and mast cells in RBPjCKO animals. Genetic depletion of these cells in RBPjCKO;CD8a/ and RBPjCKO;Kitwsh/wsh did not cause any alteration in the RBPjCKO skin phenotypes, suggesting that neither cell type is necessary to mediates the effects of TSLP (Figure S2H). In a separate set of experiments, we used previously described depleting antibodies (Rogers and Unanue, 1993; Shankaran et al., 2001) to deplete CD4+ T cells, CD8+ T cells, natural killer (NK) cells, or granulocytes in RBPjCKO animals. Although we achieved the complete depletion of these cell types in the blood, we failed to alter rejection or tumor resistance phenotypes in RBPjCKO mice (Figure S2I), most likely due to the resistance of activated T cells to depletion (Figure S2J). From these sets of experiments, we concluded either that (1) the cell type(s) mediating the effects of TSLP were resistant to antibody depletion and irradiation, such as activated/memory T cells (Jamali et al., 1992; Murphy et al., 1987), or (2) not present among the cell types depleted with the antibodies we used. To distinguish these possibilities systematically, we focused on RBPjCKO;IL7raKO and PSDCKO;TSLPRKO mice, which never establish TSLP-dependent tumor resistance and thus lack activated effector cells. Having established that wild-type BM can reconstitute the skin rejection and tumor resistance in RBPjCKO;IL7raKO and PSDCKO;TSLPRKO animals (Figures 5A and 5B), we reconstituted the immune cells of lethally irradiated RBPjCKO;IL7raKO and PSDCKO;TSLPRKO mice

Cancer Cell
TSLP Overexpression Protects from Skin Cancer

Figure 5. CD4+ T Cells Mediate the Effects of TSLP on Notch-Decient Skin Rejection and Tumor Resistance
(A and B) The skin phenotypes of RBPjCKO;IL7raKO (A) and PSDCKO;TSLPRKO (B) animals are analyzed 8 weeks after BMT from wild-type littermates, Rag2/,gc/ (Rag2gcKO) or Rag2/ (Rag2KO) animals. Note that Rag2gcKO and Rag2KO donors lack T and B cells. The level of dermal inammation, the rejection of the mutant keratinocytes, the formation of hypertrophic cysts and the development of DMBA-dependent exophytic tumors are scored in the recipient animals at 10 weeks of age. BMT from TSLP receptor-decient littermates are used as controls. *p < 0.05 compared to wild-type BM donor group. (C and D) Two-week-old RBPjCKO;IL7raKO (C) and PSDCKO;TSLPRKO (D) pups were irradiated with 450-cGy and underwent adoptive T cell transfer using wildtype CD4+, wild-type CD8+ or TSLPR-decient CD4+ T cells isolated from the spleens of their littermates as shown in the schematic diagrams. The level of dermal inammation, the rejection of the mutant keratinocytes, the formation of hypertrophic cysts, and the development of DMBA-dependent exophytic tumors are documented in the recipient animals at 10 weeks of age. *p < 0.05 compared to wild-type CD4+ T cell donor group. (E) Two-week-old RBPjCKO;IL7raKO mice were treated with depleting antibodies (anti-CD4 or anti-CD8a; black arrows), lethally irradiated 2 days later and transplanted with T cell-depleted c-Kit+ BM progenitor cells from their wild-type littermates as shown in the schematic diagram. Two days later, the animals were treated with one dose DMBA while continuing to receive a weekly intraperitoneal injection of the indicated depleting antibody for 10 weeks. Flow cytometry shows the status of CD4+ and CD8+ T cells in the peripheral blood of RBPjCKO;IL7raKO mice transplanted with c-Kit+ wild-type BM and treated with anti-CD4 or antiCD8a antibody one week after the last antibody injection. Blood CD45+ leukocytes are traced in the gure. Note that PE-conjugated anti-CD8b antibody is used to detect CD8+ cells. (F) RBPjCKO;IL7raKO mice transplanted with c-Kit+ BM cells and treated with anti-CD4, anti-CD8a, or IgG control antibodies are compared. Bar graphs show the average number of DMBA-dependent exophytic tumors in each treatment group. *p < 0.05 compared to IgG control-treated group. For all experiments, mice were followed up to 90 days of age for exophytic tumor count; at least ve mice were analyzed in each group; error bars on all bar graphs represent SD; representative pictures are shown; scale bars: 1 cm. See also Figure S2.

with BM from Rag2/ donors that lack adaptive immunity. DMBA-treated RBPjCKO;IL7raKO and PSDCKO;TSLPRKO animals transplanted with Rag2/ BM failed to reject mutant skin, which formed hypertrophic cysts and developed tumors

(Figures 5A, 5B, and S2K). This clearly demonstrates that adaptive immune cells responding to TSLP are required for the tumor resistance and mutant skin rejection in Notch signaling-decient animals.
Cancer Cell 22, 494505, October 16, 2012 2012 Elsevier Inc. 499

Cancer Cell
TSLP Overexpression Protects from Skin Cancer

CD4+ T Cells Are Both Required and Sufcient to Mediate the Effects of TSLP on Skin Rejection and Tumor Resistance Among adaptive immune cell types, activated T cells are known to be resistant to irradiation and persist after antibody depletion (Jamali et al., 1992; Murphy et al., 1987). Considering that mice lacking CD8+ CTLs retained their tumor resistance and skin rejection phenotypes (RBPjCKO;CD8a/; Figure S2H), CD4+ T cells emerged as the prime candidate mediating the effects of TSLP in our model of Notch signaling-decient skin. To test this hypothesis, 2 3 106 CD4+ or CD8+ T cells from wild-type littermates were transferred to sublethally irradiated RBPjCKO; IL7raKO and PSDCKO;TSLPRKO newborns. Adoptive transfer of wild-type CD4+ T cells to RBPjCKO;IL7raKO and PSDCKO; TSLPRKO mice re-established the tumor resistance and mutant skin rejection phenotypes in these animals, but adoptive transfer of wild-type CD8+ CTLs did not (Figures 5C, 5D, and S2L). As expected, wild-type CD4+ T cells transferred into RBPjCKO; IL7raKO animals underwent Th2 differentiation (Figure S2M). This nding demonstrates that CD4+ Th2 cells competent to receive the TSLP signal are sufcient to reconstitute the protective effects of TSLP in TSLPR and Notch signaling-decient skin. To ask if CD4+ T cells were required to initiate the effects of TSLP, lethally irradiated RBPjCKO;IL7raKO mice were transplanted with c-Kit+ BM cells from their wild-type littermates and treated continually with anti-CD4, anti-CD8, or IgG control antibodies (Figure 5E). Injecting the RBPjCKO;IL7raKO animals with depleting antibodies weekly resulted in effective and sustained depletion of targeted cell types (Figure 5E). Importantly, only the RBPjCKO;IL7raKO mice that received wild-type BM progenitors plus anti-CD4 antibody retained their mutant cells, formed hypertrophic cysts, and developed skin tumors in response to DMBA (Figure 5F), despite the possible presence of a few mature CD4+ cells, which differentiated in the donor environment. In all BMT and adoptive T cell transfer experiments, the presence of donor-derived hematopoietic cells was conrmed at the conclusion of the study (Figure S2N). Together, these data demonstrate that naive CD4+ T cells constitute the cell type receiving the TSLP signal and coordinating the immune response necessary to reject Notch signaling-decient skin and prevent tumorigenesis in this context. Epidermal TSLP Overexpression in Wild-type Skin Prevents Skin Tumorigenesis To test if TSLP overexpression can mobilize the antitumor immune response in a background free of Notch mutations, we used chemical and genetic approaches to upregulate epidermal TSLP expression in wild-type animals subjected to the multistage chemical skin carcinogenesis model. First, we used the topical application a low-calcemic Vitamin D3 analog (calcipotriol; known also as MC903 or Dovonex) to induce epidermal TSLP expression in CD1 female mice (Li et al., 2006). CD1 genetic background was chosen for these studies because the skin inammation caused by TSLP overexpression downstream of calcipotriol treatment did not result in a full-blown AD-like disease (Demehri et al., 2009a). Mice were treated with a single initiating dose of DMBA (125 mg), followed by a twice-weekly dose of TPA (4 mg) for 14 weeks. The test group was also treated with topical calcipotriol (32 nmol) and controls with carrier only
500 Cancer Cell 22, 494505, October 16, 2012 2012 Elsevier Inc.

(EtOH) ve times a week during the 14 weeks of TPA application. Strikingly, the majority of carrier-only treated animals developed papillomas, whereas only two of the calcipotriol-treated animals transiently developed papillomas during 15 weeks of follow-up (n = 10 for each group, p < 0.0001; Figures 6A and 6B). To investigate if calcipotriol application can affect existing skin tumors, we randomly divided the tumor-bearing CD1 animals from the carrier-only treatment group (Figures 6A and 6B) into two groups at the end of the TPA treatment period. One subgroup received calcipotriol ve times weekly, whereas the other received carrier only. Both were monitored for an additional 7 weeks. While tumors on the carrier-treated animals continued to grow in size, tumors on the calcipotriol-treated animals shrank over time (p < 0.05 at weeks 6 and 7, Figures 6C and 6D). It is possible that the effects observed above were related to vitamin D signaling, independent of TSLP expression in the skin. To conrm that the antitumor effects are mediated through TSLP, we examined the tumor susceptibility of transgenic animals that overexpress TSLP in basal keratinocytes. Consistent with the proposed role for TSLP, K14-Tslptg/+ (K14-TSLPtg) female animals treated once with 100 mg DMBA followed by 4 mg TPA twice weekly for 14 weeks showed signicant resistance to tumorigenesis compared to their wild-type littermates (n = 22 for K14-TSLPtg group and 21 for wild-type group, p < 0.0001 log rank test, Figures 6E and 6F). The serum TSLP measurements conrmed the overexpression of TSLP in calcipotriol-treated wild-type and K14-TSLPtg animals (Figure 6G). These results clearly demonstrate that the tumor-suppressor effect of TSLP in the skin can be extended to wild-type animals and that TSLP can prevent tumorigenesis as well as inhibit growth of existing tumors. DISCUSSION The main nding of this report is that upregulation of epidermal TSLP can generate a dominant and lasting antitumor CD4+ T cell response in a Th2 inammatory microenvironment (Demehri et al., 2009a), protecting animals from spontaneous and chemically-induced skin tumors. These cells orchestrate the recognition and elimination of proliferating precancerous cells. This is in stark contrast to the previously described protumor function of Th2-polarized inammation (Johansson et al., 2008) and of TSLP (De Monte et al., 2011; Olkhanud et al., 2011; Pedroza-Gonzalez et al., 2011). Importantly, we show that chemical induction of TSLP expression in wild-type animals with DMBA-induced papillomas results in tumor regression. This latter nding suggests that TSLP upregulation may provide therapeutic benets in treating skin tumors and perhaps for other solid tumors. TSLP is a pleiotropic cytokine involved in several immune processes (Ziegler and Artis, 2010). In the skin and the lung, TSLP is expressed in response to barrier disruption (Demehri et al., 2008), where it can skew CD4+ T cell differentiation toward a Th2 subtype and mediate allergic inammation (Al-Shami et al., 2004; Tanaka et al., 2009; Ziegler and Artis, 2010; Leonard, 2002). TSLP serum levels can be a sensitive readout for the degree of disruption in epidermal differentiation/barrier integrity (Demehri et al., 2008). Epidermal Notch1 deletion results in mild disruption in epidermal differentiation, which

Cancer Cell
TSLP Overexpression Protects from Skin Cancer

Figure 6. TSLP Creates a Tumor-Suppressing Environment in the Wild-type Skin


(A and B) DMBA/TPA treated CD1 wild-type animals were topically treated with calcipotriol or carrier (EtOH), (A) time to tumor onset (p < 0.0001, log rank test) and (B) average number of tumors among tumor-bearing animals are shown. n = 10 for each group; error bars represent SEM; *p < 0.01, Students t test. (C and D) The eight tumor-bearing carrier-treated mice shown in (A and B) were randomly divided into two groups at the end of the 15-week DMBA/TPA treatment course. The test group was treated with 32 nmol calcipotriol while the control group continued to receive carrier alone 5 times per week. (C) After an additional 7-week treatment period, the tumor burdens of the calcipotriol-treated and carrier-treated mice were compared. n = 4 in each group; *p < 0.05, Students t test; error bars represent SD. (D) The representative images show the size of the remaining tumors in calcipotriol-treated versus carrier-treated mice at the end of the 7-week follow-up period. Tumors are highlighted with circles; scale bars: 1 cm. (E and F) K14-TSLPtg female mice and their wild-type female littermates treated with DMBA/TPA are compared for (E) time to tumor onset (p < 0.0001, log rank test) and (F) average tumor number among tumor-bearing animals. n = 22 for K14-TSLPtg group and n = 21 for wild-type group; error bars represent SEM; *p < 0.01, Students t test. (G) Circulating TSLP levels in calcipotriol-treated and K14-TSLPtg animals were compared to their controls. Error bars represent SD; *p < 0.0001 compared to wild-type group, Students t test.

forms a tumor-promoting environment dominated by Gr-1+ CD11b+ myeloid suppressor cells and soluble factors (Demehri et al., 2009b). The contribution of TSLP to this tumor-promoting skin environment remains to be determined. Others have shown that TSLP is responsible for promoting tumor growth and metastasis in pancreatic and breast cancer (De Monte et al., 2011; Olkhanud et al., 2011; Pedroza-Gonzalez et al., 2011). However, once all Notch-signaling is inactivated, high TSLP levels help establish a potent tumor-suppressing response. The observation that calcipotriol treatment is capable of shrinking pre-existing tumors argues against a model in which TSLP may have a protective effect on the early stages of tumorigenesis but a promoting effect on later stages of tumor growth and metastasis, but this is still feasible. There are other possible explanations for these conicting ndings: TSLP tumor-suppressor effect may be skin-specic; alternatively, based on observations in an allelic series of Notch-decient mice, TSLP levels may determine whether it promotes or inhibits tumor development and growth. The tipping point may vary according to the strength of the protumor environment in a specic context. Support for the latter comes from the observations that higher TSLP levels are needed to establish a tumorresistant phenotype in the tumor-prone Notch mutant animals than in the otherwise wild-type K14-TSLPtg animals. Below this hypothetical threshold, TSLP will have a neutral or tumor-

promoting effect; above it, TSLP acts as a potent tumorsuppressor. These concepts need to be formally tested in future studies. Nonetheless, here we report that, in both a neutral and a protumor microenvironment, TSLP concentrations capable of creating a CD4+ T cell-mediated antitumor response can be reached. Once achieved, activated CD4+ T cells not only identify and prevent growth of transformed cells but also mediate the rejection of Notch signaling-decient cells. In the Msx2-Cre animals, large territories of mutant tissue are embedded in wild-type skin. Therefore, the global resistance of calico RBPjCKO skin to chemical carcinogens suggests that the wild-type keratinocytes harboring an activated H-Ras allele are being suppressed alongside mutant cells with three different Notch-decient genotypes (i.e., RBPjCKO, N1N2CKO or PSDCKO). Importantly, TSLP-activated CD4+ T cells do not target the normal skin cells in the same animal as they proliferate to replace the rejected mutant ones. Based on these observations, we conclude that CD4+ T cells conditioned by high local TSLP concentrations must recognize immunogenic epitope(s) arising independent of Notch signaling in cells with abnormal differentiation. This protective activity does not arise in animals that have CD4+ T cells that lack the TSLP receptor, that have low local TSLP concentrations, or that do not have CD4+ T cells. It is important to note that CD4+ T cell activation by TSLP can occur even if all other tissues and hematopoietic
Cancer Cell 22, 494505, October 16, 2012 2012 Elsevier Inc. 501

Cancer Cell
TSLP Overexpression Protects from Skin Cancer

lineages, including dendritic and Langerhans cells, are rendered blind to TSLP by germline deletion of either TSLP receptor arms (Il7ra or Crlf2/Tslpr), as shown in the context of allergic disease where TSLP acts directly on CD4+ T cells (Al-Shami et al., 2005; Rochman et al., 2007; Rochman and Leonard, 2008). These data establish that TSLP-responsive CD4+ T cells are both sufcient and required to create the tumor suppressing microenvironment, likely by recruiting several cytotoxic immune effector cells, including CTLs, NK cells, and macrophages to the skin. Considering that TSLP promotes Th2 differentiation in Notch signaling-decient mice (data not shown; Al-Shami et al., 2004; Demehri et al., 2009a; Tanaka et al., 2009; Ziegler and Artis, 2010), we propose that the CD4+ Th2 cells are initiating the effects of TSLP. The exact nature of the antigens they are reacting to, and whether they are keratinocyte-specic, remain to be addressed in future studies. Once activated, the CD4+ Th2 cells home to the skin and form a lasting pool of memory cells that could not be eradicated with anti-CD4 antibody or irradiation (Figure 4D). This observation is very exciting because it demonstrates that TSLP induces a lasting antitumor immunity that is achieved by targeting antigens specic to at least some tumor cells. This protection can be achieved by application of a TSLP-inducer (calcipotriol) to fully developed skin tumors in wild-type animals. It remains to be determined if TSLP can also stimulate the regression of other solid tumors besides the ones formed in the skin. In an accompanying paper, a similar set of observations is reported. Both studies report an antitumor function for TSLP in Notch-decient skin, both demonstrate that this function of TSLP can also be elicited in animals with intact Notch signaling, and both nd an important role for CD4+ T cells. However, whereas our colleagues report that CD8+ CTLs mediate the tumor resistance phenotype, we do not nd CD8+ T cells to be necessary in our model. In their study, Notch alleles are deleted globally in the skin after birth using an inducible basal keratinocyte Cre. Consequently, deletion occurs during the hair follicle growth phase, the hair follicle bulge remains intact and the immune system matures in the absence of TSLP or tumor antigens. This deletion paradigm precludes monitoring the rejection of mutant skin clones. In our mice, the immune system matures in the presence of TSLP, which begins to accumulate in utero at E16.5 (Demehri et al., 2008), and in the presence of the antigens presented by mutant cells and by tumor cells later on. Moreover, in our mice, the hair follicle is destroyed by P10 and the bulge never forms because the outer root sheath of the hair follicles adopts an epidermal fate right after birth (Demehri et al., 2008). It is conceivable that some of the difference between the two studies reects methodological differences, but it is hard to explain why CD8+ CTLs do not contribute to our system, yet provide the bulk of protection for their Notch-decient mice. The explanation may lie in the difference between the tumor cells of origin, the timing of the Notch signaling deletion, or the differences in background (ours is mixed, whereas theirs is pure C57/B6). The differential role of CTLs in these two experimental paradigms needs to be addressed in future studies. The accompanying study demonstrates that the cystic tumors arise via a Wnt/b-catenin-dependent mechanism from bulgederived hair follicles, whereas our tumors originate from epidermal cells that did not accumulate nuclear b-catenin prior
502 Cancer Cell 22, 494505, October 16, 2012 2012 Elsevier Inc.

to tumor formation (Demehri et al., 2009b). Taken together, however, the differences between the studies support the conclusion that TSLP can provide immunological protection against skin cancer and support the speculation that it may do the same in other types of solid tumors. In summary, this study and the accompanying paper by Di Piazza et al. identify a previously unrecognized role for TSLP in inducing a robust antitumor response in several experimental paradigms (Notch loss of function, Wnt gain of function [Di Piazza et al., 2012 in this issue of Cancer Cell] and DMBAinduced H-Ras dependent tumors). Moreover, these studies demonstrate that TSLP exerts its antitumor effects through a CD4+ Th2 inammatory environment. Our ndings may explain why some individuals who suffer from Th2-dominant allergic disorders display reduced risk of developing certain types of cancers, including nonmelanoma skin cancers (Gandini et al., 2005; Hwang et al., 2012; Prizment et al., 2007; Vajdic et al., 2009; Wang and Diepgen, 2005). Importantly, therapeutic exploitation of this mechanism seems within reach given the efcacy of calcipotriol, a Food and Drug Administrationapproved drug and a potent inducer of TSLP, in blocking DMBA-induced carcinogenesis.
EXPERIMENTAL PROCEDURES Mice The mutant animals (listed in the Supplemental Experimental Procedures section) were generated according to the methods outlined in our previous report (Pan et al., 2004). All of the mice were housed in the Washington University animal facility, and all experiments were performed in accordance with relevant institutional and national guidelines and regulations approved by the Animal Studies Committee at Washington University. The pedigreed RBPjCKO cohort was maintained in mixed FVB, C57BL/6, and CD1 genetic backgrounds, which were overall more susceptible to DMBA/TPA skin carcinogenesis. All other animals were kept in mixed C57BL/6 and CD1 genetic backgrounds and, therefore, were relatively more resistant to chemical skin carcinogens. In all cancer experiments, age-matched littermates were compared. Msx2-Cretg/+; Rosa-LSL-Eyfp mouse was imaged at P0 using Leica stereoscopic uorescence microscope with regular light (rendered magenta in Figure 1A) or enhanced yellow uorescent protein (EYFP) lter. In studies related to mutant skin rejection, spontaneous/DMBA-induced tumorigenesis and life span, mice were photographed with a digital camera weekly and monitored for onset, number, and size of tumors and any sign of failure to thrive. Moribund mice were euthanized and skin, tumors, blood, spleen, and lymph nodes were harvested. Chemical Skin Carcinogenesis Studies For RBPjCKO DMBA/TPA experiments, 3-week-old mutant mice and Crenegative wild-type littermates were treated with the standard protocol for skin chemical carcinogenesis model, as previously described (Nicolas et al., 2003). RBPjCKO mice received one dose of 25 mg DMBA (Sigma, St. Louis) followed in a week by biweekly treatment with 4 mg TPA (Sigma) for 14 weeks. In K14-TSLPtg and CD1 carcinogenesis experiments, mice were treated with 125 mg DMBA (CD1, Calcipotriol treatment) or 100 mg DMBA (K14-TSLPtg). In all of the experiments, adult mice were shaved under anesthesia 2 days prior to treatment with DMBA to ensure the hair follicles were in the second telogen. In studies where tumorigenesis was induced with one dose of DMBA, the mutant animals received a single dose of 125 mg DMBA in 100 ml of acetone during the second week of life, after the mice were subjected to any other experimental procedures including irradiation, BMT, or adoptive T cell transfer. BMT The recipient mice were lethally irradiated with 950-cGy in the second week of life and transplanted with unfractionated BM cells from their littermates,

Cancer Cell
TSLP Overexpression Protects from Skin Cancer

Rag2/;gc/ or Rag2/ animals as previously described (Zhang and Ren, 1998). All transplanted animals were maintained on antibiotics containing water to prevent infection. Adoptive T Cell Transfer For T cell isolation, splenocytes were positively selected for CD4 or CD8 surface expression using CD4 or CD8 MicroBeads, respectively, followed by negative selection to remove any CD11c+ dendritic cells using CD11c MicroBeads (Miltenyi Biotec, Auburn, CA). RBPjCKO;IL7raKO and PSDCKO; TSLPRKO mice were irradiated with 450-cGy during the second week of life and injected intravenously with 2 3 106 CD4+ or CD8+ T cells isolated from the spleen of their littermates. BMT and Antibody Depletion RBPjCKO;IL7raKO newborn mice received intraperitoneal injection of 750 mg anti-CD4 (GK1.5; Bio X Cell, West Lebanon, NH), anti-CD8 (YTS; Bio X Cell), or IgG Control antibody (Sigma) in the second week of life. Forty-eight hours later, the mutant animals received BMT using c-Kit+ BM progenitors (lacking T cells) from their wild-type littermates isolated using CD117 MicroBeads and magnetic columns according to manufacturers protocol (Miltenyi Biotec Inc., Auburn, CA). Each recipient was injected intravenously with 2 3106 c-Kit+ BM cells in 100 ml PBS+2% FBS. The mutant animals were then treated topically with one dose of DMBA and continued to receive 250 mg of the depleting antibody weekly. Mice were monitored for skin rejection and tumor formation weekly and harvested at P90. Statistical Analysis Except for calcipotriol studies performed on wild-type CD1 animals, all of the animals used in this study were kept on mixed genetic backgrounds in a pedigreed colony (i.e., all animals are logged into a database). To minimize the confounding effects of strain or family background on tumor outcomes, we only compared gender-matched littermates in each cancer study. Using the power analysis described previously (Demehri et al., 2009b), we determined the number of animals needed in each chemical skin carcinogenesis study. As the test of signicance between the study groups, we used log rank test for time-to-tumor onset and Students t test for tumor counts, serum TSLP levels, and other quantitative measurements. SUPPLEMENTAL INFORMATION Supplemental Information includes two gures and Supplemental Experimental Procedures and can be found with this article online at http://dx.doi. org/10.1016/j.ccr.2012.08.017. ACKNOWLEDGMENTS We would like to thank Drs. Robert Schreiber, Kenneth Murphy, Barry Sleckman, Emil Unanue, Wayne Yokoyama, Deepta Bhattacharya, Takeshi Egawa, Suellen Greco, Omar Jassim, and David Denardo for providing us with many valuable experimental tools, ideas, and comments on the manuscript, and the members of the Kopan laboratory for their suggestions during the course of this study. We wish to thank Dr. Freddy Radtke and his coworkers for sharing their unpublished results with us during the course of these experiments. We would like to thank Dr. Gail Martin for providing Msx2-Cretg/+ mice, Dr. Tom Gridley for Notch2ox/ox mice, Dr. Jie Shen for Ps1ox/ox mice, Dr. Tasuku Honjo for Rbpjox/ox mice, Dr. Andrew Farr for the K14-TSLPtg mice, and Dr. Warren Leonard for Tslpr/ mice. We would like to thank Rheumatic Disease Core Centers Speed Congenics Lab at Washington University for their assistance with genotyping. S.D., M.T., and R.K. are supported by Grant GM55479-16 from NIH/NIGMS. S.M. and L.J.Y. were supported by funds from the American Asthma Foundation (Grant 09-0234). A.T. was supported by funds from NIH 2 U19 AI070489-09. The ow cytometry core is supported by P30 CA091842. Received: February 8, 2012 Revised: June 13, 2012 Accepted: August 21, 2012 Published: October 15, 2012

REFERENCES Al-Shami, A., Spolski, R., Kelly, J., Fry, T., Schwartzberg, P.L., Pandey, A., Mackall, C.L., and Leonard, W.J. (2004). A role for thymic stromal lymphopoietin in CD4(+) T cell development. J. Exp. Med. 200, 159168. Al-Shami, A., Spolski, R., Kelly, J., Keane-Myers, A., and Leonard, W.J. (2005). A role for TSLP in the development of inammation in an asthma model. J. Exp. Med. 202, 829839. Blanpain, C., Lowry, W.E., Pasolli, H.A., and Fuchs, E. (2006). Canonical notch signaling functions as a commitment switch in the epidermal lineage. Genes Dev. 20, 30223035. Campisi, J. (2005). Senescent cells, tumor suppression, and organismal aging: good citizens, bad neighbors. Cell 120, 513522. Cario, G., Zimmermann, M., Romey, R., Gesk, S., Vater, I., Harbott, J., Schrauder, A., Moericke, A., Izraeli, S., Akasaka, T., et al. (2010). Presence of the P2RY8-CRLF2 rearrangement is associated with a poor prognosis in non-high-risk precursor B-cell acute lymphoblastic leukemia in children treated according to the ALL-BFM 2000 protocol. Blood 115, 53935397. Coussens, L.M., and Werb, Z. (2002). Inammation and cancer. Nature 420, 860867. De Monte, L., Reni, M., Tassi, E., Clavenna, D., Papa, I., Recalde, H., Braga, M., Di Carlo, V., Doglioni, C., and Protti, M.P. (2011). Intratumor T helper type 2 cell inltrate correlates with cancer-associated broblast thymic stromal lymphopoietin production and reduced survival in pancreatic cancer. J. Exp. Med. 208, 469478. Demehri, S., Liu, Z., Lee, J., Lin, M.H., Crosby, S.D., Roberts, C.J., Grigsby, P.W., Miner, J.H., Farr, A.G., and Kopan, R. (2008). Notch-decient skin induces a lethal systemic B-lymphoproliferative disorder by secreting TSLP, a sentinel for epidermal integrity. PLoS Biol. 6, e123. Demehri, S., Morimoto, M., Holtzman, M.J., and Kopan, R. (2009a). Skinderived TSLP triggers progression from epidermal-barrier defects to asthma. PLoS Biol. 7, e1000067, doi:1000010.1001371/journal.pbio.1000067. Demehri, S., Turkoz, A., and Kopan, R. (2009b). Epidermal Notch1 loss promotes skin tumorigenesis by impacting the stromal microenvironment. Cancer Cell 16, 5566. Di Piazza, M., Nowell, C.S., Koch, U., Durham, A.-D., and Radtke, F. (2012). Loss of cutaneous TSLP-dependent immune responses skews the balance of inammation from tumor protective to tumor promoting. Cancer Cell 22, this issue, 479493. Dumortier, A., Durham, A.-D., Di Piazza, M., Vauclair, S., Koch, U., Ferrand, G., Ferrero, I., Demehri, S., Song, L.L., Farr, A.G., et al. (2010). Atopic dermatitislike disease and associated lethal myeloproliferative disorder arise from loss of Notch signaling in the murine skin. PLoS ONE 5, e9258. Extance, A. (2010). Alzheimers failure raises questions about diseasemodifying strategies. Nat. Rev. Drug Discov. 9, 749751. Gandini, S., Lowenfels, A.B., Jaffee, E.M., Armstrong, T.D., and Maisonneuve, P. (2005). Allergies and the risk of pancreatic cancer: a meta-analysis with review of epidemiology and biological mechanisms. Cancer Epidemiol. Biomarkers Prev. 14, 19081916. Han, H., Xu, W., Headley, M.B., Jessup, H.K., Lee, K.S., Omori, M., Comeau, M.R., Marshak-Rothstein, A., and Ziegler, S.F. (2012). Thymic stromal lymphopoietin (TSLP)-mediated dermal inammation aggravates experimental asthma. Mucosal Immunol. 5, 342351. 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. He, R., Oyoshi, M.K., Garibyan, L., Kumar, L., Ziegler, S.F., and Geha, R.S. (2008). TSLP acts on inltrating effector T cells to drive allergic skin inammation. Proc. Natl. Acad. Sci. USA 105, 1187511880. Hertzberg, L., Vendramini, E., Ganmore, I., Cazzaniga, G., Schmitz, M., Chalker, J., Shiloh, R., Iacobucci, I., Shochat, C., Zeligson, S., et al. (2010). Down syndrome acute lymphoblastic leukemia, a highly heterogeneous disease in which aberrant expression of CRLF2 is associated with mutated

Cancer Cell 22, 494505, October 16, 2012 2012 Elsevier Inc. 503

Cancer Cell
TSLP Overexpression Protects from Skin Cancer

JAK2: a report from the International BFM Study Group. Blood 115, 1006 1017. Hwang, C.Y., Chen, Y.J., Lin, M.W., Chen, T.J., Chu, S.Y., Chen, C.C., Lee, D.D., Chang, Y.T., Wang, W.J., and Liu, H.N. (2012). Cancer risk in patients with allergic rhinitis, asthma and atopic dermatitis: a nationwide cohort study in Taiwan. Int. J. Cancer 130, 11601167. Jamali, I., Field, E.H., Fleming, A., and Cowdery, J.S. (1992). Kinetics of antiCD4-induced T helper cell depletion and inhibition of function. Activation of T cells by the CD3 pathway inhibits anti-CD4-mediated T cell elimination and down-regulation of cell surface CD4. J. Immunol. 148, 16131619. Johansson, M., Denardo, D.G., and Coussens, L.M. (2008). Polarized immune responses differentially regulate cancer development. Immunol. Rev. 222, 145154. Kessenbrock, K., Plaks, V., and Werb, Z. (2010). Matrix metalloproteinases: regulators of the tumor microenvironment. Cell 141, 5267. Kopan, R., and Ilagan, M.X. (2009). The canonical Notch signaling pathway: unfolding the activation mechanism. Cell 137, 216233. n, O., and Emtestam, L. (2001). Incidence of Lapins, J., Ye, W., Nyre cancer among patients with hidradenitis suppurativa. Arch. Dermatol. 137, 730734. Lee, E.B., Kim, K.W., Hong, J.Y., Jee, H.M., Sohn, M.H., and Kim, K.E. (2010). Increased serum thymic stromal lymphopoietin in children with atopic dermatitis. Pediatr. Allergy Immunol. 21, e457e460. Leonard, W.J. (2002). TSLP: nally in the limelight. Nat. Immunol. 3, 605607. Li, M., Hener, P., Zhang, Z., Kato, S., Metzger, D., and Chambon, P. (2006). Topical vitamin D3 and low-calcemic analogs induce thymic stromal lymphopoietin in mouse keratinocytes and trigger an atopic dermatitis. Proc. Natl. Acad. Sci. USA 103, 1173611741. Li, T., Wen, H., Brayton, C., Laird, F.M., Ma, G., Peng, S., Placanica, L., Wu, T.C., Crain, B.J., Price, D.L., et al. (2007). Moderate reduction of gammasecretase attenuates amyloid burden and limits mechanism-based liabilities. J. Neurosci. 27, 1084910859. Mascia, F., Denning, M., Kopan, R., and Yuspa, S.H. (2012). The Black Box Illuminated: Signals and Signaling. J. Invest. Dermatol. 132, 811819. Published online December 15, 2011. http://dx.doi.org/10.1038/jid.2011.406. Morasso, M.I., and Tomic-Canic, M. (2005). Epidermal stem cells: the cradle of epidermal determination, differentiation and wound healing. Biol. Cell 97, 173183. Murphy, W.J., Kumar, V., and Bennett, M. (1987). Acute rejection of murine bone marrow allografts by natural killer cells and T cells. Differences in kinetics and target antigens recognized. J. Exp. Med. 166, 14991509. Nguyen, B.C., Lefort, K., Mandinova, A., Antonini, D., Devgan, V., Della Gatta, G., Koster, M.I., Zhang, Z., Wang, J., Tommasi di Vignano, A., et al. (2006). Cross-regulation between Notch and p63 in keratinocyte commitment to differentiation. Genes Dev. 20, 10281042. Nicolas, M., Wolfer, A., Raj, K., Kummer, J.A., Mill, P., van Noort, M., Hui, C.C., Clevers, H., Dotto, G.P., and Radtke, F. (2003). Notch1 functions as a tumor suppressor in mouse skin. Nat. Genet. 33, 416421. Olkhanud, P.B., Rochman, Y., Bodogai, M., Malchinkhuu, E., Wejksza, K., Xu, M., Gress, R.E., Hesdorffer, C., Leonard, W.J., and Biragyn, A. (2011). Thymic stromal lymphopoietin is a key mediator of breast cancer progression. J. Immunol. 186, 56565662. Pan, Y., Lin, M.H., Tian, X., Cheng, H.T., Gridley, T., Shen, J., and Kopan, R. (2004). gamma-secretase functions through Notch signaling to maintain skin appendages but is not required for their patterning or initial morphogenesis. Dev. Cell 7, 731743. Pedroza-Gonzalez, A., Xu, K., Wu, T.C., Aspord, C., Tindle, S., Marches, F., Gallegos, M., Burton, E.C., Savino, D., Hori, T., et al. (2011). Thymic stromal lymphopoietin fosters human breast tumor growth by promoting type 2 inammation. J. Exp. Med. 208, 479490. Peschon, J.J., Morrissey, P.J., Grabstein, K.H., Ramsdell, F.J., Maraskovsky, E., Gliniak, B.C., Park, L.S., Ziegler, S.F., Williams, D.E., Ware, C.B., et al.

(1994). Early lymphocyte expansion is severely impaired in interleukin 7 receptor-decient mice. J. Exp. Med. 180, 19551960. Prizment, A.E., Folsom, A.R., Cerhan, J.R., Flood, A., Ross, J.A., and Anderson, K.E. (2007). History of allergy and reduced incidence of colorectal cancer, Iowa Womens Health Study. Cancer Epidemiol. Biomarkers Prev. 16, 23572362. rez-Losada, J., Pelorosso, F.G., Mao, J.H., Nagase, Quigley, D.A., To, M.D., Pe H., Ginzinger, D.G., and Balmain, A. (2009). Genetic architecture of mouse skin inammation and tumour susceptibility. Nature 458, 505508. Rangarajan, A., Syal, R., Selvarajah, S., Chakrabarti, O., Sarin, A., and Krishna, S. (2001). Activated Notch1 signaling cooperates with papillomavirus oncogenes in transformation and generates resistance to apoptosis on matrix withdrawal through PKB/Akt. Virology 286, 2330. Rochman, I., Watanabe, N., Arima, K., Liu, Y.J., and Leonard, W.J. (2007). Cutting edge: direct action of thymic stromal lymphopoietin on activated human CD4+ T cells. J. Immunol. 178, 67206724. Rochman, Y., and Leonard, W.J. (2008). The role of thymic stromal lymphopoietin in CD8+ T cell homeostasis. J. Immunol. 181, 76997705. Rochman, Y., Spolski, R., and Leonard, W.J. (2009). New insights into the regulation of T cells by gamma(c) family cytokines. Nat. Rev. Immunol. 9, 480490. Rogers, H.W., and Unanue, E.R. (1993). Neutrophils are involved in acute, nonspecic resistance to Listeria monocytogenes in mice. Infect. Immun. 61, 50905096. Schmitt, C.A. (2011). Immunotherapy: TSLP fuels inammation in breast and pancreatic tumors. Nat. Rev. Clin. Oncol. 8, 254. Schreiber, R.D., Old, L.J., and Smyth, M.J. (2011). Cancer immunoediting: integrating immunitys roles in cancer suppression and promotion. Science 331, 15651570. Shankaran, V., Ikeda, H., Bruce, A.T., White, J.M., Swanson, P.E., Old, L.J., and Schreiber, R.D. (2001). IFNgamma and lymphocytes prevent primary tumour development and shape tumour immunogenicity. Nature 410, 1107 1111. Shochat, C., Tal, N., Bandapalli, O.R., Palmi, C., Ganmore, I., Te Kronnie, G., Cario, G., Cazzaniga, G., Kulozik, A.E., Stanulla, M., et al. (2011). Gain-offunction mutations in interleukin-7 receptor-{alpha} (IL7R) in childhood acute lymphoblastic leukemias. J Exp Med. Sneddon, J.B., and Werb, Z. (2007). Location, location, location: the cancer stem cell niche. Cell Stem Cell 1, 607611. Tanaka, J., Watanabe, N., Kido, M., Saga, K., Akamatsu, T., Nishio, A., and Chiba, T. (2009). Human TSLP and TLR3 ligands promote differentiation of Th17 cells with a central memory phenotype under Th2-polarizing conditions. Clin. Exp. Allergy 39, 89100. nez-Maza, O., Becker, N., Vajdic, C.M., Falster, M.O., de Sanjose, S., Mart Bracci, P.M., Melbye, M., Smedby, K.E., Engels, E.A., Turner, J., et al. (2009). Atopic disease and risk of non-Hodgkin lymphoma: an InterLymph pooled analysis. Cancer Res. 69, 64826489. Vesely, M.D., Kershaw, M.H., Schreiber, R.D., and Smyth, M.J. (2011). Natural innate and adaptive immunity to cancer. Annu. Rev. Immunol. 29, 235271. Wang, B., Yang, W., Wen, W., Sun, J., Su, B., Liu, B., Ma, D., Lv, D., Wen, Y., Qu, T., et al. (2010). Gamma-secretase gene mutations in familial acne inversa. Science 330, 1065. Wang, H., and Diepgen, T.L. (2005). Is atopy a protective or a risk factor for cancer? A review of epidemiological studies. Allergy 60, 10981111. Xia, X., Qian, S., Soriano, S., Wu, Y., Fletcher, A.M., Wang, X.J., Koo, E.H., Wu, X., and Zheng, H. (2001). Loss of presenilin 1 is associated with enhanced beta-catenin signaling and skin tumorigenesis. Proc. Natl. Acad. Sci. USA 98, 1086310868. Yuspa, S.H., D1ugosz, A.A., Cheng, C.K., Denning, M.F., Tennenbaum, T., Glick, A.B., and Weinberg, W.C. (1994). Role of oncogenes and tumor suppressor genes in multistage carcinogenesis. J. Invest. Dermatol. 103 (5, Suppl), 90S95S.

504 Cancer Cell 22, 494505, October 16, 2012 2012 Elsevier Inc.

Cancer Cell
TSLP Overexpression Protects from Skin Cancer

Zhang, X., and Ren, R. (1998). Bcr-Abl efciently induces a myeloproliferative disease and production of excess interleukin-3 and granulocyte-macrophage colony-stimulating factor in mice: a novel model for chronic myelogenous leukemia. Blood 92, 38293840. Zhang, Z., Hener, P., Frossard, N., Kato, S., Metzger, D., Li, M., and Chambon, P. (2009). Thymic stromal lymphopoietin overproduced by keratinocytes in

mouse skin aggravates experimental asthma. Proc. Natl. Acad. Sci. USA 106, 15361541. Ziegler, S.F. (2010). The role of thymic stromal lymphopoietin (TSLP) in allergic disorders. Curr. Opin. Immunol. 22, 795799. Ziegler, S.F., and Artis, D. (2010). Sensing the outside world: TSLP regulates barrier immunity. Nat. Immunol. 11, 289293.

Cancer Cell 22, 494505, October 16, 2012 2012 Elsevier Inc. 505

Article

Cancer Cell

Coordinated Silencing of MYC-Mediated miR-29 by HDAC3 and EZH2 as a Therapeutic Target of Histone Modication in Aggressive B-Cell Lymphomas
Xinwei Zhang,1,4,10 Xiaohong Zhao,1,10 Warren Fiskus,5 Jianhong Lin,6 Tint Lwin,1 Rekha Rao,5 Yizhuo Zhang,4 John C. Chan,7 Kai Fu,7 Victor E. Marquez,8 Selina Chen-Kiang,9 Lynn C. Moscinski,1 Edward Seto,2 William S. Dalton,1 Kenneth L. Wright,3 Eduardo Sotomayor,1 Kapil Bhalla,5 and Jianguo Tao1,*
of Malignant Hematology and Experimental Therapeutics Program Oncology Program 3Immunology Program H. Lee Moftt Cancer Center and Research Institute, Tampa, FL 33613, USA 4Department of Immunology and Malignant Hematology, Tianjin Cancer Hospital, Tianjin 300060, China 5The Drug Discovery, Delivery and Experimental Therapeutics, University of Kansas Cancer Center, Kansas City, KS 66160, USA 6Dana-Farber Cancer Institute, Boston, MA 02115, USA 7Department of Pathology, University of Nebraska Medical Center, Omaha, NE 68198, USA 8Chemical Biology Laboratory, Center for Cancer Research, National Cancer Institute, Frederick, MD 21702, USA 9Department of Pathology, Weill-Cornell Medical College, New York, NY 10065, USA 10These authors contributed equally to this work *Correspondence: jianguo.tao@moftt.org http://dx.doi.org/10.1016/j.ccr.2012.09.003
2Molecular 1Department

SUMMARY

We investigated the transcriptional and epigenetic repression of miR-29 by MYC, HDAC3, and EZH2 in mantle cell lymphoma and other MYC-associated lymphomas. We demonstrate that miR-29 is repressed by MYC through a corepressor complex with HDAC3 and EZH2. MYC contributes to EZH2 upregulation via repression of the EZH2 targeting miR-26a, and EZH2 induces MYC via inhibition of the MYC targeting miR-494 to create positive feedback. Combined inhibition of HDAC3 and EZH2 cooperatively disrupted the MYC-EZH2-miR-29 axis, resulting in restoration of miR-29 expression, downregulation of miR-29-targeted genes, and lymphoma growth suppression in vitro and in vivo. These ndings dene a MYC-mediated miRNA repression mechanism, shed light on MYC lymphomagenesis mechanisms, and reveal promising therapeutic targets for aggressive B-cell malignancies.
INTRODUCTION c-MYC (hereinafter termed MYC) is a transcription factor that promotes oncogenesis by activating and repressing its target genes that control cell growth and proliferation (Nilsson and Cleveland, 2003). MYC is deregulated in a large proportion of aggressive B-cell lymphomas. Although MYC has been described as a dening feature and the driving oncogene for Burkitt lymphoma, the signicance of MYC has also been recognized in other non-Hodgkins lymphomas (Dave et al., 2006). MYC, which has been detected in 9%14% of diffuse large B-cell lymphomas, is associated with an adverse prognosis as a result of chemoresistance and with shortened survival. In mantle cell lymphoma (MCL), increased expression of MYC has been found to be associated with poor prognosis and MCL aggressiveness (Hartmann et al., 2008). MYC overexpression has been implicated in high-grade large cell transformation in follicular and marginal zone cell lymphomas (Slack and

Signicance Aberrant miRNA expression and miRNA oncogenic and tumor suppressive functions have been extensively investigated in many tumors, including lymphoma; however, the molecular basis for miRNA dysregulation remains unknown and emerging. Our ndings of miRNA regulation by c-MYC (hereafter termed MYC) and epigenetic deacetylation by HDAC3 and trimethylation by EZH2 present a common mechanism for repression of many other tumor suppressor miRNAs. We demonstrated a MYC-miRNA-EZH2 feed-forward pathway that leads to persistent MYC and EZH2 overexpression and miR-29 repression, thus maintaining tumorigenic potential of lymphoma cells. Restoration of miR-29 expression through epigenetic drug cotreatment resulted in enhanced inhibition of oncogenic signaling pathways, lymphoma growth in vivo, and is a therapeutic target of histone modications in aggressive B-cell lymphomas.
506 Cancer Cell 22, 506523, October 16, 2012 2012 Elsevier Inc.

Cancer Cell
MYC, HDAC, EZH2, and miRNA

Gascoyne, 2011), supporting the features of MYC in sustaining aggressive transformation of lymphomas. Despite current modes of intensive chemotherapy and radiation, survival in patients with high MYC activity is dismal. It is still unclear what direct MYC-induced transcriptional changes promote cell transformation, and the therapeutics against MYC has remained elusive. Aberrant micro-RNA (miRNA) expression and miRNA oncogenic and tumor suppressive functions have been extensively investigated in many tumors, including lymphomas (Fabbri and Croce, 2011). However, the molecular basis for miRNA dysregulation remains uncharacterized and emerging (Liu et al., 2010). Our work and that of others have indicated that the miR-29 family might function as a tumor suppressor (Fabbri et al., 2007; Zhao et al., 2010). Expression of these miRNAs inhibits cell proliferation, promotes apoptosis of cancer cells, and suppresses tumorigenicity by targeting multiple oncogenes. Loss or downregulation of these miRNAs has been reported in a variety of hematopoietic and solid tumors and has been shown to be associated with high-risk chronic lymphocytic leukemia, lung cancer, invasive breast cancer, and cholangiocarcinoma (Fabbri and Croce, 2011). These observations are consistent with our recent study demonstrating that miR-29 is downregulated in aggressive MCL (Zhao et al., 2010). MYC has been recently implicated in controlling the expression of a host of miRNAs (Chang et al., 2008). The predominant consequence of activation of MYC is widespread repression of miRNA expression. Although the mechanisms by which MYC activates transcription have been extensively studied, less is known about how MYC represses transcription of target genes as well as miRNAs. It was reported that MYC repressed target genes Id2 and Gadd153 by recruitment of histone deacetylase 3 (HDAC3) (Kurland and Tansey, 2008). More recently, our study demonstrated that MYC acts as a repressor of miRNA-15a/16 by recruiting HDAC3 (Zhang et al., 2012). These ndings suggest that histone deacetylation may be involved in MYC-mediated transcriptional repression. Further evidence has shown that histone H3 lysine 27 trimethylation, which is mediated by enhancer of zeste homolog 2 (EZH2) at the promoter of the gene, leads to silencing of gene expression (Chen et al., 2005). The polycomb-repressive complex 2 (PRC2) contains three core proteins (EZH2, SUZ12, and EED), and PRC2 is a transcriptional repressor that has a crucial function in maintaining the delicate homeostatic balance between gene expression and repression, the disruption of which may lead to oncogenesis (Sparmann and van Lohuizen, 2006). The roles of HDAC and PRC2 in miRNA regulation and dysregulation are largely unknown and have been poorly dened so far. In this study, we explored the role of MYC, HDAC, and EZH2 in miR-29 repression and the contribution of miR-29 to cell survival and growth in MYC-associated lymphomas. We examined the regulation and functional roles of miRNAs, histone modications and their interplay in MYC, EZH2 overexpression, and the tumorigenic potential of lymphoma cells. Furthermore, we tested molecular targeting strategies to restore miR-29 expression and examined whether combined inhibitors of HDAC and EZH2 cooperatively increase miR-29 expression and inhibit lymphoma growth and shorten in vivo lymphoma survival.

RESULTS MYC Is Overexpressed in Aggressive MCL and Is Inversely Correlated with Expression of miR-29 We examined MYC and miR-29 expression and their correlation using puried lymphoma cells from MCL patients and normal donors. As shown in Figure 1A, compared with normal CD19+ peripheral blood lymphocytes, miR-29a-c was signicantly downregulated and MYC was signicantly overexpressed in MCL samples. Furthermore, MCLs with higher MYC expression have signicantly lower miR-29 expression. We used the P493-6 human B-cell line as a model to examine the role of MYC in miR-29 expression. P493-6 cells bear a tetracycline (tet)repressible MYC construct such that tet withdrawal results in rapid induction of MYC followed by cell proliferation. We compared expression levels of MYC and miR-29 in tet-treated (MYC-off) and untreated (MYC-on) cells and observed an inverse correlation between miR-29 and MYC expression (Figures 1B 1D). Expression of primary miR-29 (pri-miR-29a/b1 and primiR-29b2/c) and mature miR-29 was measured by quantitative reverse-transcribed polymerase chain reaction (qRT-PCR) in P493-6 cells with and without MYC expression. We found both primary miR-29 and mature miR-29 to be remarkably lower in MYC-on B cells than in MYC-off cells, whereas MYC repression after tet treatment signicantly upregulated, miR29a-c. In addition, MCL patient samples showed strong positive correlations between primary miRNAs of miR-29 and mature miR-29 expression (Figures S1AS1C available online). MYC, HDAC3, and PRC2 Are Tethered to the miR-29 Promoter Regions as a Corepressor Complex to Downregulate miR-29 Expression through Histone Deacetylation and Trimethylation We next investigated the epigenetic regulation of MYC-induced miR-29 repression through histone acetylation and methylation. We rst examined the effects of chromatin-modifying drugs on miR-29 expression in MCL and other MYC-expressing B-cell lymphomas. Using qRT-PCR, we evaluated the effects of a pan-HDAC inhibitor (vorinostat) on both primary and mature miR-29 expression in MCL (Jeko-1) and Burkitt lymphoma cells (Ramos). Figure 2A shows that vorinostat caused a dose-dependent increase in miR-29a-c expression. miR-29 induction was also observed with another HDAC inhibitor, trichostatin A (Figure S2A), suggesting an HDAC role in miR-29 gene expression and supporting that the miR-29 family members are subject to epigenetic control in lymphoma cells. We next studied the role of PRC2 in the downregulation of miR-29 since PRC2 has been shown to be recruited to gene promoters to induce histone trimethylation and gene repression. We evaluated the effects of the PRC2 inhibitor 3-deazaneplanocin A (DZNep) on miR-29 expression. Based on our previous study of using DZNep for leukemia cells (Fiskus et al., 2009), we chose the DZNep dosage range and revealed that DZNep resultes in a dose-dependent decrease in the protein expression of EZH2 and SUZ12 (Figure 2B; Figure S2B) and caused a dose-dependent increase in pri-miR-29a/b1, pri-miR-29b2/c, and mature miR-29 expression in these lymphoma cell lines (Figure 2C; Figure S2C). Overall, the above observations implied that both HDAC and PRC2 are involved in miR-29 expression.
Cancer Cell 22, 506523, October 16, 2012 2012 Elsevier Inc. 507

Cancer Cell
MYC, HDAC, EZH2, and miRNA

A
p p
p

p p
p
p p

Figure 1. MYC Is Overexpressed in Aggressive MCLs and Is Inversely Correlated with Expression of miR-29
(A) miR-29a-c expression inversely correlated with MYC expression in primary MCL cells. Expression levels of miR-29a-c and MYC in normal B lymphocytes and primary MCL samples were measured by qRT-PCR. The high MYC group is dened as those samples in the upper quartile (25%) of MYC expression, while all others are placed in the low MYC group for the patient samples. (BD) Expression of MYC and miR29a-c in tet-treated (MYC turn-off or MYC-off) and untreated (MYC turn-on or MYC-on) P493-6 cells. (B) Western blot shows MYC expression levels in MYC-off P493-6 cells treated with tet and in MYC-on P493-6 cells after removal of tet for indicated times. (C) Pri-miR-29 expression levels in MYC-off P493-6 cells treated with tet and in MYC-on P493-6 cells after removal of tet for indicated times. (D) Mature miR-29 and MYC expression levels in MYC-off P493-6 cells treated with tet and in MYC-on P493-6 cells after removal of tet for indicated times. Pri-miR-29 level was normalized to GAPDH, and mature miR-29 expression was normalized to RNU44. Results in (B) are representative of three independent experiments. Results in (C) and (D) are means SD from at least three biological replicates. See also Figure S1.

We explored whether MYC, HDAC3, and/or EZH2 act together to be involved in miR-29 expression in MYC-expressing lymphoma cells. The role of MYC and HDAC3 in the transcriptional regulation of miR-29 gene expression was rst examined by depleting the expression of MYC and HDAC3, respectively,
508 Cancer Cell 22, 506523, October 16, 2012 2012 Elsevier Inc.

with siRNA. Expression levels of primary and mature miR-29 were analyzed after MYC or HDAC3 was knocked down. In agreement with our earlier results with vorinostat, knockdown of HDAC3 signicantly enhanced both primary and mature miR-29 gene expression (Figure 2D). Moreover, knockdown of

Cancer Cell
MYC, HDAC, EZH2, and miRNA

MYC also markedly increased miR-29 gene expression and decreased EZH2 expression (Figure 2D). When we assessed the role of PRC2 in MYC-mediated miR-29 repression, we found that depletion of EZH2 and/or SUZ12 using siRNAs also signicantly increased miR-29 gene expression and decreased MYC expression (Figure 2E), further supporting the role of EZH2/ PRC2 in miR-29 expression. Next, we examined the miR-29a/b1 and miR-29b2/c gene promoter regions for transcription factor binding sites and identied three highly conserved MYC binding sitesS1, S2, and S3in a region 5 kb upstream and in the rst intron of both human miR-29a/b1 and miR-29b2/c (Figure 3A). We investigated whether HDAC3 and EZH2/PRC2 could be recruited to the miR-29 promoters by MYC and whether HDAC3 and EZH2 mediated MYC-induced miR-29 repression using chromatin immunoprecipitation (ChIP) assays. We used primers located within the miR-29a/b1 and miR-29b2/c proximal promoter regions of MYC binding sites and revealed that antibodies against both MYC and HDAC3 efciently immunoprecipitated the miR-29 promoter regions (Figures 3A and 3B). In addition, we found that site S3 of miR-29a/b1 and sites S2 and S3 of miR-29b2/c carry binding sites for both MYC and HDAC3, indicating that both MYC and HDAC3 can bind to the miR-29 promoters (Figure 3B; Figure S3A). These bindings are specic and MYC dependent, since no signal was detected at the miR-29 distal promoter (site S4) and no HDAC3 binding was detected when MYC was not bound at miR-29 promoters. These ndings implicate the role of MYC in recruiting HDAC3 and suggest that HDAC3-mediated histone deacetylation might contribute to MYC-induced miR-29 gene repression. To conrm the requirement of MYC for HDAC3 binding, P493-6 cells were used to manipulate MYC expression levels. ChIP assays revealed HDAC3 binding in MYC-on and lack of binding in MYC-off P493-6 cells, supporting the recruitment role of MYC (Figure 3C). We further assessed the role of EZH2 in MYC-mediated miR-29 repression and investigated whether similar regulation patterns occur through recruitment of EZH2 and SUZ12 on miR-29a/b1 and miR-29b2/c promoters. ChIP assay with anti-EZH2 and anti-SUZ12 antibody showed that both EZH2 and SUZ12 directly bound to the miR-29 promoters in MYC-on but not in MYC-off P493-6 lymphocytes (Figure 3C) and was further validated in Burkitt and MCL cell lines (Figure 3D; Figure S3B). The loss of binding of these corepressors detected by ChIP is not due to loss of the proteins from the cell but most likely due to absence of MYC since 24 hr tet treatment resulted in loss of MYC but no change in EZH2, SUZ12, or HDAC3 expression in P493 cells (Figure 3C, insert). Of note, the EZH2 and SUZ12 binding sites correspond to the MYC and HDAC3 binding sites, supporting the role of MYC in EZH2 and SUZ12 recruitment and the role of PRC2 in silencing miR-29 expression. The variation of degree and site of MYC binding on miR-29a/b1 (site S3 only) and miR-29b2/c promoters (sites S2 and S3) likely contributes to different sensitivity of miR-29a/b1 and miR-29b2/c to the treatment of HDAC and EZH2 inhibitors. Furthermore, inhibition of PRC2 with DZNep degraded EZH2 and SUZ12 and decreased EZH2 and SUZ12 binding (Figure 3D). We next performed ChIP assay to validate the MYC, EZH2 and HDAC3 binding to miR29 promoters in primary lymphoma samples. Four high-MYC samples of two blastic MCLs, one Burkitt and one Burkitt-like

(double-hit) lymphoma, and two low-MYC indolent MCLs were chosen from our clinical samples and were used in this experiment. MYC expression levels in these samples were conrmed by uorescence in situ hybridization (FISH) and immunohistochemical stains (data not shown). Figure 3E reveals consistent enrichment of MYC and, to a lesser extent, HDAC3 and EZH2 in miR-29 promoter regions in MYC-associated lymphomas and supports the ndings that these interactions are operative in primary lymphoma cells. Taken together, these results conrm that MYC is required and a signicant mediator of EZH2-mediated miR-29 repression, suggesting that HDAC3 and EZH2 have coordinated effects on miRNAs such as miR-29 expression in MYC-associated lymphomas. To test whether MYC binding is functional, we generated luciferase reporter constructs carrying the two alternative promoters of miR-29a/b1 and miR-29b2/c at site S3 for miR29a/b1 and sites S2 and S3 for miR29b2/c and their mutated types (M2 and M3). The mutants were constructed to harbor mutations in the MYC binding site (E-box). Both wild-type and mutant plasmids (E-box mutants) were then transfected into P493-6 and 293T cells, and luciferase activity was measured (Figure 3F; Figure S3C). We found luciferase activities of wild-type miR29a/b1 and miR-29b2/c promoters to be signicantly repressed by MYC overexpression. Furthermore, knockdown of HDAC3 reversed MYC-mediated repression, supporting that HDAC3 is involved in MYC-driven miRNA repression. Compared with wild-type promoters, luciferase activity of mutated-type promoters was not signicantly changed by MYC overexpression and HDAC3 knockdown. Similarly, knockdown of EZH2 reversed MYC-mediated repression in wild-type but not in mutant miR-29 promoters (Figure 3F). MYC-mediated repression was not observed in M3 of the miR-29a/b1 promoter and M2 of the miR-29b2/c promoter. This is likely attributed to the dominant function of MYC binding in site S3 of miR-29a/b1 and site S2 of miR-29b2/c promoters. These results are in line with those of ChIP experiments showing the strongest binding of MYC in S3 of miR-29a/b1 and in S2 of miR-b2/c promoters (Figure 3C). Overall, these data show that both miR-29a/b1 and miR-29b2/c loci contain MYC-binding regions that are under negative control by HDAC3 and EZH2 and that histone hypoacetylation and trimethylation contribute to MYC-induced miR-29 repression. We further performed ChIP analysis to probe acetylated histone 4 (Ac-H4), trimethylated histone 3 (Me3-H3K27), and RNA polymerase II binding to miR-29 promoters. We rst revealed that histone hypoacetylation and trimethylation are dependent on the presence of MYC since enrichment of Ac-H4 was signicantly increased and Me3-H3K27 is signicantly decreased in MYC-off P493-6 cells (Figure S3D). This study also revealed that accumulation of RNA polymerase II, a hallmark of active transcription, is tightly controlled by MYC. In agreement with the epigenetic silencing effect of HDAC3 and EZH2, HDAC3 knockdown and EZH2 inhibition, respectively, increased Ac-H4 and decreased Me3-H3K27 at the miR-29 promoters (Figures S3E and S3F). Of note, increased recruitment of RNA polymerase II was also observed. These results support the nding that depletion of MYC leads to reduced recruitment of HDAC3 and EZH2 and results in increased histone acetylation, decreased H3K27 trimethylation, and RNA polymerase II recruitment.
Cancer Cell 22, 506523, October 16, 2012 2012 Elsevier Inc. 509

Cancer Cell
MYC, HDAC, EZH2, and miRNA

510 Cancer Cell 22, 506523, October 16, 2012 2012 Elsevier Inc.

Cancer Cell
MYC, HDAC, EZH2, and miRNA

We further investigated whether PRC2 and HDAC3 form a corepressor complex with MYC to repress miR-29 expression using coimmunoprecipitation (co-IP) assays. First, 293T cells were cotransfected with vectors expressing FLAG-tagged fulllength specic HDAC3 and/or with full-length MYC. When MYC and HDAC3 were cotransfected in 293T cells, the existence of MYC, HDAC3, and SUZ12, but not EZH2, was detected in the immunoprecipitates obtained with an antibody against HDAC3, and the existence of MYC and HDAC3, but not SUZ12 and EZH2, was detected in immunoprecipitates obtained with an antibody against MYC. These results indicate that MYC coimmunoprecipitated with HDAC3 and that HDAC3 coimmunoprecipitated with MYC as well as SUZ12. Next, we asked whether endogenous MYC-HDAC3-PRC2 interaction also occurred in MYC-associated lymphoma cells. Having recently demonstrated that MYC and HDAC3 formed a coimmunoprecipitate complex to regulate the miRNA expression (Zhang et al., 2012), we further examined the interaction between HDAC3 and SUZ12 in P493-6 cells. As shown in Figure 4B, cell lysates immunoprecipitated with an HDAC3specic antibody contained HDAC3 and SUZ12. The reverse endogenous coimmunoprecipitates of HDAC3 and SUZ12 with SUZ12 antibody was also demonstrated in Jeko-1 and P493-6 cells (Figures 4B and 4C). Third, we further explored how MYC interacted with SUZ12 and EZH2 by using MYC-on and MYC-off P493-6 cells. In MYC-on P493-6 cells, strong HDAC3, weak SUZ12, and no EZH2 were coimmunoprecipitated with MYC antibody; strong SUZ12, moderate EZH2, and MYC were coimmunoprecipitated with HDAC3 antibody; strong EZH2, moderate HDAC3, and weak MYC were coimmunoprecipitated with SUZ12 antibody; and strong SUZ12, weak HDAC3, and no MYC were coimmunoprecipitated with EZH2 antibody. In contrast, in MYC-off P493-6 cells, there was no endogenous co-IP of HDAC3 and SUZ12 with MYC detected and relatively low levels of interaction of HDAC3 with SUZ12 and EZH2 (Figure 4C). Overall, these results suggest that SUZ12 and EZH2 interact with HDAC3 and MYC to form a multimolecular complex. These components interact in a linear fashion, and HDAC3 bridges the interaction between MYC and SUZ12/EZH2. To prove this, we depleted HDAC3 and tested whether this would disrupt interaction between MYC and SUZ12/EZH2 in P493-6 cells. As shown in Figure 4D, SUZ12 was not detected in MYC immunoprecipitate and MYC was not detected in SUZ12 immunoprecipitate, implicating that HDAC3 bridges the interaction between MYC and SUZ12/EZH2. These data, in conjunction with results from luciferase reporter assay, support the cooperative function of HDAC3 and EZH2 as a corepressor complex in repressing miR-29 expression.

miR-29 Is Required for MYC-Mediated Oncogenic Activity by Targeting IGF-1R and CDK6 Pathways in MCL and Other MYC-Expressing B-Cell Lymphomas We investigated whether downregulation of miR-29 is necessary for cellular transformation induced by oncogenic MYC overexpression. In addition to CDK6, our bioinformatic analysis also revealed IGF-1R as a potential target of miR-29. Increased expression of miR-29 signicantly downregulated IGR-1R (Figure 5A; Figure S4A). The relative luciferase activity of the wild-type construct of IGF-1R 30 -UTR was reduced by overexpression of miR-29c and was increased when miR-29c was knocked down, whereas such effects of miR-29c on luciferase activity were not observed with the mutant construct of IGF-1R 30 -UTR (Figure 5A). These ndings support a direct and specic interaction of miR-29c on IGF-1R 30 -UTR. To conrm the relevance of the expression of IGF-1R and the relationship between miR-29 and IGF-1R, the expression of miR-29 and IGF-1R expression were assessed in a set of primary MCL tissues and normal B lymphocytes. An inverse correlation of miR-29 and IGF-1R protein expression was observed in all MCL samples by using Pearson coefcient, and correlation coefcients were calculated identifying IGF-1R as a miR-29 target in addition to the previously demonstrated CDK6 in MCL (Zhao et al., 2010; Figure 5B; Figures S4B and S4C). Given the oncogenic feature of MYC and regulatory role of MYC in miR-29 expression, we postulated that MYC-mediated miR-29 repression and subsequent miR-29 target changes contribute to MYC-driven lymphoma cell growth and proliferation. We, therefore, tested whether MYC upregulates miR-29 targets (CDK6 and IGF-1R expression) in MYC-on and MYCoff P493-6 cells. With MYC turn-on, protein levels of CDK6 and IGF-1R were signicantly increased, whereas they were signicantly downregulated with MYC turn-off (Figure 5C). In contrast, mRNA levels of CDK6 and IGF-1R were not signicantly inuenced by MYC (Figure S4D). These data suggest a posttranscriptional mechanism of CDK6 and IGF-1R expression and are in line with MYC-driven miR-29-mediated regulation of CDK6 and IGF-1R expression. Of note, when MYC is turned off (MYC-off cells), IGF-1R declines at a faster rate than CDK6. This may be related to differences in the mRNA and/or protein half-life of these two proteins. We further assessed whether miR-29 mediated MYC-driven CDK6 and IGF-1R induction and cell growth in MYC-expressing lymphoma cells. In P493-6 cells, the ectopic forced expression of miR-29 abolished MYC-induced CDK6 and IGF-1R expression, and miR-29 knockdown blocked MYC-off-mediated CDK6 and IGF-1R repression (Figure 5D; Figure S4A). Furthermore, knockdown of CDK6 and IGF-1R induced signicant inhibition of cell growth and colony formation, and the combined inhibition of CDK6 and IGF-1R resulted

Figure 2. MiR-29 Family Is Coregulated by HDAC3 and PRC2


(A) Vorinostat treatment for 48 hr dose-dependently increased primary and mature miR-29 expression levels in Jeko-1 and Ramos cells. (B) DZNep treatment for 48 hr downregulated EZH2 and SUZ12 in Jeko-1, Ramos, and HBL2 cells. (C) DZNep treatment for 48 hr dose-dependently increased primary and mature miR-29 expression in Jeko-1, Ramos, and HBL2 cells. (D) Knockdown of MYC or HDAC3 by siRNAs increased pri-miR-29 and mature miR-29 expression levels in Mino, Jeko-1, and Ramos cells. mRNA and miRNA expression levels of cells treated with siCtrl were arbitrarily set as 1. (E) Knockdown of EZH2 and SUZ12 by their siRNAs increased miR-29a-c gene expression in Mino and Ramos cells. Results in (A) through (E) are means SD from at least 3 biological replicates. See also Figure S2.

Cancer Cell 22, 506523, October 16, 2012 2012 Elsevier Inc. 511

Cancer Cell
MYC, HDAC, EZH2, and miRNA

512 Cancer Cell 22, 506523, October 16, 2012 2012 Elsevier Inc.

Cancer Cell
MYC, HDAC, EZH2, and miRNA

in a more marked inhibition of cell growth and colony formation (Figure 5E; Figure S4E). Accordingly, miR-29 overexpression as well as MYC knockdown signicantly abrogated lymphoma colony formation capacity (Figure 5F; Figures S4F and S4G). MYC-miR-26a-EZH2-miR-494 Positive Feedback Loop Sustains MYC Activity and miR-29 Repression in MCL and Other Aggressive B-Cell Lymphomas Accumulating evidence has indicated MYC-dependent regulation of EZH2, with further evidence revealing the ability of EZH2 to induce MYC expression (Sander et al., 2008; Lu et al., 2011). We speculated that a feedback loop existed between MYC and EZH2, thereby maintaining MYC overexpression and miR-29 repression in MYC-associated lymphomas. We thus explored the interaction between MYC and EZH2 and examined the role of this circuitry in sustaining miR-29 repression. First, we tested whether MYC stimulates EZH2 expression by repression of its negative regulator miRNAs. Since EZH2 was identied by TargetScan as a potential miR-26a target and was recently experimentally validated (Sander et al., 2008), we examined the effects of miR-26a on EZH2 expression. Overexpression of miR-26a reduced EZH2 and MYC protein abundance in Jeko-1 and Mino cells, and ectopic expression of miR-26a inhibited EZH2 30 -UTR luciferase reporter activity (Figure 6A), conrming that miR-26a regulates EZH2. Given that miR-26a is a reported MYC-regulated miRNA, we next used the P493-6 cells to conrm and explore the mechanism by which MYC induces EZH2 expression. Expression of miR-26a in MYC-on P493-6 cells was signicantly lower than that shown in MYC-off cells (Figure 6B). Furthermore, the effects of MYC on EZH2 expression were examined, revealing that mRNA levels of EZH2 and SUZ12 were not changed, whereas protein levels were signicantly increased in MYC-on cells and decreased in MYC-off cells (Figure 6B; Figure S5A). This result implies that MYC regulated EZH2 via posttranscriptional regulation. Moreover, ectopic expression of miR-26a blocked MYC-induced EZH2 expression in MYC-on P493-6 cells. To substantiate that miR-26a expression is responsible for MYC-induced EZH2 change, we inhibited miR-26a by using anti-miR-26a and revealed increased EZH2 expression in MYC-off P493-6 cell (Figure 6C), further supporting

the role of miR-26a in MYC-regulated EZH2 expression. We next tested whether EZH2 stimulates MYC expression. Figure 6D shows that inhibition of EZH2 by using DZNep and shRNA against EZH2 signicantly decreased MYC expression, substantiating the regulatory role of EZH2 in MYC expression and implying a positive feedback loop of MYC and EZH2. We reasoned that EZH2 induces MYC expression through repression of MYC-repressing miRNAs. Thus, we next explored the EZH2-regulated miRNAs by examining the effects of EZH2 inhibition on the expression of miRNAs. miRNA microarray was performed and the expression prole from Jeko-1 cells after 72 hr DZNep (2 mM) treatment was determined (Figure S5B). We identied a set of miRNAs that were upregulated by DZNep, were downregulated by PRC2, and are predicted to target MYC (Figures 6E and 6F). To further test whether these miRNAs target the MYC 30 -UTR directly, we cloned the full length of MYC 30 -UTR and constructed a luciferase reporter plasmid (p-miRMYC-30 -UTR-WT). The plasmid was cotransfected into 293T cells, with each of the aforementioned pre-miRNAs and luciferase activity measured. Figure S5C shows that the luciferase activity of wild-type MYC reporter were reduced by overexpression of miR-135, miR-200, and miR-374, as well as most noticeably decreased by miR-494 overexpression. With TargetScan predicting that MYC 30 -UTR contains two miR-494-binding sites, we subsequently mutated the miR-494 binding sites in MYC 30 -UTR to test whether miR-494 specically targets MYC 30 -UTR. As revealed in Figure 6F, the mutation abolished the suppressive effect of miR-494 on the luciferase reporter activity. These results demonstrated that the miR-494 specically and directly targeted the MYC gene. To determine whether miR-494 is required and a signicant mediator for EZH2mediated MYC induction, we performed qRT-PCR and validated that miR-494 was upregulated by EZH2 inhibition through DZNep treatment and shEZH2 or siEZH2 (Figures S5D and S5E). To further conrm that miR-494 is directly regulated by EZH2, ChIP assay was performed and showed the direct EZH2 binding to miR-494 promoter regions. Furthermore, this binding is inhibited by the depletion of EZH2 through DZNep treatment (Figure S5F). We next showed that overexpression of miR-494 downregulated MYC protein level (Figure 6G). Accordingly,

Figure 3. MYC Recruits HDAC3 and PRC2 to miR-29 Promoters to Repress the miR-29 Transcription through Histone Deacetylation and Trimethylation
(A) Schematic diagram showing location of MYC-binding sites of pri-miR-29a/b1 and pri-miR-29b2/c regulatory region. Sites S1, S2, and S3 represent MYC-binding site, which has an E-box sequence. S4 was used as negative control and is located in the intron 4 of pri-miR-29b2/c and without E-box in this region. Both pri-miR-29s are highly conserved in their putative promoter region and in the pre-miR-29 stem sequences, encoded in the last intron (pre-miR-29a/b1) on chr.7q32.3 and the last exon (pre-miR-29b2/c) on chr.1q32.2, respectively. (B) ChIP assay showing MYC and HDAC3 enrichment on pri-miR-29a/b1 and pri-miR-29b2/c promoters. ChIP assay was performed using MYC or HDAC3 antibody to detect binding on S1S3 regions of pri-miR-29a/b1 and pri-miR-29b2/c promoters, and S4 was used as a negative control. Percent input was calculated with 2(Ct [1% of input] Ct [ChIP]). Ct, cycle threshold. (C) ChIP assay showing MYC, HDAC3, EZH2, and SUZ12 enrichment on pri-miR-29a/b1 and pri-miR-29b2/c promoters and dependence of HDAC3 and EZH2/ SUZ12 binding on MYC in P493-6 cells with or without 24 hr tet treatment, Inserts, western blots showing protein level of MYC, HDAC3, and EZH2/SUZ12 in MYC-on and MYC-off (24 hr tet treatment) P493-6 cells. (D) ChIP assay showing EZH2 and SUZ12 enrichment on pri-miR-29a/b1 and pri-miR-29b2/c with or without DZNep treatment. (E) ChIP assay showing MYC, HDAC3, EZH2, and SUZ12 enrichment on pri-miR-29a/b1 and pri-miR-29b2/c promoters in primary lymphoma samples with high MYC expression (blastic MCLs, Burkitt, or Burkitt-like lymphomas) and no enrichment in primary samples with low MYC expression (indolent MCLs). (F) Schematic diagram of pri-miR-29a/b1 and pri-miR-29b2/c promoter luciferase reporter. Solid boxes represent point mutation of E-Box. P493-6 cells were transfected with either wild-type or mutants (M) of pri-miR-29a/b1 or pri-miR-29b2/c promoter luciferase reporter, together with siHDAC3, siEZH2, or nontargeting siRNA. The luciferase activity is normalized to b-galactosidase. Results are means SD from three biological replicates. For ChIP assays, IgG was used as a negative control. In (B) through (F), results are means SD from at least three biological replicates. Insets, western blots showing protein level. See also Figure S3.

Cancer Cell 22, 506523, October 16, 2012 2012 Elsevier Inc. 513

Cancer Cell
MYC, HDAC, EZH2, and miRNA

Figure 4. HDAC3 Bridges the Interaction between MYC and PRC2 to Form a Corepressor Complex
(A) 293T cells were transfected with MYC plasmid or FLAG-HDAC3 plasmid or cotransfected with MYC plasmid and FLAG-HDAC3 plasmid. The whole cell lysates were immunoprecipitated using an antibody against MYC, HDAC3, and control IgG, followed by western blot with an antibody against MYC, FLAG, SUZ12, and EZH2. (B) Reciprocal co-IP showing endogenous co-IP of HDAC3 and SUZ12. Whole cell extracts of Jeko-1 cells were subjected to IP with anti-HDAC3 antibody followed by western blotting for SUZ12, and similar whole cells extracts were subjected to IP with anti-SUZ12, followed by western blotting with anti-HDAC3. (C) Co-IP of MYC, HDAC3, and SUZ12/EZH2 in MYC-on and MYC-off P493-6 cells. Cell lysates of P493-6 with and without tet treatment were immunoprecipitated with MYC, HDAC3, SUZ12, EZH2, and control IgG, respectively, followed by western blotting with an antibody against MYC, HDAC3, SUZ12, and EZH2. (D) HDAC3-mediated interaction between MYC and SUZ12/EZH2. P493-6 (MYC-on) cells were transfected with HDAC3 siRNA or nontargeting siRNA to knock down HDAC3, and co-IP experiments were performed to evaluate interaction between MYC and SUZ12/EZH2. In (A) through (D), input is equivalent to 10% of the lysate used for the co-IP. Results are representative of three independent experiments.

downregulation of EZH2 was also observed, supporting the presence of a MYC-miR-26a-EZH2-miR-494 feed-forward circuit sustaining MYC activity and miR-29 repression. To address whether the above observations in MCL and Ramos cell lines are relevant to other aggressive cell lines and primary lymphoma cells, we examined the relationship between MYC, EZH2, and miR-26a as well as miR-29 expression levels in MYC-expressing lymphoma cell lines and primary MYCexpressing lymphomas. The cell lines included two transformed large B lymphoma cell lines (SUDHL4, SUDHL10); the EpsteinBarr virus (EBV)-associated lymphoma cell line SKW6.4; aggressive MCL cell lines Jeko-1, Mino-1, HBL-2, NCEB-1, Rec-1, and Z138c; and Burkitt lymphoma cell lines Raji and Ramos. The primary lymphomas included Burkitt lymphomas, high-grade transformed diffuse large cell lymphomas, and MCLs. For comparison, we also included normal control B lymphocytes as control, with MYC-on and MYC-off P493-6 cells as positive
514 Cancer Cell 22, 506523, October 16, 2012 2012 Elsevier Inc.

and negative cell lines. In line with our hypothesis, low expression levels of miR-29 family and miR-26a were correlated with high expression of MYC and EZH2, and correlation coefcients were calculated in these cell lines and primary samples by using Pearson coefcient (Figures 7A7D). When compared with normal control B lymphocytes, MYC expression was positively correlated with EZH2 expression in these primary samples. Collectively, these observations provide EZH2 and HDAC3 as potential therapeutic targets for aggressive B-cell lymphomas. Combined Inhibitors of HDAC and EZH2 Cooperatively Derepressed miR-29 and Suppressed Tumor Growth In Vitro and In Vivo in MCL and Other Aggressive B-Cell Lymphomas In light of the importance of low or absent expression of miR-29 in MCL aggressive progression and the ability of miR-29 expression to inhibit tumor cell growth, reactivation of miR-29

Cancer Cell
MYC, HDAC, EZH2, and miRNA

C F

Cancer Cell 22, 506523, October 16, 2012 2012 Elsevier Inc. 515

Cancer Cell
MYC, HDAC, EZH2, and miRNA

represents a promising therapeutic approach for this tumor type. Given that HDAC3 and EZH2 converged at miR-29 promoters to repress miR-29 expression and suppression of CDK6 and IGF-1R pathways by miR-29, we next tested whether inhibition of HDAC and EZH2 cooperatively restored miR-29 expression and subsequently inhibited CDK6 and IGF-1R to block the clonenogenic growth in soft agar and lymphoma growth in vivo. We also asked whether combined inhibitors of HDAC and PRC2 are more effective in induction of miR-29 expression, suppression of CDK6 and IGF-1R, and tumorigenicity in vivo. Compared with each agent alone, cotreatment with vorinostat and DZNep induced signicantly higher expression of pri-miR-29a/b1, pri-miR-29b2/c, and mature miR-29 than each agent alone in HBL2, Ramos, and Mino, as well as Z138c MCL cells (Figure 8A; Figures S6AS6D). Both DZNep and vorinostat resulted in enhanced inhibition of colony formation with corresponding downregulation of CDK6 and IGF-1R in HBL2 and Z138c cells (Figures 8B and 8C; Figure S6B). Next, we compared the effects of DZNep and/or vorinostat on the viability of transformed and nontransformed lymphocytes by using P493-6 cells. Figure 8D demonstrates that exposure to DZNep or vorinostat induced more loss of viability in MYCon than in MYC-off P493-6 cells. Finally, we determined whether the combination of DZNep and vorinostat would also exert increased in vivo antilymphoma activity. Figures 8E and 8F show that cotreatment with DZNep and vorinostat inhibits tumor growth and signicantly improves survival of nonobese diabetic/ severe combined immunodecient (NOD/SCID) mice bearing lymphoma xenografts. Lymphoma size was remarkably reduced and survival of NOD/SCID mice with lymphoma was signicantly higher when they were treated with DZNep plus vorinostat than when treated with vorinostat, DZNep, or vehicle alone. Furthermore, to conrm that the in vivo targets of these inhibitors were inhibited, western blot was performed and revealed that vorinostat and/or DZNep treatment resulted in signicant downregulation of EZH2, SUZ12 and downstream target IGF-1R, as well as MYC from harvested lymphoma tissues (Figure S6E). In addition, to validate the direct role of corepressors EZH2 and HDAC3 in lymphoma formation in vivo, two independent genetic approaches were used. HBL2 cells were rst transfected with siRNAs or shRNAs against EZH2 or HDAC3 to deplete their expression, and, subsequently, these cells were applied to an in vivo lymphoma formation experiment as described in Figure 8E. Figures S6F and S6G conrmed that HDAC3 or EZH2 siRNA or shRNA knocked down EZH2 or HDAC3, respectively, and signicantly abolished lymphoma growth in vivo supporting

the role of EZH2 and HDAC3 in lymphoma formation. To support that the tumor inhibition is due to decreased proliferation, the proliferation status of the tumor cells in shEZH2- and shHDAC3-treated HBL2 xenografts was measured by using proliferation marker genes Ki-67 and PCNA. Figure S6H shows that the Ki-67 and PCNA genes were indeed signicantly decreased in shEZH2 and shHDAC3 groups when compared to shCtrl group, indicating that the tumor suppression by shEZH2/HDAC3 is at least partially through proliferation inhibition. Taken together, MYC-mediated miR-29 repression through coordinated epigenetic silencing of HDAC3 and EZH2 is a important therapeutic target of histone modications in aggressive B-cell lymphomas. DISCUSSION This study was undertaken to investigate (1) the potential interplay between MYC and histone modiers HDAC3 and EZH2 and their role in miR-29 gene repression, (2) the role of the miR-29 family and their downstream targets in MYC-driven oncogenesis, (3) the underlying mechanism of persistent MYC activation in these aggressive lymphomas through a MYC-miRNA-EZH2 positive feedback loop, and (4) whether HDAC3 and EZH2 cooperatively regulate miR-29 expression and, accordingly, whether inhibitors of HDAC and EZH2 restore expression of miR-29 in MYC-transformed B lymphoma cells to signicantly inhibit tumorigenesis ex vivo and in vivo. Our ndings indicate that miR-29 repression is a result of MYC/HDAC3 and EZH2 interaction and contributes to aggressive clinical outcome of MYC-associated lymphomas. Results of this study led to the identication of a model for interplay between MYC, HDAC3, PRC2, and miRNAs and their contribution to MYC-associated lymphomagenesis and HDAC3/EZH2/miR-29 as signicant therapeutic targets for aggressive lymphomas. We showed that MYC, HDAC3, and PRC2 form a repressive complex tethered to miR-29 promoter elements to epigenetically repress miR-29 transcription in MYC-expressing lymphoma cells. Subsequent miR-29 downregulation resulted in induction of CDK6 and IGF-1R and mediated MYC-driven lymphomagenesis shown in Figure 8G. Furthermore, we demonstrated that MYC contributed to the upregulation of EZH2 via repressing EZH2-targeting miR-26a and that EZH2, in turn, induced MYC expression via MYC-targeting miR-494, thereby generating a positive feedback loop to ensure persistent high protein levels of MYC and EZH2 and further repression of miR-29, which could be involved in maintaining the malignant phenotype. In addition

Figure 5. miR-29 Is Required for MYC-Mediated Oncogenic Activity by Targeting IGF-1R and CDK6 Pathways
(A) IGF-1R is a direct target of miR-29. Overexpression of miR-29a-c downregulates IGF-1R expression and reduces luciferase activity of wild-type IGF-1R-30 UTR reporter (IGF-1R-WT) but not mutated IGF-1R-30 UTR reporter (IGF-1R-M). (B) miR-29 level is reversely correlated with IGF-1R protein expression of MCL patient samples. (C) IGF-1R and CDK6 expression in MYC-on and MYC-off P493-6 cells. (D) Overexpression of miR-29 after 48 hr pre-miR-29a-c transfection abolished MYC-induced CDK6 and IGF-1R expression and knockdown of miR-29 by anti-miR-29s (pool of anti-miR-29a-c) transfection blocked MYC-off-induced CDK6 and IGF-1R repression. (E) Knockdown of IGF1R and CDK6 by their siRNAs inhibits lymphoma cell survival measured by MTT assay and colony formation assay in HBL2 cells after transfection with siIGF-1R and siCDK6 or control siRNA. Micrographs show the appearance of colonies in methycellulose gels at low power. (F) Overexpression of miR-29 decreases the colony formation. The numbers of tumor colonies were enumerated microscopically after an incubation of 2 weeks. Results are representative of three independent experiments or means SD from at least three biological replicates. See also Figure S4.

516 Cancer Cell 22, 506523, October 16, 2012 2012 Elsevier Inc.

Cancer Cell
MYC, HDAC, EZH2, and miRNA

Myc-3'UTR

G D

Cancer Cell 22, 506523, October 16, 2012 2012 Elsevier Inc. 517

Cancer Cell
MYC, HDAC, EZH2, and miRNA

to miR-494 other miRNAs such as miR-135a, miR-186, and miR-200c are also regulated by EZH2 and, in turn, may cooperatively regulate MYC expression. Treatment with pan-HDAC inhibitor vorinostat, EZH2 inhibitor DZNep, and their specic siRNAs disrupt the MYC-miRNA-EZH2 regulatory circuitry, resulting in enhanced restoration of miR-29 expression, downregulation of miR-29 target genes CDK6 and IGF-1R, and suppression of lymphoma cell growth (Figures S6D and S6F). Moreover, several other tumor suppressors and oncogenes such as TCL-1 and MCL1 (Pekarsky et al., 2006; Mott et al., 2007) are also regulated by miR-29 and may contribute to miR-29-mediated oncogenesis. On the other hand, a recent study implicated that miR29 can function as an oncogene in indolent chronic lymphocytic leukemia (CLL), suggesting that miR-29 can function as either a tumor suppressor or an oncogene depending on the cellular context (Pekarsky et al., 2006). Our ndings indicated that miR-29 is a tumor suppressor in aggressive MCL and revealed critical mechanisms for MYC-driven miRNA suppression and rational therapeutic targets of histone modications in aggressive B-cell malignancies. These data also indicate that MYCdriven miR-29 repression through recruitment of HDAC3 and/ or EZH2 could be a generic mechanism for miRNA silencing in aggressive B-cell lymphomas. The MYC-driven miRNA repression may underlie the molecular mechanism for lymphoma aggressive transformation and can be epigenetically targeted through manipulation of histone modications. We identied a regulatory element (site S3) located 5 kb upstream from the miR-29a/b1 and two regulatory elements (sites S2 and S3) located 5 kb upstream from tthe miR-29b2/c cluster. These elements contain a MYC-binding site (or sites) that also associate with the transcriptional repressor factors EZH2 and HDAC3. Co-IP assays revealed that MYC coimmunoprecipitates with HDAC3 and HDAC3 with EZH2, likely through SUZ12 to form a MYC-HDAC3-PRC2 complex. These ndings support the notion that MYC repressed miR-29a/b1 and miR29b2/c through recruitment and interaction with HDAC3 and PRC2 as a corepressor complex. Given that miR-29a/b1 and miR-29b2/c are located in the different chromosomes, have different promoter regions and MYC-binding sites, miR-29 transcripts indeed respond differently to the presence or absence of MYC, EZH2, and HDAC3 as shown in Figures 1 and 2. Furthermore, ChIP analysis demonstrated that MYC, HDAC3, and PRC2 colocalize to the promoters of the miR-29 cluster genes and that HDAC3 and EZH2/PRC2 binding to miR-29 promoters was MYC dependent, supporting the role of MYC in the recruit-

ment of HDAC3 and PRC2 to the miR-29 promoters. Finally, luciferase reporter assays demonstrated that miR-29 is repressed by MYC acting through HDAC3 and EZH2-mediated histone deacetylation and trimethylation. Recent work has shown that MYC is involved in miR-29 gene expression regulation and that HDAC with MYC is responsible for the silencing of miR-29b in acute myeloid leukemia cells (Liu et al., 2010). Our current ndings suggest that PRC2 is an additional factor ensuring miR-29 downregulation through working in concert with HDAC3. Thus, these ndings dene a key mechanism of miRNA transcriptional repression by MYC and shed light on the poorly understood mechanisms involved in miRNA suppression in B-cell lymphomas. The MYC-miR-26a-EZH2-miR-494 positive feedback loop was observed in MYC-expressing lymphoma cell lines and primary lymphoma cells examined. We conclude that miR-26a can function as a tumor suppressor miRNA in MYC-associated lymphomas. Once MYC is activated, miR-26a is repressed; the more miR-26a is decreased, the more its target gene (such as EZH2) is activated. Inverse correlations between miR-26a, miR-29, and MYC, EZH2 expression were detected in both cell lines and primary samples supporting the presence of MYCmiRNA-EZH2 positive feedback loop. The decrease in miR-26a expression and consequent increase in EZH2 expression have been reported in a variety of aggressive tumors such as hepatocellular carcinoma, nasopharyngeal carcinoma, and Burkitt lymphoma (Sander et al., 2008). The frequent EZH2 overexpression found in human cancers is associated with more aggressive cancer phenotypes with poor prognosis (So et al., 2011). This was further supported by ndings in a larger cohort of hepatocellular carcinoma patients, where low miR-26a expression was associated with shorter overall survival (Kota et al., 2009). In addition, EZH2 was detected in the neoplastic large cells in intermediate- and high-grade B-cell lymphomas, and its expression was correlated with clinical grade and the presence of Ki-67 expression (van Kemenade et al., 2001). In MCL, EZH2 is upregulated in proliferating MCL cells, with expression levels of MYC and EZH2 being the strongest prognostic factors independent of tumor proliferation and clinical factors of MCL (Visser et al., 2001). These reports concur with our previous and current ndings that miR-29 expression is reversely correlated, controlled with MYC and EZH2, and associated with MCL aggressive progression (Zhao et al., 2010). Our signicant nding of the MYC-EZH2-miR-29 axis provides insight into how EZH2 is activated and contributes to tumor aggressive transformation,

Figure 6. MYC-miR-26a-EZH2-miR-494 Positive Feedback Loop Sustains MYC Activity and miR-29 Repression
(A) EZH2 is a direct target of miR-26a. Overexpression of miR-26a downregulates EZH2 and MYC expression and suppresses EZH2 30 -UTR luciferase activity in 293T cells. (B) miR-26a expression is regulated by MYC. miR-26a, EZH2, SUZ12, and MYC protein expression levels in MYC turn-on and MYC turn-off P493-6 cells. (C) Overexpression of miR-26a by pri-miR-26a suppresses MYC-induced EZH2 expression in MYC-on P493-6 cells, while suppression of miR-26a by anti-miR-26a increases EZH2 expression in MYC-off P493-6 cells. (D) Inhibition of EZH2 with DZNep or shRNA decreases MYC protein expression. (E) Putative MYC 30 -UTR targeting miRNAs are upregulated by EZH2 inhibition. (F) TargetScan and microCosm depicting potential binding sites for the DZNep upregulated miRNAs in MYC-30 -UTR and MYC is a direct target of miR-494. 293T cells are cotransfected with luciferase reporters, which contain the wild-type or mutant of MYC 30 -UTR, and overexpression of miR-494 inhibits MYC-30 -UTR but not mutant 30 -UTR luciferase activities. (G) Overexpression of miR-494 suppresses MYC and EZH2 expression. Results are representative of three independent experiments or means SD from at least three biological replicates. See also Figure S5.

518 Cancer Cell 22, 506523, October 16, 2012 2012 Elsevier Inc.

Cancer Cell
MYC, HDAC, EZH2, and miRNA

A
p

B
p

C
p p

D
p p

Cancer Cell 22, 506523, October 16, 2012 2012 Elsevier Inc. 519

Cancer Cell
MYC, HDAC, EZH2, and miRNA

thus revealing mechanistic links between EZH2 and its upstream and downstream signaling in MYC-associated lymphomas. The results on EZH2 regulation of MYC are in agreement with previous studies showing that EZH2 induced MYC expression and provide insight into the mechanisms of MYC activation and EZH2-driven cell proliferation. Recent studies have revealed recurrent somatic mutations of EZH2 in lymphomas and the inactivating somatic mutations of the H3K27 demethylase, UTX, in multiple cancers (van Haaften et al., 2009). These ndings suggest that deregulation of H3K27 methylation may also contribute to constitutive MYC activation in these lymphomas and EZH2 trimethylase as an ideal therapeutic target for lymphoma therapy. Here, we reveal that dynamic forces act through a feedback circuit to modulate oncogenic expression of proteins, MYC and EZH2, at the posttranscriptional level via miRNAs. This reverberating relationship ensures the signal transduction of the upstream triggering events, leading to the sustained induction of MYC and EZH2 as well as the suppression of the downstream miR-29 family. Given the role of EZH2 in MYC activation and miR-29 repression, inhibition of EZH2 will target both upstream (MYC) and downstream (miR-29, CDK6, IGF-1R) signaling events of aggressive lymphomas. The transcriptional and posttranscriptional repression of miRNAs through MYC, HDAC3, and PRC2 could be a common feature of many tumor suppressor miRNAs. Thus, our ndings provide rational to redirect therapeutic effort by reactivating these tumor suppressor miRNAs through combined inhibition of HDAC and PRC2. Convincingly, we demonstrated that the combination of HDAC and EZH2 inhibitors (vorinostat and DZNep) or their siRNAs induced more miR-29a/b1 and miR29b2/c gene expression, resulting in the synergistic reduction of protein levels of CDK6 and IGF-1R and subsequent inhibition of cell survival and colony formation in vitro. Of note, HDAC3 and EZH2 overexpression was detected in essentially all of the lymphoma cell lines and primary samples that we tested, but not in normal B lymphocytes and nontransformed B lymphocytes. This provides a strong rationale that targeting HDAC3 and EZH2 may be more effective in lymphoma cells than in normal B lymphocytes. Indeed, our study showed that vorinostat and DZNep dramatically inhibited cell growth of transformed P493-6 cells and had no or minimal effect on nontransformed P493-6 cells. Finally, in vivo studies presented in this work illustrated that, compared with treatment with each agent alone, combined treatment with DZNep and vorinostat inhibits tumor growth and signicantly improves survival of NOD/SCID mice

bearing MCL xenografts. These results strongly support further development and testing of a combination of anti-EZH2 and a specic HDAC3 inhibitor against aggressive lymphomas.
EXPERIMENTAL PROCEDURES Cell Lines, Cell Proliferation, Colony Formation Assay, and Patient Samples Cell lines and patient sample information are detailed in Supplemental Experimental Procedures. All patient tissue specimens were from fresh biopsy-derived lymphoma tissues (lymph nodes) after informed consent was obtained, in accordance with the Declaration of Helsinki and after approval by the Institutional Review Board of the University of South Florida. Details of cell proliferation and colony formation assays are also described in the Supplemental Experimental Procedures. Co-IP and ChIP For co-IP in 293T, cells were transfected with plasmids using Lipofectamine 2000 (Invitrogen). Cells were harvested 36 hr after transfection. Protein (200 mg) was immunoprecipitated with the primary antibody (2 mg) overnight at 4 C, and the immunocomplexes were resolved by SDS-PAGE followed by immunoblot analysis. For endogenous protein interaction in Jeko-1, IP was performed using the Pierce Co-IP Kit (Thermo Scientic). Six micrograms of anti-HDAC3 antibody, anti-SUZ12 antibody, or normal rabbit immunoglobulin G (IgG) was coupled to AminoLink Plus Coupling Resin according to the manufacturers protocol. Immune complexes were eluted from the resin and analyzed by SDS-PAGE followed by immunoblot analysis. For P493-6 cell line, cells were lysed in NP-40 lysis buffer. Protein (1,000 mg) was immunoprecipitated with the primary antibody (2 mg) overnight at 4 C. HDAC3 was detected using GenScript One-Hour IP-Western Kits. For the ChIP assay, 2 3 106 cells and 3 mg of antibody was used per IP. The immunoprecipitated DNA was treated with RNase (Ambion) for 30 min at 37 C and proteinase K (Roche) for an hour at 45 C.The DNA was puried with QIAGEN PCR Spin columns. Puried DNA was analyzed by real-time PCR using specic primers. Primer sequences used in ChIP assay are listed in the Supplemental Experimental Procedures. Luciferase Assays Cells transfected with indicated plasmid were harvested and subjected to luciferase reporter assay using the luciferase assay system according to the manufacturers instructions (Promega). Details of this analysis and procedure are described in the Supplemental Experimental Procedures. siRNA Knockdown and Short-Hairpin RNA-Mediated Gene Knockdown For transient trasfection of siRNA, 5 3 106 cells were transfected by electroporation using Nucleofector (Amaxa) according to the manufacturers instruction. For short-hairpin RNA-mediated gene knockdown, cells were transduced with indicated lentivirus particles followed with puromycin selection. The knockdown efciency was conrmed by western blot.

Figure 7. miR-26a and miR-29 Downregulation Are Reversely Correlated with Upregulation of MYC and EZH2 in MCL and Other Aggressive MYC-Expressing Lymphomas
(A) miR-26a and miR-29 expression levels and MYC and EZH2 protein levels in MCL and other aggressive B-cell lymphoma cell lines. Cell lines were as follows: Jeko-1, Mino, HBL-2, NCEB-1, REC-1, Z138c (MCL); Raji and Ramos (Burkitt lymphoma) SUDHL-4 (Su-4), SUDHL-10 (Su10) (transformed large B-cell lymphoma); and SKW6.2 (EBV-associated lymphoma). (B) miR-26a and miR-29 expression levels and MYC and EZH2 protein levels in primary MCL samples and other aggressive B-cell lymphoma samples. Samples were as follows: P1 and P5 (aggressive MCL); P13 and P31P35 (Burkitt lymphoma); P14, P24, P25, and P36, (high-grade transformed diffuse large B-cell lymphomas). N1N3, CD19 sorted normal B lymphocytes. miR-26a and miR-29 expression levels were measured by qRT-PCR and normalized to RNU44. MYC and EZH2 expression levels were evaluated by western blot; in (A) and (B), the relative levels of MYC and EZH2 protein were measured by quantitative densitometry and are indicated below each lane. Insert, correlation between MYC and EZH2 protein. r, correlation coefcient. (C) Correlation between MYC/EZH2 protein expressions with miR-26a/miR-29a-c level in MCL and other aggressive B-cell lymphoma cell lines. r, correlation coefcient. (D) Correlation between MYC/EZH2 protein expressions with miR-26a/miR-29a-c level in primary MCL and other aggressive B-cell lymphoma samples. r, correlation coefcient. Results are representative of three independent experiments or means SD from at least three biological replicates.

520 Cancer Cell 22, 506523, October 16, 2012 2012 Elsevier Inc.

Cancer Cell
MYC, HDAC, EZH2, and miRNA

p p p

p p p

Cancer Cell 22, 506523, October 16, 2012 2012 Elsevier Inc. 521

Cancer Cell
MYC, HDAC, EZH2, and miRNA

The details of these analysis and procedures are described in the Supplemental Experimental Procedures. qRT-PCR Analysis and miRNA Microarray Analysis For qRT-PCR analysis, total RNA was isolated from cells with Trizol reagent (Invitrogen). qRT-PCR was performed according to the manufacturers instructions (Applied Biosystems). Jeko-1 cells were treated with DZNep for 72 hr. Total RNA was extracted and reverse transcribed into cDNA using the Megaplex Primer Pools by TaqMan miRNA reverse transcription kit (Applied Biosystems). The cDNAs were used to perform the microarray analysis using TaqMan Array miRNA Cards according to the manufacturers instructions. Array data were analyzed using DataAssist Software V3.0 (Applied Biosystems). Tumorigenesis Assays Z138c cells (5 3 106) were injected into anks of NOD/SCID mice (n = 6 mice per condition). Treatment was initiated when mean tumor volume was approximately 200 mm3. Mice were treated intraperitoneally with dimethyl sulfoxide (DMSO) (vehicle), 1 mg/kg of DZNep twice per week, and/or 30 mg/kg of Vorinostat daily for 2 weeks. Tumor growth was measured by calipers every 3 days. Survival of the mice in all groups is represented by Kaplan-Meier plot. All animal studies were performed in accordance with the Kansas University Cancer Center Institutional Guidelines and Regulations for animal care and under protocols approved by the Kansas University Medical Center Institutional Animal Care and Use Committee. Statistical Analysis All of the analyses were completed with SPSS 11.0 software, with p < 0.05 considered statistically signicant. Statistical analysis for cell proliferation and tumor growth curve was carried out by an analysis of variance. A log-rank (Mantel-Cox) test was used to test the Kaplan-Meier plot. ACCESSION NUMBERS The GEO database accession number for the microarray data is GSE40019. SUPPLEMENTAL INFORMATION Supplemental Information includes six gures and Supplemental Experimental Procedures and can be found in this article online at http://dx.doi.org/10.1016/ j.ccr.2012.09.003. ACKNOWLEDGMENTS We are grateful to the Tissue Procurement and Molecular Core Laboratory at Moftt Cancer Center for providing specimens and molecular analysis. We thank Rasa Hamilton for editorial assistance. This work was supported by grants from the National Cancer Institute (Grant R01 CA137123 to J.T.), the Maher Fund (to J.T.), the Susan and John Sykes Lymphoma Research Fund (to J.T.) and the Lymphoma Research Foundation (to J.T.).

Received: November 22, 2011 Revised: March 30, 2012 Accepted: September 4, 2012 Published: October 15, 2012 REFERENCES Chang, T.C., Yu, D., Lee, Y.S., Wentzel, E.A., Arking, D.E., West, K.M., Dang, C.V., Thomas-Tikhonenko, A., and Mendell, J.T. (2008). Widespread microRNA repression by Myc contributes to tumorigenesis. Nat. Genet. 40, 4350. Chen, H., Tu, S.W., and Hsieh, J.T. (2005). Down-regulation of human DAB2IP gene expression mediated by polycomb Ezh2 complex and histone deacetylase in prostate cancer. J. Biol. Chem. 280, 2243722444. Dave, S.S., Fu, K., Wright, G.W., Lam, L.T., Kluin, P., Boerma, E.J., Greiner, T.C., Weisenburger, D.D., Rosenwald, A., Ott, G., et al; Lymphoma/ Leukemia Molecular Proling Project. (2006). Molecular diagnosis of Burkitts lymphoma. N. Engl. J. Med. 354, 24312442. Fabbri, M., Garzon, R., Cimmino, A., Liu, Z., Zanesi, N., Callegari, E., Liu, S., Alder, H., Costinean, S., Fernandez-Cymering, C., et al. (2007). MicroRNA29 family reverts aberrant methylation in lung cancer by targeting DNA methyltransferases 3A and 3B. Proc. Natl. Acad. Sci. USA 104, 1580515810. Fabbri, M., and Croce, C.M. (2011). Role of microRNAs in lymphoid biology and disease. Curr. Opin. Hematol. 18, 266272. Fiskus, W., Wang, Y., Sreekumar, A., Buckley, K.M., Shi, H., Jillella, A., Ustun, C., Rao, R., Fernandez, P., Chen, J., et al. (2009). Combined epigenetic therapy with the histone methyltransferase EZH2 inhibitor 3-deazaneplanocin A and the histone deacetylase inhibitor panobinostat against human AML cells. Blood 114, 27332743. ` ndez, V., Moreno, V., Valls, J., Herna ndez, L., Bosch, F., Hartmann, E., Ferna Abrisqueta, P., Klapper, W., Dreyling, M., Hoster, E., et al. (2008). Five-gene model to predict survival in mantle-cell lymphoma using frozen or formalinxed, parafn-embedded tissue. J. Clin. Oncol. 26, 49664972. Kota, J., Chivukula, R.R., ODonnell, K.A., Wentzel, E.A., Montgomery, C.L., Hwang, H.W., Chang, T.C., Vivekanandan, P., Torbenson, M., Clark, K.R., et al. (2009). Therapeutic microRNA delivery suppresses tumorigenesis in a murine liver cancer model. Cell 137, 10051017. Kurland, J.F., and Tansey, W.P. (2008). Myc-mediated transcriptional repression by recruitment of histone deacetylase. Cancer Res. 68, 36243629. Liu, S., Wu, L.C., Pang, J., Santhanam, R., Schwind, S., Wu, Y.Z., Hickey, C.J., Yu, J., Becker, H., Maharry, K., et al. (2010). Sp1/NFkappaB/HDAC/miR-29b regulatory network in KIT-driven myeloid leukemia. Cancer Cell 17, 333347. Lu, J., He, M.L., Wang, L., Chen, Y., Liu, X., Dong, Q., Chen, Y.C., Peng, Y., Yao, K.T., Kung, H.F., and Li, X.P. (2011). MiR-26a inhibits cell growth and tumorigenesis of nasopharyngeal carcinoma through repression of EZH2. Cancer Res. 71, 225233. Mott, J.L., Kobayashi, S., Bronk, S.F., and Gores, G.J. (2007). mir-29 regulates Mcl-1 protein expression and apoptosis. Oncogene 26, 61336140.

Figure 8. Combined Inhibition of HDAC and PRC2 Cooperatively Reactivates miR-29 Level and Suppresses Tumor Cell Growth In Vitro and In Vivo
(A) Combined treatment with vorinostat and DZNep induces a higher expression of primary miR-29a/b1, primary miR-29b2/c, and mature miR-29a-c than each agent alone in HBL-2 cells. (B and C) Combined treatment with vorinostat and DZNep induced higher (B) downregulation of IGF-1R and CDK6 protein and (C) inhibition of clonogenic growth than each agent alone in HBL-2 cells. (D) Cell proliferation assay (CCK8) showing that MYC-on P493-6 cells are more sensitive than MYC-off P496-3 cells to DZNep and vorinostat treatment. (E and F) Cotreatment with DZNep and vorinostat inhibits tumor growth and signicantly improves survival of NOD/SCID mice bearing MCL xenografts. Tumor growth was measured by calipers. Results are mean tumor volume SEM, (treatment versus vehicle control, *p < 0.001; combination versus single agent, **p = 0.0002; #p = 0.0004). Survival of mice in all groups is represented by a Kaplan-Meier plot, and a log-rank test was used (n = 6 mice per condition). (G) Model of feed-forward regulatory circuit in which MYC contributes to the upregulation of EZH2 via repressing EZH2-targeting miR-26a and that EZH2 in turn relieves MYC negative regulation via MYC-targeting miR-494, thereby generating a positive feedback loop to ensure persistently high protein levels of MYC and EZH2 and further repression of miR-29. Results are representative of three independent experiments or means SD from at least three biological replicates. See also Figure S6.

522 Cancer Cell 22, 506523, October 16, 2012 2012 Elsevier Inc.

Cancer Cell
MYC, HDAC, EZH2, and miRNA

Nilsson, J.A., and Cleveland, J.L. (2003). Myc pathways provoking cell suicide and cancer. Oncogene 22, 90079021. Pekarsky, Y., Santanam, U., Cimmino, A., Palamarchuk, A., Efanov, A., Maximov, V., Volinia, S., Alder, H., Liu, C.G., Rassenti, L., et al. (2006). Tcl1 expression in chronic lymphocytic leukemia is regulated by miR-29 and miR-181. Cancer Res. 66, 1159011593. Sander, S., Bullinger, L., Klapproth, K., Fiedler, K., Kestler, H.A., Barth, T.F., ller, P., Stilgenbauer, S., Pollack, J.R., and Wirth, T. (2008). MYC stimulates Mo EZH2 expression by repression of its negative regulator miR-26a. Blood 112, 42024212. Slack, G.W., and Gascoyne, R.D. (2011). MYC and aggressive B-cell lymphomas. Adv. Anat. Pathol. 18, 219228. So, A.Y., Jung, J.W., Lee, S., Kim, H.S., and Kang, K.S. (2011). DNA methyltransferase controls stem cell aging by regulating BMI1 and EZH2 through microRNAs. PLoS One 6, e19503. Sparmann, A., and van Lohuizen, M. (2006). Polycomb silencers control cell fate, development and cancer. Nat. Rev. Cancer 6, 846856. 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 muta-

tions of the histone H3K27 demethylase gene UTX in human cancer. Nat. Genet. 41, 521523. van Kemenade, F.J., Raaphorst, F.M., Blokzijl, T., Fieret, E., Hamer, K.M., Satijn, D.P., Otte, A.P., and Meijer, C.J. (2001). Coexpression of BMI-1 and EZH2 polycomb-group proteins is associated with cycling cells and degree of malignancy in B-cell non-Hodgkin lymphoma. Blood 97, 38963901. Visser, H.P., Gunster, M.J., Kluin-Nelemans, H.C., Manders, E.M., Raaphorst, F.M., Meijer, C.J., Willemze, R., and Otte, A.P. (2001). The Polycomb group protein EZH2 is upregulated in proliferating, cultured human mantle cell lymphoma. Br. J. Haematol. 112, 950958. Zhang, X., Chen, X., Lin, J., Lwin, T., Wright, G., Moscinski, L.C., Dalton, W.S., Seto, E., Wright, K., Sotomayor, E., and Tao, J. (2012). Myc represses miR-15a/miR-16-1 expression through recruitment of HDAC3 in mantle cell and other non-Hodgkin B-cell lymphomas. Oncogene 31, 30023008. Zhao, J.J., Lin, J., Lwin, T., Yang, H., Guo, J., Kong, W., Dessureault, S., Moscinski, L.C., Rezania, D., Dalton, W.S., et al. (2010). microRNA expression prole and identication of miR-29 as a prognostic marker and pathogenetic factor by targeting CDK6 in mantle cell lymphoma. Blood 115, 26302639.

Cancer Cell 22, 506523, October 16, 2012 2012 Elsevier Inc. 523

Article
Blockade of miR-150 Maturation by MLL-Fusion/MYC/LIN-28 Is Required for MLL-Associated Leukemia
Xi Jiang,1,9 Hao Huang,1,9 Zejuan Li,1,9 Yuanyuan Li,1 Xiao Wang,2 Sandeep Gurbuxani,3 Ping Chen,1 Chunjiang He,1 Dewen You,4 Shuodan Zhang,1 Jinhua Wang,1 Stephen Arnovitz,1 Abdel Elkahloun,5 Colles Price,1 Gia-Ming Hong,1 Haomin Ren,1 Rejani B. Kunjamma,1 Mary Beth Neilly,1 Jonathan M. Matthews,1 Mengyi Xu,1 Richard A. Larson,1 Michelle M. Le Beau,1 Robert K. Slany,6 Paul P. Liu,5 Jun Lu,7 Jiwang Zhang,4,8 Chuan He,2 and Jianjun Chen1,*
of Hematology/Oncology, Department of Medicine of Chemistry and Institute for Biophysical Dynamics 3Department of Pathology University of Chicago, Chicago, IL 60637, USA 4Oncology Institute, Cardinal Bernardin Cancer Center, Loyola University Medical Center, Maywood, IL 60153, USA 5Genetics and Molecular Biology Branch, National Human Genome Research Institute, NIH, Bethesda, MD 20892, USA 6Department of Genetics, University Erlangen, 91058 Erlangen, Germany 7Yale Stem Cell Center, Department of Genetics, Yale University, New Haven, CT 06520, USA 8Department of Pathology, Loyola University Medical Center, Maywood, IL 60153, USA 9These authors contributed equally to this work *Correspondence: jchen@medicine.bsd.uchicago.edu http://dx.doi.org/10.1016/j.ccr.2012.08.028
2Department 1Section

Cancer Cell

SUMMARY

Expression of microRNAs (miRNAs) is under stringent regulation at both transcriptional and posttranscriptional levels. Disturbance at either level could cause dysregulation of miRNAs. Here, we show that MLL fusion proteins negatively regulate production of miR-150, an miRNA widely repressed in acute leukemia, by blocking miR-150 precursors from being processed to mature miRNAs through MYC/LIN28 functional axis. Forced expression of miR-150 dramatically inhibited leukemic cell growth and delayed MLL-fusionmediated leukemogenesis, likely through targeting FLT3 and MYB and thereby interfering with the HOXA9/ MEIS1/FLT3/MYB signaling network, which in turn caused downregulation of MYC/LIN28. Collectively, we revealed a MLL-fusion/MYC/LIN28xmiR-150xFLT3/MYB/HOXA9/MEIS1 signaling circuit underlying the pathogenesis of leukemia, where miR-150 functions as a pivotal gatekeeper and its repression is required for leukemogenesis.
INTRODUCTION MicroRNAs (miRNAs), a class of small, noncoding RNAs, are important for posttranscriptional gene regulation in both health and disease (He and Hannon, 2004; Xiao and Rajewsky, 2009). The stringent control of miRNAs, at both the transcriptional and posttranscriptional levels, is critical for maintaining a variety of important biological processes, including development, differentiation, and hematopoiesis (Siomi and Siomi, 2010). After being transcribed by RNA polymerase II, the primary miRNA transcripts (pri-miRNAs) would be under a two-step biocleavage process. In the cell nucleus, the Microprocessor, containing an RNaseIII enzyme Drosha and its cofactor DGCR8, crops the pri-miRNA into a 70 nucleotide (nt) hairpin structured precursor (pre-miRNA), which is then exported to the cytoplasm and cleaved by another RNaseIII enzyme, Dicer,

Signicance Although altered expression of many miRNAs has been reported in various cancers, including leukemia, their dysregulational mechanisms and pathologic functions remain less understood. Here, we show that miR-150 is signicantly downregulated in most cases of acute myeloid leukemia (AML). In AML with MLL rearrangements, the maturation of miR-150 is inhibited by the MLL-fusion/MYC/LIN28 functional axis. Furthermore, miR-150 itself functions as a pivotal tumor suppressor in MLL-associated leukemogenesis through repressing the expression of FLT3 and MYB, two essential oncogenes, and subsequently interfering the HOXA9/MEIS1/FLT3/MYB/MYC/LIN28 signaling network. Taken together, we report here a signaling circuit in leukemogenesis, in which the posttranscriptional repression of miR-150 maturation is a critical and necessary event. Thus, our results may provide optional strategies for anti-leukemia therapy.
524 Cancer Cell 22, 524535, October 16, 2012 2012 Elsevier Inc.

Cancer Cell
miR-150 Is a Pivotal Gatekeeper in MLL Leukemia

to remove the terminal loop region, or pre-element (preE), and to yield the mature miRNA (Newman and Hammond, 2010). The transcriptional regulation of miRNAs has been extensively studied, but the current understanding of their posttranscriptional control in cancer is limited. Lin28 is an RNA-binding protein known to block the maturation of let-7 and is an important modulator in the posttranscriptional regulation of miRNA maturation (Heo et al., 2008; Newman et al., 2008; Viswanathan et al., 2008). Lin28 itself is a direct downstream target of Myc (Chang et al., 2009). However, although MYC is aberrantly overexpressed in various types of cancer including lymphoma and leukemia (He et al., 2005; Hoffman et al., 2002; ODonnell et al., 2005), the involvement of the MYC/LIN28 axis in the posttranscriptional regulation of miRNA maturation in hematopoietic malignancies, e.g., acute myeloid leukemia (AML), is poorly understood. AML is a heterogeneous group of genetically diverse hematopoietic malignancies with variable response to treatment (Chen et al., 2010). Chromosome translocations are frequently observed in AML (Rowley, 2008). AML with chromosomal rearrangements involving the mixed lineage leukemia gene (MLL) is associated with a poor prognosis, especially in infants and individuals with leukemia that arises secondary to previous chemotherapy (Behm et al., 1996; Krivtsov and Armstrong, 2007; Rowley, 2008). More than 60 different loci that translocate to the MLL locus have been identied and cloned (Rowley, 2008). The critical feature of these chromosomal rearrangements is the generation of a chimeric transcript consisting of 50 MLL and 30 sequences of the partner gene, many of which are involved in transcriptional regulation. The human AF9 gene at 9p22 is one of the most common fusion partner genes with MLL (Krivtsov and Armstrong, 2007). Several important oncogenes are known to be direct or indirect downstream targets of MLL-fusion proteins. Among those, homeobox A (HOXA) genes, MEIS1, FLT3, MYB, and MYC are frequently upregulated in MLL-associated leukemias (Armstrong et al., 2002, 2003; Schreiner et al., 2001; Zeisig et al., 2004). The HOXA cluster genes are direct targets of MLL (Milne et al., 2005a, 2005b; Yu et al., 1995) and MLL fusion proteins promote their expression by epigenetic mechanisms (e.g., H3K79 methylation) (Bernt et al., 2011; Faber et al., 2009; Krivtsov and Armstrong, 2007; Krivtsov et al., 2008). High expression of HOXA9 and its cofactor, MEIS1, was often found in MLL-associated leukemias (Armstrong et al., 2002; Bullinger et al., 2004; Faber et al., 2009). FLT3 and its downstream effectors (e.g., AKT, STAT5, and ERK) are broadly involved in multiple processes of hematopoiesis and leukemogenesis via regulating the expression of a group of targets, such as JUN and MYC (Takahashi, 2011). MYC is a transcription factor involved in cell proliferation and apoptosis and is upregulated in the FLT3-ITD-transduced CD34+ hematopoietic stem/progenitor cells (Li et al., 2007). MYB is an essential downstream target of HOXA9/MEIS1 signaling (Hess et al., 2006), and an autoregulatory feedback loop was reported recently in which MYB binds MLL through MENIN and regulates expression of HOXA9/MEIS1 directly (Jin et al., 2010). As MYC is aberrantly overexpressed in various types of cancer including lymphoma and leukemia (He et al., 2005; Hoffman et al., 2002; ODonnell et al., 2005), we tested the hypothesis

that the MYC/LIN28 axis plays an essential role in AML in which MYC functions as an important downstream target of MLL fusions and FLT3 through posttranscriptional regulation of maturation of some critical tumor-suppressor miRNAs. RESULTS Expression of miR-150 Is Downregulated in Most AML To identify potential tumor-suppressor miRNAs that are signicantly downregulated in AML, we performed a bead-based miRNA expression proling assay of 52 AML samples (45 patient samples and seven cell lines; all bearing chromosomal translocations) along with three normal control samples and an Exiqon miRNA array assay of 100 samples (including 85 AML and 15 normal control samples). In both proling assays, we found that miR-150 was the most signicantly and consistently downregulated miRNA (q < 0.01; signicance analysis of microarrays [SAM]; Tusher et al., 2001) in most of the AML samples, including those bearing t(8;21), inv(16), t(15;17), and MLL rearrangements, compared to normal controls (Figure 1). Downregulation of miR-150 Is Not Related to DNA Copy Number Changes, Methylation, or Mutations To understand how miR-150 is downregulated in AML, we rst examined the DNA copy number of the miR-150 locus at 19q13.33 in 33 samples, including 29 AML samples and four normal controls. As shown in Figure S1A (available online), there was no signicant amplication or deletion of the genomic locus of miR-150 in MLL-associated or other subtypes of leukemia samples relative to normal controls. We also assessed the DNA methylation status of a 475-bp CpG-enriched, immediate upstream region of the miR-150 gene locus in 19 samples, including 17 AML samples and two controls, using bisulfate genomic DNA sequencing methods. No signicant changes in the DNA methylation level of the miR-150 CpG-enriched region were observed in leukemia samples relative to normal controls (Figure S1B). Moreover, no mutation was found in either the precursor miRNA sequence or the CpG island of miR-150. Together, our data suggest that other mechanism(s) might be underlying the downregulation of miR-150 in AML. Expression of miR-150 Is Regulated by MLL Fusion Proteins at Both the Transcriptional and Posttranscriptional Levels In contrast to fusion proteins with transcriptional repressing functions [e.g., t(8;21), inv(16), t(15;17), etc.] (Chen et al., 2010), MLL fusion proteins have been shown to act as transcriptional activators that directly upregulate expression of a group of oncogenes (e.g., HOXA genes and MEIS1) and miRNAs (e.g., miR-1792 and miR-196b) (Faber et al., 2009; Li et al., 2012b; Mi et al., 2010; Popovic et al., 2009; Zeisig et al., 2004). Thus, it is of great interest to understand how miR-150 expression is downregulated in MLL-associated leukemia. We rst investigated whether forced expression of MLL fusion genes can cause downregulation of miR-150. Indeed, we found that the level of miR-150 was dramatically downregulated by ectopic expression of MLL-AF9 in normal mouse bone marrow (BM) progenitor cells (i.e., lineage negative; Lin) both in vitro and in vivo (Figure 2A). In HEK293T cells, the level of miR-150
Cancer Cell 22, 524535, October 16, 2012 2012 Elsevier Inc. 525

Cancer Cell
miR-150 Is a Pivotal Gatekeeper in MLL Leukemia

Figure 1. Expression of miRNAs that Are Downregulated in Most AML Relative to Normal Controls
(A) miRNAs that are signicantly downregulated (q < 0.001; FDR < 0.0001; SAM; Tusher et al., 2001) in AML samples (n = 52) compared to normal bone marrow (BM) controls (n = 3; one CD15+ myeloid progenitor and two mononuclear [MNC] cell samples) as detected by bead-based method. (B) miRNAs that are signicantly downregulated (q < 0.0001; FDR < 0.0001; SAM) in AML (n = 85) samples than in normal BM controls (n = 15; six CD34+ hematopoietic stem/progenitor, ve CD33+ myeloid progenitor, and four MNC cell samples) as detected by Exiqon arrays. Expression data was mean-centered and the relative value for each sample is represented by a color, with red and green representing a high and low expression, respectively (scale shown in the upper left). MNC, mononuclear cell; MLL, MLL-associated leukemia.

correlated with the overexpression level of MLL-AF9 in a dosedependent manner (Figure S1C). Further studies were done using the MLL-ENL-ERtm cell line, a cell line stably expressing a conditional MLL-ENL derivative (Zeisig et al., 2004). As shown in Figure 2B, the level of miR-150 continued to increase after withdrawal of 4-hydroxytamoxifen (4-OHT) for the indicated number of days, along with the decreasing level of MLL-ENL. In order to determine whether the inhibitory effect of MLL fusions on miR-150 occurs at the transcriptional level, miR-150 primary (i.e., pri-miR-150) and precursor (i.e., pre-miR-150) transcripts were measured in the MLL-ENL-ERtm cell line. Strikingly, the miRNA primary and precursor transcripts decreased to 47%67% after 4-OHT withdrawal, whereas the level of mature miR-150 increased 4.8-fold, compared with cells with 4-OHT treatment (Figure 2C), suggesting that the MLL fusions might regulate the expression of miR-150 at both the transcriptional and posttranscriptional levels. Consistently, in the BM cells of MLL-associated leukemic mice, the levels of both pri-miR-150 and pre-miR-150 were upregulated, whereas the level of mature miR-150 was decreased (Figure 2D). Similarly, analysis of leukemic patient samples showed that mature miR-150 was consistently downregulated, whereas the levels of pri-miR-150 and pre-miR-150 were upregulated in samples with MLL fusions (Figure S1D). It was reported that MLL wild-type or MLL fusion proteins regulate the transcription of a variety of target genes, such as HOX genes, by directly binding to their promoter region (Milne et al., 2005a, 2005b; Zeisig et al., 2004, 2003). As shown in Figure 2E, we determined through chromatin immunoprecipitation
526 Cancer Cell 22, 524535, October 16, 2012 2012 Elsevier Inc.

(ChIP) assay that MLL fusion proteins bind directly to an immediate upstream region (<1 kb) of the miR-150 locus, and this binding is associated with increased level of histone H3 lysine 79 di- and tri-methylation (i.e., H3K79me2/3), a marker for active transcription in MLL-associated leukemic cells (Bernt et al., 2011; Krivtsov et al., 2008). There is no enrichment of MLL-fusion binding nor H3K79me2/3 at the control locus that is >10 kb upstream of the miR-150 locus (Figure 2E). Therefore miR-150 is likely a direct downstream target of MLL fusions. MLL Fusion Proteins Inhibit miR-150 Maturation through the MYC/LIN28 Functional Axis Recently, miR-150 was identied as one of the three miRNAs other than let-7 that could be repressed by both Lin28 and Myc (Chang et al., 2009). Interestingly, Myc has been implicated as a direct target gene of MLL and is frequently upregulated in MLL-associated leukemia (Schreiner et al., 2001; Zeisig et al., 2003), whereas Lin28 is a direct target gene of Myc (Chang et al., 2009). We then sought to investigate whether MYC and LIN28 participate in MLL-fusion-protein-mediated posttranscriptional regulation of miR-150. As shown in Figure 2F, knockdown of MLL in MONOMAC-6 cells by siRNA reduced the levels of both pri-miR-150 and pre-miR-150 to 57% and 40%, respectively, whereas mature miR-150 was raised 2-fold, compared to scrambled controls. Knockdown of MYC or LIN28 resulted in effects similar to knockdown of MLL. Overexpression of either MYC or LIN28 efciently reversed the effects of MLL knockdown on these cells (Figure 2F), indicating that MYC and LIN28 might function as downstream effectors of MLL in regulating the process of

Cancer Cell
miR-150 Is a Pivotal Gatekeeper in MLL Leukemia

Figure 2. MLL-Fusion Regulates the Level of miR-150 via the MYC/LIN28 Axis
(A) Relative expression levels of mature miR-150 in MLL-AF9-transduced colony-forming cells (in vitro) and BM cells of MLL-AF9 leukemic mice (in vivo; n = 3). (B) Expression changes of miR-150 and MLLENL during withdrawal of 4-hydroxy-tamoxifen (4-OHT) (days 0, 4, and 10 are shown) in MLL-ENLERtm cells. (C) Expression changes of pri-, pre-, and mature miR-150 transcripts 10 days post 4-OHT withdrawal in MLL-ENL-ERtm cells. (D) Expression changes of pri-, pre-, and mature miR-150 in BM cells of mice with MLL-fusionmediated leukemia. (E) ChIP assays of potential binding of MLL fusion proteins and MYC to the miR-150 gene locus and the control locus (i.e., CpG island of FCGRT, a neighboring gene) in MONOMAC-6 cells. (F) Expression changes of pri-, pre-, and mature miR-150 in MONOMAC-6 cells when the expression of MLL, MYC, and LIN28 was altered. (G and H) Assessing potential enrichments of premiR-150 and pre-let-7 in the LIN28-protein-RNA complex in THP-1 and MONOMAC-6 cells by RNA pull-down assay (G) and RNA-binding protein immunoprecipitation (RIP) assay (H), coupled with RT-qPCR. Pre-miR-103 is used as a negative control. Mean SD values are shown. *p < 0.05; **p < 0.01. See also Figure S1.

miR-150 maturation. Notably, forced expression of MYC in cells with MLL knock-down also signicantly upregulated both pri- and pre-miR-150 transcripts (Figure 2F), implying that MYC itself may regulate miR-150 primary transcription directly as well as serving as a downstream effector of MLL fusions. Indeed, our ChIP assay revealed that MYC also binds directly to the immediate upstream region of miR-150 (Figure 2E). Furthermore, Lin28 is likely to target the maturation processing of precursor miR-150 directly. In MLL-ENL-inducible cell lines, knockdown of Lin28 resulted in downregulation of pri- and pre-miR-150, and upregulation of mature miR-150 (Figure S1E), whereas overexpression of Lin28 in MONOMAC-6 cells prevented the maturation of miR-150 (Figure S1F). To investigate whether the let-7 pathway is implicated in Lin28-mediated blockade of miR-150 maturation, we transfected let-7 alone or together with Lin28 into MONOMAC6 cells and found that forced expression of let-7 (4070-fold increased expression) did not

result in a signicantly increased miR150 maturation (Figure S1F), suggesting that let-7 is not an essential mediator of this blockade. To determine whether LIN28 proteins can bind to miR-150 precursor RNA, we performed both RNA pull-down assay and RNA-binding protein immunoprecipitation (RIP) assay, coupled with qPCR. We rst used Histagged recombinant LIN28 protein to pull down target RNAs from total RNA isolated from THP-1 cells and the LIN28-RNA complex was puried with Ni-NTA beads. We found that pre-miR-150 transcripts, along with let-7 precursor (as a positive control), are signicantly enriched in the LIN28-RNA complex (Figure 2G), suggesting a potential direct binding between LIN28 and pre-miR-150. Furthermore, we used an anti-LIN28 antibody to precipitate endogenous LIN28-RNA complex from MONOMAC-6 cell lyses via RIP assay, and found that both miR-150 and let-7 precursors are highly enriched in LIN28 immunoprecipitation complex (Figure 2H), indicating an association between the miRNA precursors and endogenous LIN28 protein. In contrast, as a negative control, precursor of miR-103, a miRNA whose maturation was not inuenced by Lin28 (Viswanathan et al., 2008), was not enriched in either assay (Figures 2G and 2H). Taken together, our data suggest that both MLL fusions and MYC can promote primary transcription of miR-150, whereas the MYC/LIN28 axis inhibits its posttranscriptional maturation.
Cancer Cell 22, 524535, October 16, 2012 2012 Elsevier Inc. 527

Cancer Cell
miR-150 Is a Pivotal Gatekeeper in MLL Leukemia

Figure 3. miR-150 Inhibits Cell Growth In Vitro and Leukemogenesis In Vivo


(A) Colony-forming/replating assays of mouse BM progenitor cells cotransduced with MSCV-PIG+ MSCVneo (Ctrl), MSCV-PIG+MSCVneo-MLL-AF9 (MLL-AF9), or MSCV-PIG-miR-150+MSCVneoMLL-AF9 (MLL-AF9+miR-150). (B) Analyses of viability (VBL) and apoptosis (APT) of MONOMAC-6 or THP-1 cells 48 hr posttransfection with MSCV-PIG (Ctrl) or MSCV-PIGmiR-150 (miR-150). (C) Effects of miR-150 on the proliferation of MONOMAC-6 cells. (D and E) Effects of miR-150 on viability/apoptosis (D) and proliferation (E) of mouse BM progenitor cells in the presence or absence of MLL-AF9. Mean SD values are shown. *p < 0.05; **p < 0.01. (F and G) Effects of miR-150 in MLL-AF9-mediated leukemogenesis in vivo. Kaplan-Meier curves are shown for three groups of transplanted mice including control (n = 10), MLL-AF9 (n = 12), and MLL-AF9+miR-150 (n = 11) in a primary BMT assay (F), and for two groups of transplanted mice including MLL-AF9+miR-150 (n = 12) and MLLAF9 (n = 9) in secondary BMT assay (G). See also Figure S2.

miR-150 Inhibits Cell Proliferation/Transformation In Vitro and Leukemogenesis In Vivo In order to investigate whether miR-150 has an antagonistic action on the oncogenicity of MLL fusions, we performed colony-forming/replating assays. Mouse BM progenitor cells transduced with MSCV-PIG (bearing a PGK-puromycin-IRESGFP cassette; He et al., 2005) or MSCV-PIG-miR-150, together with MSCVneo or MSCVneo-MLL-AF9, were plated on methylcellulose medium. The colonies were replated every 7 days under the same conditions. As shown in Figure 3A, cotransduction of miR-150 and MLL-AF9 caused a signicant reduction in colonies compared to transduction of MLL-AF9 alone, indicating that forced expression of miR-150 (Figure S2A) strongly inhibits the colony-forming capacity induced by MLL-AF9. We then explored the effects of ectopic expression of miR-150 on human MLL-associated leukemic cells. As shown in Figure 3B, forced expression of miR-150 (Figure S2B) signicantly decreased cell viability and increased apoptosis in MONOMAC-6/t(9;11) and THP-1/t(9;11) cells. Moreover, overexpression of miR-150 consistently inhibited the proliferation of MONOMAC-6 cells (Figure 3C). Similarly, in retrovirally transduced mouse BM progenitor cells, we found that in the presence of MLL-AF9, forced expression of miR-150 signicantly decreased cell viability and increased apoptosis (Figure 3D), as well as inhibited cell growth (Figure 3E). In order to elucidate the in vivo role of miR-150 in leukemogenesis, we performed primary BM transplantation (BMT) assay and found that forced expression of miR-150 (Figure S2C) signicantly delayed leukemogenesis mediated by MLL-AF9 (median overall survival, 110 days versus 56 days; p < 0.001, log-rank test) (see Figure 3F). We then performed secondary BMT and showed that miR-150+MLL-AF9 leukemic cells developed
528 Cancer Cell 22, 524535, October 16, 2012 2012 Elsevier Inc.

AML in secondary recipient mice remarkably slower than MLLAF9 leukemic cells (median overall survival, 70 days versus 42 days; p < 0.001; Figure 3G). The miR-150-alone group did not show visible differences compared to the control group, either in survival or in the morphology of cells and tissues (data not shown). These ndings suggest that miR-150 does play a critical tumor suppressor role in preventing MLL-associated leukemogenesis. miR-150 Targets Both MYB and FLT3 To identify potential target genes of miR-150 in MLL-associated leukemia, we performed Agilent custom-design arrays of the eight human de novo MLL-associated AML and nine normal control (including three CD34+, two CD33+, and four MNC) samples that were used in Exiqon miRNA proling assay (see Figure 1B). By correlating expression of predicted target genes and miR-150 across the above 17 samples, we identied 158 putative target genes that exhibited a signicant inverse correlation of expression with miR-150 (p < 0.05; Pearson correlation). Analysis of Affymetrix Exon arrays of 15 additional human MLL-associated leukemia samples and nine normal controls (including three each of CD34+, CD33+, and MNC) (Li et al., 2012b) revealed that 22 of the above 158 candidate target genes were signicantly overexpressed (q < 0.05; false discovery rate [FDR] < 0.01; SAM; Tusher et al., 2001) in MLL-associated leukemia samples relative to normal controls. Furthermore, in the analysis of Affymetrix gene arrays of nine MLL-AF9-mediated mouse leukemia samples and six control samples (Li et al., 2012b), we observed that four (i.e., Flt3, Myb, Hdhd2, and Letmd1) of the above 22 candidate target genes were signicantly overexpressed (q < 0.05; FDR < 0.01; SAM) in MLL-AF9 mouse leukemia samples

Cancer Cell
miR-150 Is a Pivotal Gatekeeper in MLL Leukemia

Figure 4. miR-150 Targets Both MYB and FLT3


(A) Effects of overexpression of miR-150 on endogenous expression of MYB and FLT3. (B) Luciferase reporter and mutagenesis assays. Left panel: effects of miR-150 on luciferase activity of the reporter gene bearing wild-type or mutant 30 UTR of FLT3 in HEK293T cells. Right upper panel: Putative miR-150 target site and mutant in the 30 UTR of FLT3. Right lower panel: western blot analysis of FLT3 in MONOMAC-6 cells that were transfected with MSCV-PIG (Ctrl), MSCV-PIG-miR-150 (miR-150), scrambled siRNA control oligos (siNC), or FLT3 siRNAs (siFLT3). (C and D) Effects of depletion of MYB or FLT3 by siRNAs on viability (C) and apoptosis (D) of MONOMAC-6, THP-1, or KOCL-48 cells. (E) Effects of forced expression of MYB-CDS or FLT3-ITD-CDS alone, or together with miR-150, on viability and apoptosis of MONOMAC-6 cells. (F) Effect of miR-150 on cell viability and apoptosis of Jurkat, U937 and MONOMAC-6 cells. (G) Effect of miR-150 on the proliferation of Jurkat and U937 cells. (H) Effects of knockdown of Myb or Flt3 on MLL-AF9-induced colony-formation of mouse BM progenitor cells. Mean SD values are shown. *p < 0.05; **p < 0.01. See also Figure S3.

relative to normal controls. The expression patterns of these four genes in the above three sample sets are shown in Figure S3. As FLT3 and MYB have been shown to be broadly involved in multiple processes of hematopoiesis and leukemogenesis (Ramsay et al., 2003; Takahashi, 2011) and be critical oncogenes in MLL-associated leukemia (Armstrong et al., 2002, 2003; Hess et al., 2006; Zuber et al., 2011), we focused on these two genes for further studies. Forced expression of miR-150 reduced the levels of both MYB and FLT3 to 53%74% and 40%65%, respectively, in MONOMAC-6 and THP-1 cells (Figure 4A). MYB has been reported to be a direct target gene of miR-150 (Lu et al., 2008; Xiao et al., 2007). Luciferase reporter and muta-

genesis assays showed that miR-150 also targeted the 30 UTR of FLT3 directly (Figure 4B). Knockdown of the expression of MYB or FLT3 by siRNAs can mimic the effects of miR-150 overexpression such as decreasing cell viability (Figure 4C) and increasing apoptosis (Figure 4D) in MONOMAC-6, THP-1, and KOCL-48 cells. Cotransfection of MYB or FLT3-ITD, a constitutively active mutant of FLT3, with miR-150 completely reversed the effects of miR-150 on cell viability and apoptosis (Figure 4E). Overexpression of MYB or FLT3-ITD alone did not show noticeable effects on cell viability or apoptosis, likely due to the high expression level of endogenous MYB and FLT3 in MLL-associated leukemic cells (Somervaille et al., 2009) (Figure 4E). In FLT3-expression-negative cell
Cancer Cell 22, 524535, October 16, 2012 2012 Elsevier Inc. 529

Cancer Cell
miR-150 Is a Pivotal Gatekeeper in MLL Leukemia

Figure 5. miR-150 Inhibits MLL-FusionMediated Leukemogenesis via Regulating Flt3


(A) Kaplan-Meier curves are shown for three groups of transplanted mice (n = 5 for each group) in a primary BMT assay. (B) Kaplan-Meier curves are shown for two groups of transplanted mice including MLL-AF9 (n = 9) and MLL-AF9+shFlt3 (n = 8) in an additional primary BMT assay. (C and D) Wright-Giemsa staining of mouse peripheral blood (PB) and BM, or hematoxylin and eosin (H&E) staining of mouse spleen and liver. (E and F) Flow cytometry analyses of mouse BM cells. The blast population was gated by FSC/ SSC (in a red frame; upper panel) and then the proportions of Gr1+ and/or Mac1+ cells were analyzed (lower panel).

function as direct targets of miR-150 in regulating leukemic cell self-renewal. Repression of Flt3 Activity Is Required for the Inhibitory Effect of miR-150 in MLL-Associated Leukemogenesis The requirement of Myb function in MLL-associated cell transformation and leukemogenesis has been well documented (Hess et al., 2006; Zuber et al., 2011), suggesting that the inhibitory effects of forced expression of miR-150 on leukemogenesis would be, at least in part, related to its repression of Myb expression. To determine whether repression of Flt3 expression is also required for the inhibitory effects of miR-150 in MLLassociated leukemogenesis, we performed a series of in vivo functional studies. As shown in Figure 5A, the delay in MLL-AF9-induced leukemogenesis mediated by forced expression of miR150 (Figure S2C) could be largely reversed by coexpression of FLT3-ITDCDS. Our loss-of-function study showed that knockdown of endogenous Flt3 by shRNA signicantly delayed primary leukemogenesis mediated by MLL-AF9 (median overall survival, 76 days versus 65 days; p < 0.05) (Figure 5B), though not as signicantly as did forced expression of miR-150 (see Figure 3F). Leukemia in MLL-AF9+miR-150 mice is much less aggressive than that of the MLL-AF9 group, whereas ectopic expression of FLT3-ITD remarkably reversed the phenotype caused by miR-150 overexpression (Figure 5C). Knockdown of endogenous Flt3 by shRNA resulted in reduction of leukemic blast cells in both BM and peripheral blood (PB), similar to that found in miR-150 overexpressing mice (Figure 5D). There was a decrease in the

lines (e.g., Jurkat and U937) (Shankar et al., 2007; Yao et al., 2003), overexpression of miR-150 (Figure S2B) showed slight, if any, inhibitory effects on cell viability, apoptosis, or proliferation (see Figures 4F and 4G). Results of colony-forming/replating assays showed that cotransduction of Myb or Flt3 shRNA and MLL-AF9 caused a signicant reduction in colonies compared to transduction of MLL-AF9 alone (Figure 4H), mimicking the inhibitory effect of miR-150 on colony formation (see Figure 3A). These results indicate that both Myb and Flt3
530 Cancer Cell 22, 524535, October 16, 2012 2012 Elsevier Inc.

Cancer Cell
miR-150 Is a Pivotal Gatekeeper in MLL Leukemia

Figure 6. A Schematic Model of the MLL-Fusion/MYC/LIN28xmiR150xFLT3/MYB/HOXA9/MEIS1 Regulatory Circuit in MLL-Associated Leukemia


MLL fusion proteins inhibit the maturation of miR-150 through MYC/LIN28, release miR-150 inhibition on its target genes (FLT3 and MYB), then subsequently upregulate expression of MYC, LIN28, HOXA9, and MEIS1. The upregulation of MYC/LIN28 further represses the level of mature miR-150 and enhances/maintains the upregulation of the six genes in this regulatory circuit. These changes, in turn, result in increased cell proliferation and decreased apoptosis/ differentiation, and eventually lead to leukemogenesis.

proportion of Mac-1/Gr-1 double-positive cells in MLL-AF9+ miR-150 mice, compared to the MLL-AF9 group, whereas coexpression with FLT3-ITD signicantly increased the Mac-1/Gr-1 double-positive cell ratio (Figure 5E). Knockdown of Flt3 by Flt3-shRNA (i.e., shFlt3) showed an effect similar to forced expression of miR-150 on reducing the percentage of Mac-1/ Gr-1 double-positive cells (Figure 5F). These results demonstrate that the repression of Flt3 activity by miR-150 is, at least in part, responsible for the inhibitory effects of forced expression of miR-150 on leukemogenesis. The MLL-Fusion/MYC/LIN28xmiR-150xFLT3/MYB/ HOXA9/MEIS1 Regulatory Circuit HOXA9 and its cofactor MEIS1 are two of the best studied critical downstream target genes of MLL fusion proteins (Krivtsov et al., 2008; Milne et al., 2005a; Zeisig et al., 2004). MYB has been reported to be a direct target of HOXA9/MEIS1 signaling (Hess et al., 2006), whereas a recent study showed that MYB can also regulate expression of HOXA9/MEIS1 at the transcription

level (Jin et al., 2010), suggesting there is an autoregulatory feedback loop between them. Interestingly, FLT3 has been identied as a direct target gene of MEIS1 (Wang et al., 2005) as well as an upstream regulator of MYC (Li et al., 2007; Takahashi, 2011). MYC is also a downstream target of MLL fusion proteins (Dawson et al., 2011; Schreiner et al., 2001) and an upstream regulator of Lin28 (Chang et al., 2009). These ndings together with the data we showed above suggest that there is a critical MLL-fusion/MYC/LIN28xmiR-150x FLT3/MYB/ HOXA9/MEIS1 regulatory circuit in MLL-associated leukemia (see Figure 6). In this regulatory circuit, MLL fusion proteins function as the driver. Indeed, in human leukemia samples bearing MLL rearrangements, levels of expression of all downstream genes including MYC, LIN28, HOXA9, MEIS1, MYB, and FLT3 are signicantly increased (p < 0.05, t test) relative to normal control samples (see Figure S4A). MLL-AF9 retrovirally transduced into mouse normal BM progenitor cells signicantly induced expression of all six genes both in vitro (Figure 7A) and in vivo (Figure 7B), conrming that they are downstream target genes of MLL fusion proteins. In the MLL-ENL-ERtm cell line, their expression levels were signicantly reduced (p < 0.05, t test) in cells 10 days after withdrawal of 4-OHT (Figure S4B), indicating their dependence on the presence of MLL fusion proteins. On the other hand, this circuit also highlights the importance of the posttranscriptional regulation of the miR-150 maturation process in MLL-associated leukemogenesis. Cotransduction of miR-150 with MLL-AF9 reduced the MLL-AF9-mediated induction of expression of the six genes both in vitro (Figure 7A) and in vivo (Figure 7B), leading to a signicant reduction (p < 0.05) in colony-formation (Figure 3A) and a delay in leukemogenesis (Figures 3F and 3G). Similarly, ectopic expression of miR-150 in MONOMAC-6 cells also downregulated the expression of all six genes (see Figure 7C), leading to a signicant decrease in cell viability/proliferation and an increase in apoptosis (Figures 3B and 3C). These results indicate that mature miR-150 has an inuence on the expression of all six genes in this regulatory circuit through directly targeting FLT3 and MYB, suggesting that posttranscriptional repression of miR-150 is an essential and required event in the pathogenesis of MLL-associated leukemia. In addition, cotransfected FLT3-ITD could completely reverse the effects of ectopic expression of miR-150 in MONOMAC-6 cells on the expression of the six genes (Figure 7C) and on cell viability and apoptosis (see Figure 4E). Notably, forced expression of FLT3-ITD alone could signicantly increase expression (p < 0.05) of all six genes (Figure 7C). These data indicate that FLT3 is an essential target gene of miR-150 and a critical component of this regulatory circuit. The Strong and Widely Existing Inhibitory Effect of miR-150 on Cell Transformation We have also conducted in vitro colony-forming/replating assays to investigate the functions of several other miRNAs (e.g., miR-29a, miR-29b, miR-148a, miR-181a, and miR-181b) in MLL-fusion-mediated cell transformation because these miRNAs are also signicantly downregulated in human MLL-associated AML compared to normal controls (Figure 1) or other subtypes of AML (Li et al., 2008). However, none of these
Cancer Cell 22, 524535, October 16, 2012 2012 Elsevier Inc. 531

Cancer Cell
miR-150 Is a Pivotal Gatekeeper in MLL Leukemia

Figure 7. The Effects of MLL Fusion Proteins, miR-150, and FLT3 on the Expression of the Six MLL-Fusion Downstream Genes
(A) Effects of miR-150 on expression of endogenous Flt3, Myb, Hoxa9, Meis1, Myc, and Lin28 in mouse BM progenitor cells in the presence or absence of MLL-AF9. RNAs were collected from the rst passage of colony-forming cells. (B) Effects of miR-150 on the expression of above six genes in mouse MLL-AF9 leukemic BM cells (n = 5 for each group). (C) Effects of miR-150 and/or FLT3-ITD on expression of the six genes in MONOMAC-6 cells. Mean SD values are shown. *p < 0.05, t test. See also Figure S4.

Figure 8. The Strong and Widely Existing Inhibitory Effect of miR-150 on Cell Transformation
(A) Comparison of inhibitory effect of miR-150 with that of miR-29a, miR-29b, miR-148a, miR-181a, or miR-181b on MLL-AF9-induced colony forming capacity of mouse BM progenitor cells. (B) Relative expression level (upper panel) of ectopically expressed miR-150 in mouse BM progenitor cells transduced with common leukemic fusion genes such as MLL-AF9, MLL-ENL, PML-RARA, and AML1-ETO9a/AE9a, and its effects (lower panel) on the colony-forming/replating capacity of the transduced cells. Mean SD values are shown. *p < 0.05; **p < 0.01.

miRNAs exhibited an inhibitory effect as signicant and as consistent as miR-150 in inhibiting the cell transformation (Figure 8A). In addition, as miR-150 was signicantly downregulated in almost all subtypes of AML (Figure 1), in order to test the role of miR-150 in other subtypes of AML, we performed colonyforming/replating assays and found that forced expression of miR-150 could signicantly inhibit both MLL fusion (e.g., MLL-AF9 and MLL-ENL) and non-MLL fusion (e.g., PML-RARA resulting from t(15;17) and AML1-ETO9a/AE9a resulting from t(8;21))-induced cell transformation (Figure 8B). Thus, miR-150 likely also plays a critical tumor suppressor role in other subtypes of AML.
532 Cancer Cell 22, 524535, October 16, 2012 2012 Elsevier Inc.

DISCUSSION Expression of miRNAs is under stringent regulation at both the transcriptional and posttranscriptional levels. In the present study we showed that mature miR-150 level was signicantly downregulated in most AML samples, including MLL-associated AML. Strikingly, we found that despite the decreased abundance of mature transcripts of miR-150, its primary and precursor transcript abundance is increased in human MLL-associated AML, relative to normal controls. We further showed that MLL fusion proteins could bind to the genomic locus of miR-150 and significantly promote its primary transcription, leading to increased

Cancer Cell
miR-150 Is a Pivotal Gatekeeper in MLL Leukemia

abundance of its primary and precursor transcripts. In contrast, the mature miR-150 abundance was signicantly decreased when MLL fusion genes were expressed, and the opposite is true when MLL fusion gene expression was silenced or knocked down. Thus, an important question arises: What caused such a discrepancy between primary/precursor and mature miR-150 transcript levels in cells bearing MLL fusion genes? A recent study showed that Myc binds to the promoter region of Lin28 and activates its transcription, which is essential for Myc-mediated let-7 repression (Chang et al., 2009). Interestingly, besides let-7, three additional miRNAs (miR-150, miR146a, and miR-210) can also be repressed by both Myc and Lin28. MYC, a well-known oncogene, is a direct downstream target gene of MLL fusion proteins and thereby is frequently upregulated in MLL-associated leukemia (Schreiner et al., 2001; Zeisig et al., 2003), whereas Lin28 is a direct target gene of Myc. We therefore investigated whether MYC and LIN28 participate in MLL-fusion-protein-mediated posttranscriptional repression of miR-150. Indeed, here we showed that the maturation of miR-150 is also regulated by MYC/LIN28, and revealed an interesting function of MLL fusions in the regulation of miRNA maturation through activation of the MYC/LIN28 axis. MiR-150 has been implicated as either an oncogene or a tumor suppressor in various types of solid tumors (Li et al., 2012a; Srivastava et al., 2011; Wu et al., 2010; Zhang et al., 2012). However, its function in the pathogenesis of AML is unknown. The abundance of miR-150 in different populations of normal hematopoietic cells, including CD34+ hematopoietic stem/ progenitor cells, CD33+ or CD15+ myeloid cells, and mononuclear (MNC) cells, is rather consistent, but such abundance is signicantly downregulated in almost all subtypes of AML samples, including MLL-associated leukemia, suggesting that miR-150 might be important in maintaining normal myelopoiesis and in inhibiting leukemogenesis. Indeed, we showed that miR-150 plays a signicant inhibitory role on MLL-fusion-mediated cell transformation and leukemogenesis, and the inhibitory effect of miR-150 on cell transformation is much stronger than other miRNAs. More importantly, we identied the MLL-fusion/MYC/ LIN28xmiR-150xFLT3/MYB/ HOXA9/MEIS1 regulatory circuit in MLL-associated leukemia. In this circuit, MLL fusion proteins function as the driver, and their presence leads to the signicant upregulation of all six downstream genes, MYC, LIN28, FLT3, MYB, HOXA9, and MEIS1, as well as the primary transcription of miR-150. The upregulation of MYC/LIN28 results in the blockade of the miR-150 maturation process, which in turn leads to the release of miR-150 inhibition on FLT3 and MYB expression. The release of FLT3 and MYB would enhance the expression of HOXA9, MEIS1, MYC, and LIN28, and further enhance/ maintain the blockade of miR-150 maturation. As a result, the cells reach and maintain high levels of MYC/LIN28/FLT3/MYB/ HOXA9/MEIS1, and thereby transform the cells and lead to leukemogenesis. Our nding that FLT3 is a critical target gene of miR-150 and an essential component of the regulatory circuit provides further mechanistic evidence to support the notion that FLT3 is a promising therapeutic target in the treatment of MLL-associated leukemia (Armstrong et al., 2003; Stubbs and Armstrong, 2007). On the other hand, miR-150 functions as a pivotal gate-

keeper in inhibiting cell transformation and leukemogenesis by directly targeting FLT3/MYB and thereby inactivating the positive feedback loops within this circuit. The addition of MLL fusion proteins disrupts the balance between miR-150 and the other six genes. As the repression of miR-150 maturation is an essential event in MLL-fusion-mediated cell transformation and leukemogenesis, restoration of mature miR-150 would be an attractive strategy to treat MLL-associated leukemia alone, or in combination with other strategies, in the future. Finally, as miR-150 was signicantly downregulated in almost all subtypes of AML and exhibited a signicant inhibitory effect on cell transformation mediated by various types of leukemic fusion genes, it would be important in the future to systematically investigate miR-150s pathologenic role and critical target genes/pathways in other subtypes of AML, and to reveal the relevant molecular mechanisms underlying its downregulation. It would also be important to determine whether the MYC/ LIN28xmiR-150xFLT3/MYB/HOXA9/MEIS1 regulatory circuit also exists in other subtypes of AML as a whole or at least in part. Indeed, given the pivotal oncogenic functions of the MYC/LIN28 axis in various types of cancer including AML (Chang et al., 2009; He et al., 2005; Hoffman et al., 2002; ODonnell et al., 2005; Viswanathan et al., 2008), it is highly likely that miR-150, as an essential downstream target and antagonist of MYC/LIN28, plays a critical tumor-suppressor role in not only MLL-rearranged AML but also other subtypes of AML and even in other types of cancer. Taken together, we revealed a regulatory circuit, namely MLL-fusion/MYC/LIN28xmiR-150xFLT3/MYB/HOXA9/MEIS1 in MLL-associated leukemia (Figure 6). Our ndings may advance our understanding of the complex molecular mechanisms underlying the development and maintenance of MLLassociated leukemia, and may also provide effective strategies to treat MLL-associated leukemia, a disease that is presently treatment-resistant, and probably also other subtypes of AML, or even other types of cancer that also utilize at least part of the signaling circuit we have described herein.
EXPERIMENTAL PROCEDURES Leukemic Samples and Treatment Protocols All of the AML patient samples were obtained at the time of diagnosis or relapse and with informed consent at the University of Chicago Hospital (UCH), and were approved by the institutional review board of the hospital. All patients were treated according to the protocols of the hospital. miRNA and mRNA Expression Proling Assays The miRNA expression proling assays of the 55 human (52 AML and 3 normal control) sample set and the 100 human (85 AML and 15 normal control) samples were conducted by use of a bead-based method (Li et al., 2008) and Exiqon miRCURY LNA arrays (v10.0; covering 757 human miRNAs; Exiqon, Woburn, MA), respectively. The mRNA microarrays of the eight MLL-associated AML and nine normal control sample set, the 15 MLLrearranged AML and nine normal control sample set, and the 15 mouse sample set were conducted by use of Agilents custom-design microarrays (Agilent Technologies, Santa Clara, CA), Affymetrix GeneChip Human Exon 1.0 ST arrays, and Affymetrix GeneChip Mouse Gene 1.0 ST arrays, respectively. Cell Culture and Transfection THP-1, KOCL-48, U937, and MONOMAC-6 cells were grown in RPMI medium 1640 and transfected using the Amaxa Nucleofector Technology (Amaxa

Cancer Cell 22, 524535, October 16, 2012 2012 Elsevier Inc. 533

Cancer Cell
miR-150 Is a Pivotal Gatekeeper in MLL Leukemia

Biosystems, Berlin, Germany). See Supplemental Experimental Procedures for more details. Cell Apoptosis and Viability Assay Cell apoptosis and viability were assessed 48 hr posttransfection using ApoLive-Glo Multiplex Assay Kit (Promega, Madison, WI) following the manufacturers manuals. Luciferase Reporter and Mutagenesis Assays Luciferase reporter and mutagenesis assays were conducted as described previously (Li et al., 2012b), with some modications (see Supplemental Experimental Procedures). Chromatin Immunoprecipitation Chromatin Immunoprecipitation (ChIP) assay was performed with SABiosciences Corporations ChampionChiP One-Day kit (QIAGEN, Frederick, MD) following the manufacturers protocol with some modications (see Supplemental Experimental Procedures). Colony-Forming/Replating Assays These experiments were conducted as described previously (Li et al., 2008) with some modications (see Supplemental Experimental Procedures). Primary and Secondary BMT All experiments on mice were approved by the Institutional Animal Care and Use Committee (IACUC) of the University of Chicago. For primary BMT assays, normal BM cells of B6.SJL (CD45.1) mice were retrovirally transduced with corresponding constructs, through two rounds of spinoculation (Li et al., 2012b), and then injected by tail vein into lethally irradiated (960 rads) 8- to 10-week-old C57BL/6 (CD45.2) recipient mice with 3 3 105 donor cells per mouse plus a radioprotective dose of 1 3 106 whole BM cells. For secondary BMT assays, leukemic BM cells isolated from primary recipient mice were further transplanted into lethally irradiated C57BL/6 secondary recipient mice with the same dosage as primary BMT. Flow Cytometry Cells from PB, BM, spleen, or liver were harvested for analysis of immunophenotypes. After washing with phosphate-buffered saline (PBS) and blocking unspecic binding with Afnity Puried anti-mouse CD16/32 (eBioscience, San Diego, CA), cells were stained at 4 C with various antibodies diluted in Flow Cytometry Staining Buffer (eBioscience) for 30 min. Subsequently, cells were washed with PBS and resuspended in IC Fixation Buffer for ow cytometric analysis. APC-conjugated anti-mouse CD11b (Mac-1) and PE-conjugated anti-mouse Ly-6G (Gr-1) antibodies (eBioscience) were used for the analyses. Histopathology and Immunohistochemistry Tissue samples were xed in formalin, embedded in parafn, sectioned and stained with hematoxylin and eosin (H&E). Cytospins of PB and BM were stained with Wright-Giemsa. In Vitro RNA Pull-Down and RIP Assays THP-1 and MONOMAC-6 cells were used for the RNA pull-down and RIP assays, respectively, to investigate whether Lin28 protein binds directly to precursor miR-150 (see Supplemental Experimental Procedures). ACCESSION NUMBERS The microarray data have been deposited in the Gene Expression Omnibus (GEO) repository with the accession numbers GSE30258, GSE34184, and GSE34185. SUPPLEMENTAL INFORMATION Supplemental Information includes four gures and Supplemental Experimental Procedures and can be found with this article online at http://dx.doi. org/10.1016/j.ccr.2012.08.028.

ACKNOWLEDGMENTS The authors thank Dr. Janet D. Rowley for her support and constructive suggestions/comments and Dr. Peter Breslin for critical reading and constructive comments. We are also grateful to Drs. Gregory Hannon, Scott Hammond, Lin He, Scott Armstrong, and Michael Thirman for providing retroviral constructs. This work was supported in part by the National Institutes of Health (NIH) R01 Grant CA127277 (J.C.), American Cancer Society (ACS) Research Scholar grant (J.C.), The University of Chicago Committee on Cancer Biology (CCB) Fellowship Program (X.J.), LLS Special Fellowship (Z.L.), Gabrielles Angel Foundation for Cancer Research (J.C., Z.L., and H.H.), Leukemia & Lymphoma Society (LLS) Translational Research Grant (J.D.R. and J.C.), the Fidelity Foundation (J.D.R. and J.C.), R01 HL95896 (J.Z.), and the Intramural Research Program of the National Human Genome Research Institute, NIH (A.E. and P.P.L.), NIH P01 CA40046 (M.M.L.B), and P30 CA014599 (CCSG) (M.M.L.B). Received: February 23, 2012 Revised: June 21, 2012 Accepted: August 30, 2012 Published: October 15, 2012 REFERENCES Armstrong, S.A., Staunton, J.E., Silverman, L.B., Pieters, R., den Boer, M.L., Minden, M.D., Sallan, S.E., Lander, E.S., Golub, T.R., and Korsmeyer, S.J. (2002). MLL translocations specify a distinct gene expression prole that distinguishes a unique leukemia. Nat. Genet. 30, 4147. Armstrong, S.A., Kung, A.L., Mabon, M.E., Silverman, L.B., Stam, R.W., Den Boer, M.L., Pieters, R., Kersey, J.H., Sallan, S.E., Fletcher, J.A., et al. (2003). Inhibition of FLT3 in MLL. Validation of a therapeutic target identied by gene expression based classication. Cancer Cell 3, 173183. Behm, F.G., Raimondi, S.C., Frestedt, J.L., Liu, Q., Crist, W.M., Downing, J.R., Rivera, G.K., Kersey, J.H., and Pui, C.H. (1996). Rearrangement of the MLL gene confers a poor prognosis in childhood acute lymphoblastic leukemia, regardless of presenting age. Blood 87, 28702877. 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. hner, K., Bair, E., Fro hling, S., Schlenk, R.F., Tibshirani, R., Bullinger, L., Do hner, H., and Pollack, J.R. (2004). Use of gene-expression proling to Do identify prognostic subclasses in adult acute myeloid leukemia. N. Engl. J. Med. 350, 16051616. Chang, T.C., Zeitels, L.R., Hwang, H.W., Chivukula, R.R., Wentzel, E.A., Dews, M., Jung, J., Gao, P., Dang, C.V., Beer, M.A., et al. (2009). Lin-28B transactivation is necessary for Myc-mediated let-7 repression and proliferation. Proc. Natl. Acad. Sci. USA 106, 33843389. Chen, J., Odenike, O., and Rowley, J.D. (2010). Leukaemogenesis: more than mutant genes. Nat. Rev. Cancer 10, 2336. Dawson, M.A., Prinjha, R.K., Dittmann, A., Giotopoulos, G., Bantscheff, M., Chan, W.I., Robson, S.C., Chung, C.W., Hopf, C., Savitski, M.M., et al. (2011). Inhibition of BET recruitment to chromatin as an effective treatment for MLL-fusion leukaemia. Nature 478, 529533. Faber, J., Krivtsov, A.V., Stubbs, M.C., Wright, R., Davis, T.N., van den HeuvelEibrink, M., Zwaan, C.M., Kung, A.L., and Armstrong, S.A. (2009). HOXA9 is required for survival in human MLL-rearranged acute leukemias. Blood 113, 23752385. He, L., and Hannon, G.J. (2004). MicroRNAs: small RNAs with a big role in gene regulation. Nat. Rev. Genet. 5, 522531. He, L., Thomson, J.M., Hemann, M.T., Hernando-Monge, E., Mu, D., Goodson, S., Powers, S., Cordon-Cardo, C., Lowe, S.W., Hannon, G.J., and Hammond, S.M. (2005). A microRNA polycistron as a potential human oncogene. Nature 435, 828833. Heo, I., Joo, C., Cho, J., Ha, M., Han, J., and Kim, V.N. (2008). Lin28 mediates the terminal uridylation of let-7 precursor MicroRNA. Mol. Cell 32, 276284.

534 Cancer Cell 22, 524535, October 16, 2012 2012 Elsevier Inc.

Cancer Cell
miR-150 Is a Pivotal Gatekeeper in MLL Leukemia

Hess, J.L., Bittner, C.B., Zeisig, D.T., Bach, C., Fuchs, U., Borkhardt, A., Frampton, J., and Slany, R.K. (2006). c-Myb is an essential downstream target for homeobox-mediated transformation of hematopoietic cells. Blood 108, 297304. Hoffman, B., Amanullah, A., Shafarenko, M., and Liebermann, D.A. (2002). The proto-oncogene c-myc in hematopoietic development and leukemogenesis. Oncogene 21, 34143421. Jin, S., Zhao, H., Yi, Y., Nakata, Y., Kalota, A., and Gewirtz, A.M. (2010). c-Myb binds MLL through menin in human leukemia cells and is an important driver of MLL-associated leukemogenesis. J. Clin. Invest. 120, 593606. Krivtsov, A.V., and Armstrong, S.A. (2007). MLL translocations, histone modications and leukaemia stem-cell development. Nat. Rev. Cancer 7, 823833. Krivtsov, A.V., Feng, Z., Lemieux, M.E., Faber, J., Vempati, S., Sinha, A.U., Xia, X., Jesneck, J., Bracken, A.P., Silverman, L.B., et al. (2008). H3K79 methylation proles dene murine and human MLL-AF4 leukemias. Cancer Cell 14, 355368. Li, L., Piloto, O., Kim, K.T., Ye, Z., Nguyen, H.B., Yu, X., Levis, M., Cheng, L., and Small, D. (2007). FLT3/ITD expression increases expansion, survival and entry into cell cycle of human haematopoietic stem/progenitor cells. Br. J. Haematol. 137, 6475. Li, Y.J., Zhang, Y.X., Wang, P.Y., Chi, Y.L., Zhang, C., Ma, Y., Lv, C.J., and Xie, S.Y. (2012a). Regression of A549 lung cancer tumors by anti-miR-150 vector. Oncol. Rep. 27, 129134. Li, Z., Huang, H., Chen, P., He, M., Li, Y., Arnovitz, S., Jiang, X., He, C., Hyjek, E., Zhang, J., et al. (2012b). miR-196b directly targets both HOXA9/MEIS1 oncogenes and FAS tumour suppressor in MLL-rearranged leukaemia. Nat. Commun. 3, 688. Li, Z., Lu, J., Sun, M., Mi, S., Zhang, H., Luo, R.T., Chen, P., Wang, Y., Yan, M., Qian, Z., et al. (2008). Distinct microRNA expression proles in acute myeloid leukemia with common translocations. Proc. Natl. Acad. Sci. USA 105, 15535 15540. Lu, J., Guo, S., Ebert, B.L., Zhang, H., Peng, X., Bosco, J., Pretz, J., Schlanger, R., Wang, J.Y., Mak, R.H., et al. (2008). MicroRNA-mediated control of cell fate in megakaryocyte-erythrocyte progenitors. Dev. Cell 14, 843853. Mi, S., Li, Z., Chen, P., He, C., Cao, D., Elkahloun, A., Lu, J., Pelloso, L.A., Wunderlich, M., Huang, H., et al. (2010). Aberrant overexpression and function of the miR-17-92 cluster in MLL-rearranged acute leukemia. Proc. Natl. Acad. Sci. USA 107, 37103715. Milne, T.A., Dou, Y., Martin, M.E., Brock, H.W., Roeder, R.G., and Hess, J.L. (2005a). MLL associates specically with a subset of transcriptionally active target genes. Proc. Natl. Acad. Sci. USA 102, 1476514770. Milne, T.A., Martin, M.E., Brock, H.W., Slany, R.K., and Hess, J.L. (2005b). Leukemogenic MLL fusion proteins bind across a broad region of the Hox a9 locus, promoting transcription and multiple histone modications. Cancer Res. 65, 1136711374. Newman, M.A., and Hammond, S.M. (2010). Emerging paradigms of regulated microRNA processing. Genes Dev. 24, 10861092. Newman, M.A., Thomson, J.M., and Hammond, S.M. (2008). Lin-28 interaction with the Let-7 precursor loop mediates regulated microRNA processing. RNA 14, 15391549. ODonnell, K.A., Wentzel, E.A., Zeller, K.I., Dang, C.V., and Mendell, J.T. (2005). c-Myc-regulated microRNAs modulate E2F1 expression. Nature 435, 839843. Popovic, R., Riesbeck, L.E., Velu, C.S., Chaubey, A., Zhang, J., Achille, N.J., Erfurth, F.E., Eaton, K., Lu, J., Grimes, H.L., et al. (2009). Regulation of mir196b by MLL and its overexpression by MLL fusions contributes to immortalization. Blood 113, 33143322. Ramsay, R.G., Barton, A.L., and Gonda, T.J. (2003). Targeting c-Myb expression in human disease. Expert Opin. Ther. Targets 7, 235248. Rowley, J.D. (2008). Chromosomal translocations: revisited yet again. Blood 112, 21832189. llar, M.P., Zilles, O., Greil, J., and Slany, a-Cue Schreiner, S., Birke, M., Garc R.K. (2001). MLL-ENL causes a reversible and myc-dependent block of myelomonocytic cell differentiation. Cancer Res. 61, 64806486.

Shankar, D.B., Li, J., Tapang, P., Owen McCall, J., Pease, L.J., Dai, Y., Wei, R.Q., Albert, D.H., Bouska, J.J., Osterling, D.J., et al. (2007). ABT-869, a multitargeted receptor tyrosine kinase inhibitor: inhibition of FLT3 phosphorylation and signaling in acute myeloid leukemia. Blood 109, 34003408. Siomi, H., and Siomi, M.C. (2010). Posttranscriptional regulation of microRNA biogenesis in animals. Mol. Cell 38, 323332. Somervaille, T.C., Matheny, C.J., Spencer, G.J., Iwasaki, M., Rinn, J.L., Witten, D.M., Chang, H.Y., Shurtleff, S.A., Downing, J.R., and Cleary, M.L. (2009). Hierarchical maintenance of MLL myeloid leukemia stem cells employs a transcriptional program shared with embryonic rather than adult stem cells. Cell Stem Cell 4, 129140. Srivastava, S.K., Bhardwaj, A., Singh, S., Arora, S., Wang, B., Grizzle, W.E., and Singh, A.P. (2011). MicroRNA-150 directly targets MUC4 and suppresses growth and malignant behavior of pancreatic cancer cells. Carcinogenesis 32, 18321839. Stubbs, M.C., and Armstrong, S.A. (2007). FLT3 as a therapeutic target in childhood acute leukemia. Curr. Drug Targets 8, 703714. Takahashi, S. (2011). Downstream molecular pathways of FLT3 in the pathogenesis of acute myeloid leukemia: biology and therapeutic implications. J. Hematol. Oncol. 4, 13. Tusher, V.G., Tibshirani, R., and Chu, G. (2001). Signicance analysis of microarrays applied to the ionizing radiation response. Proc. Natl. Acad. Sci. USA 98, 51165121. Viswanathan, S.R., Daley, G.Q., and Gregory, R.I. (2008). Selective blockade of microRNA processing by Lin28. Science 320, 97100. Wang, G.G., Pasillas, M.P., and Kamps, M.P. (2005). Meis1 programs transcription of FLT3 and cancer stem cell character, using a mechanism that requires interaction with Pbx and a novel function of the Meis1 C-terminus. Blood 106, 254264. Wu, Q., Jin, H., Yang, Z., Luo, G., Lu, Y., Li, K., Ren, G., Su, T., Pan, Y., Feng, B., et al. (2010). MiR-150 promotes gastric cancer proliferation by negatively regulating the pro-apoptotic gene EGR2. Biochem. Biophys. Res. Commun. 392, 340345. Xiao, C., and Rajewsky, K. (2009). MicroRNA control in the immune system: basic principles. Cell 136, 2636. Xiao, C., Calado, D.P., Galler, G., Thai, T.H., Patterson, H.C., Wang, J., Rajewsky, N., Bender, T.P., and Rajewsky, K. (2007). MiR-150 controls B cell differentiation by targeting the transcription factor c-Myb. Cell 131, 146159. Yao, Q., Nishiuchi, R., Li, Q., Kumar, A.R., Hudson, W.A., and Kersey, J.H. (2003). FLT3 expressing leukemias are selectively sensitive to inhibitors of the molecular chaperone heat shock protein 90 through destabilization of signal transduction-associated kinases. Clin. Cancer Res. 9, 44834493. Yu, B.D., Hess, J.L., Horning, S.E., Brown, G.A., and Korsmeyer, S.J. (1995). Altered Hox expression and segmental identity in Mll-mutant mice. Nature 378, 505508. llar, M.P., and Slany, R.K. (2003). a-Cue Zeisig, B.B., Schreiner, S., Garc Transcriptional activation is a key function encoded by MLL fusion partners. Leukemia 17, 359365. llar, M.P., Schreiner, S., Martin, M.E., Fuchs, a-Cue Zeisig, B.B., Milne, T., Garc U., Borkhardt, A., Chanda, S.K., Walker, J., Soden, R., et al. (2004). Hoxa9 and Meis1 are key targets for MLL-ENL-mediated cellular immortalization. Mol. Cell. Biol. 24, 617628. Zhang, J., Luo, N., Luo, Y., Peng, Z., Zhang, T., and Li, S. (2012). microRNA150 inhibits human CD133-positive liver cancer stem cells through negative regulation of the transcription factor c-Myb. Int. J. Oncol. 40, 747756. Zuber, J., Rappaport, A.R., Luo, W., Wang, E., Chen, C., Vaseva, A.V., Shi, J., Weissmueller, S., Fellmann, C., Taylor, M.J., et al. (2011). An integrated approach to dissecting oncogene addiction implicates a Myb-coordinated self-renewal program as essential for leukemia maintenance. Genes Dev. 25, 16281640.

Cancer Cell 22, 524535, October 16, 2012 2012 Elsevier Inc. 535

Article
A Mouse Model of Rhabdomyosarcoma Originating from the Adipocyte Lineage

Cancer Cell

Mark E. Hatley,1,2,5 Wei Tang,3 Matthew R. Garcia,1,2,5 David Finkelstein,6 Douglas P. Millay,1 Ning Liu,1 Jonathan Graff,3 Rene L. Galindo,1,2,4,* and Eric N. Olson1,*
of Molecular Biology of Pediatrics 3Department of Developmental Biology 4Department of Pathology University of Texas Southwestern Medical Center, 5323 Harry Hines Boulevard, Dallas, TX 75390, USA 5Department of Oncology 6Department of Biostatistics St. Jude Childrens Research Hospital, 262 Danny Thomas Place, Memphis, TN 38105, USA *Correspondence: rene.galindo@utsouthwestern.edu (R.L.G.), eric.olson@utsouthwestern.edu (E.N.O.) http://dx.doi.org/10.1016/j.ccr.2012.09.004
2Department 1Department

SUMMARY

Rhabdomyosarcoma (RMS) is an aggressive skeletal muscle-lineage tumor composed of malignant myoblasts that fail to exit the cell cycle and are blocked from fusing into syncytial muscle. Rhabdomyosarcoma includes two histolopathologic subtypes: alveolar rhabdomyosarcoma, driven by the fusion protein PAX3-FOXO1 or PAX7-FOXO1, and embryonal rhabdomyosarcoma (ERMS), which is genetically heterogeneous. Here, we show that adipocyte-restricted activation of Sonic hedgehog signaling through expression of a constitutively active Smoothened allele in mice gives rise to aggressive skeletal muscle tumors that display the histologic and molecular characteristics of human ERMS with high penetrance. Our ndings suggest that adipocyte progenitors can be a cell of origin for Sonic hedgehog-driven ERMS, showing that RMS can originate from nonskeletal muscle precursors.
INTRODUCTION Rhabdomyosarcoma (RMS) is an aggressive skeletal musclelineage malignancy and the most common soft tissue sarcoma in children (Barr and Womer, 2009). RMS is comprised of skeletal muscle precursors that fail to exit the cell cycle and are irreversibly blocked from differentiating into syncytial muscle. RMS is typically divided into two histopathologic subgroups, each with distinct clinical features: embryonal RMS (ERMS), which is the more common subtype, and alveolar RMS (ARMS), which is notoriously more aggressive. The genetic lesions that initiate ARMS are well known with three-quarters of ARMS tumors having chromosomal translocations that result in expressing fusion proteins combining the DNA binding domain of either PAX3 or PAX7 transcription factors with the transcriptional activation domain of FOXO1A (Barr, 2001). Because PAX3 and PAX7 have critical roles in normal muscle development, the fusion proteins presumably use the DNA binding domains of the PAX proteins to drive malignant myogensis-related developmental programs (Galindo et al., 2006; Keller and Capecchi, 2005; Wang et al., 2008). ERMS accounts for 75% of RMS cases and is associated with younger age of onset (typically under ten years) and demonstrates a predilection for tissues of the head and neck. The molecular underpinnings and cellular origins of ERMS remain poorly understood. Although most cases of ERMS occur as sporadic, nonheritable tumors, ERMS also associates with familial syndromes caused by mutations in prominent oncogenesis-related signaling pathways, such as p53, Ras, and Sonic hedgehog (Shh) (Estep et al., 2006; Johnson et al., 1996; Li and Fraumeni, 1969).

Signicance Rhabdomyosarcoma is the most common soft tissue malignancy in children. Despite aggressive chemotherapy, radiotherapy, and surgery, clinical outcomes for RMS have not improved for three decades, emphasizing the need to uncover the molecular underpinnings of the disease. Here, we describe a Sonic hedgehog-driven mouse model of ERMS that mimics the histopathological and molecular characteristics of human ERMS, and with notably decreased latency, high penetrance, and a restricted anatomic location when compared to previous models. In addition to yielding mechanistic insights into ERMS, this mouse model also now provides a robust preclinical platform to explore therapeutic strategies to improve RMS treatment.
536 Cancer Cell 22, 536546, October 16, 2012 2012 Elsevier Inc.

Cancer Cell
Adipocyte Origin of Rhabdomyosarcoma

ROSA26-lox Stop lox-SmoM2 ROSA26


loxP

aP2-Cre aP2 Cre

STOP
loxP

SmoM2

D
H&E
ROSA26 SmoM2

E
Massons trichrome

F
Ki67

C 100
Tumor-free survival 80 60 40 20 0

SmoM2

G
Desmin

H
MyoD

I
Myogenin

aP2-Cre;SmoM2 0 50 100 150 Days 200 250

Figure 1. ERMS Caused by Activation of the Sonic Hedgehog Pathway in the aP2 Lineage
(A) Schematic of the conditional SmoM2 allele and aP2-Cre-mediated recombination. (B) Gross view of tumors in aP2-Cre;SmoM2 mice. (C) Kaplan-Meier survival curves illustrated tumor-free survival of aP2-Cre;SmoM2/+ mice (blue line, n = 47) compared to SmoM2/+ littermate controls (red line, n = 19) (p = < 0.0001). (D) Cross-sectional histology of tumors with H&E staining. Arrow points to rhabdomyoblast and arrowhead points to strap cell. (E) Massons trichrome. (FI) Sections of tumors immunostained with Ki67, a marker of proliferating cells (F), and Desmin (G), MyoD1 (H), and Myogenin (I), diagnostic of ERMS. Scale bar, 50 mm (D, E, and GI) and 200 mm (F). See also Figure S1.

The Shh pathway is an evolutionarily conserved signaling pathway with noted roles in development and in tumorigenesis. Normally, in the absence of extracellular Shh ligand, the Shh signaling pathway is inhibited by the Patched (Ptch) transmembrane receptor, which dominantly represses the Smoothened (Smo) G protein coupled receptor. Upon binding of Shh to Ptch, Smo is freed from Ptch-mediated inhibition and activates the Gli family of transcription factors (Gli1, Gli2, and Gli3) that then drive patterns of gene expression critical for various aspects of development (Lum and Beachy, 2004). Shh-associated diseases include nevoid basal cell carcinoma syndrome (Gorlin syndrome), an autosomal dominant condition caused by heterozygous germ-line mutations in PTCH1, which drives basal cell carcinoma, medulloblastoma, and ERMS. Activation of the Shh pathway is also found in spontaneous ERMS tumors, associating with either loss of Shh pathway negative regulators, such as PTCH1 or PTCH2 or Suppressor of Fused (SUFU), or by the activation of the downstream transcriptional effector of Shh signaling, Gli (Paulson et al., 2011; Tostar et al., 2006). Consistent with human ERMS, introducing Shh-pathwayactivating mutations in mouse models induces tumorigenesis. Transgenic mice with heterozygous deletion of Ptch1 develop tumors histologically consistent with ERMS but with a low level of penetrance (10%) (Corcoran and Scott, 2001; Hahn et al., 2000). Mice with heterozygous loss of Sufu in combination with p53 loss develop ERMS with 9% penetrance (Lee et al., 2007). The most robust mouse model of ERMS utilizes a conditional,

constitutively activate Smo allele (SmoM2) controlled by a ubiquitously expressed, inducible Cre transgene, CAGGS-CreER (Mao et al., 2006). Although these models point to the involvement of the Shh pathway in the pathogenesis of ERMS, the varied anatomic location, relatively low penetrance of tumorigenesis and occurrence of other tumor types in these models limit their usefulness as a preclinical platform. We have generated a transgenic mouse model in which overexpression of the oncogenic SmoM2 allele in the adipocyte-restricted aP2 lineage induces tumors that closely resemble human ERMS. RESULTS SmoM2 Expression in the aP2 Lineage Causes ERMS We initially sought to explore the role of Shh signaling in adipocyte development and metabolism, using an adipose protein 2 (aP2)-Cre transgenic driver (Tang et al., 2008) to conditionally express constitutively activated Smoothened, SmoM2 (Mao et al., 2006) in an adipocytic-specic pattern (Figure 1A). Surprisingly, 80% of mice with the genotype aP2-Cre;SmoM2/+ developed large, aggressive tumors in the head and ventral neck by 2 months of age (Figures 1B and 1C). The tumors ranged in greatest diameter from 1 to 2 cm, were tan/pink/white and solid upon sectioning, and well demarcated from the surrounding nonneoplastic tissue. Histologic examination revealed ERMS-type morphology: the tumors displayed dense cellularity, comprised of a spectrum of cells that were small and round
Cancer Cell 22, 536546, October 16, 2012 2012 Elsevier Inc. 537

Cancer Cell
Adipocyte Origin of Rhabdomyosarcoma

aP2-Cre;SmoM2

aP2-Cre;SmoM2;R26-LacZ

B
Relative expression

70 60 50 40 30 20 10 0

SCM, SmoM2 SCM, aP2-Cre;SmoM2 Tumor, aP2-Cre;SmoM2

Smo Gli1 Gli3 Ptch1 MyoD Myogenin Pax7 Myf5

C
Relative expression

80 70 60 50 40 30 20 10 0

to cells that were elongated and spindled, with brightly eosinophilic cytoplasm, pleomorphic nuclei, and visible cross-striations (Figures 1D and 1E). Immunohistochemistry (IHC) for Ki67 showed that the tumor cells were briskly mitotic (Figure 1F). IHC also revealed strong expression of the muscle-specic intermediate lament Desmin and nuclear staining for the muscle-specic transcription factors Myogenin and MyoD1 (Figures 1G1I). These ndings are diagnostic for RMS. Extensive analyses of the aP2 promoter in transgenic mice have documented its specicity for the adipocyte lineage and its exclusion from the skeletal muscle lineage (He et al., 2003; Ross et al., 1990; Tang et al., 2008; Urs et al., 2006). Thus, it was surprising to observe RMS tumorigenesis in aP2-Cre; SmoM2 mice. To test whether aP2-Cre activation of the SmoM2 allele altered or interfered with normal skeletal muscle and/or adipose development, we performed histologic analysis of the sternocleidomastoid (SCM), quadriceps femoris, intrascapular brown fat, and inguinal white fat from 4- to 8-weekold aP2-Cre;SmoM2/+ and SmoM2/+ control mice. We observed no gross abnormalities in muscle or adipose tissue in either cohort (Figure S1 available online). Therefore, activation of the hedgehog pathway by aP2-Cre-mediated expression of the SmoM2 allele did not evoke general defects in myogenesis or adipogenesis. Activation of Hedgehog Signaling in aP2-Cre;SmoM2/+ Tumors We analyzed in vivo the expression of the Cre-responsive R26-LacZ reporter to conrm that the RMS tumors originate from cell-autonomous activation of the SmoM2 allele. Intense, homogenous b-galactosidase staining of the tumors revealed that they formed as a consequence of cell-autonomous activation by aP2-Cre (Figure 2A). Next, to conrm that the Shh pathway was activated in tumors of aP2-Cre;SmoM2/+ mice, we compared the expression of Shh target genes in control muscle and tumor tissue from both aP2-Cre;SmoM2/+ and SmoM2/+ littermates. The tumors exhibited increased expression of Smo and Shh responsive genes, including Gli1, Gli3, and Ptch1 (Figure 2B), whereas gene expression in SCM muscle was unchanged. These ndings indicate that the Shh pathway is robustly activated in the ERMS tumors but not in the nonneoplastic skeletal muscle of the aP2-Cre;SmoM2/+ mice, suggesting an origin of tumorigenesis distinct from skeletal muscle lineage precursors. aP2-Cre;SmoM2/+ Tumors Display an Embryonal Muscle Gene Signature To further determine the extent to which the aP2-Cre;SmoM2/+ tumors resembled ERMS, we isolated RNA from tumor tissue and compared the gene expression prole to that of SCM skeletal muscle of SmoM2/+ littermates. Skeletal muscle regulatory genes (MyoD1, Myogenin, Pax7, and Myf5) and embryonic/ perinatal myosins (Myh3 and Myh8) displayed markedly increased expression in the aP2-Cre;SmoM2/+ tumors when

Skeletal Tumor Skeletal Tumor Skeletal Tumor Skeletal Tumor muscle muscle muscle muscle Myh3 Relative expression 10 8 6 4 2 0 Skeletal Tumor muscle Ckm Myh8

1500 Relative expression 1250 1000 750 500 250 0

Skeletal Tumor muscle Myh1

D
Relative expression

1.5

Myh4

1.0

0.5

Skeletal Tumor muscle

Skeletal Tumor muscle

Skeletal Tumor muscle

Figure 2. Sonic Hedgehog Activation in aP2-Cre;SmoM2 Tumors


(A) b-galactosidase enzymatic staining of cross sections of aP2-Cre; SmoM2;R26-LacZ reporter mouse tumors and aP2-Cre;SmoM2 tumors illustrating broad homogeneous staining. Scale bar, 5 mm. (B) Increased expression on Sonic hedgehog pathway target genes in the aP2-Cre;SmoM2/+ tumors compared to mature SCM of aP2-Cre;SmoM2/+ and SmoM2/+ animals by real-time PCR. Data are shown as fold change of gene expression normalized to 18S and expressed relative to SCM of SmoM2/+. Results are mean SEM (n = 4). p values by two-tailed, unpaired Students t test are as follows: Smo (p = 0.0015), Gli1 (p < 0.0001), Gli3 (p < 0.0001), and Ptch1 (p < 0.0001). (C and D) Expression of embryonic muscle development genes (C) and markers of terminally differentiated skeletal muscle (D) in aP2-Cre;SmoM2/+ tumors compared to mature SCM as detected by real-time PCR. Data are shown as fold change of gene expression normalized to 18S and expressed relative to SCM of SmoM2/+. Results are mean SEM (n = 3). p values

by two-tailed, unpaired Students t test are as follows: MyoD1 (p = 0.0011), Myogenin (p = 0.0007), Pax7 (p = 0.0009), Myf5 (p = 0.0039), Myh3 (p = 0.0165), Myh8 (p = 0.0007), Myh1 (p < 0.0001), Myh4 (p < 0.0001), and Ckm (p < 0.0001).

538 Cancer Cell 22, 536546, October 16, 2012 2012 Elsevier Inc.

Cancer Cell
Adipocyte Origin of Rhabdomyosarcoma

PCA mapping (64.6%) SCM Tumor

B
-log10 (p value unequal t test)

25 20 15
Mlf1 2610042G18Rik

lgf2 Myh8 Hes6 Dlk1 Sct Dlk1 Myh4 Mybpc2 Chrna1 Sln Ahkrd1 Ncam1 Myh8 Chrng

Figure 3. Comparative Transcriptome Analysis of Mouse ERMS and Skeletal Muscle


(A) Principle component analysis (PCA). Three PCA coordinates describe 64.6% of the total data variation (PC1, 52.9%; PC2, 5.99%; and PC3, 5.71%). Red, aP2-Cre;SmoM2/+ tumors (n = 12). Green, SmoM2/+ SCM (n = 3). (B) Volcano plot of the Log10 of the p value versus the Log2 of the fold-difference in expression in the aP2-Cre;SmoM2/+ tumors compared to the normal skeletal muscle. (C) Unsupervised hierarchical clustering analysis. Each column represents a distinct sample, and each row represents a distinct gene. (D) Comparison of aP2-Cre;SmoM2 tumors to published mouse ERMS models. Gene expressed in both the published mouse ERMS models and the aP2-Cre;SmoM2/+ tumors (19,878 genes) plotted on graph. Sixty-seven percent of gene pairs show agreement in genes upregulated (red) or downregulated (green) between the previously published mouse ERMS models and the aP2Cre;SmoM2 tumors. Spearman correlation = 0.59 and AGDEX = 0.71. (E) Comparision of aP2-Cre;SmoM2/+ tumors to human ERMS. Genes expressed in both the published human ERMS and the aP2-Cre;SmoM2/+ tumors (13,282 genes) plotted on graph. Fifty-eight percent of orthologous gene pairs show agreement in genes upregulated (red) or downregulated (green) between the human ERMS and the aP2Cre;SmoM2 tumors. Spearman correlation = 0.29 and AGDEX = 0.3. See also Figure S2.

10 5 0

-5

0 Log2ratio (AS/SCM)
Quadrant of agreement

SCM

Tumor

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18

D
Logratio (ERMS/wt) San Antonio
5
Quadrant of disagreement

-5

Quadrant of agreement

Quadrant of disagreement

-2 -1 0 1 2 standard deviations

-4

-2 0 2 4 6 Logratio (ERMS/SCM) Dallas

E
Logratio (ERMS/normal muscle) human

10 5 0 -5

Quadrant of disagreement

Quadrant of agreement

-10

Quadrant of agreement

Quadrant of disagreement

-4

-2 0 2 4 6 Logratio (ERMS/SCM) mouse

compared to the SCM (Figure 2C). In contrast, markers of mature, differentiated skeletal muscle (Myh1, Myh4, and Ckm) were dramatically diminished in aP2-Cre;SmoM2/+ tumors when compared to SCM (Figure 2D). This embryonal muscle gene expression pattern is consistent with mouse and human ERMS and is indicative of an arrested skeletal myoblastic tissue. Gene Expression Proling of aP2-Cre;SmoM2 Tumors To further characterize the tumors of the aP2-Cre;SmoM2/+ mice, we used mRNA expression proling to compare the gene expression proles of aP2-Cre;SmoM2/+ tumors and SCM. The transcriptome of the aP2-Cre;SmoM2/+ tumors was distinct from that of SCM using principle component analysis to group samples based on gene expression (Figure 3A). Focusing on changes of gene expression with a log2 ratio greater than 2.5 and a p value less that 0.0001 by Bonferroni, we identied 100 genes that were differentially expressed between aP2-Cre; SmoM2/+ tumors and normal SCM of SmoM2/+ littermates (Figures 3B and 3C). Consistent with the developing muscle gene pattern, we noted that perinatal myosin heavy chain Myh8 was among the genes expressed at high levels in the tumors compared to normal SCM and mature myosin heavy

chain Myh4 was among the genes that exhibited low expression in the tumors compared to SCM (Figures 2 and 3B). These analyses also identied several other genes previously shown to play a role in RMS pathogenesis, including Igf2, Dlk1, Ankrd1, Chrng, and Ncam1 (El-Badry er et al., 1998; Ishiet al., 1990; Gattenloehner et al., 1998; Glu guro et al., 2008; Rezvani et al., 2012). Real-time PCR validation of the array is shown in Figure S2A. During muscle development, a subset of myogenic progenitors forgoes terminal differentiation and instead become mononuclear satellite cells associated with the myober basal lamina. Muscle satellite cells are responsible for the postnatal growth, repair, and maintenance of skeletal muscle (Seale and Rudnicki, 2000). Satellite cells remain quiescent until activated by muscle injury, when they serve as muscle-specic stem cells that generate myogenic precursors and repair injured muscle (Shadrach and Wagers, 2011). Mice with loss-of-function of p53 and gain-of-function of the Shh pathway through the loss of Ptch1 in myoblasts and satellite cell lineages develop tumors that resemble human ERMS and display a gene expression signature consistent with satellite cell activation (Rubin et al., 2011). Interestingly, the tumors arising from the adipocytespecic aP2-Cre driver expressed numerous genes associated with satellite cell activation (Fukada et al., 2007), including Rho GDP dissociation inhibitor gamma (Arhgdig), ankyrin repeat
Cancer Cell 22, 536546, October 16, 2012 2012 Elsevier Inc. 539

Cancer Cell
Adipocyte Origin of Rhabdomyosarcoma

100 Tumor-free survival 80 60 40 20 0

SmoM2

aP2-Cre;SmoM2 aP2-Cre;SmoM2;Cdkn2a-Flox

50

100 150 Days

200

250

son et al., 2011; Williamson et al., 2010). To test for potential cooperativity between Cdkn2a and hedgehog-driven ERMS, we intercrossed a conditional Cdkn2aFlox allele into our model (Aguirre et al., 2003). aP2-Cre;Cdkn2aFlox/Flox mice did not develop ERMS, whereas deletion of Cdkn2a in aP2-Cre; SmoM2/+;Cdkn2aFlox/Flox mice triggered decreased latency and increased penetrance of tumorigenesis with all mice having tumors by 55 days (Figure 4). These results demonstrate that the Shh pathway and Cdkn2a can actively drive ERMS oncogenicity in vivo. Sonic Hedgehog Activation in the Embryonic Muscle Lineage and Postnatal Satellite Cells We next crossed the SmoM2 allele to mice bearing several muscle lineage-restricted Cre-drivers to determine whether activation of the Shh pathway by expression of the SmoM2 oncogene in the muscle lineage would phenocopy the tumors of the aP2Cre;SmoM2/+ mice. Activation of the SmoM2 allele early in muscle development with Myf5-Cre, Pax3-Cre, or MyoD1-Cre resulted in embryonic lethality without tumor formation, whereas mice with activation of SmoM2 in terminally differentiated skeletal muscle with MCK-Cre were viable with no evidence of tumorigenesis (Figure S3). In contrast, activation of SmoM2 with Myogenin-Cre, which is specic for early muscle differentiation, resulted in tongue tumors in 100% of the mice (Figure 5A) but no tumors in the anterior neck, as seen in aP2-Cre;SmoM2/+ mice. Specically, the Myogenin-Cre;SmoM2/+ mice were runted compared to littermates and were sacriced between 19 and 39 days for lack of weight gain and ill appearance. None of the MyogeninCre;SmoM2/+ mice had visible tumors prior to sacrice; however, all Myogenin-Cre;SmoM2/+ mice had tumors visible upon necropsy throughout the tongue (Figure 5B). The MyogeninCre;SmoM2/+ tumors appeared histologically similar to the aP2-Cre;SmoM2/+ tumors and stained positive both MyoD1 and Myogenin. We detected no signicant difference in the gene expression prole of the Myogenin-Cre;SmoM2/+ tumors and the aP2-Cre;SmoM2/+ (Figures 5C and 5D). The embryonic lethality of SmoM2 expression early in muscle development with Myf5-Cre, Pax3-Cre, and MyoD1-Cre and the more aggressive phenotype with the Myogenin-Cre suggests that aP2-Cre does not globally activate SmoM2 expression in the muscle lineage. Previous studies illustrate that satellite cells are a potential origin for ERMS (Hettmer et al., 2011; Rubin et al., 2011; Tifn et al., 2003). Given the activated satellite cell signature of the aP2-Cre;SmoM2/+ tumors, we sought to formally test whether satellite cells are an origin of the aP2-Cre;SmoM2/+ tumors. The Pax7-CreERT2 allele allows for specic SmoM2 postnatal induction in both quiescent and activated satellite cells (Lepper et al., 2009). We crossed the SmoM2 mouse with the satellite cell Cre-driver, Pax7-CreERT2 and administered tamoxifen at day-of-life 1 or day-of-life 14 (Figure S3G). Gene expression from the SCM containing only a small fraction of satellite cells from the Pax7-CreERT2;SmoM2/+ and SmoM2/+ Cre-negative littermates revealed that the SmoM2 (that has a carboxy terminal YFP fusion) was activated following tamoxifen administration (Figure S3H). None of the Pax7-CreERT2;SmoM2/+ animals injected with tamoxifen at either day 1 (n = 8) or day 14 (n = 9)

Figure 4. Cdkn2a Loss Cooperates in aP2-Cre;SmoM2 Mouse Tumors


Kaplan-Meier survival curves illustrate tumor-free survival of aP2-Cre; SmoM2/+;Cdkn2aFlox mice (green line, n = 27) compared to data from Figure 1C aP2-Cre;SmoM2/+ (blue line, n = 47) and SmoM2/+ littermate controls (red line, n = 19). p = 0.0337.

domain 1 (Ankrd1), (fetal) nicotinic cholinergic receptor gamma polypeptide (Chrng), distal-less homeobox 2 (Dlx2), hairy and enhancer of split (Hes6), low density lipoprotein receptor-related protein 4 (Lrp4), Otogelin (Otog), Secretin (Sct), troponin C (Tnnc1), troponin T1 (Tnnt1), cardiac troponin T2 (Tnnt2), and unc-5 homolog B (Unc5b) (Figure S2B). Twenty percent of all the genes with increased expression of 5-fold or greater in the aP2-Cre;SmoM2/+ tumors display increased expression in activated satellite cells compared to quiescent satellite cells. Thus, despite arising from the aP2 adipocyte lineage, the aP2-Cre; SmoM2/+ tumors displayed a gene expression prole similar to other murine ERMS models and activated satellite cells. These ndings suggest that the satellite cell gene expression signature is likely reective of the ERMS tumor pathology rather than the cell type of origin of the tumor. To further test how the aP2-Cre;SmoM2/+ tumors resemble previously reported mouse ERMS models, we compared the gene expression pattern of our model to that of previously published mouse ERMS. The comparison group contained mouse tumors that developed from the conditional deletion of Ptch1ox/+ and Trp53ox/ox alleles with Myf5-Cre, Pax3-Cre, and Pax7-CreERp, respectively (Rubin et al., 2011). Overall, 67% of the 19,878 probe pairs showed agreement in gene expression between our model and the previously published models (Figure 3D). Next, we used cross-species comparison of the mouse and human ERMS transcriptomes to determine how closely the aP2-Cre;SmoM2/+ tumors resemble human ERMS. We observed a statistically signicant match between the mouse aP2-Cre;SmoM2/+ tumors and human ERMS (Figure 3E). Of the 13,282 ortholog gene pairs, 58% showed agreement in gene expression between mouse and human. Together, these data conrm that the aP2-Cre;SmoM2/+ mouse tumors accurately model the transcriptome of both previously published mouse models and human ERMS. Cooperativity between SmoM2 Activation and Cdkn2a Loss The tumor suppressor CDKN2A is one of the most commonly mutated alleles in human cancer (Beroukhim et al., 2010; Bignell et al., 2010). Homozygous deletions of CDKN2A have been reported in ERMS (Chen et al., 2007; Iolascon et al., 1996; Paul540 Cancer Cell 22, 536546, October 16, 2012 2012 Elsevier Inc.

Cancer Cell
Adipocyte Origin of Rhabdomyosarcoma

100 80 60 40 20 0

B
MCK-Cre;SmoM2

Tumor-free survival

H&E

Figure 5. Tongue ERMS in Myogenin-Cre; SmoM2 Mice


(A) Kaplan-Meier survival curves illustrated tumorfree survival of Myogenin-Cre;SmoM2/+ mice (green line, n = 12) and of MCK-Cre;SmoM2/+ (red line, n = 18) compared to tumor-free survival of aP2-Cre;SmoM2/+ from Figure 1C (blue line). (B) Tumor histology. (a) Gross cross-section of tongue with tumor outlined by white, dashed line. (b) H&E section illustrating cross striations. (c and d) Immunochemistry illustrating nuclear MyoD1 and Myogenin. Scale bars, 2 mm (a) and 50 mm (b, c, and d). (C) Principle component analysis for mouse Myogenin-Cre;SmoM2/+ tumors (green, n = 3) and aP2-Cre;SmoM2/+ tumors (red, n = 12). (D) Volcano plot of the Log10 of the p value versus the Log2 of the fold-difference in expression in the aP2-Cre;SmoM2/+ tumors compared to the Myogenin-Cre;SmoM2/+ tumors. See also Figure S3.

aP2-Cre;SmoM2 Myog-Cre;SmoM2

MyoD d

Myogenin

100

200

300 400 Days

500

600

Myogenin tumors

aP2 tumors

D
6
Ppfibp1

-log10(p value unequal var t test)

4 2 0 -3

E430014K09Rik

Mmp10 Spt1

Spt1 Muc10 Muc10

-2 -1 0 1 2 Log2ratio (AS/SMYOG)

developed tumors by 150 days, arguing against the satellite cell population as the origin of the ERMS in the aP2-Cre;SmoM2/+ model. aP2-Cre Is Active in Adipose Tissue but Not Skeletal Muscle The development of tumors that resemble human ERMS from activation of the oncogenic SmoM2 allele in the aP2 adipocyte lineage was surprising. However, despite arising from the aP2 lineage, the aP2-Cre;SmoM2/+ tumors did not express aP2 or other genes of the adipocyte lineage (Figure S4A). Consistent with previous reports, we determined that aP2 is expressed in the progenitors from the stromal vascular fraction of brown adipose tissue (BAT) and white adipose tissue (WAT), as well as mature adipocytes. However, aP2 was not expressed from satellite cells or the SCM, gastrocnemius, or quadriceps femoris skeletal muscles (Figure S4B). The aP2-Cre transgenic used in this study was previously characterized and reported not to be expressed in the muscle lineage (Tang et al., 2008). Moreover, the identical 5.4 kb aP2 promoter element was previously shown to be adipose-specic and to result in hibernoma formation, not ERMS, when fused to simian virus 40 Large T antigen (He et al., 2003; Ross et al., 1990, 1992). Another aP2-Cre transgene containing the same 5.4 kb aP2 promoter has also been shown to be inactive in the muscle lineage but to be active primarily in BAT and WAT and in the developing cartilage, vertebra, and dorsal root ganglia (He et al., 2003; Urs et al., 2006). To conrm the pattern of expression of aP2-Cre and to further dene the origins of tumors in the aP2-Cre;SmoM2/+ mice, we again turned to aP2-Cre;R26-LacZ reporter mice. The aP2Cre was activated in both BAT and WAT but showed no detectable expression in skeletal muscle (Figure 6A), suggesting that RMS in these mice arises from a nonskeletal muscle origin. We compared the gene expression in isolated intrascapular

BAT, inguinal WAT, SCM, and gastrocnemius skeletal muscle, and tumor from the aP2-Cre;SmoM2/+ and SmoM2/+ Cre-negative littermates. Although the oncogenic SmoM2 allele was activated in the fat lineage, the tumors did not express adipocyte lineage genes, such as aP2, Pparg, and Ucp1 (Figures S4C S4E). Also, the normal skeletal muscle from the aP2-Cre; SmoM2/+ mice did not display increased Smo and Shh target genes, suggesting that SmoM2 was not activated in the muscle lineage. To further test whether the aP2-Cre transgene is inactive in skeletal muscle, we bred aP2-Cre mice to an additional reporter allele, R26-YFP. A western blot for YFP of whole tissue lysates of isolated intrascapular BAT, inguinal WAT, and SCM from aP2Cre;R26-YFP and R26-YFP littermates at P28 demonstrated no evidence of YFP expression in skeletal muscle (Figures S4F and S4G). Together, these data illustrate that the aP2-Cre transgene is not expressed in skeletal muscle, consistent with the notion that the aP2-Cre;SmoM2/+ ERMS tumors originate outside the skeletal muscle lineage. ERMS in Adipose Tissue To better delineate the tissue compartment giving rise to the aP2-Cre;SmoM2 ERMS tumors, we sacriced animals at P14, before tumor masses were grossly visible, with the notion of capturing microscopic foci of RMS arising in situ. The dissected SCM of a representative animal contained a small nodule clearly distinct from adjacent normal brown and white adipose tissues and SCM skeletal muscle (Figure 6B). The developing tumor was completely surrounded by nonneoplastic adipose tissue adjacent to and clearly separated from the SCM. The tumor displays the same appearance and histological characteristics as the previously described aP2-Cre;SmoM2/+ tumors. These ndings further support the notion that the ERMS in the aP2-Cre;SmoM2/+ mice develop outside of the muscle lineage.
Cancer Cell 22, 536546, October 16, 2012 2012 Elsevier Inc. 541

Cancer Cell
Adipocyte Origin of Rhabdomyosarcoma

BAT

WAT

Gastrocnemius

SCM

Figure 6. aP2-Cre Expression in Fat Lineages and aP2-Cre;SmoM2 Tumors in Adipose Tissue Adjacent to Muscle
(A) Mice bearing Cre-drivers aP2-Cre, Myf5-Cre, and Myogenin-Cre were bred to R26-LacZ reporter mice. The BAT, WAT, gastrocnemius, and SCM were isolated from P28 mice and stained with X-gal to determine b-galactosidase activity. Scale bars, 5 mm. (B) Serial sections through neck tissue with a visible node at P14 on necropsy and stained with H&E. (Left) Level below nodule illustrating normal SCM and adjacent adipose tissue. Arrow and arrowhead point to SCM and adipose tissue, respectively. Scale bar, 200 mm. (Middle) Section illustrating the tumor isolated within the adipose tissue. The open arrowhead points to tumor. Scale bar, 200 mm. (Right) High power view of tumor. Scale bar, 20 mm. See also Figure S4.

Myog-Cre

Myf5-Cre

aP2-Cre

WT

DISCUSSION Through adipocyte-restricted activation of the hedgehog pathway, we have developed a highly efcient mouse model of ERMS. Our ndings demonstrate that constitutive activation of the hedgehog pathway by an oncogenic Smoothened allele in the aP2 adipocyte lineage of mice results in the formation of aggressive head and neck tumors that mimic human ERMS histologically and molecularly. Though we cannot completely exclude the unlikely possibility that aP2-Cre is expressed in a rare population of myogenic cells, together these ndings strongly argue that ERMS driven by the Shh pathway can originate from tissue histologically distinct from muscle. Although activation of the Shh pathway inhibits fat formation in invertebrates and mammalian cellular models acting upstream of PPARg (Suh et al., 2006), the Shh pathway has well documented roles in skeletal muscle development and in the pathogenesis of ERMS (Barlow et al., 1997; Fan and nsterberg Tessier-Lavigne, 1994; Johnson et al., 1994; Mu et al., 1995; Zibat et al., 2010). The role of the Shh pathway in the development of ERMS is also supported by human cancer predisposition syndromes that result in Shh pathway activation and recent genomic interrogation of human ERMS samples. Gorlin syndrome, or nevoid basal cell carcinoma syndrome, is a human genetic disorder resulting from germline inactivating
542 Cancer Cell 22, 536546, October 16, 2012 2012 Elsevier Inc.

mutations of PTCH1, which encodes a negative regulator of the Shh pathway and displays an increased incidence of ERMS. Increased expression of GLI1 and PTCH1 in ERMS compared to ARMS has been reported (Paulson et al., 2011; Pressey et al., 2011). Zibat et al. (2010) found that high levels of expression of Shh target genes, PTCH1, GLI1, GLI3, and MYF5 are a common feature of ERMS and fusion gene-negative ARMS and not of fusion gene-positive ARMS. High expression of PTCH1 also correlated with poor survival (Zibat et al., 2010). These ndings illustrate that the hedgehog pathway inuences RMS pathogenesis. The tumor-free survival curve of the aP2-Cre;SmoM2/+ mice displays a biphasic appearance with some mice never developing tumors. The early development of tumors with approximately 50% of mice exhibiting tumors by 28 days of life shows that activation of SmoM2 is sufcient to initiate ERMS tumorigenesis. However, the incomplete penetrance of the model suggests that other genetic hits are likely necessary for complete neoplastic transformation and/or tumor progression. For example, the tumor suppressor CDKN2A encoding p16INK4A and p14ARF (p19Arf in mice) is commonly deleted in human ERMS, which causes concomitant deregulation of the Rb and p53 pathways (Chen et al., 2007; Iolascon et al., 1996; Paulson et al., 2011). Deletion of Cdkn2a was previously shown to cooperate with a c-Met-driven ERMS mouse model (Sharp et al., 2002). The deletion of Cdkn2a in the aP2-Cre;SmoM2/+ model resulted in complete penetrance, illustrating cooperativity between the Shh pathway and Ink4a/Arf in RMS. Through mouse crosses deleting Ptch1 and p53 with multiple muscle-specic Cre drivers, Rubin et al. (2011) recently demonstrated that ERMS developed upon inactivating these loci in muscle stem cells (satellite cells), as well as proliferating and maturing myoblasts. All ERMS tumors exhibited the same gene

Cancer Cell
Adipocyte Origin of Rhabdomyosarcoma

signature despite the cell of origin of the tumors, suggesting that the gene expression prole was more reective of the nal state of the tumor rather than the lineage from which it arose (Rubin et al., 2011). Though the aP2-Cre transgene is not active in the muscle lineage, the gene expression prole of the ERMS from aP2-Cre;SmoM2/+ mice is remarkably similar to previous murine ERMS models and to that of human ERMS. Thus, in multiple and seemingly unrelated settings, activated satellite gene signatures are seen in both mouse and human ERMS. To address whether satellite cells were the origin of the aP2-Cre;SmoM2/+ tumors, we utilized the Pax7-CreERT2 allele. Pax7 along with Pax3 play a critical role in the developing skeletal muscle, where both Pax7- and Pax3-positive cells in the dermomyotome give rise to the embryonic and fetal myoblasts that form the primary and secondary muscle bers. Consistent with these observations, we show that using Pax3Cre to activate the SmoM2 allele during development causes embryonic lethality. Also, using Cre-drivers (Myf5-Cre and MyoD1-Cre) that act downstream of Pax7 during myogenesis also caused embryonic lethality. Thus, in lieu of using the Pax7-CreERT2 allele to induce SmoM2 expression embryonically, which would globally disrupt muscle patterning and cause potent embryonic lethality, we focused on tamoxifen-induced activation of the Pax7-CreERT2;SmoM2/+ alleles in newly born pups, a time point at which Pax7 becomes restricted to the satellite cell lineage. Indeed, lethality was circumvented by this approach, which allowed for viable animals to be proled for tumorigenesis. This dedicated restricted activation of the oncogenic SmoM2 allele in the Pax7+ satellite cells did not result in tumor formation. If aP2-Cre was expressed at a level that was completely undetectable in either of the lineage tracing experiments with the R26-LacZ or R26-YFP mice in the satellite cells, and the satellite cells were the origin of these tumors, one would expect that oncogene activation in the satellite cell lineage would phenocopy the aP2-Cre;SmoM2/+ tumors, which was not observed. The cell of origin of ERMS has been and continues to be ambiguous (Hettmer and Wagers, 2010). It is generally thought that ERMS develops from cells of the muscle lineage, given that ERMS express skeletal muscle markers, such as MyoD, Myogenin, and Desmin. However, an origin restricted to muscle progenitors does not explain how ERMS occurs in locations that lack striated muscle, such as the prostate, urinary bladder, and the biliary tree (Heyn et al., 1997; Spunt et al., 2000). Transdifferentiation of mesenchymal progenitor cells, including progenitors of the adipose lineage into ERMS, could account for ERMS that arise in tissues that normally lack skeletal muscle, such as the biliary and genitourinary tract. Transdifferentiation of lineages not fated to become skeletal muscle could also explain the appearance of arrested skeletal muscle development, as lineages that transdifferentiate after their fate has been initially established could lack the developmental priming necessary to terminally differentiate into striated skeletal muscle. Recent reports suggest that skeletal muscle and brown fat share a developmental ancestry (Atit et al., 2006; Kajimura et al., 2009; Seale et al., 2008). In humans, brown fat has traditionally been thought to be restricted to newborns. However, recent reports have shown that brown adipose tissue, located primarily in the supraclavicular neck, is readily activated

upon cold exposure in adult humans (Cypess et al., 2009; van Marken Lichtenbelt et al., 2009; Virtanen et al., 2009). We speculate that constitutive activation of the Shh pathway by SmoM2 derails brown adipocyte progenitors into the muscle lineage, such that the resulting cells lack the previous developmental priming and necessary modications to develop normally into mature skeletal muscle and thus display an arrested state of skeletal muscle development or ERMS. The rhabdomyosarcoma tumors in aP2-Cre;SmoM2/+ mutant mice closely resemble human rhabdomyosarcoma with respect to location, histology, and gene expression. We postulate that the relatively short latency and high penetrance of ERMS in this model now provides an efcient genetic system and tool for the analysis of RMS modiers and provides a unique platform for preclinical studies.
EXPERIMENTAL PROCEDURES Mouse Strains All mice used in this study have been previously reported: aP2-Cre (Tang et al., 2008), SmoM2 (Jackson Laboratories, Bar Harbor, ME, USA; #5130) (Mao et al., 2006), Cdkn2aFlox (Aguirre et al., 2003), Pax3-Cre (Jackson Laboratories, #5549) (Engleka et al., 2005), Myf5-Cre (Jackson Laboratories, #7893) (Tallquist et al., 2000), MyoD1-Cre (Jackson Laboratories, #14140) (Yamamoto et al., 2009), Myogenin-Cre (Li et al., 2005), MCK-Cre (Jackson Laboratories, ning et al., 1998), Pax7-CreERT2 (Jackson Laboratories, #12476) #6475) (Bru (Lepper et al., 2009) R26-LacZ (Jackson Laboratories, #3474) (Soriano, 1999), and R26-YFP (Jackson Laboratories, #6148) (Srinivas et al., 2001). aP2-Cre;SmoM2/+ mice were generated by breeding aP2-Cre male animals to SmoM2/M2 females. aP2-Cre;SmoM2/+;Cdkn2aFl/Fl mice were generated by breeding aP2-Cre;Cdkn2aFl/Fl males with SmoM2/M2;Cdkn2aFl/Fl females. All mice used in these studies were of mixed genetic backgrounds, and all comparisons were performed on littermate controls. The end point for the Kaplan-Meier tumor-free survival analysis was the rst visible sign of tumor. All experimental procedures involving animals in this study were reviewed and approved by the Institutional Animal Care and Research Advisory Committee at the University of Texas Southwestern Medical Center. Histology and Immunohistochemistry Dissected tumors were xed in 10% neutral-buffered formalin, placed in xative overnight, embedded in parafn, and sectioned at 5 mm intervals. Hematoxylin and eosin (H&E) and Trichrome stains were performed using standard procedures. All parafn-embedded sections for immunohistochemistry were deparafnized, heated in a microwave in 0.01 M sodium citrate buffer for antigen retrieval, treated with 3% H2O2 for 10 min, and rinsed in H2O and PBS. For Ki-67 immunohistochemistry, sections were blocked in 5% goat serum in PBS, followed by incubation with anti-Ki67 antibody (Abcam, Cambridge, UK; ab15580) at a concentration of 1:100. Signals were detected with a Vectastain ABC kit (Vector Laboratories, Burlingame, CA, USA) and 3,30 -diaminobenzidine (DAB) substrate (Vector Laboratories). Sections were counterstained with hematoxylin and mounted. For MyoD, Myogenin, and Desmin immunohistochemistry, the M.O.M. Immunodetection Kit (Vector Laboratories), Vectastain Elite kit (Vector Laboratories), and 3,30 -diaminobenzidine (DAB) substrate were used. The MyoD1 primary antibody (Dako, Carpinteria, CA, USA; 5.8A, M3512) was used at a concentration of 1:50, the Myogenin primary antibody (Dako, 5FD, M3559) was used at a concentration of 1:25, and the Desmin primary antibody (Developmental Hybridoma Bank, D3 supernatants) was used at a concentration of 1:100. LacZ Staining b-galactosidase expression was detected in isolated tissue xed for 1 hr in 4% paraformaldehyde, 0.1 M phosphate buffer (pH 7.4), 0.01% sodium deoxycholate, and 0.02% Nonidet P-40 and then washed twice with buffer (0.1 M phosphate buffer [pH 7.4], 0.01% sodium deoxycholate, and 0.02% Nonidet P-40). b-galactosidase was detected by overnight incubation at

Cancer Cell 22, 536546, October 16, 2012 2012 Elsevier Inc. 543

Cancer Cell
Adipocyte Origin of Rhabdomyosarcoma

room temperature in 1 mg/ml X-gal (5-bromo-4-chloro-3-indoyl-b-D-galactoside), 5 mM potassium ferricyanide, and 5 mM potassium ferrocyanide, 2 mM MgCl2, 0.01% sodium deoxycholate, and 0.02% Nonidet P-40. Tissues were rinsed twice in buffer and postxed overnight in 4% paraformaldehyde and 0.2% glutaraldehyde. RNA Purication, RT-PCR, and Real-Time PCR Total RNA was isolated from normal SCM or tumors with Trizol reagent (Invitrogen, Carlsbad, CA, USA) in accordance with the manufacturers protocol. Reverse transcription was performed using random hexamer primers and SuperScript III First-Strand Synthesis (Invitrogen). Real-time PCR was performed using SYBR Green assay on an ABI-PE Prism 7900HT sequence detection system in accordance with manufacturers protocol. The relative quantities of the gene of interest were determined using the DDCT method and were normalized to 18S and expressed relative to the value for the normal SCM or to brown adipose tissue. In all real-time PCR assays, each group contained four independent samples assayed in duplicate. Microarray Analysis RNA from aP2-Cre;SmoM2/+ tumors or SCM with quality veried with nanodrop and bioanalyzer was hybridized to the Illumina MouseWG-6 v2.0 BeadChip. Normalized signal data from Illumina Whole-Genome Gene Expression BeadChips were collated and log base 2 transformed in STATA/SE 11.2 (College Station, TX, USA). Next, the data was imported into Partek Genomic Suite 6.6 (St. Louis, MO, USA) visualized by principal components analysis (PCA) to check for consistency of samples and class variance. The log2 transformed aP2-Cre;SmoM2/+ tumor and SCM data were batch corrected and then compared. Using Partek, each probeset was compared by an unequal variance t test. The Bonferroni correction was then applied. This same process was repeated for aP2-Cre;SmoM2/+ and Myogenin-Cre; SmoM2/+ data. Probesets that passed the Bonferroni cutoff at 0.0001 in the aP2-Cre; SmoM2/+ tumor, SCM comparison, and had an absolute value of the log2ratio >2.5 were selected for the heat map. Using Spotre Decision Site 9.1.2, the data were z transformed and hierarchically clustered using Euclidean distance; data was then recolored to the xed 2 to 2 scale. The volcano plot was produced in STATA/SE 11.2 by rst log10 transforming the p value from the unequal variance t test. The transformed score was then plotted against the log2ratio of expression from the two classes. The horizontal line represents the 0.05 Bonferroni cutoff, and the colors represent select signicantly differential genes that are either elevated (red) in aP2-Cre;SmoM2/+ tumor relative to SCM or downregulated (green). The volcano plot for the aP2-Cre;SmoM2/+/Myogenini-Cre;SmoM2/+ t test results was created in the same manner. Although in this case, no genes had statistically signicant differences. Gene expression was conrmed by real-time PCR. Cross-Species and Mouse Model Expression Comparisons The human data comes from two distinct studies: the 36 ERMS samples from the Curie Institute (Williamson et al., 2010) and Vastus Lateralis muscle of ten young trained subjects from the Mayo clinic (GSE9103). Both human studies used the Affymetrix Human Genome U133 Plus 2.0 Arrays. Data was rma summarized for each group independently (Partek Genomics Suite 6.6). Next, ERMS/wt-muscle log ratios were calculated and median normalized to center at zero to correct for source effects. The 15 mouse tumors and 3 normal SCM samples from this study were assayed by Illumina MouseWG-6 v2.0 BeadChip. Log2ratios were calculated, and the cross-species cross-platform data were matched by unigene ID based on an Affymetrix cross-species unigene le derived from NCBI homologene data. Whenever multiple probes were designated for a given unigene in either species, the probe with the maximum average expression was retained yielding 13,282 logratio pairs. The Spearman Correlation, AGDEX (Johnson et al., 2010), and linear t were calculated and graphically depicted using STATA/SE 11. The mouse model comparison was within species, but the cross two version comparison above was used. The San Antonio model was performed on the Illumina mouseRef-8 v1.1 expression beadchip. For the with intraspecies comparison, four normal skeletal muscle samples and seven ERMS samples from San Antonio were compared to our model data. Probe matching was direct by Illumina ID yielding 19,878 pairs. AGDEX, Spearman

correlation, and linear t were calculated and graphed as in the crossspecies example. Statistical Analysis Results are expressed as the mean SEM. We utilized a two-tailed, unpaired Students t test for all pairwise comparisons (GraphPad Prism version 5). p values less than 0.05 were considered signicant. ACCESSION NUMBERS Illumina data for the mouse ERMS and SCM can be found in the Gene Expression Omnibus database under number GSE40359. SUPPLEMENTAL INFORMATION Supplemental Information includes four gures and Supplemental Experimental Procedures and can be found with this article online at http://dx.doi. org/10.1016/j.ccr.2012.09.004. ACKNOWLEDGMENTS We are grateful to Jose Cabrera for gure preparation and John Shelton for experimental assistance. We thank Nabeel Bardeesy for the Cdkn2aFlox mice. Work in E.N.O.s laboratory was supported by grants from the National Institutes of Health, the Leducq Foundation, the Robert A. Welch Foundation, and the Cancer Prevention and Research Institute of Texas. M.E.H. is supported by Alexs Lemonade Stand Foundation, Pediatric Scientist Development Program sponsored by the Eunice Kennedy Shriver National Institute of Child Health and Human Development (NICHD Grant Award K12-HD000850), and the American Lebanese Syrian Associated Charities (ALSAC). Funding to R.L.G. for this work was provided by the Burroughs Wellcome Fund Career Award for Medical Scientists, the American Cancer Society/Simmons Cancer Center Institutional Research Grant (ACS-IRG-02196), The CureSearch for Childrens Cancer Young Investigator Program, and an Alexs Lemonade Stand Foundation A Award. Work in J. G.s lab was supported by NIDDK Grants R01-DK066566, R01-DK064261, and R01DK088220. Received: April 20, 2012 Revised: July 17, 2012 Accepted: September 4, 2012 Published: October 15, 2012 REFERENCES Aguirre, A.J., Bardeesy, N., Sinha, M., Lopez, L., Tuveson, D.A., Horner, J., Redston, M.S., and DePinho, R.A. (2003). Activated Kras and Ink4a/Arf deciency cooperate to produce metastatic pancreatic ductal adenocarcinoma. Genes Dev. 17, 31123126. Atit, R., Sgaier, S.K., Mohamed, O.A., Taketo, M.M., Dufort, D., Joyner, A.L., Niswander, L., and Conlon, R.A. (2006). Beta-catenin activation is necessary and sufcient to specify the dorsal dermal fate in the mouse. Dev. Biol. 296, 164176. Barlow, C., Schroeder, M., Lekstrom-Himes, J., Kylefjord, H., Deng, C.X., Wynshaw-Boris, A., Spiegelman, B.M., and Xanthopoulos, K.G. (1997). Targeted expression of Cre recombinase to adipose tissue of transgenic mice directs adipose-specic excision of loxP-anked gene segments. Nucleic Acids Res. 25, 25432545. Barr, F.G. (2001). Gene fusions involving PAX and FOX family members in alveolar rhabdomyosarcoma. Oncogene 20, 57365746. Barr, F.G., and Womer, R.B. (2009). Rhabdomyosarcoma. In Oncology of Infancy and Childhood, S.H. Orkin, D.E. Fisher, A.T. Look, S.E. Lux, D. Ginsburg, and D.G. Nathan, eds. (Philadelphia: Saunders), pp. 743781. 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

544 Cancer Cell 22, 536546, October 16, 2012 2012 Elsevier Inc.

Cancer Cell
Adipocyte Origin of Rhabdomyosarcoma

landscape of somatic copy-number alteration across human cancers. Nature 463, 899905. Bignell, G.R., Greenman, C.D., Davies, H., Butler, A.P., Edkins, S., Andrews, J.M., Buck, G., Chen, L., Beare, D., Latimer, C., et al. (2010). Signatures of mutation and selection in the cancer genome. Nature 463, 893898. rsch, D., Accili, D., ning, J.C., Michael, M.D., Winnay, J.N., Hayashi, T., Ho Bru Goodyear, L.J., and Kahn, C.R. (1998). A muscle-specic insulin receptor knockout exhibits features of the metabolic syndrome of NIDDM without altering glucose tolerance. Mol. Cell 2, 559569. Chen, Y., Takita, J., Mizuguchi, M., Tanaka, K., Ida, K., Koh, K., Igarashi, T., Hanada, R., Tanaka, Y., Park, M.J., and Hayashi, Y. (2007). Mutation and expression analyses of the MET and CDKN2A genes in rhabdomyosarcoma with emphasis on MET overexpression. Genes Chromosomes Cancer 46, 348358. Corcoran, R.B., and Scott, M.P. (2001). A mouse model for medulloblastoma and basal cell nevus syndrome. J. Neurooncol. 53, 307318. Cypess, A.M., Lehman, S., Williams, G., Tal, I., Rodman, D., Goldne, A.B., Kuo, F.C., Palmer, E.L., Tseng, Y.H., Doria, A., et al. (2009). Identication and importance of brown adipose tissue in adult humans. N. Engl. J. Med. 360, 15091517. El-Badry, O.M., Minniti, C., Kohn, E.C., Houghton, P.J., Daughaday, W.H., and Helman, L.J. (1990). Insulin-like growth factor II acts as an autocrine growth and motility factor in human rhabdomyosarcoma tumors. Cell Growth Differ. 1, 325331. Engleka, K.A., Gitler, A.D., Zhang, M., Zhou, D.D., High, F.A., and Epstein, J.A. (2005). Insertion of Cre into the Pax3 locus creates a new allele of Splotch and identies unexpected Pax3 derivatives. Dev. Biol. 280, 396406. Estep, A.L., Tidyman, W.E., Teitell, M.A., Cotter, P.D., and Rauen, K.A. (2006). HRAS mutations in Costello syndrome: detection of constitutional activating mutations in codon 12 and 13 and loss of wild-type allele in malignancy. Am. J. Med. Genet. A. 140, 816. Fan, C.M., and Tessier-Lavigne, M. (1994). Patterning of mammalian somites by surface ectoderm and notochord: evidence for sclerotome induction by a hedgehog homolog. Cell 79, 11751186. Fukada, S., Uezumi, A., Ikemoto, M., Masuda, S., Segawa, M., Tanimura, N., Yamamoto, H., Miyagoe-Suzuki, Y., and Takeda, S. (2007). Molecular signature of quiescent satellite cells in adult skeletal muscle. Stem Cells 25, 24482459. Galindo, R.L., Allport, J.A., and Olson, E.N. (2006). A Drosophila model of the rhabdomyosarcoma initiator PAX7-FKHR. Proc. Natl. Acad. Sci. USA 103, 1343913444. ller-Hermelink, Gattenloehner, S., Vincent, A., Leuschner, I., Tzartos, S., Mu H.K., Kirchner, T., and Marx, A. (1998). The fetal form of the acetylcholine receptor distinguishes rhabdomyosarcomas from other childhood tumors. Am. J. Pathol. 152, 437444. er, S., Schelp, C., von Schweinitz, D., and Gerardy-Schahn, R. (1998). Glu Polysialylated neural cell adhesion molecule in childhood rhabdomyosarcoma. Pediatr. Res. 43, 145147. Hahn, H., Wojnowski, L., Specht, K., Kappler, R., Calzada-Wack, J., Potter, D., ller, U., Samson, E., Quintanilla-Martinez, L., and Zimmer, A. Zimmer, A., Mu (2000). Patched target Igf2 is indispensable for the formation of medulloblastoma and rhabdomyosarcoma. J. Biol. Chem. 275, 2834128344. He, W., Barak, Y., Hevener, A., Olson, P., Liao, D., Le, J., Nelson, M., Ong, E., Olefsky, J.M., and Evans, R.M. (2003). Adipose-specic peroxisome proliferator-activated receptor gamma knockout causes insulin resistance in fat and liver but not in muscle. Proc. Natl. Acad. Sci. USA 100, 1571215717. Hettmer, S., and Wagers, A.J. (2010). Muscling in: uncovering the origins of rhabdomyosarcoma. Nat. Med. 16, 171173. Hettmer, S., Liu, J., Miller, C.M., Lindsay, M.C., Sparks, C.A., Guertin, D.A., Bronson, R.T., Langenau, D.M., and Wagers, A.J. (2011). Sarcomas induced in discrete subsets of prospectively isolated skeletal muscle cells. Proc. Natl. Acad. Sci. USA 108, 2000220007.

Heyn, R., Newton, W.A., Raney, R.B., Hamoudi, A., Bagwell, C., Vietti, T., Wharam, M., Gehan, E., and Maurer, H.M. (1997). Preservation of the bladder in patients with rhabdomyosarcoma. J. Clin. Oncol. 15, 6975. Iolascon, A., Faienza, M.F., Coppola, B., Rosolen, A., Basso, G., Della Ragione, F., and Schettini, F. (1996). Analysis of cyclin-dependent kinase inhibitor genes (CDKN2A, CDKN2B, and CDKN2C) in childhood rhabdomyosarcoma. Genes Chromosomes Cancer 15, 217222. Ishiguro, N., Motoi, T., Araki, N., Ito, H., Moriyama, M., and Yoshida, H. (2008). Expression of cardiac ankyrin repeat protein, CARP, in malignant tumors: diagnostic use of CARP protein immunostaining in rhabdomyosarcoma. Hum. Pathol. 39, 16731679. Johnson, R.A., Wright, K.D., Poppleton, H., Mohankumar, K.M., Finkelstein, D., Pounds, S.B., Rand, V., Leary, S.E., White, E., Eden, C., et al. (2010). Cross-species genomics matches driver mutations and cell compartments to model ependymoma. Nature 466, 632636. Johnson, R.L., Laufer, E., Riddle, R.D., and Tabin, C. (1994). Ectopic expression of Sonic hedgehog alters dorsal-ventral patterning of somites. Cell 79, 11651173. Johnson, R.L., Rothman, A.L., Xie, J., Goodrich, L.V., Bare, J.W., Bonifas, J.M., Quinn, A.G., Myers, R.M., Cox, D.R., Epstein, E.H., Jr., and Scott, M.P. (1996). Human homolog of patched, a candidate gene for the basal cell nevus syndrome. Science 272, 16681671. Kajimura, S., Seale, P., Kubota, K., Lunsford, E., Frangioni, J.V., Gygi, S.P., and Spiegelman, B.M. (2009). Initiation of myoblast to brown fat switch by a PRDM16-C/EBP-beta transcriptional complex. Nature 460, 11541158. Keller, C., and Capecchi, M.R. (2005). New genetic tactics to model alveolar rhabdomyosarcoma in the mouse. Cancer Res. 65, 75307532. Lee, Y., Kawagoe, R., Sasai, K., Li, Y., Russell, H.R., Curran, T., and McKinnon, P.J. (2007). Loss of suppressor-of-fused function promotes tumorigenesis. Oncogene 26, 64426447. Lepper, C., Conway, S.J., and Fan, C.M. (2009). Adult satellite cells and embryonic muscle progenitors have distinct genetic requirements. Nature 460, 627631. Li, F.P., and Fraumeni, J.F., Jr. (1969). Rhabdomyosarcoma in children: epidemiologic study and identication of a familial cancer syndrome. J. Natl. Cancer Inst. 43, 13651373. Li, S., Czubryt, M.P., McAnally, J., Bassel-Duby, R., Richardson, J.A., Wiebel, F.F., Nordheim, A., and Olson, E.N. (2005). Requirement for serum response factor for skeletal muscle growth and maturation revealed by tissue-specic gene deletion in mice. Proc. Natl. Acad. Sci. USA 102, 10821087. Lum, L., and Beachy, P.A. (2004). The hedgehog response network: sensors, switches, and routers. Science 304, 17551759. Mao, J., Ligon, K.L., Rakhlin, E.Y., Thayer, S.P., Bronson, R.T., Rowitch, D., and McMahon, A.P. (2006). A novel somatic mouse model to survey tumorigenic potential applied to the hedgehog pathway. Cancer Res. 66, 10171 10178. nsterberg, A.E., Kitajewski, J., Bumcrot, D.A., McMahon, A.P., and Lassar, Mu A.B. (1995). Combinatorial signaling by Sonic hedgehog and Wnt family members induces myogenic bHLH gene expression in the somite. Genes Dev. 9, 29112922. Paulson, V., Chandler, G., Rakheja, D., Galindo, R.L., Wilson, K., Amatruda, J.F., and Cameron, S. (2011). High-resolution array CGH identies common mechanisms that drive embryonal rhabdomyosarcoma pathogenesis. Genes Chromosomes Cancer 50, 397408. Pressey, J.G., Anderson, J.R., Crossman, D.K., Lynch, J.C., and Barr, F.G. (2011). Hedgehog pathway activity in pediatric embryonal rhabdomyosarcoma and undifferentiated sarcoma: a report from the Childrens Oncology Group. Pediatr. Blood Cancer 57, 930938. Rezvani, G., Lui, J.C., Barnes, K.M., and Baron, J. (2012). A set of imprinted genes required for normal body growth also promotes growth of rhabdomyosarcoma cells. Pediatr. Res. 71, 3238. Ross, S.R., Graves, R.A., Greenstein, A., Platt, K.A., Shyu, H.L., Mellovitz, B., and Spiegelman, B.M. (1990). A fat-specic enhancer is the primary

Cancer Cell 22, 536546, October 16, 2012 2012 Elsevier Inc. 545

Cancer Cell
Adipocyte Origin of Rhabdomyosarcoma

determinant of gene expression for adipocyte P2 in vivo. Proc. Natl. Acad. Sci. USA 87, 95909594. Ross, S.R., Choy, L., Graves, R.A., Fox, N., Solevjeva, V., Klaus, S., Ricquier, D., and Spiegelman, B.M. (1992). Hibernoma formation in transgenic mice and isolation of a brown adipocyte cell line expressing the uncoupling protein gene. Proc. Natl. Acad. Sci. USA 89, 75617565. Rubin, B.P., Nishijo, K., Chen, H.I., Yi, X., Schuetze, D.P., Pal, R., Prajapati, S.I., Abraham, J., Arenkiel, B.R., Chen, Q.R., et al. (2011). Evidence for an unanticipated relationship between undifferentiated pleomorphic sarcoma and embryonal rhabdomyosarcoma. Cancer Cell 19, 177191. Seale, P., and Rudnicki, M.A. (2000). A new look at the origin, function, and stem-cell status of muscle satellite cells. Dev. Biol. 218, 115124. ` , A., Seale, P., Bjork, B., Yang, W., Kajimura, S., Chin, S., Kuang, S., Scime Devarakonda, S., Conroe, H.M., Erdjument-Bromage, H., et al. (2008). PRDM16 controls a brown fat/skeletal muscle switch. Nature 454, 961967. Shadrach, J.L., and Wagers, A.J. (2011). Stem cells for skeletal muscle repair. Philos. Trans. R. Soc. Lond. B Biol. Sci. 366, 22972306. Sharp, R., Recio, J.A., Jhappan, C., Otsuka, T., Liu, S., Yu, Y., Liu, W., Anver, M., Navid, F., Helman, L.J., et al. (2002). Synergism between INK4a/ARF inactivation and aberrant HGF/SF signaling in rhabdomyosarcomagenesis. Nat. Med. 8, 12761280. Soriano, P. (1999). Generalized lacZ expression with the ROSA26 Cre reporter strain. Nat. Genet. 21, 7071. Spunt, S.L., Lobe, T.E., Pappo, A.S., Parham, D.M., Wharam, M.D., Jr., Arndt, C., Anderson, J.R., Crist, W.M., Paidas, C., Wiener, E., et al. (2000). Aggressive surgery is unwarranted for biliary tract rhabdomyosarcoma. J. Pediatr. Surg. 35, 309316. Srinivas, S., Watanabe, T., Lin, C.S., William, C.M., Tanabe, Y., Jessell, T.M., and Costantini, F. (2001). Cre reporter strains produced by targeted insertion of EYFP and ECFP into the ROSA26 locus. BMC Dev. Biol. 1, 4. Suh, J.M., Gao, X., McKay, J., McKay, R., Salo, Z., and Graff, J.M. (2006). Hedgehog signaling plays a conserved role in inhibiting fat formation. Cell Metab. 3, 2534. m, M., and Soriano, P. (2000). Tallquist, M.D., Weismann, K.E., Hellstro Early myotome specication regulates PDGFA expression and axial skeleton development. Development 127, 50595070.

Tang, W., Zeve, D., Suh, J.M., Bosnakovski, D., Kyba, M., Hammer, R.E., Tallquist, M.D., and Graff, J.M. (2008). White fat progenitor cells reside in the adipose vasculature. Science 322, 583586. Tifn, N., Williams, R.D., Shipley, J., and Pritchard-Jones, K. (2003). PAX7 expression in embryonal rhabdomyosarcoma suggests an origin in muscle satellite cells. Br. J. Cancer 89, 327332. rd, R., and Tostar, U., Malm, C.J., Meis-Kindblom, J.M., Kindblom, L.G., Toftga n, A.B. (2006). Deregulation of the hedgehog signalling pathway: Unde a possible role for the PTCH and SUFU genes in human rhabdomyoma and rhabdomyosarcoma development. J. Pathol. 208, 1725. Urs, S., Harrington, A., Liaw, L., and Small, D. (2006). Selective expression of an aP2/Fatty Acid Binding Protein 4-Cre transgene in non-adipogenic tissues during embryonic development. Transgenic Res. 15, 647653. van Marken Lichtenbelt, W.D., Vanhommerig, J.W., Smulders, N.M., Drossaerts, J.M., Kemerink, G.J., Bouvy, N.D., Schrauwen, P., and Teule, G.J. (2009). Cold-activated brown adipose tissue in healthy men. N. Engl. J. Med. 360, 15001508. Virtanen, K.A., Lidell, M.E., Orava, J., Heglind, M., Westergren, R., Niemi, T., ck, S., and Nuutila, P. (2009). Taittonen, M., Laine, J., Savisto, N.J., Enerba Functional brown adipose tissue in healthy adults. N. Engl. J. Med. 360, 15181525. Wang, Q., Fang, W.H., Krupinski, J., Kumar, S., Slevin, M., and Kumar, P. (2008). Pax genes in embryogenesis and oncogenesis. J. Cell. Mol. Med. 12 (6A), 22812294. ` s, A., Pierron, G., Thuille, B., Williamson, D., Missiaglia, E., de Reynie , M., Fre neaux, P., et al. (2010). Palenzuela, G., Thway, K., Orbach, D., Lae Fusion gene-negative alveolar rhabdomyosarcoma is clinically and molecularly indistinguishable from embryonal rhabdomyosarcoma. J. Clin. Oncol. 28, 21512158. Yamamoto, M., Shook, N.A., Kanisicak, O., Yamamoto, S., Wosczyna, M.N., Camp, J.R., and Goldhamer, D.J. (2009). A multifunctional reporter mouse line for Cre- and FLP-dependent lineage analysis. Genesis 47, 107114. Zibat, A., Missiaglia, E., Rosenberger, A., Pritchard-Jones, K., Shipley, J., Hahn, H., and Fulda, S. (2010). Activation of the hedgehog pathway confers a poor prognosis in embryonal and fusion gene-negative alveolar rhabdomyosarcoma. Oncogene 29, 63236330.

546 Cancer Cell 22, 536546, October 16, 2012 2012 Elsevier Inc.

Cancer Cell

Article
Metabolic Signatures Uncover Distinct Targets in Molecular Subsets of Diffuse Large B Cell Lymphoma
Pilar Caro,1,10 Amar U. Kishan,1,4,10 Erik Norberg,1,5,10 Illana A. Stanley,1,5 Bjoern Chapuy,2 Scott B. Ficarro,1,3,6 Klaudia Polak,1 Daniel Tondera,1 John Gounarides,7 Hong Yin,7 Feng Zhou,1,3,6 Michael R. Green,2 Linfeng Chen,2 Stefano Monti,8,9 Jarrod A. Marto,1,3,6 Margaret A. Shipp,2 and Nika N. Danial1,5,*
of Cancer Biology of Medical Oncology 3Blais Proteomics Center Dana-Farber Cancer Institute, Boston, MA 02115, USA 4Harvard-MIT Division of Health Sciences and Technology, Cambridge, MA 02139, USA 5Department of Cell Biology 6Department of Biological Chemistry and Molecular Pharmacology Harvard Medical School, Boston, MA 02115, USA 7Novartis Institutes for Biomedical Research, Cambridge, MA 02139, USA 8The Broad Institute, Cambridge, MA 02142, USA 9Section of Computational Biomedicine, Boston University School of Medicine, Boston, MA 02218, USA 10These authors contributed equally to this work *Correspondence: nika_danial@dfci.harvard.edu http://dx.doi.org/10.1016/j.ccr.2012.08.014
2Department 1Department

SUMMARY

Molecular signatures have identied several subsets of diffuse large B cell lymphoma (DLBCL) and rational targets within the B cell receptor (BCR) signaling axis. The OxPhos-DLBCL subset, which harbors the signature of genes involved in mitochondrial metabolism, is insensitive to inhibition of BCR survival signaling but is functionally undened. We show that, compared with BCR-DLBCLs, OxPhos-DLBCLs display enhanced mitochondrial energy transduction, greater incorporation of nutrient-derived carbons into the tricarboxylic acid cycle, and increased glutathione levels. Moreover, perturbation of the fatty acid oxidation program and glutathione synthesis proved selectively toxic to this tumor subset. Our analysis provides evidence for distinct metabolic ngerprints and associated survival mechanisms in DLBCL and may have therapeutic implications.
INTRODUCTION Tumors often rewire their metabolism to ensure a steady supply of intermediary metabolites for synthesis of new biomass, as well as generation of ATP and reducing equivalents (Barger and Plas, 2010; DeBerardinis et al., 2008; Tennant et al., 2010). This metabolic reprogramming is a dynamic process shaped by oncogenes and tumor suppressors (Barger and Plas, 2010; Morrish et al., 2008; Yuneva et al., 2012). One of the rst metabolic alterations identied in tumors is elevated glycolysis, even in the presence of sufcient oxygen. This program, also known as the Warburg effect or aerobic glycolysis, fullls important biosynthetic needs (Barger and Plas, 2010; Koppenol et al., 2011; Vander Heiden et al., 2009). The Warburg effect has often been interpreted as an indication of impaired mitochondrial respiration (Koppenol et al., 2011). However, the relevance of

Signicance B cell receptor (BCR) survival signals are central to pathogenesis of certain DLBCLs. Among transcriptionally dened groups of DLBCL, the OxPhos subset displays increased expression of genes involved in mitochondrial oxidative phosphorylation but lacks an intact BCR signaling network, suggesting dependence on alternative survival mechanisms. However, the functional basis of the OxPhos signature has not been dened. Using an integrative approach, we have carefully dissected the patterns of fuel utilization and energy metabolism in DLBCL, with a primary focus on the OxPhos subset. We show that fatty acid oxidation and glutathione synthesis are distinct metabolic features of OxPhos-DLBCLs selectively required for their survival. Metabolic signatures may provide important insights into the molecular heterogeneity of DLBCL and reveal rational targets in these lymphomas.
Cancer Cell 22, 547560, October 16, 2012 2012 Elsevier Inc. 547

Cancer Cell
Metabolic Fingerprints in DLBCL Molecular Subsets

mitochondrial respiration in tumors is varied depending on tumor type, and evidence for an oxidative class of tumors and tumors with dual capacity for glycolytic and oxidative metabolism exists nchez et al., 2009). (Marin-Valencia et al., 2012; Moreno-Sa Moreover, the importance of mitochondria in tumor cell survival and proliferation, including utilization of alternative oxidizable substrates such as glutamine and fatty acids, has been increasingly appreciated (Le et al., 2012; Rossignol et al., 2004; Zaugg et al., 2011). The diversity of carbon substrate utilization pathways in tumors is indicative of metabolic heterogeneity that may not only be relevant across different types of cancer but also manifest within a group of tumors that otherwise share a common diagnosis. Diffuse large B cell lymphomas (DLBCLs) are a genetically heterogeneous group of tumors and the most common nonHodgkins lymphomas in adults (Abramson and Shipp, 2005; Lenz and Staudt, 2010). However, the spectrum of fuel utilization pathways and the metabolic ngerprints within DLBCL and other similarly heterogeneous groups of tumors have not been fully elucidated. To date, efforts to capture the molecular heterogeneity of DLBCL have relied on gene expression proling that has uncovered coordinate signaling and survival paradigms in distinct subsets of DLBCL. In one approach, comparison of the genetic signatures across DLBCLs using genome-wide arrays and multiple clustering algorithms captured tumor-intrinsic distinctions in three separate and reproducible clusters (Monti et al., 2005). Groups of DLBCLs identied by this consensus cluster classication (CCC) scheme are the B cell receptor (BCR)/proliferation cluster (BCR-DLBCL) displaying upregulation of genes encoding BCR signaling components, the OxPhos cluster (OxPhos-DLBCL), which is signicantly enriched in genes involved in mitochondrial oxidative phosphorylation (OxPhos), and the host response (HR) tumors largely characterized by a brisk host inammatory inltrate (Monti et al., 2005). Another classication framework known as cell of origin (COO) delineated DLBCL subsets that shared components of their transcriptional proles with normal B cell subtypes, including germinal center B cell (GCB)-like and activated B cell (ABC)like (Alizadeh et al., 2000), and a third undened category, designated type 3 (Wright et al., 2003). CCC and COO classications capture largely different molecular aspects of DLBCL (Monti et al., 2005). Unlike tumors that rely on signaling pathways downstream of the BCR, OxPhos-DLBCLs do not display active/functional BCR signaling (Chen et al., 2008). However, the nature of survival pathways in this group of tumors is not known, and beyond the original CCC assignment, the actual functional attributes of the OxPhos molecular signature have not been fully examined. This signature includes multiple subunits of mitochondrial respiratory chain complexes I (NADH dehydrogenase) and V (mitochondrial ATP synthase) that may suggest alterations in mitochondrial energy transduction. However, given the integrative aspect of cellular metabolism and the requirement of both nuclear and mitochondria-encoded genes for proper functioning of the electron transport machinery, the precise metabolic landscape of this molecular subset could not be predicted. In the present study, we conducted an integrative analysis to dissect the metabolic ngerprints of DLBCL and to delineate
548 Cancer Cell 22, 547560, October 16, 2012 2012 Elsevier Inc.

subtype-specic differences that may selectively contribute to the growth and survival of DLBCL subsets. RESULTS Subtype-Specic Differences in the DLBCL Mitochondrial Proteome The upregulation of select genes encoding for subunits of electron transport chain (ETC) complexes in OxPhos-DLBCLs predicts potential differences in mitochondrial oxidative metabolism compared with other DLBCL groups. However, as ETC activity is linked to the supply of carbon substrates and reducing equivalents, the OxPhos signature is likely part of a broader spectrum of changes in mitochondrial nutrient metabolism that may shed light on the actual functional attributes of an OxPhos program in this DLBCL subset. To search for additional components of this metabolic program, we initially performed two-dimensional differential gel electrophoresis (2D-DIGE) to compare the proteome of mitochondria puried from representative OxPhos- and BCR-DLBCL cell lines Karpas 422 and OCI-Ly1, respectively (Chen et al., 2008). Mitochondrial proteins that were R2.5 times more abundant in the OxPhos cell line were identied by mass spectrometry (Figure S1A available online). Among 2D-DIGE candidates were subunits of mitochondrial respiratory chain complex I (NADH dehydrogenase), complex II (succinate dehydrogenase, also a tricarboxylic acid [TCA] cycle enzyme), complex V (ATP synthase), subunits of the pyruvate dehydrogenase (PDH) complex and several other TCA cycle enzymes, mitochondrial reactive oxygen species (ROS) detoxication enzyme manganese superoxide dismutase (MnSOD or SOD2), and enzymes involved in mitochondrial b-oxidation of fatty acids and metabolism of ketone bodies and amino acids (Figure S1A). We next set out to interrogate the mitochondrial protein signature dened by 2D-DIGE in a larger panel of OxPhosand BCR-DLBCL cell lines. This required an independent proteomics approach amenable to multiplexingisobaric tags for relative and absolute quantication (iTRAQ) (Figure S1B) (Choe et al., 2007; Ross et al., 2004)for simultaneous quantitative comparison of multiple unique tryptic peptides per candidate mitochondrial protein across three pairs of OxPhos- and BCRDLBCL cell lines. This analysis conrmed signicant quantitative enrichment of mitochondrial b-oxidation enzymes and multiple subunits of respiratory chain complexes, as well as TCA cycle enzymes and MnSOD in OxPhos-DLBCL mitochondria (Figure 1; Table S1). Beyond conrming the differential expression of complex I and V components that were in the original OxPhos-DLBCL cluster signature (Monti et al., 2005), the 2D-DIGE and iTRAQ proteomic analyses identied additional components of the OxPhos-DLBCL mitochondrial signature that were not fully captured by the previous RNA prole analysis of total tumors. To examine whether these observations could also be substantiated in primary OxPhos-DLBCLs, we examined tumor biopsies for differences in the abundance of these newly identied components of the mitochondrial signature at both protein and RNA levels. The small en bloc primary DLBCL biopsies precluded isolation of tumor mitochondria; however, iTRAQ-based quantication in a limited number of primary cases that

Cancer Cell
Metabolic Fingerprints in DLBCL Molecular Subsets

0.8

0.6

0.4

0.2

0.0

-0.2

-0.4

1 ACADVL 2 ACAT1 3 AK2 4 ALDH18A1 5 ALDH1B1 6 ALDH2 7 ALDH5A1 8 ATP5A1 9 ATP5B 10 ATP5C1 11 ATP5H 12 ATP5I 13 ATP5J 14 ATP5L 15 ATP5O 16 CPT2 17 ECHS1 18 ETFA 19 ETFB 20 HADHA 21 HADHB

22 IDH2 23 IDH3A 24 NDUFA5 25 NDUFA8 26 NDUFA9 27 NDUFAF2 28 NDUFS1 29 NDUFS2 30 NDUFS3 31 NDUFV1 32 NDUFV2 33 NDUFV3 34 OAT 35 OGDH 36 OXCT1 37 PDHB 38 SDHA 39 SDHB 40 SHMT2 41 SOD2 42 VDAC3

TCA Cycle
Pyruvate dehydrogenase (PDH) B Isocitrate dehydrogenase (IDH) 2 (NADP+) Isocitrate dehydrogenase (IDH) 3A (NAD+) -Ketoglutarate dehydrogenase (OGDH) Succinate dehydrogenase (SDH) subunits A and B

log Ratio OxPhos/BCR

Electron Transport and OxPhos Apparatus


Complex I, subcomplex subunits (NDUFA) 5, 8, 9 Complex I, subcomplex assembly factor (NDUFAF) 2 Complex I, Fe-S protein (NDUFS) 1, 2, 3 Complex I, flavoprotein (NDUFV) 1, 2, 3 ATP synthase (ATP5) F1 complex subunits ATP synthase (ATP5) F0 complex subunits

-Oxidation
Acyl-CoA dehydrogenase (ACAD) VL Carnitine palmitoyltransferease (CPT) 2 Enoyl-CoA hydratase, short chain (ECHS) 1 Electron-transfer-flavoprotein (ETF) A and B Hydroxyacyl-CoA dehydrogenase (HADH) A and B

-0.6

Ketone Body Metabolism


-0.8 2.5 3.0 3.5 4.0 4.5 5.0 5.5 6.0 6.5

3-oxoacid CoA transferase (OXCT1) 1 Mitochondrial acetoacetyl-CoA thiolase (ACAT1)

Amino Acid and One Carbon Metabolism

log Reporter Ion Signal

Mitochondrial aldehyde dehydrogenase (ALDH) 2 Aldehyde dehydrogenase (ALDH) 1B1 Succinate semialdehyde dehydrogenase (ALDH5A1) Pyrroline-5-carboxylate synthetase (ALDH18A1) Mitochondrial serine hydroxymethyltransferase (SHMT2) Ornithine aminotransferase (OAT)

ROS Detoxification
Manganese Superoxide dismutase (SOD2)

Figure 1. Comparison of Mitochondrial Proteome in OxPhos- and BCR-DLBCLs


(A and B) Multiplex iTRAQ analysis of mitochondria derived from three OxPhos- (Karpas 422, Toledo, and Pfeiffer) and three BCR- (OCI-Ly1, SU-DHL-4, and SUDHL-6) DLBCL cell lines. Log-log plots of reporter ion abundance ratios versus reporter ion intensities for proteins detected in replicate nanoow LC-MS/MS analyses are shown in (A). Proteins within the mitochondrial signature that are enriched in OxPhos-DLBCLs are shown as red circles in (A) and grouped per metabolic pathway in (B). Red lines in (A) represent global thresholds based on a 1.5-fold change between OxPhos and BCR subtypes. See also Figure S1 and Table S1.

produced protein samples of suitable purity and grade for liquid chromatography/tandem mass spectrometry (LC-MS/MS) indicated increased expression of several TCA cycle and mitochondrial b-oxidation enzymes, as well as MnSOD (Table S1). The RNA abundance for candidate components of the mitochondrial proteome signature was also queried in two extensive primary DLBCL expression prole data sets (Lenz et al., 2008; Monti et al., 2005) with CCC designations (Tables S2 and S3) and found to be signicantly higher in primary OxPhos- than in BCR-DLBCLs in both data sets (Figure 2). These observations conrm that the differences in mitochondrial signature identied in DLBCL cell lines translate to primary tumor biopsies. The functional categories of proteins in the OxPhos-DLBCL mitochondrial signature predict differences in mitochondrial handling of fatty acids and pyruvate, ETC activity, mitochondrial energy production, and ROS content. We next set out to validate each of these predictions. Mitochondrial Substrate Oxidation in DLBCL Subsets To assess potential differences in mitochondrial carbon substrate utilization, mitochondrial oxygen consumption rate (OCR) was measured in real time in the absence or presence of exogenously added substrates such as glucose, glutamine, and palmitate, using a panel of 34 independent OxPhos and non-OxPhos DLBCL cell lines. The non-OxPhos panel included cell lines with known CCC assignment as BCR-DLBCLs (Polo et al., 2007) and COO classication as GCB-, ABC-, or

Type 3-DLBCLs (Alizadeh et al., 2000; Wright et al., 2003) (see Supplemental Experimental Procedures). Basal OCR in response to exogenously supplied palmitate was signicantly higher in OxPhos-DLBCLs relative to non-OxPhos DLBCL cell lines (no substrate [NS] versus palmitate; Figure 3A). Basal OCR values per individual DLBCL cell line are shown in Figure S2A. The non-OxPhos DLBCLs displayed either no increase in OCR or an increase that was signicantly less than that observed in the OxPhos-DLBCLs (Figures 3A and S2A). Notably, the majority of palmitate-stimulated OCR in OxPhos-DLBCLs was sensitive to the mitochondrial ATP synthase inhibitor oligomycin, indicating that mitochondrial fatty acid oxidation (FAO) in this setting is associated with ATP synthesis (Figures 3A and S2A). These observations provide quantitative evidence for increased mitochondrial FAO in OxPhos-DLBCLs. Analysis of other oxidizable substrates showed that none of the DLBCL subsets respire on glucose (NS vs. glucose; Figures 3A and S2A). However, in response to glutamine, all DLBCL subsets exhibited a comparable increase in basal and ATPassociated (oligomycin-inhibitable) OCR (NS versus glutamine; Figures 3A and S2A). Taken together, comparison of the three carbon substrates across and within DLBCL subsets indicates that palmitate is a predominant respiratory fuel in OxPhosDLBCLs. A marked increase in palmitate-induced OCR in OxPhosDLBCL compared with non-OxPhos DLBCL cell lines parallels the absence or presence of functional BCR signaling in these
Cancer Cell 22, 547560, October 16, 2012 2012 Elsevier Inc. 549

Cancer Cell
Metabolic Fingerprints in DLBCL Molecular Subsets

A
211023_at 201304_at 201740_at 205412_at 201931_at 201761_at 216841_s_at

OxPhos (n=56)

BCR (n=64)

OxPhos (n=71)

BCR (n=83)

PDHB NDUFA5 NDUFS3 ACAT1 ETFA MTHFD2 SOD2


NDUFA5 (201304_at)
400 300 200 100 0 OxPhos BCR
p<0.0001 FDR=0.0002 p<0.0001 FDR=0.0002

211023_at 201304_at 201740_at 205412_at 201931_at 201761_at 216841_s_at


ETFA (201931_at)
600
p<0.0001 FDR=0.0002

PDHB NDUFA5 NDUFS3 ACAT1 ETFA MTHFD2 SOD2


MTHFD2 (201761_at)
1500
p=0.0004 FDR=0.0006

B
800 600 400

PDHB (211023_at)
p<0.0001 FDR=0.0002

NDUFS3 (201740_at)
800 600 400 400
p<0.0001 FDR=0.0002

ACAT1 (205412_at)
1000 800 600 400
p<0.0001 FDR=0.0002

SOD2 (216841_s_at)
300
p<0.0001 FDR=0.0002

1000

200

Transcript Abundance

200

200 0 OxPhos 2000


p=0.0022 FDR=0.0028

500

100

200 0 OxPhos BCR


p<0.0001 FDR=0.0002

200 0 OxPhos
p=0.0074 FDR=0.0083

BCR

0 BCR 2000 1500 OxPhos BCR


p<0.0001 FDR=0.0002

0 OxPhos
p=0.0004 FDR=0.0006

0 BCR OxPhos BCR


p<0.0001 FDR=0.0002

2000 1500 1000 500 0

2000 1500

1500

3000

1500

1500 1000 500 0 OxPhos BCR

1000 1000 500 500 0 OxPhos BCR 0 OxPhos BCR 1000

2000

1000

1000 500 0 OxPhos BCR 0 OxPhos BCR

500

OxPhos

BCR

0 OxPhos BCR

Figure 2. Increased Abundance of Transcripts Encoding Mitochondrial Proteins in Primary DLBCL Tumor Biopsies
(A) Heat map representation of the relative mRNA levels of genes corresponding to the components of the mitochondrial proteome signature using the Monti et al. (left) and Lenz et al. (right) expression array data sets of primary DLBCL cases with OxPhos and BCR consensus cluster assignments. (B) Transcript abundance (probe intensity) of the indicated genes in primary OxPhos- and BCR-DLBCLs from the Monti et al. (top) and Lenz et al. (bottom) data sets. Differential expression was determined by a two-sided Mann-Whitney test and p values were corrected for multiple hypothesis testing using the false discovery rate (FDR) procedure. NDUFA5, NADH dehydrogenase (ubiquinone) 1a subcomplex subunit 5; NDUFS3, NADH dehydrogenase (ubiquinone) Fe-S protein 3; ETFA, electron-transferavoprotein A; ACAT1, acetoacetyl-CoA thiolase 1; MTHFD2, methylenetetrahydrofolate dehydrogenase 2; SOD2, manganese superoxide dismutase. See also Tables S2 and S3.

subsets, respectively (Chen et al., 2008). This prompted examination of a potential reciprocal relationship between active BCR signaling and mitochondrial FAO. Such a scenario would predict that inhibition of BCR signaling in non-OxPhos/BCR DLBCLs may manifest in increased mitochondrial FAO. Indeed, acute interference with BCR signaling using multiple independent shRNAs against SYK, an upstream component of this signaling axis, was associated with signicant elevation of basal palmitate-induced OCR (Figure 3B). Notably, OCR measurements were taken after 24 hr treatment with shRNAs, which enabled sufcient depletion of SYK (Figure S2B) without affecting cell viability (data not shown). These observations are suggestive of an underlying metabolic plasticity that governs the pattern of fuel oxidation in DLBCL subsets. To provide independent and parallel evidence for differential utilization of fatty acids in DLBCL subsets, a targeted 13C isotopomer approach was undertaken. A total of eight cell lines with known designation as OxPhos or BCR were cultured in media containing uniformly labeled 13C-palmitate (U13C-palmitate), and 13C enrichment was assessed in a dened set of intermediates derived from fatty acid metabolism. The complete oxidation of 13C-palmitate by mitochondria yields eight acetyl units that can donate carbons to the TCA cycle in the form of citrate labeled on two carbons (13C2-citrate), which can, in turn, lead to the formation of 13C2-a-ketoglutarate that is in
550 Cancer Cell 22, 547560, October 16, 2012 2012 Elsevier Inc.

isotopic equilibrium with 13C2-glutamate (Figure 4A). The relative level of 13C enrichment in both 13C2-citrate and 13C2-glutamate was signicantly higher in the OxPhos-DLBCL cell lines (Figure 4B). This is consistent with greater entry of palmitate carbons into the TCA cycle. In addition to oxidative metabolism for energy production, palmitate may have important biosynthetic roles, including direct incorporation into phosphatidyl choline (PC) to yield U13C-Palmitate-PC (Figure 4A). Alternatively, palmitate-derived citrate can be exported from mitochondria and its carbons subsequently incorporated into the head group of PC (13C2-acetyl-PC; Figure 4A). In both DLBCL subtypes, direct incorporation of palmitate into PC (U13C-palmitate-PC) was observed. In addition, a larger enrichment of palmitate carbons in 13C2-acetyl-PC was detected in OxPhos cell lines (Figure 4B). A preferential increase in palmitate utilization in OxPhosDLBCLs prompted examination of its effect on proliferation in these cells. Supplementation of serum-free media containing amino acids with palmitate and carnitine led to a modest but signicant and selective increase in the proliferation of OxPhos-DLBCLs (Figure 4C). In the absence of palmitate, carnitine, which is required for mitochondrial import of long-chain fatty acids, did not inuence proliferation (Figure 4C).

Cancer Cell
Metabolic Fingerprints in DLBCL Molecular Subsets

OxPhos
Karpas 422 OCI-Ly4 Pfeiffer
Type 3 GCB

BCR
SU-DHL-4 SU-DHL-6 OCI-Ly1 OCI-Ly7 OCI-Ly3

non-ABC

ABC

OCI-Ly10 HBL-1 U2932

1100 1000 900 800 700 600 500 400 300 200 100 0

*** ***
ATP-Coupled OCR (pmoles O2 / min)

800 700 600 500 400 300 200 100 0

*** ***

Basal OCR (pmoles O2 / min)

**

*** ***

**

*** ***

N lm S ita G te l G uco lu se ta m in e

N lm S i G tate l G uco lu se ta m in e

N lm S i G tate l G uco lu se ta m in e

N lm S ita G te l G uco lu se ta m in e

N lm S i G tate l G uco lu se ta m in e

** *
Basal OCR (pmoles O2 / min)
sh Ctrl sh SYK #1 sh SYK #3 sh SYK #4

non-ABC

ABC

sh Ctrl sh SYK #1 sh SYK #3 sh SYK #4

1200 1100 1000 900 800 700 600 500 400 300 200 100 0

** * *** ** * ***

**

OCI-Ly1

OCI-Ly7

U2932

Figure 3. Mitochondrial Carbon Substrate Oxidation in DLBCL Subsets and Its Regulation by BCR Signaling
(A) Basal (left) and ATP-coupled (right) OCR in DLBCL subsets. OCR values shown are the average of all cell lines per DLBCL subtype indicated on top. For each cell line, 713 independent OCR measurements were taken. NS, no substrate added exogenously. (B) Palmitate-stimulated basal OCR in non-OxPhos DLBCL cell lines after acute knockdown of SYK. Error bars, SEM. *p < 0.05; **p < 0.01; ***p < 0.001; two-tailed Students t test. See also Figure S2.

To probe the prosurvival benet of mitochondrial FAO in DLBCL subsets, this program was inhibited using 4-bromocrotonic acid (BrCA), which irreversibly inhibits mitochondrial b-oxidation of both long- and short-chain fatty acids (el-Aleem and Schulz, 1987). Acute treatment of DLBCL cell lines (4 hr) with BrCA interfered with palmitate stimulation of basal OCR in OxPhos-DLBCL cell lines (Figure S3), while longer treatment (24 hr) was selectively toxic to this subset compared with BCR-DLBCLs (Figure 4D). These results suggest that the mitochondrial FAO program provides prosurvival benets to OxPhos-DLBCLs. Programmatic Regulation of FAO and Its Relevance to OxPhos-DLBCL Survival The concomitant increase in the abundance of several mitochondrial FAO enzymes in OxPhos-DLBCLs is consistent, at

least in part, with a programmatic increase in the transcriptional regulation of this pathway. These observations warranted systematic interrogation of the primary DLBCL transcript data sets for the prevalence of a dened list of transcriptional regulators of this pathway. Nuclear receptor peroxisome proliferator-activated receptor (PPAR) g transcripts were found to be signicantly more abundant in primary OxPhos-DLBCLs in two independent cohorts of DLBCL tumors (Figure S4A). This is consistent with the observation that several enzymes identied in the mitochondrial protein signature of OxPhos-DLBCLs such as ETF, ACAD, and HADH, as well as multiple subunits of mitochondrial respiratory chain complexes I, II, and ATP synthase, are downstream targets of PPARg (Hsiao et al., 2011). To probe the functional relevance of PPARg in DLBCL subsets, we tested the effect of its depletion using three independent siRNAs and their corresponding mixture (Figures 5A and S4B).
Cancer Cell 22, 547560, October 16, 2012 2012 Elsevier Inc. 551

N lm S i G tate l G uco lu se ta m in e

Pa

Pa

Pa

Pa

Pa

Pa

Cancer Cell
Metabolic Fingerprints in DLBCL Molecular Subsets

A
CYTOPLASM

U13C-Palmitate
13

B
PC 13 C2-Acetyl-PC
13

Karpas 422

SU-DHL-4

U C-Palmitate-PC

OxPhos

OCI-Ly4 Pfeiffer Toledo

BCR

SU-DHL-6 OCI-Ly1 OCI-Ly7

C2-Citrate

Relative 13C Enrichment

13

C2-Citrate

4 3 2 1 0

MITOCHONDRION

13

C2-Acetyl-CoA

**

-Oxidation
13

C2-Citrate
13

C2-Glutamate

NAD NADH

C2-ketoglutarate 13C2-Glutamate 3.0 3.0 2.5 2.5 * 2.0 2.0 1.5 1.5 1.0 1.0 0.5 0.5 0 0
13

NAD+

TCA CYCLE
NADH FADH2 FAD

13

C2-ketoglutarate

Relative 13C Enrichment

NADH

13 C2-Acetyl-PC U13C-Palmitate-PC 4 2.0

NAD+

1.5 1.0 0.5 0

3 2 1 0

***

100

Proliferation over 24 hr % Edu Positive Cells

OxPhos
80 60 40 20 0
K a arp s4

Ctrl Carnitine Carnitine + Palmitate

BCR

Ctrl Carnitine Carnitine + Palmitate

**

**

***

**

22

I OC

-Ly

4 Pfe

iffe

r To

led

o S

D U-

HL

-4 S D U-

HL

-6 OC

I-L

y1 OC

I-L

y7

100

% Survival Annexin V and PI negative

*** *** *** ***


OxPhos BCR
Ctrl BrCA 7.5 M Ctrl BrCA 7.5 M

80 60 40 20 0

a arp

s4

22

I OC

-Ly

4 Pfe

iffe

r To

led

o SU

DH

L-4 HL-6 -Ly1 y7 I-L -D OCI OC SU

Figure 4. Palmitate Metabolism and Its Effect on DLBCL Proliferation and Survival
(A and B) 13C isotopomer analysis of uniformly labeled palmitate (U13C-Palmitate). (A) Schematics depicting the number of carbons labeled (lled circles) in a dened set of metabolites derived from palmitate. Metabolites marked in red are selectively elevated in OxPhos-DLBCL cell lines in (B). (B) 13C enrichment in palmitate-derived metabolites. For each metabolite, cumulative data obtained from all four OxPhos-DLBCL cell lines are shown relative to the mean value of that metabolite in all four BCR-DLBCL cell lines listed on top.

552 Cancer Cell 22, 547560, October 16, 2012 2012 Elsevier Inc.

Cancer Cell
Metabolic Fingerprints in DLBCL Molecular Subsets

Figure 5. Programmatic Regulation of FAO and Its Relevance to DLBCL Survival


(A) Effect of siRNA-mediated depletion of PPARg on the survival of DLBCL subsets. (B) Survival of the indicated DLBCL cell lines cultured in the absence or presence of increasing concentrations of T0070907 for 96 hr. Error bars, SEM. **p < 0.01; ***p < 0.001; twotailed Students t test. See also Figure S4.

The most robust decrease in PPARg protein levels was achieved using the siRNA mix (Figure S4C) and was accompanied by signicant apoptosis in OxPhos-DLBCLs compared with the BCR subset (Figure 5A). To provide a pharmacologic correlate to these ndings, we also tested the effect of two selective PPARg antagonists T0070907 and GW9662 (Lee et al., 2002; Leesnitzer et al., 2002). Short-term treatment of DLBCL cell lines with these compounds blocked palmitate stimulation of OCR (data not shown). Upon longer incubation periods (96 hr), these compounds proved selectively toxic to OxPhos-DLBCLs (Figures 5B and S4D). The aforementioned genetic and pharmacologic approaches to PPARg inhibition complement and extend the results shown in Figure 4D that pharmacologic inhibition of mitochondrial b-oxidation program is toxic to OxPhos-DLBCLs and are collectively congruent with the idea that a sustained mitochondrial FAO program may be relevant for the survival of OxPhos-DLBCLs.

Differential Utilization of GlucoseDerived Carbons in DLBCL Subsets Identication of PDH as a component of the mitochondrial proteome signature in OxPhos-DLBCLs predicts differential mitochondrial handling of pyruvate. Indeed, biochemical analysis of isolated mitochondria derived from a panel of DLBCL cell lines indicated signicant increase in PDH enzyme activity in OxPhos-DLBCLs (Figure 6A). Increased PDH activity would further predict diminished availability of glucose-derived pyruvate for lactate synthesis. Biochemical quantication of glucose-derived lactate secreted by OxPhos-DLBCL cell lines compared to non-OxPhos/BCR counterparts indicated that this is indeed the case (Figure 6A). These observations prompted a more detailed examination of the fate of glucose carbons in DLBCL subsets. Beyond generation of pyruvate, glucose-derived metabolites are central to several biosynthetic pathways (Figure 6B). For example, glucose-6-phosphate can enter the pentose phosphate pathway to yield ribose sugars for nucleic acid synthesis and NADPH generation for lipid synthesis and ROS detoxication (Tennant et al., 2010; Vander Heiden et al., 2009). Dihydroxyacetone phosphate, which is in equilibrium with glyceraldehyde-3-phosphate, is used in glycerol synthesis, providing a necessary backbone for membrane phospholipids. Finally, glucose-derived citrate and aspartate, a surrogate for oxaloacetate (OAA), can be used in lipid and nucleotide synthesis, respectively. To examine differences in the terminal fate of glucose-derived pyruvate and the branch points at which glucose carbons divert to biosynthetic pathways in different DLBCL subsets, we carried out a targeted 13C isotopomer analysis using uniformly labeled 13C-glucose (U13C-glucose). Glucose uptake was comparable in DLBCL cell lines as evident from the similar levels of remaining U13C-glucose in the

(C) Effect of palmitate supplementation on the proliferation of DLBCL cell lines. Control denotes serum-free media containing all amino acids except L-glutamine. (D) Survival of DLBCL subsets cultured in the absence or presence of BrCA for 24 hr. Error bars, SEM. *p < 0.05; **p < 0.01; ***p < 0.001; two-tailed Students t test. See also Figure S3.

Cancer Cell 22, 547560, October 16, 2012 2012 Elsevier Inc. 553

Cancer Cell
Metabolic Fingerprints in DLBCL Molecular Subsets

A
OxPhos
Karpas 422 OCI-Ly4 Pfeiffer Toledo

B
BCR
SU-DHL-4 OCI-Ly3 SU-DHL-6
OCI-Ly1
U13C-Alanine U13C-Glucose U13C-Lactate

non-ABC

ABC

OCI-Ly10 HBL-1 U2932

13

PC C-Glycerol-PC U13C-Glucose-6-Phosphate U13C-Pentose

13

PC C-Acetyl-PC

OCI-Ly7

PDH Activity (mOD/min)

3.0 2.5 2.0 1.5 1.0 0.5

***

***
U13C-Glyceraldehyde-3-Phosphate

CYTOPLASM

13

U13C-Alanine
13

U13C-Lactate C3-Pyruvate

C2-Citrate

MITOCHONDRION

PDH
13 13 C2-Aspartate C2-Acetyl-CoA

25

Lactate (nmoles)

* **

13

C2-Citrate
13

20 15 10

C2-Glutamate

13

C2-OAA

13

5 0

C2-ketoglutarate

C
Relative 13C Enrichment

U13C-Glucose in Media
1.5 1.0 0.5 0

U13C-Glucose-6 -Phosphate
2.0 1.5 1.0 0.5 0 0.5 0

U13C-Pentose
1.5 1.0

13

C-Glycerol-PC

***

1.5

**

1.0 0.5 0

***

Relative 13C Enrichment

U C-Lactate (Intracellular)
1.5 1.0 0.5 0
13

13

U C-Alanine (Intracellular)
1.5

13

U13C-Lactate (Extracellular) 1.5 1.5

U13C-Alanine (Extracellular)

**

1.0 0.5 0

1.0 0.5 0

**

1.0 0.5 0
13

**

Relative 13C Enrichment

C2-Citrate

13

C2-Acetyl-PC

13

1.5

C2-ketoglutarate

C2-Glutamate
2.0 1.5 1.0

13

C2-Aspartate

1.5

1.5 1.0 0.5 0

1.5 1.0 0.5 0

**
1.0 0.5 0 1.0 0.5 0

0.5 0

Figure 6. Utilization of Glucose-Derived Carbons in DLBCL Subsets


(A) PDH enzyme activity (middle) and lactate production from glucose (bottom) in DLBCL subsets. Data are cumulative from independent DLBCL cell lines listed on top. (B and C) 13C isotopomer analysis of uniformly labeled glucose (U13C-glucose) (B) Schematics depicting the number of carbons labeled (lled circles) in intermediary metabolites of glucose metabolism. (C) 13C enrichment in glucose-derived metabolites. For each metabolite, cumulative data obtained from all four OxPhos-DLBCL cell lines (Karpas 422, OCI-Ly4, Pfeiffer, and Toledo; red bars) are shown relative to the mean value of that metabolite in all four BCR-DLBCL cell lines (SU-DHL-4, SU-DHL-6, OCI-Ly1, and OCI-Ly7; black bars). Error bars, SEM. *p < 0.05; **p < 0.01, ***p < 0.001; two-tailed Students t test.

554 Cancer Cell 22, 547560, October 16, 2012 2012 Elsevier Inc.

Cancer Cell
Metabolic Fingerprints in DLBCL Molecular Subsets

media following 8 hr incubation (Figure 6C). The BCR cell lines displayed a larger contribution of glucose-derived carbons to the synthesis of pentose sugars compared with the OxPhos cell lines (U13C-pentose; Figure 6C). This is consistent with reduced levels of U13C-glucose-6-phosphate precursor observed in this subset (Figure 6C), provided that the oxidative branch of the pentose phosphate pathway is utilized to derive pentose sugars. In addition, the contribution of glucose carbons to glycerol incorporation into PC was signicantly higher in BCR cell lines (13C-glycerol-PC; Figure 6C). The overall glycolytic capacity was assessed by comparing both intracellular and extracellular (secreted) U13C-lactate and U13C-alanine (a surrogate for pyruvate). BCR cell lines had signicantly higher intracellular and secreted U13C-alanine and U13C-lactate (Figure 6C). These data are consistent with biochemical evidence that glucose-derived lactate is higher in these cells (Figure 6A), suggesting a higher glycolytic ux. Glucose-derived pyruvate can enter the TCA cycle through a PDH-catalyzed reaction yielding citrate or through a pyruvate-carboxylase-mediated anaplerotic reaction that generates OAA (Fan et al., 2009). The ratio of PDH and pyruvate carboxylase activities in tumors is variable (DeBerardinis et al., 2007; Fan et al., 2009), suggesting signicant diversity and tumor type specicity in the mode of TCA cycle entry of glucosederived carbons. The PDH reaction can be traced by the pattern of carbon labeling in citrate as 13C2-citrate. The relative 13C enrichment in this metabolite was signicantly higher in OxPhos cell lines (Figure 6C), indicating a greater overall diversion of glucose carbons to the TCA cycle. Elevated PDH-catalyzed formation of 13C2-citrate in OxPhos-DLBCLs is consistent with increased levels of PDH and enzyme activity in this subset. Among other TCA cycle intermediates measured in this analysis, 13 C enrichment in aspartate (surrogate for OAA) was relatively higher in OxPhos compared with BCR cell lines (13C2-aspartate; Figure 6C). Despite elevated PDH activity in OxPhos-DLBCLs and higher enrichment of glucose carbons in the aforementioned TCA cycle intermediates, glucose is not fully oxidized to stimulate mitochondrial OCR in these cells (Figures 3A and S2A). This may be due to greater diversion of glucose-derived TCA intermediates to biosynthetic pathways. Overall, the distinct enrichment patterns of carbons in glucose-derived metabolites indicate relative diminution of lactate production in OxPhosDLBCL and an attendant increase in the entry of glucose carbons in the TCA cycle via citrate. Contribution of Mitochondrial Metabolism to Cellular ATP Budget in DLBCL Subsets The observed differences in utilization of palmitate- and glucosederived carbons, as well as ATP generation through mitochondrial oxidation of palmitate (ATP-coupled respiration) in DLBCL subsets, warranted comparison of mitochondrial and nonmitochondrial contributions to the cellular energy budget. To this end, the portion of total cellular ATP that is sensitive to inhibition of glycolysis versus mitochondrial metabolism was assessed (Guppy et al., 2002). Compared to BCR-DLBCL, the OxPhos subset derives a signicantly higher portion of its total energy (70%) from mitochondrial oxidative metabolism than from glycolysis (Figure 7A). The higher contribution of mitochon-

dria to total cellular ATP in OxPhos-DLBCLs is concordant with increased expression of mitochondrial ATP synthase (complex V) subunits (Figures 1 and S1A; Table S1) and an elevated rate of mitochondrial ATP synthesis in OxPhos-DLBCL cell lines (Figure 7B). These observations could be potentially explained by differences in mitochondrial content in DLBCL subtypes. However, assessment of the steady-state content of mitochondria in DLBCL cell lines did not reveal any subtype-specic differences (Figure S5A). In addition, mitochondrial SNP copy number analysis in a cohort of primary biopsies with OxPhos or BCR assignments did not reveal any differences (Figure 7C). Increased contribution of mitochondria to total cellular ATP may reect not only distinct channeling of carbon substrates in mitochondria but also differential activity or efciency of mitochondrial ETC complexes. Studies in isolated mitochondria derived from DLBCL cell lines enabled direct assessment of mitochondrial respiration at the organelle level independent of the cytosolic processing of carbon substrates. Mitochondrial respiratory states were assessed using glutamate/malate and succinate as complex I- and II-linked substrates, respectively (Figures 7D and 7E). When measuring complex II activity, rotenone was included with succinate to inhibit the reverse ow of electrons to complex I. Respiratory rates in the presence of substrate alone (also known as state II respiration) were higher in OxPhos-DLBCL mitochondria (Figure 7F). The addition of ADP, which mimics a state of energy demand driving high rates of respiration (also known as state III respiration) elicited signicantly higher OCR in OxPhosDLBCL mitochondria supplied with either complex I or II substrates (Figure 7F). This increased respiration is used for ATP synthesis and is sensitive to oligomycin. Accordingly, OCR values in the presence of substrate, ADP and oligomycin (state IV respiration) were also signicantly higher in OxPhosDLBCLs compared with BCR-DLBCLs (Figure 7F). These results provide a direct link between the OxPhos signature and an actual quantitative increase in ETC activity in this DLBCL subset. Moreover, independent biochemical measurements of mitochondrial complexes I, II, and IV in immunocapture assays provided corroborative biochemical evidence for signicant elevation of these enzyme activities in OxPhos-DLBCLs (Figure S5B). In aggregate, our observations both at the level of intact cells (Figures 3A, 7A, and 7B) and isolated mitochondria (Figures 7F and S5B) demonstrate that the OxPhos signature captures a program of mitochondrial metabolism and energy transduction that is selectively activated in this DLBCL subset. The Relevance of ROS Content and Glutathione Synthesis in DLBCL Subtypes Mitochondria are a predominant source of ROS. While ROS signaling is important for a myriad of cellular functions, excessive mitochondrial superoxide can damage mitochondrial DNA, modify proteins and lipids, and inhibit aconitase activity, thus limiting oxidative phosphorylation. Elevated ETC activity in OxPhos-DLBCL, particularly of complex I, predicted increased accumulation of mitochondrial superoxide. However, MnSOD is also concomitantly elevated in this subset (Figures 1, 2, and S1A; Table S1), potentially as a mechanism to counteract increased burden of mitochondrial superoxide. Efcient clearance of ROS would also ensure that oxidative phosphorylation
Cancer Cell 22, 547560, October 16, 2012 2012 Elsevier Inc. 555

Cancer Cell
Metabolic Fingerprints in DLBCL Molecular Subsets

A
OxPhos
100 90 80 70 60 50 40 30 20 10 0

B
BCR

C
90 80 70 60 50 40 30 20 10 0 2.0

Figure 7. Contribution of Mitochondrial Metabolism to Cellular ATP and Energy Transduction in DLBCL Subsets
(A) Percent contribution of glycolysis and mitochondrial metabolism to total cellular ATP. For each subtype, cumulative data from four OxPhosDLBCL (Karpas 422, OCI-Ly4, Pfeiffer, and Toledo) and four BCR-DLBCL (SU-DHL-4, SUDHL-6, OCI-Ly1, and OCI-Ly7) cell lines are shown. (B) Mitochondrial ATP synthesis rate. Cumulative data from four OxPhos-DLBCL (red bar) and four BCR-DLBCL (black bar) cell lines are shown as in (A). (C) Average copy number of 110 mitochondrial SNP probes for 39 OxPhos-DLBCL and 33 BCR-DLBCL cases. Differences were tested using a MannWhitney U test and found to be nonsignicant. (DF) OCR in isolated mitochondria in different respiratory states. (D) Schematics of mitochondrial respiratory complexes and substrates as well as mitochondrial inhibitors used to measure their specic activities. (E) Representative OCR traces in mitochondria isolated from DLBCL cell lines indicating respiratory states examined as described in the Supplemental Experimental Procedures. (F) OCR in isolated mitochondria measured using complex I- or complex II-linked substrates. Data are derived from four cell lines per DLBCL subset listed on top. Error bars in (A), (B), and (F), SEM. **p < 0.01; ***p < 0.001; two-tailed Students t test. See also Figure S5.

% Total Cellular ATP

***

Raw Copy Number

***
RLU/min

**

1.5 1.0 0.5 0

tic drial ol ytic drial n n c ol y c Gl y itocho Gl y itocho M M

OxPhos n=39

BCR n=33

F
OxPhos
Karpas 422 OCI-Ly4
TCA Cycle
ATP ADP+Pi
Malate Succinate

BCR
SU-DHL-4 OCI-Ly3 SU-DHL-6
OCI-Ly1

Pfeiffer
O2 H2O eIV

non-ABC

ABC

OCI-Ly10 HBL-1 U2932

Toledo

OCI-Ly7

FCCP H+ + H H+
ATP Synthase I

NAD+ FAD NADH FADH2

Glutamate/Malate Succinate + Rotenone

eII

e-

Q+

eIII

cyt c

State II OCR (pMoles O2 / min)

H+

400 350 300 250 200 150 100 50 0

400

*** ***

H+

*** ***

350 300 250 200 150 100 50 0

Oligomycin

Antimycin A (AA)

Rotenone

E
State III OCR (pMoles O2 / min)

1600 1400 1200 1000 800 600 400 200 0 350

*** ***

1600 1400 1200 1000 800 600 400 200

*** ***

2291

ADP Oligomycin FCCP

AA

OCR (pMoles O2 / min)

2040 1788 1537 1285 1034 782 531 279 28 -224 0 4 9 13 17 22 26 30 34 39 43


State II State III State IV

State IV OCR (pMoles O2 / min)

300 250 200 150 100 50 0

** **

Time (Min)

could remain elevated in OxPhos-DLBCLs. Assessment of mitochondrial superoxide using MitoSOX Red revealed lower steady state levels in OxPhos-DLBCLs (Figure 8A). These observations were further integrated with the total cellular ROS levels using CM-H2DCFDA, a uorescent probe that is sensitive to oxidation by peroxyl, alkoxyl, peroxynitrite, NO2_, CO3_, and OH_ radicals and can thus serve as an indicator of overall oxidative stress. The CM-H2DCFDA signal intensity was also signicantly lower in OxPhos-DLBCLs (Figure 8A). A potential explanation for these observations is increased diversion of peroxide generated from the SOD-catalyzed reaction into the antioxidant glutathione (GSH) system. Consistent with this possibility, GSH levels were signicantly higher in OxPhosDLBCLs (Figure 8A), suggesting increased capacity for ROS detoxication. Given the quantitative differences in ROS and GSH content in DLBCL cell lines, we hypothesized that the capacity to maintain a large GSH pool may be required for the survival of
556 Cancer Cell 22, 547560, October 16, 2012 2012 Elsevier Inc.

OxPhos-DLBCLs. To test this possibility, de novo GSH synthesis was inhibited by *** 350 *** shRNA-mediated depletion of g-glutamyl 300 cysteine synthase (GCS), a rate-limiting 250 200 enzyme that catalyzes the introduction 150 of an amide linkage between the g100 carboxyl group of glutamate and cysteine 50 0 (Figure 8B). OxPhos-DLBCL cell lines were signicantly more sensitive to GCS knockdown compared with the BCR subset (Figures 8C and S6), suggesting that they may be more reliant on GSH for survival.
0

DISCUSSION Our integrative analysis using proteomics, mitochondrial respirometry, and metabolomics have unraveled metabolic distinctions in DLBCL subsets. We show that, compared with non-OxPhos/BCR DLBCLs, nutrient and energy metabolism in OxPhos-DLBCL have a signicant mitochondrial component, marked by elevated oxidative phosphorylation, increased contribution of mitochondria to total cellular energy budget, greater incorporation of fatty acid- and glucose-derived carbons into the TCA cycle, and increased lipogenesis from these carbon substrates. In comparison, the non-OxPhos DLBCLs have greater glycolytic ux. These studies also provide a clear example of heterogeneity in fuel utilization pathways even within the same disease entity.

Cancer Cell
Metabolic Fingerprints in DLBCL Molecular Subsets

OxPhos
Karpas 422 OCI-Ly4 Pfeiffer Toledo

BCR
SU-DHL-4 OCI-Ly3 SU-DHL-6 OCI-Ly1 OCI-Ly7 OCI-Ly10 HBL-1 U2932

non-ABC

ABC

MitoSOX Red/TMRE Intensity

Mean CM-H2DCFDA Intensity

0.4 0.3 0.2 0.1 0

8000 6000 4000 2000 0

**

mol GSH /g of protein

***

10000

0.05 0.04 0.03 0.02 0.01 0

**

OxPhos
sh Ctrl sh GCS #2 sh GCS #4

BCR non-ABC
sh Ctrl sh GCS #2 sh GCS #4

ABC

sh Ctrl sh GCS #2 sh GCS #4

Glutathione

ATP

Glycine

Viability (Annexin V/PI negative)

ADP

100 80 60 40 20 0

*** *** *** *** *** *** *** ***

-Glutamyl-cysteine

GCS

Cysteine
ADP ATP

Glutamate

Ka

rpa

s4

22

OC

I-L

y4

Pfe

iffe

r To

led

o SU -D

HL

-4 SU -D

HL

-6 OC

I-L

y1 OC

I-L

y7 OC

I-L

y3 OC

I-L

y1

HB

L-1

U2

93

Figure 8. Differential Contribution of ROS Detoxication to Survival of DLBCL Subsets


(A) Mitochondrial superoxide (left), total cellular ROS (middle) and GSH (right) levels in DLBCL subtypes. Data are derived from four cell lines per DLBCL subset listed on top. (B and C) De novo GSH synthesis pathway (B) and the effect of GCS depletion on DLBCL survival (C). Cell viability was assessed 72 hr after knockdown. GCS, g-glutamyl cysteine synthase. Error bars, SEM. *p < 0.05; **p < 0.01, ***p < 0.001; two-tailed Students t test. See also Figure S6.

The differential utilization of glucose- and fatty-acid-derived carbons in OxPhos versus non-OxPhos DLBCLs appears to parallel the absence or presence of functional BCR signaling, respectively. This is consistent with our observations that acute inhibition of BCR signaling upon SYK depletion is sufcient to enhance palmitate-induced mitochondrial OCR in BCR-DLBCL cell lines. BCR-derived signals are critical for growth and survival ppers, of mature B cells as well as multiple B cell lymphomas (Ku 2005). These signals also trigger glucose utilization in a PI3Kdependent manner that is marked by an initial increase in lactate production and a subsequent shift to the pentose phosphate pathway during progression to S phase (Doughty et al., 2006). Reduction in glycolytic ux and incorporation of glucose carbons into the pentose pool in OxPhos compared with non-OxPhos DLBCLs is consistent with published observations that OxPhos-DLBCLs do not display the full phosphorylation pattern of signaling intermediates following BCR crosslinking and are insensitive to inhibitors of BCR signaling (Chen et al., 2008).

The common metabolic prole of non-OxPhos DLBCLs is striking, given that these tumors rely on multiple components of BCR signaling (Chen et al., 2008; Compagno et al., 2009; Davis et al., 2010). This may suggest that proliferation and survival mechanisms in these tumors may converge on glycolysis. The clear distinction between the metabolic proles of OxPhos- and BCR-DLBCLs suggests that the specic pattern of nutrient metabolism in the former may provide an alternative survival program independent of the BCR network. Moreover, the OxPhos metabolic signature may have broader implications for other tumor types that are independent or have lost dependency on the components of BCR signaling. Several independent approaches to inhibit the mitochondrial FAO program signicantly compromised the survival of OxPhos-DLBCL. Within this context, BrCA-mediated inhibition of mitochondrial FAO and pharmacologic or genetic interference with PPARg were selectively toxic to OxPhos-DLBCL. The relevance of PPARg in multiple cancer models and the antitumor
Cancer Cell 22, 547560, October 16, 2012 2012 Elsevier Inc. 557

Cancer Cell
Metabolic Fingerprints in DLBCL Molecular Subsets

effects of its inhibition have been reported (Burton et al., 2008). Increased PPARg activity has been recently implicated in differentiation and stimulation of antibody production in normal human B-lymphocytes (Garcia-Bates et al., 2009). Whether this is accompanied by increased FAO in normal lymphocytes remains to be determined. Nevertheless, it is possible that the metabolic signature in OxPhos-DLBCL represents pathways relevant to B-lymphocytes that are further modied to fulll nutrient and energy requirements of these tumors. Findings in other cancer models have suggested that FAO may serve as an alternative survival pathway that is triggered by glucose deprivation or lack of glucose uptake (Barger and Plas, 2010; Schafer et al., 2009). Moreover, a recent report showed increased FAO in solid tumors contributes to rapamycin resistance (Zaugg et al., 2011). The survival-promoting effect of fatty acid metabolism has also been reported in a leukemia model (Samudio et al., 2010). In this model, increased fatty acid metabolism was decoupled from oxidative phosphorylation, which appears to be distinct from enhanced ATP-coupled OCR in response to palmitate in OxPhos-DLBCL. Concomitant utilization of palmitate-derived acetyl-CoA for ATP production and citrate synthesis in OxPhos-DLBCL suggests that FAO and fatty acid synthesis may be concurrent pathways in these cells. This may be surprising in light of the inhibitory effect of citrate-derived malonyl-CoA on mitochondrial transport of long-chain fatty acids through carnitine palmitoyltransferase (CPT)1. However, several possibilities may explain these observations. The potency of inhibition by malonyl-CoA differs between the two CPT1 isoforms and is likely further inuenced by their relative abundance (McGarry and Brown, 1997). CPT1A is 80 fold less sensitive to malonyl-CoA than CPT1B and the ratio of CPT1A to 1B expression is higher in OxPhosDLBCL. The relative abundance of CPT1A and the possibility that malonyl-CoA may be used rapidly prior to its accumulation may explain how mitochondrial FAO could proceed unhindered in these cells. A scenario for rapid utilization of malonyl-CoA is active fatty acid synthesis as would be expected in a proliferating cell. In addition, fatty acid synthesis may serve as a sink for excess acetyl CoA/citrate generated due to enhanced FAO and increased entry of acetyl-CoA into the TCA cycle. Concurrent FAO and fatty acid synthesis in other tumors has been observed (Ookhtens and Baker, 1979; Ookhtens et al., 1984). It is also interesting to note that increased FAO in acute myeloid leukemia has been linked to a quiescent pool of tumor-initiating cells (Samudio et al., 2010). The precise molecular basis for concomitant use of fatty acids for ATP production and citrate synthesis in OxPhos-DLBCL awaits future studies. Parallel activation of an antioxidant defense mechanism with increased mitochondrial FAO in OxPhos-DLBCL is intriguing. A link between FAO and antioxidant capacity has been previously suggested. For example, mitochondrial FAO may serve as a source of NADPH to regenerate GSH (Pike et al., 2011). On the other hand, a reduced cellular redox state may be important for completion of b-oxidation (Korge and Weiss, 2006; Schafer et al., 2009). It is also possible that elevated GSH in OxPhos-DLBCL is secondary to its increased synthesis from glutamate. This is consistent with higher enrichment of fatty-acid-derived glutamate in this subtype. While increased GSH and MnSOD levels are in agreement with lower ROS
558 Cancer Cell 22, 547560, October 16, 2012 2012 Elsevier Inc.

content, additional mechanisms contributing to ROS handling in OxPhos-DLBCL cannot be excluded. Our ndings provide functional validation of quantiable metabolic differences associated with transcriptionally dened subsets of DLBCL. These observations also indicate that the OxPhos molecular signature is a bona de metabolic program that is selectively activated in these lymphomas, providing distinct growth and survival benets. Detailed ngerprints of the DLBCL metabolome may uncover important insights into the molecular pathogenesis and underlying heterogeneity of these lymphomas as well as additional road maps to subtypespecic therapeutic targets.
EXPERIMENTAL PROCEDURES Primary DLBCL Biopsies Protein samples were prepared from frozen biopsy specimens of newly diagnosed, previously untreated primary DLBCLs with >80% tumor involvement according to Institutional Review Board (IRB)-approved protocols from two institutions (Brigham and Womens Hospital and Dana-Farber Cancer Institute). A waiver to obtain informed consent was granted by the IRBs because otherwise discarded tissue was used. The frozen biopsy specimens from which total RNA and high-molecular-weight DNA were extracted for transcript abundance and SNP analysis have been described in a recent study (Monti et al., 2012). Sample Preparation and Labeling for iTRAQ Analysis Mitochondria isolated from DLBCL cell lines or Trizol-puried primary biopsy proteins were solubilized in 7.2 M guanidine hydrochloride, 100 mM ammonium bicarbonate. Protein concentration was determined by the Bradford assay (Bio-Rad, Hercules, CA, USA), and equal amounts of protein were reduced with 10 mM dithiothreitol (DTT) for 30 min at 56 C and alkylated with 22.5 mM iodoacetamide for 30 min at room temperature in the dark. DTT was added to a nal concentration of 20 mM to quench the remaining iodoacetamide. Proteins were digested overnight with trypsin (1:20) at 37 C after addition of 100 mM ammonium bicarbonate solution to dilute the concentration of guanidine HCl to 1 M. Digests were acidied with 10% triuoroacetic acid and desalted by C18. Aliquots of peptides were stored at 80 C. Mitochondria-derived peptides (50 mg) from three OxPhos (Karpas 422, Toledo, and Pfeiffer) and three BCR (OCI-Ly1, SU-DHL-4, and SU-DHL-6) cell lines were solubilized in 100 ml of 30% 500 mM triethylammonium bicarbonate, pH 8.5/70% ethanol, and 1 U of iTRAQ 8-plex reagent was added to each sample (Karpas 422-113, Toledo-114, Pfeiffer-115, OCI-Ly1-117, SU-DHL-4-118, SU-DHL-6-119; note that 116 and 121 reagents were not used). The solution was incubated for 1 hr at room temperature, and the reactions were combined and dried by vacuum centrifugation. Labeled peptides were desalted by C18, dried again by vacuum centrifugation, and stored at 80 C. The same procedure was used to label 100-mg aliquots of primary DLBCL biopsy-derived peptides with iTRAQ 4-plex reagent [biopsy #32 (OxPhos)-114; biopsy #39 (OxPhos)-115; biopsy #42 (BCR)-116)]. iTRAQlabeled peptide samples were analyzed and data processed as described in Supplemental Experimental Procedures. Mitochondrial Respirometry OCR was measured in real time using the XF24 extracellular ux analyzer instrument and the AKOS algorithm v1.5.069 software (Seahorse Bioscience Inc., Chicopee, MA, USA). For whole cell studies, cells were seeded on XF24 V7 plates coated with Cell-Tak (BD Biosciences, Bedford, MA, USA) at 3 3 105 cells/well in 600 ml of sodium bicarbonate-free Roswell Park Memorial Institute (RPMI) medium alone or supplemented with 10 mM glucose, 2 mM glutamine or 0.2 mM palmitate. When palmitate was tested, 0.5 mM carnitine was included in the incubation/equilibration medium to ensure import of bovine serum albumin (BSA)-conjugated palmitate into mitochondria. The plates were spun at 500 rpm (breaks off) and incubated at 37 C for 10 min to ensure cell attachment. Measurements were taken for a total of 60 min, including a 12 min incubation period prior to starting baseline measurements. Within this

Cancer Cell
Metabolic Fingerprints in DLBCL Molecular Subsets

assay time, OCR was measured for 3 min periods with 5 intervals between measurements. After baseline measurements, 2.5 mM oligomycin was added in a single 75 ml injection. In all experiments, parallel samples were run in the absence of oligomycin to ensure stable baselines as a quality control parameter for the bioenergetics health of the cells. OCR in isolated mitochondria was determined as detailed in the Supplemental Experimental Procedures. C Isotopomer-Based Metabolomics Analysis Cells (10 3 106) were incubated for 8 hr at 37 C in glutamine-free unbuffered RPMI medium containing 10 mM malate, 0.5 mM carnitine, 10 mM b-hydroxybutyrate, 10% (v/v) fetal bovine serum supplemented with 10 mM glucose, 4 mM glutamine, and 0.2 mM BSA-conjugated palmitate. When tracing glucose or palmitate carbons, the media were supplemented with U-13Cglucose or U-13C-palmitate isotopomers (Cambridge Isotope Laboratories, Andover, MA, USA) to the nal concentrations indicated above. Cell pellets and 1 ml aliquots of media were frozen on dry ice and processed for LCMS/MS and gas chromatography/mass spectrometry (GC/MS) as detailed in the Supplemental Experimental Procedures. Mitochondrial Copy Number Analysis The mitochondrial copy number was assessed using the mitochondrial SNP probes within the Affymetrix HD-SNP array 6.0 data from an independent series of primary DLBCL samples (Monti et al., 2012; GEO accession number GSE34171). The presegmentation data processing was performed according to the SNParray 6.0 analytical pipeline described elsewhere (Cancer Genome Atlas Research Network, 2008). Gene expression proling of the same samples allowed assignment of consensus cluster subtypes (Monti et al., 2012; GEO accession number GSE34171). Subsequently, the average copy number of 110 mitochondrial SNP probes was computed and visualized using a box plot for the 33 BCR and 39 OxPhos samples. Differences were tested using a Mann-Whitney U test (Dawson-Saunders and Trapp, 1994). Statistics Unless otherwise indicated, statistical analysis was performed using twotailed Students t test, assuming unequal variance. Transcript abundance was visualized with box plots (median, line; 25% and 75% quartile, box; whiskers, minimum to maximum). SUPPLEMENTAL INFORMATION Supplemental Information includes six gures, three tables, and Supplemental Experimental Procedures and can be found with this article online at http:// dx.doi.org/10.1016/j.ccr.2012.08.014. ACKNOWLEDGMENTS We thank Eric Smith for manuscript preparation. We gratefully acknowledge George Rogers, Martin Brand, David Ferrick, Min Wu, and Orian Shirihai for advice on respirometry; Wei Jiang and Frank Cook for assistance with LC-MS/MS and GC/MS; and Georg Lenz and Frank Stegmeier for HBL-1 and U2932 DLBCL cell lines. A.U.K. was supported by the Alexandra Jane Miliotis Fellowship in Pediatric Oncology, the IDEA^2 Program Grant from the Harvard-MIT Division of Health Sciences and Technology, and a Medical Student Research Training Fellowship from the Howard Hughes Medical Institute. E.N. was supported by a postdoctoral fellowship from the Swedish Research Council. This work was supported in part by funding from the Novartis Institutes for Biomedical Research (N.N.D.), National Institutes of Health Grant PO1 CA092625 (M.A.S.), and the German Research Foundation DFG Ch 735/1-1 (B.C.). The authors acknowledge generous support provided through the Dana-Farber Cancer Institute Strategic Research Initiative (to J.A.M.). N.N.D. is a recipient of the Burroughs Welcome Fund Career Award in Biomedical Sciences and a consultant for the Novartis Institutes for Biomedical Research. Received: June 16, 2011 Revised: June 7, 2012 Accepted: August 20, 2012 Published: October 15, 2012
13

REFERENCES Abramson, J.S., and Shipp, M.A. (2005). Advances in the biology and therapy of diffuse large B-cell lymphoma: moving toward a molecularly targeted approach. Blood 106, 11641174. Alizadeh, A.A., Eisen, M.B., Davis, R.E., Ma, C., Lossos, I.S., Rosenwald, A., Boldrick, J.C., Sabet, H., Tran, T., Yu, X., et al. (2000). Distinct types of diffuse large B-cell lymphoma identied by gene expression proling. Nature 403, 503511. Barger, J.F., and Plas, D.R. (2010). Balancing biosynthesis and bioenergetics: metabolic programs in oncogenesis. Endocr. Relat. Cancer 17, R287R304. Burton, J.D., Goldenberg, D.M., and Blumenthal, R.D. (2008). Potential of peroxisome proliferator-activated receptor gamma antagonist compounds as therapeutic agents for a wide range of cancer types. PPAR Res. 2008, 494161. Cancer Genome Atlas Research Network. (2008). Comprehensive genomic characterization denes human glioblastoma genes and core pathways. Nature 455, 10611068. Chen, L., Monti, S., Juszczynski, P., Daley, J., Chen, W., Witzig, T.E., Habermann, T.M., Kutok, J.L., and Shipp, M.A. (2008). SYK-dependent tonic B-cell receptor signaling is a rational treatment target in diffuse large B-cell lymphoma. Blood 111, 22302237. Choe, L., DAscenzo, M., Relkin, N.R., Pappin, D., Ross, P., Williamson, B., Guertin, S., Pribil, P., and Lee, K.H. (2007). 8-plex quantitation of changes in cerebrospinal uid protein expression in subjects undergoing intravenous immunoglobulin treatment for Alzheimers disease. Proteomics 7, 36513660. Compagno, M., Lim, W.K., Grunn, A., Nandula, S.V., Brahmachary, M., Shen, Q., Bertoni, F., Ponzoni, M., Scandurra, M., Califano, A., et al. (2009). Mutations of multiple genes cause deregulation of NF-kappaB in diffuse large B-cell lymphoma. Nature 459, 717721. Davis, R.E., Ngo, V.N., Lenz, G., Tolar, P., Young, R.M., Romesser, P.B., Kohlhammer, H., Lamy, L., Zhao, H., Yang, Y., et al. (2010). Chronic active B-cell-receptor signalling in diffuse large B-cell lymphoma. Nature 463, 8892. Dawson-Saunders, B., and Trapp, R. (1994). Basic and Clinical Biostatistics (Norwalk, CT: Appleton & Lange). DeBerardinis, R.J., Mancuso, A., Daikhin, E., Nissim, I., Yudkoff, M., Wehrli, S., and Thompson, C.B. (2007). Beyond aerobic glycolysis: transformed cells can engage in glutamine metabolism that exceeds the requirement for protein and nucleotide synthesis. Proc. Natl. Acad. Sci. USA 104, 1934519350. DeBerardinis, R.J., Lum, J.J., Hatzivassiliou, G., and Thompson, C.B. (2008). The biology of cancer: metabolic reprogramming fuels cell growth and proliferation. Cell Metab. 7, 1120. Doughty, C.A., Bleiman, B.F., Wagner, D.J., Dufort, F.J., Mataraza, J.M., Roberts, M.F., and Chiles, T.C. (2006). Antigen receptor-mediated changes in glucose metabolism in B lymphocytes: role of phosphatidylinositol 3-kinase signaling in the glycolytic control of growth. Blood 107, 44584465. el-Aleem, S.A., and Schulz, H. (1987). Evaluation of inhibitors of fatty acid oxidation in rat myocytes. Biochem. Pharmacol. 36, 43074312. Fan, T.W., Lane, A.N., Higashi, R.M., Farag, M.A., Gao, H., Bousamra, M., and Miller, D.M. (2009). Altered regulation of metabolic pathways in human lung cancer discerned by (13)C stable isotope-resolved metabolomics (SIRM). Mol. Cancer 8, 41. Garcia-Bates, T.M., Baglole, C.J., Bernard, M.P., Murant, T.I., SimpsonHaidaris, P.J., and Phipps, R.P. (2009). Peroxisome proliferator-activated receptor gamma ligands enhance human B cell antibody production and differentiation. J. Immunol. 183, 69036912. Guppy, M., Leedman, P., Zu, X., and Russell, V. (2002). Contribution by different fuels and metabolic pathways to the total ATP turnover of proliferating MCF-7 breast cancer cells. Biochem. J. 364, 309315. Hsiao, G., Chapman, J., Ofrecio, J.M., Wilkes, J., Resnik, J.L., Thapar, D., Subramaniam, S., and Sears, D.D. (2011). Multi-tissue, selective PPARg modulation of insulin sensitivity and metabolic pathways in obese rats. Am. J. Physiol. Endocrinol. Metab. 300, E164E174.

Cancer Cell 22, 547560, October 16, 2012 2012 Elsevier Inc. 559

Cancer Cell
Metabolic Fingerprints in DLBCL Molecular Subsets

Koppenol, W.H., Bounds, P.L., and Dang, C.V. (2011). Otto Warburgs contributions to current concepts of cancer metabolism. Nat. Rev. Cancer 11, 325337. Korge, P., and Weiss, J.N. (2006). Redox regulation of endogenous substrate oxidation by cardiac mitochondria. Am. J. Physiol. Heart Circ. Physiol. 291, H1436H1445. ppers, R. (2005). Mechanisms of B-cell lymphoma pathogenesis. Nat. Rev. Ku Cancer 5, 251262. Le, A., Lane, A.N., Hamaker, M., Bose, S., Gouw, A., Barbi, J., Tsukamoto, T., Rojas, C.J., Slusher, B.S., Zhang, H., et al. (2012). Glucose-independent glutamine metabolism via TCA cycling for proliferation and survival in B cells. Cell Metab. 15, 110121. Lee, G., Elwood, F., McNally, J., Weiszmann, J., Lindstrom, M., Amaral, K., Nakamura, M., Miao, S., Cao, P., Learned, R.M., et al. (2002). T0070907, a selective ligand for peroxisome proliferator-activated receptor gamma, functions as an antagonist of biochemical and cellular activities. J. Biol. Chem. 277, 1964919657. Leesnitzer, L.M., Parks, D.J., Bledsoe, R.K., Cobb, J.E., Collins, J.L., Consler, T.G., Davis, R.G., Hull-Ryde, E.A., Lenhard, J.M., Patel, L., et al. (2002). Functional consequences of cysteine modication in the ligand binding sites of peroxisome proliferator activated receptors by GW9662. Biochemistry 41, 66406650. Lenz, G., and Staudt, L.M. (2010). Aggressive lymphomas. N. Engl. J. Med. 362, 14171429. Lenz, G., Wright, G., Dave, S.S., Xiao, W., Powell, J., Zhao, H., Xu, W., Tan, B., Goldschmidt, N., Iqbal, J., et al; Lymphoma/Leukemia Molecular Proling Project. (2008). Stromal gene signatures in large-B-cell lymphomas. N. Engl. J. Med. 359, 23132323. Marin-Valencia, I., Yang, C., Mashimo, T., Cho, S., Baek, H., Yang, X.L., Rajagopalan, K.N., Maddie, M., Vemireddy, V., Zhao, Z., et al. (2012). Analysis of tumor metabolism reveals mitochondrial glucose oxidation in genetically diverse human glioblastomas in the mouse brain in vivo. Cell Metab. 15, 827837. McGarry, J.D., and Brown, N.F. (1997). The mitochondrial carnitine palmitoyltransferase system. From concept to molecular analysis. Eur. J. Biochem. 244, 114. Monti, S., Chapuy, B., Takeyama, K., Rodig, S., Hao, Y., Yeda, K., Inguilizian, H., Mermel, C., Curie, T., Dogan, A., et al. (2012). Integrative analysis reveals an outcome-associated and targetable pattern of p53 and cell cycle deregulation in diffuse large B-cell lymphoma. Cancer Cell 22, 359372. Monti, S., Savage, K.J., Kutok, J.L., Feuerhake, F., Kurtin, P., Mihm, M., Wu, B., Pasqualucci, L., Neuberg, D., Aguiar, R.C., et al. (2005). Molecular proling of diffuse large B-cell lymphoma identies robust subtypes including one characterized by host inammatory response. Blood 105, 18511861. nchez, R., Rodr guez-Enr quez, S., Saavedra, E., Mar nMoreno-Sa ndez, A., and Gallardo-Pe rez, J.C. (2009). The bioenergetics of cancer: Herna is glycolysis the main ATP supplier in all tumor cells? Biofactors 35, 209225.

Morrish, F., Neretti, N., Sedivy, J.M., and Hockenbery, D.M. (2008). The oncogene c-Myc coordinates regulation of metabolic networks to enable rapid cell cycle entry. Cell Cycle 7, 10541066. Ookhtens, M., and Baker, N. (1979). Fatty acid oxidation to H2O by Ehrlich ascites carcinoma in mice. Cancer Res. 39, 973980. Ookhtens, M., Kannan, R., Lyon, I., and Baker, N. (1984). Liver and adipose tissue contributions to newly formed fatty acids in an ascites tumor. Am. J. Physiol. 247, R146R153. Pike, L.S., Smift, A.L., Croteau, N.J., Ferrick, D.A., and Wu, M. (2011). Inhibition of fatty acid oxidation by etomoxir impairs NADPH production and increases reactive oxygen species resulting in ATP depletion and cell death in human glioblastoma cells. Biochim. Biophys. Acta 1807, 726734. Polo, J.M., Juszczynski, P., Monti, S., Cerchietti, L., Ye, K., Greally, J.M., Shipp, M., and Melnick, A. (2007). Transcriptional signature with differential expression of BCL6 target genes accurately identies BCL6-dependent diffuse large B cell lymphomas. Proc. Natl. Acad. Sci. USA 104, 32073212. Ross, P.L., Huang, Y.N., Marchese, J.N., Williamson, B., Parker, K., Hattan, S., Khainovski, N., Pillai, S., Dey, S., Daniels, S., et al. (2004). Multiplexed protein quantitation in Saccharomyces cerevisiae using amine-reactive isobaric tagging reagents. Mol. Cell. Proteomics 3, 11541169. Rossignol, R., Gilkerson, R., Aggeler, R., Yamagata, K., Remington, S.J., and Capaldi, R.A. (2004). Energy substrate modulates mitochondrial structure and oxidative capacity in cancer cells. Cancer Res. 64, 985993. Samudio, I., Harmancey, R., Fiegl, M., Kantarjian, H., Konopleva, M., Korchin, B., Kaluarachchi, K., Bornmann, W., Duvvuri, S., Taegtmeyer, H., and Andreeff, M. (2010). Pharmacologic inhibition of fatty acid oxidation sensitizes human leukemia cells to apoptosis induction. J. Clin. Invest. 120, 142156. Schafer, Z.T., Grassian, A.R., Song, L., Jiang, Z., Gerhart-Hines, Z., Irie, H.Y., Gao, S., Puigserver, P., and Brugge, J.S. (2009). Antioxidant and oncogene rescue of metabolic defects caused by loss of matrix attachment. Nature 461, 109113. n, R.V., and Gottlieb, E. (2010). Targeting metabolic transTennant, D.A., Dura formation for cancer therapy. Nat. Rev. Cancer 10, 267277. Vander Heiden, M.G., Cantley, L.C., and Thompson, C.B. (2009). Understanding the Warburg effect: the metabolic requirements of cell proliferation. Science 324, 10291033. Wright, G., Tan, B., Rosenwald, A., Hurt, E.H., Wiestner, A., and Staudt, L.M. (2003). A gene expression-based method to diagnose clinically distinct subgroups of diffuse large B cell lymphoma. Proc. Natl. Acad. Sci. USA 100, 99919996. Yuneva, M.O., Fan, T.W., Allen, T.D., Higashi, R.M., Ferraris, D.V., Tsukamoto, s, J.M., Alonso, F.J., Wang, C., Seo, Y., et al. (2012). The metabolic T., Mate prole of tumors depends on both the responsible genetic lesion and tissue type. Cell Metab. 15, 157170. Zaugg, K., Yao, Y., Reilly, P.T., Kannan, K., Kiarash, R., Mason, J., Huang, P., Sawyer, S.K., Fuerth, B., Faubert, B., et al. (2011). Carnitine palmitoyltransferase 1C promotes cell survival and tumor growth under conditions of metabolic stress. Genes Dev. 25, 10411051.

560 Cancer Cell 22, 547560, October 16, 2012 2012 Elsevier Inc.

SnapShot: Breast Cancer


Kornelia Polyak and Otto Metzger Filho Dana-Farber Cancer Institute and Harvard Medical School, Boston, MA 02215 USA

Frequency of breast cancer subtypes


TNBC Triple-negative breast cancers are ER-PR-HER2- and show significant, but not complete, overlap with the basal-like subtype of breast cancer (which is defined by differentiation state and gene expression profile).

Subtype *DCIS Luminal (nonHER2+)

Stage 0 I II III IV I II III IV I II III IV

5 year OS (%) 99 98 91 72 33 98 92 85 40 93 76 45 15

10 year OS (%) 98 95 81 54 17 95 86 75 15 90 70 37 11

TNBC 10% HER2+ breast cancers have luminal features and are characterized by ERBB2 gene amplification and overexpression leading to a dependency on HER2 signaling. HER2 + 20%

Luminal, non-HER2 + 70%

Luminal (non-HER2+) tumors are typically estrogen receptor positive, displaying high ER levels. These tumors are dependent on estrogen for growth and, therefore, respond to endocrine therapy.

**HER2+

TNBC

*Preinvasive stage **Estimated overall survival (OS) using HER2-targeted therapies

Key signaling pathways in breast cancer based on somatic mutation data


USH2A IGF-1R EGFR HER2 FGFR E-cadherin

Top 21 most commonly mutated genes in breast cancer Gene TP53 PIK3CA GATA3 MAP3K1 MLL3 CDH1 USH2A All (%) 35 34 9 8 6 6 5 3 3 3 3 3 2 2 2 2 2 2 1 1 1 Luminal 26 44 13 11 8 8 4 3 4 4 3 2 3 3 2 2 2 2 2 1 1 TNBC 54 8 0 0 3 2 8 3 0 1 1 5 1 2 1 1 0 0 1 1 0

SRC NF1 RAS mTORC2 BRAF MAP3K1 mTORC1 MEK1/2 MAP2K4 AKT GSK-3 p27 Kip1 ERK1/2
CER

PI3K

PTEN

TBL1XR1 NCOR1 RUNX1 CBFB ER NCOA3 FOXA1 MYC MYB GATA3

PTEN RUNX1 MAP2K4 NCOR1 RB1 TBX3 PIK3R1 CTCF NF1 SF3B1 AKT1 CBFB FOXA1 CDKN1B

MDM2

JNK

p53

Cyclin D1/3 CDK4 p16 Ink4A pRB

10
C
S
IN
CE 200
2

AN

CITING

STMN2

TBX3

EA

RCH

YEARS OF

Colors indicate tumor suppressors (blue), oncogenes (red), or mutant genes with unclear roles (purple), and lighter shading marks pathway components in which somatic mutations have not been identified.

Mutation frequencies (%) in all tumors, or just within luminal (including HER2+) and TNBC subtypes.

EX

ER Anastrozole Estradiol Exemestane Fulvestrant Megestrol Letrozole Raloxifenea Tamoxifen Toremifene

HER2 Afatinib Canertinib Dacomitinib Lapatinib MM-121 Neratinib Pertuzumab Trastuzumab T-DM1

PI3K Pathway (PI3K, AKT, mTOR) AZD8055b BEZ235c BGT226 BKM120 BYL719 Everolimusb GDC-0032 GDC-0068d GDC-0941 GDC-0980c INK1117 INK128b MK2206b PF-04691502c PKI-587c PX-866 Temsirolimusb XL147 XL765c

IGF, IGF-1R BMS-754807 Cixutumumab Dalotuzumab Figitumumab Ganitumab Linsitinib MEDI-573

Angiogenesis (VEGFR, PDGFR, KIT) Aflibercept Axitinib Bevacizumab Brivanib Lenvatinib MEDI-575 Motesanib Nintedanib Olaratumab Pazopanib Ponatinib Sorafenib Sunitinib Semaxanib Vandetanib Vatalanib

PARP BMN-673 CEP-9722 E7016 INO-1001 MK4827 Olaparib Rucaparib Veliparib

Others (Target) Cabozantinibe, Foretinibe, Onartuzumab (MET) AZD4547, BGJ398, Dovitinib, E-3810e, HGS1036 (FGFR) AUY922, Retaspimycin, Tanespimycin (HSP90) Ruxolitinib (JAK) Denosumab (RANKL)

a Raloxifene is used for breast cancer prevention, not treatment, b mTOR inhibitor, c dual PI3K/mTOR inhibitor, d AKT inhibitor, e also inhibits VEGFR.

562 Cancer Cell 22, October 16, 2012 2012 Elsevier Inc.

DOI 10.1016/j.ccr.2012.06.021

See online version for legend and references.

SnapShot: Breast Cancer


Kornelia Polyak and Otto Metzger Filho Dana-Farber Cancer Institute and Harvard Medical School, Boston, MA 02215 USA
Breast cancer is the most commonly diagnosed cancer and the principle cause of cancer-related mortality in women worldwide. Breast tumors are highly heterogeneous and are classified based on: (1) histology into ductal or lobular carcinomas, (2) differentiation state/gene expression profiles into luminal and basal-like subtypes, and (3) the expression of estrogen and progesterone receptors and HER2 into ER+, HER2+, and ER PR HER2 (triple-negative breast cancer, TNBC) subtypes. HER2+ and ER+ tumors all have luminal features, whereas TNBCs show significant but not complete overlap with basal-like subtype. The categorization of breast tumors based on hormone receptor and HER2 status and the use of antihormonal and HER2-targeted therapy, respectively, are among the first examples for molecular-based classification and personalized cancer treatment that made a significant difference in clinical outcomes. The widespread use of mammograms has lead to increased diagnosis of early stage disease, including ductal carcinoma in situ (DCIS), which also contributed to a decrease in mortality rates. The pie chart in the adjacent figure shows frequencies of various clinically relevant breast cancer subtypes, and the top table summarizes the prognoses of these subtypes. Systematic characterization of breast cancer genomes has identified somatic mutations in several key signaling pathways depicted in the adjacent figure. The top 21 genes with the most frequent nucleotide sequence changes are summarized in the middle table. The increased activities of these signaling pathways serve both as therapeutic targets and as biomarkers guiding the selection of patients who would most likely benefit from particular therapies. Current pathways and compounds used for rationally designed targeted therapy in breast cancer are listed in the bottom table. Estrogen Receptor Alpha The estrogen receptor alpha (ER) nuclear hormone receptor, encoded by ESR1, is a ligand-dependent transcription factor. Upon estrogen binding, ER directly and indirectly activates the expression of numerous genes. ERs activity is modulated by coactivators, including NCOA3. Whereas ER is expressed in only a small subset of cells in normal breast epithelium, where it plays an important role in breast development and differentiation, 65%70% of breast tumors display high ER levels and estrogen dependency for growth. Therefore, these ER + tumors respond to endocrine therapy. Several therapeutic approaches have been designed to modulate ER activity, including competitive antagonists, downregulators of ER protein levels, and inhibitors of the aromatase enzyme that produces estrogen. HER2 HER2, encoded by the ERBB2 proto-oncogene, is a receptor tyrosine kinase (RTK) and a member of the epidermal growth factor receptor (EGFR) family. HER2 forms heterodimers with all three other members of the EGFR family, yielding different complexes with different kinase activities and physiologic functions. ERBB2 is amplified and overexpressed in ~20%25% of breast carcinomas, leading to their dependency on HER2 signaling. The number of HER2-targeted therapies has increased significantly in recent years; these therapies include simple or drug-conjugated antibodies and small-molecule inhibitors of RTK activity specific for HER2 or broader for additional EGFR family members. These therapies have significantly improved disease-specific and overall survival of patients with HER2+ breast cancer. PI3K, AKT, and PTEN PIK3CA encodes the 110 kDa catalytic subunit of a class I phosphatidylinositol 3-kinase (PI3K). The 85 kDa regulatory subunit acts as an adaptor and mediates the activation of PI3K by RTKs and other kinases. PI3K phosphorylates phosphatidylinositol 4-phosphate and phosphatidylinositol 4,5-bisphosphate to generate phosphatidylinositol-3,4,5triphosphate (PIP3). PIP3 plays a key role in regulating the activity of signaling cascades involved in cell growth, survival, proliferation, metabolism, motility, and morphology by recruiting PH-domain-containing proteins, including AKT1 and PDPK1, to the cell membrane. A large fraction of human malignancies, including ~30% of breast carcinomas, have somatic mutations in PIK3CA leading to constitutive activation of the downstream signaling pathway. The activity of PI3K is antagonized by the PTEN tumor suppressor. PTEN is a dual-specificity phosphatase that dephosphorylates PIP3, leading to inhibition of AKT kinase activity, and can also antagonize the MAP kinase pathway via its protein phosphatase function. Germline mutations in PTEN play a role in Cowden disease and Bannayan-Zonana syndrome leading to increased cancer risk. Somatic inactivation of PTEN is common in a wide range of human cancers, including breast cancer. Approximately 50% of breast carcinomas have an activated PI3K/AKT pathway due to mutations in one of its components, making this kinase signaling cascade an attractive therapeutic target. Small-molecule inhibitors targeting PI3K, AKT, mTOR, or their combination are in various phases of clinical development. MAP3K1, MAP2K4, and JNK1 MAP3K1 is a mitogen-activated protein kinase (MAPK) kinase kinase that regulates the ERK and JNK MAPK pathways, the NF- B transcription factor, and the p300 transcriptional coactivator. MAP2K4 is another member of the MAPK-JNK signaling pathway functioning between MAP3K1 and JNK. Frequent somatic mutations in MAP3K1 and MAP2K4 have recently been reported in breast cancer, identifying MAPK-JNK signaling as a key pathway in breast cancer. ACKNOWLEDGMENTS I thank Drs. Myles Brown, Eric Winer, and Olga Pustolova for their critical reading and valuable comments. Studies in the authors laboratory are supported by NIH CA080111, US Army Congressionally Directed Research W81XWH-07-1-0294, Susan G. Komen Foundation, V Foundation, the Avon Foundation, and the Breast Cancer Research Foundation. REFErENCES 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. Curtis, C., Shah, S.P., Chin, S.F., Turashvili, G., Rueda, O.M., Dunning, M.J., Speed, D., Lynch, A.G., Samarajiwa, S., Yuan, Y., et al.; METABRIC Group; Co-chairs; Writing committee; Steering committee; Tissue and clinical data source sites; University of Cambridge/Cancer Research UK Cambridge Research Institute; British Columbia Cancer Agency; University of Nottingham; Kings College London; Manitoba Institute of Cell Biology; Cancer genome/transcriptome characterization centres; University of Cambridge/Cancer Research UK Cambridge Research Institute; British Columbia Cancer Agency; Data analysis subgroup; University of Cambridge/Cancer Research UK Cambridge Research Institute; British Columbia Cancer Agency (2012). The genomic and transcriptomic architecture of 2,000 breast tumours reveals novel subgroups. Nature 486, 346352. 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. Higgins, M.J., and Baselga, J. (2011). Targeted therapies for breast cancer. J. Clin. Invest. 121, 37973803. Shah, S.P., Roth, A., Goya, R., Oloumi, A., Ha, G., Zhao, Y., Turashvili, G., Ding, J., Tse, K., Haffari, G., et al. (2012). The clonal and mutational evolution spectrum of primary triplenegative breast cancers. Nature 486, 395399. Sjblom, T., Jones, S., Wood, L.D., Parsons, D.W., Lin, J., Barber, T.D., 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. Stephens, P.J., Tarpey, P.S., Davies, H., Van Loo, P., Greenman, C., Wedge, D.C., Zainal, S.N., Martin, S., Varela, I., Bignell, G.R., et al. (2012). The landscape of cancer genes and mutational processes in breast cancer. Nature 486, 400404. TCGA (The Cancer Genome Atlas Network) (2012). Comprehensive molecular portraits of human breast tumors. Nature 490, 61-70. Wood, L.D., Parsons, D.W., Jones, S., Lin, J., Sjblom, T., Leary, R.J., Shen, D., Boca, S.M., Barber, T., Ptak, J., et al. (2007). The genomic landscapes of human breast and colorectal cancers. Science 318, 11081113.

562.e1 Cancer Cell 22, October 16, 2012 2012 Elsevier Inc. DOI 10.1016/j.ccr.2012.06.021