Sunteți pe pagina 1din 3

Features

Tech News
The Antibody Challenge
Imagine you’re just starting in a lab. Your were traced to non-reproducibility of the including protein microarrays and Western
new PI decides to test your mettle at the commercial ATF2 and fibronectin reagents. blotting to assess specificity. About 40,000
bench with a simple project: Replicate In the wake of these findings, the clinical antibodies validated in such a manner
some immunohistochemistry results from trial was put on hold, but the original paper progressed to the next phase of testing, using
a recent publication. No problem, right? stands. After all, the data were accurate, as tissue microarrays and confocal microscopy
Not necessarily. far as they go. “The data are correct … for to check tissue and subcellular distribution
Immunohistochemistry (IHC) relies that lot of antibody.” against previously published findings. If
on antibodies, and antibodies, says Anita possible, the team checks two independent
Bandrowski of the University of California, antibody preparations to ensure that the
San Diego, “are extremely messy.” Unlike A question of validity patterns match. Sometimes, they cross-
many reagents in today’s molecular biology To get a glimpse into the challenge of reference transcriptomic data, too—cells
lab, antibodies are far from the turnkey antibody validation, take a look at the Human flush with a particular transcript should
solutions commercial vendors would have Protein Atlas (HPA) project. The HPA seeks produce a stronger signal than those that
you believe. Antibodypedia.com lists 1.2 to document the tissue distribution of at least do not.
million antibodies in its database. Some one isoform for every human protein-coding About 50% of the antibodies make it
work in Western blots but not in IHC, gene. The project, with some 150 staffers, is through this heroic process, says Fredrik
others can precipitate protein complexes but cranking out antibodies by the thousands— Ponten, the HPA’s Vice Program Director,
come up empty in flow cytometry, and some 21,984 at last count, representing 16,621 but not all are of equally high quality.
don’t work at all. More than a quarter of 246 human genes. Those antibody preparations For instance, if nothing is known about a
histone modification antibodies tested in a are used to drive an IHC juggernaut, with particular protein besides its sequence, and
2011 study were found to be non-specific; each antibody applied to 144 normal tissues, the corresponding antibody produces a
of those that were specific, 22% were 216 cancers, and a ream of cell lines, 708 distinct signal in IHC, the team will approve
unsuitable for chromatin immunoprecipi- samples in all. the antibody if no evidence exists to dispute
tation (1). Perhaps more alarmingly, some it. “That’s when you want to have solid RNA
antibodies work, but recognize the wrong sequencing data and two antibodies for each
target. In that 2011 study, 4 antibodies protein,” he says.
“showed 100% specificity, but for the wrong Rimm’s own antibody validation process
[histone] peptide,” the authors reported. is similarly comprehensive, having evolved
“We have talked to a lot of researchers over the years. Initially, his team relied
who say this is actually one of the single simply on Westerns. Then they added tissue
biggest problems that they’ve experienced microarrays, cell lines that overexpress the
with reagents in the lab,” says Elizabeth protein, and siRNAs to knock expression
Iorns, cofounder and CEO of ScienceEx- down, increasing the sophistication of their
change. process as they went. Rimm published his
Just ask David Rimm, Director of Trans- validation algorithm in 2010 (3).
lational Pathology at Yale University School “That algorithm was arguably too
of Medicine. Rimm is an anatomic pathol- complex, and the flow chart we put in is
ogist who develops quantitative immuno- too hard to follow,” he concedes. But there’s
fluorescence assays for clinical applications. no one-size-fits-all solution when it comes
In 2009, after his team showed that the to validation, and assessing antibodies
staining patterns of five antibodies (ATF2, R&D Systems’ Alex Kalyuzhny penned an editorial on antibody is important since some commercial
validation in 2009 (J Histochem Cytochem, 57:1099–1101).
p21(WAF1), p16(INK4A), beta-catenin, Credit: Michael Ehlen, R&D Systems. suppliers have exceptional track records,
and fibronectin) could together predict says Bandrowski, and others, not so much.
melanoma survival, Rimm began planning Clearly, the HPA knows what it means Some, for instance, provide only a Western
to prospectively test the biomarker in the to validate an antibody. According to Jochen blot cropped to highlight the band of
clinic. First, though, he needed to prepare Schwenk, Associate Professor at the Royal interest, while others provide unedited,
the assay for CLIA lab testing. Two years Institute of Technology in Stockholm, application-specific data. Henrik Wernérus,
had passed, and the reagent stocks were Sweden and a principal investigator in the chief scientific officer at Atlas Antibodies,
depleted, so he tasked a member of his lab project, the HPA has actually generated which commercializes the HPA project
with purchasing new batches of antibodies more than 50,000 antibody preparations antibodies, says the project tested some
against the five antigens and checking to date. 5000 commercial antibodies before making
to make sure everything was working as HPA antibodies are made in rabbits them themselves. On average, about half the
expected. It wasn’t. from bacterially expressed protein fragments reagents worked, but the success rate varied
At first, Rimm assumed there was a called PrESTs (protein epitope signature wildly among suppliers. “It was basically
“hiccup” in the lab, but subsequent tests tags). When a serum comes back, it is affinity- from 0% to 100% success rate for the
produced the same negative findings, which purified and subjected to a battery of tests, different vendors.”

Vol. 56 | No. 3 | 2014 111 www.BioTechniques.com


Features

but apparently not meticulous enough. In part, researchers may have difficulty
They began to find that their antibodies replicating findings because they cannot
behaved no differently in knockout mice unambiguously determine what reagents
than in wild-type tissues, a sign they were and conditions were used to drive those
not as specific as they should be. Soon, studies. Companies come and go, product
Saper decided he’d had enough. lines change, and catalogues are renum-
In 2005, he published an editorial bered. A researcher may dig into his or her
(3) laying out submission requirements freezer and pull out an antibody from a
for future articles in JCN. Essentially, company that’s long since gone under, says
researchers would need to demonstrate that Bandrowski. In that case, even if they faith-
their antibodies really did work as adver- fully and accurately report the reagent they
tised. In particular, they had to document used in the literature, what is the research
the source of the antibody (researcher or community to do? “Researchers can’t go
company, including catalog, clone, and lot back. They can’t get in the way-back machine
numbers), the immunogen used, the nature and figure out what the catalog of an out-of-
of the preparation (polyclonal/monoclonal business company said when these authors
and species in which it was raised), its speci- actually purchased that antibody. So how
Elizabeth Iorns, cofounder and CEO of ScienceExchange.
Credit: University of Miami. ficity (i.e., that it recognizes a particular do you resolve that?”
band on a Western blot), and any necessary In 2013 Melissa Haendel of Oregon
Ultimately, says Rimm, users must take controls (such as its behavior in a knockout Health & Science University and colleagues
responsibility for validating their own tissue) (see also Reference 4). decided to document the issue by attempting
antibodies. “If you think about it, who’s “It was not a popular stand,” Saper to uniquely identify antibodies, organisms,
responsible for QC on an antibody that you recalls. “Authors were pretty annoyed.” But cell lines, constructs, and knockdown
buy?” he asks, rhetorically. “Is it the company the impact on authors was pretty minimal. reagents from 238 journal articles spanning
or is it the user?” Between 2006 and 2011, Saper says, only five segments of the biological literature (6).
Unfortunately validation is not so “two, three, or four” papers could not be In total, 54% of resources could be uniquely
straightforward, and researchers attempting published as a result of the requirements, identified. Among antibodies, the figure was
to validate their own antibodies should out of “close to 2000” papers total. Particu- only 44%. “It’s absolutely the case that you
anticipate some “muddy waters,” according larly difficult, he says, was the immunogen can’t have scientific reproducibility without
to Alex Kalyuzhny, Scientific Manager requirement; many companies initially knowing what the ingredients of the recipe
for Immunocytochemistry and Elispot refused to provide that information, were, so to speak,” says Haendel.
assays at R&D Systems. “It’s not black and regarding it as a trade secret. Bandrowski and Haendel’s answer
white—it’s grey.” An antibody may fail “If that company goes out of business— to this problem is the Research Identifi-
because it truly does not recognize the target and in the world of biotech companies they cation Initiative, which assigns unique
antigen, and that happens quite frequently. go out of business like fireflies winking out identifiers, like a Social Security number,
Clifford Saper, the James Jackson Putnam in the night—then you have no idea what to each antibody, model organism strain,
Professor at Harvard Medical School and that antibody was made against, and you and software tool used in a research study.
former editor of the Journal of Compar- can’t replicate the experiment any more,” “There is a cultural change that really
ative Neurology (JCN), recalls that when he explains. needs to happen, and that cultural change
he was a young researcher, his lab attempted Saper went to considerable effort to reach is to improve the way that we talk about
to produce antibodies to brain naturetic out to antibody vendors and get them on these things in the literature,” Bandrowski
peptide. Of 29 rabbits injected, only 3 board. In the end, “virtually” every company explains. Funded as part of the Neuroscience
produced an appropriate staining pattern, but one agreed with him. Information Framework, the RII’s Antibody
and those varied wildly in signal. “There’s Although Saper has stepped down as
a huge amount of variation, and it’s quite editor, JCN still maintains the policy. They
common to have antibodies that stain irrel- even have compiled a database of good
evant things or artifacts.” But antibodies antibodies published in the journal as an aide
also may fail because they don’t work in a to prospective authors. The current version
particular assay type, for instance, because (V.13) contains just over 7500 entries. “If
they recognize conformational epitopes that an antibody has been used in JCN, it’s been
are lost when the target is denatured in a vetted,” he explains.
Western blot. They can fail if they recognize
both the specific and nonspecific targets. Or
they may recognize everything, or nothing. Part of a bigger challenge
In short, says Kalyuzhny, “there are many Antibody validation is one facet of a larger
different levels to that definition [of ‘bad’].” problem in science at the moment: data
irreproducibility. In one widely discussed
study, C. Glenn Begley of Amgen and
Journals take a stand Lee Ellis of the University of Texas MD
In the early 2000s during his tenure at Anderson Cancer Center reported that, of
JCN, Saper noticed a growing number of 53 “landmark” studies in hematology and
research articles in his journal were being oncology, only 6 (11%) could be replicated.
retracted due to faulty antibody data. The “Even knowing the limitations of preclinical
researchers had been quite meticulous in research, this was a shocking result,” the An example badge from the ScienceExchange Independent
demonstrating their reagents’ properties, authors wrote. (5) Antibody Validation Initiative.

Vol. 56 | No. 3 | 2014 112 www.BioTechniques.com


Registry (antibodyregistry.org) currently adds; it is also a marketing tool, conveying
features 2.2 million antibodies, some of them to users that their reagent vendors take
tied into the JCN database to link reagents quality control seriously. Furthermore, the
and references. validating labs can work with the vendor (via
The Registry, Haendel says, “is like a Science Exchange) for troubleshooting. In
Genbank for antibodies”—not just because one case, an ELISA kit shipped with a bad
it assigns unique identifiers, but because it dilution buffer that was masking the target
can serve as a aggregator of disparate pieces protein signal. Once the lab figured that out
of information. Ultimately, Haendel says, and reported it back to the vendor, the kit
researchers should be able to use the resource was updated subsequently passed validation
both to identify the best reagents for a given “with flying colors,” Pellenz says.
application and also to correct the literature Antibodies that fail validation are neither
in the event an antibody is published and flagged nor removed from the antibodies-
later found to be less specific than originally online catalog, though the company will
thought. “You can retroactively correct for inform users should they ask, so Pellenz
that in the context of the data,” she says. recommends investing a few minutes to call
Such initiatives should bridge the repro- technical support prior to purchasing any
ducibility gap. Indeed, in January NIH new antibody.
Director Francis Collins and deputy director To date, 42 antibodies and ELISA kits
Lawrence Tabak penned a commentary in actually have been approved, though many
Nature to address the lack of reproducibility more than that have been tested, and web
in life science research. They noted that some site users have flagged an additional 1,200
journals, including Nature, Science, and antibodies they would like to have tested.
Science Translational Medicine, had begun Science Exchange and antibodies-online are
implementing editorial policies to encourage also validating a collection of key reagents
the detailed reporting of experimental details. on their own, “to seed the antibodies-online
(7) Nature’s new “Reporting Checklist for catalog with validated antibodies and ELISA
Life Sciences Articles,” for instance, reads in kits to establish the IV badge,” Pellenz says.
part: “To show that antibodies were profiled According to Pellenz, 8 of antibodies-
for use in the system under study (assay and online’s 10 biggest vendors have expressed
ADVAAANCED species), provide a citation, catalog number
and/or clone number, supplementary infor-
interest in the validation program, as have
about three-quarters of their vendors overall.

IN CANCER
mation or reference to an antibody validation They cannot possibly test every antibody in
profile (e.g., Antibodypedia, 1DegreeBio).” existence—there simply are too many of
them, and in theory, each lot must be tested

RESEARCH
Atlas Antibodies Advanced Polyclonals
Back to the bench
For the researchers at the bench, the funda-
mental question remains: What antibody
anew, a cost-prohibitive proposition. But it’s a
start. And given the importance of antibodies
in life science research, it’s a good one.

are all approved in IHC, with abundant should they choose for their particular assay?
Several resources exist to help, including References
validation data available from both Antibodypedia, Linscott’s Directory, and 1. Egelhofer, T.A., et al. 2011. An assessment of
normal and cancer human tissues. The more. But with millions of reagents listed histone-modification antibody quality. Nat.
advanced antigen design and antigen on some of these pages, selection remains a Struct. Mol. Biol. 18:91-93.
challenge. 2. Lukinavičius, G., et al. 2013. Commercial Cdk1
based affinity purification of the Recently, Iorns launched a new effort antibodies recognize the centrosomal protein
Cep152. Biotechniques 55:111-114.
Triple A Polyclonals provide you to help. In collaboration with antibody
3. Saper, C.B. 2005. An open letter to our readers
with the most specific, reliable and distributor antibodies-online.com, the on the use of antibodies. J. Comp. Neurol.
ScienceExchange Independent Validation 493:477-478.
versatile antibodies available. Initiative pairs vendors with independent 4. Saper, C.B. 2009. A guide to the perplexed
labs that can provide third-party validation of on the specificity of antibodies. J. Histochem.
antibody efficacy for several hundred dollars Cytochem. 57:1-5.
per test. 5. Begley, C.G. and L.M. Ellis. 2012. Raise
The initiative kicked off in July 2013 standards for preclinical cancer research.

Nature 483:531-533.
but has only gained traction in the past few
6. Vasilevsky, N.A., et al. 2013. On the repro-
months, says Iorns. “We’ve done hundreds of ducibility of science: Unique identification of
tests so far, and we’re planning to do 10,000 research resources in the biomedical literature.
this year.” PeerJ. 1:e148.
Atlas Antibodies Advanced Polyclonals. Successfully validated antibodies are 7. Collins, F.S. and L.A. Tabak. 2014. NIH plans
Triple A rated. awarded a green check mark “badge” for the to enhance reproducibility. Nature 505:612-613.
vendor’s web site. “We are trying to establish
this as kind of a sign of quality,” says Stefan Written by Jeffrey M. Perkel, Ph.D.
Pellenz, who runs the validation initiative
at antibodies-online. But companies gain BioTechniques 56:111-114 (March 2013)
more from the process than just a logo, he doi: 10.2144/000114143
Learn more today at atlasantibodies.com
114 www.BioTechniques.com

S-ar putea să vă placă și