8/18/2019 521274 A

1/3

In 2006, things were looking pretty good forDavid Rimm, a pathologist at Yale Universityin New Haven, Connecticut. He had devel-oped a test to guide effective treatment of the

skin cancer melanoma, and it promised to savelives. It relied on antibodies — large, Y-shapedproteins that bind to specified biomolecules andcan be used to flag their presence in a sample.Rimm had found a combination of antibod-ies that, when used to ‘stain’ tumour biopsies,produced a pattern that indicated whetherthe patient would need to take certain harshdrugs to prevent a relapse after surgery. He hadsecured more than US$2 million in funding tomove the test towards the clinic.

But in 2009, everything started to fall apart.When Rimm ordered a fresh set of antibodies,

his team could not reproduce the originalresults. The antibodies were sold by the samecompanies as the original batches, and weresupposed to be identical — but they did notyield the same staining patterns, even on thesame tumours. Rimm was forced to give uphis work on the melanoma antibody set. “Welearned our lesson: we shouldn’t have beendependent on them,” he says. “That was a verysad lab meeting.”

Antibodies are among the most commonlyused tools in the biological sciences — putto work in many experiments to identifyand isolate other molecules. But it is nowclear that they are among the most com-mon causes of problems, too. The batch-to-batch variability that Rimm experienced can

produce dramatically differing results. Evenmore problematic is that antibodies oftenrecognize extra proteins in addition to theones they are sold to detect. This can causeprojects to be abandoned, and waste time,money and samples.

Many think that antibodies are a majordriver of what has been deemed a ‘reproduc-ibility crisis’, a growing realization that theresults of many biomedical experiments can-not be reproduced and that the conclusionsbased on them may be unfounded. Poorlycharacterized antibodies probably contributemore to the problem than any other laboratorytool, says Glenn Begley, chief scientific officerat Tetra Logic Pharmaceuticals in Malvern,Pennsylvania, and author of a controversial

BLAME IT ON THE ANTIBODIESAntibodies are theworkhorses of biologicalexperiments, but they arelittering the field with false findings. A few evangelistsare pushing for change.

B Y M O N YA B A K E R

I L L U S T R A T I O N

B Y N I K S P E N C E R / N A T U R E

2 7 4 | N A T U R E | V O L 5 2 1 | 2 1 M A Y 2 0 1 5

FEATURENEWS

© 2015 Macmillan Publishers Limited. All rights reserved

8/18/2019 521274 A

2/3

analysis1 showing that results in 47 of 53 land-mark cancer research papers could not bereproduced.

A few scientists who have been burned bybad experiences with antibodies have begun tospeak up. Rimm’s disappointment set him on acrusade to educate others by writing reviews,hosting web seminars and raising the problem

in countless conference talks. He and othersare calling for the creation of standards bywhich antibodies should be made, used anddescribed. And some half a dozen grass-rootsefforts have sprung up to provide better waysof assessing antibody quality.

But it is too soon to call the cause a move-ment. “There are all these resources out there,but nobody uses them and many people aren’teven aware of them,” says Len Freedman, whoheads the Global Biological Standards Insti-tute, a non-profit group in Washington DCcommitted to improving biomedical research.“Most vendors have no incentive to change

what’s going on right now, even though a lot ofthe antibody reagents suck.”

BUYER BEWARE

Take the example of Ioannis Prassas, a proteo-mics researcher at Mount Sinai Hospital inToronto, Canada. He and his colleagues hadbeen chasing a protein called CUZD1, whichthey thought could be used to test whethersomeone has pancreatic cancer. They boughta protein-detection kit and wasted two years,$500,000 and thousands of patient samplesbefore they realized that the antibody in thekit was recognizing a different cancer protein,CA125, and did not bind to CUZD1 at all 2. Inretrospect, Prassas says, a rush to get going ona promising hypothesis meant that he and hisgroup had failed to do all the right tests. “Ifsomeone says, ‘Here is an assay you can use,’you are so eager to test it you can forget thatwhat has been promised is not the case.”

Most scientists who purchase antibod-ies believe the label printed on the vial, saysRimm. “As a pathologist, I wasn’t trainedthat you had to validate antibodies; I was justtrained that you ordered them.”

Antibodies are produced by the immunesystems of most vertebrates to target an invadersuch as a bacterium. Since the 1970s, scien-tists have exploited antibodies for research. Ifa researcher injects a protein of interest intoa rabbit, white blood cells known as B cellswill start producing antibodies against theprotein, which can be collected from the ani-mal’s blood. For a more consistent product, theB cells can be retrieved, fused with an ‘immor-talized’ cell and cultured to provide a theoreti-cally unlimited supply.

Three decades ago, scientists who neededantibodies for their experiments had to makethem themselves. But by the late 1990s, rea-gent companies had started to take over thechore.

Today, more than 300 companies sell over

2 million antibodies for research. As of 2011,the market was worth $1.6 billion, according toglobal consultancy Frost & Sullivan.

DEVASTATING EFFECTS

There are signs that problems with antibodiesare having broad and potentially devastatingeffects on the research record. In 2009, one

journal devoted an entire issue to assessing theantibodies that are used to study G-protein-coupled receptors (GPCRs) — cell-signallingproteins that are targeted by drugs to treat vari-ous disorders, from incontinence to schizo-phrenia. In an analysis 3 of 49 commerciallyavailable antibodies that targeted 19 signallingreceptors, most bound to more than one pro-tein, meaning that they could not be trusted todistinguish between the receptors.

The field of epigenetics relies heavily on anti-bodies to identify how proteins that regulategene expression have been modified. In 2011, anevaluation4 of 246 antibodies used in epigenetic

studies found that one-quarter failed tests forspecificity, meaning that they often bound tomore than one target. Four antibodies wereperfectly specific — but to the wrong target.

Scientists often know, anecdotally, that someantibodies in their field are problematic, butit has been difficult to gauge the size of theproblem across biology as a whole. Perhapsthe largest assessment comes from work pub-lished by the Human Protein Atlas, a Swedishconsortium that aims to generate antibodies

for every protein in the human genome. It haslooked at some 20,000 commercial antibod-ies so far and found that less than 50% can beused effectively to look at protein distributionin preserved slices of tissue5. This has led somescientists to claim that up to half of all com-mercially available antibodies are unreliable.

But reliability can depend on the experiment.“Our experience with commercial antibodiesis that they are usually okay in some applica-tions, but they might be terrible in others,” saysMathias Uhlén at the Royal Institute of Tech-nology in Stockholm, who coordinates theHuman Protein Atlas.

Researchers ideally should check that anantibody has been tested for use in particularapplications and tissue types, but the qualityof information supplied by vendors can vary

tremendously. A common complaint from sci-entists is that companies do not provide the datarequired to evaluate a given antibody’s specific-ity or its lot-to-lot variability. Companies mightship a batch of antibodies with characterizationinformation derived from a previous batch.And the data are often derived under ideal con-ditions that do not reflect typical experiments.

Antibody companies contacted for this articlesaid that it is impossible to test their productsacross all experimental conditions, but they doprovide reliable data and work with scientiststo improve antibody quality and performance.

Many academics use Google to find products,so optimizing search results can sometimesmatter more to a company than optimizing theactual reagents, says Tim Bernard, head of thebiotechnology consultancy Pivotal Scientificin Upper Heyford, UK. Christi Bird, a Frost &Sullivan analyst based in Washington DC, saysthat researchers are often more interested inhow quickly reagents can be delivered than in

searching for antibodies with appropriate vali-dation data. “It’s the Amazon effect: they want itin two or three days, with free shipping.”

Researchers who are aware of the antibodyproblem say that scientists need to be more vigilant. “Antibodies are not magic reagents.You can’t just throw them on your sample andexpect the result you get is 100% reliable with-out putting some critical thinking into it,” saysJames Trimmer, head of NeuroMab at the Uni- versity of California, Davis, which makes anti-bodies for neuroscience. Like many suppliers,NeuroMab explicitly states the types of experi-ment that an antibody should be used for, butscientists do not always follow the instructions.

Ideally, researchers would refuse to buy anti-bodies without extensive validation data orwould perform the va lidation themselves(see ‘Bad antibodies’). This is something thatRimm is passionate about: he has developeda multistep flowchart for effective validation 6,which he shares with anyone who will listen.But the process is time consuming — Rimmrecommends control experiments that involveengineering cell lines to both express and stopexpressing the protein of interest, for example.Even he acknowledges that few labs will per-form all the steps.

Some scientists buy half a dozen antibod-ies from different vendors, and then run a fewassays to see which performs best. But theymay end up buying the same antibody fromdifferent places. The largest vendors competeon catalogue size, so they often buy antibodiesfrom smaller suppliers, relabel them and offerthem for sale. Bernard says that the 2 millionantibodies on the market probably represent250,000–500,000 unique ‘core’ antibodies.

By necessity, many researchers rely onword of mouth or the published literature foradvice. But that creates a self-perpetuatingproblem, in which better-performing antibod-ies that become available later are rarely used,says Fridtjof Lund-Johansen, a proteomics

“ANTIBODIESARE NOT MAGIC

REAGENTS.”

2 1 M A Y 2 0 1 5 | V O L 5 2 1 | N A T U R E | 2 7 5

FEATURE NEWS

© 2015 Macmillan Publishers Limited. All rights reserved

8/18/2019 521274 A

3/3

researcher at the University of Oslo. “We have very good antibodies on the market,” he says,“but we don’t know what they are.” Lund-Johansen is trying to change that by developinghigh-throughput assays that could comparethousands of antibodies at once.

TESTING TIMES

In the past decade, various projects have sprungup to try to make information about antibod-ies easier to find. The online reagents portalAntibodypedia (antibodypedia.com), whichis maintained by the Human Protein Atlas,has catalogued more than 1.8 million antibod-ies and rated the validation data available for various experimental techniques. Antibodies-online (antibodies-online.com), another portal,set up a programme two years ago for independ-ent labs to do validation studies, generally atthe vendors’ expense. But out of 275 studies,less than half of the products tested have madethe cut and earned an ‘independent validation’badge. The non-profit Antibody Registry (anti-bodyregistry.org) assigns unique identifiers toantibodies and links them to other resources.Another project, pAbmAbs (pabmabs.com/wordpress), operates in a similar way to thesocial-recommendation web service Yelp, byencouraging people to review antibodies.

But none of these efforts has gained much ofa foothold in the scientific community. Manyof the scientists contacted for this article wereunaware that such resources existed.

The antibody market has grown so crowdedthat a reputation for quality is becoming partof some suppliers’ business plans. “Now thereis so much competition that you have to dif-ferentiate yourself,” says Bernard. Vendors suchas Abcam in Cambridge, UK, are encouragingusers to report their own data and rankings

on the company’s website. Abcam’s analysis ofpurchasing behaviour shows that its customerslook at data pages on average nine times beforebuying, suggesting that customers want moreinformation.

Abgent, an antibody company based in SanDiego, California, and a subsidiary of WuXiAppTec in Shanghai, China, tested all of itsantibodies about a year ago. After reviewingthe results it discarded about one-third of itscatalogue. Whether that was a good decisiondepends on whether customers will be will-ing to spend more for better reagents, saysJohn Mountzouris, site leader at the company.Already, he says, customer complaints haveplummeted.

Some scientists are calling for much moreradical change. In a Comment in Nature inFebruary 7, Andrew Bradbury of Los AlamosNational Laboratory in New Mexico and morethan 100 co-signatories proposed a massiveshift in the way antibodies are produced andsold. They suggested using only antibodies thathave been defined down to the level of the DNAsequence that produces them, and then manu-factured in engineered ‘recombinant’ cells.This would circumvent much of the variabilityintroduced by production in animals. But theproposal demands information about individ-ual antibodies that many companies considerto be trade secrets — and the antibody market-place and its millions of products would haveto be essentially demolished and reconstructed.

Uhlén, a co-signatory on the Comment,regards the plan as a distant hope. He estimatesthat the ‘recombinant antibodies’ that Bradburyhopes for would each cost 10–100 times moreto generate than the conventional sort, andthat they would not necessarily perform bet-ter. “At the end of the day, how the binder works

in the application is more important,” he says.“Having a sequence for sure doesn’t tell you ifit works.” Other efforts are under way to findcheap, fast, reliable ways of making antibod-ies without immunizing animals, for exampleby expressing and optimizing them in viruses.

The pressure to characterize currentlyavailable antibodies is surging. As part ofefforts to improve reproducibility, someresearchers have started to discuss enlistingan independent body to establish a certifica-tion programme for commercial antibodies.And several journals (including Nature ) askauthors to make clear that antibodies used intheir papers have been profiled for that par-ticular application.

The quality will creep, rather than leap,forward, says Trimmer, who hopes to see apositive-feedback loop: as scientists becomeaware of artefacts, they will be more likely tochallenge results and uncover more artefacts.Already, he says, the widespread insoucianceabout antibody validation has started to fade.“It’s turning around a little bit,” he says. “Weneed to keep talking about it.” ■

Monya Baker writes and edits for Nature inSan Francisco, California.

1. Begley, C. G. & Ellis, L. M. Nature 483, 531–533(2012).

2. Prassas, I. & Diamandis, E. P. Clin. Chem. Lab. Med. 52, 765–766 (2014).

3. Michel, M. C., Wieland, T. & Tsujimoto, G. Naunyn-Schmiedeberg’s Arch. Pharmacol. 379, 385–388(2009).

4. Egelhofer, T. A. et al. Nature Struct. Mol. Biol. 18, 91–93 (2011).

5. Berglund, L. et al . Mol. Cell. Proteom. 7, 2019–2027(2008).

6. Bordeaux, J. et al . BioTechniques 48, 197–209(2010).

7. Bradbury, A. & Plückthun, A. Nature 518, 27–29(2015).

Problem: An antibody is supposed to recognize onlyits target protein, but sometimes binds to others,depending on the proteins present in a sample.

Solution: An antibody should be tested for off-targetbinding using positive and negative controls.

BAD ANTIBODIESThe most common problems withantibodies and how to avoid them.

CROSS-REACTIVITY Problem: Separate batches of antibody can performdifferently. This happens most often when theantibody is produced from a new set of animals.

Solution: Researchers should conrm lot numbersand characterization data with vendors.

Problem: Different experiments and experimentalconditions can change a protein's folding andtherefore its binding ability.

Solution: Scientists should check supplier'srecommended applications.

VARIABILITY WRONG APPLICATION

Targetprotein

Non-targetprotein

Antibody

Bindingsite

2 7 6 | N A T U R E | V O L 5 2 1 | 2 1 M A Y 2 0 1 5

FEATURENEWS

© 2015 Macmillan Publishers Limited. All rights reserved