- Email: [email protected]
The antibody horror show: an introductory guide for the perplexed
New BIOTECHNOLOGY xxx (xxxx) xxx–xxx
Contents lists available at ScienceDirect
New BIOTECHNOLOGY journal homepage: www.elsevier.com/locate/nbt
The antibody horror show: an introductory guide for the perplexed Simon. L. Goodman Translational Biomarkers Research, Translational Innovation Platform – Oncology, Merck KGaA, Frankfurterstr. 250, 64293, Darmstadt, Germany
A R T I C L E I N F O
A B S T R A C T
Keywords: Commercial antibodies Validation User-Training Community reporting Reproducibility
The biological literature reverberates with the inadequacies of commercial research-tool antibodies. The scientific community spends some $2 billion per year on such reagents. Excellent accessible scientific platforms exist for reliably making, validating and using antibodies, yet the laboratory end-user reality is somehow depressing – because they often “don’t work”. This experience is due to a bizarre and variegated spectrum of causes including: inadequately identified antibodies; inappropriate user and supplier validation; poor user training; and overloaded publishers. Colourful as this may appear, the outcomes for the community are uniformly grim, including badly damaged scientific careers, wasted public funding, and contaminated literature. As antibodies are amongst the most important of everyday reagents in cell biology and biochemistry, I have tried here to gently suggest a few possible solutions, including: a move towards using recombinant antibodies; obligatory unique identification of antibodies, their immunogens, and their producers; centralized international banking of standard antibodies and their ligands; routine, accessible open-source documentation of user experience with antibodies; and antibody-user certification.
Today’s biological sciences are dependent on reliable antibodies. A typical biology laboratory spends from €5000–10000 a year on commercial research tool antibodies – but many spend 10′s to 100′s-fold more, and use more than a 100 antibodies routinely, and use them on a daily basis [1]. In this article I refer specifically to such commercial research tool antibodies, not to therapeutic or diagnostic antibodies: these are rigorously defined. An estimated $2.5 billion per year was spent on research tool antibodies in 2015 [1]. In a perfect world, we would like to have, off-the-shelf, affinity binders that specifically bind to all possible epitopes, as exposed in all conceivable cell biological and biochemical technologies. But the world is imperfect, so we have instead commercial research-tool antibodies (c-Abs). I recently re-discovered a bit of journalism I wrote long ago complaining about the depressing quality of reagents [2]. From my old lab notebooks, I can now reveal the reagents in question: commercial antibodies. Was this the first “official complaint” on this topic in the literature? Well, it was in a minor non-peer reviewed journal, so I am not surprised if you did not spot it. Still, I claim it as justification for this note. Since 1989, my moan has loudened to a faint but sustained scream: in short, my current status is abject terror. I am scared of c-Abs, of their existential impact on our biological endeavour, and of their effects on the “reproducibility crisis” [3]. And, to borrow from a recent tune, “I know I’m not alone”. For there is a long, long and increasingly long line of excellent and profoundly depressing comments, recommendations and studies on the various deficiencies in the validation, characterization, and in the production, supply and use of c-Abs, and how this might be – no let’s
cut to the chase – is polluting the scientific literature [4–8]. Not to mention a Dark-Web of unreported, unpublished studies classified as “did not work”, or “company confidential”, lying silent in computer files and desk drawers around the world, the harvest of dysfunctional cAbs. Much of this waste is being financed from the public purse. It all sounds like a recipe for disaster. It seems ironic that this situation has developed. For example, given the outstanding scientific quality of the Alpbach symposium 2017 (http://affinityproteomicsalpbach.com/index.php). We well know how to make and to characterize affinity protein binders. So what exactly is the problem with those c-Abs that we seek when we need, for example, to track unfamiliar biochemical pathways, localize unfamiliar target proteins, or quantitate changes in a phosphorylated site in response to an experimental manipulation? And, the target being unfamiliar and there being no colleague who can offer a reagent, we turn to The Catalogues – and – Oh joy! there are so many antibodies to help us here. In fact, Andrew Chalmers, of the antibody cataloguing site CiteAb [9,10], has collected over 3.8 million (3.8 × 106) discrete c-Abs in catalogues and cited in the scientific literature. Only 15 years ago Michaud et al. were elated to find “approximately ten thousand antibodies” available from commercial sources [11]. And even 5 years ago, The Scientist magazine reported an increase of only 5000 c-Abs per year [12]. Yet, in the 15 years since the Michaud et al. article, the increase in c-Abs has slightly beaten Moore’s law (i.e. a doubling every 18 months). Where have they all come from? Perhaps coincidentally, the start of the explosive growth coincided with the “completion” of the human
E-mail address: [email protected]. https://doi.org/10.1016/j.nbt.2018.01.006 Received 26 October 2017; Received in revised form 3 January 2018; Accepted 16 January 2018 1871-6784/ © 2018 Published by Elsevier B.V.
Please cite this article as: Goodman DPhil., Simon.L., New BIOTECHNOLOGY (2018), https://doi.org/10.1016/j.nbt.2018.01.006
New BIOTECHNOLOGY xxx (xxxx) xxx–xxx
S.L. Goodman
you using it correctly?”. This was a perfectly valid response, as a survey initiated by participants at Alpbach and the recent Asilomar-Global Biological Standards Institute (GBSI) meeting (https://www.gbsi.org/ event/antibody-validation-2016/) showed recently: only 43% of early career researchers (< 5y tenure) validated c-Abs, and 31% did not validate, or worse, saw no need to validate c-Abs at all [1]. Mercifully, the perceived and reported need to validate antibodies before using them increased with time in tenure − or perhaps it was only the people who validated their antibodies that got any further in their careers? I do not think that point has been investigated yet. The major reason for not validating antibodies was given as “it takes too much time”. So companies are entirely right to ask whether you are using the antibody correctly. However, that is clearly not the only fly in the ointment, or irrelevant band on the Western blot, as it were, because it does take a lot of effort, time and skill to validate an antibody properly [14,15]. However, I had many antibodies that did work, so obviously I was, in general, using them correctly. What is more, I had come out of a laboratory where I, and everyone else, was routinely hung out to dry if they did not demonstrate clearly that their antibodies were or were not working correctly. So, yes, I was definitely using the antibodies correctly. And, no, they were definitely not working as advertised. What was going on with the c-Abs I just cannot say. Time went by, in fact, over 30 years went by in this state. After all: it was probably just me, wasn’t it? One insect bite after the other, through the years. In fact, things got worse, because I started doing a lot of paraffin immunohistochemistry (IHC), and there the antibody situation was more of a dog-bite than a gnat bite. For example, we screened 28 monoclonal c-Abs to FoxO3a for specific reactivity on paraffin sections, without success, though things have improved recently here. Or take one of my favourites, alphavbeta6 integrin, a well-known membrane protein. To be polite, few of the c-Abs suggested to react in IHC with this target showed a staining pattern related to the distribution of the molecule. But then I happened across the Bradbury and Plückthun (and the many co-signatories) call-to-arms in Nature two years ago [16]. After carefully reading the citations, red lights started flashing. Clearly we need to do something. There appear to be several reasons for the disconnect between what happens when we generate our own affinity binders, and what seems to be the often-unhappy everyday user-experience with c-Abs. These can be listed as (1) the reagents, (2) the targets, (3) the producers, (4) the users, (5) the literature and (6) training. Not to mention filthy lucre: the c-Ab market is around $2 bn in the USA alone [12]. I am not going to reiterate the many excellent in-depth discussions of these topics: I am just going to comment.
genome project in April 2003. Suddenly there were an awful lot of open-reading frames with no known function, in need, as it were, of protein binders – and the c-Abs scuttled in. Pushing four million antibodies is really being spoiled for choice, one would think; that is one antibody-per-protein for some 150 human-proteome-size species. Sounds like a lot. Of course, we often need several antibodies for each protein: antibodies to quantify a level of phosphorylation may not bind in solution, or on paraffin sections, or on a Western blot, etc. Then there are issues of detecting differential splicing, glycosylation, conformers, and so on. Then the antibodies might be directly labelled with many different tags and for cytometry, with diverse fluorochromes, etc. At a generous 10-antibodies per-protein-per-proteome, 4 million antibodies could hypothetically provide proteome-wide binders to cover about 15 human-like species. But after us, mice, rats, Drosophila, dogs, Zebrafish and C.elegans, the give-me-a-proteome-of-binders wish-list begins to thin out substantially. Seven full genomes covered, plus small-change. c-Abs sufficient, in theory, for 5 × 105 antibodies per popular genome. However, popular targets acquire a notable aura of more-than-several c-Abs, while other targets are neglected. For example, currently 6542 antibodies in CiteAb target “EGFR”, 5140 targeting the human receptor, and 10993 target “histone H3” (see Table 1). Even allowing for primary antibodies labelled, for instance, with many different fluorochromes, targeting variously phosphorylated forms etc. this seems to really satisfy any possible market. To cut this short: there seem to be a lot of c-Abs out there, so what is the problem? We read that it is a not unusual situation for a PhD student or post-doc to find that an antibody “does not work” [6,7,13]. In fact, even in those hands, calloused and scarred by decades of cell biological and biochemical toil, many c-Abs “do not work” [4,13]. So why is that, I wonder? For myself, although I have continued to have had routine problems with c-Abs, I really did not give it too much thought. It was a bit like continually being attacked by gnats, irritating but not too serious. Just one more slice off the budget – just another c-Ab from The Catalogues. I was not really interested in antibodies per se, sorry about that, I just wanted working c-Abs. But often they did not work. Let me be more specific about that: using the techniques that I applied, the antibodies often failed to specifically bind something resembling the target they were claimed to react with − they were not fit-for-purpose (F4P). I guess possibly some one in three functioned. Reviewing my colleagues, I find this is a common experience. The antibodies I made and characterized myself, and those my colleagues extensively validated: those did work as predicted. When I got in touch with the commercial entity concerned and told them the reagent they had sold me did not work, I generally heard “Are
1) The reagents. We can expect c-Abs to be clean, selective, specific, intact, and reproducible, can we not [5,8]? Or at least identifiable [8]? Production, supply and, notably, identification sit firmly in the lap of the producer. Or, it seems, in the lap of the gods [6,7]. For sometimes, a critical reagent is not there, having been “disappeared” from the catalogue (“the rabbit died”; “we lost the clone”). For example, three most cited anti-EGFR (and EGFR-P) antibodies out of > 6300 identified on CiteAb are rabbit and goat polyclonals (Santa Cruz biotechnology: sc-03, −03-G and −12351), together cited over 1000 times, but now no longer accessible having been removed from the catalogue [17]. Does that not immediately make those 1000 publications irreproducible? And what of the remaining 2.2 million (2.2 × 106) rabbit polyclonal antibodies indexed by CiteAb (vs. 0.9 million monoclonals)?
Table 1 Some hyper-popular commercial tool-antibody targets. Antibody Search Terma
Antibodies foundb
Mab
Pab
Actin p53 Histone H3 Tubulin RAC EGFR Estrogen Receptor Fibronectin TGF-beta 1 cdk1 VEGF
18298 16662 10993 9335 8281 6542 6398 5715 5461 3989 3341
4789 4558 2423 3491 1960 2330 1484 2665 1152 991 1119
13275 11705 8406 5701 6155 3947 4788 2948 4249 2906 2175
I find even this single example is enough: Bradbury, Plückthun et al. are clearly right − the sooner we migrate towards completely transparent recombinant reagents, the better [16,18,19]. True enough, recombinant antibodies are currently costly to make. And the road will indeed be long – only ∼5500 recombinant antibodies are as yet indexed by CiteAb. But clearly it was both costly and long to develop
Individual antibodies recovered from the CiteAb [10] internet site, using the Antibody Search Term shown, at mid-October 2017. Total antibodies indexed and their distribution by monoclonal (Mab) and polyclonal (Pab) reagents is shown. Note: a) the Antibody Search Term recovered several different targets. So “actin” recovers alpha-, beta- and gamma-actin, for example. b) individual antibodies may be present in many forms, e.g. labelled with multiple fluorochromes: each form is counted.
2
New BIOTECHNOLOGY xxx (xxxx) xxx–xxx
S.L. Goodman
the > 6000 predominantly non-recombinant anti-EGFR c-Abs – not to mention wasteful of resources. So I buy neither the financial argument – nor for that matter polyclonal antibodies. Best would be an international public resource that validated and banked recombinant reference antibodies, and there have been several efforts in this direction [[20,21,22]e.g. 20,21,22]. But these are small in comparison to the magnitude of the problem we face. It is in everyone’s interest to set up an international standards organization dedicated to the validation, banking and distribution of recombinant affinity protein-binders.
their sold-on reagents, but even of original producers to users. This happens because OEM agreements, common in the antibody supplier industry, reportedly often restrict suppliers from revealing the original c-Ab manufacturers [8]. Third, transparent identifiability of validation procedure: one can find the same c-Abs in several catalogues, with luck even under the same identification number, code or clone name, all showing an identical image representing “a validation assay”, often a Western blot. And, it is the same blot, down to the background dots and spots. In short, it appears that often OEM suppliers may not re-validate the antibody they re-sell, to ensure it works as originally described by the validator. Best would be that the OEM supplier states who validated the specific batch of antibody supplied, in-house or at a CRO, and show that data. Bottom-feeders would be deterred if producers were rigorously to implement unique reagent identifiers, such as the RRID [31,32]. As an absolute minimum, in the interests of rigor and transparency, that: (1) the original producers of an affinity reagent must be identified unambiguously by OEM suppliers; (2) the original antibody identifiers must be used consistently by OEMs and users; (3) the concentration of functional antibody in the vial is stated and guaranteed; (4) the suppliers themselves validate (or organize a validation of) the batch of reagent that they are supplying; and (5) that this validation data, including working concentration of antibody, and the protocols used are supplied.
1) The targets. The creation of new c-Abs (2003–2017) has bettered Moore’s law. That is an interesting achievement. Still: there are many targets around in many experimental contexts, so we potentially need a lot of F4P c-Abs. The state of the immunogen and the screening technology tend to constrain the c-Abs which can then be recovered [5,22,23]. So in addition to establishing a c-Ab reference bank mentioned in point (1), reference banks of correctly folded protein targets, or target domains – as opposed to linear peptides – are crucial. This is an approach that the Structural Genomics Consortium (SGC), amongst others, have followed [24], and which the commercial HuProt™ array [25,26] approaches mainly for nonmembrane human proteins. At present equivalent banks of IHCprocessed and paraffin embedded targets are not available. It would be experimentally trivial and a valuable resource to produce both over-expressing and knock-out cell lines covering the human proteome, for use in IHC or biochemical screening. You get what you screen for: screen using Western blots and you find antibodies reacting with partially denatured material exposed on WBs; screen on viable cells and you will find antibodies reacting with the native proteins on viable cells [27,28], and so on. 2) The producers. The market for c-Abs is pushing $2 billion per year. Producers range from outstanding ethical companies to the semimythological – from good businesses to bottom-feeders. And for 50 μl of clear liquid at ∼$ 250 a shot, who can grumble? Well, me for a start [2,13]. It seems that some producers care less about the quality of their products than others [13,17,29,30]. There is clearly money to be made by even the most ethically-defective commercial entities on the c-Ab market place. The solution here, I think, is transparency.
1) The users. Despite these complaints, many antibody providers supply excellent characterized and validated c-Abs and the defined protocols under which the c-Abs were validated. However, they cannot be expected to validate under the precise experimental conditions in place in every user-laboratory. So it is obvious that users must validate and characterize c-Abs as F4P in the experimental context they are using. This is a complex job [4,8,32], and its implementation and interpretation should be a fundamental part of the training of all research biologists [1]. At the very least, user-validation must include titering c-Abs against appropriate positive and negative control samples – to discover the concentration giving optimal signal-to-noise. The validation should include cells expressing “normal” levels of the target, not just induced over-expression. For imaging applications, especially of human tissues, cell and tissue arrays are valuable tools for judging specificity – and quantitative image analysis of the data sets is valuable [15,33,34]. In all cases, a comparison with the literature should be made – despite this being potentially constructed from incorrectly validated antibodies in the first place! [14,35] Multiple validation techniques, including antibody-independent techniques if available, help to increase confidence in data subsequently generated using the c-Ab [8,22,36]. Beyond the issues of F4P lie deeper questions of absolute specificity and selectivity: and we can get target-independent indications of antibody binding characteristics from IP/MS, peptide, protein or tissue arrays [37–39], that can give a broad fingerprint of antibody specificity, rather than simple anti-target information. Native protein array chips can be used to scan the non-membrane human proteome for off-target reactivity [26]. Potentially, if the coupling chemistry is demonstrably sufficiently robust, such chips could also be denatured or otherwise pretreated to match a particular experimental context, e.g. processed as for a paraffin-embedded IHC section, to investigate off-target reactivity in particular contexts. Immunomicroscopy on broad panels of fixed tissue or cells in arrays provides both cellular and subcellular distributions, which together with known or predicted biology of the target can also support the validity of the antibody [4,15]. Few producers can characterize antibodies to such depths: it is simply not economically viable for every-day c-Abs. Irrespective of whether a validation effort has succeeded or failed, it is critical to inform the supplier and the community of the result, else a growing crowd of out-of-pocket and out-of-time users will continue to
First, transparent identifiability of reagents. Producers must uniquely identify both the immunogen, the reagent itself, and who, if not the supplier, manufactured and validated it. Best would clearly be general use of a registration system like the Research Reagent Identifier (RRID) [31] to unequivocally label each c-Ab. Verifiable identity of reagents (via a cDNA sequence) and reproducibility can easily be guaranteed using recombinant binders [18], despite excellent polyclonal antibodies [8] from entirely respectable quality producers [23]. But given the exponentially increasing numbers of c-Abs scuttling around, reproducibility and identifiability will become ever more important, and there, I think we all agree, polyclonals have issues. I see no conflict between publishing the antibody sequence and making money, though I know some disagree, and think that “trolls” may milk the honest producer. If this is a genuine issue, and I have my doubts, then I suggest that the sequences of validated but not patented recombinant reagents be deposited in a secure database, for release or comparison only to licensed parties – such a mechanism cannot be beyond the wit of man. Second, transparent identifiability of producers. c-Ab suppliers are often marketed and distributed under Other External Manufacturers (OEM) agreements. The suppliers may arbitrarily rename the reagent originally produced and supplied. Perhaps that is why there are, for example, over 6000 anti-EGFR antibodies, on offer [10] – because some are covertly “synonymized”? So, searching for an active c-Ab, one may inadvertently buy the same (e.g. inactive) reagent several times. And OEM suppliers may be so shy, they fail to guarantee identity not only of 3
New BIOTECHNOLOGY xxx (xxxx) xxx–xxx
S.L. Goodman
Conclusion
accumulate. Reliable suppliers benefit from your data, by having their reagents verified, or not, as the case may be, know that their work flow is functioning, plus of course, they can be expected to promptly replace defective c-Abs. Perhaps if you were to include a bill for your costs in de-validating the reagents, that might encourage response? It typically cost we the users at least over ten times the original antibody price to invalidate it. The community cannot await a publication that only mentions positive results. A very good move in this direction is the independent antibody reporting website pAbmAbs (http://pabmabs. com/wordpress/). Producer-linked reporting seems highly gameable and suspect, despite well-intended and convincing efforts by many reputable suppliers. Independent reporting sites could solve this issue and in many ways the CiteAb approach, to “count the winners” as it were, by noting publications using a given antibody, does this. However, that is some distance from the actual validation, and, as we have seen has its own problems []. It would help if the user-community, in coordination with leading producers, agreed a standard set of test protocols by consensus, for example defined on Protocols.io. In concept to follow the “five pillars” [32], which I note is two less than those classically suggested [(Prov. 9:1) [40] – potentially the addition of protein arrays [26] and peptide arrays [38] would be wiser? The producer would document and guarantee c-Ab specificity and selectivity using at least one of the standard protocols applied to a set of robust cell standards (e.g. ATCC cell lines [41]) – leaving F4P definition to be documented rigorously by the user.
In summary, many c-Abs are poorly documented both by the producers and the users − in particular they are poorly validated as fit-forpurpose in the experiments in which they are being used. This has caused and is causing massive intellectual and financial damage, reckoned in billions of dollars per year, and includes a progressive degradation of the literature in the biological sciences, resulting in a kind of c-Ab-induced Alzheimer’s – we won’t know what we don’t know. The current situation recalls the pre-millennial Monty Python sketch involving a malicious Hungarian-to-English phrase book [42] – if translation does not convert between languages with fealty, you will find yourselves in trouble! If c-Ab-binding does not unequivocally and unambiguously report the target in context, we are in serious trouble. In conclusion, it seems impossible that reagents that form a foundation of modern biology are in such a friable state. In the end, it is in the deepest interests of the user community and publishers, that only data obtained from antibodies appropriately defined and used as fit-forpurpose be accepted into the scientific literature, or we will all find ourselves rushing for a singularity − where the amount of information being destroyed by putative specific antibodies swamps the data from reliable c-Abs. Contributors SLG conceived and wrote the text. Dr. A.Chalmers (CiteAb; Bath) kindly agreed to let me quote direct results from the CiteAb antibody search engine. Dr. M. Taussig (Babraham Institute, Cambridge, UK) kindly reviewed a pre-submission version of this text.
1) The literature. Major publishers recognize the vital role they must play – and many are fighting to do so. Unequivocal identification and validation of the antibodies used in a given publication must be routinely documented, including the RRID if it exists, the unique identifying name, the batch, the lot and the supplier. The actual concentration of antibody used in a given validation must be unequivocally stated. These are hardly novel suggestions – fortunately – and publishers are increasingly implementing them in their guides to authors. But despite some highly determined efforts in convoluted applications like IHC [4], the level of detailing of assays in the biological literature remains generally minimalist in comparison to that in other areas, i.e. the chemical literature. This has led to the horror of the “castrated Western blots”, showing only “the band of interest”, that surely have no place in respectable publications and especially not in those purportedly characterizing c-Abs. Given the availability of Supplementary information on-line, there is no longer any excuse for this type of thing.
Funding This research did not receive any specific grant from funding agencies in the public, commercial or not-for-profit sectors. Conflicts of interests None. Acknowledgements I would like to thank Professor Andreas Plückthun who responded positively to my terrified notes, and invited me to join the battle, and to the many other members of the antibody community who have listened to me reiterating what they know all-to-well, and who passed on encouragement, especially Andrew Chalmers, Anita Bandrowski, Giovanna Roncador, Fridtjof Lund-Johansen, Monya Baker, Natalie de Souza and Mike Taussig. I thank my mentors Walter Birchmeier and Klaus von der Mark, and my one-time colleagues Ludwig Trejdosiewicz and Claudia Wilm for teaching me to be extremely cautious with antibodies – but blame only myself for not having always heeded their warnings.
“Those who follow the herd must get a back view”: indeed, this may also be a literature artefact driven by c-Abs whose specificity differs entirely from that claimed by the supplier. In clear-text: poorly validated c-Abs can lead to the construction of a literature-consensus on a well-studied topic, which can be entirely deviant from the reality [14]. And despite continual vigilance, unexpected commercial incompetence can sometimes get in the way. So not only does one have to independently validate the antibodies described in the literature, but one has to validate the standards one is using to do the validation! [35] If you wish to enter into a binding relationship with a devilish molecule, kindly use a long, or at least a well validated, spoon.
References [1] Freedman LP, Gibson MC, Bradbury AR, Buchberg AM, Davis D, Dolled-Filhart MP, et al. [Letter to the Editor] The need for improved education and training in research antibody usage and validation practices. Biotechniques 2016;61:16–8. [2] Goodman SL. Testing times for biologists. New Scientist 1989;1658:60–1. [3] Begley CG, Ellis LM. Drug development: raise standards for preclinical cancer research. Nature 2012;483:531–3. [4] Saper CB. An open letter to our readers on the use of antibodies. J Comp Neurol 2005;493:477–8. [5] Weller MG. Quality issues of research antibodies. Anal Chem Insights 2016;11:21–7. [6] Baker M. Antibody anarchy: a call to order. Nature 2015;527:545–51. [7] Baker M. Reproducibility crisis: blame it on the antibodies. Nature 2015;521:274–6. [8] Voskuil JLA. Commercial antibodies and their validation. F1000Res 2014;3:232. [9] Helsby MA, Leader PM, Fenn JR, Gulsen T, Bryant C, Doughton G, et al. CiteAb: a searchable antibody database that ranks antibodies by the number of times they
1) Training. There is a pressing need to routinely educate users, and most especially young users, in the correct use of antibodies, as Les Freedman and colleagues at the GBSI noted following a survey last year [1]. It should be a fundament of any modern biology training programme that students be taught how to correctly characterize, validate and use antibodies. In particular, the limitations of the various validation techniques should be emphasized. I would go further and say that a formal certification of graduate students as antibody users would be highly desirable − before they are let loose on the community, and permitted to use c-Abs. 4
New BIOTECHNOLOGY xxx (xxxx) xxx–xxx
S.L. Goodman
have been cited. BMC Cell Biol 2014;15:6. [10] Chalmers A, CiteAb. https://wwwciteabcom/. Bristol, U.K. 2018. p. Company Website. [11] Michaud GA, Salcius M, Zhou F, Bangham R, Bonin J, Guo H, et al. Analyzing antibody specificity with whole proteome microarrays. Nat Biotechnol 2003;21:1509–12. [12] Bird C, Antibodies User Survey, The Scientist: https://www.the-scientist.com/? articles.view/articleNo/32042/title/Antibodies-User-Survey/; 2012. [13] Couchman JR. Commercial antibodies: the good, bad, and really ugly. J Histochem Cytochem 2009;57:7–8. [14] Andersson S, Sundberg M, Pristovsek N, Ibrahim A, Jonsson P, Katona B, et al. Insufficient antibody validation challenges oestrogen receptor beta research. Nat Commun 2017;8(15840):1–11. [15] Goodman SL, Grote HJ, Wilm C. Matched rabbit monoclonal antibodies against αvseries integrins reveal a novel αvβ3-LIBS epitope, and permit routine staining of archival paraffin samples of human tumors. Biol Open 2012;1:329–40. [16] Bradbury AR, Pluckthun A. Getting to reproducible antibodies: the rationale for sequenced recombinant characterized reagents. Protein Eng Des Sel 2015;28:303–5. [17] Lowe D. Trouble at Santa Cruz Biotechnology. Sci Translat Med 2016;23(May) http://blogs.sciencemag.org/pipeline/archives/2016/05/23/trouble-at-santa-cruzbiotechnology2016. [18] Bradbury AM, Pluckthun A. Antibodies: validate recombinants once. Nature 2015;520:295. [19] Bradbury AM, Pluckthun A. Standardized antibodies used in research. Nature 2015;518:27. [20] Springer TA. Institute for Protein Innovation. Cambridge, MA, USA: IPI; 2017. Home page of Institute https://proteininnovationorg/antibody-initiative/. [21] Colwill K, Renewable Protein Binder Working G, Graslund S. A roadmap to generate renewable protein binders to the human proteome. Nat Methods 2011;8:551–8. [22] Roncador G, Engel P, Maestre L, Anderson AP, Cordell JL, Cragg MS, et al. The European antibody network's practical guide to finding and validating suitable antibodies for research. MAbs 2016;8:27–36. [23] Ascoli CA. Antibody basics for the production of functional antibodies. 1 ed. Limerick, PA: Rockland Inc.; 2016. p. Company Website https://rockland-inccom/ uploadedFiles/Support/Resource_Library/Antibody%20Basicspdf. [24] Zhong N, Loppnau P, Seitova A, Ravichandran M, Fenner M, Jain H, et al. Optimizing production of antigens and fabs in the context of generating recombinant antibodies to human proteins. PLoS One 2015;10:e0139695. [25] Cambridge Protein Arrays Ltd. Cambridge, UK: Cambridge Protein Arrays Ltd.; 2017. p. https://cambridgeproteinarrays.com. [26] Jeong JS, Jiang L, Albino E, Marrero J, Rho HS, et al. Rapid identification of
[27]
[28] [29] [30] [31]
[32] [33]
[34]
[35]
[36]
[37]
[38] [39] [40] [41] [42]
5
monospecific monoclonal antibodies using a human proteome microarray. Mol Cell Proteomics 2012;11. O111.016253. Imhof BA, Vollmers HP, Goodman SL, Birchmeier W. Cell-cell interaction and polarity of epithelial cells: specific perturbation using a monoclonal antibody. Cell 1983;35:667–75. Knudsen KA, Rao PE, Damsky CH, Buck CA. Membrane glycoproteins involved in cell-substratum adhesion. Proc Natl Acad Sci U S A 1981;78:6071–5. Cyranoski D. The secret war against counterfeit science. Nature 2017;545:148–50. Reardon S. US government issues historic $3.5-million fine over animal welfare. Nat News 2016:2016. Bandrowski AE, Martone ME. RRIDs: a simple step toward improving reproducibility through rigor and transparency of experimental methods. Neuron 2016;90:434–6. Uhlen M, Bandrowski A, Carr S, Edwards A, Ellenberg J, Lundberg E, et al. A proposal for validation of antibodies. Nat Methods 2016;13:823–7. Toki MI, Cecchi F, Hembrough T, Syrigos KN, Rimm DL. Proof of the quantitative potential of immunofluorescence by mass spectrometry. Lab Invest 2017;97:329–34. Anagnostou VK, Lowery FJ, Syrigos KN, Cagle PT, Rimm DL. Quantitative evaluation of protein expression as a function of tissue microarray core diameter: is a large (1.5 mm) core better than a small (0.6 mm) core? Arch Pathol Lab Med 2010;134:613–9. Quarmby V, Quan C, Ling V, Compton P, Canova-Davis E. How much insulin-like growth factor I (IGF-I) circulates? Impact of standardization on IGF-I assay accuracy. J Clin Endocrinol Metab 1998;83:1211–6. Skogs M, Stadler C, Schutten R, Hjelmare M, Gnann C, Bjork L, et al. Antibody validation in bioimaging applications based on endogenous expression of tagged proteins. J Proteome Res 2017;16:147–55. Lund-Johansen F, de la Rosa Carrillo D, Mehta A, Sikorski K, Inngjerdingen M, Kalina T, et al. MetaMass, a tool for meta-analysis of subcellular proteomics data. Nat Methods 2016;13:837–40. Breitling F, Nesterov A, Stadler V, Felgenhauer T, Bischoff FR. High-density peptide arrays. Mol Biosyst 2009;5:224–34. Richer J, Johnston SA, Stafford P. Epitope identification from fixed-complexity random-sequence peptide microarrays. Mol Cell Proteomics 2015;14:136–47. New Oxford annotated Bible: New Revised Standard Version with the Apocrypha. Oxford University Press; 2010. ATCC. https://www.atcc.org/?geo_country=uk. 2017. Chapman GC, Idle John, Jones Eric, Palin Terry, Gilliam Michael, Terry. Hungarian phrase book. In: MacNaughton I, editor. Monty Python's Flying Circus. London, U.K: BBC; 1970https://wwwyoutubecom/watch?v=vAQJHHf3i1o.
Contents lists available at ScienceDirect
New BIOTECHNOLOGY journal homepage: www.elsevier.com/locate/nbt
The antibody horror show: an introductory guide for the perplexed Simon. L. Goodman Translational Biomarkers Research, Translational Innovation Platform – Oncology, Merck KGaA, Frankfurterstr. 250, 64293, Darmstadt, Germany
A R T I C L E I N F O
A B S T R A C T
Keywords: Commercial antibodies Validation User-Training Community reporting Reproducibility
The biological literature reverberates with the inadequacies of commercial research-tool antibodies. The scientific community spends some $2 billion per year on such reagents. Excellent accessible scientific platforms exist for reliably making, validating and using antibodies, yet the laboratory end-user reality is somehow depressing – because they often “don’t work”. This experience is due to a bizarre and variegated spectrum of causes including: inadequately identified antibodies; inappropriate user and supplier validation; poor user training; and overloaded publishers. Colourful as this may appear, the outcomes for the community are uniformly grim, including badly damaged scientific careers, wasted public funding, and contaminated literature. As antibodies are amongst the most important of everyday reagents in cell biology and biochemistry, I have tried here to gently suggest a few possible solutions, including: a move towards using recombinant antibodies; obligatory unique identification of antibodies, their immunogens, and their producers; centralized international banking of standard antibodies and their ligands; routine, accessible open-source documentation of user experience with antibodies; and antibody-user certification.
Today’s biological sciences are dependent on reliable antibodies. A typical biology laboratory spends from €5000–10000 a year on commercial research tool antibodies – but many spend 10′s to 100′s-fold more, and use more than a 100 antibodies routinely, and use them on a daily basis [1]. In this article I refer specifically to such commercial research tool antibodies, not to therapeutic or diagnostic antibodies: these are rigorously defined. An estimated $2.5 billion per year was spent on research tool antibodies in 2015 [1]. In a perfect world, we would like to have, off-the-shelf, affinity binders that specifically bind to all possible epitopes, as exposed in all conceivable cell biological and biochemical technologies. But the world is imperfect, so we have instead commercial research-tool antibodies (c-Abs). I recently re-discovered a bit of journalism I wrote long ago complaining about the depressing quality of reagents [2]. From my old lab notebooks, I can now reveal the reagents in question: commercial antibodies. Was this the first “official complaint” on this topic in the literature? Well, it was in a minor non-peer reviewed journal, so I am not surprised if you did not spot it. Still, I claim it as justification for this note. Since 1989, my moan has loudened to a faint but sustained scream: in short, my current status is abject terror. I am scared of c-Abs, of their existential impact on our biological endeavour, and of their effects on the “reproducibility crisis” [3]. And, to borrow from a recent tune, “I know I’m not alone”. For there is a long, long and increasingly long line of excellent and profoundly depressing comments, recommendations and studies on the various deficiencies in the validation, characterization, and in the production, supply and use of c-Abs, and how this might be – no let’s
cut to the chase – is polluting the scientific literature [4–8]. Not to mention a Dark-Web of unreported, unpublished studies classified as “did not work”, or “company confidential”, lying silent in computer files and desk drawers around the world, the harvest of dysfunctional cAbs. Much of this waste is being financed from the public purse. It all sounds like a recipe for disaster. It seems ironic that this situation has developed. For example, given the outstanding scientific quality of the Alpbach symposium 2017 (http://affinityproteomicsalpbach.com/index.php). We well know how to make and to characterize affinity protein binders. So what exactly is the problem with those c-Abs that we seek when we need, for example, to track unfamiliar biochemical pathways, localize unfamiliar target proteins, or quantitate changes in a phosphorylated site in response to an experimental manipulation? And, the target being unfamiliar and there being no colleague who can offer a reagent, we turn to The Catalogues – and – Oh joy! there are so many antibodies to help us here. In fact, Andrew Chalmers, of the antibody cataloguing site CiteAb [9,10], has collected over 3.8 million (3.8 × 106) discrete c-Abs in catalogues and cited in the scientific literature. Only 15 years ago Michaud et al. were elated to find “approximately ten thousand antibodies” available from commercial sources [11]. And even 5 years ago, The Scientist magazine reported an increase of only 5000 c-Abs per year [12]. Yet, in the 15 years since the Michaud et al. article, the increase in c-Abs has slightly beaten Moore’s law (i.e. a doubling every 18 months). Where have they all come from? Perhaps coincidentally, the start of the explosive growth coincided with the “completion” of the human
E-mail address: [email protected]. https://doi.org/10.1016/j.nbt.2018.01.006 Received 26 October 2017; Received in revised form 3 January 2018; Accepted 16 January 2018 1871-6784/ © 2018 Published by Elsevier B.V.
Please cite this article as: Goodman DPhil., Simon.L., New BIOTECHNOLOGY (2018), https://doi.org/10.1016/j.nbt.2018.01.006
New BIOTECHNOLOGY xxx (xxxx) xxx–xxx
S.L. Goodman
you using it correctly?”. This was a perfectly valid response, as a survey initiated by participants at Alpbach and the recent Asilomar-Global Biological Standards Institute (GBSI) meeting (https://www.gbsi.org/ event/antibody-validation-2016/) showed recently: only 43% of early career researchers (< 5y tenure) validated c-Abs, and 31% did not validate, or worse, saw no need to validate c-Abs at all [1]. Mercifully, the perceived and reported need to validate antibodies before using them increased with time in tenure − or perhaps it was only the people who validated their antibodies that got any further in their careers? I do not think that point has been investigated yet. The major reason for not validating antibodies was given as “it takes too much time”. So companies are entirely right to ask whether you are using the antibody correctly. However, that is clearly not the only fly in the ointment, or irrelevant band on the Western blot, as it were, because it does take a lot of effort, time and skill to validate an antibody properly [14,15]. However, I had many antibodies that did work, so obviously I was, in general, using them correctly. What is more, I had come out of a laboratory where I, and everyone else, was routinely hung out to dry if they did not demonstrate clearly that their antibodies were or were not working correctly. So, yes, I was definitely using the antibodies correctly. And, no, they were definitely not working as advertised. What was going on with the c-Abs I just cannot say. Time went by, in fact, over 30 years went by in this state. After all: it was probably just me, wasn’t it? One insect bite after the other, through the years. In fact, things got worse, because I started doing a lot of paraffin immunohistochemistry (IHC), and there the antibody situation was more of a dog-bite than a gnat bite. For example, we screened 28 monoclonal c-Abs to FoxO3a for specific reactivity on paraffin sections, without success, though things have improved recently here. Or take one of my favourites, alphavbeta6 integrin, a well-known membrane protein. To be polite, few of the c-Abs suggested to react in IHC with this target showed a staining pattern related to the distribution of the molecule. But then I happened across the Bradbury and Plückthun (and the many co-signatories) call-to-arms in Nature two years ago [16]. After carefully reading the citations, red lights started flashing. Clearly we need to do something. There appear to be several reasons for the disconnect between what happens when we generate our own affinity binders, and what seems to be the often-unhappy everyday user-experience with c-Abs. These can be listed as (1) the reagents, (2) the targets, (3) the producers, (4) the users, (5) the literature and (6) training. Not to mention filthy lucre: the c-Ab market is around $2 bn in the USA alone [12]. I am not going to reiterate the many excellent in-depth discussions of these topics: I am just going to comment.
genome project in April 2003. Suddenly there were an awful lot of open-reading frames with no known function, in need, as it were, of protein binders – and the c-Abs scuttled in. Pushing four million antibodies is really being spoiled for choice, one would think; that is one antibody-per-protein for some 150 human-proteome-size species. Sounds like a lot. Of course, we often need several antibodies for each protein: antibodies to quantify a level of phosphorylation may not bind in solution, or on paraffin sections, or on a Western blot, etc. Then there are issues of detecting differential splicing, glycosylation, conformers, and so on. Then the antibodies might be directly labelled with many different tags and for cytometry, with diverse fluorochromes, etc. At a generous 10-antibodies per-protein-per-proteome, 4 million antibodies could hypothetically provide proteome-wide binders to cover about 15 human-like species. But after us, mice, rats, Drosophila, dogs, Zebrafish and C.elegans, the give-me-a-proteome-of-binders wish-list begins to thin out substantially. Seven full genomes covered, plus small-change. c-Abs sufficient, in theory, for 5 × 105 antibodies per popular genome. However, popular targets acquire a notable aura of more-than-several c-Abs, while other targets are neglected. For example, currently 6542 antibodies in CiteAb target “EGFR”, 5140 targeting the human receptor, and 10993 target “histone H3” (see Table 1). Even allowing for primary antibodies labelled, for instance, with many different fluorochromes, targeting variously phosphorylated forms etc. this seems to really satisfy any possible market. To cut this short: there seem to be a lot of c-Abs out there, so what is the problem? We read that it is a not unusual situation for a PhD student or post-doc to find that an antibody “does not work” [6,7,13]. In fact, even in those hands, calloused and scarred by decades of cell biological and biochemical toil, many c-Abs “do not work” [4,13]. So why is that, I wonder? For myself, although I have continued to have had routine problems with c-Abs, I really did not give it too much thought. It was a bit like continually being attacked by gnats, irritating but not too serious. Just one more slice off the budget – just another c-Ab from The Catalogues. I was not really interested in antibodies per se, sorry about that, I just wanted working c-Abs. But often they did not work. Let me be more specific about that: using the techniques that I applied, the antibodies often failed to specifically bind something resembling the target they were claimed to react with − they were not fit-for-purpose (F4P). I guess possibly some one in three functioned. Reviewing my colleagues, I find this is a common experience. The antibodies I made and characterized myself, and those my colleagues extensively validated: those did work as predicted. When I got in touch with the commercial entity concerned and told them the reagent they had sold me did not work, I generally heard “Are
1) The reagents. We can expect c-Abs to be clean, selective, specific, intact, and reproducible, can we not [5,8]? Or at least identifiable [8]? Production, supply and, notably, identification sit firmly in the lap of the producer. Or, it seems, in the lap of the gods [6,7]. For sometimes, a critical reagent is not there, having been “disappeared” from the catalogue (“the rabbit died”; “we lost the clone”). For example, three most cited anti-EGFR (and EGFR-P) antibodies out of > 6300 identified on CiteAb are rabbit and goat polyclonals (Santa Cruz biotechnology: sc-03, −03-G and −12351), together cited over 1000 times, but now no longer accessible having been removed from the catalogue [17]. Does that not immediately make those 1000 publications irreproducible? And what of the remaining 2.2 million (2.2 × 106) rabbit polyclonal antibodies indexed by CiteAb (vs. 0.9 million monoclonals)?
Table 1 Some hyper-popular commercial tool-antibody targets. Antibody Search Terma
Antibodies foundb
Mab
Pab
Actin p53 Histone H3 Tubulin RAC EGFR Estrogen Receptor Fibronectin TGF-beta 1 cdk1 VEGF
18298 16662 10993 9335 8281 6542 6398 5715 5461 3989 3341
4789 4558 2423 3491 1960 2330 1484 2665 1152 991 1119
13275 11705 8406 5701 6155 3947 4788 2948 4249 2906 2175
I find even this single example is enough: Bradbury, Plückthun et al. are clearly right − the sooner we migrate towards completely transparent recombinant reagents, the better [16,18,19]. True enough, recombinant antibodies are currently costly to make. And the road will indeed be long – only ∼5500 recombinant antibodies are as yet indexed by CiteAb. But clearly it was both costly and long to develop
Individual antibodies recovered from the CiteAb [10] internet site, using the Antibody Search Term shown, at mid-October 2017. Total antibodies indexed and their distribution by monoclonal (Mab) and polyclonal (Pab) reagents is shown. Note: a) the Antibody Search Term recovered several different targets. So “actin” recovers alpha-, beta- and gamma-actin, for example. b) individual antibodies may be present in many forms, e.g. labelled with multiple fluorochromes: each form is counted.
2
New BIOTECHNOLOGY xxx (xxxx) xxx–xxx
S.L. Goodman
the > 6000 predominantly non-recombinant anti-EGFR c-Abs – not to mention wasteful of resources. So I buy neither the financial argument – nor for that matter polyclonal antibodies. Best would be an international public resource that validated and banked recombinant reference antibodies, and there have been several efforts in this direction [[20,21,22]e.g. 20,21,22]. But these are small in comparison to the magnitude of the problem we face. It is in everyone’s interest to set up an international standards organization dedicated to the validation, banking and distribution of recombinant affinity protein-binders.
their sold-on reagents, but even of original producers to users. This happens because OEM agreements, common in the antibody supplier industry, reportedly often restrict suppliers from revealing the original c-Ab manufacturers [8]. Third, transparent identifiability of validation procedure: one can find the same c-Abs in several catalogues, with luck even under the same identification number, code or clone name, all showing an identical image representing “a validation assay”, often a Western blot. And, it is the same blot, down to the background dots and spots. In short, it appears that often OEM suppliers may not re-validate the antibody they re-sell, to ensure it works as originally described by the validator. Best would be that the OEM supplier states who validated the specific batch of antibody supplied, in-house or at a CRO, and show that data. Bottom-feeders would be deterred if producers were rigorously to implement unique reagent identifiers, such as the RRID [31,32]. As an absolute minimum, in the interests of rigor and transparency, that: (1) the original producers of an affinity reagent must be identified unambiguously by OEM suppliers; (2) the original antibody identifiers must be used consistently by OEMs and users; (3) the concentration of functional antibody in the vial is stated and guaranteed; (4) the suppliers themselves validate (or organize a validation of) the batch of reagent that they are supplying; and (5) that this validation data, including working concentration of antibody, and the protocols used are supplied.
1) The targets. The creation of new c-Abs (2003–2017) has bettered Moore’s law. That is an interesting achievement. Still: there are many targets around in many experimental contexts, so we potentially need a lot of F4P c-Abs. The state of the immunogen and the screening technology tend to constrain the c-Abs which can then be recovered [5,22,23]. So in addition to establishing a c-Ab reference bank mentioned in point (1), reference banks of correctly folded protein targets, or target domains – as opposed to linear peptides – are crucial. This is an approach that the Structural Genomics Consortium (SGC), amongst others, have followed [24], and which the commercial HuProt™ array [25,26] approaches mainly for nonmembrane human proteins. At present equivalent banks of IHCprocessed and paraffin embedded targets are not available. It would be experimentally trivial and a valuable resource to produce both over-expressing and knock-out cell lines covering the human proteome, for use in IHC or biochemical screening. You get what you screen for: screen using Western blots and you find antibodies reacting with partially denatured material exposed on WBs; screen on viable cells and you will find antibodies reacting with the native proteins on viable cells [27,28], and so on. 2) The producers. The market for c-Abs is pushing $2 billion per year. Producers range from outstanding ethical companies to the semimythological – from good businesses to bottom-feeders. And for 50 μl of clear liquid at ∼$ 250 a shot, who can grumble? Well, me for a start [2,13]. It seems that some producers care less about the quality of their products than others [13,17,29,30]. There is clearly money to be made by even the most ethically-defective commercial entities on the c-Ab market place. The solution here, I think, is transparency.
1) The users. Despite these complaints, many antibody providers supply excellent characterized and validated c-Abs and the defined protocols under which the c-Abs were validated. However, they cannot be expected to validate under the precise experimental conditions in place in every user-laboratory. So it is obvious that users must validate and characterize c-Abs as F4P in the experimental context they are using. This is a complex job [4,8,32], and its implementation and interpretation should be a fundamental part of the training of all research biologists [1]. At the very least, user-validation must include titering c-Abs against appropriate positive and negative control samples – to discover the concentration giving optimal signal-to-noise. The validation should include cells expressing “normal” levels of the target, not just induced over-expression. For imaging applications, especially of human tissues, cell and tissue arrays are valuable tools for judging specificity – and quantitative image analysis of the data sets is valuable [15,33,34]. In all cases, a comparison with the literature should be made – despite this being potentially constructed from incorrectly validated antibodies in the first place! [14,35] Multiple validation techniques, including antibody-independent techniques if available, help to increase confidence in data subsequently generated using the c-Ab [8,22,36]. Beyond the issues of F4P lie deeper questions of absolute specificity and selectivity: and we can get target-independent indications of antibody binding characteristics from IP/MS, peptide, protein or tissue arrays [37–39], that can give a broad fingerprint of antibody specificity, rather than simple anti-target information. Native protein array chips can be used to scan the non-membrane human proteome for off-target reactivity [26]. Potentially, if the coupling chemistry is demonstrably sufficiently robust, such chips could also be denatured or otherwise pretreated to match a particular experimental context, e.g. processed as for a paraffin-embedded IHC section, to investigate off-target reactivity in particular contexts. Immunomicroscopy on broad panels of fixed tissue or cells in arrays provides both cellular and subcellular distributions, which together with known or predicted biology of the target can also support the validity of the antibody [4,15]. Few producers can characterize antibodies to such depths: it is simply not economically viable for every-day c-Abs. Irrespective of whether a validation effort has succeeded or failed, it is critical to inform the supplier and the community of the result, else a growing crowd of out-of-pocket and out-of-time users will continue to
First, transparent identifiability of reagents. Producers must uniquely identify both the immunogen, the reagent itself, and who, if not the supplier, manufactured and validated it. Best would clearly be general use of a registration system like the Research Reagent Identifier (RRID) [31] to unequivocally label each c-Ab. Verifiable identity of reagents (via a cDNA sequence) and reproducibility can easily be guaranteed using recombinant binders [18], despite excellent polyclonal antibodies [8] from entirely respectable quality producers [23]. But given the exponentially increasing numbers of c-Abs scuttling around, reproducibility and identifiability will become ever more important, and there, I think we all agree, polyclonals have issues. I see no conflict between publishing the antibody sequence and making money, though I know some disagree, and think that “trolls” may milk the honest producer. If this is a genuine issue, and I have my doubts, then I suggest that the sequences of validated but not patented recombinant reagents be deposited in a secure database, for release or comparison only to licensed parties – such a mechanism cannot be beyond the wit of man. Second, transparent identifiability of producers. c-Ab suppliers are often marketed and distributed under Other External Manufacturers (OEM) agreements. The suppliers may arbitrarily rename the reagent originally produced and supplied. Perhaps that is why there are, for example, over 6000 anti-EGFR antibodies, on offer [10] – because some are covertly “synonymized”? So, searching for an active c-Ab, one may inadvertently buy the same (e.g. inactive) reagent several times. And OEM suppliers may be so shy, they fail to guarantee identity not only of 3
New BIOTECHNOLOGY xxx (xxxx) xxx–xxx
S.L. Goodman
Conclusion
accumulate. Reliable suppliers benefit from your data, by having their reagents verified, or not, as the case may be, know that their work flow is functioning, plus of course, they can be expected to promptly replace defective c-Abs. Perhaps if you were to include a bill for your costs in de-validating the reagents, that might encourage response? It typically cost we the users at least over ten times the original antibody price to invalidate it. The community cannot await a publication that only mentions positive results. A very good move in this direction is the independent antibody reporting website pAbmAbs (http://pabmabs. com/wordpress/). Producer-linked reporting seems highly gameable and suspect, despite well-intended and convincing efforts by many reputable suppliers. Independent reporting sites could solve this issue and in many ways the CiteAb approach, to “count the winners” as it were, by noting publications using a given antibody, does this. However, that is some distance from the actual validation, and, as we have seen has its own problems []. It would help if the user-community, in coordination with leading producers, agreed a standard set of test protocols by consensus, for example defined on Protocols.io. In concept to follow the “five pillars” [32], which I note is two less than those classically suggested [(Prov. 9:1) [40] – potentially the addition of protein arrays [26] and peptide arrays [38] would be wiser? The producer would document and guarantee c-Ab specificity and selectivity using at least one of the standard protocols applied to a set of robust cell standards (e.g. ATCC cell lines [41]) – leaving F4P definition to be documented rigorously by the user.
In summary, many c-Abs are poorly documented both by the producers and the users − in particular they are poorly validated as fit-forpurpose in the experiments in which they are being used. This has caused and is causing massive intellectual and financial damage, reckoned in billions of dollars per year, and includes a progressive degradation of the literature in the biological sciences, resulting in a kind of c-Ab-induced Alzheimer’s – we won’t know what we don’t know. The current situation recalls the pre-millennial Monty Python sketch involving a malicious Hungarian-to-English phrase book [42] – if translation does not convert between languages with fealty, you will find yourselves in trouble! If c-Ab-binding does not unequivocally and unambiguously report the target in context, we are in serious trouble. In conclusion, it seems impossible that reagents that form a foundation of modern biology are in such a friable state. In the end, it is in the deepest interests of the user community and publishers, that only data obtained from antibodies appropriately defined and used as fit-forpurpose be accepted into the scientific literature, or we will all find ourselves rushing for a singularity − where the amount of information being destroyed by putative specific antibodies swamps the data from reliable c-Abs. Contributors SLG conceived and wrote the text. Dr. A.Chalmers (CiteAb; Bath) kindly agreed to let me quote direct results from the CiteAb antibody search engine. Dr. M. Taussig (Babraham Institute, Cambridge, UK) kindly reviewed a pre-submission version of this text.
1) The literature. Major publishers recognize the vital role they must play – and many are fighting to do so. Unequivocal identification and validation of the antibodies used in a given publication must be routinely documented, including the RRID if it exists, the unique identifying name, the batch, the lot and the supplier. The actual concentration of antibody used in a given validation must be unequivocally stated. These are hardly novel suggestions – fortunately – and publishers are increasingly implementing them in their guides to authors. But despite some highly determined efforts in convoluted applications like IHC [4], the level of detailing of assays in the biological literature remains generally minimalist in comparison to that in other areas, i.e. the chemical literature. This has led to the horror of the “castrated Western blots”, showing only “the band of interest”, that surely have no place in respectable publications and especially not in those purportedly characterizing c-Abs. Given the availability of Supplementary information on-line, there is no longer any excuse for this type of thing.
Funding This research did not receive any specific grant from funding agencies in the public, commercial or not-for-profit sectors. Conflicts of interests None. Acknowledgements I would like to thank Professor Andreas Plückthun who responded positively to my terrified notes, and invited me to join the battle, and to the many other members of the antibody community who have listened to me reiterating what they know all-to-well, and who passed on encouragement, especially Andrew Chalmers, Anita Bandrowski, Giovanna Roncador, Fridtjof Lund-Johansen, Monya Baker, Natalie de Souza and Mike Taussig. I thank my mentors Walter Birchmeier and Klaus von der Mark, and my one-time colleagues Ludwig Trejdosiewicz and Claudia Wilm for teaching me to be extremely cautious with antibodies – but blame only myself for not having always heeded their warnings.
“Those who follow the herd must get a back view”: indeed, this may also be a literature artefact driven by c-Abs whose specificity differs entirely from that claimed by the supplier. In clear-text: poorly validated c-Abs can lead to the construction of a literature-consensus on a well-studied topic, which can be entirely deviant from the reality [14]. And despite continual vigilance, unexpected commercial incompetence can sometimes get in the way. So not only does one have to independently validate the antibodies described in the literature, but one has to validate the standards one is using to do the validation! [35] If you wish to enter into a binding relationship with a devilish molecule, kindly use a long, or at least a well validated, spoon.
References [1] Freedman LP, Gibson MC, Bradbury AR, Buchberg AM, Davis D, Dolled-Filhart MP, et al. [Letter to the Editor] The need for improved education and training in research antibody usage and validation practices. Biotechniques 2016;61:16–8. [2] Goodman SL. Testing times for biologists. New Scientist 1989;1658:60–1. [3] Begley CG, Ellis LM. Drug development: raise standards for preclinical cancer research. Nature 2012;483:531–3. [4] Saper CB. An open letter to our readers on the use of antibodies. J Comp Neurol 2005;493:477–8. [5] Weller MG. Quality issues of research antibodies. Anal Chem Insights 2016;11:21–7. [6] Baker M. Antibody anarchy: a call to order. Nature 2015;527:545–51. [7] Baker M. Reproducibility crisis: blame it on the antibodies. Nature 2015;521:274–6. [8] Voskuil JLA. Commercial antibodies and their validation. F1000Res 2014;3:232. [9] Helsby MA, Leader PM, Fenn JR, Gulsen T, Bryant C, Doughton G, et al. CiteAb: a searchable antibody database that ranks antibodies by the number of times they
1) Training. There is a pressing need to routinely educate users, and most especially young users, in the correct use of antibodies, as Les Freedman and colleagues at the GBSI noted following a survey last year [1]. It should be a fundament of any modern biology training programme that students be taught how to correctly characterize, validate and use antibodies. In particular, the limitations of the various validation techniques should be emphasized. I would go further and say that a formal certification of graduate students as antibody users would be highly desirable − before they are let loose on the community, and permitted to use c-Abs. 4
New BIOTECHNOLOGY xxx (xxxx) xxx–xxx
S.L. Goodman
have been cited. BMC Cell Biol 2014;15:6. [10] Chalmers A, CiteAb. https://wwwciteabcom/. Bristol, U.K. 2018. p. Company Website. [11] Michaud GA, Salcius M, Zhou F, Bangham R, Bonin J, Guo H, et al. Analyzing antibody specificity with whole proteome microarrays. Nat Biotechnol 2003;21:1509–12. [12] Bird C, Antibodies User Survey, The Scientist: https://www.the-scientist.com/? articles.view/articleNo/32042/title/Antibodies-User-Survey/; 2012. [13] Couchman JR. Commercial antibodies: the good, bad, and really ugly. J Histochem Cytochem 2009;57:7–8. [14] Andersson S, Sundberg M, Pristovsek N, Ibrahim A, Jonsson P, Katona B, et al. Insufficient antibody validation challenges oestrogen receptor beta research. Nat Commun 2017;8(15840):1–11. [15] Goodman SL, Grote HJ, Wilm C. Matched rabbit monoclonal antibodies against αvseries integrins reveal a novel αvβ3-LIBS epitope, and permit routine staining of archival paraffin samples of human tumors. Biol Open 2012;1:329–40. [16] Bradbury AR, Pluckthun A. Getting to reproducible antibodies: the rationale for sequenced recombinant characterized reagents. Protein Eng Des Sel 2015;28:303–5. [17] Lowe D. Trouble at Santa Cruz Biotechnology. Sci Translat Med 2016;23(May) http://blogs.sciencemag.org/pipeline/archives/2016/05/23/trouble-at-santa-cruzbiotechnology2016. [18] Bradbury AM, Pluckthun A. Antibodies: validate recombinants once. Nature 2015;520:295. [19] Bradbury AM, Pluckthun A. Standardized antibodies used in research. Nature 2015;518:27. [20] Springer TA. Institute for Protein Innovation. Cambridge, MA, USA: IPI; 2017. Home page of Institute https://proteininnovationorg/antibody-initiative/. [21] Colwill K, Renewable Protein Binder Working G, Graslund S. A roadmap to generate renewable protein binders to the human proteome. Nat Methods 2011;8:551–8. [22] Roncador G, Engel P, Maestre L, Anderson AP, Cordell JL, Cragg MS, et al. The European antibody network's practical guide to finding and validating suitable antibodies for research. MAbs 2016;8:27–36. [23] Ascoli CA. Antibody basics for the production of functional antibodies. 1 ed. Limerick, PA: Rockland Inc.; 2016. p. Company Website https://rockland-inccom/ uploadedFiles/Support/Resource_Library/Antibody%20Basicspdf. [24] Zhong N, Loppnau P, Seitova A, Ravichandran M, Fenner M, Jain H, et al. Optimizing production of antigens and fabs in the context of generating recombinant antibodies to human proteins. PLoS One 2015;10:e0139695. [25] Cambridge Protein Arrays Ltd. Cambridge, UK: Cambridge Protein Arrays Ltd.; 2017. p. https://cambridgeproteinarrays.com. [26] Jeong JS, Jiang L, Albino E, Marrero J, Rho HS, et al. Rapid identification of
[27]
[28] [29] [30] [31]
[32] [33]
[34]
[35]
[36]
[37]
[38] [39] [40] [41] [42]
5
monospecific monoclonal antibodies using a human proteome microarray. Mol Cell Proteomics 2012;11. O111.016253. Imhof BA, Vollmers HP, Goodman SL, Birchmeier W. Cell-cell interaction and polarity of epithelial cells: specific perturbation using a monoclonal antibody. Cell 1983;35:667–75. Knudsen KA, Rao PE, Damsky CH, Buck CA. Membrane glycoproteins involved in cell-substratum adhesion. Proc Natl Acad Sci U S A 1981;78:6071–5. Cyranoski D. The secret war against counterfeit science. Nature 2017;545:148–50. Reardon S. US government issues historic $3.5-million fine over animal welfare. Nat News 2016:2016. Bandrowski AE, Martone ME. RRIDs: a simple step toward improving reproducibility through rigor and transparency of experimental methods. Neuron 2016;90:434–6. Uhlen M, Bandrowski A, Carr S, Edwards A, Ellenberg J, Lundberg E, et al. A proposal for validation of antibodies. Nat Methods 2016;13:823–7. Toki MI, Cecchi F, Hembrough T, Syrigos KN, Rimm DL. Proof of the quantitative potential of immunofluorescence by mass spectrometry. Lab Invest 2017;97:329–34. Anagnostou VK, Lowery FJ, Syrigos KN, Cagle PT, Rimm DL. Quantitative evaluation of protein expression as a function of tissue microarray core diameter: is a large (1.5 mm) core better than a small (0.6 mm) core? Arch Pathol Lab Med 2010;134:613–9. Quarmby V, Quan C, Ling V, Compton P, Canova-Davis E. How much insulin-like growth factor I (IGF-I) circulates? Impact of standardization on IGF-I assay accuracy. J Clin Endocrinol Metab 1998;83:1211–6. Skogs M, Stadler C, Schutten R, Hjelmare M, Gnann C, Bjork L, et al. Antibody validation in bioimaging applications based on endogenous expression of tagged proteins. J Proteome Res 2017;16:147–55. Lund-Johansen F, de la Rosa Carrillo D, Mehta A, Sikorski K, Inngjerdingen M, Kalina T, et al. MetaMass, a tool for meta-analysis of subcellular proteomics data. Nat Methods 2016;13:837–40. Breitling F, Nesterov A, Stadler V, Felgenhauer T, Bischoff FR. High-density peptide arrays. Mol Biosyst 2009;5:224–34. Richer J, Johnston SA, Stafford P. Epitope identification from fixed-complexity random-sequence peptide microarrays. Mol Cell Proteomics 2015;14:136–47. New Oxford annotated Bible: New Revised Standard Version with the Apocrypha. Oxford University Press; 2010. ATCC. https://www.atcc.org/?geo_country=uk. 2017. Chapman GC, Idle John, Jones Eric, Palin Terry, Gilliam Michael, Terry. Hungarian phrase book. In: MacNaughton I, editor. Monty Python's Flying Circus. London, U.K: BBC; 1970https://wwwyoutubecom/watch?v=vAQJHHf3i1o.