The pharmaceutical industry has recently been reported as facing a reproducibility crisis where researchers have failed to reproduce key drug research results. Roumen Bogoev, Bio-Rad Laboratories looks at how this issue can be improved for western blotting techniques
I read your paper and I have some bad news for you. Of the two antibodies you used, the first cannot be used for IHC and it gives false positive data by western blot. The other cannot be used for western or IHC either, as that antibody detects a conformational epitope that is destroyed by fixation.’
So begin dozens of letters to researchers informing them that faulty and unreliable antibodies had tainted their studies. But the full scale of the problem stretches far beyond a few dozen scrapped papers. A study published this May suggests that about half of published preclinical research produced each year is irreproducible1. Research points to antibodies as a major contributor to this ‘reproducibility crisis’.
Predominantly used to detect proteins, antibodies are an essential tool used in drug discovery, development and characterisation studies. Western blotting, immunohistochemistry (IHC), immunofluorescence (IF), immunoprecipitation (IP), enzyme-linked immunosorbent assay (ELISA) and more rely on antibody specificity to give qualitative and quantitative protein characterisation data. Unfortunately, this dependence on antibodies has become a major hindrance to progress in pharmaceutical R&D.
In a characterisation of roughly 20,000 antibodies, the Human Protein Atlas, a Swedish consortium aiming to generate antibodies for the entire human proteome, determined that less than half were suitable for examining protein localisation2. A 2009 investigation of 49 commercial antibodies developed to recognise G-protein coupled receptors (GPCR), which are high priority drug targets, found most were non-specific, recognising multiple different receptors3.
This begs the question: with hundreds of antibody providers, many of which are unpredictable or outright unreliable, how can companies be sure the antibody they are buying will work? Out of conversations with many researchers about the frustrations they face in seeking quality antibodies, Bio-Rad has developed a series of best practices and a simple checklist for buying and using antibodies.
Every party in the antibody industry – from funding agencies to journals to researchers – contributes to the antibody irreproducibility problem, but suppliers could do a lot more to provide a reliable product.
Scientist and founder of antibody watchdog website CiteAb, Dr Andrew Chalmers, has identified the large variety of techniques in antibody use as a key challenge for vendors. ‘People use antibodies for a massive range of applications and also in a range of different species and tissues. Companies could never validate an antibody for all of those uses,’ he says.
Very little data is available to confirm a company’s claims that the antibody will reliably detect target proteins, even in a standard western blot
In addition, very little data is available to confirm a company’s claims that the antibody will reliably detect target proteins, even in a standard western blot. Mark Shulewitz, a Senior Scientist in the Content Business Development Team at Bio-Rad, notes that on vendor websites, ‘in most or all cases, vendors tend to show only the positive data and to not show samples where they don’t see any expression. They also don't show the entire western blot; they crop it to show just the region of interest.’
Perhaps the biggest indication that antibody manufacturing has a long way to go is the phenomenon of lot-to-lot variability, especially in polyclonal antibodies. These are produced in live animals and rely on the animals’ immune systems to produce antibodies to the protein of interest; however, each animal has a unique immune system and will deliver a pool of antibodies that may or may not be specific. When a lot is discovered with high specificity, some researchers will stockpile antibody lots for years of future research, aware that a new lot could contain a poor-performing pool of antibodies, despite using the same catalogue number.
Validation is arguably the most critical word in this field, but standards for validation vary wildly among manufacturers, with little transparency for customers.
A lot of vendors will validate using the same recombinant protein that they used to immunise an animal, whether it would be a mouse or rabbit
Researchers rely on antibodies that detect one specific protein within a sea of thousands of other proteins within a cell’s lysate. A good assessment for antibody precision will test its specificity within an endogenous background. Unfortunately, many vendors take shortcuts that make it appear that their antibodies are more specific than they truly are. ‘From what we’ve seen, a lot of vendors will validate using the same recombinant protein that they used to immunise an animal, whether it would be a mouse or rabbit,’ says Shulewitz. This strategy will test if the antibody detects the protein of interest, but not its specificity in the context of a complex sample.
Dr Subhash Verma, Assistant Professor of Molecular Microbiology and Immunology at the University of Nevada, Reno, has learned throughout his career to keep an eye out for validation shortcuts that vendors have made. For instance, some companies rely on over-expressing the target protein to demonstrate performance; however, this does not always translate into performance with real samples.
‘An over-expressed antibody study will certainly work, but vendors need to test their antibodies on total-cell lysates,’ Verma says. ‘If these manipulated results are published in isolation, researchers are blind to the fact that the antibody in the study may not work for their western blot with standard or low-protein expression.’
|Raising the bar in the supply antibodies for western blot applications|
|Three decades after western blots entered mainstream use, many still fail due to antibody associated problems. After troubleshooting, nearly two in three researchers find the antibody was at fault. How can a scientist negotiate these industry wildcards? Bio-Rad Laboratories, a new player in the field, has a new concept – antibodies validated specifically for western blotting on a wide range of samples.|
|Bio-Rad is an established supplier of western blot tools; however, the offering lacked a critical element of the workflow – the primary antibodies. In 2014, the company took the first step of acquiring antibody provider AbD Serotec. This June, Bio-Rad launched PrecisionAb, an antibody product line specifically validated for western blotting to address the key challenges pervading the antibody marketplace.|
|‘We have a high degree of customer trust around our western blotting and electrophoresis gels and all those parts of the V3 workflow,’ says Mark Shulewitz, a Senior Scientist in the Content Business Development Team at Bio-Rad. ‘We didn't want to move into antibodies until we were certain that we would deliver on the quality our customer base expects.’|
|Each PrecisionAb antibody is validated on whole cell lysates from up to 12 different biologically relevant cell lines – going far beyond the typical few samples – and only antibodies that detect endogenous proteins with high sensitivity and specificity are chosen for the product line. ‘As far as lot-to-lot variations or variability, we’re testing every lot that we bring in,’ Shulewitz says. ‘A new lot has to work or we don’t accept it.’|
|Instead of releasing thousands of antibodies to the widest possible market, the PrecisionAb launch will focus on antibodies that have been well-characterised and stringently screened by western blotting (see Figure 1). Bio-Rad will publish the complete testing results for each approved antibody, positive and negative. Looking at these results, researchers need to keep in mind that a negative result doesn’t mean the protein isn’t there, it is just, relatively speaking, much less abundant,’ says Shulewitz. ‘The customers have an idea of relative expression levels of their target in their sample type, so we’re also guiding that customer towards samples that work. In addition, we’re providing our own positive control so that customers can verify that the results that they are seeing are just like those we see at Bio-Rad.’|
|Each PrecisionAb antibody comes with a complete validation protocol to help the researcher streamline optimisation and validation, saving time and reagents and delivering top results. Trial sizes are also available for all targets, enabling researchers to test the antibody and optimise experimental conditions without a large upfront investment. Positive control is included to facilitate target band identification as well as troubleshooting if needed.|
|These new products raise the bar for transparency and accountability, making it possible to complete our vendor checklist in one transaction, and to overcome the challenges of uncertain antibody quality, reproducibility and lack of data for the reliability of specific western blot applications.|
Figure 1: Validation data for two antibodies. A: this carbonic anhydrase IX (CA9) mouse monoclonal antibody failed validation due to nonspecific binding and low signal-to-noise ratio; B: this thymidylate synthase (TYMS) mouse monoclonal antibody passed validation showing high specificity and sensitivity
How can scientists move towards performing reproducible experiments using reliable antibodies? First, vendors need to be effective and efficient, validating as thoroughly as possible without making the market too broad. Second, given the many factors that can’t be controlled, scientists have to accept that they will always need to perform a final validation in-house.
A leading expert on antibody reliability, Dr C. Glenn Begley, spent more than 10 years at a large biotechnology company and confirmed that his scientific teams would attempt to validate in-house. ‘We saw considerable lot-to-lot variations, so that validation was really necessary with every new batch of antibodies,’ says Dr Begley.
Ensuring high-performance antibodies requires a vendor to replicate the environment in which the antibodies will be used
Ensuring high-performance antibodies requires a vendor to replicate the environment in which the antibodies will be used. Large manufacturing facilities, however, cannot always duplicate conditions found in small academic or industry labs. Vendors can address these variables by using beta testers and real scientists in the field.
To complement the specialisation, vendors also need to broadly validate each product batch against multiple cell lines and sample types. This means testing how well their product performs with variable levels of protein expression, degradation and background antigens.
Ultimately, scientists need to buy from vendors who provide a comprehensive and complete characterisation of their products. The company website should publish not only all positive results, but also tests performed in which a signal wasn’t detected. These data allow researchers to judge whether a given antibody might compromise their studies or work poorly for a particular application or sample type. A list of samples that are negative also can provide negative controls for others, which are essential in detecting false-positive results, an important part of any properly performed experiment.
Fundamentally, good antibodies must recognise the correct target, and off-target binding should be minimal. But how can one be assured of these standards before purchasing?
First, researchers should closely scrutinise suppliers’ validation data, paying attention to the specific validation protocol and if tests were against endogenous sources. Next, an investigation of journals citing the antibody will usually give a good indication of the antibody’s ability to detect the protein of interest. Citations can sometimes be misleading, however, due to the use of unvalidated and nonspecific antibodies in many papers published in peer-reviewed journals. Readers should therefore not assume that just because multiple groups used a particular antibody, that it was properly validated.
In addition, third party websites are helping to collate the evidence. CiteAb is a popular example, with more than two million antibodies listed and 15,000 unique visitors per month. F1000, another available resource, provides scientists with antibody validation results and peer-reviewed data, but its secondary role is in highlighting the importance of the often-overlooked validation step.
Until antibody manufacturers assemble the evidence that their products work in one convenient location, it is necessary to consult multiple sources when considering a purchase.
Figure 2: PrecisionAb mouse anti-Cdk7 antibody tested on 12 whole cell lysates
No amount of research about an antibody will guarantee its use for any company’s particular application. This means that in-house validation is crucial. For example, the amount of protein loaded on a western blot and the dilution of the antibody are factors that must be worked out within individual laboratories. Fortunately, some companies offer trial sizes of antibodies for in-house validation purposes.
Antibody suppliers have to maintain a balance between rigorous validations and the associated costs
‘We do run a test first to work out whether it is picking up the target and whether it’s the right molecular weight,’ Verma says. When this doesn’t work, he encourages scientists in his lab to persevere with validation. ‘I tell them: “Load the protein in different concentrations to get a better idea about the specificity of the antibodies. If it’s not working, then try different dilutions.” Sometimes putting too much antibody in might give you too much background noise. Or if you aren’t picking up any binding, increase the amount of antibody you’re adding,’ he says.
Antibody validation takes time, and antibody suppliers have to maintain a balance between rigorous validations and the associated costs. Until companies can provide more reliable validation data, the best course of action for choosing antibodies comes down to this – avoid the many traps throughout the industry, buy from suppliers who are transparent about their data, and perform rigorous validation in-house.
1. L.P. Freedman et al. PloS Biol. 2015, DOI: 10.1371/journal.pbio.1002165
2. L. Berglund et al. Mol. Cell. Proteom. 2008, 7, 2019
3. M.C. Michel et al. Naunyn-Schmiedeberg’s Arch. Pharmacol. 2009, 379, 385
PrecisionAb is a registered trademark of Bio-Rad