Editorial: Antibody Can Get It Right: Confronting Problems of Antibody Specificity and Irreproducibility

Antibodies are indispensable research reagents. Their use in endocrinology has exploded since the original radioimmunoassays were developed to measure hormone levels. They are now routinely used in molecular endocrinology research to detect and quantitate cellular proteins and determine their distribution in tissues, in cells, and in subcellular compartments. They detect changes in protein structure produced by covalent modification during cell signaling (eg, phosphorylation), transcriptional regulation (acetylation and methylation), or protein degradation (ubiquitination). They identify and detect protein-protein and protein-DNA interactions by coimmunoprecipitation. Antibodies are used as tools to sort cells, to identify drug targets in normal and diseased tissue, and to monitor disease status and progression. And that is just the tip of the iceberg.

Unfortunately, the widespread use of antibodies has also generated enormous controversy: the inability of investigators to replicate published data often results from false-positive or false-negative results produced with antibodies that have not been properly validated (1–5). The problems are particularly intense in fields that use antibodies to analyze proteins that are expressed at low levels in cells: G protein–coupled receptors, steroid hormone receptors, ion channels, transporters, and signal-transducing enzymes such as adenylyl cyclase (6–22). In addition, ultrasensitive detection methods have compounded the problem as illustrated by studies aimed at detecting low-abundance protein–DNA interactions (eg, chromatin immunoprecipitation assays) (5, 23). So what can we do as individual scientists and as journal editors to ensure the reliability and reproducibility of the data that we generate, the data that we publish, and the data that we rely upon to formulate new hypotheses and plan future experiments?

Challenge 1—Identifying a Reliable Antibody for the Job at Hand

Investigators can either purchase antibodies or make their own, and there are major advantages and disadvantages to each choice. However, in either case, validation of antibodies is the responsibility of the user—validation of commercial antibodies by their suppliers is usually inadequate and frequently unreliable.

The advantages of purchasing antibodies are obvious. Good antibodies can take months to produce, there are substantial up-front costs, and there is no guarantee that a particular antigen will induce antibodies of the desired properties in immunized animals. The hope is that purchasing antibodies will eliminate production delays, reduce the investment of time and money in a particular experiment, and preclude the possibility of failure to generate the desired reagent. However, identifying an appropriate antibody for purchase is no simple task: there are often dozens of antibodies available to a target protein with little information provided as to their affinity or specificity.

One might think that a manufacturer's catalog number would provide a unique identifier for a particular antibody. However, this is not the case. Vendors usually assign catalog numbers for an antibody based on the immunizing antigen, the manner in which the antibody was produced (host animal, polyclonal or monoclonal, affinity purification, etc.), and the manufacturer that produced it. In the case of polyclonal antibodies, vendors may use the same catalog number not only for different blood collections from an individual immunized animal but also for blood collections from different host animals immunized with the same antigen. Because each blood sample collected from each animal provides a unique combination of antibody clones and concentrations, this practice can result in immense lot-to-lot variability in antibodies. To add further confusion, validation data provided by a vendor for an antibody may not have been generated with the current lot of that antibody. Although monoclonal antibodies would not be expected to show such lot-to-lot variability, in fact they can. For example Pozner-Moulis et al (24) demonstrated that two different lots of a monoclonal antibody to the Met tyrosine kinase receptor showed opposite staining patterns in an array of more than 600 breast cancer cases: one showed nuclear and the other membranous and cytoplasmic staining. One final word of caution: a particular antibody may be sold by several distributors under different catalog numbers. Caveat emptor.

Because the documentation provided by manufacturers is often inadequate, a number of searchable databases have been established to inventory and index antibodies from multiple vendors and to list publications that have cited each antibody (for examples, see Refs. 25–29). Such databases are very helpful: articles that carefully validate an antibody for a specific application often provide the most useful guide for antibody selection, and authors who generate such information ought to be commended and cited. However, it is important to remember that the results in such articles may have been produced using a lot that is no longer available. Unfortunately, many published articles do not provide catalog numbers or lot numbers for the antibodies used and hence cannot be included in these databases.

“It is the user's responsibility to ensure that these antibodies are sufficiently sensitive and specific to produce reliable results in the proposed studies. The literature documenting the many problems and incorrect conclusions resulting from poorly validated antibodies shows that vigorous skepticism and critical analysis are a must during this process (1–18).”

Because commercial antibodies often lack the sensitivity and specificity required for an application, user validation is essential. In fact, testing of purchased antibodies often requires a substantial investment of time and money, and there is no guarantee that a satisfactory reagent will be identified for the intended use or that it will continue to be available. When investigators choose to generate their own antibodies, they do so to ensure that the work invested in antibody validation is not wasted and will result in sufficient amounts of a carefully characterized reagent to complete a study. However, there are also some useful side benefits. Investigators can control the purity of the antigen they use for immunization, whether it is a chemically synthesized peptide or a purified protein, thus helping to ensure that the resulting antibodies will be specific. Vendors normally do not disclose the purity of their immunogens, and, in many instances, even the nature of the immunizing antigen is kept confidential. When investigators generate their own polyclonal antibody, they can produce both preimmune serum and extra immunizing antigen for subsequent validation experiments. Generating monoclonal antibodies will often yield multiple hybridoma clones that can be compared to help ensure that any observed reactivity is specific for the target protein.

Often the decision to buy an antibody or to generate one's own will be determined by whether a commercial antibody has been demonstrated by others to “work” for a desired application and whether the antibody will be used for a short-term or a long-term project.

Challenge 2—Validating Antibodies

Once one or more promising antibodies are identified, whether purchased or generated in-house, it is the user's responsibility to ensure that these antibodies are sufficiently sensitive and specific to produce reliable results in the proposed studies. The literature documenting the many problems and incorrect conclusions resulting from poorly validated antibodies shows that vigorous skepticism and critical analysis are a must during this process (1–18). All antibodies cross-react to some extent (1, 30). In any particular situation, the ratio of specific to nonspecific binding to target and off-target reactors will be determined by the relative abundance of each of the reactors in a particular preparation, the affinity of the antibody for each reactor, the antibody concentration used in the assay, and the method of sample preparation. Off-target reactivity with polyclonal antibodies can be produced by any of the antibody species present in the preparation. However, monoclonal antibodies also show nonspecific reactivity (31). In fact, because mouse monoclonal antibodies tend to have lower binding affinities than rabbit polyclonal antibodies and are often used at high concentrations, nonspecific binding can be a prohibitive problem.

A large number of thoughtful articles provide detailed guidance for proper antibody validation and discuss the limitation and interpretation of each type of control (13, 30–36). Investigators who use antibodies need to be familiar with this literature and rigorously apply the criteria recommended. It is particularly important to remember that the reactivity of an antibody with a target protein depends on the concentration of that protein in the preparation. Thus, the fact that an antibody is able to detect a protein at the high levels observed in transfected cell lines does not guarantee that it will do so in nontransfected cells or tissues. As a result, the most stringent control for antibody specificity requires comparison of antibody reactivity in wild-type tissues or cells to reactivity in knockout animals or cell lines in which endogenous protein expression has been silenced. For human tissue samples in which such a knockout control is not feasible, an alternate stringent control would be the use of multiple antibodies recognizing different epitopes in the same target protein.

Some common misconceptions have particularly impeded reliable antibody usage. A control frequently used to evaluate antibody specificity is to determine the effect that preincubation with excess antigen has on antibody reactivity: the antigen preadsorption test (31). Although this control can sometimes be useful, it is certainly not sufficient. Blocking reactivity with excess antigen demonstrates that the reactivity is produced by an antibody that recognizes that antigen. However, it does not demonstrate that the antibody is specific only for that antigen. In fact, because this control is often carried out using very high concentrations of blocking antigen, low-affinity as well as high-affinity antibodies in a preparation will be blocked, and the binding of all these antibodies to both on-target and off-target reactors will be inhibited. Thus, although the antigen preadsorption test can identify the population of antibodies responsible for observed reactivity, it cannot demonstrate that those antibodies are specific.

“The most stringent control for antibody specificity requires comparison of antibody reactivity in wild-type tissues or cells to reactivity in knockout animals or cell lines in which endogenous protein expression has been silenced. For human tissue samples in which such a knockout control is not feasible, an alternate stringent control would be the use of multiple antibodies recognizing different epitopes in the same target protein.”

A false sense of security may also be provided by a statement from vendors that their antibody works for immunocytochemistry or immunohistochemistry but does not work on immunoblots: such a statement provides an excuse not to bother with Western blot controls. However, this statement usually does not mean that an antibody produces a “clean” Western blot with no signal, implying a lack of sensitivity for the target protein, but rather that many bands are observed on immunoblots, none of which correspond to the target protein; ie, the antibody is both insensitive and lacks specificity. Although it is understandable that antibodies produced to native proteins may not recognize the corresponding denatured protein after SDS-PAGE, this is not true for antibodies produced against short peptides. Thus, rather than being reassuring, a vendor's statement that an antipeptide antibody does not work on immunoblots should be cause for concern. We as well as others (24) have found that immunoblotting provides a more sensitive and discriminating method for antibody validation than immunocytochemistry because it so clearly distinguishes cross-reacting proteins of different molecular sizes. Moreover, it is far more likely that an antipeptide antibody will “work” in Western blotting and not in immunocytochemistry than the reverse. The only exception that we have encountered occurred when only a very small fraction of cells in a tissue expressed the target protein. In that circumstance, the target protein constitutes such a small part of a tissue extract that it is not detectable in an immunoblot. Thus, all antipeptide antibodies should be tested by immunoblotting tissues and/or cells that do and do not contain the target protein to ensure (a) that the intact protein is detected with the necessary sensitivity and (b) that reactivity with off-target proteins is minimal.

Although extensive validation of many commonly used antibodies (eg, anti-ERK) is no longer required or expected because of the enormous amount of published data that demonstrate their specificity and provide reliable positive controls, this is not the case for antibodies to receptors and many other proteins involved in cell signaling. Thus, experiments that test antibody specificity with clear positive and negative controls are essential—otherwise an entire project may be based on an artifact, and valuable research resources will be wasted when others try to replicate published data (6–16). Investigators who wish to avoid the embarrassment of incorrect conclusions must adopt the mantra: validate early and often.

Challenge 3—Publishing Sufficient Information to Allow Critical Evaluation of Experimental Data and Facilitate Reproducibility

Journals have a responsibility to help ensure that the articles they publish report valid and reproducible results. Unfortunately, journals have put increasing emphasis on reducing the length of published articles. Thus, the space devoted to describing materials and methods has been dramatically reduced over the years. The source of many reagents is missing or inadequately described. Critical control experiments have been relegated to “data not shown.” Figures display Western blots with only slivers of gels containing bands of undefined molecular weights. Frequently, only summary graphs, but no original data, are shown. These widespread practices prevent assessment of either the sensitivity or the specificity of the antibodies utilized and thus endanger the conclusions reached.

The good news is that the advent of electronic publication now permits rigid space restrictions to be relaxed. Molecular Endocrinology plans to take full advantage of our change to electronic publication to ensure that the articles in our journal are reliable and reproducible. To guarantee that research reagents are clearly identified, we plan to follow the excellent example set by Endocrinology (37) and publish a full description of all antibodies used in a study. Essential information to be provided includes the following: the name of the individual or vendor who supplied the antibody, the immunizing antigen used to generate the antibody, the nature of the antibody preparation (species, polyclonal or monoclonal, and affinity purified or not), the catalog number, and, importantly, the lot number. We encourage authors to include this information in all submitted papers starting immediately and will require this information starting January 1, 2015. In addition, as a service to the research community, we encourage authors to also list antibodies that they have tested and found to be unsatisfactory.

Finally, to help ensure that results in our journal are both reliable and reproducible, we commit to publishing the experimental validation for each new antibody utilized (positive and negative controls, full gels for immunoblots showing the molecular weights of all stained bands, etc.). If antibodies have been validated in a previous publication, this must be cited, and the validating experiments and results should be summarized in the text to provide confidence in the present publication. In all instances, investigators need to describe, and preferably show, the experiments demonstrating that their antibodies are specifically detecting the target protein under their experimental conditions. Such supporting data may go into the body of the article or into a supplement, as the authors see fit.

Antibodies are extremely powerful and sensitive reagents that provide the foundation on which many different molecular studies are built. However, they are capable of producing not just misleading but completely incorrect results. By following best practices and exercising great caution with a healthy dose of skepticism, we can ensure that published data are both reliable and reproducible and that the antibodies we use “get it right.”

Agnes Schonbrunn, PhD
Associate Editor