Conditional Epitopes: Is Your Antibody Always Specific?

In an accompanying article, Josephsen et al. (1999) describe the interesting observation that a monoclonal antibody (MAb) originally generated against purified porcine vimentin (V9) reacts with a novel epitope on an unrelated protein, an epitope that is generated only after glutaraldehyde fixation. These authors point out that such a well-characterized reaction, even though it was not the original target of the MAb generated, can be a useful reagent when used in the proper setting. They suggest that this is “nonspecific” but “selective.” To me, the semantic arguments about what “specific” and “nonspecific” mean are less interesting. I would argue that such reactions are highly “specific” as well as “selective” but, more importantly, that they are “useful.”

This novel article highlights an important issue, however. Immunocytochemical methods are incredibly selective and sensitive. They enable us to detect avid reactions using the spectacularly “specific” properties of antibody molecules. The more these methods are used, the more we all must appreciate the power of these selective reactions and the resulting obligation to exercise care in their control and interpretation. I have personally had the experience of generating MAbs to complex mixtures of antigens, i.e., whole cells. This is a humbling experience because of the incredible diversity of high-avidity antibodies one can find. In the thousands of clones we isolated, I do not remember actually finding duplicates; that is, every new clone that showed reaction with something in a cell was unique. Sometimes these clones reacted with different parts of the same molecular species, such as complex elaborate carbohydrates, but each clone generated an antibody that recognized a different portion of such a molecule, i.e., a different epitope.

Fixation

Preservation of structure for immunocytochemistry using chemical fixatives is well-recognized to create problems. The major challenge with fixation has often been the subsequent lack of reaction of antigens with antibodies. In spite of a widespread interpretative dogma in the literature, most of these problems were not due to chemical alteration of antigen epitopes. They were more commonly a problem of inaccessibility (Willingham and Yamada 1979). This concept has finally gained some acceptance because of the success of antigen retrieval techniques demonstrating that the reactive epitopes were always there in the cells, but they were just inaccessible (Shi et al. 1995). Such a result is reported in the article by Josephsen et al. in their observation that high (2.5%) concentrations of glutaraldehyde resulted in no reaction in their cells, presumably because of inaccessibility of the epitope. In contrast, the generation of novel epitopes due to fixation is less common but is not without precedent. Unexpected crossreactions to other antigens are even more common.

“Conditional Epitopes”

Epitopes are small molecular assemblies, each having certain properties of size, charge, and shape. The original simplistic thinking of those of us in the field of immunocytochemistry suggested that antibodies generated to protein × should react only with protein X. However, the subsequent powerful technologies of MAb production, the generation of synthetic peptides, and the ability to discern primary sequence through molecular cloning, have taught many of us, including myself, painful lessons.

It is a common observation that MAbs generated to protein epitopes work well in some conditions but not in others. The usual observation is that some MAbs work well when the proteins are in their native conformation, such as in undenatured cell extracts, and these are useful for immunoprecipitation experiments. These same antibodies are also frequently useful for immunocytochemistry, because most fixation procedures preserve the native conformation of proteins, assuming that they are accessible. On the other hand, some MAbs react poorly with the native conformation; instead, they react well with fully denatured proteins after exposure to SDS. Such antibodies work well for Western blots. Antibodies made by immunizations with isolated peptides usually produce this type of useful Western blot reagent, antibodies that frequently do not work well in immunocytochemistry. Pragmatically, of every 10 antibodies generated to hydrophilic sequences of large proteins using synthetic peptides, only one might react with the native conformation of the protein. In addition, some of the epitopes buried in these fixed proteins may be similar in size, shape, and charge to regions present on other proteins, epitopes that are of entirely different primary sequence. Therefore, I believe that most of the epitopes we detect using MAbs are actually “conditional,” in that they are reactive only under certain conditions. We choose the appropriate MAb based on the conditions of the experiment to be performed. On the other hand, there are those occasional MAbs that work well with both denatured and native conformations. They presumably recognize peptide sequences that are exposed on the surface of the native protein but are also exposed when the proteins are denatured. For glycoproteins, many of the most immunogenic portions of the molecules are carbohydrate epitopes, or combination epitopes between carbohydrates and peptides. Without a prior knowledge of the three-dimensional structure of proteins, however, it is difficult to predict which of these types of reactive epitopes one should expect. When we employed mainly polyclonal antibodies as immunocytochemical reagents, we rarely encountered any problems of this type because multiple epitopes were recognized, and the ones reactive under the conditions employed were the ones we detected. Monoclonal antibodies depend, however, on reaction with single epitopes, and it is in this setting that “conditionality” becomes apparent. Under some conditions, an epitope may be generated that is not related in any way to the antigen we intend to detect.

One such case I personally encountered was the observation that a synthetic peptide sequence from the middle T-antigen of polyoma virus generated a wonderful antibody that reacted with a major cytoskeletal protein that had nothing to do with this virus (Ito et al. 1983). Another example I encountered was the observation that a commonly used MAb (C219) to the multidrug transporter (P-glycoprotein) reacted with great “specificity” to slow-twitch skeletal muscle myosin (Thiebaut et al. 1989). Both of these were examples of peptide sequences that generated antibodies that reacted with great specificity not only to other proteins but to proteins that were present in very high concentrations in cells. I have also observed antibodies such as PC10, an antibody reactive with PCNA, that appears to react with its antigen only after fixation (unpublished), similar to the situation described by Josephsen et al. The message that comes from such observations is that immunocytochemical interpretations are only as good as their controls.

The epitope generated by fixation in the article by Josephsen et al. was also generated by fixation of the protein directly on an immunoblot. That should not necessarily be expected for all such antigens, because some epitopes generated by fixation might require complex folding of the protein in solution to appropriately appose two reactive regions that go together to form the novel epitope. Their control experiment pointed to a logical interpretation of the result. Other epitopes are recognized by antibody in their native state and then lost after some specific treatment. For example, I have had the experience of utilizing an antibody that recognized an epitope present on the surface of a monomeric protein (pyruvate kinase), but which was buried when the molecule assumed its active tetrameric form (Ashizawa et al. 1991). Therefore, the “condition” necessary for its detection was the exposure of the epitope through dissociation of subunits. In another example of a conditional epitope, there is an epitope in human bcl-2 that resides in a large unstructured loop region of the molecule recognized by the MAbs 100 and 124 (amino acids 41-54 in human bcl-2) (Pezzella et al. 1990). This epitope is exposed when this protein is present in the cytoplasm of cells, but when present in the nucleus of cells it is inaccessible, perhaps through complexing with some other protein (Schandl et al. 1999). Yet, when this molecule is phosphorylated at sites present near this same epitope during G₂ of the cell cycle, the epitope becomes accessible and demonstrates the concentration of this protein on the surface of mitotic chromosomes (Liu et al. 1984; Willingham and Bhalla 1994). If, however, this molecule is extracted from cells at any phase of the cell cycle using detergents, it is soluble and the epitope is exposed in its native conformation. This epitope is also intact when the protein is denatured by SDS, and these antibodies react well with this epitope in Western blots.

Design of Appropriate Controls

There is no single control step that is applicable for all such conditional epitopes. The main message, however, is that control techniques must be highly sensitive, and many cases may require multiple control techniques to clearly interpret an unexpected localization. Competition of antibodies generated against synthetic peptides with excess peptide is, for example, a control for the antibody, not for the antigen with which it reacts. The technique I most often prefer, beyond immunoprecipitation, Western blotting, or multiple antibodies to different epitopes on the same protein, is the use of a gene expression control. That is, in tissue culture systems, one can show that the antibody in question reacts in a certain way only when a certain gene is introduced into the cell and not when it is absent. Unfortunately, this control is not possible in intact tissues, with the exception of knock-in and knockout transgenic animals. It is also not always a perfect control, either.

Conclusion

The selectivity of immunocytochemical techniques can be exquisite. Because these are such powerful techniques, however, the control of interpretation of these techniques must be extremely careful. The existence of many examples of unexpected and highly selective reactions to other antigens should make those of us who use these methods especially sensitive to the possibility of interpretive errors. On the other hand, when a reagent is developed that shows a highly selective pattern of localization, it can be extremely useful. That an antibody reacts with something that is unexpected should not limit its use, as long as the reaction it demonstrates can be precisely characterized with appropriate controls. I believe that it is perhaps irrelevant whether we call these reactions “specific” or “selective”; these reactions provide an important insight and, to me, they are just “cool.”^a