8/6/2019 Guidelines for Appropriate Controls for Immunohistochemistry

1/3

Editorial

Magic Peptides, Magic Antibodies:Guidelines for Appropriate Controls for

Immunohistochemistry

Since the earliest days of immunohistochemistry, thefield has been troubled by the tendency for these methodsto give spurious results. There are a number of reasonswhy various antibody preparations may provide variableand even erroneous results, and a series of steps one cantake to control for this. Unfortunately, these are not aswidely known as we feel they should be, even as we enterthe fourth decade in which these methods have beenwidely available to neuroanatomists. Hence, we providethis editorial to help guide aspiring authors through therocky shoals of immunolocalization in the nervous system.

WHAT AN ANTIBODY SEES

Antibodies are large protein molecules of about 150,000kD molecular weight. Native antibodies have two identicalbinding sites, although these can be dissociated from theremainder of the molecule and each other to form isolatedbinding domains (Fc fragments). Each antibody bindingregion recognizes a chemical epitope. The epitope may befound on any type of molecule (including entirely syntheticones, such as dinitrophenol, a classic experimental anti-gen). When the epitope is on a protein, it may be a partic-ular molecular configuration that is present in the nativeprotein (perhaps a site that is exposed only in a particulartertiary configuration) but is not present when the proteinhas been fixed with formaldehyde (or glutaraldehyde, oranother fixative). Conversely, an epitope may be exposed

in the fixed molecule that is not present in the nativeconfiguration.An antiserum is a preparation containing the immune

globulins from the serum (the supernatant after blood hasbeen allowed to clot and the cells have been centrifugedout), usually of an animal that has been immunized witha specific antigen. The animal will have a wide array ofimmune globulins with highly varied binding sites in itsblood, depending on its previous immune experience. Onlya very small proportion of the immune globulins in thatserum will be directed against the antigen used for immu-nization. These immune globulins will in turn recognizedifferent parts of the antigen. Once the antigen has beenfixed, many if not all of these antigenic sites will be de-

stroyed. In fact, when animals have been immunizedagainst small peptides, serum from many animals will notrecognize anything in the specific immunizing sequencewhen applied to fixed brain tissue. Among the antiserathat stain something in the brain, it is typical to find thatmany will stain spurious things, and often only a minoritywill stain the original molecule. The spurious stainingmay represent antibody that the animal just happened tobe making against an unrelated antigen (e.g., a virus theanimal encountered) and that cross-reacts with somethingin the brain, or it may represent antibody against theantigen used for immunization that cross-reacts withsomething else when applied to fixed tissue. When theantiserum is mixed with an excess of the immunizing

peptide, antibodies of the latter type will be adsorbed tothe peptide and will not stain. Antibodies of the formertype will continue to stain. Hence, the first step in authen-ticating a staining pattern is to carry out an adsorptioncontrol (see below).

A monoclonal antibody is a single clone of immune glob-ulins that is raised against a specific antigen. The immu-nized animal is killed, B cells that make immune globulinsare isolated, and these are fused with neoplastic B cellsfrom a myeloma tumor. The resulting cells are immortal(divide continuously), and each continuously dividing cellclone makes a single type of antibody molecule that rec-ognizes a single epitope. The antibodies made by eachclone are screened for recognition of the original antigen,and ones that do recognize the antigen are called mono-clonal antibodies against that antigen. Because this cloneof antibodies recognizes the antigen, it should always beblocked in adsorption testing (see below), although, if theantibody-secreting cells are raised in a host animal (in-stead of cultured in vitro), the ascites fluid that containsthe antibody may be contaminated by host immune glob-ulins. However, even the purest monoclonal antibodiesprovide no greater assurance than polyclonal antisera

DOI 10.1002/cne.10858

Published online the week of August 25, 2003 in Wiley InterScience(www.interscience.wiley.com).

THE JOURNAL OF COMPARATIVE NEUROLOGY 465:161163 (2003)

2003 WILEY-LISS, INC.

8/6/2019 Guidelines for Appropriate Controls for Immunohistochemistry

2/3

that the epitope that is recognized in the tissue belongs tothe original antigen. They require the same types of ad-ditional controls and careful interpretation.

Interestingly, many polyclonal sera may functionallyact as monoclonal preparations, because it is possible thatonly a single clone of antibodies (of the many present inthe serum) recognizes the fixed antigen in the tissue. The

probability that one or more clones in an animals serumwill recognize the fixed antigen is predicted by a Poissondistribution. Because the failure rate for producing anyclones that stain tissue is high (as many as 50% or more ofsera from immunized animals fail to stain anything infixed tissue and hence have zero clones capable of recog-nizing the antigen in fixed tissue), the Poisson distributionpredicts that most antisera that stain tissue probably onlyhave one or two clones of antibodies that actually providethe staining, and the rest of the antibodies in the serumare irrelevant, except for producing artifact. As a result,some investigators affinity purify antisera by runningthem over a column to which the antigen is fixed to asupport matrix. The nonbinding antibodies pass throughthe column, and then the clones of antibody that bind theantigen are released. Such antisera are less prone to ar-tifact (the adsorption test has essentially been performedin the purification step) and tend to have lower back-ground staining. However, quite often the highest affinityclones, which are difficult to elute, are lost in the process,and the other caveats that apply to monoclonal antibodiesstill are relevant.

WHAT IS AN ADEQUATE SET OFCONTROLS FOR AN ANTIBODY?

The key issue in designing controls for antibody stain-ing is this: Does the antibody actually reveal the antigenof interest, or is the binding an artifact? The gold standard

for deciding this issue is to stain tissue with and withoutthe antigen of interest. This can be accomplished by stain-ing tissue from a wild-type mouse and one in which theantigen of interest has been eliminated by transgenic en-gineering. If the antibody fails to stain the transgenictissue, then the antigen is present either on the moleculewhose gene was deleted or on some molecule that isdownstream of that protein in a metabolic or enzymaticpathway. It is still possible, therefore, for an antibody thatpasses this test to stain something other than what wasintended, but this is very unlikely. On the other hand, ifthe antibody fails this test, it cannot be used for immuno-localization in nervous tissue. There have been severalexamples of antibodies that have passed less stringenttests and yet failed the knockout test. Hence, this test is

recommended for all antibodies for antigens whose local-ization is not already well characterized. Unfortunately, itis directly applicable at this point only in mice, and stain-ing in tissues from other species may still represent otherantigens (not present in mice) and hence requires addi-tional controls.

If a knockout animal is not available for testing, a lessstringent but still reasonable set of tests can be carriedout. It is first necessary to perform a Western blot of thetissue of interest, to show that the antibody recognizesonly one antigen in the tissue and that it is of the appro-priate molecular weight. Although some antibodies thatrecognize two or more bands may identify the same targetmolecule in different states of posttranslational modifica-

tion (or in oligomeric forms), the use of such an antibodyrequires verification with antibodies against a differentpart of the same original molecule to demonstrate that thestaining is genuine.

The antibody should then be subjected to an adsorptiontest. It is necessary to perform this test carefully, andmany investigators are not aware of its nuances. The

investigator virtually never knows the absolute amount ofantigen that constitutes an excess. This is because theamount of antigen necessary to block, e.g., 50% of stainingcan be calculated only if the investigator knows the Kd ofthe antibody for both the tissue antigen and the solubleone (which may not be the same, insofar as the solubleantigen is usually not reacted with fixative). Some inves-tigators use a large amount of antigen (e.g., 10100M/ml of diluted antibody), but this may not be enough ifthe antiserum binds particularly tightly to the tissue an-tigen or loosely to the soluble antigen or if it is present inthe tissue in overwhelming abundance.

A more sensible approach to adsorption testing is todilute out the primary antiserum to the point where it juststains the sites of interest in the tissue. Most antisera willgo from dense staining to light or imperceptible stainingwith a dilution of about tenfold. If the investigator startswith the dilution that just obliterates staining and thendoubles the antibody concentration, this will provide aconcentration of antibody at which a 50% or greater re-duction in antibody availability will then eliminate stain-ing. Adding a large excess of antigen (e.g., 10 100 M/mlof diluted antibody) at this point should serve to blockstaining. It is important to note that, even for an antibodythat has been characterized extensively under other cir-cumstances, it is important to perform this test undertissue and fixation conditions identical to those used in theactual localization experiment.

It is common to accept adsorption controls as minimally

acceptable documentation in support of antibody specific-ity. However, there are circumstances in which they areclearly not sufficient. For example, if the antigen used forimmunization contains an amino acid sequence that isshared among a group of structurally related proteins orpeptides, it is possible that the antibody may recognizeany or all of the members of the group in staining tissue.In fact, instances of spurious cross-reactivity have beendocumented in cases in which peptides share just a fewamino acids in common or may even generate a tertiaryepitope that is similar. Hence, adsorbing the antibodyagainst known structurally related antigens can provideadditional security against cross-reactive staining.

Under some circumstances, such stringent controlsmight not be necessary. For example, in some cases, an

antiserum may be well characterized, and its stainingpattern may be well known in the species and tissue inwhich the experiment is performed. In that case, it mightnot be necessary to repeat all of the controls again. Underother circumstances, the antigen may be a large proteinthat is not readily available as an antigen for adsorptionstudies. In such cases, the best control that may be avail-able can be omitting the primary antiserum. In the end,the validity of the controls in a specific study will alwayshave to be decided by the referees of that paper, and it isalways wise to err on the side of using more controls thannecessary, rather than fewer. In particular, when apply-ing an antibody to a new species, especially one that is notclosely related to the one in which it was originally char-

162 EDITORIAL

8/6/2019 Guidelines for Appropriate Controls for Immunohistochemistry

3/3

acterized, it is almost always necessary to perform a fullset of controls, including immunoblots with tissue fromthat species.

MAGIC ANTIBODIES

A cornerstone of scientific publication is that the meth-

ods must contain adequate information to allow anotherinvestigator to repeat the experiment for the purpose ofverification. For investigators who use antibodies, thisentails a special duty. First, the manuscript must containadequate information concerning the antigen. Some com-panies are well-known for keeping the exact sequence ofthe immunizing peptide as a trade secret. Investigatorsshould be aware that such antisera are not fit for use inserious scientific work, nor will we allow publication ofstudies using such antibodies in the Journal of Compar-ative Neurology. The full sequence of the antigen must beavailable for the publication to be considered by this Jour-nal. This means that the author must either get thisinformation from the purveyor of the antibody or mustsequence the blocking peptide that is sold with the anti-body. Magic antibodies against magic antigens are notscience.

Second, full information must be given about the anti-serum that is used. Each antiserum is, in a sense, unique.When an animal is bled, there is a unique combination ofantibody clones expressed on that particular day. Thesame animal, bled on a different day, may express a verydifferent set of antibody clones. Hence, it is important toknow whenever possible both the code number for theanimal and the bleed. The catalog number for the anti-serum may not be sufficient. Because bleeds from rabbitsare limited in volume, a given manufacturer may run outof a particular bleed, and substitute another bleed fromthe same animal, or a bleed from a different animal,

without changing the catalog number. Quite often the newbleeds have not been checked to make sure that they stainthe same thing. In practice, it may be impossible to get full

information at this level from some manufacturers; theymay not have accurate records available. Hence, it is veryimportant for investigators to keep a small supply of what-ever antiserum they use for publications, in case questionsarise in the future because of a failure to replicate astaining pattern. This policy allows checking new antiseraagainst the old one for staining patterns. Similarly, if the

investigator prepares a new antiserum, it is important tomake samples of this available to other investigators whowant to replicate the work. Fortunately, this issue is not aproblem for monoclonal antisera, which by nature are thesame every time.

DONT BELIEVE EVERYTHING YOU SEE

The most important thing to understand about antibod-ies is that they are biological agents, not standard chem-ical reagents. Antibodies may bind to a wide variety ofantigens other than the one that they were raised torecognize, and there is no way to be sure that the patternthey stain really represents that antigen. Hence, it is

important to have sufficient controls to convince a skepti-cal observer that the staining pattern observed really doesrepresent the antigen, and wherever possible the resultsshould be checked with other techniques. Methods thatprovide verification of the expression of the same moleculein the same cells (e.g., in situ hybridization or single-cellRT-PCR) are particularly powerful. Ultimately, this willbe a matter for our referees to decide on a case-by-casebasis. In the end, all authors and readers should be awarethat antibody staining is fallible, and such staining shouldalways be considered to label antigen-like moleculesrather than staining the actual antigen.

CLIFFORD B. SAPER, MD, PhDPAUL E. SAWCHENKO, PhD

163EDITORIAL