Molecular Biology Reports 16: 149-153, 1992. 9 1992 Kluwer Academic Publishers. Printed in Belgium.
149
Methods of epitope mapping I. Pettersson Department of Medical Cell Genetics, Medical Nobel Institutet, Karolinska Institutet, Box 60 400, S-I04 01, Stockholm, Sweden
Key words: autoantibody, autoantigen, B-cell epitope
Ever since the discovery of autoantibodies efforts have been made to identify the corresponding antigens. It is now known that nucleic acids, proteins, and nucleic acid-protein complexes constitute targets for the connective tissue diseaseassociated autoantibodies. The protein moiety seems to be the prominent antigenic component in the nucleic acid-protein complexes [ 1, 2]. There are two approaches to identifying the parts of a protein that are seen by antibodies, starting from the protein and working with smaller and smaller protein fragments or starting at the peptide level working with synthetic peptides. The second alternative requires that the amino acid sequence is known. For most autoantigens, the amounts of purified protein required for either of the two approaches seemed discouraging. A report by Chambers and Keene in 1985 [3] demonstrated how this problem could be circumvented. A cDNA encoding the antigen could be cloned from an expression library using autoantibodies for selection. Armed with such c D N A clones, one can then proceed with either antigenic recombinant proteins or with synthetic peptides based on the amino acid sequence deduced from the DNA sequence. During the last few years, many investigations have dealt with the question of how autoantibodies react with specific regions of protein autoantigens. This is often referred to as epitope mapping although, strictly speaking, that level of resolution is seldom achieved. An epitope is the set of amino acid residues in a protein that are seen by a certain antibody [4]. Epitopes are often
divided in two conceptual categories and the number of amino acid residues involved is still a controversial area [5-7]. The linear or continuous epitopes can be defined with synthetic peptides 3-8 amino acid residues long [8]. In crystallographic studies conformational or discontinuous epitopes have from crystallographic studies been estimated as being composed of up to 15-22 residues [7].
Epitope mapping using recombinant proteins Many of the autoantigens were initially isolated from lambda gtl 1 libraries as beta-galactosidase fusion proteins, i.e. as chimeras between a protein encoded by the vector, beta-galactosidase, and the antigenic piece encoded by the cDNA. To identify the antigenic region(s), naturally occurring restriction sites were then used to generate a set of overlapping fragments. The corresponding recombinant proteins were assayed for remaining antigenicity. Exonuclease digestion of specific fragments was used for closer mapping of the borders of the smallest antigenic sequence. There were some inherent limitations to this method. The available restriction sites might be too few or unevenly distributed along the cDNA. It was also necessary to express fragments with different reading frames. One way of avoiding these problems was to generate random fragments from an antigen-expressing clone, redone these smaller fragments into an expression vector and isolate and characterize the new clones [9]. An alterna-
150 tive to these methods is to use PCR technology [10] to generate fragments with predetermined overlaps and in the same reading frame. The entire sequence of interest can thus be systematically explored. A further advantage is that all published clones are accessible through amplification of commercially available cDNA. More recently developed vector systems aim at a soluble product expressed as a small, cleavable fusion protein or expressed without any added sequences. This might seem more desirable from the point of view of epitope mapping but it comes at a cost. On the one hand, the disadvantages of fusion proteins are obvious. The foreign vectorencoded part could affect the antigenicity of the fused protein. It could also act as an antigen and create a high background in the assays for the autoantigen. This has been identified as a problem when screening autoimmune mouse sera with beta-galactosidase fusion proteins [ 11 ]. The possible assay methods and, thus, the spectrum of autoantibodies that can be detected are also restricted by the properties of the fusion protein. The high molecular weight beta-galactosidase fusion proteins are easily detected in crude E. coli lysates by immunoblotting but have to be extracted and solubilized in urea for use in ELI SA assays. They are not suited at all for assays requiring soluble antigens. On the other hand, an insoluble fusion protein is better than no protein at all. The amount of recombinant protein produced by vectors encoding no or only small fusion partners is highly variable from antigen to antigen. In some instances, the expression of the antigen seems to be detrimental to the plasmid or to the E. coli host. The desired protein is then obtained in very low yields or is mostly degraded or is not detectable at all. The 'expressibility' might even vary between different parts of the same antigen [ 12] and depend on exactly how the fragment is positioned in the sequence. Thus, the planned epitope mapping might not be possible to perform as the protein fragments cannot be identified or obtained in sufficient quantity. After overcoming problems of expression, the next question is whether or not the recombinant proteins actually are antigenic. On the whole, the
answer has been yes, the autoantigens produced in E. coli are good substitutes for the real eukaryotic proteins and their application in, e.g., diagnostic tests, is well underway [13, 14]. The conceptual rather than technical problems of epitope mapping are illustrated when different authors seemingly disagree in their localization of the antigenic regions. The La antigen has been mapped using both recombinant proteins and enzymatic degradation of authentic La protein. In a paper by Bini et al. [15], their results were compared with those of other investigators. Depending on the study, only carboxy-terminal fragments were antigenic, all fragments were antigenic or one amino-terminal and one carboxy-terminal antigenic region could be identified. Those latter two regions corresponded to 270 of the 353 residues in the clone under investigation. Such discrepancies could actually be expected if anti-La sera contain antibodies reacting with several more or less overlapping regions on the La protein. The exact position in the sequence of the various fragments then becomes important, especially for conformational epitopes, when results are compared. In addition, if different vectors are used, the fusion proteins differ with respect to solubility and size, even when they contain the same La protein fragment. How this influences the antigenicity has not been systematically explored. If protein conformation indeed is important for the antigenicity of a certain autoantigen, antibody recognition would depend on both the residues actually contacting the antibodies, the epitope(s), and the residues critical for maintaining the structural framework of the properly folded protein. These could be the same, partially overlapping or completely different residues. Loss of antigenicity e.g. after deletion of a certain section in an antigen-expressing clone would then have two alternative explanations. First, the deleted section encodes the amino acid residues of the epitopes. Second, the deletion has removed the residues that are important for the structure of the protein. The amino acid residues in the epitope(s) are still there but cannot assume the correct conformation. In the context of epitope mapping, it is important to be aware of these two possibilities when
151 the conclusions depend on the interpretation of the lack of antigenicity. The analysis of the U1 snRNP-associated 70K protein can be used as an example. A recombinant protein corresponding to residues 100-156 in the 70K sequence was recognized by 80~o of the anti-RNP sera in one study [ 16]. Using a naturally occurring restriction site within the D N A segment encoding this antigenic region clones containing the 5' or 3' half of the reactive clone were made. The expressed proteins were not antigenic. In another study, all anti-RNP sera reacted with 70K fusion proteins containing residues 99-167 [ 17 ]. Truncation of a couple of nucleotides at a time from either end of an antigen-producing clone would yield proteins that were recognized by fewer and fewer sera. The amino acid residues encoded by the restriction site or by the deleted nucleotides are important for the antigenicity of the 70K fragments, but are they all necessarily part of epitopes? If information about the likely protein structure is considered [18, 19], it can be argued that neither the clones made using the initially mentioned restriction site nor the truncated clones can produce a properly folded protein. The Sm antigen illustrates another type of problem, namely that the classically defined autoantibody specificity is heterogeneous. Antibodies in anti-Sm sera immunoprecipitate the (U) snRNPs and react with two (U)RNA-associated proteins, B/B' and D on immunoblots. Some of the antibodies react with only B/B' and others with only D. The major anti-Sm specificity is the antibody reacting with both proteins which also has been demonstrated with monoclonal antibodies derived from autoimmune mice [1]. Thus, when the Sm antigen is investigated, the results will depend on how that antigen is defined in the study - are epitopes present only on B/B' and only on D included or is one dealing only with the shared B / B ' - D epitopes?
Epitope mapping using synthetic peptides Synthetic peptides have several advantages over recombinant antigens as tools for epitope map-
ping. With peptides, it is possible to identify epitopes in the true immunological sense. Peptide synthesis is a controlled chemical process in contrast to the in vivo production of a recombinant protein. Peptides are stable and can be engineered to fit as antigens into different assay systems. The contribution of each residue can be systematically explored. The disadvantages are of both a practical and conceptual nature. The perceived technological or economical threshold often limits the use of synthetic peptides to a follow-up of antigenic regions identified with recombinant antigens. An educated guess is made by man or computer and a limited number of peptides are synthesized and compared with the recombinant or native autoantigen in assays. Depending on the antigen, peptides have been found to be either good [20, 21] or bad [17, 22] substitutes for larger antigenic protein fragments. In the latter case, it is tempting to conclude that further experiments with peptides is a waste of time and money. This is not necessarily true, the failure could be due to the bias introduced by selecting peptides based on previous mapping data. Synthetic peptides prove their worth when used as an independent method of epitope mapping. The entire sequence of the antigen can be represented by sets of consecutive or overlapping peptides which are then assayed for antigenicity. This peptide approach is possible to pursue even without access to a peptide synthesizer. By using commercially available kits based on small amounts of resins enclosed in polypropylene tissue or polyethylene pins in the format of a microtiter plate, many peptides can be produced simultaneously and in quantities sufficient for epitope mapping [23, 24]. The complete sequence of the Ro 60 kD protein was assayed in this manner and the antigenic peptides revealed a previously unnoticed homology with a viral protein [25]. The major limitation of the pepscan or pin strategy is that only relatively short peptides, 6-10 residues, are made, which obviously works against the detection of conformational epitopes. Short peptides are also more likely to be recognized by crossreactive antibodies [8]. As autoimmune sera often contain elevated levels of immunoglobulins, this could cause a high back-
152 ground reaction against all peptides e.g. in an ELISA assay. In that case, the specific recognition of peptides corresponding to epitopes might be difficult to distinguish from the background. Even a strong reaction with a single peptide might be misleading. The amino acid composition of that specific peptide, like the presence of many charged residues, could promote binding of antibodies not dependent on the antigen combining site. The significance of the reaction with a particular peptide can only be evaluated after testing a large number of control sera negative for the autoantibody in question. Longer peptides, 2030 residues, can be routinely made with a peptide synthesizer which allows more stringent assay conditions, and thus reduces the problems mentioned above. An antigenic peptide of this type, derived from the U1 snRNP-associated A protein sequence, was shown to be recognized by virtually all U1 A-reactive MCTD sera [26]. When the D protein of the Sm antigen was investigated in the same manner, two peptides correlated well with D protein recognition on immunoblots but none of the peptides tested was seen by anti-Sm monoclonal antibodies of the crossreactive type [27]. Interestingly, for both proteins, the peptides were located outside the antigenic regions previously indentified with recombinant antigens. This observation has also been made after immunization with other proteins and underscores that different methods will be biased for detection of different antibodies in a polyclonal response [28 ]. Results from the recombinant protein and synthetic peptide approaches could be expected to complement but not necessarily confirm each other.
Concluding remarks It is important to remember that the merits of the different approaches to epitope mapping should be judged against the purpose of the study. A peptide specifically recognized by nearly all sera containing a certain autoimmune specificity [20, 21, 26], would most likely be selected for the detection of the anti-linear/continuous epitope
fraction among those autoantibodies and could be highly useful for diagnostic purposes. This is true even if a majority of the antibodies were directed against conformational/discontinuous protein epitopes. If the purpose is to study the induction or maintenance of the autoimmune response, one would also have to look at and account for the conformation-dependent autoantibodies. There is also the possibility that some pathogenic autoantibodies could constitute a small fraction requiring the complete nucleic acidprotein complex as an antigen. In that case, the 'pathogenic' epitope would not be identified nor mapped using the techniques discussed in this review.
Acknowledgements This work was supported by a grant from the Axel och Margeret Ax:son Johnson Foundation and by grant B92-16X-09927-01A from the Swedish Medical Research Council.
References 1. Tan EM (1989) Adv. Immunol.49:93-151 2. BautzEKF, KaldenJR, HommaM, ShulmanLE & Tan EM (Eds)(1991) Molecularand CellBiologyof Autoantibodies and AutoimmunityMol. Biol. Rep. 15:95-216 3. ChambersJC & KeeneJD (1985) Proc. Natl. Acad. Sci. USA 74:5463-5467 4. Jerne NK (1960) Annu. Rev. Microbiol. 14: 341-358. 5. Van RegenmortelMHV, Briand JP, Muller S & Plau6 S (1988) Syntheticpolypeptides as antigens. Elsevier Scientific Publishers, Amsterdam 6. Van RegenmortelMHV (1989) Immunol.Today 10: 266272 7. LaverWG, Air GM, WebsterRG & Smith-GillSJ (1990) Cell 61:553-556 8. GeysenHM (1985) Immunol.Today 6:364-369 9. Habets WJ, Sillekens PTG, Hoet MH, McAllister G, Lerner MR & van VenrooijWJ (1989) Proc. Natl. Acad. Sci. USA 86:4674-4678 10. ErlichHA (Ed) (1989) PCR Technology.StocktonPress, New York 11. Pisetsky DS & Groudier JP (1989) J. Immunol. 143: 3609-3613 12. FrorathB, ScanariniM, Netter HJ, AbneyCC, Liedvogel B, LakomekHJ & NorthemannW (1991) BioTechniques 11:364-371
153 13. Paxton H, Bendele T, O'Connor L & Haynes D (1990) Clin. Chem. 36:792-797 14. Seefig HP, Ehrfeld H, Schroeter H, Heim C & Renz M (1991) J. Immunol. Methods 143:11-24 15. Bini K, Chu J-L, Okolo C & Elkon K (1990) J. Clin. Invest. 85:325-333 16. Cram DS, Fisicaro N, Coppel RL, Whittingham S & Harrison LC (1990) J. Immunol. 145:630-635 17. Netter HJ, Guldner HH, Szotstecki C & Will H (1990) Scand. J. Immunol. 32:163-176 18. Query CC, Bentley RC & Keene JD (1989) Cell 57: 89101 19. Nagai K, Oubridge C, Jessen TH, Li J & Evans PR (1990) Nature 348:515-520 20. Elkon KB, Skelly S, Parnassa AP, Danho W, Weissbach H & Brot N (1986) Proc. Natl. Acad. Sci. USA 83: 7419-7423
21. Hiness JJ, Danho W & Elkon KB (1991) Arthritis Rheum. 34:572-579 22. St Clair EW, Kenan D, Burch JA, Keene JD & Pisetsky DS (1990) J. Immunol. 144:3868-3876 23. Houghten RA (1985) Proc. Natl. Acad. Sci. USA 82: 5131-5135 24. Geysen HM, Meloen RH & Barteling SJ (1984) Proc. Natl. Acad. Sci. USA 81:3998-4002 25. Scofield RH & Harley JB (1991) Proc. Natl. Acad. Sci. USA 88:3343-3347 26. Barakat S, Briand JP, AbouafN, Van Regenmortel MHV & Muller S (1991) Clin. exp. Immunol. 86:71-78 27. Barakat S, Briand JP, Weber JC, Van Regenmortel MHV & Muller S (1990) Clin. Exp. Immunol. 81:256-262 28. Berzotsky JA (1985) Science 229:932-940
149
Methods of epitope mapping I. Pettersson Department of Medical Cell Genetics, Medical Nobel Institutet, Karolinska Institutet, Box 60 400, S-I04 01, Stockholm, Sweden
Key words: autoantibody, autoantigen, B-cell epitope
Ever since the discovery of autoantibodies efforts have been made to identify the corresponding antigens. It is now known that nucleic acids, proteins, and nucleic acid-protein complexes constitute targets for the connective tissue diseaseassociated autoantibodies. The protein moiety seems to be the prominent antigenic component in the nucleic acid-protein complexes [ 1, 2]. There are two approaches to identifying the parts of a protein that are seen by antibodies, starting from the protein and working with smaller and smaller protein fragments or starting at the peptide level working with synthetic peptides. The second alternative requires that the amino acid sequence is known. For most autoantigens, the amounts of purified protein required for either of the two approaches seemed discouraging. A report by Chambers and Keene in 1985 [3] demonstrated how this problem could be circumvented. A cDNA encoding the antigen could be cloned from an expression library using autoantibodies for selection. Armed with such c D N A clones, one can then proceed with either antigenic recombinant proteins or with synthetic peptides based on the amino acid sequence deduced from the DNA sequence. During the last few years, many investigations have dealt with the question of how autoantibodies react with specific regions of protein autoantigens. This is often referred to as epitope mapping although, strictly speaking, that level of resolution is seldom achieved. An epitope is the set of amino acid residues in a protein that are seen by a certain antibody [4]. Epitopes are often
divided in two conceptual categories and the number of amino acid residues involved is still a controversial area [5-7]. The linear or continuous epitopes can be defined with synthetic peptides 3-8 amino acid residues long [8]. In crystallographic studies conformational or discontinuous epitopes have from crystallographic studies been estimated as being composed of up to 15-22 residues [7].
Epitope mapping using recombinant proteins Many of the autoantigens were initially isolated from lambda gtl 1 libraries as beta-galactosidase fusion proteins, i.e. as chimeras between a protein encoded by the vector, beta-galactosidase, and the antigenic piece encoded by the cDNA. To identify the antigenic region(s), naturally occurring restriction sites were then used to generate a set of overlapping fragments. The corresponding recombinant proteins were assayed for remaining antigenicity. Exonuclease digestion of specific fragments was used for closer mapping of the borders of the smallest antigenic sequence. There were some inherent limitations to this method. The available restriction sites might be too few or unevenly distributed along the cDNA. It was also necessary to express fragments with different reading frames. One way of avoiding these problems was to generate random fragments from an antigen-expressing clone, redone these smaller fragments into an expression vector and isolate and characterize the new clones [9]. An alterna-
150 tive to these methods is to use PCR technology [10] to generate fragments with predetermined overlaps and in the same reading frame. The entire sequence of interest can thus be systematically explored. A further advantage is that all published clones are accessible through amplification of commercially available cDNA. More recently developed vector systems aim at a soluble product expressed as a small, cleavable fusion protein or expressed without any added sequences. This might seem more desirable from the point of view of epitope mapping but it comes at a cost. On the one hand, the disadvantages of fusion proteins are obvious. The foreign vectorencoded part could affect the antigenicity of the fused protein. It could also act as an antigen and create a high background in the assays for the autoantigen. This has been identified as a problem when screening autoimmune mouse sera with beta-galactosidase fusion proteins [ 11 ]. The possible assay methods and, thus, the spectrum of autoantibodies that can be detected are also restricted by the properties of the fusion protein. The high molecular weight beta-galactosidase fusion proteins are easily detected in crude E. coli lysates by immunoblotting but have to be extracted and solubilized in urea for use in ELI SA assays. They are not suited at all for assays requiring soluble antigens. On the other hand, an insoluble fusion protein is better than no protein at all. The amount of recombinant protein produced by vectors encoding no or only small fusion partners is highly variable from antigen to antigen. In some instances, the expression of the antigen seems to be detrimental to the plasmid or to the E. coli host. The desired protein is then obtained in very low yields or is mostly degraded or is not detectable at all. The 'expressibility' might even vary between different parts of the same antigen [ 12] and depend on exactly how the fragment is positioned in the sequence. Thus, the planned epitope mapping might not be possible to perform as the protein fragments cannot be identified or obtained in sufficient quantity. After overcoming problems of expression, the next question is whether or not the recombinant proteins actually are antigenic. On the whole, the
answer has been yes, the autoantigens produced in E. coli are good substitutes for the real eukaryotic proteins and their application in, e.g., diagnostic tests, is well underway [13, 14]. The conceptual rather than technical problems of epitope mapping are illustrated when different authors seemingly disagree in their localization of the antigenic regions. The La antigen has been mapped using both recombinant proteins and enzymatic degradation of authentic La protein. In a paper by Bini et al. [15], their results were compared with those of other investigators. Depending on the study, only carboxy-terminal fragments were antigenic, all fragments were antigenic or one amino-terminal and one carboxy-terminal antigenic region could be identified. Those latter two regions corresponded to 270 of the 353 residues in the clone under investigation. Such discrepancies could actually be expected if anti-La sera contain antibodies reacting with several more or less overlapping regions on the La protein. The exact position in the sequence of the various fragments then becomes important, especially for conformational epitopes, when results are compared. In addition, if different vectors are used, the fusion proteins differ with respect to solubility and size, even when they contain the same La protein fragment. How this influences the antigenicity has not been systematically explored. If protein conformation indeed is important for the antigenicity of a certain autoantigen, antibody recognition would depend on both the residues actually contacting the antibodies, the epitope(s), and the residues critical for maintaining the structural framework of the properly folded protein. These could be the same, partially overlapping or completely different residues. Loss of antigenicity e.g. after deletion of a certain section in an antigen-expressing clone would then have two alternative explanations. First, the deleted section encodes the amino acid residues of the epitopes. Second, the deletion has removed the residues that are important for the structure of the protein. The amino acid residues in the epitope(s) are still there but cannot assume the correct conformation. In the context of epitope mapping, it is important to be aware of these two possibilities when
151 the conclusions depend on the interpretation of the lack of antigenicity. The analysis of the U1 snRNP-associated 70K protein can be used as an example. A recombinant protein corresponding to residues 100-156 in the 70K sequence was recognized by 80~o of the anti-RNP sera in one study [ 16]. Using a naturally occurring restriction site within the D N A segment encoding this antigenic region clones containing the 5' or 3' half of the reactive clone were made. The expressed proteins were not antigenic. In another study, all anti-RNP sera reacted with 70K fusion proteins containing residues 99-167 [ 17 ]. Truncation of a couple of nucleotides at a time from either end of an antigen-producing clone would yield proteins that were recognized by fewer and fewer sera. The amino acid residues encoded by the restriction site or by the deleted nucleotides are important for the antigenicity of the 70K fragments, but are they all necessarily part of epitopes? If information about the likely protein structure is considered [18, 19], it can be argued that neither the clones made using the initially mentioned restriction site nor the truncated clones can produce a properly folded protein. The Sm antigen illustrates another type of problem, namely that the classically defined autoantibody specificity is heterogeneous. Antibodies in anti-Sm sera immunoprecipitate the (U) snRNPs and react with two (U)RNA-associated proteins, B/B' and D on immunoblots. Some of the antibodies react with only B/B' and others with only D. The major anti-Sm specificity is the antibody reacting with both proteins which also has been demonstrated with monoclonal antibodies derived from autoimmune mice [1]. Thus, when the Sm antigen is investigated, the results will depend on how that antigen is defined in the study - are epitopes present only on B/B' and only on D included or is one dealing only with the shared B / B ' - D epitopes?
Epitope mapping using synthetic peptides Synthetic peptides have several advantages over recombinant antigens as tools for epitope map-
ping. With peptides, it is possible to identify epitopes in the true immunological sense. Peptide synthesis is a controlled chemical process in contrast to the in vivo production of a recombinant protein. Peptides are stable and can be engineered to fit as antigens into different assay systems. The contribution of each residue can be systematically explored. The disadvantages are of both a practical and conceptual nature. The perceived technological or economical threshold often limits the use of synthetic peptides to a follow-up of antigenic regions identified with recombinant antigens. An educated guess is made by man or computer and a limited number of peptides are synthesized and compared with the recombinant or native autoantigen in assays. Depending on the antigen, peptides have been found to be either good [20, 21] or bad [17, 22] substitutes for larger antigenic protein fragments. In the latter case, it is tempting to conclude that further experiments with peptides is a waste of time and money. This is not necessarily true, the failure could be due to the bias introduced by selecting peptides based on previous mapping data. Synthetic peptides prove their worth when used as an independent method of epitope mapping. The entire sequence of the antigen can be represented by sets of consecutive or overlapping peptides which are then assayed for antigenicity. This peptide approach is possible to pursue even without access to a peptide synthesizer. By using commercially available kits based on small amounts of resins enclosed in polypropylene tissue or polyethylene pins in the format of a microtiter plate, many peptides can be produced simultaneously and in quantities sufficient for epitope mapping [23, 24]. The complete sequence of the Ro 60 kD protein was assayed in this manner and the antigenic peptides revealed a previously unnoticed homology with a viral protein [25]. The major limitation of the pepscan or pin strategy is that only relatively short peptides, 6-10 residues, are made, which obviously works against the detection of conformational epitopes. Short peptides are also more likely to be recognized by crossreactive antibodies [8]. As autoimmune sera often contain elevated levels of immunoglobulins, this could cause a high back-
152 ground reaction against all peptides e.g. in an ELISA assay. In that case, the specific recognition of peptides corresponding to epitopes might be difficult to distinguish from the background. Even a strong reaction with a single peptide might be misleading. The amino acid composition of that specific peptide, like the presence of many charged residues, could promote binding of antibodies not dependent on the antigen combining site. The significance of the reaction with a particular peptide can only be evaluated after testing a large number of control sera negative for the autoantibody in question. Longer peptides, 2030 residues, can be routinely made with a peptide synthesizer which allows more stringent assay conditions, and thus reduces the problems mentioned above. An antigenic peptide of this type, derived from the U1 snRNP-associated A protein sequence, was shown to be recognized by virtually all U1 A-reactive MCTD sera [26]. When the D protein of the Sm antigen was investigated in the same manner, two peptides correlated well with D protein recognition on immunoblots but none of the peptides tested was seen by anti-Sm monoclonal antibodies of the crossreactive type [27]. Interestingly, for both proteins, the peptides were located outside the antigenic regions previously indentified with recombinant antigens. This observation has also been made after immunization with other proteins and underscores that different methods will be biased for detection of different antibodies in a polyclonal response [28 ]. Results from the recombinant protein and synthetic peptide approaches could be expected to complement but not necessarily confirm each other.
Concluding remarks It is important to remember that the merits of the different approaches to epitope mapping should be judged against the purpose of the study. A peptide specifically recognized by nearly all sera containing a certain autoimmune specificity [20, 21, 26], would most likely be selected for the detection of the anti-linear/continuous epitope
fraction among those autoantibodies and could be highly useful for diagnostic purposes. This is true even if a majority of the antibodies were directed against conformational/discontinuous protein epitopes. If the purpose is to study the induction or maintenance of the autoimmune response, one would also have to look at and account for the conformation-dependent autoantibodies. There is also the possibility that some pathogenic autoantibodies could constitute a small fraction requiring the complete nucleic acidprotein complex as an antigen. In that case, the 'pathogenic' epitope would not be identified nor mapped using the techniques discussed in this review.
Acknowledgements This work was supported by a grant from the Axel och Margeret Ax:son Johnson Foundation and by grant B92-16X-09927-01A from the Swedish Medical Research Council.
References 1. Tan EM (1989) Adv. Immunol.49:93-151 2. BautzEKF, KaldenJR, HommaM, ShulmanLE & Tan EM (Eds)(1991) Molecularand CellBiologyof Autoantibodies and AutoimmunityMol. Biol. Rep. 15:95-216 3. ChambersJC & KeeneJD (1985) Proc. Natl. Acad. Sci. USA 74:5463-5467 4. Jerne NK (1960) Annu. Rev. Microbiol. 14: 341-358. 5. Van RegenmortelMHV, Briand JP, Muller S & Plau6 S (1988) Syntheticpolypeptides as antigens. Elsevier Scientific Publishers, Amsterdam 6. Van RegenmortelMHV (1989) Immunol.Today 10: 266272 7. LaverWG, Air GM, WebsterRG & Smith-GillSJ (1990) Cell 61:553-556 8. GeysenHM (1985) Immunol.Today 6:364-369 9. Habets WJ, Sillekens PTG, Hoet MH, McAllister G, Lerner MR & van VenrooijWJ (1989) Proc. Natl. Acad. Sci. USA 86:4674-4678 10. ErlichHA (Ed) (1989) PCR Technology.StocktonPress, New York 11. Pisetsky DS & Groudier JP (1989) J. Immunol. 143: 3609-3613 12. FrorathB, ScanariniM, Netter HJ, AbneyCC, Liedvogel B, LakomekHJ & NorthemannW (1991) BioTechniques 11:364-371
153 13. Paxton H, Bendele T, O'Connor L & Haynes D (1990) Clin. Chem. 36:792-797 14. Seefig HP, Ehrfeld H, Schroeter H, Heim C & Renz M (1991) J. Immunol. Methods 143:11-24 15. Bini K, Chu J-L, Okolo C & Elkon K (1990) J. Clin. Invest. 85:325-333 16. Cram DS, Fisicaro N, Coppel RL, Whittingham S & Harrison LC (1990) J. Immunol. 145:630-635 17. Netter HJ, Guldner HH, Szotstecki C & Will H (1990) Scand. J. Immunol. 32:163-176 18. Query CC, Bentley RC & Keene JD (1989) Cell 57: 89101 19. Nagai K, Oubridge C, Jessen TH, Li J & Evans PR (1990) Nature 348:515-520 20. Elkon KB, Skelly S, Parnassa AP, Danho W, Weissbach H & Brot N (1986) Proc. Natl. Acad. Sci. USA 83: 7419-7423
21. Hiness JJ, Danho W & Elkon KB (1991) Arthritis Rheum. 34:572-579 22. St Clair EW, Kenan D, Burch JA, Keene JD & Pisetsky DS (1990) J. Immunol. 144:3868-3876 23. Houghten RA (1985) Proc. Natl. Acad. Sci. USA 82: 5131-5135 24. Geysen HM, Meloen RH & Barteling SJ (1984) Proc. Natl. Acad. Sci. USA 81:3998-4002 25. Scofield RH & Harley JB (1991) Proc. Natl. Acad. Sci. USA 88:3343-3347 26. Barakat S, Briand JP, AbouafN, Van Regenmortel MHV & Muller S (1991) Clin. exp. Immunol. 86:71-78 27. Barakat S, Briand JP, Weber JC, Van Regenmortel MHV & Muller S (1990) Clin. Exp. Immunol. 81:256-262 28. Berzotsky JA (1985) Science 229:932-940
