Clin Bim'hem. Vol. 25, pp. 187-191. 1992 Printed in the USA. All rights reserved.
0009-9120/92 $5.00 + .00 Copyright ¢ 1992 The Canadian Society of Clinical Chemists.
HLA Molecules in Autoimmune Diseases WILLIAM E. BRAUN Department of Hypertension and Nephrology, Cleveland Clinic Foundation, 9500 Euclid Avenue, Cleveland, OH 44195, USA The association of certain autoimmune diseases with HLA molecules is being refined through the use of sequence-specific oligonucleotide probes and amino acid sequencing, together with continuing elucidation of the functional features of HLA molecules derived from the milestone description by Bjorkman of the HLA molecular structure. The association of insulindependent diabetes mellitus and HLA began with weak associations of Class I antigens (B8 and B15) and progressed to Class II antigens (DR3 and DR4), then to subtypes of DR4 (Dw4, 10, and 14), and now to DQ molecules including the absence of aspartic acid at position 57 of the DQ beta chain and the presence of arginine at position 52 of the DQ alpha chain. In rheumatoid arthritis (RA) the HLA antigen association remains with certain Class II molecules of the DR series (DR4 and DR1) that share amino acid sequences with a restricted number of other DR antigens seen in RA, as well as a segment of the gp 110 protein of the Epstein-Barr virus. Although ankylosing spondylitis has a strong association with the Class I antigen B27, that association is not explained by any of the B27 subtypes defined by monoclonal antibodies, by the eight variable amino acids in B27 subtypes, or by the two unique amino acids on B27. The remarkable antibody cross-reactivity among lymphocytes bearing B27, a synthetic peptide sequence (63-84) of B27, and the 188-193 sequence of K. pneumoniae nitrogenase has provided strong support for molecular mimicry being an important mechanism in the association of HLA molecules with disease. As the molecular features related to HLA and disease are being discovered, elucidating the all-important functional roles that they have in actually producing the disease will be an even greater challenge.
KEY WORDS: HLA antigens; diabetes mellitus; rheumatoid arthritis; ankylosing spondylitis. Introduction
he HLA system of antigens is controlled by genes
located on the short arm of the sixth chromosome T in a region known as the major histocompatibility
complex (MHC) that has well-defined counterparts in animals such as the mouse, rat, guinea pig, pig, dog, and horse. The HLA antigens have been divided into two major classes of molecules: Class I and Class II. Class I and Class II molecules differ in their biochemistry, cellular distribution, expression, and function. A detailed examination of these features is beyond the scope of this review and can be found in monographs and textbooks (1,2). Certain major distinctions between Class I and Class II molecules are:
Manuscript received: July 31, 1991; revised: January 28, 1992; accepted: January 31, 1992. CLINICAL BIOCHEMISTRY, VOLUME 25, JUNE 1992
Class I molecules consist of a 44,000 kD heavy chain noncovalently bound to beta-2 microglobulin with a weight of 11,700 kD, whereas Class II molecules consist of a 33,000 kD alpha chain that is relatively invariant and a 29,000 kD beta chain that is responsible for most of the polymorphisms; Class I molecules are quite ubiquitous in their distribution and appear on the surface of all nucleated cells, whereas Class II molecules are restricted to B-lymphocytes, monocytes, macrophages, dendritic cells, Langerhans cells of the skin, and activated T-cells; Class I and Class II molecules differ not only in their constitutive expression but also in their inducible expression (3); although Class I and Class II molecules are assembled in the endoplasmic reticulum, they take different intracellular pathways, with Class I molecules transporting endogenous antigens such as viral antigens through the Golgi apparatus en route to the cell surface, whereas Class II molecules transport endocytosed exogenous antigens such as bacterial antigens through the lysosomal/endosomal system; and, Class I molecules interact primarily with cytotoxic (CD8) T-cells, whereas Class II molecules interact with helper (CD4) T-lymphocytes (4). The techniques for identifying both Class I and Class II HLA antigens have been dominated by serologic methods based on complement-dependent cytotoxicity. In a limited number of laboratories, Class II antigens have also been detected by cellular techniques based on the use of homozygous typing cells (HTC) in mixed lymphocyte culture reactions. These basic serologic and cellular techniques have been d r a m a t i c a l l y improved by the use of sequencespecific oligonucleotide probes (SSOP) and amino acid sequencing. The highly polymorphic HLA system now includes more than 180 alleles defined by one or more techniques (5-7). Crystallographic structural analysis of Class I molecules have shown a configuration consisting of " . . . two sets of structurally homologous domains; one set being the membrane-proximal ~3 and ~2m domains, and the other set being the membranedistal ~1 and a2 d o m a i n s . . , the ~3 and ~2~ domains are folded into ~-sandwich structures resembling Ig constant r e g i o n s . . , the al and ~2 domains sit on top of the Ig-like domains forming a platform consisting of an eight-stranded ~-pleated sheet topped by two long m-helices that lie diagonally across the sheet 187
BRAUN
. . . the binding site for antigenic peptides is a deep cleft located between the two long a - h e l i c e s . . . " (4). The basic configuration of Class II molecules is thought to be quite similar to t h a t of Class I (8). HLA molecules and T-cell receptors permit one's i m m u n e system to distinguish self from nonself through a selection process occurring in the t h y m u s (4,9). In carrying out this general function, it appears t h a t subtle differences in the amino acid composition of HLA molecules, their varying capabilities of binding different peptide antigens in the interhelical groove, their capability of presenting these antigens to and interacting with the repertoire of T-cell receptors, as well as their aberrant expression, may influence the occurrence of autoimmune diseases. By means of basic serologic and cellular techniques, HLA alleles have been shown to be significantly associated with a number of diseases among approximately 600 investigated (10,11). Some of these associations were very strong, such as HLADR2 in narcolepsy, HLA-B27 in ankylosing spondylitis and sero-negative arthropathies, and HLADR3 with DR4 in diabetes mellitus. It is important to note t h a t these are all associations between certain HLA alleles and a particular disease, and not genetic linkage. At this point it is necessary to restate the definition of association as the nonrandom occurrence of two genetically separate traits in a population (e.g., HLA-DR3 and DR4 with diabetes mellitus), linkage as the occurrence of two loci on the same chromosome and sufficiently close together so t h a t something less t h a n completely independent assortment takes place (e.g., idiopathic hemochromatosis with HLA-A), and linkage disequilibrium as the occurrence of two alleles together more frequently t h a n would be expected on the basis of their individual gene frequencies (e.g., HLA-A1, B8, DR3, and DQ2). Associations are tested in unrelated individuals, whereas linkage studies require informative families. As provocative as m a n y of these studies were, there were nevertheless fundamental weaknesses with the serologic and cellular techniques available at t h a t time t h a t rendered them simply incapable of identifying differences in DNA and amino acid sequences. Because of such relatively insensitive techniques, as well as other factors, m a n y individuals with a particular disease appeared to lack the HLA susceptibility antigen, and conversely m a n y people with the HLA susceptibility antigen were free of the relevant disease. In this review, the primary associations between serologically and cellularly defined HLA antigens and certain autoimmune diseases will be considered along with the refinements of those HLA antigen associations based on evolving molecular genetic techniques. The steps in the molecular genetic approach to HLA and disease associations have been defined by Nepom and Erlich (12). Nepom and Erlich (12) have also described three general ways in which Class II structural polymor188
phism may mediate disease susceptibility: "Polymorphic residues in the Class II molecule might differentially bind a putative autoantigen peptide and present it to the responding T-lymphocytes. Alternatively, the spectrum of expressed T-cell receptors m i g h t be different in individuals with different Class II genotypes. Autoreactive T-cell receptors capable of recognizing an autoantigen Class II complex would more likely be selected by thymic epithelium which expressed the nonsusceptible Class II antigen. The third immunologic mechanism invoked to explain HLA disease associations is molecular mimicry" (11). The presence of Class I residues shared with a bacterial antigen (e.g., B27 and Klebsiella pneumoniae nitrogenase) or Class II residues shared with a viral pathogen could result in tolerance to such microbial epitopes leading to an altered immune response (12). "Alternatively such homologous regions could serve as an autoantigenic target with an immune response to the (bacterium or) virus leading to attack of self-cells expressing relevant Class (I or) II products" (12). Insulin-dependent diabetes mellitus (IDDM) and HLA Before the Class II HLA antigens (DR, DQ, and DP) were identified, the susceptibility to IDDM was associated with Class I molecules controlled by the B-locus, n a m e l y HLA-B8, and HLA-B15 (11) as shown in Table 1. Conversely, a resistance to IDDM was associated with HLA-B7 (11). When the serologic identification of Class II HLA-DR antigens was established, even stronger associations were found TABLE 1
Insulin-Dependent Diabetes Mellitus (IDDM): Association with HLA-DQ Beta Aspartic Acid57-Negative and Alpha Arginine52-Positive Haplotypea • Initial associations appeared with Class I antigens B8 and B15 that were in linkage disequilibrium with certain Class II HLA-DR antigens (DR3 and DR4). • DR3 and/or DR4 occurred in 90% of IDDM. • Of the five Dw subtypes of DR4, only three (Dw4, 10, and 14) were seen in IDDM. • The three Dw subtypes of IDDM were in linkage disequilibrium with DQw 3.2 (DQw8). • Key differences in the DQw8 allele (that is closely associated with IDDM) and a neutral allele DQw7 are four amino acid substitutions, including aspartic acid at position 57 of the DQ beta chain. • The best current marker for susceptibility to IDDM includes not only the absence of aspartic acid at position 57 of the DQ beta chain, but also the presence of arginine at position 52 of the DQ alpha chain. • Although the Asps7 absence and Arg 52 presence are necessary conditions for IDDM, they are not sufficient for its occurrence. aReferences (10-15). CLINICAL BIOCHEMISTRY, VOLUME 25, JUNE 1992
HLA AND DISEASE
between susceptibility to IDDM and HLA-DR3 and DR4, and the heterozygote state DR3/DR4, whereas resistance to IDDM was associated with HLA-DR2 (11). Because there is linkage disequilibrium between B8 and DR3, B15 and DR4, and B7 and DR2, it was not surprising to see the weak associations initially described with the Class I B-locus antigens assume greater strength with the Class II DR antigens. For example, the frequency of DR3 was 48% in 265 Caucasian diabetic patients compared to 23% of 776 controls (11) giving a relative risk (RR) of 3.04. DR4 occurred in 65% of 265 Caucasian diabetic patients compared to 30% in controls (RR = 4.28). For DR3/DR4 heterozygotes the RR was increased to 14.26. Resistance to IDDM (RR = 0.17) was noted for DR2 that occurred in only 4% of 265 Caucasian diabetic patients compared to the higher frequency of 26% in 776 controls (11). The strong associations of IDDM with Class II molecules were important findings that focused subsequent DNA studies on the Class II HLA molecules. In these studies it soon became apparent that susceptibility and resistance to IDDM were even more strongly related to gene products of the HLADQ locus that defined subsets of the susceptibilityrelated antigens DR3 and DR4, as well as subsets of the resistance-related DR2 antigen. For example, the DR4 subtype DQBI*0302 (DQw3.2) is present in approximately 95% of DR4-positive haplotypes of IDDM Caucasians, but it is also found in 6 0 - 7 0 % of individuals with DR4-positive haplotypes but no IDDM, and in about 27% of all Caucasians (12). DQw3.2 differs from a closely related neutral allele DQw3.1 by j u s t four amino acids (12). One key amino acid difference that has received a great deal of attention is at position 57 (12-14). The DQw3.2 allele (currently named DQw8) that lacks aspartic acid at position 57 has a gene frequency of 35.7% in IDDM compared to 10.1% in normal subjects (14), the difference being highly significant (p < 0.001). However, the simple presence or absence of aspartic acid alone is clearly not a complete explanation for resistance or susceptibility, respectively, to IDDM. Although a protective effect was associated with the DQwl.2 allele bearing aspartic acid at position 57, no protection was associated with other aspartic acid-bearing alleles such as D Q w l . 1 2 , DQw4, and DQw7 (14). Moreover, HLA-DQw.1 in the homozygous state was also protective even though it has valine rather than aspartic acid at position 57 of the DQ beta chain (14). Two reasons have been offered for the lack of utility of aspartic acid testing by itself as a predictor of risk for IDDM: 1) the HLA-DR7/DQw2 haplotype that lacks an aspartic acid at position 57 of the DQ beta chain but is protective against IDDM would erroneously classify approximately 10% of Caucasians in a high-risk category; 2) the HLA-DQw7 allele that contains aspartic acid fails to confer protection because 81% of diabetic patients heterozygous for an aspartic acidcontaining allele had the aspartic acid-positive DQw7 allele (14). CLINICAL BIOCHEMISTRY, VOLUME 25, JUNE 1992
Further resolution of this dilemma lies in the fact that an HLA-DQ~ aspartic acid 57-negative allele when found in conjunction with an HLA-DQ~ arginine 52-positive allele conferred an even stronger susceptibility to IDDM, whereas their absence constituted virtually absolute protection (15). This finding emphasizes that even the relatively invariant alpha chain of the DQ molecule may play a critical role in conjunction with the aspartic acid variation at position 57 on the beta chain, possibly as a result of the function they have in presenting antigen to the T-cell receptors (15). Khalil et al. noted that " . . . these two residues (DQ beta Asp 57 and DQ alpha Arg 52) are located at the opposite extremities of the alpha-helical sides of the antigenic groove," and " . . . the replacement of the aspartic acid by a neutral amino acid (serine, valine, and alanine) and the presence of an arginine residue might therefore modify the antigenic peptide presentation" (15). Their oligonucleotide dot-blot studies of 50 Caucasian patients with IDDM and 73 controls revealed that there was not a single case of IDDM when the susceptibility chain was either absent or present on just an alpha- or beta-chain. There appeared to be " . . . the absolute requirement of complete susceptible heterodimer expression at the cell surface for the manifestation of IDDM disease" (15). Because this combination of susceptibility alleles was also found in some healthy individuals, it implied that these factors were necessary but not sufficient for the occurrence of IDDM (15). As provocative as these findings are, there is as yet little or no information about how these amino acid changes affect specific peptide binding or presentation of the bound peptide to different T-cell receptors, or how the variation in peptide binding and presentation functions as a triggering mechanism for IDDM. Rheumatoid arthritis (RA)
In contrast to IDDM, in which HLA associations were first established with Class I antigens and then through linkage disequilibrium stronger associations were subsequently found in DR and DQ Class II antigens, RA exhibited no important Class I associations, but rather from the beginning RA was found to have strong associations with Class II antigens, Dw4 as defined by HTC, and DR4 defined serologically (Table 2). Early studies showed that Dw4 was present in 48% of 119 patients with RA compared to just 14% of 111 controls for a RR of 5.14 (11). DR4 was identified in 66% of 142 patients compared to 26% of 425 controls for a RR of 5.23. This information, derived from traditional serologic and cellular typing techniques, pointed out the key H L A specificity to be examined by more sophisticated oligonucleotide probes. From numerous studies it was well known that not all individuals with DR4 suffered from RA, and about 2 0 - 3 5 % with RA did not have DR4 (11,12). After identification of five DR4 subtypes by HTC 189
BRAUN TABLE 2
Rheumatoid Arthritis (RA): Association with HLA-DR Beta Chain and Possible Molecular Mimicry a • DR4 and/or DR1 occur in 93% of RA patients. • Of the five subtypes of DR4, only two (Dw4 and 14) occur in RA. • There is no contribution of DQ to RA because, for example, both DQw7 (DQw3.1) and DQw8 (DQw3.2) occur as frequently in controls as in RA. • Amino acids 70-74 of the DR beta chains of DR1 (Dwl), DR4 (Dw4), DR14 (Dwl4), and DR15 (Dwl5) are essentially the same as a segment of the gp 110 protein of the Epstein-Barr virus (except for a minor substitution of arginine for lysine at position 71). aReferences (16-19,21). (Dw4, Dwl0, Dwl3, Dw14, and Dw15), it became clear that Dw4 and D w l 4 were the two major subtypes associated with RA. F u r t h e r clarification came about when the DR4 subtypes, initially defined by HTC, were identified with less ambiguity by SSOP (12). In the DR4-positive Caucasian adults with seropositive RA, about 50% had the Dw4 and about 35% the D w l 4 subtype of DR4 (14). In contrast to IDDM, no further linkage disequilibriumbased associations have been found with alleles of the DQ locus. Therefore, the susceptibility alleles associated with RA appear to be derived from DR genes. The susceptibility to RA of patients who lacked DR4 and both of its subtypes, Dw4 and Dwl4, was further investigated by examining the nucleotide sequences within the D w l 4 gene (12). A D w l 4 nucleotide sequence was found to be present in two other alleles in patients with RA (DR1 and DR6, Dwl6) (12). Nepom and Erlich have summarized these findings: " . . . each of the major susceptibility genes associated with RA, in both DR4-positive and DR4negative individuals, share a common genetic element, namely a characteristic nucleotide sequence within part of the DRB1 gene. This shared sequence encompasses a functionally critical portion of the DR[3 molecule, within a s-helical loop region highly conserved among the different rheumatoid arthritis susceptibility alleles, with only a single conservative substitution distinguishing Dw4 and D w l 4 at codon 71" (12). Functionally, the region surrounding codon 71 appears to be important in T-cell recognition of DR[3 (12,16-18). Lymphocytes from seven patients with RA were all recognized by the DR4, Dwl4-specific T-cell clone (T431) and shared the Dw4-associated third hypervariable region sequence between amino acids 67 and 74, irrespective of the serological DR or the cellular Dw typing (16). As impressive as this finding was, the authors were quick to point out that " . . . this region is likely to be only one of several DR4, Dwl4-associated epitopes that may predispose to rheumatoid arthritis" (16). Their data indicated that there could be at least another three DR4, Dwl4-related epitopes predisposing to RA (16). These investigators postulated that 190
" . . . the third hypervariable region normally associated with DR4, D w l 4 could act as one of the restriction sites for the putative causative agent(s) in rheumatoid arthritis" (16). Nepom et al., however, concluded that the gene for DR4, Dw4, and the gene for DR4, D w l 4 are each individual susceptibility alleles (19). Ankylosing spondylitis (AS) HLA-B27 has a strong association with AS, particularly in Caucasians (20). This association may be based on another aspect of susceptibility, namely, molecular mimicry (Table 3). Such mimicry arises through a number of amino acid sequence homologies that have been found between HLA molecules and putative pathogens (20,21). Klebsiella pneumoniae has been implicated as a pathogen in AS (20). Consequently, molecular similarities between Klebsiella pneumoniae and B27 m a y be important, such as the six-amino acid sequence shared b e t w e e n HLA-B27 and a segment of Klebsiella pneumoniae nitrogenase (20). Other amino acid homologies between HLA molecules and potential pathogens include: 1) a cytomegalovirus protein and HLA-DR[31 alleles (21); and 2) an Epstein-Barr virus g p l l 0 fiveamino acid sequence and four different Class II alleles (DR1, D w l ; DR4, Dw4; DR14, D w l 4 ; and DR15, D w l 5 (20). Because the functional significance of these amino acid sequence changes has only begun to be explored (17,22), the precise mechanism(s) by which they modulate disease susceptibility remains unexplained. Conclusion The finding of specific amino acid s e q u e n c e s among Class II HLA antigens in individuals suffering from certain diseases reveals common fundaTABLE 3
Ankylosing Spondylitis (AS) and HLA-B27: Molecular Mimicry ~ • B27 is present in about 90% of Caucasians and 50% of Blacks with AS. • None of the six B27 variants identified by monoclonal antibodies has any unique relationship with AS. • No clear specificity for AS is provided by the 8 variable amino acids of B27 despite their facing into the antigen binding site. • No clear relationship for AS is provided by lysine (70) and asparagine (97) that are unique to B27. • B27 has 6 amino acids (72-77) identical to Klebsiella pneumoniae nitrogenase (188-193); the likelihood of this occurring by chance is 1:64,000,000. • There is cross-reactivity of antibody for B27, K. pneumoniae, a synthetic peptide sequence 63-84 of B27, and the 188-193 sequence ofK. pneumoniae. aReference (21). CLINICAL BIOCHEMISTRY, VOLUME 25, JUNE 1992
HLA AND DISEASE mental structural alterations with the potential to modulate disease susceptibility. Studies of IDDM and RA, briefly summarized herein, as well as of pemphigus vulgaris and celiac disease, have identified interesting Class II subtypic associations. Other diseases such as multiple sclerosis, systemic lupus erythematosus, and Grave's disease thus far do not appear to have specific Class II gene associations (12). In AS, amino acid sequences of HLA antigens are shared with those of a possible pathogen, Klebsiella p n e u m o n i a e . However, such molecular mimicry thus far has only circumstantial evidence to support its role in disease susceptibility. For the sake of completeness, it should be noted t h a t in Grave's disease the mere a b e r r a n t expression of Class II antigens on thyroid epithelial cells, perhaps through stimulation by I F N - g a m m a provoked by a viral infection, m a y be sufficient to i ni t i at e autoantibody production and a u t o i m m u n e injury (23).
Abbreviations IDDM RA AS HTC SSOP RR
IFN
insulin-dependent diabetes mellitus (Type 1) rheumatoid arthritis ankylosing spondylitis homozygous typing cells sequence-specific oligonucleotide probes relative risk (calculated from a comparison of the ratio of controls with and without a particular HLA antigen to the ratio of patients having a specific disease with and without the same particular HLA antigen) interferon.
References 1. Dorf ME, ed. The role of the major histocompatibility complex in immunobiology. New York: Garland STPM Press, 1981. 2. Sullivan KA, Amos DB. The HLA system and its detection. In: Rose NR, Friedman H, Fahey JL, eds. Manual of clinical laboratory immunology. Pp. 83546. Washington, DC: American Society for Microbiology, 1986. 3. Halloran PF, Wadgymar A, Autenried P. The regulation of expression of major histocompatibility complex products. Transplantation 1986; 41: 413-20. 4. Bjorkman PJ, Parham P. Structure, function, and diversity of Class I major histocompatibility complex molecules. A n n u Rev Biochem 1990; 59: 253-88. 5. WHO Nomenclature Committee. Nomenclature for factors of the HLA system, 1987. Vox Sang 1988; 55: 119-26. 6. Bodmer WF, Albert E, Bodmer JG, et al. Nomenclature for factors of the HLA system, 1987. In: Dupont B, ed. Immunobiology of HLA (Vol 1, Histocompatibility Testing 1987). Pp. 72-9. New York: SpringerVerlag 1989.
CLINICAL BIOCHEMISTRY,VOLUME 25, JUNE 1992
7. Bodmer JG, Marsh SGE, Albert ED, et al. Nomenclature for factors of the HLA system, 1990. H u m I m m u nol 1991; 31: 186-94. 8. Brown JH, Jardetzky T, Saper MA, Samraoui B, Bjorkman PJ, Wiley DC. A hypothetical model of the foreign antigen binding site of Class II histocompatibility molecules. Nature 1988; 332: 845-50. 9. Ramsdell F, Fowlkes BJ. Clonal deletion versus clonal anergy: the role of the thymus in inducing self tolerance. Science 1990; 248: 1342-8. 10. Braun WE. HLA and disease: a comprehensive review. Boca Raton, FL: CRC Press, 1979. 11. Tiwari JL, Terasaki PI. HLA and disease. New York: Springer-Verlag, 1985. 12. Nepom GT, Erlich H. MHC class-II molecules and autoimmunology. A n n u Rev Immunol 1991; 9: 493-525. 13. Nepom GT. HLA and Type I diabetes. Immunology Today 1990; ll: 314-5. 14. Baisch JM, Weeks T, Giles R, Hoover M, Stastny P, Capra JD. Analysis of HLA-DQ genotypes and susceptibility in insulin-dependent diabetes mellitus. N Engl J Med 1990; 322: 1836-41. 15. Khalil I, d'Auriol L, Gobet M, et al. A combination of HLA-DQ beta Asp 57-negative and HLA DQ alpha Arg 52 confers susceptibility to insulin-dependent diabetes mellitus. J Clin Invest 1990; 85: 1315-9. 16, Morel PA, Horn GT, Budd RC, Erlich HA, Fathman CG. Shared molecular markers of genetic predisposition to seropositive rheumatoid arthritis. H u m I m m u nol 1990; 27: 90-9. 17. Seyfried CE, Mickelson E, Hansen JA, Nepom GT. A specific nucleotide sequence defines a functional T-cell recognition epitope shared by diverse HLA-DR specificities. H u m Immunol 1988; 21: 298-9. 18. Hiraiwa A, Yamanaka K, Kwok WW, et al. Structural requirements for recognition of the HLA-Dwl4 Class II epitope - - a key HLA determinant associated with rheumatoid arthritis. Proc Natl Acad Sci USA 1990; 87: 8051-5. 19. Nepom GT, Byers P, Seyfried C, et al. HLA genes associated with rheumatoid arthritis. Identification of susceptibility alleles using specific oligonucleotide probes. Arthritis Rheum 1989; 32: 15-21. 20. Yu DTY, Choo SY, Schaack E. Molecular mimicry in HLA-B27-related arthritis. A n n Intern Med 1989; 111: 581-91. 21. Fujinami RS, Nelson JA, Walker L, Oldstone MBA. Sequence homology and immunologic cross-reactivity of human cytomegalovirus with HLA-DRB chain: a means for graft rejection and immunosuppression. J Virol 1988; 62: 100-5. 22. Rothbard JB, Gefter ML. Interactions between immunogenic peptides and MHC proteins. A n n u Rev I m m u nol 1991; 9: 527-65. 23. Bottazzo GF, Pujol-Borrell R, Hanafusa T, Feldmann M. Role of aberrant HLA-DR expression and antigen presentation in induction of endocrine autoimmunity. Lancet 1983; 2: 1115-9.
191
0009-9120/92 $5.00 + .00 Copyright ¢ 1992 The Canadian Society of Clinical Chemists.
HLA Molecules in Autoimmune Diseases WILLIAM E. BRAUN Department of Hypertension and Nephrology, Cleveland Clinic Foundation, 9500 Euclid Avenue, Cleveland, OH 44195, USA The association of certain autoimmune diseases with HLA molecules is being refined through the use of sequence-specific oligonucleotide probes and amino acid sequencing, together with continuing elucidation of the functional features of HLA molecules derived from the milestone description by Bjorkman of the HLA molecular structure. The association of insulindependent diabetes mellitus and HLA began with weak associations of Class I antigens (B8 and B15) and progressed to Class II antigens (DR3 and DR4), then to subtypes of DR4 (Dw4, 10, and 14), and now to DQ molecules including the absence of aspartic acid at position 57 of the DQ beta chain and the presence of arginine at position 52 of the DQ alpha chain. In rheumatoid arthritis (RA) the HLA antigen association remains with certain Class II molecules of the DR series (DR4 and DR1) that share amino acid sequences with a restricted number of other DR antigens seen in RA, as well as a segment of the gp 110 protein of the Epstein-Barr virus. Although ankylosing spondylitis has a strong association with the Class I antigen B27, that association is not explained by any of the B27 subtypes defined by monoclonal antibodies, by the eight variable amino acids in B27 subtypes, or by the two unique amino acids on B27. The remarkable antibody cross-reactivity among lymphocytes bearing B27, a synthetic peptide sequence (63-84) of B27, and the 188-193 sequence of K. pneumoniae nitrogenase has provided strong support for molecular mimicry being an important mechanism in the association of HLA molecules with disease. As the molecular features related to HLA and disease are being discovered, elucidating the all-important functional roles that they have in actually producing the disease will be an even greater challenge.
KEY WORDS: HLA antigens; diabetes mellitus; rheumatoid arthritis; ankylosing spondylitis. Introduction
he HLA system of antigens is controlled by genes
located on the short arm of the sixth chromosome T in a region known as the major histocompatibility
complex (MHC) that has well-defined counterparts in animals such as the mouse, rat, guinea pig, pig, dog, and horse. The HLA antigens have been divided into two major classes of molecules: Class I and Class II. Class I and Class II molecules differ in their biochemistry, cellular distribution, expression, and function. A detailed examination of these features is beyond the scope of this review and can be found in monographs and textbooks (1,2). Certain major distinctions between Class I and Class II molecules are:
Manuscript received: July 31, 1991; revised: January 28, 1992; accepted: January 31, 1992. CLINICAL BIOCHEMISTRY, VOLUME 25, JUNE 1992
Class I molecules consist of a 44,000 kD heavy chain noncovalently bound to beta-2 microglobulin with a weight of 11,700 kD, whereas Class II molecules consist of a 33,000 kD alpha chain that is relatively invariant and a 29,000 kD beta chain that is responsible for most of the polymorphisms; Class I molecules are quite ubiquitous in their distribution and appear on the surface of all nucleated cells, whereas Class II molecules are restricted to B-lymphocytes, monocytes, macrophages, dendritic cells, Langerhans cells of the skin, and activated T-cells; Class I and Class II molecules differ not only in their constitutive expression but also in their inducible expression (3); although Class I and Class II molecules are assembled in the endoplasmic reticulum, they take different intracellular pathways, with Class I molecules transporting endogenous antigens such as viral antigens through the Golgi apparatus en route to the cell surface, whereas Class II molecules transport endocytosed exogenous antigens such as bacterial antigens through the lysosomal/endosomal system; and, Class I molecules interact primarily with cytotoxic (CD8) T-cells, whereas Class II molecules interact with helper (CD4) T-lymphocytes (4). The techniques for identifying both Class I and Class II HLA antigens have been dominated by serologic methods based on complement-dependent cytotoxicity. In a limited number of laboratories, Class II antigens have also been detected by cellular techniques based on the use of homozygous typing cells (HTC) in mixed lymphocyte culture reactions. These basic serologic and cellular techniques have been d r a m a t i c a l l y improved by the use of sequencespecific oligonucleotide probes (SSOP) and amino acid sequencing. The highly polymorphic HLA system now includes more than 180 alleles defined by one or more techniques (5-7). Crystallographic structural analysis of Class I molecules have shown a configuration consisting of " . . . two sets of structurally homologous domains; one set being the membrane-proximal ~3 and ~2m domains, and the other set being the membranedistal ~1 and a2 d o m a i n s . . , the ~3 and ~2~ domains are folded into ~-sandwich structures resembling Ig constant r e g i o n s . . , the al and ~2 domains sit on top of the Ig-like domains forming a platform consisting of an eight-stranded ~-pleated sheet topped by two long m-helices that lie diagonally across the sheet 187
BRAUN
. . . the binding site for antigenic peptides is a deep cleft located between the two long a - h e l i c e s . . . " (4). The basic configuration of Class II molecules is thought to be quite similar to t h a t of Class I (8). HLA molecules and T-cell receptors permit one's i m m u n e system to distinguish self from nonself through a selection process occurring in the t h y m u s (4,9). In carrying out this general function, it appears t h a t subtle differences in the amino acid composition of HLA molecules, their varying capabilities of binding different peptide antigens in the interhelical groove, their capability of presenting these antigens to and interacting with the repertoire of T-cell receptors, as well as their aberrant expression, may influence the occurrence of autoimmune diseases. By means of basic serologic and cellular techniques, HLA alleles have been shown to be significantly associated with a number of diseases among approximately 600 investigated (10,11). Some of these associations were very strong, such as HLADR2 in narcolepsy, HLA-B27 in ankylosing spondylitis and sero-negative arthropathies, and HLADR3 with DR4 in diabetes mellitus. It is important to note t h a t these are all associations between certain HLA alleles and a particular disease, and not genetic linkage. At this point it is necessary to restate the definition of association as the nonrandom occurrence of two genetically separate traits in a population (e.g., HLA-DR3 and DR4 with diabetes mellitus), linkage as the occurrence of two loci on the same chromosome and sufficiently close together so t h a t something less t h a n completely independent assortment takes place (e.g., idiopathic hemochromatosis with HLA-A), and linkage disequilibrium as the occurrence of two alleles together more frequently t h a n would be expected on the basis of their individual gene frequencies (e.g., HLA-A1, B8, DR3, and DQ2). Associations are tested in unrelated individuals, whereas linkage studies require informative families. As provocative as m a n y of these studies were, there were nevertheless fundamental weaknesses with the serologic and cellular techniques available at t h a t time t h a t rendered them simply incapable of identifying differences in DNA and amino acid sequences. Because of such relatively insensitive techniques, as well as other factors, m a n y individuals with a particular disease appeared to lack the HLA susceptibility antigen, and conversely m a n y people with the HLA susceptibility antigen were free of the relevant disease. In this review, the primary associations between serologically and cellularly defined HLA antigens and certain autoimmune diseases will be considered along with the refinements of those HLA antigen associations based on evolving molecular genetic techniques. The steps in the molecular genetic approach to HLA and disease associations have been defined by Nepom and Erlich (12). Nepom and Erlich (12) have also described three general ways in which Class II structural polymor188
phism may mediate disease susceptibility: "Polymorphic residues in the Class II molecule might differentially bind a putative autoantigen peptide and present it to the responding T-lymphocytes. Alternatively, the spectrum of expressed T-cell receptors m i g h t be different in individuals with different Class II genotypes. Autoreactive T-cell receptors capable of recognizing an autoantigen Class II complex would more likely be selected by thymic epithelium which expressed the nonsusceptible Class II antigen. The third immunologic mechanism invoked to explain HLA disease associations is molecular mimicry" (11). The presence of Class I residues shared with a bacterial antigen (e.g., B27 and Klebsiella pneumoniae nitrogenase) or Class II residues shared with a viral pathogen could result in tolerance to such microbial epitopes leading to an altered immune response (12). "Alternatively such homologous regions could serve as an autoantigenic target with an immune response to the (bacterium or) virus leading to attack of self-cells expressing relevant Class (I or) II products" (12). Insulin-dependent diabetes mellitus (IDDM) and HLA Before the Class II HLA antigens (DR, DQ, and DP) were identified, the susceptibility to IDDM was associated with Class I molecules controlled by the B-locus, n a m e l y HLA-B8, and HLA-B15 (11) as shown in Table 1. Conversely, a resistance to IDDM was associated with HLA-B7 (11). When the serologic identification of Class II HLA-DR antigens was established, even stronger associations were found TABLE 1
Insulin-Dependent Diabetes Mellitus (IDDM): Association with HLA-DQ Beta Aspartic Acid57-Negative and Alpha Arginine52-Positive Haplotypea • Initial associations appeared with Class I antigens B8 and B15 that were in linkage disequilibrium with certain Class II HLA-DR antigens (DR3 and DR4). • DR3 and/or DR4 occurred in 90% of IDDM. • Of the five Dw subtypes of DR4, only three (Dw4, 10, and 14) were seen in IDDM. • The three Dw subtypes of IDDM were in linkage disequilibrium with DQw 3.2 (DQw8). • Key differences in the DQw8 allele (that is closely associated with IDDM) and a neutral allele DQw7 are four amino acid substitutions, including aspartic acid at position 57 of the DQ beta chain. • The best current marker for susceptibility to IDDM includes not only the absence of aspartic acid at position 57 of the DQ beta chain, but also the presence of arginine at position 52 of the DQ alpha chain. • Although the Asps7 absence and Arg 52 presence are necessary conditions for IDDM, they are not sufficient for its occurrence. aReferences (10-15). CLINICAL BIOCHEMISTRY, VOLUME 25, JUNE 1992
HLA AND DISEASE
between susceptibility to IDDM and HLA-DR3 and DR4, and the heterozygote state DR3/DR4, whereas resistance to IDDM was associated with HLA-DR2 (11). Because there is linkage disequilibrium between B8 and DR3, B15 and DR4, and B7 and DR2, it was not surprising to see the weak associations initially described with the Class I B-locus antigens assume greater strength with the Class II DR antigens. For example, the frequency of DR3 was 48% in 265 Caucasian diabetic patients compared to 23% of 776 controls (11) giving a relative risk (RR) of 3.04. DR4 occurred in 65% of 265 Caucasian diabetic patients compared to 30% in controls (RR = 4.28). For DR3/DR4 heterozygotes the RR was increased to 14.26. Resistance to IDDM (RR = 0.17) was noted for DR2 that occurred in only 4% of 265 Caucasian diabetic patients compared to the higher frequency of 26% in 776 controls (11). The strong associations of IDDM with Class II molecules were important findings that focused subsequent DNA studies on the Class II HLA molecules. In these studies it soon became apparent that susceptibility and resistance to IDDM were even more strongly related to gene products of the HLADQ locus that defined subsets of the susceptibilityrelated antigens DR3 and DR4, as well as subsets of the resistance-related DR2 antigen. For example, the DR4 subtype DQBI*0302 (DQw3.2) is present in approximately 95% of DR4-positive haplotypes of IDDM Caucasians, but it is also found in 6 0 - 7 0 % of individuals with DR4-positive haplotypes but no IDDM, and in about 27% of all Caucasians (12). DQw3.2 differs from a closely related neutral allele DQw3.1 by j u s t four amino acids (12). One key amino acid difference that has received a great deal of attention is at position 57 (12-14). The DQw3.2 allele (currently named DQw8) that lacks aspartic acid at position 57 has a gene frequency of 35.7% in IDDM compared to 10.1% in normal subjects (14), the difference being highly significant (p < 0.001). However, the simple presence or absence of aspartic acid alone is clearly not a complete explanation for resistance or susceptibility, respectively, to IDDM. Although a protective effect was associated with the DQwl.2 allele bearing aspartic acid at position 57, no protection was associated with other aspartic acid-bearing alleles such as D Q w l . 1 2 , DQw4, and DQw7 (14). Moreover, HLA-DQw.1 in the homozygous state was also protective even though it has valine rather than aspartic acid at position 57 of the DQ beta chain (14). Two reasons have been offered for the lack of utility of aspartic acid testing by itself as a predictor of risk for IDDM: 1) the HLA-DR7/DQw2 haplotype that lacks an aspartic acid at position 57 of the DQ beta chain but is protective against IDDM would erroneously classify approximately 10% of Caucasians in a high-risk category; 2) the HLA-DQw7 allele that contains aspartic acid fails to confer protection because 81% of diabetic patients heterozygous for an aspartic acidcontaining allele had the aspartic acid-positive DQw7 allele (14). CLINICAL BIOCHEMISTRY, VOLUME 25, JUNE 1992
Further resolution of this dilemma lies in the fact that an HLA-DQ~ aspartic acid 57-negative allele when found in conjunction with an HLA-DQ~ arginine 52-positive allele conferred an even stronger susceptibility to IDDM, whereas their absence constituted virtually absolute protection (15). This finding emphasizes that even the relatively invariant alpha chain of the DQ molecule may play a critical role in conjunction with the aspartic acid variation at position 57 on the beta chain, possibly as a result of the function they have in presenting antigen to the T-cell receptors (15). Khalil et al. noted that " . . . these two residues (DQ beta Asp 57 and DQ alpha Arg 52) are located at the opposite extremities of the alpha-helical sides of the antigenic groove," and " . . . the replacement of the aspartic acid by a neutral amino acid (serine, valine, and alanine) and the presence of an arginine residue might therefore modify the antigenic peptide presentation" (15). Their oligonucleotide dot-blot studies of 50 Caucasian patients with IDDM and 73 controls revealed that there was not a single case of IDDM when the susceptibility chain was either absent or present on just an alpha- or beta-chain. There appeared to be " . . . the absolute requirement of complete susceptible heterodimer expression at the cell surface for the manifestation of IDDM disease" (15). Because this combination of susceptibility alleles was also found in some healthy individuals, it implied that these factors were necessary but not sufficient for the occurrence of IDDM (15). As provocative as these findings are, there is as yet little or no information about how these amino acid changes affect specific peptide binding or presentation of the bound peptide to different T-cell receptors, or how the variation in peptide binding and presentation functions as a triggering mechanism for IDDM. Rheumatoid arthritis (RA)
In contrast to IDDM, in which HLA associations were first established with Class I antigens and then through linkage disequilibrium stronger associations were subsequently found in DR and DQ Class II antigens, RA exhibited no important Class I associations, but rather from the beginning RA was found to have strong associations with Class II antigens, Dw4 as defined by HTC, and DR4 defined serologically (Table 2). Early studies showed that Dw4 was present in 48% of 119 patients with RA compared to just 14% of 111 controls for a RR of 5.14 (11). DR4 was identified in 66% of 142 patients compared to 26% of 425 controls for a RR of 5.23. This information, derived from traditional serologic and cellular typing techniques, pointed out the key H L A specificity to be examined by more sophisticated oligonucleotide probes. From numerous studies it was well known that not all individuals with DR4 suffered from RA, and about 2 0 - 3 5 % with RA did not have DR4 (11,12). After identification of five DR4 subtypes by HTC 189
BRAUN TABLE 2
Rheumatoid Arthritis (RA): Association with HLA-DR Beta Chain and Possible Molecular Mimicry a • DR4 and/or DR1 occur in 93% of RA patients. • Of the five subtypes of DR4, only two (Dw4 and 14) occur in RA. • There is no contribution of DQ to RA because, for example, both DQw7 (DQw3.1) and DQw8 (DQw3.2) occur as frequently in controls as in RA. • Amino acids 70-74 of the DR beta chains of DR1 (Dwl), DR4 (Dw4), DR14 (Dwl4), and DR15 (Dwl5) are essentially the same as a segment of the gp 110 protein of the Epstein-Barr virus (except for a minor substitution of arginine for lysine at position 71). aReferences (16-19,21). (Dw4, Dwl0, Dwl3, Dw14, and Dw15), it became clear that Dw4 and D w l 4 were the two major subtypes associated with RA. F u r t h e r clarification came about when the DR4 subtypes, initially defined by HTC, were identified with less ambiguity by SSOP (12). In the DR4-positive Caucasian adults with seropositive RA, about 50% had the Dw4 and about 35% the D w l 4 subtype of DR4 (14). In contrast to IDDM, no further linkage disequilibriumbased associations have been found with alleles of the DQ locus. Therefore, the susceptibility alleles associated with RA appear to be derived from DR genes. The susceptibility to RA of patients who lacked DR4 and both of its subtypes, Dw4 and Dwl4, was further investigated by examining the nucleotide sequences within the D w l 4 gene (12). A D w l 4 nucleotide sequence was found to be present in two other alleles in patients with RA (DR1 and DR6, Dwl6) (12). Nepom and Erlich have summarized these findings: " . . . each of the major susceptibility genes associated with RA, in both DR4-positive and DR4negative individuals, share a common genetic element, namely a characteristic nucleotide sequence within part of the DRB1 gene. This shared sequence encompasses a functionally critical portion of the DR[3 molecule, within a s-helical loop region highly conserved among the different rheumatoid arthritis susceptibility alleles, with only a single conservative substitution distinguishing Dw4 and D w l 4 at codon 71" (12). Functionally, the region surrounding codon 71 appears to be important in T-cell recognition of DR[3 (12,16-18). Lymphocytes from seven patients with RA were all recognized by the DR4, Dwl4-specific T-cell clone (T431) and shared the Dw4-associated third hypervariable region sequence between amino acids 67 and 74, irrespective of the serological DR or the cellular Dw typing (16). As impressive as this finding was, the authors were quick to point out that " . . . this region is likely to be only one of several DR4, Dwl4-associated epitopes that may predispose to rheumatoid arthritis" (16). Their data indicated that there could be at least another three DR4, Dwl4-related epitopes predisposing to RA (16). These investigators postulated that 190
" . . . the third hypervariable region normally associated with DR4, D w l 4 could act as one of the restriction sites for the putative causative agent(s) in rheumatoid arthritis" (16). Nepom et al., however, concluded that the gene for DR4, Dw4, and the gene for DR4, D w l 4 are each individual susceptibility alleles (19). Ankylosing spondylitis (AS) HLA-B27 has a strong association with AS, particularly in Caucasians (20). This association may be based on another aspect of susceptibility, namely, molecular mimicry (Table 3). Such mimicry arises through a number of amino acid sequence homologies that have been found between HLA molecules and putative pathogens (20,21). Klebsiella pneumoniae has been implicated as a pathogen in AS (20). Consequently, molecular similarities between Klebsiella pneumoniae and B27 m a y be important, such as the six-amino acid sequence shared b e t w e e n HLA-B27 and a segment of Klebsiella pneumoniae nitrogenase (20). Other amino acid homologies between HLA molecules and potential pathogens include: 1) a cytomegalovirus protein and HLA-DR[31 alleles (21); and 2) an Epstein-Barr virus g p l l 0 fiveamino acid sequence and four different Class II alleles (DR1, D w l ; DR4, Dw4; DR14, D w l 4 ; and DR15, D w l 5 (20). Because the functional significance of these amino acid sequence changes has only begun to be explored (17,22), the precise mechanism(s) by which they modulate disease susceptibility remains unexplained. Conclusion The finding of specific amino acid s e q u e n c e s among Class II HLA antigens in individuals suffering from certain diseases reveals common fundaTABLE 3
Ankylosing Spondylitis (AS) and HLA-B27: Molecular Mimicry ~ • B27 is present in about 90% of Caucasians and 50% of Blacks with AS. • None of the six B27 variants identified by monoclonal antibodies has any unique relationship with AS. • No clear specificity for AS is provided by the 8 variable amino acids of B27 despite their facing into the antigen binding site. • No clear relationship for AS is provided by lysine (70) and asparagine (97) that are unique to B27. • B27 has 6 amino acids (72-77) identical to Klebsiella pneumoniae nitrogenase (188-193); the likelihood of this occurring by chance is 1:64,000,000. • There is cross-reactivity of antibody for B27, K. pneumoniae, a synthetic peptide sequence 63-84 of B27, and the 188-193 sequence ofK. pneumoniae. aReference (21). CLINICAL BIOCHEMISTRY, VOLUME 25, JUNE 1992
HLA AND DISEASE mental structural alterations with the potential to modulate disease susceptibility. Studies of IDDM and RA, briefly summarized herein, as well as of pemphigus vulgaris and celiac disease, have identified interesting Class II subtypic associations. Other diseases such as multiple sclerosis, systemic lupus erythematosus, and Grave's disease thus far do not appear to have specific Class II gene associations (12). In AS, amino acid sequences of HLA antigens are shared with those of a possible pathogen, Klebsiella p n e u m o n i a e . However, such molecular mimicry thus far has only circumstantial evidence to support its role in disease susceptibility. For the sake of completeness, it should be noted t h a t in Grave's disease the mere a b e r r a n t expression of Class II antigens on thyroid epithelial cells, perhaps through stimulation by I F N - g a m m a provoked by a viral infection, m a y be sufficient to i ni t i at e autoantibody production and a u t o i m m u n e injury (23).
Abbreviations IDDM RA AS HTC SSOP RR
IFN
insulin-dependent diabetes mellitus (Type 1) rheumatoid arthritis ankylosing spondylitis homozygous typing cells sequence-specific oligonucleotide probes relative risk (calculated from a comparison of the ratio of controls with and without a particular HLA antigen to the ratio of patients having a specific disease with and without the same particular HLA antigen) interferon.
References 1. Dorf ME, ed. The role of the major histocompatibility complex in immunobiology. New York: Garland STPM Press, 1981. 2. Sullivan KA, Amos DB. The HLA system and its detection. In: Rose NR, Friedman H, Fahey JL, eds. Manual of clinical laboratory immunology. Pp. 83546. Washington, DC: American Society for Microbiology, 1986. 3. Halloran PF, Wadgymar A, Autenried P. The regulation of expression of major histocompatibility complex products. Transplantation 1986; 41: 413-20. 4. Bjorkman PJ, Parham P. Structure, function, and diversity of Class I major histocompatibility complex molecules. A n n u Rev Biochem 1990; 59: 253-88. 5. WHO Nomenclature Committee. Nomenclature for factors of the HLA system, 1987. Vox Sang 1988; 55: 119-26. 6. Bodmer WF, Albert E, Bodmer JG, et al. Nomenclature for factors of the HLA system, 1987. In: Dupont B, ed. Immunobiology of HLA (Vol 1, Histocompatibility Testing 1987). Pp. 72-9. New York: SpringerVerlag 1989.
CLINICAL BIOCHEMISTRY,VOLUME 25, JUNE 1992
7. Bodmer JG, Marsh SGE, Albert ED, et al. Nomenclature for factors of the HLA system, 1990. H u m I m m u nol 1991; 31: 186-94. 8. Brown JH, Jardetzky T, Saper MA, Samraoui B, Bjorkman PJ, Wiley DC. A hypothetical model of the foreign antigen binding site of Class II histocompatibility molecules. Nature 1988; 332: 845-50. 9. Ramsdell F, Fowlkes BJ. Clonal deletion versus clonal anergy: the role of the thymus in inducing self tolerance. Science 1990; 248: 1342-8. 10. Braun WE. HLA and disease: a comprehensive review. Boca Raton, FL: CRC Press, 1979. 11. Tiwari JL, Terasaki PI. HLA and disease. New York: Springer-Verlag, 1985. 12. Nepom GT, Erlich H. MHC class-II molecules and autoimmunology. A n n u Rev Immunol 1991; 9: 493-525. 13. Nepom GT. HLA and Type I diabetes. Immunology Today 1990; ll: 314-5. 14. Baisch JM, Weeks T, Giles R, Hoover M, Stastny P, Capra JD. Analysis of HLA-DQ genotypes and susceptibility in insulin-dependent diabetes mellitus. N Engl J Med 1990; 322: 1836-41. 15. Khalil I, d'Auriol L, Gobet M, et al. A combination of HLA-DQ beta Asp 57-negative and HLA DQ alpha Arg 52 confers susceptibility to insulin-dependent diabetes mellitus. J Clin Invest 1990; 85: 1315-9. 16, Morel PA, Horn GT, Budd RC, Erlich HA, Fathman CG. Shared molecular markers of genetic predisposition to seropositive rheumatoid arthritis. H u m I m m u nol 1990; 27: 90-9. 17. Seyfried CE, Mickelson E, Hansen JA, Nepom GT. A specific nucleotide sequence defines a functional T-cell recognition epitope shared by diverse HLA-DR specificities. H u m Immunol 1988; 21: 298-9. 18. Hiraiwa A, Yamanaka K, Kwok WW, et al. Structural requirements for recognition of the HLA-Dwl4 Class II epitope - - a key HLA determinant associated with rheumatoid arthritis. Proc Natl Acad Sci USA 1990; 87: 8051-5. 19. Nepom GT, Byers P, Seyfried C, et al. HLA genes associated with rheumatoid arthritis. Identification of susceptibility alleles using specific oligonucleotide probes. Arthritis Rheum 1989; 32: 15-21. 20. Yu DTY, Choo SY, Schaack E. Molecular mimicry in HLA-B27-related arthritis. A n n Intern Med 1989; 111: 581-91. 21. Fujinami RS, Nelson JA, Walker L, Oldstone MBA. Sequence homology and immunologic cross-reactivity of human cytomegalovirus with HLA-DRB chain: a means for graft rejection and immunosuppression. J Virol 1988; 62: 100-5. 22. Rothbard JB, Gefter ML. Interactions between immunogenic peptides and MHC proteins. A n n u Rev I m m u nol 1991; 9: 527-65. 23. Bottazzo GF, Pujol-Borrell R, Hanafusa T, Feldmann M. Role of aberrant HLA-DR expression and antigen presentation in induction of endocrine autoimmunity. Lancet 1983; 2: 1115-9.
191
