HLA B27 in health and disease: a double-edged sword?

Transcription

1 Rheumatology 2002;41: Review The 2000 Michael Mason Prize Essay HLA B27 in health and disease: a double-edged sword? P. Bowness Nuffield Orthopaedic Hospital, Windmill Road, Headington, Oxford OX3 7LD, UK Abstract The strong association of the HLA class 1 allele HLA B27 with ankylosing spondylitis (AS) has been recognized for over 25 yr, however the pathogenic mechanism linking HLA B27 with AS and other spondyloarthropathies remains a mystery. We now know that the principal natural function of HLA B27 is an immunologic one, namely to bind antigenic peptides and then present them to T lymphocytes. I have shown that HLA B27 functions as an excellent antigen-presenting molecule in both spondyloarthropathy patients and healthy individuals. A working molecular model of how T cells recognize HLA B27 has been generated and tested. Evidence that T cells have a role in spondyloarthritis has also been found. First, expanded populations of T lymphocytes were found in both the blood and synovial fluid of patients with reactive arthritis (ReA). Secondly, a strong cytotoxic T-cell response to an HLA B27-restricted peptide epitope from Chlamydia trachomatis was found in a patient with ReA. This peptide, derived from a bacterium known to trigger ReA, is thus a candidate arthritogenic peptide. We have also found evidence that HLA B27 has an unusual cell biology compared with other HLA molecules. HLA B27 demonstrates an unusual ability to form heavy chain homodimers in vitro. Dimerization is dependent upon disulphide bonding through an unpaired cysteine at position 67. Remarkably these dimers lack b2 microglobulin, previously thought to be an essential component of all mature MHC class 1 molecules. HLA B27 homodimer formation has also been demonstrated in certain cell lines in vivo, and preliminary data suggest that significant numbers of T cells from patients with spondyloarthropathy express a ligand for HLA B27 homodimers. These findings have extended our understanding of the beneficial immunologic function of HLA B27, and have also led us to propose the testable new hypothesis that HLA B27 heavy chain dimerization may be involved in the pathogenesis of spondyloarthritis. KEY WORDS: HLA B27, Reactive arthritis, Ankylosing spondylitis, Immunology. Although the strong association of possession of the human leucocyte antigen HLA B27 with the spondyloarthropathies (Table 1) has been recognized for over 25 yr w1x, the pathogenic link remains elusive. Numerous hypotheses have been proposed to explain the association of HLA B27 with spondyloarthritis, and are summarized in Table 2. The theories proposed derive varying support from epidemiological studies and from animal models of disease wreviewed in 2x. However the development of a disease resembling human spondyloarthritis in HLA B27-transgenic rats and mice provides direct evidence that HLA B27 itself is involved in disease pathogenesis w3, 4x. Since the principle natural Submitted 12 April 2000; revised version accepted 7 March function of HLA class 1 molecules is to bind and present short antigenic peptides to cytotoxic T lymphocytes (CTL) w5x, Benjamin and Parham w6x have suggested that HLA B27 might predispose to spondyloarthritis by binding one or more specific arthritogenic peptides and stimulating arthritogenic T cells. The identification of different molecular subtypes of HLA B27, which differ in the amino acid composition of their peptidebinding groove w7, 8x, has led to studies of the disease association and peptide-binding specificity of different subtypes. There is some evidence that subtypes such as HLA B*2706 and *2709, which preliminary epidemiological studies suggest are not associated with ankylosing spondylitis (AS) w9x, may bind a subtly different set of peptides. If confirmed, these data would support a role for peptide binding in disease pathogenesis. 857 ß 2002 British Society for Rheumatology

2 858 P. Bowness TABLE 1. Diseases associated with HLA B27. Ankylosing spondylitis. Reactive arthritis (follows infection with species of Chlamydia, Campylobacter, Salmonella, Shigella and Yersinia). Sacroileitis associated with psoriasis. Sacroileitis associated with inflammatory bowel disease. Undifferentiated oligoarthropathy. Anterior uveitis. Aortic regurgitation together with cardiac conduction abnormality. Enthesis-related juvenile idiopathic arthritis TABLE 2. Theories explaining the association of HLA B27 with the spondyloarthropathies. HLA B27 is in genetic linkage with a disease-associated gene. HLA B27 binds and presents arthritogenic peptides to T cells. HLA B27 is involved in thymic selection of a T-cell repertoire that is susceptible to spondyloarthritis. HLA B27 has an unusual cell biology compared with other HLA class 1 molecules. The HLA B27 free cysteine at position 67 can be chemically modified, leading to an altered self. There is cross reactivity between antibodies directed at bacterial protein(s) and HLA B27. HLA B27 is a receptor for a bacterial ligand. Interaction of HLA B27 with a bacterial superantigen causes non-specific T-cell stimulation. HLA B27-derived peptides are presented by HLA class 2 molecules to CD4+ T cells Alternatively, the cell biology or immunogenicity of HLA B27 may be affected by the presence of an unpaired cysteine at position 67 of the a1 helix. We have recently provided new data suggesting that HLA B27 does indeed have an unusual cell biology (see below), and this would be consistent with data from HLA B27 transgenic mice implicating an abnormal form of HLA B27 molecules in disease. There is no clear evidence supporting antibody cross-reactivity or receptormediated mechanisms. Superantigens are microbial proteins capable of stimulating large numbers of T cells, and are capable of stimulating autoimmune reactions w10x. The bacteria implicated in reactive arthritis (ReA) have not been found to produce superantigens w11x, although interestingly another gut microorganism, Clostridium perfringens, has w12x. HLA B27 is itself capable of providing a source of peptides which in turn can be presented by HLA class 2 molecules to T cells w13, 14x. This intriguing observation undoubtedly requires further study. Strikingly, HLA B27 remains common in many populations, in a distribution which is probably not compatible with a genetic founder effect. The underlying thesis of this essay is that some particular function or functions of HLA B27 have conferred some selective advantage to HLA B27 positive individuals, at least until recent evolutionary time. This function is likely to be immunologic, very probably conferring on HLA B27 the ability to present pathogen-derived antigenic epitopes to T cells either more rapidly or more effectively than other HLA class 1 alleles. This same feature most likely also predisposes to spondyloarthritis. Thus, a proportion of HLA B27 positive individuals pay a price for carrying this super HLA allele, making this molecule a powerful double-edged sword. With the premise that by understanding the natural function of HLA B27, we may ultimately come to better understand its role in disease, I have set out to answer the following questions: (i) What determines the peptide-binding selectivity of HLA B27? How do T cells recognize this HLA B27upeptide combination and does HLA B27 function differently in patients with spondyloarthritis, compared with healthy HLA B27 (ii) positive individuals? Are T lymphocytes indeed involved in ReA, one of the spondyloarthritides? (iii) Is there evidence that T cells from individuals with ReA recognize potentially arthritogenic peptides? (iv) Is the cell biology of HLA B27 unusual, and does the unpaired cysteine at position 67 contribute to this? Understanding the natural function of HLA B27 HLA B27 selectively binds peptides and presents them to the T-cell receptor (TCR) of CTL In order to try and understand the rules governing the molecular function of HLA B27, I initially chose to study the role of HLA B27 in the immune response to influenza A virus, a common virus responsible for many millions of deaths this century. At the start of this work the natural role of MHC class 1 molecules had been shown to involve the binding of short intracellular peptide antigens and their subsequent presentation to CTL at the cell surface (reviewed in w15x). Huet et al. w16x had identified a single antigenic epitope in the flu nucleoprotein that could be recognized by HLA B27 positive individuals. This epitope, NP , comprises nine amino acids, SRYWAIRTR in single letter code. I developed an in vitro HLA B27-binding assay in which synthetic NP or substituted analogues differing at single amino acid residues were added to lysates of radiolabelled HLA B27-transfected T2 cells. This cell line lacks a functional peptide transport mechanism and is thus unable to load endogenous peptides into nascent HLA B27 molecules. However, exogenously supplied peptides can promote assembly and folding of mature HLA B27ub2mupeptide complexes. Results can be quantified following immunoprecipitation with the conformation-specific monoclonal antibody W6u32. I was able to show that binding to HLA B27 was abrogated by substituting the arginine residue at position 2 for leucine (the R2L substitution, see Table 3). The tyrosine at position 3 could be

3 HLA B27 in health and disease 859 TABLE 3. Binding to HLA B27 of NP and single amino acid substituted analogues Position HLA B27 assembly NP S R Y W A I R T R Yes Substitutions Position 1 S1F F Yes S1L L Yes Position 2 R2L L No Position 3 Y3E E No Y3F F Yes Y3L L Yes Position 4 W4A A Yes W4E E Yes W4F F Yes W4H H Yes W4K K Yes W4L L Yes W4Y Y Yes Position 7 R7A A Yes R7D D Yes R7E E Yes R7I I Yes Position 8 T8A A Yes T8E E Yes T8F F Yes T8K K Yes Position 9 R9L L Yes M peptide was added to 35S methionine-labelled lysates of T2 cells transfected with HLA B*2705. substituted to leucine (Y3L) and the resultant peptide could not only bind to HLA B27 (Table 3), but was also recognized by T cells raised against the native peptide (Table 4) w11, 17x. However the position 3 substitution to glutamic acid (Y3E) does not bind to HLA B27 and of course is not recognized by CTL. These findings were in agreement with crystallographic data showing that the peptide P2 arginine side chain was tightly bound in a pocket under the HLA B27 a1 helix (the B pocket) w18, 19x, and with the findings that pooled peptides eluted from HLA B27 w20x (but not from other MHC alleles w21x) shared an arginine residue at their second position. Based upon these data, and my results for T-cell recognition of NP peptide analogues (see below), I proposed a model predicting which side chains of a peptide antigen would be important for binding to HLA B27 and which would be critical for recognition by T cells. The latter I called flag residues w11, 17x. A schema for this model is shown in Fig. 1. The data on T-cell recognition derive from a panel of CTL clones from the blood of influenza-exposed HLA B27-positive patients and controls w11, 17x. These CTL clones specifically recognized and killed influenza-infected or NP pulsed HLA B27-positive target cells w11, 17, 22x. An example of such a CTL assay is shown in Fig. 2. The CTL clone SD17 kills HLA B27-positive target cells pulsed with the NP peptide (shown in black, dotted line). However, an analogue peptide in which the position 4 tryptophan (W) is substituted to tyrosine (W4Y) is recognized only weakly, and one with a position 4 phenyl alanine (W4F) is not recognized at all. The latter peptide only differs from the W4Y peptide by a single hydroxyl group. Thus, this CTL clone has exquisite sensitivity for even the most subtle of changes at peptide P4 residue. Data for recognition of different NP analogues by six different CTL clones are summarized in Table 4. Positions 1, 4 and 8 of NP do not significantly influence binding to HLA B27, but are critical for recognition by the TCR. This paradigm has significant implications for arthritogenic peptide models of disease pathogenesis, since it indicates rules both for peptide binding to HLA B27, and for potential cross-recognition of different peptides by the same TCR. Thus, it can be used both to identify HLA B27-binding peptides (see below) and to predict potential cross-reactive T-cell epitopes. HLA B27 is an excellent antigen-presenting molecule in both spondyloarthropathy patients and healthy individuals In order to define the molecular rules for the recognition of HLA B27 by T cells, and to look for qualitative differences in the immune responses of patients with spondyloarthritis and healthy individuals, I sequenced the TCR genes of nine HLA B27-restricted NP specific CTL clones. Four clones were derived from two patients with spondyloarthropathy and five from a healthy HLA B27-positive control w22x. TCRs, like antibodies, are generated by combinatorial recombination of germline gene segments which, together with other mechanisms including random nucleotide addition at joining regions, generates a hugely diverse T-cell

4 860 P. Bowness TABLE 4. Recognition of NP variants by NP-specific HLA B27-restricted CTL clones of defined TCR usage Secondly, the TCRs of NP specific HLA B27-restricted CTL from healthy and diseased individuals have almost identical predicted amino acid sequences w22x. This shows that the HLA B27 molecule in patients with spondyloarthritis is capable both of presenting a viral peptide and of generating a qualitatively similar T-cell response to that seen in healthy HLA B27-positive individuals. These molecular data are supported by functional recognition data both ex vivo and in vitro, with both patients and controls making strong CTL responses to influenza virus. These responses are HLA B27-restricted and always directed against NP w11, 17, 22x. I subsequently cloned the TCR from an HLA B27-restricted NP specific CTL clone (GRb), derived from a patient with spondyloarthritis, into a reporter cell line RBL-2H3. Thus, it could be formally proved that the HLA B27uNP complex was both necessary and sufficient for specific immune recognition by this TCR w11, 23x. This cell line was used to generate the monoclonal antibody to the human TCRb gene segment TCRVB7S1 w24x. The importance of the TCRVB7S1 gene segment in the human HLA B27uNP immune response has subsequently been confirmed using this antibody w25x. Unfortunately, the hope that such a reagent might prove invaluable for identifying HLA B27-restricted T cells in general (including those recognizing arthritogenic peptides) was dashed by the finding that HLA B27-restricted CTL specific for an HIV peptide use completely different TCRVB segments w11x. Thus, TCRs that recognize different peptides bound to HLA B27 use different TCR gene segments, even though a significant part of their interaction is directly with the HLA B27 a-helices. Recognition is scored as the lowest variant peptide concentration giving half of maximal killing found with NP ,10 28 M or M, i.e. equivalent or better killing than for wild-type peptide.,10 27 M;,10 26 M;,10 25 M (or 25% maximal lysis at M);, <10% cytolysis at M. Most titrations were repeated two to five times, the most frequent result is shown (or lower concentration of two equally frequent results). repertoire. This is shown schematically in Fig. 3, together with a cartoon representation of the mature TCR protein. Table 4 summarizes the TCR a- and b-chain usage of six of these clones, as well as showing the predicted amino acid usage and length of their third hypervariable regions w22x. Two important conclusions arise from comparison of these sequences. First, NP-specific HLA B27-restricted TCR use highly conserved TCRs, with the pattern of conservation (if not the particular segment used) suggesting a common rule for MHC class 1-restricted responses w22x. Thus, most TCR recognizing HLA B27uNP used a single Vb gene segment (Vb7.1 or using the revised WHO nomenclature BV7S1), and one of two Va or Ja segments. This bias is not found in peripheral blood T lymphocytes from patient GR w22x, implying antigen-driven selection. Towards a molecular model of HLA B27 recognition by T cells I was able to predict the exact orientation in which the TCR recognizes HLA B27uNP , by correlation of the fine specificity of CTL clones for P1 peptide analogues with their TCR a-chain usage w17x. This predicted orientation (i.e. with the TCR a-chain docking with the N-terminal part of bound peptide and b chain with C terminal) has subsequently been found in the two crystal structures of TCRs complexed with MHC class 1 w26, 27x. Dr E. Y. Jones has created a detailed molecular model showing how the GRb TCR docks with the HLA B27uNP complex, using these data together with the atomic coordinates HLA B27 w18, 19x, and of the 2C TCR w26x. This model is shown in Fig. 4, and suggests that the TCR a-chain plays an important role in the recognition both of the NP position 4 side chain and also of the HLA B27 a1 helix. This model must of course be regarded as hypothetical, particularly since the 2C TCR is not HLA B27 restricted. However, it is supported by experiments in which site-directed mutation of certain a-chain residues of the GRb TCR, predicted to make important interactions with either the HLA B27 helices or bound peptide, did indeed abrogate

5 HLA B27 in health and disease 861 FIG. 1. Cartoon schema showing binding of antigenic peptides to HLA B27. The side chains of anchor residues of bound peptide are located within pockets surrounding the peptide-binding groove. In contrast, the side chains of flag amino acids are orientated towards and contacted by the TCR of T lymphocytes recognizing the HLA B27upeptide combination (after Bowness et al. w22x). excellent antigen-presenting molecule in both spondyloarthropathy patients and healthy individuals. A working molecular model of HLA B27upeptide recognition by T cells has been generated. FIG. 2. Amino acid substitutions of the peptide fourth position greatly reduce or abrogate CTL recognition of NP recognition of HLA B27uNP by the transfected GRb TCR w23x. These experiments have given us a clear idea of the molecular anatomy of recognition of a particular HLA B27upeptide combination by T cells. This knowledge could be used to design potentially therapeutic peptides capable of binding tightly to HLA B27, and possibly also of antagonizing subsequent T-cell recognition. In summary, these studies show that HLA B27 binds peptides according to well defined molecular rules, and presents them to the TCRs of CTL. HLA B27 is an Are T lymphocytes involved in ReA? If the pathogenic role of HLA B27 in the spondyloarthropathies is to present arthritogenic peptides to CTL, one would expect to find evidence of expanded populations of CD8+ T cells within the joints of patients with spondyloarthritis. Whilst a few such HLA B27-restricted CTL clones have been identified w28x, most studies of synovial fluid in patients with ReA have found predominant HLA class 2-restricted CD4+ responses to the triggering organisms w29, 30x. Other groups have questioned whether T cells are involved at all in the pathogenesis of ReA. Therefore, we set out to study the T-cell repertoire of patients with ReA, using monoclonal antibodies to the TCR. This technique has the advantage of giving quantitative and reproducible ex vivo data, although it gives no information about T-cell specificity. Paired samples of peripheral blood and synovial fluid lymphocytes from 10 patients with ReA arthritis were stained with a panel of 17 monoclonal antibodies to TCR Vb families, together with antibodies to CD4 or CD8. Significant T-cell expansions, defined as being >3 S.D. above the mean value for control peripheral blood leucocytes (PBL), were observed in all patients. Table 5 shows that expansions were found in both CD4+ and CD8+ populations, and in both synovial fluid and blood w31x. The greatest number of expansions was found in the synovial CD8 compartment. Although different Vb expansions were found in different patients, we did for example find TCRBV22 expansions in the synovial CD8 compartment of five out of six patients with ReA following Chlamydia infection w31x. Although the results of studies of synovial fluid TCR repertoire are difficult to interpret (principally

6 862 P. Bowness FIG. 3. Cartoon showing rearrangement of TCR genes (A) and structure of mature protein (B). because of worries about bias in T-cell recruitment or persistence, discussed in Bowness and Bell w32x), the presence of T-cell expansions in the peripheral blood of patients with ReA is undoubtedly significant. Our results are consistent with models in which both CD4+ and CD8+ T lymphocytes are involved in disease, although they give little clue as to pathogenic mechanisms. CD8+ T-cell expansions studied further expressed both activation and memory markers. Perhaps more interestingly, some of the predicted amino acid sequences of these CD8+ synovial T cells w33x show remarkable homology with those obtained by other groups, possibly favouring an antigen-driven mechanism (P. Bowness and E. May, unpublished results). It will be important to compare the T-cell repertoire in ReA with that found in other spondyloarthropathies such as psoriatic arthritis. The finding of common patterns of TCR usage might point to common pathogenic mechanisms. FIG. 4. Molecular model showing the interaction of the TCR of an HLA B27-restricted T-cell GRb with HLA B*2705 and bound NP peptide. Views shown concentrate on the third hypervariable region of the TCRa chain. Orthogonal views are orientated: (a) along the peptide-binding groove with the TCRb chain omitted and TCR CDR3a loop highlighted; and (b) onto the peptide-binding groove. The Ca postions for key residues are indicated. The modelled side chain positions for peptide P4 residue tryptophan, MHC a1 helix residue 65 and TCR CDR3a residue 102 are shown in ball and stick representation. Components of this figure were drawn using a modified version of MOLSCRIPT. TABLE 5. Multiple T-cell expansions are found in the synovial fluid and blood of patients with ReA CD4+ T cells CD8+ T cells Controls (n = 13) 0 2 Patients (n = 10) PBL Synovial fluid a Paired samples of peripheral blood and synovial fluid lymphocytes from 10 patients were stained with a panel of 17 monoclonal antibodies to TCR Vb families, together with antibodies to CD4 or CD8. Significant expansions are defined as being >3 S.D. above the mean value for control PBL. a CD8+ T-cell expansions studied expressed both activation and memory markers.

7 HLA B27 in health and disease 863 Does Chlamydia trachomatis, a bacterium that triggers ReA, contain potential HLA B27-binding peptides and can these stimulate immune responses? As a first step to testing the arthritogenic peptide hypothesis, HLA B27-restricted CTL specific for C. trachomatis peptides were sought in patients with ReA. One hundred and eleven peptides were synthesized, with sequences derived from the first 13 C. trachomatis genes characterized. Peptides were selected according to the rules for HLA B27-binding defined previously thus, all were 8 10 amino acids in length and had R at P2, and an aromatic, hydrophobic or positively charged C terminus. Sixty-two of the peptides were found to bind to HLA B27 in in vitro binding assays w34, 35x (P. Bowness, R. L. Allen and A. McMichael, manuscript in preparation). Responses to these 62 peptides were sought in the blood and synovial fluid of 10 patients with ReAuAS and in blood from healthy HLA B27-positive controls. Figure 5 shows data for synovial fluid T cells from donor WT after 17 days in culture. Targets pulsed with different peptides are shown on the x-axis, and specific killing on the y-axis. Cultured blood and synovial fluid lymphocytes from this donor specifically recognize and kill targets pulsed with a single peptide derived from the C. trachomatis heat shock protein 70 (sequence KRKSNKIIG), but not other peptides. Recognition is HLA B27 restricted, since peptide-induced killing only occurs for targets bearing HLA B27 w34x. Chlamydia trachomatis-infected HLA B27-positive targets were also recognized (data not shown) w35x, more accurately reflecting the route by which antigenic peptides bind to HLA B27 within cells. Responses to this peptide were not observed in other patients with ReA (P. Bowness and R. Allen, unpublished results), although interestingly HLA B27- restricted CTL responses to a nonamer peptide derived from the Yersinia heat shock protein 60 gene product have recently been described in patients with Yersiniainduced ReA w36x. This finding confirms the principle FIG. 5. Synovial fluid lymphocytes from a patient with ReA were cultured for 21 days with two pools of 10 peptides each (pools nos 5 and 6). CTL activity was then assayed by specific lysis of C1R B*2705 target cells pulsed with individual peptides or with pools at an effector : target ratio of 4 : 1. Duplicates of each condition were averaged. that HLA B27 can present potentially arthritogenic peptides, derived from triggering bacteria, to CTL. However, it is important to recognize that such responses are not found in all patients, and their mere presence does not imply any pathogenic role. As stated previously CD4+ responses restricted by HLA class 2 molecules have been more frequently identified in ReA patients w29, 30, 37x. Furthermore, adoptive transfer studies in transgenic rats have suggested that CD4+ T cells are perhaps more important than CD8+ in transferring disease (w38, 39x but note these data are now qualified by the authors w40x). The findings of both CD4 and CD8 T-cell expansions in the synovial fluid of patients with ReA, and of both CD4 and CD8 T-cell responses to triggering organisms, leads us to a model, shown schematically in Fig. 6, in which both CD4 and CD8 T cells are involved in the pathogenesis of ReA w2x. One possible scenario is that CD4 T cells initiate joint inflammation and CD8 T cells play a role in its perpetuation. This is consistent with epidemiological evidence suggesting that HLA B27 may be a severity marker in ReA w41x. In summary, these studies show that C. trachomatis contains many potential HLA B27-binding epitopes, and that CTL recognizing one such epitope can be found in the blood and synovial fluid of a patient with ReA. Does Cys67 confer an unusual cell biological behaviour on HLA B27? We have recently developed the methodology to study the folding of recombinant HLA B27 together with peptides and b2 microglobulin in vitro w23, 42x. These studies were undertaken first in order to examine the role of the unpaired cysteine at position 67 of the HLA B27 heavy chain in peptide binding and HLA B27 folding, and secondly in order to generate tetrameric complexes of HLA B27, refolded with specific peptides. Such tetrameric complexes can be fluorescently tagged, permitting direct visualization of specific T-cell responses w43x. We have generated for the first time fluorescent-labelled tetrameric complexes of HLA B27 refolded with NP and b2m, and shown that these tetramers specifically stain HLA B27-restricted NP-specific T cells as well as transfected TCRs w23x. Figure 7 shows a tetrameric HLA B27 NP complex, and also specific FACS staining of CTL specific for HLA 27uNP but not HLA B27-restricted CTL recognizing a different peptide. In attempting to generate these complexes, we found that the unpaired Cys67, which lies near to the mouth of B pocket in the HLA B27 a1 helix, inhibited refolding of HLA B27 heavy chains with b2m and peptide w42, 44x. However, when Cys67 was mutated to serine, heterotrimers refolded easily and the resultant complexes were still recognized by HLA B27- restricted NP-specific TCRs w23x. This shows that Cys67 is not essential for the antigen-presenting function of HLA B27.

8 864 P. Bowness FIG. 6. Schema for role of CD4+ and CD8+ T cells in the pathogenesis of ReA. FIG. 7. HLA B27-restricted T cells can be stained with HLA B27-peptide tetrameric complexes. Here the HLA B27 Cys67 residue has been mutated to serine. Specific staining of flu NP-specific CTL by HLA B27(Ser67)uNP tetrameric complexes shows that Cys67 is not essential for recognition by T cells (right-hand panel). We then re-examined the behaviour of the extracellular domains of HLA B27, and to our surprise found that they formed homodimers under the conditions used for refolding other HLAub2mupeptide complexes. Figure 8A shows a Western blot of HLA B27 heavy chain homodimers refolded in vitro, which we have called HC-B27 w42x. Dimerization of truncated HLA B27 molecules in vitro is dependent on disulphide bonding through Cys67 w42x, since site-directed mutagenesis of this residue to a serine almost completely abolished dimerization, and since dimers are converted to heavy chain monomers under conditions that reduce disulphide bonds. Remarkably, these dimers are not associated with b2m, previously considered essential to the structural integrity of all mature HLA complexes. Despite the absence of b2m, HC-B27 maintains the conformation of a significant part of its peptide-binding groove, as evidenced by the binding of monoclonal antibodies such as W6u32, and is capable of binding peptide w42x. Expression of HC-B27 homodimers also occurs in certain transfected cell lines (Fig. 8B) w42x. However standard human cell lines and tissues cannot yet be studied, since the antibodies currently used to detect HC-B27 dimers also react with other HLA alleles.

9 HLA B27 in health and disease 865 A) B) FIG. 8. HLA B27 dimer formation (A) in vitro and (B) in vivo. (a) SDS PAGE and HC10 Western blot of HC-B27 homodimer refolded in vitro (lanes 1 and 3) and HLA B27 (Ser67 mutant)ub2mupeptide heterotrimer complexes (lanes 2 and 4). (B) T2 HLA B27 lysates immunoprecipitated with W6u32, analysed under reducing (1) and non-reducing (2) conditions by Western blot with HC10. FIG. 10. Cartoon of HLA B27 homodimer. FIG. 9. Staining of peripheral blood T lymphocytes from a patient with spondyloarthritis with HC-B27 tetrameric complexes. The PBL were double-stained with PE-conjugated HC-B27 (x-axis) and FITC-stained antibodies to CD3 (A) or CD16 (B) (y-axis). Tetramer+ CD3+ cells are in the upper right quadrant of (A). We are currently trying to generate a monoclonal antibody specific for HC-B27, in order to study its expression in healthy HLA B27-positive individuals and patients with spondyloarthritis. We have also used fluorescent-tagged tetrameric complexes of HC-B27 to obtain intriguing preliminary data showing that the T cells of patients with spondyloarthritis express a ligand for HC-B27. Figure 9 shows an experiment in which tetrameric HC-B27 complexes specifically bind 5 8% of peripheral blood CD3+ T lymphocytes from two patients with spondyloarthritis. T cells from an HLA B27-positive control were also stained. One possible mechanism by which expression of HC-B27 could lead to joint inflammation would be by the presentation of peptides to either CD4+ or CD8+ T cells. Peptide presentation by MHC class 1 molecules to CD4+ T cells has only rarely been reported, but might be favoured for HC-B27 by partial unwinding of the a1 helix (making this complex more similar to a class 2 molecule, see model below). Alternatively, since the complexes used in Fig. 9 were refolded around an HIV-derived peptide, and the patients and control had not been exposed to HIV, another mechanism might be involved. For example HC-B27 might be a ligand for natural killer receptors w45x or other receptors. In order to investigate these observations further, a hypothetical model of the HLA B27 dimer, HC-B27, was built using computer graphics and molecular dynamics, based on the crystal structure of the extracellular regions of HLA B27 w19x. Figure 10 shows in cartoon form a side view of HC-B27, and Fig. 11 shows

10 866 P. Bowness FIG. 11. Molecular model of the HC-B27 HLA B27 heavy chain homodimer. The HC-B27 homodimer is viewed as if looking down towards the cell surface. The a1 and a2 domains of the two HLA B27 heavy chains are depicted in ribbon representation. The disulphide bond between the two Cys67 residues is shown and the (putative) bound peptides are depicted schematically in tube form. The model was created using HLA B27 crystal structure coordinates, the interactive computer graphics program O and the molecular dynamics program Xplor. a top view of the molecular model (N. Zaccai and E. Y. Jones, unpublished results). Figure 11 shows the a1 and a2 helices of two HLA B27 heavy chains viewed looking down towards the cell surface. One heavy chain is shown in green with bound peptide as a ribbon, and the other in blue with peptide in ball and stick form. Such a model shows that dimerization through Cys67 is energetically feasible. Work is now underway to identify conditions that lead to HC-B27 dimer formation in vivo. We hypothesize that dimer formation might only occur under particular conditions. These could include a relative lack of b2m (as in the b2m-deficient mice) or of suitable peptides (as in cell lines with defects in their antigen-presenting apparatus). Similar conditions could occur during infection with intracellular bacteria, some of which are known to trigger ReA. We also propose that cell-surface expression of HC-B27 results in a local inflammatory response, through either cellular or humoral mechanisms. If HC-B27 is indeed a proinflammatory ligand, its production in certain tissues at certain times might explain the pathogenic role of HLA B27. Future experiments can readily be designed to test this hypothesis. For example, it will be crucial to know if HC-B27 dimers are produced in diseased transgenic rodents, and in patients with spondyloarthropathy. Lastly, if expression of an aberrant form of HC-B27 does have a pathogenic role in disease, therapeutic strategies could be developed to inhibit HC-B27 production, or to directly target the molecule, for example, using monoclonal antibodies. Conclusions HLA B27 plays an admirable, perhaps outstanding role in the immune response to viruses, however, it is also directly involved in the pathogenesis of the spondyloarthropathies. Although the exact mechanism is unknown, we have shown that HLA B27 is capable of presenting potentially arthritogenic peptides to CTL. Whilst peptide binding may be important in disease pathogenesis, we have also recently shown that the cell biology of HLA B27 is unusual. A testable new hypothesis, involving production of pro-inflammatory HC-B27 dimers, is proposed to explain disease pathogenesis. Acknowledgements I am particularly grateful to the supervision and encouragement of Professor Andrew McMichael, in whose laboratory the work described was carried out, and to the Medical Research Council who have supported me with a Clinical Training Fellowship, and myself and Rachel Allen with a Clinician Scientist Fellowship. References 1. Brewerton DA, Caffrey M, Hart FD, James DCO, Nichols A, Sturrock RD. Ankylosing spondylitis and HL-A27. Lancet 1973;i: Bowness P. T cell responses in reactive arthritis. Curr Med Lit Rheumatol 1996;15: Hammer RE, Maika SD, Richardson JA, Tang JP, Taurog JD. Spontaneous inflammatory disease in transgenic rats expressing HLA-B27 and human b2m: an animal model of HLA-B27-associated human disorders. Cell 1990;63: Khare SD, Luthra HS, David CS. Spontaneous inflammatory arthritis in HLA-B27 transgenic mice lacking beta 2-microglobulin: a model of human spondyloarthropathies. J Exp Med 1995;182:

11 HLA B27 in health and disease Townsend A, Bodmer H. Antigen recognition by class I-restricted cytotoxic T lymphocytes. Annu Rev Immunol 1989;7: Benjamin R, Parham P. Guilt by association: HLA B27 and ankylosing spondylitis. Immunol Today 1990;11: Breur-Vriesendorp S, Dekker-Says A, Ivanyi P. Distribution of HLA-B27 subtypes in patients with ankylosing spondylitis: the disease is associated with a common determinant of the various B27 molecules. Ann Rheum Dis 1987;46: Rudwaleit M, Bowness P, Wordsworth P. The nucleotide sequence of HLA-B*2704 reveals a new amino acid substitution in exon 4 which is also present in HLA- B*2706. Immunogenetics 1996;43: Lopez-Larrea C, Sujirachato K, Mehr N et al. HLA-B27 subtypes in Asian patients with ankylosing spondylitis. Evidence for new associations. Tissue Antigens 1995; 45: Choi YW, Kotzin B, Herron L, Callahan J, Marrack P, Kappler J. Interaction of Staphylococcus aureus toxin superantigens with human T cells. Proc Natl Acad Sci USA 1989;86: Bowness P. Recognition of antigen and superantigen by cytotoxic T lymphocytes. DPhil thesis. Oxford University, Bowness P, Moss PAH, Tranter H, Bell JI, McMichael AJ. Clostridium perfringens enterotoxin is a superantigen reactive with human T cell receptors Vb 6.9 and Vb 22. J Exp Med 1992;176: Davenport M. The promiscuous B27 hypothesis (letter). Lancet 1995;346: Marker HE, Meyer BK, Wildner G. HLA-B27-derived peptides as autoantigens for T lymphocytes in ankylosing spondylitis. Arthritis Rheum 1997;40: Townsend A, Bodmer H. Antigen recognition by class I-restricted T lymphocytes. Annu Rev Immunol 1989;7: Huet S, Nixon DF, Rothbard J, Townsend ARM, Ellis SA, McMichael AJ. Structural homologies between two HLA B27 restricted peptides suggest residues important for interaction with HLA B27. Int Immunol 1990;2: Bowness P, Allen RL, McMichael, AJ. Identification of T cell receptor recognition residues for a viral peptide presented by HLA B27. Eur J Immunol 1994; 24: Madden DR, Gorga JC, Strominger JL, Wiley C. The structure of HLA B27 reveals nonamer self-peptides bound in an extended conformation. Nature 1991;353: Madden DR, Gorga JC, Strominger JL, Wiley DC. The three dimensional structure of HLA-B27 at 2.1 Å resolution suggests a general mechanism for tight peptide binding to MHC. Cell 1992;70: Jardetzky TS, Lane WS, Robinson RA, Madden DR, Wiley DC. Identification of self peptides bound to purified HLA-B27. Nature 1991;353: Rotzschke O, Falk K, Deres K, Schild H, Rammensee HG. Isolation and analysis of naturally processed viral peptides as recognized by cytotoxic T cells. Nature 1990;348: Bowness P, Moss PAH, Rowland-Jones SL, Bell JI, McMichael AJ. Conservation of T-cell receptor usage by HLA B27-restricted influenza-specific cytotoxic T-lymphocytes suggests a general pattern for antigen-specific major histocompatibility complex class-i-restricted responses. Eur J Immunol 1993;23: Bowness P, Allen RL, Barclay DN, Jones EY, McMichael AJ. Importance of a conserved TCR J alpha-encoded tyrosine for T cell recognition of an HLA B27upeptide complex. Eur J Immunol 1998; 28: Callan MF, Reyburn HT, Bowness P et al. A method for producing monoclonal antibodies to human T-cell-receptor beta-chain variable regions. Proc Natl Acad Sci USA 1993;90: Callan MFC, Reyburn HT, Bowness P, Rowland JS, Bell JI, McMichael AJ. Selection of T cell receptor variable gene-encoded aminoacids on the third binding site loop: a factor influencing variable chain selection in a T cell response. Eur J Immunol 1995; 25: Garboczi DN, Ghosh P, Utz U, Fan OR, Biddison WE, Wiley DC. Structure of the complex between human T-cell receptor, viral peptide and HLA-A2 wcommentx. Nature 1996;384: Garcia KC, Degano M, Stanfield RL et al. An alphabeta T cell receptor structure at 2.5 A and its orientation in the TCR-MHC complex wsee commentsx. Science 1996;274: Hermann E, Yu DT, Meyer BK, Fleischer B. HLA-B27- restricted CD8 T cells derived from synovial fluids of patients with reactive arthritis and ankylosing spondylitis wsee commentsx. Lancet 1993;342: Sieper J, Braun J, Wu P, Kingsley G. T cells are responsible for the enhanced synovial cellular immune response to triggering antigen in reactive arthritis. Clin Exp Immunol 1993;91: Hassell AB, Reynolds DJ, Deacon M, Gaston JSH, Pearce JH. Identification of T-cell stimulatory antigens of Chlamydia trachomatis using synovial fluid-derived T-cell clones. Immunology 1993;79: Allen RL, Gillespie GM, Hall F et al. Multiple T cell expansions are found in the blood and synovial fluid of patients with reactive arthritis. J Rheumatol 1997;24: Bowness P, Bell JI. T-cell receptors and rheumatic disease: approaches to repertoire analysis. Br J Rheumatol 1992;31: Bowness P, Wordsworth BP, Moss PAH, McMichael AJ, Bell JI. T cell receptor usage of synovial lymphocytes from a patient with reactive arthritis. Br J Rheumatol 1992;31:S Bowness P, Allen RA, McMichael AJ et al. HLA B27- restricted peptide presentation to cytotoxic T cells in reactive arthritis. Arthritis Rheum 1995;38:S Bowness P, Allen RA, McMichael AJ. Chlamydia-specific cytotoxic T cell responses in reactive arthritis. Br J Rheumatol 1997;36:S Ugrinovic S, Mertz A, Wu P, Braun J, Sieper J. A single nonamer from the Yersinia 60-kDa heat shock protein is the target of HLA-B27-restricted CTL response in Yersinia-induced reactive arthritis. J Immunol 1997;159: Gaston JSH, Deane KHO, Jecock RM, Pearce JH. Identification of 2 Chlamydia trachomatis antigens recognized by synovial fluid T cells from patients with chlamydia induced reactive arthritis. J Rheumatol 1996;23: Breban M, Fernandez SJ, Richardson JA et al. T cells, but not thymic exposure to HLA-B27, are required for the inflammatory disease of HLA-B27 transgenic rats. J Immunol 1996;156:

12 868 P. Bowness 39. Breban M, Hammer RE, Ricardson JA, Taurog JD. Transfer of the inflammatory disease of HLA-B27 transgenic rats by bone marrow engraftment. J Exp Med 1993;178: Taurog JD, Fox MF, Maika SD et al. Inflammatory disease in HLA-B27 transgenic rats. Immunol Rev 1999;169: Leirisalo M, Skylv G, Kousa M et al. Followup study on patients with Reiter s disease and reactive arthritis, with special reference to HLA-B27. Arthritis Rheum 1982;25: Allen RL, O Callaghan CA, McMichael AJ, Bowness P. Cutting edge: HLA-B27 can form a novel b2-microglobulinfree heavy chain homodimer structure. J Immunol 1999;162: Altman JD, Moss P, Goulder P et al. Phenotypic analysis of antigen-specific T lymphocytes. Science 1996;274: Allen RA, Bowness P, McMichael AJ. Expression and folding of HLA B27 in vitro. Br J Rheumatol 1998;37(Suppl. 1): Lanier LL. Follow the leader: NK cell receptors for classical and nonclassical MHC class I. Cell 1998;92:705 7.