Possession of HLA–B27 is strongly associated with development of spondylarthritides, a group of related diseases including ankylosing spondylitis (AS) and reactive arthritis (ReA). Despite intensive research, the pathogenic role of HLA–B27 remains unclear (for review, see ref. 1). The natural immunologic function of HLA–B27 is to bind antigenic peptides together with β2-microglobulin (β2m) for presentation to the T cell receptor (TCR) of CD8+ cytotoxic T lymphocytes. However, certain features of disease in HLA–B27 transgenic rat (2) and mouse (3) models of spondylarthritis have suggested a possible pathogenic role for HLA–B27 heavy chains independent of β2m. Thus, murine disease requires expression of HLA–B27 in the absence of murine β2m (mβ2m), and can occur in animals with extremely few CD8+ T cells (3). Furthermore, disease onset is delayed and severity reduced by administration of the monoclonal antibody (mAb) HC-10 (4, 5). HC-10 recognizes free human HLA class I heavy chains (6). These results have led to the suggestion that HLA–B27 heavy chains may be directly involved in disease pathogenesis (3). Disease in the rat requires a high copy number of the HLA–B27 transgene (7), and disease cannot be transferred by CD8+ T cells alone (8).
We recently described the formation of β2m-free disulfide-bonded HLA–B27 heavy-chain homodimers, termed HC-B27 (9). Dimerization in vitro is dependent on the presence of the free cysteine at position 67 of the HLA–B27 heavy-chain α1 helix. The β2m-free HLA–B27 heavy chains could also be detected on the surface of HLA–B27–transfected cells (9). If expressed at the cell surface, HLA–B27 heavy-chain homodimers and multimers could play a role in the pathogenesis of spondylarthritis through interaction with either cell-mediated or humoral receptors.
In addition to direct cognate interactions with the TCR, mature class I complexes have been shown to bind several other immunomodulatory molecules, including members of the killer cell immunoglobulin-like receptor (KIR) family, and the immunoglobulin-like transcripts (ILT; also known as leukocyte Ig-like receptors, or LIR ). KIRs are expressed on certain natural killer (NK), T, and NKT cells (for review, see ref. 11). KIRs are polymorphic and demonstrate allele-specific recognition, with the cognate KIR for HLA–B27 being the 3-domain KIR3DL1. ILT/LIRs have a somewhat different expression pattern, with ILT2 expressed on B cells, as well as NK, T cells, and monocyte/macrophages (12). ILT4 is more selectively expressed on dendritic cells, monocytes, and macrophages. ILT2 and ILT4 receptor family members have a broader specificity, with ILT2 recognizing all of the class I alleles previously studied (12). ILT4 binds to most HLA–A and B alleles studied, as well as to the nonclassic HLA–G (13, 14).
Here we show that both HLA–B27 heavy-chain homodimers and receptors for HLA–B27 homodimers are expressed on populations of peripheral blood and synovial monocytes and B and T lymphocytes from patients with spondylarthritis. Control subjects also express receptors for HLA–B27 heavy-chain homodimers. KIR3DL1, KIR3DL2, and ILT4 and at least one additional receptor, but not ILT2, are capable of binding to HLA–B27 heavy-chain homodimers. Such interactions could contribute to joint inflammation and disease pathogenesis in the spondylarthritides.
- Top of page
- PATIENTS AND METHODS
We have shown that patients with spondylarthritis express cell-surface HLA–B27 heavy-chain homodimers (HC-B27) and also cellular receptors for HC-B27. A role for free HLA–B27 heavy chains in the pathogenesis of spondylarthritis has previously been suggested following studies of HLA–B27–transgenic β2m-deficient mice (3). HLA–B27+, β2m−/− murine peripheral blood lymphocytes (PBLs) express cell-surface HLA–B27 heavy chains following concanavalin A treatment (3), and heavy-chain expression can also be detected in the thymus and PBLs of B27+, mβ2m−/− huβ2m+ mice (4). Furthermore, disease onset can be delayed by administration of HC-10, an antibody with specificity for HLA–B27 heavy chains, but not by ME1, an mAb that recognizes HLA–B27 associated with β2m and peptide (4, 5). HLA–B27 heavy chains have been found to preferentially form disulfide-bonded homodimers in vitro (9), and β2m-free HLA–B27 heavy chains can be detected at the cell surface of certain HLA–B27–transfected cell lines with defects in antigen presentation (9).
Here we have shown that patients with spondylarthritis express HC-10–reactive class I heavy chains on monocytes and lymphocytes ex vivo, and that HC-B27 homodimers are expressed on PBMCs. We have also shown that some heavy chains in SFMCs as well as PBMCs are expressed in the form of homodimers. Expression of β2-free heavy chains has been described previously (19), and activated human lymphocytes are known to express conformationally distinct free heavy chains on the cell surface (20), where clustering has been observed (21). Thus, cell-surface heavy-chain expression is not confined to patients with spondylarthopathy or indeed to HLA–B27. However, cell-surface expression of HC-B27 homodimers and subsequent interaction with heavy-chain receptors could differ quantitatively or qualitatively from other class I molecules and contribute to the pathogenesis of the spondylarthropathies. The ability of HLA–B27 to form disulfide bonds through cys 67 may be important in this respect.
We have used fluorescent tetrameric complexes of HC-B27 homodimers to show that patients with spondylarthritis express HC-B27 receptors on the cell surface of populations of monocytes, lymphocytes, and NK cells. Although numbers of HC-B27 tetramer-staining cells in peripheral blood did not differ significantly between patients and controls, we found evidence of increased staining levels for patient synovial fluid T lymphocytes. HLA–B27 homodimers were constructed using a mixture of His-tagged and biotin-tagged heavy chains followed by 2 purification steps. This not only minimized the possibility of contamination with bacterial proteins, but also ensured consistent and optimum tetramerization, since each heavy-chain dimer carried only a single biotin residue. If heavy-chain dimers bearing 2 biotin sites were used, variable interactions with streptavidin could occur. Two biotin residues on a single homodimer could occupy 2 of the 4 binding sites on a single avidin molecule, substantially reducing the avidity of the tetramer reagent, and preventing detection of low-affinity interactions. Alternatively, 2 biotin residues on a single homodimer could bind to different streptavidin molecules, generating multimolecular aggregates capable of artifactual binding.
Differences in tetramer generation may explain the discrepancy between our results for normal PBMCs and those of a recent study using tetramers of HLA–B27 heavy chains, where the only significant binding observed was to monocytes (22). In this study we did not include patients with spondylarthritis, where we have seen particularly strong staining of synovial fluid T cells (see Figure 4C). We have shown that HC-B27 tetramers can bind to KIR3DL1, KIR3DL2, and ILT4 receptors, and almost certainly an additional receptor or receptors expressed on monocytes and lymphocytes, but not to ILT2. Although KIR3DL1 is known to bind to HLA–B27/β2m/peptide complexes (14), it was somewhat unexpected that a β2m-free HLA heavy-chain structure could bind to its cognate KIR. We believe that this interaction is likely to be functionally relevant in vivo, since binding of HLA–B27/β2m heterodimers to KIR-3DL1–transfected cells could be blocked by HC-B27. KIR3DL1 recognition is dependent on residues 77–83 of the class I α1 helix (23). Our data thus suggest that the HC-B27 structure closely resembles that of HLA–B27/β2m/peptide complexes in this region (see below).
Binding of HC-B27 to KIR3DL2 was unexpected but consistently observed in repeated experiments. The level of staining of a transfected cell line was comparable with that observed for DX31, a KIR3DL2-specific mAb. KIR3DL2 did not bind to a number of different HLA–B27/β2m/peptide complexes, consistent with previous studies showing specificity for HLA–A allotypes (18). We suggest that a conformational change upon HC-B27 dimerization permits binding to KIR3DL2. Interestingly, KIR3DL2 is itself usually expressed as a disulfide-bonded homodimer (P140) (18).
HC-B27 tetramers bound to both CD4+ and CD8+ T cells ex vivo. Our results suggest that this binding, as well as that seen at lower levels for NK cells, was due to interaction with KIR3 and related receptors rather than with the TCR. Thus, HC-B27 binding to T cells correlated with KIR expression, and mAb against KIR3DL1 and KIR3DL2 blocked HC-B27 staining of a KIR3DL1-expressing NK line. Interestingly, HC-B27 tetramers stained only a limited proportion of the total number of KIR3DL1/2-expressing cells ex vivo. This finding is consistent with the observation that binding of standard tetrameric complexes to such receptors on lymphocytes is not commonly observed (Bowness P, et al: unpublished observations). HC-B27 FACS staining of a KIR3DL1-expressing NK cell line was, however, increased following stimulation with IL-12 or IL-15. It is likely that binding is dependent on activation-induced changes in KIR expression, possibly including conformational change or a local aggregation of KIR. Similar phenomena have been described for other tetrameric complexes, with HLA–G tetramer staining of CD14+,ILT4+ monocytes significantly enhanced by the presence of an ILT4-binding mAb (14). Recently, virus-specific CD8+ T lymphocytes whose ability to bind cognate MHC/peptide tetrameric complexes varies without any change in the level of cell-surface TCR have also been described (24). There is evidence that this phenomenon may be due to association in lipid rafts (25).
Taken together with these findings, our results also suggest that HLA–B27 homodimers and multimers might be more potent than standard HLA complexes in their interactions with cells expressing non-TCR HLA receptors. We believe that alternative explanations for our data, such as changes in expression of different allelic forms of KIR3DLl (26), are unlikely, although it is possible that we observed stronger staining of purified cell populations because of a reduction in competition between multiple receptors for the same ligand (HC-B27) with a mixed population. Finally, the failure of mAb to KIR3DL1 and KIR3DL2 to block HC-B27 binding to T cells suggests that some of the binding observed was due to expression of an additional heavy-chain receptor or receptors such as LIR-6 (22).
Tetrameric complexes of HC-B27 heavy-chain homodimers bound to T cells of both patients with spondylarthritis and controls, although binding to CD8+ T lymphocytes was not significantly higher in the small group of patients with AS studied. What might be the functional consequences of binding of HLA–B27 heavy chains to KIR receptors expressed on T cells? Although the role of KIR expression on T cells is not yet clearly defined, there is accumulating evidence that KIR expression on CD8+ memory T cells is associated with increased T cell survival (27), and that in vitro engagement of KIRs on T cells can inhibit activation-induced cell death (28). In the context of spondylarthritis, one attractive and testable hypothesis is that engagement by HLA–B27 heavy chains of KIRs on T cells at sites of inflammation, such as the joints, could promote the survival of proinflammatory T cell clones. There is some evidence for KIR expressing CD8+ T cell clones recognizing self-peptides. In this respect, engagement of both KIR3DL1 and KIR3DL2 by HC-B27 could promote the survival of not only T cells expressing the cognate KIR for heterodimers of HLA–B27 but also T cells expressing the KIR for HLA–A3 and HLA–A11.
HC-B27 heavy-chain tetramers were also found to stain CD19+ B lymphocytes in repeated experiments. Because HC-B27 does not bind to ILT2 transfectants, this cannot be explained by binding to ILT2 receptors, which are known to be expressed on B cells and bind all mature HLA molecules studied previously (12). Furthermore, HC-B27 staining of CD19+ B cells ex vivo could not be blocked with an excess of mAb recognizing ILT2. This result is consistent with those recently obtained using HLA–B27 heavy-chain tetramers constructed differently and ILT2 transiently expressed in 293 T cells (22). A likely explanation for our results is that a subpopulation of CD19+ cells expresses a different ILT family member capable of binding HC-B27. Transcription of a number of ILT/LIRs has been demonstrated in B cells (and monocytes; see below), some of which are potentially stimulatory (12, 29). Although the physiologic role of ILT/LIR receptors is not well understood, ILT2 ligation has been shown to inhibit both NK- and T cell–mediated cytotoxicity. Thus, it is possible that, in the presence of stimulatory interactions with other heavy-chain receptors, the failure of HC-B27 to interact with ILT2 could have an overall proinflammatory action.
Since ILT2 is thought to recognize regions on the class I α3 domain (30, 31), loss of recognition by ILT2 of HC-B27 suggests that HC-B27 has a different α3 domain conformation to standard HLA–B27/β2m/peptide complexes (in contrast to the similar interaction with KIR3DL1, suggesting a similar α1/α2 conformation).
We have also shown that HC-B27 homodimers bind to ILT4, an inhibitory receptor expressed on monocytes, macrophages, and dendritic cells (13). ILT4 is known to bind to a number of classic MHC molecules and also the nonclassic molecule HLA–G (13, 14), and is thought to transduce a negative signal (13). Interestingly, HC-B27 binding to monocytes could be only partially inhibited using an mAb recognizing ILT4. A likely explanation for our findings is that HC-B27 binds both to ILT4 and to an additional receptor or receptors expressed on monocytes.
How might HC-B27 homodimers trigger or perpetuate spondylarthritis? One possibility is that infection (for example, with intracellular bacteria known to trigger ReA) could increase cellular expression of HLA–B27 homodimers. Under these circumstances, the balance between normal HLA–B27 heterodimer and homodimer expression could be important in determining disease outcome. ILT2 is broadly expressed on the majority of CD4+ and CD8+ T cells, and receptor ligation can inhibit T cell activation (12). Since homodimers did not bind to ILT2, lack of interaction with this receptor could contribute to pathogenesis by reducing the threshold required for T cell activation, particularly if more homodimer were expressed at sites of inflammation. Alternatively, interaction of HC-B27 with an immunoreceptor that does not recognize standard HLA–B27 heterodimers, such as KIR3DL2 (expressed on both T and NK cells), could have an immunomodulatory action.
Could the interaction of free MHC heavy chains with MHC receptors explain the development of arthritis in animal models of spondylarthropathy? Support for such a general mechanism comes from a preliminary report showing that the introduction into T cells of NKB1, an allelic form of KIR3DL1, produces particularly severe arthritis in HLA–B27–transgenic mice (32). Although rodents do not have KIR receptors, they express a variety of MHC receptors, including the paired inhibitory receptors (33, 34), orthologs of the human ILT/LIR family. Finally, either murine or human free heavy chains expressed at high levels could adopt a related conformation to cell-surface B27 homodimers and interact with similar receptors. This could explain the observations that certain β2m− mice strains can develop arthritis even in the absence of HLA–B27 (35) and that transgenic rats expressing the ser67 mutant of HLA–B27 still develop arthritis albeit at a reduced frequency (36). Moreover, our results do not exclude, but rather complement, existing hypotheses such as the role of “arthritogenic peptides” and “ER misfolding” in disease and suggest an additional level at which molecular characteristics of HLA–B27 may contribute to inflammation through immunoreceptor recognition.
In conclusion, our data show that patients with spondylarthritis express cell-surface HLA–B27 heavy-chain homodimers (HC-B27). We also show that HC-B27 homodimers can interact with class I receptors such as KIR and ILT, and that such receptors are expressed on T and B lymphocytes, NK cells, and myelomonocytic cells. We propose that dimerization may promote the creation of a conformation of HLA–B27 capable of binding to an overlapping but distinct repertoire of immunoreceptors compared with the standard trimeric complex of HLA–B27, β2m, and peptide. These findings raise the novel possibility that interaction of HLA–B27 heavy chains with immunoreceptors on cells of the myelomonocytic cell lineage or lymphocytes might be involved in the pathogenesis of spondylarthritis.