Lymphoblastoid cells express HLA-B27 homodimers both intracellularly and at the cell surface following endosomal recycling

Authors

  • Lucy A. Bird,

    Corresponding author
    1. MRC Human Immunology Unit, Weatherall Institute of Molecular Medicine, John Radcliffe Hospital, Headington, Oxford, GB
    • MRC Human Immunology Unit, Weatherall Institute of Molecular Medicine, John Radcliffe Hospital, Headington, Oxford, OX3 9DS, GB Fax: +44-1865-222502
    Search for more papers by this author
    • The first two authors contributed equally to this work.

  • Chen Au Peh,

    1. MRC Human Immunology Unit, Weatherall Institute of Molecular Medicine, John Radcliffe Hospital, Headington, Oxford, GB
    Search for more papers by this author
  • Simon Kollnberger,

    1. MRC Human Immunology Unit, Weatherall Institute of Molecular Medicine, John Radcliffe Hospital, Headington, Oxford, GB
    Search for more papers by this author
  • Tim Elliott,

    1. Cancer Sciences Division, University of Southampton School of Medicine, Southampton General Hospital, Southampton, GB
    Search for more papers by this author
  • Andrew J. McMichael,

    1. MRC Human Immunology Unit, Weatherall Institute of Molecular Medicine, John Radcliffe Hospital, Headington, Oxford, GB
    Search for more papers by this author
  • Paul Bowness

    1. MRC Human Immunology Unit, Weatherall Institute of Molecular Medicine, John Radcliffe Hospital, Headington, Oxford, GB
    Search for more papers by this author

Abstract

The MHC class I allele HLA-B27 is very strongly associated with development of autoimmune spondyloarthritis, although the disease mechanism remains unknown. Class I molecules classically associate in the endoplasmic reticulum (ER) with β2-microglobulin (β2m) and antigenic peptides for cell surface expression and presentation to T cells. We have previously shown that HLA-B27 is capable of forming β2m-free disulfide-bonded homodimers in vitro. Here we show that HLA-B27 forms disulfide-bonded homodimers in vivo by two distinct pathways. HLA-B27 homodimers form in the ER but appear unable to egress to the cell surface in human cells. Cell surface HLA-B27 homodimers are abundantly expressed in a variety of lymphoid cell lines. Experiments with inhibitors indicate that HLA-B27 homodimers can arise from cell-surface heterodimers via an endosome-dependent recycling pathway. HLA-B27 homodimer expression on the cell surface of 721.220 is dependent on the unpaired cysteine67 and is inhibited by restoration of tapasin function or by incubation with peptides that bind strongly to HLA-B27 heterodimers. Cell surface expressed HLA-B27 homodimers are likely to be immunologically reactive ligands for NK family immunoreceptors and, hence, could play a pathogenic role in spondyloarthritis.

Abbreviations:
β2m:

β2-Microglobulin

BFA:

Brefeldin A

AS:

Ankylosing spondylitis

1 Introduction

MHC class I molecules are expressed on most nucleated cells where they serve to present antigens to cytotoxic T lymphocytes (CTL) 1. They are transmembrane glycoproteins usually comprising a 45-kDa heavy chain (HC), non-covalently associated with β2-microglobulin (β2m) and peptide antigens of 8–10 amino acids. Class I assembly is assisted by an elaborate complex of proteins including calnexin, calreticulin, ERp57 and tapasin 2. Tapasin bridges the class I complex to the transporters associated with antigen processing (TAP) and plays a critical role in class I assembly, as in its absence cell surface MHC class I levels are decreased 3. Tapasin also enhances TAP expression and increases peptide translocation 4. Once loaded with appropriate peptide, the trimeric class I heavy chain/β2m/peptide complex traverses the ER-Golgi apparatus, undergoes further glycosylation and is expressed on the cell surface.

Possession of HLA-B27 (B27) confers susceptibility to a group of closely related arthritic diseases collectively referred to as spondyloarthropathies. Although this association suggests directinvolvement of B27 in disease pathogenesis, the mechanism remains unclear despite intense investigation (reviewed in 5). One possible hypothesis is that arthritogenic peptides are presented by B27 heterodimers to auto-reactive T cells 6. Alternatively, a direct pathogenic role for class I heavy chains has been suggested by a mouse model of spondyloarthritis that requires expression of B27 in the absence of β2m 7. Moreover, disease onset is delayed and severity reduced by treatment with the heavy chain-specific mAb HC-10 but not by an mAb detecting intact β2m-associated B27 molecules (ME-1) 8, 9.

An increasing number of studies have suggested that B27 has abnormal cell biology, which may be of pathogenic significance. For example, B27 molecules are less dependent upon tapasin than other class I molecules for cell surface expression 10. Furthermore, peptide-receptive forms of B27 capable of binding exogenous peptides have been implicated at the cell surface of antigen-presentation proficient B cells 11, and unusually long peptides (up to 33 amino acids in length) have been eluted from B27 12, 13. It has been recently described that, as a result of its unique B pocket architecture, newly synthesized B27 is slow to fold and associate with β2m, and has a tendency to misfold 14. Finally, soluble recombinant B27 heavy chains have been shown to refold to form disulfide-bonded homodimers in the absence of β2m 15. Mutagenesis of the soluble heavy chain molecule showed that dimerization in vitro is dependent on the presence of an unpaired cysteine residue at position 67 (Cys67) 15. We have suggested that cell surface homodimer expression could stimulate inflammatory immune responses and play a pathogenic role in spondyloarthritis 16. Since Allen et al. 17 recently described a number of NK receptors and related immunoreceptors (KIR3DL1, LILRB2, and LIR6) bind B27 homodimers, a key question is whether B27 homodimers are indeed expressed at the cell surface.

Here we show that disulfide-bonded B27 heavy chain homodimers (and multimers) are abundantly expressed on the cell surface and can form following the dissociation of unstable heterodimeric complexes. Cys67 plays a critical role in cell surface homodimer expression.

2 Results

2.1 Cell surface B27 homodimers are expressed in 721.220.B*2705 and C1R.B*2705

The assembly and cell surface expression of B27 have been reported to be relatively independent of tapasin 10. Thus, B27 is expressed at high levels on the cell surface when transfected into the tapasin-deficient cell line 721.220. Surprisingly, when 721.220.B*2705 cell transfectants were analyzed by cell surface flow cytometry, not only did they show strong staining with conformation-dependent antibodies, they also showed staining with the mAb HC-10 that recognizes free class I heavy chains 10, 18. In view of evidence showing that HC-10 also recognized soluble B27 homodimers in vitro15, we wished to determine whether the HC-10 cell surface staining observed on 721.220.B*2705 cells could in part be due to homodimer formation. B27 molecules from surface radio-iodinated 721.220. B*2705 cells were immunoprecipitated with HC-10 and analyzed by SDS-PAGE under reducing and nonreducing conditions. Fig. 1a (left panel) shows that, in the absence of DTT, precipitates contained not only a 45-kDa band corresponding to free heavy chain, but also both a 90-kDa band and higher molecular weight multimers (lane 2). The 90-kDa molecule, which represents 35–45% of the total radioactivity as determined by densitometry, and the higher molecular mass species, largely resolve to a 45-kDa band upon reduction (lane 4). Neither the 90-kDa band nor the larger molecular mass bands were detected in untransfected 721.220 cell lysates (lanes 1 and 3). Fig. 1a (right panel) shows that B27 homodimers are also abundantly expressed on the surface of CIR.B*2705 cells, which have no reported defects in antigen presentation. Although it has been reported that HC-10 can immunoprecipitate heavy chains weakly associated with β2m in certain conditions 19, no β2m was precipitated with HC-10-reactive material in our experiments (Fig. 1a, 4a and data not shown).

2.2 Two-dimensional gel electrophoresis confirms that the 90-kDa band is a disulfide-bonded heavy chain homodimer

Two-dimensional (2D) gel electrophoresis allows the separation of proteins according to their isoelectric point in the first dimension, and molecular weight in the second. Lysates from 721.220.B*2705 cells were subjected to 2D gel electrophoresis followed by HC-10 Western blot. Fig. 1b shows the presence of HC-10-reactive heavy chain spots at 45 and 90 kDa with the same pI (A). This it is unlikely that the 90-kDa components are the product of a B27 monomer complexed to some other molecule besides another B27 monomer. Furthermore, the 90-kDa spots were reduced to 45 kDa in presence of DTT (B). B27 homodimers could be detected in PBMC from a patient with ankylosing spondylitis (AS) (Fig. 1b, C), but not from healthy individuals carrying B39 and B15 (D and E), which also contain an unpaired Cys67.

Figure 1.

Characterization of HLA-B27 homodimer formation. (a) In 721.220.B*2705 and C1R.B*2705 cells most cell surface B27 is expressed as heavy chain homodimers and multimers. Left hand gel; 721.220 cells (lanes 1 and 3) and 721.220.B*2705 (lanes 2 and 4) were surface-iodinated, lysed and immunoprecipitated with HC-10. Precipitates were run on 10% SDS-PAGE in nonreducing (NR) (lanes 1 and 2) and reducing (R) conditions (lanes 3 and 4). Right hand gel; C1R and C1R.B*2705 cells were surface-iodinated, lysed and immunoprecipitated with HC-10 or W6/32. Precipitates were resolved by NR (lanes 1-4) and R (lanes 5-8) 10% SDS-PAGE. A faint 12-kDa band, corresponding to β2m, co-precipitated with HC-10-reactive material in CIR.B*2705 cells but not 721.220.B*2705 cells. (b) 2D gel electrophoresis of whole cell lysates. 721.220.B*2705 cells (A and B) or freshly isolated PBMC (C–E) were lysed and analyzed on a NR (A, C–E) and R (B) 2D gel. Followed by HC-10 Western blot. Arrowheads indicate 45 and 90-kDa species. The multiple spots correspond to different numbers of sialic acid residues present. (c) Dimer formation does not occur post lysis. 721.220.B*2705 were metabolically labeled (B2705*). An aliquot was removed and lysed, heavy chains were precipitated with HC-10 and run on reducing SDS-PAGE (lane 1). The remaining labeled 721.220.B*2705 cells were mixed with an equal number of unlabeled HCT.B2705HA before lysis. The lysate was immunoprecipitated with HC-10 (lane 2) or anti-HA tag rat mAb (lane 3). In lane 4 HCT.B*2705HA were metabolically labeled and immunoprecipitated with anti-HA tag. The figure shows only the 45-kDa portion of the reducing gel. Note that addition of the HA tag increases the molecular weight of the heavy chain (compare lanes 1 and 4). Gels are representative of repeated experiments.

2.3 Dimer formation is not a post lysis event

Although cells were lysed in presence of 10 mM iodoacetamide, we wished to exclude the possibility that dimerization was occurring as a post lysis artifact. We therefore generated an HA-tagged B27 molecule (B*2705HA) and mixed unlabeled HCT.B*2705HA cells with metabolically labeled 721.220.B*2705 cells. Following lysis, proteins were immunoprecipitated with either HC-10 or anti-HA antibody and analyzed by SDS-PAGE under reducing conditions. Fig. 1c shows that labeled heavy chains are precipitated by HC-10 (lane 2) but not by anti-HA antibody (lane 3) in the mixed cell lysate. As a control, lane 4 shows that the anti-HA antibody is capable of precipitating radiolabeled HA-tagged heavy chains. B*2705HA homodimers could be detected under nonreducing conditions (data not shown), proving that the presence of the HA-tag does not itself prevent dimer formation. Thus, dimerization does not occur post lysis under the conditions used in this study.

2.4 Cell surface B27 homodimers are increased in the absence of tapasin

We next asked whether tapasin played a role in cell surface homodimer expression. Fig. 2a shows that in 721.220.B*2705 cells with restored tapasin function cell surface W6/32-reactive heterodimers are substantially increased and HC-10-reactive homodimers are reduced. In fact the ratio of HC-10-reactive homodimers to W6/ 32-reactive monomers decreases twofold in presence on functional tapasin as determined by band densitometry (Fig. 2a, right panel). Also of note, a faint band at 90 kDa was immunoprecipitated with W6/32, suggesting that a proportion of B27 homodimers are recognized by W6/ 32. 721.221.B*2705 cells, which have no defects in antigen processing, expressed greatly reduced levels of HC-10-reactive material.

Figure 2.

Role of tapasin and β2m in HLA-B27 homodimer expression. (a) B27 heavy chain homodimer expression is reduced in presence of functional tapasin. 721.220.B*2705, 721.220.B*2705.htapasin and 721.221.B*2705 cells were surface iodinated, lysed, and immunoprecipitated with HC-10 (lanes 1, 3 and 5) or W6/32 (lanes 2, 4 and 6), and analyzed on a nonreducing SDS-PAGE gel. The right panel indicates band densities, which have been normalized for the background, and the ratio of HC-10-reactive homodimer band to W6/32-reactive monomer band. (b) B27 homodimers form in β2m-negative human cells but are not detected on the cell surface. Untransfected HCT cells (lane 1) and B27 transfected cells (HCT.B*2705) (lane 2) were lysed in 1% Triton X-100, 10 mM iodoacetamide and resolved on nonreducing 10% SDS-PAGE, then immunoblotted with HC-10, followed by anti-mouse IgG-HRP. Arrows indicate 45 and 90-kDa markers. The same cells lines were stained with HC-10, W6/32 or isotype control on ice, followed by anti-mouse IgG-FITC and analyzed by FACSCalibur. Note that cells were cultured in presence of FCS. Results are consistent with repeated experiments.

2.5 B27 homodimers that form intracellularly do not egress to the cell surface

To determine whether β2m plays a role in the formation and expression of B27 homodimers, we transfected the β2m-negative human colon carcinoma cell line HCT with B27. Fig. 2b (left panel) shows several bands around 90 kDa (lane 2) are detected in HCT.B27 whole cell lysates. These reduce to a single band with DTT (data not shown). However, no HC-10 reactivity is detected on the surface by FACS analysis (Fig. 2b, right panel). Thus, in HCT cells cultured under standard conditions, B27 (but not the endogenous HLA class I alleles) forms intracellular homodimers, but these do not egress to the cell surface.

We next returned to study the 721.220.B*2705 cells. In pulse-chase experiments HC-10-reactive B27 homodimers were detected 5 min after pulse labeling, but remained sensitive to endo-H treatment 180 min after the pulse, suggesting they remain entirely in the ER (data not shown). This has also been confirmed very recently by Dangoria and colleagues 20. In contrast, Fig. 3a shows that HC-10-reactive cell surface homodimers are Endo-H resistant (compare lane 2 to lane 1 of Fig. 3a). As expected, cell surface W6/32-reactive β2m-associated B27 heavy chains were also endo-H resistant (compare lane 4 to lane 3). We therefore hypothesized that surface B27 homodimers arise from endo-H-resistant B27 heavy chains that have dissociated from β2m.

2.6 Cell surface B27 homodimers can arise from heterodimers in the presence of brefeldin A

To follow the fate of cell surface heterodimers (and homodimers) we next made use of the protease papain. Papain alone has been shown to strip β2m-free class I heavy chains from the cell surface 21. However, in the presence of 10 mM cysteine, both β2m-free heavy chains and β2m-associated heterodimers are removed 21. Fig. 3b shows that following papain treatment of 721.220.B*2705 cells ME-1 staining of B27 heterodimers is retained (A, dotted line), while surface HC-10-reactive heavy chains are removed (B, dotted line). Following a 2-h incubation, HC-10 cell surface staining is once again detected (B, bold line). This re-acquisition still occurs even if further egress of class I molecules is blocked with brefeldin A (BFA) (D, bold line), albeit at a reduced level. In contrast, if all ME-1-reactive class I molecules are also stripped from the surface in presence of cysteine (E and G, dotted lines), the presence of BFA completely blocks re-accumulation of HC-10-reactive molecules (H, bold line). Thus the HC-10-reactive B27 molecules seen in D, which include homodimers (as shown in Fig. 4c), are derived from cell surface heterodimers.

Figure 3.

Cell surface B27 homodimers are endo-H resistant and can arise from heterodimers. (a) Surface iodinated 721.220.B*2705 cells were lysed and immunoprecipitated with HC-10 or W6/32, digested with or without endo-H and separated by nonreducing SDS-PAGE. The β2m band in W6/32 immunoprecipitates is not shown. (b) 721.220.B*2705 cells were treated with the protease papain in standard RPMI/10% FCS to remove β2m-free heavy chains only (A–D), or stripped of all cell surface class I with activated papain by the addition of 10 mM Cysteine (E–H). Cells were incubated in RPMI/10% FCS (R10) with or without 10 μg/ml BFA for 45 min prior to papain treatment, during papain treatment and for the 2-h recultivation period. Surface staining of ME-1 or HC-10 was assayed by flow cytometry pre-papain treatment (filled histogram), immediately post-papain treatment (dotted line) and following 2-h recultivation at 37°C (thick line). Results are fully reproducible in repeated experiments.

Figure 4.

 Incubation with B27-binding peptide decreases steady state surface homodimer expression and re-acquisition of HC-10-reactivity following stripping of surface heavy chains. (a) Cells were incubated overnight with or without 50 μM EBV-derived peptide epitope. Surface iodinated cells were lysed and immunoprecipitated with HC-10 or W6/32. Samples were resolved on nonreducing SDS-PAGE. (b) 721.220.B*2705 cells were incubated overnight with or without peptides (50 μM). Cells were then stripped of HC-10-reactive heavy chains with papain treatment. Re-acquisition of HC-10-reactive heavy chains was then assayed following 2 h of further incubation in presence of B27 binding peptide epitopes from HIV (dashed line) or EBV (thin line), an A2-restricted `flu epitope (dot-dash line) or in the absence of peptide (thick line). The filled histogram indicates HC-10 staining immediately after papain treatment. The dotted histogram shows HC-10 staining following incubation in presence of B27-binding EBV peptide and BFA. (c) 721.220.B*2705 cells were incubated overnight with or without B27-restricted EBV peptide. BFA was added for 1 h before stripping with papain where indicated. Following treatment with papain, cells were either incubated on ice or at 37°C in presence of BFA ± peptide for 90 min. Cells were surface iodinated and lysates immunoprecipitated with HC-10 or W6/32. Complexes were resolved on nonreducing SDS PAGE. Data are representative of repeated experiments.

2.7 Addition of external peptide stabilizes W6/32-reactive heterodimeric complexes and decreases cell surface expression of HC-10-reactive homodimers

In the absence of tapasin, cell surface B27 complexes decay more rapidly 22, and have reduced thermal stability 23. This is thought to be due to the presence of a greater proportion of suboptimal, low-affinity peptides loaded into B27 complexes. Our observations confirm these findings (data not shown). Based on these results we attempted to stabilize cell surface heterodimeric complexes by culturing 721.220.B*2705 cells with a known high-affinity B27-binding peptide epitope. 721.220.B*2705 cells were incubated overnight in the presence (50 μM) or absence of the B27-restricted EBV-derived peptide RRIYDLIEL. Surface radiolabeled molecules were immunoprecipitated with either HC-10 or W6/32 and analyzed by SDS-PAGE. Incubation with peptide ligand increased the steady-state level of W6/32-reactive heterodimers (densitometry measurements of W6/32 heavy chain band increased from 2115.6 to 2734.7), and substantially decreased the level of surface HC-10-reactive homodimers (2245.7 to 1572.5) (Fig. 4a). Interestingly, a small increase in W6/32-reactive homodimers was observed, suggesting the possibility that these homodimers form loaded with EBV-derived peptide. Fig. 4b shows that, after stripping cell surface heavy chains, incubation of 721.220.B*2705 with high-affinity B27 epitopes inhibits restoration of HC-10 reactivity. Irrelevant peptides did not inhibit HC-10 re-expression. Furthermore, the combination of specific peptide and BFA completely prevents formation of HC-10-reactive molecules. Surface iodination and immunoprecipitation experiments confirmed that the re-acquired HC-10 surface reactivity is composed of heavy chain monomers, homodimers and multimers (Fig. 4c).

2.8 Inhibitors of endosomal acidification prevent re-expression of cell surface HC-10 reactivity following papain stripping

Previous studies have shown that cell surface class I molecules, like class II molecules, can enter the endosomal pathway and be recycled to the cell surface 21, 24, 25. We therefore wished to determine the role of this pathway in homodimer formation. Chloroquine, a lysosomotrophic agent reported to neutralize the pH in endosomal and lysosomal compartments 26, and Bafilomycin, a proton pump inhibitor, were used in experiments with 721.220.B*2705 (data not shown), C1R.B*2705 and the B27 homozygous B cell line Hom2.BCL. Re-acquisition of surface HC-10-reactivity following papain stripping was consistently inhibited by Chloroquine or Bafilomycin (Fig. 5).

Figure 5.

 Inhibitors of endosomal acidification prevent re-expression of cell surface HC-10 reactivity following papain stripping. C1R.B*2705 and Hom2 cells were incubated with BFA and Chloroquine or Bafilomycin for 1 h before and 2 h after papain treatment. An aliquot was incubated on ice immediately post papain treatment (dotted histogram), the remainder was incubated at 37°C for 2 h in R10 + BFA ± inhibitors. Cells were then stained with ME-1 or HC-10 and assayed by flow cytometry.

2.9 Cys67 is necessary for cell surface HLA-B27 homodimer expression

Soluble HLA-B27 molecules refolded in vitro required Cys67 for homodimer formation 15. However, full-length B27 contains two additional unpaired cysteine residues, position 308 in the transmembrane domain, and position 325 in the cytoplasmic tail. We constructed a series of mutants, exchanging cysteines 67, 308, and 325 for serines, singularly and in combination. Stable transfectants were made in 721.220 cells and all mutants gave comparable high expression of cell surface B27 (data not shown). Cell surface radiolabeling of these mutant cell lines (Fig. 6a) shows that Cys67 is required for expression of surface B27 homodimers. Thus, when only Cys67 is retained, and in the absence of cysteines 308 and 325 (lanes 1 and 3), the majority of surface heavy chain is present as dimers or high molecular weight multimers. However, when all three cysteines were abolished at once (lanes 2 and 4), dimers were not detectable on the cell surface. Interestingly, in the latter mutant an increase in free heavy chain monomers was observed and expression of multimers was unchanged. These results show that Cys67 is necessary and, by itself, adequate for cell surface expression of B27 heavy chain homodimers. This contrasts with dimers of mouse alleles, which lack an unpaired cysteine in the α1 helix, but dimerize via cytoplasmic cysteines 27.

We next determined if this pattern was present in intracellular heavy chain forms. In HC-10 Western blot analysis of 721.220.B*2705 lysates a number of bands (three most prominent) around 90 kDa are detected. The variation in size could not be accounted for by the differing glycosylation states of the heavy chains, since upon reduction all bands collapsed to a single band at 45 kDa. It is therefore more likely that the bands correspond to homodimers whose differing patterns of intramolecular disulfide bonds result in slightly different mobilities through the gel. Mutating the cysteine residues in B27 altered the pattern of dimer bands, as seen in a nonreducing Western blot (Fig. 6b, upper panel). Several features were notable. The complete spectrum of bands as seen in wild-type B27 was present only when Cys67 was preserved (compare lane 4 to lanes 2 and 3, Fig. 6b, upper panel). Loss of Cys67 led to fewer dimer bands (lanes 3 and 4). However, these three cysteines cannot be exclusively involved in dimerization since faint dimer bands were visible even when all three cysteine residues had been substituted (lane 3), suggesting the involvement of other cysteine residues in the α2 and α3 domains, which are usually involved in intra-domain disulfide bonds. Therefore, althoughall cysteines can be involved in intracellular dimer disulfide links, Cys67 is critical for formation of cell surface expressed B27 homodimers.

Figure 6.

 Cys67 is required for cell surface expression of B27 homodimers in 721.220.B*2705 cells. (a) 721.220 cells expressing mutant B*2705 constructs (the unpaired Cys67 is retained in 721.220.B*2705.C308S.C325S, all unpaired cysteines are mutated to serine in 721.220.B*2705.C67S.C398S.C325S) were surface iodinated and immunoprecipitated with HC-10. Complexes were resolved on 10% SDS-PAGE in nonreducing (NR) and reducing (R) conditions. (b) HC-10 Western blot analysis of 721.220 cell lysates expressing wild-type B*2705 or cysteine to serine mutatants.

3 Discussion

In this study we have shown that β2m-free B27 homodimeric complexes are abundantly expressed in lymphoblastoid cells. We have identified two distinct pathways of homodimer formation, and provide evidence that cell surface homodimers arise following dissociation of cell surface heterodimeric forms. Such dimers are likely to be of biological relevance. Allen et al. 17 recently demonstrated that HLA-B27 homodimers are ligands for a number of receptors, including KIR3DL1, and LILRB2. We have confirmed and extended these findings to show that KIR3DL2 is aligand for B27 homodimers, but not heterodimers 28. Furthermore, HLA-B27 homodimer tetramers bound to significant populations of T and B lymphocytes and cells of the myelomonocytic lineage, with enrichment of homodimer binding cells in the synovial fluid of a patient with spondyloarthropathy 28.

Homodimerization of other MHC alleles has been reported previously 27, 29; however, the HLA-B27 homodimers described here are almost certainly both qualitatively and quantitatively different given the key involvement of Cys67 (see below) and other B pocket residues 20. Interestingly, cell surface expression of β2m-free class I heavy chains has previously been described for several cell types, including activated and resting T cells [3032, EBV-transformed B cell lines 33, 34, and spleen cells from HLA-B27/human β2m transgenic mice 8. However, these studies did not look for evidence of homodimerization. Very recently, Malik et al. 35 described a population of peptide-containing β2m-free HLA-B27 monomers on the surface of transfected lymphoblastoid cell lines thatreact with the mAb MARB4. The MARB4-reactive species are distinct from the HC-10-reactive molecules and are contained within the ME-1-reactive cohort. It is thus likely that at least two forms of β2m-free B27 heavy chains are expressed at the cell surface.

Intracellular HC-10-reactive homodimers that form soon after heavy chain synthesis in the ER do not obtain a mature phenotype and are likely to be targeted for degradation. In agreement with our findings, Dangoria et al. 20 have also recently described B27 homodimer formation within the ER in association with B27 misfolding, and have proposed that this may lead to a pro-inflammatory stress response 14, 36.

In contrast, we provide several lines of evidence indicating that cell surface expressed B27 homodimers arise from cell surface heterodimers: (1) intracellular B27 homodimers do not egress to the cell surface in β2m-negative HCT cells; (2) after stripping of cell surface B27 homodimers and in the presence of BFA, cell surface heterodimer expression is both necessary and sufficient for homodimer re-expression; and (3) in the tapasin-deficient cell line 721.220 surface homodimer expression is increased, but can be inhibited by incubation with high-affinity B27-binding peptide epitopes. In the absence of tapasin, B27 complexes are less thermostable consistent with loading of suboptimal peptide ligands 23. Thus, we propose that unstable but transport-competent B27/β2m/peptide complexes act as a major source of free heavy chains. It has been reported that incubation of B27-expressing murine cells with B27 ligands increased the half-life of W6/32-reactive molecules 37. It is therefore probable that exogenous peptides prevented the dissociation of B27 heterodimers, thus decreasing the generation of B27 free heavy chains and homodimers. Alternatively, exogenous peptides could have induced the refolding of free heavy chains with serum β2m 38. Our data raise the possibility that B27-binding peptides could have a hitherto unsuspected role in disease pathogenesis through modulation of homodimer expression levels. Interestingly, the incidence of arthritis in B27 transgenic rats isreduced by co-expression of a minigene encoding a high-affinity B27 peptide 39. It was suggested that this peptide blocked the presentation by B27 heterodimers of a putative arthritogenic peptide to T cells. However, our results, together with the finding that B27 transgenic rats express HLA-B27 homodimers (Bird et al., unpublished data), suggest an alternative interpretation. Thus, co-expression of a high-affinity peptide might reduce disease indirectly by stabilizing B27 heterodimers and hence inhibiting B27 homodimer expression.

How and where might surface-expressed B27 heavy chains dimerize in vivo? Our data suggest that cell surface B27 heterodimeric complexes may undergo dissociation at the cell surface ormore likely within endocytic compartments, followed by formation of disulfide bonds between free heavy chains and subsequent re-expression. Inhibition of cell surface homodimer expression with Bafilomycin and Chloroquine suggest that endosomal acidification and recycling may indeed be required. Recycling of class I via endosomal compartments has been well described 21, 24, 25. Cresswell et al. 40 have recently described an IFN-γ-inducible lysosomal thiol-oxidoreductase (GILT) in late endocytic compartments of APC. Although this enzyme is predominantly concerned with reduction of disulfide bonds, it is not inconceivable that this or a similar enzyme could promote disulfide bond formation under appropriate conditions. Since spondyloarthritis is frequently triggered by infection with intracellular bacteria, oxidizing conditions within activated macrophages infected with such pathogens may promote the formation of disulfide bonds between heavy chains. Alternatively, arthritogenic pathogens could stimulate homodimer formation by interfering with tapasin or other proteins of the peptide-loading complex. Interestingly, Boisgerault et al. 41 showed that Shigella infection induced clustering of B27 molecules during macropinocytosis and alterations in peptide presentation.

Cys67 in the B27 α1 helix has been shown to have an important role in the formation of soluble recombinant homodimers 15. Here show that Cys67 is necessary and sufficient to enable the formation of B27 cell surface homodimers. Since we have shown a critical role for endosomal recycling in cell surface homodimer formation, it may be of relevance that, of the three unpaired cysteines, only Cys67 is accessible to the endosomal milieu. The B27 mutant with only Cys67 retained was capable of generating at least three intracellular disulfide-bonded forms, comparable to wild type. These findings imply that while Cys67 can form disulfide bonds with intra-chain cysteines of the α2 and α3 domains, these misfolded forms do not exit the ER in human cells. Cys67 has previously been shown to be highly chemically reactive, perhaps in part due to polarization from a neighboring Lys7042. In this respect we have shown that presence of Cys67 alone in B15 and B39 heavy chains is not sufficient to form homodimers. However, Dangoria et al. 20 recently showed that B7 heavy chains can dimerize. Recent reports indicate that C67S mutation decreases the stability of B27 molecules, and promotes unfolding to high molecular weight species 43, 44. Although cells with the C67S mutation express HC-10-reactive heavy chain monomers and high molecular weight multimers, these mutant heavy chains were unable to form stable homodimers at the cell surface.

We have shown that B27 forms both intracellular and cell surface heavy chain homodimers. We propose that cell surface B27 homodimers and multimers may be physiologically significant and important to the pathogenesis of spondyloarthropathy. Our results could explain the recent observation that patients with spondyloarthritis express significantly elevated levels of HC-10-reactivity by FACS analysis 45. We have shown that, for one patient with AS, this HC-10-reactive material includes HLA-B27 homodimers 28. While cell surface heavy chain homodimer and multimer expression has been described for other MHC alleles, we propose that it is quantitatively and/or qualitatively different for HLA-B27. The key role of Cys67 would support the latter. In this regard qualitative differences in receptor engagement of B27 homodimeric and heterodimeric complexes have been described. Thus B27 homodimers do not engage the inhibitory immunoglobulin-like receptor LILRB1, while both B27 homodimers and heterodimers to bind KIR3DL1, LILRB2, and LIR6 receptors 17, 28. Furthermore, we have shown that populations of lymphocytes, monocytes and NK cells carry receptors for B27 homodimers 28. Since infection with certain intracellular pathogens is known to trigger spondyloarthritis, one possible scenario is that such an infection in a susceptible individual leads to increased expression of B27 homodimers and multimers. Our data suggests novel mechanisms by which this could be brought about, for example by reduction in the quality of peptide cargo loaded onto heterodimeric B27 (thus promoting generation of homodimers from cell surface heterodimers) or by alteration of the redox conditions in endocytic compartments.

4 Materials and methods

4.1 Cell lines

The cell line 721.220 was previously derived by mutagenesis of a human LCL in which HLA-A, HLA-B are deleted, and low levels of a mutant tapasin are expressed, although HLA-Cw1 remains intact on one chromosome 46. 721.220 transfected with HLA-B2705 alone or transfected with both HLA-B2705 and human tapasin have been described previously 10. C1R.B27 is a B lymphocyte cell line transfected with HLA-B27 lacking endogenous expression of class I–A and –B alleles. The β2m-negative HCT human colon carcinoma cell line 47was kindly provided by Dr. M. Zuniga (University of California, Santa Cruz). RSV.5neo.B2709 cDNA was kindly provided by Dr. Lopez de Castro. Hom2 BCL (HLA Type: A3, B*2705, Cw1, DRB1*01, DRB1*05). Cells were maintained at 37°C, 5% CO2 in RPMI 1640 (Gibco) supplemented with 10% fetal calf serum (Gibco), 2 mM glutamine, 50 U/ml penicillin, 50 U/ml streptomycin (R10).

4.2 Antibodies

The mAb HC-10 (IgG2a) reacts with a determinant on β2m-free heavy chains of HLA-B, -C and some -A alleles 18 and was used in Western blots and immunoprecipitations. W6/32 is specific for HLA class I heavy chains associated with β2m 48. ME-1 recognizes HLA-B27, -B7, -B42, -B67, -B73, and -Bw22 49. Antibodies were purified from hybridoma supernatants by protein A-Sepharose affinity column as described previously 50. In immunoprecipitation experiments with rat anti-HA tag antibody (clone 3F10. Roche) protein-G-Sepharose was used.

4.3 Cell lysis and Western blot analysis

Cell lysis (5×106cells/sample) was carried out in 500 μl lysis buffer [20 mM Tris pH 8.0, 150 mM NaCl, 1% Triton X-100, 10 mM iodoacetamide, 2 mM phenylmethylsulfonyl fluoride (PMSF)] on ice for 30 min. Nuclei and cell debris were pelleted (13,000 rpm 10 min at 4°C). A 20-μl aliquot of lysate was boiled (3 min) in SDS sample buffer (50 mM Tris-HCl, 2% SDS, 10% v/v glycerol, and bromophenol blue) with or without reducing agent dithiothreitrol (DTT) (100 mM) and run on a 10% SDS-PAGE gel 51. Proteins were transferred onto nitrocellulose membranes (Amersham Hybond-CTM Super) using Hoefer SemiPhor Unit (Amersham Pharmacia Biotech). The membranes were blocked overnight at 4°C in blocking buffer [PBS containing 5% (w/v) skimmed milkpowder, and 0.1% Tween-20 (Sigma)]. Heavy chains were detected using HC-10 diluted in blocking buffer, incubated for 1 h at room temperature. Following washes with blocking buffer, binding was detected by incubating membranes with peroxidase-conjugated goat anti-mouse Ig (Dako). Binding was visualized using ECL reagent (Amersham) and exposure to film (Biomax MR-1, Kodak).

4.4 2D gel electrophoresis

Cell lysates were resolved by isoelectric focusing in the first dimension using IPGphor dry strips pH 4–7 or 3–11 according to manufacturers instructions. Followed by 10% SDS-PAGE.

4.5 Cell surface radiolabeling and immunoprecipitation

For radioiodination of cells (1×107 cells/condition) iodogen pre-coated tubes (Pierce) were used. Cells were labeled in Dulbecco's phosphate-buffered saline (DPBS) with 500 μCi 125I at room temperature for 15 min. Washes were performed with DPBS containing 5 mM KI. The cell pellets were lysed and cleared as above, plus rotation at 4°C with 80 μl protein A-Sepharose beads (10% v/v, Sigma) for 1 h. Heavy chains were then immunoprecipitated with 10 μg/ml HC-10 or W6/32 for 1 h, followed by protein A-Sepharose beads (60 μl) for an additional 1 h (4°C). Beads were washed three times with lysis buffer, and a final wash with 10 mM Tris. mAb-Ag complexes were eluted from the beads by boiling (5 min) in SDS-PAGE sample buffer, and supernatants (10–20 μl) were analyzed on 10% SDS-PAGE gels. For Endoglycosidase H (endo-H) digestion Ab-Ag complexes were eluted in 50 mM sodium citrate, 0.2% SDS buffer (pH5.5) and incubated overnight at37°C with 1 μl endo-H (Roche). Gels were fixed in 20% methanol, 10% acetic acid and dried down on 3MM paper (Whattman) to be visualized by autoradiography on Kodak X-OMAT film or using a phosphoimager.

4.6 Metabolic labeling and immunoprecipitations

Cells (1×107 per condition) were starved in RPMI (Gibco) lacking methionine and cysteine and supplemented with 5% dialyzed FBS (Sigma). After 1 h at 37°C, cells were incubated with [35S]methionine (100 μCi/ml) for 1 h at 37°C. Unincorporated [35S]methionine was removed by washing once in RPMI containing 1 mM cold methionine/cysteine. For the mixing experiment labeled cells were mixed with unlabeled cells before lysis. Immunoprecipitations were carried out as above.

4.7 Protease treatment of cells and re-expression experiments

Papain treatment was achieved by incubating cells with 150 U/ml papain from Carica papaya (Roche) for 45 min at 37°C in RPMI 1640/10% FCS medium. Peptides used in re-expression experiments were HLA-B*2705-restricted EBV EBNA 3C peptide (RRIYDLIEL), and HIV-1 gag peptide (KRWIILGLNK) and HLA-A2-restricted flu matrix peptide (GILGFVFTL). Peptides were dissolved in DMSO and incubated at 50 μM final concentration pre and post papain treatment. BFA (Sigma) was used at 10 μg/ml. Chloroquine (Sigma) and Bafilomycin (Sigma) were used at final concentrations of 100 μM and 0.01 μM, respectively, pre- and post-papain treatment.

4.8 Generation and expression of mutant HLA-B27

Mutagenesis experiments were carried out on full-length B27 cDNA contained within the RSV5.neo plasmid using Stratagene QuikChangeTM Site-Directed Mutagenesis Kit according to manufactures instructions. Constructs were checked by restriction digests and DNA sequencing of both strands (Sequencing facility, Biochemistry, Oxford University). 721.220 cells were transfected with the mutated cDNA constructs by electroporation (220 V, 950 μF) at 4°C. Cells were left to recover in RPMI/10% FCS for 36 h before adding antibiotic G418 at 0.6 mg/ml for selection. Successful transfection was demonstrated on G418-resistant cell colonies by flow cytometry for the expression of B27 transgenes with mAb 116.5.28 anti-Bw4 epitope (SaxonEurope, Suffolk, GB). If necessary cultures were enriched with positively expressing cells using magnetic beads coated with goat anti-mouse IgG (Dynabeads®, Dynal Ltd.) and anti-Bw4 antibody so that comparable expression levels of all mutant HLA-B*2705 constructs was achieved. Media for transfectants were supplemented with 0.5 mg/ml G418 (Gibco).

4.9 Flow cytometry

Cells were stained (5×105–1×106 cells/condition) on ice with 1–10 μg/ml of antibody for 30 min. mAb HC-10, W6/32 and ME-1 were used as primary antibodies followed by goat anti-mouse IgG conjugated to fluorescein isothiocyanate (FITC) (Dako). Washes were performed with PBA (PBS, 1 mg/ml BSA, 0.05% sodium azide) at 4°C. Cells were resuspended in PBA or fixedin 2% paraformaldehyde before cytometric analysis on a Becton Dickinson FACSCalibur using CellQuest software.

Acknowledgements

This research was funded by the MRC (UK) (L.A.B., C.AP., P.B., A.J.McM.), the Arthritis Research Campaign (S.K.) and the Welcome Trust (T.E.).

Footnotes

  1. 1

    WILEY-VCH

  2. 2

    WILEY-VCH

  3. 3

    WILEY-VCH

  4. 4

    WILEY-VCH

  5. 5

    WILEY-VCH

  6. 6

    WILEY-VCH

Ancillary