Folding of HLA–B27 subtypes is determined by the global effect of polymorphic residues and shows incomplete correspondence to ankylosing spondylitis

Authors

  • Begoña Galocha,

    1. Consejo Superior de Investigaciones Cientificas, Madrid, Spain
    Search for more papers by this author
  • José A. López de Castro

    Corresponding author
    1. Consejo Superior de Investigaciones Cientificas, Madrid, Spain
    • Centro de Biología Molecular Severo Ochoa, Universidad Autónoma, 28049 Madrid, Spain
    Search for more papers by this author
    • Dr. López de Castro has received speaking fees and honoraria (less than $10,000) from the Spanish Society of Rheumatology.


Abstract

Objective

To investigate the maturation and folding of HLA–B27 subtypes and the relationship of these features to ankylosing spondylitis (AS).

Methods

Stable transfectants expressing B27 subtypes and site-directed mutants were used. Maturation/export rates were measured by acquisition of endoglycosidase H resistance. Folding efficiency was estimated from the ratio of unfolded heavy chain to folded heavy chain, which was immunoprecipitated with specific antibodies, in pulse-chase experiments. Association with calnexin was analyzed in coprecipitation experiments. Cytosolic dislocation was estimated by immunoprecipitation of deglycosylated heavy chain after proteasome inhibition. The level of heavy chain expression on unstimulated or interferon-γ (IFNγ)–stimulated cells was quantified by Western blotting.

Results

There was no correlation between the export rate and the association of HLA–B27 subtypes with AS. Three of the 4 AS-associated B27 subtypes showed inefficient folding, but B*2707 folded with the same high efficiency as the non–disease-associated subtypes. Some individual mutations that mimicked subtype polymorphism profoundly influenced folding, but in a context-dependent way. The differences in export and folding rates among B27 variants were unrelated to levels of heavy chain expression in the corresponding transfectants, as indicated by the lack of correlation between the two parameters and by heavy chain up-regulation with IFNγ. Misfolded heavy chain was inefficiently cleared from the endoplasmic reticulum, based on the marginal increase in levels of deglycosylated heavy chain, which resulted from loss of the glycan moiety after cytosolic dislocation, following proteasome inhibition.

Conclusion

HLA–B27 subtype folding is determined by the overall heavy-chain structure, since the effect of a given polymorphism depends on its structural context. Heavy chain misfolding does not explain the association of B*2707 with AS.

The pathogenetic mechanism underlying the association of HLA–B27 with ankylosing spondylitis (AS) (1, 2) remains unknown. HLA–B*2705 shows slow folding and a tendency to misfold, resulting in its accumulation in the endoplasmic reticulum (ER) (3). This suggests that misfolded B27 heavy chain might induce ER stress and activate the unfolded protein response (UPR) and the overload response, leading to activation of NF-κB and inflammation (4). Support for this hypothesis came from observations in transgenic rats, which with the appropriate genetic background and with high copy numbers of the B27 transgene, develop a disease with similarities to human spondylarthropathies, including arthritis and gut inflammation (5, 6). The UPR was found to be activated in macrophages from animals with active disease, and morbidity correlated with UPR activation (7). In cells from humans as well as from transgenic rats, misfolded B*2705 heavy chain forms disulfide-linked homodimers and multimers, which bind the ER stress sensor BiP (8–10).

The role of B27 misfolding in the joint pathology of transgenic rats was challenged by the observation that both the accumulation of misfolded B27 heavy chains and UPR induction were significantly reduced upon overexpression of human β2-microglobulin (β2m) (11). Although gut inflammation was absent in these animals, arthritis, spondylitis, and enthesitis were maintained or even exacerbated (11).

The molecular determinants of B*2705 misfolding are largely unknown. Residues in the B pocket, particularly E45, play a significant role (3, 8). C67, located in this pocket and involved in heavy-chain homodimerization, also influences the assembly/export rate of B27, as determined with the C67S mutation (9). However, the effect may depend on the mutation, since C67A had no influence on folding (8).

The misfolding hypothesis should explain the differential association of B27 subtypes with AS. B*2705, B*2702, and B*2704 are strongly associated with disease (12). B*2707 is usually, but not always, associated with AS (13, 14). B*2706, which is frequent in populations in Southeast Asia, and B*2709, which is prevalent only in Sardinia, are not associated with AS (15–19), although 2 patients with AS who were positive for B*2709 have been described (20, 21) and 1 study suggested that the B*2709 haplotype in Sardinia might include other protective genes (22).

In this study, we analyzed the folding and export of AS-associated and non-associated HLA–B27 subtypes to test the pathogenetic role of B27 misfolding. Using B*2705 mutants, we also determined the influence of residues that are polymorphic among B27 subtypes on these features.

MATERIALS AND METHODS

Cell lines and antibodies.

HMy2.C1R (C1R) is an HLA–A–negative human B cell line that expresses low levels of HLA–B35 (23). Stable C1R transfectants expressing 6 HLA–B27 subtypes (B*2702, B*2704, B*2705, B*2706, B*2707, and B*2709) and 11 B*2705 mutants mimicking subtype polymorphism (Figure 1) were used. These B27 subtypes and mutants have been described previously (24–26). Mutants are designated with the single-letter code of the amino acid(s) being introduced, followed by the corresponding position number(s). Cells were cultured in RPMI 1640 supplemented with 2 mML-glutamine and 10% fetal bovine serum (FBS) (Gibco Life Technologies, Paisley, UK). ME1 (IgG1) is an anti-HLA–B7/B27/B22 monoclonal antibody (mAb) that recognizes heavy chain/β2m/peptide complexes (27). HC10 (IgG2a) is a mAb that recognizes β2m-free class I major histocompatibility complex (MHC) heavy chain (28), both in monomeric and oligomeric forms (8, 29). AF8 (IgG1) is a mAb that recognizes human calnexin (30). GTU88 (Sigma-Aldrich, Steinheim, Germany) is an anti–γ-tubulin mAb (IgG1).

Figure 1.

Export rates of HLA–B27 variants. A, C1R cells expressing B27 subtypes were metabolically labeled for 15 minutes and chased for the indicated times. B27 was recovered from the lysates with the monoclonal antibody ME1. Half of the immunoprecipitates (IP) were digested with endoglycosidase H (Endo H) (lanes 2, 4, 6, 8, and 10) and the other half were not (lanes 1, 3, 5, 7, and 9) prior to analysis by sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE) under reducing conditions. Endo H–resistant (heavy chain [HC] + carbohydrate [CHO]) and Endo H–sensitive (heavy chain – CHO) forms of the B27 heavy chain are indicated. B, C1R transfectants were labeled for 1 minute and chased for the indicated times. B27 was immunoprecipitated with ME1 and analyzed by SDS-PAGE. Results are from a representative experiment of at least 3 independent experiments performed. C, C1R cells expressing the indicated subtypes or mutants were labeled for 15 minutes and chased for up to 4 hours. B27 was immunoprecipitated, digested with Endo H, and analyzed by SDS-PAGE as in A. Radioactive signals of glycosylated and deglycosylated heavy chain were quantified by densitometry and used to calculate the time required to obtain 50% of Endo H–resistant (Endo Hr) heavy chain (top). Numbers at the top of the bars are the actual percentages represented. Values are the mean and SD of at least 3 experiments. Residue changes among subtypes are shown at the bottom; – indicates identity with B*2705.

Western blot analysis.

Approximately 5 × 105 cells were lysed in 0.5% Nonidet P40 (NP40), 50 mM Tris HCl, pH 7.4, 5 mM MgCl2, containing a cocktail of protease inhibitors (Complete Mini; Roche, Mannheim, Germany). After sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE; 10% gels) of whole lysates under reducing conditions, HLA class I heavy chain and γ-tubulin were revealed with HC10 or anti–γ-tubulin mAb, respectively, using peroxidase-conjugated sheep anti-mouse IgG antibody NA931 (Amersham, Little Chalfont, UK). Whenever indicated, C1R cells were stimulated with 100–200 units/ml of IFNγ for 3–6 days.

Metabolic labeling, pulse-chase analysis, and immunoprecipitation.

Cells were incubated with L-Met/L-Cys–free Dulbecco's modified Eagle's medium (Gibco Life Technologies) supplemented with 10% FBS and 2 mML-glutamine for 45 minutes at 37°C. Cells were then pulse-labeled with 500–1,000 μCi/ml of 35S-labeled Met/Cys (Amersham) at 37°C and chased with RPMI 1640 supplemented with 1 mM cold L-Met and L-Cys at 37°C. At each time point, cells were centrifuged, resuspended in 50 μl of phosphate buffered saline, frozen in liquid nitrogen, and stored at −80°C. Cells were lysed in 0.5% NP40, 50 mM Tris HCl, pH 7.4, 5 mM MgCl2, containing a cocktail of protease inhibitors (Complete Mini), centrifuged (16.000g for 10 minutes at 4°C), precleared 3 times at 4°C for 60 minutes each with Sepharose CL-4B beads (Sigma-Aldrich) and 3 μl of normal mouse serum, and then immunoprecipitated with specific mAb and protein A–Sepharose beads (Sigma-Aldrich).

Immunoprecipitates were normalized to equal amounts of trichloroacetic acid–precipitable 35S-labeled protein, washed 3 times with 0.5% NP40, 50 mM Tris HCl, pH 7.4, 150 mM NaCl, and 5 mM EDTA, and then analyzed by 12.5% SDS-PAGE under reducing conditions. In some experiments, the proteasome inhibitor epoxomicin (Calbiochem, Schwalbach, Germany) was added 45 minutes prior to pulse labeling at a concentration of 2 μM. Whenever indicated, preincubation with epoxomicin was performed for up to 3 hours, and the class I heavy chain was analyzed by Western blotting of cell lysates. Tunicamycin (Sigma-Aldrich), an inhibitor of glycosylation, was added at 10 μg/ml overnight and again 45 minutes before pulse labeling. Endoglycosidase H (Endo H; New England Biolabs, Beverly, MA) was added to the immunoprecipitates according to the manufacturer's instructions. Whenever indicated, C1R cells were stimulated with 100–200 units/ml of IFNγ for 3–6 days. Radioactivity in the samples was visualized by fluorography after treatment with DMSO/diphenyloxazole. The autoradiograms were scanned and quantified using Tina 2.09e image analyzer software (Raytest Isotopenmessgeräte, Straubenhardt, Germany).

RESULTS

Export rate of HLA–B27 variants and its relationship to heavy chain structure and AS association.

The intracellular transport of B27 subtypes and mutants was compared by examining their acquisition of Endo H resistance. This enzyme cleaves only the high-mannose moiety acquired by the heavy chain in the ER that has not been further processed in the Golgi apparatus. C1R transfectants were labeled for 15 minutes, chased for 4 hours, and HLA–B27 was immunoprecipitated with ME1. Samples were either left untreated or were treated with Endo H and then analyzed by SDS-PAGE. For B*2705, B*2702, B*2704, and B*2709, Endo H–sensitive heavy-chain bands were observed for up to 2 hours or more of the chase period. For B*2706 and B*2707, these bands were not observed after 1 hour (Figure 1A).

To exclude the possibility that differences in the acquisition of Endo H resistance among subtypes were a consequence of different assembly rates in the ER, we examined the generation of B27 heterodimers at early time points. C1R transfectants were labeled for 1 minute, chased for 0–15 minutes, immunoprecipitated with ME1, and analyzed by SDS-PAGE. Properly folded class I complexes were observed as early as 1 minute during the chase period in all cases (Figure 1B).

Conversion to 50% Endo H–resistant forms occurred for B*2706 and B*2707 within 14 and 17 minutes, respectively. This took ∼30 minutes for B*2702, B*2704, B*2709, and B*0702. B*2705 showed the slowest export rate (Figure 1C). Thus, subtype differences in this feature do not correlate with the AS association.

The molecular determinants of the export rate are complex (Figure 1C). Some mutations resulted in significantly increased time for acquisition of Endo H resistance relative to B*2705 (i.e., A81 and D114), whereas others resulted in faster export rates (i.e., N77I80 and I80A81). The effect of a given mutation was dependent on its structural context (compare A81, N77A81, and I80A81). Thus, the export rate of HLA–B27 variants depends on their specific structure and not on a particular polymorphic residue.

Incomplete correspondence between HLA–B27 subtype folding and AS.

To analyze the formation of B27–peptide complexes relative to the total amount of heavy chain, C1R transfectants were pulse-labeled for 15 minutes. One half of each sample was immunoprecipitated with ME1 and the other half with HC10, and the immunoprecipitates were analyzed by SDS-PAGE (Figure 2A). Large differences in the amount of heavy chain precipitated with either mAb were apparent among the subtypes. The ratio of unfolded heavy chain precipitated with HC10 to the properly assembled heavy chain precipitated with ME1 (hereinafter referred to as the HC10:ME1 ratio) was high for B*2705, B*2702, and B*2704 and was much lower for B*2706, B*2707 and B*2709. This suggested a large but incomplete correspondence between folding efficiency and subtype association to AS.

Figure 2.

Accumulation of HLA–B27 heavy chain in the endoplasmic reticulum in 3 of 4 ankylosing spondylitis–associated subtypes. A, C1R transfectants were labeled for 15 minutes. HLA–B27 heterodimers and free heavy chain were immunoprecipitated (IP) from the lysates with ME1 (lanes 1–6) and HC10 (lanes 7–12), and analyzed by sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE). The arrow and asterisk indicate the B27 heavy chain and a nonspecific band, respectively. Results are from a representative experiment of 3 independent experiments performed. B, C1R cells expressing B27 subtypes or the B*0702 control were labeled for 15 minutes and chased for the indicated times. B27 heterodimers and free heavy chain were recovered with ME1 (lanes 1–5) and HC10 (lanes 6–10), and analyzed by SDS-PAGE. Arrows and asterisks indicate the B27 heavy chain and nonspecific polypeptides, respectively. Results are from a representative experiment of at least 3 independent experiments performed. C, Radioactive signals of assembled heavy chain (ME1) and free heavy chain (HC10) were quantified by densitometry. Results are expressed as the HC10:ME1 ratio. Values are the mean and SD of at least 3 experiments. D, Half-life (T1/2) of the free heavy chain of B27 subtypes were calculated from the pulse-chase experiments for HC10 shown in B. Values are the mean and SD of at least 3 experiments. Numbers at the top of the bars are the actual percentages represented.

To further characterize these differences, C1R cells were labeled as above, chased for 4 hours, immunoprecipitated with ME1 and HC10, and analyzed by SDS-PAGE. The association of heavy chain with β2m was examined by concomitantly measuring the disappearance of free heavy chain and the appearance of assembled class I molecules with time (Figure 2B). Most of the B27 molecules were present as β2m-free heavy chain along the course of the chase period in B*2705, B*2702, and B*2704 (lanes 6–10). For B*2706, B*2709, and B*0702, most of the heavy chain was complexed with β2m at the initiation of the chase (lane 1), with only low amounts of free heavy chain right after the pulse (lane 6), which became very low or undetectable at later times. These results indicate that the 2 non–AS-associated subtypes fold much more efficiently than do the 3 AS-associated subtypes. However, B*2707 folding was similar to that of B*2706 and B*2709.

The HC10:ME1 ratio along the chase period was higher for B*2702, B*2704, and B*2705 than for B*2706, B*2707, and B*2709 (Figure 2C). The half-life of unfolded heavy chain was ∼5–10 times lower for the subtypes not associated with AS and B*2707 than for the other AS-associated subtypes B*2702, B*2704, and B*2705 (Figure 2D). As expected, the unfolded heavy chain of these latter subtypes remained Endo H–sensitive (Figure 3A) and were associated with calnexin (Figure 3B) throughout the chase, indicating that it remained in the ER.

Figure 3.

HLA–B27 polymorphism determination of heavy chain misfolding. A, C1R cells expressing B*2705, B*2702, and B*2704 were labeled for 15 minutes and chased for the indicated times. Free heavy chain was immunoprecipitated (IP) from the lysates with HC10. Endoglycosidase H (Endo H)–resistant (heavy chain [HC] + carbohydrate [CHO]) and Endo H–sensitive (heavy chain – CHO) forms were distinguished by sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE). The asterisk denotes a nonspecific polypeptide. B, C1R cells expressing B27 subtypes and nontransfected cells were labeled for 15 minutes and chased for the indicated times. Unfolded B27 heavy chain was recovered by coimmunoprecipitation with the anticalnexin (CNX) monoclonal antibody AF8 and analyzed by SDS-PAGE. Results are from a representative experiment of at least 3 independent experiments performed. B*2705, B*2702, and B*2704, but not other subtypes, remained associated with calnexin even after 4 hours (results not shown). C, C1R transfectant cells were labeled for 15 minutes and chased for up to 4 hours. B27 complexes and free heavy chain were immunoprecipitated from the lysates with ME1 and HC10, respectively, analyzed by SDS-PAGE, and quantified by densitometry. Densitometry data were used to calculate the HC10:ME1 ratio (solid bars) and the half-life (T1/2) of unfolded heavy chain (open bars). Values are the mean and SD of at least 3 experiments. B27 variants are ordered according to increasing heavy chain expression levels (see Figure 5).

Figure 5.

Different heavy chain expression among C1R transfectants does not account for the maturation and folding properties of HLA–B27 variants. Heavy chain protein expression was determined by Western blotting, quantified by densitometry, and normalized to the respective γ-tubulin control. Heavy chain (HC) levels, relative to B*2705, are compared with A, the time of acquisition of endoglycosidase H (Endo H) resistance (50% Endo Hr), B, the assembled heavy chain to free heavy chain ratio (HC10:ME1), and C, the half-life (T1/2) of unfolded heavy chain. B27 variants are ordered and classified (groups I–III) according to increasing heavy chain protein expression levels. Values are the mean and SD of at least 3 experiments.

Modulation of HLA–B27 folding by subtype polymorphism.

The influence of B27 polymorphism on heavy chain folding was assessed from the HC10:ME1 ratio and the half-life of the free heavy chain of B*2705 mutants, as calculated from pulse-chase experiments. The following observations resulted from these data (Figure 3C). First, individual mutations had distinct effects on folding, relative to B*2705. N77 and D114 had little influence, S77, I80, Y116, H116 (B*2709), and E152 improved folding, and A81 had no effect on the HC10:ME1 ratio, but increased the half-life of the heavy chain. Second, the folding features of double mutants and subtypes could not be predicted from the effects of single mutations, indicating interactive/compensatory effects. For example, the effect of I80 on improved folding was dominant in N77I80 and I80A81, but not in B*2702 (N77I80A81). Moreover, whereas S77 and E152 improved folding, their joint presence in B*2704 resulted in decreased folding efficiency. That the A211G polymorphism of B*2704 might have an additional influence was not ruled out because the S77 and E152 mutants are structurally identical to B*2705 at position 211.

B*2706 differs from B*2705 in the same 3 changes as B*2704 plus D114 and Y116. Among the single mutants that mimic B*2706 polymorphism, S77, Y116, and E152 showed the same efficient folding as B*2706, but D114 folded as inefficiently as B*2705. This was not or was only slightly compensated for in the D114Y116 mutant.

Thus, subtype polymorphism has a profound influence on HLA–B27 folding, and this is determined by the composite effect of multiple changes, rather than by the dominant influence of a single change.

Inefficient clearing of HLA–B27 heavy chains by dislocation to the cytosol.

Misfolded polypeptides in the ER are dislocated to the cytosol, deglycosylated, and then degraded by the proteasome, a process known as ER-associated degradation (ERAD). The deglycosylated class I molecules are revealed upon proteasome inhibition as 40-kd bands that are distinguished from the N-glycosylated heavy chain (43 kd) by SDS-PAGE. To determine whether misfolded B27 heavy chain was subjected to ERAD, C1R transfectants were pulse-labeled for 15 minutes, chased for 1 hour in the presence or absence of epoxomicin, immunoprecipitated with HC10, and analyzed by SDS-PAGE (Figure 4A).

Figure 4.

Inefficient endoplasmic reticulum–association degradation of the HLA–B27 heavy chain. A, C1R transfected and control cells (nontransfected and B*0702-transfected C1R) were labeled for 15 minutes and chased for 1 hour in the presence or absence of epoxomicin. Free heavy chain was immunoprecipitated (IP) from the lysates with HC10, and analyzed by sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE). Epoxomicin and tunicamycin (TN) were used as described in Materials and Methods. Glycosylated (heavy chain [HC] + carbohydrate [CHO]) and deglycosylated (heavy chain – CHO) heavy chains are indicated. Results are from a representative experiment of at least 3 independent experiments performed. Asterisks denote nonspecific polypeptides. Different gels and exposure times are shown; thus, band intensities are not comparable among gels. B, Western blotting of nontransfected and B*2705-transfected C1R cells incubated with 2 μM epoxomicin for the indicated times was performed. Cell lysates were subjected to SDS-PAGE and analyzed with HC10 or anti–γ-tubulin monoclonal antibody. Results are from a representative experiment of 3 independent experiments performed. The mean intensity of the 40-kd band (heavy chain – CHO) in the B*2705 transfectant, relative to the tubulin control in the 3 experiments, was increased a mean ± SD of 1.6 ± 0.2 and 1.3 ± 0.3 after 2 hours and 3 hours of proteasome inhibition, respectively, as compared with the nontransfected C1R cells.

With the inhibitor, the majority of the B27 heavy chain remained glycosylated. Only a faint band of 40 kd, never above the background levels in the nontransfected cells, was detected in all cases. This band might represent the low amount of endogenous class I molecules in C1R cells. When B*2705-C1R transfectants were analyzed in their steady state by Western blotting of whole lysates, a <2-fold increase in the 40-kd band relative to nontransfected cells was observed after 2–3 hours of preincubation with 2 μM epoxomicin (Figure 4B). Thus, misfolded B*2705 heavy chain remains in the ER and only a limited amount, if any, is dislocated to the cytosol and subjected to ERAD in C1R cells.

Differences in heavy chain expression do not account for the differential folding of HLA–B27 variants.

Since stable transfectants with different levels of B27 expression were used in this study, it was critical to determine whether the export and folding differences among B27 variants were related to their respective heavy chain protein expression levels. These were quantified by Western blotting. The transfectants were classified into 3 groups (groups I–III) whose heavy chain expression relative to B*2705 was 0.3–0.5, 0.7–1, and >1.5, respectively.

There was no correlation between the export rate and levels of heavy chain expression (Figure 5A). Some transfectants with similar expression showed significantly different export rates (B*2707 versus D114Y116, D114 versus N77I80, N77A81 versus I80A81, and B*2705 versus A81). Conversely, transfectants with very different levels of heavy chain expression showed similar export rates (S77 and I80A81).

There was also no correlation between the HC10:ME1 ratio and heavy chain expression (Figure 5B). For example, the expression level of B*2709 was ∼3 times higher than that of B*2706 and 2 times higher than that of B*2707, but was similar to that of B*2705, B*2702, and B*2704. However, the HC10:ME1 ratio of B*2709 was similar to those of B*2706 and B*2709 and was much lower than those for the other 3 subtypes. Moreover, this ratio was similar for B*2706 and I80A81, although heavy chain expression of the mutant was ∼10 times higher. Conversely, some variants with similar heavy chain expression levels showed significantly different HC10:ME1 ratios (D114Y116 versus B*2707, D114 and N77 versus B*2709 and N77I80, and N77A81 versus I80A81). Thus, the folding efficiency of HLA–B27 variants in C1R cells is unrelated to levels of heavy chain expression. With few exceptions (B*2704, D114Y116, and A81), the half-life of unfolded heavy chain paralleled the HC10:ME1 ratio and showed no correlation with levels of heavy chain expression (Figure 5C).

Thus, differences in heavy chain protein expression do not provide a trivial explanation for the distinct export rates and folding features of HLA–B27 variants.

Lack of increase in HLA–B27 misfolding upon IFNγ stimulation of C1R cells.

Folding differences among B27 variants depend on their specific structure and do not correlate with the amounts of heavy chains. However, different expression levels of the same variant might influence misfolding, as has been observed for B*2705 in transgenic rats (7, 31). To address this issue, C1R transfectants expressing B*2706, B*2707, and B*2709 were stimulated with IFNγ for up to 3 days or 6 days, and heavy chain expression was quantified by Western blotting. The response of each transfectant to IFNγ was variable among experiments, so that only those resulting in a >5-fold increase in heavy chains were considered. Thus, the stimulation levels cannot be compared among transfectants.

The folding efficiency of B27 was assessed in unstimulated and up-regulated cells by comparing the respective HC10:ME1 ratios after immunoprecipitation of pulse-labeled cells for 15 minutes, at a chase time of 0 hours (Figure 6). For B*2706, B*2707, and B*2709, the maximum increase in heavy chain levels was 11-fold, 14.5-fold, and 7-fold, respectively. However, no increase in the HC10:ME1 ratio was observed in either case. These results confirm that in C1R cells, a significant increase in heavy chain protein levels did not result in increased misfolding of 3 B27 subtypes associated (B*2707) or not associated (B*2706 and B*2709) with AS. Since β2m expression is also up-regulated by IFNγ, this might favor the formation of heavy chain/β2m heterodimers and explain the decrease in the HC10:ME1 ratio that was observed in stimulated cells in 1 experiment.

Figure 6.

Lack of increase in B*2706, B*2707, and B*2709 misfolding after interferon-γ (IFNγ)–mediated stimulation of C1R cells. Heavy chain (HC) induction was estimated by Western blotting of whole lysates with HC10 and anti–γ-tubulin monoclonal antibody and expressed as the increased value of the assembled heavy chain to free heavy chain ratio (HC10:ME1) at a given time (3 or 6 days after stimulation), relative to the same ratio in unstimulated cells. Misfolded B27 heavy chain was estimated as the HC10:ME1 ratio of the heavy chain immunoprecipitated (IP) with HC10 and ME1, respectively, from pulse-labeled cells. For each transfectant, Western blots and sodium dodecyl sulfate–polyacrylamide gel electrophoresis identification of immunoprecipitated heavy chain from individual experiments are shown at the left. The corresponding quantification of the relevant bands is shown at the right. Numbers at the top of the bars are the fold induction relative to unstimulated cells (open bars) and absolute HC10:ME1 ratios (solid bars).

DISCUSSION

Hypotheses about the pathogenetic role of HLA–B27 must account for the differential association of B27 subtypes with AS. The search for biochemical properties that correlate with disease initially focused on the peptide specificity (14, 32–35), assuming that differential presentation of a self-ligand mimicking an external antigen could be a critical pathogenetic event (36). Recently, other subtype features have been comparatively examined. Whereas those studies failed to find a relationship between the surface expression or export rate of B27 subtypes and AS (26, 37, 38), a larger correspondence exists with peptide specificity. Subtypes not associated with AS show a high restriction for ligands with nonpolar C-terminal residues, whereas AS-associated subtypes, except B*2707, accept other residues at this position. The restriction to nonpolar C-terminal residues in B*2706, B*2707, and B*2709 is determined by their lack of D116.

In this study, we show that 1) there is no correlation between the export rate of B27 subtypes and an association with AS, 2) there is a correlation between inefficient folding and AS susceptibility, except for B*2707, 3) there is a strict correlation between inefficient folding and the presence of D116 among the subtypes analyzed, but this residue is not sufficient to determine that behavior, 4) B27 folding is dramatically influenced by polymorphisms outside the B pocket, 5) the export and folding of B27 variants are unrelated to their levels of heavy chain expression, 6) IFNγ-mediated up-regulation of 3 B27 subtypes did not result in increased misfolding, and 7) the B27 heavy chain is inefficiently dislocated to the cytosol for proteasomal degradation.

Our study was performed with stable C1R transfectants expressing different levels of B27, which may potentially affect the extent of misfolding. Therefore, great care was taken to quantify heavy chain expression and to analyze other parameters in that context. The export rate and folding efficiency of the B27 variants did not correlate with, and could not be explained by, differences in heavy chain expression among transfectants. Furthermore, IFNγ stimulation of 3 subtypes that folded very efficiently showed that misfolding was not increased after substantial heavy chain up-regulation.

The lack of correlation between the maturation/export rate and AS was previously reported based on the finding that B*2706, but not B*2709, matured faster than did B*2704 and B*2705 (37). Our study confirms these results and further shows that whereas B*2702 has an export rate comparable to that of the other AS-associated subtypes, B*2707 was as fast as B*2706. Among the subtypes analyzed, a fast export rate (B*2706 and B*2707) correlated with both the absence of H114 and the presence of Y116, 2 positions that influence interactions with the peptide-loading complex (PLC) (37–46). Acidic residues at position 114 determine tapasin dependency in various class I molecules (37, 43). Tyr116 instead of a polar residue was shown to correlate with a strong interaction with the PLC among HLA–B35 and HLA–B15 subtypes (39–41). However, this effect may be context-dependent, since D116 abrogated chaperone interactions in HLA–A68 (42), but had the opposite effect on an HLA–B*0702 mutant (40). Among HLA–B44 subtypes differing only by the presence of D116 (B*4402) or Y116 (B*4405), B*4405 shows both tapasin-independence and a faster export rate. This has been attributed to the hydrophobicity of the F pocket and its influence on peptide binding (45, 46).

Thus, peptide loading, tapasin dependency, and export rate are interrelated and are influenced by residues 114 and 116 in a context-dependent way. This would explain the complex effect of HLA–B27 polymorphisms on the export rate. For example, the D114 mutant showed a slower export rate than did B*2705, as has also been observed on the B*2704 background (37). This is consistent with an effect of this change on increasing tapasin dependency, but delayed export did not take place in the concomitant presence of Y116 (B*2706 and D114Y116). Unlike in HLA–B44, the D116Y change alone in B*2705 did not result in faster export rate of the Y116 mutant. Thus, the relationship of these and other polymorphisms to the maturation kinetics of B27 is not straightforward. Whereas subtype polymorphism influences the export rate, this feature is determined by the joint effect of multiple changes in the structural context of each subtype, rather than by a critical polymorphic residue. In our study, tapasin dependency was not analyzed, but previous studies have examined this issue for some B27 variants (37, 38).

Except for B*2707, there was a correlation between folding efficiency and subtype association with AS. To assess the significance of this finding, it is important to analyze the molecular determinants of B27 folding and to explain why B*2707 folds as efficiently as non–AS-associated subtypes. Our results showed that B27 folding is influenced by subtype polymorphism in a context-dependent way. There was a strict correlation between inefficient folding (B*2705, B*2702, B*2704) and the presence of D116 among the subtypes analyzed, and the Y116 or H116 (B*2709) changes dramatically increased folding efficiency. This is consistent with the effect of D116 on peptide binding and tapasin dependency in HLA–B44 (45, 46). However, D116 was not sufficient to determine inefficient folding in HLA–B27, since multiple mutants conserving this residue (S77, I80, N77I80, I80A81, E152) folded very efficiently, as did other mutants in previous studies (3, 8). Thus, the effect of a polymorphism at position 116 on B27 folding is critically dependent on the structural context of this residue.

B*2705 and B*2709 have been reported not to differ in heavy chain misfolding and intracellular accumulation (47). In that study, the nonlymphoid cell line HeLa was used, and the heavy chain was expressed as a fusion protein with fluorescent reporters, which might affect cell physiology (48). Subtype misfolding might be distinctly affected by cell-dependent differences. Since unrelated cell lines and B27 constructs were used in both studies, a comparison is difficult, and the basis for the discrepancy is unclear.

Misfolding is closely related to the formation of homodimers and multimers in the ER (8, 9) that bind BiP (10), leading to UPR activation (7). Homodimers and UPR were not examined in our study. Nevertheless, although not formally confirmed, it is reasonable to assume that B27 variants with low folding efficiency should be similar in both of these features. First, all the variants analyzed, independently of their folding behavior, have identical Cys residues, some of which are critical for the formation of homodimers (8, 9, 29). Second, this feature is a consequence, rather than a cause, of misfolding, since homodimers can be induced in HLA–A2 by decreasing its rate of assembly at 26°C (9).

Peptide loading and interaction with the PLC are major determinants of MHC class I folding and export, but are probably not the only ones. Otherwise, one would expect a parallel effect of B27 polymorphism on both parameters, which was not observed. The intrinsic tendency of B27 to misfold is likely determined by its primary structure prior to peptide loading. Presumably, a fraction of the newly synthesized heavy chain fails to progress to heterodimer formation and remains in the ER in a misfolded state. The export rate, as measured by acquisition of Endo H resistance among ME1-reactive heterodimers, was not concerned with the fraction of previously misfolded heavy chain. The same applies to the assembly rate, which in our study used ME1 to detect the early formation of B27 heterodimers, but not of misfolded heavy chain.

Low folding efficiency correlated in general with high half-life values of the free heavy chain, with a few exceptions. For example, although B*2704 folding, as assessed from the HC10:ME1 ratio and coprecipitation with calnexin, was more efficient than for B*2702 and B*2705 folding, the half-life of the B*2704 heavy chain was comparable or even larger than the half-life of these 2 subtypes. A similar trend was observed with the D114Y116 and A81 mutants. The basis for this behavior is unclear. However, the HC10:ME1 ratio and the half-life of the unfolded heavy chain reflect different molecular features. Whereas the former parameter indicates the fraction of the heavy chain that folds to heterodimer, the half-life measures the disappearance of heavy chain, either through progression to heterodimer or through degradation/clearing. For example, the clearing rate of unfolded heavy chain might depend, to some extent, on intrinsic molecular features of the B27 variants, so that in some cases, although comparatively less heavy chain fails to progress to heterodimer, this progression might be slower or the misfolded heavy chain might be more resistant to clearing.

It was previously reported that misfolded B*2705 is dislocated to the cytosol for proteasomal degradation in C1R cells (3). In our experience, a limited amount of cytosolic B27 heavy chain, only slightly above the background of nontransfected cells, was detected in the presence of epoxomicin. The background presumably corresponds to the unproductive HLA–B35 heavy chain in C1R (23). The small increase in the deglycosylated heavy chain in the B*2705 transfectants was observed only after prolonged inhibition of the proteasome. Therefore, at least in C1R cells, dislocation of misfolded B27 heavy chain is inefficient and cannot alleviate its accumulation in the ER.

In conclusion, our comparative analysis of HLA–B27 subtypes indicates that inefficient folding alone cannot explain their differential association with AS, due to the efficient folding of B*2707. We speculated that this subtype might be a weaker susceptibility factor for AS, whose influence on disease outcome might be more susceptible to modulation by genetic or environmental factors than other AS-associated subtypes (14). Thus, further studies on B*2707 might shed light on the putative relevance of B27 folding in the pathogenesis of the spondylarthropathies.

In Sardinia, some MHC markers mapping at natural killer (NK)–related genes are increased in healthy individuals carrying B*2705 or B*2709 as compared with patients carrying B*2705 (22). Furthermore, 2 AS patients who were found to be positive for B*2709 have been described (20, 21). These findings apparently challenge the concept that B*2709 is not a predisposing factor for AS (49). However, several circumstances concurring in these studies might prevent us from reaching definitive conclusions. The putatively protective NK-related genes (22) were assigned based on relatively few individuals, and their effect might require confirmation in other populations in which these protective genes should also be found along with B*2705. Of the 2 B*2709-positive AS patients, one also carried the AS-associated allele B*1403 (20), whereas the other (21) had ulcerative colitis, suggesting the presence of additional genes that predispose to AS.

Thus, 2 alternatives remain. First, the specific molecular features of B*2709 would not be relevant. Non-B27 MHC genes would protect B27-positive individuals from disease regardless of the subtype, and their absence would account for the reported B*2709-positive AS patients. This alternative could explain the lack of association of B*2707 with AS in 1 population (13). It would also obviate the concept of B27 misfolding as a pathogenetic feature, explaining why B*2707 is generally AS-associated, as well as the prevalence of arthritis in transgenic rats, in which misfolding was greatly diminished (11). Second, the molecular features may affect the pathogenetic role of B*2709 and other subtypes. The effect of non-B27 genes on disease would be subtype-dependent. This alternative accounts for the fact that in the reported B*2709-positive AS patients, there are other predisposing factors, and it may also account for the lack of association of B*2707 in the Greek Cypriot population (13). Typing these individuals for the protective markers described in the Sardinian population might help to distinguish between the two alternatives. The complex etiology of AS underscores the need for further understanding of both the genetic epidemiology and the molecular biology of B27 subtypes. This, along with identification of other susceptibility genes and their interplay with HLA–B27, may finally explain the pathogenetic role of this molecule.

AUTHOR CONTRIBUTIONS

Dr. López de Castro had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study design. Galocha, López de Castro.

Acquisition of data. Galocha.

Analysis and interpretation of data. Galocha, López de Castro.

Manuscript preparation. Galocha, López de Castro.

Statistical analysis. Galocha.

Acknowledgements

We thank Michael B. Brenner (Brigham and Women's Hospital, Harvard Medical School, Boston, MA), for kindly providing the anticalnexin mAb.

Ancillary