Differential Association of HLA-B*2705 and B*2709 to Ankylosing Spondylitis Correlates with Limited Peptide Subsets but Not with Altered Cell Surface Stability*

In contrast to HLA-B*2705, B*2709 is weakly or not associated to ankylosing spondylitis. Both allotypes differ by a single D116H change. We compared the B*2705- and B*2709-bound peptide repertoires by mass spectrometry to quantify the effect of B*2709 polymorphism on peptide specificity. In addition, shared and differentially bound ligands were sequenced to define the structural features of the various peptide subsets. B*2705 shared 79% of its peptide repertoire with B*2709. Shared ligands accounted for 88% of the B*2709-bound repertoire. All B*2705 ligands not bound to B*2709 had C-terminal basic or Tyr residues. Most B*2709-bound peptides had C-terminal aliphatic and Phe residues, but two showed C-terminal Arg or Tyr. The B*2709-bound repertoire included 12% of peptides not found in B*2705. These had aliphatic C-terminal residues, which are also favored in B*2705. However, these peptides bound weakly B*2705 in vitro, indicating distinct contribution of secondary anchor residues in both subtypes. Differences in peptide binding did not affect the ratio of native to β2-microglobulin-free HLA-B27 heavy chain at the cell surface. Our results suggest that weaker association of B*2709 with ankylosing spondylitis is based on differential binding of a limited subset of natural ligands by this allotype.

In contrast to HLA-B*2705, B*2709 is weakly or not associated to ankylosing spondylitis. Both allotypes differ by a single D116H change. We compared the B*2705and B*2709-bound peptide repertoires by mass spectrometry to quantify the effect of B*2709 polymorphism on peptide specificity. In addition, shared and differentially bound ligands were sequenced to define the structural features of the various peptide subsets. B*2705 shared 79% of its peptide repertoire with B*2709. Shared ligands accounted for 88% of the B*2709-bound repertoire. All B*2705 ligands not bound to B*2709 had Cterminal basic or Tyr residues. Most B*2709-bound peptides had C-terminal aliphatic and Phe residues, but two showed C-terminal Arg or Tyr. The B*2709-bound repertoire included 12% of peptides not found in B*2705. These had aliphatic C-terminal residues, which are also favored in B*2705. However, these peptides bound weakly B*2705 in vitro, indicating distinct contribution of secondary anchor residues in both subtypes. Differences in peptide binding did not affect the ratio of native to ␤ 2 -microglobulin-free HLA-B27 heavy chain at the cell surface. Our results suggest that weaker association of B*2709 with ankylosing spondylitis is based on differential binding of a limited subset of natural ligands by this allotype.
The association of HLA-B27 with ankylosing spondylitis (AS) 1 and other spondyloarthropathies (Spa) ranks among the strongest of any MHC antigen with a human disease, although the mechanism of this association remains unknown. The differential linkage of HLA-B27 subtypes to AS revealed by population studies is expected to provide unique insight into the pathogenetic role of HLA-B27. Whereas B*2705, B*2702, B*2704, and B*2707 are all associated to AS (1), B*2706 and B*2709 are weakly or not associated to this disease (2)(3)(4)(5). In particular, B*2709 is not associated to AS in Sardinia, the only population where it has been found at significant frequency (5). In continental Italy, where this allotype is found at much lower frequency (about 3% of HLA-B27 individuals), a patient with undifferentiated Spa without axial involvement was initially reported (6). More recently, 2 of 47 individuals with Spa in southern continental Italy were found to be B*2709. Again, both patients showed undifferentiated Spa without axial involvement (7). In an independent study (8) a B*2709 individual with sacroiliitis and oligoarthritis was reported. The finding of a few individuals with Spa, but not AS, supports a lower association of B*2709 with this disease, while suggesting some involvement of this subtype with related forms of arthritis. B*2709 differs from B*2705 by a single amino acid substitution, D116H (9). This change, which is unique among known HLA-B27 subtypes, was reported to restrict peptide binding specificity to ligands with C-terminal nonpolar residues (10). This is in contrast with B*2705, which binds peptides with C-terminal basic, aliphatic, and aromatic residues (11). Previous studies did not address the effects of B*2709 polymorphism on other anchor positions of natural ligands or, more importantly, the overall effect of altered binding specificity on the constitutive peptide repertoire.
In the present study we carried out a systematic comparison of the B*2705-and B*2709-bound peptide repertoires to determine the percentage of natural ligands shared by both allotypes. In addition, we sequenced a large number of peptides either common to both subtypes or selectively presented by one of them, to identify the structural features determining differential binding to both subtypes. Finally, we compared the expression of ␤ 2 -microglobulin-free B*2705 and B*2709 heavy chains to test whether differential peptide binding might result in altered cell surface stability MATERIALS AND METHODS DNA Mutagenesis-HLA B*2709 genomic DNA was obtained from a B*2705 genomic construct (12), substituting the XcmI-BssHII fragment (nucleotides 360 -610) by the same fragment containing the 418 G 3 C mutation (resulting in the D116H) obtained from a cDNA coding for B*2709 (a kind gift from R. Sorrentino). The resulting genomic construct was thoroughly sequenced to ensure that it contained only the desired mutation.
Cell Lines and Monoclonal Antibodies (mAb)-HMy2.C1R (C1R) is a human lymphoid cell line with low expression of its endogenous class I antigens (13,14). B*2705-C1R transfectant cells were described elsewhere (12). The B*2709-C1R transfectant cell line was made as described (12) using the B*2709 gene. For some experiments a C1R cell line transfected with B*2709 cDNA was also used (15). C1R cell lines were cultured in Dulbecco's modified Eagle's medium supplemented with 7.5% fetal bovine serum or, for flow cytometry, in RPMI 1640 * This work was supported by Grants SAF99/0055 from the Plan Nacional de IϩD, Grant PM99-0098 from the Ministry of Science and Technology, and an institutional grant to the Centro de Biología Molecular Severo Ochoa from the Fundación Ramón Areces. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Mass Spectrometry (MS) Analysis and Sequencing-The peptide composition of HPLC fractions was analyzed by MALDI-TOF MS using a calibrated Kompact Probe instrument (Kratos-Schimadzu) operating in the positive linear mode, as previously described (25). Dried fractions were resuspended in 5 l of methanol/water (1:1) containing 0.1% formic acid, and 0.5-l aliquots of the sample were deposited onto the stainless steel MALDI probe and allowed to dry at room temperature. Then 0.5 l of matrix solution (saturated ␣-cyano-4-hydroxycinnamic acid in 33% aqueous acetonitrile and 0.1% trifluoroacetic acid) was added and again allowed to dry at room temperature.
Peptide sequencing was carried out by quadrupole ion trap nanoelectrospray MS/MS in an LCQ instrument (Finnigan ThermoQuest, San Jose, CA), exactly as detailed elsewhere (26,27). One peptide was sequenced by MALDI-TOF/TOF using the Applied Biosystems 4700 Proteomics Analyser with TOF/TOF TM optics. The corresponding HPLC fraction was dried and redissolved in methanol/formic acid/water as above. A 1-l aliquot was mixed with 1 l of matrix (␣-cyano-4-hydroxycinnamic acid: 2.5 mg/ml in acetonitrile/0.1% trifluoroacetic acid/water, 1:1). Full scan MS was carried out by MALDI-TOF, and the selected ion peak was subjected to MS/MS. Assignment of the isobaric Ile/Leu or Lys/Gln residues was done on the basis of unambiguous matching with known sequences in the data base. In one case (TRIPKIQKL), assignment of Lys 8 was not possible by this procedure, and the residue was assigned on the basis of identity of retention time in HPLC of the corresponding synthetic peptide, but not of the peptide with Gln 8 , with the natural ligand.
Synthetic Peptides-These were synthesized using standard Fmoc (N-(9-fluorenyl)methoxycarbonyl) chemistry, and purified by HPLC. The correct molecular mass of purified peptides was established by MALDI-TOF MS, and their correct composition and quantification was done by amino acid analysis after hydrolysis in 6 M HCL using a 6300 Amino Acid Analyzer (Beckman Coulter, Palo Alto, CA).
Epitope Stabilization Assay-The epitope stabilization assay used to measure peptide binding was performed as described (28) with minor modifications. Briefly, B*2705 RMA-S transfectants were incubated at 26°C for 22 h in RPMI 1640 medium supplemented with 10% heatinactivated fetal bovine serum. These were then washed three times in AIM-V serum-free medium (Invitrogen), incubated for 1 h at 26°C with various peptide concentrations in the same medium, transferred to 37°C, and collected for flow cytometry after 4 h. HLA-B27 expression was measured using 50 l of hybridoma culture supernatant containing the mAb ME1. Binding of the RRYQKSTEL peptide, used as positive control, was expressed as C 50 , which is the molar concentration of the peptide at 50% of the maximum fluorescence obtained at the concentration range 100 -0.01 M. Binding of other peptides was assessed as the concentration of peptide required to obtain the fluorescence value at the C 50 of the control peptide. This was designated as EC 50 .

Shared and Distinct Subsets of B*2705 and B*2709 Ligands in Vivo-
The B*2705-and B*2709-bound peptide pools were isolated from the corresponding C1R transfectant cells after immunoprecipitation of HLA-B27 and acid extraction. The peptide pools were fractionated by HPLC under identical conditions, and the peptide composition of individual fractions was analyzed by MALDI-TOF MS. The spectrum of each HPLC fraction from one subtype was compared with the correlative, previous and following fractions from the other subtype. In this comparison only ion peaks showing 5% or more of the maximum intensity in the corresponding HPLC fraction of at least one subtype were taken into account. Ion peaks with the same (Ϯ 1) mass-to-charge ratio (m/z) and retention times were assigned as identical ligands in both subtypes. Although because of the complexity of the peptide mixture analyzed this is not always the case, it is reasonable to assume that the overwhelming majority of peptides with the same molecular mass and retention time from these peptide pools are identical. This was confirmed by sequencing such peptides from both subtypes in multiple instances. Ion peaks detected from only one subtype in two independent preparations were assigned as peptides differentially bound to that subtype.
In a number of cases a prominent ion peak in one subtype had a much smaller counterpart in the other subtype. In all these cases the peptides were assigned as shared ligands. When the prominent ion peak showed Ͼ50% of the maximal intensity in the MALDI-TOF MS spectrum, and the intensity of its counterpart in the other subtype was at least 10-fold lower in two independent preparations, the corresponding peptide was assigned as a quantitative difference. This type of assignment must be considered with caution because the ion intensity of a given peptide in complex mixtures may have significant variations and is not an accurate indication of peptide amount. However, with the conservative criteria adopted concerning signal intensity differences and reproducibility we attempted to compensate in part for this limitation.
A total of 1221 and 1094 molecular species were compared from the B*2705-and B*2709-bound peptide pools, respectively

FIG. 2. Amino acid sequence of natural ligands from B*2705 and B*2709.
Peptides found in the peptide pools from both molecules (shared) or only in one of them (B*2705-or B*2709-specific ligands) are indicated. Peptides labeled with one (*) or two (**) asterisks were assigned as quantitative differences predominant in B*2705 or B*2709, respectively. The human proteins and their accession numbers with which complete match of the ligands was found, and the subtype(s) from which each ligand was sequenced are indicated. Most ligands are reported here for the first time. When this is not the case, the corresponding reference is given (11, 30, 31, 39 -42). Peptide sequences were obtained by quadrupole ion trap nanoelectrospray MS/MS except KRFKEANNFL, which was sequenced by MALDI-TOF/TOF MS/MS. B*2709 polymorphism confers to this subtype distinct binding capacity, so that 12% of the peptide repertoire from this subtype is absent from B*2705, and at least 16% of shared ligands are significantly more abundant in B*2709.
B*2709 Ligands Are Not Totally Restricted to Nonpolar Cterminal Residues-A total of 64 peptides were sequenced, mainly by nanoelectrospray MS/MS, from the B*2705-and/or B*2709-bound peptide pools. They included 32 shared ligands, of which 7 were quantitative differences predominant in B*2705 (2) or B*2709 (5), 23 ligands found only in B*2705, and 9 ligands found only in B*2709 (Fig. 2). All the peptides found only in B*2705 had C-terminal basic (Arg, Lys) or Tyr residues, whereas shared ligands, except quantitative differences, had C-terminal nonpolar residues, including aliphatic ones (Ala, Leu, Val, Met) or Phe. These results confirm previous studies (10) showing that B*2709 polymorphism restricts binding to peptides with C-terminal hydrophobic residues. In addition, two novel findings arise from our results. First, this restriction was not absolute because two peptides with C-terminal Tyr and Arg, respectively, were found as natural ligands of B*2709, albeit in much lower amount than in B*2705. The identity of the peptide with a C-terminal Tyr in B*2709 was formally confirmed by MS/MS sequencing (Fig. 2). Second, a number of peptides with C-terminal aliphatic residues (Leu, Ile, Val, Cys) were found only or predominantly in B*2709, indicating that, although these residues are also favored in B*2705, some such peptides do not bind or bind inefficiently to this subtype in vivo. However, no obvious differential usage of particular residues was observed at positions other than the C-terminal one between peptides found only or predominantly in B*2705, and those differentially bound to B*2709.
B*2709-specific Ligands Have Moderate Affinity for B*2705-To confirm that B*2709 ligands not found in B*2705 were indeed poor binders to this allotype, three such peptides were tested for binding to B*2705 in an epitope stabilization assay using RMA-S transfectant cells (Fig. 3) (Fig. 4). The ratio of ME1-to HC10-associated fluorescence was about 18: 1 and 21:1 for B*2705 and B*2709, respectively (Table II), suggesting that about 5% of the molecules of each subtype were expressed at the cell surface as ␤ 2 -microglobulin-free heavy chains.

DISCUSSION
The HLA-B27 prototype, B*2705, binds with few exceptions (29) peptides with an Arg 2 and C-terminal basic, aliphatic, or aromatic residues (11). Whereas Arg 2 is a major motif shared with other HLA-B27 subtypes, these differ among each other in the nature of accepted C-terminal residues. Because B*2702, which like B*2705 is associated to AS presents only peptides with C-terminal aliphatic or aromatic residues (30), it was suggested that HLA-B27 ligands with C-terminal basic residues may not be relevant for susceptibility to AS. The initial reports that B*2706 and B*2709 did not bind peptides with C-terminal Tyr, a residue that is accepted by disease-associated subtypes B*2705, B*2702, and B*2704, suggested that putative arthritogenic peptides might carry this motif (10, 31). This view was challenged by the finding that the AS-associated subtype B*2707 failed to show a C-terminal Tyr motif upon pool sequencing (32). Failure to bind peptides with C-terminal Tyr was recently confirmed for B*2706 (33) and, in the present study, for B*2709, but several novel aspects were revealed by our results now. First, although B*2709 specificity is highly restricted to peptides with C-terminal aliphatic or Phe residues two peptides with C-terminal Arg and Tyr, respectively, were found in the B*2709-bound peptide pool. Although both peptides were much more abundant in B*2705 their amount in B*2709 could be immunologically significant because our MS techniques were less sensitive than CTL, which are known to recognize minimal amounts of peptides presented at the cell surface. Thus, exclusion of peptides with C-terminal basic or Tyr residues by B*2709 is not absolute. Second, some B*2709 ligands with C-terminal aliphatic residues were not found in B*2705 and bound this subtype in vitro with moderate efficiency although B*2705 accepts peptides with these motifs. Because there was no obvious differential features arising from the sequences of B*2709-specific ligands, selective binding of some peptides with C-terminal nonpolar residues to B*2709 may be because of a combination of suboptimal motifs at multiple secondary anchor positions. This would be compatible   (15). This suggests that many shared ligands adopt similar conformation and antigenic features when bound to both subtypes.
The relevance of our results for the pathogenetic role of HLA-B27 in Spa is obviously dependent on the extent to which B*2709 may determine susceptibility for this group of diseases. Because this subtype has been found with significant frequency only in Sardinia, the strongest evidence against involvement of this allotype in AS comes from the absence of B*2709 AS patients in this population (5). That none of the four reported B*2709 individuals with Spa in continental Italy showed typical AS or axial pathology, except for sacroiliitis in one case (7), strongly supports that B*2709 is weakly or not associated with AS but also suggests that this subtype may influence susceptibility to related forms of arthritis.
Without excluding alternative mechanisms (34 -36), our results are compatible with involvement of some peptide(s) belonging to a limited set of the B*2705-bound repertoire in the pathogenesis of AS, as proposed by the arthritogenic peptide hypothesis (37). Because peptides with C-terminal basic residues are not bound to B*2702, our results suggest, in agreement with previous studies (10,31,33), that putative AS-related peptide(s) might have C-terminal Tyr. The capacity of B*2709 to bind low amounts of some peptides with this motif might explain the apparent relationship of this subtype with undifferentiated Spa. The apparent contradiction that the AS-associated B*2707 subtype does not bind peptides with C-terminal Tyr, based on sequencing by Edman degradation (32) might require reassessment using the much more sensitive MS sequencing, to determine whether such peptides might bind B*2707 in small amounts as found here for B*2709.
There is an alternative interpretation to the relationship between the B*2705-and B*2709-bound peptide repertoires and the differential association of these subtypes to AS. It is conceivable that a putative arthritogenic peptide might actually bind better to a subtype not associated to AS than to a disease-associated subtype, because thymic deletion of autoreactive T-cells against such peptide would be favored for the subtype capable to present it more efficiently. For B*2705 and B*2709, this possibility was suggested by a previous study showing that a synthetic peptide derived from a self-protein that elicited CTL responses in B*2705-positive AS patients better than on healthy individuals actually bound better B*2709 than B*2705 (38). According to this view, putative arthritogenic peptides should be among those with C-terminal nonpolar residues that are selectively or predominantly presented by B*2709, such as those found in this study. Discerning between both alternatives requires further research, but our results may help to focus on the limited subsets of HLA-B27bound peptides that are not overlapping among subtypes differentially associated to AS.