Identification of novel HLA-B27 ligands derived from polymorphic regions of its own or other class I molecules based on direct generation by 20 S proteasome.

HLA-B27 is strongly associated with ankylosing spondylitis. Natural HLA-B27 ligands derived from polymorphic regions of its own or other class I HLA molecules might be involved in autoimmunity or provide diversity among HLA-B27-bound peptide repertoires from individuals. In particular, an 11-mer spanning HLA-B27 residues 169-179 is a natural HLA-B27 ligand with homology to proteins from Gram-negative bacteria. Proteasomal digestion of synthetic substrates demonstrated direct generation of the B27-(169-179) ligand. Cleavage after residue 181 generated a B27-(169-181) 13-mer that was subsequently found as a natural ligand of B*2705 and B*2704. Its binding to HLA-B27 subtypes in vivo correlated better than B27-(169-179) with association to spondyloarthropathy. Proteasomal cleavage generated also a peptide spanning B*2705 residues 150-158. This region is polymorphic among HLA-B27 subtypes and class I HLA antigens. The peptide was a natural B*2704 ligand. Since this subtype differs from B*2705 at residue 152, it was concluded that the ligand arose from HLA-B*3503, synthesized in the cells used as a source for B*2704-bound peptides. Thus, polymorphic HLA-B27 ligands derived from HLA-B27 or other class I molecules are directly produced by the 20 S proteasome in vitro, and this can be used for identification of such ligands in the constitutive HLA-B27-bound peptide pool.

posed of the MHC heavy chain, ␤2m, and peptide, is then exported to the cell surface where it can be recognized by cytolytic T lymphocytes.
Proteasomes are multicatalytic complexes located in the nucleus and cytosol. In eukaryotic cells their catalytic core, or 20 S proteasome, consists of 28 subunits organized in heptameric sets to build a four-ring barrel structure. Each of the external rings contains seven noncatalytic ␣ subunits, whereas each of the internal rings contains seven ␤ subunits, three of which, ␤1, ␤2 and ␤5, are catalytic (2). In vertebrates these subunits can be cooperatively replaced by homologous interferon-␥-induced subunits ␤1i, ␤2i, and ␤5i, to form the immunoproteasome (3). The 20 S catalytic core, when bound to the PA700 (19 S) activator, results in the 26 S proteasome, which is involved in ATP-dependent digestion of ubiquitinated proteins (4 -6). The 20 S proteasome can also interact with the PA28 (11 S) regulator, which increases dual cleavage of polypeptide chains (7)(8)(9)(10) and antigen presentation (11). The 20 S proteasome exhibits several protease activities. Some of these appear to be associated with a single subunit: trypsin-like with ␤2, chymotrypsin-like with ␤5, and postglutamyl activity with ␤1 (12). The other two activities, a "branched chain amino acid-preferring" and a "small neutral amino acid-preferring," are not associated with a single subunit (13,14). Cleavage specificity is modulated by amino acid residues in the vicinity of cleavable peptide bonds (15)(16)(17)(18)(19).
Although proteasomes are the major proteases involved in generation of MHC class I-bound peptides, the precise processing of these ligands remains unclear. An important issue is whether the proteasome directly generates the MHC ligands or peptide precursors that are subjected to further trimming. Direct generation of class I ligands has been reported previously (20 -24), but there is also evidence for exopeptidase activity both in the cytosol and the ER (25)(26)(27)(28)(29). It has been suggested that proteasomes tend to generate the exact C-terminal ends of class I ligands, but are less precise at the N terminus, thus generating N-terminally extended precursors (30,31). Another issue is that proteasomes may cleave peptide bonds internal to the sequence of natural ligands, leading to their destruction (18,32,33). The balance between cleavage leading to generation or destruction of a given peptide may determine its presence or influence its amount in the class I-bound pool.
HLA-B27 has special interest for its strong association with ankylosing spondylitis (AS) and reactive arthritis (ReA) (34,35). Gram-negative bacteria, including species of Salmonella, Yersinia, Chlamydia, and Campylobacter, are known pathogenetic agents for ReA (36). The mechanism involving HLA-B27 and bacteria in this disease is unknown. A classical hypothesis invoked molecular mimicry between bacterial and self-peptides presented by HLA-B27 as a source of a B27-directed autoreactive cytolytic T lymphocytes response that would be a primary pathogenetic event (37). Alternative mechanisms remain open (38 -42). Although more than 90% of patients with AS and about 70% of those with ReA are B27-positive, most HLA-B27positive individuals remain healthy. Additional genetic factors modulate susceptibility to these diseases (43), but their identity remains unknown.
This study addressed the generation of HLA-B27 ligands derived from HLA-B27 or other class I heavy chains by the 20 S proteasome. Misfolded class I polypeptides are dislocated to the cytosol (44,45) and degraded by proteasomes (46), and HLA-B27 seems to misfold more than other HLA class I proteins (39). However, it is not known what peptides derived from class I molecules are HLA-B27 ligands, whether these peptides are directly produced by the proteasome, or whether proteasomal cleavage of class I molecules can be used to predict novel natural ligands of HLA-B27. To address these issues we focused on two regions of the ␣2 domain: around residues 169 -179 and around residues 150 -158. The first region has homology with protein sequences from Gram-negative bacteria (47). It was postulated that molecular mimicry between bacterial proteins and a peptide derived from this region of the HLA-B27 molecule and presented by HLA-B27 could elicit autoreactivity upon bacterial infection and play a role in the pathogenesis of ReA and other spondyloarthropathies. A natural ligand of HLA-B27, spanning residues 169 -179 of its own molecule, herein designated as B27-(169 -179), was subsequently found in B*2705 and other HLA-B27 subtypes (48,49). The second region is polymorphic among class I molecules and could provide information about polymorphic HLA-B27 ligands derived from other class I proteins. These peptides would be presented by some, but not all, HLA-B27-positive individuals and could be a source of antigenic diversity of HLA-B27 dependent on the non-B27 HLA class I type of the individual.

MATERIALS AND METHODS
Synthetic Peptides-Synthetic peptides were synthesized using standard Fmoc (N-(9-fluorenyl)methoxycarbonyl) chemistry and purified by HPLC. The correct molecular mass of purified peptides was established by matrix-assisted laser desorption/ionization time of flight (MALDI-TOF) mass spectrometry (MS), and their quantification was done by amino acid analysis after hydrolysis in 6 M HCl. In cysteinecontaining peptides this residue was incorporated as carboxymethyl-Cys during synthesis.
Cell Lines and Isolation of HLA-B27-bound Peptides-HMy2-C1R (C1R) is a class I-deficient human plasma cell line with low expression of its endogenous HLA-B*3503 and -Cw4 class I antigens (50). Transfectants expressing high levels of B*2705 or other HLA-B27 subtypes were used as the source of HLA-B27-bound peptides. These cells were cultured in DMEM with 7.5% heat-inactivated fetal calf serum (both from Life Technologies, Paisley, United Kingdom).
Isolation of HLA-B27-bound peptides was done as described previously (51), with minor modifications. Briefly, 1-1.5 ϫ 10 10 HLA-B27 transfectant cells were lysed at 4°C in 20 mM Tris/HCl buffer, 150 mM NaCl, and 1% Nonidet P-40 (pH 7.5) with a mixture of protease inhibitors. After centrifugation, cell lysates were subjected to affinity chromatography using the W6/32 monoclonal antibody (IgG2a, specific for a monomorphic HLA-A, HLA-B, and HLA-C determinant (52). HLA-B27bound peptides were eluted with 0.1% aqueous trifluoroacetic acid at room temperature, filtered through Centricon 3 (Amicon, Beverly, MA), and concentrated to 100 l for HPLC fractionation. This was conducted in a Waters Alliance system (Waters, Milford, MA) using a Vydac C18 (0.21 ϫ 25 cm) 5-m particle size column (Vydac, Hesperia, CA) at a flow rate of 100 l/min, as follows: isocratic conditions with buffer A (0.08% trifluoroacetic acid in water) for 15 min, followed by a linear gradient of 0 -44% buffer B (80% acetonitrile and 0.075% trifluoroacetic acid in water) for 90 min, and a linear gradient of 44 -100% buffer B for another 35 min. Peptide fractionation was simultaneously monitored at 210 and 280 nm. Fractions of 50 l were collected and stored at Ϫ20°C.
Purification of 20 S Proteasome and Digestion of Synthetic Substrates-The 20 S proteasome was purified from B*2705-C1R cell ly-sates by ion-exchange chromatography and centrifugation in glycerol gradient as described previously (22). These preparations consisted of a mixture of 20 S proteasome and immunoproteasome, as determined by two-dimensional gel electrophoresis and Western blot analysis (data not shown).
Peptide substrates were incubated at 37°C and 125 g/ml with purified 20 S proteasome at an enzyme:substrate ratio of 1:10 (w/w) in 20 mM Hepes buffer (pH 7.6). Digestion was stopped by adding 0.20 volume of 0.4% aqueous trifluoroacetic acid. Digestion mixtures were dried down to 100 l in a SpeedVac and fractionated by HPLC using exactly the same conditions as for HLA-B27-bound peptides.
Mass Spectrometry-The peptide composition of individual HPLC fractions was determined by MALDI-TOF MS as described previously (22). Dried fractions were resuspended in 5 l of methanol/water (1:1) containing 0.1% formic acid, and 0.5 l was used for analyses. When required, 1 l of these samples was subjected to peptide sequencing by quadrupole ion trap nanoelectrospray MS/MS, as described previously (53,54).
Alternatively, peptide sequencing was done by post-source decay MALDI-TOF MS. A 0.5-l aliquot of the sample was 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) were added and again allowed to dry at room temperature. The post-source decay MALDI spectrum was measured on a Bruker Reflex TM III MALDI-TOF mass spectrometer (Bruker-Franzen Analytic GmbH, Bremen, Germany) equipped with the SCOUT TM source in positive ion reflector mode using delayed extraction. The spectrum was recorded in 14 segments, each successive segment representing a 20% reduction in reflector voltage. The precursor ion was selected by FAST TM deflecting pulses. About 200 shots were averaged per segment, and the segments were pasted, calibrated, and smoothed with Bruker XTOF 5.0.3 software. Data analysis was performed using Bruker BioTools 2.0 software.

Generation of B27-(169 -179) and a C-terminally Extended
13-mer by the 20 S Proteasome-We first addressed the generation of B27-(169 -179), RRYLENGKETL, by the 20 S proteasome from a synthetic 30-mer with the sequence of B*2705 residues 158 -187, designated as B27-(158 -187). The digestion mixture was fractionated by HPLC (Fig. 1A), and fractions corresponding to absorbance peaks were analyzed by MALDI-TOF and, sometimes, also by quadrupole ion trap nanoelectrospray MS. The yield of individual digestion products was estimated on the basis of their absorbance at 210 nm, normalized to take into account peptidic length differences. When various peptides co-eluted, the percentage of each peptide in the absorbance peak was estimated on the basis of their respective ion peak signal intensities in the MALDI-TOF spectra. This is only an approximation, because ion peak intensity may not strictly correlate with peptide abundance.
About 50% of the B27-(158 -187) substrate was digested after 24 h. Of 21 digestion products obtained with Ͼ0.1% yield, 13 resulted from cleavage at a single peptide bond, and 8 were internal fragments resulting from dual cleavage (Fig. 1B). Internal fragments accounted for only 5% of the total digestion products. Thus, the relationship between proteasomal cleavage and HLA-B27 ligands must be established not only from the internal fragments observed, but mainly from the analysis of cleaved bonds in the synthetic substrate (Table I). Cleavage was observed immediately after all four Leu residues: Leu-160, Leu-168, Leu-172, and Leu-179. Cleavages after Leu-168 and Leu-179 are those involved in the generation of the natural B27-(169 -179) ligand, although this peptide was not found as a digestion product of B27-(158 -187). Cleavage was also observed immediately after all three Arg residues: Arg-169, Arg-170, and Arg-181. The latter peptide bond was the major cleavage point of the B27-(158 -187) substrate (Table I). Dual cleavage after Leu-168 and Arg-181 generated an internal 13mer, B27-(169 -181), with a yield of 0.7% of the total digest (Fig. 1B). This peptide has anchor motifs typical of HLA-B27 ligands (Arg-2, Tyr-3, Arg-13), suggesting that it could exist as a natural ligand of HLA-B27. Cleavage also occurred after Tyr-171, Asn-174, Thr-178, Gln-180, and Asp-183. Overall, cleavage of B27-(158 -187) occurred mainly at two clusters of peptide bonds: Leu-168-Glu-173 and Thr-178-Pro-184.
These results indicate that the 20 S proteasome cleaves B27-(158 -187) at the precise peptide bonds required for direct generation of the natural B27-(169 -179) ligand. In addition, they suggest the existence of a hitherto unknown B27 ligand from the same region: B27-(169 -181). Significant cleavage at internal peptide bonds within these sequences was also observed.
Influence of the Positioning of the B27-(169 -179) Sequence in the Substrate on Proteasomal Cleavage-Synthetic substrates commonly used to analyze proteasomal cleavage and generation of peptide epitopes are frequently designed so that the sequence of interest is located approximately in the middle, as in B27-(158 -187). Since neighboring residues can influence proteasome specificity, the register in which the sequence of interest is placed within the precursor might affect cleavage patterns and lead to unreliable assessment of proteasomal peptide products. To address this issue we designed two additional substrates, B27-(165-194) and B27-(154 -183), in which the B27-(169 -179) sequence was placed near the N-terminal or the C-terminal end, respectively. Both substrates were di-gested by the 20 S proteasome and the digestion products analyzed as in the previous paragraph.
B27-(165-194) was digested almost completely (99%) within 24 h ( Fig. 2A). A total of 44 digestion products, including 23 single-cleavage and 21 internal fragments, were obtained with Ͼ0.1% yield (Fig. 2B). Internal fragments accounted for 17% of the total digestion products. Three main observations were made. First, essentially the same peptide bonds as in B27-(158 -187) were cleaved in B27-  in the region in which both substrates overlap (Table I) B27-(154 -183), in which the B27-(169 -179) sequence was displaced toward the C-terminal end, was digested with 75% yield in 24 h (Fig. 3). A major cleavage point, after Ala-158, dominated the digestion. The total yield of peptide fragments resulting from cleavage at this bond was about 74% of the digest. Other neighbor peptide bonds were also cleaved: after Tyr-159, Leu-160, and Gly-162 (Table I). Overall, 22 peptides, including 15 single-cleavage and 7 internal fragments, were produced with Ͼ0.1% yield upon digestion of B27-(154 -183). Cleavage around the N-terminal end of the B27-(169 -179) sequence was similar, as in the other two substrates ( Fig. 3B and Table I), including cleavage after Leu-168, at the precise N terminus of B27-(169 -179), and within this sequence: after Arg-169, Arg-170, Tyr-171, Leu-172. In contrast, substantial differences were observed around the C-terminal region of B27-(169 -179): cleavage did not occur after Leu-179 or Gln-180, bonds that were significantly cleaved in the other two precursors, and was observed with much lower yield after Arg-181 (Fig. 3B, Table I). These alterations are presumably explained by the proximity of these bonds to the substrate C terminus, which might impair proteasomal cleavage in its vicinity.
In conclusion, proteasomal cleavage in and around the B27-(169 -179) sequence was little influenced by its location in the substrate, except near the substrate C terminus. Cleavage at the precise N-and C-terminal ends indicated that B27-(169 -179) can be directly produced by the proteasome, as observed with one of the substrates. Cleavage after Arg-181 generated a 13-mer, B27-(169 -181), with structural features typical of HLA-B27 ligands.
B27-(169 -181) Is a Natural Ligand of B*2705-The B*2705bound peptide pool was isolated from B*2705-C1R transfectant cells and fractionated by HPLC (Fig. 4A). Fractions collected around the retention time of B27-(169 -181) were analyzed by MALDI-TOF MS. An ion peak at mass/charge (m/z) 1662.4, compatible with the molecular mass of the 13-mer, was found in HPLC fraction number 132 (Fig. 4B). The sequence of the corresponding peptide was determined by post-decay MALDI-TOF MS (Fig. 4C) and shown to be B27-(169 -181). Thus, this peptide is a natural B*2705 ligand. In the same experiment, B27-(169 -179) eluted as the main peptide component of a major absorbance peak (Fig. 4A). On the basis of the absorbance and peptide composition of the corresponding HPLC    transfectants were fractionated by HPLC exactly as for B*2705. HPLC fractions around the retention time of B27-169-181) were analyzed by MALDI-TOF MS. Alignment of correlative fractions from different subtypes was confirmed by the presence of co-eluting peptides common to different subtypes (Fig. 5). An ion peak at m/z 1662.4, corresponding to B27-169-181), was found in fraction number 132 from B*2705. In this particular chromatography B27-(169 -181) co-eluted with many other peptides. The purity of this ligand in its corresponding HPLC fraction, as assessed by MALDI-TOF MS, was variable among different HPLC runs, ranging from being the predominant peptide (Fig. 4B) to eluting in a rather complex peptide mixture (Fig. 5A). The 13-mer was detected only as a very weak signal in the corresponding fraction from B*2709 (Fig. 5A) and was not seen in adjacent HPLC fractions from this subtype (data not shown). Whereas B27-(169 -179) was found in B*2709 similarly as in B*2705, that is 4% of the B*2709-bound peptide pool, B27-(169 -181) accounted only for 10 Ϫ4 %. Thus, the 11-mer/13-mer ratio in B*2709 was about 40,000:1 (Table II).
An ion peak corresponding by molecular mass (m/z: 1662.2) and retention time to B27-(169 -181) was also found in B*2704, but not in the corresponding fraction of B*2706 (Fig. 5B) or adjacent ones (data not shown). B27-(169 -181) was less abundant in B*2704 (0.03%) than in B*2705, but still in the range of many natural class I-bound ligands (62). This lower abundance relative to B*2705 correlates with the lower suitability of the C-terminal Arg residue of this peptide for B*2704 (63,64). Although in B*2705/B*2709 and B*2704/B*2706 there is a correlation between binding of B27-(169 -181) in vivo and subtype association to AS, this correlation was not complete, since the 13-mer was not found in B*2702. B27-(169 -179) was abundant in all five subtypes (Table II).
B27-(150 -158) Is a Natural B*2704 Ligand Arising from a Non-B27 Class I Molecule in the Same Cell-A search for the B27-(150 -157/158/159/160) peptides in the B*2705-bound pool was carried out by MALDI-TOF MS of HPLC fractions around the retention times of these peptides. This screening failed to reveal any major ion peak corresponding to the molecular mass of any of them (data not shown). Using the same approach, B27-(150 -158) was found in the B*2704-bound peptide pool from B*2704-C1R transfectant cells. An ion peak at m/z 1013.8, compatible with the molecular mass of this peptide, was found in HPLC fraction number 132 from this allotype (Fig. 7A). The peptide was identified as B27-(150 -158) by quadrupole ion trap electrospray MS/MS (Fig. 7B). However, the sequence of this peptide, ARVAEQLRA, which accounted for 0.02% of the B*2704-bound pool, could not come from B*2704, since this subtype has Glu instead Val at position 152. C1R cells synthesize a mutant form of B*3503 with reduced translation and cell surface expression (50). B*3503, but not Cw4 also present in these cells, is identical to B*2705 at residues 150 -158, and in the whole 139 -163 region, corresponding to the substrate used for in vitro digestion, except at position 163. Therefore it is very likely that B*3503 is the source of the ARVAEQLRA ligand of B*2704. This peptide is an example of a natural HLA-B27 ligand whose expression is dependent on the non-B27 class I molecules present in the cell, and therefore on the HLA type of each individual. DISCUSSION This study addressed the proteasomal processing of HLA-B27 and other class I molecules, leading to generation of poly-  This was estimated on the basis of the absorbance and peptide composition of the corresponding HPLC peak, relative to the total absorbance of the peptide pool, as described for proteasomal digestion products (see text).
c Data obtained from Edman degradation (49). morphic HLA-B27 ligands. First, we analyzed whether the 20 S proteasome cleaved synthetic substrates at the precise N-and C-terminal ends of the natural B27-(169 -179) ligand or rather cleavage at the N-terminal end was less precise, favoring generation of N-terminally extended precursors. Second, we assessed internal cleavage within the sequence of the natural ligand and analyzed the influence of substrate structure on cleavage patterns. Third, we applied this analysis to identification of novel HLA-B27 ligands derived from its own or other class I molecules. This approach was recently applied to identification of tumor-associated antigens (65).
Three overlapping synthetic precursors with the sequence of B*2705 in which the B27-(169 -179) sequence was placed in N-terminal, central, and C-terminal registers were used to determine cleavage of this region by the 20 S proteasome. Several observations arose from these experiments. First, the 20 S proteasome cleaved efficiently at the precise N-and Cterminal residues of the B27-(169 -179) ligand, as also observed for two other HLA-B27 ligands derived from non-HLA proteins (22), and with other MHC class I-bound peptides (8,20,24). Second, significant cleavage within the B27-(169 -179) sequence may limit the amount of this peptide available for binding to HLA-B27. The cleavage efficiency at peptide bonds leading to generation of this ligand, relative to cleavage of internal bonds might be different in vivo or subjected to differential regulation by PA28. However, the fact that B27-(169 -179) was a major component of the B27-bound peptide pool suggests that cleavage of internal bonds does not predominate in vivo over cleavage leading to direct generation of the ligand. Efficient transport into the ER, or its high affinity for HLA-B27 (49), may also influence the abundance of this peptide in the B27-bound pool. Third, proteasomal cleavage after Arg-181 allowed the prediction and identification of the B27-(169 -181) 13-mer as a novel HLA-B*2705 ligand containing a sequence homologous to proteins from Gram-negative bacteria (47). Thus, proteasomal cleavage in vitro reproduces, at least in some cases, in vivo processing to the point that it can be used to identify unknown MHC class I ligands.
Significantly, cleavage after Lys-176 was not observed. This would have generated a nonamer, B27-(168 -176), or an octamer, B27-(169 -176), with B*2705 anchor motifs. The former peptide was initially proposed as the putative peptide mediating molecular mimicry with bacterial peptides in the context of HLA-B27 (47), but it has not been found as a natural B27 ligand. The cleavage pattern observed in this region explains this absence and suggests that the presence of these two peptides in the B27-bound pool is unlikely.
Polymorphism at flanking positions may substantially alter proteasomal cleavage in and around a given peptide sequence (15)(16)(17)(18)(19). An important aspect of in vitro digestions that has seldom been specifically addressed is to what extent the location of a particular sequence within the precursor substrate may influence cleavage patterns. Our results demonstrated that proteasomal specificity around the sequence of a given ligand was little influenced by its precise location within the substrate. The main difference was observed in peptide bonds close to the C-terminal end of some substrates. For instance, impaired cleavage at residues 179 -181 in B27-(154 -183) relative to B27-(158 -187), and at residues 158 -160 in B27-(139 -163) relative to B27-(154 -183), is presumably due to the proximity of the negatively charged C terminus (9).
The pathogenetic significance of presentation by HLA-B27 of two peptides derived from its own molecule containing a sequence with homology to proteins from Gram-negative bacteria is unclear. That B27-(169 -179) is a prominent natural ligand of HLA-B27 subtypes associated and not associated to AS suggests that it is not relevant to this disease. However, the possibility that it may be arthritogenic only in the context of some subtypes cannot be ruled out. In addition, autoreactive CD8 ϩ cytolytic T lymphocytes with specificity for B27-(168 -176) have been detected in AS patients. 2 Although, as mentioned above, there is no evidence for this peptide being a natural B27 ligand, nor is it produced by the 20 S proteasome in vitro, the finding raises the possibility that natural HLA-B27 ligands containing this sequence may play a role in disease. If so, the finding of B27-(169 -181) might have some significance. This peptide has a C-terminal Arg residue, which is disfavored for binding to B*2706 and B*2709, subtypes not associated with AS. Indeed, this 13-mer was not found in B*2706 and was in very low amount in B*2709, whereas it was prominent in the disease-associated B*2705 and B*2704. However, correlation of this peptide with association to AS, although better than for B27-(169 -179), was not complete, since it was apparently absent from the disease-associated B*2702 subtype.
The region spanning residues 150 -158 includes three positions that are polymorphic among class I molecules: 152, 156, and 158 (66). Of these, position 152 is either Val or Glu, both among class I proteins and HLA-B27 subtypes (67). That B27-(150 -158) was directly generated from a synthetic precursor and found as a B*2704 ligand in C1R transfectant cells again shows that proteasome cleavage in vitro is suitable for predicting novel class I MHC ligands. As noted above, this peptide probably arose from B*3503, expressed at reduced levels in C1R cells (50). Thus, it is a polymorphic HLA-B27 ligand whose expression is genetically determined and dependent on the concomitant presence of certain non-B27 HLA class I molecules. Peptides such as this one introduce diversity among B27-bound peptide repertoires from different individuals.
Proteasomal degradation of misfolded class I polypeptides after dislocation from the ER to the cytosol is probably a physiological quality control process (68), but becomes especially relevant in situations that favor class I misfolding (46). These might be, for instance, intracellular bacterial infections or stimulation of class I protein synthesis during inflammation. If polymorphic peptides derived from other class I molecules and presented by HLA-B27 would play a pathogenetic role in HLA-B27-associated disease, this could contribute to explain that only a fraction of the B27-positive individuals develop spondyloarthropathies. If potentially pathogenetic peptide sequences were encoded by few class I HLA alleles, these would probably show up as additional genetic markers for these diseases. However, if such peptide sequences were encoded by multiple non-B27 class I alleles, their association with spondyloarthropathy would be much more difficult to detect by conventional genetic analysis. Indeed, besides a significant contribution of non-MHC genes (Ͼ50%), an additional contribution of non-B27 genes within the MHC to AS is supported by genetic studies (43).
In conclusion, three natural HLA-B27 ligands, derived from the B27 molecule itself or other class I proteins, were directly generated by the 20 S proteasome in vitro. Although protein processing in vivo may be different to some extent due to substrate differences and to involvement of PA28, cleavage of synthetic precursors in vitro explains the presence of the natural ligands analyzed in this study. The correspondence be-tween in vitro cleavage patterns and generation of natural ligands was illustrated by prediction and finding of novel class I-derived HLA-B27 ligands.