Presentation of Cytosolically Stable Peptides by HLA-B27 Is Not Dependent on the Canonic Interactions of N-Terminal Basic Residues in the A Pocket¹

HLA-B27 is a major susceptibility factor for ankylosing spondylitis and other spondyloarthropathies (1, 2). Although the mechanism of this association is not known, the peptide-binding properties of the molecule are thought to be critical, either through presentation of foreign peptide Ags showing molecular mimicry with self-ligands (3), or through a more general influence of peptide ligands in determining folding and stability of the native molecule (4). Additionally, HLA-B27 is a very efficient Ag-presenting molecule (5), determining good protective responses against a variety of infections, including HIV (6, 7, 8, 9) and hepatitis C (10). These observations justify the interest of characterizing the peptide-binding specificity of HLA-B27 and its molecular basis.

A major feature of HLA-B27 ligands is their almost absolute restriction for R at position (P)³ 2 (11, 12), which is determined by the structure of the B pocket, of which a critical feature is the E45 residue (13, 14). HLA-B27 ligands show higher heterogeneity at the N-terminal (P1) position, although basic (R, K) and small residues (G, A, S) are predominant; for example, in a published registry of 108 constitutive B*2705 nonamer ligands (12), the frequency of basic and small residues was 27.8% (R1, 20.4%; K1, 7.4%) and 41.6% (G1, 23.1%; A1, 10.2%; S1, 8.3%), respectively. P1 residues bind in the A pocket. Their interactions include a conserved network of hydrogen bonds established with the peptidic N terminus within the pocket itself (13), as well as contacts of residues at the surface of the pocket with the P1 side chains. The influence of P1 residues on binding to HLA-B*2705 was assessed in one study with poly-A analogs containing the R2 motif and changes at P1 (15). In this study R1 was the residue that made the highest contribution to binding.

X-ray analyses have determined the binding mode of the peptidic R1 to HLA-B27 and provided an explanation for its contribution to peptide affinity (16). Three residues at the surface of the A pocket interact with the R1 side chain of HLA-B27 ligands: R62, E163, and W167. E163 forms a salt bridge with the guanidinium group of the peptidic R1. Additionally, this residue interacts with R62 through π-π stacking of their guanidinium groups and with the indole ring of W167 through van der Waals interactions of the aliphatic R1 side chain. Thus, R62 and W167 form a tightly packed sandwich with the peptidic R1, allowing for its stable anchoring to HLA-B27.

Although structural considerations might explain the prevalence of R1 among HLA-B27 ligands, an alternative, not mutually exclusive, explanation was recently proposed on the basis that peptides with N-terminal dibasic sequences (i.e., RR, KR) were particularly resistant to amino-peptidase-mediated cytosolic degradation (17). Thus, such peptides, even if generated in relatively low amounts, would advantageously compete for survival in the cytosol and reach the MHC class I loading pathway. HLA allotypes, such as HLA-B27, with specificity for peptides with R2 would then preferentially bind peptides with N-terminal basic residues due to their increased availability, relative to other ligands.

The present study was undertaken to assess the role of the A pocket and, in particular, of the structural determinants of the peptidic R1 binding, in the preference of HLA-B27 for peptides with dibasic N-terminal sequences. It was reasoned that if the preference of HLA-B27 for R1 was essentially determined by A pocket interactions, presentation of peptides with N-terminal basic residues would be severely compromised by mutating the contact residues at the surface of the pocket. Conversely, if the contribution of P1 anchoring to the overall peptide affinity were not critical, even severe disruption of the interactions that stabilize the peptidic R1 side chain would not affect the preferential binding of peptides with basic P1 residues to HLA-B27, due to their relatively high availability. Thus, we generated two HLA-B*2705 mutants lacking only E163 (E163T, herein designated as T163) or both this residue and W167 (E163T-W167S, herein designated as T163S167) and compared their peptide repertoires, the P1 residue frequencies among their constitutive peptide ligands, and the molecular stability of these mutants, relative to B*2705.

HMy2.C1R (C1R) is an HLA-A-negative human lymphoid cell line with low expression of its endogenous HLA-B35 and -Cw4 molecules (18). C1R transfectants expressing B*2705 and the T163 mutant were obtained using genomic DNA and were previously described (19, 20). The T163S167 mutant was obtained from the full-length cDNA of T163 cloned into pcDNA3 (Invitrogen) by PCR-mediated site-directed mutagenesis using a QuickChange site-directed mutagenesis kit (Stratagene) with the following primers: 5′-CACGTGCGTGGAGTCGCTCCGCAGATACC3-′ and 3′-GGTATCTGCGGAGCGACTCCACGCACGTG5-′. After confirming the correct sequence of the double mutant, it was transfected into 10⁷ C1R cells by electroporation of 10 μg DNA at 260 V and 960 μF. The transfectants were selected with increasing concentrations of G418 (Invitrogen), ranging from 0.5 to 1 mg/ml. All the cell lines were cultured in RPMI 1640 medium supplemented with 2 mM l-Gln and 10% FBS (Invitrogen). The mAbs used were W6/32 (IgG2a, specific for a monomorphic HLA class I determinant of the native heterodimer) (21), ME1 (IgG1, specific for HLA-B27, B7, B22) (22), HC10 (IgG2a), which recognizes free class I H chain (HC) (23), and the anti-γ-tubulin mAb (IgG1) GTU88 (Sigma-Aldrich). Flow cytometry was performed as previously described (24).

Cells (∼10⁶) were lysed at 4°C for 20 min in 0.5% Igepal CA-630 (Sigma-Aldrich), 50 mM Tris-HCl (pH 7.4), and 5 mM MgCl₂ containing a cocktail of protease inhibitors (Complete Mini, Roche). After centrifugation of whole lysates and SDS-PAGE, HLA class I HC and γ-tubulin were revealed with HC10 or the anti-γ-tubulin mAb, respectively, using human-absorbed HRP-conjugated goat anti-mouse Ig (SouthernBiotech) as secondary Ab.

HLA-B27-bound peptides were isolated from ∼10¹⁰ C1R transfectant cells as previously described (25). Briefly, cells were lysed in 1% Igepal CA-630 (Sigma-Aldrich), 20 mM Tris-HCl buffer, and 150 mM NaCl (pH 7.5) in the presence of a cocktail of protease inhibitors. After ultracentrifugation, cell lysates were subjected to affinity chromatography using the W6/32 mAb. HLA-B27-bound peptide pools were eluted with 0.1% aqueous trifluoroacetic acid (TFA) at room temperature, filtered through Centricon 3 (Millipore), concentrated, and subjected to reverse-phase HPLC fractionation. This was conducted in a Waters Alliance system using a Vydac 218TP52 column at a flow rate of 100 μl/min, as previously described (26). Fractions of 50 μl were collected.

The peptide composition of HPLC fractions was analyzed by MALDI-TOF MS using an Autoflex MALDI-TOF mass spectrometer (Bruker Daltonik) in positive ion reflector mode. Dried fractions were resuspended in 0.5 μl of 33% aqueous acetonitrile and 0.1% TFA. They were deposited onto an AnchorChip 600/384 target plate (Bruker Daltonik) and allowed to dry at room temperature. Then, 0.5 μl of matrix solution (α-cyano-4-hydroxycinnamic acid in 33% aqueous acetonitrile and 0.1% TFA) at 2 mg/ml were added and allowed to dry again at room temperature. MS spectra were processed using the MoverZ software (version 2001.02.03: www.bioinformatics.genomicsolutions.com/moverZDL.html).

In some instances individual HPLC fractions were reanalyzed using a MALDI-TOF/TOF instrument (4800 Proteomics Analyzer, Applied Biosystems). In these cases, the dried HPLC fractions were reconstituted with 0.6 μl of 33% aqueous acetonitrile and 0.1% TFA, loaded onto an Opti-TOF 384-well MALDI insert (Applied Biosystems) and allowed to dry at room temperature. Then, 0.6 μl of the matrix solution (α-cyano-4-hydroxycinnamic acid in 33% aqueous acetonitrile and 0.1% TFA) was added at 3 mg/ml. These mass spectra were acquired in reflector positive mode and processed using the 4000 Series Explorer software version 3.5.

Peptide sequencing was conducted by quadrupole ion trap nanoelectrospray MS/MS in a Thermo Finnigan LCQ DECA-XP instrument, as previously described (27). Alternatively, a Surveyor HPLC system coupled online to an LTQ LIT instrument (Thermo Finnigan) was used. Peptides were concentrated in a reverse-phase precolumn (0.32 × 30 mm, BioBasic 18, Thermo Electron) and eluted on an analytical column (0.18 × 150 mm, BioBasic 18, Thermo Fisher Scientific) using a 72- to 86-min gradient from 5% to 40% solvent B (solvent A: 0.1% formic acid; solvent B: 80% acetonitrile, 0.1% formic acid) at a flow rate of 2 μl/min. Normalized collision energy was set to 35% and a 3-Da mass window was used to fragment selected parental ions.

Some peptide sequences were obtained using an esquire3000^plus ion trap mass spectrometer (Brucker) after online chromatographic separation of samples as follows. Samples were dissolved in buffer A (0.5% acetic acid in water) and loaded onto a column (100 mm × 100 μm internal diameter) packed with 5-μm Kromasil C18 beads (EKA Chemicals) and fractionated in a Famos-Switchos-Ultimate chromatographic system (LCPackings) with a 45-min linear gradient of 5–30% buffer B (90% acetonitrile, 0.5% acetic acid in water) at 500 nl/min. MS/MS spectra were acquired by automatic switching between MS and MS/MS mode using dynamic exclusion. Some samples were analyzed using the multiple reaction monitoring mode, which allows specific peptide masses to be selected.

Interpretation of the mass spectra was done manually but assisted by various software tools as follows. Manual inspection of the spectrum usually allowed us to determine a partial sequence. This information, together with the mass-to-charge (m/z) ratio of the parent ion (in all cases, charge 2 parent ions were selected for sequencing in this study), was used as input data for a Mascot (version 2.2) search (www.matrixscience.com) in the human protein entries of the MSDB database (release 2006/31/08, Imperial College, London), using a window of 0.8 m/z units. Of the 20 output sequences showing the highest scores in this preliminary search, those few showing the canonical R2 motif of HLA-B27 ligands and absence of “prohibited” residues for HLA-B27 binding, such as N-terminal or C-terminal Pro, were selected. From each of these sequences, a list of theoretical fragment ions was generated using the MS product tool of the ProteinProspector package (version 4.27.2 basic: prospector.ucsf.edu/) as an assistance to match the putative candidate sequences to our experimental MS/MS spectrum. When one single proper match was obtained, the peptide sequence was assigned.

To facilitate the analysis of residue frequencies, a software tool was developed, designated as MSearcher, which recognizes peptide series written as unformatted text databases. These can be directly generated in a text editor with data transferred from conventional word processors, such as Microsoft Office applications. The software calculates several parameters from single or multiple (up to four) peptide positions. These parameters include the residue, residue type, or residue size frequencies, and the deviation from the mean values of these parameters in the proteome. The software was implemented in a user-friendly graphical interface and is freely available upon request. For statistical analysis the Fisher’s exact test was used.

The thermostability assay was performed as detailed elsewhere (28). Briefly, C1R transfectants were pulse-labeled for 15 min and chased at 0, 2, and 4 h. The lysates were incubated for 1 h at various temperatures, ranging from 4°C to 50°C, immunoprecipitated with W6/32, and analyzed by SDS-PAGE. The amount of HLA-B27 heterodimer precipitated at each temperature at any given chase time was expressed as a percentage of the amount precipitated at 4°C, and it was plotted, for each time point, as a function of the temperature.

The expression of HLA-B*2705, T163, and T163S167 on the surface of C1R transfectant cells was analyzed by flow cytometry with the ME1 and HC10 mAb, which recognize the folded heterodimers and the unfolded HC, respectively (Fig. 1, A and B). The surface expression of B*2705 and T163 heterodimers was very similar and ∼3-fold higher than for the double mutant. Similar results were obtained with W6/32 (data not shown), indicating that the lower expression of the T163S167 mutant was not due to modification of the ME1 epitope. Indeed, the HC protein expression of the T163S167 mutant, as estimated by Western blot, was significantly lower than for the other two variants (Fig. 1,C). The fluorescence ratio obtained with HC10 and ME1 (HC10/ME1 ratio) was also similar for B*2705 and T163, but was ∼2-fold higher for T163S167 (Fig. 1 B). This might be consistent with higher cell surface dissociation of the double mutant, compared with B*2705 or T163.

The B*2705-bound peptide pool was compared with those of T163 and T163S167 on the basis of identity of molecular mass and chromatographic retention time of the corresponding peptide species (Table I). The strategy used was the same as in previous studies from our laboratory (29, 30, 31, 32). The peptide pools to be compared were isolated from the corresponding C1R transfectant cells after immunoaffinity purification with W6/32 and acid extraction. They were fractionated by HPLC in consecutive runs under identical conditions, and each of the peptide-containing fractions was analyzed by MALDI-TOF MS. The MS spectrum of each HPLC fraction from one variant was compared with the correlative previous and following fractions from the other variant to take into account small shifts in the retention time of peptides that may occur even in consecutive chromatographic runs. Ion peaks with the same (±0.8) m/z in this comparison were considered to reflect identical ligands of both variants. This assumption was validated for significantly overlapping peptide pools in previous analyses of HLA-B27 subtypes (29, 30, 31) and, in this study, by sequencing the relevant peptides from multiple variants. Ion peaks found in only one peptide pool were considered to reflect variant-specific ligands.

With this criterion, T163 shared with B*2705 ∼85% of its peptide repertoire. It is likely that some of the differences may actually consist of peptides that are found in lower amounts, and below the detection limit of the MALDI-TOF instrument, in one of the variants (see below). Thus, the observed peptide sharing is probably a minimum estimation.

In the comparison with T163S167, twice as many ion peaks were detected for B*2705 than for the double mutant. This is probably due to the lower expression level of the corresponding transfectant. Thus, only the peptide sharing of the mutant relative to the wild type was taken into account, since differential ion peaks in this comparison cannot be explained by lower protein expression. T163S167 shared with B*2705 44.5% of its peptide repertoire. Again, the peptide sharing observed is probably a minimum estimation determined by the sensitivity of the MS instrument used. For instance, of 30 ion peaks found in B*2705, but not in the double mutant, in this comparison using a conventional MALDI-TOF instrument, 17 (56.7%) were detected using the much more sensitive (∼20- to 100-fold) MALDI-TOF/TOF analyzer. If the differences found in the comparison of the peptide repertoires in Table I were corrected by this percentage, the peptide overlap of T163 and T163S167 with B*2705 would be 93.5% and 76%, respectively.

These results indicate that the single T163 mutation has a limited effect on peptide specificity, allowing for binding of a large majority of B*2705 ligands to this mutant. The double mutation has a larger effect, but it is still compatible with binding of a substantial fraction of the B*2705-bound peptide repertoire.

To determine the effect of the T163 mutation on residue usage at P1, the presence of two peptide sets in the T163-bound peptide pool was analyzed. Set 1 (Table II) consisted of 54 B*2705 ligands of known sequence that give significant signals in the MALDI-TOF MS spectrum (33) and included some prominent components of the B*2705-bound peptide pool. Set 2 (Table III) consisted of 41 ion peaks of diverse intensity, or even not detected by MALDI-TOF MS, which were sequenced in this study from B*2705, the mutants, or both. Some of these peptides were already known B*2705 ligands (12, 31). The predominant P1 residues among set 1 peptides were basic (R, 26.4%; K, 9.4%) and small (G, 15.1%; A, 11.3%; S, 11.3%). These frequencies reflected approximately those of a wider series of B*2705 ligands (12). Among the peptides in set 2, the most notorious difference was the almost 5-fold lower frequency of R1 among the B*2705 ligands, relative to those in set 1 (Fig. 2 A).

Of a total of 90 B*2705 ligands in both peptide sets, including all 54 from set 1 and 36 from set 2, 26.7% had R or K and 43.4% had small residues (G, A, S) at P1 (Fig. 2,B). Only four (4.4%) of these B*2705 ligands, all from set 2, were not found in T163, and none of them had R1 or K1 (Table III). Thus, the E163 residue in B*2705 is dispensable for binding peptides with basic P1 residues.

To examine the possibility that the T163 mutation could have a quantitative effect on the peptides with basic P1 residues bound to the mutant, the amounts of 15 shared ligands with R1 recovered from B*2705 and T163 were estimated from the corresponding ion peak intensities in the MALDI-TOF spectra. This is only an approximate estimation since MALDI-TOF MS is not a quantitative technique and peak intensities may be influenced by multiple factors, including the composition of the sample. With this limitation in mind, a pattern of relative peptide yields consistent with a variant-related bias was not observed (Table IV). The intensity of individual ion peaks from T163, relative to those from B*2705, ranged from ∼25% to 259%, suggesting that peptides with R1 bound the mutant with similar efficiency as did the wild type, with some peptide- to-peptide differences. The only exception was RRYLENGKETLQR, whose yield from the mutant was <1% relative to the wild type. This peptide is derived from residues 169–181 of the HLA-B27 HC. The possibility that the T163 mutation may influence the generation of this ligand seems unlikely, since the related peptide RRYLENGKETL was recovered from the mutant and the wild type with similar yields (Table IV).

The presence of the set 1 and set 2 peptides in the T163S167-bound peptide pool was analyzed by MALDI-TOF MS and their sequences were determined by nano-ESI MS/MS as above. The lower expression of the T163S167 mutant resulted in lower peptide yields. To account for the possibility that the lack of detection of B*2705 ligands in the double mutant was due simply to low recovery, those ion peaks found in B*2705 but not in the double mutant upon MALDI-TOF analysis were reanalyzed using a significantly more sensitive MALDI-TOF/TOF instrument. Only when a given ion peak from the double mutant was not detected after this second analysis was it assigned as a differentially bound peptide.

Of the 54 B*2705 ligands in set 1, 45 (83%) were found in the double mutant (Table II). These shared ligands included all 5 peptides with K1 and 11 of the 15 peptides (73.3%) with R1. These results indicate that, despite the severe disruption of the A pocket environment in T163S167, this mutant still retains a significant capacity to bind peptides with basic P1 residues.

Of the 41 peptides in set 2 (Table III), 25 (61%) were found in both peptide pools, 11 (26.8%) were found in B*2705 but not in the double mutant, and 5 (12.2%) were found in the double mutant but not in B*2705. All four peptides with R1 or K1 were found among the shared ligands. Small P1 residues (G, A, S) showed a lower joint frequency among shared, relative to B*2705-specific ligands in this set (40% and 81.9%, respectively). This was due mainly to the higher frequency of G1 and S1 among B*2705-specific ligands compared with the shared subset (36.4% vs 8% and 27.3% vs 8%, respectively). These differences did not reach statistical significance, although the increased frequency of G1 among B*2705-specific ligands approached this limit (Fig. 3 A).

Of the five peptides sequenced from the double mutant but not found in B*2705, four showed acidic or N residues at P1, which are absent or found with low frequency among B*2705 ligands (12). The fifth peptide, GRVNLNVLR, was not detected by MALDI-TOF MS in any of the variants. Thus, although it was not found in the wild type by nanoelectrospray MS/MS, it is difficult to rule out its presence in the B*2705-bound pool.

Overall, of 90 B*2705 ligands in both sets 1 and 2, 70 (77.8%) were also found in the double mutant and 20 (22.2%) were not. The frequency of peptides with basic P1 residues was similar or higher among the subset of shared ligands (R1, 18.6%; K1, 10%) than among the B2705-specific ones (R1, 20%; K1, 0%), confirming that binding of peptides with these motifs is largely maintained in the absence of both E163 and W167. As noted in the analysis of set 2, the overall frequency of small P1 residues (G, A, S) among shared ligands was smaller than among differentially bound ones (38.5% and 60%, respectively). The most notorious difference was the increased frequency of G1 among B*2705-specific ligands, relative to shared ones (30% vs 11.4%), which reached statistical significance (Fig. 3 B). This suggests that the absence of W167 may disfavor binding of peptides with G1.

The effect of the T163 mutation on the stability of HLA-B27 was analyzed by thermostability analyses in pulse-chase experiments. Lysates from pulse-labeled cells were incubated at various temperatures and immunoprecipitated with the conformation-sensitive mAb W6/32 at different chase times. The thermostability of the molecule was measured as the percentage of HLA-B27 HC immunoprecipitated at a given temperature, relative to the amount recovered at 4°C (Fig. 4). The double mutant could not be analyzed in these experiments because its low expression level precluded a clear distinction of the T163S167 HC from that of the endogenous class I molecules of C1R cells. Both B*2705 and T163 showed a significant stability in these experiments, with ∼70% recovery of W6/32-precipitated material at 50°C at 0 h chase time. B*2705 increased its stability with time, reaching 100% thermostability at 50°C at 4 h chase time. The T163 mutant failed to do so, suggesting that this mutation decreased the overall stability of HLA-B27/peptide complexes. The time-dependent increase in the stability of B*2705 occurred equally in the presence of brefeldin A (data not shown), indicating that it takes place in the endoplasmic reticulum.

The specificity of Ag presentation by HLA class I molecules is determined essentially by the structural complementarity of peptide ligands with the Ag-binding groove and, in particular, with the side chain-binding pockets (34). For HLA-B27 ligands, the primary anchor positions are P2, which is restricted almost exclusively to R, and the C-terminal position, which for B*2705 is predominantly restricted to basic, aliphatic, and aromatic residues. Other positions, particularly P1, P3, and P7, make a significant contribution to peptide binding, but they show a larger allowance for different residues. They are called secondary anchor positions. Basic and small P1 residues jointly account for 69.4% of the nonamers and 82% of the decamers in a published registry of B*2705 ligands (12). This restriction, together with substantial interactions of the peptidic R1 side chain in the A pocket of HLA-B27 (16), seemed to suggest that the frequency of basic P1 residues, particularly R, among HLA-B27 ligands is determined by structural complementarity, and that the peptidic R1 is a significant anchor residue, as also suggested by in vitro binding assays with poly-A analogs (15).

In addition to the binding features, Ag presentation depends on the availability of appropriate ligands. This, in turn, is determined by the efficiency of their generation along the Ag-processing pathway, and by the resistance of the peptide to cytosolic degradation. It has been estimated that ∼2000 protein molecules must be processed for each peptide copy that binds to an MHC class I molecule (35, 36). This is due to the extraordinary amino-peptidase activity in the cytosol to which peptides are exposed before binding to TAP, which results in a very short half-life of peptides in vivo (37). Thus, peptides resistant to amino-peptidase-mediated degradation would need lower thresholds for their generation to reach TAP, and they would compete advantageously for binding to MHC class I molecules. A consequence of this would be that such ligands might be presented and act as Ags even if they are produced in relatively low amounts or from nonabundant proteins in the cell. For this reason, the demonstration that peptides with N-terminal dibasic sequences were particularly resistant to cytosolic amino peptidases (17) was potentially very relevant for the immunological properties of HLA-B27 and a few other allotypes with specificity for R2 residues. On the basis of these observations it was proposed that the high frequency of HLA-B27 ligands with basic P1 residues would be significantly determined by their increased cytosolic stability and availability, rather than being a consequence of structural complementarity with the A pocket environment. A problem with this interpretation was that the frequency of R2 among ligands of non-B27 HLA class I molecules is not necessarily associated with a correspondingly high frequency of basic P1 residues, suggesting a critical role of the structural complementarity between the MHC protein and the peptide ligand in determining P1 preferences. For instance, of 17 HLA-B14 (B*1402 or B*1403) ligands with R2 from one study (32), none showed R or K at P1. Of 39 B*3901 ligands with R2 or K2 listed in the SYFPEITHI database (www.syfpeithi.de), none showed R or K at P1. Similarly, none of nine B*3909 reported ligands with R2 (38) showed a basic P1 residue. Finally, of five natural ligands of Gogo*0101 with R2, only one had K1 (39). All three HLA-B14, HLA-B39, and Gogo*0101 have T instead of E at position 163, but they have R62 and W167. Thus, in the absence of an appropriate structure of the peptide binding site, it seemed that the increased cytosolic stability of peptides with N-terminal dibasic sequences was not sufficient for their prevalent presentation by HLA class I molecules.

Our study sought to determine the role of the A pocket in HLA-B27 in the remarkable restriction of this molecule for basic and small P1 residues. In particular, the importance of residues E163 and W167 in determining the preference of HLA-B27 for R1 was investigated by mutating either E163 alone or both E163 and W167 to other naturally occurring polymorphisms at these positions. Contrary to our expectations, the T163 change had a limited effect on P1 residue usage, including R1. It is possible that, in the presence of this mutation, R1 residues can still bind with moderate loss of peptide stability. Support for this explanation can be found, for instance, in the crystal structure of H-K^b-VSV8 peptide complex (40), in which the peptidic R1 adopts an alternative binding mode with R62 and W167, possibly imposed by the presence of T163 in H-K^b (16). Thus, the lack of basic P1 residues among the peptide ligands of HLA-B14 or HLA-B39 cannot be simply due to their lack of E163.

Although the cumulative effect of the T163 and S167 mutations had a more significant effect on peptide specificity, the impact of the presumably severe disruption of the A pocket surface induced by these mutations on the frequency of basic P1 residues was much smaller than might have been anticipated. These results strongly suggest that the native structure of the A pocket surface in HLA-B27 is not a major determinant of the preference of this molecule for basic P1 residues. It is conceivable that peptide anchoring to HLA-B27 mediated by positions other than P1 may provide sufficient stability even if A pocket interactions are severely disrupted, obviating the requirement for an optimal anchor residue at P1. It is also possible that, even in the double mutant, basic P1 residues might adopt an alternative binding mode that provides significant stabilization of the P1 side chain. Although this is difficult to rule out in the absence of an x-ray diffraction model of the mutant, it would appear less likely, due to the drastic alteration of the A pocket surface induced by the double mutation. Note that the overall contribution of the P1 side chain to binding is not only a consequence of the alteration of nonbinding interaction patterns, but also of the changes induced by the mutations in the conformational entropy of binding, a component whose importance has been recently highlighted both in the process of molecular recognition by proteins (41) and for P1 and other anchor residues of MHC class I molecules (42). Since loss of conformational entropy counteracts the favorable contribution of nonbonding interactions, the mutations introduced in the A pocket surface of HLA-B27 may not only disrupt such contacts, which contribute to the enthalpic component of the Gibbs free energy of P1 interaction, but they also decrease the loss of conformational entropy by allowing a larger conformational space of the P1 side chain, thus partially compensating for the disrupting effect on the enthalpic component. We cannot formally rule out this possibility, but it seems unlikely that entropic effects alone might explain binding of peptides with basic P1 residues to pocket A surface mutants with significantly altered nonbonding interactions.

If HLA-B27 ligands can bypass the requirement of the canonic pocket A interactions to bind peptides with R1, this would support the claims by Herberts et al. (17) that the predominance of HLA-B27 ligands with R or K at P1 is favored by their resistance to amino-peptidase-mediated degradation in the cytosol, even if this feature is not reflected in a similar predominance among other HLA class I molecules, for which pocket A interactions might possibly be more determinant.

The only statistically significant effect induced by the double mutation on P1 residue frequencies concerned the increased number of peptides with G1 among those bound by B*2705 but not by T163S167. The high frequency of HLA-B27 ligands with G1 has not yet been satisfactorily explained, since this residue is thermodynamically disfavored compared with R1 (16), and the susceptibility of peptides with N-terminal GR sequences to cytosolic degradation has not been determined. However x-ray diffraction analyses showed that, in the presence of a peptidic G1 residue, the side chain of W167 in HLA-B27 folds down onto the peptide N terminus, which binds in the A pocket through a conserved network of hydrogen bonds, shielding it from the solvent (16, 43). To the extent to which this conformation may provide some stability, compensating in part for the loss of interactions with the peptidic R1 side chain, the absence of W167 in the double mutant may disfavor binding of peptides with G1, which would explain our results. Alternatively, an increased allowance of the double mutant for P1 residues less favored in B*2705 might also result in a decreased frequency of G1 among T163S167-bound peptides.

Thermostability analyses showed that both the wild type and the T163 mutant bind highly stable peptide repertoires. Some time-dependent increase of the global thermostability was observed for B*2705, which occurred equally in the presence of brefeldin A (data not shown). Therefore, the effect is probably due to the tapasin-mediated optimization of the B*2705-bound peptide repertoire in the endoplasmic reticulum, as reported in a previous study (44). This optimization was not observed in the T163 mutant. Thus, although peptide specificity was little affected by this mutation, it induced a certain loss of stability on its bound peptide repertoire. Although the low expression level of the T163S167 mutant precluded a reliable estimation of its thermostability, the larger cell surface dissociation observed by flow cytometry suggests that the overall stability of the T163S167/peptide complexes might be lower than for the wild type or the T163 mutant. If so, although the double mutation has a limited effect on the peptide repertoire and, in particular, on the frequency of peptides with basic P1 residues, the molecular stability of the complexes might be affected.

In conclusion, our results indicate that the native structure of the A pocket surface in HLA-B27 is largely dispensable for binding peptides with N-terminal basic residues to this molecule, although W167 may be significant in determining the high frequency of G1 among HLA-B27 ligands. That peptides with R or K at P1 bind to B27 mutants with a severely altered A pocket surface strongly suggests that HLA-B27 ligands are well stabilized in the peptide binding groove by other contacts. Therefore, in agreement with the suggestion by Herberts et al. (17), the high frequency of HLA-B27 ligands with N-terminal basic residues may be more determined by their cytosolic stability than by the canonic interactions of these residues with those at the A pocket surface. The potential consequences of this fact for the physiological function of HLA-B27 as an Ag-presenting molecule were properly noted in that previous study: HLA-B27 would present peptides generated in lower amounts than required for presentation by other MHC molecules, because such peptides would be less destroyed in the cytosol. The relevance of this feature for the pathogenetic role of HLA-B27 in spondyloarthropathies is less clear, since both disease-associated (i.e., B*2705) and non-disease-associated subtypes (i.e., B*2706) have identical structures in the A pocket and its vicinity, as well as similar high frequencies of ligands with basic P1 residues (12). However, individual peptides may be pathogenetically relevant in the context of only some subtypes, due to subtype-specific binding or antigenicity.

We thank the staff of the Proteomics facilities at the Centro de Biología Molecular “Severo Ochoa” and Centro Nacional de Biotecnología, Madrid (both members of the ProteoRed network) for help in mass spectrometry. Special thanks are given to our colleagues Miguel Marcilla for help in the anaysis of MS/MS data and to Andreas Ziegler and Barbara Uchanska-Ziegler (Institut fur Immungenetik, Charite-Universitatsmedizin Berlin) for critical reading of the manuscript and valuable insights.

The authors have no financial conflicts of interest.