The rheumatoid arthritis–associated allele HLA–DR10 (DRB1*1001) shares part of its repertoire with HLA–DR1 (DRB1*0101) and HLA–DR4 (DRB*0401)

Authors


Abstract

Objective

To identify the peptide anchor motif for the rheumatoid arthritis (RA)–related HLA allele, DR10, and find shared natural ligands or sequence similarities with the other disease-associated alleles, DR1 and DR4.

Methods

The HLA–DR10–associated peptides were purified, and a proportion of these natural ligands were de novo sequenced by mass spectrometry. Based on crystallographic structures, the complexes formed by peptide influenza virus hemagglutinin HA306–318 with DR1, DR4, and DR10 were modeled, and binding scores were obtained.

Results

A total of 238 peptides were sequenced, and the anchor motif of the HLA–DR10 peptide repertoire was defined. A large proportion of the DR10-associated peptides had the structural features to bind DR1 and DR4 but were theoretical nonbinders to the negatively associated alleles DR15 and DR7. Among the sequenced ligands, 10 had been reported as ligands to other RA-associated alleles. Modeling data showed that peptide HA306–318 can bind DR1, DR4, and DR10 with similar affinities.

Conclusion

The data show the presence of common peptides in the repertoires of RA-associated HLA alleles. The combination of the shared epitope present in DR1, DR4, and DR10 together with common putative arthritogenic peptide(s) could influence disease onset or outcome.

Rheumatoid arthritis (RA) is a chronic, inflammatory, autoimmune rheumatic disease with genetic and environmental components (for review, see ref. 1). The association between RA susceptibility and HLA has been well known for a long period of time (2, 3). HLA–DR alleles, such as DR1 (DRA*0101 and DRB1*0101), some subtypes of DR4 (DRA*0101, DRB1*0401, DRB1*0404, DRB1*0405, and DRB1*0408), and DR10 (DRA*0101 and DRB1*1001), are all associated with both disease susceptibility and disease severity. HLA–DR4 is the main allele associated with RA in Northern Europe (4), whereas the frequency of other DR4 subtypes and of DR1 and DR10 alleles is also increased in patients with RA from southern Europe and in other populations (5–12). DR10 has been described as conferring the highest susceptibility to RA in a Spanish population (13).

A major feature shared by these HLA–DR alleles is the presence of a basic consensus sequence in the third hypervariable region of DRB1, spanning residues 70–74 of the DRβ chain. This sequence is QRRAA for DR*0101, *0404, *0405, and *0408, QKRAA for DR*0401, and RRRAA for DR*1001. The presence of this basic “motif” led to the proposal of the “shared epitope” hypothesis (14), which postulates that the side chains of several of these amino acids could be involved in disease pathogenesis, by either defining the peptide preference or directly interacting with the T cell receptor (TCR), influencing selection of the T cell repertoire. In addition, a putative “protective epitope” has been defined for the same region, with the sequence DERAA, corresponding to DRB1*0402, *1102, *1301, *1302, and *1304, and is associated with less severe disease (15).

The mechanisms responsible for the autoimmune process in RA, including the autoantigens inducing or maintaining the autoimmune response in situ, remain unidentified. Some putative antigens have been proposed; one of them, type II collagen, is used to induce several experimental arthritis models in mice. Human type II collagen can also induce arthritis in mice transfected with human DR1 or DR4 (16, 17).

As a consequence of the sequence homology of the presenting molecules, similarities in the peptides presented by RA-associated alleles can be expected. The peptide repertoires for DR1 and DR4 have been described, and their specific peptide-binding motifs are known (18–20). In addition, the 3-dimensional (3-D) structures of these 2 alleles complexed with different peptides have been resolved, and the capacity of both alleles to bind the same peptides has been demonstrated. These include one of the best-studied antigenic peptides, influenza virus hemagglutinin HA306–318, which binds to DR4 and DR1. In addition, a T cell clone has been described that recognizes this peptide in the context of both DR1 and DR4, indicating that they can bind and present peptides in a similar manner (21).

The repertoire of peptides associated with DR10 has not been analyzed. So far, only 2 DR10-restricted epitopes have been reported (22, 23), and, therefore, the DR10 peptide-binding motif remains unknown. In this report, we describe such a motif after the analysis of 238 DR10 natural ligands using mass spectrometry (MS) and compare this repertoire with those of DR1 and DR4.

MATERIALS AND METHODS

Cell lines and antibodies.

BEN-DR10 is an HLA–DR10 (DRA*0101, DRB1*1001) homozygous lymphoblastoid cell line with the class I genotype HLA–A*0101, A*2301, B*3701, B*5101, Cw*01BH, and Cw*06BF. It was derived by Epstein-Barr virus (EBV) transformation of B lymphocytes from fresh peripheral blood mononuclear cells following the standard procedure (24). Cells were grown in roller bottles in RPMI (Sigma-Aldrich, St. Louis, MO) with 10% fetal bovine serum (FBS) (Invitrogen, Carlsbad, CA), washed twice in cold phosphate buffered saline, and stored at −70°C. The HLA–DR–specific monoclonal antibody B8.11.2 (IgG2b) (25) was a kind gift from Dr. F. Koning (Leiden University, Leiden, The Netherlands).

HLA–DR purification and MS.

Peptide α/β–HLA–DR complexes were purified as described previously (26), with modifications. Briefly, ∼2 × 109 cells were lysed in 20 mM Tris HCl (pH 7.4), 150 mM NaCl, 1% Nonidet P40 (NP40) with protease inhibitors (5 μg/ml aprotinin, 1 mM EDTA, 5 mM iodoacetamide, 5 μg/ml leupeptin, 0.05% [weight/volume] NaN3, 5 μg/ml pepstatin, and 1 mM phenylmethylsulfonyl fluoride) at 4°C for 1 hour. Lysates were centrifuged for 15 minutes at 1,500g, and supernatants were ultracentrifuged for 1 hour at 100,000g. Supernatants were precleared using a Tris-blocked Sepharose column, and the flow-through was loaded on a B8.11.2 Sepharose column.

Columns were washed with 200 volumes of NaCl–EDTA–Tris (NET) buffer (50 mM Tris HCl, pH 7.4, 150 mM NaCl) with 10% saturated NaCl and 0.5% NP40, followed by a second wash with NET, 5% saturated NaCl, and 0.5% NP40, and finally with NET buffer. Peptide–HLA complexes were eluted with 0.1% trifluoroacetic acid (TFA), and 1-ml fractions were collected. Protein-containing fractions were pooled and concentrated in a SpeedVac (Thermo Fisher Scientific, Waltham, MA), and peptides were purified by ultrafiltration in a Centricon-10 device (Amicon, Beverly, MA) and concentrated for chromatography. The purity of the immunoprecipitated HLA–DR α-chains and β-chains was controlled by sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE) of the >10-kd material retained in the Centricon-10 device.

In one experiment, the HLA–DR–associated peptide pool was fractionated by reverse-phase high-performance liquid chromatography (HPLC), as described previously (27), in a 25 × 0.21–cm C18 Tracer analytical column (ODS-A, 5 μm) (Teknokroma, Barcelona, Spain), using an Ettan HPLC system (GE Healthcare, Fairfield, CT). The gradient was 0–10% eluent B in 5 minutes and 10–60% eluent B in 90 minutes, where eluent A = 0.1% TFA in water, and eluent B = 80% acetonitrile in A. Fractions were collected every 0.5 minute. One microliter of each fraction was analyzed by matrix-assisted laser desorption ionization−time-of-flight (MALDI-TOF) MS on an Ultraflex spectrometer (Bruker, Bremen, Germany), and 20 μl was dried and resuspended in 20 μl 0.1% formic acid in water for nano–electrospray ionization (nanoESI) analysis. Fractions were analyzed on an Esquire HCT ion trap mass spectrometer (Bruker), coupled to a nano-HPLC system (UltiMate; LC Packings, Amsterdam, The Netherlands).

Peptide fractions were first concentrated on a 300-μm inner diameter (ID) × 1 mm PepMap nanotrapping column and then loaded onto a 75-μm ID × 15 cm PepMap nanoseparation column (LC Packings). Peptides were then eluted by an acetonitrile gradient (0–60% eluent B in 35 minutes and 60–80% eluent B in 5 minutes, where A = 0.1% formic acid in water and B = 80% acetonitrile in A; flow rate ∼250 nl/minute) through a PicoTip emitter nanospray needle (New Objective, Woburn, MA) onto the nanospray ionization source of the ion trap mass spectrometer. MS/MS fragmentation (1.9 seconds, 100–2,800 mass/charge [m/z]) was performed on 2 of the most intense ions, as determined from a 1.2-second MS survey scan (310–1,500 m/z), using a dynamic exclusion time of 1.2 minutes for precursor selection. Automated optimization of MS/MS fragmentation amplitude, starting from 0.60V, was used.

In a second experiment, peptides were fractionated by strong cationic exchange–reverse-phase 2-D chromatography prior to on-line MS analysis. The peptide mixture was injected onto a 300-μm ID × 10 cm Poros 10S strong ion exchange (SCX) column (LC Packings) equilibrated with 0.1% formic acid, 5% acetonitrile, at a flow rate of 30 μl/minute, during 5 minutes. Peptides were then eluted from the SCX column by successive injections of 20-μl solutions of ammonium bicarbonate in increasing concentrations in 0.1% formic acid, 5% acetonitrile (0, 10, 20, 50, 100, 200, 400, 800, 1,200, 1,600, and 2,000 mM ammonium bicarbonate). The unretained fraction and each of the salt-eluted fractions were successively transferred to a 300-nm ID × 1 mm PepMap reverse-phase nanotrapping column for concentration and desalting, at a flow rate of 30 μl/minute. Peptides in each fraction were separated by nano–reverse-phase chromatography and analyzed on-line by electrospray MS, exactly as in the first experiment but using an extended gradient, going from 0–60% eluent B in 150 minutes. The whole process was automated, using an autosampler (Famos; LC Packings) and a 2-valve switching module (Switchos; LC Packings) controlled with HyStar version 2.3 software (Bruker).

MS data were processed using the Data Analysis 3.2 and BioTools 2.1 software packages (Bruker). Database searches were performed using the Mascot search engine (Matrix Science, London UK).

Statistical analysis.

The statistical analysis was performed as previously described (28). Briefly, the frequency of a given amino acid residue in a given peptide position was compared with the mean frequency of that residue among human proteins (http://www.ebi.ac.uk/integr8/StatsAminoPage.do?orgProteomeId=25), using the chi-square test with Yates' correction. Bonferroni correction was used (in this case, multiplying the P value by 20). P values less than 0.05 before and after Bonferroni correction were considered.

Theoretical binding.

The ProPred program (http://www.imtech.res.in/raghava/propred/) (29) was used to calculate the theoretical binding score. A 3% threshold was used to select positive from negative binding.

Structure modeling.

Modeling of the DR10–HA306–318 structure and assignment of a binding score relative to DR1 and DR4 were achieved using a recently developed simulation protocol (Muixí L, et al: unpublished observations). For a summary of the method and the results, see Supplemental Data, available on the Arthritis & Rheumatism Web site at http://www.mrw.interscience.wiley.com/suppmat/0004-3591/suppmat/.

RESULTS

Description of the DR10-bound peptide repertoire.

Analysis of the DR10 repertoire was done by de novo sequencing of DR10-bound peptides using ion-trap MS after HPLC fractionation. Two independent experiments were performed. First, 180 fractions were collected and analyzed by MALDI-TOF MS and electrospray ion-trap MS after the DR10-bound peptide pool reverse-phase HPLC fractionation. Second, 2-D reverse-phase SCX chromatography was performed, and eluted peptides were analyzed by MS/MS fragmentation, using an electrospray ion trap connected online to the HPLC system. An example of the MALDI-TOF MS spectrum of one of the HPLC fractions obtained in the first experiment is shown in Figure 1A. All peptide-containing fractions showed peaks, with sizes within the range of HLA–DR ligands. In addition, SDS-PAGE of the Centricon-10–retained material confirmed the purity of the immunoprecipitation (data not shown). De novo peptide sequencing was accomplished by using electrospray MS/MS. The spectra and fragment assignation of 2 peptides are shown in Figures 1B and C.

Figure 1.

Examples of matrix-assisted laser desorption ionization−time-of-flight (MALDI-TOF) and nano–electrospray ionization mass spectrometry (MS) spectra. A, MALDI-TOF MS spectrum of the reverse-phase high-performance liquid chromatography fraction 79. One microliter of each 100-μl fraction was loaded in an AnchorChip plate and analyzed by MALDI-TOF MS. B, Electrospray MS/MS spectrum of an ion peak at 679.0 mass/charge (m/z). Detected fragment ions of the main b and y series and the deduced peptide sequence are indicated. The sequence corresponds to residues 680–696 of transferrin receptor 1. C, Electrospray MS/MS spectrum of an ion peak at 783.8 m/z. Detected fragment ions of the main b and y series and the deduced peptide sequence are indicated. The sequence corresponds to residues 39–52 of elongation factor 1. a.u. = arbitrary units.

A total of 238 natural DR10 ligands were identified. Most consisted of nested sets, as defined for other HLA class II alleles. Thus, all peptides in the DR10-bound peptide pool could be grouped into 88 different sets, of which 43 were composed of a single peptide, and 45 contained 2–25 different peptides (see Supplemental Table 1, available on the Arthritis & Rheumatism Web site at http://www.mrw.interscience.wiley.com/suppmat/0004-3591/suppmat/). The peptide size showed a normal distribution, with molecular weights ranging from 1,193.6 daltons to 2,876.7 daltons (average molecular weight 1,828.3 daltons) (Figure 2A). Peptides arose from 84 different proteins: 83 human proteins and bovine transferrin derived from the FBS added to the culture medium. Peptides were derived from proteins from different cell compartments. Although some peptides could be localized in <1 cell compartment, we used the most common localization for the analysis, to avoid redundancy. As expected, most were external or internal membrane proteins or residents of the endocytic pathway. Localization of the rest of the proteins followed a pattern common to other EBV cell line repertoires (30) (Figure 2B).

Figure 2.

Peptide size distribution and intracellular origin of parental proteins. A, Peptide size distribution of the DR10-associated peptides. Values are the total number of peptides in a molecular weight range between 1,000 daltons (Da) and 3,000 daltons, in groups of 100 daltons. B, Intracellular distribution of the DR10-associated peptide parental proteins. Values are the percentage of proteins for each intracellular localization, as described in the Swiss-Prot database (http://www.expasy.org/). M = membrane; C = cytosol; G = Golgi apparatus; E/L = endosome/lysosome compartments; ER = endoplasmic reticulum; S = secreted; MIT = mitochondria; ND = not determined.

Core residue use among DR10 natural ligands.

DR10 and DR1 are different in 16 residues, and DR10 and DR4 are different in 19 residues. Figure 3A shows the polymorphic changes between DR10 and DR1 or DR4. Differences localized in the binding groove are far from the pocket where the position 1 (P1) core residue is anchored (Figure 3B). Consequently, peptides bound to DR1 and DR4 contain the same aromatic or hydrophobic residues at P1 (Tyr, Phe, Trp, Leu, Ile, Ala, Val, or Met) (http://www.syfpeithi.de/Scripts/MHCServer.dll/FindYourMotif.htm), and therefore the same should apply to DR10. Thus, to define the anchor motif in DR10, an aromatic or hydrophobic residue was assigned to P1 of the peptide core in all DR10-associated peptides, considering the same core sequence for all peptides conforming to any given nested set (see Supplemental Table 1, available on the Arthritis & Rheumatism Web site at http://www.mrw.interscience.wiley.com/suppmat/0004-3591/suppmat/).

Figure 3.

HLA–DR1, DR4, and DR10 sequences. A, Primary sequences of DR1, DR4, and DR10 molecules. Variable residues for DR10 and DR1 or DR4 are indicated. B, Localization of DR10 polymorphic residues in the peptide-binding groove of DR1 and DR4. The α-chain is shown in pink, the β-chain is shown in orange, and the peptide scaffold is shown in green. Differential residues between DR1 and DR10 (left) and between DR4 and DR10 (right) are shown in blue.

A frequency analysis of the amino acids occupying each core position was performed, considering only the 88 different peptide cores, independently of the number of peptides per sequence. The relative frequency of all amino acids was obtained for each of the 9 core positions. This experimental frequency was normalized over the mean frequency of the same residue among human proteins (http://www.ebi.ac.uk/integr8/StatsAminoPage.do?orgProteomeId=25). Figure 4A shows a summary of the relative frequency of each amino acid in each of the 9 positions and the increase or decrease in frequency respective to their relative abundance in the human proteome.

Figure 4.

Description of the DR10 peptide anchor motif. A, Residue use among DR10 natural ligands. For each of the peptide core positions (P), the percent frequency of each amino acid (residue frequency [RF]) was calculated. Normalized values were calculated as the number of times that each RF was increased or decreased relative to the mean frequency of the corresponding residue among human proteins (deviation from the mean in the proteome [DMP]). DMP values are also presented as histograms. Red bars show DMP values that were significantly increased (P < 0.05) after Bonferroni correction, yellow bars show DMP values that were significantly increased before, but not after, the correction, and black bars show nonsignificant DMP values. B, DR10 peptide-binding motif. Residues that were statistically significantly increased after Bonferroni correction (top) were considered to be main anchor residues. Also shown are secondary residues, including those that were significantly increased before Bonferroni correction (DMP >1) (bottom).

The data showed statistically significant differences of at least 1 residue in every single core position before Bonferroni correction, except for P7 (Figure 4A). The uncorrected data revealed a list of favored residues in each position but not a valuable structural motif, which could be defined only with statistically significant differences after the correction. These were observed for residues occupying P1, P4, P5, P6, P8, and P9. Thus, P1 was mainly occupied by Phe and Tyr (10.6-fold and 12.1-fold increase), Leu was preferred in P4 (5.9-fold increase), Lys in P5 (2.8-fold increase), Ala and Pro in P6 (2.3-fold and 3.7-fold increase, respectively), Arg and Tyr in P8 (4.8-fold and 3.9-fold increase, respectively), and Asn and Ser in P9 (10.0-fold and 2.3-fold, respectively). The remaining favored residues constituted the secondary residues for the structural binding motif (Figure 4B). Comparing the resulting motif with those from associated and nonassociated alleles, the pattern showed that the DR10 motif shared P1 and at least 1 of the other 2 main anchor positions (P4 or P9) with DRB1*0101 (P4), *0401 (P9), and *0405 (P4 and P9). Only DRB1*0404 did not share the P1 residues but had compatible P4 and P9 content. In contrast, a maximum of 1 position was shared between the DR10 motif and those of nonassociated alleles DRB1*0402 (P9), *0701 (P1), or *1501 (none) (data not shown).

Theoretical binding of DR10 natural ligands to RA-associated alleles.

HLA–DRB1*0101, *0102, *0401, *0404, *0405, and *0408 are alleles that are positively associated with RA and carry the shared epitope. Negatively associated alleles are DRB1*0701 and *1501. Other alleles such as DRB1*0402, *1102, *1301, *1302, and *1304 contain the protective epitope and are associated with less severe disease. We used the PropPred program to theoretically define how many of the DR10 ligands could be bound to those alleles (see Supplemental Table 2, available on the Arthritis & Rheumatism Web site at http://www.mrw.interscience.wiley.com/suppmat/0004-3591/suppmat/). A total of 79% of the peptides showed some binding to ≥1 of the alleles analyzed. Between 20% and 54% were potential good binders to at least 1 positively associated allele (group 1), and <2% were potential good binders to negatively associated alleles (group 2). Among the protective epitope alleles, the results were more heterogeneous: >20% of the peptides were potential good binders to *1302, but only 4–14% were potential good binders to DRB1*0402, *1102, *1301, and *1304 (Figure 5). Thus, at least theoretically, DR10 could share part of its repertoire with the main RA-associated alleles as well as with one of the protective epitope carriers, DRB1*1302.

Figure 5.

Theoretical binding of DR10 ligands to rheumatoid arthritis (RA)–related HLA–DR alleles. Data are represented as the percentage of DR10 peptides that are theoretical good binders to RA-associated alleles containing the “shared epitope” (solid bars), negatively associated alleles (open bars), and alleles containing the “protective epitope” DERAA (shaded bars).

Search of DR10 natural ligands in the peptide repertoires of other RA-associated alleles.

To determine whether this theoretical overlap could be confirmed by our data, a search was performed to find any known DR association of each DR10 ligand. Ten peptides that were identical to or length variants of the DR10-bound peptides had previously been reported as natural ligands for RA-associated HLA molecules (2 to DRB1*0101, 3 to DRB1*0401, and 5 to DRB1*0405) (Table 1). Thus, the data confirm that the DR10 peptide repertoire overlaps with other RA-associated alleles. In addition, 1 peptide had been described as being associated with DRB1*1101.

Table 1. HLA–DR10–associated peptides reported as natural ligands for other alleles*
SequenceAlleleRef.
  • *

    Boldface indicates part of the sequence corresponding to the core. Italics indicate a part of the sequence that is not totally shared between the DR10 peptides and peptides from other alleles.

EDPAFYMLKGKLQYELDRB1*010141
RVEYHFLSPYVSPKESPDRB1*010142
DGKDYIALNEDLRSWTAADRB1*040143
VDYFLNVATAQGRYYDRB1*040143
IEKFEKEAAEMGKGDRB1*040130
KPPQYIAVHVVPDQLMDRB1*040544
KPPQYIAVHVVPDQLDRB1*040544
KPPQYIAVHVVPDQDRB1*040544
FREFKLSKVWRDQHDRB1*040545
REFKLSKVWRDQHDRB1*040545
DLHSYVVMNHGRSYTAISDRB1*110141

Modeling.

To test whether a peptide capable of binding DR1 and DR4 may also bind DR10 with favorable energetics, a computer simulation study was performed. The antigenic peptide derived from HA306–318 (31, 32) was chosen as a test case, given the availability of crystallographic structures for its complex with DR1 and DR4 (Protein Data Bank entries 1DLH and 1J8H, respectively). Two models of DR10 were generated (DR101DLH and DR101J8H), using the DR1 and DR4 crystallographic structures as templates. Then, the interaction of HA306–318 with DR101DLH and DR101J8H was modeled (DR10–HA1DLH and DR10–HA1J8H), again using the known DR1 and DR4 complex structures as templates. The modeling procedure delivers a relative binding score in energy units. To make the comparison possible, the interaction of HA306–318 with DR1 and DR4 was remodeled following the same procedure (DR1–HA1DLH and DR4–HA1J8H). This remodeling exercise was performed not to reproduce the original crystallographic structure but to measure the work required to build the side chains of HA306–318 within the binding groove of DR1 and DR4.

The relative binding scores obtained for the different complexes were −208 kJ/mole for DR10–HA1DLH, −197 kJ/mole for DR10–HA1J8H, −239 kJ/mole for DR1–HA1DLH, and −329 kJ/mole for DR4–HA1J8H, suggesting that 1) the 2 models for DR10, although structurally significantly different (given the different templates used), show a difference in the relative binding score for HA306–318 that is within the uncertainty of the method; 2) the relative scores for binding of HA306–318 to DR10 and DR1 are very close; and 3) binding to DR4 appears to be most favored. In both the crystal structures and the 2 DR10 models, Tyr308 from the HA peptide interacted with the side chains of the same HLA residues and was almost identically positioned, supporting the assignment of the P1 core residue for DR10. We cannot draw a conclusion regarding the manner in which the residues conforming the shared epitope are exposed to the TCR, because they are not fixed residues. However, as in the DR1 and DR4 templates (see Supplemental Data, available on the Arthritis & Rheumatism Web site at http://www.mrw.interscience.wiley.com/suppmat/0004-3591/suppmat/), the side chain from the P4 residue (Gln4) in the 2 models can be visualized at enough distance from P71 of the major histocompatibility complex β-chain to generate a hydrogen bond between the 2 chains.

DISCUSSION

Part of the peptide repertoire that naturally associates with DR10 has been described. Although the association of RA with DR10 in several populations is well documented (5–12), no natural ligands associated with DR10 are known, unlike the other RA-associated HLA alleles (mostly DR1 and DR4, with well-defined peptide repertoires), and only 2 DR10-restricted T cell epitopes have been described (22, 23).

We sequenced 238 DR10-associated peptides by nanoESI ion trap MS. As expected, most of the proteins from which the ligands were derived were located in membrane compartments, including plasma membrane, Golgi apparatus, endosomes, lysosomes, endoplasmic reticulum, and secretion vesicles. Several proteins (nearly 20% of the total) arose from cytosolic proteins, consistent with previous findings for other alleles (30). However, several parental cytosolic proteins (6 of 15) could be alternatively located in membrane compartments.

HLA–DRB1 alleles present a Gly–Val dimorphism in the DRβ86 position that influences P1 specificity (33–35). The strong predominance of Phe1 and Tyr1 in DR1 and DR4 natural ligands has been related to the presence of Gly in the DRβ86 position, compared with other alleles (36–38). All residues that could directly interact with the P1 peptide were identical for DR1, DR4, and DR10, including the Glyβ86 position, suggesting that very similar hydrophobic residues should be found in P1. All of the sequenced peptides contained at least 1 hydrophobic or aromatic residue that could be selected as P1, and we used this feature to define the peptide core. Supporting this, the modeling experiment showed that Tyr308 from HA306–318 accommodates in the P1 pocket of DR10 in the same way, i.e., interacting with the same HLA side chains and adopting the same orientation as in the P1 pockets of DR1 and DR4.

The P4 peptide of the DR10 anchor motif looks more like P4 in DR1 than in DR4. Although all 3 alleles can bind peptides with Leu4 or Ile4, these are main residues only for DR1 and DR10, while for DR4 peptides, acid residues are more abundant in P4. Acid residues in P4 were not favored but were permitted in DR10, as demonstrated by the sequencing of 3 peptides containing Glu4 or Asp4. In the crystal structure of the complex DR4–type II collagen (CII)1168–1180, Glu4 of the peptide interacted with Lys in DRβ71 (36). The recently resolved structure of DR1 complexed with immunogenic CII259–273 revealed that its Glu4 (39) similarly interacted with Arg in DRβ71. HLA–DR10 also contains Arg in P71, which may act as in DR1. In any case, the binding affinity of CII259–273 to DR1 and DR4 improved when Glu4 was substituted by Ala4 (40), suggesting that peptides with acidic residues in P4 may not be the best binders for any of the RA-related HLA molecules. Nevertheless, self peptides are not necessarily expected to be the best binders.

In our analysis, the P8 peptide residue for DR10 was one additional unexpected anchor position, with basic residues (mostly Arg), and Tyr and Val as the most favored amino acids. According to a recent report, a DR10-restricted melanoma-specific T cell clone recognized a minimum epitope with the decamer sequence YFAAELPPRN (23), where the 2 last residues were Arg and Asn, as in P8 and P9 of the DR10 anchor motif (Figure 4). Both the N-terminal and C-terminal residues, Tyr and Asn, respectively, were necessary for T cell recognition (23), i.e., the whole decamer was required. Because both Tyr and Phe at the N-terminal end of the sequence could occupy the P1 core, 2 possibilities arise: either Phe is the P1 anchor residue and Tyr is required for TCR interaction, or Tyr is the P1 anchor residue and the peptide core is a decamer. Although in our analysis only nonamers were considered to form the peptide cores, a few peptides could better fit to the motif if a decamer was located in the core. An example is the class I–derived set with assigned core YIAL(K/N)EDL(R/S), which would better fit to the motif if the core was instead the decamer YIAL(K/N)EDL(R/S)S.

In addition, the unique DR10-restricted T cell epitope YFAAELPPRN contained Ala4 (nonamer) or Glu4 (decamer). Although these were not the preferred P4 residues in our analysis, they were permitted, because 3.4% and 2.3% of the peptide cores in the peptide pool contained Ala4 or Glu4, respectively. Thus, our motif is compatible with the only DR10-restricted epitope described. Another reported T cell epitope that could be DR10 restricted was recognized much better in the context of DR11 (DRB1*1101) (22).

The possibility of common peptides for the RA-associated DR10, DR1, and DR4 repertoires was also suggested by our data. Indeed, 10 of the sequences had previously been reported as natural ligands for DRB1*0101 (n = 2), *0401 (n = 3), and *0405 (n = 5). These data are consistent with theoretical binding data (see Supplemental Tables 1 and 2, available on the Arthritis & Rheumatism Web site at http://www.mrw.interscience.wiley.com/suppmat/0004-3591/suppmat/) and with the motif comparison between all of these alleles, where that of DRB1*0405 was the most similar to that of DR10, sharing P1, P4, P6, and P9. One of the peptides was also a putative good binder to DRB1*1101.

We used HA306–318 to model DR10–peptide interactions, because it is the only peptide that has been crystallized, forming a complex with either DR1 or DR4 (31, 32). In our model, the nonamer YVKQNTLKL was the binding core for DR1 and DR4 and also for DR10. HA306–318 is a good binder for many alleles, and the modeling data suggested that its physical properties make it a good binder for DR10 as well. The data also supported our assumption of a hydrophobic or aromatic residue in P1 of DR10, because the interaction of this single peptide with P1 in the 3 alleles is almost identical. There are differences in other positions, but these differences are not greater than those observed between the DR1–HA and DR4–HA crystals (see Supplemental Data, available on the Arthritis & Rheumatism Web site at http://www.mrw.interscience.wiley.com/suppmat/0004-3591/suppmat/). The relevance of these differences for T cell recognition is unknown, but the differences do not prevent recognition of both complexes by the same T cell clone (31). Therefore, although we cannot draw conclusions in this regard from the modeling data, the possibility remains that T cells recognize the same peptide presented by the 3 alleles.

AUTHOR CONTRIBUTIONS

Dr. Jaraquemada 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. Alvarez, Jaraquemada.

Acquisition of data. Alvarez, Collado, Daura, Colomé, Canals.

Analysis and interpretation of data. Alvarez, Jaraquemada.

Manuscript preparation. Alvarez, Daura, Jaraquemada.

Statistical analysis. Alvarez.

Generation and typing of HLA–DR10 homozygous EBV cell line. Rodríguez-García, Gallart.

Acknowledgements

We thank Eduard Palou for HLA typing and J. A. López de Castro for his help with the statistical analysis. Dr. Daura acknowledges the Port d' Informació Cientifica and the Òliba project of Universitat Autònoma de Barcelona for providing storage and computational resources.

Ancillary