The interplay between citrullination and HLA-DRB1 polymorphism in shaping peptide binding hierarchies in rheumatoid arthritis

The HLA-DRB1 locus is strongly associated with rheumatoid arthritis (RA) susceptibility, whereupon citrullinated self-peptides bind to HLA-DR molecules bearing the shared epitope (SE) amino acid motif. However, the differing propensity for citrullinated/non-citrullinated self-peptides to bind given HLA-DR allomorphs remains unclear. Here, we used a fluorescence polarization assay to determine a hierarchy of binding affinities of 34 self-peptides implicated in RA against three HLA-DRB1 allomorphs (HLA-DRB1*04:01/*04:04/*04:05) each possessing the SE motif. For all three HLA-DRB1 allomorphs, we observed a strong correlation between binding affinity and citrullination at P4 of the bound peptide ligand. A differing hierarchy of peptide-binding affinities across the three HLA-DRB1 allomorphs was attributable to the β-chain polymorphisms that resided outside the SE motif and were consistent with sequences of naturally presented peptide ligands. Structural determination of eight HLA–DR4–self-epitope complexes revealed strict conformational convergence of the P4-Cit and surrounding HLA β-chain residues. Polymorphic residues that form part of the P1 and P9 pockets of the HLA-DR molecules provided a structural basis for the preferential binding of the citrullinated self-peptides to the HLA-DR4 allomorphs. Collectively, we provide a molecular basis for the interplay between citrullination of self-antigens and HLA polymorphisms that shape peptide–HLA-DR4 binding affinities in RA.

The HLA-DRB1 locus is strongly associated with rheumatoid arthritis (RA) susceptibility, whereupon citrullinated self-peptides bind to HLA-DR molecules bearing the shared epitope (SE) amino acid motif. However, the differing propensity for citrullinated/non-citrullinated self-peptides to bind given HLA-DR allomorphs remains unclear. Here, we used a fluorescence polarization assay to determine a hierarchy of binding affinities of 34 self-peptides implicated in RA against three HLA-DRB1 allomorphs (HLA-DRB1*04:01/*04:04/*04:05) each possessing the SE motif. For all three HLA-DRB1 allomorphs, we observed a strong correlation between binding affinity and citrullination at P4 of the bound peptide ligand. A differing hierarchy of peptide-binding affinities across the three HLA-DRB1 allomorphs was attributable to the ␤-chain polymorphisms that resided outside the SE motif and were consistent with sequences of naturally presented peptide ligands. Structural determination of eight HLA-DR4 -self-epitope complexes revealed strict conformational convergence of the P4-Cit and surrounding HLA ␤-chain residues. Polymorphic residues that form part of the P1 and P9 pockets of the HLA-DR molecules provided a structural basis for the preferential binding of the citrullinated self-peptides to the HLA-DR4 allomorphs. Collectively, we provide a molecular basis for the interplay between citrullination of selfantigens and HLA polymorphisms that shape peptide-HLA-DR4 binding affinities in RA.
Rheumatoid arthritis (RA) 5 is an autoimmune disease of the synovial joints. A characteristic of RA is the presence of anticitrullinated protein antibodies (ACPA) in sera, for which ϳ70% of all patients are seropositive (1). ACPA target proteins that have undergone citrullination, a post-translational modification (PTM) process driven by a family of enzymes known as peptidyl-arginine deiminases (PAD), convert arginine to citrulline (2)(3)(4). PAD type 2 (PAD-2) and PAD-4 expression is closely associated with synovial joint inflammation in RA patients (5). Citrullination creates neo-self-antigens, and given the high rate of ACPA in RA patients, these antigens are considered the prime generators of the autoimmune CD4 ϩ T cell response in ACPAϩ RA. Citrulline-specific, antigen-experienced CD4 ϩ T cells are found in HLA-DR4 (DRA1*01:01/HLA-DRB1*04:01) RA patients as well as in HLA-DR4 (DRA1*01:01/HLA-DRB1*04:01) transgenic mice primed with citrullinated selfantigens (6 -11). Furthermore, citrulline-specific Th1 and Th17 cells are increased in the HLA-DRB1*04:01 ϩ RA patients, and pro-inflammatory cytokines are produced by CD4 ϩ T cells in response to citrullinated self-antigens (8 -11). Sources for the primary RA-associated autoantigens may be from the site of disease, including articular cartilage and synovial fluids (12- 14), but others may be derived from blood plasma or surrounding mucosal tissues that are susceptible to inflammation (2). These proteins could undergo PTMs during numerous physiologic processes, including infection, apoptosis, and cellular stress. Some of the best-characterized autoantigens that bind ACPAs are citrullinated vimentin, fibrinogen, ␣-enolase, and type II collagen, which are present at high levels in the joint synovium (3,15).
One of the key inherited risk factors that contribute to ACPA-positive RA is the human leukocyte antigen (HLA) class II loci, namely HLA-DRB1, which encodes the HLA class II antigen-presenting molecules (16 -21). The antigen-binding groove of the HLA class II molecule can accommodate peptide ligands that vary in length, but the main pockets that interact most strongly with the bound peptide are P1, P4, P6, P7, and P9, which can accommodate the side chains of the peptide residues 1, 4, 6, 7, and 9 (22,23). A conserved amino acid sequence QKRAA, QRRAA, or RRRAA in position 70 -74 of the HLA-DRB1 chain, known as the shared epitope (SE) motif, is highly prevalent (ϳ90%) among ACPA seropositive patients (11,16). This SE motif defines the P4 pocket of the high-risk HLA-DRB1 RA-associated allomorphs. Subsequent genome-wide association studies have shown that two polymorphisms encoding ␤-chain residues at positions 11 and 13 at the base of the P4 pocket are also strongly associated with RA susceptibility (16). We have previously reported the structural basis for the association of the SE, citrullination, and RA by showing the size and charge of the P4 pocket accommodate antigens with a citrulline residue in the P4 position but prevents the binding of peptide ligands with the natively-encoded positively-charged P4-Arg (24).
Although citrullinated peptides have generally shown enhanced affinity to several SE HLA-DRB1 allomorphs relative to native peptides (25), the affinity of individual peptides has not been compared, with respect to polymorphisms outside the SE motif. Here, we show the affinity of citrullinated peptides across three different HLA-DRB1 allomorphs and reveal the key interactions between RA-associated antigens with RA-susceptible HLA-SE allomorphs. The crystals structures of eight citrullinated peptides bound to HLA-DRB1*04:01/*04:04/ *04:05 demonstrate the close convergence of the binding modes of the P4-Cit, with polymorphisms outside this pocket accounting for the hierarchy of binding of citrullinated selfpeptides to the three HLA-DR4 allomorphs. These specificities were also revealed in the natural repertoire of bound peptides, suggesting similar selective pressure operates on the self-peptide repertoire (immunopeptidome). Accordingly, we provide a molecular understanding of the interplay between the citrullination of self-antigens and HLA polymorphism that shape peptide-HLA binding affinity in RA.

P1 and P9 specificities distinguish the shared epitope-positive HLA-DR4 peptidomes
The extent to which polymorphisms outside the P4 pocket contribute to the natural selection of peptides was established using monoallelic HLA-DR4ϩ antigen-presenting cells. For this purpose, we generated transfectants of the T2 cell line (class II-deficient) that expressed HLA-DM and either HLA DRB1*04:01, 04:04, or 04:05. We have previously reported the immunopeptidomes of the HLA-DRB1*04:01ϩ and 04:04ϩ cells (24). Each of these datasets contained over 1000 -3000 unique and high confidence peptides, and although earlier reports have identified some HLA-DRB1*04:05 peptide ligands (26), here we report a large dataset of peptide ligands (n ϭ 2935) from the same parental T2 cell line (Table S1). These endogenous peptide sequences determined from multiple peptide elu-tion experiments were identified with high confidence using strict bioinformatic criteria that includes the removal of common contaminants (27). The motif for HLA-DRB1*04:05 generated using this approach (Fig. 1A) was in general agreement with previously determined motifs (28,29) and, like the other SEϩ DR4 allomorphs, disfavored Arg at P4 and showed strong preferences at P1 (Phe, Tyr, and Ile), P6 (Asp, Asn, and Thr), and P9 (Asp and Glu). To compare these preferences to the other SEϩ HLA-DR4 allomorphs examined in this study, we generated a heat map of difference matrices to highlight differences in their naturally selected ligands (Fig. 1B). These heat maps show the difference in amino acid usage at each position of the core 9-mer of peptides bound to HLA-DRB1*04:05 compared with HLA-DRB1*04:01 or HLA-DRB1*04:04. Of note, HLA-DRB1*04:05 selected peptides with similar P1 usage to HLA-DR*04:01 but a quite different selection of peptides with acidic P9 residues. In contrast, P1 and P9 differences were observed when HLA-DR*04:05-bound peptides were compared with HLA-DR*04:04-bound peptides, with an increase in aromatic amino acid selection at P1 and increased acidic residue selection at P9.
Peptides with relative IC 50 Ͼ250 M and Ͻ5 M were considered non-binding and moderate binders to the HLA allomorphs, respectively. Peptides with relative IC 50 Ͻ1 M (Table  1) for the HLA allomorphs were considered to have high affinity. The fluorescence polarization values were converted to percentage of binding using the values for fully bound (ϳ150 -250 mP) and free fluorescence (ϳ50 mP). Our aim was to determine the following: (a) whether the enhanced capacity of HLA-SE allomorphs to present self-antigens is restricted to P4-Cit or HLA DRB1 polymorphism and rheumatoid arthritis whether additional positions influence HLA binding; (b) whether a citrullinated epitope could bind to all HLA-SE with high affinity despite the polymorphisms in the peptidebinding cleft; and (c) whether there is a conserved structural basis for binding of citrullinated peptides to all three HLA allomorphs.

HLA DRB1 polymorphism and rheumatoid arthritis
P4-Cit was increased by at least 4-fold in comparison with P4-Arg in the native sequence when bound to HLA-DRB1*04:01 (Table 1   Underlined are the predicted core binding region from P1 to P9 pocket. Citrulline is labelled as X. Mean values represent IC 50 obtained from three independent experiments. Binding affinity of Ͼ250 M is considered no binding (NB).

HLA DRB1 polymorphism and rheumatoid arthritis
citrullination outside the P4 pocket does not impact the peptide affinity for the given HLA allomorphs.

Overlapping sequences of vimentin bind to HLA-SE allomorphs
Citrullinated vimentin peptides stimulate a dose-dependent response in PBMCs from RA patients (6) as well as CD4 ϩ T cells isolated from immunized HLA-DR4 transgenic mice (25).   (Table 1 and Fig. 2C). These differing binding preferences suggest that the HLA-DR ␤-chain polymorphism can impact antigen hierarchies in disease development.

Structural comparison of HLA-DR4 with RA-associated autoantigens
Next, we ascertained whether the presence of multiple citrullinated residues in a given epitope caused a register shift when bound to a given HLA-DRB1 allomorph and how the

Citrulline in position P2 and P3 has minimal impact on peptide binding
The presence of a citrulline in the P2 position does not affect peptide affinity for the HLA-SE tested (Table 1 and Fig. 2). Indeed, the P2 pocket of the HLA-DRB1*04:01 could accommodate either arginine or citrulline as reflected in the fluorescence polarization data (Table 1 and Fig. 2). To address how an epitope with a P2-Cit could be accommodated, we determined the crystal structure of the HLA-DRB1*04:01 in complex with CII-1240Cit(1237-1249) that has a P2-Cit (Fig. 4F). The refined structure was very similar to that of HLA-DRB1*04:01 in complex with native CII(1237-1249) peptide (32). The bound CII peptide is extended across the peptide-binding site in a linear manner with Met-1239 occupying the P1 non-polar pocket, Asp-1242 in the P4 pocket, and Ala-1244 and Gly-1247 in the P6 and P9 pockets, respectively. The side chain of the P2-Cit in peptide CII-1240Cit(1237-1249), like the P2-Arg, projects away from the peptide-binding site of HLA-DRB1*04:01 and is freely accessible for TCR recognition.
Collectively, citrulline in position P2 or P3 in an epitope has minimal influence on peptide affinity for HLA-SE as the side chains of these residues project outward from the peptidebinding groove, and thus both arginine and citrulline can be accommodated without changing the overall peptide affinity.

Presentation of peptides with amino acid residues other than citrulline in the P4 pocket of HLA-SE
The SE motif creates an electropositive P4 pocket encompassing a positively charged residue at position 71␤. Consistent with our previous findings, these SE allomorphs allowed binding of autoantigen-derived peptides with polar or acidic residues at P4 but disallowed peptides with a P4 arginine due to electrostatic repulsion (24). In the HLA-DRB1*04:01-CII-

HLA DRB1 polymorphism and rheumatoid arthritis
1240Cit(1237-1249) complex, the side chain of the negatively charged P4-Asp, a preferred P4-anchor residue in the natural repertoire, is oriented upward similar to P4-Cit and forms a salt bridge with Lys-71␤ in HLA-DRB1*04:01 (Fig. 4F). In the HLA-DRB1*04:04 crystal structures in complex with either the native or citrullinated histone2B(68 -82), Ile was placed in the P4 pocket of the HLA-DRB1*04:04 molecule. Subsequently, the Arg-71␤ in the HLA-DRB1*04:04 becomes available to form a salt bridge with Asp-28␤ and water-mediated hydrogen bond with the P7-Glu of the peptide (Fig. 4, G and H). Alignment of C␣ backbone of the peptides bound to HLA-SE allomorphs highlights the conserved feature of citrulline and other non-positively charged residues. Thus, there is sufficient plasticity within the P4 pocket to accommodate differing amino acids.

HLA polymorphism dictates peptide-binding affinity
The crystal structures of the HLA-DR4 -peptide complexes show that the peptides bind as a straight extended chain within the binding groove with no extensive conformational changes in the HLA-SE molecule itself. Three side chains of the selfpeptides are accommodated by polymorphic P1, P4, and P9 pockets in the HLA-DRB1*04:01/*04:04/*04:05-binding site. These pockets determine the peptide specificity of the HLA-DRB1*04:01/*04:04/*04:05.

Discussion
Although HLA-DR4 molecules can present a range of citrullinated self-peptides that are associated with RA, the binding hierarchy of such peptides against a given HLA-SE molecule was unclear. A PTM peptide may generate a novel T cell epitope in two ways: (i) increased MHC class II binding affinity by modifying a key anchor residue, or (ii) modification of a non-anchor but solvent-exposed residue for TCR contact (34). A recent study on RA-associated HLA-DRB1*14:02 that encodes two polymorphic residues at the base of the P4 pocket (V11S and H13S) showed that, in contrast to DRB*04:01, both citrullinated and non-citrullinated peptides can be accommodated within the HLA-DRB1*14:02 molecule and stimulate a T cell response among Indigenous Native American RA patients (21). Here, we addressed the impact of HLA polymorphism and peptide citrullination on the HLA-SE binding hierarchy of self-peptides implicated in RA.
The shared epitope residues are critical in selecting specific amino acids at position P4 of peptide that will bind to DR4 as reflected in the fluorescent polarization assay carried out across 34 RA-related self-peptides. In general, P4-Cit increased peptide affinity for the majority of the citrullinated peptides for HLA-DRB1*04:01/*04:04/*04:05 and is accommodated in a highly conserved manner in the P4 pocket of the HLA-SE as shown in the series of crystal structures solved here and in ear-

HLA DRB1 polymorphism and rheumatoid arthritis
lier studies (11,24). Our data also showed that not all HLA allomorphs bind to citrullinated antigens with the same affinity and that citrulline outside of the P4 pocket has minimal impact on the peptide-HLA affinity compared with their native arginine counterpart.
The HLA-DRB1*04:01/*04:04/*04:05 bound to an overlapping but not an identical set of citrullinated peptides. This is most likely attributable to polymorphic residues within the peptide-binding cleft of these HLA-SE allomorphs and is also reflected in the selection of naturally presented peptides by these different HLA-DR4 allomorphs. Moreover, the majority of these self-antigens were identified in studies utilizing either HLA-DRB1*04:01 RA patients PBMC or DR4/IE transgenic mice (Table 1), which may explain why certain citrullinated self-peptides bound better to HLA-DRB1*04:01 but not HLA-DRB1*04:04/*04:05 (35). Thus, there is a need for further epitope discovery in the context of these other HLA-DR4-SE allomorphs. It also appears that each HLA allomorph has its own binding hierarchy for these RA-associated T cell epitopes. Whether the hierarchy of peptides reflects the immunodominant epitopes responsible for RA disease development in patients carrying specific RA-susceptible alleles will depend on T cell studies in vivo and structural analysis with the TCR. There are some parallels with other autoimmune-like diseases, such as celiac disease, where PTM peptides could bind to HLA allomorphs with higher affinity and subsequently form more stable peptide-MHC complexes and prolong antigen presentation to the autoreactive T cells at the site of inflammation (36 -39). Further studies aimed at determining whether a correlation between the rank order of RA-associated epitopes and disease pathogenesis exists will help to identify key autoantigens involved in RA development.

Peptides
Panels of 13-15-mer peptides with overlapping sequences spanning sections of the vimentin, aggrecan, fibrinogen, cartilage-intermediate proteins (CILP), collagen type II, and LL37 were synthesized by GL Biochem (Shanghai, China). Arginine in the peptides was systematically replaced with citrulline where it was appropriate (Table 1). For competition binding assay, the N-terminally acetylated HA(306 -318) analog Ac-PRFVK(Tamra)QNTLRLAT was labeled with TAMRA through the primary amine Lys-310 (GL Biochem, Shanghai, China). This peptide was shown to bind to HLA-DRB1* alleles in a fluorescent polarization assay (40). Lysine to arginine modification at positions 307 and 315 was made to the peptide to limit fluorescent labeling occurring at Lys-310 only. Each peptide was dissolved in water, TBS (10 mM Tris-Cl, pH 8, 150 mM NaCl), or DMSO at 1 mM and subsequently diluted as needed. Peptide concentration was determined by 215 nM peptide carbon backbone measurements.

Fluorescence polarization assay
Various concentrations of each test peptide (range from 500 to 0 M) were incubated in competition with 20 nM fluorescent reference peptides to bind with 100 nM recombinant HLA-DR protein in the presence of 20 nM HLA-DM in Assay Buffer (100 mM trisodium citrate, pH 5.4, 50 mM NaCl, 5 mM EDTA), as described previously (40). Fluorescent polarization was measured after 24 or 72 h of incubation at 37°C using PHERAstar microplate reader (BMG LABTECH). Peptide-binding curves were simulated by non-linear regression with Prism software (Version 7.01, GraphPad Software Inc.) using a sigmoidal dose-response curve. IC 50 binding values were calculated as the peptide concentration needed for 50% inhibition of reference peptide binding. Relative binding values (%) were calculated as the IC 50 value of the substituted peptide divided by the IC 50 value of the non-substituted peptide at a given concentration (ϫ100).

Peptide loading of HLA-DR4
HLA-DRB1*-Strep-CLIP was incubated with factor Xa (New England Biolabs) in the presence of 2 mM CaCl 2 for 6 h at room temperature to remove the covalent linked Strep-CLIP peptide. This invariant peptide was exchanged with test peptides in solution at 20-fold molar excess, catalyzed by HLA-DM at 1:5 DM/HLA-DRB1* ratio. The exchange was done between 24 and 72 h at 37°C in 50 mM tri-sodium citrate, pH 5.4, and 5 mM EDTA. The peptide-loaded HLA-DRB1* was separated from unloaded HLA-DRB1*-CLIP and HLA-DM using StrepTactin-Sepharose (IBA, Göttingen, Germany). The unbound fraction, containing HLA-DRB1* loaded with test peptides, was concentrated for structural studies.

Protein crystallization and structural determination
For crystallization, HLA-DRB1* was buffer-exchanged to 25 mM Tris, pH 7.6, 50 mM NaCl, and 2 mM CaCl 2 . The Fos/Jun zippers were removed by an overnight incubation with enterokinase (New England Biolabs) at room temperature. Subsequently, the extracellular domain of the heterodimer HLA-DRB1* loaded with the peptide of interest was purified using anion-exchange chromatography (HitrapQ, GE Healthcare). HLA-DRB1* in complex with peptides was crystallized in the presence of 16 -30% PEG3350, 200 mM potassium nitrate, 100 mM Bistris propane, pH 7.6. Plate-like crystals were exposed to cryoprotectant, 16% ethylene glycol for 30 s before being flash-HLA DRB1 polymorphism and rheumatoid arthritis cooled in liquid nitrogen. A Morpheus crystallization condition (42) containing 100 mM imidazole/MES, pH 6.5, 20 mM NPS mix, 12.5% MPD, 8 -12% PEG3350, 12.5% PEG1000 were used for the HLA-DRB1*-peptide complex that did not crystallize in the previous condition described. Data sets were collected on the MX1 or MX2 beamline of the Australian Synchrotron.
Data were integrated with XDS or iMOSFLM and scaled and merged in Aimless. Phases were obtained using molecular replacement in PhaserMR, CCP4 suite (43). The PDB entry 4MDI (chain A and B, without glycans or peptides) was used as search model for HLA-DRB1* for molecular replacement. The structures were built and refined in Coot and REFMAC/PHENIX (44). Final model was validated in Molprobity (45).

Repertoire analysis of HLA-DR4 allomorphs
T2-HLA-DRB1*04:05 cells expressing HLA-DM were generated via retroviral transduction of the parental T2 line as described previously (46). Cells were expanded in RPMI 1640 medium, 10% FCS and pellets of 10 9 cells snap-frozen in liquid nitrogen. Cells were ground under cryogenic conditions and resuspended in lysis buffer (0.5% IGEPAL, 50 mM Tris, pH 8, 150 mM NaCl, and protease inhibitors) as described previously (47,48). Cleared lysates were passed over a protein A pre-column followed by an affinity column cross-linked with a monoclonal antibody specific for HLA-DR (LB3.1). Peptide-MHC complexes were eluted from the column by acidification with 10% acetic acid. Peptides were isolated using reversed-phase HPLC (Chromolith C18 Speed Rod, Merck) on an Akta Ettan HPLC system (GE Healthcare). Fractions were concentrated and run on an AB SCIEX 5600ϩ triple TOF high-resolution mass spectrometer as described (27). Acquired data were searched against the human proteome (Uniprot/SwissProt version 2016_12) using ProteinPilot TM software version 5 (AB SCIEX). The resulting peptide identities were subject to strict bioinformatic criteria, including the use of a decoy database to calculate the false discovery rate (FDR). A 5% FDR cutoff was applied, and the filtered dataset was further analyzed manually to exclude redundant peptides and known contaminants. To generate motifs, the minimal core sequences found within nested sets were extracted, and the resulting list of peptides were aligned using MEME, where motif width was set to 9 -15 and motif distribution set to "one per sequence" (49). Peptides derived from HLA or immunoglobulin molecules were not included in the final motif analysis. Motifs were submitted to iceLogo for visualization using the frequencies of amino acids in the human proteome as a reference set (50). Comparison was made to previously published HLA-DR*04:01 and 04:04 datasets using standard statistical tools (24).