Arthritogenic peptide binding to DRB1*01 alleles correlates with susceptibility to rheumatoid arthritis

Journal of Autoimmunity 72 (2016) 25e32

Contents lists available at ScienceDirect

Journal of Autoimmunity journal homepage: www.elsevier.com/locate/jautimm

Arthritogenic peptide binding to DRB1*01 alleles correlates with susceptibility to rheumatoid arthritis Christina L. Roark a, *, Kirsten M. Anderson a, b, Michael T. Aubrey a, Edward F. Rosloniec c, Brian M. Freed a, b a b c

ClinImmune Labs, 12705 E. Montview Blvd, Suite 250, Aurora, CO 80045, USA Department of Immunology and Microbiology, University of Colorado AMC, 12800 E. 19th Ave., Aurora, CO 80045, USA Research Service, 151 Veterans Affairs Medical Center, 1030 Jefferson Avenue, Memphis, TN 38104, USA

a r t i c l e i n f o

a b s t r a c t

Article history: Received 29 March 2016 Received in revised form 20 April 2016 Accepted 24 April 2016 Available online 30 April 2016

Genetic susceptibility to rheumatoid arthritis (RA) is often defined by the presence of a shared epitope (QKRAA, QRRAA, or RRRAA) at positions 70e74 in HLA-DRb1. However, DRb1*01:01 and 01:02 contain the same QRRAA epitope, but differ considerably in their susceptibility to RA. The purpose of this study was to determine if this difference could be explained by their ability to bind three arthritogenic peptides that we have previously shown to bind to the archetypal RA-susceptible allele, DRb1*04:01, but not to the resistant DRb1*08:01 allele. Binding of type II collagen258-272, citrullinated and native vimentin66-78, and citrullinated and native a-enolase11-25 were measured on cell lines expressing either DRb1*01:01, *01:02 or *01:03 in association with DRa1*01:01. DRb1*01:01 and *01:02 both exhibited a 6.5-fold preference for citrullinated vimentin66-78 compared to native vimentin. However, DRb1*01:01 also exhibited a 1.7fold preference for citrullinated a-enolase11-25 and bound collagen258-272, while DRb1*01:02 bound neither of these peptides. Consistent with its known resistance to RA, DRb1*01:03 preferentially bound native vimentin66-78 and a
enolase11-25 over the citrullinated forms of these peptides, and also failed to bind collagen258
272. Site-directed mutagenesis was performed to determine which amino acid residues were responsible for the differences between these alleles. Mutating position 86 in DRb1*01:01 from glycine to the valine residue found in DRb1*01:02 eliminated binding of both citrullinated a-enolase11-25 and collagen258-272, thereby recapitulating the peptide-binding profile of DRb1*01:02. The difference in susceptibility to rheumatoid arthritis between DRb1*01:01 and *01:02 thus correlates with the effect of position 86 on the binding of these arthritogenic peptides. Consistent with their association with RA resistance, positions I67, D70 and E71 all contributed to the inability of DRb1*01:03 to bind these arthritogenic peptides. © 2016 Elsevier Ltd. All rights reserved.

Keywords: Histocompatibility Leukocyte Antigen (HLA) Rheumatoid arthritis Genetic susceptibility Peptide binding

1. Introduction Rheumatoid arthritis (RA) is a systemic autoimmune disease that is highly associated with the HLA-DRB1 locus. The predominant HLA-DRB1 alleles associated with RA are DRB1*04:01, *04:04 and *04:05, and to a lesser extent, DRB1*10:01 and *01:01 [1e8]. It has long been hypothesized that these disparate alleles confer susceptibility via a “shared epitope” comprised of the amino acids

* Corresponding author. E-mail addresses: [email protected] (C.L. Roark), kirsten.anderson@ ucdenver.edu (K.M. Anderson), [email protected] (M.T. Aubrey), [email protected] (E.F. Rosloniec), [email protected] (B.M. Freed). http://dx.doi.org/10.1016/j.jaut.2016.04.006 0896-8411/© 2016 Elsevier Ltd. All rights reserved.

QKRAA, QRRAA or RRRAA in positions 70e74 [3], which presumably mediate presentation of arthritogenic peptides to specific T cells that cause disease. Although the presence of this epitope is statistically associated with RA, multiple studies have suggested that the shared epitope hypothesis does not fully explain RA susceptibility [9e13]. For example, a distinct hierarchy exists between DRB1 alleles with respect to disease susceptibility [14]. Considerable differences can be observed in RA-susceptibility among patients with DRB1*04:01, *04:03, *04:04, *04:05 and *04:07 alleles, despite the presence of the shared epitope [15]. Furthermore, DRB1*01:01 is associated with RA but DRB1*01:02 is not, even though the QRRAA epitope is present in both. These observations suggest that the shared epitope by itself cannot explain disease susceptibility.

26

C.L. Roark et al. / Journal of Autoimmunity 72 (2016) 25e32

Investigations into the role of HLA in rheumatoid arthritis have been hampered by the lack of a functional assay to assess peptide binding to a large numbers of HLA protein alleles. In order to solve this problem, we recently developed an HLA multiplex assay which can be used to assess peptide binding to all common DRb1 protein alleles simultaneously [16]. We found that RA-susceptibility correlates with the preferential binding of citrullinated peptides compared to their native forms, while resistance is associated with a preference for native peptides. We also observed a strong correlation between RA susceptibility and the binding of an immunodominant collagen peptide [16]. In the present study, we examined the role of the QRRAA shared epitope in the context of DRb1*01:01 and *01:02, which differ only at amino acid positions 85 and 86. We have previously shown that a G86V mutation dramatically reduced the preference of DRb1*04:01 for citrullinated peptides and collagen [16], and we hypothesized that a similar effect could explain the difference in RA susceptibility between DRB1*01:01 and *01:02. We therefore analyzed the binding of three arthritogenic peptides to DRb1*01:01 and *01:02, as well as the RA-resistant DRb1*01:03. In order to dissect the individual effects of amino acid differences on arthritogenic peptide binding, site-directed mutagenesis was performed to change amino acids in DRb1*01:01 to corresponding residues present in either DRb1*01:02 or DRb1*01:03. 2. Material and methods

streptomycin, and 10% fetal bovine serum (FBS).

2.3. Mutagenesis The DRB1*01:01 plasmid was used as a template for sitedirected mutagenesis (Agilent) at amino acid positions 67, 70, 71, 85, and 86 to residues found in either DRb1*01:02 or *01:03 (Fig. 1). All cell lines were sorted for high expression of HLA-DR by flow cytometry (MFI ¼ 14,338 ± 678 SEM) and expression levels were verified for all 8 cell lines at the time of peptide binding using a pan-specific anti-HLA-DRa monoclonal antibody (clone LN3, eBioscience). The following single amino acid mutations were made to DRb1*01:01: L67I, Q70D, R71E, V85A, and G86V. In addition, the double mutant V85A þ G86V (DRb1*01:02) and triple mutant L67I þ Q70D þ R71E (DRb1*01:03) were generated to represent the other two DRb1*01 alleles. The primers utilized were as follows (forward primer listed, mutated nucleotides in bold): L67I (gc cag aag gac atc ctg gag cag agg cgg), Q70D (g gac ctc ctg gag gac agg cgg gcc gcg), R71E (g gac ctc ctg gag cag gag cgg gcc gcg), V85A (gc aga cac aac tac ggg gct ggt gag agc), G86V (gc aga cac aac tac ggg gtt gtt gag agc), V85A-G86V (DRB1*01:02) (gc aga cac aac tac ggg gct gtt gag agc), and L67I-Q70D-R71E (DRB1*01:03) (g gac atc ctg gag gac gag cgg gcc gcg). MSCV vectors were used to produce viral supernatant and T2 cell lines transduced as described above.

2.1. Peptide design and synthesis Citrullinated and native vimentin66-78 SAVRLRSSVPGVR, citrullinated and native a-enolase11-25 IFDSRGNPTVEVDLF [17], and unmodified type II collagen258-272 PGIAGFKGEQGPKGE [18,19], were synthesized with a biotinylated PEG3 linker on the N-terminus to >98% purity by Proimmune (Oxford, United Kingdom). The underscored R indicates the native arginine residues that were replaced with citrulline in the citrullinated form of the peptide. Biotinylated HA306-318 (PKYVKQNTLKLAT) was also synthesized as a control for HLA class II binding. Citrullinated and native vimentin6678 and citrullinated and native a-enolase11-25 peptides were chosen as arthritogenic peptides because of their ability to elicit T cell responses in RA patients and DR4-IE transgenic mice [17,20e22]. Non-modified collagen258-272 was also chosen because it is capable of inducing T cell activation and collagen-induced autoimmune arthritis in DR1-transgenic mice [23]. All peptides were resuspended in Dulbecco's phosphate buffered saline (DPBS) (Life Technologies) at 400 mM concentration and kept frozen at
20  C until use in peptide binding and T cell studies. 2.2. Cell lines RNA was isolated from an individual expressing DRB1*01:01 and cDNA cloned into a murine stem cell virus (MSCV) plasmid for retroviral transduction of the human class II negative T2 cell line. The HLA-DRa and HLA-DRb1 chains were individually packaged as retrovirus by transient transfection of Phoenix 293T cells with GFPþMSCV plasmids as previously described [24]. The retrovirus was used to transduce 105 T2 cells with DRA1*01:01 and DRB1*01:01 MSCV viral supernatant and sorted for high expression of HLA-DRþ/GFPþ three days post transduction. Post-sort, RNA was isolated from each cell line to verify the HLA sequences for both wild type and mutated HLA alleles by Sanger sequencing. The generation of T cell hybridomas restricted to DRb1*01:01 and specific for human type II collagen was previously reported [19,25]. All cell lines were grown in IMDM-GlutaMAX (Life Technologies) supplemented with sodium pyruvate, thio-penicillin/

2.4. Peptide binding to cells T2 cell lines expressing HLA-DRab1 were harvested in log-phase growth, washed with Dulbecco's PBS (DPBS), and resuspended in media (IMDM-GlutaMAX, 10% FBS, thio-pen/strep and sodium pyruvate) at 4  106 cells/mL. In a 96-well round-bottom plate, 50 mL of resuspended cells, 50 mL of 400 mM biotinylated stock peptide and 100 mL DPBS were combined. Negative control wells contained 50 mL resuspended cells and 150 mL DPBS. Plates were incubated overnight at 37  C. Plates were washed twice with DPBS to remove unbound peptide, then resuspended in 100 mL 1X Zombie Aqua (BD Biosciences) for 15 min at room temperature. Cells were lightly fixed for five minutes in 1% formaldehyde in DPBS to prevent loss of peptide from the cell surface. To detect peptide binding, 1X PElabeled streptavidin (One Lambda) was added for 30 min at 4  C. Prior to acquisition on the LSRII flow cytometer (BD Biosciences), cells were again lightly fixed. Data were analyzed by FlowJo Version X (Tree Star) and the average MFI ± SEM for 3 independent experiments was determined. PE-streptavidin staining of transfected T2 cell lines in the absence of peptide (89 ± 25 MFI) or peptide binding on T2 cells transfected with only the HLA-DRa chain (35 ± 35 MFI) were used to determine background levels. A ratio paired Student t-test was performed for statistical comparison of citrullinated versus native peptide binding for each DRb1 allele using GraphPad Prism software version 6.0 (Graph Pad).

70

80

90

DRB1*01:01 WNSQKDLLEQ RRAAVDTYCR HNYGVGESFT DRB1*01:02 ---------- ---------- ----AV---DRB1*01:03 ------I--D E--------- ---------Fig. 1. HLA sequence alignment for RA susceptible (DRb1*01:01), neutral (DRb1*01:02) and resistant (DRb1*01:03) alleles for amino acid positions 61e90. There are only two amino acid differences between DRb1*01:01 and *01:02 and three amino acid differences between DRb1*01:01 and *01:03. These differences are found in the peptide binding region.

C.L. Roark et al. / Journal of Autoimmunity 72 (2016) 25e32

2.5. T cell hybridoma assays DRb1*01:01-restricted, collagen-specific T cell hybridomas (105) and wild type or mutated HLA-DRb1*01:01-expressing T2 cell lines (105) were co-cultured overnight with various concentrations of collagen258-272. Supernatants were harvested the following day and IL-2 production measured using the murine IL-2 Ready-Set-Go ELISA kit (eBioscience). Data shown are the mean ± SEM. 3. Results 3.1. Comparison of arthritogenic peptide binding to DRb1*01:01, DRb1*01:02 and DRb1*01:03 DRb1*01:01 is associated with susceptibility to RA, while *01:02 is neutral and *01:03 is associated with resistance [14,15]. DRb1*01:01 and *01:02 differ only at positions 85 and 86 in pocket 1 of the peptide-binding groove (Fig. 1). DRb1*01:03 also shares extensive sequence homology with *01:01, differing only at positions 67, 70 and 71 in pocket 4. We hypothesized that these amino acid differences affect RA susceptibility by their effects on the binding of arthritogenic peptides. To test our hypothesis, MHCdeficient T2 cells were retrovirally transduced to express DRa1 with either DRb1*01:01, *01:02 or *01:03. A strong HLA-DR binding peptide, influenza HA306-318 was used as a reference peptide on each cell line to confirm that the HLA-DRab1 protein alleles were functional. Binding of citrullinated and native vimentin66-78, citrullinated and native a-enolase11-25 and collagen258-272 were then measured on wild type and mutant HLA-expressing cell lines using flow cytometry. As shown in Fig. 2, DRb1*01:01 demonstrated a 6.5-fold preference for citrullinated vimentin66-78 (598 ± 215 MFI) and a 1.7-fold preference for citrullinated a-enolase11-25 (141 ± 33 MFI) compared

27

to the native forms of these peptides (93 ± 17 and 82 ± 10 MFI, respectively). DRb1*01:01 bound collagen258-272 2.4-fold over background (184 ± 53 MFI). DRb1*01:02 also exhibited a 6.3-fold preference for citrullinated vimentin66-78 (1048 ± 346 MFI) compared to the native vimentin (166 ± 29 MFI), but did not bind either citrullinated a
enolase11-25 or collagen258-272. In contrast, DRb1*01:03 exhibited a distinct preference for the native forms of vimentin66-78 (1179 ± 357 MFI) and a-enolase11-25 (239 ± 76 MFI) compared to the citrullinated forms of these peptides (761 ± 260 and 137 ± 28, respectively) and a marked reduction in binding of collagen258-272 (115 ± 18 MFI) (Fig. 2). 3.2. Mutagenesis of crucial amino acid residues in DRb1*01:01 to those in DRb1*01:02 In order to assess the effect of individual amino acid differences between DRb1*01:01 and *01:02, we mutated the DRb1*01:01 valine at position 85 to alanine (V85A) or the glycine at position 86 to valine (G86V). The V85A mutation did not change the binding preference for citrullinated vimentin66-78 or citrullinated aenolase11-25, nor did it affect binding of collagen258-272 (Fig. 3). In contrast, the G86V mutation reduced the preference for citrullinated vimentin66-78 from 6.5-fold to 5-fold by increasing binding of native peptide, and eliminated binding of citrullinated a-enolase1125 and collagen258-272 (Fig. 3). 3.3. Mutagenesis of crucial amino acids between DRb1*01:01 and DRb1*01:03 DRb1*01:03 differs from *01:01 only at positions 67, 70, and 71 (Fig. 1), but is highly resistant to RA. To determine the contribution of each amino acid difference, site-directed mutagenesis was performed to generate the cell lines expressing L67I, Q70D or R71E

Fig. 2. Arthritogenic peptide binding to DRb1*01:01, *01:02, and *01:03. Binding of biotinylated citrullinated and native vimentin66-78, citrullinated and native a-enolase11-25, and collagen258-272 was measured on T2 cells transduced with HLA-DRb1*01:01, *01:02, or *01:03. Influenza HA306-318 binding was measured as a reference control. The data represent the MFI ± SEM of three independent experiments. A ratio paired t-test was used to measure statistical differences between citrullinated and native peptide binding for each DRb1 allele. *p < 0.05, **p < 0.01.

28

C.L. Roark et al. / Journal of Autoimmunity 72 (2016) 25e32

Fig. 3. The binding preference for arthritogenic peptides is reduced by mutagenesis of DRb1*01:01 susceptible amino acids to amino acids found in *01:02. Citrullinated peptide in dark gray, native peptide in light gray, and negative control (SA-PE alone) in white. Columns indicate peptide, rows indicate DRb1 allele or mutation. Data represents the average MFI of 3 independent experiments. For vimentin66-78 and a-enolase11-25, the MFI ratio of binding of the citrullinated peptide over the native is shown in the bottom right corner. For collagen258-272 and HA306-318, the ratio of peptide binding to the negative control (no peptide) is shown in the bottom right corner.

mutants in DRb1*01:01. The L67I mutation reduced the binding preference for citrullinated vimentin66-78 from 6.5-fold to 2-fold and virtually eliminated the preference for citrullinated aenolase11-25. Collagen258-272 binding was also reduced to background levels (Fig. 4). The Q70D mutation increased the binding of the native forms of vimentin66-78 (454 ± 150 MFI) and a-enolase1125 (147 ± 38 MFI) without affecting binding of the citrullinated forms (458 ± 166 and 137 ± 32 MFI, respectively). The net result of the Q70D mutation was a loss of preference for both citrullinated vimentin66-78 and a-enolase11-25 (ratios equal to 1.0). However, the Q70D mutation did not affect binding of collagen258-272 or HA306318 . R71E by itself dramatically increased the preference for the

native peptides over the citrullinated peptides, but also increased collagen258-272 and HA306-318 binding (Fig. 3). Thus, the lack of preference of DRb1*01:03 for arthritogenic peptides appears to be due to the combined effects of I67, D70 and E71. 3.4. The role of individual amino acids in presentation of collagen to DRb1*01:01 restricted, collagen-specific T cell hybridomas In order to verify that the peptide binding measured on the transfected cells lines represented functional peptide/MHC complexes, we quantitated the effects of the individual point mutations on T cell responses by four DRb1*01:01-restricted, collagen-specific

C.L. Roark et al. / Journal of Autoimmunity 72 (2016) 25e32

29

Fig. 4. The binding preference for arthritogenic peptides is also reduced by mutagenesis of DRb1*01:01 susceptible amino acids to resistant amino acids found in *01:03. Citrullinated peptides are shown in dark gray, native peptides in light gray, and the negative controls (SA-PE alone) in white. Columns indicate peptide, rows indicate HLA type or DRb1*01:01 mutation. Data represents the average MFI of 3 independent experiments. For vimentin66-78 and a-enolase11-25, the MFI ratio of binding of the citrullinated peptide over the native is shown in the bottom right corner. For collagen258-272 and HA306-318, the ratio represents binding of the peptide over the negative control (SA-PE alone).

T cell hybridomas. Only collagen-specific hyrbridomas were used

because we do not yet have T cell hybridomas specific for

C.L. Roark et al. / Journal of Autoimmunity 72 (2016) 25e32

citrullinated peptides. As would be expected from MHC-restricted T cells, collagen258-272 elicited a strong IL-2 response from all four T cell hybridomas when presented by DRb1*01:01, but not when presented by DRb1*01:02 or *01:03 (Fig. 5A and B). The G86V mutation reduced the T cell responses to collagen258-272 in all four hybridomas by >99%, commensurate with the reduction in collagen258-272 binding (Fig. 3). In contrast, the V85A mutation did not affect T cell responses to collagen258-272 (Fig. 5A), corresponding with its lack of effect on peptide binding (Fig. 3). These results suggest that G86 plays the critical role in the increased arthritogenic potential of DRb1*01:01 as compared to *01:02. The L67I mutation reduced the response of all four T cell hybridomas by >99% (Fig. 5B), even though collagen258-272 binding was only reduced by 57% (Fig. 4). The Q70D mutation exhibited a wide range of effects on the responses of T cell hybridomas (10%e 90% inhibition). These differences are likely due to the differential role of Q70 in binding the four different T cell receptors expressed

102

101 3

4

104

G86V

0

*01:02 1

2

3

G86V *01:02

101

102

10

1

2

3

4

Q70D R71E L67I *01:03

DR1.CII.E174

G86V *01:02

101 3

4

104

102

2

3

*01:01

0

V85A *01:01

1

2

Q70D R71E L67I *01:03

103

4

103

0

1

1

DR1.CII.E174

104

102

104

IL-2 (pg/ml)

102

3

4

DR1.CII.22.7

10

2

3

*01:01

0

V85A *01:01

1

2

R71E L67I *01:03

103

4

3

0

1

101

DR1.CII.22.7

104

IL-2 (pg/ml)

V85A *01:01

102

101

IL-2 (pg/ml)

101

103

0

*01:01 Q70D

102

DR1.CII.21

DR1.CII.21

104

IL-2 (pg/ml)

2

103

G86V *01:02

4

CII peptide ( g/ml)

IL-2 (pg/ml)

IL-2 (pg/ml)

V85A *01:01

1

DR1.CII.16 104

103

0

We recently reported that susceptibility to rheumatoid arthritis correlates strongly with a preference of HLA-DRb1 protein alleles for citrullinated arthritogenic peptides [16]. Using site-directed mutagenesis, we demonstrated that Q70 and K71 were largely

B

DR1.CII.16

104

4. Discussion

IL-2 (pg/ml)

A

on these cell lines. The R71E mutation abolished T cell responses to all four T cell hybridomas, despite enhancing collagen258-272 binding by 1.8-fold (Fig. 4). These results suggest that R71 is a critical contact residue for all four T cell receptors, although it is possible that the peptide conformation may have been altered by this mutation. However, the fact that D70 and E71 did not block collagen258-272 binding suggests that other T cells could react to collagen presented in this manner. The inability of DRb1*01:03 to present collagen258-272 appears to be solely due to isoleucine in position 67.

IL-2 (pg/ml)

30

*01:01

103

Q70D

102

101 0

1

2

3

4

R71E L67I *01:03

CII peptide ( g/ml)

Fig. 5. The IL-2 response of DRb1*01:01-restricted collagen258-272-specific T cell hybridomas to collagen258-272 presented by wild type DRb1*01:01 or mutated cell lines expressing amino acid differences found in DRb1*01:02 (A) or DRb1*01:03 (B). T cell hybridomas were stimulated overnight with collagen258-272 peptide presented by wild type or mutant HLA-expressing T2 cells. IL-2 production was measured by ELISA. N ¼ 3 independent experiments. Results depicted represent the mean ± SEM.

C.L. Roark et al. / Journal of Autoimmunity 72 (2016) 25e32

responsible for this phenomenon, but that G86 and A74 also contributed significantly. In contrast, we observed that D70, E71 and V86 promoted binding of arginine residues found in native peptides and correlated with resistance to RA. Since DRb1*01:01 and *01:02 differ only at positions 85 and 86, we postulated one or both of these residues would act in a similar manner and thereby reduce the specificity of DRb1*01:02 for citrullinated peptides and collagen. Consistent with this theory, DRb1*01:01 exhibited a significant binding preference for citrullinated vimentin66-78 and a-enolase1125 compared to the native forms of these peptides. However, the DRb1*01:01 preference for these citrullinated peptides was not as pronounced as previously shown for DRb1*04:01 [16], which correlates with the reduced association between DRB1*01:01 and RA. DRb1*01:02 exhibited preference for citrullinated vimentin66-78 similar to that of DRb1*01:01, but did not bind either citrullinated a-enolase11-25 or collagen258-272. The G86V mutation responsible for this phenomenon produced the exact same effect on peptide binding when introduced in the context of DRb1*04:01 [16]. These differences may account for the diminished association between DRB1*01:02 and RA, although this hypothesis needs to be tested with a larger set of arthritogenic peptides. Finally, DRb1*01:03 exhibited a clear preference for native peptides over the citrullinated forms and failed to bind collagen258-272, and appeared nearly identical to the arthritogenic peptide-binding profile of another resistant allele, DRb1*08:01 [16]. The role of position 86 in the binding of arthritogenic peptides is not clear. Position 86 is involved in the formation of pocket 1, and as native and citrullinated peptides do not differ in their P1 residues, one would expect that the G86V mutation would have little effect on the binding of these peptides. However, the G86V mutation eliminated binding of a-enolase11-25, which we have also seen in the context of DRb1*04:01 [16]. The G86V mutation also eliminated binding of collagen258-272 in the context of DRb1*01:01 and *04:01 [16], which corresponded to a loss of response by DRb1*1:01restricted collagen-specific T cell hybridomas and DRb1*04:01restricted hybridomas [16]. This phenomenon is likely to be dependent on the peptide anchor residues in these particular peptides rather than the citrulline in P4. However, the lack of association between DRB1*01:02 and RA, and the reduced association between DRB1*04 alleles that have V86 (e.g. DRB1*04:04 and *04:05) and RA, argues in favor of G86 playing a role in disease susceptibility. We also compared peptide binding and T cell responses between DRb1*01:01 and the resistance allele, DRb1*01:03, which differ by three amino acids (L67I, Q70D and R71E). The IDE epitope is also found in the RA-resistant allele DRb1*04:02. The effect of this epitope on the loss of binding preference for citrullinated vimentin66-78 and a-enolase11-25 and the loss of collagen258-272 binding was virtually identical in DRb1*01 alleles as we have previously seen in the DRb1*04 alleles [16]. The mechanism for the loss of preference for citrullinated peptides was the same in the DRb1*01 alleles as it was in the context of DRb1*04 alleles; the DE epitope in positions 70 and 71 promotes the binding of arginine residues found native peptides. The QK (DRb1*04:01) and QR (DRb1*01:01, *04:03, *04:04, *04:05) and RR (*10:01) epitopes produce a positive charge in pocket 4 that repulses arginine residues in native vimentin66-78 and a-enolase11-25 but allows binding of the neutral citrulline residue. The binding of these three arthritogenic peptides to DRb1 protein alleles exhibits a remarkable correlation with susceptibility to RA. DRb1*01:01 exhibited a 6.5-fold preference for citrullinated vimentin66-78 and a 1.7-fold preference for citrullinated aenolase11-25, compared to a 12.7-fold preference for both peptides by DRb1*04:01 [16]. This weaker preference for citrullinated

31

peptides correlates with a lower susceptibility to RA by DRb1*01:01 as compared to DRb1*04:01 [8,15]. DRb1 alleles that bind citrullinated and native forms of these peptides equally (e.g. DRb1*04:02) are neutral with respect to disease susceptibility, and alleles that preferentially bind native peptides (e.g. DRb1*01:03 and 08:01) are highly resistant to RA. Furthermore, the fact that DRb1*01:02 retains a preference for citrullinated vimentin66-78 (but not citrullinated a-enolase11-25) while DRb1*04:02 does not may explain why *01:02 is slightly more susceptible than *04:02 [15]. Funding and conflict of interests No financial support or other benefits from commercial sources or any other financial interests from any of the authors exist that could create a potential conflict of interest or the appearance of a conflict of interest with regard to this work. Acknowledgements All authors were involved in drafting the article or revising it for important intellectual content. All authors approved the final version to be published. Study conception and design. Roark, Anderson, Aubrey, and Freed. Acquisition of data. Roark, Anderson, and Rosloniec. Analysis and interpretation of data. Roark, Anderson, Aubrey and Freed. References [1] D. Carthy, W. Ollier, C. Papasteriades, H. Pappas, W. Thomson, A shared HLADRB1 sequence confers RA susceptibility in Greeks, Eur. J. Immunogenet. Off. J. Br. Soc. Histocompat. Immunogenet. 20 (1993) 391e398. [2] S.H. Chan, Y.N. Lin, G.B. Wee, W.H. Koh, M.L. Boey, HLA class 2 genes in Singaporean Chinese rheumatoid arthritis, Br. J. Rheumatol. 33 (1994) 713e717. [3] P.K. Gregersen, J. Silver, R.J. Winchester, The shared epitope hypothesis. An approach to understanding the molecular genetics of susceptibility to rheumatoid arthritis, Arthritis Rheum. 30 (1987) 1205e1213. [4] G.T. Nepom, P. Byers, C. Seyfried, L.A. Healey, K.R. Wilske, D. Stage, et al., HLA genes associated with rheumatoid arthritis. Identification of susceptibility alleles using specific oligonucleotide probes, Arthritis Rheum. 32 (1989) 15e21. [5] R. Ploski, O.J. Mellbye, K.S. Ronningen, O. Forre, E. Thorsby, Seronegative and weakly seropositive rheumatoid arthritis differ from clearly seropositive rheumatoid arthritis in HLA class II associations, J. Rheum. 21 (1994) 1397e1402. [6] P. Stastny, E.J. Ball, P.J. Dry, G. Nunez, The human immune response region (HLA-D) and disease susceptibility, Immunol. Rev. 70 (1983) 113e153. [7] P. Stastny, E.J. Ball, M.A. Khan, N.J. Olsen, T. Pincus, X. Gao, HLA-DR4 and other genetic markers in rheumatoid arthritis, Br. J. Rheum. 27 (Suppl. 2) (1988) 132e138. [8] D. van der Woude, B.A. Lie, E. Lundstrom, A. Balsa, A.L. Feitsma, J.J. HouwingDuistermaat, et al., Protection against anti-citrullinated protein antibodypositive rheumatoid arthritis is predominantly associated with HLADRB1*1301: a meta-analysis of HLA-DRB1 associations with anticitrullinated protein antibody-positive and anti-citrullinated protein antibody-negative rheumatoid arthritis in four European populations, Arthritis Rheum. 62 (2010) 1236e1245. [9] S.T. du Montcel, L. Michou, E. Petit-Teixeira, J. Osorio, I. Lemaire, S. Lasbleiz, et al., New classification of HLA-DRB1 alleles supports the shared epitope hypothesis of rheumatoid arthritis susceptibility, Arthritis Rheum. 52 (2005) 1063e1068. [10] E. Genin, M.C. Babron, M.F. McDermott, B. Mulcahy, F. Waldron-Lynch, C. Adams, et al., Modelling the major histocompatibility complex susceptibility to RA using the MASC method, Gen. Epidemiol. 15 (1998) 419e430. [11] J.M. Meyer, J. Han, R. Singh, G. Moxley, Sex influences on the penetrance of HLA shared-epitope genotypes for rheumatoid arthritis, Am. J. Hum. Gen. 58 (1996) 371e383. [12] A.S. Rigby, A.J. MacGregor, G. Thomson, HLA haplotype sharing in rheumatoid arthritis sibships: risk estimates subdivided by proband genotype, Gen. Epidemiol. 15 (1998) 403e418. [13] S. Tezenas du Montcel, D. Reviron, E. Genin, J. Roudier, P. Mercier, F. ClergetDarpoux, Modeling the HLA component in rheumatoid arthritis: sensitivity to DRB1 allele frequencies, Gen. Epidemiol. 19 (2000) 422e428. [14] N. de Vries, H. Tijssen, P.L. van Riel, L.B. van de Putte, Reshaping the shared

32

[15]

[16]

[17]

[18]

[19]

[20]

C.L. Roark et al. / Journal of Autoimmunity 72 (2016) 25e32 epitope hypothesis: HLA-associated risk for rheumatoid arthritis is encoded by amino acid substitutions at positions 67-74 of the HLA-DRB1 molecule, Arthritis Rheum. 46 (2002) 921e928. S. Raychaudhuri, C. Sandor, E.A. Stahl, J. Freudenberg, H.S. Lee, X. Jia, et al., Five amino acids in three HLA proteins explain most of the association between MHC and seropositive rheumatoid arthritis, Nat. Gen. 44 (2012) 291e296. K.M. Anderson, C.L. Roark, M. Portas, M.T. Aubrey, E.F. Rosloniec, B.M. Freed, A molecular analysis of the shared epitope hypothesis binding of arthritogenic peptides to DRbeta1*04 alleles, Arthritis Rheumatol. (2016), http://dx.doi.org/ 10.1002/art.39636 [Epub ahead of print]. PMID: 26866513. E.A. James, M. Rieck, J. Pieper, J.A. Gebe, B.B. Yue, M. Tatum, et al., Citrullinespecific Th1 cells are increased in rheumatoid arthritis and their frequency is influenced by disease duration and therapy, Arthritis Rheumatol. 66 (2014) 1712e1722. H.Y. Kim, W.U. Kim, M.L. Cho, S.K. Lee, J. Youn, S.I. Kim, et al., Enhanced T cell proliferative response to type II collagen and synthetic peptide CII (255-274) in patients with rheumatoid arthritis, Arthritis Rheum. 42 (1999) 2085e2093. E.F. Rosloniec, K.B. Whittington, D.M. Zaller, A.H. Kang, HLA-DR1 (DRB1*0101) and DR4 (DRB1*0401) use the same anchor residues for binding an immunodominant peptide derived from human type II collagen, J. Immunol. 168 (2002) 253e259. J.A. Hill, S. Southwood, A. Sette, A.M. Jevnikar, D.A. Bell, E. Cairns, Cutting edge:

[21]

[22]

[23]

[24]

[25]

the conversion of arginine to citrulline allows for a high-affinity peptide interaction with the rheumatoid arthritis-associated HLA-DRB1*0401 MHC class II molecule, J. Immunol. 171 (2003) 538e541. S.W. Scally, J. Petersen, S.C. Law, N.L. Dudek, H.J. Nel, K.L. Loh, et al., A molecular basis for the association of the HLA-DRB1 locus, citrullination, and rheumatoid arthritis, J. Exp. Med. 210 (2013) 2569e2582. O. Snir, M. Rieck, J.A. Gebe, B.B. Yue, C.A. Rawlings, G. Nepom, et al., Identification and functional characterization of T cells reactive to citrullinated vimentin in HLA-DRB1*0401-positive humanized mice and rheumatoid arthritis patients, Arthritis Rheum. 63 (2011) 2873e2883. E.F. Rosloniec, D.D. Brand, L.K. Myers, K.B. Whittington, M. Gumanovskaya, D.M. Zaller, et al., An HLA-DR1 transgene confers susceptibility to collageninduced arthritis elicited with human type II collagen, J. Exp. Med. 185 (1997) 1113e1122. N.A. Bowerman, M.T. Falta, D.G. Mack, J.W. Kappler, A.P. Fontenot, Mutagenesis of beryllium-specific TCRs suggests an unusual binding topology for antigen recognition, J. Immunol. 187 (2011) 3694e3703. X. He, E.F. Rosloniec, L.K. Myers, W.L. McColgan 3rd, M. Gumanovskaya, A.H. Kang, et al., T cell receptors recognizing type II collagen in HLA-DRtransgenic mice characterized by highly restricted V beta usage, Arthritis Rheum. 50 (2004) 1996e2004.