Antibodies directed against citrullinated proteins (ACPAs) are highly specific for rheumatoid arthritis (RA). The production of ACPAs is most likely dependent on the presence of T cells, since ACPAs undergo isotype switching and are associated with the shared epitope (SE)–containing HLA–DRB1 alleles. Vimentin is a likely candidate protein for T cell recognition, since >90% of patients positive for ACPAs that are reactive with (peptides derived from) citrullinated vimentin carry SE-containing HLA–DRB1 alleles. The aim of this study was to identify citrullinated vimentin peptides that are presented to HLA–DRB1*0401–restricted T cells.
HLA–DR4–transgenic mice were immunized with all possible citrulline-containing peptides derived from vimentin, and T cell reactivity was analyzed. Peptides recognized in a citrulline-specific manner by T cells were selected and analyzed for their ability to be processed from the entire vimentin protein. A first inventory of the selected epitopes recognized by T cells was performed using peripheral blood mononuclear cells (PBMCs) from ACPA+, HLA–DR4+ patients with RA.
A citrulline-specific response was observed for 2 of the peptides analyzed in DR4-transgenic mice. These peptides were found to be naturally processed from the vimentin protein, since citrullinated vimentin was recognized by peptide-specific T cells. T cell reactivity against these peptides was also observed in cultures of PBMCs from RA patients.
This study identifies, for the first time, 2 naturally processed peptides from vimentin that are recognized by HLA–DRB1*0401–restricted T cells in a citrulline-specific manner. These peptides can be recognized by T cells in ACPA+, HLA–DR4+ patients with RA, as shown in a first inventory.
Rheumatoid arthritis (RA) is a chronic, systemic inflammatory autoimmune disease that is characterized by the presence of autoantibodies. Among the autoantibodies described in RA, antibodies directed against citrullinated proteins (ACPAs) are highly specific and predictive of the development of RA (1–3) and can be detected in ∼70–80% of patients with longstanding RA. Antigens recognized by ACPAs are present in the inflamed joint (4, 5). The production of ACPAs is most likely dependent on the presence of T cells, since ACPAs undergo isotype switching.
HLA class II alleles are the most important genetic risk factor for RA. Notably, HLA–DRB1 molecules sharing a common epitope, R(Q)K(R)RAA, at position 70–74 in the third hypervariable region of the DRB1 chain, the so-called shared epitope (SE), are associated with both susceptibility to and severity of RA (6–9). It has been shown that the SE-containing HLA–DRB1 alleles predispose to ACPA+ disease, but not to ACPA− RA (10), and that they are associated with the production of ACPAs (11). This latter evidence provides an additional indication for the existence of T cell responses underlying the production of ACPAs.
Several human proteins were found to be citrullinated in vivo, one of which is vimentin, also known as the Sa antigen (12–15). Citrullinated vimentin has been shown to be present in the synovial fluid of RA patients and to be recognized by ACPAs in ∼40% of RA patients (13, 16–18). Furthermore, we have shown that in >90% of patients with ACPA+ RA in whom a citrullinated peptide derived from human vimentin is recognized (19), and in 80% of patients with ACPA+ RA in whom the Sa antigen is recognized (Department of Rheumatology, Leiden University Medical Center: unpublished observations), at least 1 SE-containing HLA–DRB1 allele is present. These observations not only show that the SE-containing HLA–DRB1 alleles influence the specificity of the ACPA response, but also suggest that vimentin could be a protein involved in the recruitment of T cell help for ACPA-producing B cells.
In the present study, we examined whether CD4+ T cells that are specific for citrullinated peptides from the human vimentin protein, presented in the context of the most frequent SE-containing HLA–DRB1 allele, HLA-DRB1*0401, could be identified. To this end, we used an unbiased approach, by testing the capacity of all possible citrullinated peptides derived from human vimentin to induce T cell responses in DR4-transgenic mice. Epitope mapping in HLA-transgenic mice, both for class I HLA and for class II HLA, has been shown to be a reliable method for the identification of T cell epitopes that are also recognized by human T cells (20–24), and therefore we anticipated that the epitopes identified in HLA-transgenic mice would be prime candidates for recognition by human T cells. We identified 2 naturally processed peptides of citrullinated vimentin that were recognized by HLA–DRB1*0401–restricted T cells in a citrulline-specific manner. Reactivity against these 2 peptides was also observed in cultures of peripheral blood mononuclear cells (PBMCs) from ACPA+, HLA–DRB1*04+ patients with RA.
These studies identified, for the first time, human citrulline–specific T cell responses against naturally processed epitopes from an autoantigen present in the inflamed joint. These results thus provide a rationale for more comprehensive analyses in patients with RA.
MATERIALS AND METHODS
DR4-transgenic mice (HLA–DRB1*0401/HLA–DRA1*0101/human CD4–transgenic mice) lacking the endogenous class II major histocompatibility complex (MHC) were kindly provided by L. Fugger (25). The mice were bred in our in-house mouse facility at Leiden University Medical Center.
Patients and controls.
Blood samples were obtained from HLA–DRB1*04+ healthy human donors after the subjects had provided their informed consent, and PBMCs were isolated from the blood by Ficoll-Paque. Patients carrying at least 1 HLA–DRB1*04 allele and who were positive for ACPAs were recruited from the Leiden Early Arthritis Clinic (EAC) cohort (26). The characteristics of the 10 patients are shown in Table 1.
Peripheral blood mononuclear cells from 10 anti–citrullinated protein antibody–positive patients with rheumatoid arthritis were tested for the presence of T cell responses against the identified citrullinated vimentin epitopes. The age (in years), sex, and HLA–DRB1 alleles of each patient are shown. Among the patients tested for anti-Sa antibodies, 66% were positive.
The vimentin gene was amplified by polymerase chain reaction and cloned using Gateway technology (Invitrogen, San Diego, CA) in a bacterial expression vector containing an N-terminal histidine tag. The protein was overexpressed in Escherichia coli BL21(DE3) and purified by immobilized metal chelate affinity chromatography on nickel–nitrilotriacetic acid beads, as described previously (27). The protein was citrullinated in a 0.1M Tris, pH 7.6, solution by adding peptidyl arginine deiminase type 2 (20 units/ml) (from rabbit skeletal muscle; Sigma-Aldrich, St. Louis, MO) and 10 mM CaCl2 for 3 hours at 55°C.
The peptides were chemically synthesized in the peptide facility at Leiden University Medical Center and dissolved in phosphate buffered saline/0.05% DMSO. Every peptide was designed with a citrulline in the middle of the peptide, with a total length of 19 amino acids. In total, the vimentin protein contains 43 arginine residues, but 33 peptides were synthesized in citrullinated form (vim1–33) (Table 2), since some peptides contain 2 citrulline residues in close proximity. Peptides able to induce a T cell response in DR4-transgenic mice were also synthesized in noncitrullinated form. The citrullinated peptides were grouped in 6 pools of 5 peptides each and 1 pool of 3 peptides.
Table 2. Peptides synthesized from the vimentin protein*
Amino acid sequence
X = citrulline; R = arginine.
Immunization protocol and epitope mapping.
DR4-transgenic mice were injected with 100 μl of the peptide pools or individual peptides (100 μg/peptide) emulsified in Freund's complete adjuvant (CFA; Difco, Detroit, MI), administered subcutaneously in the base of the tail. On day 21, the mice received a booster injection with the same peptide pool or individual peptide (100 μg/peptide) emulsified in Freund's incomplete adjuvant (IFA; Difco), administered subcutaneously in the flank. On days 42–49 after the first immunization, spleen cells were isolated and restimulated once with the immunizing antigen (10 μg/ml peptide) at a density of 4 × 106 cells per well in 24-well plates in culture medium (Iscove's modified Dulbecco's medium [IMDM] with 8% fetal calf serum, penicillin, streptomycin, and 0.02 mM β-mercaptoethanol). Four days later, the cells were harvested with 2 mM EDTA and centrifuged on a Ficoll-Paque gradient. The cells were then rested for another 3 days at a density of 106 cells per well in the presence of 3 cU/ml recombinant interleukin-2 (rIL-2).
On day 7, the cells were harvested and tested at various cell concentrations in round-bottomed, 96-well plates. Irradiated spleen cells (100,000 per well) from a naive mouse were added to each well, along with the different peptide pools/peptides (10 μg/ml/peptide) or recombinant (citrullinated) vimentin (20 μg/ml). As a positive control, either rIL-2 (20 cU/ml) or phytohemagglutinin (1 μg/ml) was used. Every condition was tested in triplicate. Four days later, 3H-thymidine was added to the wells for 16 hours. The plates were harvested in a Tom-Tec Mach3 harvester (PerkinElmer, Groningen, The Netherlands) and the T cell count was determined using a 1450 Microbeta counter (PerkinElmer). To determine the restriction of the T cell response, blocking antibodies against HLA–DR (B8.11.2) (28) were used.
Enzyme-linked immunosorbent assay (ELISA).
Supernatants from the stimulated spleen cells were removed before the addition of 3H-thymidine, and the levels of interferon-γ (IFNγ) in these separate supernatants were measured using a standard sandwich ELISA. Rat anti-mouse coating and detection antibodies were purchased from BD PharMingen (San Diego, CA). Streptavidin–horseradish peroxidase (Sanquin, Amsterdam, The Netherlands) and avidin–biotinylated enzyme complex (Sigma-Aldrich) were used as the enzyme and substrate, respectively. Stimulation indices (SIs) were calculated by dividing the amount of IFNγ produced upon antigenic stimulation by the amount produced by unstimulated cells. The results are expressed as the mean ± SEM.
Intracellular cytokine staining.
PBMCs from healthy individuals or patients with RA from the Leiden EAC cohort were isolated, and 3 × 106 cells/well were cultured for 2 hours in 24-well plates with the different citrullinated peptides (10 μg/ml), or with Memory Mix (mix of Candida albicans [0.005%], tetanus toxoid [0.75 Lf/ml], and tuberculin purified protein derivative [5 μg/ml]) as a positive control. After removal of the antigen, the cells were cultured for 4 days in IMDM with 5% pooled human serum, penicillin, and streptomycin. On day 4, 2 × 106 autologous PBMCs were plated in 24-well plates, and nonadherent cells were removed after 2 hours of incubation. Cultured cells (1–2 × 106) were added to the adherent antigen-presenting cells and restimulated with antigen overnight. For the last 14 hours of stimulation, 3 μg/ml brefeldin A (Sigma-Aldrich) was added to the wells. The cells were then stained for the cell surface markers CD3, CD4, and CD45RA (for 20 minutes on ice), and intracellular IFNγ was stained using a Cytofix/Cytoperm fixation/permeabilization solution kit (BD Biosciences, Alphen aan de Rijn, The Netherlands) according to the manufacturer's instructions. Antibodies both for surface markers and for intracellular markers were purchased from BD Biosciences. Data acquisition and data analyses were performed on an LSRII with FACS DIVA software (BD Biosciences).
For statistical comparisons of the peptide and protein responses, paired t-tests were performed in GraphPad Prism, version 4.0 (GraphPad Software, San Diego, CA). P values less than 0.05, taking into account the 95% confidence intervals, were considered significant.
Peptide pools inducing antigen-specific T cell responses.
Based on several lines of evidence (12–17, 19), we hypothesized that vimentin represents a relevant candidate autoantigen recognized by HLA–SE–restricted T cells. To identify the vimentin epitopes recognized by T cells, we chose to utilize an unbiased approach in which all possible citrullinated peptides of human vimentin (Table 2) were analyzed for their ability to induce a T cell response in DR4-transgenic mice. To enable efficient analyses of the large number of peptides generated, we made a first selection of potential epitopes by immunizing DR4-transgenic mice with 7 peptide pools. Peptide pool–specific T cell responses were observed repeatedly in the bulk cultures obtained from mice immunized with peptide pools 1 and 7 (Figure 1A). No response against these peptide pools was observed when spleen cells from naive mice were tested (results not shown). These results indicate that the immunogenic citrullinated T cell epitopes were among the peptides contained in peptide pools 1 and 7.
Characterization of immunogenic peptides.
To identify the individual peptides responsible for the induction of T cell responses by the respective peptide pools, DR4-transgenic mice were immunized with peptide pool 1 or peptide pool 7, and the T cell reactivity to each peptide pool, as well as to the individual peptides from the pool, was analyzed. A dose-dependent T cell response of spleen cells from mice immunized with peptide pool 1 was observed after stimulation with peptide 5 (vim26–44) (Figure 1B). Likewise, peptide 31 (vim415–433) was consistently recognized by spleen cells from mice immunized with peptide pool 7 (Figure 1C). Taken together, these results indicate that citrullinated vimentin peptides 5 and 31 are immunogenic in HLA–DR4-transgenic mice.
Furthermore, we observed that the responses against peptides 5 and 31 were citrulline-specific and HLA–DR restricted. To determine whether the response against these 2 citrullinated vimentin peptides was citrulline-specific, mice were immunized with these peptides, and spleen cells from the immunized mice were restimulated in vitro with these same peptides. Subsequent proliferation of the spleen cells against either the citrullinated or the noncitrullinated form of the peptides was assessed. A dose-dependent T cell response was observed against citrullinated vim5, whereas no responses could be observed in response to the noncitrullinated counterpart (Figures 2A and B). On average, the SI of the cultures restimulated with citrullinated vim5 was ∼6 times higher than that in cultures stimulated with the noncitrullinated control peptide (Figure 2B, inset). Similar results were obtained in cultures restimulated with citrullinated vim31 (Figures 2C and D). In the latter case, the T cell response was, on average, 9 times higher upon stimulation with the citrullinated peptide as compared with that with the noncitrullinated peptide (Figure 2D, inset). No IFNγ was detected in spleen cells from sham-immunized mice (i.e., immunized with CFA/IFA without peptide) upon restimulation with either of the citrullinated peptides (results not shown). Moreover, we confirmed by intracellular cytokine staining that CD4+ T cells were producing IFNγ in the spleen cells of mice immunized and challenged with the citrullinated peptides, which was not evident in the spleen cell cultures with their noncitrullinated counterparts (results not shown).
As expected, peptide-specific responses were impaired in the presence of anti–HLA–DR antibodies, confirming that the responses against citrullinated vim5 and citrullinated vim31 were HLA–DR restricted (Figures 3A and B). Taken together, these findings indicate that the 2 peptides identified induce a citrulline-specific T cell response in an HLA–DRB1*0401–restricted manner.
Natural processing of the immunogenic peptides from citrullinated vimentin protein.
To analyze whether the citrullinated peptides recognized by T cells were naturally processed and presented from the entire vimentin protein, we next tested the reactivity of spleen cells from peptide-immunized mice against the recombinant human vimentin, in either the citrullinated or the noncitrullinated form. In 3 independent experiments, a significant response against the citrullinated vimentin protein, as compared with that in medium alone or that with the noncitrullinated protein, was observed in spleen cells from mice immunized with citrullinated vim5 (Figure 3C). Similar results were obtained in spleen cells from mice immunized with vim31 (Figure 3D). No reactivity was observed when the spleen cells of naive mice were tested against the citrullinated vimentin protein (results not shown). These results indicate that the epitopes identified can be naturally processed from citrullinated vimentin.
Recognition of both vim5 and vim31 by T cells in RA peripheral blood.
We next analyzed whether the 2 citrullinated vimentin peptides, vim5 and vim31, that were identified as DRB1*0401-restricted T cell epitopes in DR4-transgenic mice could be recognized by T cells from patients with RA. Because the presence of IgG ACPAs in patients with RA implies the existence of memory T helper cell responses that provide help to ACPA-producing B cells, we investigated, as a first inventory, the presence of T cells with a memory phenotype specific for vim5 or vim31 in 10 ACPA+, HLA–DRB1*04+ patients with RA from the Leiden EAC cohort (see Table 1 for the patients' characteristics). The PBMCs from these patients were tested by intracellular cytokine staining for IFNγ production after stimulation with citrullinated vim5 or citrullinated vim31.
After gating of the CD3+CD4+CD45RA−lymphocytes, cells from 3 of the patients with RA responded to citrullinated vim5 but showed no reactivity against noncitrullinated vim5 (Figures 4A and B [left]). In addition, the PBMCs from 5 healthy controls were also tested. The percentages of CD3+CD4+CD45RA− cells producing cytokines were much lower in healthy controls compared with that in patients with RA, and no significant differences between the 2 different culture conditions were observed in the control PBMC cultures (Figure 4B [right]).
Among the 9 patients with RA whose PBMCs were tested for citrullinated vim31 reactivity, a marginal, but detectable, response was observed in the PBMCs of 3 patients (Figures 4C and D [left]). Among these 3 patients, the PBMCs from 1 of the patients also responded to the noncitrullinated vim31 peptide, although to a lesser extent than that to the citrullinated form. In contrast, no responses against either the citrullinated or the noncitrullinated vim31 were observed in the healthy control PBMCs (Figure 4D [right]). All PBMCs from both the patients and the healthy controls showed T cell reactivity against the control antigen (Memory Mix; results not shown). These findings suggest that both of the identified vimentin epitopes can be recognized by T cells from ACPA+, HLA–DRB1*04+ patients with RA.
At present, only limited data have been provided on the potential T cell epitopes that can be recognized in a citrulline-dependent manner by T cells from patients with RA. Therefore, an unbiased inventory focusing on the relevant autoantigens recognized by ACPAs, as in the present experiments, is highly relevant. In this study, we examined whether CD4+ T cells specific for citrullinated peptides derived from human vimentin could be identified. In total, 33 peptides were synthesized in their citrullinated form and tested for T cell reactivity in DR4-transgenic mice. A citrulline-specific response was observed against 2 of the peptides, vim5 (vim26–44) and vim31 (vim415–433). Our results indicated that these peptides are naturally processed epitopes of human vimentin. Moreover, we provided evidence to indicate that these epitopes can be recognized by T cells with a memory phenotype in patients with RA.
The 2 T cell epitopes identified in this study have not been previously described as being involved in either B cell or T cell responses. Previous studies identified T cell reactivity against a vimentin-derived peptide in HLA–DR4–positive mice (29). Although the peptide sequence used by Hill et al (29) was similar, it was not homologous to the sequence present in vimentin (30). A (large) leucine present in vimentin (position 69) was replaced by a small alanine, thereby possibly influencing the binding capacity to the HLA molecule (31). Our study identifies 2 vimentin epitopes that can be recognized by HLA–DRB1*0401–restricted T cells without the apparent requirement for additional amino acid changes, such as was described by Hill et al (29). Furthermore, the peptide in the study by Hill et al was selected on the basis of a prediction program focusing on the ability of arginine/citrulline to bind the anchor region shared by the SE-containing HLA–DRB1 alleles. We, however, used an unbiased approach in which all possible peptides from the vimentin protein were tested by positioning every arginine present in the middle of a peptide. Although the arginine is centered, it can bind to every anchor position of the binding pocket, since the peptide is 19 amino acids long.
The observed T cell responses in HLA–DR4–transgenic mice are HLA restricted, as was shown using an HLA–DR–blocking antibody. However, these IFNγ responses were only partially blocked in the presence of the anti–HLA–DR antibody. This is probably due to the high abundance of HLA–DR molecules on the cell surface of antigen-presenting cells or to the relatively low affinity of the anti-DR antibody to HLA–DRB1*0401, since the capacity of this antibody to inhibit T cell responses was previously shown to vary depending on the DR molecule involved (32).
Because binding of the peptides to the HLA–DR4 molecule is indispensable for induction of a T cell response, we tested both the citrullinated and the noncitrullinated forms of vim5 and vim31 for their binding to HLA–DRB1*0401 in a competitive binding assay, using a biotinylated influenza peptide (HA309–320) as a competitor. Only weak inhibition of the biotinylated peptide could be observed at high peptide concentrations (results not shown). Therefore, no conclusions could be drawn about the differences in binding capacity between the identified citrullinated and noncitrullinated versions of the vimentin peptides to the HLA–DR4 molecule. This observation is in line with other observations showing that low-affinity peptides can also efficiently induce T cell responses, as was described for an insulin peptide in NOD mice (33) and, recently, for a dominant gluten peptide in DQ8-transgenic mice (34). A low net binding value to the class II MHC can be a consequence of both the association and the dissociation rate, with the MHC molecule being high, while the T cell response is readily observed (34–36). Although our results might be counterintuitive, they are very intriguing, since it has been proposed that low-affinity peptides play an important role in the induction of autoimmunity by their ability to escape tolerance induction (37, 38).
The IFNγ production observed in PBMCs from RA patients was rather low. However, this would be consistent with the view that the expected precursor frequency of T cells reacting with citrullinated vimentin peptides is low. Even the T cell fraction reactive to recall antigens (i.e., a mix of tetanus toxoid, Candida albicans, and tuberculin purified protein derivative) was, on average, only 3% after restimulation. Furthermore, it has been shown that PBMCs from RA patients produce less IFNγ compared with PBMCs from healthy individuals in response to recall antigens, probably due to the effects of immunosuppressive drugs (39).
In this study, we have performed an inventory of T cell responses against the identified epitopes in 10 ACPA+, HLA–DRB1*04+ patients with RA and 5 HLA–DRB1*04+ healthy controls. To obtain a comprehensive view of the pattern of reactivity of citrullinated vimentin–specific T cells, several different aspects remain to be elucidated in a larger cohort of patients and controls. Future studies should include assessment of the recognition of these T cell epitopes in ACPA+, as well as ACPA−, patients, and should further investigate the requirement for SE-containing HLA–DRB1 alleles for this recognition. Likewise, a more extensive characterization of the cytokine profile of these T cells would be informative, since it is conceivable that the cytokine profile changed from a regulatory type in healthy controls to proinflammatory in RA patients, similar to previous findings with another RA candidate autoantigen (39). Furthermore, since our approach focused specifically on the identification of citrulline-specific T cells, the possibility cannot be excluded that T cells reacting against noncitrullinated peptides from the vimentin protein or another protein that is internalized and presented together with citrullinated vimentin might exist. These studies would involve immunization with noncitrullinated peptides from vimentin or other proteins that could be associated with vimentin in vivo.
All of these aspects imply the necessity for further, more detailed studies. Nonetheless, our study is the first to identify citrulline-specific T cell responses in humans, characterized by T cell recognition of epitopes from an autoantigen present in the inflamed joint of RA patients. As such, these results provide a valuable basis for future, more extensive studies.
All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be published. Dr. Ioan-Facsinay 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 conception and design. Feitsma, Huizinga, Toes, Ioan-Facsinay.
Acquisition of data. Feitsma, van der Voort, Franken, Bannoudi, Elferink, Drijfhout, Ioan-Facsinay.
Analysis and interpretation of data. Feitsma, van der Voort, de Vries, Toes, Ioan-Facsinay.
We are very grateful to Dr. M. Wiesner and Dr. F. Koning (from the Department of Immuno-hematology and Blood Transfusion at Leiden University Medical Center) for providing technical assistance and materials to perform the competitive binding assay. We would also like to thank N. Klar-Mohamad (from the Department of Nephrology at Leiden University Medical Center) for assisting with the purification of the anti–HLA–DR antibodies. Finally, we would like to thank Dr. R. Brand for statistical advice.