Proinflammatory responses to self HLA epitopes are triggered by molecular mimicry to Epstein-Barr virus proteins in oligoarticular juvenile idiopathic arthritis

Abstract

Objective

To evaluate whether abnormal T cell recognition may be generated by exposure to exogenous antigens presenting sequence homology with epitopes contained in self HLA alleles, and if such recognition may be part of the mechanisms that fuel inflammation in autoimmune diseases associated with certain HLA alleles.

Methods

Cytotoxic responses of peripheral blood mononuclear cells to 9-mer peptides derived from HLA molecules (DRB1*1101, DRB1*0801, or DPB1*0201) associated with oligoarticular juvenile idiopathic arthritis (JIA) or homologous peptides derived from Epstein-Barr virus (EBV) proteins (Bolf1 or Balf2) were analyzed in patients with oligoarticular JIA and in healthy controls matched for HLA–DRB1*1101, DRB1*0801, or DPB1*0201. Production of proinflammatory cytokines in culture supernatants was determined by enzyme-linked immunosorbent assay.

Results

T cell cytotoxic responses and production of proinflammatory cytokines in response to stimulation with self HLA–derived peptides were found only in patients with oligoarticular JIA, and not in controls. Patients with oligoarticular JIA, but none of the healthy controls, had EBV–self HLA cross-reactive T cells.

Conclusion

Our data suggest a disease- and allele-specific mechanism of autoimmunity in oligoarticular JIA. This mechanism may be part of the pathogenesis of the disease, and could be the basis of one of the likely multiple candidates for antigen-specific immunotherapy approaches in the future.

Oligoarticular juvenile idiopathic arthritis (JIA) is the most common form of childhood chronic arthritis. It is a relatively homogeneous clinical entity characterized by onset often before 6 years of age, chronic arthritis involving ≤4 joints during the first 6 months of disease, circulating antinuclear antibodies, and an elevated risk of developing chronic iridocyclitis. The etiopathogenesis of the disease is believed to be autoimmune in nature. Similarly to other autoimmune diseases, oligoarticular JIA is strongly associated with several HLA class II alleles, namely, HLA–DRB1*1101, DRB1*0801, and DPB1*0201. Each of these 3 alleles represents an independent risk factor for the development of the disease (1, 2). Although association with HLA alleles is a rather common characteristic of autoimmune diseases, the functional significance of these associations in etiopathogenesis has not yet been established. In this study we explored the hypothesis that T cell cross-recognition of exogenous and self HLA–derived antigens generates an abnormal circuit which maintains and expands T cells, which may participate in autoimmune inflammation by generation of proinflammatory cytokines.

We identified sequence homologies between 2 Epstein-Barr virus (EBV) proteins and the HLA–DRB1*1101, DRB1*0801, and DPB1*0201 alleles associated with oligoarticular JIA. Bolf1, an EBV protein of unknown function, shares an amino acid stretch with the third hypervariable region (HVR3) of the β1 chain of HLA–DPB1*0201, and Balf2, an EBV single-stranded DNA–binding protein (3), contains two different amino acid sequences, one sharing homology with the HVR3 of the β1 chain of HLA–DRB1*1101 and the other with the same region of DRB1*0801. Balf2 is one of the early lytic cycle proteins expressed during EBV replication, and is a target of CD8+ memory T cell–mediated cytotoxic responses and of primary T lymphocyte cytotoxicity in patients with infectious mononucleosis (4, 5).

We report here that patients with oligoarticular JIA, but not controls matched for HLA–DRB1*1101, DRB1*0801, or DPB1*0201, show cytotoxic T cell responses to synthetic self HLA–derived peptides encompassing the above-mentioned homologous sequences. Such responses are disease and allele specific, insofar as they are not found in healthy controls, they are limited to only one self allele, and they do not hinder alloreactivity, namely recognition of peptides from non-self HLA alleles. Self reactivity may be, at least in part, the consequence of initial activation of cross-reactive T cells by homologous EBV-derived epitopes. In fact, we found EBV–self HLA cross-reactive T cells in oligoarticular JIA patients, but not in controls. Most importantly, interferon-γ (IFNγ), a proinflammatory cytokine, was produced by patients with oligoarticular JIA, but not controls, upon stimulation with the same self peptides, thus supporting the concept that these autoimmune responses may have an active part in the likely multiple antigen-specific pathways that trigger and sustain inflammation in oligoarticular JIA.

PATIENTS AND METHODS

Patients.

Seventeen patients (ages 3–24 years) with oligoarticular JIA (6) were included in the study. All patients were antinuclear antibody positive (by standard immunofluorescence). None were HLA–B27 positive (by serologic typing) or rheumatoid factor positive (by standard latex fixation test). At the time of blood sampling all patients exhibited active disease, as defined according to the presence of synovitis upon examination, and were receiving nonsteroidal antiinflammatory drugs. None of the patients was treated with oral glucocorticoids. All patients had serologic evidence of previous EBV infection, as determined by titration of IgG to viral capsid antigen.

The control group consisted of subjects who 1) had serologic evidence of previous EBV infection (determined as above) and 2) carried of at least one of the HLA class II alleles associated with oligoarticular JIA. Fifteen of 70 healthy controls who had been HLA typed and tested for previous EBV infection (ages 18–26 years) met the above criteria and were therefore included in the study. For ethical reasons, the selection of additional, age-matched controls was performed with a different approach. In children undergoing venipuncture for chemistry studies prior to minor surgical procedures, an extra amount of blood was obtained. Part of the sample was used to establish cell cultures, and part was used for the HLA typing procedure (see below). Among 30 subjects, 5 (ages 5–14 years) were HLA–DRB1*1101+ with evidence of previous EBV infection, and were therefore included in the study. Permission for drawing extra blood during routine venipuncture was obtained from parents of all children.

HLA typing.

All subjects were HLA typed by standard serologic techniques. DR5 and DR8 serotypes were assigned according to hybridization patterns obtained with subtype-specific probes after specific amplification of the DR locus (Dynal, Oslo, Norway). HLA–DPB1 alleles were assessed by the polymerase chain reaction reverse dot-blot technique, using an Inno-Lipa DPB1 typing kit (Innogenetics, Zwijndrecht, Belgium).

Identification of putative epitopes.

Sequence homologies among the HVR3 of the β1 chain of the HLA alleles associated with oligoarticular JIA and proteins from EBV were obtained by scanning nonredundant NCBI databases using BLAST and FASTA software. The search was performed by comparing HLA-derived amino acid stretches from the HVR3 of the β1 chain with the known pathogen-derived protein database. This search was originally performed with the standard stringency criteria. Candidate sequences of interest were then screened with increased stringency criteria (i.e., word size, gap cost, matrix). Sequences of interest were then scanned for HLA class I universal major binding motifs, utilizing an algorithm kindly provided by Dr. Alessandro Sette (Epimmune, San Diego, CA). Scores were confirmed using a computerized prediction of peptide binding motifs to individual HLA alleles, based on the BIMAS site (www.bimas.dcrt.nih.gov). Both programs assign a score to each amino acid residue in each position, and then provide a combined HLA binding score. Sequences are shown in Table 1. Two unrelated HLA class I binder 9-mer peptides derived from the human skeletal myosin heavy chain (Myo111–119: EFQKMRRDL, Myo114–122: KMRRDLEEA) were used as controls.

Table 1. Amino acid sequences of the peptides of human or viral origin, and the corresponding peptide codes*

	Homologous sequence	Peptide	Peptide code		Binding avidity†
			EBV-derived	HLA-derived
* EBV = Epstein-Barr virus. † Ratio to the binding obtained with an influenza matrix–derived pan–class I binder peptide (GILGFVFTL), used as positive control. Values are the mean ± SEM.
β1 chain HLA–DRB1*1101	YWNSQKDLLEQKRQVD	EYWNSQKDL		DR1101/I	0.63 ± 0.1
EBV protein Balf2*	YWQLNQNLLERLSRLG	TYWQLNQNL	Balf2/1101/I		0.70 ± 0.1
β1 chain HLA–DRB1*0801	RPSAEY	TELGRPSAEYL		DR0801/I	0.59 ± 0.2
EBV protein Balf2	RDDAEY	SLTRDDAEYL	Balf2/0801/I		0.54 ± 0.3
β1 chain HLA–DPB1*0201	EEERAV	DILEEERAV		DP0201/I	0.43 ± 0.1
EBV protein Bolf1	EEEEAV	ATEEEEEAV	Bolf1/0201/I		0.30 ± 0.15

HLA–peptide binding.

HLA class I molecules, purified by immunoaffinity chromatography as previously described (7), were added to a solution of degased phosphate buffered saline (PBS) containing biotinylated peptides of 9 amino acid residues at an HLA:biotinylated peptide ratio of 1:10. HLA–peptide mix (50 μl) was incubated in a 96-well plate (Fisher, Los Angeles, CA) for a minimum of 2 hours at room temperature. After incubation, anti–HLA class I antibody (PharMingen, San Diego, CA) was added at a 1:1 molar ratio with HLA–peptide preparation and incubated for a minimum of 2 hours at room temperature, then dialyzed in a microdialyzer cassette (Cellulose Ester Membrane Frames for Microdialyzers; Fisher) for 24 hours with 3 degased PBS changes, to remove excess unbound antibodies and peptides. After dialysis, neutravidin–horseradish peroxidase (HRP) diluted 1:1,000 in PBS (Streptavidin HRP; Sigma, St. Louis, MO) was added, and the mix was dialyzed for 24 hours with 3 degased PBS changes with Spectra/Por Microdialyzer (Fisher).

Once dialysis was complete, a 50-μl test sample was transferred into a 96-well enzyme-linked immunosorbent assay (ELISA) plate (Costar, Cambridge, MA) and developed with the appropriate substrate solution. To determine background levels, the following conditions were also tested: HRP only, primary antibody with HRP, biotinylated peptide only with HRP, HLA class I molecules without biotinylated peptide but with HRP, and PBS only. The assay was stopped by addition of equal volume of 1M H₃PO₄ and the solution read at 450–650 nm.

Generation of peptide-specific cytotoxic T cell lines.

Peripheral blood mononuclear cells (PBMCs) from patients or healthy subjects were stimulated in 24-well plates at 2 × 10⁶ cells/ml (Costar, Cambridge, MA) with 10 μg/ml of peptide in RPMI 1640 medium supplemented with 10% human type AB serum (Sigma), 2 mML-glutamine, and 50 mg/ml gentamicin (Gibco, Grand Island, NY). Recombinant interleukin-2 (rIL-2) was added at 10 units/ml, 3 days after initiation of the culture. The cultures were subsequently fed every 3 days with medium containing 10 units/ml rIL-2, and restimulated on day 7 with irradiated (3,000 rad) autologous PBMCs at 2 × 10⁶ cells/ml, pulsed with the same peptide (10 μg/ml).

Cytotoxicity assay.

The cytotoxic assay was performed on day 14. Autologous EBV-transformed lymphoblastoid B cell lines were preincubated overnight with (10 μg/ml) or without peptide and ⁵¹Cr labeled to be used in a 4-hour assay, as previously described (8). The assays were performed at effector:target cell ratios of 80:1, 30:1, and 10:1. The results are shown as lytic units (LU)/10⁶ cells. A lytic unit is the number of effector cells that gives 10% specific lysis with 3,000 labeled target cells. To evaluate the effect of blocking of CD8 function, effector cells were preincubated with an anti-CD8 monoclonal antibody (mAb) (25 μg/ml; Becton Dickinson, Mountain View, CA) for 30 minutes at room temperature, and the subsequent 4-hour cytotoxic assay performed in the presence of the same concentration of the anti-CD8 mAb.

Cytokine measurement.

Supernatants of PBMCs from patients or controls, stimulated for 12 days in the absence or presence of the appropriate EBV- or HLA-derived peptide, were collected and stored at −20°C to be used in solid-phase immunoenzymatic assays for the measurement of IL-6 (R&D Systems, Minneapolis, MN), IFNγ, IL-4, and IL-12 (Bender MedSystems, Vienna, Austria).

Statistical analysis.

Data were analyzed using the Mann-Whitney U test for unpaired samples, Wilcoxon's test for matched pairs, or Spearman's rank correlation test, unless otherwise indicated. P values less than 0.05 were considered significant.

RESULTS

Peptide selection.

Using a database search, we found regions of homology between the HVR3 of the β1 chain of the oligoarticular JIA–associated DR*1101, DR*0801, and DP*0201 alleles and the EBV proteins Balf2 and Bolf1 (Table 1). Because EBV is an intracellular pathogen target of class I–restricted T cell responses, we selected appropriate peptides by scanning these regions of homology to identify putative HLA class I agretopic supermotifs. The search yielded 3 sets of paired peptides derived either from the EBV proteins Bolf1 or Balf2 (EBV-derived peptides) or from the HVR3 of the β1 chain of the HLA class II molecules associated with oligoarticular JIA (HLA-derived peptides) (Table 1). Avidity of the peptides for class I HLA molecules was measured by a direct HLA–peptide binding assay. Affinity-purified class I molecules from 3 unrelated subjects were incubated with biotinylated relevant peptides or a pan–HLA class I binder (HA-1: GILGFVFTL), which was used as a standard. Binding of the tested peptides was expressed as a ratio to that found with the standard. Results of these experiments showed that the peptide pairs had comparable binding avidity to HLA class I molecules, in accordance with predictions for HLA class I binding supermotifs (Table 1). Hence, it may be argued that differences in immunogenicity within and between the pairs should not be ascribed to different HLA class I molecule avidity.

EBV-derived peptides are targets of physiologic cytotoxic responses in patients with oligoarticular JIA and in controls.

To evaluate cytotoxicity to the EBV-derived peptides, PBMCs from DRB1*1101+ subjects were stimulated with the Balf2/1101/I peptide, while those from DRB1*0801+ or DPB1*0201+ subjects were stimulated with the Balf2/0801/I or Bolf1/0201/I peptides, respectively. Autologous lymphoblastoid cell lines, pulsed with the same peptides, were used as target cells in the cytotoxicity assay. As shown in Figure 1, there were some differences in cytotoxic activity between patients and controls, although statistical significance was not reached. No statistically significant differences between groups were found with respect to the production of IL-6, IFNγ, IL-4, and IL-12, measured in collected culture supernatants on day 12 (data not shown). Taken together, these results validate the peptides selected as antigens, presumably among the targets of physiologic immune response to EBV, a ubiquitous pathogen.

Figure 1. Cytotoxic activity of peripheral blood mononuclear cells from HLA–DR*1101+ (A), DR*0801+ (B), or DP*0201+ (C) patients with oligoarticular juvenile idiopathic arthritis (o-JIA) or controls (CTRL), stimulated and tested with the Epstein-Barr virus–derived peptides Balf2/1101/I, Balf2/0801/I, or Bolf1/0201/I, respectively. Dashed lines indicate the detection limit of the assay (4 lytic units/10⁶ cells); solid lines indicate mean values.

HLA-derived peptides homologous to EBV-derived peptides are targets of oligoarticular JIA–specific immune responses.

Effector T cells for cytotoxic assays were generated by stimulating PBMCs from patients or controls in a 14-day culture in the presence or absence of the DR1101/I, DR0801/I, or DP0201/I HLA-derived peptides. The lysis of target cells presenting the DR1101/I peptide was significantly higher (P = 0.006) in DRB1*1101+ patients than in DRB1*1101+ controls (Figure 2A). Similar results were obtained with PBMCs from DPB1*0201+ patients and DPB1*0201+ controls stimulated and tested against target cells pulsed with the DP0201/I peptide (P < 0.03) (Figure 2C). Findings with PBMCs from DRB1*0801+ patients and controls with the DR0801/I peptide were also similar, but did not reach statistical significance (Figure 2B). Cytotoxicity to self HLA–derived peptides was undetectable in the controls tested. To investigate the possibility that the cytotoxic activity toward autologous target cells presenting self HLA–derived peptides might be age related, we selected 5 additional DRB1*1101+ healthy controls comparable in age to the patients with oligoarticular JIA (see Patients and Methods). Similar to the results obtained with adult controls, PBMCs from the healthy DRB1*1101+ children, after incubation with DR1101/I peptide, showed undetectable cytotoxicity (<4 LU/10⁶ cells), therefore suggesting that such self reactivity was not age related.

Figure 2. Cytotoxic activity of peripheral blood mononuclear cells from HLA–DR*1101+ (A), DR*0801+ (B), or DP*0201+ (C) patients with o-JIA or controls, stimulated and tested with the HLA-derived peptides DR1101/I, DR0801/I, or DP0201/I, respectively. Dashed lines indicate the detection limit of the assay (4 lytic units/10⁶ cells); solid lines indicate mean values. See Figure 1 for definitions.

Next, in the same patients, we correlated cytotoxic responses to HLA-derived peptides with those to the homologous EBV-derived peptides (Figure 3). A significant difference (P < 0.03 by Wilcoxon's matched pairs test), with a higher response to the HLA-derived peptide, was observed. In contrast, in the control group, we found that cytotoxic responses to the HLA peptides were significantly lower than those to the homologous EBV-derived peptides (P < 0.02) (data not shown). Taken together, these findings show that self reactivity is present in patients with oligoarticular JIA, but not in healthy controls expressing the HLA alleles associated with oligoarticular JIA.

Figure 3. Comparison of cytotoxic response to the Epstein-Barr virus (EBV)- or the homologous HLA-derived peptide in peripheral blood mononuclear cells from patients with oligoarticular juvenile idiopathic arthritis.

Cytokine production was tested in culture supernatants from patient and control PBMCs stimulated with the self HLA–derived peptides. The production of IL-4, IL-6, and IL-12 was comparable in supernatants from PBMCs of patients and controls (data not shown). IFNγ production in response to HLA-derived peptides was significantly higher (P < 0.05) in patients (mean ± SEM 36.9 ± 13.7 pg/ml) than in controls (5.2 ± 3.9 pg/ml) (Figure 4). Only 1 of 9 controls produced detectable levels of IFNγ, compared with 9 of 15 patients (P = 0.03 by Fisher's exact test). Moreover, IFNγ production was significantly correlated (r = 0.670, by Spearman's rank correlation test, P = 0.023) with specific lysis, expressed as LU/10⁶ cells.

Figure 4. Interferon-γ (IFNγ) production in culture supernatants from peripheral blood mononuclear cells of patients with o-JIA or controls, stimulated with self HLA–derived peptides. Dashed line indicates the detection limit of the assay (4 pg/ml); solid lines indicate mean values. See Figure 1 for other definitions.

Patients with >1 oligoarticular JIA–associated allele show a preferential pattern of self reactivity. Coexpression of at least 2 of the HLA–DRB1*1101, DRB1*0801, and DPB1*0201 alleles has been demonstrated to represent an independent risk factor for oligoarticular JIA (1). Hence, it was important to evaluate whether cells from patients with >1 oligoarticular JIA–associated allele would identify a preferential target in their responses. Interestingly, in all of these patients, the strongest reactivity was directed toward 1 of the 2 or 3 alleles expressed (Table 2).

Table 2. Cytotoxic activity of peripheral blood mononuclear cells from individual patients with oligoarticular juvenile idiopathic arthritis expressing >1 of the alleles associated with the disease*

* Results are expressed as lytic units/10^6 cells. Detection limit of the assay is 4 lytic units/10^6 cells. ND = not determined.
DRB1*1101+DPB1*0201+ patients (n = 6)
DR1101/I
Patient 1	<4
Patient 2	<4
Patient 3	8.9
Patient 4	8.6
Patient 5	6.1
Patient 6	41
DP0201/I
Patient 1	9.2
Patient 2	62.5
Patient 3	21
Patient 4	<4
Patient 5	31.8
Patient 6	125
DRB1*0801+DPB1*0201+ patients (n = 2)
DP0201/I
Patient 1	19.7
Patient 2	76
DR0801/I
Patient 1	<4
Patient 2	<4
DRB1*0801+1101+DPB1*0201+ patients (n = 3)
DR1101/I
Patient 1	144.9
Patient 2	12.2
Patient 3	43.2
DP0201/I
Patient 1	ND
Patient 2	10.8
Patient 3	6.9
DR0801/I
Patient 1	27.7
Patient 2	56.6
Patient 3	<4

Triggering of cross-reactive self cytotoxicity by recognition of viral peptides in patients but not in controls.

Based on the results described above, we investigated whether the differences in self reactivity between patients and controls could be ascribed, at least in part, to cross-recognition of EBV and self HLA–derived peptides. Effector cells were generated by 14-day incubation of PBMCs in the presence or absence of the EBV-derived peptides, homologous to the peptides derived from the HVR3 on the β1 chain of the DR*1101, DR*0801, or DP*0201 alleles. The cytotoxic activity of these cells was then tested against autologous target cells pulsed overnight with the corresponding homologous HLA-derived peptide. As shown in Figure 5, the cytotoxic response of PBMCs from DRB1*1101+ patients, stimulated with the Balf2/1101/I EBV-derived peptide, to target cells pulsed with the homologous DR1101/I HLA-derived peptide, was significantly higher (P < 0.04) than that observed with PBMCs from controls. Similar results were obtained with DRB1*0801+ patient PBMCs stimulated with the Balf2/0801/I EBV-derived peptide and tested against the homologous DR0801/I HLA-derived peptide and with DPB1*0201+ patient PBMCs stimulated with the Bolf1/0201/I EBV-derived peptide and tested against the homologous DP0201/I HLA-derived peptide.

Figure 5. Cytotoxic activity of peripheral blood mononuclear cells from HLA–DR*1101+ (A), DR*0801+ (B), or DP*0201+ (C) patients with o-JIA or controls stimulated with the Epstein-Barr virus–derived peptides Balf2/1101/I, Balf2/0801/I, or Balf1/0201/I, respectively (effector cells [E]) and tested against autologous target cells (T) pulsed with the homologous HLA-derived peptides DR1101/I, DR0801/I, or DP0201/I, respectively. Dashed lines indicate the detection limit of the assay (4 lytic units/10⁶ cells); solid lines indicate mean values. See Figure 1 for other definitions.

In representative experiments (2 DPB1*0201+ and 2 DRB1*1101+ patients), anti-CD8 mAb efficiently blocked cytotoxicity to either the HLA-derived (mean ± SEM 73.7 ± 37.1% inhibition) or the EBV-derived (80.4 ± 27.6% inhibition) peptides. This indicates that the responses were CD8-mediated.

Preservation of alloreactive cytotoxic responses to HLA-derived peptides in patients with oligoarticular JIA.

To analyze the specificity of the cytotoxic responses to self HLA–derived peptides, we performed the cytotoxicity assay with PBMCs from patients with oligoarticular JIA (n = 5) and controls (n = 13). The PBMCs were cultured and tested in the presence of autologous lymphoblastoid cells pulsed with peptides derived from HLA alleles that were still part of the pool tested in this study, but different from those expressed on the subjects' own cells. Comparable results were found in oligoarticular JIA patients and controls (mean ± SEM 14.0 ± 16.3 LU/10⁶ cells in patients and 21.0 ± 47.7 LU/10⁶ cells in controls; P = 0.80), suggesting that HLA peptides derived from alleles associated with oligoarticular JIA, but not the same as those expressed by the subject being tested, are recognized as exogenous peptides by both patients and controls.

Self-reactive cytotoxicity is not aimed at antigens irrelevant to the disease process.

We evaluated cytotoxic activity of 6 oligoarticular JIA patients' PBMCs, which were incubated and tested against 2 unrelated self peptides derived from the human skeletal myosin heavy chain (Myo111–119: EFQKMRRDL and Myo114–122: KMRRDLEEA). No lysis was detectable (<4 LU/10⁶ cells for both Myo111–119 and Myo114–122), similar to findings in 5 controls.

DISCUSSION

EBV is a pandemic infectious agent that affects people at a young age. It has the intriguing ability to mold the immune responses of the host in order to ensure its own long-term survival. Different mechanisms to blunt the host protective immune responses have been identified, often based on functional molecular mimicry (9–11). These characteristics have made EBV a longstanding prime suspect as an etiopathogenic agent in various autoimmune diseases, including adult rheumatoid arthritis (RA) and systemic lupus erythematosus (12–14). The concept that EBV plays a role in disease triggering is particularly appealing in oligoarticular JIA since in the general population the encounter with the virus usually occurs during the first years of life, and in the vast majority of cases the onset of oligoarticular JIA occurs before 6 years of age.

In this study we identified sequence homologies between the 2 EBV proteins, Bolf1 and Balf2, and the HVR3 of the β1 chain of the DRB1*1101, DRB1*0801, and DPB1*0201 HLA class II alleles associated with oligoarticular JIA. We explored the hypothesis that abnormal T cell cross-recognition of these shared sequences is present in patients with oligoarticular JIA and has a possible role in initiating and/or maintaining a circuit of autoimmune inflammation.

We first identified putative class I agretopic motifs, bearing the homologous sequences, shared among the EBV and self proteins. We then tested these peptides in a population of patients with oligoarticular JIA, classified according to HLA type, and compared the immune responses in a population of healthy controls expressing the HLA–DRB1*1101, DRB1*0801, or DPB1*0201 alleles. We found that the EBV-derived peptides are targets of cytotoxic responses, associated with production of proinflammatory cytokines, without significant differences between patients and controls. These responses are consistent with anamnestic physiologic responses to an infectious agent, since all subjects studied tested seropositive for previous EBV infection.

When we evaluated cytotoxic responses to the self HLA–derived peptides, we found that healthy controls with evidence of previous EBV infection did not show a T cell cytotoxic response to these peptides. Of note, the absence of cytotoxic response was found both in healthy adults and in healthy children, therefore excluding a possible relationship of the recognition of those peptides to age. In contrast, T cell cytotoxic responses directed against self-derived HLA epitopes could be found in patients with oligoarticular JIA. These responses were accompanied by a significant increase, compared with controls, in the production of IFNγ. Patients carrying >1 susceptibility allele showed a preferential response toward 1 of them, providing functional results consistent with the well-established concept that each HLA allele associated with oligoarticular JIA represents an individual risk factor. T cell responses to the self HLA–derived peptides were not the consequence of generalized and nonspecific self reactivity. This is supported by the finding of selective reactivity against only 1 of the epitopes in patients expressing >1 allele associated with oligoarticular JIA, and by the lack of T cell responses against human skeletal myosin heavy chain–derived peptides, used as controls. Alloreactivity, namely the capability of reacting against non-self HLA peptides, was conserved in oligoarticular JIA patients and comparable with that in controls.

Immune recognition of HLA-derived epitopes has been described in other autoimmune diseases. HLA–B27+ patients with spondylarthropathies show abnormal T and B cell responses to the epitopes shared by the HLA–B27 and peptides derived from Klebsiella pneumoniae and Shigella flexneri (15–18). In addition, in patients with RA, HLA class I–restricted CD8+ T cell clones reactive with self HLA–DQ2 have been described (19). Moreover, IgM rheumatoid factor has been shown to react with HLA class I molecules (20).

Further investigation of the self reactivity to HLA-derived peptides in patients with oligoarticular JIA showed that expansion of EBV-specific T cells led to the generation of self HLA–directed cross-reactive responses. The cross-recognition appears to be a disease-specific phenomenon, not related to genetic background, since controls matched for HLA–DRB1*1101, 0801, or DPB1*0201 haplotypes did not show cross-reactive responses. This observation in controls is consistent with observations reported by Burrows et al (21). In their study, an EBV peptide was able to induce a cross-reactive response against the HLA–B*4402 allele, whereas they found that in HLA–B*4402+ subjects the EBV peptide did not induce T cell responses to the HLA–B*4402 allele. These observations show that in healthy subjects, the response to self HLA epitopes is abrogated and tolerance to the self epitope is present.

Taken together, our data from patients with oligoarticular JIA support a scenario in which the encounter with EBV antigens triggers cytotoxic T cell responses and production of proinflammatory cytokines. This type of reaction should be self-limited in a physiologic immune response, and it may be perpetuated in a disease- and allele-specific manner by the presence of T cell reactivity to homologous class II self HLA–derived peptides. In this context, HLA would not act as an antigen-presenting molecule, but as an antigen itself. Sustained T cell reactivity to self peptides may be responsible for continued production of proinflammatory cytokines, which may participate in the autoimmune damage in vivo, in the context of a class I–restricted immune recognition of self HLA epitopes. The scenario described here has many elements in common with other models of abnormal immune responses to self and exogenous antigens, as described by us and others. Of particular relevance are the analogies with reactive arthritis in HLA–B27+ individuals and with “shared epitope”–positive RA (7, 18, 22–28). In some instances, such models have evolved from the stage of study of pathogenesis to a basis for investigations into the development of immunotherapy (29–34). Further work is in progress to characterize the T cell events described herein at the individual cell level. This model may contribute to elucidation of the complex pathogenesis of oligoarticular JIA and identification of appropriate tools for antigen-specific immunotherapy.

Acknowledgements

We wish to thank Dr. Alessandro Sette for his expert assistance in identification of the candidate agretopes. We also wish to thank Nicole Lewon for her excellent work in editing the manuscript.