Rheumatoid arthritis is characterized by synovial joint infiltration of activated CD4+ T cells and MHC class II+ APC, and is linked to specific HLA-DR alleles. Candidate autoantigens in synovial fluid and cartilage include type II collagen (CII) and cartilage gp39 (HCgp39). Using preparations of native Ag and T cells derived from Ag-immunized DR4-transgenic mice, we determined that human ex vivo differentiated DR4+ dendritic cells (DC) and macrophages (Mφ) can mediate MHC class II presentation of CII or HCgp39 epitopes. The form of the Ag (soluble, partially degraded, or particulate) delivered to the APC influenced its presentation by DC and Mφ. DC efficiently presented partially degraded, but not native CII α-chains, while Mφ presentation was most efficient after phagocytosis of bead-conjugated CII. Both DC and Mφ presented soluble HCgp39, and activated Mφ from some donors presented epitopes derived from endogenously synthesized HCgp39. When synovial fluid from rheumatoid arthritis patients was used as a source of Ag, DC presentation of HCgp39 and CII epitopes was efficient, indicating that synovial fluid contains soluble forms of CII and HCgp39 amenable to internalization, processing, and presentation. These data support the hypothesis that CII and HCgp39 are autoantigens and that their class II-mediated presentation by DC and Mφ to T cells in vivo has a critical role in the pathogenesis of human rheumatoid arthritis.
Rheumatoid arthritis (RA)3 is characterized by chronic synovial joint inflammation and infiltration of activated CD4+ T cells and APC, e.g., dendritic cells (DC) and macrophages (Mφ) (1). Matrix metalloproteinases (MMPs) secreted by activated macrophages and fibroblasts degrade cartilage proteins, resulting in increased availability of extracellular matrix proteins for internalization by APC (2). As a result of inflammation, tissue damage, or bacterial infection, extracellular and intracellular proteolytic degradation of normally sequestered cartilage proteins may produce peptides that bind to MHC class II (MHCII) molecules in activated APC, leading to initiation of autoreactive T cell responses. The observation that specific HLA class II alleles, including subtypes of DR4 (DRB1*0401, *0404, *0405) and DR1 (DRB1*0101), confer susceptibility to RA supports this hypothesis (3). Each of these DR molecules contains a conserved sequence within aa 67–74 of the peptide-binding groove of the DRβ1 domain, which is absent in DR molecules not linked to RA. The structural features of this shared epitope affect both peptide binding and T cell recognition.
Although the target Ags for disease initiation or maintenance in humans remain unknown, candidate RA autoantigens have been identified. Type II collagen (CII) is a unique component of articular cartilage, and several studies have demonstrated T or B cell immunity to CII in RA patients (4, 5, 6, 7). The significant amounts of degraded human CII (hCII) in rheumatoid cartilage, which correlate with elevated production of collagenases MMP-1 and MMP-13, are likely to be accessible for processing and presentation by joint APC (8, 9). Expression of the DRB1*0101 or DRB1*0401 transgenes in arthritis-resistant mouse strains confers susceptibility to collagen-induced arthritis, and DR-restricted T cell responses to immunodominant CII epitopes, including 259–273, are generated (10, 11, 12, 13). These murine studies suggest that MHCII+ APC displaying self peptides are involved in RA pathogenesis.
A second candidate RA autoantigen is human cartilage glycoprotein 39 (HCgp39, or YKL-40), typically expressed by cells in rheumatoid synovium (14). Synthesis and secretion of HCgp39 occur during monocyte to macrophage differentiation, and is increased in individuals experiencing active arthritis (15, 16). Although serum levels of HCgp39 are elevated in several inflammatory joint diseases, increased HCgp39 production correlates with the degree of joint destruction and disease activity only in RA (17, 18, 19). Immunization of susceptible strains of mice with HCgp39 induces a chronic, relapsing arthritis and T cell responses to multiple immunodominant epitopes, including 263–275 (20). Peripheral blood T cells from DR4+ RA patients and healthy adults responded to these same immunodominant HCgp39 epitopes in vitro (21).
Although these studies suggest hCII and HCgp39 may be RA autoantigens, few studies have characterized the ability of human APC, particularly DC or Mφ, to process and present these self proteins. DC and Mφ are present in significant numbers in the rheumatoid joints of RA patients. In rheumatoid synovial tissue and fluid, 20–45% of non-T mononuclear cells are CD33+CD14dim DC (22). These DC and CD14+CD68+ Mφ express high cell surface MHCII, and a subset expresses CD86, a molecule that signifies full costimulatory function. In inflamed synovial tissue, activated DC and Mφ are found in clusters with activated CD4+ T cells (23, 24).
To study CII and HCgp39 presentation by human DC and Mφ, we generated T cell hybridomas by immunizing DR4-transgenic mice with peptides corresponding to CII 259–263 and HCgp39 263–275. These T cells were used to determine whether these epitopes resulted after human DR4+ blood monocyte-derived DC and Mφ were incubated with native human CII, HCgp39, and synovial fluid (SF) from RA patients. Our results show that ex vivo differentiated human DC and Mφ, phenotypically similar to RA synovial joint APC, are capable of generating and displaying immunodominant epitopes from two autoantigens found in inflamed synovial joints of RA patients. When RA SF was used as a source of Ag, DC presentation of HCgp39 and CII epitopes was very efficient, indicating that SF contains soluble forms of these Ag amenable to internalization, processing, and presentation. Data presented in this study support the hypothesis that CII and HCgp39 are autoantigens presented to T cells by DC and Mφ during human RA.
Materials and Methods
Generation of T cell hybridomas specific for CII 259–273 and HCgp39 263–275
Mice carrying the DRA and DRB1*0401 transgenes (kindly provided by D. Zaller, Merck Research Laboratories, Rahway, NJ) were immunized in the hind footpads and base of the tail with synthetic peptides (50 μg) in CFA. After 10 days, T cells from draining LN were restimulated in vitro with peptide for 3 days and fused with the TCRα/β−/− variant of the BW5147 thymoma (25). T cell hybrids were tested for DR4 restriction and peptide specificity by incubation of 1 × 105 T cells with variable amounts of cognate or irrelevant peptides and DR4+ or DR4− mouse splenocytes (3 × 105), or a human DRB1*0401 B-lymphoblastoid cell line (B-LCL) Priess (1 × 105) (26). IL-2 production by the T cell hybrids was assessed by proliferation of HT-2 cells, an IL-2-dependent cell line (25), using an Alamar blue colorimetric assay; results are expressed as arbitrary units of OD at 570 vs 600 nm, average of duplicate wells. Synthetic peptides were made in the peptide synthesis core facility of the City of Hope Medical Center and purified by HPLC, and the sequences were confirmed by mass spectrometry.
Peripheral blood was obtained from normal healthy volunteers with their informed consent, according to the City of Hope Institutional Review Board (IRB) guidelines. SF samples were obtained from RA patients with active synovitis who fulfilled the American College of Rheumatology criteria (27). Patient samples were collected with informed consent, according to the University of Southern California School of Medicine IRB guidelines.
Generation of human DRB1*0401 monocyte-derived DC and Mφ
DR4+ healthy adult donors were identified by FACS analyses of peripheral blood with a pan anti-DR4 mAb (359-13F10) obtained from S. Radka (Ribozyme Pharmaceuticals, Boulder, CO) (28). Genomic DNA was extracted from DR4+ PBMC and subtyped for DRB1*04 alleles by PCR using sequence-specific primers (Dynal Biotech, Lake Success, NY).
Blood from DRB1*0401 donors was Ficoll gradient separated, and PBMC were plated for differential adherence at 50 × 106 cells/10-cm plate in 10 ml of RPMI 1640 containing 10% FCS, 2 mM glutamine, 100 U penicillin/0.1 mg streptomycin/ml, 10 mM HEPES buffer, and 1 mM sodium pyruvate. After 2 h, the nonadherent cells (consisting primarily of lymphocytes) were removed by gently washing the plates three times with warm culture medium. The adherent monocytes were cultured an additional 12 h, after which time they detached from the plate. For immature DC, monocytes were cultured in human rGM-CSF (800 U/ml; Immunex, Seattle, WA) and human rIL-4 (500 U/ml; PeproTech, Rocky Hill, NJ) at 1 × 106 cells/ml for 8 days (29). For Mφ, monocytes were cultured in human rM-CSF (50 ng/ml; PeproTech) and human rIL-6 (20 ng/ml) at 3 × 105 cells/ml for 6 days (30). DC and Mφ were fed by replacing one-half of the culture medium with fresh medium and cytokines. DC were matured with LPS (2 μg/ml) for 17 h, while Mφ were activated by human rIFN-γ (100 U/ml; PeproTech) for 14 h, followed by removal of the IFN-γ, and an additional 10-h incubation in medium containing fresh M-CSF and IL-6. Mφ were then treated with LPS (0.1 μg/ml) for 14 h (31). Mφ were removed from the plate with trypsin/EDTA (1 ml/well for 5 min), or with warm PBS for FACS analysis of trypsin-sensitive CD33 expression.
FACS analyses of DC and Mφ
FcR were blocked on DC and Mφ with human IgG (1 μg/1 μl), except when using IgG-Alexa Fluor 488 to detect FcR. Cells (2 × 105/100 μl) were stained in PBS, 5% newborn calf serum, and 0.1% sodium azide. DC and Mφ were incubated with anti-CD86 FITC, anti-CD14 FITC, anti-CD33 PE, or biotinylated mAb L243 (directed against monomorphic determinants on HLA-DR), followed by streptavidin-CyChrome. All mAb were from BD PharMingen (San Diego, CA), except L243, which was purified from hybridoma supernatant. To measure receptor-mediated endocytosis, DC were incubated with FITC-dextran (1 mg/ml) (32) at 4°C (surface binding) or at 37°C (surface binding and endocytosis via mannose receptors). To measure binding of IgG to FcR, DC were stained at 4°C with human IgG-Alexa Fluor 488. After staining, all cells were fixed with 1% paraformaldehyde and analyzed on a FACSCalibur using CellQuest software (BD Biosciences, San Jose, CA).
Purification of HCgp39 from MG-63 cell supernatant
The HCgp39-producing MG-63 osteosarcoma (American Type Culture Collection, Manassas, VA) was cultured at confluence in Mφ serum-free medium (Life Technologies Invitrogen, Carlsbad, CA) for 30 days, and the HCgp39 in conditioned medium was affinity purified on a heparin Poros HE 4.6 × 100 column using a BioCad workstation (PerSeptive Biosystems, Framingham, MA) (14, 33, 34). Briefly, MG-63-conditioned medium was loaded onto the Poros HE column and washed extensively with 10 mM sodium phosphate, 50 mM sodium chloride, pH 7.5. Bound material was eluted with a NaCl gradient (from 50 mM to 2 M) in 10 mM sodium phosphate, pH 7.5. HCgp39 (∼40 kDa) was detected in concentrated fractions by immunoblotting with an anti-YKL-40 mAb (Quidel, San Diego, CA).
Human CII and generation of cyanogen bromide-derived fragments
CII was extracted and purified from adult human femoral condylar cartilage by differential salt precipitation, as described (8). For some experiments, CII was purchased from Biogenesis (Kingston, NH) and used with similar results. To mimic degraded CII, CII was cleaved at methionine residues with cyanogen bromide (CNBr or CB) (35, 36). CII was dissolved in 70% formic acid at 5 mg/ml to which was added 12 mg/ml of CNBr. The tubes were flushed with nitrogen, sealed, and kept at 26°C. To ensure the most complete cleavage possible, the reaction was continued for 18–20 h and was terminated by a 10-fold dilution with distilled water. After lyophilization, the CII CNBr fragments were stored at −20°C. Purity and characterization of the CII fragments used in these experiments were determined by SDS-PAGE and were identical with published results (8). Amino acid analysis was used to calculate the total amount of protein in CII preparations, and hydroxyproline content was used to determine the actual concentration of CII.
Analyses of CII and HCgp39 by immunoblotting
Aliquots of purified HCgp39, hCII, and hCII CNBr fragments (1 μg), SF (25 μl), or Nonidet P-40-solubilized cell lysates of Mφ (105 cell equivalents) were mixed with SDS sample buffer and heated to 60°C (CII) or boiled (HCgp39) for 10 min before 10% SDS-PAGE and transfer to nitrocellulose. Immunoblotting was accomplished using mAb MAB 1330 and MAB 8887 (Chemicon, Temecula, CA) to detect hCII and hCII CNBr fragments, respectively, and an anti-YKL-40 mAb or polyclonal anti-YKL-40 antiserum (Quidel) to detect HCgp39. Binding of primary mAbs was detected using peroxidase-conjugated F(ab′)2 goat anti-mouse (or anti-rabbit) IgG (Jackson ImmunoResearch Laboratories, West Grove, PA) and a chemiluminescent substrate for peroxidase, followed by exposure to film.
Ag presentation assays with human DC and Mφ
DRB1*0401 DC were incubated with various Ag, including native bovine or human CII, hCII CNBr fragments, conditioned medium containing HCgp39 from MG-63 cells, purified HCgp39, or cell-free RA SF. A YKL-40-specific ELISA kit (Quidel) was used to quantitate the amount of HCgp39 present in heparin-purified fractions and RA SF. DC (3 × 105 cells/ml) were incubated overnight either in the absence or presence of Ag, washed, and matured with LPS for 17 h before incubation with T cells. For T cell assays, 1 × 105 DC were incubated with 1 × 105 T cells, or DC were titrated in wells before the addition of T cells. Synthetic cognate peptides were added to DC during the T cell assay. Cultures were incubated for 20 h, and T cell IL-2 production was assessed by proliferation of HT-2 cells, as described above.
DRB1*0401 Mφ (3 × 105 cells/ml) were incubated in the absence or presence of purified gp39, or native or CNBr fragments of bovine CII, 2 h before, and during, treatment with IFN-γ. Ag and IFN-γ were removed by washing, and Mφ were cultured an additional 10 h in fresh medium with fresh cytokines before LPS addition. For CII presentation assays, Mφ were fixed with 0.2% paraformaldehyde and extensively washed before incubation with T cells. For T cell assays, 1–10 × 104 Mφ were incubated with 1 × 105 T cells for 20 h. Synthetic cognate peptides were added to Mφ during the T cell assay.
Proteins were coupled to polystyrene beads (diameter, 3 μm; Polysciences, Warrington, PA) either by a covalent amino bond (hCII) or by passive adsorption (human IgG, human serum albumin (HSA)), according to the manufacturer’s directions. The amount of protein bound to the beads (typically 0.1–0.2 mg/ml) was determined based on the difference in OD of the protein solution before and after the linkage. Activated human Mφ (1 × 105/well) were incubated with titrated amounts of proteins (soluble or bead linked), and after 3 h, T cell hybrids (1 × 105/well) were added and cultures were incubated for 20 h.
T cell hybrids to CII and HCgp39 are specific for cognate peptide on DR4+ human APC
T cell hybrids were generated by immunizing DR4+ transgenic mice with synthetic peptides corresponding to immunodominant epitopes of CII (259–273) and HCgp39 (263–275). To test for DR4 and Ag specificity, T cells were incubated with DR4+ and DR4− mouse splenocytes in the presence or absence of cognate peptide (Fig. 1). T cell hybrids specific for CII 259–273 and HCgp39 263–275 on DR4+ mouse splenocytes were also specific for cognate peptide/DR4 complexes presented by the human B-LCL Priess, indicating that these T cells respond to human APC (Fig. 1). No responses could be detected when T cell hybrids were incubated with DR4− mouse splenocytes and cognate peptide (Fig. 1), nor when incubated with DR4+ APC and irrelevant DR4-binding peptides (unpublished data).
Phenotypic characterization of human monocyte-derived DC and Mφ used for Ag presentation assays
We next sought to determine whether human DC and Mφ could present CII 259–273 and HCgp39 263–275 to T cell hybrids after intracellular processing of native CII or HCgp39. These epitopes only would be arthritogenic in humans if human APC were able to process native protein and present similar peptides. In addition, it was important to confirm that the CII- and HCgp39-specific T cell hybrids recognize cognate peptide/DR4 complexes resulting from Ag processing, as some T cell hybrids generated by immunization with peptides do not recognize peptide/MHCII complexes resulting from processing of native protein (21).
Ag presentation assays were performed using human monocyte-derived DC and Mφ generated from healthy DRB1*0401 donors of peripheral blood. CD14+ monocytes cultured in GM-CSF and IL-4 for 8 days differentiated into nonadherent clusters of cells that exhibited typical DC morphology. Flow cytometry was used to determine that these cells were CD14−CD33+, a phenotype characteristic of DC (Fig. 2,B). Exposure of DC to LPS overnight increased cell surface MHCII and CD86 expression, events associated with DC maturation (Fig. 2, C and D). Mature DC exhibited reduced uptake of FITC dextran at 37°C (a molecule endocytosed via mannose receptors) compared with immature DC (Fig. 2,E). DC maturation also resulted in decreased cell surface FcR, as measured by incubation of DC with fluoresceinated human IgG at 4°C (Fig. 2 F).
Mφ were obtained by culturing CD14+ monocytes in M-CSF and IL-6 for 6 days, which produced an adherent CD14+CD33+ cell population (Fig. 2,H). Mφ were activated by priming with IFN-γ and subsequent stimulation with LPS, resulting in increased surface MHCII and CD86 expression (Fig. 2, I and J). Exposure to either IFN-γ or LPS alone did not result in significant increases in MHCII or CD86 expression on Mφ (unpublished data). Thus, the DC and Mφ obtained by these differentiation and activation regimens exhibited high surface MHCII and CD86; these profiles are similar to activated DC and Mφ found in SF of RA patients.
Human DC can efficiently present an immunodominant epitope derived from native HCgp39
To determine whether DRB1*0401 DC could present the HCgp39 263–275 epitope, immature DC were cultured with native Ag overnight and, after LPS maturation, incubated with HCgp39-specific T cell hybrids. Native HCgp39, present in conditioned serum-free medium from the osteosarcoma MG-63, was affinity purified using a heparin column, according to published protocols (34). One protein of ∼40 kDa was detected in purified fractions of HCgp39 using an anti-YKL-40 mAb (Fig. 3,A). This protein was present in the culture medium after, but not before, extended (30 day) incubation of MG-63 cells. Mature DRB1*0401 DC could efficiently present HCgp39 263–275 to T cells after incubation with MG-63-conditioned medium or with heparin-purified HCgp39 (Fig. 3, B and C). Interestingly, the amount of HCgp39 protein required to detect a T cell response after incubation with DC was less than amounts typically found in RA patient SF (see below). Paraformaldehyde-fixed DC did not present the epitope after incubation with the MG-63-conditioned medium, indicating that intracellular HCgp39 processing was required (Fig. 3 B).
Activated human Mφ present HCgp39 263–275 in the absence of exogenously supplied HCgp39
To determine whether DRB1*0401 Mφ could present the HCgp39 263–275 epitope to T cells, Mφ were incubated in the presence or absence of heparin-purified soluble HCgp39, activated, and cultured with HCgp39-specific T cells. Efficient presentation of HCgp39 263–275 by activated Mφ occurred in the absence of exogenously added HCgp39, and presentation was augmented by soluble HCgp39 (Fig. 4,A). Resting Mφ presented soluble HCgp39 poorly, but could present the HCgp39 263–275 peptide via surface MHCII, suggesting that resting Mφ lack machinery for efficient Ag processing (Fig. 4,A). Presentation of endogenous HCgp39 by activated Mφ is consistent with the observation that Mφ synthesize HCgp39 during their differentiation from monocytes (16). To discard the possibility that Mφ can nonspecifically stimulate T cell hybrids in the absence of Ag, activated Mφ were cultured with an IgGκ-specific T cell hybrid in the presence or absence of cognate peptide (IgGκ 188–203). T cells were stimulated only when the IgGκ peptide was added (Fig. 4 A).
To confirm differences in presentation of endogenous HCgp39 by DC and Mφ, monocytes from the same donor were differentiated into either DC or Mφ and incubated in the presence or absence of soluble HCgp39 before activation and incubation with T cells. Only Mφ presented HCgp39 263–275 in the absence of exogenously supplied HCgp39 (Fig. 4 B), consistent with the finding that DC synthesize significantly less HCgp39 mRNA than Mφ (37).
Interestingly, the ability of activated Mφ to present endogenously synthesized HCgp39 varied significantly, depending on the donor of normal blood monocytes (Fig. 4,C). Immunoblots of Mφ (resting and activated) from these donors showed that HCgp39 production by resting Mφ increased upon activation, and that the amount of HCgp39 protein in cell lysates correlated with the extent of Ag presentation (Fig. 4,C). Activated Mφ from donor 1 exhibited the highest level of endogenous HCgp39 production and presentation via MHCII. Activated Mφ from donor 1 were tested twice several months apart, and, in both experiments, efficiently presented endogenously synthesized HCgp39. We could not detect HCgp39 on immunoblots of resting or activated Mφ from donors that did not appreciably present endogenously synthesized HCgp39 (Fig. 4 C). Addition of cognate HCgp39 peptide to resting and activated Mφ from each donor activated the HCgp39-specific T cells, indicating that all Mφ tested were MHCII+ and capable of stimulating T cells (unpublished data). This is the first evidence that Mφ production of HCgp39 leads to MHCII-mediated presentation of the immunodominant HCgp39 epitope after exposure to activating stimuli. The reason that the extent of HCgp39 synthesis by activated Mφ varies among normal donors, yet is consistently synthesized by one donor over time, is unclear.
DC efficiently present an immunodominant CII epitope only after partial CII degradation
To determine whether DRB1*0401 DC could present the CII 259–273 epitope to CII-specific T cell hybrids, immature DC were incubated with native bovine or human CII or hCII CNBr-derived fragments. The epitope 259–273 is contained within the >30-kDa CB11 fragment (Fig. 5,A). Paraformaldehyde-fixed DC incubated with the hCII CNBr fragments were unable to present CII 259–273, indicating that the fragments required intracellular processing for generation of this epitope (Fig. 5,B). CNBr fragments of hCII were presented significantly better than either native bovine or human CII α-chains, suggesting that partially degraded CII is more efficiently internalized, or processed intracellularly, by DC (Fig. 5,B). Titration of hCII CNBr fragments and highly purified native hCII showed that while DC efficiently presented hCII CNBr fragments, higher concentrations of native hCII were presented poorly (Fig. 5,C). Titration of DC after incubation with native hCII (34 μg/ml) or hCII CNBr fragments (15 μg/ml) indicated that as few as 6 × 103 DC were required for presentation of both hCII CNBr fragments and cognate peptide with comparable high efficiency to CII-specific T cell hybrids (Fig. 5 D). These results indicate that a T cell response can be elicited from very low numbers of DC incubated with partially degraded CII, but not native CII.
Human Mφ present soluble native CII and CB fragments of CII poorly
To assess Mφ generation of the CII 259–263 epitope, DRB1*0401 Mφ were incubated in the presence or absence of native bovine CII or bovine CII CNBr fragments, activated, and cultured with CII-specific T cells. In preliminary experiments, we did not detect presentation of native CII or CII CNBr fragments to T cells by resting or activated Mφ, yet cognate peptide, CII 259–273, was presented very efficiently (unpublished data). To determine whether soluble factors released by Mφ could suppress T cell responses, we paraformaldehyde fixed Mφ after incubation with Ag and activation, before incubation with T cells. Poor presentation of CII and CII CNBr fragments was observed for both activated and resting fixed Mφ, although the hCII 259–273 peptide was efficiently presented (Fig. 6). Compared with DC, Mφ presentation of soluble CII CNBr fragments was markedly inefficient. Because flow cytometry analyses did not reveal significant differences between DC and Mφ MHCII surface expression (unpublished data), the poor presentation of soluble CII by Mφ may be due to inefficient CII internalization and/or proteolytic degradation.
Presentation of hCII by human Mφ is increased when hCII is attached to polystyrene beads
Various studies have demonstrated that presentation of Ag by Mφ is greatly improved if particulate Ags are acquired via phagocytosis (38). To determine whether phagocytosis of hCII could increase Mφ presentation of CII, we coupled hCII to polystyrene beads. Light microscopy confirmed efficient internalization of beads by Mφ. Initial experiments with IgG- and HSA-coated beads showed that presentation of these proteins was significantly increased when Mφ were incubated with bead bound in contrast with soluble IgG and HSA (Fig. 7,A). Mφ also generated CII epitopes more efficiently when provided with bead-bound hCII, as compared with soluble hCII (Fig. 7 B). Thus, Mφ presentation of CII 259–273 is increased when CII is available in a particulate form.
Human DC can present CII 259–273 and HCgp39 263–275 after incubation with SF from patients with RA
To determine whether hCII and HCgp39 are present in RA SF in forms accessible to APC for Ag processing and DR4-mediated presentation, normal donor DRB1*0401 DC were incubated with cell-free RA SF, before LPS maturation and incubation with HCgp39- or CII-specific T cell hybrids. DC pulsed with SF (50% v/v of culture medium) efficiently presented the CII 259–273 epitope to T cell hybrids (Fig. 8,A). The superior ability of DC to present CII epitopes after incubation with SF, as compared with purified native CII α-chains (Fig. 5), suggests that partially degraded CII is present within these SF samples. Partially degraded CII fragments in SF were detected using an antiserum specific for CII neoepitopes that are generated upon MMP cleavage (8) (unpublished data).
DC also presented HCgp39 epitopes after incubation with SF from RA patients (Fig. 8,B). Quantitation of HCgp39 amounts in SF by ELISA showed that SF contains 3–4 μg/ml HCgp39, and immunoblots of SF from RA patients demonstrated that HCgp39 was not degraded (Fig. 8 C). DC were incubated overnight in 25% v/v SF, with a final HCgp39 concentration of 0.7–1 μg/ml. DC presentation of purified HCgp39 (0.7 μg/ml) was superior to presentation of comparable amounts of HCgp39 present in SF, suggesting that these SF may contain other molecules that can depress APC function and/or T cell activation, such as IL-10. An HCgp39-specific T cell response was elicited from low numbers of DC (3–6 × 103) that had been exposed to SF. Addition of SF to paraformaldehyde-fixed DC did not result in HCgp39 presentation, indicating intracellular Ag processing was required for SF proteins (unpublished data).
These data show that SF from RA patients contains HCgp39 and CII in forms amenable to uptake and presentation by normal human DR4+ DC, and that these Ag are present in RA SF in sufficient quantities for presentation by DC. Our results suggest that very few Ag-exposed DC in SF may be required to activate Ag-specific T cells.
Linkage of genes encoding certain HLA-DR alleles with RA suggests that MHCII-mediated presentation of autoantigens is one factor that contributes to disease pathogenesis. Activated MHCII+ DC and Mφ colocalize with CD4+ T cells in the inflamed synovial joints of RA patients. Soluble forms of CII and HCgp39 also are found in RA SF and are targets of immune responses in RA patients and murine models of arthritis. However, despite these observations, little was known about the ability of human DC or Mφ to present epitopes derived from processing of cartilage proteins. The objective of our experiments was to determine whether human DR4+ DC and Mφ mediate MHCII presentation of CII or HCgp39 epitopes. We have shown that human ex vivo differentiated DC and Mφ, exhibiting activated phenotypes similar to synovial joint APC, are capable of generating CII and HCgp39 MHCII epitopes previously defined as immunodominant in mice. These data suggest that processing of these epitopes is not strictly dependent upon APC uniquely conditioned in an inflamed joint environment. The form of the Ag delivered to the APC influenced its presentation by DC and Mφ, perhaps due to variation in optimal mechanisms of Ag internalization and localization within endocytic vesicles, or optimal substrates for proteolytic degradation.
Importantly, SF obtained from RA patients contains soluble forms of HCgp39 and CII in quantities that are readily internalized, degraded, and presented to T cells by DC, indicating that, in vivo, SF DC will most likely display these peptide/MHCII complexes. Experiments with titrated numbers of APC suggest that very few Ag-exposed DC in SF will be required to activate Ag-specific T cells. These data support the hypothesis that CII and HCgp39 are the targets of autoreactive T cell responses and that their presentation by DC and Mφ in vivo contributes to the pathogenesis of human RA.
In RA SF, elevated levels of HCgp39 correlate with the presence of HCgp39-producing cells, including activated Mφ and CD16+ monocytes (15). Human DC efficiently presented the HCgp39 263–275 epitope to T cells after incubation with <1 μg/ml soluble HCgp39. This is in contrast to previous studies with human B-LCL and unfractionated PBMC, which required incubation with greater amounts of purified HCgp39 for stimulation of T cells (21, 39). Additionally, HCgp39-specific T cells were activated in response to low numbers of DC after incubation with SF (25% v/v) from RA patients, suggesting that exposure of low numbers of joint DC to SF in vivo would be sufficient for efficient presentation of HCgp39.
Activated Mφ from some donors presented endogenously synthesized HCgp39, and activated, but not resting, Mφ also presented soluble HCgp39, indicating that both exogenous and endogenous pathways for MHCII presentation of HCgp39 are operational only in activated Mφ. In vivo, Mφ activation by inflammatory cytokines, such as IFN-γ and TNF-α present in RA SF (40), or by LPS or other Toll-like receptor ligands during a bacterial or viral infection, may lead to Mφ presentation of HCgp39.
Synthesis and presentation of HCgp39 by activated Mφ were not detected in all normal donors examined, suggesting that HCgp39 synthesis is regulated by undetermined polymorphic factors. In addition to HLA DRB1 genes, a propensity for elevated HCgp39 production by activated Mφ may predispose individuals to develop RA. Because HCgp39 is synthesized by Mφ during differentiation and after activation by common inflammatory stimuli, and found in both serum and SF, it is not surprising that T cells specific for HCgp39 are present in blood of healthy individuals as well as RA patients (21). Although HCgp39-specific T cells in healthy individuals may be unresponsive due to mechanisms of peripheral tolerance in vivo, a destructive HCgp39-specific T cell response could develop in RA patients due to high levels of HCgp39 synthesis by Mφ in the inflamed synovial joint environment.
Highly purified native CII α-chains were presented poorly by DC, whereas large CNBr fragments of CII were presented as efficiently as peptides, suggesting that extracellular degradation of extracellular matrix could potentiate CII internalization or processing in vivo. Similar results were obtained with murine DC (41, 42). RA synovial tissue and fluid contain pathologically excessive amounts of MMP and their activating enzymes that are produced by synovial cells, including Mφ, upon exposure to IL-1α, IL-1β, or TNF-α (9, 43). MMP-1, MMP-8, and MMP-13 cleave nondenatured CII once within the triple helical domain, generating three-quarter and one-quarter fragments; these cleavages are the rate-limiting step, after which CII is susceptible to digestion by other extracellular proteases (44). Enzymatically active forms of the cysteine proteases cathepsins B, L, and S also are elevated in RA SF and have been shown to cleave collagens (45, 46, 47). Extracellar degradation of Ags by tissue proteases leading to Ag presentation by APC is not unprecedented, because it has been observed in the retina microenvironment (48).
Phagocytosis of bead-conjugated CII improved Mφ presentation of CII 259–273, compared with the poor presentation of soluble native or CNBr fragments of CII. In contrast, soluble native CII was well presented by murine peritoneal Mφ, perhaps due to differences in CII preparations (41, 42). Our data suggest that human Mφ optimally present CII after facilitated internalization by phagocytosis or receptor-mediated endocytosis. In vivo, CII may be internalized via FcR as immune complexes or after phagocytosis of apoptotic CII-producing chondrocytes (49).
Disruption of peripheral self tolerance leading to tissue-specific autoimmunity may occur by several mechanisms involving Ag presentation (50). APC that are recruited and activated in response to proinflammatory cytokines in joints may display epitopes of cartilage proteins that are qualitatively or quantitatively distinct from constitutive epitopes, due to altered proteolytic degradation of Ag by APC. Second, joint-infiltrating APC may be exposed to cartilage-specific Ags that normally are sequestered, but made accessible to APC by inflammation or tissue damage. Our data with CII support this second hypothesis. We show that purified native CII α-chains are not an optimal substrate for the MHCII-processing pathway, indicating that APC exposure to native CII in cartilage would not lead to significant display of CII epitopes. Although constitutively present in small amounts due to normal protein turnover, degraded CII is greatly increased upon cytokine induction of MMP. Thus, partial extracellular degradation of CII makes accessible the optimal substrate for DC internalization or processing, leading to display of CII epitopes in quantities potentially sufficient for activation of naive T cells. Moreover, we have shown that DC presentation of CII present in SF is superior to presentation of native CII, consistent with the presence of degraded CII fragments in SF. Such altered Ag presentation capability, coupled with acquisition of T cell costimulatory function (22), suggests a critical role for DC in the disruption of T cell self tolerance to CII.
Together with studies of the specificity of human T cells in arthritic joints (4, 5, 6, 51), the study of MHCII-mediated autoantigen presentation by DC and Mφ will aid in the definition of human RA autoantigens. T cell hybrids specific for CII and HCgp39 will be useful for detecting naturally processed CII and HCgp39 epitopes generated in vivo and presented by ex vivo APC isolated from synovial joints of RA patients. These studies provide a mechanism for how different populations of MHCII+ APC may contribute to the autoimmune response during RA, and increase our understanding of the role of these two autoantigens in RA immunopathology.
We thank Leticia Cano, Jennifer DiCesare, and Stuart Mackenzie for their expert technical advice and assistance. We thank Dr. Dennis Zaller for kindly providing the DR4-transgenic mice, and Dr. Jose Alberola-Ila for helpful discussions.
E.C.T. was supported by a National Institutes of Health National Research Service Award. S.K. was supported by grants from the Arthritis Foundation (Investigator Award), the Southern California chapter of the Arthritis Foundation, and the Arthritis National Research Foundation. G.R.D. was supported in part by the Nemours Foundation and a grant from the National Institutes of Health, National Institute of Arthritis and Musculoskeletal and Skin Diseases (AR-45404).
Abbreviations used in this paper: RA, rheumatoid arthritis; B-LCL, B-lymphoblastoid cell line; CII, type II collagen; CNBr or CB, cyanogen bromide; DC, dendritic cell; hCII, human CII; HCgp39, human cartilage gp39; HSA, human serum albumin; Mφ, macrophage; MMP, matrix metalloproteinase; SF, synovial fluid.