Generation of neoantigenic epitopes after posttranslational modification of type II collagen by factors present within the inflamed joint

Abstract

Objective

Collagen-induced arthritis is a commonly accepted model of rheumatoid arthritis (RA). However, it has been difficult to substantiate the involvement of autoimmunity to type II collagen (CII) in the pathogenesis of RA. The aim of this investigation was to determine if CII, modified by reactive oxidant species present within the inflamed joint, could generate neoantigenic epitopes.

Methods

Oxidants that play a role in acute and chronic inflammation and are present in the rheumatoid joint (hydroxyl radical, hypochlorous acid, and peroxynitrite) were used for modification of native CII. In addition, CII was glycated with ribose, since nonenzymatic oxidative reactions by glycation are evident in RA. Modifications were analyzed by sodium dodecyl sulfate–polyacrylamide gel electrophoresis and 3-dimensional fluorescence followed by enzyme-linked immunosorbent assay (ELISA) and Western blotting, using, as probes, sera from patients with RA and from patients with other inflammatory and noninflammatory joint diseases.

Results

Only 1 RA serum sample showed strong binding to native CII. In contrast, binding to modified CII was increased in 14 of 31 RA sera, of which 7 were strong binders and 7 were moderate binders. Among the non-RA serum samples, only 1 yielded a strong reaction to modified CII and 5 of 41 were moderate binders. Samples that showed the strongest binding to modified CII in ELISA also showed strong binding to various fragmented or aggregated forms of CII in Western blots, as well as strong binding to fragmented CII present in RA synovial fluid.

Conclusion

When modified by conditions found within the inflamed joint, CII acts as an autoantigen in RA.

Rheumatoid arthritis (RA) is an autoimmune disease characterized by chronic inflammation of the joints, which is associated with synovitis and erosion of the cartilage and bone. This damage has been linked to the action of proinflammatory cytokines (1), free radicals (2), and matrix metalloproteinases (MMPs) (3). Inflammatory cells such as T and B cells, macrophages, and neutrophils infiltrate the inflamed synovial membrane. These cells consume increased amounts of oxygen, resulting in the generation of reactive oxidants such as superoxide radical (O), hydrogen peroxide (H₂O₂), hydroxyl radical (^·OH), hypochlorous acid (HOCl), and nitric oxide (NO^·), which are, in turn, involved in acute and chronic inflammation (4, 5).

The fast reaction between O and NO^· forms peroxynitrite (ONOO⁻), which inhibits the activity of tissue inhibitor of metalloproteinases 1 (6), resulting in elevated activities of MMP-1 (7) and MMP-3 (8). In addition, oxidative stress is associated with sequential oxidative reactions, which generate advanced glycation end products (AGEs) that have damaging effects on proteins (9). This nonenzymatic glycation is evident in RA, despite the absence of hyperglycemia (10). Nonenzymatic glycation has been demonstrated by an increase in the level of AGE compounds such as pentosidine and carboxymethyllysine in the serum, synovial fluid, and urine of RA patients (11, 12).

The role of chemical and enzymatic posttranslational modifications in the acquisition of neoantigenicity by type II collagen (CII) is unknown. Although CII is the principal component of human articular cartilage and one of the most-characterized autoantigens in RA (13–16), autoimmune reactivity against CII modified by oxidative reactions has not yet been reported. However, autoantibodies against IgG-AGE (10) and a T cell response against IgG modified by HOCl and peroxynitrite (17) have been observed.

Distinct from nonenzymatic posttranslational modification, the relevance of enzymatic posttranslational modification in modulating the immune response to CII was demonstrated by observations in collagen-induced arthritis (CIA), a model of RA that is induced in a susceptible strain of mice (DBA/1) following intradermal immunization with CII emulsified in adjuvant. In this murine model, most anti-CII CD4+ T cells specifically recognize a glycosylated peptide of CII (amino acids 256–270) (18). Recently, much attention has been given to another enzymatic posttranslational modification involving peptidylarginine deiminase (PAD), an enzyme that catalyzes the conversion of arginine to citrulline (19). The most-characterized protein target for PAD is the non–joint-specific antigen, filaggrin. In fact, the presence of antibodies against deiminated filaggrin has become an important diagnostic measure in patients with RA (20). Moreover, Burkhardt and coworkers recently reported the occurrence of a humoral response to a citrullinated CII–immunodominant epitope encompassing amino acids 359–369 (21).

In the DBA/1 mouse model of RA, the 2 arms of adaptive immunity, T and B cells, contribute to the pathogenesis of CIA, but their relative importance in priming of both immune activation and joint destruction is still unclear (for review, see ref. 22). With regard to the immune responses to CII in RA, there is no clear consensus as to whether native CII is either a target for T cells or a target for autoantibodies (23–27). Although antibodies to native CII and CII-responsive T cells have been found in RA patients, their clinical significance is unclear. These autoantibodies may be directly related to the pathogenesis of the disease or could represent a bystander effect induced by chronic inflammation (28). It has therefore been difficult to substantiate the involvement of CII in the immunopathogenesis of RA.

In the present study, we tested the reactivity of RA sera to CII that had been chemically modified in vitro. Specifically, we investigated the effect of chemical posttranslational modifications by reactive oxidants in inducing neoantigenicity within CII.

PATIENTS AND METHODS

Chemical modification of CII.

Bovine CII was prepared from bovine cartilage using the method described by Miller (29). Bovine CII (2 mg/ml) in phosphate buffered saline (PBS) was incubated overnight at 37°C with the following systems for generating reactive oxygen species: 1) 1 mM CuCl₂ and 2 mM H₂O₂ (Fenton reaction [30]), which was used to modify CII with hydroxyl radical; 2) 1 mM HOCl (31, 32) (BDH Chemicals, Oxford, UK); 3) 2 mM peroxynitrite (supplied as a 200-mM stock solution dissolved in 4.7% NaOH; Calbiochem, Beeston, Notts, UK) with 200 mM phosphate buffer, pH 7.2, added to adjust the pH to 7.2 (a control CII sample in 0.047% NaOH with 200 mM phosphate buffer, pH 7.2, was set up in parallel); and 4) glycation by incubating CII with 2M ribose (Sigma, Dorset, UK) (33). Bovine serum albumin (BSA; Sigma) was also modified as above and was used as control antigen.

Modification of CII was monitored by 7.5% reducing sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE) followed by staining with Coomassie (Sigma) and silver (Pierce, Tattenhall, Cheshire, UK). We also obtained fluorescence spectra of modified CII samples. Three-dimensional (3-D) scanning fluorescence spectra were obtained using a Hitachi F-4500 spectrofluorometer (Tokyo, Japan). Samples were briefly centrifuged prior to scanning, to remove aggregated material. Simultaneous excitation (250–450 nm) and emission (250–600 nm) spectra were recorded.

Serum and synovial fluid samples.

The serum samples used were from the Guy's Hospital rheumatology serum bank. Thirty-one of the serum samples were from patients with RA (34). As controls, we used samples from patients with other inflammatory joint diseases (n = 23) as well as patients with back pain (n = 7), osteoporosis (n = 8), and gout (n = 2). Samples were collected from patients of varying ages whose disease was in various stages and who were receiving different treatments. Synovial fluid was collected from 6 RA patients and 1 patient with osteoarthritis (OA). The age range was 65–93 years, and patients were equally distributed between those who were rheumatoid factor (RF) seronegative and those who were RF seropositive.

Approval was granted by the appropriate local ethics committee (Guy's Research Ethics Committee) for the taking and storing of serum from the patients included in this study. Patients' consent was obtained before the collection of samples.

CIA serum samples were collected from male DBA/1 mice injected with 100 μl of 2 mg/ml bovine CII mixed 1:1 in Freund's complete adjuvant. Mice developed CIA within 3 weeks after the injection, and were killed after 8 weeks at the time when serum samples were collected (13).

Enzyme-linked immunosorbent assay (ELISA).

An ELISA was performed using the modified and native CII as targets. Briefly, ELISA plates (Nunc, London, UK) were coated with 10 μg/ml of modified or native protein in PBS for incubation at 4°C overnight. Plates were then washed 3 times with PBS. After blocking for 2 hours with 2% Marvel in PBS, 100 μl of 1:200-diluted serum samples in 2% Marvel–PBS was added to each well, followed by a 2-hour incubation at 37°C. Plates were then washed with PBS plus 0.1% Tween, followed by 3 washes with PBS. Anti-human IgG–horseradish peroxidase (HRP) (Sigma) was then added at 1:1,000 dilution in 2% Marvel–PBS for another 2-hour incubation. The ELISA plates were washed, and 100 μg/ml 3,3′,5,5′-tetramethylbenzidine substrate (Sigma) in 100 mM sodium acetate, pH 6.0, was added. Subsequently, the reaction was stopped with 1M sulfuric acid.

The optical density (OD) was measured at 450 nm using a GENios plate reader (TECAN, Theale Court, Reading, UK) and Magellan software (TECAN). The ELISA that was used to investigate serum samples from DBA/1 mice with CIA was carried out in a similar manner, except that serum samples were diluted 1:2,000 and binding was detected by anti-mouse HRP conjugate (Sigma).

Western blotting.

Modified CII as target antigen.

Modified and native CII (2 μg of each) were run on a 7.5% denaturing SDS gel and electroblotted onto a nitrocellulose membrane (BDH Chemicals). After blocking with 10% Marvel–0.1% Tween–PBS, membranes were incubated with a 1:200 dilution of serum samples in 10% Marvel–0.1% Tween–PBS for 2 hours at room temperature, followed by incubation with anti-human IgG-HRP (Sigma) reagents. Membranes were washed extensively with 0.1% Tween–PBS before development with enhanced chemiluminescence (Pharmacia, Milton Keynes, UK).

Synovial fluid as target.

Synovial fluid was incubated with protein A–Sepharose for 1 hour at room temperature, to exclude IgG contamination. This was followed by treatment with 1% SDS for 30 minutes at room temperature. Samples were then mixed with SDS standard loading buffer, followed by 5 minutes of boiling. Samples were run on reducing 7.5% denaturing SDS–polyacrylamide gels, followed by blotting and probing with serum samples as described above.

Statistical analysis.

Analyses of data were performed using the GraphPad Prism software package (GraphPad, San Diego, CA). Results are expressed as the mean and SD. The Mann-Whitney U test was used to compare the different binding of serum samples to various modified forms of CII.

RESULTS

SDS-PAGE analysis of chemically modified CII.

With SDS-PAGE, native CII showed a major electrophoretic band that migrated to the region below the 150-kd protein marker, as well as traces of higher molecular weight aggregates (Figure 1, lane 5); the major band that was evident at ∼130 kd corresponds to the constituent α-chains of CII (actual molecular mass of ∼95 kd), and the apparent molecular mass of ∼130 kd reflects the well-known slow migration of the collagen α-chains relative to the globular protein markers. Glycation by ribose involved a shift in the position of the CII α-chain band to a region of slightly higher molecular mass, as well as fragmentation to bands in the electrophoretic region of 50–150 kd (Figure 1, lane 1).

Figure 1. Sodium dodecyl sulfate (SDS)–polyacrylamide gel electrophoresis analysis of chemically modified type II collagen (CII). Samples containing CII were incubated overnight at 37°C with or without the various free radical generators. Equal amounts of protein were boiled with loading buffer containing β-mercaptoethanol for 3 minutes before loading on the SDS–polyacrylamide gel and electrophoresis. Lane 1, Glycation by ribose caused a shift in the CII α-chain band to a position of slightly higher molecular weight, as well as fragmentation. Lane 2, Hydroxyl radical caused a loss of the intact CII polypeptide, together with both fragmentation and formation of high molecular weight aggregates that did not migrate from the sample loading slot. Lane 3, Hypochlorous acid treatment resulted in both aggregation and fragmentation. Lane 4, Peroxynitrite treatment resulted in a loss of intact CII polypeptide, together with fragmentation. Lane 5, The α-chain polypeptide of native untreated CII. The position of the molecular weight markers (in kd) is shown on the right.

Exposure of CII to the hydroxyl radical–generating system caused extensive aggregation, with aggregates trapped in the loading slot and in the stacking gel. Loss of intact CII polypeptides and fragmentation into 2 major products in the electrophoretic region of 120–150 kd were also observed (Figure 1, lane 2). Treatment with hypochlorous acid resulted in small CII polypeptide fragments in the region ranging between those of native CII and those at the 25-kd marker, and a higher percentage of crosslinked CII migrating slower than the 250-kd marker, as well as aggregates that were observed in the stacking gel or in the loading slot (Figure 1, lane 3). The degree of fragmentation varied according to the concentration and freshness of HOCl (results not shown). Treatment with peroxynitrite resulted in a loss of intact CII in conjunction with a smear of protein through the entire lane, indicating extensive fragmentation (Figure 1, lane 4).

Fluorescence profile of native and modified CII.

The modification of proteins by reactive oxygen species is known to cause the formation of fluorescent species. We performed a 3-D fluorescence profile study of both native and modified CII (results shown in Figure 2). Some native fluorescence was detectable in the native CII, with a relative fluorescence intensity (RFI) of 18.8 at a maximum excitation (Ex_max) of 326 nm and maximum emission (Em_max) of 386 nm. Modification of CII with hypochlorous acid resulted in only a slight increase in the RFI, to 21, at the same excitation and emission wavelengths. Treatment with ribose caused not only an increase in fluorescence intensity (RFI 33.4), but also a shift in the Ex_max and Em_max to 332 nm and 404 nm, respectively. This is characteristic of pentosidine formation. We detected no fluorescence for CII modified with hydroxyl radical or with peroxynitrite. For all CII samples, extensive light scattering, as indicated by the straight lines on the spectra, was detected; this is attributable to collagen fibers and aggregates that were enhanced after collagen modification, especially with hydroxyl radical– and peroxynitrite-induced changes.

Figure 2. Three-dimensional fluorescence profile of modified and native type II collagen (CII). For native CII, a relative fluorescence intensity (RFI) of 18.8 at a maximum excitation (Ex_max) and maximum emission (Em_max) of 326 nm and 386 nm, respectively, was detected. Modification of CII with HOCl resulted in an increase in the RFI to 21 at the same excitation and emission maxima. The highest value of fluorescence was detected for CII treated with ribose, with an RFI of 33.4 at an Ex_max of 332 nm and Em_max of 404 nm, as a result of pentosidine formation. No fluorescence was detected for CII modified with hydroxyl radicals or with peroxynitrite. Extensive light scattering was detected in all CII samples, as indicated by the straight lines on the spectra, which could be attributed to collagen fibers and aggregates being more extensive after collagen modification.

Antibody binding to modified CII in patients with RA and in control individuals.

Table 1 summarizes the ELISA results for binding to CII in 31 RA and 41 non-RA serum samples. Although there were no significant differences in the age or sex distribution between the 2 patient groups, there was a striking difference in the immunoreactivity toward modified CII. While only 1 RA serum sample (3.2%) showed strong binding to native CII, 45.2% of the RA serum samples (14 of 31) were positive binders to modified CII, of which 22.6% (7 of 31) were considered strong binders. In contrast, in the non-RA group, there was only 1 strong binder to modified CII (2.4%), but no strong binders to native CII. Five other serum samples (2 from patients with back pain, 1 from an OA patient, 1 from a patient diagnosed as having OA and psoriatic arthritis, and 1 from a patient with ankylosing spondylitis) exhibited moderate binding to either HOCl- or peroxynitrite-treated CII (9.7%). Figure 3A shows a representative ELISA profile of 21 RA serum samples and 26 non-RA serum samples.

Table 1. Binding of serum samples to native and modified type II collagen (CII)*

	RA	Non-RA
* Rheumatoid arthritis (RA) and non-RA samples were tested in enzyme-linked immunosorbent assay (ELISA) for their ability to bind to native or modified CII. The non-RA group comprised other inflammatory or noninflammatory joint diseases, including psoriatic arthritis (n = 3), systemic lupus erythematosus (n = 1), ankylosing spondylitis (n = 4), palindromic arthritis (n = 2), scleroderma (n = 1), Behçet's disease (n = 1), primary Sjögren's syndrome (n = 2), fibromyalgia (n = 1), inflammatory arthritis (n = 1), tendinitis (n = 1), reactive arthritis (n = 1), psoriasis (n = 1), osteoarthritis (n = 4), back pain (n = 7), osteoporosis (n = 8), and gout (n = 2). Positive binders were defined as samples with ELISA optical density (OD) values of at least 3 times above background levels. Strong binders to modified CII were samples with 3–5 times higher binding to modified CII than to native CII. Moderate binders were samples that showed 2–3 times more binding to modified than to native CII. † By Mann-Whitney U test.
Age, mean (range) years	60.4 (38–83)	58.7 (37–78)
No. female:no. male	2.4:1	1.62:1
No. strong binders to native CII (mean ± SD OD)	1 (1.14 ± 0.20)	0
Binding to native CII, mean ± SD OD	0.20 ± 0.24	0.113 ± 0.056
No. strong binders to modified CII (mean ± SD OD)	7 (0.83 ± 0.29)	1 (0.89 ± 0.06)
No. moderate binders to modified CII (mean ± SD OD)	7 (0.38 ± 0.04)	5 (0.48 ± 0.09)
Total no. binders to modified CII/total no. subjects	14/31	6/41
P versus binding to native CII†	0.0001	0.0004

Figure 3. Binding of serum samples to modified type II collagen (CII), as assessed by enzyme-linked immunosorbent assay (ELISA). A, ELISA detection of antibodies to modified and native CII in serum samples from patients with rheumatoid arthritis (RA) (n = 21) and patients with other inflammatory joint diseases (non-RA) (n = 26). A significant increase in the binding to modified CII is detected in all RA sera when compared with binding to native CII (∗ = P < 0.0023), except for CII modified by hydroxyl radical (∗∗ = P = 0.19 versus binding to native CII). Box plots show the 25th and 75th percentile, and the horizontal line shows the median (50th percentile). Bars outside the box indicate the minimum and maximum value. B, Representative ELISA pattern demonstrated by 2 RA samples (from patients 33 and 47) and 1 non-RA sample (from patient 67, a patient with back pain), while sample 66 (from a patient with osteoarthritis) is a negative control. Bars show the mean and SD. C, Binding of serum samples from DBA/1 mice with collagen-induced arthritis (CIA). The ELISA was done with CIA serum samples diluted 1:2,000. Binding of murine samples was higher to native CII than to modified CII. O.D. = optical density.

A significant increase in binding to HOCl- and peroxynitrite-modified CII in comparison with binding to native CII was seen by ELISA in all RA sera (P < 0.0001), with a mean ± SD OD of 0.20 ± 0.24, 0.45 ± 0.32, and 0.35 ± 0.21 for binding to native CII, HOCl-modified CII, and peroxynitrite-modified CII, respectively. Glycated CII also showed a small, but significant, increase in binding (OD 0.29 ± 0.30 for binding to glycated CII; P = 0.002 versus binding to native CII). The difference between binding to native CII (OD 0.20 ± 0.24) and that to hydroxyl radical–modified CII (OD 0.17 ± 0.10) was not significant (P = 0.19). Binding of HOCl-modified CII was higher in RA sera than in non-RA sera (OD 0.45 ± 0.32 versus 0.25 ± 0.10; P = 0.01).

Figure 3B shows typical ELISA results. RA sample 33 did not exhibit binding to native CII, but did show strong binding to glycated and HOCl-modified CII, whereas the binding to peroxynitrite-modified CII was more moderate. In RA sample 47, strong binding to both native and modified CII was detected. Patient 66 had OA, and this patient's serum failed to show binding to any form of CII. The same pattern of binding as that shown for RA sample 33 was seen for serum sample 67, which was from a patient with back pain. None of these samples bound to native or modified BSA (typical ELISA OD values were lower than 0.1).

We also assessed serum samples collected from the same RA patient at different time points. For example, RA serum sample 70, which was collected earlier than serum sample 41 from the same patient, did not show binding to native CII (mean ± SD OD 0.21 ± 0.06), but did exhibit binding to HOCl- and peroxynitrite-modified CII (OD 0.77 ± 0.1 and 0.79 ± 0.10, respectively). In serum sample 41, however, neither binding to native CII nor binding to modified CII were detected (OD values were within the range of background values). In general, HOCl modification was the most efficient in generating neoantigens, since binding to HOCl-treated CII was stronger for most serum samples tested, and there were more positive binders than with the other treatments.

Antibody binding to modified CII in sera from mice with CIA.

Since CIA in genetically susceptible strains of mice is a commonly utilized model of RA, we also tested the pattern of binding to modified CII in CIA murine sera. As seen in Figure 3C, binding of CIA serum samples from DBA/1 mice was significantly stronger to native CII than to modified CII, with OD values in the majority of samples that were twice as high for binding to native CII compared with that to the CII modifications (mean ± SD OD 0.67 ± 0.15 for native CII binding versus 0.36 ± 0.14, 0.44 ± 0.11, 0.60 ± 0.15, and 0.37 ± 0.13 for glycated, ^·OH-treated, HOCl-treated, and peroxynitrite-treated CII binding, respectively; P < 0.001 for each comparison, except for P > 0.05 for binding to HOCl-treated CII). The binding to CII modified with HOCl was, in some cases, the same as that to native CII, but higher than that to other modifications of CII. Interestingly, the sera from mice that were injected with CII but did not develop disease exhibited an even lower binding to modified CII (murine samples 6, 8, and 31). For example, sample 31, which was from a mouse that did not develop CIA, showed binding to native CII at a mean ELISA OD of 0.51, compared with mean OD values of 0.10, 0.18, 0.19, and 0.11 for binding to glycated, ^·OH-treated, HOCl-treated, and peroxynitrite-treated CII, respectively.

Analysis of serum reactivity to CII by Western blotting.

We next used Western blotting to test a range of serum samples that exhibited the strongest binding to hypochlorous acid– or peroxynitrite-modified CII in ELISA. The results demonstrated a correlation between strong ELISA binders and samples that showed strong binding to various fragments and aggregates of CII in Western blots. Figure 4 shows the results obtained from 1 RA serum sample (sample 47), using the same modified CII preparation as in Figures 1, 2, and 3B. In the case of glycated CII (Figure 4, lane 1), serum sample 47 showed binding to 1) forms of the CII α-chains with lower mobility in comparison with the native CII bands, 2) a CII fragment in the region of 100 kd, and 3) a range of CII fragments in the region of 25–100 kd. In the case of CII treated with hydroxyl radical (Figure 4, lane 2), serum sample 47 showed binding to high molecular weight aggregates, which were observed in the stacking gel, and to a range of CII fragments in the region of 25–150 kd. Serum sample 47 also exhibited binding of CII modified by HOCl to high molecular weight complexes (higher than 250 kd), as well as to a range of fragments in the region between 25 kd and 150 kd (Figure 4, lane 3). A smear of binding to peroxynitrite-treated CII was also observed (Figure 4, lane 4).

Figure 4. Binding of serum samples to modified type II collagen (CII), as determined by Western blotting. Modified and native bovine CII (2 μg) were run on a 7.5% sodium dodecyl sulfate gel under reducing conditions, blotted onto a nitrocellulose membrane, and probed with a 1:200 dilution of sera followed by incubation with anti-human IgG–horseradish peroxidase reagents. Representative results are shown for the pattern of binding to native and modified CII, using serum sample 47 (lane 1, glycated CII; lane 2, ^·OH-modified CII; lane 3, HOCl-modified CII; lane 4, peroxynitrite-modified CII; lane 5, native CII). Sample 47 showed binding to a range of fragmented CII, after treatment with any of the systems for generating reactive oxygen species, and showed binding to high molecular weight aggregates after the treatment of CII with ^·OH or HOCl. The position of molecular weight markers (in kd) is shown on the right.

In general, patterns of binding to either aggregated or fragmented CII were different from one serum sample to another. Of note, serum samples that had exhibited low binding to CII by ELISA, as well as commercial anti-CII antibodies, displayed binding to only the electrophoretic band that corresponded to the intact native CII α-chain polypeptide.

Binding of serum to synovial fluid proteins.

Serum samples from patients with RA were found to exhibit binding to fragmented CII in synovial fluid. Figure 5A shows the results of binding in synovial fluid from RA patients (lanes 1, 3, and 4), compared with the findings in synovial fluid from an OA patient (lane 2). RA samples 33 and 47, which had previously demonstrated strong serum binding to modified CII (see Figure 3B), were also observed to show strong binding to 2 protein bands in RA synovial fluid, in the molecular weight region of 50 kd, and to an additional band at ∼20 kd. The binding pattern was different in OA synovial fluid, since only the lower band in the region of 50 kd was detected and the 20-kd band was not detected.

Figure 5. Binding of serum to synovial fluid proteins. Synovial fluids were pretreated with protein A–Sepharose and then with 1% sodium dodecyl sulfate (SDS) for 30 minutes at room temperature, mixed with SDS standard loading buffer under reducing conditions, and run on 7.5% SDS–polyacrylamide gels, followed by blotting and probing with serum samples as in Figure 4. A, Lanes 1, 3, and 4, Binding results in synovial fluid samples from rheumatoid arthritis (RA) patients. Lane 2, Results in a sample from a patient with osteoarthritis (OA). RA serum samples 33 and 47 show binding to 2 fragments of type II collagen (CII) in the region of 50 kd and to an additional fragment at 20 kd. In the OA synovial fluid, only the lower band in the region of 50 kd is detected and the 20-kd fragment is not detected. B, An anti-CII single-chain variable fragment (scFv) detects the same pattern of fragments as that detected in samples 33 and 47 for RA synovial fluid samples (lanes 1, 3, and 4). In the OA synovial fluid (lane 2), the pattern is different: the upper band at ∼50 kd, as well as an additional CII fragment at ∼80 kd, are detected with the scFv. C, Serum from a mouse with collagen-induced arthritis recognizes the same 50-kd fragments.

To investigate the possibility that these immunoreactive protein bands correspond to CII, rather than to other synovial fluid proteins, the same synovial fluid samples were probed with a phage display single-chain variable fragment (scFv) specific to CII (Figure 5B). Phage display antibody libraries are made by fusing the antibody gene fragments to the bacteriophage-coat protein genes. As a result of this fusion, a repertoire of antibody phage is built, which subsequently is used to raise specific antibody fragments. The anti–modified CII scFv was selected by panning using HOCl-modified CII as a target for selection, using phage display human scFv. This scFv was observed to bind to modified, but not to native, CII (Nissim A, et al: unpublished observations).

The anti-CII scFv, when used to probe electrophoretically separated RA synovial fluid samples, detected the same pattern of fragments as that detected in serum samples 33 and 47 (Figure 5B, lanes 1, 3, and 4). In the OA synovial fluid (Figure 5B, lane 2), the pattern was different: only the upper band at ∼50 kd as well as an additional CII fragment at ∼80 kd were detected. This 80-kd fragment was detected in RA serum sample 72. Addition of anti-human IgM-HRP to probe the human serum sample did not change the band pattern (results not shown).

Furthermore, a serum sample from a CIA mouse, which showed relatively better binding to modified CII as determined by ELISA (mean OD values of 0.65, 0.57, 0.62, 0.66, and 0.50 for binding to native, glycated, ^·OH-treated, HOCl-treated, and peroxynitrite-treated CII, respectively) than that displayed by the other murine samples, recognized the same 50-kd fragments (Figure 5C). This is consistent with the conclusion that these fragments correspond to fragmented CII present in human synovial fluid. The CIA serum sample also showed binding to the 20-kd fragment, although this cannot be seen clearly in Figure 5C due to the low signal intensity.

DISCUSSION

In RA patients, antibodies to native CII have been reported since the 1970s (35). However, the clinical significance of this observation has been unclear, since the incidence of anti–native CII in patients with RA varies widely (36, 37). Moreover, anti–native CII antibodies have also been found in other diseases and in healthy controls (38). In the present study, a further step was taken to explore the potential involvement of CII in autoimmunity by exposing CII in vitro to protein-modifying factors that are generated during chronic inflammation in the RA joint. We wanted to ascertain whether posttranslational modifications induced in CII could play a role in autoimmunity against CII.

CII was exposed to glycation and oxidation reactions known to be involved in acute and chronic inflammation (4, 5). We monitored the degree of CII modification by both SDS-PAGE and 3-D fluorescence, and observed changes in both protein mobility and fluorescence (Figures 1 and 2, respectively) that were consistent with the findings in earlier reports on CII glycation (33, 39). We also observed fragmentation and aggregation of CII, as has been reported for BSA or CII treated with ^·OH/H₂O₂ (40–42), HOCl (43–45), and ONOO⁻ (46).

Of the 31 samples analyzed, only 3 RA sera (9.7%) showed binding to native CII. One RA serum sample was a strong binder and the other 2 were moderate binders, consistent with previous reports of frequencies of 3% and 88% of anti–native CII binders in RA sera (36, 37). The percentage of samples that exhibited binding increased 4-fold when tested against modified CII, such that there were 7 strong binders and 7 moderate binders (45% of all samples) (Table 1). CII treated with HOCl was the most reactive, followed by CII treated with peroxynitrite, glycation, and hydroxyl radical, respectively (Figures 3A and B). In contrast, only 1 non-RA sample showed strong binding to modified CII.

To eliminate the possibility that binding in the sera could be attributed to either 1) covalent bond formation as a result of a reactive group or groups within the treated CII or 2) specific binding to a given modified amino acid, we tested 2 additional target proteins for binding. None of the serum samples showed binding to native or chemically modified BSA. Moreover, we used an additional known target autoantigen, heavy-chain binding protein (BiP) (47), and performed the same modifications as for CII. In general, binding to native BiP was greater than that to modified BiP, in direct contrast to our results on CII (Corrigall V, et al: unpublished observations). Therefore, the anti–modified CII immune recognition is specific and is directed against new CII conformational epitopes induced by the chemical treatments used.

Surprisingly, the quantitative differences in the binding of RA sera to native CII and that to modified CII were reversed when we tested samples from CIA mice at the terminal stage of the disease. All CIA samples showed stronger binding to native CII (Figure 3C). Nevertheless, the binding of samples from CII-injected mice that did not develop CIA was much lower to modified CII compared with that to native CII. This observation might be explained by the fact that the establishment of CIA in mice has a much shorter time scale than the typical time course of the presentation and diagnosis of RA in humans, such that there is insufficient time for the immune system to respond to the modified CII.

In contrast to the low binding or absence of binding to native CII in RA serum samples, as detected by ELISA, all samples consistently exhibited binding to native CII when assessed by Western blotting. However, those samples that showed binding to modified CII in ELISA bound additional fragmented and aggregated CII bands on Western blots (e.g., sample 47 in Figure 4). We believe that this can be attributed to the fact that proteins in SDS gels are under strong denaturing conditions, and therefore the binding seen in the Western blots would be to denatured epitopes. ELISA is therefore a better method to test total binding to modified proteins, although the conformation of proteins might also be changed partially on binding to plastic. Nevertheless, the importance of obtaining results from Western blotting was to evaluate the distribution of binding epitopes and to observe binding to high molecular mass aggregates, since, in ELISA, we could not be sure that these aggregates would immobilize to plastic.

It could be argued that these findings are of little relevance to the disease of RA itself and are merely an interesting in vitro phenomenon. However, we hypothesize that some of the larger aggregates can be retained within the joint, although it is possible that the smaller CII fragments can diffuse out of the joint, since it is known that the protein permeability of the RA joint is increased (48). We found that RA or OA synovial fluid samples, when subjected to SDS-PAGE and Western blotting with serum samples 33 or 47 (which had shown strong binding to modified CII in vitro) or control phage display scFv selected in vitro against CII, exhibited 2 immunoreactive bands in the region of 50 kd. An additional small fragment of ∼20 kd was also detected. This result suggests that the serum samples were indeed capable of binding to CII modified by the inflammatory process taking place in the joint (Figure 5A).

In conclusion, the present findings support the possibility that chemical modification of self antigens, in RA in particular and in inflammation in general, is the cause of formation of neoepitopes. We propose that the oxidative modification of CII, possibly in combination with proteolysis, creates a CII autoantigen. It is possible that in the inflamed joint, the abnormally high fluxes of reactive oxygen/nitrogen species (4, 5) give rise to chemical reactions that modify CII. Doyle and Mamula (49) speculated that the breakdown of tolerance occurs because antibodies against modified self protein are promiscuous and bind both the modified and unmodified self antigen; this process is commonly characterized as epitope spreading. Our hypothesis, which extends this proposal, is that these posttranslational modifications of CII may contribute to the vicious circle of chronicity by providing additional neoepitopes to which the immune system is not tolerized, resulting in stimulation of the immune response against self antigens.

Acknowledgements

We thank Mrs. G. Adams for providing the serum samples from DBA/1 mice with CIA.