Human Monoclonal Rheumatoid Synovial B Lymphocyte Hybridoma with a New Disease-Related Specificity for Cartilage Oligomeric Matrix Protein¹

Rheumatoid arthritis (RA)⁴ is an autoimmune disease in which a complex interaction among synovial macrophages, fibroblasts, T cells, and B cells generates a coordinated immune attack on the integrity of the joints. Experimental data indicate that Ags trigger an oligoclonal expansion of T cells (1, 2) and the local affinity maturation of B cells in the RA synovium (3, 4, 5, 6). Cartilage-specific Ags such as collagen types II, IX, and XI or proteoglycans (7, 8) have been considered as Ags that could continuously fuel autoimmune responses directed to the joints. The support for their potential role in the pathogenesis of autoimmune arthritis is mainly derived from experimental models of inflammatory joint disease. Nevertheless, the nature of Ags inducing intrasynovial T and B cell expansion in RA in humans is enigmatic, and to date no single joint-specific self-Ag that gives rise to the arthritogenic immune response has been identified.

One possible way to characterize tissue-specific pathogenic Ags (arthritogenic Ags) is to define the reactivity of rheumatoid B cell hybridomas with IgV gene characteristics of Ag activated B cells. Using this approach our group was able more recently to identify a mitochondrial Ag possibly involved in the pathogenesis of RA (9).

In the present study the human synovial B cell hybridoma ELB13/3-56 (4) was analyzed for its specific recognition of cartilage Ags. A heptameric peptide of cartilage oligomeric matrix protein (COMP) could be defined as the target structure of the somatically mutated IgG produced by the synovial hybridoma ELB13/3-56. Moreover serum IgGs specific for the COMP peptide were detectable at statistically significant higher levels in a cohort of RA compared with osteoarthritis (OA) patients, suggesting that this specificity is RA associated and could be used for the development of new diagnostic tools.

Synovial tissues were obtained at synovectomy and arthroplasty from patients (n = 5) with confirmed seropositive RA (10) and from patients (n = 5) with definite idiopathic OA (11). Human hyaline cartilage of normal joints (knee) were obtained from autopsy of patients (n = 3) without joint diseases. Serum samples were obtained from 22 RA patients with confirmed seropositive RA (disease duration between 1.5 and 7.5 yr), 24 patients with definite idiopathic OA, and 20 age-matched healthy controls. All RA patients were receiving antirheumatic medication (gold, methotrexate, and sulfasalazine).

In all cases representative parts of the synovial tissue (∼50%) were fixed in formalin and embedded in paraffin (Giemsa, hematoxylin/eosin staining) to be used for diagnosis and scoring of the degree of inflammation (inflammation score according to Krenn et al. (4)). Molecular analysis was exclusively performed on tissue material exhibiting unequivocal macroscopic signs of inflammation. All patients gave their informed consent, and the ethics review committee of the University of Wurzburg approved the study.

Histopathological evaluation exhibited in all cases (n = 4) the characteristic morphology of long-standing RA with villous hypertrophy of synovial tissue with marked edema, enlargement of synovial lining, and a variable degree of inflammatory infiltration ranging from 3 to 5 according to the inflammatory score reported by Krenn et al. (4, 12).

Histopathological evaluation exhibited a moderate villous hypertrophy with moderate enlargement of synovial lining, moderate fibrosis, moderate incorporation of cartilage fragments, and variable degrees of inflammatory infiltration of 1 and 3 according to the inflammatory score reported by Krenn et al. (12).

The human monoclonal rheumatoid synovial B cell hybridoma ELB13/3-56 was established in our laboratory as previously described (4).

The hybridoma supernatant (250 ml) was loaded on a 1-ml HiTrap rProtein A column (Pharmacia Biotech) following the supplier’s instructions. After a thorough washing procedure (20 mM sodium-phosphate buffer, pH 7), the bound Ab was eluted in 3 column vol of 100 mM sodium citrate buffer, pH 4.0, and collected in 1-ml aliquots. The concentration of purified Ab was determined by the Bradford protein quantification methods using the Rotiquant system (Roth, Karlsruhe, Germany).

Genomic DNA was prepared as previously described (6). Briefly, 10- × 2-μm cryosections of synovial membrane were incubated at 56°C for 1 h with proteinase K (Roche, Mannheim, Germany), which was inactivated by heating at 95°C. The amplification of V_H genes was conducted in a 25-μl volume containing 1.75 mM MgCl₂, 0.4 pM primer, 1 U of Taq polymerase (MBI Fermentas, St. Leon-Rot, Germany), and 200 μM of each dNTP. The cycle profile for amplification consisted of DNA denaturation at 94°C for 90 s, followed by 40 cycles of 94°C for 90 s, primer annealing at 62°C for 60 s, and extension at 72°C for 90 s. The following primers specific for the V_H genes of ELB13/3-56 (4) were used (given in the 5′-3′ direction): ELB-forward, 5′-GAC CCT GTC CCT CAC CTG C(AG)C TGT C-3′; and ELB-reverse, 5′-GTA GAC AAA ATA ACT CCC CGA ATT AAA TG-3′.

PCR products were separated by electrophoresis in 2% agarose gels (Roth, Karlsruhe, Germany) and subsequently extracted using a High-Pure DNA gel extraction kit (Roche). Cloning of PCR fragments was performed with pGEM-T Easy Cloning kit (Promega, Mannheim, Germany). Positive clones were sequenced using the DyeDeoxy Termination Cycle Sequencing Kit (Applied Biosystems, Weiterstadt, Germany) and were analyzed by an automated DNA sequencer (ABI Prism 373). Both strands were sequenced using T7 primers. The sequences were analyzed using DNAMan for Windows (Lynon Software, Montreal, Canada), GenBank, and v-Base databases.

Indirect immunoperoxidase staining was conducted on 7-μm cryosections of human cartilage (knee) as previously described (12) using 100 μl of ELB13/3-56 culture supernatant as primary Ab in a 1/50 dilution. Negative controls were always run in parallel by replacing the primary mAb with PBS.

For the immunoblotting experiments either 50 μl of cartilage extracts or purified COMP were diluted in sample buffer (containing 50 mM Tris-Cl (pH 6.8) and 2% SDS) under reducing (2% (v/v) 2-ME) and nonreducing conditions and applied to 12% polyacrylamide gels and run according to the method of Laemmli (13). The samples were transferred to a nitrocellulose membrane by electrophoresis in a semidry transfer cell (Bio-Rad, Munich, Germany) according to the method of Towbin et al. (14). The development of Ag-specific staining was performed using either the ELB13/3-56 Ab or a polyclonal rabbit anti-bovine COMP Ab (positive control) or an isotype-matched human rheumatoid mAb with specificity for a mitochondrial Ag (negative control) (9), a peroxidase-coupled secondary Ab (Dako, Hamburg, Germany), and chemiluminescence enhancement (Pierce, Rockford, IL).

Fifty micrograms of the purified ELB13/3-56 Ab was incubated overnight at 48°C with 200 μg of cartilage extract, 10 μl of protein A agarose (Bio-Rad), 400 μl of IP buffer (containing 1% Triton X-100, 150 mM NaCl, and 50 mM Tris-Cl, pH 8.0), and 500 μl of distilled water. After centrifugation for 30 s at 13,000 rpm, the supernatant was discarded, and the pellet was washed three times with IP washing buffer (containing 0.1% Triton X-100, 150 mM NaCl, and 50 mM Tris-Cl, pH 8.0). Then the pellet was resuspended in 50 μl of reducing sample buffer (containing 50 mM Tris-Cl (pH 6.8), 2% SDS, and 2% 2-ME) and incubated for 5 min at 95°C. After centrifuging for 1 min at 13,000 rpm, the supernatant was transferred into a fresh tube.

The sample obtained by IP was run in a 12% polyacrylamide gel according to the method of Laemmli (13). The gel was silver stained using a modification of the method described by Shevchenko et al. (15) to reduce the background staining and to enhance the staining of smaller bands. Briefly, the gel was fixed for 60 min in a 25% methanol/25% ethanol/5% C₂H₃NaO₂ solution followed by two washing steps, first with 25% methanol/25% ethanol and second with distilled water for 30 min, respectively. Subsequently, the gel was sensitized for 5 min with 0.02% Na₂S₂O₃ and washed twice for 15 min each time in distilled water. The staining procedure was performed in 0.1% AgNO₃/0.02% formaldehyde for 30 min. Following an additional washing step with distilled water for 15 min, the bands were developed using 2% Na₂CO₃/0.04% formaldehyde. Finally, the reaction was stopped in distilled water. The whole procedure was conducted at room temperature under gentle shaking. The specific band was cut off and stored at 4°C in distilled water until sequencing.

The protein was identified by nanoelectrospray mass spectrometry (16) after tryptic digestion as previously described (17). The peptide mixture was dried down and dissolved for mass spectrometric work in 2.5 μl of methanol/water/formic acid (50/49/1, v/v/v).

The mass spectrometry measurements were performed with a Q-Tof (Micromass, Manchester, U.K.) equipped with a nanoflow Z spray ion source. To identify the protein we used the sequence tag program, which combines partial manual spectrum interpretation of about three amino acids (sequence tag) with the residual mass N- and C-terminals of the interpreted region and the peptide mass to search in a nonredundant translated nucleotide database (18).

The PhD-7 phage display peptide library kit (New England Biolabs, Beverley, MA) was used for mapping the epitope(s) recognized by ELB13/3-56. Two rounds of biopanning were conducted according to the supplier’s instructions. Briefly, petri dishes were incubated overnight at 48°C with 100 mg/ml of ELB13/3-56, and after being blocked with the blocking solution (0.1 M NaHCO₃, 5 mg/ml BSA, and 0.02% NaN₃) they were incubated with 2 × 10¹¹ phage from the original library. The bound phages were then eluted with 0.2 M glycine-HCl (pH 2.2) and amplified in the Escherichia coli strain ER2537. With the amplified phage a new round of biopanning was conducted. At the end of this second round the bound phage were transfected into the ER2537 E. coli, and the heptameric peptide was determined by automated cycle sequencing (ABI Prism 373, Applied Biosystems). The consensus peptide sequence was compared with the entire COMP sequence to determine the location of the epitope within the protein.

Ninety-six-well plates were coated overnight at room temperature in a dark, humid chamber with 100 μl/well of the heptameric COMP peptide (KDPRNVG) in PBS (100 μg/ml) and were subsequently blocked with RPMI medium. For quantification of COMP-specific IgG, serum samples derived from RA (n = 22) and OA (n = 24) patients or healthy controls (n = 20) were added to the COMP peptide-coated microtiter wells at a standard dilution in PBS (200 μg/ml) for 1 h at room temperature. Ab binding was detected using an HRP-conjugated rabbit anti-human IgG (Dako, Copenhagen, Denmark) and substrate solution (4 OPD-Tabs (Dako), 14 ml of citrate buffer (pH 2), and 25 μl of H₂O₂). The color development was stopped after 10 min by the addition of 50 μl of 3 M H₂SO₄. Finally, the absorbance was determined at 490 nm in a microplate reader (model 550, Bio-Rad). The same procedure was repeated using purified ELB13/3-56 in plates coated either with the proposed COMP epitope or with the heptameric control peptide PVGNDRK. The control heptamer was chosen for the same composition but a different order of amino acid residues compared with the selected COMP peptide to ensure that differences in Ab binding could be exclusively referred to sequence-specific variations in physicochemical properties.

A 96-well plate was coated overnight at room temperature with the COMP-heptamer KDPRNVG (200 μl/well of 0.1 mg/ml in 4 M guanidine-HCl/Na₂CO₃, pH 10) and blocked with 2 mg/ml BSA preceded and followed by thorough washing procedures with 0.09% NaCl/0.05% Tween 20. On a second BSA-blocked 96-well plate, 110 μl of ELB13/3-56 or RA sera (diluted 1/400 in 4% Triton/10 mM NaH₂PO₄, pH 7.4) were preincubated with 110 μl of purified COMP at different concentrations (from 0–3.12 μg/ml PBS) overnight. Following the overnight preincubation period the content of the wells (200 μl) was transferred to the COMP heptamer-coated plate and incubated for 1 h. Subsequently, the assay procedure for detection of the peptide-bound Abs concerning washings, use of secondary Abs, color development, and recording followed the same instructions as those given for the conventional ELISA.

ELB13/3-56 is a hybridoma producing an IgG2λ Ab that carries a high number of somatic mutations with high R/S values in the complementarity-determining regions, indicating that the IgVH genes have undergone an Ag-induced affinity maturation (4). We designed primers specific for ELB13/3-56 heavy chain genes and performed PCR amplifications with the synovial membrane of RA and OA patients.

From the tissue of all five RA patients, but from none of the respective OA-derived material, we were able to amplify heavy chain genes highly homologous (98 ± 1%) to the ELB13/3-56 sequences (Fig. 1). This indicates that B cells closely related to ELB13/3-56 and eventually of the same specificity are frequently present in the inflamed synovial membrane of RA patients in a disease-specific manner.

Binding of ELB13/3-56 to the human hyaline cartilage could be demonstrated in vitro by an immunohistologic analysis that revealed a strong and diffuse staining pattern in the interterritorial matrix between the chondrons. In the chondrocytes and the pericellular/perichondroid matrexes no staining could be detected (Fig. 2,a). This staining pattern is in accordance with the immunohistochemical data obtained by DiCesare et al. (19) using a rabbit antiserum specific for human COMP. The specificity of the immunohistochemical results obtained with the ELB13/3-56 in our study was proven by the absence of any staining in the respective negative controls (Fig. 2 b).

Immunoblotting experiments of cartilage extracts under reducing and nonreducing conditions revealed the specificity of ELB13/3-56; staining of three distinct bands at 60, 70, and 90 kDa is clearly visible under unreduced conditions. In addition, a smear at 200 kDa and a band at 500 kDa that is considerably weaker in the negative control are apparently related to the immunoreactivity of ELB13/3-56 (data not shown). The staining pattern of ELB13/3-56 is reminiscent of the electrophoretic mobility of the COMP protein under nonreducing conditions as described by Neidhart et al. (20). Whereas the 200-kDa signal corresponds to the characteristic oligomeric smear obtained for COMP, the 500-kDa band could represent the pentameric form of this extracellular matrix protein. The bands between 60 and 90 kDa are within the range of electrophoretic mobility that has been described for the α and β bands and for a low m.w. fragment of COMP (20). Under reducing conditions no specific bands were detectable.

Although immunoblotting experiments suggested binding of ELB13/3-56 to COMP, definite experimental proof of Ag specificity was still missing. Therefore, cartilage extracts were immunoprecipitated with ELB13/3-56 resulting in a 40-kDa band that was made visible by silver staining (Fig. 3,A) and was further analyzed by sequencing using nanospray tandem mass spectrometry (Table I). The sequence information from tryptic fragments of the immunoprecipitated protein revealed identity with COMP (Fig. 3, B and C) in consistency with the earlier immunohistochemical and immunoblot results. A definite proof for the binding of ELB13/3-56 to COMP was obtained by immunoblotting the Ab against purified COMP (Fig. 4).

Subsequent to the identification of Ag specificity of ELB13/3-56 we focused on characterization of the respective epitope by application of a mimotope strategy that was originally described by Geysen et al. (21). This approach is based on the fact that in some cases discontinuous epitopes on proteins can be mimicked by short peptides. Hence peptides can be effectively used to define Ab specificity (22, 23, 24, 25). Therefore, purified ELB13/3-56 was immobilized on petri dishes and repetitively panned against a heptameric phage library. With every round of panning there was an enrichment of sequences giving rise to specific recognition by ELB13/3-56. Table II lists the amino acid sequences of 7-mer peptides with the best binding capacity to ELB13/3-56. All sequences were aligned and yielded a consensus motif XSPPNVP. This consensus motif was most closely related to the COMP sequence KDPRNVG (aa residues 669–675).

Further support for the proposed epitope specificity of ELB13/3-56 was derived from more efficient binding of the mAb to the synthetic COMP peptide (KDPRNVG) than to a control heptamer (PVGNDRK) with an identical composition but a different order of amino acid residues as shown in Fig. 5. This clearly indicates that mAb binding is critically dependent on sequence specific physicochemical characteristics of the identified COMP peptide.

The identification of a heptameric COMP peptide as an epitope of a human synovial B cell hybridoma led us to determine whether its recognition by IgG autoantibodies is possibly disease specific. Thus, serum samples of RA (n = 22) and OA (=24) patients and from healthy donors (n = 20) were tested on ELISA plates coated with the COMP heptamer. As shown in Fig. 6 the serum samples taken from RA patients bound with significantly higher efficiency to the COMP heptamer than the OA sera (p < 1 × 10⁻⁴, by Students t test) or healthy control sera (p < 1 × 10⁻⁴, by Student’s t test), indicating a disease-specific autoantibody response. The specificity of the circulating autoantibodies in RA patients for the heptameric COMP epitope was further supported by the inhibition of IgG binding to COMP protein-coated microtiter plates by the respective synthetic peptide as shown for representative serum samples and the mAb Elb13/3 56 in Fig. 7.

The destruction of joint cartilage and tendon is a key feature of RA. Histopathology and molecular analysis of lymphocyte receptors for clonality indicate that T and B cells are expanded in an Ag-dependent manner in the inflamed synovial tissue, and the inflammatory reaction is related to local disease activity (26). However, Ags of pathogenic relevance, especially those of tissue specificity (arthritogenic Ags), have not yet been identified. In the present study a rheumatoid factor-negative human rheumatoid synovial B cell hybridoma was characterized for the specific recognition of COMP, an extracellular matrix protein restricted in its expression to cartilage and tendons (19, 27). Hence, this human COMP-specific Ab targets structures in the joint that are preferentially affected by the inflammatory process in RA and can therefore be considered a prototype of a tissue-specific autoantibody.

The success of the epitope-mapping strategy for the ELB13/3-56 mAb in the present study stresses the usefulness of phage libraries displaying small peptides as powerful tools to investigate Ab fine specificities. The usefulness of this approach has been proven by other groups, e.g., Ditzel et al. (28), who applied a phage display library strategy as an alternative to conventional epitope mapping for the study of a recombinant form of the HIV type 1 surface glycoprotein gp120. Another example of the potential of the phage technology is the characterization of binding sites for four mAbs on SHP-1, an SH2 domain-containing protein tyrosine phosphatase, by Murthy et al. (29). Phage display libraries have even been used to mimic microbial polysaccharides, thus allowing the production of Abs to carbohydrate epitopes (30).

However, the delineation of a short peptide consensus sequences and the deduced identification of a heptameric COMP epitope raise the question of whether the identified motif represents the entire epitope or, instead, a continuous part within a discontinuous determinant. X-ray analysis of Ab-Ag complexes by Barlow et al. (31) estimated that most Abs are probably raised against discontinuous determinants consisting of 15–22 residues on several surface loops. In contrast, semiquantitative estimations of Gibbs free energy changes by Novotny et al. (32) have predicted that few of the determinant residues contribute actively to the binding energetics, and the surrounding residues allow structural complementarity. In addition to these physicochemical considerations, which are in line with our proposal that the identified COMP heptamer might represent a part of an epitope accessible on the surface of the native COMP protein, are several precedent studies that came to similar conclusions. Thus, Geysen et al. identified Ab binding sites on the myohemerythrin using linear peptide homologues of the protein sequence (33). In this study the epitopes could be preferentially localized in exposed areas of convex surface shape and often negative electrostatic potential by comparison of the mapping results with the known three-dimensional protein structure (33). Another group studying rheumatoid factors revealed that the Ab bound continuous epitopes. These epitopes did not exhibit sequence homology, but showed a conservation in the modeled secondary structures, leading to the conclusion that the main chain atoms rather than the side chains contributed most directly to the conformational similarities between the recognized peptides (34). An additional example of a successful application of a linear peptide approach in a mapping strategy of conformational epitopes is the thorough analysis of anti-Ro/SSA-specific autoantibodies characteristically occurring in systemic lupus erythematosus and Sjögren’s syndrome (35, 36, 37). We regard these precedent investigations as strong support for our hypothesis that the identified heptameric COMP peptide mimics structures also involved in Ab recognition of the native COMP protein. Moreover, the potency of the linear synthetic COMP peptide to almost completely inhibit IgG binding to the purified COMP protein in our study (Fig. 7) provides additional direct experimental evidence of the accessibility of a similar continuous binding site at the surface of the native protein.

In RA, germinal center-like structures are detectable in the inflamed synovial tissue, suggesting a locally generated B cell response to tissue-specific (auto)Ags (38, 39). In this respect characterization of a COMP-specific synovial B cell hybridoma provides the first experimental evidence of the nature of the Ag(s) possibly involved. The B cell hybridoma is specific for a tendon- and cartilage-specific Ag and carries somatically mutated IgV_H genes with high R/S values in the complementarity-determining regions, characteristics of Ag-driven affinity maturation (38, 40).

Moreover, the clonotypic sequences of the COMP-specific hybridoma could be detected in synovectomy material derived from RA patients, but in none of the investigated OA cases. This indicates that the occurrence of B cells with somatically muted V_H genes nearly identical with those coding for a COMP-specific Ag receptor are characteristic of the chronic synovitis of RA. Hence, B cell activation in RA and OA follow different patterns (12). The vigor of the local immune response to COMP in the joints is also reflected by the detection of circulating IgG autoantibodies with specificity for the identified heptameric COMP determinant in the serum of RA patients with different disease durations.

Further evidence of a potential role of COMP as an arthritogenic autoantigen in RA is derived from recent animal experiments. Thus, COMP has been shown to induce arthritogenic immune responses in rodents (8). The autoimmune arthritis induced by immunization with COMP in rats shares some striking clinical features with RA, e.g., polyarthritic involvement and histopathologic characteristics.

Autoimmunity to COMP could contribute to the pathogenesis of RA in keeping active the inflammatory process in the joints. In this respect synovial B cells could be crucial for the chronicity of the arthritic process in RA similar to their role in experimental disease models in rodents (7). Especially the local production of complement-fixing autoantibodies to cartilage components (e.g., the IgG2λ monospecific ELB13/3-56) such as COMP could lead to a continuous immune complex formation with the cartilage matrix, thereby attracting polymorphonuclear leukocytes and macrophages. The resulting engulfment of the immune complexes is inevitably accompanied by proteolytic damage of the cartilage matrix, as, for example, matrix metalloproteinases 19 and 20 that cleave aggrecan and COMP (41). Finally, the liberation of tissue-specific autoantigens (e.g., collagens, aggrecan, or COMP) from the injured cartilage could fuel the vicious cycle, leading to the perpetuation and amplification of RA joint inflammation.

We thank K. Born for help preparing the manuscript, and E. Wozniak and E. Schmitt for expert technical assistance. This work received the approval of the University of Wurzburg ethics committee.