The Major Synovial Targets of the Rheumatoid Arthritis-Specific Antifilaggrin Autoantibodies Are Deiminated Forms of the α- and β-Chains of Fibrin¹

Rheumatoid arthritis (RA)⁴ is the most frequent of human systemic autoimmune diseases. It is characterized by the formation, in synovial membranes, of an inflammatory and invasive tissue, the rheumatoid pannus, capable of eroding adjacent cartilage and bone, subsequently leading to joint destruction. Both cellular and humoral autoimmune mechanisms have been implicated in the still poorly understood initiating and perpetuating events of RA pathogenesis. Reflecting the humoral autoimmune processes occurring during RA is the presence of autoantibodies of diverse specificities in the serum of patients. Among these, antifilaggrin autoantibodies (AFA) are the most disease specific and are increasingly recognized as a useful diagnosis tool (1, 2, 3, 4, 5). AFA were originally described as IgG labeling the cornified layer of rat esophagus epithelium by indirect immunofluorescence and formerly called “anti-keratin Abs” (AKA) (6). The recognized Ag is also present in the cornified layer of human epidermis (7). For the past 10 years, our group has gathered proofs that the epithelial Ags recognized by AKA are all related to (pro)filaggrin, that is either to filaggrin, a protein putatively involved in the aggregation of cytokeratin intermediate filaments, or to its precursor, profilaggrin. In human epidermis, the Ag was identified as a 37- to 40-kDa neutral/acidic isoform (isoelectric point (pI) from 5.8 to 7.4) of filaggrin (8), and, in the rat esophagus epithelium, the Ags were characterized as three (pro)filaggrin-related proteins of 210, 120–90, and 130–60 kDa, respectively (9). Antiperinuclear factor (APF), designating serum IgG that decorate perinuclear granules in the superficial cells of the human buccal mucosa epithelium by indirect immunofluorescence, had also been demonstrated to be specifically associated to RA (10). We showed that APF detect (pro)filaggrin-related proteins of buccal epithelial cells and provided direct immunological evidence of the large identity between AKA and APF that we consequently proposed to name AFA (11).

Further characterization of the AFA-targeted epithelial (pro)-filaggrins recently allowed us to show that they correspond to posttranslationally modified proteins, of which arginine residues are converted into citrulline residues following a deimination mediated by a peptidylarginine deiminase (PAD) (12). Moreover, we and others showed that the citrulline residues are an essential constituent of the (pro)filaggrin epitopes recognized by AFA, but only in the context of some specific (pro)filaggrin amino acid sequences (12, 13). In particular, we demonstrated that deimination of a recombinant human filaggrin by PAD is an absolute prerequisite to its recognition by AFA and identified two citrulline-substituted filaggrin-derived peptides bearing major AFA epitopes (12).

In addition to their close association to the disease, AFA have been shown to appear precociously and even to precede the clinical symptoms of RA (14, 15, 16). Their presence and titer are related to the disease activity and severity (1, 2, 17), their ratio to global IgG is increased in the synovial membrane interstitium with regard to the serum or the synovial fluid, and they are synthesized by plasmocytes of the rheumatoid pannus (18). The above points strongly suggest that AFA are implicated in the pathophysiology of RA. Moreover, the concentration of AFA and the presence of AFA-secreting plasmocytes within the main site of immunopathological activity in RA suggest the presence of a target for the autoantibodies in diseased synovial joints. It also raises the possibility that this so far elusive articular autoantigen may drive the humoral antifilaggrin response. Here, we provide the first report on major autoantigens targeted by AFA in the rheumatoid synovium and give evidence of their identity with deiminated α- and β-chains of fibrin. The pathophysiological implications of these findings are discussed.

Human sera were obtained from 165 patients with RA according to the criteria of the American Rheumatism Association (19), from 24 patients with other inflammatory or noninflammatory rheumatic diseases, and from 10 healthy blood donors. Semiquantitative titration of AFA in sera was performed by indirect immunofluorescence on rat esophagus cryosections (so-called AKA) and by immunoblotting on neutral/acidic filaggrin-enriched epidermis extracts as previously described (2, 5). Sera classified as AFA positive contained autoantibodies detectable by both methods over the threshold corresponding to a diagnostic specificity of 0.95, whereas both reactivities were absent in AFA-negative sera. All non-RA sera were AFA negative. Samples of synovial tissue were obtained from RA patients at wrist synovectomy or arthroplasty. They all showed disease-characteristic lesions as assessed by histological analysis of hematoxylin-eosin-stained sections.

Immunoperoxidase staining experiments were performed on 4-μm tissue sections using the Histostain Plus kit (Zymed, San Francisco, CA) according to the manufacturer’s recommendations. Before analysis with an Ab to modified citrulline (purified rabbit IgG, 0.25 μg/ml) (20), synovial tissue and skin fragments were fixed overnight at 4°C in a Bouin’s solution and successively dehydrated for 24 h in 7.5, 15, and 30% sucrose solutions before embedding in Tissue-Tek medium. Then cryosections were postfixed for 15 min at 4°C with 4% paraformaldehyde and 2.5% glutaraldehyde in PBS and then treated to modify citrulline residues, as previously described (21). A mouse mAb to fibrin(ogen) (311, 5 μg/ml; American Diagnostica, Greenwich, CT) was used to probe synovial tissue using both cryosections from Bouin-fixed fragments and cryosections from snap-frozen fragments postfixed by 100% acetone at −30°C for 15 min. Cryosections of human skin were probed with biotin-conjugated affinity-purified human autoantibodies to deiminated fibrinogen (5 μg/ml; see below). In all cases, the specific Ab was visualized either by the broad-spectrum biotin-conjugated secondary Ab provided in the kit followed by peroxidase-conjugated streptavidin or only by this second step when analyzing the biotin-conjugated autoantibodies. All of the sections were counterstained with hematoxylin. Lastly, unfixed cryosections of rat esophagus were probed with the purified autoantibodies to deiminated fibrinogen (400 μg/ml) and their binding was visualized by indirect immunofluorescence using FITC-conjugated goat Ig Fab to human IgGγ (Biosys, Compiègne, France).

In all cases, negative controls included omission of the specific Ab and use of a species and/or isotype-matched Ab(ies). Moreover, for the Ab to modified citrulline, the citrulline modification step was omitted.

Snap-frozen synovial tissue fragments were sequentially extracted (five times in each buffer) with an Ultra-Turrax homogenizer (T25 basic; IKA Labortechnik, Staufen, Germany) in 6 ml/g of tissue of ice-cold 40 mM Tris-HCl-based buffers (pH 7.4) containing first 150 mM NaCl (TBS extract), then 8 M urea deionized by incubation with a mixed bed resin (AG 501-X8; Bio-Rad, Hercules, CA)-(urea extract), then 8 M deionized urea and 50 mM DTT (urea-DTT extract), and finally 6 M guanidine-HCl (guanidine extract). All buffers were supplemented with 20 mM EDTA, 0.02% sodium azide, 2 μg/ml aprotinin, 10 mM N-ethylmaleimide, and 1 mM PMSF (all from Sigma, St. Louis, MO). The pellet was then extracted twice in 0.5 M acetic acid (acetic acid extract) and separated in two parts. One part was solubilized overnight at 4°C by partial digestion with 0.1 mg/ml pepsin (Sigma) in 0.5 M acetic acid (pepsin digest). The second part was thoroughly rinsed in H₂O, then partially digested overnight at 37°C by 100 U/ml purified bacterial collagenase (Type VII; Sigma) in 100 mM Tris-HCl (pH 7.4) containing 200 mM NaCl and 10 mM CaCl₂ (collagenase digest). After each extraction or digestion, the homogenate was clarified by centrifugation at 4°C for 20 min at 15,000 × g. This sequential extraction method using increasingly powerful denaturing agents and then buffers and enzymes commonly used for the solubilization of extracellular matrix components, such as proteoglycans and collagens, allowed solubilization of all of the intracellular and extracellular materials present in a rheumatoid synovial membrane to be achieved. No pellet was observed at the end of the sequential extraction procedure, which was therefore exhaustive. The urea, urea-DTT, guanidine, and acetic acid extracts and the pepsin partial digest were dialyzed against H₂O before analysis.

Proteins of the various sequential synovial membrane extracts and digests and human fibrinogen were separated by SDS-PAGE on 7.5 or 10% polyacrylamide gels. Human epidermis extracts enriched in neutral/acidic filaggrin and rat esophagus epithelium extracts containing the AFA-targeted antigenic proteins were separated by SDS-PAGE and nondenaturing PAGE, respectively, using a PhastSystem (Pharmacia Biotech, Uppsala, Sweden) as previously described (3, 5). Proteins were then electrotransferred onto reinforced nitrocellulose membranes (Hybond-C extra; Amersham, Little Chalfont, U.K.). Membrane strips were probed with human sera (1:100), a pool of AFA purified from 45 RA sera (see below; 10 μg/ml), affinity-purified autoantibodies to deiminated fibrinogen (see below; 10 μg/ml), the 311 mAb to fibrin(ogen) (5 μg/ml), sheep antisera directed against either the Aα- or the γ-chain of human fibrinogen (1:4,000; Cambio, Cambridge, U.K.), and a rabbit antiserum directed to the Bβ-chain of human fibrinogen (1:200,000; Cambio). When the membranes were probed with the Ab to modified citrulline (0.125 μg/ml), citrulline residues were previously chemically modified (21). Peroxidase-conjugated secondary probes were used for the detection of all of the primary Abs: protein A (Sigma), sheep Abs to mouse IgG (Biosys), goat Fab to rabbit IgG (Biosys), and rabbit F(ab′)₂ to sheep IgG (Southern Biotechnology Associates, Birmingham, AL) for the detection of human, murine, rabbit, and sheep IgG, respectively. Peroxidase activity was visualized using ECL Western blotting reagents (Amersham) following the procedure suggested by the manufacturer. The same protocol was applied for probing polyvinylidene fluoride (PVDF)-blotted proteins of the urea-DTT extracts after two-dimensional electrophoresis (see below).

NH₂-terminal amino acid sequence analysis was performed on the synovial antigenic proteins electrotransferred onto a PVDF membrane after two-dimensional electrophoresis performed using a PhastSystem. Proteins of the urea-DTT extracts from synovial membranes were precipitated by acetone, redissolved in the urea-DTT buffer (15-fold concentrated), and then separated in the first dimension by isoelectrofocusing (IEF) using precast IEF PhastGels that were previously washed, dried, and rehydrated in deionized 8 M urea, 0.5% Nonidet P-40, and ampholytes generating a 3–10 pH gradient (Pharmacia Biotech). SDS-PAGE was performed in the second dimension using precast 7.5% polyacrylamide gels. Proteins were then electrotransferred onto PVDF ProBlott membranes (Applied Biosystems, Foster City, CA) in 50 mM Tris and 50 mM boric acid. Amido black-colored membrane areas, previously characterized as corresponding to the two AFA-reactive antigenic proteins, were excised. They were separately sequenced for up to 10 Edman degradation cycles on an Applied Biosystems 494A sequencer following the manufacturer’s specifications in the Laboratoire de Microséquençage des Protéines of the Institut Pasteur (Paris, France). Standard single-letter codes are used for amino acid residue abbreviation.

Plasminogen-depleted human fibrinogen (95% pure; Calbiochem, Meudon, France) was incubated at 0.86 mg/ml with rabbit skeletal muscle PAD (7 U/mg fibrinogen; Sigma) in 0.1 M Tris-HCl (pH 7.4), 10 mM CaCl₂, and 5 mM DTT for 2 h at 50°C.

AFA were separately purified from 45 high-titered AFA-positive RA sera by affinity chromatography on neutral/acidic human epidermal filaggrin and subsequently pooled as previously described (12, 18). Autoantibodies to deiminated fibrinogen were purified by affinity chromatography from the IgG fraction of a pool of 38 RA sera with high AFA titers obtained by standard affinity chromatography on a protein G column (HiTrap Protein G; Pharmacia Biotech). IgG (1.6 mg/ml in a phosphate buffer (pH 7.4) containing 0.65 M NaCl) were loaded onto a 1-ml NHS-HiTrap column (Pharmacia Biotech) coupled with 1.5 mg of deiminated fibrinogen. After a 2-h incubation at room temperature, the column was washed with 8 volumes of 20 mM phosphate buffer (pH 7.0) containing 2 M NaCl and bound Abs were eluted with 0.2 M glycine-HCl (pH 2.5), immediately neutralized by addition of 2 M Tris, and then desalted. Biotin conjugation was performed on a portion of the Abs by incubation with N-hydroxysuccinimidobiotin (20 mol/mol; EZ-Link NHS-Biotin; Pierce, Rockford, IL) in 100 mM NaHCO₃.

The presence of deiminated proteins in the rheumatoid synovial membranes was demonstrated in seven of eight RA patients by immunoperoxidase staining of tissue sections with an Ab to modified citrulline. The Ab stained more or less abundant interstitial amorphous deposits (Fig. 1, G–I) and labeled the cytoplasm of numerous mononuclear cells. Both deposits and cells were located in the lining layer (Fig. 1, A–F) and/or in the deep synovium (Fig. 1, G–I). The cells morphologically evoked macrophages and fibroblasts, both cell types being observed in some sections. Moreover, the Ab labeled the cytoplasm of smooth muscle cells in arteriole walls (Fig. 1 J).

To biochemically characterize the deiminated proteins of the RA synovial membranes, fragments of synovial tissue from four RA patients were sequentially extracted successively using a TBS, a urea, a urea-DTT, and a guanidine buffer, then acetic acid. The final extraction pellet was divided in two parts and partially digested with pepsin and with collagenase. When the protein extracts and digests were analyzed by immunoblotting with the Ab to modified citrulline (Fig. 2), several deiminated proteins of various molecular mass were detected in the TBS, urea, urea-DTT, and guanidine extracts. By contrast, no immunoreactive proteins were detected either in the acetic acid extract or in the pepsin and collagenase partial digests.

To determine whether any of the proteins present in the rheumatoid synovial membranes was specifically recognized by AFA, the protein extracts and digests were analyzed by immunoblotting with affinity-purified AFA, AFA-positive and AFA-negative RA sera, and AFA-negative control sera obtained from patients with various non-RA rheumatic diseases and from healthy blood donors. No reactivity related to AFA, i.e., observed with both purified AFA and a large majority of AFA-positive RA sera, was detected either in the TBS or urea extracts or, as expected given the absence of reactivity with the Ab to citrulline, in the acetic acid extract or in the pepsin or collagenase partial digests (data not shown). However, in the urea-DTT (a subgroup of sera is shown in Fig. 3,A) and guanidine extracts (data not shown), two protein bands of 64–78 kDa (p64–78) and of 55–61 kDa (p55–61), comigrating with deiminated proteins (Figs. 2 and 3 A), were recognized by the purified AFA and by 25 of 28 AFA-positive RA sera. None of the 22 AFA-negative RA or non-RA sera recognized the urea-DTT- or guanidine-soluble p64–78 or p55–61. The p64–78 and p55–61 Ags were immunodetected in the urea-DTT and guanidine synovial extracts from the four RA patients, the labeling intensity of the Ags being variable from one patient to another.

Immunoblotting analysis of the urea-DTT synovial extracts separated by two-dimensional IEF/SDS showed that the two major proteins of around 70 and 60 kDa they contain exhibit extensive charge heterogeneity (pI ∼5.85–8.45). The two proteins were recognized by the Ab to citrulline, the purified AFA, and three AFA-positive RA sera but not by two AFA-negative control sera (Fig. 3,B and data not shown). These deiminated proteins therefore correspond to the p64–78 and p55–61 Ags identified after monodimensional SDS-PAGE of urea-DTT extracts. From an identical amido black-colored PVDF blot of the two-dimensionally separated urea-DTT extract, three areas bearing the most basic, intermediate, and most acidic variants of p64–78 and p55–61, respectively, were excised. The proteins borne by each membrane fragment were separately characterized by NH₂-terminal amino acid sequence analysis. From the intermediate isoelectric variants of p64–78 and p55–61, the sequences obtained were GPRVVERHQS and GHRPLDKKRE, respectively. The sequences perfectly matched those of the 10 NH₂-terminal residues of the α- and β-chain of human fibrin, respectively. Shorter sequences obtained from the most basic and most acidic variants of p64–78 and p55–61 confirmed these results. The apparent molecular masses of the p64–78 and p55–61 Ags present in the urea-DTT and guanidine extracts of RA synovial membranes are compatible to the apparent molecular masses of the α- and β-chains of human fibrin, respectively. In addition, p64–78 and p55–61 were highly reactive with antisera specific for the Aα- or Bβ-chains of human fibrin(ogen), respectively, and with the mouse mAb 311 directed to fibrin(ogen) (Fig. 3 C), further demonstrating their identity with the α- and β-chains of human fibrin.

Human fibrinogen is a large plasma glycoprotein that plays a major role in the final phase of blood coagulation. It is a dimeric molecule with each half-molecule composed of three different polypeptides designated Aα, Bβ, and γ. The coagulation enzyme thrombin triggers the conversion of fibrinogen to fibrin by cleaving the fibrinopeptides A and B from the Aα- and Bβ-chains of fibrinogen, yielding the α- and the β-chain of fibrin, respectively (22). To check that deimination generates the AFA-targeted epitopes on the α- and β-chains of fibrin, human plasma fibrinogen was deiminated in vitro with a rabbit skeletal muscle PAD using conditions previously shown to generate AFA epitopes on human recombinant filaggrin (12) The reactivity to the various undeiminated and deiminated fibrinogen chains was analyzed by immunoblotting with purified AFA, 37 AFA-positive RA sera with decreasing immunoblotting reactivity toward the neutral/acidic epidermal filaggrin, 10 AFA-negative RA sera, and 19 control sera from patients with nonrheumatoid inflammatory or noninflammatory rheumatic diseases (illustrated in Fig. 4). The efficiency of deimination of the three fibrinogen chains was clearly demonstrated by their reactivity with the Ab to citrulline. With human sera, reactivity toward the undeiminated fibrinogen chains was absent or very weak and not associated with any serum subgroups. When deiminated, both the Aα- and Bβ-chains became reactive with purified AFA and with 35 of 37 and 33 of 37 AFA-positive RA sera, respectively. By contrast, the deiminated γ-chain remained largely unreactive. Reactivity to the Aα-and/or Bβ-chains was also observed with some of the 10 AFA-negative RA sera. Globally, the staining intensity of the deiminated Aα and Bβ polypeptides was found to be related to the serum AFA titer. None of the 19 control sera from nonrheumatoid patients (except one that faintly labeled the Bβ-chain) recognized the deiminated Aα-, Bβ-, or γ-chains.

The 311 mAb was used to probe sections of rheumatoid synovial membranes from six RA patients by immunoperoxidase staining. The Ab stained more or less abundant interstitial amorphous deposits located in the lining and/or in the deep synovium (Fig. 5, A and B) and very rare macrophage-like mononuclear cells (Fig. 5,C). The comparison of serial sections of synovial tissue stained with the 311 mAb (Fig. 5) and the Ab to citrulline (Fig. 1) showed that 1) the interstitial amorphous deposits stained by the latter Ab correspond to fibrin and 2) most of the intracellular deiminated material does not correspond to fibrin since the number of cells stained by the 311 mAb was far smaller than the number of cells stained by the Ab to citrulline. These results confirmed the presence of extracellular and intracellular fibrin(ogen) material in rheumatoid synovium (23, 24) and, in perfect agreement with the immunochemical data, indicated that at least part of the fibrin found in RA synovial membranes is deiminated.

The sequencing and immunoblotting data presented above strongly suggested the identity or at least the large overlap between the Abs targeting the p64–78 and p55–61 synovial Ags and those recognizing the in vitro deiminated Aα- and Bβ-chains of human fibrinogen. To confirm this, autoantibodies to deiminated fibrinogen were purified from a pool of 38 RA sera with high AFA titers by affinity chromatography onto in vitro deiminated fibrinogen. When tested by immunoblotting (Fig. 6,A), the purified autoantibodies were not only highly reactive to the deiminated Aα- and Bβ-chains of human fibrinogen but also intensely stained the urea-DTT- and guanidine-soluble p64–78 and p55–61 Ags. The overlap between AFA and autoantibodies to the deiminated Aα- and Bβ-chains of fibrinogen, suggested by the correlations of the Abs titers in the series of patients, was confirmed by the immunoblotting reactivity of purified AFA to the in vitro deiminated Aα- and Bβ-chains of fibrinogen (Fig. 4). Furthermore, the autoantibodies purified onto deiminated fibrinogen were reactive with all of the Ags defined by AFA in human epidermis and rat esophagus epithelium. Indeed, both the neutral/acidic filaggrin from human epidermis and the three profilaggrin-related antigenic proteins from rat esophagus epithelium were recognized by immunoblotting (Fig. 6,A). Moreover, the stratum corneum (cornified layer) of the rat esophagus epithelium and that of the human epidermis were both strongly labeled immunohistologically (Fig. 6, B and C).

Given that deiminated (pro)filaggrins were known to be the epithelial Ags recognized by AFA and since synovial joints but not squamous epithelia are the site of rheumatoid inflammation, it was natural to check whether (pro)filaggrin was or was not expressed in RA synovial membranes. Our immunohistological and immunochemical analyses of synovial membranes from 9 RA patients with 29 anti-(pro)filaggrin mouse monoclonal IgG Abs that recognize at least 20 different epitopes borne by the various forms of (pro)filaggrin (Ref. 25 and our unpublished observations) definitively showed that (pro)filaggrins are not expressed in the tissue (data not shown). Therefore, since the epitopes targeted by AFA are generated by deimination of (pro)filaggrins (12, 13), we postulated that AFA were directed to one or several cross-reactive deiminated antigenic protein(s) present in the rheumatoid synovial membranes.

Sequential extractions of synovial membranes from RA patients using increasingly powerful denaturing agents, then buffers and enzymes commonly used for solubilization of extracellular matrix components such as proteoglycans and collagens (26, 27, 28, 29), allowed us to achieve exhaustive solubilization of the intracellular and extracellular proteins present in the tissue. Immunoblotting analysis of the extracted proteins with a probe specific for citrulline residues demonstrated that several deiminated proteins with various solubility properties and molecular masses are present in the rheumatoid synovial membranes. Among them, only two proteins were shown to be significantly and specifically targeted by AFA-positive RA sera and purified AFA. Using protein sequencing, these AFA-targeted deiminated proteins were identified as the α- and β-chains of fibrin and that was confirmed using various Abs specific for human fibrin(ogen) chains. The fact that only two among the deiminated proteins extracted from RA synovium were detected by AFA confirms that not only citrulline but also other specific residues near citrulline are necessary to generate the epitopes recognized by AFA (12, 13). Taken together, these results clearly demonstrate that targets of AFA exist in rheumatoid synovial membranes and identify the major targets as deiminated forms of the α- and β-chains of fibrin.

A part of the antigenic deiminated fibrin was extracted using urea-DTT and another part could be solubilized only after further extraction using 6 M guanidine-HCl. The latter part probably corresponds to fibrin molecules linked to the extracellular matrix by association with fibronectin and other components of the synovial connective tissue, including collagens and proteoglycans (22, 30). Indeed, these molecules are known to require high concentrations of strong denaturing agents for their solubilization (26).

The reactivity of AFA to the deiminated α- and β-chains of fibrin was clearly confirmed by analyzing the reactivity of rheumatoid sera and purified AFA toward human plasma fibrinogen deiminated in vitro. By immunoblotting, the deiminated Aα- and/or Bβ-chains of fibrinogen were recognized by all of the AFA-positive and by a significant part of the AFA-negative RA sera, whereas they were unreactive with the 19 control sera coming from non-RA patients (except one that moderately labeled only the Bβ-chain). This showed that, as expected, the immunoreactivity is specific for RA and moreover that the deiminated fibrinogen chains are targeted by a larger proportion of RA sera than (pro)-filaggrin. Consequently, diagnostic immunoassays using as immunosorbent deiminated fibrin(ogen) or peptides derived from this protein will probably display a higher diagnosis sensitivity than those using filaggrin or filaggrin-derived peptides. Moreover, this result reinforces the hypothesis that deiminated fibrin is the true in vivo target of the RA-specific autoantibodies and (pro)filaggrins are only cross-reactive Ags. Immunopurification from a pool of RA sera of autoantibodies to deiminated fibrinogen showed that a large overlap between AFA and the autoantibodies to the deiminated fibrin(ogen) exists. Indeed, all of the previously identified targets of AFA in human epidermis and rat esophagus epithelium were recognized both immunochemically and immunohistologically by the autoantibodies immunopurified onto deiminated fibrinogen. Comparative analysis of the reactivity of large series of RA sera to neutral/acidic filaggrin and to deiminated fibrinogen, completed by immunoabsorption experiments, will allow the actual extent of the overlap between AFA and the autoantibodies to deiminated fibrin(ogen) to be more precisely determined.

The immunohistological analysis of rheumatoid synovial membranes with the Ab to citrulline residues confirmed the presence of deiminated proteins in the synovium and showed that these proteins are mainly located in the cytoplasm of fibroblast-like and macrophage-like mononuclear cells and in interstitial amorphous deposits. The labeling obtained with a mAb to human fibrin(ogen) confirmed that the protein is largely present in the tissue and the comparison with the labeling pattern of the Ab to citrulline confirmed that at least part of this fibrin is deiminated. This comparison also indicated that most of the intracellular deiminated proteins stained by the Ab to citrulline do not correspond to fibrin. They most probably correspond to the other deiminated proteins found in the rheumatoid synovial membrane extracts. Their specific targeting by autoantibodies associated with RA remains to be thoroughly evaluated. Now, an important question is whether deimination of the synovium fibrin and, more generally, of synovial proteins, is or is not specific for rheumatoid inflammation. We can already affirm that at least the presence of intracellular deiminated proteins is specific for RA, since the related labeling is absent from synovium in various other inflammatory and noninflammatory joint diseases.⁵ These recent preliminary data support the hypothesis that the presence of deiminated proteins in the synovial membrane could be specific for RA.

Fibrin deposition is usual during tissue inflammation but the presence of fibrin-related extracellular deposits and intracellular (phagocytosed?) material has long been known to be particularly prominent in rheumatoid inflammation of synovial tissue (23, 24). In the RA joints, accumulation of fibrin is thought to result from an altered balance between coagulation and fibrinolysis. It could contribute to chronicity and progressivity of the disease through the proinflammatory and degradative effects of various factors participating in and resulting from its formation and elimination (31, 32, 33, 34, 35, 36). A more disease-specific role for fibrin in the pathogenesis of RA is very probably through its action as an autoantigen and perhaps as an immunogen. Indeed, deimination of fibrin could make it antigenic, the deiminated fibrin eliciting or, at least, sustaining specific autoantibody production. Interestingly, induction of a chronic arthritis in rabbit by intra-articular injection of heterologous and, to a lesser degree, autologous fibrin, into previously sensitized animals, was described nearly 40 years ago (37). The pathophysiological significance of these findings probably needs to be re-examined in light of identification of deiminated fibrin as the major target of AFA.

Although it is clear that protein deimination interferes in several in vitro observed protein-protein interactions and enzymatic processes (38, 39), the biological function of PADs has not yet been fully elucidated. Because PADs were shown to be expressed by many cell types including hemopoietic cells (40), the synthesis and secretion of one or several PADs mediating deimination of fibrin, either by inflammatory infiltrating cells or by synoviocytes of the rheumatoid synovial membranes, is highly probable. Confirmation of the presence of a PAD(s), identification of its (their) type and cell origin, and elucidation of the mechanisms leading to its (their) activation are the next questions to answer. Whether deimination of fibrin indirectly belongs to the apoptosis-related posttranslational modifications of self proteins (41, 42) recently proposed to be responsible for generation of antigenic targets in several autoimmune diseases (43, 44, 45, 46, 47), also remains to be determined.

Considered as a whole, these results constitute a strong additional argument for the involvement of AFA in the pathogenesis of RA. Indeed, both the autoantibodies, which are produced locally (18), and their specific targets are present at the site of rheumatoid inflammation and that highly suggests that the immunological conflict actually occurs in the synovial tissue. If the B autoimmune reaction to deiminated fibrin is as critical in the development and/or maintenance of human RA as B reaction to glucose-6-phosphate isomerase is critical in the development of the recently described RA-like joint disorder of the K/B × N TCR-transgenic mouse line (48, 49), then specific modulation of this autoimmune response, e.g., by target peptides or analogs, may be a new approach for the therapy of human RA.

We thank Prof. B. Fournié (Service de Rhumatologie, Hôpital Purpan, Toulouse, France), Prof. B. Mazières, and Prof. A. Cantagrel (Service de Rhumatologie, Hôpital Rangueil, Toulouse) for providing patient data and sera. We also thank Prof. M. Mansat and Dr. M. Rongières (Service de Traumatologie et Orthopédie, Hôpital Purpan) for providing samples of rheumatoid synovial membranes, Dr. R. Durroux (Laboratoire d’Anatomie Pathologique, Hôpital Rangueil) and Dr. S. Valmary (Laboratoire de Biologie Cellulaire et Cytologie, Hôpital Purpan) for advice in histopathological analyses of synovial tissues, and Prof. M. Costagliola and Prof. J.-P. Chavoin (Service de Chirurgie Plastique, Hôpital Rangueil) for providing samples of human skin. The technical assistance of M. P. Henry, M. F. Isaïa, and S. Wambergue is gratefully acknowledged.