To address the ongoing debate concerning the specificity of synovial citrullinated proteins for rheumatoid arthritis (RA) and to analyze their pathophysiologic relevance to the induction or perpetuation of the RA-specific anti–citrullinated protein antibodies (ACPAs).
Synovium of 19 RA patients and 19 non-RA controls was immunostained for the presence of citrullinated proteins with a mouse monoclonal antibody (F95), for the citrullinating enzyme peptidyl arginine deiminase type 2 (PAD-2), and for the free citrulline–producing enzyme inducible nitric oxide synthase (iNOS). Extending the RA cohort to 61 patients, the findings of anticitrulline staining in synovium were related to serum and synovial fluid ACPA levels, as measured by enzyme-linked immunosorbent assay.
F95 staining indicated the presence of synovial intracellular citrullinated proteins in 53% of RA samples versus 5% of control samples, whereas extracellular staining was not RA specific. Immunoblotting and inhibition experiments confirmed that the antibody recognized citrullinated proteins but not free citrulline. Accordingly, iNOS was equally found in RA and control synovium and in intracellular citrullinated protein–positive and intracellular citrullinated protein–negative samples. In contrast, intracellular citrullinated proteins colocalized with PAD-2, which was found in 59% of RA samples versus 17% of control samples. Independent of local disease activity, the presence of the RA-specific synovial intracellular citrullinated proteins was associated with significantly higher systemic and local ACPA levels and with local ACPA production in the joint.
These data confirm the presence of RA-specific intracellular citrullinated proteins in synovium. The link with PAD-2 and local and systemic ACPA levels emphasizes their pathophysiologic relevance for RA-specific humoral autoimmunity.
Anti–citrullinated protein antibodies (ACPAs) are highly specific for rheumatoid arthritis (RA) (1–6). The occurrence of ACPAs early in the disease course and even before the onset of clinical symptoms (7, 8), as well as the correlation with disease severity and prognosis (9, 10), suggests that ACPAs are not only valuable clinical biomarkers but are also pathophysiologically involved in the disease. However, epithelial (pro)filaggrin, which is the first identified antigenic target of ACPAs (11–13), is not expressed in the joint and RA does not affect filaggrin-containing epithelial tissues, such as skin and buccal mucosa. Based on crucial observations that ACPAs can be produced locally in the joint (14) and that posttranslational modification of arginine into citrulline residues in the context of specific amino acid sequences is essential for recognition by ACPAs (15–17), the assumption was made that citrullinated filaggrin could be a cross-reacting in vitro antigen rather than a genuine in vivo antigenic target of ACPAs and, subsequently, that distinct citrullinated proteins present in the inflamed synovium may be involved in the induction or perpetuation of the RA-specific humoral autoimmune responses.
First, Masson-Bessiere and coworkers demonstrated the presence of several deiminated proteins in RA synovium, of which the deiminated forms of the α- and β-chains of fibrin appeared to be major antigenic targets of ACPAs (18). Moreover, in vitro deiminated fibrinogen is a highly specific and sensitive substrate for the detection of serum ACPAs (19). However, the mere presence of deiminated fibrin in synovium does not fully explain the RA-specific induction of ACPAs, since citrullinated fibrin was detected equally in inflamed synovium in non-RA patients and in animal arthritis, where deiminated fibrin does not induce ACPAs (20–22). Second, Despres and coworkers described a 50-kd antigen (Sa antigen) in extracts from human placenta, spleen, and RA synovium (23). Anti-Sa antibodies are specific for RA and recognize deiminated vimentin, which was previously shown to be present in apoptotic macrophages in vitro (24, 25). Third, we demonstrated the presence of intracellular citrullinated proteins that colocalized with ACPA reactivity in RA synovium (26). Although the biochemical nature of these intracellular citrullinated proteins remains ill-defined, the major interest in comparison with both above-mentioned potential antigenic targets is that they are specifically present in RA synovium (20, 26, 27).
Since the recent observations might indicate a pathophysiologic role in the induction of ACPAs in vivo, in this study we attempted to confirm the RA specificity of synovial intracellular citrullinated proteins, with special reference to the presence of the citrullinating enzyme peptidyl arginine deiminase (PAD) (28, 29). Considering the presence of several deiminated proteins in inflamed synovium, we also sought direct evidence that these RA-specific synovial intracellular citrullinated proteins are genuine antigenic targets of the humoral autoimmune process by analyzing their influence on local and systemic ACPA levels.
PATIENTS AND METHODS
Patients and samples
For the analysis of intracellular citrullinated proteins, PAD type 2 (PAD-2), and inducible nitric oxide synthase (iNOS), we studied 19 RA patients who fulfilled the American College of Rheumatology (ACR; formerly, the American Rheumatism Association) criteria (30). All patients had active disease, as characterized by a mean ± SD swollen joint count (SJC) of 7.5 ± 6.2, with at least 1 swollen knee joint as the indication for needle arthroscopy, a serum C-reactive protein (CRP) level of 78 ± 71 mg/liter, and an erythrocyte sedimentation rate (ESR) of 55 ± 28 mm/hour. The control group (n = 19) consisted of 8 patients with spondylarthropathy who fulfilled the European Spondylarthropathy Study Group criteria (31), 6 patients with osteoarthritis (OA), 4 patients with juvenile chronic arthritis, and 1 patient with villonodular synovitis. As in the RA group, all patients had active joint disease, with at least 1 swollen knee joint as an indication for needle arthroscopy. For the analysis of the relationship of intracellular citrullinated proteins to ACPA titers, 42 patients who fulfilled the ACR criteria were added to the RA cohort (30). Similar to the first cohort, all patients had at least 1 swollen knee joint and had active disease, with a mean ± SD SJC of 6.6 ± 4.9, a serum CRP level of 48 ± 37 mg/liter, and an ESR of 46 ± 23 mm/hour. All patients signed a letter of informed consent before inclusion in the study, which was approved by the Ethics Committee of Ghent University Hospital.
Synovial tissue samples were obtained from clinically involved knee joints of all patients by needle arthroscopy, as previously described (32). Paired synovial fluid (SF) and serum samples were collected.
Synovial intracellular citrullinated proteins were detected using a well-described and validated mouse IgM monoclonal antibody (mAb), F95, which was raised against a decacitrullinated peptide linked to the carrier protein, keyhole limpet hemocyanin (33–35). For the detection of the citrullinating PAD enzymes, we focused on PAD-2, based on the presence of both PAD-2 and PAD-4, but not PAD-1 and PAD-3, subtypes in synovial extracts (36), the expression of PAD-4 in the nucleus rather than the cytoplasm (37–39), and the absence of correlation between PAD-4 polymorphisms and intracellular citrullinated proteins (40). A single-chain protein against human PAD-2, similar to a light chain of an IgG antibody, was developed by phage display technology. It was directed against the PAD-2 antigen VEK-NNP-KKA-SWT-WGP-EGQ-GA, which is 100% homologous for human, rat, and mouse PAD-2 and has <50% homology with peptides from other PADI subtypes. Since this single-chain protein contained the antigenic marker FLAG, rabbit anti-FLAG IgG were loaded with the Fab ends against PAD-2 and used for immunohistochemical detection of PAD-2 in synovium. Blot experiments using recombinant PADI confirmed that only PAD-2 was recognized (41). The following commercially available antibodies were also used: rabbit anti–L-citrulline polyclonal antibody (pAb) (Biogenesis, Poole, UK) (26) and mouse anti-iNOS mAb (R&D Systems, Minneapolis, MN).
Frozen sections of synovial tissue were stained by incubation with the primary antibody, followed by, sequentially, a biotinylated second antibody, a streptavidin–peroxidase complex, and aminoethylcarbazole as substrate (LSAB+ system; Dako, Glostrup, Denmark), as previously described (26, 27). For F95, the same protocol was used as described in detail previously for rat and human brain tissue (34, 35). For all stainings, isotype- and concentration-matched irrelevant antibodies were used as negative controls. For tissue-blocking experiments, F95 was incubated for 24 hours with 0.9 μM decacitrullinated peptide or free citrulline and subsequently used for immunohistochemistry.
Global synovial inflammation was assessed by the degree of inflammatory infiltration and synovial vascularity and the number of CD3+ T lymphocytes, CD20+ B lymphocytes, and plasma cells, as previously described (26, 27). The stained sections were masked with regard to diagnosis and were evaluated on a 0–3 semiquantitative scale by 2 independent observers (LDR, DB), using the mean score from the 2 observers in case of discrepancy (26, 27).
Dot blots containing 25 μg of decacitrullinated peptide, free citrulline, or mouse brain extracts were dried on polyvinylidene difluoride membrane (Perkin Elmer, Boston, MA), which was then cut lengthwise to transect each blot. The top half of the membrane was blocked with 3% bovine serum albumin for 2 hours, washed, and incubated with F95 (1:50 dilution) for 1 hour. After washing, binding of F95 was detected with horseradish peroxidase–conjugated anti-mouse Igm (Jackson ImmunoResearch, West Grove, PA) using the Western lighting chemiluminescence kit (Perkin Elmer). The bottom half of the membrane was used as a positive control to detect either free amino acids (free citrulline), by staining with 0.25% ninhydrin in 10% acetone and 10% acetic acid, or peptide bounds (decacitrullinated peptide and brain extract), by staining with biuret reagent (Sigma, St. Louis, MO) and Folin's reagent (Sigma).
Double immunofluorescence experiments were performed with the antibodies noted above. Detection was performed using tetramethylrhodamine isothiocyanate–conjugated goat anti-rabbit Ig (Dako), fluorescein isothiocyanate–conjugated goat anti-mouse Ig (Dako), or Alexa Fluor 555–conjugated goat anti-rabbit Ig (Molecular Probes, Eugene, OR), as appropriate.
ACPA levels in serum and SF were measured by the commercially available anti–citrullinated peptide 2 (anti–CCP-2) enzyme-linked immunosorbent assay (ELISA) containing synthetic CCPs as substrate (Immunoscan RA, mark 2; Euro-Diagnostica, Arnhem, The Netherlands). To compare ACPA in serum and SF, the titers were corrected for total IgG as measured by the human IgG ELISA quantitation kit (ImTec Diagnostics, Antwerp, Belgium). Both ELISAs were performed according to the manufacturers' instructions.
Comparisons between the 2 groups were performed using the Mann-Whitney U test. The chi-square test was used for comparison of proportions. Correlations were assessed by Spearman's rho coefficient. For ACPA levels, the median values are provided. P values less than 0.05 were considered significant.
Detection of RA-specific intracellular citrullinated proteins in the inflamed synovium. When we investigated whether intracellular citrullinated proteins were present in RA and control synovium, it was found that the mouse anticitrulline mAb F95 stained single mononuclear cells in RA synovium (Figure 1A). Control staining with an irrelevant concentration and isotype-matched primary antibody was completely negative (Figure 1B). Intracellular staining was observed in 10 of 19 RA samples and in only 1 of the control samples (P = 0.004) (Figure 1C). Consistent with these findings, F95 staining correlated significantly with the results obtained with rabbit anticitrulline pAb, which we previously used (26) to indicate the RA specificity of intracellular citrullinated proteins in synovium (r = 0.943, P < 0.001) (Figure 1D). Interestingly, with F95 there was extracellular staining of extrasynovial deposits infiltrated by high numbers of polymorphonuclear cells on the surface of the lining layer, and occasionally the deeper synovium (Figures 1E and F). In sharp contrast to the intracellular staining, however, the extracellular staining was not RA specific since it was also seen in 9 of the 19 control samples. To exclude the possibility that the staining observed with F95 was due to the detection of free citrulline, additional immunoblotting experiments were performed, which indicated that F95 recognized decacitrullinated peptide and mouse brain extract but not free citrulline (Figure 1G). Moreover, decacitrullinated peptide could inhibit the F95 staining in RA synovium, whereas free citrulline could not (Figures 1H and I).
Relationship of RA specificity of intracellular citrullinated proteins to PAD-2. During the study of potential mechanisms related to the RA specificity of synovial intracellular citrullinated proteins, we noticed that immunohistochemical staining demonstrated the pronounced presence of the citrullinating enzyme PAD-2 in RA synovium. PAD-2 showed a predominant cellular expression pattern both in synovium and in extrasynovial deposits, occasionally with associated extracellular staining (Figures 2A–C). PAD-2 expression was significantly higher in RA (median score 1, range 0–3) than in control synovial tissue (median 0, range 0–2) (P = 0.041), with PAD-2 expression in 59% of the RA samples versus 17% of the controls (P = 0.026). In RA synovium, expression of intracellular citrullinated proteins was higher in the PAD-2–positive samples (median 2, range 0–3) than in the PAD-2–negative samples (median 0, range 0–1) (P = 0.054). Furthermore, double immunofluorescence showed that intracellular citrullinated proteins always colocalized with PAD-2 staining in RA synovium, whereas not all PAD-2–positive cells stained for intracellular citrullinated proteins (Figures 2D–F), which is consistent with the theory that not the mere presence, but rather the activation, of PAD-2 leads to deimination.
In sharp contrast, cellular expression and occasional extracellular staining of iNOS were found in the lining layer, the sublining layer, and the vascular endothelium of both RA (71% iNOS positive) and non-RA (95% iNOS positive) synovia (Figures 2G–I). There was no difference in iNOS expression levels between RA samples (median 1.5, range 0–3) and control samples (median 2, range 0–3). Within the RA group, there was no significant association with synovial intracellular citrullinated proteins.
Determination of systemic ACPA levels by synovial intracellular citrullinated proteins. Based on the RA specificity, the pathophysiologic link with distinct synovial PAD-2 expression, and the previously demonstrated colocalization with ACPA reactivity in RA synovium (26), we next assessed whether the synovial intracellular citrullinated proteins were pathophysiologically relevant antigenic determinants of ACPAs or cross-reactive substrates such as is assumed for epithelial filaggrin. Serum ACPA levels were significantly higher in patients with (27 of 61) versus patients without (34 of 61) intracellular citrullinated proteins in synovium (mean 865 units/ml versus 324 units/ml; P = 0.017) (Figure 3A). Interestingly, the previously described link between ACPAs and the HLA–DR shared epitope (SE) (6, 10, 42–45) was clearly more pronounced in the synovial intracellular citrullinated protein–positive patients (1,105 units/ml in SE-positive patients versus 167 units/ml in SE-negative patients; P = 0.050) than in the synovial intracellular citrullinated protein–negative patients (446 units/ml in SE-positive patients versus 110 units/ml in SE-negative patients; P not significant) (Figures 3B and C).
Since both ACPAs and the HLA–DR SE are associated with disease severity (6, 9, 45, 46), we assessed whether the presence of intracellular citrullinated proteins in RA synovium might be merely the consequence of higher levels of local inflammation. Parameters of synovial inflammation (infiltration, vascularity, CD3+ T cells, CD20+ B cells, and plasma cells) did not correlate with serum ACPA levels or SE status. In contrast, vascularity (P = 0.026) and CD20+ B lymphocytes (P = 0.025) were even higher in SE-negative patients. Accordingly, there were no significant differences in the parameters of inflammation between intracellular citrullinated protein–positive versus intracellular citrullinated protein–negative synovium, with the exception of vascularity, which was higher in the former group (P = 0.012). Finally, synovial biopsies obtained from an additional cohort of RA patients without knee synovitis showed intracellular citrullinated protein in 5 of 16 cases (data not shown). Taken together, these data indicated that the presence of synovial intracellular citrullinated proteins is associated with ACPA levels independent of local inflammation.
Local ACPA levels are determined by synovial intracellular citrullinated proteins. Considering the previous demonstration of local production of ACPAs in the synovial membrane (14) and the indicated link between serum ACPA levels and synovial intracellular citrullinated proteins, we next investigated ACPA levels in the SF of intracellular citrullinated protein–positive and –negative subjects. We found that the difference in ACPA levels between the groups that were positive and negative for intracellular citrullinated protein was even more pronounced in SF (mean 858 units/ml versus 127 units/ml; P = 0.013) than in serum. After correction of the ACPA levels for total IgG in serum and SF, the resulting relative amount of IgG directed against citrullinated proteins was higher in SF than in serum in the intracellular citrullinated protein–positive group (median ratio 1.49) but lower in SF than in serum in the intracellular citrullinated protein–negative group (median ratio 0.91) (P = 0.024). The increase of local ACPA levels specifically in those joints characterized by the presence of synovial intracellular citrullinated proteins suggested local ACPA production rather than trapping of systemically produced ACPAs, because inhibition experiments indicated that free ACPAs, but not ACPAs complexed with citrullinated proteins such as citrullinated fibrinogen, could be identified in the anti–CCP-2 ELISA (data not shown).
Considering the high specificity of ACPAs and their occurrence before the clinical onset of RA (3–8), identification of the molecular and cellular players involved in the citrullinated protein/ACPA conflict is of major interest to unravel the pathogenesis of RA. Previous data indicated that not all in vivo or in vitro substrates reactive with ACPA are pathophysiologically relevant, as illustrated by deiminated filaggrin, and that the relevant antigenic target(s) are probably present in the inflamed joint (14, 47). Therefore, the present study addressed the pathophysiologic relevance of synovial targets of ACPA by directly analyzing human synovial tissue and paired serum and SF samples.
Staining with the F95 anticitrulline monoclonal antibody indicated the presence of RA-specific intracellular citrullinated proteins in synovium, which is consistent with the results of our previous study (26). Whereas the results of our first study could have been biased by the detection of free citrulline or by the nonspecific binding of the anti–L-citrulline antibody to plasma cells due to its rabbit origin (48), the F95 antibody is of mouse origin and does not recognize free citrulline. Taken together with results of previous experiments that demonstrate the specificity of F95 for citrullinated but not native proteins (ref. 33, and Cantaert T: unpublished observations), these data confirm that the RA-specific targets recognized by F95 in synovium are intracellular citrullinated proteins. We also demonstrated that the RA-specific intracellular staining with the F95 antibody correlates strongly with the data obtained with the rabbit anticitrulline pAb.
In contrast to the RA specificity of the intracellular staining, F95 also stained extracellular citrullinated proteins that are not RA specific. These observations are consistent with findings of a recent study using different phage display antibodies and purified human ACPAs (49) as well as with our findings on deiminated fibrin using an antibody recognizing chemically modified citrulline epitopes independently of the amino acid context (20). The detection of different synovial deiminated proteins, some of which are RA specific, by distinct antibodies is not contradictory but rather indicates that not only citrulline, but also neighboring amino acids, determine the specific epitopes recognized by the different anticitrulline antibodies, mimicking the in vivo situation with ACPAs (16). Of interest, the extracellular versus intracellular localization of deiminated antigens may be of crucial importance since it was demonstrated that the abundant extracellular presence usually leads to B cell tolerance, whereas intracellular localization as such may be sufficient to convert a tolerogenic self antigen into a potent autoimmunogen (50). Biochemical studies are ongoing to determine if the RA-specific intracellular citrullinated proteins are distinct intracellular molecules or rather epitopes processed from deiminated extracellular proteins such as fibrin.
Considering the presence of several deiminated proteins in RA synovium (18, 20, 24, 26), we continued to investigate the repeated finding that only the intracellular citrullinated proteins described in the present study are highly specific for RA synovium (20, 26, 27). In sharp contrast to iNOS, we demonstrated that both the number of samples containing PAD-2 and the degree of PAD-2 expression were significantly higher in RA than in control synovium and that intracellular citrullinated proteins colocalized with the citrullinating enzyme PAD-2. However, it should be taken into account that not the mere presence, but also the activation, of PAD-2 is pathophysiologically relevant, as illustrated by the fact that not all PAD-2–positive cells contained citrullinated proteins in RA synovium and that PAD-2 was also found in some control samples. Studies are being conducted to assess whether this is related to programmed cell death, as demonstrated in vitro (25, 28, 29), or to other mechanisms leading to the increase of the intracellular calcium levels needed for the activation of the citrullinating PADI enzymes in human RA synovium.
Of interest, the presence of not only PAD-2, but also PAD-4, was previously demonstrated in human synovial tissue, and both subtypes were also found in SF mononuclear cells (29, 36). However, the association between functional haplotypes of PADI4 and RA remains controversial (40, 51–53). Furthermore, in contrast to PAD-2, the PAD-4 protein was extensively distributed throughout RA synovium, did not colocalize selectively with intracellular citrullinated proteins, and was also found in OA synovium (54). It would be interesting to assess if PAD-2 and PAD-4 have different cellular origins, synovial localization, and substrate specificities in the joint, and if this relates to the difference between the RA specificity of intracellular citrullinated proteins and the more ubiquitous presence of other deiminated proteins in synovium.
Besides the RA specificity and the link with the distinct presence of PAD-2 in RA synovium, the higher ACPA titers in synovial intracellular citrullinated protein–positive versus intracellular citrullinated protein–negative patients is a third major argument provided by this study for the pathophysiologic importance of these distinct proteins. The relevance of this finding is emphasized by several facts. First, previous studies failed to demonstrate a link between ACPAs and non–RA-specific citrullinated proteins, including fibrin (20, 49). Second, our findings indicated that the presence of intracellular citrullinated proteins is independent of the degree of synovial inflammation and that the demonstrated link with ACPAs is thus not merely a secondary phenomenon due to higher ACPA levels leading to enhanced inflammation. Third, the link between ACPAs and intracellular citrullinated proteins was related to the presence of the HLA–DR SE, corresponding with increasing evidence of an interaction between the HLA–DR genetic background and anticitrullinated protein reactivity (42–45). Finally, the relationship between intracellular citrullinated proteins and ACPAs was stronger locally in SF than systemically in serum, with higher SF than serum ACPA levels in intracellular citrullinated protein–positive but not intracellular citrullinated protein–negative joints. Since this cannot be fully explained by immobilization of ACPAs by complex formation with citrullinated targets in the joints, these data are consistent with the previous demonstration of local ACPA production (14, 49) and also suggest that this could be driven by specific synovial deiminated proteins.
In conclusion, the present results indicate the RA-specific presence in synovium of intracellular citrullinated proteins, which is, at least partially, related to the distinct presence of the citrullinating PAD-2 enzyme. Additionally, the association of synovial intracellular citrullinated proteins with systemic and local ACPA levels provides direct evidence for the pathophysiologic relevance of these proteins as RA-specific antigenic targets of the humoral autoimmune process in vivo.