Anti–citrullinated protein antibodies (ACPAs), the main serologic hallmark of rheumatoid arthritis (RA), are autoantibodies directed toward epitopes in which arginine amino acids have been posttranslationally modified toward citrulline by peptidyl arginine deiminases (1–3). ACPAs are intimately linked to the development of RA, as these antibodies are highly specific for RA (4), can appear years before the onset of clinical disease (5, 6), are associated with persistent and destructive disease (7, 8), and define a genetically and immunologically distinct set of RA patients (9, 10) who, despite having similar systemic and articular disease activity (11), show a better response to B cell depletion with rituximab (12).
Although these findings provide circumstantial evidence for the involvement of ACPA reactivity in the disease process, it remains to be determined in detail how, and at what stage and to what extent, ACPAs contribute to disease development. Ex vivo studies with human macrophages and mast cells have clearly indicated that ACPA-related immune complexes can trigger these cells to produce and release proinflammatory mediators (13, 14). However, it remains unknown whether ACPAs also play a role in the initiation phase of the disease, are just a mere bystander of the autoimmune process, or, on the contrary, even represent an abortive attempt to control and/or prevent the onset of RA (15).
The study of appropriate animal models may help us to gain a deeper insight into the potentially pathogenic role of ACPAs. Consistent with the findings from the above-mentioned studies with human macrophages and mast cells (13, 14), anti–citrullinated protein–specific antibodies have been reported to enhance tissue injury in experimental arthritis (16, 17). However, both the presence of genuine ACPAs and the arthritogenic role of these ACPAs in the initiation of experimental arthritis remain highly debated (18–23). The discrepancies between these reports may be due to the use of different citrullinated antigens for immunization, different mouse strains, and different detection methods. Therefore, in the present study, we aimed to revisit this issue by assessing, systematically, whether citrullinated proteins can induce a genuine anti–citrullinated protein response in mice and whether this response modulates clinical arthritis.
- Top of page
- MATERIALS AND METHODS
- AUTHOR CONTRIBUTIONS
- ADDITIONAL DISCLOSURES
ACPAs are important diagnostic and prognostic biomarkers in human RA, but the pathologic relevance of these antibodies remains largely unknown. Studies with human macrophages and mast cells have clearly indicated that the immune complexes that are formed by ACPAs can trigger these cells to produce and/or release proinflammatory mediators (13, 14). Interpreting the observations from experimental models, however, remains challenging, because the data on the presence and arthritogenicity of ACPAs in murine arthritis have been conflicting (16–23).
One of the major reasons for the discrepancies in the published reports may be the different methods used for the detection of ACPAs in mice. In contrast to the findings in human serum, in mouse serum, the antibodies binding the CCPs used in the commercially available ELISAs were reported to bind with similar avidity and affinity to the arginine counterpart, indicating that these antibodies are not specific for the citrulline residue, but rather are specific for the surrounding amino acids making up the immunogenic epitope (19). Our findings extend this concept by demonstrating, in 3 different mouse strains, that reactivity to the citrullinated peptide was clearly detected in CIA, but the immunoreactivity was similar upon immunization with citrullinated CII and that upon immunization with native CII, was not specific for the citrullinated epitope, as similar reactivity was observed with the arginine epitope, and, in contrast to experiments with human serum, could not be inhibited by preincubation with citrullinated epitopes. These results confirm that the commercially available anti-CCP assays are not reliable in the detection of ACPAs in mice. The possibility that other citrullinated peptides, such as epitopes from vimentin or α-enolase, might be more reliable substrates in rodents remains to be investigated.
To avoid this potential for false-positivity, ELISA methods using whole citrullinated proteins as coating antigen have been developed by several groups (18, 20–22, 26). However, one caveat, which has not always been taken into account, is the contamination of the in vitro citrullinated protein with PAD in the protein mixture used for immunization and/or in the substrate used for detection of ACPAs. Herein, we clearly showed that all investigated mouse strains developed a strong immunoreactivity against the rabbit PAD-2 enzyme used to citrullinate the antigens in vitro, leading to false-positive results in the ELISAs using in vitro citrullinated proteins as substrate. Moreover, the PAD protein itself can become citrullinated (results not shown) (28), which further complicates the interpretation of these data.
We circumvented this issue by developing a Western blot assay in which we utilized citrullinated fibrinogen as the substrate, since this substrate contains a wide variety of epitopes that can be citrullinated and specifically recognized by human ACPAs (29–31), and since the γ-chain can clearly be distinguished from contaminating PAD based on the molecular weight. Whereas the immunoreactivity of human ACPAs is largely restricted to 3 dominant epitopes of the α- and β-chain (32), we demonstrated specific immunoreactivity toward the citrullinated, but not native, γ-chain of fibrinogen in Balb/c mice that were hyperimmunized not only with citrullinated fibrinogen but also with other citrullinated proteins, such as citrullinated MBP or CII.
Using this robust and reliable detection method, we investigated the immunogenicity of different citrullinated proteins in 3 different mouse strains: Balb/c mice, which are well known for strong humoral responses but are not susceptible to CIA, SJL mice, which show a mild form of arthritis after CII injection, and DBA/1 mice, which are the most susceptible to CIA. In all 3 strains, ∼80% of the animals developed ACPAs after multiple immunizations with citrullinated fibrinogen in Freund's complete adjuvant/incomplete adjuvant, which is consistent with previous observations in experiments with Balb/c mice, DR4-IE–transgenic B6 mice, and Lewis and Brown-Norway rats (21, 22, 26). Interestingly, 2 of 6 Balb/c mice and 4 of 6 DBA/1 mice immunized multiple times with native fibrinogen also developed some ACPA reactivity, suggesting that the native protein injected for immunization either 1) already contains some citrullinated epitopes (23) or 2) can be citrullinated in vivo by mouse PAD. The former hypothesis is unlikely, as blotted human native fibrinogen is not recognized by the Senshu antibody (results not shown) (33). The latter concept, however, fits with the results reported in several studies indicating that ACPAs can develop upon immunization of mice with other native proteins (16, 17), and that, when endogenous PAD enzymes are chemically blocked, citrullinated synovial protein levels and severity of CIA are reduced (34).
Balb/c mice also developed ACPAs when immunized with citrullinated MBP or with citrullinated collagen, a substrate which is pathogenetically relevant but contains less epitopes that can be citrullinated (17). In SJL mice, ACPA reactivity was detected to some extent upon immunization with citrullinated MBP (results not shown), but not upon immunization with citrullinated CII.
Finally, in DBA/1 mice, ACPAs did develop upon immunization with citrullinated collagen, but this was observed in only a small minority of the animals. However, the use of citrullinated fibrinogen as the detection substrate in Western blot analysis may not detect all ACPAs, and antibodies against other citrullinated epitopes that are not present in citrullinated fibrinogen could be present in the sera of SJL and DBA/1 mice.
These results thus indicate that all 3 different strains can develop ACPAs, but that this humoral response depends on the strain (with Balb/c mice being most susceptible), the immunogen (with collagen being poorly immunogenic in comparison with fibrinogen or MBP), and, most probably, also the immunization protocol. Indeed, in Balb/c mice, the appearance of ACPAs was most pronounced after multiple immunizations with citrullinated CII; a similar protocol could not be used in DBA/1 and SJL mice due to development of severe arthritis after 2 injections with CII. In combination with the use of different detection methods, the variability in ACPA development related to these different factors could possibly explain many of the discrepancies reported in the literature.
The ability to induce and detect murine ACPAs also allowed us to reassess the pathogenic role of these antibodies in these models. First, we investigated whether the presence of ACPAs can aggravate experimental arthritis. It was previously shown that transfer of monoclonal ACPAs was insufficient to induce arthritis in DBA/1 mice, but did aggravate disease activity in mice with CAIA (16). In contrast, ACPAs raised by immunization with citrullinated fibrinogen did not modulate the course of subsequent adjuvant-induced arthritis in Lewis rats (21). Similarly, in our experiments, the presence of ACPAs induced by active immunization of Balb/c mice failed to affect the incidence or severity of CAIA. Although this discrepancy between the results of the former study (16) and the results of the latter study (21) and our own study may be attributed to the specific features of specific monoclonal ACPAs (17), overall these findings indicate that naturally raised ACPAs do not consistently affect the course of experimental arthritis.
Second, we assessed whether the response raised by immunization with native or citrullinated fibrinogen was arthritogenic, as has been reported previously not only in DR4-IE–transgenic B6 mice but also in wild-type SJL mice (22, 23). In contrast with the previous findings in SJL mice in the study by Ho et al (23), our study found that none of the Balb/c, SJL, or DBA/1 mice immunized up to 5 times with either native or citrullinated fibrinogen developed arthritis. This could be due to differences in the immunization protocol used, the mildness of the arthritis observed in the study by Ho et al (mean arthritis score of 4, maximum score of 12), or the sample size in those experiments (n = 10 mice). Our observations are in accordance with those from previous studies in rats (21) and mice (18), and further confirm that the presence of circulating ACPAs alone is not sufficient to induce arthritis in rodents.
Finally, even if we were not able to detect robust ACPAs during the course of CIA in SJL and DBA/1 mice, we assessed whether the use of citrullinated CII, rather than native CII, for immunization modulated the clinical course of CIA, since a previous report indicated that this immunization protocol had a modest effect on arthritis incidence and day of onset of disease in rats (20). We did not observe any difference in incidence or severity of arthritis in both SJL and DBA/1 mice. In one experiment in SJL mice, we observed a slightly accelerated onset of disease when citrullinated CII was used, but this effect was not reproducible in subsequent experiments. Moreover, the anti-CII antibody levels were not augmented by citrullination of CII; on the contrary, similar experiments in Balb/c mice even showed lower IgG anti-CII levels when citrullinated CII was used for immunization.
Two caveats should be considered when interpreting these data. First, the disease severity scores were relatively low in these experiments in mice with CIA, which may be explained by partial denaturation of CII during the 2 hours of incubation at 37°C needed for the in vitro citrullination reaction. Second, it needs to be emphasized that all of our experiments were performed with heterologous proteins (either human or bovine). Whether immunization with citrullinated autologous proteins, as was performed by Duplan and colleagues in a previous study (21), yields similar results needs to be further assessed in comparative studies.
Collectively, our findings indicate that mice can develop genuine ACPAs, but the detection of ACPAs is highly dependent on the strain, immunogen, immunization protocol, and, most importantly, the use of a reliable detection method. However, these ACPAs were not overly pathogenic, since neither the presence of preexisting ACPAs nor the use of citrullinated collagen could modulate the clinical course of arthritis.