Rheumatoid arthritis (RA) is the most prevalent systemic rheumatic disorder, affecting ∼1% of the world population. It is characterized by chronic inflammation of the joints, eventually causing irreversible joint damage that can lead to severe disability. The inflamed joints are infiltrated by large numbers of activated mononuclear cells that contribute to the destruction of articular cartilage.
One of the major features of RA is that many circulating autoantibodies directed to self antigens, including the well-known rheumatoid factor (RF) antibodies, can be found in patient sera (for review, see ref.1). The most specific autoantibody system for RA is the family of autoantibodies directed to citrulline-containing proteins (for review, see ref.2), including antiperinuclear factor (APF) (3), antikeratin antibody (4), antifilaggrin autoantibodies (5), anti-Sa (6), and anti–cyclic citrullinated peptide (anti-CCP) (5, 7). The essential part of the antigenic determinant recognized by these antibodies is the unusual amino acid citrulline (7, 8). Citrulline can be generated by posttranslational modification of arginine residues (guanido group ureido group) (Figure 1). This modification is catalyzed by peptidylarginine deiminase (PAD) enzymes, of which 4 mammalian isotypes have been described (9). Antibodies to citrullinated proteins can be detected in up to 80% of patients with RA, with >98% specificity (2, 10).
Figure 1. Enzymatic conversion of peptidylarginine to peptidylcitrulline by peptidylarginine deiminase in the presence of Ca2+.
Download figure to PowerPoint
Many murine models of arthritis exist, and 2 such models were studied here. Type II collagen (CII)–induced arthritis (CIA) is a widely used model of arthritis, based on autoreactivity against cartilage CII. Immunization of susceptible strains of mice with foreign CII consequently leads to cross-reactivity of the T cell response and an antibody-mediated immune reaction to homologous CII. After a delayed onset (∼4 weeks), arthritis progresses to a chronic stage, characterized by severe erosions of cartilage and bone (11, 12). Streptococcal cell wall (SCW)–induced arthritis in mice can be produced by intraarticular injection of bacterial fragments. An acute, macrophage-initiated inflammation is induced, with severe joint swelling accompanied by the release of chemokines, leading to rapid infiltration of predominantly polymorphonuclear neutrophils (PMNs) and macrophages into the joint. This acute local inflammation does not lead to a chronic or polyarticular arthritis; uninjected joints remain unaffected (13, 14).
In animal models, RF has been reported in MRL/lpr mice (15) and was recently observed in interleukin-1 receptor antagonist–deficient (16) and HLA–DQ8–transgenic mice with CIA (17). There are no reports on the presence of antibodies to citrullinated proteins or citrullinated synovial antigens in mouse models. This is the first report on the synovial expression of PAD enzymes and the presence of citrullinated proteins in CIA and SCW-induced arthritis.
- Top of page
- MATERIALS AND METHODS
In mouse models of both acute and chronic arthritis, PAD2 mRNA is present in the synovium but is not translated into PAD2 protein. In contrast, PAD4 mRNA, although absent from naive synovium, is readily transcribed and translated in PMNs infiltrating the synovial tissue during inflammation. As a consequence, several synovial proteins are subjected to citrullination. One of these proteins was identified as fibrin, which has been reported to be citrullinated also in the synovium of patients with RA. Although in mice the generation of citrullinated antigens during synovial inflammation was evident, no anticitrullinated protein antibodies could be detected.
Mouse PAD2 mRNA expression was substantial in both naive and arthritis-affected mice. The primers used to amplify mPAD2 are selected on separate exons, so there is no risk of possible artifacts caused by the presence of genomic DNA in the RNA isolates. Nevertheless, mPAD2 protein could not be detected in the synovial tissue, by either Western blotting or immunohistochemical analysis. In contrast, mPAD2 protein was abundantly present in the muscle tissue surrounding the synovium. These observations strongly indicate that the synthesis of PAD2 protein in synovial cells is strictly regulated at the translational level. We assume that additional, as yet unknown, factors are needed for the translation of mPAD2 mRNA. The exact nature of such factors is not clear at present, but the results presented here indicate that mPAD2 protein is not involved in the citrullination of synovial proteins in these mouse models of arthritis. Mouse PAD4 is expressed by PMNs infiltrating the synovium during inflammation. In the synovium of naive mice, no mPAD4 expression could be detected, whereas in the inflamed synovium mPAD4 was found at both the RNA and the protein levels. The amount of mPAD4 present in the synovium was consistent with the degree of inflammation: weak expression in mildly inflamed synovium and stronger expression in severely inflamed synovium. The immunolocalization showed a clear overlap of mPAD4 expression with PMNs, but not with monocytes or macrophages.
Expression of the human homolog of rodent PAD4 in human PMNs has been reported (37, 39). This human homolog was previously known as PAD5. The human enzyme shows slightly different reaction kinetics toward artificial substrates (40). This is why the human enzyme was first believed to be a novel PAD, hence its name PAD5. Recent data (for review, see ref.45), however, all indicate that human PAD5 corresponds to the rodent PAD4 (37, 39). Sequence data show that human PAD4/5 is more strongly conserved to rodent PAD4 (71% identical amino acids) than to other PAD isotypes (∼50% identical amino acids). Its position within the PAD gene cluster (all PADs are encoded on human chromosome 1p36, mouse chromosome 4E1, and rat chromosome 5q36) is conserved. Furthermore, PAD4/5 contains a conserved nuclear localization signal motif, which has been shown to be required and sufficient for nuclear localization, whereas all other PAD isotypes are located in the cytosol (39). Based on these data, our suggestion to change the name of human PAD5 to PAD4 was recently approved by the Human Genome Organisation Gene Nomenclature Committee (45, 46).
PAD4 expression has been observed in human and mouse peripheral blood granulocytes (refs.37 and47, and Vossenaar E, et al: unpublished observations) and in HL60-derived granulocyte-like cells (40). In both mouse models, infiltration of PMNs into the synovium is a crucial step in the onset of joint inflammation. These results strongly suggest that the infiltrating PMNs are responsible for the synovial expression of mPAD4 protein.
For the activation of mPAD4 protein, Ca2+ (and possibly other factors) is needed. PAD enzymes are present in the cytosol or nucleoplasm (39, 48), where the Ca2+ concentration usually is very low (∼10−7M) and strictly regulated. The threshold Ca2+ concentration for PAD activation is approximately 10−5M (ref.49 and Nijenhuis S, et al: unpublished observations); the cytosolic Ca2+ concentration is thus too low for activation of PAD. Several studies have shown that PAD enzymes can be activated by stimulation of PAD-containing cells with a calcium ionophore (39, 44, 47, 50). Also in human buccal mucosa cells (which contain the APF antigen filaggrin) (3), PAD enzymes are activated only very late in the terminal differentiation of the cells (41, 51). By that time, the integrity of the plasma membrane of the cells is compromised, causing influx of Ca2+ from the extracellular space and subsequent activation of PAD. In the inflamed synovium, many cells undergo apoptosis or necrosis. Especially PMNs have a very short lifespan, estimated to be only a couple of days. When the membrane integrity is lost during the death of these cells, Ca2+ can freely enter the cell and activate PAD enzymes that are already present. An alternative possibility is that PAD enzymes could leak out of the dying cells, become activated (the extracellular Ca2+ concentration is ∼10−3M), and induce the citrullination of extracellular proteins such as fibrin. Citrullinated proteins could be detected in the leukocyte infiltrate in the inflamed synovium (Figure 6). This area contains large numbers of mPAD4-positive PMNs, which corroborates the idea that PAD4 is responsible for the citrullination of synovial proteins during inflammation.
In sections of inflamed synovial tissue, large deposits of extravascular fibrin could be seen (Figures 6E and F). These fiber-like structures can be visualized with antibodies against fibrin(ogen) but are known to contain other proteins as well (such as fibronectin and collagen) (52, 53). These fiber-like structures are also intensively decorated with Senshu antibodies, indicating that they contain citrullinated proteins. Western blots confirm that fibrin is one of these citrullinated proteins (Figure 7), although other citrullinated proteins appear to be present as well. The identity of these proteins is the subject of our further investigations.
Our finding that extravascular fibrin is citrullinated during joint inflammation in mouse models of arthritis is very interesting, because the presence of citrullinated fibrin has also been reported in human RA (29). It therefore appears that the generation of citrullinated fibrin in both mice and humans is the result of synovial inflammation and is not a particularly human RA–specific phenomenon.
Although citrullinated proteins were present during synovial inflammation in mice, no antibodies directed to them could be detected, either in the serum or in the synovial fluid. In fact, in none of the many animal models of arthritis tested (including rodent, canine, and primate models) does anti-CCP appear to be positive (van Boekel M: personal communication). There are 2 likely explanations for this phenomenon.
First, there is a huge difference in the duration of the human disease as compared with the mouse models. In the SCW model, joint inflammation is induced within 24 hours, while in the CIA model onset of inflammation occurs within a few weeks. The development of RA takes many years, possibly more than a decade. Two independent studies have recently shown that former blood donors in whom RA later developed became RF positive and anti-CCP positive years before appearance of the first clinical symptoms of RA (54, 55). Anti-CCP production and, consequently, citrullination are clearly very early processes in the development of RA. Nevertheless, it could take many months or even years before anti-CCP autoimmunity develops. The duration of disease in the animal models (a maximum of 10 weeks in the CIA model) may well be too short for the evolution of anti-CCP antibodies.
Second, the role of genetic factors (e.g., HLA alleles as a risk factor for RA) has been known for more than 25 years. The HLA–DR4 (HLA–DRB1*0401 and *0404) phenotypes are especially important in this respect (56). Recent molecular modeling data indicate that peptides containing citrulline, but not arginine, can be bound by *0401 major histocompatibility complex molecules (57). This citrulline-specific interaction might be the basis of a citrulline-specific immune response. T cell proliferation assays with HLA–DRB1*0401–transgenic mice showed that stimulation with citrullinated peptides, not arginine peptides, induced proliferation and activation of T cells (58). These data suggest that the specific structure of HLA–DR4 molecules plays an important role in the induction of autoimmunity to citrullinated proteins. The mice used in the CIA model (DBA/1) and the SCW-induced arthritis model (C57BL6) do not have the arthritis-prone alleles. Their respective HLA haplotypes, H-2q and H-2b, are not associated with a risk of developing RA.
Taken together, these factors are compatible with the danger model hypothesis (59): an initial small inflammation caused by an external pathogen can cause the death of cells that cause or participate in inflammation, and consequently induce the citrullination of synovial proteins. In a certain genetic environment, this may lead to presentation of citrullinated peptides by antigen-presenting cells and consequently to activation of T cells. If the right conditions are present, this initial immune response can snowball into a systemic disease that is manifested many years later.
Citrullination of synovial proteins is not a process specific to RA. In the models of both acute and chronic arthritis, several synovial proteins, including fibrin, are subjected to citrullination during inflammation. Although these mouse models can be very useful in understanding the arthritis process, they do not exactly mimic RA (especially not the SCW model, which is a model of acute arthritis). The humoral response that is so characteristic of RA is absent in these mice. We conclude that the generation of citrullinated antigens is an inflammation-associated process, but that the antibody response to citrullinated proteins is highly specific to RA and possibly is involved in the perpetuation of the human disease.