Rheumatoid arthritis (RA) is a systemic, chronic inflammatory disease primarily affecting the joints. The synovial inflammation leads to cartilage destruction, bone erosion, joint deformity, and loss of joint function [1]. This disease is autoimmune in nature and characterized by the infiltration of T cells, B cells, macrophages, and neutrophils into the synovial lining and fluid of the periarticular spaces [2]. The infiltrating cells express adhesion molecules and produce a variety of inflammatory cytokines and chemokines to contribute to the complex pathogenesis of RA. The etiopathogenesis of this disease has not yet been completely elucidated.

Using gene-manipulating techniques, we have established two mouse models for RA: human T-cell leukemia virus type I (HTLV-I)-transgenic (Tg) mice and interleukin-1 receptor antagonist (IL-1Ra)-knockout (KO) mice [3, 4]. HTLV-I is the causative agent of adult T-cell leukemia. The virus encodes a transcriptional transactivator, Tax, within the pX region that activates multiple cellular genes, including those for cytokines, cytokine receptors, and immediate early transcriptional factors, via activation of enhancers such as cAMP-responsive enhancer, nuclear factor kappa B-dependent enhancers, or serum-responsive elements [5, 6]. Tg mice carrying the tax gene spontaneously develop autoimmune arthritis, likely due to overexpression of proinflammatory cytokines and increased T-cell resistance to Fas-induced apoptosis [2, 3, 7]. IL-1Ra is a negative regulator of IL-1 which competes for the binding of IL-1α and IL-1β to its cognate receptors. Because the three isoforms of IL-1Ra protein, which possess inhibitory activity against IL-1, are synthesized by alternative splicing of a single gene, we produced mice deficient in all three isoforms of IL-1Ra. These IL-1Ra-KO mice also spontaneously develop autoimmune arthritis, due to excess T-cell activation [2, 4, 8].

Although the etiology of the arthritis differs between these mice, the histopathologies of the lesions are very similar. These lesions exhibit marked synovial and periarticular inflammation, with articular erosion caused by the invasion of granulation tissues, which closely resembles RA in humans. Osteoclast activation is obvious at the pannus, and the infiltration of inflammatory cells, including neutrophils, lymphocytes, and macrophages, can be detected in synovial tissues. Both of these mouse models develop autoimmunity with elevated antibody titers against immunoglobulin (Ig) G and type II collagen. Given that the histopathology observed in these models closely resembles that seen in RA in humans, pathogenic mechanisms similar to those operating in these models are likely functioning in human RA. Actually, an etiological correlation was suggested between HTLV-I and RA in Japan [9, 10]. In addition, an association was suggested between IL-1Ra polymorphism and RA [11, 12]. We took advantage of these mouse models of RA to analyze comprehensively the gene expression patterns functioning in this condition, using high-density oligonucleotide arrays.

In this analysis, we focused on genes that exhibited similar changes in both of the disease models. This approach should efficiently identify the genes involved in the pathogenesis of arthritis irrespective of the etiology, and these genes should include those that function in the effector phase of inflammation or in the bone erosion process. To determine the genomic distribution of the arthritis-related genes, we assigned these genes into the whole genome. The members of the same gene family often form clusters on the chromosomes [13, 14]. Furthermore, because relatively wide genomic regions form open complex structures upon activation [15, 16], we expected that genes in the same cluster might be functionally related. Using this analysis, we identified several arthritis-related gene clusters in the specific genomic regions, and some of the genes were successfully detected as a cluster whose expression changes are only marginal.

To identify genes involved in the pathogenesis of arthritis, we used two mouse models of RA. We compared the gene expression profiles between arthritic and normal joints using these models and high-density oligonucleotide arrays, in which approximately 36,000 genes and ESTs were analyzed. We analyzed whole synovial tissues instead of using specific cell types because we are interested in not only expression level changes of a gene in a cell but also cell population changes in the synovial tissues [26]. In this report, we examined two mouse models of RA on the same genetic background and with the same disease severity. These models, however, had different etiologies; one is caused by the action of HTLV-I-tax, whereas the other is caused by excessive IL-1 signaling. We extracted common genes that were involved in the pathogenesis of both models irrespective of etiology. We expected to identify genes involved in the effector phase of the disease, given that the molecular mechanisms of the initial phase are likely to be different between the two models. With RNA samples derived from the arthritic joints of either animal model, 1,467 clones on the array (approximately 4%) changed at least threefold. The SAM method demonstrated that, of these genes, 554 independent genes changed significantly in both models. These results suggest that a large proportion of the genes functioned in the pathogenesis of arthritis in both models. These common genes may function during the effector phase, whereas those specific to the individual models may function during the initiation phase in a manner dependent on the disease etiology. We analyzed those genes whose changes in expression levels were common to both models.

We found that several interesting genes, including Saa3, which encodes serum amyloid A3, were activated in these models. This gene was upregulated to the greatest extent in arthritic mice in comparison with normal mice. IL-1β induces Saa3 expression; SAA3, in conjunction with SAA1/SAA2, the expressions of which were also elevated in our RA animal models, induce the transcription of matrix metalloproteinases (MMPs) [27]. Mmp-3 and Mmp-9, two such MMPs, were also upregulated in these models. We also observed the upregulation of chemokines, such as Cxcl5 and Cxcl1, which recruit neutrophils to inflammatory sites through interactions with their receptor, Cxcr2, which was also elevated and confirmed by northern blot analysis. Cxcr4 may be involved in the chemotaxis of naïve and memory T cells. Ccr1 and Ccr2 are expressed on memory T cells, whereas Ccr5 is expressed on Th1 cells. Cxcl12, the ligand for Cxcr4, is produced in the synovial tissues of patients with RA [28]. Ccr6, Ccr1,Ccr2, and Ccr5 are specifically expressed by immature dendritic cells (DCs). Cxcl16, a T-cell chemoattractant, is expressed by both DCs and macrophages. Ccl9, Ccl6, and Cxcl13 are produced by macrophages, whereas Cxcl14 is produced by fibroblasts. In addition, proinflammatory cytokines and their cognate receptors, such as IL-1^β, IL-1RI, TNF-α R, IL-6Rα, IL-2R^γ, and IL-17R, were also significantly elevated. Thus, the augmented expression of chemokines, cytokines, and their receptors indicates the importance of these molecules in the pathogenesis of arthritis. It should be noted, however, that the augmentation of the expression of these genes does not necessarily mean that they are actually activated in synovial tissues. The infiltrated cells to the inflammatory sites may also contribute to the expression pattern. Elevated expression of serum amyloid proteins, metalloproteinases, chemokines, and cytokines is frequently seen in the joints of patients with RA [27, 29], indicating that the gene expression profiles obtained from these RA models well represent those of patients with RA. In addition, many of these genes function during the elicitation of inflammation, supporting our assumption that the genes augmented in both models may function during the effector phase.

The gene expression profiles of the well-known collagen-induced arthritis (CIA) model [30] and proteoglycan-induced arthritis (PGIA) model [31] for RA have already been reported. We compared our data set with those previously reported gene expression profiles. Approximately 60% of the genes that changed in our models also changed more than two times at the early phase of CIA. Moreover, approximately 63% of the genes corresponded to those found in PGIA at the acute phase and approximately 50% at the initiation and chronic phases. Thus, many of the genes that changed in our models also changed in other RA models, suggesting similar mechanisms function in common in those RA models.

We next mapped these arthritis-related genes into chromosomes. Working under the assumption that functionally related genes form clusters, we attempted to detect the activation of genes as a cluster, despite the fact that small individual changes were not prominent enough to be detected in our initial analysis. The most significant peak was detected at chromosome 17, 33- to 35-Mb locus (#1–3), which corresponds to the H-2 gene cluster. Other significant peaks corresponded to chromosome 11, 81- to 83-Mb locus (#4–6), which includes members of the Ccl and Slfn families. Immunoglobulin kappa chain cluster in chromosome 6, 68- to 70-Mb locus, and heavy chain cluster in chromosome 12, 112- and 113-Mb locus were also clearly detected. Moreover, chromosome 6, 124-Mb locus, chromosome 15, 79-Mb locus, and chromosome 19, 11-Mb locus were included in those attractive gene clusters. The contribution of individual genes in these clusters was relatively small; the significance of a number of genes was recognized only after summation of genes in a specific region. These gene clusters, however, clearly contained a number of genes important in the development of arthritis. For more than two decades, the MHC gene cluster has been known to affect susceptibility to a variety of autoimmune diseases [32, 33]. This region encodes the MHC class I and class II genes and other immune-related genes. MHC molecules are required for antigen recognition by lymphocytes, ultimately leading to activation and progression of immune responses. In addition, the genes encoding complement components are located within this cluster. The levels of antibodies against IgG and type II collagen were elevated in these RA models [4, 7]. In these animal models, the classical pathway components of the complement system, C2, C4, encoded within this cluster (Chr. 17, 34 Mb: [#2]), and C1q, C1s, and C1r, found within other clusters, were upregulated, suggesting that immune complexes are involved in this enhancement of gene expression. The expression of Factor B, encoded by the H2-Bf gene within this cluster, was also augmented. Factor B is an essential component of the alternative pathway of the complement system. This pathway is also critical in K/BxN mice, another animal model for RA [34]. Although the alternative pathway typically activated by microbial surface antigens, immune complexes formed by autoreactive IgGs initiate the alternative pathway in these models to facilitate the development of arthritis.

A similar study was reported using CIA and PGIA models, identifying gene clusters in chromosomes 2, 3, 11, and 17 [35]. Their microarray chips contained a total of 9,500 known genes and EST clones, and 203 selected genes were mapped into the chromosome using a 1.5-fold differential expression threshold level. In the present study, 36,000 oligonucleotide probe sets were included in microarray chips, and 550 selected common significant genes were mapped. Interestingly, we detected the same clusters in chromosomes 2, 11, and 17 as were found in their study, suggesting that the same genes are involved in the pathogenesis of arthritis regardless of the etiology. Although they found a cluster on chromosome 3 containing 11 genes within an 8-Mb-long region, this region was not assigned as an arthritis-related gene cluster in our study, because of broad distribution of genes among cytobands. Increased expressions of MHC class I and class II genes, complement genes, and chemokine genes were already reported using microarray analysis ofsynovium from patients with RA [36] or streptococcal cell wall-induced arthritis in rats [37]. Our results in this report using mouse arthritis models are consistent with these results, suggesting that these genes are commonly involved in the pathogenesis at the elicitation phase.

The gene density peak on chromosome 11, 81- to 83-Mb locus (#4–6) includes the Ccl and Slfn genes. CCL2, CCL7, and CCL8, members of the chemokine family, recruit monocytes to sites of injury and infection. CCL2 influences innate immunity through this effect on monocytes and modulates adaptive immunity via control of Th2 polarization. An antagonist of CCL2 suppresses arthritis in the MRL-lpr mouse model [38]. CCL5, a T-cell and monocyte chemoattractant, plays an important role in the development of adjuvant-induced arthritis [39]. CCL6 was expressed in experimental inflammatory demyelinating disorders that promote recruitment of macrophages [40]. CCL9 recruits CD11b⁺ DCs and promotes osteoclast differentiation and survival [41, 42]. Thus, these chemokines likely play important roles in the development of arthritis. Schlafen proteins have been implicated in the regulation of cell growth and T-cell development [43]. These proteins form a big family localized to a specific genomic cluster adjacent to the Ccl gene cluster. This region coincides with a number of autoimmune susceptibility loci, including the Idd4, Eae7, and Orch3 loci [44]. The syntenic region of this locus on 17q of the human genome is associated with several autoimmune disorders [45]. Therefore, this genomic region on chromosome 11 is likely to be critical in the development of autoimmune arthritis.

The chromosome 6, 124-Mb (#12) locus is one of the non-MHC susceptibility loci linked to complex autoimmune diseases. The mouse 6 F2 cytoband, the syntenic rat 4q42 and human 12p13 regions, have been implicated in several inflammatory diseases, including arthritis [46–48], systemic lupus erythematosus, spontaneous diabetes, atheroscrelosis, encephalomyelitis, asthma or airway hyper-responsiveness, and allergy. This locus contains complement component genes and C-type lectin superfamily genes, suggesting the possible involvement of these protein products in the development of arthritis. Chromosome 15, 79-Mb locus (#13) contains GM-CSF receptor, Csf2rb1, and Csf2rb2. It is known that GM-CSF plays an important role in the effector phase of arthritis [49]. Ms4a family molecules, CD20 (Ms4a1), and high-affinity IgE receptor ^β chain (FcΣ RIb; Ms4a2) tended to be upregulated in these animals, but these genes were not significantly altered in this study. Although the upregulation of Ms4a7, Ms4a4d, Ms4a6d, and Ms4a6c expression in chromosome 19, 11-Mb locus (#14) was also observed in these models of RA, the roles of these molecules in the development of arthritis are not yet known [50].

We succeeded in identifying several arthritis-related genes by transcriptome mapping, but we might have failed to detect some potentially important genes. Because we analyzed mRNA preparations from whole joints, but not from a single species of cell, we could not detect the change of gene expression in a single species of cell when the content of this single species of cell was low in the whole joint. Furthermore, we could not detect the MMP gene cluster consisting of MMP-1a, MMP-1b, MMP-3, and MMP-13 on chromosome 9, which were suggested to be involved in the pathogenesis [51]. In this case, we demonstrated upregulation of MMP-3 and MMP-9 in both HTLV-I-Tg and IL-1Ra-KO mice. However, MMP-8 and MMP-13 expressions were increased more than three times in IL-1Ra-KO mice, but not in HTLV-I-Tg mice, excluding the MMP-8 and MMP-13 genes from significantly activated genes in this analysis. Unfortunately, MMP-1a and MMP-1b were not mounted on the GeneChip we used (Murine Genome U74v2 Set). Nonetheless, the present data are important for understanding the transcriptome change as a whole, reflecting the gene expression levels not only in a single cell but also in the cell population in the joints during arthritis.

We have performed a comprehensive transcriptome analysis of two etiologically different RA models by using microarrays and identified many genes that were commonly activated in both models. Because the initial mechanism of the pathogenesis should be different between two models, these genes are likely to function during the effector phase of the disease. Moreover, by mapping genes with marginal expression changes into chromosomes, we succeeded in detecting several gene clusters that may be involved in the pathogenesis of arthritis. A similar approach, using microarrays, will likely be useful in detecting genes related to multiple disease processes.