α₉ Integrin and Its Ligands Constitute Critical Joint Microenvironments for Development of Autoimmune Arthritis¹

Rheumatoid arthritis (RA)⁴ is a complex, multisystem disease. The role of inflammatory cells in the pathogenesis of RA has been well established. However, the significance of resident cells within the synovial membrane has only recently begun to be fully appreciated (1, 2). One important characteristic of synovial cells is their attachment to the cartilage extracellular matrix (ECM) proteins, facilitating their invasion into cartilage and bone tissues to cause joint destruction. The interaction of those synovial cells with ECM proteins is mediated by adhesion molecules such as integrins. Integrins mediate not only cell attachment but also intracellular signaling pathways (3, 4). Osteopontin (OPN) has been characterized as an ECM protein. However, OPN is now recognized as one of the matricellular proteins that, unlike classical ECM proteins, are soluble proteins and exert proinflammatory responses (5, 6, 7). Indeed, overexpression of OPN has been associated with various inflammatory disorders in human and mice including Crohn’s disease, multiple sclerosis, RA, and autoimmune arthritis (8, 9, 10, 11, 12, 13, 14, 15, 16). Another matricellular protein, tenascin-C (TN-C), is also significantly up-regulated at the pathologic foci of RA (17). The lack of OPN and TN-C abrogates the development of Th1 type immunity-mediated diseases, including autoimmune arthritis and hepatitis in mice (18, 19, 20). OPN and TN-C each bind to multiple receptors but share a common receptor, the α₉β₁ integrin (21). It has been shown that the α₉ integrin is involved in the transendothelial migration of neutrophils into inflammatory foci in human and in the generation of osteoclasts in both human and mouse (22, 23). However, it is unclear whether the interaction between the α₉ integrin and its ligands, OPN and TN-C, within the joint tissue microenvironment plays any mechanistic role for the development of inflammatory autoimmune arthritis. Collagen Ab-induced arthritis (CAIA) has recently been used extensively to study the inflammatory phase of arthritis in which the critical involvement of not only IL-1 but also OPN has been documented (18, 24, 25). Therefore, we examined how OPN and TN-C and their receptor, the α₉ integrin, can be integrated into the pathogenesis of arthritis.

In this study we provide evidence that the α₉ integrin and its ligands are critical intrinsic regulators that control the development of autoimmune arthritis. In the arthritic microenvironment, both synovial fibroblasts and macrophages express the α₉ integrin. There are more fibroblasts than macrophages in arthritic joints, and fibroblasts are the major sources of locally produced OPN and TN-C. Importantly, the α₉ integrin differentially regulates the production of cytokines and chemokines by each cell type. α₉-mediated production of these cytokines and chemokines then presumably contributes to the recruitment of inflammatory cells into the synovium, osteoclast activation, synovial hyperplasia, pannus formation and, cartilage and bone destruction. Importantly, inhibition of the α₉ integrin function with an inhibitory anti-α₉ integrin Ab significantly reduces the production of arthrogenic cytokines and chemokines and ameliorates ongoing arthritis. These results suggest that the α₉ integrin serves as a novel therapeutic target for the treatment of autoimmune disorders, including RA.

Anti-human α₄ and α₉ integrin Abs were purchased from Chemicon. PE- or FITC-Mac-1, FITC-Gr-1, FITC-VCAM-1, PE-TLR-4, biotinylated anti-class II, and streptavidin-allophycocyanin were purchased from BD Pharmingen. Anti-murine α₄ integrin Ab and FITC-F4/80 from eBioscience and Serotec, respectively. Anti-vimentin and anti-smooth muscle actin (SMA) Abs were from Sigma-Aldrich. PE-anti-human Mac-1 was purchased from Boehringer Mannheim. An Ab (D7-FIB) that reacts to human fibroblasts (26) was obtained from Serotec. Anti-human OPN, anti-mouse OPN, and anti-human TN-C Abs used for immunohistology were purchased from Immuno-Biological Laboratories.

A murine α₉ integrin was cloned into the pcDNA3.1 vector from cDNA derived from B16-BL6 cells using a pair of primers, α₉ integrin-5′ (5′-GCTAAGCTTCTGGGGATGGGCGGCCCGGCT-3′) and α₉ integrin-3′ (5′-CCATCTAGATCACTGGTTTTTCTGGACCCA-3′). A murine α₄ integrin was cloned into the pcDNA3.1 vector from cDNA derived from B16-BL6 cells using the primer pair α₄ integrin-5′ (5′-CGTGGATCCGAGCGCATGGCTGCGGAAGCGAGGTGC-3′) and α₄ integrin-3′ (5′-CAGCTCGAGTCAGTCATCATTGCTTTTGCTGTTGAC-3′). These constructs were transfected to Chinese hamster ovary (CHO)-K1 cells or NIH3T3 cells with Lipofectamine 2000. Stable transfectants were selected by G418 (Sigma-Aldrich). CHO-K1 cells or NIH3T3 cells, transfected with murine α₉ (or α₄) integrin were referred to as α₉/CHO or α₉/NIH (α₄/CHO or α₄/NIH), respectively.

Armenian hamsters were immunized with α₉/CHO cells and splenocytes were fused with X63-Ag8-653 mouse myeloma cells according to the standard procedure. Hybridoma cells, secreting IgG Abs specifically bound to α₉/NIH cells but not to α₄/NIH cells, were recovered.

Total RNA was isolated from ankle joints, synovial fibroblasts, or macrophages from ankle joints using TRIzol (Invitrogen) and first-strand cDNA was generated with a first-strand cDNA synthesis kit (GE Healthcare Biosciences). Real-time quantitative PCR was performed using the LightCycler FastStart DNA Master SYBR Green I system (Roche Diagnostics). Some of the specific primers using in this study were shown in supplemental Table I.⁵ The expression level of mRNA was calculated by the calibration curve method using LightCycler software version 3. Data were standardized by G3PDH.

Arthritis in BALB/c mice was induced by using an arthritogenic mAb cocktail kit (Immuno-Biological Laboratories) and treatment protocol was depicted in supplemental Fig. 1. The clinical severity of arthritis was graded in each of the four paws on a 0–4 scale. The disease severity was recorded for each limb as follows: 0, normal; 1, focal slight swelling and/or redness in one digit; 2, moderate swelling and erythema of more than two digits; 3, marked swelling and erythema of the limb; and 4, maximal swelling, erythema, deformity and/or ankylosis. The appearance ratio of arthritic limb was calculated by following formula; (the number of arthritic limb/the number of all limbs) × 100.

Histological assessment of the arthritic joints was conducted using sections stained with Fast Green/safranin-O or H&E. The degrees of synovial proliferation, leukocyte infiltration, and cartilage degeneration were graded as follows: 0, normal; 1, mild proliferation of synovium and minor leukocyte infiltration into synovium or minor erosion of the cartilage; 2, invasion of synovium into the joint space and moderate leukocyte infiltration or mild pannus erosion of the cartilage; and 3, extensive leukocyte infiltration into the joint space, fibrous ankylosis of the joints, and peripheral and subchondral cartilage erosion. In some experiments, content of the cartilage matrix proteoglycan was evaluated by safranin-O staining of joint sections and scored arbitrarily as 0 when normal or as 1–3 according to the degree of depletion (loss of staining). In other experiments, serial frozen sections were made from joints both treated with normal hamster IgG (NHG) or anti-α₉ integrin Ab. Sections were fixed with cold acetone and were subjected to tartrate-resistant acid phosphatase (TRAP) staining. The expression of OPN or TN-C in mouse or human synovial tissues was assessed by immunostaining of synovial tissues using specific Abs according to the methods described previously (8, 20).

Murine IL-6, IL-1α, and TNF-α production in culture supernatant was measured by using ELISA kits as specified by the manufacturers (BD Bioscience or R&D Systems). Murine OPN, thrombin-cleaved form of OPN (OPN N half), and TN-C expression were measured by ELISA (Immuno-Biological Laboratories).

An amino-terminal half of human OPN (nOPN; I17–R168) in which RGD was mutated to RAA (nOPN/RAA) (27) was cloned into the pGEX6P-1 vector from the full-length OPN cDNA as a template using the following pair of primers: OPN-5′ (5′-TCTGGATCCATACCAGTTAAACAGGCT GAT-3′) and OPN(RAA)-3′ (5′-GCTCTCGAGTTACCTCAGTCCATAAACCACACTAGCAGCTCG-3′. A third fibronectin type III repeat domain of human TN-C (R⁷⁹⁷–P⁸⁹⁶) in which RGD was mutated to RAA (TN-Cfn3/RAA) (28) was cloned into pGEX6p-1 from cDNA derived from HT-1080 cells using the following pair of primers; TN-C-5′ (5′-GTTGGATCCAGGGTGACCACCACACGCTTG-3′ and TN-C (RAA)-3′ (5′-TCTCTCGAGTTAGGGAGCATCGAGGCCTGTTGTGAAGGTCTCTTTGGCTGGGTTGCTTGACATGGCAGCTCT-3′). TNCΔ was cloned into the pGEX6p-1 vector from TN-Cfn3/RAA construct as a template using the following pair of primers: TNCΔ-5′ (5′-GCCTTGATCACCTGGTTCAAGCCCACCTACGGCATCAAAGACGTGCC-3′) and TNCΔ-5′ (5′-GGCACGTCTTTGATGCCGTAGGTGGGCTTGAACCAGGTGATCAAGGC-3′). pGEX plasmid were transformed to JM109-competent cells and the GST fusion proteins were purified with glutathione-Sepharose 4B beads. The proteins were then cut off from GST with PreScission protease (GE Healthcare).

Cell binding assays were performed as described previously (29). Briefly, the 96-well plates were precoated with BSA or mutated forms of human nOPN/RAA or TN-Cfn3/RAA (1.25 μg/ml) or, in some experiments, with the synthetic peptides SVVYGLR, AEIDGIEL, or GRGDS (10 μg/ml), followed by blocking with 1% BSA in PBS for 1 h at room temperature. After washing with PBS, synovial fibroblasts or α₉/NIH cells (1 × 10⁴) were added to the wells and incubated at 37°C for 1 h. The adherent cells were fixed and stained with 0.5% crystal violet in 20% methanol for 30 min and then lysed with 20% acetic acid. The resultant supernatants from each well were analyzed by an immunoreader and absorbance at 590 nm was measured to determine the relative number of cells adhered to wells.

Murine synovial fibroblasts and macrophages (3 × 10³) were cultured on plates coated with mutated forms of human nOPN/RAA, TN-Cfn3/RAA, or TN-CΔ (2.5–10 μg/ml) for 48 h in FCS-free medium. Cells were lysed with TRIzol reagent (Invitrogen) and subjected to the analysis of cytokine gene expression using real-time PCR. Culture supernatants were subjected to the measurement of cytokine and OPN or TN-C production by ELISA.

In Fig. 1,D, first the ankle joint portion of a leg was surgically cut and separated from other area and then the dermal, subcutaneous, tendinous, and muscle tissues were surgically taken off from joints under a dissecting microscope. The remaining soft tissues were surgically separated from bone tissues and were used as synovial tissues. In Figs. 2 C and 5–7, these soft tissues were treated with 0.1% type II collagenase (Worthington Biochemical) at 37°C by vigorous stirring. After washing, synovial cells were cultured in 10% FCS containing medium at 37°C overnight. Attached cells were regarded as “adherent cells” and unattached cells were regarded as “nonadherent cells” and were used in Fig. 5 for expression analysis of cell surface Ag. For functional analysis, synovial cells were fractionated into TLR-4⁺ or TLR-4⁻ fractions by using MACS (Miltenyi Biotec). The TLR-4⁺ fraction contained ∼90% Mac-1⁺Gr-1⁻ cells and was thus used as a macrophage in Fig. 7 and supplemental Fig. 4. The TLR-4⁻ fraction was cultured overnight and nonadherent cells were discarded. Adherent cells were enriched for Mac-1⁻Gr-1⁻ cells (∼90%) and used as fibroblasts in Fig. 6 and supplemental Fig. 4. Human synovial specimens were obtained through synovectomy from patients who fulfilled the American College of Rheumatology criteria for RA. Informed consent was obtained from all study subjects before sample collection. The study protocol was approved by the institutional review board at the Hokkaido University Graduate School of Medicine. Human synovial cell suspension was made by the method described above for mouse synovial cells.

Synovial tissues obtained from normal or arthritic joints were pulverized by a homogenizer (Polytoron PT1600E; Kinematica) in 12.75 μl/mg PBS containing a protease inhibitor cocktail (Roche Diagnostics). The suspension of synovial tissues mixed with 1.5 μl/mg tissue 10% Triton and 0.75 μl/mg 10% Nonidet P-40 were vortexed and incubated on ice for 30 min. Then, the suspension were centrifuged at 15,000 rpm for 30 min and supernatant was collected. OPN and TN-C concentrations were measured by ELISA.

Murine synovial cells were stained with PE-Mac-1 and FITC-Gr-1. After washing, cells were further stained with biotinylated anti-mouse α₄ or α₉, followed by streptavidin-allophycocyanin. In other experiments, cells were stained with FITC-Mac-1 and PE-TLR-4, FITC-Mac-1 and PE-VCAM-1, or PE-Mac-1 and FITC-F4/80. In some experiments, cells were stained with FITC-Mac-1 and biotinylated anti-mouse class II, followed by streptavidin-allophycocyanin. Human synovial cells were stained with FITC-D7-FIB and PE-Mac-1. After washing, cells were further stained with biotinylated anti-human α₉ integrin, followed by streptavidin-allophycocyanin as described above. All analyses were performed on a FACSCalibur flow cytometer (BD Biosciences) with FlowJo software (Tree Star).

Statistical evaluation was performed based on the Student’s t test to compare the differences between groups. For calculation of the appearance ratio of arthritic joint, χ² statistics were derived using the CHIDIST function of Microsoft Excel. Single (∗) and double asterisks (∗∗) indicate that values are significantly different between control and experimental group, with values of p < 0.05 or p < 0.01, respectively.

The murine model of RA was induced by i.v. administration of a mixture of four anti-type II collagen mAbs followed by i.p. administration of LPS. We surgically dissected arthritic joints and carefully obtained synovial tissues. Gene expression of both the α₉ integrin and its ligands, OPN and TN-C, was significantly up-regulated in synovial tissues of arthritis joints after LPS injection (Fig. 1,A). α₄ integrin gene expression was also induced. OPN and/or TN-C proteins were clearly expressed by hyperplastic synovial tissues obtained not only from arthritic mice but also from RA patients (Fig. 1, B and C, respectively). To confirm that synovial tissues produce OPN and TN-C, synovial tissues were obtained at the indicated time points and OPN and TN-C contents were examined by ELISA. Compared to normal synovial tissues, synovial tissues obtained on days 3 and 6 produced significantly more OPN and TN-C. OPN production was ∼2- to 3-fold higher than production of TN-C. Of note, the thrombin-cleaved form of OPN (designated as OPN N half), which is capable of interacting with α₉ integrin (27), was increased in arthritic joints at days 3 and 6 but was not found in normal joints (Fig. 1 D).

A detailed functional study of the α₉ integrin in autoimmune arthritis has been hampered by the early postnatal mortality of the α₉ knockout (30) and the lack of specific mAbs recognizing the murine α₉ integrin. Thus, we generated both inhibitory and noninhibitory mAbs the specifically react to the murine α₉ integrin (supplemental Fig. 2). Anti-α₉ integrin Abs could react with murine fibroblasts transfected with the α₉ integrin (α₉/NIH) but not with those transfected with the α₄ integrin (α₄/NIH). In addition, those Abs could detect endogenous α₉ integrins expressed by the B16-BL6 murine malignant melanoma cells. Importantly, one of the anti-α₉ integrin Abs (55A2C) could inhibit the binding of α₉/NIH cells to the synthetic peptides SVVYGLR and AEIDGIEL but not to the GRGDS peptide. It has been shown that SVVYGLR or AEIDGIEL are specific ligands for the α₉ integrin and correspond to the internal sequences of OPN and TN-C, respectively (27, 28). The binding of α₉/NIH cells to the GRGDS peptide is mediated by RGD-recognizing integrins and, thus, the anti-α₉ integrin Ab is unable to inhibit it. Thus, a specific anti-α₉ integrin blocking Ab, 55A2C, allows us to evaluate the role of α₉ integrin in the pathogenesis of inflammatory arthritis.

Arthritis was induced by the injection of a cocktail of Abs against type II collagen on day −3 and was followed by LPS injection on day 0. Mice were treated twice with anti-α₉ integrin Ab before the onset of arthritis at days −4 and 0 (supplemental Fig. 1). The joint swelling became evident at day 2 and reached its peak at day 6. The severity of arthritis was significantly reduced by an inhibitory anti-α₉ integrin Ab (55A2C), but not by a noninhibitory Ab (18R18D) (Fig. 2,A). In addition, the appearance ratio of an arthritic limb was also reduced by 55A2C treatment (Fig. 2,B). We then analyzed the cellular composition of arthritic joints at different time points after arthritis induction. Under normal conditions there were four times as many fibroblasts as macrophages, and neutrophils were not detected. Fibroblasts increased ∼2- and then 3-fold at days 3 and 6, respectively, as compared with a normal joint (Fig. 2,C). Nevertheless, macrophages and neutrophils increased at days 3 and 6. However, fibroblasts remained the dominant cell type in arthritic joints up to day 6, when arthritis reached its peak. Of note, the numbers of all cells were significantly reduced by 55A2C treatment (Fig. 2,C). Consistent with the cell counts, histological analysis performed at the early phase of arthritis on day 3 after LPS injection demonstrated the presence of acute inflammation as evidenced by marked interstitial edema and moderate leukocyte accumulation without evident cartilage and bone degeneration. Acute inflammation was significantly attenuated by 55A2C treatment (Fig. 2,D). Expressions of CCL2 and CCL4, chemotactic factors involved in leukocyte recruitment, were significantly augmented at day 1, when arthritis was not yet clinically evident and was significantly reduced by 55A2C treatment (Fig. 2 E).

In the late phase of arthritic joints at day 14 there was marked synovial hyperplasia and inflammatory cell infiltration consisting of mainly macrophages (marked by closed arrowheads; Fig. 3,A) and neutrophils (marked by arrows; Fig. 3,A). We also noticed the pannus formation and the appearance of irregularity on bone surfaces (marked by arrows; Fig. 3,B), indicating the presence of prominent bone absorption. The appearance of those pathological changes was significantly inhibited by 55A2C (Fig. 3, A–C). Joint histology on day 14 clearly demonstrated the presence of erosion and cartilage degeneration (marked by closed arrowheads; Fig. 3,D) and the loss of safranin-O staining (marked by open arrowheads; Fig. 3,D) of a joint section, which indicate the loss of proteoglycan in arthritic joints. Joint degeneration was clearly attenuated by 55A2C (Fig. 3,D). To further evaluate bone absorption, we immunohistochemically detected TRAP-positive multinuclear cells, presumably activated osteoclasts. There were numerous TRAP-positive multinucleated cells along the periosteal surface of the metatarsal bones in arthritic joints (Fig. 3 E). However, those cells were hardly detected in Ab-treated mice, which is consistent with a role for the α₉ integrin in the differentiation and activation of osteoclasts as previously suggested (23, 24).

To mimic a therapeutic intervention as might be relevant to human RA patients, we injected 55A2C after the clinical symptom became evident on day 3, resulting in a marked reduction of the clinical score of arthritis (Fig. 4,A) that was consistent with the improvement of joint histology such as synovial hyperplasia, inflammatory cell infiltration, pannus formation, and bone and cartilage destruction at day 14 (Fig. 4, B and C). The results indicated that the α₉ integrin is also involved in late phase of arthritis where sustained leukocyte infiltration, synovial hyperplasia, and cartilage and bone destruction dominate local pathology.

To elucidate molecular mechanisms that might account for the attenuation of autoimmune arthritis by therapeutic administration of an anti-α₉ integrin Ab, we examined the expression of α₉ integrin and its ligands in arthritic joints at day 9. The expression of the α₉ integrin and its ligands was not significantly altered by 55A2C treatment, whereas the expression of the α₄ integrin was significantly reduced (Fig. 4,D). Therefore, it is likely that 55A2C interferes with the ligand-receptor interaction without affecting expression of the α₉ integrin and its ligands. In arthritic joints, the expressions of various cytokines that are involved in the pathogenesis of inflammatory arthritis are up-regulated at day 9 (Fig. 4,E). Importantly, the augmented expressions of IL-1β, IL-6, and TGF-β in arthritic joints were significantly down-regulated by 55A2C treatment. In contrast, the expressions of IFN-γ and TNF-α were not inhibited by 55A2C treatment (Fig. 4,E). We also evaluated the expressions of various chemokines that are chemotactic for monocytes, macrophages, and neutrophils at day 9 (Fig. 4,F). The expressions of CCL2 and CXCL2 were not augmented in arthritic joints at day 9 (note that CCL2 was significantly expressed at day 1; see Fig. 2 E). However, the expression of CXCL5, which shares the common receptor CXCR with CXCL2, was augmented at day 9 and was significantly reduced by 55A2C treatment. The expressions of CCL3 and CCL4, which share the common receptor CCR5, were augmented at day 9 and were significantly reduced by Ab treatment. In addition, the expressions of CXCL12 and CXCL14 were also augmented in arthritic joints and were again reduced by Ab treatment. Taken together, our results indicated that an anti-α₉ integrin Ab inhibits the production of cytokines and chemokines in arthritic joints, leading to the inhibition of macrophage and neutrophil recruitment into joints.

We then examined which cell components within arthritic joints express the α₉ integrin. Synovial tissues obtained at day 6 were fractionated into adherent and nonadherent fractions. The majority of adherent cells were Mac-1⁻Gr-1⁻ cells (Fig. 5,A) and were positive for vimentin and negative for anti-SMA and exhibited morphological features of synovial fibroblasts (Fig. 5,B). Those synovial fibroblasts express the α₉ but not the α₄ integrin (see “R3” in Fig. 5,A). Approximately 20% of adherent cells are Mac-1⁺Gr-1⁻ macrophages and express both α₉ and α₄ integrins (see “R1” in Fig. 5,A). There is a small fraction of Mac-1⁺Gr-1⁺ neutrophils. Unexpectedly, we found that synovial neutrophils express the α₄ integrin but do not express the α₉ integrin (see “R2” in Fig. 5,A). The nonadherent cell population also consisted of Mac-1⁺Gr-1⁻ macrophages, Mac-1⁻Gr-1⁻ fibroblasts, and Mac-1⁺Gr-1⁺ neutrophils, which represent a major population (Fig. 5,A). The integrin expression profiles are indistinguishable from those of the corresponding subsets in the adherent cell population (data not shown). We next fractionated the adherent cell population into Mac-1⁺ (see “R1” in Fig. 5,C) and Mac-1⁻ (see “R2” in Fig. 5,C) populations. Mac-1⁺ adherent cells expressed F4/80, and thus those cells were enriched for macrophages. Synovial macrophages express class II and TLR-4 but little if any VCAM-1 (Fig. 5,C). In contrast, the Mac-1⁻ cell population was negative for F4/80 and thus those cells were enriched for fibroblasts. Synovial fibroblasts clearly expressed VCAM-1 but little if any TLR-4 (Fig. 5,C). We also examined whether human arthritic synovial cells express the α₉ integrin. Synovial cells obtained from RA patients were fractionated into Mac-1⁺D7-FIB⁺ synovial macrophages (se “R1” in Fig. 5,D) and Mac-1⁻D7-FIB⁺ synovial fibroblasts (see “R3” in Fig. 5,D). Both human synovial macrophages and fibroblasts express significant surface levels of the α₉ integrin (Fig. 5 D). It should be noted that both α₉ and α₄ integrins specifically bind to VCAM-1 (21, 31). Thus, our data indicate that VCAM-1 expressed by synovial fibroblasts may provide an anchor for cell attachment and the recruitment of inflammatory cells that express its counter receptor, the α₄ integrin (neutrophils and macrophages) and/or the α₉ integrin (macrophages).

Synovial fibroblasts are the predominant cell components within arthritic joints (Fig. 2,C). Therefore, we first attempted to clarify the role of synovial fibroblasts in the pathogenesis of inflammatory arthritis. Murine synovial fibroblasts obtained from arthritic joints bound to recombinant OPN (nOPN/RAA) and TN-C (TN-Cfn3/RAA) proteins, which are ligands for α₉ integrins (Fig. 6,A). The recombinant OPN and TN-C fragments used in these experiments were genetically modified so that the RGD sequence, which can be recognized by an RGD-recognizing integrin, was replaced by RAA, thus allowing the study of α₉ integrin-dependent cell adhesion without the confounding effects of RGD-binding integrins (27, 28). In addition, the binding was specifically inhibited by an inhibitory anti-α₉ integrin Ab. These data indicate that the α₉ integrin can mediate the cell adhesion of synovial fibroblasts to OPN and TN-C, enabling the strong interaction of fibroblasts with a joint ECM, which culminates in the destruction of cartilage and bone. We also found that the detachment of fibroblasts, both NIH3T3 transfected with the murine α₉ integrin (α₉/NIH) and the parent NIH3T3, resulted in the significant cell death in vitro (detachment induced-cell death or anoikis) and the anoikis was significantly inhibited in α₉/NIH but not in α₉-negative fibroblasts (NIH3T3) by the synthetic SVVYGLR peptide, which corresponds to the α₉ integrin recognition sequence within OPN (27) (supplemental Fig. 3). The rescue from anoikis of α₉/NIH fibroblasts by the SVVYGLR peptide was reversed by the addition of an anti-α₉ integrin Ab, indicating that the rescue from anoikis is α₉ integrin dependent. Of note, the thrombin-cleaved form of OPN, which exposes the SVVYGLR sequence, is elevated in arthritic joints (Fig. 1 D) and in RA patients, and detached fibroblasts are frequently seen in the synovial fluid of RA patients (2, 32, 33). Thus, our data favor the hypothesis that synovial hyperplasia is enhanced by the impairment of synovial fibroblast cell death through the interaction between the α₉ integrin and its ligands.

We next asked whether the α₉ integrin on synovial fibroblasts contributes to the production of the proinflammatory cytokines and chemokines detected at arthritic joints (Fig. 4, E and F). The stimulation of synovial fibroblasts by the α₉ integrin ligands OPN and TN-C resulted in augmented gene expression of proinflammatory cytokines such as IL-1α and IL-6 (Fig. 6,B). These cytokines were previously shown to be involved in the activation and recruitment of inflammatory cells (34, 35). However, the expressions of IFN-γ, TNF-α, and IL-1β were not augmented. Among the seven chemokines tested, expressions of CCL2, CXCL5, and CXCL12 were induced by both OPN and TN-C stimulation, while CCL4 expression was slightly induced by TN-C stimulation (Fig. 6,C). We also found that matrix metalloproteinase (MMP)-9 expression was induced by OPN and TN-C stimulation of synovial fibroblasts. In contrast, MMP-2 expression was significantly reduced by both OPN and TN-C (Fig. 6,D). MMP-9 has been shown to activate various cytokines including TGF-β and IL-1β (36, 37), thus contributing to the pathogenesis of inflammatory arthritis, whereas MMP-2 inactivates chemokines including CCL7 (38), thus down-regulating inflammatory arthritis (39). Production of IL-6 protein was also clearly induced by OPN and TN-C stimulation, and IL-6 production was significantly inhibited by 55A2C treatment (Fig. 6 E). Thus, our data demonstrate that interaction between the α₉ integrin on synovial fibroblasts and its ligands within joints dramatically modulates fibroblast gene expression in a pattern predicted to contribute to the severity of inflammatory arthritis.

We finally asked whether synovial macrophages also contribute to the production of cytokines and chemokines. Similar to the case of synovial fibroblast stimulation as shown in Fig. 6,B, the expressions of IL-6 and IL-1α by synovial macrophages were induced upon stimulation with OPN and TN-C (Fig. 7,A). In contrast to case of the synovial fibroblasts, TNF-α and IL-1β expressions by synovial macrophages were also augmented. In addition, TGF-β expression was also increased in synovial macrophages derived from arthritic joints (Fig. 7,B). We also found that expressions of CCL2, CCL3, CCL4 CXCL2, and CXCL5, but not those of CXCL12 and CXCL14, were augmented by OPN and TN-C stimulation in synovial macrophages (Fig. 7,C). Note that CCL2 and CXCL5 can be produced by both synovial fibroblasts and macrophages. CCL3, CCL4, and CXCL2 were derived from synovial macrophages, but not from fibroblasts. In contrast, CXCL12 was derived from synovial fibroblasts, but not from macrophages (see Fig. 6,C). We confirmed that not only gene expressions but also protein expressions of IL-6, IL-1α, and TNF-α were induced by OPN and TN-C stimulation of synovial macrophages (Fig. 7,D). The induction of the IL-6 protein by synovial macrophages was mediated by α₉ integrin stimulation, because an Ab specific for the α₉ integrin could inhibit the production of IL-6 (Fig. 7,E) and a mutated form of recombinant TN-C lacking the domain responsible for interaction with the α₉ integrin was unable to induce the IL-6 protein (Fig. 7 F). Taken together, these data support the notion that synovial fibroblasts and macrophages contribute differently to the production of cytokines and chemokines through the interaction between α₉ integrin with its ligands.

Integrins make up a large family of cell surface receptors composed of at least 24 different α/β heterodimers that mediate cell-to-cell and cell-to-matrix interactions and signal transduction (3, 4). Although most members of the integrin family are capable of mediating cell migration (40), the α₄ integrin has been shown to be involved in directing leukocyte trafficking to sites of inflammation (41). Importantly, in vivo studies using blocking mAbs and antagonistic peptides have successfully demonstrated that antagonism of the α₄ integrin holds promise as an effective therapeutic approach to patients suffering from asthma, RA, multiple sclerosis, and inflammatory bowel diseases (42, 43). The α₄ integrin is structurally most closely related to the α₉ integrin (44). Interestingly, the α₉ and α₄ integrins share common ligands such as VCAM-1 and OPN (21, 22, 31, 45). Nevertheless, except for the interaction between VCAM-1 and the α₄ integrin, the physiological and pathological importance of the interaction between α₉ integrins and their ligands are not fully appreciated.

CAIA has recently been used extensively to study the inflammatory phase of arthritis. Pathogenesis of the CAIA model used in this study has been primarily explained by the activation of complement by an immune complex and the subsequent C5a-mediated neutrophil and macrophage infiltration (46). We demonstrated that OPN and/or TN-C were locally produced (Fig. 1,D) and that synovial fibroblasts expressed the α₉ integrin (Fig. 5). Thus, we reasoned that the autocrine and paracrine interactions of OPN and/or TN-C and its receptor, α₉ integrin, on synovial fibroblasts trigger an early phase of inflammatory events leading to the development of arthritis. Once OPN and/or TN-C is produced locally, autocrine and paracrine interaction of the α₉ integrin with its ligands further stimulate fibroblasts to produce more OPN and/or TN-C (supplemental Fig. 4), leading to exacerbation of arthritis.

The α₉ integrin plays a critical role in the recruitment of inflammatory cells in many ways. Firstly, the interaction of the α₉ integrin on synovial cells with OPN and TN-C expressed at the site of inflamed joints results in the production of chemokines that are responsible for the recruitment and activation of inflammatory cells. Importantly, synovial fibroblasts and macrophages contribute differently for the production of various chemokines such as CCL2, CCL3, CCL4, CXCL2, CXCL5, and CXCL12 (Figs. 6 and 7). Other than chemotactic effect, these chemokines are involved in angiogenesis and osteoclastgenesis, which are crucial for the development of arthritis (47, 48, 49). CXCL12 could stimulate vasculogenesis by the recruitment of endothelial progenitor cells, and CCL2 may also promote neovascularization in inflamed tissues (49). Additionally, CCL2 and CCL3 induce the fusion of preosteoclasts, leading to osteoclast formation and activation (50, 51). CXCL12 has key role in the recruitment and activation of the osteoclast precursor (52). Secondly, the α₉ integrin acts as an anchor molecule for cell adhesion. Synovial fibroblasts obtained from arthritis joints express VCAM-1, which serves as a counter-receptor for both α₉ and α₄ integrins (22, 31). Thus, macrophage infiltration may be due to the interaction between α₉ integrin on macrophages and VCAM-1 on synovial cells (Fig. 5,C) and endothelial cells (22). In addition, activated synovial fibroblasts interact with inflammatory cells via VCAM-1, and this interaction favors not only recruitment but also the survival of inflammatory cells by impairing apoptosis (2, 53, 54). It should be noted that we found that synovial neutrophils do not express α₉ integrins (se “R2” in Fig. 5,A). Thus, recruitment of neutrophils into synovial tissues in mice seems to be α₉ integrin independent. Thirdly, the interaction of the α₉ integrin with its ligand leads to the production of various proinflammatory cytokines that are produced differently by synovial fibroblasts and macrophages. IL-1 and IL-6 were derived from both synovial fibroblasts and macrophages, whereas TNF-α and TGF-β are derived from synovial macrophages (Figs. 6 and 7). These cytokines should accelerate the accumulation of inflammatory cells in the synovium by the induction of chemotactic factors or adhesion molecule expression on leukocytes or endothelial cells (34, 35, 55).

Therapeutic agents that target several cytokines were recently developed and extensively used in the therapy of autoimmune disorders including RA (55). In this study, the blocking of the α₉ integrin by an Ab dramatically inhibited ongoing arthritis with the attenuation of not only inflammation but also of bone destruction (Fig. 4). In addition, expressions of proinflammatory molecules including cytokines, chemokines, and protease by synovial cells were induced by an α₉ integrin-mediated signal (Figs. 6 and 7). Of note, the expression of TNF-α was not inhibited by an anti-α₉ integrin Ab, although clinical and histological scores of arthritis were significantly reduced (Fig. 4). Thus, TNF-α production in inflamed joints was independent of α₉ integrin-mediated signaling. It will be important to examine in future studies whether α₉ integrin-mediated signaling plays a critical role in the pathogenesis of arthritis in RA patients who are not responding to the anti-TNF-α treatment.

In conclusion, our proposed model for the role of the α₉ integrin in the development of arthritis is illustrated in Fig. 8. The autocrine and paracrine interaction of OPN and/or TN-C with its receptor, α₉ integrin, on synovial cells results in the production of various proinflammatory molecules. IL-6, IL-1α, CCL2, and CXCL5 could be produced by both synovial fibroblasts and macrophages. IL-1β, TNF-α, CCL3, CCL4, and CXCL2 could be derived from synovial macrophages, but not from fibroblasts. CXCL12 could be derived from synovial fibroblasts, but not from macrophages. Those cytokines and chemokines induce inflammation, promote synovial cell growth, and activate osteoclasts, thus playing a critical role in the development of autoimmune arthritis (34, 56, 57). Importantly, there is cross-talk between synovial fibroblasts and macrophages. MMP-9 expressed by synovial fibroblasts activates TGF-β and IL-1 (36, 37), which are produced by synovial macrophages. Of note, TGF-β stimulates the production of the ECM, including OPN and the induction of integrins by fibroblasts (58, 59). Thus, it is likely that the interaction between α₉ integrin and its ligands activates synovial component cells and controls the inflammation and subsequent bone destruction via the expression of various inflammatory molecules.

We express sincere appreciation to Prof. Dean Sheppard at University of California San Francisco for critical reading of the manuscript and for providing valuable suggestions.

The authors have no financial conflict of interest.