MT1-MMP is a crucial promotor of synovial invasion in human rheumatoid arthritis

INTRODUCTION

Rheumatoid arthritis (RA) is a systemic autoimmune disease, and one of its key pathological features is the destruction of joint tissues, including cartilage, by inflamed synovial pannus leading to a loss of joint function. Invading synovial pannus tissue produces proteinases that degrade components of the joint extracellular matrix (ECM) (1). Cartilage is a unique tissue comprised of a small number of resident chondrocytes and their abundant surrounding ECM, and its function is heavily dependent on the nature of the ECM (2, 3). Cartilage ECM is predominantly composed of type II collagen and aggrecan with other minor components including small leucine rich proteoglycans and type IX and XI collagens (2, 3). Aggrecan is considered to be lost first in the early stages of RA, followed by collagen degradation. Aggrecan loss is primarily due to two types of metalloproteinases; aggrecanases that belong to the ADAMTS (a disintegrin and metalloproteinase with thrombospondin motifs) family and matrix metalloproteinases (MMPs)(1, 4). As aggrecan maintains water in the cartilage, it is an important ECM component endowing cartilage with compressive resistance properties. Loss of aggrecan thus abrogates cartilage function. The proteinases responsible for collagen degradation are members of the MMP family including collagenases (MMPs-1, -8, -13), gelatinase A (MMP-2) and membrane-type 1 MMP (MT1-MMP/MMP14) (5). Type II collagen is the major collagen in cartilage, accounting for 90 % of the total collagen (2). Monomeric type II collagen is composed of three alpha 1(II) chains in a triple helical structure which form a fibrillar meshwork by cross-linking with each other and with other minor collagens. The collagen meshwork provides the tissue with structure and tensile strength. Thus, degradation of collagen results in irreversible structural and functional damage to the cartilage.

Fibrillar collagens including type II collagen are resistant to most proteinases due to their triple helical structure, but five MMPs, namely MMP-1, MMP-2, MMP-8, MMP-13, and MT1-MMP, can degrade type II collagen (5). Among them, MT1-MMP has been shown to be an essential collagenolytic enzyme during development (6, 7). MT1-MMP null mice show skeletal defects and general fibrosis which are thought to be caused by a lack of cellular collagenolytic activity (6). Among the above collagenases MT1-MMP is the only one that can promote cellular invasion into a collagen matrix (8). MT1-MMP, as a cellular collagenase, has also been shown to play an essential role in angiogenesis (7, 9), tumour invasion (10–12) and growth (13) and smooth muscle cell migration in the arterial wall (14). Besides collagen II, MT1-MMP degrades collagens I and III, fibronectin, laminins 1 and 5, fibrin, and aggrecan core protein (15–17). MT1-MMP also activates proMMP-2 (11) and proMMP-13 (18) and it processes ECM receptors including CD44 (19) and syndecan 1 (20), promoting cell migration. MT1-MMP was reported to be expressed by the lining cells of human RA synovium (21, 22) and cells at the cartilage pannus junction (22), and potential involvement of MT1-MMP in cartilage invasion by RA synovial cells by activation of proMMP-2 or by itself has been reported (21–23). However, the exact role of MT1-MMP in RA pathogenesis and the mechanism of synovial pannus invasion are not clear at present.

We have previously shown that homo-dimer formation of MT1-MMP is essential for cell surface collagen degradation (24) and proMMP-2 activation (25), and that disruption of homo-dimerization by over-expressing a catalytic domain deletion mutant of MT1-MMP (MT1-ΔCat) effectively inhibits both collagenolytic activity and proMMP-2 activation (24, 25). Thus, MT1-ΔCat is a useful tool to examine the biological significance of MT1-MMP. In this study, we found that MT1-MMP is highly expressed in synovial cells from patients with RA, particularly in the cells actively invading into cartilage. Using an adenoviral MT1-ΔCat over-expression system, we demonstrated that MT1-MMP is a key enzyme in synovial invasion in rheumatoid arthritis.

Materials and Methods

RESULTS

Cartilage invasion by pannus requires type II collagen degradation (1). Since one of the membrane-bound MMPs, MT1-MMP has been shown to promote cellular invasion into a collagen matrix (8–10, 27), we examined whether human rheumatoid synovial cells from patients with RA express MT1-MMP by Western blot analysis using a specific anti-MT1-MMP antibody. As shown in Figure 1A, all of the tested primary synovial cells expressed high levels of MT1-MMP (n=6). HT1080, human fibrosarcoma cells, were taken as a positive control as they are well known to express MT1-MMP (11, 28). Band patterns seen in synovial cells are the same as in HT1080, with the active form of MT1-MMP (50 kDa), a catalytic domain-processed form of around 40 kDa, and further degraded forms of lower molecular weights being visible. It is noteworthy that the level of MT1-MMP expressed by the synovial cells is equivalent to or higher than that in HT1080, an invasive cancer cell line. Next, we tested if MT1-MMP expression correlates with cartilage invasion by immunohistochemical analysis of rheumatoid joint specimens stained with the same anti-MT1-MMP antibody. As shown in Figure 1B, MT1-MMP was highly expressed in pannus tissue, particularly at the pannus-cartilage junction, i.e. at the invasion edge of the pannus (see Figure 1B a, d arrows). In these specimens, cartilage completely lacks aggrecan as indicated by the lack of Safranin O staining (Figure 1B c, f). We have analysed joint specimens from 24 different RA patients and similar results were obtained with all specimens (data not shown). Some RA joint specimens were also stained with anti-MT1-MMP and anti-CD68 antibodies. Most cells invading into cartilage, especially those invading deeply into the tissue, were CD68-negative and spindle shaped (Figure 1C, deep invasion). However, in some regions, CD68 positive cells were present in the pits formed on the cartilage surface (Figure 1C, superficial invasion). Nevertheless, both cell types highly express MT1-MMP at the invasion front.

To test the functional contribution of MT1-MMP to synovial pannus invasion, we developed an ex vivo invasion model. Freshly isolated RA synovial tissues were diced into small pieces about 2 mm³, embedded into a 3-D collagen gel and cultivated for 5 days. In the initial 24 h, small T-cell-like cells migrated into the collagen matrix, but they were undetectable after 48 h. Numerous large spindle-shaped fibroblastic cells started to invade into collagen after 48–72 h (Figure 2). Because the invasion edge of synovial tissue consists of large fibroblastic cells (Figure 1C), the fibroblastic cell invasion into collagen gel observed in this ex vivo model is likely to represent in vivo pannus invasion. Next, we performed this invasion assay in the presence of tissue inhibitor of metalloproteinases (TIMP)-1 (0.5 µM), TIMP-2 (0.5 µM), or GM6001 (10 µM). It is known that the activity of transmembrane-type MT-MMPs including MT1-, MT2-, MT3-, and MT5-MMP cannot be inhibited by TIMP-1 but is inhibited by TIMP-2 and GM6001 (29, 30). As shown in Figure 2, invasion was completely inhibited by TIMP-2 and GM6001, but not by TIMP-1. We have performed this invasion assay using joint synovial tissues from 5 different patients, and the results were identical (data not shown). Among TIMP-1-insensitive MT-MMPs, MT1-MMP and MT2-MMP have been shown to promote cell invasion into a type I collagen matrix (8). However, we could not detect MT2-MMP in any of the synovial cells we tested by Western Blotting (data not shown) which supports previous reports (21, 31) showing either no expression or inconsistent low expression of MT2-MMP compared to MT1-MMP. Therefore, these data suggest that MT1-MMP may be a key collagenolytic enzyme promoting synovial invasion.

We have recently reported that MT1-MMP requires homo-dimerization to degrade collagen, and inhibition of this interaction by expression of a catalytic domain deletion mutant of MT1-MMP (MT1F-ΔCat) effectively inhibited collagenolytic activity on the cell surface (24). Therefore, we constructed an adenoviral vector for expression of MT1F-ΔCat (Ad-MT1F-ΔCat) to test the role of MT1-MMP in synovial fibroblast invasiveness (Figure 3A). We also made Ad-MT1F to over-express the enzyme as well (Figure 3A). When isolated RA synovial fibroblasts were cultured on a collagen film, they degraded collagen leaving unstained clear areas, and this activity was inhibited by TIMP-2 and GM6001 but not by TIMP-1 (Figure 3B). This suggests that collagen degradation by these cells was caused by MT1-MMP, and not by any soluble MMPs including MMP-1, MMP-2, MMP-8 and MMP-13. We then tested the effect of adenoviral expression of MT1F and MT1F-ΔCat. As shown in Figure 3C, these exogenous genes were overexpressed in synovial fibroblasts. It was noted that most of MT1F expressed in the cells processed to a 40 kDa species and lost its catalytic domain containing FLAG tag (Figure 3C). Such processing has been reported previously and suggests that functional activity is high in these cells (30, 32, 33). The collagen-degrading activity was not affected by infection of the cells with different MOI of mock viruses (Ad-Mock), but could be inhibited in a dose-dependent manner by Ad-MT1F-ΔCat (Figure 3D). Degradation was further accelerated in a dose-dependent manner by infection with Ad-MT1F. Taken together, the data suggest that MT1-MMP is the cellular collagenase in human synovial cells.

We next examined whether MT1F-ΔCat can inhibit synovial cell invasion using an ex vivo collagen invasion assay. To establish an effective adenoviral infection of human RA synovial tissue, we carried out dose-dependent infection with a recombinant adenovirus that expresses green fluorescent protein and found that a virus titre of 10⁸ pfu/ml was minimally toxic and maximally effective for expression of the gene, although, in this system, it is inevitable that there will be some non-infected cells especially in deeper areas of the tissue. As shown in Figure 4, cells from RA synovial tissue infected with Mock viruses invaded into the collagen gel, and this invasion was largely suppressed by GM6001. In contrast, pannus infected with Ad-MT1F-ΔCat showed significantly reduced invasion, and addition of GM6001 further inhibited invasion. Infection with Ad-MT1-F slightly enhanced invasiveness in terms of invasion distance compared to Ad-Mock infected cells, and GM6001 again strongly inhibited invasion. In order to obtain more quantitative measurement of synovial cell invasion, we carried out microcarrier bead invasion assays using isolated synovial fibroblasts. As shown in Figure 5, synovial cells attached to microcarrier beads invaded from the beads into the collagen gel very efficiently. This invasion was inhibited by TIMP-2 and GM6001 but not by TIMP-1 (Figure 5A). TIMP-1 rather enhanced invasion in this particular experiment, but this effect was observed in one out of three experiments using different synovial cells. When these cells were infected with Ad-MT1F-ΔCat prior to the assay, invasion was almost completely inhibited, whereas the expression of MT1-F enhanced the invasion (Figure 5B). These data suggest that MT1-MMP is indeed a crucial collagenolytic proteinase enabling these synovial cells to invade into a collagen matrix.

We next examined the invasion of synovial fibroblasts into the cartilage matrix. For this assay, we chose bovine nasal cartilage as it is possible to dissect out pieces of reproducible size and quantity for different treatments. Fresh bovine nasal cartilage was frozen and thawed three times to kill the chondrocytes and diced into 3 × 5 × 5 mm pieces. Human synovial cells were then cultured on the cartilage surface for up to 4 weeks. As shown in Figure 6A, synovial cells invaded into the cartilage in a time-dependent manner. This invasion was not inhibited by TIMP-1, but was completely inhibited by TIMP-2 and GM6001. We have attempted adenoviral infection in this experimental system, but were hampered by the 4-week length of the assay since adenoviral transgene expression lasts only up to 10 days.

In RA cartilage, aggrecan is thought to be lost in the early stages of the disease, and indeed there is not much aggrecan left at the late stage of the disease, as shown in Figure 1B. Pratta et al. (34) reported that aggrecan protects collagen against degradation by MMP-1 in vitro. Therefore we next examined whether removal of aggrecan from the cartilage would affect synovial cell invasion. To remove aggrecan, live cartilage was treated with 10 µM retinoic acid for 7 days to stimulate aggrecanase production and aggrecan removal. The cartilage was then subjected to freezing and thawing three times to kill the chondrocytes. As shown in Figure 6B, this treatment depleted aggrecan from the cartilage as indicated by the lack of Safranin O staining. Untreated cartilage maintained staining although some loss of staining was noted at the edge of the tissue. In both treatments, synovial cells clearly invaded into the cartilage in 4 weeks. However, synovial fibroblasts invaded much deeper into retinoic acid-treated cartilage. To analyze this quantitatively, we counted a number of the cells at different depths from the surface of the cartilage. As shown in Figure 6C, in untreated cartilage, most synovial cells invaded up to 90 µm from the cartilage surface, whereas in retinoic acid-treated cartilage more cells were present in deeper areas. The proportion of cells that had invaded deeper than 90 µm was 17 % for non-treated cartilage and 39 % for retinoic acid-treated cartilage. These data suggest that aggrecan indeed protects collagen in cartilage, and aggrecan loss may be an important initial step in promoting synovial invasion into cartilage in RA.

DISCUSSION

The synovial pannus is formed as a result of inflammation, which causes proliferation of synovial lining cells and infiltration of monocytes and lymphocytes that produce high levels of inflammatory mediators such as TNFα and IL-1 (35). It is thought that the major cause of sustained inflammation is over-production of cytokines, and indeed anti-TNFα therapy has been proven to be a very effective treatment for RA (36). However, expression of MT1-MMP does not correlate with the area where inflammatory cells reside but rather with the invasive edge, suggesting that inflammatory cytokines may not stimulate expression of MT1-MMP. In fact, IL-1 and TNFα treatment did not modify MT1-MMP expression in isolated synovial fibroblasts (M-C. Miller and Y. Itoh, unpublished data). This is not surprising, since the promotor region of MT1-MMP lacks AP-1 and NFκB responsible elements and a TATA box, which are present in many inflammatory cytokine-inducible genes (37, 38). It has been shown that concanavalin A (39), cytochalasin D (40), collagen (41) and v-src transformation of epithelial cells (42) induce MT1-MMP expression. However, it is not known what regulates MT1-MMP gene expression under pathological conditions such as in RA. Nevertheless, there appeared to be a strong correlation between high levels of MT1-MMP expression and synovial invasion in RA.

To investigate whether there is a functional link between MT1-MMP expression and pannus invasion, we examined invasion of freshly isolated synovial tissue or isolated synovial fibroblasts into collagen gels or cartilage. In every assay, our results indicate that synovial cells utilize MT1-MMP as a collagenolytic enzyme to invade into the matrix. Invasion was almost completely inhibited by TIMP-2 and GM6001 but not by TIMP-1. The concentration of these inhibitors used in the experiments were over excess to achieve almost complete inhibition of the enzymes (0.5 µM for TIMPs and 10 µM for GM6001). Thus soluble collagenases, including MMP-1, MMP-2, MMP-8 and MMP-13, are unlikely to be involved. These results agree with the previous report by Hotary et al.(8) that over-expression of these soluble collagenases does not enhance cell invasion. Furthermore, collagen invasion was effectively inhibited by adenoviral expression of MT1F-ΔCat, suggesting that MT1-MMP is the responsible enzyme. However, a number of reports in the past have shown that MMP-1 and MMP-13 are highly up regulated in RA synovial fibroblasts with cytokine treatment and in RA joint tissues (43–50). So what are they doing? These soluble collagenases are likely to degrade collagens and other matrix components in cartilage, such as aggrecan (51), in a wider area, which may in turn compromise the integrity of the tissue and promote pannus invasion indirectly. Nonetheless, further work is required to understand the exact roles of these soluble collagenases in the pathogenesis of RA.

MT1-MMP plays an important role in postnatal development, as evidenced by the phenotype of MT1-MMP-null mice (6, 7). Lack of the MT1-MMP gene causes skeletal dysplasia accompanied by severe runting, osteopenia and fibrosis of soft tissue, and the mice die around 7–12 weeks after birth (6, 7). An interesting additional phenotype is the development of severe generalised arthritis accompanied by the presence of hyper-cellular and vascularised synovial tissues degrading articular cartilage in all joints (6, 7). This apparently contrasts with our finding that MT1-MMP promotes synovial invasion into cartilage in RA. However, the cell types involved in cartilage degradation appear to be different in each case. In the MT1-MMP-null mice, cells degrading cartilage are tartrate-resistant acidic phosphatase (TRAP)-positive osteoclast-like giant cells, which dominate the inflamed synovium (6, 7). We found no apparent TRAP-positive osteoclast-like cells at the cartilage-pannus junction in the human RA joints we examined. We did occasionally find CD68-positive macrophages in pits on the cartilage surface, but cells invading deeply into the cartilage were Factor VIII- (endothelial marker), CD3- (T-cell marker) and CD68-negative (Figure 1C and unpublished observations). Thus it is likely that fibroblast-like cells are the major cartilage-invading cell type in human RA (52). Taken together, the arthritic phenotype of the MT1-MMP null mice is likely to be caused by a different mechanism.

Besides type II collagen, aggrecan is the other major component of cartilage, endowing cartilage with compressive resistance properties by retaining water molecules in the tissue (3). In RA and osteoarthritis (OA), aggrecan is depleted from cartilage by the action of aggrecanases belonging to the ADAMTS family (1, 4). It has been reported that deletion of the ADAMTS-5 gene in mice protects against aggrecan depletion and cartilage degradation in experimental OA and inflammatory arthritis models (53, 54). Pratta et al. (34) have reported that removing aggrecan makes cartilage more susceptible to collagen degradation by MMP-1 in vitro, and Little et al (55) have recently shown that mutation of the aggrecanase cleavage sequence in the interglobular domain of aggrecan by gene knock-in prevents aggrecan loss and cartilage erosion. These reports suggest that aggrecan may protect against collagen degradation. In this study, we confirmed this protective effect of aggrecan. Removing aggrecan by pre-treating cartilage with retinoic acid made cartilage more susceptible to synovial cell invasion. As aggrecan is completely lost in advanced RA, this may be one of the mechanisms promoting synovial pannus invasion. The mechanism by which aggrecan protects against collagenolytic action remains unclear. It is possible that the negatively charged sulfated polysaccharide chains of aggrecan that fill the collagen fibril gaps may prevent enzyme-collagen interactions. Without retinoic acid treatment, synovial cells were still able to invade into cartilage, but they appeared to be unable to migrate further than 60 µm from the cartilage surface. During the course of the experiments, we found that safranin O staining near the cartilage surface became weaker during incubation period of 2–4 weeks, even in the absence of viable cells (data not shown). This is presumably due to diffusion of aggrecan into culture media. It is thus possible that the loss of aggrecan up to 60 µm might have been sufficient for synovial cells to be able to invade.

In summary, we have provided evidence which supports that MT1-MMP is a key proteinase that promotes pannus invasion into cartilage. MT1-MMP also plays an important role in monocyte transmigration through the endothelium (56) and in angiogenesis (7, 9, 57), both of which contribute to RA progression. We thus believe that MT1-MMP can be a target molecule for potential therapeutic intervention for RA.