Tissue Inhibitor of Matrix Metalloproteinase-2 Regulates Matrix Metalloproteinase-2 Activation by Modulation of Membrane-type 1 Matrix Metalloproteinase Activity in High and Low Invasive Melanoma Cell Lines*

Activation of pro-matrix metalloproteinase (MMP)-2 on the surface of malignant cells by membrane-bound MT1-MMP is believed to play a critical role during tumor progression and metastasis. In this study we present evidence that MT1-MMP plays a key role for the in vitro invasiveness of malignant melanoma. Melanoma cell lines secreted latent MMP-2 when cultured on plastic. However, when cells were grown in floating type I collagen lattices, only high invasive melanoma cells activated proMMP-2. Activation could be inhibited by antibodies against MT1-MMP, by addition of recombinant tissue inhibitor of metalloproteinases (TIMP)-2 and by inhibition of MT1-MMP cleavage. MT1-MMP protein was detected as an inactive protein in all cell lines cultured as monolayers, whereas in collagen gels, active MT1-MMP protein was detected in the membranes of both high and low invasive melanoma cells. Production of TIMP-2 was about 10-fold higher in low invasive cells as compared with high invasive melanoma cells and was further increased in the low invasive cells upon contact to collagen. Thus, in melanoma cells TIMP-2 expression levels might regulate MT1-MMP-mediated activation of proMMP-2. High invasive melanoma cells displayed increased in vitro invasiveness, which was inhibited by TIMP-2. These data indicate the importance of these enzymes for the invasion processes and support a role for MT1-MMP as an activator of proMMP-2 in malignant melanoma.


Activation of pro-matrix metalloproteinase (MMP)-2 on the surface of malignant cells by membrane-bound MT1-MMP is believed to play a critical role during tumor progression and metastasis. In this study we present evidence that MT1-MMP plays a key role for the in vitro invasiveness of malignant melanoma. Melanoma cell lines secreted latent MMP-2 when cultured on plastic. However, when cells were grown in floating type I collagen lattices, only high invasive melanoma cells activated proMMP-2. Activation could be inhibited by antibodies against MT1-MMP, by addition of recombinant tissue inhibitor of metalloproteinases (TIMP)-2 and by inhibition of MT1-MMP cleavage. MT1-MMP protein was detected as an inactive protein in all cell lines cultured as monolayers, whereas in collagen gels, active MT1-MMP protein was detected in the membranes of both high and low invasive melanoma cells. Production of TIMP-2 was about 10-fold higher in low invasive cells as compared with high invasive melanoma cells and was further increased in the low invasive cells upon contact to collagen. Thus, in melanoma cells TIMP-2 expression levels might regulate MT1-MMP-mediated activation of proMMP-2. High invasive melanoma cells displayed increased in vitro invasiveness, which was inhibited by TIMP-2. These data indicate the importance of these enzymes for the invasion processes and support a role for MT1-MMP as an activator of proMMP-2 in malignant melanoma.
Local degradation of connective tissue in the vicinity of the cell surface is thought to be essential for tumor cell invasion and metastasis. Neoplastic cells that invade surrounding tissues and metastasize through the blood stream or lymphatic system must penetrate several barriers, including basement membranes and the interstitial connective tissue (1,2). Degradation of extracellular matrix components is accomplished through the combined action of different proteases, primarily of the matrix metalloproteinase (MMP) 1 (3) and serine protease families (4).
Because of the importance of MMP activity for initiating efficient matrix degradation, MMP expression and activity is tightly regulated and is subject to several levels of control, including gene transcription, post-translational extracellular activation, and inhibition by soluble inhibitors (3,5). Basal production of metalloproteinases is typically low or absent and requires an inducing factor, such as a cytokine, phorbol ester, or contact with components of the extracellular matrix to increase their expression (3). Enzyme activation is achieved by removal of the N-terminal propeptide domain through exogenous or autocatalytic cleavage. In vitro, serine proteases, such as plasmin or trypsin, have been shown to activate most MMPs (3). However, MMP-2 is unique among the MMPs in that its expression is constitutive and its activation can be achieved in a membrane-associated manner as has been shown for fibroblasts (6), for endothelial cells (7,8), and for tumor cells (9 -11). A novel membrane-type matrix metalloproteinase (MT1-MMP) was originally cloned from a breast cancer cDNA library and was characterized as an integral membrane MMP that can activate proMMP-2. Thus far, four different membrane-bound MMPs have been described (12)(13)(14)(15). Although MT2-MMP and MT3-MMP were also shown to activate proMMP-2 in vitro (16,17), MT1-MMP is the one with the best documented correlation to the invasive phenotype of different types of cancer, such as breast and lung carcinomas (9,18).
In vitro, MT1-MMP expression is up-regulated by concanavalin A, by monensin, cytochalasin D, and by fibrillar type I collagen (19,20); the latter is, to date, the only known physiological inducer (21). Like other members of the metalloproteinase family, MT1-MMP is synthesized as a proform. The mechanism of MT1-MMP activation and whether cleavage of the N-terminal peptide of proMT1-MMP is a prerequisite for its activity are still a matter of discussion, but both the intracellular protease furin (22) and extracellular plasmin (23) have been reported to be capable to cleave the propeptide. Several studies have demonstrated that activation of proMMP-2 by MT1-MMP depends upon the presence of low amounts of TIMP-2, which is believed to be required for the formation of a membrane-bound ternary complex consisting of MT1-MMP, TIMP-2, and latent MMP-2 (24,25). A second "free" MT1-MMP located in proximity to the ternary complex may, in a second step, cleave proMMP-2 bound to the MT1-MMP/TIMP-2 receptor. On the other hand, at high concentrations, TIMP-2 inhibits MMP-2 activation, presumably by blocking the activity of MT1-MMP (25)(26)(27).
MMP-2 (28) and MMP-9 (29) have been shown to be overexpressed in many different types of tumors. Furthermore, the expression and activation of these enzymes has been correlated to the invasive and metastatic phenotype of tumors like nonsmall-cell lung carcinoma (30), gastric and breast carcinomas (31,32), and squamous cell carcinomas (33), where it is thought to have a major impact on tumor progression. Further, Mac-Dougall and co-workers (34) have shown that MMP-2 is constitutively expressed in malignant melanoma cell lines and Vaisanen et al. (35) have demonstrated that expression of MMP-2 increases with enhanced atypia and dedifferentiation in melanocytic lesions. However, little is known about the role of MMP-2 in neoplastic progression of human melanomas and about the mechanisms underlying the activation of proMMP-2 in vivo.
To examine the hypothesis that variable expression and/or activity of MT1-MMP, MMP-2, and TIMP-2 might be different in melanoma cells of high and low metastatic potential, we analyzed the expression and activation pattern of MMP-2 and MT1-MMP in different melanoma cell lines. Here we report that active MT1-MMP protein is detected in the cell membranes of melanoma cells lines of both high and low invasive potential when cultured in collagen lattices. Activation of proMMP-2, however, is only observed in high invasive melanoma cell lines, which produced significantly lower amounts of TIMP-2 when compared with low invasive cells. Activation of proMMP-2 by high invasive melanoma cells was completely inhibited by antibodies against MT1-MMP, by a synthetic furin inhibitor, as well as by the addition of recombinant TIMP-2. These data indicate that MT1-MMP is responsible for MMP-2 activation in highly invasive melanoma cell lines and that this process is highly dependent on the expression level of TIMP-2.

MATERIALS AND METHODS
Melanoma Cell Lines and Cell Culture-The following melanoma cell lines were used: MV3 and BLM, known to be highly metastatic with early and frequent formation of metastasis after subcutaneous inoculation in nude mice. The nonmetastatic melanoma cell lines IF6 and 530 showed low invasive abilities and no organ metastasis unless injected intravenously (36,37). The cells were cultured in RPMI 1640 supplemented with 10% fetal calf serum, 2 mM glutamine, and 100 units/ml each of penicillin and streptomycin and then passaged by trypsinization. Three-dimensional collagen type I gel cultures were prepared as described previously (38). Briefly, cells were seeded into collagen gels (3.5 ϫ 10 6 cells/ml) containing 1 mg/ml porcine skin type I collagen (DGF Stoess AG, Eberbach, Germany).
Preparation of Conditioned Media-Cells were cultured as monolayer for 48 h, washed three times with PBS, followed by trypsinization and counting. 3.5 ϫ 10 6 cells/ml were seeded into collagen gels, and the medium was changed daily. After 48 h, the gels were washed three times with PBS and the medium replaced by serum-free RPMI; 24 h later, the medium was collected. Alternatively, conditioned media and cells grown as monolayers or in collagen gels were combined and homogenized by sonication. In the inhibition experiments, cell suspensions were preincubated for 20 min with the synthetic furin inhibitor decanoyl-Arg-Val-Lys-Arg-chloromethylketone (CMK; Bachem Biochemicals, Heidelberg, Germany) or aprotinin (Roth, Karlsruhe, Germany), and then seeded in collagen gels either with furin inhibitor concentrations varying from 0 to 75 M or aprotinin in concentrations from 1 to 10 M or methanol, the CMK solvent. After 24 h, cultures were washed three times with PBS and media replaced by serum-free media containing fresh inhibitors. After another 24 h, culture samples were processed as described above. Fibroblast-conditioned medium was prepared by incubating confluent fibroblast cultures with serum-free medium for 48 h. Conditioned media were stored at 4°C until use.
In Vitro Invasion Assays-Melanoma cell invasion was assayed as described previously (40) using a modified Boyden chamber equipped with polycarbonate filters (8-m pore size, Corning Costar, Bodenheim, Germany). The filters were coated with 25 l of Matrigel (Becton Dickinson, Heidelberg, Germany) and placed above the lower compartment, which contained either 200 l of serum-free RPMI 1640 culture medium (random migration) or fibroblast conditioned media. Melanoma cells were suspended in serum-free RPMI 1640 (5 ϫ 10 5 /ml) and seeded into the upper compartment of the chamber (total volume 0.7 ml). The chambers were incubated at 37°C for 36 h. Cells attached to the upper side of the filter were mechanically removed. The filters were then stained with eosin and thiazine dyes (Dade Diff-Quick, Dü dingen, Switzerland). Invasion was determined by counting the cells that had migrated to the lower surface of the filter (100-fold magnification). For each filter, the number of cells in 5 randomly chosen microscope fields was determined and averaged. Each experiment was performed in triplicate. Invasion assays were also performed in the presence of human recombinant TIMP-2 (5.4 nM) or of 100 l of medium conditioned by the non-invasive melanoma cell line IF6, which contains 80 ng of TIMP-2, corresponding to a final concentration of 5.4 nM.
Preparation of Plasma Membranes-For preparation of crude plasma membranes, the collagen gels were disrupted by mechanical shearing through a 10-ml syringe and subsequent incubation with bacterial collagenase D (Roche Molecular Biochemicals, Mannheim, Germany) at a concentration of 1 mg/ml in PBS, pH 7.4, at 37°C for approximately 10 min until no collagen fibers were visible in the solution. The digestion was stopped by the addition of 0.5 volume of fetal calf serum, and cells were pelleted by centrifugation (2,000 ϫ g, 10 min, 4°C). All subsequent steps for preparation of crude plasma membranes were carried out at 4°C. After centrifugation the cells were resuspended in 20 mM Tris-HCl, pH 7.4, containing 8.7% (w/v) sucrose and homogenized by Dounce homogenization. The solution was layered on top of a 38.5% (w/v) sucrose cushion and centrifuged in a swing-out rotor (100,000 ϫ g, 1 h). The plasma membranes were collected from the interface and pelleted by centrifugation (100,000 ϫ g, 1 h). Plasma membranes were then resuspended in 50 mM Tris-HCl, pH 8, 5 mM CaCl 2 . The protein concentration was determined by a commercial protein assay (Bio-Rad, Munich, Germany) and adjusted to 2 mg/ml. For proMMP-2 activation experiments, 10 g of crude plasma membranes were incubated with 20 l of fibroblast conditioned medium containing proMMP-2 for 24 h at 37°C in the presence of either 0.3 g of anti-MT1-MMP monospecific peptide antibodies (raised against a peptide corresponding to residues 160 -173 of human MT1-MMP, clone 114-1F2; Fuji Chem., Toyama, Japan) or of 3.8 nM human recombinant TIMP-2 (Calbiochem-Novabiochem, Bad Soden, Germany). 15 l of the samples was analyzed by gelatin zymography.
Northern Blot Analysis-Total RNA was isolated from melanoma cells grown for 48 h as monolayers or within collagen gels by direct lysis into guanidine thiocyanate followed by phenol-chloroform extraction (41). 10 g of total RNA were resolved by formaldehyde-agarose gel electrophoresis, blotted onto Hybond-Nϩ membranes (Amersham Pharmacia Biotech, Braunschweig, Germany), and hybridized with the random-primed 32 P-labeled cDNA probe for MT1-MMP (kindly provided by H. Sato, Kanazawa University, School of Medicine, Kanazawa, Japan).
Immunochemical Assays-For Western blot analysis, crude plasma membrane preparations (30 to 60 g) were separated on 10% SDSpolyacrylamide gels under reducing conditions and transferred to nitrocellulose membranes (Hybond C; Amersham Pharmacia Biotech, Freiburg, Germany) in a semidry blotting chamber. After blockage of nonspecific binding sites with 5% nonfat milk in PBS containing 0,5% Tween (v/v) blots were incubated with 1 g/ml anti-MT1-MMP monospecific peptide antibody (114-1F2; Fuji Chemical, Toyama, Japan) for 3 h at room temperature. After extensive washing, the nitrocellulose membrane was incubated with anti-mouse IgG conjugated with horseradish peroxidase (Dako, Hamburg, Germany; final concentration 1:2,000) for 1 h at room temperature. Bound antibodies were detected with ECL Western blotting detection reagents (Amersham Pharmacia Biotech) according to the manufacturer's instructions.
For quantification of TIMP-1 and TIMP-2 protein levels, conditioned media were combined with sonicated cells or collagen gels, subjected to enzyme-linked immunoabsorbent assay multiwell plates (Amersham Pharmacia Biotech) and measured following the manufacturer's in-structions. Briefly, samples (100 l) were incubated in multiwell plates precoated with specific antibodies against TIMP-1 or TIMP-2. Antigenantibody complexes were measured after incubation with a horseradish peroxidase-conjugated anti-TIMP-1 and anti-TIMP-2 Fab fragment. Tetramethylbenzidine was used as a substrate to evaluate peroxidase activity. Serial dilutions of human recombinant TIMP-1 and TIMP-2 were used as an internal standard.

Effect of Cell-Matrix Interactions on the Activity of MMP-2 in Melanoma Cell Lines of High and Low Invasive Potential-
Conditioned media prepared from the melanoma cell lines MV3, BLM, IF6, and 530, grown as monolayers and within three-dimensional collagen lattices, were analyzed by gelatin zymography with respect to their production/activation of MMP-2 and MMP-9. In monolayer culture, all cell lines expressed variable amounts of latent MMP-2, whereas activated forms of this enzyme and MMP-9 were not detectable. When cultured in three-dimensional native type I collagen lattices, all melanoma cell lines showed an increase of proMMP-9 and proMMP-2 secretion (Fig. 1). In addition, upon contact with collagen, the high metastatic melanoma cells MV3 and BLM gained the ability to activate proMMP-2 into its 62/59-kDa active forms, whereas the cells of low invasive potential, 530 and IF6, continued to show MMP-2 only as latent proenzyme.
Since activation of proMMP-2 has been reported to occur by a plasma membrane dependent mechanism, we prepared membrane fractions from melanoma cells grown under both culture conditions. As shown in Fig. 2, plasma membranes purified from the two high invasive melanoma cell lines grown in contact to collagen showed activation of exogenous proMMP-2 when plasma membranes were incubated with fibroblast-conditioned medium containing latent MMP-2. In contrast, no activation was obtained by membranes isolated either from the low invasive melanoma cells grown in collagen gels or by membranes purified from any of the melanoma cells cultured on plastic dishes (Fig. 2).
Expression of MT-MMP mRNA in Melanoma Cells Grown as Monolayers or in Collagen Gels-MMP-2 is unique among the MMPs in that it is not activated by serine proteases such as plasmin (43). The putative physiological activators of proMMP-2 are the membrane-type matrix metalloproteinases, e.g. MT1-MMP (12,24). In order to elucidate the role of MT1-MMP for the activation of proMMP-2 in the high invasive melanoma cell lines, we analyzed MT1-MMP transcript levels in the melanoma cell lines under both culture conditions. All melanoma cells constitutively produced MT1-MMP mRNA in different amounts in monolayer cultures. Upon contact with collagen, a slight increase of the specific mRNA levels was observed with the high invasive melanoma cells (Fig. 3).
In addition to MT1-MMP, MT2-and MT3-MMP have also been reported in vitro to activate proMMP-2 (16,17), whereas the substrate specificity of MT4-MMP is still unclear. We therefore analyzed mRNA levels for the different MT-MMPs in high and low invasive melanoma cells by semiquantitative RT-PCR analysis. All four MT-MMP mRNAs were detectable by RT-PCR in monolayer as well as collagen cultures. Semiquantitative analysis revealed a slight up-regulation of the transcripts for MT1-MMP in the high invasive melanoma cells cultured in collagen gels as already shown by Northern blot analysis. However, as compared with MT1-MMP, transcript levels for MT2-, MT3-, and MT4-MMP were low (data not shown).

Role of MT1-MMP and TIMP-2 in Collagen-induced Activation of proMMP-2 in High and Low Invasive Melanoma Cell
Lines-Western blot analysis of MT1-MMP in crude membrane preparations displayed a 63-kDa band corresponding to the unprocessed zymogen (44) in all melanoma cell lines grown as monolayers (Fig. 4A). When cells were grown within threedimensional type I collagen lattices, high as well as low invasive cell lines showed an additional immunoreactive protein band of 60 kDa, corresponding to furin-activated MT1-MMP (44). This indicates that both high and low invasive cells display activated MT1-MMP protein on their cell surface upon growth in contact with a collagen matrix, although the high invasive melanoma cells only showed activation of proMMP-2.
Preincubation of the invasive melanoma cell line MV3 with increasing concentrations of the synthetic furin inhibitor CMK resulted in complete suppression of collagen-induced proMMP-2 activation (Fig. 4B), whereas the serine protease inhibitor aprotinin did not affect proMMP-2 activation. In addition, detection of MT1-MMP protein in membranes isolated from MV3 cells that were treated with 50 M CMK for 24 h displayed a significant increase of the latent 63-kDa protein, whereas the amounts of 60-kDa protein were reduced. These data suggest that treatment with the furin inhibitor CMK results in an accumulation of the unprocessed membranebound MT1-MMP protein form, providing strong evidence that Golgi-associated furin is mainly responsible for proMT1-MMP processing in melanoma cells. Furthermore, preincubation of membranes with a monospecific MT1-MMP antibody resulted in a strong reduction of proMMP-2 activation (Fig. 5). These invasive potential were cultured either as monolayers (M) or seeded into three-dimensional type I collagen gels (G). After 48 h, monolayers and collagen gels were washed with PBS to remove serum and the culture was continued for 24 h with serum-free medium. Conditioned media were combined with cells and collagen gels, homogenized by sonication, and analyzed by gelatin zymography. Molecular size standards are indicated to the left. Positions of proMMP-9, proMMP-2, and MMP-2 are marked. The data presented here are representative of three different experiments.
FIG. 2. Activation of proMMP-2 by plasma membranes from melanoma cell lines. High (BLM, MV3) or low (530, IF6) invasive melanoma cell lines were grown as monolayers or within three-dimensional collagen lattices. Cells were collected either by scraping them off the dishes or by digesting the collagen gels with bacterial collagenase (1 mg/ml, 10 min, 37°C) to liberate the cells. Crude membranes were prepared as described under "Materials and Methods." 10 g of the crude plasma membrane preparations from these cells were incubated with 20 l of fibroblast conditioned medium containing latent MMP-2 (co) for 24 h at 37°C. Activation of proMMP-2 by the plasma membranes was assayed by gelatin zymography.
observations support the theory that MT1-MMP plays a substantial role in proMMP-2 activation in melanoma cells.
As shown above, membranes from high as well as low invasive melanoma cell lines contain activated MT1-MMP as components of their plasma membranes, but only the high invasive cells showed activation of proMMP-2 upon contact with fibrillar collagen. Since no difference in the activation of MT1-MMP was found between melanoma cells of high and low invasiveness, we asked whether variable secretion of TIMP-2 by the melanoma cells could account for the differences in proMMP-2 activation between the two cell types. Quantification of TIMP levels revealed that the non-invasive cell lines IF6 and 530 produced about 10-fold higher amounts of TIMP-2 under both culture conditions as compared with the high invasive BLM and MV3 cells (Fig. 6). Furthermore, upon contact with fibrillar type I collagen, both low invasive melanoma cell lines further increased their TIMP-2 production, whereas the high invasive BLM and MV3 cells showed a significant decrease of TIMP-2 secretion. In contrast, TIMP-1, which has been reported not to inhibit proMMP-2 activation (26), was detected in variable amounts in the different melanoma cell lines under both culture conditions, showing no correlation with the reported in vivo invasiveness of the cells (Fig. 6). Additional evidence for the role of TIMP-2 in regulating MT1-MMP-mediated activation of proMMP-2 in melanoma cells came from the finding that the activation process was also inhibited in the presence of 3.8 nM human recombinant TIMP-2 ( Fig. 5) or medium conditioned by the low invasive IF6 cells (data not shown). This observation is in accordance with reports showing that an excess of TIMP-2 inhibits proMMP-2 activation by blocking the activity of MT1-MMP (6,25). Altogether, these data suggest that in non-invasive melanoma cells the strongly increased TIMP-2 protein levels affect MT1-MMP-mediated proMMP-2 processing in collagen gel cultures.

Role of MT1-MMP, MMP-2, and TIMP-2 for the in Vitro Invasiveness of Melanoma Cell Lines of High and Low Invasive
Potential-In order to evaluate whether activated MT1-MMP and MMP-2 contribute to the invasive behavior of melanoma cells, we analyzed the ability of the cells to penetrate a barrier composed of Matrigel in Boyden chamber assays. In contrast to the melanoma cells IF6 and 530 with low penetration of the Matrigel-coated filters, the high invasive BLM and MV3 melanoma cells showed a high rate of penetration (Fig. 7A). The invasive capacity of MV3 cells was reduced by about 50% when the cells were preincubated with 50 M amounts of the synthetic furin inhibitor CMK, which was shown above to inhibit MT1-MMP processing and to abolish proMMP-2 activation (Fig. 7B). These data suggest a direct involvement of active MT1-MMP in the invasion process. Invasion of Matrigel by MV3 cells was reduced by a similar extent when the invasion assays were performed in the presence of medium conditioned by the non-invasive melanoma cell line IF6, indicating inhibition of Matrigel invasion by soluble factors released by the non-invasive melanoma cells, e.g. TIMP-2. The involvement of TIMP-2 was corroborated by the addition of equivalent concentrations of recombinant TIMP-2 (5.4 nM) to the cell suspension into the Boyden chamber, which resulted in a comparable inhibition of the invasion process (Fig. 7B). DISCUSSION Several reports have shown that overexpression of MMP-2 by tumor cells and stromal cells can be correlated with the invasive and metastatic behavior of various tumors (9, 29 -33). However, the activity of MMP-2 is tightly regulated, and not only is it dependent on de novo synthesis of the latent enzyme, it also appears to be rather determined by the activation of the zymogen and by the presence of the specific inhibitor TIMP-2. The membrane-associated matrix metalloproteinase, MT1-MMP, which was identified as a potential physiological activator of proMMP-2 (12, 24), could be localized in different tumors and is thought to play a key role in tumor invasion (45). Activation of proMMP-2 and concomitant induction of MT1-MMP gene expression was also found in fibroblasts, endothelial cells, and breast carcinoma cell lines treated with the phorbol ester 12-O-tetradecanoylphorbol-13-acetate (31,44) or concanavalin A (46,47) or cultivated in collagen gels (19,20).
During invasion of the dermal connective tissue, melanoma cells interact with different components of the extracellular matrix, including type I collagen. As it has been shown by several investigators that cell-matrix interactions modulate de novo synthesis of matrix-degrading metalloproteinases in different cell types (3,9,19,38), we compared synthesis and activation of proMMP-2 in melanoma cells grown on plastic dishes and within three-dimensional gels composed of fibrillar type I collagen.
In the present study, we used melanoma cell lines, which have been previously characterized by their different invasiveness after subcutaneous injection into nude mice (36,37). When grown on plastic dishes, both high and low invasive melanoma cells secreted MMP-2 in its latent protein form. Upon contact with the collagenous matrix, only the high, but not the low invasive, melanoma cells gained the capability to convert the latent proenzyme into its 62/59-kDa active forms, a process that could be attributed to the cell surface. This process was completely abolished by the addition of a specific MT1-MMP antibody, indicating that other MT-MMPs, although low transcript levels could be detected by RT-PCR analysis, are not involved in the activation process.
Conversion of latent MMP-2 to its 62/59-kDa active forms by plasma membranes isolated from high invasive cells grown in collagen lattices could be completely abolished by preincubation of the cells with the synthetic furin inhibitor CMK (48), which prevents intracellular activation of proMT1-MMP (49). In addition, crude membranes prepared from furin inhibitortreated melanoma cells contained MT1-MMP mainly as unprocessed 63-kDa form. Interestingly, similar observations were reported by Maquoi et al. (50), showing that the reduced activation of proMMP-2 by HT1080 fibrosarcoma cells upon treatment with furin inhibitor is paralleled by suppressed levels of processed MT1-MMP protein. However, in contrast to the melanoma cells used in our studies, in HT1080 cells, proMMP-2 activation was not completely abolished even at higher concentrations (100 M). Although Cao et al. (51) have shown by transfection studies that MT1-MMP can activate proMMP-2 without cleavage by a furin-like proteases, our data strongly suggest that proteolytic cleavage of the N-terminal domain of proMT1-MMP is a prerequisite for proMMP-2 activation in melanoma cells. Previously published studies have also demonstrated that the serine proteases plasmin (23) and uPA (52) can extracellularly activate MT1-MMP. Indeed, the melanoma cells MV3 and BLM have been shown to produce uPA (53). However, activation of proMMP-2 is unlikely to be caused by uPA cleavage, as aprotinin added to the culture system did not prevent proMMP-2 processing. In conclusion, our data convincingly demonstrate that furin-activated MT1-MMP mediates proMMP-2 activation in high invasive melanoma cells.
Our data clearly show that interactions of the cells with the surrounding collagen matrix are a prerequisite for the induc-tion of MT1-MMP-mediated proMMP-2 activation. Under monolayer culture conditions, the lack of proMMP-2 activation correlates with the production of the unprocessed 63-kDa MT1-MMP protein form in all melanoma cell lines. In the floating collagen gel cultures, membrane preparations of all melanoma cell lines contain an additional protein form of MT1-MMP, which corresponds to the activated 60-kDa MT1-MMP described by others (44,50). Increased activation of proMMP-2 upon contact with fibrillar type I collagen also has been described for other cell types, including fibroblasts (19,20,54) and microvascular endothelial cells (8). In contrast to the collagen-induced MT1-MMP synthesis observed in fibroblasts and endothelial cells, which was paralleled by increased proMMP-2 activation in melanoma cells, we found increased intracellular processing of latent MT1-MMP by the collagenous matrix, whereas synthesis of MT1-MMP mRNA was barely affected. In addition, although active MT1-MMP was also detected in the membranes from low invasive melanoma cells, these cells failed to activate exogenous supplied proMMP-2. These data suggest that the presence of active MT1-MMP per se is not sufficient to explain the differences of proMMP-2 activation observed between high and low invasive melanoma cells. Therefore, other factors are likely to be responsible for the differences in the activation process.
As pointed out by Zucker et al. (27) and Butler et al. (25), the activation of proMMP-2 by the membrane activator MT1-MMP strongly depends on the presence of TIMP-2, which is bound via the interaction of the N-terminal domains of MT1-MMP. This MT1-MMP/TIMP-2 "receptor" in turn binds the C-terminal domain of proMMP-2 via the C-terminal portion of TIMP-2. Cleavage of MMP-2 might then occur by the concerted action of a second, not complexed, MT1-MMP molecule close to this complex (24,25,27). We observed that collagen-induced acti-  vation of proMMP-2 by the high invasive melanoma was inhibited by the addition of recombinant TIMP-2, suggesting that TIMP-2 is indeed involved in this process.
To investigate the role of TIMP-2 in proMMP-2 activation in high and low invasive melanoma cells, production of TIMP-2 by the melanoma cell lines was quantified. It turned out that TIMP-2 levels were significantly higher in the low invasive as compared with the high invasive cells. Additionally, we found an increase of TIMP-2 protein production by the low invasive cells in collagen gels, whereas the highly invasive cell lines reduced TIMP-2 production upon collagen contact. Since activated MT1-MMP protein was detected in low invasive melanoma cells upon contact with collagen, the increased TIMP-2 level in these cells might be responsible for the inhibition of proMMP-2 activation by MT1-MMP. Therefore, in low invasive melanoma cells, an altered balance of TIMP-2 and MT1-MMP probably contributes to the inhibition of proMMP-2 activation by blockage of the activator MT1-MMP, and finally leading to a lower invasive potential in the low invasive melanoma cells. On the other side, although TIMP-2 secretion is reduced in the high invasive cells upon contact with collagen, these levels seem to be sufficient to mediate the formation of a ternary complex between MT1-MMP, TIMP-2, and proMMP-2, and to allow proMMP-2 activation. These data are in agreement with the TIMP-2 studies performed by others (17,24,27) and underline once more the dual effect of TIMP-2 on proMMP-2 activation (24,51). To further clarify the role of TIMP-2 in the activation of MMP-2, we treated the high invasive melanoma cells with either recombinant TIMP-2 or with media conditioned by IF6 cells shown to contain high amounts of TIMP-2. As expected, the increased levels of TIMP-2 in the culture medium resulted in suppressed proMMP-2 activation by the high invasive cells.
Furthermore, to test whether MT1-MMP-mediated activation of proMMP-2 might contribute to the invasive ability of malignant melanoma, we analyzed the in vitro invasiveness of high and low invasive melanoma cells using Matrigel as barrier. The results showed that the ability of MV3 cells to penetrate this barrier was greatly reduced in the presence of the furin inhibitor. This reduction is probably a consequence of the reduced MT1-MMP cleavage. Addition of TIMP-2 or of IF6 conditioned media also resulted in a partial inhibition of the invasive capacity of MV3 cells. Taken together, these data indicate that proMMP-2 activation by MT1-MMP significantly contributes to the invasiveness of these melanoma cells, a process that strongly depends on the levels of TIMP-2.
Therefore, as shown in our studies, differences of the invasive phenotype might not become obvious unless more complex in vitro culture systems such as three-dimensional matrices are used.
FIG. 6. TIMP-1 and TIMP-2 production by melanoma cell lines. Highly (BLM, MV3) and low (530, IF6) invasive melanoma cell lines were grown as monolayers (open bars) or within three-dimensional collagen lattices (black bars) for 48 h. Then, the culture was continued with serum-free medium. After 24 h, the cells and collagen gels were combined with the media and homogenized by sonication. TIMP-2 and TIMP-1 levels were determined using an enzyme-linked immunoabsorbent assay as described under "Materials and Methods." The data are means Ϯ S.D. of triplicate experiments.