Activation of Matrix Metalloproteinase-9 (MMP-9) via a Converging Plasmin/Stromelysin-1 Cascade Enhances Tumor Cell Invasion*

Matrix metalloproteinase-9 (MMP-9) may play a critical catalytic role in tissue remodeling in vivo, but it is secreted by cells as a stable, inactive zymogen, pro-MMP-9, and requires activation for catalytic function. A number of proteolytic enzymes activate pro-MMP-9 in vitro, but the natural activator(s) of MMP-9 is unknown. To examine MMP-9 activation in a cellular setting we employed cultures of human tumor cells (MDA-MB-231 breast carcinoma cells) that were induced to produce MMP-9 over a 200-fold concentration range (0.03–8.1 nm). The levels of tissue inhibitors of metalloproteinase (TIMPs) in the induced cultures remain relatively constant at 1–4 nm. Quantitation of the zymogen/active enzyme status of MMP-9 in the MDA-MB-231 cultures indicates that even in the presence of potential activators, the molar ratio of endogenous MMP-9 to TIMP dictates whether pro-MMP-9 activation can progress. When the MMP-9/TIMP ratio exceeds 1.0, MMP-9 activation progresses, but through an interacting protease cascade involving plasmin and stromelysin 1 (MMP-3). Plasmin, generated by the endogenous urokinase-type plasminogen activator, is not an efficient activator of pro-MMP-9, neither the secreted pro-MMP-9 nor the very low levels of pro-MMP-9 associated with intact cells. Although plasmin can proteolytically process pro-MMP-9, this limited action does not yield an enzymatically active MMP-9, nor does it cause the MMP-9 to be more susceptible to activation. Plasmin, however, is very efficient at generating active MMP-3 (stromelysin-1) from exogenously added pro-MMP-3. The activated MMP-3 becomes a potent activator of the 92-kDa pro-MMP-9, yielding an 82-kDa species that is enzymatically active in solution and represents up to 50–75% conversion of the zymogen. The activated MMP-9 enhances the invasive phenotype of the cultured cells as their ability to both degrade extracellular matrix and transverse basement membrane is significantly increased following zymogen activation. That this enhanced tissue remodelling capability is due to the activation of MMP-9 is demonstrated through the use of a specific anti-MMP-9 blocking monoclonal antibody.

Degradation of connective tissue extracellular matrix (ECM) 1 and dissolution of epithelial and endothelial basement membrane are remodeling processes that occur during tumor invasion and metastasis. A family of proteolytic enzymes that have been functionally linked to these remodeling processes are the matrix metalloproteinases (MMPs) (1,2). The members of this family are multidomain, zinc-containing, neutral endopeptidases and include the collagenases, stromelysins, gelatinases, and membrane-type metalloproteases (3,4). Different MMPs can be induced in a variety of embryonic, and adult cell types and distinct subsets are often found to be elevated in tumor tissue and malignant cells in culture (5)(6)(7). Each MMP has a preferred substrate specificity toward individual matrix proteins, but there is overlapping specificity within the whole family (3,4), and thus a few MMPs acting in tandem have the potential to catalyze the complete degradation of the proteinaceous components of the basement membrane and ECM. However, the catalytic manifestations of MMP enzymes are highly regulated. First, the MMPs are expressed as inactive zymogens and require distinct activation processes to convert them to active enzymes (8,9), and second, a family of proteins, the tissue inhibitors of metalloproteinases (TIMPs), are correspondingly widespread in tissue distribution and function as highly effective MMP inhibitors (K i ϳ10 Ϫ10 M) (3). How the invading tumor cells that utilize the MMP's degradative capacity circumvent these negative regulatory controls is not well understood.
Within the cytokine-and oncoprotein-enriched tumor tissue environment, MMP expression more than likely will be upregulated, and thus the limiting reaction in MMP-mediated tumor invasion may be zymogen activation. Activation of MMP zymogens involves disruption of a coordination bond formed between a highly conserved unpaired cysteine in the aminoterminal propeptide of the pro-MMP molecule and the zinc ion at the active center (8). Disruption of this bond can be mediated by chemical and/or proteolytic mechanisms, often leading to an autoproteolytic event, resulting in the removal of the cysteinecontaining propeptide domain, generating a lower molecular weight, catalytically active form of the MMP. If an active MMP is involved in the initial proteolytic step or if an autolytic processing event is part of the final step, then the presence of TIMPs in the surrounding milieu will prevent zymogen activation. MMP activation can occur through intracellular, extracellular, and cell surface-mediated proteolytic mechanisms. Intracellular activation of stromelysin 3 (MMP-11) occurs in the Golgi network and is mediated by the intracellular serine pro-tease furin (10). Activation of interstitial collagenase (MMP-1) and stromelysin (MMP-3) occurs extracellularly and can be mediated by the serine protease, plasmin (11,12). Cell surface activation of an MMP can occur when gelatinase A (MMP-2) is brought into contact with a membrane-associated MMP, MT-1 MMP (13,14). Interestingly, gelatinase B (MMP-9), a close structural homologue of MMP-2 does not appear to be activated by the same mechanism, as it remains a zymogen under identical cellular conditions where a majority of co-expressed pro-MMP-2 is activated via the cell surface mechanism (15,16).
MMP-9 is produced by mesenchymal, epithelial, and hematopoietic cells and also by distinct tumor cell types (17)(18)(19)(20). While MMP-2 appears to be constitutively expressed by many cell types in culture, MMP-9 expression is induced by cytokines (21,22), growth factors (23), and cell/stroma interactions (19,20,24). MMP-9 expression has been correlated with a number of physiological and pathological processes, including trophoblast implantation (25), bone resorption (26), inflammation (18,19), and arthritis (27). Tumor cell invasion in a number of instances also has been linked with MMP-9 activity (5)(6)(7)20). Furthermore, Muschel and colleagues (28,29) have shown by both transfection and ribozyme-based approaches that MMP-9 is directly involved in tumor metastasis, and more recently MMP-9 activity has been linked with the process of tumor cell intravasation (30). Thus in these malignant cell systems, the regulatory controls that maintain MMPs as inactive zymogens were circumvented, since conversion of pro-MMP-9 to active enzyme had clearly occurred. The exact mechanism of MMP-9 activation in malignant tissue, however, has not been defined. A number of purified proteases, including trypsin (31,32), chymase (33), MMP-2 (34), tissue kallikrein (35), trypsin-2 (36), plasmin (37,38), MMP-7 (39), MMP-13 (40), and MMP-3 (31-33, 37, 38, 41, 42) have been reported to activate pro-MMP-9 in vitro. MMP-3, based on in vitro kinetic and catalytic parameters, appears to be the most efficient activator of pro-MMP-9 (32) and may be a natural activator in vivo. It is not clear, however, if MMP-3 will always be present in the tumor tissues where pro-MMP-9 is expressed. Furthermore, even if MMP-3 is available in the tissues, it is produced as a zymogen and also requires activation. In addition, the inhibitory potential of the TIMPs must be circumvented in the tumor tissue to allow for catalytic manifestations of both the activating MMP-3 and the activated MMP-9.
To demonstrate a possible tumor-associated mechanism of MMP-9 activation and the circumvention of TIMP-mediated control, we have examined cultures of a breast carcinoma cell line, MDA-MB 231, which expresses both MMP-9 and TIMP-1 and TIMP-2. MDA-MB-231 cultures do not appear to express MMP-2, also a potent gelatinase. Since MMP-9 activation is monitored by the generation of gelatinase activity, the absence of any interfering MMP-2 gelatinase activity in the cultures was critical for quantitating zymogen activation. Our results indicate that MMP-9 produced by the breast tumor cells is a stable zymogen even when exceptionally high levels of MMP-9 are induced and when active plasmin is generated in the cultures. However, when pro-MMP-3 is introduced into the system at concentrations that stoichiometrically exceed the endogenous TIMP levels, plasmin activates the MMP-3 which in turn efficiently processes and activates the pro-MMP-9. The activated MMP-9 contributes significantly to tumor cell-mediated ECM degradation and basement membrane invasion, and the specificity of these mechanisms is demonstrated through the use of neutralizing anti-MMP-9 monoclonal antibodies.

MATERIALS AND METHODS
Cell Culture-MDA-MB-231 cells, HT1080 cells, and MDA-MMP-9 cells were cultured at 37°C in 8% CO 2 in DMEM containing 10% heat inactivated fetal bovine serum (FBS) supplemented with 0.1 mM nonessential amino acids, 2 mM glutamine, 1 mM sodium pyruvate, and 100 units/ml penicillin with 100 g/ml streptomycin. For harvesting serumfree conditioned medium, the cell cultures were washed with DMEM and incubated for 24 h with serum-free DMEM that included the above mentioned supplements.
MMP-9 Enzyme-linked Immunosorbent Assay-Microtiter 96-well plates were precoated with an anti-MMP-9 monoclonal antibody (6-6B IgG 2b ) (1 g/ml, 50 l/well) in PBS for 1 h at room temperature and blocked with 1% bovine serum albumin/PBS (100 l/well) for 1 h at room temperature. Conditioned medium (50 l/well) was added for 1 h at 37°C followed by washing and the addition of a secondary anti-MMP-9 monoclonal antibody, 7-11C IgG 1 , (1 g/ml, 50 l/well) in 1% bovine serum albumin/PBS for 1 h at 37°C. The cells were washed with PBS, 0.5% Tween, incubated with alkaline phosphatase-conjugated goat anti-mouse IgG (1:1000 in 1% bovine serum albumin/PBS), followed by washing and the addition of p-nitrophenyl phosphate substrate and the development of color according to manufacturer (Kirkegaard and Perry, Gaithersburg, MD) instructions.
Western Blot Analysis-Samples of conditioned medium were first separated by SDS-polyacrylamide gel electrophoresis and then electrophoresed onto nitrocellulose membranes (Millipore, Burlington, MA). The membranes were blocked in 5% non-fat milk in Tris-buffered saline, 0.5% Tween, washed, and incubated overnight with purified monoclonal antibodies (1-5 g IgG/ml). The blots were then washed and incubated for 2-4 h with secondary antibody, horseradish peroxidaseconjugated goat anti-mouse IgG (Kirkegaard and Perry) at 1:1000 dilution. The blots were developed using the ECL chemiluminescence detection system (Amersham Pharmacia Biotech).
Activation of Pro-MMP-9 in Culture-The MDA-MB-231 cells and MDA-MMP-9 cells were plated at 1 ϫ 10 5 cells/cm 2 in DMEM, 10% FBS. After 24 h the cells were rinsed once with 2 mM 6-amino-n-caproic acid in serum-free DMEM to remove any cell surface plasminogen. The cells were rinsed twice with serum-free DMEM and incubated in the same medium in the presence or absence of the following components: plasminogen (25 nM), pro-MMP-3 (2-16 nM), active MMP-3 (5 nM), aprotonin (40 g/ml), TIMP-1 (20 nM), anti-MMP-9 IgG (200 nM), and normal mouse IgG (200 nM) added singularly or in the indicated combinations. The cell cultures were incubated for 24 -48 h and the conditioned medium collected and stored at Ϫ70°C before analysis. Pro-MMP-3 was expressed as a recombinant protein in Chinese hamster ovary K-1 cells and purified from conditioned medium as described previously (44). Active MMP-3 (MMP-3⌬C) was a C-terminal truncated derivative of MMP-3 that spontaneously activated and yielded a specific activity equivalent to native MMP-3. It was expressed in Escherichia coli and purified as described previously (45). Anti-MMP-9 monoclonal antibody 7-11C blocks activation of pro-MMP-9 and was purified from hybridoma conditioned medium as described previously (46).
Preparation of Cell Extracts-MDA-MMP-9 cells were plated and grown to confluence in 100-mm dishes (1.5-2 ϫ 10 7 cells/plate in DMEM, 10% FBS. Each plate was washed twice with serum-free DMEM with aprotinin 40 g/ml) and once with serum-free DMEM alone. The cells were incubated in 5 ml of serum-free DMEM either alone, with 2 g/ml plasminogen, with 16 nM pro-MMP-3, or in combination for 20 h at 37°C. At the end of this time, the conditioned media were collected, and 40 g/ml aprotinin was added. The cells were washed twice with phosphate-buffered saline and extracted at 4°C with constant shaking in 1 ml of lysis buffer (0.1 M Tris, pH 8.0, 0.5% Triton X-100) with aprotinin (40 g/ml). The extracts were centrifuged at 12,000 rpm for 10 min at 4°C. The resulting supernatants were each incubated for 1 h at 4°C in an end-over-end mixer with 100 l of gelatin-Sepharose that had been equilibrated with 50 nM Tris, pH 7.25, 200 nM NaCl, 10 nM CaCl 2 , 0.05% Brij-35. The beads were washed five times in 1 ml of equilibration buffer, and the total cellular MMP-9 was eluted from the beads by resuspending the slurry twice with 75 l equilibrium buffer containing 10% Me 2 SO, centrifuging twice, and recovering and combining the supernatants (150 l).
ECM Degradation Assays-[ 3 H]Proline-labeled rat smooth muscle cell ECM was prepared as described by Jones et al. (51). Degradation experiments were carried out on ECM predigested with trypsin (5 g/ml for 2 h at 37°C in PBS). Following digestion, trypsin activity was inhibited with soybean trypsin inhibitor (10 g/ml), and the matrices were washed three times with DMEM (500 l). For the matrix degradation assays, the cells were seeded on radiolabeled ECM-coated 48well plates at 1 ϫ 10 5 cells/cm 2 in 250 l of DMEM. After 24 h the cells were rinsed once with serum-free DMEM and incubated in the same medium (250 l) in the presence or absence of the following components: plasminogen (25 nM), pro-MMP-3 (16 nM), aprotonin (40 g/ml), 7-11C anti-MMP-9 IgG (200 nM), control IgG (200 nM), and TIMP-1 (20 nM). To follow the progressive degradation of the ECM, aliquots (50 l) of the culture supernatants were harvested after 24, 48, and 72 h, and the soluble radioactivity quantitated in a scintillation counter.
Basement Membrane Invasion Assay Using Matrigel-Matrigelcoated filter inserts (8 m pore size) that fit into 24-well invasion chambers were obtained from Becton Dickinson (Bedford, MA). The MDA-MMP-9 cells to be tested for invasion were detached from tissue culture plates with a nonenzymatic cell dissociation solution (Sigma), washed, and resuspended in DMEM (5 ϫ 10 4 cells/200 l) and added to the upper compartment of the invasion chamber in the presence or absence of the indicated components. Culture medium (500 l) was added to the lower compartment of the invasion chamber. The Matrigel invasion chambers were incubated at 37°C for 48 h in 8% CO 2 . After incubation the filter inserts were removed from the wells, and the cells on the upper side of the filter were removed using cotton swabs. The filters were fixed, mounted, and stained according to the manufacturer's instructions (Becton Dickinson). The cells that invaded through the Matrigel and were located on the under side of the filter were counted. Three to five invasion chambers were used per condition. The values obtained were calculated by averaging the total number of cells from three filters.

Enhanced Expression of MMP-9 from Human Tumor Cells
That Do Not Express MMP-2-To examine the activation of pro-MMP-9 in a cellular setting that is devoid of MMP-2, cultures of the human breast carcinoma cell, MDA-MB-231, were treated with cytokines known to stimulate the production of pro-MMP-9. Gelatin substrate zymography, a method that can detect subnanogram amounts of gelatinase, was used to monitor levels of MMP-9 in the conditioned medium harvested from the cultures. A parallel culture of HT1080, a human fibrosarcoma cell line that produces both MMP-2 and MMP-9 was used as control. Treatment of MDA-MB-231 cells with PMA and interleukin-1 resulted in increased levels of MMP-9 expression while platelet-derived growth factor treatment yielded no change in MMP-9 levels ( Fig. 1). There was no evidence of MMP-2 expression in the treated cultures. Although PMA stimulation resulted in a substantial increase in MMP-9 expression, an 82-kDa form was not observed, indicating that the induced enzyme was not activated but remained in the 92-kDa zymogen form.

Generation of Active Stromelysin (MMP-3) in MDA-MB-231
Cultures Fails to Activate Pro-MMP-9 -To determine if a previously proposed MMP-3-dependent mechanism of pro-MMP-9 activation (41) could occur in a complex cell culture system, a proteolytic cascade that would yield active MMP-3 was initiated in the MDA-MB-231 cultures. These cultures produce uPA at levels sufficient to catalyze the conversion of plasminogen to plasmin. Plasmin is a known activator of pro-MMP-3 (12), and active MMP-3 had been shown to activate pro-MMP-9 under in vitro conditions using purified preparations of MMP-3 and pro-MMP-9 (32,38). The proteolytic cascade was initiated by supplementation of the PMA-treated MDA-MB-231 cultures with pro-MMP-3 (16 nM) and plasminogen (25 nM). The activation of pro-MMP-3 was monitored by immunoblot analysis and activation of pro-MMP-9 was measured by zymography (Fig. 2). In the absence of plasminogen, the addition of pro-MMP-3 had no effect on the status of the pro-MMP-9 as it remained in the 92-kDa zymogen form (Fig. 2, zymograph, lane 2). The pro-MMP-3 also was maintained as the 55-kDa zymograph (Fig. 2, immunoblot, lane 2). Addition of plasminogen plus pro-MMP-3 caused a conversion of the 55-kDa pro-MMP-3 to the 45-kDa active MMP-3 (immunoblot, lane 3) but, unexpectedly, little or no activation of pro-MMP-9 occurred; only trace levels of an 82-kDa form of MMP-9 appeared in the zymograph (lane 3), and no gelatinase activity was detectable. Plasmin was generated in the culture system since plasmin activity as measured by a specific peptide hydrolysis assay was detected only in the plasminogen-containing cultures (data not shown). The plasmin that was generated in this system was responsible for the conversion of pro-MMP-3 because the addition of aprotinin, a The absence of conversion of pro-MMP-9 to active MMP-9 in the presence of a functioning proteolytic cascade, which clearly had generated active MMP-3, was further investigated. The relative concentrations of pro-MMP-9, TIMP-1, and TIMP-2 in the cultures were determined (Table I). Interestingly, the levels of pro-MMP-9 in the PMA-stimulated cultures were increased 8-fold over the levels in the unstimulated culture (0.25 nM versus 0.03 nM), but the endogenous TIMP concentrations still remained 10 -20-fold higher than the MMP-9 concentration in these cultures (3-4 nM versus 0.25 nM). These data suggested that TIMP levels were controlling the progression to active MMP-9 and that activation might only occur when the total MMP concentration exceeded the TIMP concentration in the immediate environment. Within the MDA-MB-231 cell culture system, the molar concentration of pro-MMP-9 would have to be increased at least 20-fold to exceed the endogenous TIMP levels.
An MDA-MB-231-transfected Cell Line Expressing Increased Levels of MMP-9 -To achieve increased levels of MMP-9 expression, MDA-MB-231 cells were transfected with an MMP-9 cDNA construct (28), selected, cloned, and analyzed for MMP-9 expression by zymography and immunoblot analysis. A stable cell line that secreted substantially increased levels of MMP-9 was isolated and designated MDA-MMP-9 (Fig. 3, A and C). Like the parent MDA-MB-231 cells, this cell line did not express MMP-2 (Fig. 3, B and C), MMP-3 (Fig. 3D), or MMP-1 (Fig. 3E) and did express significant levels of uPA (Fig. 3H). A direct enzyme assay indicated that the uPA concentration in the cell culture supernatant was 0.1-0.2 g/ml, which could very effectively activate microgram quantities of plasminogen. Although this cell line expressed increased levels of MMP-9, the TIMP levels were the same as the parent cell line and slightly less than the TIMP levels expressed by the PMAtreated MDA-MB-231 cells (Fig. 3, F and G). Quantitation of MMP-9 levels expressed by the MDA-MMP-9 cell line showed that the MMP-9 concentration now exceeded the concentration of both TIMP-1 and TIMP-2 (8 nM versus ϳ3 nM) ( Table I, third  column). However, the MMP-9 in the transfected cultures remained in the 92-kDa zymogen form despite the significant increase in the MMP-9:TIMP ratio (Fig. 3, A and C), and no gelatinase activity was manifested by the transfected cultures (data not shown). Furthermore, when the transfected cultures were incubated in the presence of plasminogen, which was rapidly converted to plasmin, the 92-kDa zymogen form was maintained (see Fig. 4, inset, lane 1).
MMP-3-dependant Activation of the Pro-MMP-9 Secreted by the MDA-MMP-9 Cells-The activation of pro-MMP-9 in the tumor cell culture system was re-analyzed using the trans-   fected MDA-MMP-9 cells. By adding plasminogen (2.5 g/ml) and pro-MMP-3 at increasing concentrations to the cell culture system, the activation of MMP-9 occurred in a dose-dependent manner (Fig. 4). The appearance of a gelatinolytic zone with an apparent molecular mass of 82 kDa was observed upon the addition of 8 -16 nM pro-MMP-3 (Fig. 4, zymograph inset), and this change in molecular mass was accompanied by a corresponding increase in solution phase gelatinolytic activity (Fig.  4, line graph). In the presence of 16 nM pro-MMP-3, more than half of the 92-kDa form of the enzyme had been processed, and significant gelatinolytic activity was observed.
The presence in the cell culture system of active MMP-3 generated from the added pro-MMP-3 was measured using a fluorogenic peptide assay that is specific for MMP-3 (49). Enzyme activity above background was observed only in cultures containing pro-MMP-3 at concentrations Ն8 nM (Fig. 4, bar graph), which corresponded with the appearance of both the 82-kDa form of MMP-9 and gelatinase activity. At concentrations less than 8 nM pro-MMP-3, MMP-3 activity as well as conversion of MMP-9 and gelatinase activity were at background levels. These data suggested that the effective generation of gelatinolytic activity was dependent on concentrations of MMP-3 and MMP-9 that exceeded the endogenous TIMP concentrations.
To determine the pro-MMP-3 concentration necessary to activate MMP-9 in the absence and presence of stoichiometric levels of TIMP-1, the plasmin-induced proteolytic cascade was reconstituted in vitro using purified components (Fig. 5). Purified pro-MMP-9 was added at 8 nM, the concentration present in the MDA-MMP-9 cell culture system (Table I). Plasmin was added at 10 nM, the concentration determined to be generated from 25 nM plasminogen in a uPA-containing culture, and pro-MMP-3 was added at increasing concentrations (0 -16 nM). Plasmin alone failed to convert pro-MMP-9 (Fig. 5A, lane 1). The conversion of pro-MMP-9 to the 82-kDa form of the enzyme was observed with as little as 2 nM of pro-MMP-3 in the absence of TIMP (Fig. 5A, lanes 2-5). However, the addition of purified TIMP-1 at a concentration of 4 nM prevented the conversion of pro-MMP-9 until 8 nM pro-MMP-3 was added and then the 82-kDa form is observed (Fig. 5B). Almost complete conversion of the enzyme occurred at a pro-MMP-3 concentration of 16 nM, which yields an effective concentration of 6 -8 nM active MMP-3 (40 -50% conversion). This activation pattern of MMP-9 closely resembled the pattern observed in the crude cell culture system (Fig. 4) and provided additional evidence that the activation of MMP-9 was regulated by the levels of TIMP in the system.
The requirements of the proteolytic cascade for MMP-9 activation and the contribution of the individual components was evaluated by the addition of specific inhibitors to the supplemented MDA-MMP-9 cell cultures. Pro-MMP-3 and plasminogen were employed at concentrations that would yield 50 -75% conversion of pro-MMP-9, and samples of culture supernatants were analyzed by zymography for conversion of pro-MMP-9 and solution phase gelatinase activity for appearance of active enzyme (Fig. 6). Only background levels of gelatinase activity were generated when either plasminogen or pro-MMP-3 was added. Interestingly, some processing of pro-MMP-9 to an intermediate 84 -86-kDa form was present (zymograph, lanes 2 and 3), but this form was not enzymatically active in solution  (bar graph, lanes 2 and 3). Substantial gelatinase activity was A partner monoclonal antibody, 6-6B, isolated from the same fusion as 7-11C, had been shown previously to block organo-mercurial mediated activation of pro-MMP-9 (46).
The presence in the cultures of both plasmin and pro-MMP-3 are required for active MMP-9 to be generated, and neither one alone is sufficient (Fig. 6). Plasmin efficiently activates pro-MMP-3 (Fig. 2), and presumably the active MMP-3 directly converts 92-kDa pro-MMP-9 to 82-kDa active MMP-9. However, it was possible that plasmin, in addition to activating pro-MMP-3, could proteolytically modify the pro-MMP-9, thereby allowing it to be efficiently processed and activated by MMP-3. Plasmin thus could have a dual requirement in the system. To address this possibility, serum-free conditioned medium from MDA-MMP-9 cells containing endogenous pro-MMP-9 (8 nM) was preincubated in the absence or presence of plasminogen (2 g/ml). After 16 h of incubation, the plasminogen-containing sample had a measured concentration of active plasmin of 0.9 g/ml (10 nM), a significant level of active enzyme. Aprotinin was added to the samples, completely inhibiting any further action of plasmin, and then 5 nM active MMP-3 was added to both samples and incubation at 37°was continued. At various times, aliquots were withdrawn from both samples and analyzed by gelatin zymography to monitor the pro-MMP-9 conversion process and by radiolabeled gelatin degradation to monitor the appearance of active MMP-9. Fig. 7 demonstrates that the processing of pro-MMP-9 by MMP-3 is nearly identical for plasmin-pretreated or -untreated samples (Fig. 7, A and B). Plasmin pretreatment also did not enhance the rate of appearance of active MMP-9 and in fact slightly retarded the appearance of gelatinase activity (Fig. 7C). These results also illustrate that prior to MMP-3 addition, plasmin alone at physiological concentrations does not generate any detectable gelatinase activity or conversion to 82-kDa forms (Fig. 7B, lane 1). Furthermore, the requirement for plasmin in the activation process appears to be solely to activate pro-MMP-3, since 5 nM active MMP-3 alone can replace both plasminogen and the 16 nM pro-MMP-3 in generating full conversion to the 82-kDa gelatinase (Fig. 7A, lane 6). Interestingly, 2.5 nM MMP-3, in contrast to 5 nM MMP-3, was unable to generate any active gelatinase in the treated or untreated conditioned media (data not shown), consistent with the need to overcome the 3-4 nM TIMP present in the conditioned medium.
Activation of the Cell-associated MMP-9 Appears to Have Similar Plasmin/MMP-3 Requirements-It has been reported that MMP-2 and MMP-9 are associated with the cell surface and can be activated independently of other MMPs directly via a cell surface uPA/plasmin cascade (52). The association of pro-MMP-9 with cells, and its activation requirements were examined in the overexpressing MDA-MMP-9 cultures. The cells were incubated for 24 h in serum-free medium in the absence or presence of either plasminogen, pro-MMP-3, or both together. The conditioned medium (5 ml) was harvested from the four cultures, and a total cell and membrane extract was prepared from the same harvested cultures. The extract was passed over and eluted from gelatin-Sepharose in order to concentrate all the cell-associated MMP-9 free of any extraneous cellular proteins and inhibitors (0.15 ml). Aliquots (1 l) from each conditioned medium and extract were examined by zymography, and aliquots (75 l) from each were tested for active gelatinase (Fig. 8). The total gelatinase activity in each sample was calculated to provide a measure of how much active MMP-9 was cell-associated compared with the amount of secreted, soluble MMP-9. The zymographic analysis indicates that a lower molecular mass form of MMP-9 (ϳ80 -85 kDa) exists in the cells from the untreated cultures (lane 5), but it would not appear to be the 82-kDa active gelatinase, since no enzymatic activity can be detected assaying as much as 50% (75 l) of the total cell extract. The 80 -85-kDa band may be a variant glycosylated form of pro-MMP-9 found associated only with cells and not secreted (53). Such a cell-associated MMP-9 variant would exist as a zymogen and thus would manifest little or no soluble enzymatic activity but yield a zone of lysis in the zymograph. Alternatively, the 80 -85-kDa species could be an active form of MMP-9 that is tightly complexed with a cell-associated inhibitor and thus in solution unable to express gelatinase activity, but following dissociation in SDS-polyacrylamide gel electrophoresis would yield a zone of lysis in the zymograph. This band is present in all of the cell extracts; however an enhanced zymographic signal at 80 -85 kDa appears in the extracts prepared from cells treated with pro-MMP-3 and plasminogen (Fig. 8, lane 8). This zymograph enhancement is accompanied by a small but significant increase in the soluble gelatinase activity of the sample. The cell-associated gelatinase activity (659 cpm) is much lower than the activity (14,056 cpm) manifested by a similar 82-kDa band generated in the corresponding conditioned medium (lane 4), possibly indicating the presence of an inhibitor in the cell extracted sample. When the gelatinase activity present in the total cell extract is calculated, the amount of active enzyme present (0.30 unit) represents only 0.15% of the total activity of the culture: 99.8% of the activity (197.6 units) is present in the conditioned medium. Thus only a very small level of active MMP-9 is associated with the tumor cells, possibly on the cell surface. However, even that small level of active enzyme appears to require the combined action of plasmin and pro-MMP-3 for its generation. With no pro-MMP-3 present, the presence of plasmin alone causes little or no increase in cellassociated active MMP-9 (Fig. 8, lane 7).
MMP-9 Activation via the Plasmin/MMP-3 Cascade Enhances ECM Degradation-The unique reactivity of the monoclonal antibody, 7-11C, to human MMP-9 (46) and that it could prevent the generation of active MMP-9 (Fig. 6) indicates that the antibody could be used to assess the specific involvement of MMP-9 activation in human tumor cell invasive behavior. To assess the functional activity of the MMP-9 generated in the MDA-MMP-9 tumor cell culture system, progressive matrix degradation by these cells was analyzed in the absence and presence of the monoclonal antibody. The ECM is coated with glycoproteins that protect the fibrillar collagens from the catalytic activity of MMPs; thus examination of the role of specific MMPs in ECM degradation requires removal of this glycoprotein layer (54). To evaluate the role of MMP-9 activity in ECM degradation, MDA-MMP-9 cells were cultured on a smooth muscle cell matrix that was radiolabeled and partially depleted of glycoproteins by trypsin treatment (Fig. 9). Incubation of the cells alone for 3 days resulted in a low level of ECM degradation (Fig. 9, condition A), and addition of plasminogen to this system resulted in a small but significant increase of released radiolabeled ECM fragments over the three days (condition B). The addition of plasminogen and pro-MMP-3, however, enhanced the ECM degradation 3-fold over the culture of cells alone (C). This enhanced ECM degradation was due to MMP-9 activation as shown by the resulting inhibition using the specific anti-MMP-9 monoclonal antibody (condition D) and TIMP-1 (condition E). The enhanced solubilization of radiolabeled ECM that is sensitive to the anti-MMP-9 antibody is apparently the result of active MMP-9 degrading the collagenous components of the ECM. Since the smooth muscle ECM does not contain type IV or type V collagen, the known native substrates of MMP-9, the observed degradation is likely the result of MMP-9 acting catalytically on unfolded or partially denatured collagens I and III, the major collagens found in this ECM.
MMP-9 Activation Results in Enhanced Tumor Cell Invasion of Basement Membrane-Transmigration of cells across a complex basement membrane is often used as an indicator of the invasive behavior of cells. The migratory and invasive ability of MDA-MMP-9 cells were examined in a series of Matrigel invasion assays to determine the relative contribution of MMP-9 activation to this translocation process (Fig. 10). MMP-9 conversion and the generation of MMP-9 gelatinase activity were monitored using samples of culture medium that were removed directly from the upper compartment of the invasion chamber during the assay (Fig. 10, top). In the absence of plasminogen and pro-MMP-3, 2600 cells, approximately 5% of the inoculated MDA-MMP-9 cells, transmigrated across the filter in 48 h (Fig.  10, column 1). In the presence of either plasminogen or pro-MMP-3, 7-8% of the inoculated cells transmigrated across the filter (columns 2 and 3). These levels of cellular invasion were accompanied by a small generation of gelatinase activity (418 and 595 cpm, respectively) above background levels (267 cpm). When plasminogen and pro-MMP-3 were both added to the culture chamber, 6000 cells or ϳ12% of the inoculated cells transmigrated across the filter in 48 h (column 4). This enhanced transmigration was accompanied by conversion of pro-FIG. 7. Conversion and activation of pro-MMP-9 by active MMP-3 following pretreatment with plasmin. Serum-free conditioned medium was collected after 24 h from 1.5 ϫ 10 7 MDA-MMP-9 cells yielding a concentration of pro-MMP-9 of 8 -10 nM. The cell-free conditioned medium (3 ml) was incubated at 37°C for 16 h in the absence (A) or presence (B) of plasminogen (2 g/ml). Plasminogen conversion to plasmin was monitored, and at the end of the preincubation, active plasmin was present at 0.9 g/ml (10 nM). Aprotinin (40 g/ml) was added to the conditioned media to block any further activity of plasmin, and then 5 nM active MMP-3 was added along with radiolabeled gelatin (20 g/ml, 2000 cpm/g), and incubation was continued at 37°C. At the indicated times, 1-l aliquots were removed for gelatin zymography and 100-l aliquots removed for measuring [ 3 H]gelatin degradation (C).
MMP-9 to the 82-kDa form of the enzyme (Fig. 10, zymograph) and a substantial increase in gelatinase activity (3056 cpm). The cellular transmigration and gelatinolytic activity were inhibited by the anti-MMP-9 monoclonal antibody (column 5) and TIMP-1 (column 6), indicating a specific role for MMP-9 and its activation in the invasion process. DISCUSSION In the present study, we have utilized a cell culture model in an attempt to recapitulate some of the biochemical events that might take place in an invasive tumor when active MMP-9 is generated. A critical feature of the model is that the natural regulators of MMP function, the TIMPs, are present endogenously in nanomolar amounts. A second feature of the model is that it has the capacity to generate an additional protease system, namely the uPA/plasmin cascade, that has been closely linked to the migratory, invasive phenotype (7) and also linked functionally to the activation of a number of MMPs including MMP-9 (55). Our initial analysis of this culture system (Figs. 1 and 2) illustrates the inherent stability of the pro-MMP-9 zymogen and its apparent resistance to activation even when the levels of MMP-9 are increased by cytokine treatment, when plasminogen is added to the cultures and when active MMP-3 is generated. Since active plasmin was demonstrated to be present in the cultures and at sufficient catalytic levels to activate pro-MMP-3, plasmin did not appear to be an effective activator of pro-MMP-9. However, MMP-3 had been shown in vitro to be an efficient pro-MMP-9 activator (32), and the question arose as to why the newly generated MMP-3 did not activate the available pro-MMP-9. Determining the total TIMP (4 nM) present in the cultures and comparing it with the amount of available pro-MMP-9 (0.25 nM) indicated that sufficient TIMP-1 was present to complex all of the pro-MMP-9 and retard its activation. Furthermore, excess TIMP-1 and TIMP-2 also would be present to inhibit the nanomolar levels of the plasmin-activated MMP-3. Even if the newly generated MMP-3 exceeded the TIMP levels, it was possible that pro-MMP-9 at subnanomolar levels tightly complexed with TIMP-1 and in equilibrium with other TIMPs in the cellular milieu was unable to interact efficiently with the excess MMP-3 to progress through activation. Thus the dominating molar excess of the TIMPs was one compelling reason why MMP-9 activation could not progress even in the presence of a potential activator.
In order to overcome the suppressive effect of the TIMPs, transfection of MDA-MB-231 cells was employed to up-regulate the MMP-9 levels. The transfected cells were unchanged in terms of expression of all the proteases and protease inhibitors tested (Fig. 3) with the exception of MMP-9, which was increased 200-fold over the parental cultures and 30-fold over that of the PMA-treated cultures. However, overriding the TIMP levels with excess MMP-9 expression was not sufficient by itself to initiate activation of pro-MMP-9. The MMP-9-overexpressing cultures needed to be supplemented with both plasminogen and pro-MMP-3 (Figs. 4 and 6). Supplementation with plasminogen alone did not bring about activation, even though active plasmin was generated via the endogenous tumor cell uPA. This result documents the inherent stability of the pro-MMP-9 zymogen in a cellular setting and further indicates that plasmin is not a direct activator of pro-MMP-9 under the cellular conditions employed herein. Studies from other laboratories also had concluded that plasmin was not an efficient activator of pro-MMP-9 (31,37), although a number of reports had implicated plasmin as a direct activator of pro-MMP-9 (38,57,58,62). Supplementation of the transfected cultures with pro-MMP-3 alone also did not yield activated MMP-9, as the pro-MMP-3 in the absence of plasminogen remained a 55-kDa zymogen, further demonstrating that plasmin was responsible for the conversion of 55-kDa pro-MMP-3 to the 45-kDa active MMP-3. When both plasminogen and pro-MMP-3 were added to the cultures, the tumor cell uPA activated the plasminogen, the generated plasmin activated the pro-MMP-3, and then active MMP-3, when it exceeded the concentration of TIMP, converted 92-kDa pro-MMP-9 to 82-kDa active MMP-9.
In studies of zymogen activation by proteases, it would seem important to distinguish true enzyme activation from proteolytic conversion of the proenzyme to smaller but catalytically inactive forms. Enzyme activity measurements in solution were critical in the present study, since distinct lower molecular mass forms of MMP-9 were generated upon addition of plasminogen alone or pro-MMP-3 alone (Fig. 6). These processed forms were active in gelatin substrate gels (zymographs) but were not active in solution, and thus true activation of MMP-9 did not occur under these conditions. In addition, a lower molecular mass form of MMP-9 was observed in tumor cell extracts (Fig. 8), and its similar zymographic position to the 82-kDa active MMP-9 suggested that it represented a cellassociated form that had been activated. However, enzyme activity measurements on the isolated cellular MMP-9 indicated that this lower molecular mass form was not an active enzyme but behaved like a zymogen. Other laboratories also have shown that certain lower molecular weight forms of MMP-9 are not enzymatically active (41,42,53). Toth et al. (53) demonstrated that a cell-associated 85-kDa species of MMP-9 was active in zymographs but enzymatically inactive in solution. The 85-kDa form represented a different glycosylated variant of pro-MMP-9. Ogata et al. (41) showed that an 86-kDa processed form of MMP-9 was zymographically active but not enzymatically active in solution. This 86-kDa form represented a species that had been proteolytically processed at the Glu 41 -Met 42 site in pro-MMP-9, 47 residues upstream of the actual activation site. A number of studies (38, 56 -58), however, have relied mainly on zymographic or immunoblot detection methods and implied that the appearance of distinct lower molecular mass forms of MMP-9 represented activation of the zymogen. In some of these studies plasmin was implicated as a direct activator of pro-MMP-9 (57,58). In the culture system described herein, plasmin indeed can generate lower molecular mass forms of MMP-9 but does not yield activated enzyme. The plasmin involvement in MMP-9 activation in this system is only indirect; it functions within an interacting cascade to generate active MMP-3. This indirect role of plasmin and the apparent requirement for plasminogen in the cascade is more clearly illustrated when already activated MMP-3 is added directly to the tumor cell culture medium (Fig. 7). The requirement for plasminogen and pro-MMP-3 in the cascade is circumvented as active MMP-3 directly generates the 82-kDa MMP-9 and in turn generates significant gelatinase activity.
In contrast to the present studies, Mazzieri et al. (52), using HT1080 cultures, concluded that plasmin could process and activate pro-MMP-9 (and also pro-MMP-2) in the absence of any detectable MMP-3. The plasmin that appeared to be responsible for this MMP activation was cell surface-activated plasmin, and the gelatinases that were activated were apparently associated with the cell surface. Since most of our studies involve the secreted, soluble pro-MMP-9, which could be readily activated by MMP-3 and could not be activated at all by plasmin, the apparent contrasting results may reflect cell-associated activation versus solution phase activation. However, when the tumor cell-associated MMP-9 was examined (Fig. 8), it was found that little if any active MMP-9 was associated with the cells, that plasmin did not enhance the level of active MMP-9 that became cell-associated, and that when a detectable level of active MMP-9 was found associated with the cells After a 48-h incubation the total number of cells that invaded and migrated to the underside of the filters was counted (bar graph). After 24 h of invasion, an aliquot of the conditioned medium from the upper chambers was removed and analyzed by gelatin substrate zymography (zymograph inset). The conditioned medium also was assayed for gelatinase activity by monitoring the degradation of 3 H-labeled gelatin (mean counts/min released is indicated above each lane of the zymograph).
(Ͻ1%), the specific requirements for its activation appeared to be the same as that for activation of the secreted, soluble MMP-9. Thus it is not yet resolved whether significant levels of pro-MMP-9 become bound to the cell surface and whether such cell-associated zymogen has a distinct mode of activation.
A possible limitation of this cell culture model for MMP-9 activation is that the source of active MMP-3 is via exogenous addition of purified pro-MMP-3. The cultured breast carcinoma cells do not produce MMP-3 (Fig. 3), and indeed a majority of human tumor cells may not manifest elevated expression of MMP-3 (59). The major source of MMP-3 in tumor tissue appears to be not from the tumor cells but from stromal cells, namely fibroblasts and macrophages (60,61). In a study of breast tumor tissue, Heppner et al. (19) indicated that the enhanced production of a number of MMPs, including MMP-3, was not from the tumor cells but from adjacent inflammatory and stromal cells responding to the presence of tumor. Therefore we envision the exogenous addition of pro-MMP-3 as representing a stromal source of this zymogen. Indeed the concentration of pro-MMP-3 (2-16 nM) employed in the model system was higher than that which might be found in normal stromal tissue or in standard cultures of fibroblasts and macrophages. However, the elevated levels of MMP-3 often found with inflammation, wound repair, and tumor infiltration (1-3) might approach nanomolar levels. Thus the concentration of MMP-3 employed in the present study may not be physiologic but rather pathologic.
It should be emphasized that MMP-3 may not be the sole natural activator of pro-MMP-9. In the injured arteries of homozygous MMP-3-deficient mice, some MMP-9 activation occurs approximately equal to the level observed in the injured arteries of wild type mice, and it was concluded that MMP-9 activation does not depend solely on MMP-3 (62). It should be noted, however, that MMP-9 activation in (62) was monitored only by the appearance of lower molecular mass forms of MMP-9. Nevertheless, MMP-3 probably will not be found in all the tissues where MMP-9 activation occurs. However, since very favorable catalytic parameters describe MMP-3's efficient MMP-9 activating ability in vitro (32), and since active MMP-3 can be generated by a widely distributed serine protease cascade (Fig. 6), and since MMP-3 can effectively generate a processed form of MMP-9 that functions catalytically in tissue invasion processes (Figs. 9 and 10), MMP-3 would appear to be a prime candidate as a natural activator of pro-MMP-9 in vivo. Other pro-MMP-9 activators also may function in vivo (and in MMP-3-deficient mice). Procollagenase-3 (MMP-13), like MMP-3, exhibits favorable kinetics of pro-MMP-9 activation, can itself be activated by a widely distributed protease system and is co-expressed with MMP-9 in a number of situations that involve tissue and skeletal remodeling (40,63).
The TIMPs are clearly an important controlling element in any MMP-mediated function. Their regulatory capacity in the MMP-9 activation model described in this study is multifold. They can inhibit MMP-3, the activator of the process; one of them, TIMP-1, can bind specifically to the MMP-9 zymogen and dampen its activation; and finally the TIMPs can inhibit the activated MMP-9 from functioning catalytically. The regulatory ability of TIMPs appears to be based mainly on stoichiometry, and thus a mechanism for circumventing TIMP control of cellular MMP-9 activation would be to titer out the endogenous TIMP. In our model system this was accomplished by inducing overexpression of pro-MMP-9 in the tumor cells and by providing nanomolar levels of pro-MMP-3. In vivo such titering or saturation of TIMPs may occur when MMP-9, MMP-3, or other MMP expression in tissues is induced by endogenous cytokines, growth factors, or oncoproteins. The titering of TIMPs by se-creted MMPs may occur only in the tissues in very focalized areas such as in regions of active wound repair, or at sites of inflammatory infiltrates, or at the edge of the tumor/stromal interface where locally a high concentration of secreted enzymes can develop. However, such upward shifts in the overall MMP/TIMP balance in the case of MMP-9 may not be sufficient to initiate zymogen activation. MMP-9, being a relatively stable zymogen, appears to require exogenous catalytic intervention. Based on the model system we have described, that exogenous intervention involves the convergence of a highly regulated serine protease cascade with that of a distinctly different but also highly regulated metalloproteinase cascade.