Individual Timp Deficiencies Differentially Impact Pro-MMP-2 Activation*

Membrane-type matrix metalloproteinases (MT-MMPs) have emerged as key enzymes in tumor cell biology. The importance of MT1-MMP, in particular, is highlighted by its ability to activate pro-MMP-2 at the cell surface through the formation of a trimolecular complex comprised of MT1-MMP/tissue inhibitor of metalloproteinase-2 (TIMP-2)/pro-MMP-2. TIMPs 1-4 are physiological MMP inhibitors with distinct roles in the regulation of pro-MMP-2 processing. Here, we have shown that individual Timp deficiencies differentially affect MMP-2 processing using primary mouse embryonic fibroblasts (MEFs). Timp-3 deficiency accelerated pro-MMP-2 activation in response to both cytochalasin D and concanavalin A. Exogenous TIMP-2 and N-TIMP-3 inhibited this activation, whereas TIMP-3 containing matrix from wild-type MEFs did not rescue the enhanced MMP-2 activation in Timp-3-/- cells. Increased processing of MMP-2 did not arise from increased expression of MT1-MMP, MT2-MMP, or MT3-MMP or altered expression of TIMP-2 and MMP-2. To test whether increased MMP-2 processing in Timp-3-/- MEFs is dependent on TIMP-2, double deficient Timp-2-/-/-3-/- MEFs were used. In these double deficient cells, the cleavage of pro-MMP-2 to its intermediate form was substantially increased, but the subsequent cleavage of intermediate-MMP-2 to fully active form, although absent in Timp-2-/- MEFs, was detectable with combined Timp-2-/-/-3-/- deficiency. TIMP-4 associates with MMP-2 and MT1-MMP in a manner similar to TIMP-3, but its deletion had no effect on pro-MMP-2 processing. Thus, TIMP-3 provides an inherent regulation over the kinetics of pro-MMP-2 processing, serving at a level distinct from that of TIMP-2 and TIMP-4.

Matrix metalloproteinases (MMPs) 2 are fundamental to biological processes because of their ability to cleave and remodel the extracellular matrix (ECM). MT1-MMP is one of six cell surface membrane-type MMPs (MT-MMP). Its activity and regulation have been widely studied in the context of cell surface MMP activation, cell migration, and inva-sion (1)(2)(3), independently and in conjunction with other cell adhesion molecules (4 -10). A key function of MT1-MMP is to process pro-MMP-2, whose activity also correlates with an invasive propensity in several cancers and is predictive of poor survival (reviewed in Refs. [11][12][13][14]. Understanding the regulation of pro-MMP-2 processing by MT1-MMP and other members of the MT-MMP family is central to defining their role in cancer biology. The classical model for the cell surface activation of MMP-2 is through the formation of a trimolecular complex comprising MT1-MMP, TIMP-2, and pro-MMP-2 (15). The transmembrane MT1-MMP interacts via its N-terminal domain to the N terminus of TIMP-2, forming a "receptor" onto which pro-MMP-2 (72 kDa) binds. Pro-MMP-2 is initially cleaved to its intermediate form (64 kDa) by an adjacent active MT1-MMP. The second stage of MMP-2 processing, resulting in its fully active form (62 kDa), involves an autocatalytic event that requires an MMP-2 molecule in trans (16). It is known that of the six MT-MMPs, MT2-, 3-, 5-, and 6-MMP (17)(18)(19)(20)(21) also have the capacity to activate pro-MMP-2. Alternative mechanisms of MMP-2 processing at the plasma membrane, such as the urokinase plasminogen system (22,23) or TIMP-independent processing involving MT2-MMP, have also been suggested (24). Although the dynamics of the trimolecular complex have been well studied, new insights into its regulation and control are continually being discovered. More recently, cell surface processing of pro-MMP-2 is reported to occur through formation of a trimolecular complex comprised of MT3-MMP, TIMP-3, and pro-MMP-2 (25), although TIMP-3 is known not to form similar complexes with MT1-MMP (26).
TIMPs are the naturally occurring inhibitors of metalloproteinase activity. There are four members of the TIMP family, and each has a specific niche with respect to function. Studies have focused on the dual role of TIMP-2 in regulating the processing of pro-MMP-2. A threshold level of TIMP-2 is required in relation to MT1-MMP to construct the trimolecular complex, which still leaves sufficient MT1-MMP uninhibited to cleave pro-MMP-2. At higher concentrations, TIMP-2 prevents MMP-2 processing by inhibiting all free MT1-MMP (27)(28)(29). The presence of TIMP-2 was initially considered necessary to achieve any form of MMP-2 processing (30,31), but recently Timp-2 Ϫ/Ϫ cells were shown capable of some processing to the intermediate form (32,33). Although TIMP-1 is known not to inhibit MT1-MMP and does not associate as strongly with MMP-2 (26, 34 -36), TIMP-4, the most recently discovered member of the family (37,38), is able to associate with and inhibit both MT1-MMP and MMP-2. It is, however, not able to replace TIMP-2 in trimolecular complex formation with MMP-2 and MT1-MMP (32,39,40). The role of TIMP-3, which can inhibit both MT1-MMP and MMP-2, is currently less well understood with respect to trimolecular complex function (25,26,34,39).
TIMP-3 is unique among the TIMP family in that it is bound to the ECM rather than remaining a freely soluble protein (41,42). Additionally, it has a broader inhibition profile that extends to members of the ADAM (a disintegrin and metalloproteinase domain) and ADAM-TS families, proteases that have the potential to control the bioactivity of many growth factors and cytokines (reviewed in Ref. 43). TIMP-3 is also implicated in the regulation of apoptosis (44 -46). In the Timp-3-deficient mouse we have previously observed greater matrix degradation, indicating increased MMP activity, as well as unscheduled MMP-2 activation (44,47,48). In the present study we used primary mouse embryonic fibroblasts from Timp-deficient mice to determine the physiological roles of individual TIMPs in pro-MMP-2 processing. Timp-3 Ϫ/Ϫ mouse embryonic fibroblasts (MEFs) displayed an accelerated rate of pro-MMP-2 processing, whereas Timp-4 Ϫ/Ϫ MEFs showed no alteration. We further investigated the necessity of TIMP-2 during the activation process in Timp-3 Ϫ/Ϫ cells using double deficient Timp-2 Ϫ/Ϫ / Timp-3 Ϫ/Ϫ MEFs. This study provides a comprehensive parallel assessment of each TIMP in cell surface processing of pro-MMP-2 and highlights a unique regulatory function of TIMP-3.

MATERIALS AND METHODS
Experimental Animals-Timp-3 Ϫ/Ϫ mice were generated as previously described (47) and backcrossed at least six times into either FVB or C57/Bl6 strains. Timp-2 Ϫ/Ϫ and Timp-1 Ϫ/Ϫ mice were generated as previously reported (30,49) and are in the C57/Bl6 background. Timp-4 Ϫ/Ϫ mice 3 generated in Dr. E. Vuorio's laboratory (University of Turku, Finland) were maintained in C57/Bl6 background. Controls included mice with identical backgrounds to the experimental animals. All animals were cared for in accordance with guidelines of the Canadian Council for Animal Care.
MEF Isolation and Maintenance of Cells-Primary MEFs were isolated from embryos at day 13.5-15.5 of gestation. Cells were maintained in DMEM ϩ 10% fetal bovine serum. All experiments were performed using non-immortalized MEF cultures maintained for 4 -6 passages. Cell stimulation experiments were carried out in 24-well culture dishes, and cells were seeded the night prior to the experiment at a density of 1 ϫ 10 5 cells/well in 1 ml of DMEM ϩ 10% fetal bovine serum. The following day the subconfluent monolayers were washed twice with serum-free DMEM prior to treatment with concanavalin A (con A; 10 or 50 g/ml) or cytochalasin D (cyto D; 1 g/ml) (Sigma) in serum-free DMEM supplemented with insulin transferrin, sodium-selenite (Sigma) in a 250-l volume. At a given time point conditioned medium and lysates (prepared in radioimmune precipitation assay buffer containing 50 mM Tris, pH 7.4, 150 mM NaCl, 0.1% SDS, 1% deoxycholate, 1% nonident P40, 1% Triton X-100 ϩ inhibitors) were collected.
TIMP Purification and Inhibition Experiments-Human recombinant TIMP-1 was expressed in Chinese hamster ovary K-1 cells and purified from the conditioned medium (50). Human recombinant TIMP-2 was expressed in 293-EBNA cells (Invitrogen) using the mammalian expression vector pCEP4 (Invitrogen). TIMP-2 was purified from the condition medium as described previously (51). Recombinant N-terminal domain of TIMP-3 (N-TIMP-3) was prepared as described by Kashiwagi et al. (52). Each TIMP was added at 10, 25, 50, 75, or 100 nM 1 h prior to the addition of cyto D. N-TIMP-3 was dissolved in a glycerol-containing buffer (0.125%), and this was included as an exper-imental control (data not shown). Synthetic metalloproteinase inhibitors used included GM6001 (Chemicon) and PD166793 (Pfizer Inc.) (48,53), both at 25 M 1 h prior to the addition of con A or cyto D.
Gelatin Zymography-Cell-conditioned medium and lysates were analyzed for MMP-2 activity by gelatin zymography. Samples were prepared in non-reducing conditions in 5ϫ sample buffer (0.625 M Tris-HCl, 10% glycerol, 2% SDS, 2% Bromphenol Blue and resolved on 8% SDS-PAGE gels containing 0.5 mg/ml of gelatin Type A from porcine skin (Sigma). The gels were washed twice in 2.5% (v/v) Triton X-100 at room temperature before incubation in 100 mM Tris-HCl, pH 7.9, 30 mM CaCl 2 , and 0.02% sodium azide (TAB). Incubation conditions were either overnight at room temperature or at 37°C, depending on gelatinase activity in the samples. The gels were stained with Coomassie Brilliant Blue for 15 min and destained with MeOH/acetic acid to reveal zones of gelatinase activity (54).
Reverse Zymography-Extracellular matrix retrieved from MEF cultures was analyzed for the presence of TIMP-3 by reverse gelatin zymography. After the removal of cells from the culture dish with 10 mM EDTA/EGTA in phosphate-buffered saline, pH 8.0, the underlying matrix was dissolved with non-reducing sample buffer and the sample loaded onto a 10% SDS-PAGE gel containing 1 mg/ml of gelatin and conditioned medium from BHK cells (42,55). The gel was washed once in 2.5% Triton X-100 for 15 min before an overnight incubation at room temperature in TAB containing 2.5% Triton X-100. The following day the buffer was changed to TAB alone and the gel incubated for 16 h at 37°C. The gel was stained with Coomassie Brilliant Blue and destained as above. A parallel 12% SDS-PAGE gel was silver stained to compare loading and position of major protein bands. Staining of proteins was achieved using a Silver Stain Plus kit (Bio-Rad).
Matrix Swap Assay-MEFs were seeded onto 6-well plates at a density of 2.4 ϫ 10 5 cells/ml in 3 ml of DMEM ϩ 10% fetal bovine serum. Cells were left in culture for 7-10 days to allow for matrix deposition before detachment from the underlying matrix with 10 mM EDTA/10 mM EGTA in phosphate-buffered saline, pH 8.0. Fresh cultures of wildtype or Timp-3 Ϫ/Ϫ cells were then seeded onto matrix of the same genotype or "swapped" at a density of 5 ϫ 10 5 cells/well in 3 ml of DMEM ϩ insulin transferrin, sodium-selenite (Sigma). Cells were left to attach for 1.5 h before the monolayers were washed twice with DMEM and 1.5 ml of DMEM ϩ ITS with or without cytochalasin D (1 g/ml). After 4 h conditioned medium was collected and MMP-2 activity analyzed by gelatin zymography.
Quantitative Taqman Real-time PCR-Quantification of expression of MT1-MMP, MT2-MMP, MT3-MMP, TIMP-2, TIMP-3, TIMP-4, and MMP-2 was performed using the Applied Biosystems ABI prism 7700 sequence detection system (Taqman) (15). Briefly, 1 g of total RNA extracted from primary MEF cultures with or without the addition of cytochalasin D (1 g/ml) using SV Total RNA isolation system (Promega) was reverse transcribed using Superscript II (Invitrogen) using random hexamers (Amersham Biosciences). Taqman reactions were carried out in 96-well plates using 0.5% cDNA, 12.5 l of 2ϫ Taqman universal PCR master mix, 100 M probe, and 200 M of each primer and water to a final volume of 25 l. 18S rRNA was used as an endogenous control. Primer sequences for murine MMPs and TIMPs were as described (56).
MT1-MMP Activity Assay-Cell membranes were prepared from seven subconfluent T75 culture flasks after 6 h of culture with or without cyto D (1 g/ml) for each genotype following the protocol described by Ward et al. (57). A 1:250 dilution of each preparation was used to assess the MT1-MMP activity using a Matrix Metalloproteinase-14 Biotrak Activity Assay system (Amersham Biosciences). Experiments were performed according to the manufacturer's instructions.
Statistical Analysis-Error bars are S.E. of the mean. Two-way analysis of variance with post-test and t-tests was performed to assess the significant difference between data sets. *, p Ͻ0.05.

Loss of TIMP-3 Accelerates Processing of Pro-MMP-2-Primary
MEFs were treated with cyto D or con A, which are known to induce pro-MMP-2 processing (57-60). MMP-2 activation was studied over the course of 24 h using SDS-PAGE gelatin zymography of cell lysates and conditioned medium. Comparable pro-MMP-2 levels were present in unstimulated wild-type (WT) and Timp-3 Ϫ/Ϫ MEFs. Over time, both the WT and Timp-3 Ϫ/Ϫ MEFs responded to cyto D and con A stimulation by the appearance of processed forms (intermediate and active) of MMP-2. In the conditioned medium from WT cells, active MMP-2 was first detectable after 3 h ( At the lower concentration of con A (10 g/ml), the process of pro-MMP-2 activation was much slower but still accelerated in Timp3 Ϫ/Ϫ MEFs. Specifically, greater MMP-2 activation was now seen at 24 h (data not shown). With respect to MMP-9, we observed that pro-MMP-9 was induced over 24 h following con A and cyto D treatments; however, MMP-9 levels remained comparable between WT and Timp-3 Ϫ/Ϫ groups.
In the Timp-3 null MEF cultures, pro-MMP-2 processing was clearly accelerated. Comparing the conditioned medium from WT and Timp-3 Ϫ/Ϫ cells, enhanced kinetics of pro-MMP-2 activation was evident. Both the intermediate and fully processed enzyme species were present at higher levels in 3-h conditioned medium after treatment with cyto D, and the robust appearance of the intermediate form was apparent following con A treatment (Fig. 1A, lanes 5 and 6). The processing of pro-MMP-2 to its fully active form following cyto D treatment was more pronounced at 6 h (Fig. 1A, lane 11), and full processing after con A treatment was also evident (Fig. 1A, lane 12). By 24 h similar levels of fully activated MMP-2 were seen in WT and Timp-3-deficient cells (lanes 14 and 15, 16 and 17). Increased MMP-2 activation was not observed in the unstimulated primary Timp-3 Ϫ/Ϫ MEF cultures. This effect of accelerated MMP-2 processing due to the loss of TIMP-3 was mirrored in analyses performed with cell lysates (Fig. 1B). In fact, as early as 1 h post-stimulation, processing of pro-MMP-2 was observed in the Timp-3 Ϫ/Ϫ MEFs lysates, whereas none was apparent in the wildtype cells (data not shown). These data indicate an accelerated rate of MMP-2 processing in the absence of TIMP-3.
Enhanced MMP-2 Processing in the Absence of TIMP-3 Is Inhibited by TIMP-2 and TIMP-3, but Not TIMP-1-We found that active-to-pro ratio of MMP-2 was consistently higher (ϳ1.8-fold) in Timp-3 Ϫ/Ϫ MEFs. To investigate whether increased activation of MMP-2 in Timp-3 Ϫ/Ϫ MEFs involved the same cell surface mechanism, we added recombinant TIMP proteins to MEF cultures ( Fig. 2A). TIMP-1 was not able to inhibit the processing of pro-MMP-2 induced after the addition of cyto D in either the WT or Timp-3 Ϫ/Ϫ cells. It was noted, however, that increasing concentrations of TIMP-1 did slightly decrease the ability of WT cells to process the intermediate form of MMP-2 to the fully active, the stage of MMP-2 processing previously shown to be sensitive to inhibition by TIMP-1 (16). This effect was, however, not seen in the Timp-3 Ϫ/Ϫ cells, and high doses of TIMP-1 up to 100 nM were unable to reduce processing ( Fig. 2A, top panel). The addition of either TIMP-2 (middle panel) or N-TIMP-3 (bottom panel) to both WT and  APRIL 14, 2006 • VOLUME 281 • NUMBER 15

JOURNAL OF BIOLOGICAL CHEMISTRY 10339
Timp-3 Ϫ/Ϫ cells inhibited MMP-2 processing induced by cyto D. However, although the WT cells could be inhibited by the lowest dose (10 nM) of either TIMP-2 or N-TIMP-3, total inhibition of MMP-2 processing in Timp-3 Ϫ/Ϫ cells was only achieved with a much higher concentration (50 nM). These results further strengthen our observations of a more robust and less regulated processing of MMP-2 in the absence of TIMP-3.
To investigate whether the processing we observed in the absence of TIMP-3 is through increased MMP or ADAM activity, we used a broad spectrum metalloproteinase inhibitor, GM6001, as well as an MMPspecific inhibitor, PD166793, at 25 M 1 h prior to con A or cyto D stimulation. The addition of either GM6001 or PD166793 similarly inhibited the processing of pro-MMP-2 in Timp-3 Ϫ/Ϫ cells at 3 and 6 h (Fig. 2B), suggesting the enhanced pro-MMP-2 processing is through an MMP-specific mechanism.
TIMP-3 Present in the Matrix of WT Cells Does Not Reduce MMP-2 Processing in Timp-3 Ϫ/Ϫ MEFs-TIMP-3 is unique among TIMP proteins in that it is secreted and bound to the ECM rather than remaining soluble in the cell milieu (41,42). We investigated the role of matrixassociated TIMP-3 in MMP-2 processing. The presence of functional TIMP-3 in the matrix secreted by WT cells was confirmed by reverse zymography. A band of inhibition at 24 kDa indicated the presence of TIMP-3 (Fig. 3A) along with a weaker higher molecular mass band indicating the glycosylated form of TIMP-3. As expected, these bands were absent in the matrix derived from Timp-3 Ϫ/Ϫ MEFs. A parallel silver stain gel showed equivalent protein levels in all lanes (data not shown). We asked whether TIMP-3 present in the matrix of WT cells would be sufficient to regulate MMP-2 processing and, conversely, whether the absence of TIMP-3 in the matrix would allow accelerated pro-MMP-2 activation by the WT cells. We performed the matrix-cell swap experiments as described under "Materials and Methods." Briefly, MEFs of defined genotypes were allowed to deposit extracellular matrix over 10 days before the removal of the original cell monolayer. A fresh preparation of MEFs was then plated either on their own genotypematched extracellular matrix or swapped. Stimulation with cyto D was performed 1.5 h later for a period of 4 h. TIMP-3-containing matrix deposited by the WT cells was not able to decrease the pro-MMP-2 processing in Timp-3 Ϫ/Ϫ cells following cyto D stimulation (Fig. 3B, lane 8) because the processing was comparable with that of Timp-3 Ϫ/Ϫ cells plated on their own matrix (lane 11). Further, the absence of TIMP-3 in the matrix deposited by the Timp-3 Ϫ/Ϫ cells failed to enhance pro-MMP-2 activation in WT cells (lane 12). Control experiments performed in parallel included MEFs plated on tissue culture dishes devoid of matrix, which again revealed greater processing of pro-MMP-2 in Timp-3 Ϫ/Ϫ cells compared with the WT counterparts (lanes 1-4).
We also tested whether equivalent numbers of cells attached to the underlying extracellular matrices and whether the cells remained adherent during the time frame of this assay. We observed comparable levels of attachment across genotypes, matrices, and treatments (Fig.  3C). Together, these data indicate that WT matrix is unable to rescue the accelerated pro-MMP-2 processing in Timp-3 Ϫ/Ϫ cells.
Expression of the MMP-2 Cell Surface Activation Molecules Remains Unaltered in the Absence of TIMP-3-The classical model proposed for cell surface processing of MMP-2 involves the formation of a trimolecular complex composed of pro-MMP-2, a member of the MT-MMP family, usually MT1-MMP, and TIMP-2. We first determined whether the expression of these molecules was comparable between WT and Timp-3 Ϫ/Ϫ MEFs, both in the resting state and after stimulation with cyto D. Real-time Taqman RT-PCR for MT1-MMP, MMP-2, and TIMP-2 was performed on samples collected at 0, 3, 6, and 24 h poststimulation. Shown as representative are the 6-and 24-h expression profiles of these genes (Fig. 4, A and B). TIMP-2 and MMP-2 expression levels were comparable in WT and Timp-3 Ϫ/Ϫ MEFs before or up to 6 h after cyto D treatment, with a trend of increase in TIMP-2 expression after 24 h of treatment. At 6 h, the expression of MT1-MMP significantly increased upon cyto D treatment in both the WT and Timp-3 Ϫ/Ϫ cells over their respective untreated controls. Similarly, we found no difference in MT1-MMP levels as a function of genotype (Fig. 4A). At 24 h a significant increase in MT1-MMP expression was found only after cyto D treatment in Timp-3 Ϫ/Ϫ cells compared with the untreated control (Fig. 4B). MT2-MMP, which is also implicated in MMP-2 proc-

Timp-3 Deficiency Accelerates Pro-MMP-2 Activation
essing (24,61), was also unaltered by a lack of Timp-3 or treatment with cyto D at 6 h (Fig. 4A). Recently, MT3-MMP was highlighted in the context of MMP-2 processing involving TIMP-3 (25). MT3-MMP expression was also comparable between Timp-3 Ϫ/Ϫ and WT cells at 6 h (Fig. 4A). In fact, the addition of cyto D led to a trend of decreased MT3-MMP expression at 6 and 24 h and of MT2-MMP at 24 h (Fig. 4,  A and B). This trend was similar between Timp3 Ϫ/Ϫ and WT cells.
Next, we determined any corresponding changes in protein levels by Western blotting of cell lysates (Fig. 4C). An increase in MT1-MMP level was apparent 6 h post cyto D treatment in both WT and Timp-3 Ϫ/Ϫ MEFs, consistent with its increased RNA expression, but there was no difference in MT1-MMP as a function of TIMP-3 loss. An HT1080 cell lysate stimulated with con A served as a positive control for the identification of MT1-MMP (Fig. 4C). An enzyme-linked immunosorbent-based assay for MT1-MMP activity has recently become available and was used to determine MT1-MMP activity in cell membranes prepared after 6 h of cyto D treatment (Fig. 4D). As expected, MT1-MMP activity significantly increased in both WT and Timp-3 Ϫ/Ϫ MEFs after cyto D treatment, and the magnitude of increase in MT1-MMP activity was notably higher in Timp-3 Ϫ/Ϫ than in WT cells. For instance, MT1-MMP activity rose by 1.8-fold in Timp-3 Ϫ/Ϫ but 1.5-fold in WT cells. Thus, the accelerated and enhanced pro-MMP-2 activation following the deletion of Timp-3 resulted from increased MT1-MMP activity rather than a significant alteration in expression of MMP-2, MT1-MMP, MT2-MMP, MT3-MMP, or TIMP-2.
TIMP-3 Is an External Regulator of the Trimolecular Complex-TIMP-2 also functions as a critical linker molecule within the trimolecular complex as well as being an inhibitor (27-29, 62, 63). Cells deficient in TIMP-2 were initially reported as defective in their ability to process pro-MMP-2 (30, 31). More recent work has highlighted that, although full MMP-2 processing is impaired, processing to the intermediate form does occur in the absence of TIMP-2 (32,33). Because Timp-3-deficient cells exhibit enhanced MMP-2 processing to the intermediate and fully active forms, we asked whether the role of TIMP-3 in MMP-2 processing is in an inhibitory capacity or whether it can promote MMP-2 processing independently of TIMP-2-dependent formation of the trimolecular complex.
Panels of MEFs deficient in individual (Timp-1 Ϫ/Ϫ , Timp-2 Ϫ/Ϫ , Timp-3 Ϫ/Ϫ ), and double deficient (Timp-2 Ϫ/Ϫ /-3 Ϫ/Ϫ ) TIMPs, along with wild-type controls, were stimulated for 3, 6, and 24 h in culture with con A and cyto D. Both conditioned medium and cell lysates were collected and their MMP-2 status analyzed by gelatin zymography (Fig.  5). Conditioned medium samples from Timp-2 Ϫ/Ϫ cells did not display any processing until 6 h when barely detectable processing to the intermediate form first became evident with cyto D treatment (Fig. 5A, middle panel, lane 8). This processing to the intermediate form was more substantial by 24 h with both cyto D and con A treatment (Fig. 5A,  bottom panel, lanes 8 and 9). Comparison of MMP-2 species between Timp-2 Ϫ/Ϫ and Timp-2 Ϫ/Ϫ /Timp-3 Ϫ/Ϫ MEFs revealed a far greater rate of processing in the latter. Con A induced processing to the intermediate form within 3 h (Fig. 5A, top panel, lane 12), and both inducers resulted in higher levels of intermediate MMP-2 at 6 h in the double deficient cells (middle panel, lanes 11 and 12). There was even the appearance of fully processed MMP-2 in the conditioned medium of Timp-2 Ϫ/Ϫ /Timp-3 Ϫ/Ϫ cells by 24 h after con A stimulation (bottom panel, lane 12). The processing of pro-MMP-2 in the cell lysates further emphasized the observations seen using conditioned medium. Further, the appearance of the fully active form of MMP-2 was more apparent in   1 and 2) and Timp-3 Ϫ/Ϫ (lanes 3 and 4) cells, control (lanes 1 and 3), or treated with cyto D (1 g/ml) (lanes 2  and 4). Both full-length and the 45-kDa form of MT1-MMP are indicated. HT1080 treated with con A is run as a positive control. D, MT1-MMP activity assay. Membrane preparations from WT and Timp-3 Ϫ/Ϫ MEFs were assessed for their MT1-MMP activity using a Biotrak Activity Assay Sytem as described under "Materials and Methods." Open triangles, WT unstimulated; closed triangles, Timp-3 Ϫ/Ϫ unstimulated; open squares, WT cyto D (1 g/ml); closed squares, Timp-3 Ϫ/Ϫ (1 g/ml).

Controls for these experiments included parallel analyses with WT and
Timp-3-deficient MEFs (lanes 1-6), where the pattern of heightened MMP-2 processing in Timp-3 Ϫ/Ϫ compared with WT MEFs was as that described for Fig. 1. The inclusion of Timp-1 Ϫ/Ϫ MEFs further confirmed the lack of TIMP-1 involvement in the processing of MMP-2, with a pattern of activation similar to that seen in WT cells (Fig. 5, A and  B, all panels, lanes 1-3 compared with lanes 13-15).
Of the two original reports on Timp-2-deficient mice, one included a Northern blot depicting higher MMP-2 and MT1-MMP RNA levels (30). Using Taqman RT-PCR, we tested whether these molecules were altered in our single and double deficient cells (Fig. 5C). We observed a trend of increased MMP-2 and MT1-MMP expression in Timp-2 Ϫ/Ϫ cells, although this was not statistically significant. Interestingly, the double deficient Timp-2 Ϫ/Ϫ /Timp3 Ϫ/Ϫ cells showed a trend of decreased expression of these molecules. Altogether, the combined absence of TIMP-2 and TIMP-3 enabled processing of MMP-2 to the intermediate form, but efficient conversion to the fully active form required TIMP-2. This suggests that TIMP-3 functions externally of the trimolecular complex by regulating MT1-MMP activity.

TIMP-4 Deficiency Does Not Alter Pro-MMP-2 Processing-TIMP-4
has properties similar to TIMP-3 with respect to its interactions with MT1-MMP and MMP-2. We therefore hypothesized that a TIMP-4 deficiency would mimic that of TIMP-3 deficiency during pro-MMP-2 processing although it is not matrix bound. After confirmation that WT MEFs express TIMP-4 (Fig. 6A), we stimulated Timp-4 Ϫ/Ϫ MEF cultures with cyto D for 3 or 6 h. Acceleration of pro-MMP-2 processing did not occur in Timp-4 Ϫ/Ϫ MEFs compared with their WT controls (Fig. 6B). Timp-3 Ϫ/Ϫ MEFs with their respective WT controls again showed enhanced MMP-2 processing (Fig. 6B). It is possible that the lack of accelerated MMP-2 activation in Timp-4 Ϫ/Ϫ MEFs arises from compensation by other TIMPs such as TIMP-2 and/or TIMP-3. Taqman RT-PCR revealed no significant difference in the expression of TIMP-2 or TIMP-3 following 6 h of stimulation with cyto D (Fig. 6C).
Modeling the Contribution of TIMPs to Pro-MMP-2 Processing-At the cell surface, MT1-MMP binds TIMP-2, forming a "receptor" complex for pro-MMP-2 docking and subsequent cleavage to intermediate MMP-2 by an adjacent active MT1-MMP (Fig. 7A). The efficiency of this initial stage is in part governed by the concentration of free TIMP-2, TIMP-3, and TIMP-4 in the cell milieu. The second stage of MMP-2 processing to the fully active form is through an autocatalytic event. It is  APRIL 14, 2006 • VOLUME 281 • NUMBER 15 as yet unclear whether this occurs at the cell surface or by soluble MMP-2. Fig. 7B depicts the activation events in the absence of TIMP-2 as proposed by Lafleur et al. (33) whereby pro-MMP-2 is tethered to cell surface independently of MT1-MMP/TIMP-2 and MMP-2 processing is stalled at the intermediate stage. However, the necessity of this cell surface anchor remains open at present. In the absence of TIMP-3 (Fig.  7C), TIMP-2 is available for trimolecular complex formation and we observe accelerated MMP-2 processing through both stages of activation, resulting in fully active form. It likely arises from the removal of TIMP-3, an external inhibitor resulting in more TIMP-free MT1-MMP available for trimolecular complex formation. With a combined TIMP-2/TIMP-3 deficiency (Fig. 7D), MMP-2 processing culminates in a large increase in intermediate MMP-2 and yields a low level of fully active MMP-2. This processing is likely initiated by more TIMP-free MT1-MMP available for generating the intermediate MMP-2, leading to a chance meeting of MMP-2 molecules and their autocatalytic processing to the fully active form. In contrast to Timp-2 and Timp-3, deletion of either Timp-1 or Timp-4 does not affect MMP-2 processing.

DISCUSSION
MT1-MMP has emerged as a key metalloproteinase in cancer progression. This membrane-bound protease has the capacity to degrade ECM proteins and is known for its central role in the cell surface activation of the constitutively expressed soluble MMP, MMP-2. This study has undertaken a comprehensive comparison of pro-MMP-2 activation as a function of individual Timp gene deficiencies. We demonstrated that TIMP-3 provides an important inherent control over the kinetics of cell surface activation of MMP-2 by MT1-MMP, serving an inhibitory function distinct from TIMP-2 and TIMP-4. This effect is consistently seen in independent Timp3 Ϫ/Ϫ mouse chimeras generated from two independent embryonic stem cell clones, multiple independent primary MEF clones, and in different strains of Timp3 Ϫ/Ϫ mice, including C57BL6 and FVB. Timp-3 Ϫ/Ϫ mice have greater MMP-2 activation during physiological and pathological events, such as during mammary gland involution, and in a model of heart disease (44,48). Increased MMP-2 activation at the cell surface may be a key mechanism by which TIMP-3 loss contributes to aberrant ECM remodeling.
The addition of con A or cyto D to cell cultures stimulates MMP-2 processing by increasing MT1-MMP activity that is effectively inhibited by the addition of soluble TIMP-2 or N-TIMP-3 (34,64). Timp-3-deficient primary MEFs show accelerated kinetics of pro-MMP-2 processing, and inhibition requires ϳ5-fold higher level of exogenous TIMP-2 or N-TIMP-3 than the WT controls. This demonstrates a more efficient activation mechanism in the absence of TIMP-3. Soluble TIMP-3 (N-TIMP-3), but not TIMP-3 bound to the ECM, rescues the heightened MMP-2 activation in Timp-3 Ϫ/Ϫ MEFs. The increased level of MMP-2 processing in Timp-3 Ϫ/Ϫ cells is not due to altered cell proliferation or apoptosis compared with WT controls (data not shown). Further, the increased MMP-2 processing in Timp-3 Ϫ/Ϫ MEFs does not arise from a significant alteration in expression of TIMP-2, MMP-2, MT1-MMP, MT2-MMP, or MT3-MMP, which have each been shown to play a role in MMP-2 processing (15,24,25,61). However, 24 h of treatment with cyto D stimulates a trend of increased TIMP-2 expression in both WT and Timp-3 Ϫ/Ϫ MEFs. As a regulator of MMP-2 activation, this increase in TIMP-2 production may underlie the comparable levels of active MMP-2 observed at this time point in the two groups. At the protein level, MT1-MMP increases as a function of the cyto D treatment but not as a function of TIMP-3 deletion. The lack of mouse antibodies to the other above components precluded their assessment at the protein level. Biochemical studies show that TIMP-3 rapidly interacts with MMP-2, although this interaction is weaker than that of TIMP-2 (26) and it regulates pro-MMP-2 processing by inhibiting MT1-MMP (34). There is evidence to show direct TIMP-3 binding with cell surface proteoglycans (26), and it has recently been noted by proteomic screening of MT1-MMP-asscociating proteins that TIMP-3 associates with MT1-MMP at the surface. 4 Altogether, these results demonstrate that TIMP-3 provides an important inherent control over the kinetics of cell surface activation of MMP-2 by MT1-MMP, and we postulate that TIMP-3 closely associated with the cell surface is able to regulate the trimolecular complex.

Timp-3 Deficiency Accelerates Pro-MMP-2 Activation
The loss of TIMP-3 may affect MMP-2 processing through a number of mechanisms. It may simply result in the removal of an inhibitor of both MT1-MMP and MMP-2, thereby increasing their activity and subsequent efficiency of pro-MMP-2 processing. Alternatively, the lack of TIMP-3 may facilitate generation of trimolecular complexes of MT1-MMP with TIMP-2 and pro-MMP-2 by increasing the amount of TIMP-free MT1-MMP at the cell surface. It also remains possible that the lack of TIMP-3 increases the availability of free MT3-MMP, which can activate pro-MMP-2 through TIMP-2 (25). Interestingly, although MT1-MMP expression significantly increases upon cyto D treatment, that of MT3-MMP decreases. Other possibilities such as alterations in MT-MMP trafficking or MMP-2 clearance, although not investigated in this study, should not be ruled out.
Conversion of pro-MMP-2 to its fully activated state is a two-step process via an intermediate form of MMP-2. We used a genetic approach to map the relative roles of TIMP-2 and TIMP-3 in the trimolecular complex. There was no active MMP-2 in cyto D-treated Timp-2 Ϫ/Ϫ cells, but a minimal level existed in the lysates of con A-treated cells. Bigg et al. (32) and Lafleur et al. (33) have reported that TIMP-2 deficiency results in loss of fully activated MMP-2 with the generation of detectable intermediate MMP-2 species. When TIMP-3 deficiency was superimposed on TIMP-2 loss, we observed a pattern of activation similar to TIMP-2, but the kinetics of MMP-2 processing was again accel-erated. There was not only a precipitous generation of the intermediate MMP-2 form, but the fully active MMP-2 species was more pronounced in con A-treated cell lysates (6 and 24 h) as well as conditioned medium (24 h). This specific increase observed after con A stimulation may be due to cross-linking of MMP-2 and MT1-MMP at the cell surface leading to autocatalysis or inhibition of MT1-MMP constitutive endocytosis (65,66). Thus, the extent and pattern of pro-MMP-2 activation in the double TIMP-2/TIMP-3 mutant MEFs reflected features of single deficiencies of TIMP-2 or TIMP-3.
In this study we also examined the effect of TIMP-4 deficiency on pro-MMP-2 processing. TIMP-3 and TIMP-4 interact with MMP-2 and MT1-MMP in a comparable fashion. Both exhort inhibition after binding to the N termini of MMP-2 or MT1-MMP (26,32,34), and each can also interact with the C-terminal domain of MMP-2 (26,32,40). The latter property opens the potential of their participation in trimolecular complex assembly, although neither has been shown to do so (26,32,67,68). Despite the similarities between TIMP-3 and TIMP-4, we observed accelerated pro-MMP-2 activation only in Timp-3 Ϫ/Ϫ and not Timp-4 Ϫ/Ϫ cells. This failure to accelerate pro-MMP-2 processing is not due to a compensatory increase in the levels of TIMP-2 or TIMP-3 in Timp-4 Ϫ/Ϫ cells. This highlights the need to more fully understand the role of these TIMPs beyond biochemical testing, in more biological cell-based systems.