Tissue Inhibitor of Metalloproteinase-2 (TIMP-2) Binds to the Catalytic Domain of the Cell Surface Receptor, Membrane Type 1-Matrix Metalloproteinase 1 (MT1-MMP)*

It has been proposed that tissue inhibitor of metalloproteinase-2 (TIMP-2), in stoichiometric concentrations, serves as an intermediate in progelatinase A activation by binding to activated membrane type 1-matrix metalloproteinase 1 (MT1-MMP) on the plasma membrane. An MT1-MMP-independent cell surface receptor for TIMP-2 has also been postulated. To clarify TIMP-2 binding, we have performed 125I-TIMP-2 binding studies on transfected COS-1 cells and endothelial cells. Specific receptors for TIMP-2 were identified on COS-1 cells transfected with MT1-MMP cDNA, but not on vector-transfected cells. Treatment of MT1-MMP transfected COS-1 cells with a hydroxamic acid inhibitor of MMPs, CT-1746, but not an inactive stereoisomer, CT-1915, produced dose-dependent inhibition of specific TIMP-2 binding comparable with that noted with excess unlabeled TIMP-2. This result suggests that TIMP-2 binds to the zinc catalytic site of MT1-MMP. As demonstrated by the limited competition for binding of C-terminal deleted TIMP-2, the C-terminal domain of TIMP-2 participates in binding to MT1-MMP. Cross-linking studies followed by immunoprecipitation using antibodies to MT1-MMP were employed to identify 125I-TIMP-2·MT1-MMP complexes in MT1-MMP-transfected COS-1 cell membrane extracts. TIMP-2 receptors were also identified on concanavalin A-treated human umbilical vein endothelial cells; inhibition of TIMP-2 binding with CT-1746 was demonstrated.

epithelial and endothelial cells under basal conditions (1,2), has an important role in turnover of basement membrane type IV collagen and other matrix proteins during angiogenesis, tissue remodeling, and repair (3,4). Since progelatinase A secretion is not induced by most cytokines that regulate other MMPs (5), the final activation step appears to exert a more important influence in controlling tissue gelatinase A activity than with other MMPs. Activation of progelatinase A further differs from other MMPs by involving a cell surface activation mechanism (1, 6 -8) that requires the participation of a 63-kDa integral plasma membrane MMP (membrane type-MMP: MT1-MMP) (9). Based on studies in activated tumor cells, Strongin et al. (8) proposed that TIMP-2, contributes to the proteolytic activity by binding to activated MT1-MMP in the plasma membrane; this bimolecular complex then binds progelatinase A. Cleavage of the Asn 67 -Leu 68 peptide bond 2 of progelatinase A then occurs to generate an intermediate form that undergoes autoactivation (8,10,11). Despite the presence of TIMP-2 in this system, MT1-MMP has been implicated in the initial cleavage event. Higher concentrations of TIMP-2 inhibit progelatinase A activation. Debate has arisen as to whether TIMP-2 binds initially to MT1-MMP or an unidentified receptor on the cell surface prior to progelatinase A activation (8,12). Independent of its MMP inhibitory effect, TIMP-2 also functions as a regulator of cell proliferation (4,13,14).
Coexpression of MT1-MMP, gelatinase A, and TIMP-2 cDNA in mesenchymal cells, coinciding with the activation of progelatinase A during embryogenesis, suggests that MT1-MMP functions physiologically to initiate tissue remodeling on the cell surface (15). Expression of MT1-MMP in tumors also suggests a critical role for this protein in cancer dissemination (9,16).
Transfection of cDNAs for MT1-MMP, TIMP-2, and progelatinase A into COS cells that lack an endogenous system for progelatinase A activation has provided a useful model to explore cell surface protein interactions. MT1-MMP-transfected COS-1 cells or isolated plasma membranes, but not pcDNA3 vector-transfected cells or membranes deficient in MT1-MMP, readily activated exogenous or cell secreted progelatinase A (9,17,18). Consistent with the hypothesis of Strongin et al. (8), low level secretion of TIMP-2 by COS-1 cells was associated with MT1-MMP-induced progelatinase A activation; higher concentrations of TIMP-2, however, interfered in this process (18). In contrast to these findings, using radiolabeled progela-tinase A as the ligand, Sato et al. (19) demonstrated by autoradiography that binding of progelatinase A to COS-1 cells did not require the presence of TIMP-2. However, in another report from the same research group, Imai et al. (20) reported that COS-1 cells transfected with mutant MT1-MMP cDNA lacking the transmembrane domain (⌬MT1-MMP) secreted ⌬MT1-MMP in a complex with TIMP-2, which is then able to bind to progelatinase A. Other investigators reported that high concentrations of soluble forms of MT1-MMP (lacking the transmembrane domain) directly activate progelatinase A (21)(22)(23)(24), thereby bypassing the cell surface activation mechanism. TIMP-2 does not participate in this activation, but can function as an inhibitor of this process. Although stimulated tumor cells have been reported to secrete prodomain deleted MT1-MMP in vitro, the physiologic relevance of this mechanism remains to be determined (20).
The purpose of the current study was to: 1) determine whether MT1-MMP itself serves as a surface receptor for TIMP-2 in transfected COS-1 cells and 2) examine the physiologic receptor binding mechanism in human umbilical vein endothelial cells (HUVEC). In addition to examining the binding of radiolabeled TIMP-2 in HUVEC, 125 I-labeled TIMP-2 binding was examined in COS-1 cells transfected with MT1-MMP cDNA or pcDNA3 vector in the presence of synthetic inhibitors of MMPs. The results indicate that the functional catalytic site of MT1-MMP serves as the binding site for TIMP-2 in both MT1-MMP transfected COS-1 cells and HUVEC.
Construction of Plasmids and COS-1 Transfections-MT1-MMP cDNA, isolated from endothelial cells, encoding an open reading frame from amino acid residues methionine 1 to valine 582, was cloned in a pcDNA3 expression vector employing a cytomegalovirus promoter as we have previously described (17,18). DNA sequence was confirmed by extensive restriction analysis and sequence analysis using an automated DNA sequencer (Applied Biosystems PRISM dye terminator cycle sequencing core kit, Perkin-Elmer).
On the day of transfection, cultivated COS-1 cells were washed with phosphate-buffered saline (PBS), pH 7.4, followed by the addition of DMEM containing 10% NuSerum, 300 g/ml DEAE-dextran, 100 M chloroquine, and 1.25 g/ml DNA (17,18). The cells were then incubated for 4 h at 37°C in an atmosphere of 5% CO 2 and 95% air. The cells were washed once with DMEM and incubated for 2 min in 10% Me 2 SO in Ca 2ϩ /Mg 2ϩ -free PBS at room temperature and then washed twice with PBS. Finally, the cells were incubated for 1 day in DMEM containing 10% fetal calf serum. For receptor binding studies, COS-1 cells were propagated to confluence in 24-well dishes (Becton Dickinson, Lincoln Park, NJ) and then switched to serum-free conditions. The strategy for generating a truncated form of recombinant TIMP-2 with deletion of the C-terminal region extending from Cys 128 to Pro 194 (⌬TIMP-2) has been described (27).
Endothelial Cell Cultivation-Endothelial cells were routinely cultivated on gelatin-coated plates in M199 medium supplemented with 20% heat-inactivated fetal calf serum, 10 ng/ml endothelial growth factor (Life Technologies, Inc.), penicillin (100 units/ml), and streptomycin (100 g/ml) in 5% CO 2 at 37°C. For receptor binding studies, HUVEC were propagated in gelatin-coated 24-well dishes and employed in experiments when Ͼ90% confluence was achieved. Concanavalin A (40 g/ml) was added for the final 18 h of HUVEC culture in serum-free M199 or human endothelial serum-free medium.
Immunoblotting and Gelatin Substrate Zymography-Immunoblotting was performed using protein A affinity-purified polyclonal antibodies to human MT1-MMP as described previously (1). Molecular weights were determined using prestained protein standards.
Zymography was performed in 10% polyacrylamide gels that had been cast in the presence of 0.1% gelatin as described previously (1). After electrophoresis, SDS was replaced by Triton X-100, thus renaturing gelatinases, followed by incubation in a Tris-based buffer for 24 h. Gels were stained with Coomassie Brilliant Blue, and gelatinolytic activity was detected as a clear band in the background of uniform staining.
Preparation of Plasma Membrane-enriched Cell Fractions-Cell membranes were prepared from endothelial cells following nitrogen cavitation as described previously (28). The post-nuclear supernatant (770 ϫ g ϫ 10 min) was collected, and heavy organelles were removed by centrifugation at 6,000 ϫ g for 15 min. This supernatant was centrifuged at 100,000 ϫ g for 1 h at 4°C to recover the plasma membrane enriched lighter cell organelles in the pellet. Cell organelles were characterized by electron microscopy as described previously (28). The experimental design included incubating cell organelles with either buffer (to determine spontaneous release of membrane-associated gelatinase) or with 72-kDa recombinant progelatinase A for 18 h at 37°C to determine membrane-induced activation of progelatinase A. After incubating membranes with progelatinase A, conditioned medium was recovered and subjected to gelatin zymography. Protein determinations were made using the bicinchoninic acid reagent as per the manufacturer's instructions (Pierce BCA protein assay reagent).
Cell Surface Binding of TIMP-2-Recombinant TIMP-2 (rTIMP-2) was iodinated to a specific activity of 5.5 ϫ 10 10 dpm/mg by adding 0.25 mCi of Na 125 I to a tube containing 10 g of rTIMP-2 and 100 g of chloramine T. After 5-min incubation at 23°C, 200 g of sodium metabisulfite was added, and 125 I-labeled TIMP-2 in PBS containing 0.1% BSA was separated from free 125 I by chromatography over a G-25 Sephadex column (Pharmacia Biotech. Inc.). In an independent study using the Nanorange protein quantitation kit (Molecular Probes, Eugene, OR), this radiolabeling procedure resulted in Ͼ95% recovery of 125 I-TIMP-2. Biological activity of radiolabeled TIMP-2 was confirmed by measuring the ability of 125 I-labeled TIMP-2 to inhibit the activation of progelatinase A as compared with unlabeled TIMP-2 (demonstrated by zymography). Binding of 125 I-labeled TIMP-2 to cells propagated in 24-well dishes (Corning Costar, Wilmington, MA) to Ͼ90% confluence was performed in duplicate (ϳ10% variation between duplicates) as follows. Cells were washed thoroughly and treated with 3 M glycine buffer in 0.9% saline, pH 3, for 3 min to dissociate preformed receptorligand complexes. Cells were then washed with cold PBS with 0.1% BSA. For equilibrium binding experiments, dilutions of 125 I-labeled TIMP-2 (0.1-16 nM) in PBS-BSA buffer were added to cells in 200 l of serum-fee medium (total volume) in the presence or absence of excess of unlabeled rTIMP-2 or rTIMP-1 at 4 and 22°C. After 30 -240 min of incubation, supernatant fluid was collected and dishes were washed three times with PBS; washes were collected and added to the unbound 125 I-TIMP-2 fraction. Cell monolayers were then lysed in 0.1% SDS in 0.5 M NaOH and collected as the bound fraction. Bound and unbound 125 I were measured by ␥ counting. The residual radioactivity associated with cells in the nonspecific binding experiment (50-fold excess TIMP-2) was subtracted from the total bound fraction (no unlabeled TIMP-2) to give specific binding. Scatchard plot analysis of binding data employed best-fit curves using the Sigma Plot program (Jandel Scientific, San Rafael, CA). In competition experiments, cells were preincubated with hydroxamic acid inhibitors or TIMPs for 30 min prior to the addition of 125 I-TIMP-2.
Cross-linking Experiments-The 100,000 ϫ g plasma membraneenriched fraction from COS-1 was isolated as described above for HU-VEC. Membrane proteins were extracted using 0.25% (final concentration) polyoxyethylene ether (W-1) in 10 mM HEPES buffer, pH 7.5, containing 150 mM KCl and 1 mM CaCl 2 for 1 h at 4°C. Following centrifugation at 13,000 ϫ g for 15 min, extracted proteins were incubated for 3 h at 4°C with 125 I-labeled TIMP-2. Cross-linking was performed in 10 l of total volume containing indicated amounts of radiolabeled ligand, competing ligand, and detergent-extracted membrane proteins in HEPES buffer. The reaction was incubated with 2 mM BS 3 for 1 h at 4°C to allow cross-linking (covalent amide bond formed when the N-hydroxysuccinimide ester conjugation reagent reacts with primary amines); the cross-linking was quenched by the addition of 1 M Tris for 10 min. Affinity-purified rabbit polyclonal antibodies to MT1-MMP (1) were added to the reaction mixture and incubated at 4°C for 16 h to immunoprecipitate MT1-MMP and MT1-MMP-containing complexes. The mixture was then incubated with protein A-coated Sepharose beads (Pharmacia Biotech Inc.) for 2 h with constant rocking at 4°C. The beads were then washed in HEPES buffer five times by centrifugation until release of radiolabeled 125 I was minimal. Beads were then added to SDS-PAGE sample buffer containing ␤-mercaptoethanol. Samples were heated for 2 min in a boiling water bath prior to loading on a 8 -12% gradient gel. The gels were subjected to SDS-PAGE followed by autoradiography.

Characterization of Binding of 125 I-Labeled TIMP-2 to
Transfected COS-1 Cells-The time course for binding of 125 Ilabeled TIMP-2 to MT1-MMP-transfected COS-1 cells showed that equilibrium was reached after 2 h (Fig. 1A). In contrast, 125 I-TIMP-2 binding to pcDNA3 vector-only-transfected COS-1 cells revealed low level binding that peaked at 30 -60 min and decreased thereafter. The concentration dependence of TIMP-2 binding to MT1-MMP-transfected COS-1 cells demonstrated specific and saturable binding characteristics at 22°C (Fig. 2, A  and B); similar 125 I-labeled TIMP-2 binding was demonstrated at 4°C (data not shown). Addition of unlabeled TIMP-2, but not TIMP-1, displaced 125 I-labeled TIMP-2 binding to MT1-MMPtransfected COS-1 cells in a dose-dependent fashion with Ͼ80% decrease in binding at 100-fold molar excess of unlabeled TIMP-2; C-terminal-deleted TIMP-2 (⌬TIMP-2) was Ͻ50-fold as effective as a competitor as wild-type TIMP-2 (Fig. 3). Treatment of MT1-MMP-transfected COS-1 cells with CT-1746, a hydroxamic acid inhibitor of MMPs with specificity for gelatinases in preference to interstitial collagenases, but not the inactive stereoisomer CT-1915, produced dose-dependent inhibition of specific cell binding of 125 I-TIMP-2 comparable with that noted with unlabeled TIMP-2 (Fig. 3). CT-1847, a broad spectrum MMP inhibitor, was also an effective inhibitor of 125 I-labeled TIMP-2 binding to MT1-MMP-transfected COS-1 cells (data not shown).
Scatchard plot analysis of binding data demonstrated the existence of a high affinity binding site with K d ϭ 1.39 nM and 450,000 sites per MT1-MMP-transfected COS-1 cell (Fig. 2B). A mixture of low dose unlabeled TIMP-2 and CT-1847 resulted in a comparable degree of inhibition of 125 I-TIMP-2 binding as compared with expectation with a higher dose of either inhibitor alone (data not shown). This result suggests that both the hydroxamic acid inhibitor and TIMP-2 bind to the same site in the catalytic domain of MT1-MMP. Pretreatment of 125 I-labeled TIMP-2 with recombinant human progelatinase A did not alter the binding characteristics of TIMP-2 on MT-transfected COS-1 cells (data not shown). No specific binding of 125 I-labeled C-terminal-deleted TIMP-2 (⌬TIMP-2) to HUVEC was demonstrable (data not shown). No significant binding of 125 I-TIMP-2 to pcDNA3 vector-transfected COS-1 cells was demonstrated by Scatchard plot analysis; nonspecific binding exceeded specific binding (data not shown).
Cross-linking experiments using 125 I-labeled TIMP-2 and detergent-extracted crude plasma membranes of COS-1 cells transfected with MT1-MMP cDNA demonstrated the presence of three specific products that migrated as broad bands with apparent molecular masses of ϳ44, ϳ80, and ϳ160 kDa and a 22-kDa 125 I-labeled TIMP-2 band (Fig. 4, lane 4). The forma-tion of these cross-linking products is competitively inhibited in the presence of excess unlabeled TIMP-2 as demonstrated in Fig. 4, lane 5. The reaction product at ϳ80 kDa is consistent with a TIMP-2⅐MT1-MMP complex. The ϳ160-kDa 125 I-labeled product is consistent with dimer formation of the 80-kDa complex. Omission of the BS 3 cross-linking reagent or the antibody to TIMP-2 from the reaction mixtures did not lead to formation of the specific products (Fig. 4, lanes 2 and 3). A ϳ44-kDa band (Fig. 4, lane 4) is consistent with the presence of TIMP-2 dimers. The three specific products (44, 80, and 160 kDa) were not detectable in pcDNA3 vector transfected COS-1 cells (Fig.  4, lanes 6 -9). The radiolabeled 22-kDa TIMP-2 band (free) in experiments lacking the BS 3 cross-linking reagent was considerably larger in MT1-MMP cell extracts as compared with Binding of 125 I-TIMP-2 to Endothelial Cells-Specific 125 Ilabeled TIMP-2 binding to concanavalin A pretreated HUVEC revealed similar binding characteristics to those described with MT1-MMP transfected COS-1; binding equilibrium was achieved between 2-3 h (Fig. 1B). Unlabeled TIMP-2, but not TIMP-1, displaced 125 I-labeled TIMP-2 binding to concanavalin A-pretreated HUVEC (Fig. 5A). Scatchard plot analysis of binding data demonstrated the existence of a high affinity binding site with K d ϭ 0.77 nM (Fig. 5B). The number of receptors per endothelial cell was 183,000 per cell. The binding of 125 I-labeled TIMP-2 to untreated HUVEC was considerably less and was not studied in detail. The hydroxamic acid inhibitor of MMPs, CT-1746, was as effective an inhibitor of 125 I-TIMP-2 binding to HUVEC (91% inhibition at 60 nM concentration) as compared with 100% inhibition with 60 nM unlabeled TIMP-2; the stereoisomer, CT-1915, was 100-fold less active than CT-1746. Complex formation between 125 I-TIMP-2 and progelatinase A did not affect the K d or the rate (measured at 30, 60, 120, 180, and 240 min) of 125 I-TIMP-2  6 -9). Lanes 2,4,5,7,8,and 9 were incubated with the cross-linking reagent BS 3 for 1 h at 4°C. Lanes 3-8 were treated with specific polyclonal antibodies to MT1-MMP followed by binding to protein A-Sepharose beads for 16 h. Unlabeled ligand, TIMP-2 (10-fold excess to radiolabeled TIMP-2), was added to the reaction in lanes 5 and 8 for competition. After incubation, the reactions were analyzed by SDS-PAGE (8 -12% gradient gel) under reducing conditions, followed by autoradiography. All lanes are from the same gel.
binding to HUVEC (data not shown). No specific binding of 125 I-labeled C-terminal deleted TIMP-2 to HUVEC was demonstrable (data not shown).
Characterization of MT1-MMP in COS-1 Cells and Endothelial Cells-We have previously used immunoblotting and immunoprecipitation techniques to demonstrate the presence of a 63-kDa MT1-MMP protein band in cell lysates of COS-1 cells transfected with MT1-MMP cDNA; pcDNA3 vector transfected cells failed to display MT1-MMP (17,18).
Using immunoblotting, MT1-MMP was identified in the plasma membrane fraction of untreated HUVEC primarily as a 63-kDa protein; a weak MT1-MMP band was identified in conditioned medium. Following concanavalin A treatment of HUVEC, MT1-MMP was identified in the plasma membrane-enriched fraction as both 63-(proMT1-MMP) and 57-kDa protein bands (presumably activated MT1-MMP); the conditioned medium contained a 53-kDa immunoreactive band, which may represent the activated form of secreted MT1-MMP (Fig. 6B).

DISCUSSION
In this study we have employed radiolabeled recombinant TIMP-2 to demonstrate specific TIMP-2 receptors on MT1-MMP transfected COS-1 cells and human endothelial cells. The absence of specific TIMP-2 binding to pcDNA3 vector transfected COS-1 cells identifies MT1-MMP as the TIMP-2 receptor. The K d of 1.39 and 0.77 nM (derived from Scatchard plot analysis) for the TIMP-2 receptor on COS-1 cells and concanavalin A treated-endothelial cells, respectively, is similar to previous data derived from phorbol 12-myristate 13-acetatetreated wild-type HT-1080 fibrosarcoma cell and MCF-7 breast cancer cell lines, but the number of receptor sites identified per MT1-MMP transfected COS-1 cell (450,000) and per endothelial cell (183,000) are considerably higher than previously reported with cancer cells (25,000 -40,000/cell) (8,30). These differences may be due to our use of a low pH buffer to strip surface receptors of bound ligand prior to adding 125 I-TIMP-2, which would have the effect of increasing the number of avail- To further characterize MT1-MMP as the TIMP-2 receptor, 125 I-TIMP 2 binding to MT1-MMP transfected COS-1 cells was performed in the presence of CT-1746, a hydroxamic acid inhibitor with greater activity against gelatinases. CT-1746, but not the inactive stereoismomer CT-1915, inhibited 125 I-TIMP 2 binding to cells in a dose-dependent fashion that was comparable with that achieved with unlabeled TIMP-2. CT-1847, an MMP inhibitor with broad spectrum activity, likewise readily inhibited 125 I-TIMP 2 binding to cells. These data are consistent with the concept that TIMP-2 binds directly to the catalytic site of MT1-MMP on the cell surface. The fact that both a general and a more specific hydroxamic acid-based inhibitor displayed approximately equivalent activity in blocking TIMP 2 binding to MT1-MMP suggests that the catalytic site of MT1-MMP differs considerably from other well known MMPs. The N-terminal domain of TIMP-2 that consists of the first three disulfide loops and has an OB barrel-like structure (homologous to that seen in proteins of the oligosaccharide/oligonucleotide binding (OB) fold family) (32) has been shown to behave as a fully active inhibitor of several MMPs, including intestitial collagenase (27), MMP-7, stromelysin-1, and gelatinase A (33). These data therefore suggest that the N-terminal domain of TIMP-2 contains the domain interacting with the active catalytic domain of these MMPs. However, in the case of MT-MMP, our data suggest that the C-terminal domain of TIMP-2 positively influences the association between the Nterminal domain of TIMP-2 and the catalytic domain of MT-MMP. Of relevance for this possibility is the observation of Nguyen et al. (34) who demonstrated that the C-terminal domain of TIMP-1 and TIMP-2 act to increase the association constant by binding to the C-terminal domains of gelatinase A or the N-terminal domain of stromelysin-2. Additional studies of the C-terminal domain structures of TIMP-2 and TIMP-1 (32) will be helpful in explaining their differences in inhibiting MT1-MMP. Of note, competition for binding to active collagenase (soluble) has likewise been observed between TIMP-1 and low molecular weight synthetic inhibitors that are directed at the catalytic zinc of collagenase (35).
To characterize the physical interaction between 125 I-labeled TIMP-2 and MT1-MMP-transfected COS-1 cells, cross-linking studies using a water-soluble, homobifunctional reagent were performed on detergent-solubilized plasma membranes. Using a specific antibody to MT1-MMP and protein A-Sepharose beads to isolate antibody complexes, specific products of ϳ80 and ϳ160 kDa were isolated in the immunoconjugates of detergent extracts of MT1-MMP cDNA-transfected cells. The broad bands on SDS-PAGE may be partially due to intramolecular cross-linking (31). The ϳ80-kDa product is consistent with the complex formed by TIMP-2 (22 kDa) and MT1-MMP (60 kDa); the ϳ160-kDa product appears to represent a dimer of the complex (Fig. 4). Additional studies wil be required to characterize dimer formation. In other cross-linking experiments employing extracted membranes from MT1-MMP cDNA-transfected COS-1 cells and nonradiolabeled TIMP-2, antibodies to MT1-MMP were used in Western blotting to document MT1-MMP complex formation with TIMP-2 (data not shown). In cross-linking studies employing phorbol-treated tumor cell extracts, but lacking an immunoconjugation step that we employed, Strongin et al. (8) also identified radiolabeled TIMP-2 complex formation with a 60-kDa membrane-associated activator (presumably MT-MMP), yielding a specific product of ϳ82 kDa.
Whereas MT1-MMP-transfected COS-1 cells activate progelatinase A directly in vitro, endothelial cells require pretreatment with concanavalin A, thrombin, or phorbol esters (1, 2, 29, 36) to activate progelatinase A. However, plasma membranes isolated from untreated endothelial cells activate progelatinase A on coincubation (Fig. 6A); pretreatment of cells with concanavalin A enhanced membrane-induced progelatinase A activation. The explanation for the enhanced progelatinase A-activating function of plasma membranes versus intact cells remains to be determined.
In addition to the 63-kDa proMT1-MMP band noted in HU-VEC plasma membranes, treatment of endothelial cells with concanavalin A resulted in the additional appearance of a 57-kDa MT1-MMP immunoreactive band in the enriched plasma membranes (presumably activated MT1-MMP) and a 53-kDa band in the conditioned medium (possibly a secreted form). Enhanced 125 I-TIMP2 binding to concanavalin A-treated endothelial cells, therefore, may reflect binding to "activated MT1-MMP" or concanavalin A-induced structural alterations of the plasma membrane that enhance receptor function. Based on current limitations in evaluation of membrane bound MT1-MMP, resolution of this issue may be difficult.
The presence of cell receptors for gelatinase A that function independently of TIMP-2 receptors has been disputed (8,30,37,38). Although earlier experiments by Emonard et al. (37) and Ward et al. (38) demonstrated cell surface binding of 125 Iprogelatinase A, these investigators did not consider the participation of TIMP-2 as an intermediary in the binding process. More recently, Emmert-Buck et al. (30) demonstrated that progelatinase A⅐TIMP-2 complexes, but not progelatinase A alone, bound to the surface of tumor cells in a TIMP-2-dependent manner (complex was displaced by unlabeled TIMP-2).
The mechanism of progelatinase A activation following formation of the triplex between MT1-MMP, TIMP-2, and progelatinase A on cell surfaces remains speculative. Based on our data, we propose that the N-terminal domain of TIMP-2 binds to the zinc-containing catalytic site of MT1-MMP, leaving the C-terminal domain of TIMP-2 available for binding to progelatinase A. The dilemma raised by this model is that the catalytic site of MT1-MMP is occupied by TIMP-2 and, therefore, would not be available to attack the complexed progelatinase A molecule. One potential explanation is that an adjacent active and unoccupied MT1-MMP molecule is required to attack the progelatinase A molecule (22). This might explain the biphasic effect of TIMP-2 on progelatinase A activation (8,18).
The role of integrin ␣ v ␤ 3 in progelatinase A binding to cells (39) and its interaction with MT1-MMP and TIMP-2 remains to be clarified, especially in endothelial cells and tumor cells which express high levels of these heterodimers. Finally, although this report has focused on the TIMP-2 receptor role in progelatinase A activation, it remains to be determined whether the receptor property of MT1-MMP functions in the growth regulatory effects of TIMP-2.