Plasma Membrane-bound Tissue Inhibitor of Metalloproteinases (TIMP)-2 Specifically Inhibits Matrix Metalloproteinase 2 (Gelatinase A) Activated on the Cell Surface*

The cell-surface activation of pro-matrix metalloproteinase 2 (pro-MMP-2) is considered to be critical for cell migration and invasion. Treatment of human uterine cervical fibroblasts with concanavalin A activates pro-MMP-2 on the cell surface by converting it to the 65-kDa form with a minor form of 45 kDa. However, the 65-kDa MMP-2 was inactivated by tissue inhibitor of metalloproteinases (TIMP)-2 that was bound to the plasma membrane upon concanavalin A treatment. TIMP-2 binds to the plasma membrane through its N-terminal domain by two different modes of interaction as follows: one is sensitive to a hydroxamate (HXM) inhibitor of MMPs and the other is HXM-insensitive. TIMP-2 bound to the membrane in a HXM-insensitive manner, comprising about 40–50% of TIMP-2 on the membrane, is the inhibitor of the cell surface-activated MMP-2. It, however, does not inhibit MMP-3, MMP-9, and the 45-kDa MMP-2 lacking the C-terminal domain. The inhibition of the 65-kDa MMP-2 by TIMP-2 is initiated by the interaction of their C-terminal domains. Subsequently, the MMP-2·TIMP-2 complex is released from the membrane, and the activity of MMP-2 is blocked by TIMP-2. In the presence of collagen types I, II, III, V, or gelatin, the rate of inhibition of the 65-kDa MMP-2 by the membrane-bound TIMP-2 decreased considerably. These results suggest that the pericellular activity of MMP-2 is tightly regulated by membrane-bound TIMP-2 and surrounding extracellular matrix components.

Matrix metalloproteinases (MMPs), 1 also called matrixins (1), are synthesized and secreted from a number of cell types, and they play a central role in degradation of extracellular matrix macromolecules. In excess, MMPs are thought to participate in accelerated breakdown of connective tissue matrix associated with a number of diseases such as arthritis, atherosclerosis, tissue ulcerations, and tumor cell invasion and metastasis (see Refs. 1-4 for review). The activities of matrixins are regulated by endogenous tissue inhibitors of metalloproteinases (TIMPs) and the plasma proteinase inhibitor ␣ 2 -macroglobulin (␣ 2 M). It is postulated that TIMPs regulate MMP activity in the tissue or cell periphery, whereas ␣ 2 M is most likely the regulator in the fluid phase (1)(2)(3). Currently four distinct TIMPs (TIMPs-1-4) have been identified in humans (5)(6)(7)(8)(9). TIMP-1 and TIMP-2 are readily found in the conditioned medium of many cell types, but TIMP-3 is tenaciously bound to the extracellular matrix after synthesis and secretion from the cells (7,8). TIMP-1 and TIMP-2 can form a specific complex with the zymogens of MMP-9 (gelatinase B) and MMP-2 (gelatinase A), respectively, through the C-terminal domain of each molecule (6,10).
MMP-2 digests a number of extracellular macromolecules such as collagen types I, IV, V, VII, and XI, laminin, elastin, proteoglycans, and entactin, as well as gelatins (1-3, 11, 12). Like many other matrixins, MMP-2 is secreted from the cells as an inactive zymogen, pro-MMP-2. Thus, the activation of pro-MMP-2 is one of the critical steps involved in controlling its activity. Although mercurial compounds, such as 4-aminophenylmercuric acetate (APMA) and HgCl 2 , readily activate the zymogen to an active 65-kDa MMP-2, most of the proteinases that activate other matrixins lack an ability to activate pro-MMP-2 (12). A physiological activation of pro-MMP-2 is thought to take place on the cell surface as suggested by experiments with the fibroblasts or neoplastic cells that are treated with concanavalin A (Con A) or phorbol ester (13)(14)(15)(16). More recently, membrane-type MMPs containing a transmembrane domain have been molecularly cloned, and MT1-MMP (17), MT2-MMP (18) and MT3-MMP (19) were shown to activate pro-MMP-2 (17,19,20). It was reported that TIMP-2 was recruited to the cell surface of HT-1080 cells when these cells were treated with a phorbol ester (21,22). Strongin et al. (21) reported that binding of TIMP-2 was mediated through MT1-MMP expressed on the cell surface, and they suggested that the interaction of pro-MMP-2 with the TIMP-2 on the cell surface was a critical step for activation. However, the exact mecha-nism of pro-MMP-2 activation and the fate of the TIMP-2 upon activation of pro-MMP-2 are not clearly understood.
In this report, we have investigated the activation processes of pro-MMP-2 using human uterine cervical fibroblasts treated with ConA. This treatment induced the activation of pro-MMP-2 and accumulation of TIMP-2 on the cell surface. The TIMP-2 bound to the plasma membrane specifically inhibits MMP-2 upon activation by the same membrane. We also report that there are at least two modes of binding of TIMP-2 to the cell surface, i.e. hydroxamate inhibitor of MMP (HXM)-sensitive and -insensitive binding. The TIMP-2 bound to the plasma membrane in an HXM-insensitive manner specifically inhibits MMP-2 upon activation. Our study suggests that the TIMP-2 bound to the membrane plays an important role in controlling the pericellular activity of MMP-2.
Purification of Pro-MMP-2 and TIMP-2-Pro-MMP-2 free from TIMP-2 was purified from the conditioned medium of human uterine cervical fibroblasts by gelatin-Sepharose 4B and, subsequently, by gel permeation chromatography on Sephacryl S-200 as described previously (28).
Purification of Active MMP-3 and MMP-9 -Recombinant pro-MMP-3 was purified from the culture medium of Chinese hamster ovary cells transfected with human MMP-3 cDNA as described by Benbow et al. (29). Purified pro-MMP-3 was then activated with chymotrypsin, and the 45-kDa MMP-3 was isolated according to Ogata et al. (30). To obtain active MMP-9, pro-MMP-9⅐TIMP-1 complex was purified from the conditioned medium of HT-1080 cells according to Ogata et al. (31). The purified complex was activated by a 2 M excess of active MMP-3 at 37°C for 2 h. The active MMP-9 was then separated by passing the sample through anti-TIMP-1 affinity and anti-MMP-3 affinity columns and by gel permeation chromatography on Sephacryl S-200. The concentration of these active MMPs was determined by titrating with a known amount of purified TIMP-1.
Preparation of Plasma Membrane-Human uterine cervical fibroblasts were cultured in DMEM containing 10% fetal calf serum. After confluence, the cells were washed and treated with 50 g/ml ConA in the serum-free DMEM supplemented with 0.2% lactalbumin hydrolysate for 36 h. The cells were then scraped with a rubber policeman and used for the plasma membrane preparation as described previously (14,16) with modifications. Briefly, the cells were suspended in chilled 250 mM sucrose, 100 mM Tris-HCl (pH. 8.0), 0.02% NaN 3 , 2 mM diisopropyl phosphorofluoridate, 10 M E-64, then homogenized on ice, and the whole lysate was centrifuged at 16,000 ϫ g for 20 min at 4°C. The supernatant was then centrifuged at 100,000 ϫ g for 1 h at 4°C. Precipitates were washed with TNC buffer and centrifuged again at 100,000 ϫ g for 1 h at 4°C. Precipitates were then suspended in TNC buffer and kept at Ϫ20°C until used for experiments. The concentration of the membrane was determined by measuring the A 280 of the sample dissolved in 50 mM NaOH taking A 1 cm, 280 nm 1% value as 10. Zymography and Reverse Zymography-Zymography was conducted with SDS-polyacrylamide gel containing gelatin (0.8 mg/ml) (32). The samples were mixed with SDS-PAGE sample buffer without reducing agent and were subjected to electrophoretic analysis at room temperature. Enzymic activity was visualized as negative staining with Coomassie Brilliant Blue R-250.
Reverse zymography was conducted with SDS-polyacrylamide gel containing gelatin and pro-MMP-2 (1.2 g/ml) as described previously (33). The samples were mixed with SDS-PAGE sample buffer without reducing agent and subjected to electrophoresis at room temperature. After electrophoresis, gels were washed with the buffer containing 2.5% Triton X-100 to renature pro-MMP-2 in the gels by removal of SDS. The activity of TIMPs was visualized as inhibition of gelatin lysis by MMP-2; thus, positive staining with Coomassie Brilliant Blue R-250 is observed where TIMPs locate. The intensity of positive staining is linear in the range of 0.1-1 ng of TIMP-2 per lane.
Binding of 125 I-TIMP-2 to Human Uterine Cervical Fibroblasts-Purified TIMP-2 (50 g/ml) was labeled with 125 I Ϫ using IODO-GEN (PIERCE) in TNC buffer according to the manufacturer's instructions. Free 125 I Ϫ was removed by gel filtration. Human uterine cervical fibroblasts were cultured in 96-well plates until confluent in the DMEM containing 10% fetal calf serum. The cells were washed with DMEM supplemented with 0.2% lactalbumin hydrolysate and antibiotics once and treated with 50 g/ml ConA in the same medium for 16 h. Cells were washed with HBSS three times, and different concentrations of 125 I-labeled TIMP-2 or 125 I-labeled N-TIMP-2 were incubated in the presence or absence of a 100-fold excess unlabeled TIMP-2 or N-TIMP-2, respectively, at 4°C for 30 min in the DMEM containing 0.1% (w/v) bovine serum albumin. The cells were washed three times with chilled HBSS, and the radioactivity associated with the cells was measured by a gamma counter.
Enzyme Assays-All enzyme assays were carried out in TNC buffer containing 0.05% Brij 35. MMP-3 activity was measured using S-3 Hcarboxymethylated transferrin as described previously (34). The activity of MMP-2 activated by the plasma membrane was measured using a fluorogenic peptide, Mca-Pro-Leu-Gly-Leu-Dpa-Ala-Arg-NH 2 (23), in the presence of 10 M phosphoramidon to inhibit phosphoramidonsensitive peptidases present in the membrane. The activity of MMP-2 in the mixture was calculated by subtracting the fluorescence generated by membrane alone from the fluorescence generated by the sample mixture. Gelatin containing samples were diluted to make a final gelatin concentration of 1 g/ml in the assay tube; at this concentration gelatin did not affect fluorogenic peptide digestion activity of MMP-2.

Inhibition of MMP-2 by TIMP-2 Bound to the Cell Surface-
Incubation of pro-MMP-2 with the plasma membrane from human uterine cervical fibroblasts treated with ConA activated the zymogen and converted it to 65 kDa with a minor component of 45 kDa as reported previously (14,16). We then examined whether these species were proteolytically active or not by incubating the activated sample with ␣ 2 M prior to zymographic analysis (Fig. 1A). ␣ 2 M can bind and form a large ␣ 2 M-proteinase complex only with an active endopeptidase but not with an inactive enzyme (35). ␣ 2 M⅐MMP complexes run near the top of the gel after SDS-PAGE (27,28). Although a large excess of ␣ 2 M was introduced, the majority of the 65-kDa species failed to bind to ␣ 2 M, suggesting that this form of MMP-2 is not proteolytically active. On the other hand, the 45-kDa MMP-2 bound to ␣ 2 M. These results suggest that the activated 65-kDa MMP-2 is specifically inhibited by endogenous inhibitors that may be present in the plasma membrane. Reverse zymography of the same samples revealed that the membrane preparation contained a readily detectable amount of TIMP-2 but not TIMP-1 (Fig. 1B). However, TIMP-2 was not found in the membrane of unstimulated control cells (see Fig. 3). The failure of the 65-kDa MMP-2 to bind to ␣ 2 M suggests that MMP-2, upon activation, was specifically inhibited by TIMP-2 on the membrane.
Orientation of TIMP-2 on the Plasma Membrane and the Specific Inhibition of MMP-2 by the Membrane-bound TIMP-2-We then examined whether the TIMP-2 in the plasma membrane inhibits APMA-activated MMP-2 and other MMPs. Fig. 2 shows a dose-dependent inhibition of two different preparations of MMP-2 by membranes prepared from ConA-treated cells but not by those from control cells. It is notable, however, that the inhibition of MMP-2 was observed up to about 60 -80%, but the rest of 20 -40% was resistant to inhibition even with an increased amount of membrane. The TIMP-2-resistant activity was due to the presence of the 45-kDa MMP-2 in the APMA-activated MMP-2 (see below). Thus, two different preparations of active MMP-2 gave a slightly different inhibition curve depending on the amount of the 45-kDa MMP-2 generated.
In contrast, TIMP-2 on the membrane failed to inhibit MMP-3 and MMP-9, whereas free TIMP-2 in solution inhibited MMP-3 (Table I). These results suggest that the N-terminal inhibitory domain of TIMP-2 is not readily available to other MMPs when it is bound to the membrane. If so, the reaction of MMP-2 and the membrane-bound TIMP-2 is likely to be triggered by their initial interaction through their unique C-terminal domains. To determine the orientation of TIMP-2 in the membrane, the membrane was dot-blotted on a nitrocellulose filter without denaturation and reacted with monoclonal antibodies raised against the peptide (Asp 30 -Gln 44 ) in the N-terminal domain (clone 68 -6H4) and the C-terminal tail (Tyr 178 -Asp 193 ) of TIMP-2 (clone 67-4H11) (36). As shown in Fig. 3, the ConA-membrane preparation reacted with monoclonal antibody 67-4H11 but not with -6H468. Isolated TIMP-2 reacted with both antibodies, whereas TIMP-2 of the pro-MMP-2⅐TIMP-2 complex reacted only with 68 -6H4. The control membrane did not react with either antibody. These data suggest that TIMP-2 binds to the membrane mainly through its N-terminal domain.
To investigate further the possible interaction of the N-terminal domain of TIMP-2 to the membrane, the binding experiments of full-length TIMP-2 and the C-terminal truncated recombinant TIMP-2 (N-TIMP-2) to the cell surface were conducted. Scatchard plot analysis (data not shown) indicated that both TIMP-2 and N-TIMP-2 bound to ConA-stimulated cells with a similar affinity with K d values of 1.4 and 1.6 nM, respectively. The numbers of the binding sites of two TIMP-2 species were 1.92 ϫ 10 4 /cell and 1.91 ϫ 10 4 /cell, respectively, almost identical to each other.
Release of the MMP-2⅐TIMP-2 Complex and the Free 45-kDa MMP-2 from the Membrane upon Activation-MMP-2, upon activation on the cell surface, was inhibited by TIMP-2 on the membrane. We then examined whether the MMP-2⅐TIMP-2 complex remains on the cell surface or is released from the cell. This was investigated by separating the membrane components from the supernatant by ultracentrifugation at 100,000 ϫ g after incubating the membrane with pro-MMP-2 and analyzing both fractions by zymography and reverse zymography (Fig. 4). Pro-MMP-2 was fully activated by the membrane within 4 h. This treatment also generated the 45-kDa MMP-2. A small portion of the 65-kDa MMP-2 was associated with the membrane at a 1-h time point, but most MMP-2 was released from the membrane after 4 h. The 45-kDa MMP-2 was never found in the membrane fraction. When MMP-2 activity in the supernatant was measured against Mca-Pro-Leu-Gly-Leu-Dpa-Ala-Arg-NH 2 , the enzyme activity correlated with the formation of the 45-kDa MMP-2 with the highest activity at 4 h. The intensity of the 65-kDa form in the supernatant and in the total mixture did not change even after a 20-h incubation. The amount of TIMP-2 in the supernatant also increased in a time-dependent manner, suggesting that TIMP-2 bound to the membrane specifically reacts with the cell surface-activated MMP-2.
Cell Surface-bound TIMP-2 Does Not Inhibit the 45-kDa  3 and 4) or without (lanes 1 and 2) crude plasma membrane (Mb) (final concentration of 2.0 mg/ml) prepared from the ConA-treated human uterine cervical fibroblasts. Cont., pro-MMP-2 alone. The samples were then divided into two portions; one portion was further incubated with ␣ 2 M (100 g/ml) at 37°C for 30 min, and the other was incubated with TNC buffer as indicated. Plasma membrane (Mb) alone was run as control (lanes 5 and 6). The samples were then analyzed by zymography (A) and reverse zymography (B) under nonreducing conditions. MMP-2-Although the activation of pro-MMP-2 by the membrane generates predominantly the 65-kDa form, a small amount of the 45-kDa MMP-2 is also generated, and only the 45-kDa species bound to ␣ 2 M (see Fig. 1). The 45-kDa MMP-2 lacks the C-terminal domain (37), which may be the reason why the membrane-bound TIMP-2 cannot inhibit the 45-kDa MMP-2. To test this further, MMP-2 containing a large portion of 45-kDa MMP-2 was prepared by activating pro-MMP-2 with APMA in the presence of 5% Me 2 SO. These conditions slow down the autodegradation of the activated enzyme, and the 45-kDa MMP-2 can be visualized more easily. When this enzyme preparation was reacted with ␣ 2 M, both 65-and 45-kDa species bound to ␣ 2 M and migrated at the top of the zymogram (Fig. 5). When this enzyme preparation was reacted first with the membrane at 37°C for 30 min then with ␣ 2 M, only the 45-kDa species bound to ␣ 2 M. These data further support that the C-terminal domain of MMP-2 is essential for the interaction and subsequent inhibition by TIMP-2 bound to the membrane.
Characterization of TIMP-2 Binding Modes to the Membrane-Strongin et al. (21) reported that TIMP-2 binds to the active MT1-MMP expressed on the cell surface. If this reaction is through the active site of the enzyme, a peptidyl hydroxamic acid inhibitor of the MMPs (HXM) should compete with the binding of TIMP-2 (38). We thus tested the effect of two synthetic inhibitors, Matlystatin A (25) and GM6001X (26). As the membrane-bound TIMP-2 The 45-kDa active MMP-3 (7 nM) was reacted with various concentrations of the isolated TIMP-2 or the plasma membrane prepared from human uterine fibroblasts treated with ConA at 37°C for 24 h. The plasma membrane at a protein concentration of 0.8 mg/ml contains 7 nM TIMP-2 as judged by reversed zymography. The residual MMP-3 activity was measured using S 3 H-carboxymethylated transferrin (33). MMP-9 (3 nM) was reacted with the plasma membrane (2 mg/ml) at 37°C for 1 h. The residual activity was measured using synthetic peptide substrate Mca-Pro-Leu-Gly-Leu-Dpa-Ala-Arg-NH 2  Pro-MMP-2 (final concentration of 1 g/ml) was mixed with the plasma membrane from ConA-treated fibroblasts (final concentration of 3.5 mg/ml) (ϩMb) and incubated at 37°C for the indicated periods. Portions of the samples were then centrifuged at 100,000 ϫ g for 1 h to separate the supernatant (Sup) and the membranes (Mb). The samples before and after separation were analyzed by zymography (A) and reverse zymography (B). Pro-MMP-2 treated with 1 mM APMA for 1 h at 37°C is shown as a standard. The activity of MMP-2 in the supernatant was measured against Mca-Pro-Leu-Gly-Leu-Dpa-Ala-Arg-NH 2 , and the relative activities are indicated taking the APMAactivated MMP-2 as 100%. Membrane alone exhibited a small amount of MMP-2 activity in zymography but not against the peptide substrate, suggesting that it was inhibited by TIMP-2. shown in Fig. 6, ConA treatment of the cells increased the specific binding of TIMP-2 6.8-fold compared with nontreated control cells. When the binding experiments were performed in the presence of an HXM inhibitor, the binding of TIMP-2 to the cells was inhibited in a dose-dependent manner down to the level of control cells. This suggests that the exogenously added TIMP-2 binds to the active MMPs on the cell surface, presumably to MT1-MMP whose expression is induced upon the ConA treatment of the cells.
The specific binding of TIMP-2 to the ConA-treated cell surface was also examined by reverse zymography. As shown in Fig. 7A, both TIMP-1 and TIMP-2 were found in the culture medium. Upon ConA treatment of the cells the amount of TIMP-2 in the medium, but not that of TIMP-1, decreased, and in turn a significant amount of TIMP-2 was found in the cell fraction. Although 50 M Matlystatin A reduced TIMP-2 binding to the ConA-treated cell to the level of control cells (Fig. 6), reverse zymography revealed that the cellular fraction obtained from the cells cultured in the presence of ConA and Matlystatin A for 30 h retained a considerable amount of TIMP-2 (Fig. 7A). Western blotting and its densitometric analyses for TIMP-2 in the two plasma membrane preparations indicated that about 50% TIMP-2 was associated with the membrane even after culturing in the presence of 10 M GM6001-X. These results indicate that TIMP-2 binds to plasma membrane by two different mechanisms as follows: one is an HXM-sensitive manner and the other is HXM-insensitive.
Both ConA-treated membranes and ConA/GM6001-Xtreated membranes contained a very similar amount of MT1-MMP (see Fig. 8, inset). The blocking of TIMP-2 binding to the plasma membrane by HXM increased the peptidase activity of the membrane against Mca-Pro-Leu-Gly-Leu-Dpa-Ala-Arg-NH 2 about 1.4-fold. This peptidase activity was inhibited by TIMP-2 in a dose-dependent manner, but not by TIMP-1, suggesting the activity is derived from MT1-MMP. Thus, the increase of this peptidase activity is most likely due to an increased amount of active MT1-MMP on the membrane. Nonetheless, the ability of the membrane from the ConA/ GM6001-X-treated cells to activate pro-MMP-2 was approximately 50% that of the ConA membrane (Fig. 8B). These results indicate that an increase of MT1-MMP activity alone does not correlate with the activation of pro-MMP-2.

TIMP-2 Bound to the Membrane in an HXM-insensitive Manner Inhibits the 65-kDa MMP-2-Plasma
membranes were prepared from the ConA-treated cells and the ConA/GM6001-Xtreated cells, and their ability to inhibit MMP-2 was compared. Fig. 9 shows that both membrane preparations inhibited MMP-2 in an identical fashion. This indicates that only the TIMP-2 bound to the membrane in an HXM-insensitive man- Activation of Pro-MMP-2 by a Plasma Membrane in the Presence of a Matrix Protein-An interaction of the activated 65-kDa MMP-2 with membrane-bound TIMP-2 has suggested that MMP-2 may not exhibit full proteolytic activity in vivo when activated on the cell surface. On the other hand, MMP-2 binds to extracellular matrix components such as collagens through the fibronectin type II-like domain (39,40) and to fibronectin and heparin through the C-terminal domain of the enzyme (41). Therefore, we examined whether the reaction rate of MMP-2 with membrane-bound TIMP-2 decreases in the presence of matrix components. Pro-MMP-2 was reacted with the ConA-treated cell membrane in the presence of gelatin, collagen I, or fibronectin for various periods, and MMP-2 activity of the mixture was measured against Mca-Pro-Leu-Gly-Leu-Dpa-Ala-Arg-NH 2 (Fig. 10A). When pro-MMP-2 was reacted with the membrane without matrix components, the activity reached a maximum after a 4-h incubation and then gradually decreased. In the presence of gelatin or collagen I, the activity detected was about twice as much as the control throughout the incubation time. Fibronectin did not show any effect on the activity. Therefore, we concluded that the enhancing effect on the apparent MMP-2 activity is specific for proteins that interact with fibronectin type II-like domains of MMP-2. Similar results were observed when pro-MMP-2 was treated with APMA in the presence of gelatin or collagen I. A higher, stable MMP-2 activity was observed for the first 4 h, whereas APMA alone rapidly inactivated MMP-2 (Fig. 10B). It is notable that 50 -63% of the full activity still remained even after a 20-h incubation at 37°C when gelatin or collagen I was present. Fibronectin was not effective.
When the membrane-activated products were analyzed by zymography, it became evident that the decrease of activity was not due to autodegradation but due to inhibition by TIMP-2. As shown in Fig. 10C, pro-MMP-2 was completely converted to 65 kDa by the membrane within 4 h in all cases in a very similar manner, suggesting that none of the matrix components influenced the activation process. Importantly, although a higher activity was detected in the presence of gelatin or collagen I, there were no significant differences in the intensity of 65-and 45-kDa bands at 1-and 4-h time points in these experiments. This suggests that binding of MMP-2 to gelatin or collagen I slows down the rate of inactivation of MMP-2 by TIMP-2 on the membrane. Similarly, collagens II, III, and V also reduced the rate of TIMP-2-MMP-2 interaction in a similar manner (data not shown). At later time points, compared with controls, a higher intensity of the 45-kDa band was noted in the presence of those extracellular matrix components. Since this species of the enzyme is not inhibited by the membrane-bound TIMP-2, the elevated enzymic activity in the presence of gelatin and collagen I may be in part attributed to this 45-kDa species.
On the other hand, when pro-MMP-2 was activated with APMA in the presence of gelatin or collagen I, the 65-and 45-kDa bands were more intense in zymography, suggesting that these matrix components stabilize the activated forms of MMP-2 and prevent them from autodegradation. Fibronectin (Fig. 10) and bovine serum albumin (data not shown) had no effect on the pro-MMP-2 activation and autolysis by either APMA or the membrane, further supporting that this event was specific to the components that react with the fibronectin type II-like domains of MMP-2 such as gelatin and collagens. DISCUSSION The results presented here demonstrate that fibroblasts treated with ConA accumulate TIMP-2 on the cell surface and that TIMP-2 on the membrane specifically inhibits MMP-2 activated on the cell surface. We also have shown that TIMP-2 binds to the plasma membrane by at least two mechanisms, i.e. an HXM-sensitive and an HXM-insensitive manner, and that the level of bound TIMP-2 by two different interaction modes are similar in the case of human uterine cervical fibroblasts. HXM-sensitive binding suggests that TIMP-2 binds to the active site of MMP, presumably MT1-MMP on the cell surface by interacting through its inhibitory domain. Recently, Zucker et FIG. 8. Increase of Mca-Pro-Leu-Gly-Leu-Dpa-Ala-Arg-NH 2 peptidase activity in the plasma membrane from ConA/HXM-treated cells. A, the plasma membranes were prepared from ConA-treated or ConA/GM6001-X-treated cells as described under "Experimental Procedures," and the peptidase activity was measured against Mca-Pro-Leu-Gly-Leu-Dpa-Ala-Arg-NH 2 . The ability of TIMP-1 or TIMP-2 to inhibit the activity was also tested. Mean values Ϯ S.D. from triplicate samples are shown. A 40-g portion of each membrane preparation was subjected for Western blotting using rabbit anti-(human MT1-MMP) antibody (inset). B, pro-MMP-2 (1 g/ml) was reacted with the above membranes (0.5 mg/ ml) at 37°C for 2 h. The samples were subjected to zymographic analysis under nonreducing conditions. Cont., pro-MMP-2 alone. al. (42) and Butler et al. (43) reported that the binding of TIMP-2 to the membrane was through an interaction between the inhibitory site of TIMP-2 and the catalytic site of MT1-MMP. However, TIMP-2 that is bound to active MT1-MMP is unable to inhibit MMP-2 (Ref. 44 and this work). In this report we have shown that the TIMP-2 bound to the membrane in an HXM-insensitive manner has the ability to interact with and inhibit MMP-2 on the cell surface and regulates pericellular MMP-2 activity. This inhibition is specific for the 65-kDa form of MMP-2. The membrane-anchored TIMP-2 could not inhibit MMP-3 and MMP-9 and most likely other matrixins as well. The reaction of the membrane-bound TIMP-2 and the 65-kDa MMP-2 probably takes place in at least three steps as follows: step 1, the interaction through C-terminal domains of both MMP-2 and TIMP-2; step 2, the dissociation of TIMP-2 from the binding molecule located on the membrane, which exposes the N-terminal domain of TIMP-2 for inhibition; and step 3, the binding of the inhibitory domain of TIMP-2 and the catalytic domain of MMP-2. The specific interaction of the C-terminal domains of MMP-2 and TIMP-2 is critical for this reaction. Thus, the 45-kDa MMP-2 that lacks the C-terminal domain and other matrixins are not inhibited by the membrane-bound TIMP-2.
Whereas the TIMP-2 bound to the plasma membrane is an important regulator of MMP-2 activated on the cell surface, the rate of the reaction slows down considerably in the presence of collagen I or other collagen types but not fibronectin. This protection is due to the interaction of these macromolecules with the fibronectin type II-like domains of MMP-2, since a similar effect was observed with gelatin. Fibronectin-like domains are found in MMP-2 and MMP-9, and they probably allow these enzymes to be located on the matrix before they become activated. Such interactions appear to be important not only for targeting the enzymes to specific locations in the tissue but also for stabilizing the active enzyme. Based on computer molecular modeling, Bá nyai et al. (45) have proposed that fibronectin-like domains in MMP-2 are located near the end of the active site cleft, close to the S 1 Ј pocket of the enzyme. They are therefore thought to be crucial for a proper orientation of the substrate in relation to the catalytic site of the enzyme. Although the three-dimensional structure of MMP-2 is not known, it may be postulated that the binding of collagen or gelatin to these domains may influence the rate of interaction between MMP-2 and TIMP-2. Furthermore, our activation studies of pro-MMP-2 by APMA in the presence of collagen I or gelatin have provided some insights into the stabilization of the activated MMP-2. When pro-MMP-2 is activated by APMA, the activated MMP-2 is rapidly degraded, but in the presence of macromolecules that interact with the fibronectin-like domains of MMP-2 about 50 -60% of MMP-2 remains active even after 20 h at 37°C. In our activation assay system gelatin in the activation mixture was completely degraded into small peptides after 8 h, 2 but the activity of MMP-2 was sustained beyond this point, suggesting that small peptides derived from gelatin are sufficient to stabilize the MMP-2 molecule. Thus, the half-life of MMP-2 in collagen-rich tissues is probably much longer than what has been postulated from the in vitro activation study with APMA.
Our present study does not address whether or not the interaction of pro-MMP-2 and TIMP-2 bound to the cell surface in an HXM-insensitive manner is critical for the activation of pro-MMP-2. However, it is notable that removing TIMP-2 from MT1-MMP by HXM significantly decreased the activation activity of the membrane by 50%, although it increased TIMP-2sensitive peptidase activity that is thought to be MT1-MMP (Fig. 8). These data suggest that an increase in MT1-MMP activity does not necessarily enhance the activation of pro-MMP-2. A certain amount of TIMP-2 bound to MT1-MMP appears to be required for efficient activation of pro-MMP-2. This supposition agrees with the hypothesis by Strongin et al. (21) that the formation of a ternary complex of MT1-MMP⅐TIMP-2⅐pro-MMP-2 on the cell surface is critical for the activation of pro-MMP-2. It is not clear, however, from our experiments if the TIMP-2 bound to the membrane in an HXMinsensitive way participates in the activation of pro-MMP-2. Nonetheless, our studies indicate that TIMP-2 plays an important role in regulating MMP-2 activity on or near the cell surface. The activated 65-kDa MMP-2 can be further converted to the 45-kDa form by removing the C-terminal domain autolytically or by other proteinases, and this reaction is likely to be enhanced in the presence of extracellular matrices as shown in our in vitro experiments. Once this takes place, the membranebound TIMP-2 can no longer inhibit the activity of MMP-2. Olson et al. (46) have recently reported that when active MMP-2 loses its C-terminal domain, TIMP-1 can no longer inhibit the activity, and even TIMP-2 inhibits poorly by increasing the K i value about 38-fold in the solution. The effective regulator of the 45-kDa form would therefore be only ␣ 2 M in this case. Recently, Brooks et al. (46) have reported the Cterminal hemopexin-like domain of MMP-2 can inhibit MMP-2-mediated angiogenesis. They have also shown that at the site of angiogenesis, C-terminal hemopexin-like domains accumulated. Although the presence of the 45-kDa MMP-2 was not reported at that site, it may be speculated that the active 45-kDa species might be present in such tissues rich in collagens as demonstrated by in vitro experiments in this paper. In such tissues even if TIMPs are abundant, cells may utilize the 45-kDa MMP-2 to degrade the surrounding matrix. Taken together, the expression of MMP-2 activity in the periphery of cells is precisely regulated by the level of expression of cellsurface activators, the amount of TIMP-2 bound to the cell surface, and extracellular matrix components.