Mutation Analysis of Membrane Type-1 Matrix Metalloproteinase (MT1-MMP) THE ROLE OF THE CYTOPLASMIC TAIL CYS 574 , THE ACTIVE SITE GLU 240 , AND FURIN CLEAVAGE MOTIFS IN OLIGOMERIZATION, PROCESSING, AND SELF-PROTEOLYSIS OF MT1-MMP EXPRESSED IN BREAST CARCINOMA CELLS*

Membrane type-1 matrix metalloproteinase (MT1-MMP) is a key enzyme in the activation pathway of matrix prometalloproteinase-2 (pro-MMP-2). Both activation and autocatalytic maturation of pro-MMP-2 in trans suggest that MT1-MMP should exist as oligomers on the cell surface. To better understand the functions of MT1-MMP, we designed mutants with substitutions in the active site (E240A), the cytoplasmic tail (C574A), and the RRXR furin cleavage motifs (R89A, ARAA, and R89A/ ARAA) of the enzyme. The mutants were expressed in MCF7 breast carcinoma cells that are deficient in both MMP-2 and MT1-MMP. Our results supported the existence of MT1-MMP oligomers and demonstrated that a disulfide bridge involving the Cys 574 of the enzyme’s cytoplasmic tail covalently links MT1-MMP monomers on the MCF7 cell surface. The presence of MT1-MMP oligomers also was shown for the enzyme naturally expressed in HT1080 fibrosarcoma

MT1-MMP 1 (MMP-14) is a member of a large family of zinc endoproteinases, matrixins or matrix metalloproteinases (MMPs) (1,2). There are several structural features such as the modular domain structure and the existence of an N-terminal propeptide domain, a zinc-coordinating active site domain, and a C-terminal hemopexin-like domain that are characteristic for most MMPs (1)(2)(3). A subfamily of membrane type (MT)-MMPs including MT1-MMP is distinguished by a relatively short transmembrane domain and a cytoplasmic tail, which associate these enzymes with discrete regions of the plasma membrane and the intracellular compartment. MT1-MMP expression has been documented in many tumor cell types and strongly implicated in malignant progression (3,4). In addition to its ability to directly cleave certain components of the extracellular matrix (5,6), MT1-MMP initiates the activation pathway of the most widespread MMP, MMP-2, by converting pro-MMP-2 into an activation intermediate that further undergoes autocatalytic conversion to generate the mature enzyme of MMP-2 (7)(8)(9). Structure-function relationships of MT1-MMP (10 -16) and the mechanisms of pro-MMP-2 activation to the mature enzyme (9,(17)(18)(19)(20) are not understood in detail (21)(22)(23). An immediate proximity of at least two molecules of MT1-MMP (an "activator" and a "receptor") on the cell surface is required for in trans activation of MMP-2 to the mature form (17,19,20,24). However, there is no direct biochemical evidence to support the existence of MT1-MMP oligomers on cell surfaces. In addition, mechanisms involved in activation and trafficking of MT1-MMP are not well elucidated and remain controversial (10,(13)(14)(15)(25)(26)(27)(28). Thus, furin, a serine proteinase of the trans-Golgi network, has been earlier assumed to function as a unique activator of MT1-MMP (25). However, evidence is emerging that there could be alternative pathways of MT1-MMP activation (27,28). In this respect, it is not possible to rule out certain autocatalytic steps in MT1-MMP activation such as those involved in the activation pathway of pro-MMP-2 and pro-MMP-9 (8, 29 -31).
To better understand functions of MT1-MMP, we con-structed mutant MT1-MMPs and evaluated cell surface expression of the wild type and mutant enzymes in MCF7 breast carcinoma cells deficient in MT1-MMP and MMP-2. This allowed us to specifically identify the direct effects of MT1-MMP on cell locomotion. Here, we report novel mechanisms that may control dimerization, processing, and self-inactivating proteolysis of MT1-MMP in breast carcinoma cells.
Cell Transfection-Human MCF7 breast carcinoma cells were stably transfected with MT1-MMP-wt, -C574A, -E240A, and -ARAA using LipofectAMINE according to the manufacturer's recommendations (Life Technologies, Inc.). Cell clones resistant to 0.6 -0.8 mg/ml of zeocin were further selected for cell surface MT1-MMP by flow cytometry as described (9,22,33). Briefly, cells were incubated with 5 g/ml control rabbit IgG or rabbit anti-hinge and further with a fluorescein isothiocyanate-conjugated F(abЈ) 2 fragment of goat anti-rabbit IgG (Sigma). Viable cells were analyzed on a FACStar flow cytometer (Becton Dickinson, Mountain View, CA). To avoid any clone-specific effects, transfected cell lines were generated as corresponding pools of positive cell clones (3-5 clones for each cell line). Control cells transfected with the original pcDNA3-zeo plasmid were generated as a pool of zeocin-resistant cells. Transfected cells were routinely grown in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal calf serum (FCS) and 0.2 mg/ml zeocin.
In addition, a series of cells transiently transfected with the control zeo-plasmid, the wild type, and R89A and R89A/ARAA mutant MT1-MMPs were constructed to facilitate the studies involving the role of furin in MT1-MMP processing. For these purposes, MCF7 cells were seeded at 1 ϫ 10 6 cells/well of a six-well plate in DMEM supplemented with 10% FCS. After incubation for 18 h, cells were transfected with the respective recombinant plasmids (2 g each) mixed with Lipo-fectAMINE Plus reagent (Life Technologies, Inc.) according to the manufacturer's recommendations. After incubation for 48 h, the efficiency of transiently transfected cells was analyzed by gelatin zymography and immunocapture to evaluate MMP-2 activation and the levels and processing of MT1-MMP, respectively (see "Activation of MMP-2").
Assay of Gelatinolytic Activity-To measure the proteolytic activity of MMP-2, we used biotin-labeled gelatin as a substrate (34). For these purposes, MT1-MMP-wt, -C574A, -E240A, and -ARAA cells were plated at 2.5 ϫ 10 5 cells/well of a 24-well cluster in 0.5 ml of serum-containing DMEM. After an overnight incubation, cells were washed with serumfree DMEM and then incubated for 1 h with 750 ng of pro-MMP-2 in 0.15 ml of serum-free DMEM to fully saturate the TIMP-2⅐MT1-MMP complexes existing on cell surfaces. Next, cells were extensively washed with DMEM and 0.1% BSA-DMEM to remove soluble unbound pro-MMP-2, and biotinylated gelatin was added to each well (75 pg in 0.5 ml of 0.1% BSA-DMEM). After incubation at 37°C for 4 h, aliquots were taken from each well and mixed with EDTA (final concentration of 10 mM) to stop the reaction. Further, the amounts of degraded gelatin were quantified in each sample as described earlier (34).
Immunocapture of MT1-MMP-Cells were surface-biotinylated for 1 h on ice with 0.1 mg/ml sulfo-N-hydroxysuccinimide-LC-biotin (Pierce). Where indicated, cells were incubated with protease inhibitors for 48 h prior to labeling. Labeled cells were solubilized at 5 ϫ 10 6 Ϫ 1 ϫ 10 7 cells/ml in PBS, pH 7.4, containing 50 mM n-octyl-␤-D-glucopyranoside or 1% Triton X-114, 1 mM CaCl 2 , 1 mM MgCl 2 , 1 mM phenylmethylsulfonyl fluoride, 1 g/ml leupeptin, 1 g/ml pepstatin, and 1 g/ml aprotinin for 1 h on ice. Insoluble material was removed by centrifugation. Next, supernatants were precleared for 2 h on ice with Protein A-agarose beads (Calbiochem). Aliquots (each containing 1 mg of total protein) of precleared supernatants were incubated overnight at 4°C with 1-3 g of anti-hinge and 30 l of a 50% Protein A-agarose slurry. Following washes, the beads were boiled with 2ϫ SDS sample buffer with or without 50 mM DTT for 5 min. Eluted proteins were separated by electrophoresis on 10% acrylamide gels and transferred onto an Immobilon-P membrane (Millipore Corp., Bedford, MA). Bands containing biotin-labeled proteins were visualized by using avidin-horseradish peroxidase (Sigma) and TMB/M (Chemicon) as a substrate.
Flow Cytometry Analysis of MT1-MMP Expression-Fluorescenceactivated cell sorting analyses were performed as previously described (9). All staining procedures were done on ice in Dulbecco's PBS supplemented with 1 mM CaCl 2 , 1 mM MgCl 2 , and 1% BSA (DPBS/BSA), pH 7.2. Cells were stained with 2 g/ml anti-hinge antibody. Further, cells were incubated with a fluorescein isothiocyanate-conjugated F(abЈ) 2 fragment of goat anti-rabbit IgG (Sigma). Population gates were set by using cells incubated with normal rabbit IgG.
Flow Cytometry Analysis of TIMP-2 Binding-Cells were first incubated for 1 h with or without 5 g/ml TIMP-2, washed with DPBS/BSA, and then incubated with 10 g/ml TIMP-2-specific mAb T2-101 for 2 h followed by incubation with fluorescein isothiocyanate-conjugated F(abЈ) 2 fragment of sheep anti-murine IgG (1:100) for 30 min. After removal of unbound antibodies, cells were resuspended in DPBS/BSA supplemented with 3 g/ml propidium iodide (Sigma), and viable cells were analyzed on a FACScan flow cytometer (Becton Dickinson). Population gates were set by using cells incubated with normal murine IgG.
Cell Adhesion-Cell adhesion was performed in the wells of a high binding 96-well plate (Corning Glass) precoated with 1 g/ml collagen type I (Vitrogen 100; Cohesion, Palo Alto, CA) overnight at 4°C. Plates were washed with PBS and blocked for 1 h at 37°C with 1% BSA in DMEM supplemented with 10 mM HEPES, pH 7.2 (DMEM/BSA). Cells were incubated overnight in serum-containing DMEM, detached with enzyme-free buffer (Specialty Media, Lavalette, NJ), washed, and resuspended in DMEM/BSA. Cells were plated at 5 ϫ 10 4 /0.1 ml in DMEM/BSA for 1 h at 37°C. After three washes with DPBS, adherent cells were fixed and stained with Crystal Violet in 10% ethanol. Following washing with DPBS, the incorporated dye was extracted with a 1:1 mixture of 100 mM sodium phosphate and 50% ethanol, pH 4.5, and the absorbance was measured at 540 nm.
Cell Migration in Transwells-The directional migration of cells in Transwells (Costar, Cambridge, MA) was analyzed under serum-free conditions as previously described (9,22,33). The undersurface of a 6.5-mm insert membrane with an 8-m pore size was coated overnight at 4°C with 20 g/ml collagen type I, washed with PBS, and blocked with 1% BSA. Cells were cultured overnight in DMEM plus 10% FCS and then detached with enzyme-free buffer. A total of 7.5 ϫ 10 5 cells were plated in 0.15 ml of AIM-V medium (Life Technologies) per insert. The outer chamber was filled with 0.6 ml of AIM-V medium. Following incubation for 48 h, cells that migrated to the membrane's undersurface were detached with trypsin/EDTA and counted.
In Vitro Cell Invasion-Cell invasion assays were performed in serum-free AIM-V in 6.5-mm Transwells with the 8-m pore size membranes. The undersurface of the Transwell membrane was precoated with collagen type I at 20 g/ml overnight at 4°C. After washing with PBS, 50 l of PBS containing 3 g of Matrigel (Becton Dickinson, Bedford, MA) were dried overnight on the upper surface of the membrane at room temperature. Matrigel was reconstituted in PBS at 37°C for 2 h. Cells were seeded at 3 ϫ 10 5 in 0.1 ml of AIM-V medium into the inner chamber of the Transwell. The outer chamber was filled with 0.5 ml of AIM-V medium. Cells were allowed to invade Matrigel for 44 -48 h at 37°C in a CO 2 incubator. Then the upper surface of the inserts was wiped. To harvest cells that migrated onto the membrane's undersurface, the inserts were incubated in 0.5 ml of trypsin/EDTA. Cells were collected, pelleted, resuspended in trypan blue solution, and counted with a hemocytometer.
Immunofluorescence-To visualize MT1-MMP, cells (4 ϫ 10 4 cells/ well) were plated onto Lab-Tek II chamber slides (Nalge Nunc International, Naperville, IL) precoated with 10 g/ml fibronectin at 4°C overnight. Following incubation for 48 h, cells were fixed for 20 min with 4% paraformaldehyde in PBS and permeabilized with 0.2% Triton X-100 for 5 min. Cells were stained with 10 g/ml rabbit anti-hinge and further with 10 g/ml goat anti-rabbit IgG conjugated with Alexa Fluor 568 (Molecular Probes, Inc., Eugene, OR). Following washing with PBS, the slides were mounted in SlowFade Light Antifade solution (Molecular Probes). Confocal images were collected with differential-interference contrast and epifluorescence optics on a Bio-Rad confocal microscope. The recorded images were processed with Adobe Photoshop software (San Jose, CA). The immunocapture studies revealed that reduced MT1-MMP-wt was represented by the 63-kDa (the proenzyme), 60-kDa (the enzyme), 42-kDa, and 39-kDa protein bands (degradation products; these bands are the most prominent). Under nonreducing conditions, a relatively low amount of the protein material in the 39 -42-kDa region was present, while the significant quantities of several high molecular weight MT1-MMP-wt forms (78 -85 kDa and 120 kDa (the two major forms) and 170 kDa and 220 kDa (the two minor forms)) were observed (Fig. 1A, wt). MT1-MMP-wt was undetectable in the lysates of cells transfected with the original pcDNA3-zeo plasmid (Fig.  1A, zeo). No specific bands were detected in the samples immunocaptured with control rabbit IgG (data not shown).

Dimerization of MT1-MMP on
To further address a question of whether the oligomers of MT1-MMP-wt preexisted on the cell surface or formed in cell lysates, cells expressing MT1-MMP-wt were surface-biotinylated and lysed in buffer containing 10 mM iodoacetamide. Incubation was carried out for 1 h to complete the modification of free cysteines. Residual iodoacetamide was blocked by excess cysteine (100 mM). When MT1-MMP-wt was immunocaptured and analyzed, the pattern of alkylated MT1-MMP was identical to that shown on Fig. 1A. Since alkylation during lysis had no effect, these findings confirmed that the observed MT1-MMP-wt oligomers preexisted on the cell surface.
To verify that the 78 -85-kDa species of MT1-MMP-wt revealed under nonreducing conditions contained the 39-and 42-kDa MT1-MMP-wt monomers, we excised the 78 -85-kDa band from the nonreducing gel and extracted the protein in PBS plus 0.1% SDS. Streptavidin-coated Dynabeads (Dynal, Lake Success, NY) were used to capture the biotin-labeled proteins from the extract. After washings, the captured proteins were eluted with 1% SDS, reduced with DTT, and rerun on the gel followed by Western blotting and developing by avidin-horseradish peroxidase (Fig. 1A, eluate). Two protein forms with apparent molecular masses of 39 and 42 kDa were identified after these procedures. These findings confirmed that the 39-and 42-kDa degradation products of MT1-MMP-wt form dimers and explained the broad width of the 79 -85-kDa MT1-MMP-wt band that might include three combinations of disulfide-linked 39-and 42-kDa species. Accordingly, the 120-kDa MT1-MMP-wt form observed under nonreducing conditions probably corresponds to a dimer consisting of the 60-or 63-kDa MT1-MMP-wt monomers. Low nanogram (or high picogram) amounts of the 120-, 170-, and 220-kDa MT1-MMP-wt forms greatly complicate their direct isolation and analysis similar to that performed with the 78 -85-kDa species.
Pretreatment of cells with Ilomastat, a specific hydroxamate inhibitor of MMPs (32), abrogated both activation and proteolysis of MT1-MMP-wt. Consequently, the 63-kDa band of the MT1-MMP proenzyme was the major species observed in reduced samples. Under nonreducing conditions, treatment with the inhibitor revealed increased levels of the 170-and 220-kDa MT1-MMP-wt bands ( Fig. 1A; wt ϩ Ilomastat). In contrast, if cells were pretreated with the inhibitors of serine, aspartic, and cysteine proteinases (phenylmethylsulfonyl fluoride, leupeptin, pepstatin, and aprotinin) or a competitive inhibitor of furin (50 M decanoyl-Arg-Val-Lys-Arg-chloromethylketone), the pattern of MT1-MMP-wt remained unchanged relative to that of untreated MT1-MMP-wt cells (data not shown).
To evaluate whether the same forms of MT1-MMP exist in another cell type, we analyzed HT1080 fibrosarcoma cells (Fig.  1B). The HT1080 cell line is widely used in MMP studies and known to naturally express substantial levels of MT1-MMP (7-9, 13, 20, 35). To avoid any misinterpretation of the MT1-MMP forms, HT1080 and MCF7 samples were run side-by-side on the same gel (Fig. 1B). Under reducing conditions, the 60-kDa MT1-MMP was the only form of the enzyme observed in HT1080 cells ( Fig. 1B; ϩDTT, HT1080). No 39 -42-kDa proteolyzed forms were observed. This correlates well with earlier reports indicating stabilization of the activated MT1-MMP enzyme by the relatively high levels of TIMP-2 existing in HT1080 cells (14).
Although the relative amount of MT1-MMP in HT1080 cells was lower than in MT1-MMP-transfected MCF7 breast carcinoma, the same 60-kDa (reduced) and 120-kDa (nonreduced) mature enzyme forms were present in both cell lines. These findings suggest that a significant fraction of naturally expressed MT1-MMP may exist on the cell surface as dimer and/or multimers.
Site-directed Mutagenesis of MT1-MMP-The data on transfected MCF7 cells showed that extensive maturation, dimerization, and degradation of cell surface MT1-MMP occurred. To specifically address the processing and dimerization of this proteinase, we designed and expressed mutant MT1-MMPs in MCF7 cells.
The expression plasmids encoding cDNA for mutant MT1-MMPs were constructed using a polymerase chain reactionbased QuickChange mutagenesis system (Stratagene) and cloned into the pcDNA3-zeo plasmid. MCF7 cells were transfected with mutant MT1-MMPs. According to flow cytometry (Fig. 3A) and efficiencies of the respective transfectant cells in MMP-2 activation (Fig. 4) and immunocapture (Figs. 1A and 3B), the levels of MT1-MMP-wt, -ARAA, -R89A, -R89A/ARAA, -E240A, and -C574A expression were highly similar. Further, these results were supported also by immunofluorescence and TIMP-2 binding studies (Fig. 5, A  and B; Table I).  (Fig. 3B, C574A). Evidently, a disulfide bridge covalently links the wild type enzyme monomers on the cell surface, thereby creating stable dimers of MT1-MMP-wt. The mutation of the Cys 574 residue of the cytoplasmic tail abrogated the ability of MT1-MMP-C574A monomers to form this disulfide bridge.
Catalytically Inactive MT1-MMP-E240A Is Incapable of Selfproteolysis-Next, we analyzed molecular forms of the catalytically inactive MT1-MMP-E240A mutant. Immunocapture and the subsequent analysis of the E240A protein showed the 60-kDa mature MT1-MMP as the major band in reduced samples (Fig. 3B, E240A). The existence of the dominant 60-kDa MT1-MMP-E240A protein excludes autocatalytic mechanisms of MT1-MMP activation. Since there is a complete absence of degraded forms (39 -42-kDa reduced, 78 -85-kDa non-reduced), we concluded that the E240A construct is incapable of self-proteolysis ( Fig. 4A; lane 1).
Under nonreducing conditions, the E240A mutant showed relatively significant amounts of the 170-and 220-kDa MT1-MMP-specific bands (Fig. 3B, E240A). Similar high molecular weight forms were also observed in the nonreduced wild type plus Ilomastat samples from MCF7 cells (Fig. 1, A and B) and in the MT1-MMP samples from HT1080 fibrosarcoma cells (Fig. 1B).

Furin Cleavage Is Not Essential for Activation of MT1-MMP in Breast Carcinoma
Cells-Recent controversial studies implicated furin, a serine protease of the trans-Golgi network, in the processing of the latent 63-kDa MT1-MMP proenzyme to the active enzyme by cleaving either the 108 RRKR 111 , the 89 RRPR 92 , or both sequences in the propeptide domain (25)(26)(27)(28)36). To evaluate the effects of furin motif cleavage in the processing and dimerization of MT1-MMP, we constructed and analyzed MT1-MMP-R89A and MT1-MMP-ARAA mutants, each exhibiting a single respective modified furin motif, and the double MT1-MMP-R89A/ARAA mutant with no sites susceptible to furin cleavage. Immunocapture demonstrated that the pattern of MT1-MMP-ARAA, -R89A, and -R89A/ARAA on cell surfaces was highly similar to that of MT1-MMP-wt (Fig.  3B). Specifically, the mutants showed the 60-, 42-, and 39-kDa bands when reduced and the 78 -85-kDa (major) and 120-kDa (minor) MT1-MMP-specific bands under nonreducing conditions. There is no evidence of a 63-kDa MT1-MMP proenzyme. A complete conversion of MT1-MMP-R89A/ARAA to the 60-kDa enzyme demonstrated that there is a furin-independent alternative pathway of MT1-MMP activation in these breast carcinoma cells. Since Ilomastat inhibited the processing of the 63-kDa MT1-MMP-wt to the 60-kDa mature forms in MCF7 cells (Fig. 1A, wt ϩ Ilomastat), a putative pro-MT1-MMP-processing enzyme appears to be a matrixin-like metalloproteinase.
To additionally support our findings, we evaluated the effects of Ilomastat on MT1-MMP expressed in HT1080 cells. The samples of HT1080 and MCF7 cells were run side-by-side to facilitate the comparison of MT1-MMP forms (Fig. 1B). Ilomastat induced the accumulation of the 63-kDa proenzyme in MT1-MMP-wt cells (Fig. 1B, compare ϩDTT, MCF7-wt with MCF7-wt ϩ Ilomastat). This correlated with the presence of higher levels of the 170-and 220-kDa species of MT1-MMP revealed under the nonreducing conditions (Fig. 1B, ϪDTT, MCF7-wt ϩ Ilomastat). In contrast, Ilomastat failed to affect naturally expressed MT1-MMP in HT1080 (Fig. 1B). These   -, MT1-MMP-wt-, MT1-MMP-E240A-, MT1-MMP-ARAA-, and  MT1-MMP-C574A-transfected MCF7 breast carcinoma cells were incubated with and without 5 g/ml TIMP-2. After extensive washing of unbound TIMP-2, cell-associated TIMP-2 was detected by flow cytometry using 10 g/ml TIMP-2-specific mAb T2-101. Mouse IgG was used as negative control. The cytometry data from one of three independent experiments have been presented. MFI, mean fluorescence intensity. results are not surprising, since activation and processing of intrinsic MT1-MMP in HT1080 cells were specifically shown to involve furin (37). However, this does not rule out an existence of the furin-independent pathway(s) of MT1-MMP processing in other cell types (27,28) including breast carcinomas. The furin-independent mechanisms involved in MT1-MMP activation in cancer cells are yet to be elucidated.
To evaluate the mutants in more detail, we assessed the time course of pro-MMP-2 activation by cells expressing the wild type, ARAA, and C574A constructs. Aliquots of medium were withdrawn in 0.5-8 h and analyzed by gelatin zymography. Fig. 4B shows that the wild type, ARAA, and C574A constructs (upper, middle, and bottom panels, respectively) were similarly efficient in activating pro-MMP-2.
To quantitatively confirm that activation of pro-MMP-2 by cells expressing mutant MT1-MMP results in gelatinolytic activity, we employed activity assay using biotinylated gelatin as a substrate (34). Since MCF7 cells do not produce any detectable gelatinolytic activity in serum-free conditions (Fig. 4A), cells were supplemented with exogenous pro-MMP-2. For these purposes, cells were incubated with excess pro-MMP-2 to fully saturate the available MT1-MMP⅐TIMP-2 surface receptors. Next, cells were washed to remove unbound soluble pro-MMP-2 and any traces of the MMP-2 enzyme and free TIMP-2 that might have preexisted in the proenzyme samples. This significantly reduced the background activity and allowed us to follow the activation of pro-MMP-2 associated with the MT1-MMP⅐TIMP-2 surface receptors. Biotin-labeled gelatin was added to cells to examine the gelatinolytic activity of MMP-2 converted into the active enzyme by the MT1-MMP⅐TIMP-2 complexes. The gelatinolytic activity of MMP-2 generated by the cells expressing MT1-MMP-wt, -ARAA, and -C574A (Fig.  4C) correlated well with the results of zymography (Fig. 4B). Thus, the C574A mutant was almost as efficient in generating MMP-2's gelatinolytic activity as the ARAA mutant. As expected, MT1-MMP-E240A failed to demonstrate any gelatinolytic activity (Fig. 4C).
To evaluate whether MT1-MMP mutants were capable of TIMP-2 binding, transfected MCF7 cells were pretreated with excess TIMP-2 followed by staining with anti-TIMP-2 mAb T2-101 and flow cytometry. Without TIMP-2 pretreatment, none of the cells were capable of binding anti-TIMP-2 mAb (Table I). In turn, if wild type, ARAA, and C574A cells were pretreated with TIMP-2, the levels of cell-associated TIMP-2 significantly increased relative to those of mock-transfected cells (Table I). These findings agreed with the results of gelatin zymography (Fig. 4, A and B), activity measurements (Fig. 4C), and immunocapture studies (Fig. 3B), confirming that there were no significant differences in the levels of active MT1-MMP expressed on the surface of wild type, ARAA, and C574A cells. Evidently, ARAA and C574A mutations did not affect TIMP-2 binding. In contrast, the E240A mutation in the enzyme's active site abolished the ability of MT1-MMP to bind TIMP-2. It is clear from the crystal structure of the MT1-MMP⅐TIMP-2 complex (38) and TIMP-1 binding studies with the Glu mutant of ministromelysin-1 (39) that the interaction of TIMPs with active MMPs does not rely on the Glu in the active site. Accordingly, our data suggest that the E240A mutation abolished TIMP-2 binding by significantly perturbing the overall structure of the enzyme's active site. The immunocapture of a 60-kDa form of the E240A mutant (Fig. 3B) suggests that the mutation did not affect the N-terminal processing of MT1-MMP. The recent data of Valtanen et al. (40), who have experimentally documented the proper processing of MT1-MMP-E240A mutant, support our suggestion.
Cell Surface Localization of Mutant MT1-MMP-To analyze the localization of MT1-MMP, cells expressing wild type, ARAA, C574A, and E240A constructs were plated on fibronectin-coated glass slides, fixed, and subjected to immunofluorescence staining with rabbit anti-hinge antibodies followed by fluorescence and confocal microscopy. We specifically employed permeabilized cells in these experiments to identify if there was any difference in both the intracellular and plasma membrane pools of mutant MT1-MMPs as compared with MT1-MMP-wt. A comparison of the phase contrast and fluorescence images indicated that endogenous expression of MT1-MMP in mock-transfected cells was not sufficient to generate any detectable specific fluorescence (Fig. 5A, zeo; upper right panel). Staining of any tested cells with control rabbit IgG was also negative (data not shown). In cells transfected with MT1-MMP-wt and MT1-MMP-C574A, the protein products were mainly localized to the cell surface. Cell localization and distribution across the plasma membrane of MT1-MMP-C574A was similar to that of the wild type enzyme (Fig. 5A). Cells expressing MT1-MMP-ARAA and -E240A exhibited a pattern of MT1-MMP staining similar to that of the wild type or C574A constructs (data not shown). ZX sections of stained cells (Fig.  5B) confirmed the cell surface localization of MT1-MMP in cells expressing the wild type enzyme and the C574A mutant.
However, there was a significant difference in the morphology of cells expressing MT1-MMP-C574A relative to cells expressing the wild type MT1-MMP. Under routine cell culture conditions, MT1-MMP-wt cells plated on plastic were well spread and demonstrated cell protrusions and ruffling, i.e. displaying a motile phenotype. C574A cells remained more round and appeared as cell clusters with smooth edges and almost no ruffling or spreading (Fig. 5C), thereby suggesting a lower migratory potential and indicating alterations in the cytoskeleton.
MT1-MMP-C574A Does Not Support Cell Adhesion, Migration, and Invasion-To analyze the effects of mutant MT1-MMPs on cell locomotion, we evaluated cells expressing the wild type construct and the mutants in a series of adhesion, migration, and invasion assays (Fig. 6). Expression of the wild type enzyme or MT1-MMP-E240A did not affect adhesive characteristics of cells. In contrast, expression of the C574A mutant significantly reduced the adhesive efficiency of cells onto type I collagen (Fig. 6A). Similar results were obtained when fibronectin and vitronectin were used as the substrates for cell attachment (data not shown).
Further, we evaluated the migratory efficiency of cells expressing wild type, E240A, and C574A constructs on collagencoated surfaces. The expression of the wild type enzyme increased collagen-mediated migration of cells at least 2.5-fold compared with that of mock-transfected cells (Fig. 6B). The C574A and the catalytically inactive E240A mutants failed to facilitate cell migration.
To analyze the effects of mutant MT1-MMPs on the ability of cells to invade through basement membranes, we employed the Transwell cell invasion assay. Relative to mock-transfected cells, cell invasion through Matrigel was strongly enhanced by the expression of either MT1-MMP-wt or MT1-MMP-ARAA (Fig. 6C). Significant inhibition of cell invasion by Ilomastat additionally supported a direct role of MT1-MMP in cell locomotion. In contrast, both the C574A and catalytically inactive E240A mutants did not stimulate cell invasion. Since the C574A mutation affected the adhesive efficiencies of the transfected cells, low migration and invasion of the C574A mutant were not surprising. These findings indicate a significant functional role of the cytosolic portion of MT1-MMP in stimulating cell motility. DISCUSSION Given the central role of MT1-MMP in diverse aspects of malignancy (41)(42)(43)(44), the localization of this enzyme to specific cell surface sites such as the invasive front and invadopodia (13,23,(45)(46)(47)(48) can efficiently regulate matrix proteolysis in the vicinity of cell surfaces. MT1-MMP has transmembrane and cytoplasmic domains, which target the enzyme to invasive front (3,13,45,47,49). In addition to its ability to directly degrade the extracellular matrix (3), MT1-MMP initiates activation of MMP-2 and MMP-13 (8,50). These activation mechanisms are not understood in detail (9,21,51,52). The in trans mechanisms of pro-MMP-2 activation implicate at least two molecules of MT1-MMP, a "receptor" molecule in a complex with TIMP-2 and an "activator" TIMP-2-free molecule. Accordingly, these two molecules of MT1-MMP should be co-localized in immediate proximity on the plasma membrane in order to bring together the binding and the activation of pro-MMP-2 (12,17,19,20). Dimerization of MT1-MMP could accomplish this co-localization. However, direct evidence for dimerization of MT1-MMP has been missing.
To Our observations suggest that MT1-MMP is capable of oligomerization on cell surfaces. Homodimerization was most evident for the enzyme's autolytic ectodomain forms. These inactive forms of MT1-MMP, 39 kDa (presumably, starting from Gly 285 ) and 42 kDa (presumably, starting from Ile 256 ) both lacking the zinc-binding catalytic site domain were identified and characterized in previous reports of other groups (14,49). While our manuscript was in preparation, dimerization was demonstrated for MT1-MMP naturally expressed by platelets , and -C574A (5 ϫ 10 4 cells/well) were allowed to adhere for 1 h to the wells of a 96-well plate coated with 1 g/ml collagen type I. Adherent cells were fixed and stained with Crystal Violet in 10% ethanol. The incorporated dye was extracted, and the absorbance was measured at 540 nm. B, mock-transfected cells (zeo) and cells expressing MT1-MMP-wt, -E240A, and -C574A were plated into the Transwells (53). Apparently, the Cys 574 residue of the cytoplasmic tail is involved in an intermolecular disulfide bond linking monomers of the wild type MT1-MMP.
Since the C574A mutant was quite proficient in MMP-2 activation, we suggest that this mutation and, accordingly, the absence of a covalent link between monomers, does not completely abolish dimerization of the enzyme. The existence of self-proteolyzed forms as well as efficacy of the mutant in pro-MMP-2 activation indirectly supports the presence of non-S-S dimers on the surface of cells expressing the MT1-MMP-C574A construct. Association of the homodimer is likely to be initiated by the motif involving the PRXXLYC 574 XRSXXXXV sequence of the cytoplasmic tail. This motif is fully conserved in MT1-, MT2-, and MT3-MMPs while MT5-MMP lacks several essential residues of this motif (54). MT4-and MT6-MMPs are entirely missing the motif (55)(56)(57). It cannot be excluded that protein-disulfide isomerase activity (58) is involved in the mechanisms that facilitate a disulfide bridge formation and stabilization of MT1-MMP dimers.
Further, there is evidence of the extensive self-proteolysis of MT1-MMP-wt, -R89A, -ARAA, -R89A/ARAA, and -C574A in our cell system that is devoid of MMP-2 (22). The catalytically inactive MT1-MMP-E240A protein was incapable of self-proteolysis. In agreement, a hydroxamate inhibitor, Ilomastat, blocked autolytic cleavage of MT1-MMP. Autolysis of MT1-MMP that occurs under deficiency of TIMP-2 (14) is likely to be a mechanism for negative regulation of MT1-MMP. Our observations suggest that the soluble activity of MMP-2 is not a prerequisite for the degradation of MT1-MMP on cell surfaces (11,13,49,59).
Recent studies suggested that furin might be a physiologically relevant activator of MT1-MMP (25)(26)(27)(28). However, evidence is emerging that there could be alternative pathways of MT1-MMP activation (27,28). Our data confirmed the hypothesis that furin cleavage of both putative RRXR motifs of MT1-MMP is not necessary for the processing of MT1-MMP and the subsequent activation of pro-MMP-2 in breast carcinoma cells. In our experiments, MT1-MMP-ARAA, -R89A, -R89A/ARAA, and -wt displayed similar, if not identical pattern in immunocapture and MMP-2 activation studies. Resistance of the double MT1-MMP-R89A/ARAA mutant to furin cleavage did not cause any accumulation of the respective proenzyme in MCF7 cells. However, MCF7 cells accumulated the MT1-MMP proenzyme in the presence of Ilomastat. A putative matrixin-like proteinase involved in activation of MT1-MMP remains to be identified. These findings extend the physiological implications of the recent report that furin-independent pathway of MT1-MMP activation exists in rabbit dermal fibroblasts (27). In addition, we expressed MT1-MMP in furin-deficient LoVo lung carcinoma cells. Our studies correlate well with the observations of Yana and Weiss (36) and indicated that LoVo cells were fully capable of MT1-MMP activation (data not shown). These data support the existence of furin-independent cellular pathways involved in the processing of the full-length membraneanchored MT1-MMP proenzyme.
MT1-MMP-wt, -ARAA, and -C574A were efficient in TIMP-2 binding and, with the exception of MT1-MMP-C574A, facilitated migration and invasion of the respective cells through basement membrane-like matrices. The catalytically inactive E240A construct failed to promote cell locomotion. In agreement, Ilomastat inhibited invasion of cells expressing MT1-MMP-wt. Thus, our results indicate that MT1-MMP is directly involved in cell invasion and migration and support our earlier report that, in functional cooperation with integrin ␣ v ␤ 3 , MT1-MMP facilitated migration of MCF7 cells devoid of MMP-2 (9, 22). In addition, our studies extend the recent observations that MT1-, MT2-, and MT3-MMP confer invasion-incompetent Madin-Darby canine kidney cells with the ability to penetrate collagen type I matrices (43). Hence, the previously underestimated function of MT1-MMP to support cell locomotion appears to be a general phenomenon (44,60).
Intriguingly, the proteolytically active mutant MT1-MMP-C574A failed to stimulate migration and invasion of transfected cells. In contrast to all other MT1-MMP constructs, the expression of C574A also negatively affected the adhesive ability of the respective cells. Poor adhesion of C574A cells may result in their inefficient migration and invasion. Immunofluorescence, flow cytometry, TIMP-2 binding, and MMP-2 activation studies demonstrated that the expression levels of this mutant were similar to those of MT1-MMT-wt. However, MT1-MMP-C574A cells were unable to efficiently accomplish adhesion and locomotion. Similarly, a chimeric MT1-MMP protein containing the interleukin-2 receptor ␣ chain transmembrane and cytoplasmic domains failed to localize to invadopodia and to facilitate invasion of melanoma cells (45). Recent results of Lehti et al. (13), who reported that a truncation of 10 amino acids that included the Cys 574 decreased the invasion activity of melanoma cells by 30%, have pointed out that the middle portion of the cytoplasmic tail had an important role in cell invasion. In contrast, Hotary et al. (43) observed that truncation of the MT1-MMP cytoplasmic domain did not affect the invasive phenotype of Madin-Darby canine kidney cells stimulated with hepatocyte growth factor. However, the assays of Hotary et al. (43) were not strictly quantitative. Alternatively, Urena et al. (15) and Nakahara et al. (45) demonstrated that the cytoplasmic tail is critically involved in trafficking of MT1-MMP to discrete regions of the cell surface. In addition, our most recent finding clearly indicates that expression of either the C574A construct or the MT1-MMP mutant missing the entire cytoplasmic tail does not affect the locomotion of extremely migratory U-251 glioma cells (data not shown). Thus, although our results are not identical to what has been observed previously, we used a significantly different cell system that could account for the apparent disparity in findings.
Apparently, there are two distinct mechanisms that affect cell locomotion and involve MT1-MMP: the first where the proteolytic activity of MT1-MMP facilitates cell motility and the second where the cytoplasmic tail of the enzyme communicates with the putative intracellular components. Thus, the expression of the C574A mutant is likely to modify specifically the interactions of the MT1-MMP's cytoplasmic tail with the intracellular milieu, thereby affecting cell morphology, adhesion, and migration.
Hypothetically, translocations across the cell surfaces in migrating versus stationary cells indicate the direct critical interactions of MT1-MMP with the intracellular milieu (13,23,48,61,62). Since the putative cytoplasmic components that associate with cell surface MT1-MMP would not be biotinylated, they are not seen in our immunocapture experiments. Our most recent results indicate that the peptide derived from the cytoplasmic tail of MT1-MMP is capable of binding specifically with the p32/gC1q-R multifunctional protein (63). This protein may be a compartment-specific partner of MT1-MMP. The p32 is likely to be involved in directional trafficking of MT1-MMP from the Golgi network to the plasma membrane. 2 Further studies are needed to confirm direct interactions of MT1-MMP with the intracellular milieu via the enzyme's cytoplasmic domain.
In summary, we would like to emphasize that the existence of dimers and possibly, higher oligomers of MT1-MMP on cell surfaces correlates well with the mechanisms of pro-MMP-2 activation. Further, our data point to an important function of the cytosolic portion of the MT1-MMP molecule in modulating cell adhesion and locomotion. There is growing evidence that MT1-MMP is a key enzyme involved in cancer cell invasion. Mutant MT1-MMPs characterized in this report may find further applications in structure-function analyses of MT-MMPs and other cancer-related studies.