Complex pattern of membrane type 1 matrix metalloproteinase shedding. Regulation by autocatalytic cells surface inactivation of active enzyme.

Membrane type 1 matrix metalloproteinase (MT1-MMP) is a type I transmembrane MMP shown to play a critical role in normal development and in malignant processes. Emerging evidence indicates that MT1-MMP is regulated by a process of ectodomain shedding. Active MT1-MMP undergoes autocatalytic processing on the cell surface, leading to the formation of an inactive 44-kDa fragment and release of the entire catalytic domain. Analysis of the released MT1-MMP forms in various cell types revealed a complex pattern of shedding involving two major fragments of 50 and 18 kDa and two minor species of 56 and 31-35 kDa. Protease inhibitor studies and a catalytically inactive MT1-MMP mutant revealed both autocatalytic (18 kDa) and non-autocatalytic (56, 50, and 31-35 kDa) shedding mechanisms. Purification and sequencing of the 18-kDa fragment indicated that it extends from Tyr(112) to Ala(255). Structural and sequencing data indicate that shedding of the 18-kDa fragment is initiated at the Gly(284)-Gly(285) site, followed by cleavage between the conserved Ala(255) and Ile(256) residues near the conserved methionine turn, a structural feature of the catalytic domain of all MMPs. Consistently, a recombinant 18-kDa fragment had no catalytic activity and did not bind TIMP-2. Thus, autocatalytic shedding evolved as a specific mechanism to terminate MT1-MMP activity on the cell surface by disrupting enzyme integrity at a vital structural site. In contrast, functional data suggest that the non-autocatalytic shedding generates soluble active MT1-MMP species capable of binding TIMP-2. These studies suggest that ectodomain shedding regulates the pericellular and extracellular activities of MT1-MMP through a delicate balance of active and inactive enzyme-soluble fragments.

Release of the extracellular portion of type I transmembrane proteins, referred to as ectodomain shedding, has been estab-lished as a major regulatory mechanism to control the activity of a variety of membrane-bound proteins on the cell surface (1). Recent evidence suggests that ectodomain shedding is also characteristic of the membrane type matrix metalloproteinases (MT-MMPs), 1 a subfamily of membrane-anchored MMPs by means of a transmembrane domain or a glycosylphosphatidylinositol anchor (2,3). The MT-MMPs are major mediators of proteolytic events on the cell surface, including turnover of extracellular matrix components (4,5), cleavage of various surface adhesion receptors (6 -8), and initiation of zymogen activation cascades (9,10). Uncontrolled MT-MMP activity contributes to abnormal development (11) and is a key determinant in cancer metastasis and tumor angiogenesis (12)(13)(14). To control the extent of pericellular activity, the MT-MMPs are inhibited by the tissue inhibitors of metalloproteinases (TIMPs), a family of natural protein MMP inhibitors. In addition, MT-MMPs have a unique regulatory mechanism in which the active enzyme undergoes a series of processing steps, either autocatalytic (15)(16)(17) or mediated by other proteases (18), that regulate the activity and nature of the enzyme species at the cell surface and at the pericellular space. Previous studies have shown that active MT1-MMP is autocatalytically processed on the cell surface to an inactive membrane-tethered ϳ44-kDa species lacking the entire catalytic domain (17). This processing is inhibited by TIMP-2, TIMP-4, and synthetic MMP inhibitors consistent with being an intermolecular autocatalytic event (19,20). Inhibition of MT1-MMP processing induces accumulation of the active enzyme on the cell surface, and, as a consequence, net MT1-MMP-dependent proteolysis is enhanced. Indeed, we have shown that, under certain conditions, inhibition of MT1-MMP autocatalysis by synthetic MMP inhibitors enhances pro-MMP-2 activation by MT1-MMP in the presence of TIMP-2 (20,21). Thus, although the presence of inhibitors will stabilize MT1-MMP on the cell surface, the absence or reduced levels of inhibitors will facilitate autocatalysis. As a membrane-anchored protein, the autocatalytic processing of active MT1-MMP on the cell surface raises questions as to the fate of the ectodomain and its functional consequences. Accumulating evidence suggests that catalytic domain shedding may represent a general characteristic of several members of the MT-MMP family, and both autocatalytic and non-autocatalytic mechanisms of shedding have been described, e.g. autocatalysis was implicated in the shedding of MT1-MMP in cells transfected to overexpress MT1-MMP (22). Other studies reported shedding of MT1-MMP from a breast carcinoma cell line after treatment with concanavalin A (ConA) (23)(24)(25), which was not inhibited by TIMP-2 (23); therefore, autocatalysis could not be involved. MT5-MMP sheds its catalytic domain in a process that appears to be mediated by a pro-convertase that removes the ectodomain intracellularly (26). Pre-mRNA splicing was reported to be involved in the generation of a soluble form of MT3-MMP, which retained catalytic activity and sensitivity to TIMP-2 inhibition (27). Thus, although different mechanisms of shedding may exist, collectively, these data suggest a unique property of MT-MMPs: the ability to generate soluble fragments by a process of ectodomain shedding, which may possess important functional consequences for pericellular proteolysis in normal and malignant processes. Here we have identified the major soluble forms of MT1-MMP and characterized the major autocatalytic fragment. We demonstrated that the autocatalytic shedding mechanism of MT1-MMP is likely to have evolved to terminate MT1-MMP-dependent proteolysis by hydrolyzing the enzyme at specific and vital sites.
Cell Surface Biotinylation-HT1080 cells in six-well plates were untreated or treated with 100 nM TPA or 10 g/ml ConA in 1 ml of serum-free medium overnight. The cells were rinsed with cold phosphate-buffered saline containing 0.1 mM CaCl 2 and 1 mM MgCl 2 and then biotinylated with 0.5 mg/ml sulfo-NHS-biotin as described (38).
Cloning, Expression, and Isolation of Recombinant MT1-MMP Mutants-A catalytically inactive mutant of MT1-MMP was generated by replacing Glu 240 with Ala (E240A-MT1) using the QuikChange TM sitedirected mutagenesis kit (Stratagene, La Jolla, CA). A cytosolic domain (CD) deletion mutant (⌬CD-MT1) was constructed by introducing a termination codon at Arg 563 by polymerase chain reaction (PCR) using specific primers and wild-type MT1-MMP cDNA as the template. The amplified E240A-MT1 and ⌬CD-MT1 cDNA fragments were cloned into the pTF7EMCV-1 vaccinia expression vector using appropriate restriction sites to generate the respective expression vectors pTF7-E240A-MT1 and pTF7-⌬CD-MT1, as described (28,33). The sequence of the inserts was verified by DNA sequencing. Expression of the MT1-MMP mutants was carried out in BS-C-1 cells by the infection/transfection procedure (28,32,33). Briefly, BS-C-1 cells were grown in 100-mm culture dishes to 80% confluence and infected with 5 plaque-forming units/cell vTF7-3 virus in infection medium for 45 min, as described (32). The cells were washed with phosphate-buffered saline and then transfected with 2 g/dish pTF7-MT1 (wild type MT1-MMP), pTF7-E240A-MT1, or pTF7-⌬CD-MT1 DNA plasmids using Effectene transfection reagent (Qiagen, Valencia, CA), as described by the manufacturer. Control cells were infected but received no plasmid DNA. Four h after transfection, the cells were metabolically labeled (12 h, 37°C) with 100 Ci/ml [ 35 S]methionine. The media were collected, clarified by centrifugation, and concentrated to ϳ0.5 ml followed by immunoprecipitation, as described below.
To express the 21-kDa (Tyr 112 -Gly 284 ) and 18-kDa (Tyr 112 -Ala 255 ) fragments of MT1-MMP, we utilized high fidelity PCR to amplify the respective cDNA fragments, which were cloned into the NdeI and Hin-dIII restriction sites of the pET-24a(ϩ) expression vector (Novagen, Madison, WI). The recombinant plasmid vectors were introduced into recipient Escherichia coli BL21(DE3) by transformation. For protein expression, 5 ml of the bacterial cultures were induced overnight at 37°C with 0.4 mM isopropyl-␤-D-galactopyranoside. Cells were pelleted by centrifugation, resuspended in 0.5 ml of collagenase buffer, and sonicated. Inclusion bodies were collected by centrifugation and dissolved in collagenase buffer containing 8 M urea. The solubilized proteins were diluted 10-fold in collagenase buffer supplemented with 50% glycerol and dialyzed overnight against collagenase buffer with 10% glycerol. The MT1-MMP fragments were resolved by 15% SDS-PAGE and stained with Coomassie Blue. The same samples were also analyzed for enzymatic activity and ability to bind TIMP-2 as described below.
Immunoaffinity Purification and Microsequencing of the 18-kDa MT1-MMP Fragment-BS-C-1 cells in 150-mm tissue culture dishes were infected to express MT1-MMP as described (17). Three h after infection, the cells were rinsed with serum-free media and incubated overnight with 15 ml/dish serum-free DMEM. The media (ϳ300 ml) were collected, clarified by centrifugation, and concentrated ϳ15-20fold with Centricon Plus-80. The concentrated medium was incubated (12 h, 4°C) with pAb 160/Protein A-agarose beads to immunoprecipitate the MT1-MMP forms. The beads were washed twice with serumfree DMEM and twice with HNTG buffer (50 mM Tris/HCl, pH 7.5, 150 mM NaCl, 0.1% Nonidet P-40, and 10% glycerol). The immunoprecipitated proteins were eluted with reducing sample buffer, boiled and subjected to 15% SDS-PAGE, transferred to a polyvinylidene difluoride membrane, and stained with Coomassie Blue R-250. As reference, an aliquot of the eluate was also subjected to immunoblot analysis using pAb 160. The Coomassie Blue-stained 18-kDa protein was cut out from the polyvinylidene difluoride membrane and was submitted for N-terminal microsequencing by Edman-based chemistry to ProSeq (Boxford, MA). The amino acid sequence of the 18-kDa protein was also determined by mass spectrometry. To this end, the 18-kDa form bound to pAb 160/Protein A-agarose beads was eluted with 100 mM glycine/ HCl, pH 2.5, and the collected fractions were neutralized with 1.88 M Tris, pH 8.8. An aliquot of each fraction was subjected to immunoblot analysis using pAb 160 to identify the 18-kDa protein. The fractions with the 18-kDa fragment were concentrated and subjected to 4 -20% SDS-PAGE followed by Coomassie Blue R-250 staining. The protein was cut out from the gel and sent to the Harvard Microchemistry Facility for sequence analysis by microcapillary reverse-phase high performance liquid chromatography nano-electrospray tandem mass spectrometry on a Finnigan LCQ DECA quadrupole ion trap mass spectrometer.
Computational Modeling-A full computational model (minus the hinge region) for the ectodomain of pro-MT1-MMP was developed for the studies of the various aspects of the biochemistry of this enzyme. The primary sequence of MT1-MMP was obtained from Swiss-Prot data bank (code MM14_HUMAN). An initial model of MT1-MMP was generated by homology modeling with the aid of COMPOSER software, implemented in the SYBYL package version 6.7, and it was further refined with energy minimization procedures. The catalytic and propeptide domains were constructed based on the structure of stromelysin-1 (Protein Data Bank identification code 1slm) (41) following similar procedures that were used for construction of the hemopexin-like domain (42). Individual domains of MT1-MMP were thus constructed using homology modeling and three-dimensional structure alignment, except for the catalytic domain, which was based on the published x-ray structure of the catalytic domain of MT1-MMP (Protein Data Bank identification code 1bqq) (43). The hinge region of MT1-MMP was not modeled because of the lack of any homologous protein that could serve as a three-dimensional template. Energy minimization of the complete MT1-MMP complex was carried out using the SANDER module of the AMBER 5.0 suite of programs (44). The force field of Cornell et al. (45) was used to model the enzyme. The parameters for bonds, angles, and van der Waals interactions involving zinc atoms were taken from Massova et al. (46). The enzyme was immersed in a 97 ϫ 96 ϫ 98-Å 3 box of TIP3P-water (47). Water molecules present in the x-ray crystallographic structure were retained in the model. A 10-Å cutoff was applied to the model, and the nonbonded list was updated every 50 cycles. A total of 20,000 energy minimization cycles were carried out, which consisted of 300 steepest descent steps, followed by conjugate gradient minimization.
Enzyme Kinetic and Inhibition Studies-All enzymatic assays were carried out using the fluorescence substrate MOCAcPLGLA 2 pr(Dnp)A-RNH 2 (Peptides International, Louisville, KY) in a buffer consisting of 50 mM HEPES, pH 7.5, 150 mM NaCl, 5 mM CaCl 2 , 0.01% Brij-35, and 1% Me 2 SO (buffer R). Substrate hydrolysis was monitored with a Photon Technology International spectrofluorometer at excitation and emission wavelengths of 328 and 393 nm, respectively. The synthetic peptide MOCAcPLG (Peptides International) was used to calibrate the assays as described by Knight (48). Concentrations of MT1-MMP cat (Tyr 112 -Gly 288 ) and that of the recombinant 21-kDa (Tyr 112 -Gly 284 ) fragment were determined by titration with TIMP-2. The TIMP-2 concentration was determined using a molar extinction coefficient of 39,600 M Ϫ1 cm Ϫ1 (49). Concentration of the 18-kDa (Tyr 112 -Ala 255 ) fragment was determined by the BCA protein assay (Pierce) relative to a calibration curve established with the recombinant 21-kDa (Tyr 112 -Gly 284 ) fragment as a standard. The kinetic parameters k cat and K m for the reaction of the recombinant 18-kDa (Tyr 112 -Ala 255 ) and 21-kDa (Tyr 112 -Gly 284 ) MT1-MMP species with the fluorogenic substrate were determined by computer fitting the substrate concentration dependence of the initial rates of substrate hydrolysis to the Michaelis-Menten equation using the program Scientist (MicroMath, Salt Lake City, UT). The fluorogenic substrate concentration was varied between 0.1 and 11 M, where the extent of trivial quenching of the substrate is insignificant. Inhibition studies were carried out as previously described (20). The determination of k off was attempted from the enzyme activity recovered after dilution of a pre-formed enzyme-inhibitor complex. The dissociation of the MT1-MMP⅐TIMP-2 complex, however, was too slow for the direct analysis of the k off parameter for TIMP-2. The k off value was estimated based on a 10-fold difference observed between the slopes of the linear portions of the dissociation curves for the complexes of MT1-MMP cat with a C-terminal deletion TIMP-2 mutant (⌬CTD-TIMP-2) (steady state rate) and wild type TIMP-2, as previously described (20).
MT1-MMP-TIMP-2 Interactions-Binding of soluble MT1-MMP forms to TIMP-2 was examined using various approaches. (i) Serumfree 35 S-labeled media (1 ml) from BS-C-1 cells expressing or not MT1-MMP were incubated (4 h, 4°C) with or without 100 ng of either recombinant TIMP-2 or TIMP-1. The samples that received TIMP-2 or TIMP-1 were immunoprecipitated with anti-TIMP-2 or anti-TIMP-1 antibodies and Protein G-agarose beads. The samples without TIMP addition were immunoprecipitated with anti-MT1-MMP antibodies and Protein A-agarose beads. The immunoprecipitates were resolved by reducing 15% SDS-PAGE followed by autoradiography. (ii) Conditioned medium of BS-C-1 cells infected to express MT1-MMP in the presence of 1 M marimastat, to induce the appearance of the 31-35-kDa species, was subjected to TIMP-2-affinity binding using immobilized TIMP-2 on an Affi-Gel 10 matrix, prepared as previously described (17). Briefly, the medium was concentrated (ϳ80-fold), and ϳ0.4 ml of the concentrated medium (0.3 M final marimastat concentration) was incubated (12 h, 4°C) with Affi-Gel 10-TIMP-2 matrix by continuous rotation. After a brief centrifugation, the supernatant containing the unbound proteins was collected. The bound proteins were eluted with reducing Laemmli sample buffer (40). The bound and unbound fractions were resolved by reducing 15% SDS-PAGE followed by immunoblot analysis. (iii) Purified recombinant 18-and 21-kDa MT1-MMP fragments were incubated (1 h, 4°C) with TIMP-2 (1:1 molar ratio) in 50 l of collagenase buffer followed by immunoprecipitation with anti-TIMP-2 or anti-MT1-MMP antibodies. The complexes were detected by immunoblot analysis.
Immunoprecipitation-For immunoprecipitation of soluble MT1-MMP forms, serum-free media from 35 S-labeled cells expressing recombinant or natural MT1-MMP were immunoprecipitated with pAb 160/ Protein A-agarose beads under nondenaturing conditions as described (15). In some experiments, the 35 S-labeled media were concentrated (10-fold) before immunoprecipitation. To immunoprecipitate denatured samples, the concentrated media were supplemented with 10ϫ harvest buffer and boiled. The samples then received 2.5% Triton X-100 (final concentration) followed by addition of either pAb 437 or pAb 160 and Protein A-agarose beads as previously described (17). MT1-MMP⅐TIMP-2 complexes in 35 S-labeled medium samples or in mixtures of recombinant MT1-MMP fragments and TIMP-2 were co-immunoprecipitated with mAb 101 to TIMP-2 under nondenaturing conditions as previously described (33). The immunoprecipitates were resolved by reducing 15% SDS-PAGE, followed by autoradiography or by immunoblot analysis.
Gelatin Zymography and Immunoblot Analysis-Gelatin zymography was performed using 10% or 15% Tris-glycine SDS-polyacrylamide gels containing 0.1% gelatin as described (38). For immunoblot analysis, the serum-free conditioned media were collected, clarified by centrifugation, and concentrated (ϳ80-fold) on a Centricon Plus-20 concentrator (Fisher, Itasca, IL). An aliquot was resolved by reducing 15% SDS-PAGE followed by transfer to nitrocellulose membrane. The membranes were incubated with the appropriate antibodies as described (38). The immunocomplexes were detected by ECL according to the manufacturer's instructions (Pierce).

Membrane-bound and Soluble Forms of MT1-MMP-We
have previously shown that active MT1-MMP (57 kDa) is autocatalytically processed to a major membrane-bound 44-kDa species starting at Gly 285 (17). This processing should release the entire catalytic domain of MT1-MMP. We therefore examined the surface and extracellular distribution of MT1-MMP in various cell lines known to express natural MT1-MMP and in cells engineered to express recombinant MT1-MMP as a model system. In cells expressing natural MT1-MMP, these studies were carried out either without stimulation or after stimulation with TPA or ConA, two agents known to induce MT1-MMP expression (10,16,29,50). Surface biotinylation followed by immunoprecipitation with pAb 160 to the catalytic domain or pAb 437 to the hemopexin-like domain demonstrated that untreated HT1080 cells display the 57-kDa species of MT1-MMP as the major enzyme form on the cell surface (Fig. 1A, lanes 1  and 2). Treatment with either TPA (Fig. 1A, lanes 3 and 4) or ConA (Fig. 1A, lanes 5 and 6) induced the appearance of a 44-kDa species on the cell surface, which was detected only with the pAb 437 (Fig. 1A, lanes 3 and 5), indicating that this species represents a membrane-inserted form lacking the catalytic domain, in agreement with our previous studies using a vaccinia expression system (17). Thus, both TPA and ConA promote the processing of natural MT1-MMP (57 kDa) into the inactive 44-kDa form.
We next examined the serum-free conditioned media of various cell lines (untreated or treated with TPA or ConA) and Timp-2 (Ϫ/Ϫ) mouse fibroblasts stable transfected to express recombinant MT1-MMP for soluble MT1-MMP forms by immunoblot analysis (Fig. 1B) and immunoprecipitation (Fig. 1C). As shown in Fig. 1B, the medium of untreated HT1080 cells contains three proteins of 56, 50, and 18 kDa, which were recognized by a mAb to the catalytic domain of MT1-MMP (Fig. 1B,  lane 7). TPA (Fig. 1B, lane 8) and ConA (Fig. 1B, lane 9) treatment of HT1080 cells enhanced the levels of these forms in the media and resulted in the appearance of an additional soluble MT1-MMP form of ϳ31-35 kDa, which was particularly evident with ConA (Fig. 1B, lane 9). Media of untreated HFL1 fibroblasts (Fig. 1B, lane 10) and U-87 glioblastoma cells (data not shown) showed presence of the 50-kDa species. Media of ConA-treated HFL1 (Fig. 1B, lane 11) and U-87 (Fig. 1B, lane  12) cells contained the 50-and 18-kDa species and very low levels of the ϳ31-35-kDa species. Media of ConA-treated MDA-MB-231 contained mostly the 50-kDa form, as determined by immunoblot analysis (Fig. 1B, lane 13) or immunoprecipitation (Fig. 1C, lane 16). With the exception of the 18-and the 31-35-kDa species, the 56-and 50-kDa species were recognized by pAb 437 to the hemopexin-like domain, indicating that they comprise most of the ectodomain (data not shown).
Metabolic labeling of TPA-treated HT1080 cells followed by immunoprecipitation with pAb 160 yielded the 56-, 50-, 31-35-, and 18-kDa species (Fig. 1C, lane 14), whereas the same procedure in Timp-2 (Ϫ/Ϫ) cells expressing MT1-MMP yielded mostly the 50-and 18-kDa species (Fig. 1C, lane 17). No signal was observed in samples precipitated with Protein A-agarose beads without antibody (Fig. 1C, lanes 15 and 18). Considering that the 18-kDa species contains only one methionine residue (based on sequencing data, as shown below), the results of the immunoprecipitation of the 35 S-labeled media indicate that the 18-kDa species, compared with the other forms, exhibits a relatively higher specific activity and hence represents the major soluble form of MT1-MMP.
The concentrated serum-free media of untreated HT1080 cells and BS-C-1 cells expressing MT1-MMP were subjected to ultracentrifugation (100,000 ϫ g for 1 h at 4°C) to assess the distribution of the released MT1-MMP species. Under these conditions, membrane fragments and their associated proteins and large protein aggregates go to the pellet, whereas membrane-free soluble species distribute mostly in the supernatant (51,52). Immunoblot analysis showed that the 18-, 31-35-, and 50-kDa species were detected in the supernatant, whereas the 56-kDa species remained in the pellet (data not shown). These results suggest that, with the exception of the 56-kDa species, the other MT1-MMP species are true soluble forms.
Complex Regulation of MT1-MMP Shedding-We next investigated the effects of various protease inhibitors on the profile of MT1-MMP forms present in the media. To this end, we used TPA-treated HT1080 cells ( Fig. 2A), which express natural MT1-MMP and the Timp2 (Ϫ/Ϫ) cells (Fig. 2C) expressing recombinant MT1-MMP. In addition, we used BS-C-1 cells infected with vaccinia virus expressing MT1-MMP, as we have previously reported (17,20). The profile of soluble MT1-MMP species found in the vaccinia expression system (shown in Fig. 2B) was the same as that observed in cells expressing natural MT1-MMP (Fig. 1) and hence is not a consequence of overexpression of recombinant enzyme or cell lysis. Thus, this experimental system recapitulates the natural pattern of MT1-MMP shedding. The protease inhibitor studies showed that presence of the 18-kDa species in the media was specifically inhibited by TIMP-2, marimastat, and TIMP-4 in all cells tested (Fig. 2, A-C). TIMP-1, an extremely poor MT1-MMP inhibitor (53) (Fig. 2B, lane 7), and SB-3CT, a mechanismbased synthetic inhibitor specific for the gelatinases (36) (Fig.   FIG. 1. Membrane-bound and soluble MT1-MMP forms. A, biotinylation of membrane-bound MT1-MMP forms. HT1080 cells were untreated (lanes 1 and 2) or treated with either 100 nM TPA (lanes 3 and 4) or 10 g/ml ConA (lanes 5 and 6) in serum-free media overnight. The cells were then surface-biotinylated, as described (38). The biotinylated MT1-MMP forms were immunoprecipitated with pAb 437 (lanes 1, 3, and 5) or with pAb 160 (lanes 2, 4, and 6).  15 and 18). The immunoprecipitates were resolved by reducing 15% SDS-PAGE followed by autoradiography.
2C, lanes 14 -16), had no effect. Aprotinin (Fig. 2, A (lane 2) and B (lanes 8 and 9)) and leupeptin (40 g/ml; data not shown), two serine protease inhibitors, and E64 (10 M), an aspartic protease inhibitor (data not shown), had no effect on the shedding of the 18-kDa species. None of the inhibitors tested had a significant effect on the levels of the 50-kDa form and in fact, the levels of this species were somewhat increased in the presence of TIMP-2 and marimastat in BS-C-1 cells (Fig. 2B, lanes 2 and 6, respectively) but not in HT1080 cells (Fig. 2A, lane 3). Interestingly, we also found that, both in HT1080 cells (Fig. 2A,  lane 3) and in BS-C-1 cells expressing MT1-MMP (Fig. 2B,  lanes 2-6), the ϳ31-35-kDa species accumulated in the presence of TIMP-2 and marimastat but not in the presence of aprotinin (Fig. 2, A (lane 2) and B (lanes 8 and 9)).
The inhibitor profile studies suggested that shedding of the 18-kDa species is an autocatalytic event, whereas shedding of the 50-kDa species is not. To further investigate this process, we generated a catalytically inactive mutant of MT1-MMP by replacing Glu 240 with Ala (E240A-MT1). We also examined the role of the cytosolic domain of MT1-MMP in shedding. To this end, we constructed a truncated MT1-MMP lacking the cytosolic domain (⌬CD-MT1) by introducing a stop codon at Arg 563 . Wild type MT1-MMP and the E240A-MT1 and ⌬CD-MT1 mutants were expressed in BS-C-1 cells using the infection-transfection procedure followed by metabolic labeling as described under "Experimental Procedures." The 35 S-labeled conditioned media were immunoprecipitated with pAb 160. As shown in Fig. 3, wild type MT1-MMP shed the 50-, 31-35-, and 18-kDa species (Fig. 3, lane 1). In contrast, the E240A-MT1 catalytic mutant shed the 50-kDa species and a ϳ28-kDa form but not the 18-kDa fragment (Fig. 3, lane 2) consistent with the autocatalytic shedding of the latter species. The ⌬CD-MT1 mutant showed a shedding pattern similar to that observed with the wild type enzyme (Fig. 3, lane 3).
Structure and Characterization of the 18-kDa MT1-MMP Soluble Form-The relatively higher amounts of the 18-kDa fragment allowed its purification and characterization. N-ter-minal sequencing by Edman degradation revealed that the 18-kDa species displays Tyr 112 in the N terminus, in agreement with the N terminus displayed by membrane-tethered active MT1-MMP (57 kDa) (9,17). Mass spectrometry analysis of tryptic digests was performed to determine the C terminus of the 18-kDa fragment. As shown in the table of Fig. 4A (inset), a total of 32 peptides were isolated and their sequence determined. Three peptides demonstrated a C terminus ending at Ala 255 , indicating that the 18-kDa form extends from Tyr 112 to Ala 255 and therefore comprises most of the catalytic domain. Indeed, SDS-PAGE analysis demonstrated an ϳ3-kDa difference between the shed 18-kDa fragment (Fig. 4B, lane 2) and a commercially available recombinant catalytic domain of MT1-MMP (MT1-MMP cat ), which is known to extend from Tyr 112 to Gly 288 (Fig. 4B, lane 1).
Considering that the 44-kDa membrane-tethered species of MT1-MMP starts at Gly 285 (17) and Ile 256 and another between Gly 284 and Gly 285 . To determine the relative location of these sites in the catalytic domain of MT1-MMP, we used an energy-minimized computational model of the ectodomain of human pro-MT1-MMP that was recently constructed in our laboratory. 2 The only missing piece of structural information in this model pertained to the hinge between the catalytic and the hemopexin-like domains. The spatial location of the hemopexin-like domain in the computational model (data not shown) was based on that seen in the x-ray structure of MMP-2 (54). From this model, Fig. 5A shows only a view of the catalytic domain extending from Tyr 112 to Ser 287 . In the event that the hinge actually would not dislocate the hemopexin-like domain in MT1-MMP, the Ala 255 -Ile 256 bond is sheltered by the hemopexin-like domain, leaving the cleavage site Gly 284 -Gly 285 as the only likely candidate for the first hydrolytic cleavage event. However, even if the hemopexin-like domain of MT1-MMP is dislocated away from the catalytic domain, leaving the surface regions shown in Fig. 5B fully exposed to the milieu, still the cleavage site Gly 284 -Gly 285 is more readily accessible than is Ala 255 -Ile 256 because of the nature of the secondary structures in the protein. Therefore, shedding of the 18-kDa fragment is likely to be initiated at the Gly 284 -Gly 285 peptide bond, followed by a second cleavage at the Ala 255 -Ile 256 site. Fig. 5C depicts a diagram of active MT1-MMP (Tyr 112 -Val 582 ), showing the two cleavages at the Gly 284 -Gly 285 and Ala 255 -Ile 256 sites leading to the formation of the inactive 44-kDa species, which has been isolated and characterized from plasma membranes (17), and the soluble 18-kDa species. To shed the 18-kDa species, this process would have also generated a ϳ21-kDa intermediate fragment (Fig. 5C, dashed bracket) extending from Tyr 112 to Gly 284 . However, a soluble fragment of ϳ21 kDa, the putative precursor of the 18-kDa species, was not detected. Additionally, hydrolysis at the Ala 255 -Ile 256 peptide bond predicted impaired catalytic activity of the 18-kDa species because of the proximity of this site to the conserved methionine residue (Met 257 ) of the so-called methionine turn (55) and to the consensus sequence of the catalytic zinc ion binding site (Fig. 5C).
To gain insight into the biochemical properties of the 18-kDa (Tyr 112 -Ala 255 ) fragment and the putative 21-kDa (Tyr 112 -Gly 284 ) intermediate species, these proteins were expressed in bacteria and purified to homogeneity for further analyses. Activity assays demonstrated that, whereas the 21-kDa (Tyr 112 -Gly 284 ) fragment exhibited gelatinolytic (Fig. 6B, lane  4) and pro-MMP-2-activating activities (Fig. 6C, lane 6), the 18-kDa (Tyr 112 -Ala 255 ) fragment was catalytically inactive (Fig. 6, B (lane 3) and C (lane 5)). To obtain quantitative data, the recombinant fragments were examined for their ability to hydrolyze a fluorogenic peptide substrate as a function of time. MT1-MMP cat was used as a positive control. As summarized in Table I, k cat and K m values of ϳ1 s Ϫ1 and 10 M, respectively, were obtained yielding k cat /K m values of 10 5 M Ϫ1 s Ϫ1 , which reflects the high reactivity of MT1-MMP cat and the 21-kDa (Tyr 112 -Gly 284 ) enzymes toward the synthetic peptide substrate used. Moreover, indistinguishable values were obtained for these two MT1-MMP species. In contrast, no enzyme concentration dependence of the rate of substrate hydrolysis was detected with the 18-kDa (Tyr 112 -Ala 255 ) fragment with concentrations up to 235 nM. In fact, the hydrolysis rate of the substrate in the presence of the enzyme was essentially indistinguishable from the background hydrolysis detected in buffer only. A comparable concentration of the 21-kDa (Tyr 112 -Gly 284 ) species (54 nM) yielded an increase in fluorescence that rapidly exceeded the detection limit of the instrument used. Together, these studies indicate that autocatalytic processing of MT1-MMP at the Ala 255 -Ile 256 site obliterates catalytic competence resulting in an inactive soluble form of 18 kDa.
Although the 18-kDa species of MT1-MMP is inactive, the other soluble species may maintain enzymatic activity. Unfortunately, the paucity of these enzyme species in the media precluded purification and characterization. Thus, to investigate the activity of the soluble MT1-MMP species, we used conditioned medium of BS-C-1 cells infected to express MT1-MMP and examined its ability to promote pro-MMP-2 activation after addition of exogenous recombinant pro-MMP-2. This cell expression system was chosen because it is devoid of MMP-2 (33) and because it releases the 50-and 18-kDa form of MT1-MMP into the media (Fig. 2B, lane 1). As a control, we used conditioned media of BS-C-1 cells infected only with the T7 RNA polymerase-expressing vaccinia vTF7-3 virus (33). As shown in Fig. 7, the conditioned media derived from the MT1-MMP-expressing cells promoted the generation of the intermediate form of MMP-2 (Fig. 7, lane 6), consistent with the twostep model of surface activation of pro-MMP-2 by MT1-MMP, as previously proposed (53,56). In contrast, no processing of pro-MMP-2 was observed with the control media (Fig. 7, lane  4), demonstrating the specificity of the reaction. Because the 18-kDa species is inactive, these studies suggest that MT1-MMP fragments other than the 18-kDa form, such as the 50-kDa species, are likely to be the enzyme species responsible in the media for the processing of pro-MMP-2.
Interactions with TIMP-2-We have previously shown that, on the cell surface, TIMP-2 binds to the active 57-kDa form of MT1-MMP but not the 44-kDa inactive species (17), indicating that binding is mostly mediated by the catalytic domain. Here we examined the ability of the soluble MT1-MMP forms to bind TIMP-2. To this end, media of 35 S-labeled BS-C-1 cells infected to express MT1-MMP were incubated with or without exogenous recombinant TIMP-2 or TIMP-1 followed by immunoprecipitation. As shown in Fig. 8A, MT1-MMP-expressing BS-C-1 cells shed the 50-and 18-kDa species of MT1-MMP, as determined after immunoprecipitation with pAb 160 (Fig. 8A, lane   1). No signal was detected without antibody (Fig. 8A, lane 2). After addition of exogenous TIMP-2 and immunoprecipitation with mAb 101, only the 50-kDa species of MT1-MMP and some endogenously produced 35 S-TIMP-2 were detected in the coprecipitate (Fig. 8A, lane 3). Indeed, BS-C-1 cells produce very low levels of endogenous TIMP-2 (20). No signal was detected in samples that received exogenous TIMP-1 and anti-TIMP-1 pAb (Fig. 8A, lane 4). Accordingly, the 50-kDa species could not be detected with pAb 160 after TIMP-2 addition because of epitope occupancy in the enzyme-inhibitor complex as this pAb is directed to the catalytic domain (data not shown). These results indicate that TIMP-2, but not TIMP-1, binds to the soluble 50-kDa species via the catalytic domain, in agreement with the known TIMP-binding profile of MT1-MMP (53). In contrast, the soluble 18-kDa fragment cannot form a stable complex with TIMP-2.
A TIMP-2 affinity binding procedure was carried out to assess whether the 31-35-kDa species could bind TIMP-2. To this end, we used conditioned medium from BS-C-1 cells that were infected to express MT1-MMP in the presence of 1 M marimastat, which induces the appearance of these species, as shown in Fig. 2B. The media were subjected to TIMP-2 affinity binding using immobilized TIMP-2, and the bound and unbound fractions were analyzed by immunoblot analysis as described under "Experimental Procedures." As shown in Fig. 8B, the 31-35-and the 50-kDa species were detected in the bound fraction (Fig. 8B, lane 3), albeit a significant amount of these species remained in the unbound fraction (Fig. 8B, lane 2) when compared with the load (Fig. 8B, lane 1), suggesting that they exhibit a relatively low affinity for TIMP-2, under the experimental conditions. In contrast, the 18-kDa species did not bind to the TIMP-2 affinity matrix, in agreement with the immunoprecipitation data. Although these studies are not quantitative, the poor binding of the soluble MT1-MMP species to the immobilized TIMP-2 is unlikely to be the result of the presence of marimastat because the affinity matrix is saturated with TIMP-2 and the final concentration of marimastat was less than 0.3 M.
We next examined the ability of the recombinant MT1-MMP fragments to bind TIMP-2 by co-immunoprecipitation. These studies demonstrated that only the 21-kDa (Tyr 112 -Gly 284 ) species was able to form a stable complex with the inhibitor (Fig.  8C, lane 3). Thus, loss of the 29 amino acids between Ala 255 and Gly 284 during the formation of the 18-kDa fragment strongly compromised TIMP-2 interactions. Consistently, inhibition studies demonstrated that TIMP-2 was an effective slow-binding inhibitor of the 21-kDa (Tyr 112 -Gly 284 ) species when compared with MT1-MMP cat (Table II). Fitting the data to a slowbinding inhibition model yielded k on and k off values of 10 6 M Ϫ1 s Ϫ1 and 2 ϫ 10 Ϫ4 s Ϫ1 , respectively, resulting in sub-nanomolar K i values.

DISCUSSION
The stability of active MT1-MMP on the cell surface is a complex process involving a balance between autocatalytic processing (17) and enzyme internalization (57,58). Both processes can regulate the amount of active enzyme available for pericellular proteolysis and appear to be regulated in part by the presence of TIMPs. The major pathway of active MT1-MMP processing on the cell surface is an autocatalytic event that generates a 44-kDa membrane-anchored fragment starting at Gly 285 and thus lacks the entire catalytic domain (17). This process may switch the proteolytic machinery from the cell  1 and 3). The immunoprecipitates were resolved by 15% SDS-PAGE followed by immunoblot (IB) analysis with the same antibodies.

TABLE II Kinetic parameters for inhibition of the 21-kDa MT1-MMP fragment by TIMP-2
The enzyme (0.2-0.8 nM) was added to a solution of MOCAcPLGLA 2 pr(Dnp)ARNH 2 (10 M) and increasing concentrations of TIMP-2 in buffer R. Substrate hydrolysis was monitored for 20 min. The kinetic parameters were evaluated as described under "Experimental Procedures."  5 and 6). Samples were incubated for 0 min (lanes 1, 3, and 5) and 22 h (lanes 2, 4, and 6). Aliquots from the samples were subjected to gelatin zymography. Lane 7 shows APMA-activated pro-MMP-2 (62 kDa) as a positive control. MT1-MMP species including two major species of 50-and 18 kDa and a series of minor fragments of 56, and 31-35 kDa, which are differentially regulated by TPA, ConA, and MMP inhibitors. With the exception of the 56-kDa species, which was retained in the pellet after ultracentrifugation and thus, may be associated with membrane fragments (52), all other MT1-MMP species were soluble and thus, represent true shedding. Shedding of the 50-and 18-kDa species occurred without external stimulation, indicating that they represent a normal process of MT1-MMP turnover under basal conditions. However, exposure of cells to either TPA or ConA, two nonphysiological agents known to induce MT1-MMP expression and pro-MMP-2 activation (29,50,59), resulted in increased levels of all the soluble forms in the media. Based on the protease inhibitor profile, both autocatalytic and non-autocatalytic processes are involved in MT1-MMP shedding. High affinity natural (TIMP-2 and TIMP-4, but not TIMP-1) and synthetic MMP inhibitors, known to stabilize active MT1-MMP on the cell surface by inhibiting the processing of MT1-MMP to the 44-kDa form (17, 19 -21), inhibited shedding of the 18-kDa species. Additionally, the E240A-MT1 catalytic mutant enzyme, which cannot be processed to the 44-kDa species (22,60), failed to shed this fragment. Thus, shedding of the 18-kDa species is the product of the autocatalytic processing of active MT1-MMP (57 kDa) on the cell surface, which yields a major inactive membrane-tethered species of 44 kDa (17). Consistently, TPA and ConA treatments, which promote MT1-MMP expression and processing on the cell surface, stimulate shedding of the 18-kDa fragment. The ability of cells to elicit autocatalytic shedding depends on the expression level of MT1-MMP on the cell surface and the levels and availability of TIMPs. High levels of TIMP-2 and/or presence of other TIMP-2-binding MMPs will alter the autocatalytic pathway by modifying TIMP-2 availability as shown in MDA-MB-231 cells, which, as opposed to HT1080 cells, do not produce MMP-2; therefore, the autocatalytic shedding (release of the 18-kDa species) is restricted.
A battery of protease inhibitors including metalloprotease, serine, and aspartic protease inhibitors failed to reduce the levels of the 50-and 31-35-kDa species. Additionally, these species were observed in the media of cells expressing the E240A-MT1 catalytic mutant. Thus, production of these soluble MT1-MMP fragments is a non-autocatalytic event. Interestingly, TPA and ConA, which promote autocatalytic processing, enhanced the levels of these species, suggesting an additional level of regulation by these agents (50). The identity of the protease(s) responsible for the non-autocatalytic shedding of MT1-MMP remains to be determined. Our evidence and previous evidence (23,24) suggest that, in the case of the 50-kDa species, the protease(s) must cleave within the juxtamembrane (stem) region of MT1-MMP causing the release of the entire ectodomain. Consistently, the soluble 50-kDa species was able to form a complex with TIMP-2, in agreement with early studies (24). Whether cleavage at the stem region takes place at the cell surface or intracellularly, as shown with MT5-MMP (18), remains to be determined. However, it is unlikely that a furin-like enzyme is involved in this process because, in contrast to MT5-MMP, a specific furin-recognition motif was not found in the ectodomain of MT1-MMP (18). Based on the pattern of MT1-MMP forms present on the cell surface (the 57and 44-kDa species) and the high levels of the 18-kDa species, as determined by the immunoprecipitation experiments, the non-autocatalytic pathway is likely to comprise a minor aspect of the shedding process. However, this process may produce functional enzyme fragments, such as the 50-kDa species (24), which would promote pro-MMP-2 processing, as demonstrated here with the conditioned media of BS-C-1 cells expressing MT1-MMP, and as shown previously in gelatin zymography assays (24). Thus, soluble ectodomain fragments with catalytic activity may extend MT1-MMP-dependent proteolysis beyond the cell surface environment by promoting the hydrolysis of a variety of substrates including extracellular matrix components (4). In addition, these fragments by binding TIMP-2 may deprive the membrane-tethered enzyme of inhibitor regulation.
An interesting observation of this study was the appearance of a 31-35-kDa soluble species that was induced either by ConA treatment or high levels of TIMP-2 or marimastat. The appearance of the 31-35-kDa species in the presence of TIMP-2 or marimastat correlated with a decrease in the levels of the 18-kDa species, suggesting the possibility that the formation of these species may be related. For example, it is possible that, in addition to the autocatalytic shedding, there is a non-MMP-dependent shedding mechanism that releases the 31-35-kDa fragment, which in turn is processed to the 18-kDa species via a metalloprotease-dependent process, as suggested by the accumulation of the 31-35-kDa species in the presence of TIMP-2 or marimastat. Alternatively, TIMP-2 binding to the 31-35-kDa species may prevent a non-metalloprotease from accessing the Ala 255 -Ile 256 site, thus resulting in accumulation. Indeed, the 31-35-kDa species binds TIMP-2, albeit with an apparent low affinity. Another possibility is that the accumulation of the 31-35-kDa species in the presence of TIMP-2 or marimastat represents shedding of MT1-MMP catalytic domain-inhibitor complexes. Stabilization of the membrane-anchored enzyme by formation of enzyme-inhibitor complexes (17) may induce conformational changes, which may predispose the enzyme to a non-metalloprotease-mediated ectodomain shedding. However, the lack of a readily detectable counterpart to the 31-35-kDa species on the plasma membrane suggests the possibility that this fragment(s) is not a shedding product of the membranebound enzyme but a result of the turnover of the larger MT1-MMP soluble species, like the 50-kDa species, via a TIMP-2insensitive process. Structural data and studies in cellular systems with defined proteolytic backgrounds will help to distinguish between these possibilities. Although the potential contribution of the 31-35-kDa species to the formation of the 18-kDa species cannot be disregarded, this is likely to be minimal when compared with the formation of the 18-kDa species generated by the autocatalytic processing of MT1-MMP on the cell surface. Taken together, these observations further underscore the complexity of the MT1-MMP shedding process and the unexpected consequences that TIMP-2 and synthetic MMP inhibitors may have on the regulation of MT1-MMP on the cell surface, as we have previously documented (20,21).
Previous studies reported that the cytosolic domain of MT1-MMP plays a role in the stabilization of MT1-MMP on the cell surface by altering the rate of enzyme internalization (58) and is also involved in enzyme homodimerization (60), a process thought to favor autocatalytic turnover (61). Here we have shown that the pattern of MT1-MMP shedding was essentially unaltered in enzymes lacking the cytosolic domain. This finding suggests that homotypic physical interactions mediated by the cytosolic domain (60) are not essential for MT1-MMP autocatalytic and non-autocatalytic shedding.
The autocatalytic pathway of MT1-MMP shedding concludes with the release of an 18-kDa fragment that extends from Tyr 112 to Ala 255 , 29 amino acid residues upstream of the Gly 285 displayed at the N terminus of the membrane-tethered 44-kDa species (17). Therefore, shedding of the 18-kDa fragment would require cleavage at both the Ala 255 -Ile 256 and the Gly 284 -Gly 285 peptide bonds. The computational model of MT1-MMP shows that the Ala 255 -Ile 256 peptide bond near the methionine turn is sheltered and thus is less accessible to proteolysis. This sug-gests that cleavage at the Gly 284 -Gly 285 peptide bond, which is on the surface, is likely to precede that at the Ala 255 -Ile 256 site. However, at present, it is unclear whether cleavage at the Gly 284 -Gly 285 site predisposes for hydrolysis at the second Ala 255 -Ile 256 site. Data from the recombinant 21-kDa fragment, which contains the Ala 255 -Ile 256 site but ends at Gly 284 , clearly indicates that disruption of the Gly 284 -Gly 285 site does not disturb the integrity and functionality of the Ala 255 -Ile 256 peptide bond. Indeed, we have shown that the 21-kDa fragment is stable and catalytically competent. Furthermore, the 21-kDa fragment was not hydrolyzed at the Ala 255 -Ile 256 site when was incubated alone or with HT1080 cells or their conditioned media (data not shown). We posit that the sequence of events leading to the cleavage of the Ala 255 -Ile 256 peptide bond after cleavage at the Gly 284 -Gly 285 site occur only within membranetethered active MT1-MMP molecules. Conceivably, cleavage at the Gly 284 -Gly 285 site within a membrane-anchored enzyme destabilizes the structure yielding the Ala 255 -Ile 256 peptide bond susceptible for subsequent hydrolysis. The inhibitor profile suggests that cleavage at the Gly 284 -Gly 285 site is an autocatalytic event because high affinity MT1-MMP protease inhibitors like TIMP-2 and TIMP-4 (19,53,62) prevented formation of the 44-kDa species starting at Gly 285 (17). Additionally, a catalytic mutant of MT1-MMP was not processed to the 44-kDa species, as previously reported (22,60). In regard to the Ala 255 -Ile 256 site, the data suggest that cleavage at that site is most likely to be also autocatalytic because TIMP-1 does not prevent shedding of the 18-kDa species. In addition, if it were mediated by another metalloprotease or a serine protease, the presence of TIMP-1, marimastat, or serine protease inhibitors should have resulted in the appearance of the 21-kDa fragment extending from Tyr 112 to Gly 284 , which was not detected. According to the N terminus of the membrane-bound 44-kDa species (Gly 285 ) and the C-terminal end of the soluble 18-kDa fragment (Ala 255 ), shedding of the catalytic domain should proceed via an intermediate species of ϳ21 kDa. However, such a species could not be detected on the cell surface or in the media of cells expressing recombinant or natural MT1-MMP. A plausible explanation for the absence of a soluble 21-kDa fragment during the shedding of the 18-kDa species is that the cleavages at the Ala 255 -Ile 256 and Gly 284 -Gly 285 sites occur rapidly and sequentially and thus would preclude accumulation of a 21-kDa inter-mediate form and thus ends with the 18-kDa fragment as the final product. Attempts to induce accumulation of the intermediate 21-kDa species by generating A255V or A255I substitutions at the Ala 255 -Ile 256 cleavage site were unsuccessful as these mutants failed to undergo activation and processing, demonstrating the importance of this site for catalytic competence. 3 The 18-kDa fragment ends just one residue upstream of the conserved Met 257 known to be part of the methionine turn, a structural feature characteristic of all members of the MMP family and of the super family of metzincins (55). Topologically, the methionine turn is positioned near the three histidines that coordinate with the catalytic zinc ion and is on the opposite side of these residues with respect to the active site cavity. Thus, the methionine turn is thought to be critical for catalysis, based on its close proximity to the coordination site for the catalytic zinc ion. Furthermore, the side chain of Ile 256 , upstream of Met 257 , forms a portion of the S1Ј pocket of this enzyme. The loss of the 29-amino acid fragment during the formation of the 18-kDa MT1-MMP species would by necessity truncate the S1Ј pocket; hence, it has the ability to potentially impair or alter substrate binding properties. Furthermore, the proximity of the surface loop that bears the methionine residue to the TIMP-2 binding region (43) indicates that a disorder in this location would likely impair TIMP-2 binding. Our results with the 18-kDa fragment support this notion and provide experimental documentation of the importance of the methionine turn for MMP-mediated catalysis and TIMP binding. It is worth noting that, as far as we know, the proteolytic inactivation at the methionine turn as it occurs during MT1-MMP processing has not been reported for other members of the MMP family including soluble MMPs, despite the conserved nature of this motif. This suggests the possibility that MT1-MMP specifically developed a self-controlling mechanism to allow the enzyme to function principally as a membrane-anchored protease, and any perturbation in its cellular localization would result in specific enzyme inactivation. This may explain why a transmembrane-deleted soluble MT1-MMP expressed in HT1080 cells was processed at the Ala 255 -Ile 256 site, 3 P. Osenkowski and R. Fridman, manuscript in preparation.  (16,17,19,21,62). In the non-autocatalytic shedding of MT1-MMP (on right side of figure), a yet unknown protease releases the ectodomain (ϳ50 kDa, no structural information yet), which possess catalytic activity (24). This species can bind TIMP-2. Not shown here is the minor ϳ31-35-kDa soluble fragment, which may be related to these pathways of shedding. Single-letter amino acid codes are used. possibly by the endogenous MT1-MMP (15). Together, these data reveal the importance of the Ala 255 -Ile 256 site for the maintenance of catalytic competence in MT1-MMP. Sequence alignment of the transmembrane MT-MMPs (MT1-, MT2-, MT3-, and MT5-MMP) (26) reveals a complete homology around the A-I peptide bond and the residues near the methionine turn (depicted in Fig. 5C). Presently, the shedding mechanisms of the MT-MMP family members have not been completely elucidated. It would be interesting to determine whether cleavage at the conserved Ala-Ile peptide bond represents a common and specific mechanism designed to terminate MT-MMP-dependent catalysis and TIMP interactions at the cell surface.
In summary, we have identified the major soluble forms of MT1-MMP and demonstrated the complexity of MT1-MMP shedding and its regulation by natural and synthetic MMP inhibitors. Fig. 9 depicts the autocatalytic processing of MT1-MMP on the cell surface leading to the release of the inactive 18-kDa species and the non-autocatalytic shedding leading to the release of the entire ectodomain by a yet unknown protease. Inhibitors of MT1-MMP block autocatalytic shedding and thus stabilize the active enzyme on the cell surface (17). The autocatalytic shedding terminates MT1-MMP-dependent pericellular proteolysis, independently of endogenous inhibitors, by specific hydrolysis at vital conserved sites (methionine turn). On the other hand, the non-autocatalytic shedding, as represented by the 50-kDa species (24) and possibly the 31-35-kDa species (not shown in Fig. 9), may still contribute to pericellular proteolysis and partly compensate for the removal of enzyme from the cell surface by shifting the proteolytic machinery to a new front, possibly with different substrates and functional consequences. Finally, the shed ectodomain of MT1-MMP may bind TIMPs (24) and hence alter the enzyme-inhibitor balance at the cell surface and/or may play new unexpected roles (3).