The Inactive 44-kDa Processed Form of Membrane Type 1 Matrix Metalloproteinase (MT1-MMP) Enhances Proteolytic Activity via Regulation of Endocytosis of Active MT1-MMP*

Membrane type 1 (MT1) matrix metalloproteinase (MMP-14) is a membrane-tethered MMP considered to be a major mediator of pericellular proteolysis. MT1-MMP is regulated by a complex array of mechanisms, including processing and endocytosis that determine the pool of active proteases on the plasma membrane. Autocatalytic processing of active MT1-MMP generates an inactive membrane-tethered 44-kDa product (44-MT1) lacking the catalytic domain. This form preserves all other enzyme domains and is retained at the cell surface. Paradoxically, accumulation of the 44-kDa form has been associated with increased enzymatic activity. Here we report that expression of a recombinant 44-MT1 (Gly285–Val582) in HT1080 fibrosarcoma cells results in enhanced pro-MMP-2 activation, proliferation within a three-dimensional collagen I matrix, and tumor growth and lung metastasis in mice. Stimulation of pro-MMP-2 activation and growth in collagen I was also observed in other cell systems. Expression of 44-MT1 in HT1080 cells is associated with a delay in the rate of active MT1-MMP endocytosis resulting in higher levels of active enzyme at the cell surface. Consistently, deletion of the cytosolic domain obliterates the stimulatory effects of 44-MT1 on MT1-MMP activity. In contrast, deletion of the hinge turns the 44-MT1 form into a negative regulator of enzyme function in vitro and in vivo, suggesting a key role for the hinge region in the functional relationship between active and processed MT1-MMP. Together, these results suggest a novel role for the 44-kDa form of MT1-MMP generated during autocatalytic processing in maintaining the pool of active enzyme at the cell surface.

The importance of MT1-MMP for pericellular proteolysis demands a tight control of its catalytic activity at the cell surface. This is partly achieved by the action of endogenous protease inhibitors, the tissue inhibitors of metalloproteinases (TIMPs), which bind to the active site inhibiting catalysis (15). In addition, by virtue of being a membrane-anchored protease, MT1-MMP developed a unique set of regulatory mechanisms that control enzymatic activity independently of TIMPs. These processes include the targeting of active enzyme to specific plasma membrane locations, endocytosis, and autocatalytic processing (1, 16 -19). Together, these distinct processes determine the level of active enzyme on the cell surface. However, how these processes are integrated to control the pool of active MT1-MMP is not understood.
Although MT1-MMP processing is thought to terminate activity on the cell surface, the structural characteristics of the remnant 44-kDa product suggest that it may play a more complex role in enzyme regulation. The 44-kDa species is composed of the hinge region, the hemopexin-like domain, and the complete anchoring apparatus with its cytosolic tail. Furthermore, the 44-kDa fragment is retained on the cell surface, and like active MT1-MMP it is also cleared from the cell surface by endocytosis (28,48). Previous studies showed that expression of recombinant species of MT1-MMP lacking the catalytic domain, analogous to the 44-kDa fragment, inhibited MT1-MMP-dependent pro-MMP-2 activation (57)(58)(59), collagen degradation (60), in vitro tumor cell migration and invasion (61,62), and in vivo tumor growth (62). These studies suggested that the 44-kDa species of MT1-MMP behaves as a dominant negative regulator of enzyme function. Paradoxically, and contrary to these observations, a plethora of studies have shown a positive correlation between enhanced processing of MT1-MMP and enzymatic activity (21,29,30,36,37,46,(52)(53)(54)(63)(64)(65). Given these conflicting observations, we set forth to investigate the functional relationship between active MT1-MMP and its processed 44-kDa form in various cellular settings and its effect on enzyme function in vitro and in vivo. We present evidence that the 44-kDa form of MT1-MMP positively influences the level and activity of surface MT1-MMP by regulating enzyme internalization. This positive effect of the processed form depends on the presence of the hinge region. We propose a model in which the 44-kDa form of MT1-MMP regulates the endocytosis of the active protease to preserve a viable level of MT1-MMP on the cell surface.

EXPERIMENTAL PROCEDURES
Cell Culture-Human fibrosarcoma HT1080 cells, human breast carcinoma MDA-MB-231 cells, human breast carcinoma T47D cells, and monkey kidney epithelial BS-C-1 cells were obtained from the American Type Culture Collection (ATCC, Manassas, VA). HT1080, MDA-MB-231, and BS-C-1 cells were cultured in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal bovine serum (FBS) and antibiotics. T47D cells were cultured in RPMI 1640 medium supplemented with 10% FBS, insulin, and antibiotics. Human glioma U87 cells were a gift from Dr. Bonnie Sloane (Wayne State University, Detroit, MI) and were cultured in minimum essential media supplemented with 10% FBS and antibiotics. Human breast carcinoma BT549 cells, a kind gift from Dr.
Hyeong-Reh Choi Kim (Wayne State University, Detroit, MI), were cultured in DMEM/F-12 supplemented with 10% FBS and antibiotics. HeLa cells (CCL-2.2) were obtained from the ATCC and were cultured in Spinner medium (Quality Biologicals Inc., Gaithersburg, MD) supplemented with 5% horse serum and antibiotics.
Recombinant Proteins and Antibodies-Human recombinant pro-MMP-2, TIMP-2, and TIMP-1 were expressed in HeLa cells infected with the appropriate recombinant vaccinia viruses and purified to homogeneity, as described previously (67). The rabbit polyclonal antibody (pAb) 437, raised against residues 437-454 of the hemopexin-like domain of human MT1-MMP was described previously (6 Fig. 1A), the entire catalytic domain of MT1-MMP (Tyr 112 to Gly 284 ) was deleted by mutagenic PCR from pro-MT1-MMP using the QuikChange TM site-directed mutagenesis kit (Stratagene, La Jolla, CA), specific primers, and wild-type pro-MT1-MMP cDNA as the template. This procedure generated a cDNA fragment in which the furin-recognition motif ending at Arg 111 was positioned immediately upstream of Gly 285 at the beginning of the hinge region. This cDNA fragment was cloned into the pcDNA3.1 (Invitrogen) expression vector for stable transfection, the pTF7-EMCV-1 vector for transient expression in the vaccinia expression system (66), and the pIND vector for inducible expression in response to ponasterone A (Invitrogen). In addition, we generated a series of cDNA constructs to express 44-MT1 lacking the cytosolic tail (ending at Arg 563 and referred to as 44⌬CT; Fig. 1A) and 44-MT1 lacking the hinge region (residues 285-315 and referred to as 44⌬H; Fig. 1A). These cDNA constructs were inserted into the pcDNA3.1 expression vector. The correct sequence of all the cDNA constructs was verified by DNA sequencing of both strands.
Stable Transfections-Cells (HT1080, MDA-MB-231, U87, and T47D) were stably transfected with empty pcDNA3.1 vector (referred to as EV cells) or pcDNA3.1/44-MT1 vector (referred to as 44-MT1 cells) using the Effectene transfection reagent (Qiagen, Valencia, CA) as described by the manufacturer. T47D breast carcinoma cells were also transfected with wild-type MT1-MMP in pcDNA3.1 vector. Geneticin (G418 sulfate, Invitrogen)-resistant pooled populations were selected using conventional approaches. HT1080 and U87 cells were also transfected with 44⌬CT and 44⌬H in pcDNA3.1, and both clones and pooled populations were obtained by Geneticin selection. For most studies, both pooled populations and representative single clones were used. For in vivo studies, two representative clones with each construct, which varied in detected amounts of recombinant protein expression, were pooled (pooled clones).
Inducible Expression System-BT549 breast carcinoma cells were double-transfected with the pVgRXR vector (Invitrogen), which expresses the heterodimeric ecdysone receptor and the retinoid X receptor, and the pIND-inducible expression vector that expresses a modified ecdysone-response element upstream of a minimal heat shock promoter and a multiple cloning site. In the presence of ponasterone A, heterodimeric receptors binding to a modified ecdysone-response element induce expression of the 44-MT1 cDNA that was previously inserted into the pIND-inducible expression plasmid (Invitrogen).
Infection-Transfection-BS-C-1 cells seeded in 6-well plates were infected with 1 plaque-forming unit/cell of vTF7-3 vaccinia virus for 45 min in DMEM containing 2.5% FBS. The media were removed, and the infected cells were co-transfected with pTF7-EMCV-1 vector expressing wild-type MT1-MMP alone or 44-MT1 at 1:1 and 1:10 molar ratios (MT1-MMP:44-MT1) or with empty vector alone, up to a total of 3 g of plasmid DNA/well using Effectene. The next day, the cells were incubated with 10 nM recombinant pro-MMP-2 for 30 min at 37°C in serum-free media. The lysates and media were collected and subjected to immunoblot analysis and gelatin zymography, respectively, as described below.
Immunoblot Analyses-Cells were lysed with lysis buffer (25 mM Tris/HCl, pH 7.5, 1% Nonidet P-40, and 100 mM NaCl) supplemented with EDTA-free protease inhibitor mixture (Roche Applied Science), on ice for 1 h. The lysates were then clarified by centrifugation (13,000 rpm, 30 min). The supernatants were collected, and protein concentration was determined using the BCA procedure (Pierce) as described by the manufacturer. Equal amounts of protein were mixed with reducing Laemmli SDS sample buffer, boiled, and resolved by SDS-PAGE, followed by transfer to a nitrocellulose membrane. Detection of antigen was carried out with the appropriate antibody followed by enhanced chemiluminescence according to the manufacturer's instructions (Pierce). ␤-Actin was used as a loading control.
Gelatin Zymography-Serum-free conditioned media of stable or transiently transfected cells were collected after 16 h of incubation, clarified by centrifugation, and mixed with Laemmli sample buffer without reducing agents and heating. In some experiments, the stably transfected cells were treated with 10 g/ml of concanavalin A (conA, Sigma) overnight in serum-free media. In other experiments, HT1080 transfectants suspended in complete media were seeded in 12-well plates, which were uncoated or previously coated with 1.5 mg/well of rat tail collagen I (Trevigen, Gaithersburg, MD). Six h later, the media were aspirated and replaced with 0.3 ml/well serum-free media followed by an overnight incubation. Pro-MMP-2 activation was monitored in the media by gelatin zymography, as described previously (70).
Cell Surface Biotinylation-Cells in 6-well plates were cooled down for 5 min, rinsed with cold phosphate-buffered saline (PBS), pH 7.4, containing 0.1 mM CaCl 2 and 1 mM MgCl 2 (PBS-CM), and then biotinylated with 0.5 mg/ml EZ-link sulfo-NHSbiotin (Pierce) for 30 min at 4°C, as described previously (24). As a control, a parallel plate of cells received PBS without biotin. Biotin was quenched with 50 mM NH 4 Cl in PBS-CM for 10 min at 4°C. The cells were lysed with 0.2 ml/well of lysis buffer and centrifuged (13,000 rpm, 20 min), and the supernatant was incubated with 50 l of streptavidin beads (Pierce) overnight at 4°C. The beads were washed four times with harvest buffer (0.5% SDS, 60 mM Tris/HCl, pH 7.5, 2 mM EDTA) supplemented with 2.5% Triton X-100 (final concentration). The bound biotinylated proteins were eluted with reducing Laemmli SDS sample buffer, boiled, and resolved by 10% SDS-PAGE, followed by transfer to a nitrocellulose membrane. The biotinylated MT1-MMP forms were detected with the pAb 437, pAb 815, or mAbLEM2/15 MT1-MMP antibody. The blot was reprobed with transferrin receptor antibodies as a control for loading.
Semiquantitative RT-PCR-Total RNA of HT1080 cells was extracted with RNeasy mini kit (Qiagen, Valencia, CA). RT-PCR was performed with 1 g of each total RNA sample using SuperScript TM III reverse transcriptase (Invitrogen) and subsequently HotStar Taq master mix kit (Qiagen) following the manufacturer's instructions. The sequences of the specific primers (IDT, Coralville, IA) for human TIMP-2 used are as follows: forward, 5Ј-GAA ACG ACA TTT ATG GCA ACC-3Ј; reverse, 5Ј-GCT GGA CCA GTC GAA ACC-3Ј. Thirty cycles of PCR were performed. The housekeeping gene GAPDH was also amplified and used as an internal control. The sequences of the human GAPDH primers (IDT) are as follows: forward, 5Ј-CCA CCC ATG GCA AAT TCC ATG GCA-3Ј; reverse, 5Ј-TCT AGA CGG CAG GTC AGG TCC ACC-3Ј. The amplified fragments were resolved by 1% agarose gels and detected by ethidium bromide staining.
TIMP-2 ELISA-Cells in 6-well plates were incubated at 37°C with serum-free media. Forty eight hours later, the media were collected, and the cells were rinsed twice with PBS, solubilized with cold lysis buffer, and centrifuged. The conditioned media were collected and clarified by centrifugation and concentrated with a Microcon YM-10 centrifugal filter device. Concentrated media and the lysate samples were analyzed for TIMP-2 protein by ELISA (QIA40, Oncogene Research Products, San Diego) as described by the manufacturer's instructions.
Growth of Cells within a Three-dimensional Collagen I Matrix-Cells suspended in 50 l of serum free-media were added to a solution of 3.0 mg/ml rat tail collagen I, which was previously neutralized with 0.1 N NaOH. The mixture of cells and collagen was poured into the wells (0.5 ml/well, 10 4 cells/ well for HT1080 and MDA-MB-231 cells, and 2 ϫ 10 4 cells/well for U87 cells) of a 12-well tissue culture plate in the absence or presence of recombinant 12.5 g/ml TIMP-1 or 5 g/ml TIMP-2, and the plates were incubated at 37°C for 1 h to allow polymerization. Then each well received 3 ml of complete media (DMEM ϩ 10% FBS ϩ antibiotics and Geneticin). After 11 days incubation for HT1080 cells and 14 days for MDA-MB-231 and U87 cells, the media were removed, and the collagen matrix was washed twice with PBS. Then 10 units of bacterial collagenase (Sigma) per 1 mg of collagen were added to the wells and incubated for 15 min at 37°C. After the collagen gel was completely dissociated, complete media were added to stop collagenase activity, and the solution was centrifuged (1,000 rpm) for 5 min and washed twice with PBS. The cells were counted in the presence of trypan blue. Triplicate wells were used for each of the stable transfectant cell lines. Each experiment was performed in triplicate, and quantification values represent the means of three independent experiments Ϯ S.D. Statistical significance was determined using the Student's t test.
RNA Interference-Gene silencing of endogenous MT1-MMP in pooled populations of HT1080 cells transfected with 44-MT1 was performed using sequence-specific small interfering RNA (siRNA), which were pre-designed by Ambion (Austin, TX). siRNAs used here to specifically target the human sequence of catalytic domain of MT1-MMP were as follows: forward, 5Ј-CGG AGA AUU UUG UGC UGC CTT-3Ј; reverse, 5Ј-GGC AGC ACA AAA UUC UCC GTG-3Ј. As a negative control, scramble and GAPDH siRNAs (Ambion) were used. Briefly, HT1080 cells were seeded in 12-well plates (8 ϫ 10 4 cells/well) in complete media. The next day, the cells were transiently transfected with several concentrations of the MT1-MMP and control siRNAs using Effectene. Twenty four h posttransfection, the media were diluted 2-fold with fresh serumfree media, and the cells were incubated an additional 24 h before analyses.
Indirect Immunofluorescence Microscopy-Cells grown on 22-mm 2 glass coverslips were rinsed with PBS and acid-washed (5 min on ice) with a buffer composed of 50 mM glycine-HCl, pH 3.0, and 150 mM NaCl to dissociate TIMPs bound to the cell surface. The cells were then washed several times with ice-cold PBS-CM, and nonspecific sites were blocked with 5% normal donkey serum (Jackson ImmunoResearch, West Grove, PA) in PBS-CM buffer. Following several washes with ice-cold PBS-CM, the cells were incubated (1-2 h, 4°C) with 73 g/ml pAb 198 against the catalytic domain of MT1-MMP. After washing with PBS-CM, the cells were incubated (1 h, 4°C) with a 1:50 dilution of FITC-conjugated donkey anti-rabbit IgG antibody (Jackson ImmunoResearch) in PBS supplemented with 1% bovine serum albumin. Following repeated washes with icecold PBS-CM, the cells were fixed with ice-cold 4% paraformaldehyde in PBS, pH 7.4, washed with PBS supplemented with 100 mM glycine and incubated (10 min, room temperature) with DAPI (Molecular Probes, Eugene, OR) for nuclei staining. Finally, the coverslips were mounted with anti-fade reagent (Molecular Probes).
To detect total MT1-MMP (intracellular and surface), the cells were acid-washed as described above and then fixed with ice-cold 4% paraformaldehyde in PBS followed by permeabilization with 0.5% Triton X-100in PBS at room temperature. The cells were then incubated with primary and secondary antibodies as described above with the exception that all of the incubations were performed at room temperature. The nuclei were stained with DAPI, and the coverslips were mounted with 50% glycerol in PBS. Negative controls included cells incubated only with secondary antibody or cells incubated with normal rabbit IgG as a primary antibody. These conditions produced no specific signals (data not shown).
The samples were examined and photographed using a 60ϫ oil-immersion objective with a Zeiss LSM-510 META laser scanning confocal microscope at the Microscopy and Imaging Resources Laboratory at Wayne State University School of Medicine. DAPI was excited with a 745 nm multiphoton laser, and FITC was excited with a 488 nm argon laser. To compare the intensity of the fluorescence signal on the cell surface among the various transfectants, all the parameters for brightness and contrast were held consistent during the experiment. Average intensity of fluorescence signal per cell (as identified by DAPI staining) was measured and analyzed using the Metamorph TM software, version 6.3, revision 2 (Molecular Devices, Sunnyvale, CA). Each experiment was performed in triplicate, and quantification values represent the means of three independent experiments Ϯ S.D. Statistical significance was determined using the Student's t test.
Quantitative Fluorescence-based MT1-MMP Antibody Uptake Assay-This assay was performed as described (54,71). Briefly, HT1080 transfectants (EV, 44-MT1, 44⌬CT, and 44⌬H) were seeded (7 ϫ 10 5 cells per well) on 22-mm 2 glass coverslips placed on 6-well plates. Cells were cooled down for 5 min on ice and acid-washed with 50 mM glycine-HCl, 0.1 M NaCl, pH 3, for 5 min on ice followed by several washes with ice-cold PBS-CM, as described above. The cells were then incubated with 73 g/ml pAb 198 to MT1-MMP in ice-cold PBS-CM for 2 h at 4°C. Negative controls included cells incubated with equal amount of normal rabbit IgG as a primary antibody or only with secondary antibody. These conditions produced no specific signals (data not shown). Endocytosis was then allowed to proceed by shifting the cells to 37°C for 10 min in warm PBS-CM. Endocytosis was stopped by the addition of 0.01% NaN 3 , pH 7.2, in PBS-CM until the end of experiment. Cells were washed several times with ice-cold PBS-CM containing 0.01% NaN 3 , pH 7.2, and then incubated with normal donkey serum (5% in PBS-CM) for 1 h at 4°C. Following several washes with ice-cold PBS-CM at 4°C, the cells were incubated with FITC-conjugated anti-rabbit IgG antibody (1:50 dilution in ice-cold PBS-CM) for 1 h at 4°C. After washing with ice-cold PBS-CM, the cells were fixed with ice-cold 4% paraformaldehyde for 20 min at 4°C and washed with PBS supplemented with 100 mM glycine. The nuclei were stained with DAPI, and the coverslips were mounted with anti-fade reagent (Molecular Probes). In some experiments, the cells were seeded on collagen I-coated plates (50 g collagen I/well) before the antibody uptake assay. Samples were examined and photographed with a Zeiss LSM-510 META laser scanning confocal microscope using a 60ϫ oil-immersion objective. The intensity of the fluorescence signal on the cell surface was analyzed as described above. Each experiment was performed in triplicate, and quantification values represent the means of three independent experiments Ϯ S.D. Statistical significance was determined using the Student's t test.
Reversible Biotinylation Assay-We followed the procedure of Wu et al. (72). To this end, we used HT1080 transfectants (EV and 44-MT1) and parental HT1080 cells seeded in 6-well plates. In the case of HT1080 parental cells, the cells were treated overnight at 37°C with conA (20 g/ml) in the presence or absence of 10 M GM6001 (Chemicon) in serum-free media. The next day, the cells were rinsed and cooled down for 5 min with cold PBS-CM and then biotinylated with 0.5 mg/ml disulfide-cleavable EZ-link sulfo-NHS-SS-biotin (Pierce) for 30 min. Excess biotinylating reagent was quenched with 50 mM NH 4 Cl in PBS-CM for 10 min at 4°C followed by two washes of the cells with cold PBS-CM. To initiate internalization, the wells received pre-warmed serum-free media (3 ml/well), and the plates were immediately incubated at 37°C for various times (0 -120 min). After each time point, the plates were cooled down in ice to halt internalization, and the media were aspirated and the cells washed with cold PBS-CM. Cell surfacebound biotinylating reagent was stripped by incubating the cells (20 min on ice, twice) in 2 ml/well of reducing solution (150 mM NaCl, 1 mM EDTA, 1% bovine serum albumin, 20 mM Tris, pH 8.6, supplemented with 40 mM glutathione). In each experiment, a plate of biotinylated cells was maintained at 4°C for 0 -120 min, as a control of total surface MT1-MMP. These plates labeled "Total" did not receive reducing solution; instead they received PBS-CM. After addition of the reducing solution (endocytosed MT1-MMP) or PBS-CM (total MT1-MMP), the cells were washed twice with cold PBS-CM and lysed with 0.2 ml/well of Nonidet P-40 lysis buffer. The lysates were centrifuged (13,000 rpm, 20 min), and the supernatants were collected. Protein concentration was then determined in each supernatant using the BCA method. An equal amount of protein from each lysate was then incubated (overnight at 4°C) with 50 l of immobilized neutravidin protein beads (Pierce) to pull down biotinylated proteins. The beads were washed four times with harvest buffer (0.5% SDS, 60 mM Tris/HCl, pH 7.5, 2 mM EDTA) supplemented with 2.5% Triton X-100 (final concentration). The bound biotinylated proteins were eluted with reducing Laemmli SDS sample buffer, boiled, and resolved by 10%, reducing SDS-PAGE followed by transfer to a nitrocellulose membrane. The total and endocytosed MT1-MMP forms were detected with pAb 815 to the hinge region of MT1-MMP. The blots were then reprobed with a mAb to human transferrin receptor. Densitometric analysis of the bands of internalized active MT1-MMP in the same blot was performed with the US-SCAN-IT program (Silk Scientific Inc., Orem, UT). The density (area per square unit) was plotted against time and expressed in arbitrary units.
Tumorigenicity and Experimental Metastasis Assays-Tumorigenicity assays were performed on 4 -6-week-old male homozygous CB-17 SCID/SCID mice and female NCr nude mice (Taconic Farms, Germantown, NY) according to the Animal Welfare Regulations at Wayne State University. HT1080 cell transfectants (pool populations and pooled clones) were harvested and resuspended in serum-free DMEM. Each mouse was inoculated subcutaneously with 5 ϫ 10 5 cells in 100 l of serum-free DMEM. The volume of the tumor was measured twice a week with a caliper, using the formula V ϭ (LW 2 ) ϫ 0.4 (where V is the volume (mm 3 ); L is the longest diameter (mm), and W is the shortest diameter (mm)). Data were analyzed for statistical significance by unpaired t test with Welch correction using the GraphPad InStat version 3.0 (GraphPad Software, San Diego). The differences were considered to be statistically significant at p Ͻ 0.05. Tumor incidence represents the percentage of mice bearing a tumor larger than 100 mm 3 at day 28. The mice were euthanized 4 weeks after tumor cell inoculation, and the tumors were harvested and frozen at Ϫ80°C until use. Experimental metastasis assays were conducted in CB-17 SCID/SCID mice. Briefly, each mouse was inoculated in the tail vein with 1 ϫ 10 6 HT1080-EV or HT1080-44-MT1 cells (pooled populations) suspended in 100 l of serum-free DMEM. Twelve days later the mice were sacrificed, and the lungs were harvested for assay of metastatic burden by real time PCR and counting of lung colonies in hematoxylin and eosin sections as described below.
Determination of Lung Metastatic Burden-Twelve days after tail vein cell inoculation, the mice were sacrificed and the lungs were harvested. Equal portions from each lung were snapfrozen in liquid nitrogen or formalin-fixed. Frozen lung tissues were used for quantitative analyses of human alu sequences by quantitative real time PCR, as described (73). Briefly, genomic DNA was extracted from harvested mouse lung tissues using the Puregene DNA purification kit (Genta Systems, Minneapolis, MN). The primers specific for the human alu sequences were as follows: sense, 5Ј-ACG CCT GTA ATC CCA GCA CTT-3Ј, and antisense, 5Ј-TCG CCC AGG CTG GAG TGC A-3Ј. Real time PCR was performed using the SYBR Green QRT-PCR master mix kit according to the manufacturer's protocol (Stratagene). PCR conditions included polymerase activation at 95°C for 10 min followed by 40 cycles at 95°C for 30 s, 63°C for 30 s, and 72°C for 30 s. Each assay included a negative control (diethyl pyrocarbonate water, Ambion), a positive control (human genomic DNA), a no-template control, and the experimental samples in duplicate. A quantitative measurement of mouse DNA was obtained through amplification of the mouse GAPDH genomic DNA sequence with mouse GAPDH primers (sense, 5Ј-AAC GAC CCC TTC ATT GAC-3Ј, and antisense, 5Ј-TCC ACG ACA TAC TCA GCA C-3Ј) using the same PCR conditions described for alu. The threshold cycle (Ct) of each sample was recorded as a quantitative measure of the amount of PCR product in the sample using a Stratagene Mx 4000 TM PCR machine. The alu signal was normalized against the relative quantity of mouse GAPDH and expressed as ⌬Ct ϭ (Ct alu Ϫ Ct GAPDH ). The changes in alu signal relative to the total amount of mouse genomic DNA were expressed as ⌬⌬Ct ϭ ⌬Ct treatment Ϫ ⌬Ct control . Relative changes in metastasis were then calculated as 2 Ϫ⌬⌬Ct . Data processing and statistical analyses were performed using Microsoft Excel (Microsoft Corp., Redmond, WA).
Formalin-fixed lung tissues were sectioned and processed for hematoxylin and eosin staining using conventional procedures. Lung colonies were visualized and counted in serial sections under a light microscope. At least three different slides from each tumor and five different tumors from each cell line were evaluated for the number of metastatic colonies.

RESULTS
Characterization of Transfectants-To investigate the relationship between active MT1-MMP (57 kDa) and its processed 44-kDa form in regulation of MT1-MMP activity, we engineered a cDNA construct that produces a 44-kDa recombinant protein extending from Gly 285 to Val 582 when expressed in cells (Fig. 1A). We also generated two deletion mutants lacking either the cytosolic tail (44⌬CT) or the entire hinge region (44⌬H), as described under "Experimental Procedures." Purposely, none of the constructs included N-or C-terminal tags to avoid any possible alterations in protein conformation (Fig.  1A). The role of the 44-MT1 constructs on MT1-MMP activity was investigated in four human cancer cell lines (HT1080, MDA-MB-231, BT549, and U87 cells), which express endogenous MT1-MMP. HT1080, MDA-MB-231, and U87 cells were stably transfected to constitutively express 44-MT1, and BT549 cells were transfected with an inducible vector to express 44-MT1 in response to ponasterone A treatment. HT1080 cells and U87 cells were also transfected to stably express 44⌬CT and 44⌬H. T47D breast carcinoma cells, which have no detectable levels of endogenous MT1-MMP, were transfected with wild-type MT1-MMP or 44-MT1, and thus they were used to examine the effects of MT1-MMP and 44-MT1 independently of each other. BS-C-1 cells were used to transiently express wild-type MT1-MMP with or without 44-MT1 using the vaccinia expression system (74). The expression of the recombinant MT1-MMP proteins in the various cell types and their localization on the cell surface were confirmed by immunoblot analyses (Fig. 1, B-D, only HT1080, BT549, and BS-C-1 cells are shown) and surface biotinylation (supplemental Fig. 1A, only HT1080, BT549, and U87 cells are shown), respectively. All of the recombinant 44-kDa forms were detected on the cell surface. In the case of 44-MT1, its level of expression in the transfectants did not differ much from the levels of the 44-kDa species generated after stimulation of parental HT1080 cells with conA, as shown in supplemental Fig. 1B or observed in other studies with cells stimulated with a variety of factors (24,30,40,50). Basal levels of endogenous MT1-MMP (57 kDa) in the various transfectants did not show significant changes when compared with control (EV) cells, when determined by immunoblot analyses (Fig. 1, B and C) or surface biotinylation (supplemental Fig. 1A). Consistently, no changes in MT1-MMP mRNA levels were detected by semiquantitative RT-PCR (data not shown). In addition, HT1080 transfectants did not differ in expression of any other mRNAs of MT-MMPs as well as MMP-1, MMP-2, MMP-9, and MMP-13 mRNAs. Also, levels of all four TIMPs mRNA did not vary among stable pools and clones (data not shown).
Collagen I was shown to promote pro-MMP-2 activation in HT1080 cells (30). As reported, HT1080-EV cells on collagen I exhibited a partial but consistent activation of pro-MMP-2, mostly to the intermediate form, when compared with cells seeded on uncoated dishes (Fig. 2C). In contrast, full pro-MMP-2 activation was detected in HT1080-44-MT1 cells cultured on collagen I-coated dishes. Thus, 44-MT1 expression is associated with enhanced pro-MMP-2 activation even in the presence of collagen I.
Previous studies demonstrated that expression of an N-terminally tagged 44-kDa form inhibits MT1-MMP-mediated pro-MMP-2 activation (57,59,76). These species, as opposed to 44-MT1, lack significant parts of the hinge region, which may affect the experimental outcome (see "Discussion"). To address this possibility, we deleted the entire hinge (Gly 285 -Gly 315 ) of 44-MT1 to generate 44⌬H (Fig. 1A) and expressed this form in HT1080 cells (Fig. 1B). As shown in Fig. 2B, two clones expressing 44⌬H did not exhibit constitutive (unstimulated) activation of pro-MMP-2. In fact, 44⌬H expression partially inhibited conA-induced pro-MMP-2 activation, and only the intermediate form of MMP-2 was detected, when compared with HT1080-EV cells (Fig. 2B). Thus, removal of the hinge region is associated with a negative effect of 44-MT1 on conAstimulated pro-MMP-2 activation. Together, these results show that 44-MT1 expression (extending from Gly 285 to Val 582 ) can promote unstimulated pro-MMP-2 activation in HT1080 cells, and this effect is reduced by the absence of the hinge region.
To determine whether the constitutive stimulation of pro-MMP-2 activation observed in HT1080-44-MT1 cells is MT1-MMP-dependent, we used RNA interference to specifically knock down endogenous MT1-MMP expression without affecting the expression of 44-MT1. As shown in Fig. 2I  44-MT1 expression was barely affected. Zymography of the media revealed that the MT1-MMP siRNA-transfected HT1080-44-MT1 cells lacked activation of pro-MMP-2 (Fig. 2I,  upper panel). In contrast, constitutive pro-MMP-2 activation was evident in scrambled siRNA-transfected HT1080-44-MT1 cells. These results suggest that the enhanced activation of pro-MMP-2 by 44-MT1 is mediated by interactions between active and processed MT1-MMP. Consistent with this assertion, expression of 44-MT1 in T47D cells, which do not express endogenous MT1-MMP, did not result in pro-MMP-2 activation (Fig. 2H). However, active MMP-2 was readily detected in T47D cells transfected to express wild-type MT1-MMP, as expected (Fig. 2H).
Because activation of pro-MMP-2 depends on the level of TIMP-2 (6), we measured TIMP-2 mRNA and protein in the HT1080 transfectants. All of the transfectants exhibited similar levels of TIMP-2 mRNA (supplemental Fig. 1D). However, ELISA data showed a differential distribution of TIMP-2. In HT1080-44-MT1 and HT1080-44⌬CT cells, TIMP-2 was mostly associated with the cell lysate, whereas in HT1080-EV and HT1080-44⌬H cells most of the TIMP-2 was present in the conditioned media (supplemental Fig. 1C). The differential distribution of TIMP-2 observed in these transfectants correlated with their pattern of pro-MMP-2 activation (Fig. 2B). Because the cell association of TIMP-2 is mainly mediated by MT1-MMP, these results suggest that different levels of surface active MT1-MMP among the transfectants may explain pro-MMP-2 activation results, as is shown below.
We next examined the effects of 44-MT1 on pro-MMP-2 activation using a ponasterone-inducible system in BT549 cells, which express endogenous MT1-MMP and pro-MMP-2, and in BS-C-1 cells infected-transfected to express MT1-MMP with or without 44-MT1. These expression systems allow functional evaluation of the gene of interest without issues of selection. Exposure of BT549 cells to ponasterone induced expression of 44-MT1 (Fig. 1C), which resulted in pro-MMP-2 activation (Fig. 2D). Likewise, co-expression of wild-type MT1-MMP with 44-MT1 in BS-C-1 cells also induced activation of pro-MMP-2, which was more evident at higher levels of 44-MT1 protein ( Fig.  1D and Fig. 2E).
Treatment of MDA-MB-231-EV and U87-EV cells with conA induced pro-MMP-2 activation, as expected (Fig. 2, F and  G). However, expression of 44-MT1 in these cells had no effect on activation of pro-MMP-2 regardless of conA treatment (Fig.  2, F and G). Furthermore, expression of 44⌬CT or 44⌬H in U87 cells had no effect on conA-stimulated pro-MMP-2 activation when compared with U87-EV cells (Fig. 2G).
44-MT1 Effect on Cell Growth within Three-dimensional Collagen I-MT1-MMP was described previously as conferring tumor cells with the ability to grow within a three-dimensional collagen I matrix (77). Therefore, we examined the effects of 44-MT1 expression on the proliferation of HT1080 and U87 cells within three-dimensional collagen I. As shown in Fig. 3, both HT108-44-MT1 and U87-44-MT1 cells exhibited enhanced growth within three-dimensional collagen I when compared with control cells. In contrast, 44-MT1 expression had no detectable effect on the rate of growth of MDA-MB-231 cells in three-dimensional collagen I (Fig. 3). Expression of 44⌬CT or 44⌬H in HT1080 or U87 cells had no significant effect on cell growth when compared with EV cells (Fig. 3), further demonstrating that both the cytosolic domain and the hinge region mediate the biological effect of 44-MT1. Next, we treated HT1080-44-MT1 cells growing within three-dimensional collagen I with either TIMP-2 or TIMP-1. As shown in Fig. 3, TIMP-2 but not TIMP-1 was associated with inhibited growth of HT1080-44-MT1 cells, consistent with this growth effect mediated by MT1-MMP. Thus, 44-MT1 expression promotes growth of two cell lines within a three-dimensional collagen I matrix. No differences in cell growth were detected when the HT1080 or U87 cell transfectants were cultured on uncoated dishes or on a two-dimensional collagen I matrix (data not shown).
Expression of 44-MT1 in HT1080 Cells Correlates with High Levels of Active MT1-MMP on the Cell Surface-The enhanced MT1-MMP activity associated with 44-MT1 expression, and the higher levels of cell-associated TIMP-2 in HT1080 cells raised the possibility that these effects were in part mediated by an increased level of surface MT1-MMP, which could not be resolved by conventional surface biotinylation. We therefore examined HT1080 transfectants for expression of surface-active MT1-MMP by indirect immunofluorescence. To detect only the endogenous active enzyme, we used pAb 198, which recognizes the catalytic domain of MT1-MMP (68) and does not react with 44-MT1. To increase the level of detection, the cells were briefly acid-washed to dissociate any potential TIMP bound to the active site of MT1-MMP. As shown in Fig. 4, nonpermeabilized HT1080-44-MT1 cells displayed prominent labeling of active MT1-MMP at the plasma membrane when compared with control cells or cells expressing 44⌬CT or 44⌬H. Quantification of microscopy data showed a 4-fold increase in active MT1-MMP in HT1080-44-MT1 cells when compared with HT1080-EV cells (Fig. 4A, surface). In contrast, no significant differences in surface MT1-MMP were found between HT1080-EV, HT1080-44⌬CT, and HT1080-44⌬H cells (Fig. 4, A and B, surface). Immunofluorescence of permeabilized cells with pAb 198 showed that all the transfectants expressed similar levels of total MT1-MMP protein (Fig. 4, A  and B, total). Incubation of cells with normal rabbit IgG as a primary antibody produced no positive signal (data not shown).
Expression of 44-MT1 Delays the Internalization of MT1-MMP in HT1080 Cells-To examine whether the increase in surface MT1-MMP expression in HT1080-44-MT1 cells was because of a reduced rate of endocytosis, we monitored and quantified the percentage of active MT1-MMP remaining at the cell surface following antibody cross-linking mediated internalization. We compared the relative amounts of active MT1-MMP remaining at the cell surface after 10 min of incubation at 37°C with that of cells maintained at 4°C. Note that comparisons of fluorescence signal (remaining active MT1-MMP) were conducted within the same cell transfectant because the initial levels of surface MT1-MMP varied between EV and 44-MT1 cells, as shown in Figs. 4 and 5B. Fig. 5B shows the actual differences in signal between HT1080-EV and HT1080-44-MT1 cells at 0 min when imaged under identical contrast and brightness conditions. Fig. 5A shows confocal images of cells during the antibody uptake assay photographed using identical contrast and brightness conditions in each group, which reveals the difference in levels of surface MT1-MMP between 0 and 10 min within the same cell transfectant. Thus, the differences in signal (surface MT1-MMP) between control and 44-MT1-expressing cells are not seen in Fig. 5A. Quantitative analyses of surface fluorescence of confocal images (Fig. 5A) showed that nearly 15% of the MT1-MMPbound antibody remained at the cell surface of HT1080-EV cells after 10 min of incubation at 37°C when compared with cells maintained at 4°C (Fig. 5C, p Ͻ 0.01). In contrast, ϳ40% of active MT1-MMP remained at the surface of HT1080-44-MT1 cells after 10 min at 37°C when compared with the same cells maintained at 4°C (Fig. 5C, not statistically significant relative to 4°C). Under the same experimental conditions, HT1080-44⌬CT and HT1080-44⌬H cells had ϳ22% (p Ͻ 0.05) and ϳ10% (p Ͻ 0.01) of MT1-MMP remaining at the cell surface after 10 min at 37°C, respectively, when compared with cells maintained at 4°C (0 min). The relative amount of active MT1-MMP remaining on the surface of these cells did not differ from that remaining on HT1080-EV cells. Thus, expression of 44-MT1 in HT1080 cells is associated with a reduced internalization of active MT1-MMP. This effect is not seen with expression of 44⌬CT or 44⌬H and in HT1080-EV cells indicating that deletion of the cytosolic tail or the hinge region abolishes the ability of 44-MT1 to delay the internalization of active MT1-MMP.
Because collagen I was shown to regulate MT1-MMP internalization (54,61), the antibody uptake assay was also conducted in HT1080 cell transfectants seeded on collagen I-coated coverslips, as described under "Experimental Procedures." These data also showed a delayed internalization of active MT1-MMP in HT1080-44-MT1 cells when compared with control EV cells or cells expressing 44⌬CT or 44⌬H (supplemental Fig. 2). These results indicate that the process of MT1-MMP regulation by endocytosis is not influenced by the presence of collagen I.
We also followed the internalization of MT1-MMP using a reversible biotinylation assay (72). This assay was performed using a cleavable, membrane-permeable biotin. After incubations at 37°C, residual surface biotin was stripped with gluta- thione, a membrane-impermeable reducing agent, and internalized biotinylated MT1-MMP was captured by avidin beads and detected by immunoblot analyses. Fig. 6A shows a representative blot depicting the endocytosed and total biotinylated active endogenous MT1-MMP and 44-MT1 form in the HT1080 transfectants. The data of Fig. 6B depict the densitometric analysis of internalized active MT1-MMP as a function of time. Using this approach, we observed more endocytosed active MT1-MMP (glutathione-resistant) in HT1080-EV cells when compared with HT1080-44-MT1 cells under the same experimental conditions. These data further demonstrated a delayed internalization of active MT1-MMP as a function of 44-MT1 expression, in agreement with the antibody uptake assay.
To further examine the relationship between natural active and processed MT1-MMP in regulation of MT1-MMP endocytosis, we used parental HT1080 cells treated with conA to induce processing of endogenous MT1-MMP. In addition, the cells were treated with or without GM6001, a broad spectrum MMP inhibitor, to inhibit conversion of the active MT1-MMP to the 44-kDa species. Endocytosis of MT1-MMP was then fol-lowed by reversible biotinylation (Fig. 6, C and D). This study showed that autocatalytic processing of MT1-MMP to the 44-kDa species (in the absence of GM6001) is associated with reduced endocytosis of active MT1-MMP. Under the same conditions, inhibition of processing by GM6001 correlated with enhanced MT1-MMP internalization (Fig. 6, C and D). These results suggest a direct relationship and a balance between processing (and consequently generation of the 44-kDa species) and endocytosis, which serves to maintain the pool of active enzyme at the cell surface. It should be noted, however, that under these experimental conditions the relative contribution of processing and the 44-kDa species on active MT1-MMP internalization could not be separated because they are mutually dependent. However, the data with the HT1080-44-MT1 transfectants (Fig. 5 and Fig.  6A) demonstrate the role of the 44-kDa species on MT1-MMP internalization independently of the processing event. In that sense, the studies with the HT1080 parental and transfectant cells are complementary and support the notion of a direct effect of the 44-kDa species on the rate of internalization of active MT1-MMP.
Effect of 44-MT1 Forms on Tumor Growth and Metastasis-The HT1080 cell transfectants were inoculated subcutaneously in athymic mice, and tumor growth was monitored for 28 days. Initially we used pooled populations and SCID mice (Fig. 7, A and B). These studies showed that HT1080-44-MT1 cells formed significantly larger tumors than HT1080-EV cells (Fig. 7, A and B). In contrast, expression of 44⌬CT had no significant effect on tumor growth. Surprisingly, we found that 7/8 mice inoculated with HT1080-44⌬H cells showed no evidence of tumor formation (Fig. 7, A and B). We repeated this experiment using a mixture of two clones of each transfectant, which were inoculated in NCr mice to account for potential differences because of clonal variation and mouse strain. These studies essentially reproduced the previous results with the exception that the HT1080-44⌬CT clones grew faster than control clones (Fig. 7, C and D). Taken together, these results demonstrate that although 44-MT1 expression is associated with enhanced growth of HT1080 cells in mice, expression of 44⌬H is strongly correlated with decreased tumorigenic potential. These results suggest that the presence of the hinge regulates the activity of the processed form of MT1-MMP in vivo. We next examined the HT1080 transfec- tants for differences in their ability to form lung colonies after intravenous inoculation in SCID mice. Mice were sacrificed 12 days after cell inoculation and lungs harvested, and tumor burden was determined by a quantitative real time PCR of genomic DNA using human alu sequences and mouse GAPDH as an internal control (73). As shown in Fig. 8A, the results of alu PCR indicated that mice inoculated with HT1080-44-MT1 cells contained significantly greater amounts of human genomic DNA in their lungs when compared with lungs of mice injected with HT1080-EV cells. As observed with the subcutaneous tumor assay, expression of 44⌬H also inhibited lung colonization after intravenous inoculation, as determined by alu PCR (Fig. 8A). These results were confirmed by counting tumor colonies in the lungs of mice (Fig. 8B). Thus, 44-MT1 expression in HT1080 cells is associated with enhanced tumor growth and lung colonization.

DISCUSSION
Once on the cell surface, the pool of active MT1-MMP is regulated by a complex array of processes, including autocatalytic degradation (78). This process leads to the cleavage of active MT1-MMP to an inactive 44-kDa membrane-tethered form lacking the catalytic domain but retaining key enzyme domains, and thus has the potential to directly and indirectly influence the function of neighboring active MT1-MMP mole-cules. Here we provide evidence that the product of MT1-MMP processing, the 44-kDa species, may regulate the amount of active enzyme at the cell surface and consequently influence enzymatic activity and cell behavior both in vitro and in vivo. Our data show that 44-MT1 expression is generally associated with enhanced pro-MMP-2 activation and growth in three-dimensional collagen I in various cell types. It also correlates with enhanced tumor growth and lung colonization of HT1080 cells in immunodeficient mice. In vitro, the stimulatory action of 44-MT1 was dependent on the interactions between 44-MT1 and active MT1-MMP because inhibition of endogenous MT1-MMP expression or activity abolished the observed effects. Furthermore, in the absence of endogenous MT1-MMP expression, such as in T47D cells, 44-MT1 expression had no effects on pro-MMP-2 activation and growth in three-dimensional collagen. In vivo, however, the contribution of an MT1-MMP-independent effect of 44-MT1 on tumor growth and metastasis cannot be ruled out considering recent evidence showing the ability of catalytically inactive MT1-MMP to promote tumor growth in a process that was suggested to involve the cytoplasmic tail (79). Here we found that HT1080 cells expressing 44⌬CT produced smaller tumors than cells expressing 44-MT1, consistent with a role of this domain in vivo.
Not all MT1-MMP-mediated activities were equally affected by expression of 44-MT1, and its effects were dependent on the structural organization of the 44-kDa form, the type of activity assayed, and the cell type, just to mention a few. At present, the reasons why different cell types respond differently to 44-MT1 expression in selected activities are unclear. Regardless, in none of the cellular systems used here did we observe a dominant negative effect of 44-MT1 on MT1-MMP function, including in vivo systems. Contrary to our findings, previous studies showed that expression of a recombinant 44-kDa form inhibited MT1-MMP activity (57,59,60). This inhibitory function was ascribed to a competitive interference of the hemopexin and cytosolic domains of the 44-kDa form with the tendency of MT1-MMP to undergo homophilic interactions (1,57,59,62) and to cleave collagen I (60,76). According to these studies, the 44-kDa species acts in a manner consistent with being a dominant negative regulator of MT1-MMP activity. The results presented here, however, do not support the universality of this model because pro-MMP-2 activation, growth within collagen I, in vivo tumor growth, and lung metastasis were not disrupted by 44-MT1 expression. Consequently, our observations do not support the notion that homophilic interactions of active MT1-MMP are critical for enzymatic activity, if indeed a function of the 44-kDa form is to disrupt these interactions via its hemopexin-like domain and/or the cytosolic tail, as proposed earlier (57-59, 62, 76). In fact, our data show that the processed form of MT1-MMP, which only lacks the catalytic domain, does not disrupt enzyme activity. Interestingly, Hotary et al. (80) showed that MT1-MMP devoid of the hemopexin-like domain confers cells with the ability to degrade basement membranes suggesting that this domain is dispensable for this activity of MT1-MMP. Nevertheless, even if 44-MT1 forms were to disrupt homodimeric-oligomeric complexes of active MT1-MMP, the evidence presented here would suggest a lesser importance of these complexes for enzyme function than previously considered (1). Possibly, dimeric and oligomeric forms of MT1-MMP only represent a minor fraction of the pool of active enzyme on the cell surface (59), and thus their presence may not be essential for proper enzyme function.
Why do our results differ from those published earlier? The majority of the studies reporting a dominant negative effect of the 44-kDa form used recombinant 44-kDa-like proteins lacking most of the hinge region. In addition, these forms included either a FLAG or a 25-kDa glutathione S-transferase tag at the N terminus (57)(58)(59)76). The FLAG-tagged, hinge-deleted 44-kDa species was found to be processed to a smaller form (57), possibly because of the lack of O-glycosylation, as discussed below (25,36). The glutathione S-transferase-tagged 44-kDa construct (59) lacked the signal sequence and the prodomain of MT1-MMP, which are considered to be essential for proper trafficking and membrane insertion. Thus, significant deletions within the hinge region and/or presence of N-terminal tags may affect the function of the recombinant 44-kDa protein and consequently its effects on active MT1-MMP. Consistent with this possibility, we found that deletion of the hinge in 44-MT1 was associated with inhibition of conAstimulated pro-MMP-2 activation and tumorigenicity in HT1080 cells. Interestingly, 44⌬H had no effect on growth within the three-dimensional collagen I indicating a selective effect of this fragment on cell behavior. The fact that 44⌬H lacks the hinge region suggests that this domain may play a role in regulating the function of the 44-kDa processed form. Accumulating evidence suggests that the hinge of MT1-MMP plays an important role in enzyme activity and stability because of its high mobility and O-glycosylation, respectively (25,36,60,81). Indeed, 44-MT1 is O-glycosylated, whereas 44⌬H is not (data not shown). Interestingly, a FLAGtagged 44-kDa species lacking the hinge, which inhibited pro-MMP-2 activation in HT1080 cells, was found to be processed to a smaller form (57), possibly because of the lack of O-glycosylation. Because the hinge is the main target of autocatalytic processing (81), its presence in the 44-kDa form may also contribute to the stabilization of active MT1-MMP by interfering with processing, as proposed previously (59). On the other hand, significant deletions within the hinge region and/or presence of N-terminal tags may cause conformational or structural changes in the 44-kDa protein, which in turn may disrupt active MT1-MMP-44-kDa interactions leading to inhibition of enzyme function. Although the mechanism(s) by which these hinge-truncated 44-kDa species inhibit MT1-MMP activity is unclear, it is worth noting that under physiological processing conditions, the resultant major 44-kDa form would include the hinge region. We posit that this species should be expected to behave similarly to the recombinant 44-MT1 form used here. Regardless, our results suggest a novel role for the hinge in regulating the function of the 44-kDa form of MT1-MMP. In addition to differences in the nature of the 44-kDa form as discussed above, careful analyses of previous data suggest that use of cell lines with low levels of endogenous MT1-MMP expression (59,60,62) with highly expressed recombinant 44-kDa proteins may also contribute to the described dominant negative effects.
Our data indicate that expression of 44-MT1 in HT1080 cells correlates with higher levels of surface MT1-MMP because of a slower rate of MT1-MMP endocytosis, as determined in two distinct assays. The relatively higher levels of surface-active MT1-MMP correlated with higher levels of TIMP-2 associated with the cell lysates of HT1080-44-MT1 cells (supplemental Fig. 1C). Together, these findings suggest that generation of the 44-kDa form by autocatalytic turnover serves to preserve a basal level of functional protease at the cell surface by regulating the rate of endocytosis of active MT1-MMP. Given that 44-MT1 maintains the cytosolic tail and thus possesses the same endocytic signal of the full-length protease, this process may be due to competitive interactions between the cytosolic tails of the MT1-MMP forms for components of the endocytic machinery (82). Indeed, expression of 44⌬CT had no effect on the rate of active MT1-MMP internalization. This suggests the possibility that components of the endocytic pathway, including those mediated by di-leucine motifs, as present in the cytosolic domain of MT1-MMP, are saturable and thus prone to competition within the same type of receptors, as established in previous studies with several transmembrane receptors (82)(83)(84). Competition for the endocytotic machinery between MT1-MMP species may be due to the accumulation of the 44-kDa form as a consequence of autocatalytic processing. Alternatively, preferential internalization of the processed form cannot be ruled out. At present, the data presented here do not allow us to discriminate between these possibilities. This will require determining the rate of internalization of each species alone and in combination. However, this study is complicated by the need to express comparable levels of each species in cells without endogenous MT1-MMP and to prevent processing of the wild-type enzyme, which further complicates this type of analyses. Nevertheless, our data strongly suggest that there is a direct relationship between processing and endocytosis which serves to maintain a viable level of protease at the cell surface.
Based on the results presented here, we propose a model for the regulation of active MT1-MMP by its processed form. Conditions that increase the level of active MT1-MMP at the cell surface augment the probability of autocatalytic processing (78). This leads to the formation of a membrane-anchored 44-kDa degradation product (Gly 285 -Val 582 ) that lacks the entire catalytic domain. Processing decreases the amount of active protease, and the overall activity of MT1-MMP at the cell surface is reduced. The net amount of active MT1-MMP can be recovered by de novo synthesis, enhanced trafficking, and/or reduced turnover and endocytosis. MT1-MMP processing also leads to accumulation of the 44-kDa product. Consequently, the 44-kDa form may saturate the local endocytic pathway resulting in delayed internalization of active MT1-MMP, which would eventually lead to recovery and preservation of enzymatic activity. Although the precise mechanism(s) by which the 44-kDa species delays the internalization of active MT1-MMP needs to be elucidated, this model provides a plausible explanation to the positive association between high levels of processed MT1-MMP (43-45-kDa forms) and increased enzymatic activity reported in many studies. We also propose that the regulation of active MT1-MMP by its processed 44-kDa form is also analogous to the stabilization model of active MT1-MMP by TIMPs and synthetic MMP inhibitors, which posits that under certain conditions these can also promote MT1-MMP activity by inhibiting autocatalytic processing (17,85). The findings of this study should facilitate the re-evaluation of the contribution of MT1-MMP autocatalytic processing and its degradation product to enzyme function.