Phosphorylation of Membrane Type 1-Matrix Metalloproteinase (MT1-MMP) and Its Vesicle-associated Membrane Protein 7 (VAMP7)-dependent Trafficking Facilitate Cell Invasion and Migration*

Background: Intracellular trafficking of MT1-MMP is essential for its role in tumor cell invasion. Results: Mutation of Thr567 in MT1-MMP altered its internalization and recycling and associated biochemical signaling. Conclusion: Phosphorylation of MT1-MMP at Thr567 regulates its intracellular trafficking, which is coupled to integrin trafficking and ERK phosphorylation. Significance: Phosphorylation of MT1-MMP may be a regulatory point for control of tumor cell invasion. In multicellular organisms, uncontrolled movement of cells can contribute to pathological conditions, such as multiple sclerosis and cancer. In highly aggressive tumors, the expression of matrix metalloproteinases (MMPs) is linked to the capacity of tumor cells to invade surrounding tissue and current research indicates that the membrane-anchored membrane type 1-matrix metalloproteinase (MT1-MMP) has a central role in this process. Endocytosis and trafficking of MT1-MMP are essential for its proper function, and here we examine the phosphorylation, internalization, and recycling of this enzyme, and the associated biochemical signaling in HeLa and HT-1080 fibrosarcoma cells. Activation of protein kinase C with phorbol 12-myristate 13-acetate resulted in phosphorylation of endogenous MT1-MMP at Thr567 in vivo. Mutation of Thr567 to alanine (to mimic non-phosphorylated MT1-MMP) reduced internalization of MT1-MMP, whereas mutation of Thr567 to glutamic acid (to mimic phosphorylation) resulted in decreased levels of MT1-MMP on the cell surface. The endosomal trafficking and recycling of MT1-MMP was found to be dependent upon Rab7 and VAMP7, and blocking the function of these proteins reduced cell migration and invasion. Intracellular trafficking of MT1-MMP was observed to be coupled to the trafficking of integrin α5 and phosphorylation of ERK that coincided with this was dependent on phosphorylation of MT1-MMP. Together, these results reveal important roles for MT1-MMP phosphorylation and trafficking in both cell signaling and cell invasion.


In multicellular organisms, uncontrolled movement of cells can contribute to pathological conditions, such as multiple sclerosis and cancer. In highly aggressive tumors, the expression of matrix metalloproteinases (MMPs) is linked to the capacity of tumor cells to invade surrounding tissue and current research indicates that the membrane-anchored membrane type 1-matrix metalloproteinase (MT1-MMP) has a central role in this
process. Endocytosis and trafficking of MT1-MMP are essential for its proper function, and here we examine the phosphorylation, internalization, and recycling of this enzyme, and the associated biochemical signaling in HeLa and HT-1080 fibrosarcoma cells. Activation of protein kinase C with phorbol 12-myristate 13-acetate resulted in phosphorylation of endogenous MT1-MMP at Thr 567 in vivo. Mutation of Thr 567 to alanine (to mimic non-phosphorylated MT1-MMP) reduced internalization of MT1-MMP, whereas mutation of Thr 567 to glutamic acid (to mimic phosphorylation) resulted in decreased levels of MT1-MMP on the cell surface. The endosomal trafficking and recycling of MT1-MMP was found to be dependent upon Rab7 and VAMP7, and blocking the function of these proteins reduced cell migration and invasion. Intracellular trafficking of MT1-MMP was observed to be coupled to the trafficking of integrin ␣5 and phosphorylation of ERK that coincided with this was dependent on phosphorylation of MT1-MMP. Together, these results reveal important roles for MT1-MMP phosphorylation and trafficking in both cell signaling and cell invasion.
The ability of cells to invade into and migrate through their surrounding environment is a crucial component in the development and homeostasis of all multicellular organisms and can be a primary contributor to the progression of disease (1). The extracellular matrix (ECM) 3 acts as both a supportive scaffold and a physical barrier, maintaining tissue architecture (2), and degradation of ECM proteins disrupts this structure, enabling cell invasion through the matrix. Matrix metalloproteinases (MMPs) constitute a family of zinc-dependent proteolytic enzymes that function to degrade the ECM (3,4). MMP family members can be subdivided into membrane type MMPs and soluble MMPs. Membrane type MMPs contain either a transmembrane domain or glycosylphosphatidylinositol anchors, which tether them to the plasma membrane (5), whereas soluble MMPs are secreted by cells into the extracellular space, where they can diffuse into the ECM.
The expression of MMPs has been intimately linked with highly aggressive cancers, such as ovarian, melanoma, breast, and colorectal (6 -9), and recent research indicates that the membrane-anchored membrane type 1 matrix metalloproteinase (MT1-MMP) has a central role in the aggressive behavior of some tumor types (2,8,10,11). MT1-MMP can directly degrade the ECM by proteolytic cleavage of collagen types I, II, and III, laminins 1 and 5, fibronectin, vitronectin, fibrin, and aggrecan and can indirectly degrade the ECM through activating pro-MMP-2 (12). Activation of MT1-MMP is accomplished intracellularly by the Golgi-associated enzyme furin, which recognizes the unique amino acid sequence (RRKR) located between the prodomain and catalytic domain (7,12). Once activated, MT1-MMP is trafficked to the plasma membrane through a Rab8-dependent pathway (13). At the plasma membrane, MT1-MMP-mediated proteolytic degradation of ECM proteins facilitates cell invasion. The proteolytic activity of MT1-MMP can be attenuated through internalization, which occurs through both clathrin-mediated and caveolinmediated pathways (14,15). It has also been demonstrated that internalized MT1-MMP can be recycled back to the plasma membrane (5). Importantly, deletion of the cytoplasmic tail of MT1-MMP inhibits internalization, and this can perturb cell * This work was supported by the Natural Sciences and Engineering Research invasion, demonstrating the importance of recycling for proper enzyme function (16).
Trafficking of cellular material between the plasma membrane and intracellular compartments is dependent upon SNARE (soluble N-ethylmaleimide-sensitive factor-activating protein receptor) proteins, which facilitate the fusion of vesicles with target membranes. Several recent studies support a role for SNAREs in MT1-MMP localization; SNAP23, syntaxin 13, VAMP3 (17), TI-VAMP/VAMP7 (18), and syntaxin4 (19) have been shown to be involved in trafficking of MT1-MMP and cell invasion. These functions of SNAREs complement their established roles in cell-ECM adhesion and migration (20 -22).
Having characterized the requirement for SNAREs in trafficking of MT1-MMP and efficient tumor cell invasion, in the present study we examine the internalization of MT1-MMP and its trafficking from endosomal compartments back to the plasma membrane. We report that MT1-MMP is phosphorylated at Thr 567 in a protein kinase C-dependent manner and that this phosphorylation is required for MT1-MMP internalization. Once internalized, MT1-MMP is recycled through the late endosome in a VAMP7-dependent manner. Inhibiting either the internalization or the recycling of MT1-MMP significantly reduced the ability of cells to invade through a Matrigel matrix. Interestingly, mutation of Thr 567 to alanine resulted in increased cell surface expression of MT1-MMP but impaired cellular invasive and migratory abilities. This suggests that the internalization and recycling of MT1-MMP functions to promote invasion and migration through mechanisms beyond ECM degradation. Consistent with this, we found evidence that trafficking/recycling of MT1-MMP is coupled to trafficking of ␣5 integrin and stimulates phosphorylation of ERK.
Immunofluorescence Microscopy-Cells were grown on glass coverslips coated with fibronectin-coated (20 g/ml), serumstarved, and treated with PMA where indicated and subse-quently fixed and permeabilized with 2% (w/v) paraformaldehyde, 0.2% Triton X-100 in PBS. Samples were then blocked with 5% (w/v) bovine serum albumin (BSA) powder/PBS before staining with primary and secondary antibody, followed by washing and mounting. Samples were imaged using a 63ϫ (numerical aperture 1.4) lens on a Leica DM-IRE2 inverted microscope with a Leica TCS SP2 system (Leica, Heidelberg, Germany). Images were captured, and three-dimensional reconstructions were performed using the Leica Confocal Software package.
Immunoprecipitation-Cyanogen bromide-activated Sepharose beads (Sigma) were coated with antibody as per the manufacturer's instructions. Cells were lysed in 1% Nonidet P-40, 2 mM EDTA, 10% glycerol, 137 mM NaCl, 20 mM Tris-HCl, pH 8.0, 10 mM NaF, 10 mM Na 4 P 2 O 7 , 0.2 mM Na 3 VO 4 , and protease inhibitor mixture (Sigma). Lysate was incubated with antibody-bound beads overnight at 4°C and washed three times with lysis buffer followed by four washes with 0.1% Tween/PBS. Bound proteins were eluted using 2.5ϫ SDS running buffer, heated to 100°C. Proteins were separated using SDS-PAGE and analyzed by Western blotting.
MT1-MMP Trafficking-Cells were grown on fibronectincoated glass coverslips and serum-starved for 3 h in DMEM. 500 nM PMA was added to cells to induce trafficking of MT1-MMP to the cell surface. Cells were washed with ice-cold PBS and incubated with 1% BSA/PBS for 20 min at 4°C on ice to prevent internalization of MT1-MMP. Anti-MT1-MMP antibodies were added to the cells at 8 g/ml for 1 h and/or anti-␣5 antibodies at 5 g/ml. Cells were rinsed with 0.2 M glycine/HCl, pH 5.0, followed by successive rinses with PBS to remove any unbound antibody and either fixed and permeabilized with 2% (w/v) paraformaldehyde, 0.2% Triton X-100 in PBS, or slowly warmed in serum-free medium to 37°C to monitor internalization. After fixation, cells were incubated with AlexaFluor647 and/or AlexaFluor594 secondary antibody in 1% BSA/PBS for 1 h. MT1-MMP surface expression was examined using the 63ϫ (numerical aperture 1.4) lens of a Leica inverted micro-scope with constant gain and pin hole parameters. Cell surface expression was quantified for 15-25 cells per experiment using ImageJ software. For removal of TIMP-2 (tissue inhibitor of matrix metalloproteinase-2) from the cell surface, cells grown on coverslips were washed with 150 mM NaCl, and 50 mM glycine/HCl, pH 3.0 (5 min on ice). Cells were then rinsed with PBS and incubated with TIMP-2 antibody for 1 h at 4°C prior to fixation with 4% PFA. After fixation, cells were incubated with rhodamine-phalloidin and/or AlexaFluor488 secondary antibody in 1% BSA/PBS for 1 h. MT1-MMP surface expression was examined using the 63ϫ (numerical aperture 1.4) lens of a Leica inverted microscope with constant gain and pin hole parameters.
Cell Invasion Assay-Cell culture inserts, in 24-well dishes (Costar), were prepared with and without (control) Matrigel. The bottom chamber was coated with 20 g/ml fibronectin, and the upper chamber was coated with 0.125 mg/ml growth factor-reduced Matrigel (BD Biosciences). HT-1080 or HeLa cells were transfected for 12 h, at which point they were lifted and seeded onto Matrigel-coated and control (without Matrigel) upper surface in 0.1% BSA/serum-free medium (80,000 cells/well). The cells that invaded toward chemoattractant (10% FBS, 0.1% BSA) in the lower chamber and penetrated the Matrigel were fixed with 4% paraformaldehyde, stained with DAPI, and counted. Cells that did not invade were removed with a cotton swab prior to fixation of sample. Ten fields of cells per  membrane were counted. The data are presented as the number of cells that invaded through the Matrigel divided by the number of cells that migrated through the control insert (setting mock-treated, GFP-transfected cells at 100%.) Cell Migration Assays-Boyden transwell migration chambers (Costar) were coated with 20 g/ml fibronectin on the bottom chamber. Transfected cells were counted, and 20,000 cells/well in serum-free medium were added to the top well. Cells were serum-starved for 3 h, and then the lower wells were filled with DMEM, 10% FBS, and cells were allowed to migrate overnight. The top and bottom of the membrane were fixed in 4% PFA, stained with DAPI, and mounted on coverslips. 10 fields of cells on the membrane were counted per experiment, using fluorescence microscopy. The data are presented as the number of transfected cells that migrated to the bottom membrane divided by the number of cells that remain on the top membrane.
Sucrose Gradient Centrifugation-A method was established to isolate endosomal compartments using sucrose gradient centrifugation modified from previous work (23). Briefly, cells were kept on ice, washed three times in ice-cold PBS, scrapped, and pelleted at 200 ϫ g for 5 min. The pellet was resuspended in a homogenization buffer (250 mM sucrose, 1 mM EDTA, 3 mM imidazole) and pelleted at 1,300 ϫ g for 10 min. The pellet was resuspended in the homogenization buffer supplemented with 0.03 M cycloheximide and passed through a 22-gauge needle 5-10 times and pelleted at 2,000 ϫ g for 10 min. The postnuclear supernatant was then loaded on top of a sucrose gradient (10 -40%) and centrifuged using a SW41 rotor at 100,000 ϫ g for 18 h. All centrifugation was done at 4°C. 500-l fractions were extracted from the gradient, and 100 l was added to 5ϫ SDS running buffer, heated to 100°C. Proteins were separated using SDS-PAGE and analyzed by Western blotting.

RESULTS
PMA Stimulates MT1-MMP Phosphorylation-Phorbol 12-myristate 13-acetate (PMA) induces the activation of conventional (␣, ␤, and ␥) and novel (, , ⑀, and ␦) isoforms of protein kinase C (PKC) and is one of the most potent and widely used tumor promoters (24). When stimulated by PMA, isoforms of PKC regulate the machinery that controls the intracellular trafficking and transport of proteins; for example, PKC⑀ stimulates membrane trafficking through binding of ␤'COP, a coatomer protein, and regulates the function of cytoskeletal components (25). We and others have shown that PMA stimulates the trafficking of MT1-MMP to the plasma membrane (17,26). Trafficking of MT1-MMP occurs rapidly because cell surface levels of MT1-MMP transiently increase within 20 min of PMA treatment (17,26). Recently, it was shown using an in vitro kinase assay that MT1-MMP is phosphorylated in a PKCdependent manner (27). Here, we tested whether MT1-MMP is phosphorylated in vivo upon PMA treatment and examined the effects of this phosphorylation on MT1-MMP trafficking, cell migration, and cell invasion.
HT-1080 cells were treated with PMA and extracted, and MT1-MMP immunoprecipitates were probed for phospho-Ser/Thr (Fig. 1A). Phosphorylated, endogenous MT1-MMP was detected 10 min after PMA treatment, and this increased significantly within 20 min of treatment (Fig. 1A). Extracts from HT-1080 cells (with or without PMA treatment) were also immunoprecipitated with anti-phospho-Ser/Thr and probed for MT1-MMP. In these experiments, an increase in MT1-MMP phosphorylation was also observed after treatment of cells with PMA (Fig. 1B). To determine the site of phosphorylation, we mutated the only threonine residue, Thr 567 , in the cytoplasmic tail of MT1-MMP to alanine (T567A-MT1-MMP), a site previously shown to be phosphorylated in vitro (27). HT-1080 cells transfected with GFP-tagged wild-type MT1-MMP (GFP-WT-MT1-MMP) or T567A mutant MT1-MMP (GFP-T567A-MT1-MMP) were treated with or without PMA and extracted, and GFPtagged MT1-MMP was immunoprecipitated with antibodies to GFP. The immunoprecipitates were subsequently probed  DECEMBER 16, 2011 • VOLUME 286 • NUMBER 50 JOURNAL OF BIOLOGICAL CHEMISTRY 43409 by Western blot for phospho-Ser/Thr. Fig. 1C shows that the mutation of Thr 567 to alanine abolishes the PMA-dependent phosphorylation of MT1-MMP.

MT1-MMP Phosphorylation and Trafficking
TIMP-2, a member of the TIMP family of MMP inhibitors, can directly bind and modulate MT1-MMP activity on the cell surface (28,29). TIMP-2 is expressed in HT-1080 cells, and we examined the possibility that the interaction of MT1-MMP with TIMP-2 could influence the phosphorylation of MT1-MMP. Therefore, we removed cell surface-associated TIMP-2 by acid wash, as described previously (30), prior to inducing MT1-MMP phosphorylation. Removal of TIMP-2 from the cell surface (Fig. 1E) had no effect on the PMA-induced phosphorylation of MT1-MMP (Fig. 1D).
To test whether the observed changes in cell surface MT1-MMP levels were the result of altered internalization of MT1-MMP, transfected cells were treated with PMA, cell surfacelabeled with MT1-MMP antibodies at 4°C, and then warmed to 37°C to allow internalization. Under these conditions, internalized WT-MT1-MMP and T567E-MT1-MMP could be detected within 15 min, whereas T567A-MT1-MMP remained abundant on the cell surface and could not be detected in obvious endosomal structures at this time point (Fig. 3, A and B). Internalization was assessed by monitoring the presence of MT1-MMP, which had been labeled on the cell surface, within small vesicular structures that were located more centrally within the cell, away from the F-actin network near the edge of the cells (Fig. 3). These results are consistent with the notion that T567A-MT1-MMP is more abundant on the cell surface (Fig. 2) because it is not internalized as efficiently as WT-MT1-MMP (note the lack of MT1-MMP-containing vesicles in the sample transfected with T567A-MT1-MMP compared with the sample transfected with WT-MT1-MMP in Fig. 3). T567E-MT1-MMP appeared to be internalized efficiently (Fig. 3).
Phosphorylation Mutants of MT1-MMP Affect Cell Migration and Invasion-To test the role of phosphorylation of MT1-MMP in cell migration and invasion, HeLa cells were transfected with WT-, T567A-, or T567E-MT1-MMP and analyzed using transwell migration assays and Matrigel invasion assays. HeLa cells were used again here because of their low endogenous level of MT1-MMP expression, resulting in the fact that their capacity to invade in vitro is dependent upon transgenic expression of MT1-MMP (31). Expression of T567E-MT1-MMP increased both cell migration (by 24%) and cell invasion (by 63%) compared with WT-MT1-MMP (Fig. 4, A and B). Cells expressing T567A-MT1-MMP exhibited a decrease in both migration (36% reduction) and invasion (42% reduction) compared with WT-MT1-MMP ( Fig. 4A and B). The observed changes to cell migration were consistent whether migration assays were conducted for 12 h (Fig. 6B) or 6 h (Fig. 6C). It is interesting that expression of T567A-MT1-MMP reduced migration and invasion, given that expression of this construct caused an increase in cell surface expression of the MT1-MMP (Fig. 2). These observations suggest that MT1-MMP internalization and trafficking are essential for this enzyme to promote cell migration and invasion.
MT1-MMP Is Recycled through the Late Endosome in a VAMP7-dependent Manner-To elucidate the pathway through which MT1-MMP is internalized and recycled in response to PMA-stimulated phosphorylation, we used antibodies to label cell surface MT1-MMP and subsequently follow its intracellular trafficking. This approach has been used effectively by others to monitor recycling of MT1-MMP (5). HT-1080 cells were used for these experiments to exploit their high level of endogenous MT1-MMP expression. MT1-MMP was observed on the surface of cells, but internal MT1-MMP was not detected where GFP-tagged Rab5 served as a marker of the early endosome (Fig. 5, A-C). After PMA stimulation under serum-free conditions, MT1-MMP was internalized into a Rab5 compartment within 20 min (Fig. 5, D-F), and 40 min later, some MT1-MMP remained in the Rab5 compartment (Fig. 5, G, H, and J), whereas some was localized in a Rab7  DECEMBER 16, 2011 • VOLUME 286 • NUMBER 50 compartment (Fig. 5, G, I, J, and K-M). By 2 h, a substantial portion of MT1-MMP was localized in the Rab7 compartment (Fig. 5, N-P). Consistent with the localization of MT1-MMP in the late endosome, the distributions of MT1-MMP and Rab7 were found to partly overlap with the endosomal SNARE VAMP7 (Fig. 6, A-D). This pathway of MT1-MMP trafficking was consistently observed, including in experiments in which cell surface TIMP-2 had been washed away as described in the legend to Fig. 1 (data not shown).

MT1-MMP Phosphorylation and Trafficking
Internalized MT1-MMP was not trafficked exclusively into a Rab7 compartment, and some could be detected in a compartment containing the lysosomal marker LAMP-2 but not Rab7 (Fig. 6, E-H). This localization might correspond to a population of internalized MT1-MMP that is targeted for degradation. The MT1-MMP that is delivered to the Rab7/VAMP7 compartment does not appear to be destined for degradation as some of the MT1-MMP co-localizing with Rab7 is not localized with (Fig. 6, E-H). Also, we detected MT1-MMP within this compartment up to 4 h postinternalization (data not shown). Furthermore, if trafficking was restimulated with PMA, we observed that MT1-MMP recycled back to the plasma membrane, as indicated by decreased signal in the late endosome and increased signal at the plasma membrane; compare Fig. 6D (after internalization) with Fig. 6K (after restimulation with PMA). Expression of dominant negative mutants of VAMP7 (VAMP7C) and Rab7 (Rab7DN) inhibited this trafficking back to the plasma membrane (Fig. 6, compare N with K and compare T with Q). Recycling of MT1-MMP was quantified in these experiments by measuring the amount of MT1-MMP at the plasma membrane after PMA treatment to restimulate MT1-MMP trafficking out of Rab7 endosomes. Image stacks like those in Fig. 6 were analyzed using ImageJ to determine levels of MT1-MMP at the plasma membrane (dotted outline in Fig. 6, J, M, P, and S) as a percentage of total MT1-MMP (Fig. 6U). MT1-MMP was never observed to appear at the plasma membrane uniformly but rather was delivered to restricted regions of the membrane, possibly at sites where membrane protrusions are forming. Overall, the results are consistent with a model of MT1-MMP trafficking in which both VAMP7 and Rab7 are involved in a recycling pathway.
We also examined the endosomal markers Rab8 (Fig. 7, A-C) and Rab11 (Fig. 7, D-F) and found no significant overlap of these two markers with internalized MT1-MMP. In all experiments with antibody-labeled MT1-MMP, the distribution of the enzyme closely resembled that of unlabeled endogenous MT1-MMP. As reported by others (15,32), there is a large pool of intracellular MT1-MMP, shown in Fig. 7, G-I, co-localizing with the Golgi SNARE VAMP4. Following PMA treatment, a portion of this MT1-MMP is transported to the cell periphery, where it no longer co-localizes with VAMP4 (Fig. 7, J-L). To ensure that antibody labeling of MT1-MMP was not altering its normal trafficking pattern, we assessed total MT1-MMP trafficking under identical conditions. Confirming our observations made with antibody-labeled MT1-MMP, staining of total endogenous MT1-MMP revealed that, during PMA-induced recycling, MT1-MMP co-localizes first with Rab5 (Fig. 7, M-O) and subsequently with Rab7 (Fig. 7, P-R).
To verify the subcellular compartmentalization of MT1-MMP following PMA stimulation, we used sucrose gradient centrifugation to isolate subcellular organelle fractions. Fig.  8A shows that, 30 min post-PMA treatment, MT1-MMP was abundant in fractions 18 -20, which are enriched in Rab5. 2 h post-PMA treatment, MT1-MMP is abundant in fractions 14 -17, which are enriched in Rab7 (Fig. 8B). These observations are consistent with the distributions of MT1-MMP documented using confocal microscopy above. Inhibition of late endosomal trafficking, through expression of either VAMP7C or Rab7DN, reduced migration and invasion in transfected cells (Fig. 9). This is consistent with results observed after transfection of T567A-MT1-MMP and suggests that trafficking of MT1-MMP is essential for normal function of this enzyme during cell migration and invasion.
PMA Induces Association and Co-trafficking of MT1-MMP with Integrin ␣5 and Phosphorylation of ERK-The impaired migration and invasion of cells expressing T567A-MT1-MMP, VAMP7C, or Rab7DN led us to investigate if PMA treatment was possibly enabling an association between MT1-MMP and ␣5 integrin. MT1-MMP has been shown to process a number of integrin subunits, including ␣3, ␣5, and ␣v (33,34), and ␣5␤1 has been implicated in cell migration and enhanced cell invasion (35,36). We examined immunoprecipitates of MT1-MMP for the presence of ␣5 integrin by Western blot. 20 min after PMA treatment, we detected an increase in the amount of ␣5 integrin pulled down with MT1-MMP (Fig. 10). This observation suggested the possibility that MT1-MMP and ␣5 integrin associate and are trafficked together, and to further investigate this possibility, we examined the localization of these proteins in HT-1080 cells. Cell surface MT1-MMP as well as intracellular MT1-MMP co-localized with ␣5 integrin (Fig. 11, A-D). When internalization and trafficking of cell surface MT1-MMP and ␣5 integrin were monitored following PMA treatment, it was observed that MT1-MMP and ␣5 co-localize at the cell surface (Fig. 11, F-H) and internalize together within 20 min (Fig. 11, J-L), and after 2 h, they are observed together in a compartment consistent with a late endosome (Fig. 11, M-P). This was confirmed using sucrose gradient centrifugation to assess the intracellular localization of MT1-MMP and ␣5. 30 min post-PMA treatment, MT1-MMP and ␣5 integrin were found to be abundant in fractions 17-20, enriched in Rab5, and 2 h post-PMA treatment, MT1-MMP and ␣5 integrin were found in fractions 14 -17, enriched in Rab7 (Fig. 8).
It has previously been demonstrated that ERK phosphorylation is linked with MT1-MMP-dependent cell migration (37), and increased ERK signaling enhances tumorigenesis (38). We therefore analyzed changes to ERK phosphorylation during PMA-stimulated MT1-MMP phosphorylation and internaliza-   tion. In this system, ERK phosphorylation was increased 20 min after PMA treatment (Fig. 12, A and B), coinciding with MT1-MMP phosphorylation (Fig. 1) and internalization (Fig. 3). To examine the dependence of ERK phosphorylation on MT1-MMP phosphorylation, HeLa cells were transfected with GFPtagged WT-, T567A-, or T567E-MT1-MMP, and ERK phosphorylation was measured. ERK phosphorylation was dramatically increased in cells transfected with T567E-MT1-MMP, compared with untransfected control cells or cells transfected with WT-or T567A-MT1-MMP (Fig. 12, C and D). These results are consistent with the notion that phosphorylation-dependent internalization and recycling of MT1-MMP is required for efficient cell migration and invasion, possibly because this trafficking is needed for an association with ␣5 integrin that contributes to activation of ERK.

DISCUSSION
Studies on the endocytosis and trafficking of MT1-MMP reveal the importance of these processes to both cellular invasion and migration (11,16,27,39,40). It has been established that MT1-MMP is phosphorylated at Tyr 573 in a Src-dependent manner and that this event regulates the endocytosis of MT1-MMP in cells stimulated with epidermal growth factor (EGF) (27,32). MT1-MMP has also been shown to be phosphorylated at Thr 567 using an in vitro kinase assay (27), and here we observed the phosphorylation of MT1-MMP at Thr 567 in a PMA-dependent manner in vivo. Mutation of Thr 567 to alanine (T567A), creating a non-phosphorylatable form of MT1-MMP, significantly impaired invasion and migration in cells expressing this mutant, although cell surface levels of this MT1-MMP were increased. Interestingly, T567E-MT1-MMP (which mimics phosphorylated MT1-MMP) was less abundant on the cell surface and appeared to be endocytosed faster than WT-MT1-MMP. These results implicate phosphorylation of Thr 567 , in addition to that of Tyr 573 , in regulating internalization and trafficking of MT1-MMP. The observed phosphorylation of MT1-MMP at Thr 567 in a PKC-dependent manner is consistent with the presence of a consensus sequence (TXR) for PKC as predicted from sequence analysis using the Web tool NetPhos (41). The internalization of MT1-MMP following PMA stimulation led to delivery of the enzyme to a Rab5-containing compartment, followed by transition into a late endosomal compartment, as characterized by the presence of both Rab7 and VAMP7. Moss et al. (27) reported a lack of colocalization between MT1-MMP and Rab7 in the context of EGF-stimulated Tyr 573 phosphorylation, whereas Remacle et al. (15) showed MT1-MMP localization in the late endosome after PMA treatment, in this case characterized by Lamp-1. Collectively, these findings suggest the possibility of alternative trafficking routes that might be influenced by specific phosphorylation events on MT1-MMP. We monitored trafficking up to 4 h post-MT1-MMP internalization and found that after 1 h, the majority of MT1-MMP is localized to a late endosomal compartment and can be stimulated to recycle back to the plasma membrane. This suggests that trafficking to the late endosome is not a means of down-regulating MT1-MMP expression, but rather it is a recycling pathway, allowing for redelivery of MT1-MMP to the plasma membrane. The delivery of MT1-MMP to the plasma membrane was dependent upon Rab7 and VAMP7, and inhibiting this step using dominant negative forms of Rab7 and VAMP7 reduced both invasion and migration, similar to the results seen in the T567A mutant. To our knowledge, this is the first published report of a role for Rab7 in a recycling pathway in mammalian cells. Although we cannot formally exclude the possibility that inhibition of Rab7 or VAMP7 might have more general effects on intracellular trafficking, the results of these experiments do support a model wherein Rab7 and VAMP7 are involved in trafficking of MT1-MMP during cell migration and invasion. Consistent with previous studies showing that VAMP7 is required for MT1-MMP localization to invadopodia (18), our findings support the notion that VAMP7 is a key regulator of MT1-MMP trafficking because we observed dramatic effects on invasion following its inhibition.
We have found that intracellular trafficking and recycling of MT1-MMP to the plasma membrane is crucial for this enzyme to promote not only cell invasion but migration as well. It has been established that MT1-MMP colocalizes with integrin and has a role in processing ␣3, ␣5, and ␣v integrin subunits (34,42,43). Here, we observed that MT1-MMP co-localizes with ␣5 and co-traffics with ␣5␤1 to the late endosome, and we then considered the possibility that the observed decrease in migration and invasion of cells transfected with T567A or dominant negative Rab7 and VAMP7 was due, in part, to changes in integrin-MT1-MMP association. We assessed the co-localization of MT1-MMP and ␣5 as well as the trafficking of ␣5␤1 and found an association between MT1-MMP and ␣5 that increased upon PMA stimulation and coincided with the phosphorylation of Thr 567 . It is thus possible that phosphorylation of Thr 567 results in an increased association between MT1-MMP and ␣5 integrin. Interestingly, we also observed an increase in ERK phosphorylation correlating with MT1-MMP phosphorylation and internalization as well as in T567E mutants; thus, it is reasonable to speculate that MT1-MMP and ␣5 integrin may interact to promote ERK activation. There is an established link between ERK and MT1-MMP; for example, it has been demonstrated that MT1-MMP-dependent cell migration requires ERK activation (33). The ERK pathway has also been shown to increase transcription of MT1-MMP (44). Furthermore, it has been shown that binding of TIMP-2 by MT1-MMP can activate ERK and up-regulate cell migration (45). This activation of ERK was found to be dependent on the cytoplasmic tail but not the catalytic activity of MT1-MMP and required the 573 YCQR 576 sequence. Tyr 573 of MT1-MMP is phosphorylated in an Src-dependent manner, and it is thus possible that this phosphoryla- tion event could affect ERK activation. It is further plausible that MT1-MMP phosphorylation at Thr 567 may function in a similar manner through binding of integrin.
MT1-MMP is one of the best characterized MMPs and has been extensively studied since its discovery in 1994 (46). Overexpression of MT1-MMP has been shown in vitro and in vivo to promote invasion, migration, and metastasis of tumor cells (47). Nonetheless, the regulation of MT1-MMP activity, as well as other MMPs, in cancer is complex and not fully understood (48). Much recent research has gone into the development of pharmacological inhibitors to target MMP synthesis, secretion, and activity (49), and a more detailed understanding of the molecular mechanisms underlying MMP trafficking will benefit this effort. The findings here show, for the first time, the in vivo phosphorylation of MT1-MMP at Thr 567 , which is accompanied by a decrease in MT1-MMP cell surface expression, possibly due to an increase in endocytosis and trafficking. We also highlight the importance of MT1-MMP trafficking through a late endosome as part of a recycling pathway rather than a degradative pathway. Moreover, MT1-MMP association with integrin ␣5 and subsequent internalization may be a means to up-regulate the ERK pathway and increase cell migration and invasion. Collectively, this work points to the importance of MT1-MMP phosphorylation in regulating MT1-MMP activity at the cell surface and the activation of intracellular signaling.