Matrix-dependent Proteolysis of Surface Transglutaminase by Membrane-type Metalloproteinase Regulates Cancer Cell Adhesion and Locomotion*

Cell invasion requires cooperation between adhesion receptors and matrix metalloproteinases (MMPs). Membrane type (MT)-MMPs have been thought to be primarily involved in the breakdown of the extracellular matrix. Our report presents evidence that MT-MMPs in addition to the breakdown of the extracellular matrix may be engaged in proteolysis of adhesion receptors on tumor cell surfaces. Overexpression of MT1-MMP by glioma and fibrosarcoma cells led to proteolytic degradation of cell surface tissue transglutaminase (tTG) at the leading edge of motile cancer cells. In agreement, structurally related MT1-MMP, MT2-MMP, and MT3-MMP but not evolutionary distant MT4-MMP efficiently degraded purified tTG in vitro. Because cell surface tTG represents a ubiquitously expressed, potent integrin-binding adhesion coreceptor involved in the binding of cells to fibronectin (Fn), the proteolytic degradation of tTG by MT1-MMP specifically suppressed cell adhesion and migration on Fn. Reciprocally, Fn in vitro and in cultured cells protected its surface receptor, tTG, from proteolysis by MT1-MMP, thereby supporting cell adhesion and locomotion. In contrast, the proteolytic degradation of tTG stimulated migration of cells on collagen matrices. Together, our observations suggest both an important coreceptor role for cell surface tTG and a novel regulatory function of membrane-anchored MMPs in cancer cell adhesion and locomotion. Proteolysis of adhesion proteins colocalized with MT-MMPs at discrete regions on the surface of migrating tumor cells might be controlled by composition of the surrounding ECM.

Integrin adhesion receptors dynamically regulate cell-matrix interactions by the binding to matrix proteins and inside-out signaling (30,31). This allows cells to discriminate any subtle alteration of the environment and to adjust cell locomotion accordingly. Direct interactions with multiple transmembrane and cell surface proteins (32) including integrin-associated protein-50 (33), TM4SF proteins (tetraspanins) (34) and tTG (35) further attenuate adhesive and signaling efficiency of integrins.
Cell surface tTG (protein-glutamine ␥-glutamyltransferase, EC 2.3.2.13) promotes integrin-dependent adhesion and spreading of cells. By both direct associations with multiple ␤ 1 and ␤ 3 integrins and the binding with Fn, tTG independently mediates the interactions of integrins with Fn (35). The high affinity binding of tTG with Fn specifically involves the 42 kDa gelatin-binding domain of the Fn molecule, which consists of modules I 6 II 1,2 I 7-9 (36). The enhancement of integrin-mediated adhesion and spreading of cells on Fn is independent from the enzymatic activity of surface tTG (35). Intriguingly, reduced expression of tTG has been linked to aggressiveness and high metastatic potential of tumors, whereas overexpression of tTG in fibrosarcomas inhibited primary tumor growth (37,38). Proteolysis of tTG at the normal tissue/tumor boundary was observed in invasive tumors (38).
Here, we report that depending on the structure of the ECM, MT-MMPs are capable of both positively and negatively regulating locomotion of cancer cells. Matrix-dependent proteolysis of surface tTG by MT1-MMP occurs on tumor cells of a diverse tissue origin, thereby representing a general phenomenon and a novel MT-MMP function. Our data suggest an existence of an unexpected link between tumor cell locomotion, the ECM and membrane-anchored MMPs. Regulatory proteolysis of cell surface adhesion proteins by the adjacent MT-MMP molecules is likely to play a significant functional role in cancer cell invasion.
Proteolysis of tTG in Vitro and the NH 2 -terminal Sequencing of the Proteolytic Fragments-Purified tTG (10 -20 g) was incubated with the individual catalytic domain of MT-MMPs (0.2 g each) for 0.5-12 h at 37°C in 0.05 M Tris-HCl buffer, pH 7.5, containing 50 mM NaCl, 1 mM CaCl 2 , and 10 M ZnCl 2 . The reaction was stopped by SDS-sample buffer. After electrophoresis in 12% gels and transfer to a polyvinylidene difluoride membrane, protein bands were stained, excised, and subjected to the NH 2 -terminal microsequencing at the Protein Chemistry Facility of Washington University (St. Louis, MO).
Flow Cytometry-To determine surface levels of MT1-MMP, live untreated or GM6001-treated HT-vector and HT-MT cells were stained with 10 g/ml rabbit anti-MT1-MMP antibody followed by fluoresceinlabeled goat anti-rabbit IgG. For measurements of cell surface tTG, live nonpermeabilized transfectants were each incubated for 2 h at 37°C without or with 20 M GM6001, 1 g/ml TIMP-1, and 1 g/ml TIMP-2 and then stained with 10 g/ml rabbit polyclonal anti-tTG antibody and secondary fluorescein-conjugated IgG (35). Cells were analyzed by a FACScan flow cytometer (Becton Dickinson). At least three independent experiments were performed for each cell line.
Zymography-To evaluate the status of MMP-2, nonreduced aliquots of medium conditioned with HT-vector and HT-MT cells were analyzed by gelatin zymography. Zymography was performed in 0.1% gelatin and 10% polyacrylamide gels as described previously (29). After electrophoresis, SDS was replaced by Triton X-100, followed by incubation in a Tris-based buffer overnight. Gels were stained with Coomassie Brilliant Blue, and gelatinolytic activity was detected as clear bands in the background of uniform staining.
Measurement of Transglutaminase Activity-Transglutaminase activity expressed on the surface of HT-vector and HT-MT cells was determined as reported earlier (35). For these purposes, cells were detached by EDTA. Live cells (2 ϫ 10 6 ) were resuspended in 0.5 ml of phosphate-buffered saline containing 2 mM Ca 2ϩ and 10 mg/ml N,Ndimethylcaseine. Further, cells were incubated with 10 Ci of [ 3 H]putrescine (35.7 Ci/mmol, PerkinElmer Life Sciences) for 1 h at 37°C on a rotator. To evaluate the incorporation of [ 3 H]putrescine, N,N-dimethylcaseine was precipitated from cell-free supernatants with ice-cold 10% trichloroacetic acid. Excess label was removed by successively washing the pellets with 5% trichloroacetic acid, ethanol, and acetone. Finally, the pellet was dried and redissolved in 100 l of 1% SDS. Protein-incorporated radioactivity was determined by scintillation counting.
Binding of the 42-kDa Fragment of Fn to Cells-EDTA-detached HT-vector and HT-MT cells resuspended at 2 ϫ 10 6 cells/200 l of Tyrode's buffer, 10 mM HEPES, 150 mM NaCl, 2.5 mM KCl, 2 mM NaHCO 3 , 2 mM MgCl 2 , 2 mM CaCl 2 , 1 mg/ml bovine serum albumin, 1 mg/ml dextrose, pH 7.4, were incubated with 1 M 125 I-labeled 42 kDa Fn fragment (specific activity 1.6 ϫ 10 6 cpm/g) for 1 h at 37°C on a rotator. Further, cells were layered on 0.5 ml of 20% sucrose in Tyrode's buffer and centrifuged for 5 min at 10,000 rpm. 125 I radioactivity associated with the cell pellets was quantified in a gamma counter. Measurements were performed in triplicates. Values were corrected for nonspecific binding determined in presence of excess (10 M) unlabeled 42-kDa Fn fragment.
Immunoprecipitation of Cell Surface MT1-MMP and tTG-To determine the expression and the status of surface MT1-MMP and tTG, EDTA-detached cells were surface biotinylated for 15 min in phosphatebuffered saline containing 0.1 mg/ml sulfo-NHS-biotin. MT1-MMP and tTG were immunoprecipitated from the lysates of surface-biotinylated cells with MT-MMP-and tTG-specific polyclonal antibodies, respectively. The resulting immune complexes were separated by SDS-PAGE and transferred to a membrane support. Biotin-labeled proteins were visualized by probing the membrane with neutravidin-horseradish peroxidase.
Coprecipitation Studies-To analyze coprecipitation of tTG with ␤ 1 and ␤ 3 integrins, HT-MT cells were plated for 12 h in serum-free medium supplemented with 20 g/ml cycloheximide on plastic coated with Fn or collagen type I. Where indicated, 20 M GM6001 was added to the medium. Next, cells were lysed in RIPA buffer, containing 1% Triton X-100, 0.55 sodium deoxycholate, 0.1% SDS, 150 mM NaCl, 50 mM Tris-Cl, pH 7.5, with 0.5 mM phenylmethylsulfonyl fluoride, 0.5 mM benzamidine, 10 g/ml leupeptin, and 10 g/ml aprotinin (35). ␤ 1 integrins, ␤ 3 integrins, and tTG were each immunoprecipitated from the aliquots (0.5 mg of total protein) of the RIPA lysates with mAb P4C10, mAb LM609, and rabbit polyclonal anti-tTG antibody, respectively. The precipitated samples were separated by SDS-PAGE and transferred to a membrane. To identify specifically tTG in the precipitated samples the membrane was probed with mAb TG100 against tTG.
To induce proteolytic degradation of surface tTG by MT1-MMP, cells were detached by EDTA and then incubated in serum-free medium at 37°C for 2 h. Then, cells were plated for 3 h on Fn in serum-free medium containing 20 g/ml cycloheximide. Three antibodies (murine anti-tTG, rat anti-␤ 1 integrin, and rabbit anti-MT1-MMP) were used to visualize simultaneously tTG, ␤ 1 integrin, and MT1-MMP on live nonpermeabilized cells. After fixation, the proteins were detected with goat anti-rabbit IgG, goat anti-mouse IgG, and goat anti-rat IgG conjugated with blue Alexa-Fluor 350, green Alexa-Fluor 488, and red Alexa-Fluor 594 (Molecular Probes, Eugene, OR) with nonoverlapping emission spectra, respectively. The cells were analyzed and photographed by using Nikon Eclipse E800 epifluorescence microscope and Spot RT digital camera (Molecular Diagnostics).
Cell Adhesion and Migration Assays-For adhesion studies, a suspension of 35 S-labeled U-vector or U-MT cells in serum-free medium was incubated for 2 h without or with 20 M GM6001, 1 g/ml TIMP-1, or 1 g/ml TIMP-2 in the presence of 20 g/ml cycloheximide. 1 h before plating on plastic coated with 10 g/ml Fn, the 110-kDa or the 42-kDa Fn fragments and blocked with bovine serum albumin, function-blocking anti-tTG antibody (10 g/ml) or control nonimmune IgG (10 g/ml) was added to the corresponding cell samples. The cells were plated for 1 h without or with GM6001 in serum-free medium supplemented with 20 g/ml cycloheximide (35). Adherent cells were washed with phosphate-buffered saline and lysed in 1% SDS. The bound radioactivity was counted and converted in the number of adherent cells by referring to the levels of 35 S incorporation/10 3 cells.
Directional migration of cells (5 ϫ 10 4 cells/insert) in Transwells (Costar, Cambridge, MA) with the membrane undersurface coated with collagen I, Fn, the-42 kDa or 110-kDa Fn fragments (10 g/ml each), was analyzed under serum-free conditions (27,32). U-vector and U-MT cells were incubated for 2 h without or with 20 M GM6001, 1 g/ml TIMP-1, and 1 g/ml TIMP-2 in serum-free AIM-V medium (Life Technologies, Inc.). 1 h prior to plating, function-blocking anti-tTG antibody (10 g/ml) or control nonspecific IgG (10 g/ml) was added to the respective samples. Both the upper and the lower chambers of the inserts were supplemented with 10 M GM6001, 1 g/ml TIMP-1, 1 g/ml TIMP-2, 10 g/ml function-blocking anti-tTG antibody, or 10 g/ml control IgG. After incubation for 12 h, cells transmigrated to the membrane undersurface were detached and counted. At least three independent experiments were performed with each reagent and cell type.

RESULTS
tTG Is Degraded and Inactivated on the Surface of Cancer Cells Expressing MT1-MMP-HT1080 fibrosarcoma and U251 glioma cells that naturally express relatively low levels of MT1-MMP were transfected with the MT1-MMP cDNA (GenBank U41078) to up-regulate MT1-MMP and create HT-MT and U-MT pools of cells, respectively (27)(28)(29)32). To avoid any undesirable clonal effects, our studies were performed with the corresponding pools of stable transfectants. Highly elevated levels of MT1-MMP in U-MT cells relative to U-vector control were confirmed previously by immunoprecipitation, flow cytometry, and functional assays (27-29, 32, 40). To corroborate these findings further, we evaluated the levels of MT1-MMP in HT-MT cells by immunoprecipitation, flow cytometry, and by analysis of the ability of cells to initiate the activation of the secretory MMP-2 proenzyme. Immunoprecipitation and flow cytometry studies employed rabbit polyclonal AB815 antibody directed against the hinge region of MT1-MMP. Immunoprecipitation confirmed an increase in the amounts of 60-kDa enzyme and the 42-kDa ectodomain form of MT1-MMP expressed on the surface of HT-MT cells, relative to those of HT-vector cells (Fig. 1a). Earlier, it has been shown that TIMP-2 regulates the amount of active 60-kDa MT1-MMP enzyme on the cell surface, whereas in the absence of TIMP-2 MT1-MMP undergoes autocatalysis to the 42-kDa form (41,42). Apparently, our immunoprecipitation experiments showing high levels of the 42-kDa form point to an insufficiency of TIMP-2 relative to MT1-MMP in HT-MT transfectants (43).
Flow cytometry confirmed the immunoprecipitation data and demonstrated high cell surface levels of MT1-MMP expressed in HT-MT cells (Fig. 1b). As determined by gelatin zymography of conditioned medium samples, up-regulation of MT1-MMP correlated directly with the ability of HT-MT cells to activate the MMP-2 proenzyme (44). The intermediate and active forms of MMP-2 enzyme were present in the conditioned medium from HT-MT but not from HT-vector cells (Fig. 1c, middle and bottom arrows). Apparently, MT1-MMP transfection shifts a balance among the preexisting levels of TIMP-2, MT1-MMP, and MMP-2, thus inducing the activation of pro-MMP-2 by the cells (23). GM6001, a hydroxamate inhibitor of MMPs (45), failed to affect the total amounts of MT1-MMP expressed by HT-MT and HT-vector cells (Fig. 1b). In contrast, GM6001, by binding to the active site and, thereby inactivating MT1-MMP, completely abolished the MT1-MMP-induced activation of pro-MMP-2 (Fig. 1c). These findings correlate well with the results of our previous studies with U-MT and U-vector control cells (40) and demonstrate high levels of functionally active MT1-MMP enzyme expressed on the surface of cancer cells stably transfected with MT1-MMP cDNA. Flow cytometry analyses demonstrated a 4-fold decrease in surface tTG expression in HT-MT cells relative to that of HTvector control cells (Fig. 2a). GM6001 at 20 M restored the expression of tTG on the surface of HT-MT cells to the control levels. Because GM6001 is a general inhibitor of MMP activity, we employed TIMP-2, an efficient inhibitor of MT1-MMP, and TIMP-1, which is a poor inhibitor of MT1-MMP, to distinguish the MT1-MMP effect on tTG. Importantly, TIMP-2 fully restored surface tTG on HT-MT cells. In turn, TIMP-1 was significantly less potent relative to TIMP-2 and GM6001 in its ability to restore the tTG expression on HT-MT cells (Fig. 2a). Any inhibitor had no effect on surface tTG of HT-vector cells.
These findings indicate that MT1-MMP resistant to TIMP-1 inhibition is primarily involved in proteolysis affecting cell surface tTG in HT-MT cells.
To confirm the proteolytic degradation of surface tTG by MT1-MMP in cultured cells, the HT-vector and HT-MT transfectants were labeled with membrane-impermeable sulfo-NHSbiotin. The surface-biotinylated tTG was immunoprecipitated from cell lysates. In contrast to intact tTG (molecular mass ϳ80 kDa) found in HT-vector cells, its proteolytic fragments of ϳ70, ϳ53, ϳ41, and ϳ32 kDa were identified in untreated HT-MT cells (Fig. 2b). Inhibitors such as GM6001, TIMP-1, and TIMP-2 failed to affect the status of tTG in HT-vector cells. On the contrary, pretreatment with GM6001 completely abolished the degradation of surface tTG in HT-MT cells. In these assays, TIMP-2 was slightly less efficient compared with GM6001, whereas TIMP-1 failed to protect tTG from proteolysis in HT-MT cells (Fig. 2b). Together, our observations point to the main role of TIMP-1-resistant MT1-MMP in the proteolysis of cell surface tTG in HT-MT cells. However, MMP-2 and possibly other MMPs may also contribute to proteolytic degradation of surface tTG.
To demonstrate that degradation of cell surface tTG in HT-MT cells abolished its functional activities, we measured both the transglutaminase enzymatic (cross-linking) activity and the ability of surface tTG to bind specifically the 42-kDa gelatin-binding fragment of Fn (Fig. 2, c and d). We observed only a low residual level of transglutaminase activity on the surface of untreated HT-MT cells compared with that of control HT-vector cells. GM6001 restored the enzymatic activity of surface tTG in HT-MT cells to the level observed in HT-vector cells (Fig. 2c). The fact that the associations of surface tTG with Fn specifically involve the 42-kDa Fn fragment was established in our earlier work (35). Accordingly, the efficiency of HT-MT cells to bind the 42-kDa Fn fragment directly was drastically reduced relative to that of control HT-vector cells. Again, inhibition of MMP activity by GM6001 fully restored the binding efficiency of HT-MT cells and failed to affect that of HT-vector cells (Fig. 2d). In agreement with our previous results (35), preincubation of HT-vector and HT-MT cells with polyclonal antibody against tTG abolished the binding of the 42-kDa fragment of Fn to cell surfaces (data not shown).
Immunofluorescent staining confirmed low levels of surface tTG in untreated nonpermeabilized HT-MT cells. In contrast, immunofluorescence of surface tTG was far more intense in HT-vector cells, particularly at the leading edge (Fig. 2e, arrows). Together, our results show that expression of MT1-MMP by cells strongly contributes to the degradation of surface tTG. The degradation of tTG correlates with the inactivation of the both enzymatic and coreceptor functions of this cell surface protein.
Proteolysis of Purified tTG in Vitro-To support our findings, we used enzymatically active individual catalytic domains of MT1-MMP (Fig. 3a) and related MT2-MMP, MT3-MMP, and MT4-MMP (Fig. 3b) to cleave the purified tTG protein in vitro. MT1-MMP, MT2-MMP, and MT3-MMP efficiently and specifically degraded tTG in vitro. The proteolytic fragments of tTG generated in vitro were similar to those detected on the cell surface (Figs. 2b, 3a, and 4a). This provides a key evidence that the MT1-MMP activity is essential for the proteolysis of surface tTG observed in cultured cells.
Extensive proteolysis of tTG by the catalytic domain of MT1-MMP resulted in four major (molecular masses of ϳ70, ϳ53, ϳ41, and ϳ32 kDa) and several minor cleavage products (Fig.  3a, marked by arrows and numbered). In contrast, the catalytic domain of MT4-MMP, a glycosylphosphatidylinositol-linked membrane enzyme that is structurally and functionally distant from the MT1/MT2/MT3-MMP subfamily (20), failed to degrade tTG in vitro (Fig. 3b).
The 53-, 41-, and 32-kDa fragments were partially sequenced to identify the MT1-MMP cleavage sites (Fig. 3, b and d, arrows). The MT1-MMP cleavage at the two close FTR 458 -AN and ANH 461 -LN sites generated the 53-and 32-kDa fragments that represented the NH 2 -and the COOH-terminal parts of tTG, respectively. The NH 2 terminus of both the intact tTG polypeptide and the 53-kDa fragment is blocked by acetylation. The cleavage of tTG at PVP 375 -VR splits the protein in half providing the COOH-terminal 41-kDa fragment with the NH 2 -terminal V 376 RAIK sequence. The PVP 375 -VR sequence partially fits the PXP(G)L consensus cleavage site of MT1-MMP elucidated by using phage display libraries (46). Importantly, MT1-MMP cleavage at any of these three sites should eliminate both the receptor and enzymatic activity of tTG by separating its NH 2terminal Fn binding and the COOH-terminal integrin binding domains (35,47) and by inactivating its second (catalytic) domain (Fig. 2, c and d). Thus, the mapping of cleavage sites on tTG in vitro is in agreement with the results of our enzymatic and coreceptor binding studies of surface tTG in HT-MT cells.
Fn Inhibits Proteolysis of tTG-We next assessed whether Fn, a matrix ligand of cell surface tTG, and laminin and collagen type I, other common ECM proteins but not the ligands of tTG (35), affect the degradation of tTG by MT1-MMP. Excess Fn blocked proteolysis of tTG in vitro. In contrast, excess laminin or denatured collagen type I had no effect on the tTG proteolysis (Fig. 3c). To confirm these effects in cultured cells, we plated HT-MT and HT-vector cells in serum-free medium on plastic coated with Fn and collagen type I. Further, cells were detached, labeled with biotin, and surface tTG was analyzed by immunoprecipitation (Fig. 4a). Without GM6001, HT-vector cells on collagen expressed intact surface tTG. On the contrary, surface tTG was degraded in HT-MT cells plated on collagen type I but intact in cells cultured on Fn (Fig. 4a). To elaborate further on the effects of Fn, we explored our earlier findings that cell surface tTG coprecipitated with ␤ 1 and ␤ 3 integrins (35). For these purpose, we immunoprecipitated ␤ 1 and ␤ 3 integrins from RIPA lysates and analyzed the precipitated samples for tTG. Large amounts of undegraded (ϳ80 kDa) tTG, which comprised ϳ15-25% of total cellular tTG, coprecipitated with ␤1 and ␤ 3 integrins from the lysates of HT-MT cells cultured on collagen in the presence of GM6001 or on Fn without this hydroxamate inhibitor (Fig. 4b, middle and bottom  panels, respectively). In contrast, only low levels of undegraded tTG, which accounted for ϳ1.8 -2% of total cellular tTG, coprecipitated with ␤ 1 or ␤ 3 integrins if HT-MT cells were grown on collagen I without GM6001 (Fig. 4a, top panel, arrow). These results show that Fn specifically inhibited proteolysis of its own adhesion receptor, cell surface tTG, by MT1-MMP. A failure of collagen type I, a substrate more susceptible to MT1-MMP degradation than Fn (8,45), to inhibit proteolysis of tTG, emphasizes a functional significance of the specific inhibitory effect of Fn.
Localization of tTG, ␤ 1 Integrins, and MT1-MMP on Cell Surfaces-Our earlier results indicated that tTG was colocalized with ␤ 1 /␤ 3 integrins at discrete regions of the cell surface (35). Apparently, surface tTG could be degraded by MT1-MMP only if this membrane-anchored proteinase is proximal to the tTG⅐integrin clusters. To demonstrate that tTG degradation by MT1-MMP might occur on cancer cells of a diverse tissue origin, we examined tTG, MT1-MMP, and ␤ 1 integrins on the surface of U-MT and U-vector cells derived from parental U251 human glioma cells. The status of MT1-MMP and MMP-2 in U-MT and U-vector cells has been characterized extensively in our previous work (39). When U-MT cells were plated for 24 h in serum-containing medium on Fn, that was shown to prevent proteolysis of tTG (Fig. 4, a and b), an extensive colocalization of tTG, MT1-MMP, and ␤ 1 integrins became evident at the periphery of U-MT cells (Fig. 5a, merge in yellow).
To corroborate these findings, we preincubated U-MT and U-vector cells in suspension for 2 h without or with GM6001 and then plated the cells for 3 h on Fn in serum-free medium containing cycloheximide. No MT1-MMP was detected on Uvector cells, whereas ␤ 1 integrins and tTG were colocalized at the cell periphery (Fig. 5b, left panels, arrows). In U-MT cells, MT1-MMP accumulated at the edges of lamellipodia (Fig. 5b,  top panels, arrow, arrowhead). In the absence of GM6001 the peripheral staining of tTG in U-MT cells was consistently very weak or absent, pointing to degradation of the cell surface protein (Fig. 5b, middle panels, arrowhead). Localization of ␤ 1 integrins at the periphery of untreated U-MT cells was partially disrupted (Fig. 5b, arrowhead). In contrast, GM6001 restored both the peripheral staining and the colocalization of tTG with MT1-MMP and ␤ 1 integrin at the leading edge of U-MT cells (Fig. 5b, right panels, arrows).
Functional Significance of the Regulatory Proteolysis of Cell Surface tTG-To evaluate the functional significance of the proteolytic degradation of surface tTG by MT1-MMP, we employed cell function in vitro tests such as adhesion and migration assays. Surface tTG interacts with Fn via binding to its 42-kDa gelatin binding domain (35). Accordingly, we used intact Fn and its proteolytic gelatin-binding fragment of ϳ42 kDa in our adhesion and migration experiments. As a control, we employed the ϳ110-kDa Fn fragment that is incapable of binding tTG but associates with integrins (35).
Adhesion of untreated U-MT cells to Fn was ϳ30% lower relative to that of control U-vector cells (Fig. 6a). In the presence of MMP inhibitors such as GM6001 or TIMP-2, adhesion of U-MT cells was restored to the control levels. In contrast, TIMP-1, a poor inhibitor of MT1-MMP activity, showed only a weak effect in restoring adhesion of U-MT cells to Fn. Because FIG. 4. Fn specifically protects surface tTG from proteolysis by MT1-MMP. HT-vector and HT-MT cells were plated for 12 h in serumfree medium supplemented with 20 g/ml cycloheximide on plastic coated with Fn or collagen type I. Where indicated, GM6001 was added to the medium. Panel a, Fn inhibits proteolysis of tTG. 12 h after plating on substrates, cells were detached with EDTA, surface biotinylated, and lysed. tTG was immunoprecipitated from the cell lysates, separated by SDS-PAGE, and analyzed by blotting with neutravidinhorseradish peroxidase. Note the degradation of surface tTG in HT-MT cells on collagen I and the inhibition of the tTG proteolysis by Fn and GM6001. Panel b, Fn increases the levels of tTG coprecipitated with ␤ 1 and ␤ 3 integrins in HT-MT cells. 12 h after plating on substrates, cells were lysed with RIPA buffer. ␤ 1 integrins, ␤ 3 integrins, and tTG were immunoprecipitated from aliquots of cell lysates (0.5 mg of total protein) with specific mAb P4C10, mAb LM609, and mAb TG100, respectively. The precipitated samples were separated by SDS-PAGE and transferred to a membrane. Precipitated tTG was detected in the samples by probing the membrane with anti-tTG polyclonal antibody. Note a reduction of integrin-associated surface tTG in cells plated on collagen I. the interaction of U-251 glioma cells with Fn involves multiple integrin receptors (27)(28)(29), elimination or blocking of a single receptor such as tTG cannot affect the overall adhesive properties of cells strongly. Indeed, we observed a moderate effect of tTG proteolysis on the adhesion of U-MT cell to Fn. To emphasize specifically the effect of tTG degradation on cell adhesion, we evaluated further the adhesion of U-vector and U-MT cells to the 42-kDa fragment of Fn (Fig. 6, b and d).
A more striking difference between U-MT and U-vector cells was identified when cells were allowed to adhere on this fragment of Fn. U-MT cells were about 7-fold less efficient in adhesion to this fragment compared with U-vector control. GM6001 and TIMP-2 each restored adhesion of U-MT cells to U-vector control levels, whereas TIMP-1 displayed a weak effect (Fig. 6b). All of the inhibitors of MMP activity had no effect on U-vector cells (Fig. 6, a and b).
A direct role of tTG in facilitating cell adhesion was confirmed in experiments that employed function-blocking antibody against tTG. When this antibody was used in adhesion assays without GM6001, it did not affect adhesion of U-MT cells to Fn or its 42-kDa fragment (Fig. 6, c and d). In contrast, the use of antibody against tTG abolished stimulatory effects of GM6001 on adhesion of U-MT cells to both of these substrates. Neither MMP inhibitors nor anti-tTG affected adhesion of U-MT and U-vector cells to the 110-kDa fragment of Fn (data not shown).
Further, we evaluated migration of U-vector and U-MT cells in Transwells with the undersurface coated with Fn (Fig. 7a). The stimulatory effects of GM6001, TIMP-2 and TIMP-1 on U-vector cells were insignificant. In contrast, GM6001 and TIMP-2 each strongly enhanced migration of U-MT cells whereas TIMP-1 had relatively low effect (Fig. 7a). Finally, we assessed migration of U-vector and U-MT cells in Transwells with the undersurface coated with collagen type I, Fn, or its 42and 110-kDa fragments. GM6001 did not alter migration of control U-vector cells on Fn and collagen I significantly (Fig.  7a). Compared with U-vector cells, U-MT cells displayed significantly higher levels of migration on collagen type I, but much lower migration on Fn. GM6001 caused an opposite effect on U-MT cells migrating on collagen type I and Fn. Thus, this inhibitor of MMP activity reduced migration of U-MT cells on collagen but enhanced migration of these cells on Fn to the levels characteristic of U-vector cells (Fig. 7, a and b). Migration of U-MT cells on the 42-kDa Fn fragment also strongly increased in the presence of GM6001 (Fig. 7d). Blocking anti-tTG antibody abolished stimulatory effect of GM6001 on U-MT cells migrating on Fn or its 42-kDa fragment, whereas control antibody had little or no effect (Fig. 7, c and d). As expected, GM6001 and anti-tTG antibody did not alter migration of cells on the 110-kDa fragment of Fn (data not shown). DISCUSSION MT-MMP activity that is mostly expressed in aggressive tumors (18,19) greatly contributes to invasion of many tumor cell types (3,12,13). MT1-MMP has been thought to be exclusively involved in the breakdown of the ECM components including collagens (8) and laminin-5 (11), and in the activation pathway of soluble MMPs, i.e. MMP-2 (9, 23) and MMP-13 (24). However, our report presents evidence that membrane-anchored MT-MMPs in addition to the breakdown of the ECM To protect surface tTG from proteolysis, cells were plated in serum-containing medium for 24 h on Fn-coated coverslips. Next, live nonpermeabilized cells were double stained with anti-tTG mAbs CUB7402/TG100 and polyclonal anti-MT1-MMP antibody AB815 (left panels), anti-␤ 1 integrin mAb 9EG7 and polyclonal anti-MT1-MMP antibody AB815 (middle panels), and anti-tTG mAbs CUB7402/TG100 and anti-␤ 1 integrin mAb 9EG7 (right panels). Note the colocalization (lower panels, merge in yellow) of these proteins at the edge of lamellipodia. Bar, ϳ40 m. Panel b, MT1-MMP expression causes a disappearance of surface tTG resulting from proteolytic degradation. U-vector and U-MT cells were detached with EDTA and incubated for 2 h prior to plating in cycloheximide-containing serum-free medium without or with GM6001 (left and two right panels, respectively). Because cycloheximide by significantly blocking de novo protein synthesis did not allow cells to restore the surface tTG pool, degradation of surface tTG by MT1-MMP became evident in the absence of the MMP inhibitor. 3 h after plating on Fn, nonpermeabilized cells were triple labeled with rabbit anti-MT1-MMP antibody AB815 (blue), murine mAbs CUB7402/TG100 against tTG (green), and rat mAb 9EG7 against ␤ 1 integrin (red). Note the colocalization of tTG with ␤ 1 integrins at the periphery of untreated U-vector and GM6001-treated U-MT cells (left and right panels, arrows). MT1-MMP degradation caused the loss of tTG on U-MT cells cultured without GM6001 (middle panels, arrowheads). Bar, ϳ25 m. may be directly engaged in the regulatory proteolysis of adhesion receptors on tumor cell surfaces.
Here we report that MT1-MMP and its structural homolog (MT2-MMP and MT3-MMP) specifically degrade tTG, a ubiquitous integrin-binding cell surface coreceptor for Fn (35). Our studies with inhibitors of MMP activity including a hydroxamate inhibitor GM6001 (45), TIMP-2, and TIMP-1, showed that MT-MMPs are likely to have a primary role in the degradation of surface tTG under the conditions of cell culture. Although our results were obtained with cancer cells overexpressing MT1-MMP, they demonstrate that tTG is a novel substrate for MT1-MMP and a potential target for cancer cell MT1-MMP in vivo. Further work is needed to assess a relative contribution of individual MMPs to the proteolytic degradation of surface tTG in vivo. Proteolysis of tTG by MT1-MMP specifically decreases adhesion and migration of cells on Fn. This underscores the significance of the coreceptor function of surface tTG on cancer cells. Our findings identified an unexpected function of MT-MMPs and a novel means of regulating cell locomotion. Trafficking and subsequent clustering of MT1-MMP with the tTG⅐integrin receptor complexes at the invasive front of tumor cells might be involved in the temporal and spatial control of cell locomotion. Importantly, Fn inhibited the proteolytic degradation of its own receptor, cell surface tTG by MT1-MMP, thereby supporting and enhancing cancer cell locomotion. This suggests that individual components of the ECM may reciprocally regulate proteolysis of their specific adhesion receptors on cell surfaces.
Recently, we reported that up to 40% of total cellular ␤ 1 integrins including ␣ 1 ␤ 1 and other collagen-binding integrins, are directly associated with tTG on cell surfaces (35). Apparently, proteolysis by MT1-MMP of the tTG component of the cell surface tTG⅐integrin complexes is likely to alter the pattern of cell matrix recognition and to switch cells from the binding to Fn to a more efficient interaction with collagens or other ECM proteins. These versatile adjustment mechanisms may regulate migration of cells within composite matrices including connective tissue and tissue barriers such as the basement membrane.
Why do cancer cells, instead of aggressively invading the tissue, down-regulate their locomotion? Degradation of adhesion proteins including surface tTG by MT1-MMP may allow cells to avoid invasion of deficient, degraded matrices and, on the contrary, to trigger detachment and subsequent metasta- sis. In this regard, our data explain previous findings that the reduced expression and proteolysis of tTG observed in vivo are linked to high metastatic potential of tumor cells (37,38,48).
A coordinated interplay of adhesion receptors, proteolytic enzymes, the ECM, and the cytoskeleton is likely to control spatially and temporary invasion and the focalized proteolysis by tumor cells (3). Our work identifies a novel functional link between cell motility and the ECM composition. This link establishes a dual regulation of cell locomotion imposed by MT-MMPs. Evidently, depending on the structure and composition of the ECM, these proteinases may serve as both positive and negative regulators of cell motility. In accordance with the continually changing environment, MT-MMPs might be capable of differentially regulating cell locomotion by shifting molecular gears on the surface of migrating cancer cells. The regulatory proteolysis involving membrane-anchored MT-MMPs is likely to be critical to the status of adhesion proteins such as tTG and, possibly integrins (32). Reciprocally, proteolysis of ECM proteins and a deposition of tumor-specific components can modify the ECM structure. Jointly, these complex mechanisms are likely to control cancer cell invasion efficiently.
The existence of the ancestral membrane-anchored MMPs in plants (49) that have no ECM similar to that of mammals also indicates that matrix breakdown may not be a primary function of MT-MMPs. The cleavage specificity of MT-MMPs is not unique enough (2,8,9) to define an exquisite role of these membrane-anchored proteinases in cell locomotion (11)(12)(13).
In contrast, targeting of MT-MMPs and the tTG⅐integrin complexes to the leading edge of migrating cancer cells where new transient cell matrix adhesive contacts are formed and most of the protrusive activity occurs, can modulate the efficiency of the regulatory proteolysis of adhesion proteins by MT-MMP. This localized proteolytic control of cell adhesion and migration is not possible for soluble MMPs. This hypothesis may partially explain a unique functional role of MT1-MMP, MT2-MMP, and MT3-MMP in tumor cell invasion (12).