O-Glycosylation Regulates Autolysis of Cellular Membrane Type-1 Matrix Metalloproteinase (MT1-MMP)*

Membrane-tethered MT1-MMP,² the most abundant member of the membrane-type (MT) matrix metalloproteinase subfamily, is distinguished from soluble MMPs by a short transmembrane domain and a cytoplasmic tail (,1, 2). MT1-MMP functions in cancer cells as an important mediator of proteolytic events on the cell surface (3, 4), and it is directly engaged in the cleavage of cell surface receptors and the pericellular proteolysis of the extracellular matrix components (5–7). MT1-MMP expression is associated with a variety of pathophysiological conditions and especially with cell locomotion, tumor growth, and metastasis (4, 5, 8–11).

MT1-MMP and related membrane-tethered MMPs are regulated, both as proteinases and as membrane proteins, at the transcriptional and post-transcriptional levels by multifaceted coordinated mechanisms (12–22). The trafficking and the internalization of MT1-MMP have been identified as two additional mechanisms that regulate its biological functions. Both clathrin-coated pits and caveolae are involved in the internalization of MT1-MMP (12, 19, 23–28).

To exercise its proteolytic activity, MT1-MMP requires the proteolytic removal of its N-terminal prodomain sequence (29). Because the prodomain part of MT1-MMP has the furin-cleavage motif, furin and related proprotein convertases (PCs) are the physiological activators of latent MT1-MMP (13, 30–32). Proteolytic processing by the PCs leads to the activation of the latent MT1-MMP zymogen, which occurs primarily in the trans-Golgi network during the secretory passage of MT1-MMP (30, 33). The levels of expression of naturally synthesized, cell surface-associated, MT1-MMP are quite low, normally in the range of 100,000–200,000 sites/cell (34). These low expression levels make it difficult to characterize the biochemical and functional parameters of the protease. In most of the earlier studies involving MT1-MMP, the levels of cellular MT1-MMP were increased by transfecting the cells, either transiently or stably, with the MT1-MMP gene with the result that MT1-MMP was overexpressed (6, 8, 15, 35). Recent findings, however, indicate that MT1-MMP also functions as a self-convertase, and, when overexpressed, the autolytic pathways are initiated, which contributes to the activation and subsequent autolytic degradation of the protease (23, 36–38). In agreement with the requirements of the activation mechanisms and the subsequent autolysis, multiple molecular forms of MT1-MMP were detected in cells/tissues (15, 20, 23, 39). As of this writing, the origin and the molecular nature of the multiple forms of MT1-MMP are not completely known nor understood. Cell type-specific glycosylation may affect the functionality of cellular MT1-MMP and contribute to the generation of its multiple molecular forms (39). Most of the plasma membrane proteins are glycoproteins. An analysis of the peptide sequence of MT1-MMP (GenBank™ P50281), using the NetOGlyc and NetNGlyc software (www.cbs.dtu.dk), suggests that the proline-rich hinge region of MT1-MMP exhibits five potential O-glycosylation sites at Thr²⁹¹, Thr²⁹⁹, Thr³⁰⁰, Ser³⁰¹, and Ser³⁰⁴. Wu et al. (39) has presented experimental evidence of glycosylation of four potential target sites (Thr²⁹¹, Thr²⁹⁹, Thr³⁰⁰, and Ser³⁰¹) in the hinge region of MT1-MMP, thus, providing a solid foundation for the studies reported here.

In this report, we reinvestigate the roles of glycosylation and of the furin-mediated processing in the autolytic degradation of cellular MT1-MMP. Our experimental data provide evidence that glycosylation plays a significant role in regulating the sensitivity of MT1-MMP to self-proteolysis. We conclude that in multiple cancer cell types glycosylation and self-proteolysis of MT1-MMP co-exist in a delicate balance and that an aberration of this fine balance by overexpressing the protease results in its extensive degradation and self-inactivation.

MATERIALS AND METHODS

Antibodies and General Reagents—The reagents were purchased from Sigma unless indicated otherwise. GM6001 (a subnanomolar range hydroxamate inhibitor of MT1-MMP), TIMP-2, the MT1-MMP antibody AB815 and a TMB/M substrate were purchased from Chemicon (Temecula, CA). Decanoyl-Arg-Val-Lys-Arg-chloromethylketone (dec-RVKR-cmk; a nanomolar range inhibitor of PCs), and protease inhibitor mixture set III were obtained from Bachem Bioscience (King of Prussia, PA) and Calbiochem, respectively. Sulfosuccinimidyl-2-(biotin-amido) ethyl-1,3-dithiopropionate (EZ-Link sulfo-NHS-SS-biotin) and sulfosuccinimidyl-6-(biotinamido) hexanoate (EZ-Link sulfo-NHS-LC-biotin) were purchased from Pierce. Gelatin-Sepharose 4B was purchased from Amersham Biosciences. Marimastat-tethered Sepharose 6B was a kind gift of Dr. S. Mobashery (University of Notre Dame) (40, 41).

Cell Lines—Human glioma U251, breast carcinoma MCF7, fibrosarcoma HT1080, embryonic kidney HEK 293, Chinese hamster ovary (CHO), and colon carcinoma LoVo cells, stably transfected with the full-length human MT1-MMP gene, and therefore, overexpressing the protease (U-MT, MCF-MT, HT-MT, HEK293-MT, CHO-MT, and LoVo-MT cells, respectively), were constructed and characterized earlier (23, 36, 42–44). The corresponding control cells were each transfected with the original pcDNA3-zeo plasmid (Invitrogen, Carlsbad, CA) without the MT1-MMP insert (mock cells). The catalytically inactive MT1-MMP mutant (MT-E240A) was generated by replacing the essential Glu-240 of the enzyme active site with Ala. MCF7 cells which stably express MT-E240A were described earlier (36). Glioma U251 cells (U-MT/PDX cells), co-expressing MT1-MMP with α₁-antitrypsin Portland (PDX), were isolated and described in our previous work (45).

To facilitate the isolation of cellular MT1-MMP, the MT1-MMP-His-tagged construct (MT-His) was created by inserting the (His-His)₅ tag between the C-terminal Val⁵⁸² and the stop codon. The MT-His construct was next recloned on the pcDNA3-zeo vector. Breast carcinoma MCF7 cells were transfected with the MT-His construct using Lipofectamine and selected by Western blot analysis from the antibiotic-resistant clones. The expression of the MT-His construct was confirmed by flow cytometry and immunostaining analyses of MCF-MT-His cells.

Deglycosylation of MT1-MMP—Cells were routinely maintained in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal bovine serum (FBS) and gentamicin (10 μg/ml). In the deglycosylation experiments, cells were incubated for 24 h with 2 mm benzyl 2-acetamido-2-deoxy-α-d-galactopyranoside (BGN) alone to inhibit O-glycosylation of MT1-MMP and then co-incubated for an additional 24 h with BGN and GM6001 (25 μm).

MMP-2 Activation Assays—Cells were seeded in wells of a 12-well plate containing 1 ml DMEM-10% FBS (5 × 10⁴ cells/well). In 24 h, cells were replenished with fresh serum-free DMEM (0.8 ml/well) and incubated for an additional 18–24 h. The medium was next separated from the cells by centrifugation. Medium aliquots (10 μl) were analyzed by gelatin zymography to detect gelatinolytic activity of the 68-kDa proenzyme, the 64-kDa activation intermediate, and the 62-kDa mature enzyme of MMP-2 (46). Where indicated, cells were incubated for 24 h with or without BGN alone and co-incubated for an additional 18–24 h with or without BGN and either GM6001 (1–100 μm) or TIMP-2 (1–100 nm) to inactivate MT1-MMP.

Immunoprecipitation and Western Blotting—Cells were grown for 12 h in DMEM-10% FBS. Cells were then washed with PBS and surface biotinylated with sulfo-NHS-LC-biotin according to the manufacturer's instructions. Next, cells were washed with ice-cold PBS and lysed with 50 mm N-octyl-β-d-glucopyranoside (Amresco, Solon, OH) in PBS supplemented with 1 mm CaCl₂, 1 mm MgCl₂, and protease inhibitor mixture set III. The lysates were precleared with protein G-agarose beads (Calbiochem). The samples of cell lysates, each containing 1.0 mg of protein, were mixed with the MT1-MMP antibody AB815 (1 μg) and protein G-agarose, and the mixture was incubated at 4°C overnight. After extensive washings, immune complexes were released by boiling the beads for 5 min in a 2× SDS sample buffer (125 mm Tris-HCl, pH 6.8, 4% SDS, 0.005% bromphenol blue, and 20% glycerol) containing 50 mm dithiothreitol. Solubilized proteins were subjected to Western blotting with Extravidin conjugated with horseradish peroxidase (HRP) and a TMB/M substrate. Where indicated, cells were co-incubated for 24 h with GM6001 (25 μm) and dec-RVKR-cmk (50 μm) to inactivate cellular MT1-MMP and furin, respectively.

Biotinylation and 2-Mercaptoethane Sulfonic Acid (MESNA) Treatment of Cells—Cells grown to an 80–90% confluence were washed twice with an ice-cold Soerensen Buffer (SBS), containing 14.7 mm KH₂PO₄, 2 mm Na₂HPO₄, and 120 mm sorbitol, pH 7.8, and then incubated for 10 min in ice-cold SBS. Cell surface-associated MT1-MMP was biotinylated by incubating cells for 20 min on ice with SBS supplemented with membrane-impermeable EZ-Link NHS-SS-biotin (0.3 mg/ml). Excess biotin was removed by washing the cells in ice-cold SBS. The residual amounts of biotin were quenched by incubating the cells for 10 min in SBS containing 100 mm glycine. Quenched cells were washed in SBS and incubated for 15 min at 37°C in L-15 medium supplemented with a 1% insulin/transferrin/selenium to allow the internalization of biotin-labeled MT1-MMP. To remove the residual cell surface biotin, cells were incubated for 25 min on ice in SBS containing membrane-impermeable MESNA (150 mm). Cells were extensively washed with ice-cold SBS, and lysed in an RIPA buffer (20 mm Tris-HCl, 150 mm NaCl, 0.1% SDS, 1% Triton X-100, 1% deoxycholate, 1% IGEPAL, pH 7.4) containing a protease inhibitor mixture set III, and, in addition, 1 mm phenylmethylsulfonyl fluoride and 10 mm EDTA. MT1-MMP was precipitated from cell lysates using streptavidin-agarose beads. The precipitated samples were analyzed by Western blotting with the MT1-MMP AB815 antibody followed by goat anti-rabbit IgG-conjugated with HRP and a TMB/M substrate (47).

Pulldown of MT1-MMP on Con A-Agarose Beads—U-MT/PDX, HT-mock, and HT-MT cells grown to 80% confluence were lysed in an RIPA buffer containing a protease inhibitor mixture set III supplemented with 1 mm phenylmethylsulfonyl fluoride, 2 mm CaCl₂ and 2 mm MgCl₂. Cell lysates (0.7 mg of total protein) were then incubated overnight at 4°C with Con A-agarose beads (30 μl of 50% slurry) equilibrated in an RIPA buffer containing 2 mm CaCl₂ and 2 mm MgCl₂. After extensive washings, the precipitated material was released by boiling the beads for 5 min in a 2× SDS sample buffer containing 50 mm dithiothreitol. Solubilized proteins were analyzed by Western blotting using the MT1-MMP AB815 antibody as described above.

Purification of MT1-MMP using Marimastat-tethered Sepharose—HT-mock and HT-MT cells were lysed in an RIPA buffer supplemented with a protease inhibitor mixture set III and 1 mm phenylmethylsulfonyl fluoride. The cells were passed 10 times through a 0.26-gauge needle. The cell debris was removed by centrifugation, and the supernatant fraction was used for further experiments. To remove gelatin-binding MMP-2 and MMP-9, the supernatant aliquot (1.5 ml with a total protein of 1 mg/ml) was incubated for 16 h at 4°C with gelatin-Sepharose 4B beads (150 μl of a 50% slurry). The samples were spun to pellet the beads and the supernatant samples were dialyzed against 50 mm Tris-HCl buffer, pH 7.5, containing 150 mm NaCl, 5 mm CaCl₂, and 0.2% Brij-35 and supplemented with an inhibitor mixture set III and 1 mm phenylmethylsulfonyl fluoride. The dialyzed samples were then mixed with the Marismatat-tethered Separose 6B beads (60 μl of a 50% slurry). After incubation for 16 h at 4°C, the resin was washed twice with the above buffer, and then three times with the same buffer containing 1 mm NaCl. MT1-MMP was eluted from the beads using reducing 5× SDS sample buffer (125 mm Tris-HCl, pH 6.8, 10% SDS, 0.005% bromphenol blue and 50% glycerol). The samples were separated by SDS-PAGE, transferred onto a polyvinylidene difluoride membrane (Millipore, Bedford, MA) and analyzed for the presence of carbohydrates using a GlycoProfile III Fluorescent Glycoprotein Detection kit (Sigma). The N-terminal peptide sequence of the MT1-MMP species isolated from HT-MT cells was determined by N-terminal microsequencing at ProSeq (Boxford, MA).

Enzymatic Desialylation of MT1-MMP—Enzymatic desialylation with neuraminidase (Sigma) was performed according to the manufacturer's instructions. Briefly, MCF-MT-His cells were lysed in an RIPA buffer containing a protease inhibitor mixture set III, supplemented with phenylmethylsulfonyl fluoride (1 mm) and GM6001 (10 μm). The MT-His protein was immunocaptured from cell lysates using a murine monoclonal antibody to polyhistidine immobilized on agarose beads. The immunocaptured samples were extensively washed with RIPA buffer and then equilibrated in a neuraminidase buffer containing 100 mm sodium acetate, pH 5.5, supplemented with 2 mm CaCl₂. The MT-His samples were incubated for 24 h at 37°C in the buffer alone and in the buffer supplemented with neuraminidase (0.5 unit). The digest samples were analyzed by Western blotting with the MT1-MMP antibody followed by goat anti-rabbit IgG conjugated with HRP and a TMB/M substrate. In addition, the samples were analyzed using HRP-labeled peanut lectin.

Metabolic Labeling and the Degradation Kinetics of Cellular MT1-MMP—LoVo-MT cells were metabolically labeled in methionine-deficient DMEM with [³⁵S]methionine Translabel (10 mCi/1 × 10⁷ cells) (ICN, Irvine, CA) (23). After a 15-min pulse with [³⁵S]methionine, cells were chased for 0.5–4 h in DMEM-5% FBS supplemented with 10 mm l-methionine. Immediately after chase, 0.01% sodium azide was added to the cells to block cellular metabolic and synthetic activity as well as to prevent protein trafficking. Cells were then surface-biotinylated with membrane-impermeable biotin and lysed as described above. Cell lysates were immunoprecipitated with protein G-agarose beads, and the AB815 antibody to MT1-MMP to isolate total cellular [³⁵S]MT1-MMP. After washing, the bound material was solubilized by incubating the beads in 100 μl of 0.05% trifluoroacetic acid. Following centrifugation, 1 m Tris was added to the supernatant fractions to neutralize the acid and to reach pH 7. Streptavidin-agarose beads (50 μl of a 50% slurry) were added to each of the supernatant fractions to capture the biotinylated, cell surface-associated, [³⁵S]MT1-MMP. After washing, the captured proteins were released by boiling the beads in 50 μl of reducing 2× SDS sample buffer. Capture of the biotinylated molecules on streptavidin-agarose allowed us to separate the cell surface-associated biotin-labeled [³⁵S]MT1-MMP from the total cellular [³⁵S]MT1-MMP. The samples of the total cell and the cell surface-associated, biotin-labeled pools were each examined by SDS-PAGE followed by radioautography.

In Vivo Tumorigenicity Assay and the Analysis of MT1-MMP in Tumor Xenografts—To generate tumor xenografts, LoVo-mock, and LoVo-MT (1 × 10⁷ cells in 0.1 ml PBS each), and HT-mock and HT-MT cell suspension (1 × 10⁶ cells in 0.1 ml of PBS each) were injected s.c. in athymic female 4-week-old immunodeficient BALB/c nu/nu mice (Benton & Kingman, San Francisco, CA). Five animals were used per each group. Tumor xenograft growth was monitored every 3–7 days by caliper measurements of two perpendicular diameters of xenografts (D₁ and D₂). Tumor volume was calculated by the formula p/6(D₁ × D₂)^3/2, and was expressed as mean tumor volume ± S.E. (in mm³). At the end of the experiments at 21 days (for HT-mock and HT-MT xenografts) and at 40 days (for LoVo-mock and LoVo-MT xenografts) after the cell injection, mice were sacrificed according to National Institutes of Health guidelines. Tumors were excised free of connective tissue, washed and cut. Tumor pieces (about 5–10 mm³ each) were extracted with 2× SDS sample buffer (1:2 w/v). After extraction for 2 h at ambient temperature, samples were diluted 2-fold with a 1× SDS sample buffer and homogenized by sequential passages through 18- and 23-gauge needles. The solubilized material was separated from the pellet by centrifugation at 14,000 × g for 30 min. Aliquots of supernatants (50 μg of total protein each) were analyzed by Western blotting with the MT1-MMP antibodies.

RESULTS

Overexpressed MT1-MMP Is Autolytically Degraded in Cancer Cells—We determined the expression levels and the molecular forms of cell surface-associated MT1-MMP naturally synthesized by MDCK-mock, U-mock, HT-mock, LoVo-mock, MCF-mock, CHO-mock, and HEK293-mock cells by surface labeling with membrane-impermeable biotin. By using the antibody AB815, MT1-MMP was captured from cell lysates and its biotin-labeled molecular forms were detected by Western blotting with Extravidin conjugated with HRP. We specifically used the AB815 antibody in these experiments because it recognizes the full-length and the degraded, membrane-tethered, species of the proteinase, which lack the catalytic domain (36). Naturally synthesized MT1-MMP was represented by minor levels of full-length MT1-MMP in mock-transfected cells (Fig. 1). In contrast, an extensive degradation of overexpressed MT1-MMP and, as a result, the accumulation of the 40–45 kDa degraded, cell surface-associated, species was observed in the stably transfected MDCK-MT, U-MT, HT-MT, LoVo-MT, MCF-MT, CHO-MT, and HEK293-MT cells. Degradation of the protease was absent in MCF-MT-E240 cells, which expressed the catalytically inert E240A mutant. Similarly, the hydroxamate inhibitor GM6001 blocked both the proteolytic activity and the degradation of MT1-MMP in MCF-MT cells (Fig. 1).

We confirmed that the extensive self-proteolysis of MT1-MMP that we earlier observed in cultured cells also exists in tumors by generating tumor xenografts in immunodeficient nude mice. For these purposes, mice were s.c. injected with LoVo-mock, LoVo-MT, HT-mock, and HT-MT cells. The incidence of tumors was 100% in all cell types. Twenty-one days and forty days after cell injection, HT-MT and LoVo-MT cells generated ∼2.5-times and 3-times larger tumor xenografts than HT-mock and LoVo-mock cells, respectively. These results suggested that MT1-MMP strongly promoted tumor growth, and they were consistent with our earlier observations and the results of others (3, 8, 28, 43). We next used Western blotting to determine the status of MT1-MMP in tumor xenografts. In agreement with the degradation of recombinant MT1-MMP that we observed in cultured cells, MT1-MMP was significantly degraded in both LoVo-MT and HT-MT tumor xenografts. The degradation of MT1-MMP naturally expressed by HT-mock and LoVo-mock xenografts was non-existent. In LoVo xenografts, MT1-MMP was predominantly represented by the proenzyme species, while in HT1080 xenografts we observed the mature enzyme of MT1-MMP. We corroborated these observations and also determined the kinetics of the MT1-MMP degradation by performing pulse-chase experiments in the LoVo-MT cells followed by isolating and analyzing the cell surface-associated MT1-MMP and the total cell MT1-MMP pool.

Rapid Autolytic Degradation of Overexpressed MT1-MMP—We determined the kinetics of the degradation of MT1-MMP through the use of pulse-chase techniques with l-[³⁵S]methionine. Radiolabeling of LoVo-MT cells was followed by cell surface biotinylation and the capture of biotin-labeled, cell surface-associated [³⁵S]MT1-MMP on streptavidin-agarose. The use of this method allowed us to distinguish the biotin-labeled, cell surface-associated, [³⁵S]MT1-MMP pool from total cell [³⁵S]MT1-MMP pool (Fig. 2).

Our data showed a rapid degradation and a 2-h half-life of MT1-MMP at the cell surface. The 15-min pulse time was sufficient to detect trafficking of MT1-MMP to the plasma membrane, and 1 h was enough to observe a significant autolytic degradation of newly synthesized MT1-MMP. During the next 2–3 h, MT1-MMP was degraded completely and, as a result, only the degraded forms were observed on the cell surface. Because the cells had been preincubated prior to cell surface biotinylation with sodium azide to block cellular metabolism, it was not possible to observe endocytosis of MT1-MMP in this cell system. GM6001 added to the cells during a pulse-chase period blocked the autolytic degradation of cell surface-associated MT1-MMP, whereas dec-RVKR-cmk did not affect the efficiency of MT1-MMP degradation. Our current results agree well with our earlier findings which indicated a high rate of both exocytosis and decay of newly synthesized MT1-MMP (23). Consistent with earlier observations (15, 48), these results also suggested that autolysis was the major mechanism involved in the degradation of cell surface-associated, overexpressed, MT1-MMP in our cell systems. We hypothesize that other post-translational mechanisms, in addition to inhibition by TIMPs and especially by TIMP-2 (29, 49), are involved in a multifaceted control of cellular MT1-MMP activity.

PDX Inhibits Furin and Regulates MT1-MMP Processing—To identify these putative regulatory mechanisms and to evaluate the molecular forms of MT1-MMP, we inhibited the intracellular activation of MT1-MMP in glioma U251 cells. We co-transfected U251 cells with MT1-MMP and PDX (α₁-antitrypsin variant Portland). PDX is a picomolar range inhibitor of furin and furin-like PCs (32, 50). Furin cleavage of the ¹⁰⁸RKPR ↓ Y¹¹² motif, which is localized in the prodomain sequence of MT1-MMP, is known to be involved in the processing and activation of MT1-MMP (Fig. 3).

We determined the identity of the molecular forms of MT1-MMP by examining U-MT and U-MT/PDX cells. Cell surface biotinylation followed by an analysis of the biotin-labeled proteins captured on streptavidin-beads was used to distinguish the forms of MT1-MMP presented on the cell surface. Western blotting of the total cell lysates was used to determine all of the cellular MT1-MMP forms. Where indicated, the hydroxamate GM6001 and the furin inhibitor, dec-RVKR-cmk, were added to the cells to block the cell surface MT1-MMP and the furin activity, respectively.

Gelatin zymography was used to determine MMP-2 activation in U-MT and U-MT/PDX cells. No activation of MMP-2 was observed in U-mock cells, in which the levels of naturally synthesized, cellular MT1-MMP were insignificant when compared with those in U-MT cells. As expected, U-MT/PDX cells were less potent in activating MMP-2 when compared with U-MT cells (Fig. 3). This finding suggested that PDX inhibited cellular furin and furin-like PCs and that this event repressed the activation of MT1-MMP in U-MT/PDX cells.

An analysis of MT1-MMP revealed the presence of the 63-, 58-, and 55-kDa full-length forms (most probably, the proenzyme, the intermediate and the enzyme, respectively) and, in addition, the 45-kDa degraded forms of MT1-MMP, in total cell extract of both U-MT and U-MT/PDX cells (Fig. 3). The amounts of the 58-kDa intermediate were similar in both cell types whereas the amounts of the 55-kDa mature enzyme, and especially the degraded products, were significantly elevated in U-MT cells when compared with U-MT/PDX cells. In U-MT/PDX cells, the 63-kDa latent proenzyme was also easily visible. GM6001, by inhibiting MT1-MMP itself, reduced its autolytic degradation, blocked the accumulation of the 45-kDa degraded forms of MT1-MMP and, consequently, stimulated the accumulation of the 55-kDa MT1-MMP species. GM6001 had no effect on the 58-kDa intermediate. A furin inhibitor dec-RVKR-cmk increased, albeit insignificantly, the levels of the putative 63-kDa proenzyme and the 58-kDa intermediate in U-MT/PDX cells and in U-MT cells. These results suggest that the MT1-MMP precursor is quantitatively processed in its secretory pathway by furin and, as a result, the fully processed, activated, MT1-MMP form arrives at the cell surface. Subsequently this active species of MT1-MMP autocatalytically degrades in U-MT cells. In turn, the expression of PDX in the intracellular compartment leads to the loss of the furin-related PC activity and rescues the MT1-MMP precursor. These events stimulate the delivery of the MT1-MMP proenzyme jointly with the processed, activated, form of MT1-MMP to the surface of U-MT/PDX cells. In addition, it is evident that even in the presence of PDX and GM6001 the inhibition of MT1-MMP processing was incomplete in the intracellular milieu of U-MT/PDX cells. This residual intracellular activity of MT1-MMP was sufficient to support autolysis which appears to begin in the course of the trafficking of de novo synthesized MT1-MMP through the cell compartment to the plasma membrane.

An additional analysis of cell surface-associated MT1-MMP showed the presence of the 63-kDa proenzyme and, mostly, the mature 55-kDa enzyme and degraded products in U-MT/PDX cells. In U-MT cells, cell surface MT1-MMP was represented by the enzyme and, predominantly, by the degraded forms. No proenzyme was detected on the surface of U-MT cells. In U-MT/PDX and U-MT cells, GM6001 inhibited the autocatalytic degradation of MT1-MMP and, consequently, stimulated the accumulation of the cell surface-associated mature enzyme in U-MT cell and both the proenzyme and the enzyme in U-MT/PDX cells. No significant, additional, accumulation of the proenzyme and the enzyme was found in the U-MT cells co-incubated with dec-RVKR-cmk. In contrast, dec-RVKR-cmk inhibited the conversion of the MT1-MMP proenzyme into the enzyme in U-MT/PDX cells. This event, however, did not significantly increase the relative concentrations of the cell surface 63-kDa proenzyme in U-MT/PDX cells when compared with those observed with GM6001 alone, suggesting the efficient endocytosis of the proenzyme and its clearance from the cell surface. In addition, our recent results showed that glioma U251 cells express significant levels of PC7 in addition to furin and PC5/6 (51). PC7 is known to be insensitive to PDX (52). Our results observed in U-MT/PDX cells support the role of PC7 in the activation of MT1-MMP.

The 58-kDa intermediate was not observed on the surface of either U-MT or U-MT/PDX cells. These data suggest that this intermediate accumulates predominantly inside the cell compartment and, in contrast to the proenzyme and the mature enzyme of MT1-MMP, is not efficiently transported to the plasma membrane.

Because glycosylation is known to regulate the intracellular trafficking of cellular proteins (53–56) and because we suspected that the intracellular intermediate represented the incompletely glycosylated species of the protease, we closely studied the mechanisms of glycosylation of MT1-MMP.

O-Glycosylation Regulates the Efficiency of Self-proteolysis of MT1-MMP—MT1-MMP exhibits several O-glycosylation sites (39) and no N-glycosylation sites. BGN (benzyl 2-acetamido-2-deoxy-α-d-galactopyranoside), a well-known, synthetic N-acetylgalactosamine analogue, competitive, inhibitor of O-glycosylation, was used to competitively inhibit O-glycosylation of MT1-MMP in U-MT and U-MT/PDX cells (57). Cell surface biotinylation followed by the capture of the biotin-labeled proteins on streptavidin-beads was used to isolate cell surface-associated MT1-MMP. Western blotting of the total cell lysates was used to determine all of the cellular MT1-MMP forms. Cells were co-incubated for 24 h with BGN alone to inhibit glycosylation and then for an additional 24 h with BGN and GM6001 to suppress the autocatalytic activity and, consequently, the degradation of cellular MT1-MMP.

Fig. 4 shows that, despite the presence of GM6001 in the samples, BGN significantly stimulated self-proteolysis of MT1-MMP in the intracellular milieu. In addition, inhibition of O-glycosylation decreased the apparent molecular weight of the MT1-MMP forms. A close analysis of the data in Fig. 4 suggests that the 58-kDa intermediate indeed represents an incompletely glycosylated proenzyme form of MT1-MMP. The results of our analysis of cell surface-associated MT1-MMP in U-MT and U-MT/PDX cells agreed well with our tentative conclusions. Our results also suggest that glycosylation is not a prerequisite for the delivery of the proenzyme-enzyme forms of MT1-MMP to the plasma membrane. The differentially glycosylated 58-kDa intermediate, however, was observed only inside the cells and it was not detected on the cell surface. Similar mechanisms were reported for many other membrane proteins including DPP-1V, CEA, MUC1, and CD44 (53, 58). As expected, tunicamycin (an inhibitor of N-glycosylation) had no effect on MT1-MMP. In agreement, the treatment of the MT1-MMP samples with peptide-N-glycosidase F did not affect the mobility of the protease forms.

To confirm that naturally synthesized MT1-MMP is also glycosylated, we evaluated MT1-MMP in HT-mock cells and, for comparison, in HT-MT cells, in which the protease was overexpressed (Fig. 4). The 55-kDa mature enzyme represented the full-length MT1-MMP in HT-mock and HT-MT cells. Similar to its effect in U-MT and U-MT/PDX cells, BGN decreased the apparent molecular weight of the MT1-MMP enzyme in HT-mock and HT-MT cells thus suggesting that naturally produced MT1-MMP was also glycosylated in HT-mock cells. To corroborate these results, we chromatographed the cell lysate samples on a Con A column. Con A effectively binds both N-glycans and O-glycans (59, 60). U-MT/PDX, HT-mock, and HT-MT cells were each lysed, and MT1-MMP was captured on Con A-agarose beads. The captured material was eluted and evaluated by Western blotting with the MT1-MMP antibody AB815. The proenzyme, the intermediate and, less efficiently, the enzyme, were capable of interacting with the lectin in U-MT/PDX cells (Fig. 4). The data are consistent with the data of Wu et al. (39) who reported that in COS cells the MT1-MMP enzyme was less efficient in binding to Con A when compared with the proenzyme. Fig. 4 shows that MT1-MMP isolated from HT-mock cells and also from HT-MT cells, was glycosylated. The presence of a carbohydrate moiety in MT1-MMP from HT-mock cells has indicated that naturally synthesized protease is glycosylated.

To additionally confirm glycosylation of MT1-MMP in HT-mock cells, we used the Fluorescent Glycoprotein Detection kit. This kit was designed for the selective staining of glycoproteins using a modification of the periodic acid-Schiff method (61). We used the Marimastat-tethered resin to isolate MT1-MMP from the MMP-2/MMP-9-depleted lysates of HT-mock and HT-MT cells. As we expected, the isolated samples of MT1-MMP were efficiently stained with the Glycoprotein Detection kit and the 55-kDa enzyme of MT1-MMP was readily observed in the isolated samples (Fig. 4). The N-terminal microsequencing determined the N-terminal peptide sequence (YAIQGLKWQHN) of the isolated 55-kDa MT1-MMP enzyme, which corresponded to the known ¹¹²YAIQGLKWQHN N-terminal sequence of the mature enzyme and thus validated its authenticity. Because of extensive self-degradation of the overexpressed enzyme in either U-MT or U-MT/PDX cells, we and other laboratories (32) were unable to isolate the 58-kDa intermediate in amounts sufficient for N-terminal sequencing.

Glycosylation Affects the Uptake of MT1-MMP by the Cells—To determine if glycosylation affects the internalization of MT1-MMP in U-MT/PDX cells, we examined the uptake of biotin-labeled MT1-MMP followed by treatment of the cells with membrane-impermeable MESNA. MESNA treatment releases the biotin moiety from the cell surface-associated proteins, whereas the internalized proteins are protected from MESNA treatment (27). Where indicated, BGN was used to inhibit O-glycosylation of cellular MT1-MMP (Fig. 5). Co-incubation of all cell samples with GM6001 was used to suppress the autolysis of cellular MT1-MMP. The inhibitor was then included in all incubation buffers. Immediately after co-incubation with GM6001, the cells were incubated for 10 min at 0°C and then cell surface biotinylated with membrane-impermeable EZ-Link sulfo-NHS-SS-biotin. After removal of the excess biotin, the cells were then incubated at 37°C for 15 min to allow biotin-labeled MT1-MMP to be internalized. The cells were next put on ice, to arrest further internalization, treated with MESNA and lysed. The subsequent capture of biotin-labeled proteins on streptavidin-agarose beads and the Western blot analysis with the MT1-MMP antibody allowed us to determine the residual levels of biotin-labeled, internalized, MT1-MMP in the samples. Fig. 5 shows that BGN treatment changed the uptake rate of the MT1-MMP proenzyme when compared with that of the enzyme in U-MT/PDX cells. In the absence of BGN, however, the proenzyme was internalized more rapidly than the enzyme. To quantitatively assess the results, the gels were scanned, and the images were digitized. The density of the individual proenzyme (P) and the enzyme (E) bands of MT1-MMP was expressed as a percentage of the total MT1-MMP (p + E = 100). The P/E ratio was 40:60 in the MESNA-untreated U-MT/PDX cells, but after the MESNA treatment of these cells the P/E ratio was 60:40. In the presence of BGN, the rate of internalization of the enzyme and the proenzyme was similar.

Terminal Sialylation Plays an Important Role in Regulating Self-proteolysis of MT1-MMP—In the course of glycosylation, sialic acid frequently terminates the carbohydrate moiety. To determine if the binding of the terminal sialic acid is important for the regulation of MT1-MMP, we used MCF-MT-His cells. Similar to MT1-MMP in glioma U251 cells, MT-His was represented by the putative proenzyme, the intermediate and the enzyme species of MT1-MMP in MCF-MT-His cells. The treatment of MCF-MT-His cells with BGN led to the deglycosylation of MT1-MMP, to a shift in the mobility of the MT1-MMP forms and to an increase in the degradation of MT1-MMP in the intracellular compartment (Fig. 6). These effects, however, were relatively insignificant. We corroborated these data by performing deglycosylation experiments in MCF-MT cells. Because the shift and the degradation observed in breast carcinoma cells were similar to the results in glioma cells, we concluded that glycosylation of MT1-MMP is a common phenomenon.

The MT-His protein was pulled down from cell lysates by using an antibody to polyhistidine. The immunoprecipitated samples were treated with Clostridium perfringens neuraminidase. This neuraminidase cleaves α-2,3-linked sialic acid residues most efficiently. After the treatment, the samples were analyzed by Western blotting with the MT1-MMP antibody. Fig. 6 shows that the neuraminidase treatment significantly stimulated the autocatalytic degradation of MT1-MMP and, as a result, the degraded forms of MT1-MMP were observed in the cell samples. Neuraminidase treatment alone was sufficient to induce levels of degradation of MT1-MMP similar to those induced by BGN.

In addition, we used peanut lectin with its well defined carbohydrate binding specificity to confirm the presence of sialic acid in the carbohydrate moiety of MT1-MMP. This lectin binds to galactose-N-acetylgalactosamine but this interaction is abolished if the galactose residue is sialylated. Consistent with its binding characteristics, peanut lectin was incapable of binding to the MT-His protein. Neuraminidase treatment of the MT-His samples, however, removed the sialic acid moiety and restored the ability of peanut lectin to interact with the full-length and degraded MT1-MMP forms (Fig. 6). These findings suggest that the terminal α-linked sialic acid, a negatively charged, terminal residue of the cell surface oligosaccharides, significantly contributes to the stability of the MT1-MMP molecule. The results of other laboratories suggest that MT1-MMP undergoes autolysis to generate a degraded form which displays an N terminus either at Gly²⁸⁵ (48) or at Gly²⁸⁴ (37). We believe that the presence of the strongly negatively charged sialic acid that terminates the oligosaccharides linked to the Thr²⁹¹-Ser³⁰⁴ hinge region sequence restricts the access of the active site of MT1-MMP to the autolytic cleavage site at Gly²⁸⁵, thereby serving as a buffer zone. Conversely, the absence of the normal glycosylation, and especially the absence of the terminal sialic acid, stimulates the autocatalytic degradation of MT1-MMP, a phenomenon that has been observed in many laboratories and that is shown in our Fig. 1.

Glycosylation of MT1-MMP Does Not Affect the Processing of MMP-2—To determine if glycosylation affects the MMP-2-activating function of MT1-MMP, we evaluated the efficiency of MMP-2 activation by U-MT cells (Fig. 7). Glycosylation of cellular MT1-MMP was blocked by co-incubation of the cells with BGN for 48 h. Where indicated, TIMP-2 and GM6001 were added for 18–24 h to block MMP-2 activation by the cells. We did not observe any significant differences in the efficiency of MMP-2 activation by normally glycosylated MT1-MMP when compared with deglycosylated protease. As expected, TIMP-2 and GM6001 blocked the activation reactions. Again, we did not observe any significant differences in the sensitivity of glycosylated MT1-MMP versus deglycosylated protease to these inhibitors. These results suggested that in our cell setting glycosylation did not significantly affect the interactions of MT1-MMP with either TIMP-2 or GM6001.

To further analyze the effect of BGN on MT1-MMP in more detail, we examined how the increasing concentrations of GM6001 inhibited the degradation of MT1-MMP in U-MT cells in the presence and in the absence of BGN (Fig. 7). In the presence of BGN, deglycosylated MT1-MMP was destabilized and, therefore, it was significantly more resistant to GM6001 inhibition. As a result, the degraded forms of MT1-MMP were still observed in the total cell extract even in the presence of GM6001 at 25 μm. In contrast, in the absence of BGN, GM6001 at 1–2 μm was sufficient to totally abolish MT1-MMP autolysis and MMP-2 activation. These data are not surprising because GM6001 is more efficient in accessing the cell surface-associated MT1-MMP than the intracellular pool of the protease. We conclude that O-glycosylation regulates the stability of the MT1-MMP enzyme.

Altogether our data suggest that glycosylation of MT1-MMP, and especially the presence of the terminal α-linked sialic acid in the oligosaccharide chain, stabilizes the protease. Conversely, the absence of the normal glycosylation induces an extensive self-degradation of MT1-MMP, a phenomenon observed in many recombinant cell types, which overexpress MT1-MMP.

DISCUSSION

MT1-MMP is an archetypal plasma-membrane-tethered, membrane proteinase from the MMP family. MT1-MMP plays a highly significant role in cell motility, in the regulation of cell adhesion signaling receptors, in the activation of soluble proteinases and in the cleavage of the extracellular matrix proteins. Because of its high functional significance, MT1-MMP has been the subject of investigation by multiple studies. Unfortunately, the natural levels of cellular MT1-MMP are very low and these low natural levels greatly complicate the biochemical studies of cellular MT1-MMP. As a result, volumes of studies have been performed with cells that have been transfected with MT1-MMP and which, therefore, overexpress the proteinase. In contrast with the naturally expressed, cellular MT1-MMP which is represented by its full-length, enzyme-proenzyme forms, overexpression stimulates self-degradation of MT1-MMP and the generation of membrane-tethered, catalytically inactive, ectodomain forms. The catalytically inactive form missing the catalytic domain and commencing at either Gly²⁸⁴ or Gly²⁸⁵ is the main product of the MT1-MMP autolytic pathway (21, 37). Earlier work experimentally demonstrated that MT1-MMP is post-translationally modified by O-glycosylation of the Thr²⁹¹-Ser³⁰¹ hinge region sequence (39). In the peptide sequence of MT1-MMP, these sites are adjacent to the autolytic cleavage site at Gly²⁸⁵. O-Linked glycosylation occurs at a later stage during protein processing in the Golgi apparatus (62).

On the basis of these data, we hypothesize that glycosylation contributes to the stability of cellular MT1-MMP and that incomplete glycosylation promotes autolytic degradation of the recombinant, overexpressed, MT1-MMP enzyme. Our experimental results are consistent with this hypothesis and also with earlier data presented by other laboratories (39). In the course of our study, we initially confirmed that overexpressed MT1-MMP was significantly degraded in many cell systems, including breast carcinoma, colon carcinoma, glioma, fibrosarcoma, Madin-Darby canine kidney, embryonic kidney, and Chinese hamster ovary cells. We also determined that glycosylation, rather than the furin cleavage of the prodomain part of the MT1-MMP proenzyme at the ⁸⁹RRPR ↓ C⁹³ sequence, is the main reason for generating the intermediate molecular form of MT1-MMP. The ⁸⁹RRPRC⁹³ sequence includes the active site Zn²⁺ binding Cys⁹³ and this region is shielded from the external proteases by the prodomain sequence (63–67). In agreement, our recent data confirmed that the ⁸⁹RRPRC⁹³ site is inaccessible to cleavage by furin and furin-like PCs in the MT1-MMP proenzyme (51).

A careful analysis of our data (Figs. 4 and 6) suggests that there is a noticeable difference in glycosylation of MT1-MMP in different cells. Thus, approximately one third part of the amounts of the MT1-MMP proenzyme was hypo-glycosylated in U-MT/PDX cells (Fig. 4A; arrow on the left points to the hypo-glycosylated MT1-MMP zymogen) and in MCF-MT-His cells (Fig. 6, A and B), whereas at least a half of the proenzyme pool was hypo-glycosylated in MCF-MT cells (Fig. 6A). All of these cell types overexpress the recombinant protease. In contrast, MT1-MMP was uniformly glycosylated and the hypo-glycosylated species were not observed in HT-mock cells, which naturally express the protease (Fig. 4A).

We next determined that BGN, a competitive inhibitor of O-glycosylation, significantly promoted the autocatalytic degradation of MT1-MMP. Our results also demonstrated that glycosylation affects the uptake rate of MT1-MMP and, especially the MT1-MMP proenzyme when compared with the uptake rate of the enzyme. We also identified a previously undetected glycosylated intermediate of MT1-MMP. This intermediate resided predominantly in the intracellular compartment and it was inefficiently transported to the plasma membrane when compared with the normally glycosylated forms of MT1-MMP.

Our data suggest that autolytic degradation begins early in the course of trafficking of newly synthesized, incompletely glycosylated, MT1-MMP through the cell compartment to the plasma membrane and then, following the arrival of MT1-MMP at the cell surface, self-proteolysis continues. In contrast to the earlier results reported by Wu et al. (39), our data suggested that glycosylation of MT1-MMP was not required for the activation of MMP-2. According to Wu et al. (39), glycosylation affected the recruitment of TIMP-2 to the cell surface, resulting in the defective formation of the MT1-MMP/TIMP-2/proMMP-2 trimeric activation complex. The differences between our and earlier results may be explained, at least partly, by the use of the distinct cell systems. In addition, an analysis of the data presented in the Wu publication (Ref. 39, Fig. 8) suggests that because of extensive degradation of glycosylation-deficient mutants the levels of the functionally active MT1-MMP enzyme were low in his cell system. These low levels of the MT1-MMP enzyme rather than interference with TIMP-2 binding led to the inefficient activation of MMP-2 in his study. Autolysis of hypo-glycosylated mutants observed by Wu et al. (39) agrees well with our data and support our suggestion that deglycosylation significantly accelerates the autolytic degradation of MT1-MMP. Our results correlate also well with the earlier observations of Kinoshita et al. (68) who used Escherichia coli-derived, non-glycosylated, MT1-MMP in MMP-2 activation and TIMP-2-binding experiments and who demonstrated the ability of this recombinant MT1-MMP to be potent in both reactions.

Our results also suggested that either the removal of the terminal sialic acid of the oligosaccharide moiety or the hypo-O-glycosylation is a prerequisite for the extensive self-degradation of MT1-MMP. Sialic acid is likely the most important functional component of the oligosaccharide cassette in MT1-MMP and its presence, probably, restricts the access of the catalytic domain to the hinge region and to the autolytic cleavage site. The stabilization effect of glycosylation was also reported for MMP-9 (69) and, in general, represents a common mechanism involved in the stabilization of multiple distinct proteins (70). In addition, the absence of both a sialylation pathway and the post-translational sialylation of yeast proteins (71) explains the extensive self-proteolysis of MT1-MMP derived from yeasts Pichia pastoris that we observed earlier (37). Overall, our data are in agreement with and significantly extend the observations of many other laboratories and present a model where glycosylation regulates the autocatalysis of the enzyme and through this mechanism, the presentation and the functional activity of MT1-MMP in many cancer cell types.