An alternative processing of integrin alpha(v) subunit in tumor cells by membrane type-1 matrix metalloproteinase.

Membrane type-1 matrix metalloproteinase (MT1-MMP) and alpha(v)beta(3) integrin are both essential to cell invasion. Maturation of integrin pro-alpha(v)chain (pro-alpha(v)) involves its cleavage by proprotein convertases (PC) to form the disulfide-bonded 125-kDa heavy and 25-kDa light alpha chains. Our report presents evidence of an alternative pathway of pro-alpha(v) processing involving MT1-MMP. In breast carcinoma MCF7 cells deficient in MT1-MMP, pro-alpha(v) is processed by a conventional furin-like PC, and the mature alpha(v) integrin subunit is represented by the 125-kDa heavy chain and the 25-kDa light chain commencing from the N-terminal Asp(891). In contrast, in cells co-expressing alpha(v)beta(3) and MT1-MMP, MT1-MMP functions as an integrin convertase. MT1-MMP specifically cleaves pro-alpha(v), generating a 115-kDa heavy chain with the truncated C terminus and a 25-kDa light chain commencing from the N-terminal Leu(892). PC-cleavable alpha(3) and alpha(5) but not the PC-resistant alpha(2) integrin subunit are also susceptible to MT1-MMP cleavage. These novel mechanisms involved in the processing of integrin alpha subunits underscore the significance and complexity of interactions between MT1-MMP and adhesion receptors and suggest that regulation of integrin functionality may be an important role of MT1-MMP in migrating tumor cells.

The integrins are a family of heterodimeric transmembrane receptors that mediate dynamic interactions between the extracellular matrix (ECM) 1 and the cytoskeleton (1)(2)(3). By linking the ECM with the cytoskeleton, integrins regulate cell adhesion, motility, contractility, and invasion (4,5). The integrins are composed of noncovalently associated ␣ and ␤ subunits that combine to form about 25 different receptors. Among all known integrins, ␣ v ␤ 3 plays a unique functional role in tumor angiogenesis and metastasis (6).
It has been established that certain mature integrin subunits including ␣ 3 , ␣ 4 , ␣ 5 , ␣ 6 , ␣ 7 , ␣ 8 , ␣ 9 , ␣ v , ␣ IIb , and ␣ E are generated by post-translational endoproteolytic cleavage of the respective precursors. The cleavage at the highly conserved pairs of basic amino acids is a redundant function of proprotein convertases (PC) from the subtilisin/kexin family. PC cleavage converts single chain ␣ precursors into respective mature subunits consisting of an N-terminal heavy chain and a C-terminal light chain connected by a disulfide bridge (7). Furin, a member of the PC family, has been specifically implicated in the cleavage of pro-␣ 5 , -␣ 6 , and -␣ v chains (7,8). The role of this cleavage in integrin function is unclear. Arguably, post-translational maturation by proteolytic cleavage of the ␣ chain is not required for ligand binding and cell adhesion but is essential to "outside-in" signal transduction by integrins (9,10). Thus, inhibition of pro-␣ v cleavage by overexpression of ␣ 1 -antitrypsin Portland, a furin inhibitor, impaired integrin ␣ v ␤ 5 -mediated signal transduction and spreading in adenocarcinoma HT29-D4 cells (9).
In addition to integrins, tumor cells are believed to exploit matrix metalloproteinases (MMPs) to cross the ECM barriers. MMPs are a family of zinc-dependent endopeptidases shown to degrade the ECM (11). MT1-MMP belongs to a subfamily of MMPs, which are distinguished by a transmembrane domain that anchors the molecule to the plasma membrane. It has been suggested that the principal function of MT1-MMP is to mediate the activation pathway of soluble MMPs including MMP-2 and MMP-13 (12)(13)(14). Although MT1-MMP is detectable in normal tissue, elevated functional activity of MT1-MMP is associated with malignant and metastatic tumors (11,15,16).
Up-regulation of MMPs in invasive cells has been frequently associated with integrin expression (17)(18)(19)(20)(21)(22). There is increasing evidence that integrins interact with MMPs on the surface of normal and tumor cells (23). Thus, ␣ v ␤ 3 integrin has been demonstrated to bind and localize activated MMP-2 to discrete regions of invasive melanoma cells (24). In agreement, expression of integrin ␣ v ␤ 3 was repeatedly linked to the activation of MMP-2 in multiple tumor cell types (25)(26)(27). In human glioma U251 and melanoma BML cells, ␣ v ␤ 3 was shown to co-localize and interact with MT1-MMP (16,27).
Our study demonstrated that in addition to the breakdown of the ECM and activation of soluble MMPs, MT1-MMP functions as an integrin convertase. We show that in breast carcinoma MCF7 cells co-expressing ␣ v ␤ 3 integrin and MT1-MMP, the protease is directly involved in the endoproteolytic cleavage of pro-␣ v . This alternative processing of pro-␣ v generates ␣ v ␤ 3 integrin that is superior relative to conventional ␣ v ␤ 3 integrin in promoting cell adhesion, migration, and focal adhesion kinase phosphorylation (43). Furthermore, other integrin ␣ chains, which undergo PC-mediated maturation, were also sensitive to cleavage by MT1-MMP. Our results suggest that MT1-MMP-mediated integrin processing may be a general mechanism by which cells selectively regulate the functionality of integrins. This novel regulatory mechanism of integrin function underscores the significance and complexity of interactions between MT1-MMP and cell adhesion receptors in tumor cells.
In co-cultivation experiments, ␤3/zeo cells were incubated for 48 h in DMEM/FCS with MCF7 cells expressing MT-WT or MT-E240A alone. For these purposes, parental ␣ v ␤ 3 -and MT1-MMP-deficient MCF7 cells were transfected with the MT-WT and MT-E240A constructs to generate MT-WT and MT-E240A cells, respectively (31).
Flow Cytometry-Cells were stained with 5 g/ml murine anti-␣ v ␤ 3 mAb LM609 or rabbit anti-MT1-MMP antibodies in ice-cold Dulbecco's PBS, pH 7.2 (DPBS; Invitrogen), supplemented with 1 mM CaCl 2 , 1 mM MgCl 2 , 1% BSA, and 0.02% sodium azide (all from Sigma). After incubation with the corresponding secondary sheep anti-mouse or goat anti-rabbit IgG F(abЈ) 2 conjugated with fluorescein isothiocyanate (Sigma), cells were analyzed on a FACStar flow cytometer using CellQuest software (Becton Dickinson). Population gates for negative controls were set using cells stained with murine or rabbit IgG (both from Sigma).
Immunoprecipitation and Western Blotting-Cells were plated in DMEM/FCS. Where indicated, cells were incubated for 12-72 h with hydroxamate inhibitors, AG3340 or Ilomastat, or left untreated. After incubation, cells were washed with DPBS and surface-biotinylated with sulfo-N-hydroxysuccimide long chain-biotin according to the manufacturer's instructions (Pierce). Following incubation for 60 min, cells were washed with ice-cold DPBS and lysed with 50 mM N-octyl-␤-D-glucopyranoside (Amresco, Solon, OH) in Tris-buffered saline supplemented with 1 mM CaCl 2 , 1 mM MgCl 2 , and protease inhibitor mixture containing 1 mM phenylmethylsulfonyl fluoride and 2 g/ml each of aprotinin, pepstatin, and leupeptin (OG buffer). The lysates were precleared with Pansorbin (Calbiochem). The samples of cell lysates each containing 1.0 mg of protein were mixed with 2-3 g of anti-integrin or anti-MT1-MMP antibodies and Protein G-agarose. Following incubation for 14 h at 4°C, agarose beads were washed with the OG buffer and then with DPBS supplemented with 0.05% Tween 20 (DPBS/Tw) and 0.5 M NaCl and with DPBS/Tw. Immune complexes were released by boiling the beads for 5 min in 2ϫ SDS sample buffer (20 l). Where indicated, the samples were reduced with 50 mM dithiothreitol. After centrifugation, solubilized proteins were subjected to SDS-PAGE. Separated proteins were transferred to an Immobilon P membrane (Millipore Corp., Bedford, MA). Following blocking with 1% casein in DPBS/Tw, the membrane was probed with Extravidin-HRP (Sigma), and the bound HRP activity was visualized with TMB/M substrate (Chemicon).
To identify the MT1-MMP-processed forms of ␣ 3 and ␣ 5 integrin subunits, the lysates of cells surface labeled with biotin were adsorbed on avidin-agarose beads (Pierce) and washed as described above. Adsorbed proteins were then released by boiling with 1ϫ SDS-PAGE buffer and separated on an 8% SDS-PAGE gel followed by transfer to a membrane. Integrin chains were visualized by Western blotting.
Purification and the NH 2 -terminal Sequencing of ␣ v ␤ 3 Integrin-Integrin ␣ v ␤ 3 was purified from ␤3/zeo, ␤3/MT-WT and ␤3/MT-E240A cells on a mAb LM609 column as previously reported (29). Briefly, the OG lysate of 1.6 ϫ 10 9 cells was passed over the column. After washing the column with the OG buffer, adsorbed ␣ v ␤ 3 integrin was eluted with 0.05% trifluoroacetic acid, pH 2.5, in 50 mM N-octyl-␤-D-glucopyranoside. The samples were quickly neutralized, dialyzed against 5 mM Tris, pH 7.5, containing 0.1% SDS, concentrated 20-fold in a Speed Vac concentrator, separated by SDS-PAGE, and transferred to an Immobilon membrane. After staining with Coomassie Blue, the integrin bands were excised and subjected to the NH 2 -terminal microsequencing.
To analyze whether MT1-MMP co-purifies with ␣ v ␤ 3 integrin, the samples of purified ␣ v ␤ 3 integrin were separated by SDS-PAGE and transferred to a membrane. Immunoblotting was performed using rabbit antibody AB1932 against ␤ 3 integrins and antibody AB815 against MT1-MMP. Bound primary antibodies were visualized with the HRPconjugated secondary anti-rabbit antibodies and the TMB/M substrate.
Furin-deficient human colon carcinoma LoVo cells (32) were used as a source of pro-␣ v . Pro-␣ v integrins were partially purified from LoVo cells on a mAb L230 column. Pro-␣ v integrins (0.2 g) were digested for 16 h at 37°C using rMT-cat (0.1 g; 5 pmol) in 20 mM Tris-HCl, pH 7.5, 150 mM NaCl, 50 M ZnCl 2 , 5 mM CaCl 2 . The digest was subjected to SDS-PAGE. Immunoblotting was performed with 2 g/ml rabbit antibody AB1930 specific to the cytoplasmic tail of the ␣ v integrin chain.
Confocal Microscopy-Cells were plated at 1-1.5 ϫ 10 4 cells/well of an eight-well LabTek II glass chamber (Nalge Nunc, Naperville, IL). Following incubation for 48 h, cells were washed with PBS and fixed for 20 min with an ice-cold, 1:1 methanol/acetone mixture. Nonspecific binding was blocked by incubation for 30 min at room temperature in DPBS containing 10% goat serum and 5% BSA. Cells were stained with 10 g/ml mAb LM609 or mAb L230 and further with 20 g/ml goat anti-mouse IgG conjugated with Alexa 568 (Molecular Probes, Inc., Eugene, OR). Next, cells were stained with 10 g/ml rabbit anti-MT1-MMP antibodies and 20 g/ml goat anti-rabbit IgG conjugated with Alexa 488 (Molecular Probes). After washing in DPBS, cells were embedded into VectaShield (Vector, Burlingame, CA) and examined using a scanning confocal microscope (MRC-1024; Bio-Rad). Acquisition and processing of images were performed with the Lasersharp (Bio-Rad) and AdobePhotoshop software.
Gelatin Zymography-Cells were plated in DMEM/FCS with or without increasing concentrations of Ilomastat. After incubation for 24 h, cells were washed and further incubated for 24 h with 50 ng/ml MMP-2 in serum-free DMEM supplemented with the same concentration of Ilomastat. Medium was mixed with an equal volume of 2ϫ SDS sample buffer. Aliquots (10 l) were run on gelatin-gels (Novex, Carlsbad, CA). To visualize gelatinolytic bands, the gels were processed as reported (16).

Expression of ␣ v ␤ 3 Integrin and MT1-MMP in MCF7
Cells-We specifically selected breast carcinoma MCF7 cells for our studies, since the parental cell line was deficient in MT1-MMP, MMP-2, and the ␤ 3 integrin subunit but expressed ␣ v integrins. Previously, we have characterized MCF7 cells stably transfected with ␤ 3 integrin subunit (␤3 cells) and wild type MT1-MMP (MT-WT cells) (16). To extend our studies, we generated MCF7 cell lines co-expressing ␣ v ␤ 3 integrin with either the wild type or the enzymatically inactive mutant MT1-MMP (MT-E240A), in which the essential Glu 240 of the catalytic center was replaced by the Ala residue (MT-E240A cells). To generate ␤3/zeo control cells, ␤3 cells were additionally transfected with the original pcDNA-zeo plasmid. The ␤3/zeo, ␤3/MT-WT, and ␤3/MT-E240A cells were characterized by flow cytometry, gelatin zymography, immunoprecipitation, and Western blotting.
To further prove that MT-E240A has no enzymatic activity, we used gelatin zymography to analyze the ability of cells to activate exogenous pro-MMP-2 (Fig. 2B). The ␤3/zeo and ␤3/  2B). These findings confirmed that MT-E240A was catalytically inactive.
MT1-MMP-dependent Processing of ␣ v ␤ 3 Integrin-Co-expression of MT1-MMP and ␣ v ␤ 3 integrin enhanced migration of MCF7 cells on vitronectin, suggesting specific interactions between the protease and the integrin (16). To evaluate this possibility, we immunoprecipitated ␣ v ␤ 3 integrin from cells surface labeled with biotin and analyzed the precipitates by Western blotting. In MT1-MMP-deficient cells (␤3/zeo) and in cells expressing the mutant (␤3/MT-E240A) or the wild type enzyme inhibited by the hydroxamate inhibitor AG3340 (␤3/ MT-WT), ␣ v ␤ 3 integrin consisted of the 150-kDa ␣ v and the 90-kDa ␤ 3 chains (nonreducing conditions). Under reducing conditions, the 150 kDa pro-␣ v , the 125-kDa heavy ␣ v chain, and the 105-kDa ␤ 3 chain were observed in these cell lines. Nonreduced ␣ v ␤ 3 integrin from ␤3/MT-WT cells had an additional 140-kDa ␣ v band. Reduced ␣ v ␤ 3 integrin from ␤3/MT-WT cells consisted of the 125-kDa heavy ␣ v chain, an ␣ v heavy chain with an unusually low molecular weight (115 kDa), and the 105-kDa ␤ 3 chain. No pro-␣ v was observed in ␤3/MT-WT cells (Fig. 2C). These data strongly suggested a correlation between the presence of the 115-kDa ␣ v form and MT1-MMP activity in ␤3/MT-WT cells. In addition, our results indicated that the light ␣ v chain was still disulfide-bonded to the heavy ␣ v chain in ␤3/MT-WT cells.
To confirm a correlation between the MT1-MMP activity and the presence of the 115-kDa ␣ v form, we inhibited the MMP proteolytic activity by incubating ␤3/zeo (control) and ␤3/MT-WT cells for 48 h with increasing concentrations of Ilomastat, a potent hydroxamate inhibitor of MMPs. Following incubation, the cells were surface-labeled with biotin and lysed, and the lysates were immunoprecipitated with anti-␣ v ␤ 3 mAb LM609 followed by Western blotting. The inhibitor did not affect the composition of integrin ␣ v ␤ 3 in ␤3/zeo cells (Fig. 3A). On the contrary, Ilomastat at increasing concentrations gradually diminished the amounts of the 115-kDa ␣ v heavy chain and reciprocally increased the levels of the 150-kDa pro-␣ v in ␤3/ MT-WT cells (Fig. 3B, upper panel). At the 10 -50 M range of Ilomastat concentrations, the status of ␣ v ␤ 3 integrin in ␤3/ MT-WT cells was similar to that observed in ␤3/zeo and ␤3/ MT-E240A cells. The inhibition of the pro-␣ v processing by Ilomastat correlated well with the inactivation of the autolytic activity and the ability of MT1-MMP to activate exogenous pro-MMP-2 (Fig. 3B, middle and lower panels, respectively). Inhibitors of serine, cysteine, and aspartic proteases failed to affect either the pro-␣ v processing or self-degradation of MT1-MMP in ␤3/MT-WT cells (data not shown), making the involvement of these proteases unlikely.
To further elaborate on these observations, we evaluated the turnover rates of MT1-MMP and ␣ v ␤ 3 integrin. For these purposes, ␤3/MT-WT cells were incubated for 12-72 h with 10 M AG3340 followed by biotinylation and immunoprecipitation of ␣ v ␤ 3 integrin and MT1-MMP. In cells incubated with the inhibitor for 12 h, the 60-kDa MT1-MMP form fully replaced the autolytic forms of the protease, suggesting a short half-life (about 2.5 h) and, consequently, a high turnover rate of MT1-MMP on the cell surface (Fig. 3C, lower panel). In contrast, 48 -72 h were needed to substitute the 115-kDa ␣ v form with the 150-kDa pro-␣ v on the cell surface (Fig. 3C, upper panel). These results suggest a relatively long half-life (ϳ15 h) of cell surface ␣ v ␤ 3 integrin and agree with the earlier estimations for other integrins (33). Overall, our findings support the hypothesis that MT1-MMP is directly involved in the processing of pro-␣ v in ␤3/MT-WT cells.
Processing of Pro-␣ v by MT1-MMP-To identify the effects of MT1-MMP on pro-␣ v , we inhibited the maturation of ␣ v ␤ 3 integrin by furin by incubating ␤3/zeo and ␤3/MT-WT cells with 100 M dec-RVKR-cmk, a furin inhibitor. Following incubation with the inhibitor for 48 h, cells were surface-biotinylated, and the cell lysates were immunoprecipitated with anti-␣ v ␤ 3 mAb LM609. Although in ␤3/zeo cells the inhibitor suppressed the furin-dependent cleavage of pro-␣ v , it failed to significantly affect the processing of pro-␣ v in ␤3/MT-WT cells (Fig. 3D).
Further, we used pro-␣ v integrins purified from furin-deficient colon carcinoma LoVo cells and mature ␣ v ␤ 3 integrin isolated from human placenta to reproduce in vitro the effects of MT1-MMP observed in cells. For these purposes, both integrin samples were digested with rMT-cat, the recombinant catalytic domain of MT1-MMP. Although rMT-cat was highly active and readily degraded ␣ 1 -antitrypsin, a known substrate of MT1-MMP (data not shown), no cleavage of the placental ␣ v heavy chain was observed (Fig. 4A). In turn, rMT-cat digested LoVo pro-␣v and generated a 16-kDa C-terminal fragment recognizable on Western blots by antibody AB1930 specific to the C-terminal cytoplasmic tail of the integrin ␣ v subunit. AG3340 added to the reaction completely blocked the proteolysis. Thus, the in vitro cleavage studies confirm that pro-␣ v is susceptible to proteolysis by MT1-MMP.
Cleavage Site Sequence-To identify the MT1-MMP cleavage sequence, we isolated ␣ v ␤ 3 integrin from ␤3/zeo and ␤3/MT-WT cells, subjected the purified integrins to SDS-PAGE, and determined the N-terminal peptide sequence of all integrin bands separated under reducing and nonreducing conditions. Our results are schematically represented in Fig. 4D. The ␤ 3 chain from ␤3/zeo or ␤3/MT-WT cells had the same N-terminal 27 GP-NIXTT sequence. In ␤3/zeo samples, both 150-and 125-kDa bands had the same N-terminal 31 FNLDVD sequence, whereas the 25-kDa band of the light ␣ v chain (not shown) had the N-terminal 891 DLAL sequence. These data, consistent with the structure of ␣ v ␤ 3 integrin (34), confirmed the processing of the 150-kDa pro-␣ v by furin-like PC into the 125-kDa heavy chain and the 25-kDa light chain in ␤ 3 /zeo cells. Intriguingly, the ␣ v subunit from ␤3/MT-WT cells was represented by the ␣ v light chain with the N-terminal 892 LALSEG sequence and by the two species (125 and 115 kDa) of the heavy chain, both exhibiting the N-terminal 31 FNLDVD sequence. These findings indicate that MT1-MMP is capable of cleaving pro-␣ v at least at two sites, thereby generating the C-terminally truncated ␣ v heavy chain with the molecular mass of 115 kDa and the 25-kDa 892 LAL ␣ v light chain that is one residue shorter from its N terminus relative to the PC-processed 891 DLAL chain.
MT1-MMP Cleaves ␣ 3 and ␣ 5 but Not ␣ 2 -Since MT1-MMP cleavage of pro-␣ v occurs within a region of high homology with other ␣ integrin chains, we examined if other ␣ subunits could also be processed by MT1-MMP. For these purposes, cells were surface-labeled with biotin, and biotin-labeled proteins were isolated with avidin-agarose beads and analyzed by Western blotting using ␣ 2 , ␣ 3 , and ␣ 5 anti-integrin mAbs (Fig. 4C). Both ␣ 3 and ␣ 5 integrin chains appeared to be processed by MT1-MMP in ␤3/MT-WT cells but not in ␤3/zeo cells. In contrast, the pattern of the ␣ 2 integrin subunit, known to be resistant to furin-like convertases, was highly similar in ␤3/MT-WT and ␤3/zeo cells. These observations suggest that MT1-MMP is likely to be capable of proteolytic modification of integrin ␣ subunits, which mature by PC cleavage.
Proximity of ␣ v ␤ 3 Integrin and MT1-MMP on the Cell Surface-To confirm that pro-␣ v is accessible to MT1-MMP cleavage, we investigated whether the proteinase and ␣ v ␤ 3 integrin are in close proximity on the cell surface. We used several approaches to examine this issue. First, ␤3/MT-WT cells were doubly stained with anti-MT1-MMP antibodies and anti-␣ v ␤ 3 mAb LM609 or anti-␣ v mAb L230. Co-localization of integrins with MT1-MMP was clearly seen at multiple cell surface sites and, specifically, at the cell protrusions (Fig. 5). Staining with control IgG was negative.
Second, co-precipitation of MT1-MMP with ␣ v ␤ 3 integrin confirmed the immediate proximity of these proteins and the existence of complexes between them on the cell surface. Thus, ␣ v ␤ 3 integrin from ␤3/MT-WT but not from ␤3/zeo cells coprecipitated with the characteristic 42-kDa autolytic species of MT1-MMP (Fig. 6A).
Third, MT1-MMP was co-purified with ␣ v ␤ 3 integrin from  1 g). The samples were separated by reducing SDS-PAGE and stained with Coomassie Blue. Positions of molecular weight markers are indicated on the left. B, purified LoVo pro-␣ v (0.2 g) was incubated alone or with rMT-cat (0.1 g) and AG3340 (1 M). The reduced samples were separated by SDS-PAGE in a 4 -20% gradient gel followed by Western blotting with antibody AB1930 specific to the cytoplasmic tail of ␣ v integrin and a secondary HRP-conjugated antibody. C, ␤3/zeo and ␤3/MT-WT cells were surface labeled with biotin and lysed. The biotinylated proteins were precipitated from the cell lysates with avidin-agarose and separated by SDS-PAGE under nonreducing conditions. Integrin ␣ v , ␣ 2 , ␣ 3 , and ␣ 5 subunits were identified by Western blotting using mAb AV1, AB1936, AB1920, and AB1949, respectively. D, schematic representation of pro-␣ v processing by MT1-MMP. The cleavage site is enlarged. The first MT1-MMP cleavage generates the 25-kDa light ␣ chain that is one residue shorter from its N terminus relative to the PC-processed ␣ chain. The C-terminal truncation distinguishes the 115-kDa heavy ␣ chain of the 140-kDa ␣ v subunit from the conventional 125-kDa heavy ␣ chain of the 150-kDa ␣ v subunit. The putative second MT1-MMP cleavage site is localized downstream from the Cys 852 . Glycosylation of the Cys 852 -Cys 904 loop is depicted by an asterisk.
Finally, co-cultivation of cells expressing either MT1-MMP or ␣ v ␤ 3 integrin failed to promote the cleavage of pro-␣ v . Thus, the ␤3/zeo cells were co-cultured for 48 h with the ␤ 3 -deficient MT-WT or MT-E240A cells. Then ␣ v ␤ 3 integrin was precipitated from the cell lysates and evaluated by immunoblotting (Fig. 6C). Evidently, co-cultivation did not affect the status of ␣ v ␤ 3 integrin in mixed cultures. It appears that simultaneous expression of both molecules in the same cell is required for the processing of pro-␣ v by MT1-MMP.
To study the functional implications of our findings, we analyzed whether the RGD ligand binding of ␣ v ␤ 3 integrin was affected in cells expressing MT1-MMP. For these purposes, ␣ v ␤ 3 integrin expressed by ␤3/zeo, MT-WT, and MT-E240A cells was immunoprecipitated using a recombinant Fab fragment, Fab-9, designed as an RGD ligand highly specific for ␤ 3 integrins (29,35). As shown in Fig. 7, this chimeric ligand was similarly efficient relative to anti-␣ v ␤ 3 mAb LM609 in precipitating all ␣ v ␤ 3 species.
Our further studies demonstrated that the alternative processing of pro-␣ v by MT1-MMP resulted in ␣ v ␤ 3 integrin that was superior relative to conventional ␣ v ␤ 3 in promoting cell adhesion, migration, and focal adhesion kinase phosphorylation (43). DISCUSSION Emerging evidence indicates that membrane-tethered MMPs such as MT1-MMP are directly involved in endoproteo-lytic modifications of cell surface receptors, including CD44, tissue transglutaminase, and ␣ v ␤ 3 integrin (16, 36 -38), enabling cells to adjust the receptor profile in the continually changing ECM environment. Our work identifies a novel functional link between MT1-MMP and ␣ v ␤ 3 integrin and demonstrates for the first time that, similarly to furin, MT1-MMP is able to specifically cleave pro-␣ v in tumor cells. This cleavage appears to occur at two distinct sites localized within a loop between the disulfide-bonded Cys 852 and Cys 904 of pro-␣ v . Cleavage at these sites was found to generate a 115-kDa heavy chain and a light chain commencing from the N-terminal Leu 892 . Relative to the PC-processed ␣ v subunit species, the MT1-MMP-processed heavy ␣ chain is C-terminally truncated, and the light ␣ chain is one residue shorter from its N terminus. Accordingly, PC processing of pro-␣ v at Asp 891 should prevent MT1-MMP cleavage at Leu 892 and vice versa. The loop bordered by Cys 852 and Cys 904 contains a putative glycosylation site at the Asn 874 . The loss of the glycosyl component in addition to the deletion of the C-terminal peptide sequence can explain the 10-kDa size difference between the 115-kDa heavy chain processed by MT1-MMP and the 125-kDa PC-processed heavy chain of the ␣ v integrin subunit.
It appears that MT1-MMP is capable of processing other PC-cleavable ␣ integrin subunits such as ␣ 3 and ␣ 5 . The ␣ chains that are resistant to PC cleavage (such as the ␣ 2 integrin chain) are also resistant to MT1-MMP (Table I). Relatively broad cleavage specificity of MT1-MMP is largely defined by the presence of a hydrophobic residue at the P 1 Ј position of the scissile bond (39). This may explain the ability of MT1-MMP to hydrolyze those multiple ␣ integrin chains that bear a hydrophobic residue at the P 1 Ј position. Processing of ␣ integrin subunits by MT1-MMP observed in MCF7 cells was confirmed in glioma U251 cells co-expressing ␣ v ␤ 3 integrin and MT1-MMP. The processing of pro-␣ v by MT1-MMP did not affect the ability of ␣ v ␤ 3 integrin to efficiently bind the RGD ligand. Our findings support the hypothesis that in migrating aggressive tumor cells overexpressing MT1-MMP novel, additional portions of the mature ␣ v integrins generated from pro-␣ v via the MT1-MMP pathway will co-localize with this proteinase, apparently, at the invasive cell front and the cell protrusions, the right place to additionally promote cell invasiveness.
Adhesive function of integrins has been associated with the transduction of biochemical signals into the interior of the cell (40). Integrin ligation normally induces outside-in signaling and phosphorylation of cytoplasmic tyrosine kinases, including focal adhesion kinase (41) that specifically modulates integrinmediated cell migration (42). In agreement with these observations, MT1-MMP-mediated processing of pro-␣ v -stimulated outside-in signal transduction through focal adhesion kinase (43). Accordingly, co-expression of MT1-MMP and ␣ v ␤ 3 integrin promoted cell adhesion and migration on vitronectin.
Overall, our findings identified a novel, MT1-MMP-dependent pathway of ␣ v ␤ 3 integrin maturation. This pathway is likely to be functionally important in aggressive tumor cells overexpressing MT1-MMP. The data also suggest that matrix breakdown may not be the primary function of MT1-MMP in tumor cells. In turn, localized proteolytic control of cell receptors is unlikely for soluble MMPs, while the proteolytic regulation of cell receptors including integrins may be an important function of MT1-MMP in migrating cells. This hypothesis may partially explain a unique functional role of MT1-MMP in tumor cell migration and invasion (15). However, since the available cell systems naturally express relatively low levels of MT1-MMP activity, pro-␣ v , or both, the question as to whether FIG. 6. Proximity of ␣ v ␤ 3 integrin and MT1-MMP in cells. A, the lysates of surface biotinylated ␤3/zeo and ␤3/MT-WT cells were precipitated with anti-␣ v ␤ 3 mAb LM609 and anti-MT1-MMP antibodies. The precipitates were analyzed by SDS-PAGE under reducing conditions, followed by Western blotting with Extravidin-HRP. Positions of molecular weight markers are indicated on the right. B, the lysates of ␤3/zeo, ␤3/MT-WT and ␤3/ E240A cells were subjected to affinity chromatography on a LM609 column as described under "Materials and Methods." The purified samples were analyzed by reducing SDS-PAGE, followed by Western blotting with rabbit antibodies against the ␤ 3 integrin subunit (upper panel) and rabbit anti-MT1-MMP antibodies (lower panel). MT1-MMP from glioma U251 cells transfected with the MT-WT construct (U-MT-WT cells) was used to assure the identification of the protease species. Positions of molecular weight markers are indicated on the right. C, ␤3/zeo cells were cultured alone or with MT-WT and MT-E240A cells. After incubation for 48 h, the lysates of surface-biotinylated cells were precipitated with anti-␣ v ␤ 3 mAb LM609. Precipitates were analyzed by reducing SDS-PAGE in 8% gel, followed by Western blotting with Extravidin-HRP. For comparison, ␣ v ␤ 3 integrin immunoprecipitated from ␤3/MT-WT cells is shown in the left lane.

TABLE I
The putative cleavage site of MT1-MMP in ␣ integrins s indicates the scissile bond hydrolyzed by MT1-MMP in the ␣ v integrin chain. 2 indicates the putative scissile bond susceptible to MT1-MMP in the sequence of PC-cleavable ␣ integrins. According to Fig. 4C, PC-resistant ␣ 2 appears to be insensitive to MT1-MMP. Hydrophobic residues at P 1 Ј are underlined.

Cleavage site Integrin
the cleavage of pro-␣ v occurs in untransfected cells remains to be answered. To specifically address this question, the experiments employing transfection of cells with ␣ 1 -antitrypsin Portland that inhibits processing of precursors mediated by proprotein convertases are currently under way in our laboratory.