High Affinity Binding of Latent Matrix Metalloproteinase-9 to the α2(IV) Chain of Collagen IV*

Association of matrix metalloproteinases (MMPs) with the cell surface and with areas of cell-matrix contacts is critical for extracellular matrix degradation. Previously, we showed the surface association of pro-MMP-9 in human breast epithelial MCF10A cells. Here, we have characterized the binding parameters of pro-MMP-9 and show that the enzyme binds with high affinity (K d ∼22 nm) to MCF10A cells and other cell lines. Binding of pro-MMP-9 to MCF10A cells does not result in zymogen activation and is not followed by ligand internalization, even after complex formation with tissue inhibitor of metalloproteinase-1 (TIMP-1). A 190-kDa cell surface protein was identified by ligand blot analysis and affinity purification with immobilized pro-MMP-9. Microsequencing and immunoblot analysis revealed that the 190-kDa protein is the α2(IV) chain of collagen IV. Specific pro-MMP-9 surface binding was competed with purified α2(IV) and was significantly reduced after treatment of the cells with active MMP-9 before the binding assay since α2(IV) is hydrolyzed by MMP-9. A pro-MMP-9·TIMP-1 complex and MMP-9 bind to α2(IV), suggesting that neither the C-terminal nor the N-terminal domain of the enzyme is directly involved in α2(IV) binding. The closely related pro-MMP-2 exhibits a weaker affinity for α2(IV) compared with that of pro-MMP-9, suggesting that sites other than the gelatin-binding domain may be involved in the binding of α2(IV) to pro-MMP-9. Although pro-MMP-9 forms a complex with α2(IV), the proenzyme does not bind to triple-helical collagen IV. These studies suggest a unique interaction between pro-MMP-9 and α2(IV) that may play a role in targeting the zymogen to cell-matrix contacts and in the degradation of the collagen IV network.

The degradation of ECM 1 components is partly achieved by proteolytic enzymes closely associated with discrete areas of cell-matrix contacts. These include the plasminogen/plasmin system (1), cathepsins (2), and MMPs (3)(4)(5)(6). The gelatinases A (MMP-2) and B (MMP-9) are two members of the MMP family that have been shown to play a central role in many normal and pathological conditions involving ECM degradation, including wound healing, angiogenesis, embryogenesis, arthritis, and tumor metastasis (3)(4)(5)(6). Like other members of the MMP family, the gelatinases are produced in a latent form (pro-MMP) 2 that requires activation to become proteolytically active. Thus, activation is a critical step in the regulation of MMP-dependent proteolytic activity. Considerable evidence has associated the gelatinases with the ability of tumor cells to metastasize due to their ability to degrade basement membrane collagen IV and to their elevated expression in malignant tumors (5,6). Previous studies have shown that tumor cells contain gelatinase activity in the plasma membranes consistent with cell surface association (7)(8)(9)(10). We have shown a surface localization of both gelatinases in carcinoma cells of breast tumors (11,12) and epithelial cells of fibrocystic breast disease (12), and others have shown the localization of pro-MMP-9 in the tumor basement membrane zone of skin tumors (13) colocalizing with collagen IV. 3 In recent years, much information has been gained on the cell surface association of pro-MMP-2. It has been shown that pro-MMP-2 binds to the cell surface via membrane type 1-MMP (MT1-MMP), a subclass of MMPs bound to plasma membranes (14,15). Interestingly, the binding of pro-MMP-2 to MT1-MMP requires the participation of tissue inhibitor of metalloproteinase-2 (TIMP-2) (14). The trimer complex allows for pro-MMP-2 activation on the cell surface, possibly by another MT1-MMP molecule. A later study suggested that the cell surface association of pro-MMP-2 can also occur through the binding of the C-terminal domain of the enzyme to integrin ␣ v ␤ 3 (16). Thus, several mechanisms may play a role in the surface localization of pro-MMP-2.
Pro-MMP-9 is structurally similar to pro-MMP-2, with both enzymes containing three tandem copies of a 58-amino acid residue fibronectin type II-like module (17,18), known as the gelatin-binding domain, which plays a role in binding to extracellular matrix components (19 -21). Furthermore, both latent enzymes can bind a TIMP molecule in the C-terminal domain of the zymogen with pro-MMP-9 binding to TIMP-1 (17) and pro-MMP-2 to TIMP-2 (4,5). Pro-MMP-9 also contains a 54amino acid proline-rich extension of unknown function that is similar to the ␣2(V) chain of collagen V and in addition is a glycosylated enzyme (17). In contrast to pro-MMP-2, little is known about the interactions of pro-MMP-9 with the cell surface, yet pro-MMP-9 has been found to be present on the cell surface. Recent studies by Partridge et al. (22) showed that pro-MMP-9 is present in the plasma membrane and focal contacts of cultured endothelial cells. Pro-MMP-9 was also detected in the plasma membranes of human fibrosarcoma HT1080 cells (23), and the enzyme could be activated by a plasmin-dependent mechanism (24). We have recently shown that, upon induction with 12-O-tetradecanoylphorbol-13-acetate, pro-MMP-9 binds to the surface of human breast epithelial MCF10A cells (25). In this report, we further examined the binding of pro-MMP-9 to a variety of cell lines and demonstrated the existence of a single high affinity (K d ϳ22 nM) binding site. Using immobilized pro-MMP-9, ligand blot analysis and co-immunoprecipitation experiments with surface biotinylated cells, we have identified the ␣2(IV) chain of collagen IV as the major pro-MMP-9-binding protein. These studies provide novel evidence on the interactions of pro-MMP-9 with ␣2(IV) that may play a role in the localization of zymogen at cell-matrix contacts and degradation of basement membranes.

Expression and Purification of Recombinant Gelatinases and TIMP-1-
Human recombinant pro-MMP-9, pro-MMP-2, and TIMP-1 were all expressed in HeLa cells using a recombinant vaccinia virus expression system and purified to homogeneity as described previously (26,27). To obtain 35 S-pro-MMP-9 and 35 S-pro-MMP-2, expression of recombinant enzymes in infected HeLa cells was carried out in the presence of 15 Ci/ml [ 35 S]methionine. The specific activities of 35 S-pro-MMP-2 and 35 S-pro-MMP-9 were 0.0159 Ci/pmol and 0.0106 Ci/pmol, respectively.
Iodination of Pro-MMP-9 and TIMP-1-Pro-MMP-9 or TIMP-1 (50 -100 g) in 100 l of collagenase buffer (50 mM Tris-HCl (pH 7.5), 5 mM CaCl 2 , 150 mM NaCl, and 0.02% Brij-35) were placed in a vial coated with IODOGEN (Pierce) as described by the manufacturer and allowed to incubate for 1 min at 25°C. Na 125 I (500 Ci) was added to each vial, and the iodination reaction was allowed to continue for 3 min at 25°C. The reaction was quenched by the addition of 200 g (100 l) of bovine serum albumin (BSA) and 2 mM NaI. Unincorporated Na 125 I was removed by a 1-ml Sephadex G-25 (fine) column equilibrated with collagenase buffer. The specific activity of 125 I-pro-MMP-9 and 125 I-TIMP-1 was determined after trichloroacetic acid precipitation and quantitation by densitometry of the proteins in Coomassie Blue-stained SDSpolyacrylamide gels relative to a standard curve of unlabeled purified proteins. Typically, the specific activities of 125 I-TIMP-1 and 125 I-pro-MMP-9 were 0.049 Ci/pmol and 0.322 Ci/pmol, respectively. No detectable autocatalytic/degradation forms of MMP-9 were observed in the iodinated enzyme as determined by both gelatin-zymography and autoradiography.
Cells-Human immortalized breast epithelial MCF10A cells (31) were grown as described previously (25). MDA-MB-231 breast cancer cells and PC3 prostate cancer cells were provided by Dr. Fred Miller (Karmanos Cancer Institute, Detroit, MI) and grown in DMEM supplemented with 10% fetal bovine serum (FBS) and RPMI 1640 medium supplemented with 10% FBS, respectively. Human HT1080 fibrosarcoma (CCL-121) cells were obtained from the American Type Culture Collection (Rockville, MD) and grown in DMEM supplemented with 10% FBS. Rat vascular endothelial (RVE) cells derived from brain capillaries (32) and mouse lung microvascular endothelial (CD3) cells were provided by Dr. Diglio (Department of Pathology, Wayne State University) and grown in DMEM supplemented with 10% FBS.
Binding Assays-Cells (MCF10A, MDA-MB-231, PC3) grown to 80% confluence (3-5 days after seeding) in 12-well (22-mm) plates were rinsed with PBS and incubated (15 min at 4°C) with cold binding medium (25 mM Hepes (pH 7.5) with 0.5% BSA in DMEM). The medium was aspirated, and various concentrations (1-40 nM) of 125 I-pro-MMP-9 diluted in binding medium were added to each well (300 l) in the presence or absence of 80-fold excess unlabeled pro-MMP-9. After a 45-min incubation at 4°C, the medium was aspirated and the cells were washed three times with cold PBS containing 0.1% BSA. The cells were lysed with 0.5 ml/well of 0.5 M NaOH for determination of radioactive counts in a ␥ counter (Packard model 5650), and the results were expressed as the mean of the values obtained from triplicate samples. The number of cells in each well was determined in quadruplicate wells. Time-course experiments were similarly done, except that the concentration of 125 I-pro-MMP-9 for each well was kept constant at 18 nM and the cells were harvested after various times at 4°C. The nonspecific binding of 125 I-pro-MMP-9 was determined in the presence of 80-fold excess unlabeled pro-MMP-9. Typically, specific binding represented an average of approximately 20 -30% of total bound 125 I-pro-MMP-9. The association rate constant (k on ) of pro-MMP-9 was determined from the time-course experiment following logarithmic transformation of the amount of specifically bound 125 I-pro-MMP-9 versus time. Binding of 125 I-pro-MMP-9 to the cell surface follows a second-order binding isotherm. Thus, the k on is determined from the slope of a line plotted as ln([LReq]/([LReq] Ϫ [LR])) versus time as described previously (33), where LReq is the quantity of 125 I-pro-MMP-9 bound at 120 min and LR is the quantity of 125 I-pro-MMP-9 bound at the times 0, 2, 15, and 30 min. The slope of this line was determined by linear regression analysis using Microsoft Excel™, and the error represents the standard deviation of the slope. The equilibrium binding constant (K d ) and the number of binding sites per cell were determined by nonlinear curve-fitting analysis using the GraphPad Prism™ software version 2.0 and by Scatchard analysis. The slope and the intercept from the Scatchard analysis were determined by linear regression analysis using Microsoft Excel™. For the latter, the error represents the standard deviation of the slope and intercept. For HT1080 and RVE cells, binding was performed with 125 I-pro-MMP-9 (3.6 nM) in the presence and absence of 80-fold excess unlabeled pro-MMP-9 for 45 min at 4°C and the amount of specific ligand bound (fmol/cells) was determined. Competition of 125 I-pro-MMP-9 binding with pro-MMP-2 was carried out in a ligand binding assay as described above, except that 80-fold excess unlabeled pro-MMP-2 was used instead of unlabeled pro-MMP-9.
Cellular Distribution of Bound 125 I-Pro-MMP-9 -MCF10A cells were incubated with cold binding medium and then incubated with 18 nM/ well of 125 I-pro-MMP-9 in triplicate wells for 45 min at 4°C. The medium was aspirated, and the cells were washed four times with cold binding medium. Each well then received 0.5 ml of prewarmed (37°C) binding medium, and the plates were incubated at 37°C for various times. At each time point, the medium was recovered and the cells were washed with PBS, followed by the addition of 0.5 ml/well of 0.25% Pronase E (Sigma) in PBS. The cells were incubated at 4°C for 30 min, and the monolayer was dislodged by gentle pipetting and transferred to a microcentrifuge tube. The samples were centrifuged for 5 min at 2000 ϫ g, and the supernatant (cell surface-bound fraction) was transferred to a new tube. The pellet (internalized fraction) was washed once with PBS and then resuspended in 0.5 ml of PBS. The radioactivity of the three fractions, in triplicate, was measured in a ␥ counter. Internalization studies in the presence of TIMP-1 were performed similarly, except that the cells were incubated with a 125 I-pro-MMP-9⅐TIMP-1 complex that was previously formed by incubating 125 I-pro-MMP-9 with TIMP-1 for 30 min at 22°C.
Competition of 125 I-Pro-MMP-9 Binding with Purified ␣2(IV)-125 I-Pro-MMP-9 (3.6 nM) was incubated for 1.5 h at 4°C with ϳ2-fold molar excess of affinity-purified ␣2(IV) at a final concentration of 7 nM in the presence of 5 mM EDTA. Binding assays with MCF10A cells were carried out as described above.
Binding of 125 I-Pro-MMP-9 to Cells Treated with MMP-9 -MCF10A cells were incubated for 45 min at 37°C in binding medium with 1.2 pmol/well (final concentration of 4 nM) of either purified 82-kDa active species of MMP-9 or MMP-9 that was previously incubated with TIMP-1 (2.4 pmol) for 30 min at 22°C. After treatment, the cells were washed twice with cold binding medium and incubated for 15 min in binding medium at 4°C. Binding assays were carried out as described above.
Coupling of Pro-MMP-9 to Affi-Gel 10 -One milligram of recombi-nant pro-MMP-9 in 50 mM Hepes (pH 7.5), 5 mM CaCl 2 , 150 mM NaCl, and 0.02% Brij-35 was allowed to bind to 1 ml of Affi-Gel 10 (Bio-Rad) for 5 h at 4°C with rotation in the presence of 60 mM CaCl 2 . After coupling, the matrix (Affi-Gel 10-pro-MMP-9) received a 150-l volume of 1 M ethanolamine (pH 8) and incubated with rotation for 1 h at 4°C. The matrix was allowed to settle, and the supernatant was subjected to SDS-PAGE to determine the amount of uncoupled pro-MMP-9. The Affi-Gel 10-pro-MMP-9 matrix was washed four times with collagenase buffer. The immobilized pro-MMP-9 maintained its capability to bind TIMP-1, as determined by binding of 125 I-TIMP-1 compared with soluble enzyme. Affinity Purification of the 190-kDa Protein-MCF10A cells were lysed with (0.8 ml/150-mm plate) ice-cold lysis buffer (25 mM Tris (pH 7.5), 100 mM NaCl, 1% Nonidet P-40, 10 g/ml aprotinin, 1 g/ml leupeptin, 5 mM benzamidine, and 1 mM PMSF). After a centrifugation (20 min, 14,000 rpm), the supernatant was collected and incubated (4°C) with Affi-Gel 10-pro-MMP-9 batchwise overnight, poured into a column (Polyprep™, Bio-Rad), and the flow-through fraction collected. The matrix was washed with 20 ml of 25 mM Tris (pH 7.5), 500 mM NaCl, 0.1% Nonidet P-40, 2 mM PMSF, 5 mM benzamidine, 10 g/ml leupeptin, and 10 g/ml aprotinin (wash 1), followed by 10 ml of the same buffer as described above but containing 150 mM NaCl (wash 2). The 190-kDa protein was eluted from the column with 4.5 ml of 50 mM Tris (pH 7.5), 150 mM NaCl, 5 mM CaCl 2 , 20% Me 2 SO, 2 mM PMSF, 5 mM benzamidine, 10 g/ml leupeptin, and 10 g/ml aprotinin. Three 1.5-ml fractions of eluate were collected. Forty microliters of the load, flow-through, wash 1, wash 2, and eluate fractions were analyzed by silver-stained SDS-PAGE (34) and ligand blot as described below. The protein concentrations of each column fraction were determined by the BCA protein assay (Pierce) reagent. The same procedure was used with lysates of surface biotinylated cells, except that 0.5-1 ml of lysates were incubated with 50 l of Affi-Gel 10-pro-MMP-9 or uncoupled Affi-Gel 10 matrix. After binding, the matrix was washed with 1 ml of cold HNTG buffer (50 mM Tris (pH 7.5), 150 mM NaCl, 0.1% Nonidet P-40, 10% glycerol) followed by one wash in the same buffer containing 500 mM NaCl and three washes in HNTG buffer. After a brief centrifugation, the bound proteins were eluted with 40 l of collagenase buffer with 10% Me 2 SO and subjected to SDS-PAGE, blotting, and detection by streptavidin-HRP or ligand blot as described below.
Microsequencing-The affinity-purified 190-kDa protein was subjected to 7% SDS-PAGE under reducing conditions, followed by staining with 0.25% Coomassie Brilliant Blue. The band containing the 190-kDa protein was excised from the gel, and the pieces were washed three times (15 min each) with Millipore water. The gel slices were washed three times (5 min each) in 50% acetonitrile (Aldrich, HPLC grade), frozen in dry ice, and then sent to Dr. William Lane at the Harvard Microchemistry sequencing facility (Cambridge, MA).
Cell Surface Biotinylation-Surface proteins were biotinylated with sulfo-NHS-biotin (Pierce) as described previously (25). The biotinylated cells were lysed with 2 ml/dish of ice-cold lysis buffer, the lysates incubated for 1 h on ice, and the supernatant collected after a 15-min centrifugation (13,000 ϫ g) at 4°C. The supernatants were analyzed immediately for the presence of pro-MMP-9-binding proteins by pro-MMP-9-affinity chromatography or co-immunoprecipitation with pro-MMP-9 and detection with streptavidin-HRP as described below.
Preparation of Conditioned Medium-Confluent cultures of MCF10A cells were incubated with serum-free DMEM/F-12 (15 ml/150-mm dish) for 24 h at 37°C. The medium was collected, centrifuged, and concentrated (6-fold) with a Centriprep-10 (Amicon). The medium was analyzed by immunoblot analysis as described above. In some experiments, cells were grown in the presence of daily additions of ascorbate (75 g/ml) and the conditioned medium was obtained as described.
Co-immunoprecipitation of ␣2(IV) with Pro-MMP-9 -Lysates of surface-biotinylated or non-biotinylated cells in lysis buffer or samples of concentrated serum-free conditioned medium were incubated with 0.5-1 g/ml purified pro-MMP-9 for 1 h at 4°C. In some experiments, the lysates received active MMP-9, pro-MMP-2, active MMP-2, or TIMP-1 in the presence or absence of 10 mM EDTA. Five micrograms of the appropriate antibody or control IgG were added for another 16-h incubation at 4°C. Each sample received 25 l of protein G-Sepharose 4 Fast Flow matrix (Amersham Pharmacia Biotech) and incubated for 3 h at 4°C with continuous rocking. The matrix was washed with ice-cold HNTG buffer, followed by one wash in the same buffer supplemented with 500 mM NaCl and three additional washes with HNTG buffer. The samples were boiled in the presence of 20 l of 2ϫ Laemmli sample buffer (35) with reducing agent, and the supernatants were subjected to SDS-PAGE and the proteins transferred to a nitrocellulose membrane for immunoblot analysis or ligand blot analysis as described below.
Immunoblot and Ligand Blot Analysis-Samples resolved by SDS-PAGE were transferred to a BA-S 85 nitrocellulose membrane (Schleicher & Schuell). The membrane was blocked (12 h at 4°C) with Blotto (3% BSA and 3% nonfat dry milk in 100 mM Tris-HCl (pH 7.5), 150 mM NaCl, and 0.02% NaN 3 ) and washed twice with T-TBS (20 mM Tris (pH 7.5), 137 mM NaCl, and 0.1% Tween 20). For immunoblot analysis, the membranes were incubated with the appropriate primary antibody diluted in T-TBS containing 0.5% nonfat dry milk, washed three times with T-TBS and incubated with the HRP-conjugated secondary antibody. The immunodetection of the antigen was performed using ECL (Amersham Pharmacia Biotech) according to the manufacturer's instructions. For biotinylated samples, the membranes were incubated with streptavidin-HRP (Amersham Pharmacia Biotech) and developed by ECL. For ligand blot analysis, the membranes were incubated (1 h at 25°C) with 1 g/ml pro-MMP-9 in T-TBS containing 0.5% nonfat dry milk and 1% Nonidet P-40, followed by three washes with T-TBS. After washing, the membranes were incubated (1 h at 25°C) with anti-MMP-9 antibodies diluted in T-TBS containing 0.5% nonfat dry milk. After three washes in T-TBS, the membranes were incubated (1 h at 25°C) with the HRP-conjugated secondary antibody in T-TBS, followed by three washes with T-TBS. The proteins were detected as described above.
Binding of Mouse-EHS Collagen IV to Immobilized Pro-MMP-9 -EHS native collagen IV (a generous gift from Dr. Hynda Kleinman, National Institutes of Health, Bethesda, MD) was diluted (10 g/ml) in lysis buffer (0.5 ml) and incubated (12 h at 4°C) with 30 l of Affi-Gel 10-pro-MMP-9 matrix. After incubation, the beads were centrifuged (13,000 ϫ g, 1 min at 4°C) and the supernatant was collected (unbound fraction). The beads were washed with ice-cold HNTG buffer, twice with HNTG buffer containing 500 mM NaCl and a final wash in HNTG buffer. The beads were resuspended in 50 l of 10% Me 2 SO in collagenase buffer and mixed for 30 min at 4°C, followed by a brief centrifugation to obtain the supernatant (bound fraction). The bound and unbound fractions were mixed with sample buffer under reducing conditions and subjected to SDS-PAGE, followed by ligand blot analysis as described above.
Determination of the Affinity of ␣2(IV) for Pro-MMP-2 and Pro-MMP-9 -Affinity-purified ␣2(IV) (9.5 nM) was allowed to complex with varying concentrations of either 35 S-pro-MMP-2 (20 -750 nM) or 35 S-pro-MMP-9 (2.5-250 nM) for 1 h at 4°C. The samples were then subjected to gel filtration using a Superose 6 column (10/30) (Amersham Pharmacia Biotech) equilibrated with 50 mM Tris (pH 7.5), 150 mM NaCl, 5 mM EDTA, and 0.02% Brij-35 at a flow rate of 0.2 ml/min, and 70 fractions (0.3 ml) were collected. Aliquots (0.2 ml) of each fraction were placed in 5 ml of scintillation fluid and counted to determine the amount (pmol) of enzyme bound. Complex formation was evaluated based on the relative inclusion volume of pro-MMP-2, pro-MMP-9, or ␣2(IV) alone or in complex. The dissociation constant (K d ) was determined from the binding profiles (picomoles of pro-MMP bound versus pro-MMP concentration) by extrapolating the point where 50% maximal binding was obtained.

Characteristics of 125 I-Pro-MMP-9 Binding to MCF10A
Cells-Previously, we showed that exposure of MCF10A cells to 12-O-tetradecanoylphorbol-13-acetate results in the secretion of pro-MMP-9 that associates with the cell surface (25). To measure the binding of pro-MMP-9, we carried out binding assays of radioiodinated enzyme to untreated MCF10A cells to avoid interference with the endogenously produced pro-MMP-9 (25). Fig. 1A shows time-dependent binding of 125 I-pro-MMP-9 to MCF10A cells at 4°C. Logarithmic transformation of the data revealed a k on value of 6.94 Ϯ 0.67 ϫ 10 4 M Ϫ1 s Ϫ1 (Fig. 1A,  inset). Incubation of the cells with increasing concentrations (1-40 nM) of 125 I-pro-MMP-9 in the presence of 80-fold excess of unlabeled enzyme demonstrated a specific saturable binding (Fig. 1B). Scatchard analysis revealed a K d value of 2.16 Ϯ 0.16 ϫ 10 Ϫ8 M and 1.3 ϫ 10 5 sites/cell (Fig. 1B, inset). Similar values were obtained after nonlinear curve-fitting analysis of the data (K d ϭ 2.49 Ϯ 0.27 ϫ 10 Ϫ8 M and 1.4 ϫ 10 5 sites/cell). We found no evidence of cell surface activation of 125 I-pro-MMP-9, as determined by SDS-PAGE followed by autoradiography or by gelatin zymography (data not shown).
Binding specificity for pro-MMP-9 was determined in MCF10A cells using pro-MMP-2 as a competitor as described under "Experimental Procedures." These studies showed that pro-MMP-2 caused a 27% decrease in specific 125 I-pro-MMP-9 binding with 3.3 fmol/1.3 ϫ 10 5 cells in the presence of unlabeled pro-MMP-2 compared with 4.5 fmol/1.3 ϫ 10 5 cells with pro-MMP-9 alone. Thus, pro-MMP-2 slightly competes with pro-MMP-9 for the same binding site on MCF10A cells.
Cellular Distribution of 125 I-Pro-MMP-9 and a 125 I-Pro-MMP-9⅐TIMP-1 Complex after Binding to MCF10A Cells-We investigated the fate of pro-MMP-9 after binding using ligand internalization studies as described under "Experimental Procedures." Fig. 2A shows that the amount of 125 I-pro-MMP-9 associated with the cell pellet (internalized ligand) reached a maximum of 21% of the total ligand bound after 10 min of incubation at 37°C. Later time points showed little or no change (Ͻ5%) in the radioactivity of the pellet fraction. Concurrently, after a 10-min incubation, ϳ80% of the total cellassociated radioactivity was recovered by Pronase E digestion (representing cell surface-bound enzyme). This fraction exhibited a decline in radioactivity consistent with dissociation of cell surface-bound ligand and its appearance in the supernatant fraction ( Fig. 2A). SDS-PAGE analysis demonstrated that the radioiodinated enzyme in the supernatant fraction remained in the latent form at all times examined. In contrast, the internalized enzyme was degraded to low (Ն20 kDa) molecular mass fragments (data not shown). Thus, binding of pro-MMP-9 to MCF10A cells is not followed by significant ligand internalization.
Binding studies with urokinase plasminogen activator demonstrated that binding of this activator to its receptor in the presence of plasminogen activator inhibitor-1 results in rapid internalization of the enzyme/inhibitor complex (36). To determine if a similar process could occur with pro-MMP-9, we examined the cellular partitioning of a 125 I-pro-MMP-9⅐TIMP-1 complex after binding to MCF10A cells as described above. These experiments showed no significant differences in the cellular distribution of the free ( Fig. 2A) versus that of the complex (Fig. 2B) ligand after incubation at 37°C for up to 2 h. Furthermore, the association of pro-MMP-9 with TIMP-1 did not alter the rate of ligand dissociation into the supernatant fraction as Ͼ90% of the radioactivity in this fraction was tri-chloroacetic acid-precipitable. Additionally, autoradiograms of the samples after SDS-PAGE showed no differences in cellular distribution between free and complexed zymogen (data not shown).
Identification of a 190-kDa Cell Surface Pro-MMP-9-binding Protein-The above binding data were consistent with the existence of a single pro-MMP-9-binding component on the cell surface. To identify the putative pro-MMP-9-binding protein, we carried out an affinity purification using immobilized pro-MMP-9 (Affi-Gel 10-pro-MMP-9). Lysates of surface-biotinylated MCF10A cells were incubated with Affi-Gel 10-pro-MMP-9 matrix or uncoupled Affi-Gel 10 and the bound proteins were eluted with 10% Me 2 SO. The eluted proteins were then detected by either streptavidin-HRP or ligand blot analysis as described under "Experimental Procedures." Fig. 3 shows that MCF10A cells express a major 190-kDa protein that specifically and consistently binds to the Affi-Gel 10-pro-MMP-9 matrix (Fig. 3, A, lane 2; and B, lanes 2, 4, and 5) but not to the uncoupled matrix (Fig. 3, A and B, lanes 1 and 3). Several minor biotinylated proteins (ϳ70 -90 kDa) were found to bind, albeit inconsistently, to the pro-MMP-9 affinity matrix (Fig.  3A, lane 2). The 190-kDa protein was detected by streptavidin-HRP (Fig. 3A, lane 2) consistent with biotinylation and consequently, cell surface localization. The streptavidin detection was specific since blots of samples derived from lysates of non-biotinylated cells and incubated with Affi-Gel 10-pro-MMP-9 matrix were negative (Fig. 3A, lane 4). The presence of the 190-kDa protein in the non-biotinylated cells was confirmed by ligand blot analysis (Fig. 3B, lane 4). Electrophoretic migration of the 190-kDa protein was similar under nonreducing (Fig. 3B, lane 5) or reducing conditions (Fig. 3, A, lane 2,  and B, lanes 2 and 4). In contrast, under nonreducing conditions, pro-MMP-9 exhibited the presence of monomer and dimer forms (Fig. 3B, lane 6), as expected (37).
To determine whether the 190-kDa protein forms a precipitable complex with pro-MMP-9, lysates of MCF10A cells were incubated with exogenous pro-MMP-9 and subjected to a coimmunoprecipitation experiment with anti-MMP-9 antibodies as described under "Experimental Procedures." The immunoprecipitates were analyzed by ligand blot analysis. This experiment showed that the 190-kDa protein co-precipitates with pro-MMP-9 (Fig. 3C). In the absence of enzyme, the 190-kDa protein was not detected, demonstrating the specificity of the interaction. Taken together, these results demonstrate that MCF10A cells express a 190-kDa cell surface protein that specifically binds to pro-MMP-9.
Affinity Purification of the 190-kDa Protein and Hydrodynamic Studies-The 190-kDa protein was purified from lysates of MCF10A cells using the Affi-Gel 10-pro-MMP-9 matrix as described under "Experimental Procedures." Fig. 4 shows the analysis of the column fractions in a silver-stained SDS-PAGE (Fig. 4A) and by ligand blot analysis (Fig. 4B). The 190-kDa protein was detected in the 20% Me 2 SO eluate (Fig. 4, A and B,  lane 5). Analysis of the protein content in each fraction indicated a ϳ3300-fold purification. The affinity-purified 190-kDa protein was subjected to hydrodynamic studies for determination of the sedimentation coefficient and Stokes radius to determine the native molecular mass. As shown in Fig. 5A, the 190-kDa protein was detected in six fractions derived from the gradient as shown by ligand blot analysis (Fig. 5A, inset). A sedimentation coefficient of 7.9 S was calculated from two determinations using known protein standards. The Stokes radius of the 190-kDa protein was determined by gel filtration using protein standards (Fig. 5B) and was calculated to be 52.6 Å. By combining the sedimentation and Stokes radius (38), a native molecular mass of 192,000 was calculated. These data and the electrophoretic mobility of the 190-kDa protein on SDS-PAGE (Fig. 3) indicate that this protein is monomeric.
The 190-kDa Pro-MMP-9-binding Protein Is the ␣2(IV) Chain of Collagen IV-The affinity-purified 190-kDa protein was submitted for microsequencing as described under "Experimental Procedures." Analyses of three HPLC-purified peptides  1-4) or nonreducing (lanes 5 and 6) conditions and transferred to a nitrocellulose membrane. Detection of the pro-MMP-9-binding protein was performed by either streptavidin-HRP (A) or ligand blot analysis (B) with pro-MMP-9 as a probe followed by detection with anti-MMP-9 antibody (CA-209) and anti-mouse IgG-HRP using ECL. Lane 6 shows the monomer and dimer forms of purified pro-MMP-9 (10 ng). C, co-precipitation of the 190-kDa protein with pro-MMP-9. A lysate (0.5 ml) of MCF10A cells was incubated (ϩ) or not (Ϫ) with pro-MMP-9 (1 g/ml), followed by addition of rabbit polyclonal anti-MMP-9 antibodies (5 g) and then immunoprecipitated with protein G-Sepharose beads as described under "Experimental Procedures." The immunoprecipitates were subjected to 4 -12% SDS-PAGE under reducing conditions, blotting to nitrocellulose and detection by ligand blot. obtained after tryptic digestion revealed the following amino acid sequences: GVSGFPGADGIPGHPGQGGP, DGYQGPDG-PRG, and KIAIQPGTVGPQG, which correspond to residues 109 -128, 325-335, and 1396 -1408, respectively, of the human ␣2(IV) chain of collagen IV (39). Immunoblot analysis of the 190-kDa protein demonstrated reactivity with three different mAbs (H22, H25, and H21) against the human ␣2(IV) chain (Fig. 6A). Furthermore, co-immunoprecipitation of the 190-kDa protein from an MCF10A cell lysate with exogenous pro-MMP-9 using an anti-MMP-9 antibody followed by immunoblot analysis with H22 mAb further demonstrated that the coprecipitated 190-kDa protein is the ␣2(IV) chain (Fig. 6B).
Expression of ␣2(IV) in Cultured Cells and Binding to Pro-MMP-9 -We examined the expression of ␣2(IV) in cell extracts and on the surface of MCF10A, MDA-MB-231, HT1080, RVE, and CD3 cells by surface biotinylation, immunoblots and ligand blot analyses. Fig. 7 shows that all the cells examined express ␣2(IV) that specifically bound to the pro-MMP-9-affinity matrix (Fig. 7, A-D) or co-precipitated with pro-MMP-9 (Fig. 7E). Surface biotinylation followed by Affi-Gel 10-pro-MMP-9 purification demonstrated that ␣2(IV) is readily detected on the surface of MCF10A (Fig. 7B,  lane 1), HT1080 (Fig. 7B, lane 3) and RVE (Fig. 7D) cells but could not be detected on the surface of MDA-MB-231 cells (Fig. 7B, lane 2). However, in the latter, ␣2(IV) was identified in the cell lysate (Fig. 7A, lane 2) consistent with the ligand binding data demonstrating a reduced number of pro-MMP-9 binding sites/cell in the MDA-MB-231 cells. Expression of ␣2(IV) was also detected in mouse endothelial CD3 cells by co-immunoprecipitation with pro-MMP-9 and detection by ligand blot analysis (Fig. 7E, panel 5). Thus, these studies demonstrate that monomeric ␣2(IV) is expressed by a variety of cells and can be detected on the cell surface.
Binding of Pro-MMP-9 to Collagen IV-The identification of ␣2(IV) and its binding to pro-MMP-9 raised the question about interactions of this enzyme with triple-helical collagen IV. Since the most abundant collagen IV is a heterotrimeric protein composed of two ␣1(IV) and one ␣2(IV) chains (40), we examined whether MCF10A cells produce trimeric collagen IV and whether it binds to pro-MMP-9. Concentrated serum-free conditioned medium of MCF10A cells was subjected to SDS-PAGE under nonreducing conditions, followed by immunoblot analysis using various antibodies to collagen IV. As shown in Fig. 8, mAbs to either the ␣2(IV) (Fig. 8, lane 1, mAb 1910) or ␣1(IV) (Fig. 8, lane 2, H11) chains recognized a protein of Ͼ450 kDa, likely to represent trimeric collagen IV. In contrast, H22 antibodies against ␣2(IV) did not react with trimeric collagen IV (Fig. 8, lane 3) suggesting that its epitope is not exposed under nonreducing conditions. A protein of ϳ160 kDa was also detected, although to a lower extent, with both H11 and H22 antibodies. The nature of this protein is unknown. It should be noted that, under these conditions, monomeric ␣2(IV) chain  could not be detected in the media of MCF10A cells by immunoblot and ligand blot analysis. Furthermore, growth of MCF10A cells in the presence of daily added ascorbate (75 g/ml) did not alter the localization and amounts of the ␣2(IV) chain (data not shown).
The ability of pro-MMP-9 to bind collagen IV was tested by a co-immunoprecipitation experiment after the addition of pro-MMP-9 to the media of MCF10A cells. As control, pro-MMP-9 was also added to a cell lysate to co-precipitate ␣2(IV). These studies demonstrated that, although the cell-associated ␣2(IV) coprecipitated with pro-MMP-9 (Fig. 9A, lane 2), the secreted trimeric collagen IV did not (Fig. 9A, lane 1). Additionally, the collagen IV in the conditioned medium did not bind to the pro-MMP-9-affinity matrix (data not shown). The interaction of pro-MMP-9 with native collagen IV was further investigated using purified EHS collagen IV and the pro-MMP-9-affinity matrix as described under "Experimental Procedures." After the affinity step, the bound and unbound fractions were analyzed by ligand blot analysis under reducing conditions for the presence of mouse collagen IV chains. Fig. 9B shows that mouse collagen IV did not bind to the affinity matrix (Fig. 9B,  lane 2). However, the ␣1(IV) and ␣2(IV) chains were readily detected in the unbound fraction (Fig. 9C, lane 2). It should be noted that, under the conditions of the ligand blot analysis, the mouse ␣1(IV) chain was detected. Taken together, these results suggest that pro-MMP-9 binds to monomeric ␣2(IV) with an affinity greater than that to triple-helical collagen IV.
Affinity of Pro-MMP-9 and Pro-MMP-2 for ␣2(IV)-Since pro-MMP-9 bears a high degree of sequence similarity with pro-MMP-2 and both possess a gelatin-binding domain (41), we determined the ability of pro-MMP-2 to compete for the binding of ␣2(IV) to pro-MMP-9. The ␣2(IV) was co-imunoprecipitated with pro-MMP-9 and anti-MMP-9 antibodies in the absence or presence of pro-MMP-2 (equimolar or 5-fold molar excess, relative to pro-MMP-9) and in the presence of EDTA to prevent degradation of ␣2(IV) by any trace of MMP-2. The immunoprecipitates were resolved by ligand blot analysis. As shown in Fig.  11, ␣2(IV) co-precipitates only in the presence of pro-MMP-9 (Fig. 11A, lanes 2-4). No significant differences in the amounts of ␣2(IV) co-precipitating with pro-MMP-9 were detected in the presence of equal molar amounts (Fig. 11A, lane 3) or excess (5-fold) molar amounts of pro-MMP-2 (Fig. 11A, lane 4), suggesting that under these conditions only pro-MMP-9 bound to ␣2(IV). In agreement with these results, ␣2(IV) incubated with pro-MMP-2 alone did not form a co-precipitable complex as determined using anti-MMP-2 antibodies and ligand blot analysis (Fig. 11A, lane 6). Consistently, pro-MMP-2 was detected in the immunoprecipitates when the same blot was developed with anti-MMP-2 antibodies (Fig. 11B, lane 8).
To obtain a quantitative measurement of the relative affinities of pro-MMP-9 and pro-MMP-2 for the ␣2(IV) chain, we examined the ability of ␣2(IV) to form a complex with either pro-MMP-2 or pro-MMP-9 by co-chromatography and the dissociation constants (K d ) were determined as described under "Experimental Procedures." Fig. 12 shows that both pro-MMP-9 and pro-MMP-2 bind ␣2(IV) as a function of proenzyme concentration. Under these conditions, maximal binding was observed at concentrations of enzymes of ϳ150 nM for pro-MMP-9 and Ͼ750 nM for pro-MMP-2. From these data, the K d values of ␣2(IV) for pro-MMP-9 and pro-MMP-2 were calculated to be approximately 45 nM and Ն350 nM, respectively.

DISCUSSION
Previously, we have shown the surface association of pro-MMP-9 in 12-O-tetradecanoylphorbol-13-acetate-treated MCF10A cells (25). Other studies have described the presence of pro-MMP-9 on plasma membranes of HT1080 cells (23) and in focal contacts of endothelial cells (22). Here, we have characterized the binding parameters of pro-MMP-9 and demonstrated that the proenzyme binds with high affinity (K d ϳ20 -30 nM) to the surface of a variety of cell types. In MCF10A cells, internalization studies demonstrated that binding of pro-MMP-9, free or in complex with TIMP-1, is not followed by ligand internalization. Furthermore, under the experimental conditions, surface binding of pro-MMP-9 does not result in zymogen activation, possibly due to the lack of activators in the cell culture system. However, recent studies demonstrated that surface-associated pro-MMP-9 can be activated by a plasmindependent mechanism (23,24). Taken together, these findings suggest that the surface association of pro-MMP-9 is not directly associated with activation and/or internalization but plays a role in confinement of proenzyme in areas of cell-matrix contacts, where it could become the target for potential pro-MMP-9-activating proteinases.
We identified a 190-kDa cell surface protein in various cell lines that specifically bound to a pro-MMP-9-affinity matrix  1 and 5), TIMP-1 (0.33 g/ml) (lanes 2 and 6), a pro-MMP-9⅐TIMP-1 complex (lanes 3, 4, 7, and 8) or with MMP-9 (0.5 g/ml) (lanes 9 and 10) in the absence (lanes 1-9) or presence (lane 10) of 5 mM EDTA. The samples were immunoprecipitated with mAbs to either pro-MMP-9 (CA-209) (lanes 1, 3, 5, 7, 9, and 10) or TIMP-1 (IM 322) (lanes 2, 4, 6, and 8). and formed a co-precipitable complex with pro-MMP-9. Microsequencing of the affinity-purified protein revealed a complete sequence homology to the ␣2(IV) chain of human collagen IV (39). Additionally, immunoblot analysis with chain-specific antibodies (30) corroborated that the 190-kDa protein is ␣2(IV). Several observations indicate that the cell surface association of pro-MMP-9 is mediated by ␣2(IV). (i) Preincubation of pro-MMP-9 with affinity-purified ␣2(IV) significantly inhibited ligand binding. (ii) Pretreatment of MCF10A cells with MMP-9 reduced by 80% the binding of proenzyme, likely due to the degradation of surface-associated ␣2(IV) as observed in experiments with soluble ␣2(IV) and MMP-9. Accordingly, treatment of cells with MMP-9 and TIMP-1 resulted in the recovery of the majority of ligand binding; (iii) Binding of pro-MMP-9 to the cells exhibited an affinity (K d ϳ22 nM) that was in close agreement with the affinity (K d ϳ45 nM) of the enzyme for purified ␣2(IV). In addition, there was a good correlation between the number of pro-MMP-9 binding sites determined in the binding assays and the level of expression of surface-associated ␣2(IV). For example, the breast cancer MDA-MB-231 cells that exhibited 10-fold less binding sites than MCF10A cells showed undetectable levels of surface-associated ␣2(IV). The reason for the low levels of ␣2(IV) on the surface of these malignant cells is unknown but may be related to their inability to retain ␣2(IV) on the cell surface since ␣2(IV) was detected intracellularly. Thus, in vitro, the surface association of pro-MMP-9 would depend on the ability of each cell type to express and retain ␣2(IV) on the cell surface. Whether pro-MMP-9 and pro-MMP-2 bind also to monomeric ␣1(IV) is yet unknown and remains to be determined. However, the fact that only a single polypeptide, ␣2(IV), was specifically and consistently co-precipitated with pro-MMP-9 or bound to immobilized proenzyme suggests a unique and preferential interaction between pro-MMP-9 and ␣2(IV).
Co-immunoprecipitation experiments using a pro-MMP-9⅐TIMP-1 complex or MMP-9 demonstrated that neither the C-terminal nor the N-terminal domains of pro-MMP-9 appeared to be critical for interactions with ␣2(IV). However, we could not precipitate the ␣2(IV)/pro-MMP-9 complex with gelatin-agarose beads, suggesting a role for the gelatin-binding domain. In this regard, it was interesting to observe that pro-MMP-2, when compared with pro-MMP-9, exhibited a lower affinity for ␣2(IV) as determined by the co-immunoprecipitation and gel filtration experiments. Since pro-MMP-2 contains a similar gelatin binding domain (41), this suggests that the interactions of ␣2(IV) with pro-MMP-9 are also influenced by sites/domains other than the gelatin binding domain and/or by the three-dimensional conformation of pro-MMP-9. It would be of interest to determine whether the 54-amino acid proline-rich ␣2(V)-like extension that is uniquely present in pro-MMP-9 (17,41), plays any role in the binding of the enzyme to ␣2(IV). In the cell binding assays, pro-MMP-2 slightly competed with pro-MMP-9 binding, suggesting that surface-associated ␣2(IV) is also available for pro-MMP-2 binding. However, given the lower affinity of pro-MMP-2 for ␣2(IV) and the existence of alternate high affinity surface binding sites for pro-MMP-2 (14), this enzyme would be expected to bind to the cell surface via a different mechanism (14,16).
The binding of gelatinases to collagen IV has been examined in previous studies (21,42); however, the sites of interaction were not defined. The results presented here suggest that a major high affinity binding site for pro-MMP-9 in collagen IV must reside within the ␣2(IV) chain. However, our data also indicate that neither triple-helical collagen IV secreted by MCF10A cells nor EHS collagen IV bound to pro-MMP-9 under conditions of ␣2(IV) binding. In agreement with these results, we have recently observed a weak affinity (K d ϭ 2.15 M) of pro-MMP-9 for EHS collagen IV as determined by surface plasmon resonance 4 . Thus, whereas monomeric ␣2(IV) forms a tight complex with pro-MMP-9 (K d nM), trimeric collagen IV binds pro-MMP-9 with very low affinity (K d M). This is also consistent with the studies of Steffensen et al. (21), who showed a weaker affinity of a recombinant gelatin-binding domain of pro-MMP-2 for trimeric collagen IV compared with that for denatured collagen IV. This has led to the suggestion that binding sites for the gelatin-binding domain in denatured collagen IV may be masked in triple-helical collagen IV (21). Therefore, it is conceivable that the binding of pro-MMP-9 to ␣2(IV) is mediated by sites that are cryptic in collagen IV and that are exposed after partial denaturation and/or degradation of the collagen IV molecule. If so, these sites would have to be conserved after partial proteolysis to allow binding of pro-MMP-9. Such a scenario would probably involve a partial degradation of the collagen IV network by a protease(s) other than the gelatinases as the catalytic efficiency of MMP-2 and MMP-9 against native collagen IV has been reported to be limited (43)(44)(45). After collagen IV degradation, secretion of pro-MMP-9 by pro-MMP-9 producing-cells in close association with the basement membrane would then facilitate binding of pro-MMP-9 to ␣2(IV). After activation of the ␣2(IV)-bound pro-MMP-9, the enzyme would then contribute to the complete degradation of the collagen IV network consistent with its ability to degrade denatured collagens (41).
Another possible scenario for the results observed here may involve binding of pro-MMP-9 to a surface-associated ␣2(IV). Although the origin and binding mechanism of ␣2(IV) chains to the cell surface remain to be elucidated, our in vitro studies with a variety of cell lines (epithelial, endothelial and fibrosarcoma) clearly show that, although some ␣2(IV) chains are assembled with ␣1(IV) into trimeric collagen IV molecules, a portion of single ␣2(IV) chains are secreted and deposited on the cell surface. The surface-associated ␣2(IV) chains appear to be stable as determined by surface biotinylation and pulsechase analysis. Indeed, we have found that ␣2(IV) chains can be detected in the cellular compartment of both MCF10A and HT1080 cells and in the supernatant of the latter even after a 4 M. Olson and R. Fridman, unpublished results. 10-h chase period without any evidence of processing and/or degradation. 5 These findings are in contrast with the general notion that ␣(IV) chains that fail to form trimeric collagen molecules are unstable and/or rapidly hydrolyzed. This is generally true if active proteinases are present in the experimental system, as they can hydrolyze non-triple-helical collagen chains as shown here with ␣2(IV) and MMP-9. Thus, the integrity and eventually the detection of ␣(IV) chains would depend on the presence of active proteinases and/or inhibitors in each particular culture system and tissue. Detection of monomeric ␣(IV) chains may also depend on the extraction procedures and/or on the antibodies used since some antibodies may only recognize ␣(IV) chains in triple-helical conformation. Here we have used immobilized pro-MMP-9 and chain-specific mAbs raised against synthetic peptides, which independently allowed for the detection of native and stable ␣2(IV) chains on the cell surface. Although the processing, localization and stability of monomeric ␣2(IV) warrants further in vitro and in vivo studies, it is tempting to speculate that surface-associated ␣2(IV) may anchor pro-MMP-9 on the cell surface after autocrine or paracrine secretion. After activation by pro-MMP-9-activating enzymes, MMP-9 would then play a role in localized surface proteolysis. Since the binding of pro-MMP-9 to the ␣2(IV) chain is of high affinity and involves the zymogen form, it will be important in future studies to address the activation of pro-MMP-9 and its interactions with TIMP-1 in the context of pro-MMP-9 bound to ␣2(IV).