The Involvement of the Fibronectin Type II-like Modules of Human Gelatinase A in Cell Surface Localization and Activation*

Recombinant collagen-binding domain (rCBD) comprising the three fibronectin type II-like modules of human gelatinase A was found to compete the zymogen form of this matrix metalloproteinase from the cell surface of normal human fibroblasts in culture. Upon concanavalin A treatment of cells, the induced cellular activation of gelatinase A was markedly elevated in the presence of the rCBD. Therefore, the mechanistic aspects of gelatinase A binding to cells by this domain were further studied using cell attachment assays. Fibroblasts attached to rCBD-coated microplate wells in a manner that was inhibited by soluble rCBD, blocking antibodies to the β1-integrin subunit but not the α2-integrin subunit, and bacterial collagenase treatment. Addition of soluble collagen rescued the attachment of collagenase-treated cells to the rCBD. As a probe on ligand blots of octyl-β-d-thioglucopyranoside-solubilized cell membrane extracts, the rCBD bound 140- and 160-kDa protein bands. Their identities were likely procollagen chains being both bacterial collagenase-sensitive and also converted upon pepsin digestion to 112- and 126-kDa bands that co-migrated with collagen α1(I) and α2(I) chains. A rCBD mutant protein (Lys263 → Ala) with reduced collagen affinity showed less cell attachment, whereas a heparin-binding deficient mutant (Lys357 → Ala), heparinase treatment, or heparin addition did not alter attachment. Thus, a cell-binding mechanism for gelatinase A is revealed that does not involve the hemopexin COOH domain. Instead, an attachment complex comprising gelatinase A-native type I collagen-β1-integrin forms as a result of interactions involving the collagen-binding domain of the enzyme. Moreover, this distinct pool of cell collagen-bound proenzyme appears recalcitrant to cellular activation.

The plasma membrane of various human cancer cells contains high levels of collagenolytic and gelatinolytic proteinases (1, 2) with a positive correlation shown between the expression of the matrix metalloproteinase (MMP) 1 gelatinase A and in-vasive potential (3). Moreover, certain tumor cell lines, which do not express gelatinase A, can bind the enzyme to their cell membranes by a membrane-associated receptor in trans (2,4). Activation of progelatinase A by cell membranes of concanavalin A (ConA)-stimulated (5,6) or 12-O-tetradecanoyl-phorbol-13-acetate-stimulated (7,8) normal cells requires a specific mode of enzyme-cell interaction that utilizes the COOH-terminal domains of gelatinase A and the tissue inhibitor of MMPs, TIMP-2 (8 -10). Four membrane type (MT)-MMPs possessing a hydrophobic transmembrane domain have been shown to activate progelatinase A at the cell surface (11,12) in an activation complex comprising progelatinase A, TIMP-2, and MT-MMP (12,13). Here, the active site of MT-MMP functions as a receptor for the inhibitory NH 2 domain of TIMP-2, leaving the TIMP-2 COOH domain free to interact with progelatinase A. Recent site-directed mutagenesis studies have mapped the TIMP-2-binding site on gelatinase A to the junction of the outer rim of ␤-blades III and IV of the hemopexin-like COOH-terminal domain (C domain) 2 . However, alternative interactions of the gelatinase A C domain with TIMP-4 (14) and cell surface components such as the ␣ v ␤ 3 integrin receptor (15), fibronectin (16), and heparin (16 -18) have also been identified.
The C domain of MMPs is involved in several important protein-protein interactions. In gelatinase B the C domain binds TIMP-1, whereas interstitial and neutrophil collagenases utilize the C domain for binding and cleavage of native type I collagen (19). However, the gelatinase A C domain does not bind collagen (16,20). Instead, a different collagen-binding domain (CBD) is found in gelatinases A and B consisting of three fibronectin type II-like modules inserted in the catalytic domain (21,22). In addition to binding denatured type I collagen (23)(24)(25), our characterization of recombinant human gelatinase A CBD (rCBD) showed that this domain accounts for all of the binding properties of the enzyme to native and denatured collagen types I, V, and X and elastin and also contains a heparin-binding site (17,25). 3 The importance of these functions is shown by CBD deletion, which reduces gelatinase A cleavage of denatured type I collagen by 90% (20) and abolishes elastin binding and cleavage (26).
The gelatinase A CBD may also serve to localize the enzyme to matrix components in tissues (17,20,25). These properties may similarly provide another mode of cell binding to membrane-associated matrix proteins, including collagen and hepa-ran sulfate proteoglycans, and thus may play a role in gelatinase A activation (18) and its physiological function on the cell surface. Here we report experiments that establish that the fibronectin-like CBD localizes gelatinase A to fibroblast cell surfaces by the formation of a gelatinase A-type I collagen-␤ 1integrin complex. Notably, this distinct pool of cell-bound enzyme shows a lowered cellular activation potential compared with soluble progelatinase A. This finding has important implications for the role of cell membrane-bound stromal gelatinase A on tumor cells.

EXPERIMENTAL PROCEDURES
Recombinant Gelatinase A Domains and Antibodies-rCBD (Val 191 -Gln 364 ) and the rC domain (Gly 417 -Cys 631 ) of human gelatinase A were expressed in Escherichia coli and purified by Zn 2ϩ -chelate and gelatin-Sepharose chromatography as appropriate (14,25). Electrospray mass spectrometry of the recombinant proteins was performed on a SCIEX API 300 (Perkin-Elmer) mass spectrometer. The convention used in this paper to distinguish between the recombinant protein comprised of the gelatinase A triple fibronectin type II-like repeat and the domain present in the natural enzyme will be to refer to the recombinant collagenbinding domain as the rCBD and to the domain in the enzyme as the CBD (no r).
Rabbit polyclonal antibody (␣CBD) was raised against rCBD injected with sarcosyl-extracted rCBD inclusion bodies and was then affinity purified over rCBD-AffiGel 10 (Bio-Rad) columns. Anti-peptide antibody (␣His 6 ) to the NH 2 -terminal His 6 fusion tag on the recombinant proteins was affinity purified as before (16).
Competition Experiments-Fibroblasts in 96-microwell tissue culture plates were treated with soluble rCBD (1.0 ϫ 10 Ϫ4 to 1.0 ϫ 10 Ϫ8 M) or rC domain (5.6 ϫ 10 Ϫ6 to 1.0 ϫ 10 Ϫ8 M) for 24 -28 h during and/or after ConA treatment (20 g/ml) (5) of quiescent cells in serum-free conditions. Conditioned medium and cell extracts were analyzed by zymography on 10% polyacrylamide/40 g/ml gelatin SDS-PAGE gels (27). To determine whether progelatinase A could bind unstimulated cells by the CBD, quiescent cells were thoroughly rinsed with PBS to remove unbound secreted enzyme. Gelatinase A was then competed from cell surfaces by incubation of the cell layers with 1.2 or 12 ϫ 10 Ϫ6 M rCBD in serum-free ␣-MEM at 22°C for 5 min only. This short time was selected to minimize contributions from newly synthesized enzyme to the medium during the incubation. After medium harvesting, the remaining cell-associated enzyme was assessed after lysis of the cell layer with SDS-PAGE sample buffer.
Cell Attachment Assay-Tissue culture surface treated 96-microwell plates were coated with 2-fold serially diluted rCBD (50 -0.25 g/ml) in 100 l PBS/well for 18 h at 4°C. After blocking with 10 mg/ml heatdenatured bovine serum albumin (BSA) for 30 min, 4 ϫ 10 4 fibroblasts were added per well in serum-free ␣-MEM (to avoid cell attachment from serum proteins) and incubated for 90 min at 37°C. Cells were then thoroughly rinsed with PBS and fixed with 4% formaldehyde in PBS. The attached cells were stained with 0.1% crystal violet in 200 mM boric acid, pH 6.0 (28). After extensive rinses, cellular stain was dissolved in 10% acetic acid, and cell numbers were quantitated by measurement of the optical density at 590 nm in a microplate reader. Positive control wells were coated with fibronectin (Chemicon) or acid soluble type I collagen prepared from rat tail collagen (25) or were nonblocked wells. Any cell attachment to BSA-blocked wells served to adjust for nonspecific attachment. Experiments were performed in duplicate or triplicate and repeated several times, but results were only compared for experiments on the same plate.
Cell Morphology and Spreading Characterization-For scanning electron microscopy, cells were seeded and grown in serum-free ␣-MEM on rCBD-coated glass coverslips (1 cm 2 ) blocked with BSA. After 1 or 2 h, cells were rinsed and fixed with 2.5% glutaraldehyde in PBS. Slides were stained with 1% osmium in PBS, treated with 2% tannic acid, dried by critical point drying, and sputter-coated with gold for analysis on a Stereoscan 260 (Cambridge Instruments) scanning electron micro-scope. Phase contrast microscopy was used to quantitate cell spreading at different time points after seeding 5 ϫ 10 3 cells on rCBD-or fibronectin-coated wells. Cells were fixed in 4% formaldehyde for 30 or 60 min at 22°C, and cell spreading, as judged by the appearance of lamellar cytoplasm, was then quantitated.
Mechanisms of Cell Attachment-Harvested cells were treated with 0.075-7.5 units/100 l highly pure bacterial collagenase (clostridiopeptidase A, Type III, fraction A (EC 3.4.24.3), Sigma) or 0.01 and 0.1 units/ml highly pure heparinase (Flavobacterium heparinum heparinase, Seikagaku Corporation) in ␣-MEM with 10 mM Ca 2ϩ acetate and 0.1% BSA for 15-30 min at 37°C. Enzymes were then removed by repeated cell sedimentation (120 ϫ g, 5 min) and washes in serum-free ␣-MEM prior to seeding in rCBD (25 g/ml)-coated wells. Attachment of bacterial collagenase-treated cells to native type I collagen bound to rCBD-coated wells was also quantitated. In addition, cells were seeded in the presence of blocking monoclonal antibody mAb13 (0.6 -20 g/ml) to the ␤ 1 -integrin subunit (kindly provided by Dr. K. Yamada, NIDR, National Institutes of Health) or ascites fluid antibody (P1E6, Life Technologies, Inc.) to the ␣ 2 -integrin subunit diluted 1:10 to 1:100. Affinity purified ␣CBD and ␣His 6 antibodies served as controls in the 90-min incubations. The effect of 1 or 10 g of heparin (Sigma) in 100 l of PBS added to rCBD-coated wells for 1 h prior to seeding was also assessed.
Ligand Blot Analyses-Confluent fibroblast cultures were rinsed thoroughly with PBS and then treated with 50 mM octyl-␤-D-thioglucopyranoside (Sigma) in PBS for 30 min at 15°C (29). After clarification at 10,000 ϫ g for 15 min at 22°C, detergent-solubilized cell membrane protein was precipitated at 0°C and then collected by centrifugation at 10,000 ϫ g for 10 min at 0°C. The protein pellet was dissolved in PBS, separated under nonreducing or reducing (65 mM DTT) conditions by 7.5% SDS-PAGE, and transferred to Immobilon-P polyvinylidene difluoride membrane (Millipore). The blots were BSA-blocked and then incubated with 20 g/ml rCBD in 150 mM NaCl, 10 mM Tris, pH 7.2, with 0.2% BSA for 1 h at 22°C. After washes, rCBD bound to the blotted proteins was detected using ␣CBD antibody and enhanced chemiluminescence reagents (Amersham Pharmacia Biotech). The rCBD-binding proteins were characterized by digestion with pepsin (0.1 mg/ml (Sigma) for 3 h at 15°C, pH 2.0) or highly pure bacterial collagenase (4 units/100 l for 18 h at 37°C, pH 7.0). An aliquot of the pepsin-treated sample was adjusted to pH 7.0 and incubated with bacterial collagenase for 18 h at 37°C. The efficiency and specificity of the enzyme digestions was verified using BSA, type I collagen and rCBD as control substrates.

RESULTS
Recombinant Protein Expression-The rCBD mass was measured by electrospray mass spectrometry to be 21,218 Da, confirming NH 2 -terminal methionine processing of the recombinant protein (predicted mass 21, 212 Da), fidelity of expression, and homogeneity of the protein preparation. The typical yield of purified rCBD from 3.6 liters of culture was 120 mg.
The Collagen-binding Domain Mediates Binding of Gelatinase A to Cells-When rCBD was incubated with human fibroblasts for 24 h during and after ConA treatment (Fig. 1A) or for 24 h after ConA treatment only (not shown), an increase in gelatinase A activation was apparent in six separate experiments. At high rCBD concentrations, essentially all the soluble gelatinase A was converted to the 59-kDa (ϪDTT) activated form (5). Although quantitation of enzyme levels from zymograms is only semiquantitative, less than ϳ3% of the total soluble gelatinase A remained as the 66-kDa (ϪDTT) zymogen form in the presence of 100 M rCBD (lane 100 ϩ) compared with ϳ28 -34% in those cells not treated with rCBD (lanes 0 ϩ). This trend was also apparent at 50 M rCBD. In contrast, recombinant gelatinase A C domain reduced cellular activation of the enzyme as before (17) (not shown).
Cell lysates containing gelatinase A that was bound to cells via the C domain of the enzyme or was intracellular in the cell secretory pathway were also prepared after rCBD treatment. Unlike the effect of rCBD on gelatinase A levels in the medium (Fig. 1A), addition of rCBD to cells during and/or after ConA treatment did not alter the ratios of latent (66 kDa) to active (59 kDa) gelatinase A in the lysates (Fig. 1B). As estimated from enzyme levels per microliter, the total enzyme recovered in the lysates of ConA-activated cells was ϳ10-fold less than that in the medium. In other experiments, zymography also demonstrated that cell-bound progelatinase A (the 66-kDa zymogen form) was competitively displaced from unstimulated cells that had not been ConA-treated. This was found even after a short 5-min pulse of the rCBD intended to minimize accumulations of newly secreted progelatinase A during the experiment (Fig. 1C). Extraction of the cell layer with SDS-PAGE sample buffer revealed that additional gelatinase A remained associated with the cells that was either not fully released by the short exposure to the rCBD or was bound by the C domain or was intracellular. That the increased gelatinase A activation upon ConA addition combined with rCBD treatment was not because of a direct cellular response to binding rCBD was shown in cultures incubated in the absence of ConA where rCBD addition for 24 h did not induce gelatinase A activation (not shown). Moreover, neither gelatinase A expression nor activation was altered in cells that were attached to rCBD-coated plates (see "The Collagen-binding Domain of Gelatinase A Mediates Cell Attachment") without ConA treatment (Fig. 1D).
Thus, these data show that in addition to interactions involving the C domain, progelatinase A can bind to cells via another domain of the enzyme, the CBD. Because only latent and not active gelatinase A was displaced in unstimulated cultures by the rCBD, these competition experiments also show that cell binding via the CBD of progelatinase A is not sufficient for enzyme activation. Indeed, because gelatinase A activation upon ConA treatment increases in the presence of excess rCBD, we conclude that cellular progelatinase A bound by the CBD has a lower cellular activation potential than the soluble enzyme in the medium. Hence, displacement of CBD-bound progelatinase A by the rCBD in ConA-treated cells may facilitate entry of the latent enzyme into the cellular activation pathway.
The Collagen-binding Domain of Gelatinase A Mediates Cell Attachment-The mechanistic aspects of gelatinase A cell binding via the CBD were further investigated by adaptation of cell attachment assays. Fibroblasts attached to rCBD-coated microwells in a concentration-dependent manner ( Fig. 2A), but this was less efficient than cell attachment to fibronectin (Fig.  2B). Incubation of fibroblasts with soluble rCBD prior to seed-inginhibitedattachmenttorCBD-coatedwellsinaconcentrationdependent manner, confirming binding specificity (Fig. 2C). Attachment was not observed in wells coated with 10 mg/ml BSA, whereas cell attachment to tissue culture-treated plastic alone or to type I collagen-coated wells was similar to that on fibronectin under saturating conditions. As assessed by phase contrast microscopy significantly fewer fibroblasts displayed cytoplasmic spreading on rCBD coated at 10 g/ml (23%) compared with fibronectin (50%) after 30 min. Greater differences in cell spreading were apparent between rCBD and fibronectin using 25 g/ml coated protein with 23 and 90%, respectively, of the cells spreading after 30 min. Although the kinetics of cell attachment and spreading differed at these early time points, spreading of cells on both substrates plateaued at 80 -90% of the attached cells by 60 min. Scanning electron microscopy confirmed both cell attachment to rCBD protein and these differences. After 1 and 2 h on fibronectin (Fig. 3, A and C, respectively), cells demonstrated typical cytoplasmic spreading (arrows) with a diameter of ϳ100 m. In contrast, cells on rCBD were smaller (diameter of ϳ50 m) and more rounded after 1 h (Fig. 3B) with limited spreading and extension of only delicate filopodia (arrowheads) after 2 h (Fig. 3D). Thus, this novel use of cell attachment assays confirmed the potential for gelatinase A binding to cells via the CBD of the enzyme.
␤ 1 -Integrins Are Involved in Cell Attachment to rCBD-A role for ␤ 1 -integrins in CBD-mediated gelatinase A cell binding was demonstrated using mAb13, an anti-␤ 1 -integrin blocking monoclonal antibody. At 2.5 g/ml antibody, more than 50% of the cell attachment to rCBD-coated wells was inhibited (Fig. 4). This inhibition increased to 90% at antibody concentrations Ͼ5 g/ml. In comparison, ␣ 2 -integrin blocking antibody and affinity purified ␣CBD and ␣His 6 control antibodies showed no significant blocking effects at these concentrations.
Ligand blotting was performed to identify cell proteins that may interact with the rCBD. On polyvinylidene difluoride blots of octyl-␤-D-thioglucopyranoside solubilized cell membrane proteins, rCBD bound two distinct protein bands having apparent masses of 140 and 160 kDa under reducing conditions (Fig. 5) in the approximate positions of ␣and ␤-integrin subunits or procollagen chains. However, both bands were degraded by bacterial collagenase. The 140-and 160-kDa bands were also partially pepsin-sensitive, being degraded to pepsin-resistant, but collagenase-sensitive, 112-and 126-kDa proteins. These co-migrated with collagen ␣1(I) and ␣2(I) chains that were also bound by the rCBD (Fig. 5). Thus, these data exclude the identity of the 140-and 160-kDa protein bands as integrin chains. Rather, the data provide strong evidence that the rCBD can interact with procollagen chains in cell membrane protein extracts. Nonetheless, other proteins, including those that do not renature on these blots or that require subunit interactions, might also be involved in the CBD interaction.
The Role of Pericellular Collagen in Cell Attachment to rCBD-In addition to any direct interaction with other cell membrane proteins, the ligand blots indicated that binding of gelatinase A CBD to native cellular collagen might represent one mode of gelatinase A cell binding. To test this, rCBD-coated wells were incubated with 10 g of soluble type I collagen in 100 l of PBS/well to saturate rCBD collagen-binding sites prior to cell seeding. On the rCBD-collagen complexes, cell attachment levels approached that for wells coated with 1.0 g/well collagen alone (Fig. 6). Cell attachment diminished with decreasing amounts of collagen bound to the rCBD.
In other experiments, cells pretreated with bacterial collagenase showed a concentration-dependent decrease in cell attachment to rCBD-coated wells (Fig. 7A). In control experiments, 0.75 unit of collagenase completely digested 100 g of purified native type I collagen after a 15-min incubation at 37°C and did not cleave rCBD or BSA (not shown). The digestions were not cytotoxic as evident from unaltered attachment of the treated cells to fibronectin or uncoated tissue culture wells (not shown). That the reduced cell attachment was because of pericellular collagen removal and not to integrin degradation was supported by antibody blocking experiments. Incubation of bacterial collagenase-treated cells (0.75 units/100 l) with ␤ 1 -integrin blocking antibody further reduced cell attachment to rCBD (not shown) or to the coated rCBD collagen complexes by ϳ75% (Fig. 7B). In positive control experiments, binding of collagenase-treated cells to wells coated with colla-

FIG. 4. Cell attachment to rCBD is inhibited by anti-␤ 1 -integrin antibodies.
To 96-well plates coated with 25 g/ml rCBD, fibroblasts were seeded in the presence of blocking monoclonal antibodies to the ␤ 1 -integrin subunit (mAb13) and to the ␣ 2 -integrin subunit (P1E6), and as controls, affinity purified ␣CBD and ␣His 6 antibodies or medium alone was used. The concentration of the P1E6 monoclonal antibody was estimated from the ascites fluid protein concentration. Cell attachment was quantitated after 90 min. Means of duplicate measurements are shown (n ϭ 3).
FIG. 5. Ligand blotting. Detergent-solubilized cell proteins were digested with pepsin or bacterial collagenase followed by SDS-PAGE on 7.5% gels and subsequent transfer to polyvinylidene difluoride membranes. Blots were probed with 20 g/ml rCBD for 1 h at 22°C. Control membranes were not probed (Control). rCBD bound to the blotted proteins was detected with affinity-purified ␣CBD antibody and enhanced chemiluminescence reagents. St, type I collagen standard; CB, Coomassie Blue-stained detergent solubilized protein. The amount (in kDa) of marker proteins and the collagen ␣-chain, pro ␣-chain, and ␤-components are shown. gen alone was also blocked with the ␤ 1 -integrin blocking antibody (Fig. 7B). However, it should be noted that cell attachment had already been greatly diminished by treatment with bacterial collagenase, and so this result does not necessarily show a direct interaction between ␤ 1 -integrins and the rCBD, but it is a possibility that cannot be totally excluded. Alternatively, the further minor contribution to the rCBD cell interaction by ␤ 1 -integrins may have been through molecules not degraded by bacterial collagenase that were bound to this integrin class and the rCBD. As further evidence for the role of the interaction of the rCBD with ␤ 1 -integrin-bound collagen, cell attachment to bacterial collagenase-treated cells could be rescued in a concentration-dependent manner by addition of soluble native type I collagen to the rCBD-coated wells before cell seeding (Fig. 7A). Rescue was not complete, due either to release of endocytosed collagenase (30) or to nonrescued cell interactions involving other collagen types.
Cell Attachment to rCBD Is Not Heparan Sulfate-dependent-To determine the potential for rCBD to interact with cell membrane heparan-sulfate proteoglycans through the low affinity heparin-binding site on the rCBD 3 (25), Lys 357 3 Ala, a mutant of rCBD that shows complete loss of heparin binding, 3 did not display any differences in mediating cell attachment compared with the wild type rCBD (Fig. 8A). In contrast, a rCBD mutant (Lys 263 3 Ala), characterized as having reduced type I collagen binding affinity, 3 showed reduced cell attachment properties compared with the wild type rCBD (Fig. 8A).
In other experiments, cells were treated with heparinase prior to plating, but even at concentrations as high as 0.1 units/ml there was no alteration in cell attachment to the rCBD (Fig.  8B). Lastly, heparin was added to rCBD-coated wells prior to cell seeding to block heparin-binding sites, but this too did not reduce attachment levels from controls (not shown). Collectively, these results show that pericellular collagen was a ratelimiting component of cell attachment to rCBD-coated wells and that heparan-sulfate proteoglycans were not involved. Therefore, this reveals the potential for gelatinase A binding to cells via interactions involving the CBD of the enzyme and native cellular collagen that is cell associated by ␤ 1 -integrins. DISCUSSION By studying fibroblast cell attachment to the recombinant CBD of human gelatinase A we have developed a novel approach to mechanistically explore cell-binding mechanisms of gelatinase A. Our data indicate that the interaction between rCBD and cells in the attachment assays is representative of gelatinase A utilizing this domain to bind cell surfaces. Notably, the potential for gelatinase A interaction with cells via the CBD of the enzyme was shown by the competitive release of progelatinase A from unstimulated fibroblasts by the rCBD. The importance of the gelatinase A C domain-TIMP-2 C domain interaction for activation by MT-MMPs is also thereby demonstrated because cell surface localization of progelatinase A through the CBD was not sufficient for activation. This confirms previous reports using CBD and C domain deletion mutants of the enzyme (20). Although the rCBD binds heparin (25), we found no evidence of rCBD binding to cell membrane heparan sulfate proteoglycans. Rather, our studies overall in- FIG. 6. Cell attachment to rCBD-type I collagen complex. 96well plates were coated with either rCBD (1.5 or 25 g/ml) or native type I collagen (1, 3, or 10 g/100 l PBS coated in each well). After blocking with BSA, rCBD-coated wells were then incubated with soluble native type I collagen (1, 3, or 10 g/100 l PBS/well). Tissue culture treated plastic (P) or wells blocked with BSA (B) served as positive and negative controls (Cont), respectively. Human fibroblasts were seeded, and cell attachment was analyzed after 90 min. Data points are the means of duplicate wells (n ϭ 3).

FIG. 7.
Effect of bacterial collagenase treatment on cell attachment to rCBD. Panel A, fibroblasts were treated with highly pure bacterial collagenase (0.075, 0.75, or 7.5 units/100 l) in serum-free ␣-MEM for 15 min at 37°C. Collagenase-treated cells (4 ϫ 10 4 ) were then seeded in wells coated either with 25 g/ml rCBD alone or subsequently bound with native type I collagen (1, 3, or 10 g/100 l PBS in each well). Panel B, collagenase (0.75 units/100 l)-treated fibroblasts were plated for 90 min in the presence of anti-␤ 1 -integrin antibody/ serum-free ␣-MEM in wells coated with either 25 g/ml rCBD subsequently complexed with native type I collagen (10 g/100 l PBS in each well) or native type I collagen alone (0.5 g/well). Means (n ϭ 3) and S.D. bars are shown. dicate that the gelatinase A CBD mediates cell surface localization of the enzyme to fibroblasts by binding pericellular collagen that in turn is anchored to cell membrane ␤ 1 -integrins, among which ␣ 1 ␤ 1 , ␣ 2 ␤ 1 , and ␣ 3 ␤ 1 are type I collagen receptors (31).
Direct binding of rCBD to ␤ 1 -integrins as an integrin-associated protein or via the integrin ligand-binding site is a possibility not completely ruled out by our studies. However, no RGD sequence occurs in the gelatinase A CBD. Interestingly, the gelatinase B CBD contains a RGD sequence that aligns with RSDG at positions 339 -342 in gelatinase A. In support of a ␤ 1 -integrin-collagen-gelatinase A CBD bridge model, detergent solubilized fibroblast cell surface proteins that were bound by rCBD on ligand blots were fully degraded by bacterial collagenase. Furthermore, these proteins were also pepsin-resistant, a characteristic of the native triple helical collagen domain. This points to native collagen or procollagen mediating the cell binding interactions with rCBD. Cell attachment to the rCBD-collagen complexes showed that rCBD occupancy of the major type I collagen-binding site on the collagen telopeptides (25) did not sterically block ␤ 1 -integrin attachment to the collagen. This indicates that the integrin receptors and rCBD recognize different binding sites on collagen.
Other evidence supporting this mode of cell binding was that bacterial collagenase treatment of fibroblasts greatly reduced cell attachment to rCBD, and this was rescued in a concentrationdependent manner by adding native type I collagen to the coated rCBD films. That collagen was a rate-limiting component in attachment of untreated cells was also shown by binding type I collagen to coated rCBD-coated wells prior to cell seeding. This produced a concentration-dependent increase in cell attachment ultimately approaching that of cells attached to collagen alone.
Our demonstration of cell binding by the gelatinase A CBD reveals a distinct mode of gelatinase A cell localization additional to that involving the hemopexin-like C domain (5-7, 10, 13). We have previously proposed that gelatinase A cellular activation involves TIMP-2 bridging the C domain of progelatinase A and MT-MMP with activation occurring by a second MT-MMP (17). Indeed, activation does not occur with C domain deletion mutants of gelatinase A (10) or in the absence of TIMP-2 (13) and can be inhibited by adding rC domain to ConA-treated cells (17) or membranes (13). However, the mechanism of release of active gelatinase A is enigmatic (17) given the K d values of TIMP-2 binding the C domain of gelatinase A (14) and that of the inhibitory N domain of TIMP-2 binding the active site of MT-MMP. Possibly, release of active gelatinase A occurs upon MT-MMP degradation from the active 60-kDa form to a truncated membrane-bound 42-kDa form of MT-MMP that was recently reported (32). Despite convincing evidence for the trimolecular complex of gelatinase A-TIMP-2-MT-MMP, alternative cell-binding mechanisms for gelatinase A are indicated from other studies. Because TIMP-2 as well as the progelatinase A-TIMP-2 complex can bind to cell surfaces, a specific TIMP-2 receptor, possibly distinct from active MT-MMPs, may also mediate enzyme binding (33). Indeed, signal transduction events and growth factor effects can be elicited upon TIMP-2 cell binding (33), and gelatinase A can bind to cells not expressing MT1-MMP (34). Binding of the gelatinase A C domain, which has homology to vitronectin, to the ␣ v ␤ 3 integrin vitronectin receptor can also occur (15). Vitronectin binding to ␣ v ␤ 3 integrins also enhances gelatinase A expression and cell FIG. 9. Model of the influence of cell surface collagen on progelatinase activation at the cell membrane after ConA stimulation. Panel A, in unstimulated cells, secreted progelatinase A accumulates extracellularly (1) or binds (2) to cellular collagen (3), which is bound to the cell membrane (4) via ␤ 1 -integrins (5). The TIMP-2 (6) C domain (shaded square) interacts with the progelatinase A C domain to form progelatinase A-TIMP-2 complexes (7). Synthesis and secretion of progelatinase A is indicated by the arrows originating from within the cell. Collagen binding by the progelatinase A is mediated through the CBD of the enzyme in an equilibrium with the much larger pool of soluble progelatinase A. Panel B, ConA treatment of cells induces the expression of active MT-MMP (8), which is anchored to the cell membrane by a transmembrane domain and cytoplasmic extension (9). The TIMP-2-progelatinase A complex (7) binds to the active site of the MT-MMP via the NH 2 -terminal "inhibitory" domain of TIMP-2 (shaded half-circle). A second active MT-MMP molecule (10) then cleaves (asterisk) the prodomain of the gelatinase A to generate the 62-kDa (ϪDTT) activation intermediate. Full gelatinase A activation then proceeds autocatalytically by another active gelatinase A molecule (not shown) to generate active gelatinase A that is bound to the MT-MMP on the cell surface (7). The 59-kDa (ϪDTT) active gelatinase A (11) is then released from the MT-MMP, possibly by MT-MMP degradation as discussed in the text. Active proteinases are indicated by a line in the open circle representing the active cleft in the catalytic domain after removal of the prodomain (shaded circle). Ongoing synthesis and secretion of progelatinase A replenishes the pools of progelatinase A extracellularly and that bound to integrin-linked collagen on the cell surface. Panel C, in unstimulated cells, the addition of recombinant CBD protein (13) competes for cell surface collagen binding with the natural CBD of the collagen-bound progelatinase A (2). As experimentally shown in Fig.  1C, a short incubation time with rCBD displaces progelatinase A (14) only from cell surfaces. Panel D, addition of recombinant CBD (13) with ConA competes off collagen-bound progelatinase A (2) in proximity to the MT-MMP-TIMP-2 activation complexes on the cell membrane (8,10). This increases the amount of progelatinase A activation and release of active enzyme (11). The presence of recombinant CBD also competes for any binding of newly activated gelatinase A to the cell-bound colla-gen. Together with the ongoing activation of soluble progelatinase A (1), this results in the accumulation of active gelatinase A extracellularly relative to the zymogen form of the enzyme over time, as shown experimentally in the medium in Fig. 1A. penetration of basement membranes (35), emphasizing the important role of integrins in gelatinase A function. Thus, gelatinase A may localize to cell surfaces by a number of distinct mechanisms including the CBD, the C domain via the TIMP-2-MT-MMP complex, a distinct TIMP-2 receptor, and the C domain via the ␣ v ␤ 3 integrin receptor.
As shown in Fig. 1, when rCBD was added to ConA-treated cells this produced an elevated activation of the progelatinase A in the medium, but not in the cell layer, over that seen by ConA alone as first described by Overall and Sodek (5). Because the total amount of enzyme recovered in the cell lysates, which also includes proenzyme in the secretory pathway, was approximately 10-fold less than that found in the medium, the total cellular response to the rCBD was one characterized by a marked elevation in ConA-induced gelatinase A activation. The explanation we favor for this new finding is presented in Fig. 9. The displacement of progelatinase A by the rCBD in cells treated with ConA would promote enzyme activation before release to the medium because of the proximity of the released enzyme with the cell membrane and MT-MMPs. This is likely to be the mechanism because the active enzyme accumulated in the medium rather than in the cell layer, which retained relatively unaltered levels of latent and active gelatinase A. The progelatinase A on the cell-bound collagen would be replenished from newly synthesized enzyme bound at the time of secretion. Thus, the relative proportions of latent to active gelatinase A in the collagen-bound pool would not necessarily alter significantly upon rCBD addition. Together with the ongoing activation of soluble progelatinase A, the presence of rCBD would also compete for binding of newly activated soluble active gelatinase A to cell-bound collagen. This would result in the accumulation of active gelatinase A in the medium relative to the zymogen form of the enzyme over time. Thus, these data strongly indicate that the pool of gelatinase A that is cell-bound via ␤ 1 -linked collagen resists entry into the MT-MMP activation pathway.
We propose that CBD-mediated cell binding of progelatinase A may provide a means of maintaining a pool of latent enzyme at the cell membrane. Cell binding by the CBD also has the potential to target progelatinase A from one cell to another in trans, a mechanism thought to be important for increasing the proteolytic potential of tumor cells. However, our data indicate that in these cells MT-MMPs would not necessarily readily activate enzyme so targeted unless subsequently released from the collagen. Of note, MT1-MMP can cleave native type I collagen (36,37). Therefore, on MT-MMP induction, release of CBD-bound progelatinase A from the MT-MMP degraded cellbound collagen may provide the means for entry of this pool of progelatinase A into the C domain-TIMP-2-MT-MMP activation pathway. Activated gelatinase A would thereby be localized at sites of ongoing cell matrix degradation and gelatinolysis.