Pro-collagenase-1 (matrix metalloproteinase-1) binds the alpha(2)beta(1) integrin upon release from keratinocytes migrating on type I collagen.

In injured skin, collagenase-1 (matrix metalloproteinase-1 (MMP-1)) is induced in migrating keratinocytes. This site-specific expression is regulated by binding of the alpha(2)beta(1) integrin with dermal type I collagen, and the catalytic activity of MMP-1 is required for keratinocyte migration. Because of this functional association among substrate/ligand, receptor, and proteinase, we assessed whether the integrin also directs the compartmentalization of MMP-1 to its matrix target. Indeed, pro-MMP-1 co-localized to sites of alpha(2)beta(1) contacts in migrating keratinocytes. Furthermore, pro-MMP-1 co-immunoprecipitated with alpha(2)beta(1) from keratinocytes, and alpha(2)beta(1) co-immunoprecipitated with pro-MMP-1. No other MMPs bound alpha(2)beta(1), and no other integrins interacted with MMP-1. Pro-MMP-1 also provided a substrate for alpha(2)beta(1)-dependent adhesion of platelets. Complex formation on keratinocytes was most efficient on native type I collagen and reduced or ablated on denatured or cleaved collagen. Competition studies suggested that the alpha(2) I domain interacts with the linker and hemopexin domains of pro-MMP-1, not with the pro-domain. These data indicate that the interaction of pro-MMP-1 with alpha(2)beta(1) confines this proteinase to points of cell contact with collagen and that the ternary complex of integrin, enzyme, and substrate function together to drive and regulate keratinocyte migration.

loproteinase, in human cutaneous wounds. In response to injury, collagenase-1 is induced in basal epidermal cells (keratinocytes) as the cells move off of the basement membrane and contact native type I collagen in the underlying dermis (1), and this inductive response is specifically controlled by the collagen-binding integrin ␣ 2 ␤ 1 (2). As we demonstrated in various experiments, the catalytic activity of collagenase-1 is required and sufficient for keratinocyte migration on complex matrices containing type I collagen. For example, keratinocytes plated on mutant, collagenase-resistant type I collagen do not migrate, even in the presence of fibronectin and vitronectin; yet they express MMP-1 at levels equivalent to those released by cells on wild-type collagen (2). Keratinocyte migration is also completely inhibited by anti-collagenase-1 antibodies, which block the catalytic activity of the enzyme, and by anti-␣ 2 ␤ 1 blocking antibodies (2). Thus, MMP-1, collagen, and ␣ 2 ␤ 1 function together in migrating keratinocyte during re-epithelialization of cutaneous wounds.
It is becoming increasingly clear that extracellular proteolysis is a cell-regulated process. After all, cells do not release proteases indiscriminately, especially enzymes like MMP-1 with such a defined substrate specificity. Rather they rely on precise interactions to accurately degrade, cleave, or process specific substrates in the pericellular space. Indeed, an emerging concept is that metalloproteinases, as for some serine and cysteine proteinases (3)(4)(5)(6), are anchored to the cell membrane, thereby targeting their catalytic activity to specific substrates within the pericellular space. In recent years, specific cell-MMP interactions have been reported, such as the binding of gelatinase-A (MMP-2) to cell membranes (7) and to the integrin ␣ v ␤ 3 (8), gelatinase-B (MMP-9) to CD44 (9), and matrilysin (MMP-7) to heparan sulfate proteoglycans (10). The membrane-type MMPs (MMP-14, -15, -16, -17, -24, and -25) are single-pass transmembrane proteins that are fixed and active at the cell surface and, in addition to acting as proteinases, may provide docking sites for other MMPs. Indeed, pro-MMP-2 also interacts with MMP-14 and TIMP-2 (tissue inhibitor of metalloproteinase 2) on the cell surface, and this trimeric complex is essential for activation of this gelatinase (11)(12)(13)(14). It is likely that other MMPs are also attached to cells via specific interaction to membrane proteins, and determining these anchors will lead to identifying activation mechanisms and relevant substrates.
Because the ␣ 2 ␤ 1 integrin regulates MMP-1 expression by binding the substrate of the enzyme, we assessed whether MMP-1 interacts with this collagen receptor on the surface of keratinocytes. We report here that pro-MMP-1 specifically binds the ␣ 2 ␤ 1 integrin on keratinocytes migrating on type I collagen. Formation of the ␣ 2 ␤ 1 -pro-MMP-1 complex may provide a mechanism to increase the localized concentrations of enzyme, in turn facilitating the cleavage of type I collagen and keratinocyte migration across the dermal wound bed.

EXPERIMENTAL PROCEDURES
Enzymes and Substrates-Human pro-MMP-1 was purified by twostep chromatography from conditioned medium of phorbol-treated U937 cells as described (15). Matrilysin was purchased from Chemicon International, Inc. (Temecula, CA). To isolate prodomains and activate metalloproteinases, autolysis of purified zymogens was induced with 1 mM 4-aminophenylmercuric acetate (Sigma) for 1 h at room temperature, and the two fragments were separated by size selection dialysis in 0.5 ml of 0.05 M Tris, 0.01 M CaCl 2 using a 10,000 Molecular Weight Cut-Off Slide-A-Lyzer (Pierce). The purity and integrity of these fragments was verified by SDS-polyacrylamide gel electrophoresis. Cleaved collagen was prepared by incubating 500 g of bovine type I monomeric collagen (Vitrogen; Collagen Corp., Palo Alto, CA) with 25 g of 4-aminophenylmercuric acetate-activated MMP-1 overnight at room temperature. MMP-1 was removed by dialysis in phosphate-buffered saline for 4 -6 h at 4°C. Gelatin was made by heating type I collagen at 80°C for 10 min.
Explants and Cell Culture-Normal human adult skin was obtained from patients undergoing reductive mammoplasty or lateral abdominoplasty. Four-millimeter, full thickness punch biopsies were attached upright to the surface of dry 6-well cluster dishes and were then submerged in Dulbecco's modified Eagle's medium containing 5% fetal bovine serum. After an overnight (16 -20 h) incubation at 37°C, skin biopsies were fixed in Histochoice (Amresco, Solon, OH) and processed for paraffin embedding. Keratinocytes were isolated from human skin and grown in Dulbecco's modified Eagle's medium (1.8 mM Ca 2ϩ ) as described (18). The cells were plated on dishes or on slides precoated with a 500 g/ml solution of bovine type I monomeric collagen (Vitrogen), 500 g/ml collagenase-cleaved collagen, or 500 g/ml gelatin.
In some experiments, keratinocytes were treated with 30 ng/ml epidermal growth factor (R & D Systems, Minneapolis, MN). Anti-␣ 2 blocking antibody (6F1 or MAB1950) was added to a final concentration of 10 g/ml. To prevent cell detachment, keratinocytes were plated 24 h before addition of integrin-blocking antibodies. The HaCaT human keratinocyte cell line (19) was provided by Dr. Norbert Fusenig (German Cancer Research Center) and were grown as described (2).
In Situ Hybridization and Immunohistochemistry-MMP-1 mRNA was detected in 5-m, deparaffinized sections of cultured skin biopsies by hybridization with 2.5 ϫ 10 4 cpm/l 35 S-labeled antisense RNA as described (1). Tissue sections were processed for immunohistochemistry using alkaline phosphatase as described (1). Endogenous peroxidase activity was blocked by incubation in 0.3% H 2 0 2 for 30 min at room temperature. Anti-human ␣ 2 antibody AB-1936 was diluted 1:1000. Bound antibody was detected using a Vectastain ABC Elite kit (Vector Laboratories, Burlingame, CA) following the manufacturer's instructions. Peroxidase activity was detected using 3,3Ј-diaminobenzidine tetrahydrochloride as chromogenic substrate. The sections were counterstained with Harris hematoxylin. For negative controls, sections were processed with preimmune serum.
Immunofluorescence-Primary human keratinocytes were plated on chamber slides precoated with a 500 g/ml type I collagen and incubated for 6 -24 h in Dulbecco's modified Eagle's medium/5% fetal calf serum. The chamber slides were washed three times with cold phosphate-buffered saline, and the cells were fixed in Histochoice for 1 min and dehydrated through graded ethanol. Fixed cells were incubated in with anti-human MMP-1 mouse monoclonal antibody (IM35L) and anti-␣ 2 rabbit polyclonal antibody (AB-1936) at the manufacturer's rec-ommended concentration. The primary antibody-antigen complexes were detected with fluorescein isothiocyanate-conjugated anti-rabbit IgG antibodies and tetramethylrhodamine isothiocyanate-conjugated anti-mouse antibody (Sigma) and viewed by laser confocal microscopy.
Immunoprecipitation-Equal numbers of primary keratinocytes (from 1 ϫ 10 5 to 2 ϫ 10 6 ) were allowed to adhere to collagen-or gelatin-coated dishes for 24 h. The conditioned medium was collected, and the cells were washed three times with ice-cold phosphate-buffered saline and then lysed with 1 ml of 1.5% Triton X-100, 10 mM Tris, pH 7.5, 150 mM NaCl, 1 mM MgCl 2 , 1 mM MnCl 2 , 2 mM phenylmethylsulfonyl fluoride, 20 g/ml aprotinin, and 12.5 g/ml leupeptin. The cell lysates and conditioned media were precleared by incubating for 2 h at 4°C in an equal volume of IPB-BSA-1 (0.8% Triton, 200 mM EDTA, 100 mg of BSA) containing protein A-Sepharose (PAS; Zymed Laboratories Inc., South San Francisco, CA). PAS was removed by centrifugation, and the preclearing procedure was repeated. Following the second centrifugation, the desired primary antibody was added to the supernatants, and the samples were incubated overnight at 4°C with gentle movement. Antibody-antigen complexes were precipitated with PAS, washed twice with IPB-BSA-2 (0.4% Triton, 10 mM EDTA, 100 mg/ml BSA), and eluted from PAS by heating at 60°C for 20 min in 60 l of nondenaturing electrophoresis sample buffer. For cells treated with an anti-␣ 2 blocking antibody (6F1 or MAB1950), complexes were formed with a rabbit polyclonal antibody (AB-806 for MMP-1 or AB-1936 for ␣ 2 ) and were precipitated from solution with Magna Bind Polyclonal Rabbit IgG (Pierce) and eluted with heat. The precipitating agents (PAS or Magna Bind) were pelleted, and the supernatants were removed.
Platelet Adhesion-96-well microtiter plates were coated with native type I collagen, MMP-1, or matrilysin (MMP-7) as described above. Nonspecific protein binding sites were blocked by incubating with 0.5% BSA in Tris-buffered saline, pH 7.4, at room temperature for 2 h. The platelets were prepared and washed as described (20), and 100 l of platelet suspension were permitted to adhere to the substrate for 1 h at room temperature. Nonadherent platelets were removed by several washings. Adhesion was quantified by measuring the hexose aminidase activity in the adherent platelets (21).
Surface Labeling-Proteins on the surface of keratinocytes were labeled using EZ-LinkTM Sulfo-NHS-LC-Biotinylation Kit (Pierce). Equal numbers (2 ϫ 10 6 ) of primary keratinocytes were plated, and 24 h later, cells were rinsed three times with serum-free medium and then incubated for 45 min in medium containing 1 mg/ml of Sulfo-NHS-LC-Biotin. The cells were then washed three times with medium to remove unbound biotin. Fresh medium was then added with various concentrations of MMP-1 or MMP-1 prodomain for 4 h. Cell lysates were collected, and labeled surface proteins were precipitated using streptavidin beads and resolved by electrophoresis. Surface-associated MMP-1 was detected by immunoblotting. Alternatively, MMP-1 was directly immunoprecipitated from cell lysates and detected on gels using horseradish peroxidase-linked streptavidin.

RESULTS
As is well established (22)(23)(24)(25), the ␣ 2 ␤ 1 integrin is localized circumferentially on basal keratinocytes in intact human skin (Fig. 1A). However, in response to injury, this integrin gathers along the basal surface of migrating keratinocytes at a wound edge (Fig. 1B). As we reported previously in studies of actual human wounds (1, 2), MMP-1 is selectively expressed by basal keratinocytes at the most forward edge of migration, and this same spatially confined pattern of expression was seen in ke-ratinocytes that have migrated from the epidermal edge of skin biopsies place in the culture (Fig. 1, C and D). The same population of cells that were positive for MMP-1 were the same as those that showed the most overt redistribution of the ␣ 2 ␤ 1 integrin. Because the interaction of ␣ 2 ␤ 1 with native type I collagen mediates induction of MMP-1 in migrating keratinocytes and because this metalloproteinase is required for cell migration on collagen (2), we reasoned that the integrin and the secreted enzyme would be found in the same pericellular compartment.
To assess this idea, we plated primary adult human keratinocytes on native type I collagen (hereupon called collagen). Reflecting the phenotype of basal cells involved in re-epithelialization in vivo, MMP-1 mRNA is expressed only in keratinocytes migrating from the colonies of proliferating and differen-tiating cells (2,26). In solitary cells examined 6 h post-plating, prominent immunofluorescence for the ␣ 2 subunit, which represents the ␣ 2 ␤ 1 heterodimer, was seen along the periphery of the cells, likely indicating the initial contacts of the spreading membrane with the collagen substratum (Fig. 1, E and F, red). The signal for MMP-1 protein appeared as clumps slightly internal to the membrane borders (Fig. 1, E and F, green and  merged yellow). Co-localization of the ␣ 2 ␤ 1 integrin and MMP-1 was seen in many close contact points just behind the edge of newly formed adhesions (Fig. 1, E and F, yellow). These observations suggest that MMP-1 is targeted to and released at sites of established cell-matrix contacts.
The overlap of immunofluorescence signals for the ␣ 2 ␤ 1 integrin and MMP-1 raised the possibility that these proteins may interact on the cell surface. Indeed, we found that MMP-1 selectively co-immunoprecipitated with ␣ 2 ␤ 1 from membrane preparations of keratinocytes plated on collagen (Fig. 2). In addition, essentially all of the co-immunoprecipitated enzyme migrated along with authentic pro-MMP-1 (Fig. 2, A-C), indicating that the zymogen form of the proteinase was associated with the integrin. In agreement with this conclusion, immuno-FIG. 1. Co-localization of ␣ 2 ␤ 1 and MMP-1 in migrating keratinocytes. A-D, biopsies (4 mm) of viable human skin were placed in culture overnight and were then fixed and processed for paraffin embedding. A, sections were stained for the ␣ 2 integrin subunit using alkaline phosphatase. In the center of the sample, where the skin is intact, signal for ␣ 2 ␤ 1 was seen on the basal-lateral surface of basal keratinocytes (arrows). B, at the sample edge, keratinocytes had migrated down the exposed side of the biopsies along the dermis. At the forward edge of the epidermal front, intense staining for ␣ 2 ␤ 1 was seen along the basal surfaces of keratinocytes (arrow). C and D, in situ hybridization revealed that MMP-1 mRNA is expressed by keratinocytes at the forward extend of and just behind the epidermal front (arrow), within the same population of cells showing marked redistribution of ␣ 2 ␤ 1 . Shown are paired bright field (C) and dark field (D) micrographs. E and F, primary human keratinocytes were isolated from normal adult skin, plated on slides coated with native type I collagen, and processed 6 h later for immunofluorescence using antibodies against MMP-1 and ␣ 2 ␤ 1 . A prominent signal for ␣ 2 ␤ 1 (red) was seen along the periphery of spreading cells and at the more central close contacts, where co-localization with MMP-1 was evident (merged signal, yellow). Cell-associated fluorescence for MMP-1 (green) was also seen apart from ␣ 2 ␤ 1 . The two confocal images were taken at 0.5 (E) and 2.0 (F) m above the substratum.
FIG. 2. Pro-MMP-1 is bound to ␣ 2 ␤ 1 on the surface of keratinocytes. Keratinocytes were plated on collagen-coated dishes, and 24 h later specific proteins were immunoprecipitated from cell lysates as described under "Experimental Procedures." In the panels shown, the antibody used for precipitation is indicated at the top of each lane. The antibodies used for immunoblotting are indicated along the side of each panel. In A, purified pro-MMP-1 was included as a migration standard (Std) and, the migration of molecular mass standards is shown on the right (in kDa). In E, proteins were immunoprecipitated from membrane extracts of HaCaT keratinocytes. As described under "Experimental Procedures," different antibodies were used if samples were immunoprecipitated and immunoblotted for the same antigen.
precipitation of TIMP-1, which is expressed by keratinocytes (27) and which does not bind pro-MMP-1, did not bring down any MMP-1 from the surface of keratinocytes (Fig. 2B). MMP-1 did not co-immunoprecipitate with the ␣ 3 ␤ 1 integrin, another collagen-binding receptor on keratinocytes (24,25), or with the ␣ 5 ␤ 1 or ␣ v ␤ 5 integrins ( Fig. 2A), which are expressed in migrating keratinocytes in vivo and in culture (24,25,28). Nearly identical levels of pro-MMP-1 were precipitated from keratinocyte membrane lysates incubated with either anti-␣ 2 or anti-MMP-1 antibodies (Fig. 2C), suggesting that essentially all cell surface-associated pro-MMP-1 is complexed with the ␣ 2 ␤ 1 integrin. In contrast, the signal for immunoblotted ␣ 2 was stronger in samples precipitated for ␣ 2 than in those precipitated for MMP-1, and a much more dramatic difference was seen in samples precipitated for the ␤ 1 subunit (Fig. 2D). These data indicate that only a subset of the ␣ 2 ␤ 1 receptors are occupied with pro-MMP-1, consistent with our confocal findings (Fig. 1,  E and F). As for primary keratinocytes, pro-MMP-1 produced by HaCaTs, a line of spontaneously transformed keratinocytes, immunoprecipitated with ␣ 2 ␤ 1 , and ␣ 2 ␤ 1 was brought down with pro-MMP-1 (Fig. 2E).
In agreement with these data, MMP-3 and MMP-9, which are expressed by keratinocytes in wounded skin (29,30), did not co-immunoprecipitate with ␣ 2 ␤ 1 from keratinocytes membrane extracts (Fig. 3, A and B), and precipitation of MMP-3 did not bring down ␣ 2 ␤ 1 (Fig. 3C). Plating keratinocytes on heat-denatured collagen (gelatin) or stimulating their production of MMPs with epidermal growth factor (31) did not promote an association of MMP-3 or MMP-9 with ␣ 2 ␤ 1 (Fig. 3, A  and B), nor did these conditions affect the levels of this integrin (Fig. 3C). We did find, however, that higher levels of MMP-9 were released from keratinocytes on native collagen compared with the levels secreted from cells on gelatin. The matrixmediated stimulation of MMP-9 expression in keratinocytes is a novel observation.
We also used a platelet adhesion assay to assess direct binding of pro-MMP-1 to native, membrane-integrated ␣ 2 ␤ 1 integrin (32). Platelets bound to dishes coated with collagen in the presence of Mn 2ϩ , and this interaction was inhibited with EDTA or 6F1 (Fig. 4), a blocking antibody that binds the I domain of ␣ 2 ␤ 1 (33). Platelets also bound to pro-MMP-1 (Fig. 4), but they did not bind to matrilysin (MMP-7), another MMP expressed by migrating epithelial cells (34). Thus, by three distinct assays: co-immunoprecipitation, platelet adhesion, and solid phase binding to ␣ 2 integrin I domain (35), we demonstrated that pro-MMP-1 and active MMP-1 binds to the ␣ 2 ␤ 1 integrin.
Increasing concentrations of 6F1 displaced pro-MMP-1 from the surface of keratinocytes (Fig. 5A), whereas a blocking antibody to ␣ 3 ␤ 1 did not (Fig. 5B). For these studies, 6F1 was added 24 h post-plating to prevent cell detachment, and 2 h later, cells and media were harvested. Because 6F1 recognizes an epitope within the I domain, these results suggests that pro-MMP-1 interacts with this region of the ␣ 2 subunit. This conclusion was verified in various experiments reported in our accompanying paper (35).
The ability of pro-MMP-1 to interact with ␣ 2 ␤ 1 was dependent on the nature of the collagen substratum. On gelatin, pro-MMP-1 did not co-immunoprecipitate with ␣ 2 ␤ 1 , even though appreciable levels of the enzyme were released into the medium (Fig. 5C). Gelatin did not influence the levels of ␣ 2 ␤ 1 on the cell surface (Fig. 5C). In some experiments, plating on gelatin barred an interaction between MMP-1 and ␣ 2 ␤ 1 (Fig.  5C), whereas in others, reduced levels of pro-MMP-1 were co-immunoprecipitated with ␣ 2 ␤ 1 (Fig. 5D). Because each experiment was done with keratinocytes from skin of a different individual, these results may reflect person-to-person variability. Regardless, our data show that pro-MMP-1 interacts poorly, if at all, with ␣ 2 ␤ 1 if keratinocytes are plated on denatured collagen. Similarly, reduced levels of pro-MMP-1 were co-immunoprecipitated with ␣ 2 ␤ 1 from keratinocytes on enzymatically cleaved collagen (Fig. 5D).
A competition assay was used to begin to determine the site on pro-MMP-1 that interacts with ␣ 2 ␤ 1 . Keratinocyte surface proteins were labeled with biotin, and cultures were then incubated for 4 h with equimolar amounts of purified active MMP-1 or its pro-domain. Biotinylated surface proteins were precipitated using streptavidin beads, and membrane-associated pro-MMP-1 was detected by immunoblotting. Cell-bound pro-MMP-1 was progressively displaced with increasing con-

FIG. 3. ␣ 2 ␤ 1 does not interact with MMP-3 or MMP-9.
A and B, keratinocytes were plated on gelatin-or collagen-coated dishes; some cells on collagen were treated with epidermal growth factor. Cells and media were collected 24 h later, and ␣ 2 ␤ 1 was immunoprecipitated (IP) from cell lysates. Medium samples and immunoprecipitates were resolved by electrophoresis and transferred to membranes for immunoblotting (IB) with antibodies against gelatinase-A (MMP-9) or MMP-3. C, in another experiment, lysates of keratinocytes on collagen, with or without epidermal growth factor, were immunoprecipitated for ␣ 2 ␤ 1 or MMP-3 and immunoblotted for ␣ 2 ␤ 1 . As described under "Experimental Procedures," different antibodies were used if samples were immunoprecipitated and immunoblotted for the same antigen. Washed platelets were permitted to adhere to the substrate for 1 h at room temperature. After washing, adhesion was quantified by measuring the hexose aminidase activity in the adherent platelets.
centrations of active MMP-1, with Ͼ50% removed with 3.6 M active MMP-1 (Fig. 6). Equal concentrations of the pro-domain had no effect on the recovery of cell-associated pro-MMP-1. We obtained similar results if pro-MMP-1 was immunoprecipitated from biotinylated cell lysates and detected on gels using horseradish peroxidase-linked streptavidin (data not shown). The efficiency of pro-MMP-1 displacement was possibly hampered by poor accessibility of the relatively large competitors to points of cell-substratum contacts. These findings suggest that pro-MMP-1 is anchored to the ␣ 2 ␤ 1 integrin via its catalytic, linker, or hemopexin domains, a conclusion that is more thoroughly refined in our accompanying paper (35). DISCUSSION Cell-matrix contacts provide an unambiguous signal informing cells which matrix protein they have encountered and, in turn, which proteinase is needed and where the enzyme should be delivered and released. During migration, cells attach to, extend over, and then release from a matrix substrate, and repeating these steps allows the cell to continue moving (36). Following injury to the epidermis, basal keratinocytes at the wound edge quickly move off of the basement membrane onto the underlying dermal matrix, which is rich in type I collagen.
In addition to being a mechanism for attachment and spreading, the interaction of keratinocytes with dermal collagen provides a site-specific signal that, along with other processes, such as altered cell-cell contacts, initiates the epithelial response to wounding. This response is characterized, in part, by the prominent and invariable expression of pro-MMP-1 (37) and is regulated by the ligation of the ␣ 2 ␤ 1 integrin, which is constitutively expressed on keratinocytes (38), with dermal collagen (2).
Here, we report that pro-MMP-1 specifically binds the ␣ 2 ␤ 1 integrin on enzyme-expressing keratinocytes plated on native type I collagen. Co-localization of this proteinase with this collagen-binding integrin was seen by confocal immunofluorescence, and direct binding of pro-MMP-1 to the ␣ 2 ␤ 1 integrin was demonstrated by co-immunoprecipitation and platelet adhesion. In our accompanying paper, we demonstrate by solid phase assays that pro-MMP-1 interacts with the I domain of the ␣ 2 ␤ 1 integrin (35). The ternary complex of the ␣ 2 ␤ 1 integrin, type I collagen, and pro-MMP-1 would spatially confine proteolysis and selectively direct catalysis to points of cellmatrix contacts. There are several examples of cell-directed proteolysis, some of which are cited in the Introduction. As discussed by Owen and Campbell (39) and as demonstrated by their studies on serine proteinases in neutrophils (4,40,41), anchoring enzymes to the cell surface provides both a mechanism to increase enzyme concentration at sites of proteolysis and a pericellular barrier to interference by natural proteinase inhibitors, which are abundant in tissue fluids. In addition to these functions, we propose that the ␣ 2 ␤ 1 -pro-MMP-1 complex functions also as a molecular motor controlling and driving keratinocyte migration over a dermal collagen.
Because the formation of ␣ 2 ␤ 1 -collagen contacts would precede the biosynthesis and secretion of pro-MMP-1, not all ␣ 2 ␤ 1contacts would be complexed with enzyme. Indeed, our coimmunoprecipitation data demonstrate that pro-MMP-1 binds to a subset of the ␣ 2 ␤ 1 integrin receptors (Fig. 2D), and our confocal observations indicate that pro-MMP-1 binds to established ␣ 2 ␤ 1 -collagen contacts (Fig. 1E). This spatial association suggests that MMP-1 functions to dissociate the integrin-collagen contacts, as suggested in our previous studies (2), rather than to break down matrix barriers that might be encountered at the forward extent of cell extension. The ␣ 2 ␤ 1 integrins bind native collagen with high affinity (42), and thus, clustering this FIG. 5. Pro-MMP-1/integrin association is reduced by anti-␣ 2 ␤ 1 antibodies and on keratinocytes grown on gelatin. A, keratinocytes were plated on type I collagen for 24 h and were treated for the last 2 h with increasing concentrations of 6F1, an anti-␣ 2 ␤ 1 blocking antibody to an epitope within the I domain. The cell were washed of medium, the lysates were immunoprecipitated (IP) for MMP-1, and the resolved products were blotted (IB) for MMP-1 protein. In this experiment, both pro-MMP-1 and active MMP-1 were immunoprecipitated, but the association of both forms with the cell surface was reduced by 6F1. B, in a similar experiment with different primary keratinocytes, 6F1 again reduced the recovery of pro-MMP-1 from cell lysates, whereas a blocking antibody against ␣ 3 ␤ 1 , another collagen-binding integrin, did not. C, keratinocytes were plated on heat-denatured collagen (gelatin) or on native fibrillar collagen with or without antibody 6F1 during the last 2 h. The cell layers were immunoprecipitated for ␣ 2 ␤ 1 and immunoblotted for or MMP-1 or ␣ 2 ␤ 1 . Medium samples were immunoprecipitated and immunoblotted for MMP-1. D, keratinocytes were plated on gelatin, enzymatically cleaved collagen (Clvd Col), or native fibrillar collagen. The cell lysates were harvested, lysed, and immunoprecipitated for ␣ 2 ␤ 1 and immunoblotted for MMP-1. Recovery of integrin-associated pro-MMP-1 was reduced from keratinocytes on the denatured substrates. Different antibodies were used if samples were immunoprecipitated and immunoblotted for the same antigen.
FIG. 6. Pro-MMP-1/␣ 2 ␤ 1 integrin interaction is disrupted by active MMP-1 and not by its pro-domain. Equal numbers of keratinocytes (2 ϫ 10 6 ) were plated on collagen, and 24 h later, surface proteins were biotinylated. The cells were washed, and fresh medium was added with he indicated concentrations of MMP-1 or purified prodomain for 4 h. The cell lysates were collected, and labeled proteins were precipitated using streptavidin beads and resolved by electrophoresis. Surface-associated MMP-1 was detected by immunoblotting. The migration of molecular mass standards is shown on the left (in kDa). receptor at contact points would tether keratinocytes to the dermis, rendering them unable to migrate. MMP-1 makes a single, site-specific cleavage through the triple helix about three-quarters of the length from the N terminus. The resultant TC A and TC B fragments are thermally unstable at body temperature and spontaneously unwind into gelatin (43), which binds ␣ 2 ␤ 1 with a much lower affinity than does native collagen (42). Thus, by simply making a single cut through the type I collagen helix, MMP-1 effectively loosens the tight contacts established by keratinocytes with the dermal matrix. The cells, in turn, would move forward by forming new contacts with uncleaved collagen on the open, superficial plane of the viable wound bed. We also detected the integrin-pro-MMP-1 complex in two lines of human colon carcinoma cells (WiDr and SW620; data not shown). Thus, this pericellular proteolytic mechanism may be operative during re-epithelialization of other tissues.
The importance of collagen being in its native, fibrillar state is highlighted by some key findings and concepts. As mentioned, the expression of pro-MMP-1 is induced by contact with native type I collagen; heat-denatured and proteolyzed collagen do not support or only weakly mediate synthesis of pro-MMP-1 (18,44). Thus, by cleaving type I collagen, MMP-1 creates a substrate that no longer supports its expression. Furthermore, pro-MMP-1 did not bind to the ␣ 2 ␤ 1 integrin on keratinocytes plated on gelatin or cleaved collagen (Fig. 5, C and D). These observations suggest that MMP-1 cleavage of type I collagen not only affects integrin-mediated signaling but also affects the ability of the integrin to bind the proteinase. Thus, in addition to affecting its expression, MMP-1 cleavage of collagen may also provide a mechanism to control its proteolytic activity. If disassociated from the cell surface, free pro-MMP-1 would assume its favored zymogen conformation or, if in an active conformation, would be readily neutralized by soluble inhibitors. The pro-MMP-1 detected in medium may reflect enzyme that was previously bound to cells.
Of interest, we found that pro-MMP-1, the zymogen form of this collagenase, preferentially, if not exclusively bound the ␣ 2 ␤ 1 integrin (Figs. 2, 5, and 6). Similarly, in these ( Fig. 5C) and previous studies (26), we have detected or isolated only pro-MMP-1 from keratinocyte-conditioned medium. In a few cultures, active MMP-1 co-immunoprecipitated with the ␣ 2 subunit, but the levels of the processed enzyme were always much less than the levels of the pro-form (Figs. 2B and 5A). The small amount of active MMP-1 we recovered from keratinocyte lysates may indicate that some degree of autoactivation occurred during the immunoprecipitation and washing steps. Although pro-MMP-1 bound to the ␣ 2 ␤ 1 integrin, this interaction is not dependent on the pro-domain. As suggested in this paper (Fig. 6) and as demonstrated thoroughly in our accompanying paper (35), the pro-domain of pro-MMP-1 does not bind to the ␣ 2 ␤ 1 integrin; this interaction is conferred by the linker and hemopexin motifs.
Although the zymogen form of MMP-1 bound the ␣ 2 ␤ 1 integrin and was the only form present in the medium, we demonstrated by several approaches that the catalytic activity of this collagenase is required for keratinocyte migration on collagen matrices (2). Although these data seemingly create a discrepancy, removal of the pro-domain is not the only way that an MMP can be activated. The critical step in activation of pro-MMPs is disruption of the interaction of the conserved cysteine in the prodomain with the zinc ion of the catalytic center, and this process does not necessarily require proteolysis (45). Thus, a conformational change induced by binding the ␣ 2 ␤ 1 integrin in the presence of its substrate, namely type I collagen, may shift pro-MMP-1 to an active state. Our ongoing studies are centered on assessing this hypothesis. Overall, our findings indicate that the ␣ 2 ␤ 1 integrin, type I collagen, and pro-MMP-1 form a trimeric complex that confines proteolysis and regulates gene expression and, in turn, keratinocyte movement during repair of cutaneous wounds.