Collagen Dissolution by Keratinocytes Requires Cell Surface Plasminogen Activation and Matrix Metalloproteinase Activity*

Matrix metalloproteinase-14 is required for degradation of fibrillar collagen by mesenchymal cells. Here we show that keratinocytes use an alternative plasminogen and matrix metalloproteinase-13-dependent pathway for dissolution of collagen fibrils. Primary keratinocytes displayed an absolute requirement for serum to dissolve collagen. Dissolution of collagen was abolished in plasminogen-depleted serum and could be restored by the exogenous addition of plasminogen. Both plasminogen activator inhibitor-1 and tissue inhibitor of metalloproteinase blocked collagen dissolution, demonstrating the requirement of both plasminogen activation and matrix metalloproteinase activity for degradation. Cell surface plasmin activity was critical for the degradation process as aprotinin, but not 2-antiplasmin, prevented collagen dissolution. Keratinocytes with single deficiencies in either urokinase or tissue plasminogen activator retained the ability to dissolve collagen. However, collagen fibril dissolution was abolished in keratinocytes with a combined deficiency in both urokinase and tissue plasminogen activator. Combined, but not single, urokinase and tissue plasminogen activator deficiency also completely blocked the activation of the fibrillar collagenase, matrix metalloproteinase-13, by keratinocytes. The activation of matrix metalloproteinase-13 in normal keratinocytes was prevented by plasminogen activator inhibitor-1 and aprotinin but not by tissue inhibitor of metalloproteinase-1 and -2, suggesting that plasmin activates matrix metalloproteinase-13 directly. We propose that plasminogen activation facilitates keratinocyte-mediated collagen breakdown via the direct activation of matrix metalloproteinase-13 and possibly other fibrillar collagenases.

Matrix metalloproteinase-14 is required for degradation of fibrillar collagen by mesenchymal cells. Here we show that keratinocytes use an alternative plasminogen and matrix metalloproteinase-13-dependent pathway for dissolution of collagen fibrils. Primary keratinocytes displayed an absolute requirement for serum to dissolve collagen. Dissolution of collagen was abolished in plasminogen-depleted serum and could be restored by the exogenous addition of plasminogen. Both plasminogen activator inhibitor-1 and tissue inhibitor of metalloproteinase blocked collagen dissolution, demonstrating the requirement of both plasminogen activation and matrix metalloproteinase activity for degradation. Cell surface plasmin activity was critical for the degradation process as aprotinin, but not ␣ 2 -antiplasmin, prevented collagen dissolution. Keratinocytes with single deficiencies in either urokinase or tissue plasminogen activator retained the ability to dissolve collagen. However, collagen fibril dissolution was abolished in keratinocytes with a combined deficiency in both urokinase and tissue plasminogen activator. Combined, but not single, urokinase and tissue plasminogen activator deficiency also completely blocked the activation of the fibrillar collagenase, matrix metalloproteinase-13, by keratinocytes. The activation of matrix metalloproteinase-13 in normal keratinocytes was prevented by plasminogen activator inhibitor-1 and aprotinin but not by tissue inhibitor of metalloproteinase-1 and -2, suggesting that plasmin activates matrix metalloproteinase-13 directly. We propose that plasminogen activation facilitates keratinocyte-mediated collagen breakdown via the direct activation of matrix metalloproteinase-13 and possibly other fibrillar collagenases.
The matrix metalloproteinases (MMPs) 1 constitute a large family of structurally related matrix degrading proteases that have pivotal roles in development, tissue remodeling, and can-cer (1)(2)(3)(4)(5). The MMPs share a number of common structural and functional features. All MMPs have essential zinc and calcium ions, are synthesized as zymogens, and are inhibited by tissue inhibitors of metalloproteinases (TIMPs) in a 1:1 enzyme-inhibitor complex. MMPs have multiple domains that control their secretion, specificity, and substrate binding. The function of the MMPs is also tightly regulated at the level of gene expression, zymogen activation, enzyme activity, and cell surface localization (1)(2)(3)5). An intramolecular complex between an unpaired Cys residue in the propeptide domain and the active site zinc inherently blocks the catalytic activity of all MMPs, and this propeptide is removed by endoproteolytic cleavage during MMP activation (6). As is the case with most zymogen cascades, many aspects of the mechanisms leading to the activation of pro-MMPs under physiological conditions are still incompletely understood. Membrane-type (MT) MMPs and MMP-11 (stromelysin-3) are equipped with a furin recognition site and are activated by furin or furin-like enzymes in the secretory pathway or on the cell surface (7). Most MMPs, however, do not contain a furin recognition site, and their activation appears to be dependent on propeptide cleavage by MT-MMPs, MMP-11, plasmin (see below), and possibly other serine proteases (8 -10).
The plasminogen activation (PA) system represents a second extracellular proteolytic system with pleiotropic functions in physiological and pathological tissue remodeling (11)(12)(13). The PA system is a complex system of serine proteases, protease inhibitors, and protease receptors that governs the conversion of the abundant protease zymogen, plasminogen (Plg), to the active serine protease, plasmin. Plg is predominantly produced by the liver and is present in high concentrations in plasma as well as most extravascular fluids (14). Plasmin is formed by the proteolytic cleavage of Plg by either of two Plg activators, the urokinase Plg activator (uPA) and the tissue Plg activator (tPA). Activation of Plg appears to be strictly associated with the cell surface via the binding to specific receptors, as well as other surfaces that present kinetically favorable circumstances for Plg activation, such as the fibrin thrombus. Surface-generated plasmin, unlike plasmin in solution, is relatively protected from its primary physiological inhibitor ␣ 2 -antiplasmin (␣ 2 -AP) (15)(16)(17). Cell surface Plg activation by uPA and tPA is regulated by two physiological inhibitors, Plg activator inhibitor-1 and -2 (PAI-1 and PAI-2) (18 -20), each forming a 1:1 complex with uPA and tPA.
The functional association between the MMPs and the PA system, in particular the role of plasmin as a pro-MMP activator, has generated substantial attention in the context of both physiological and pathological tissue remodeling. The MMPs and components of the PA system are co-expressed with remarkable consistency during development, tissue remodeling, tissue repair, and in multiple diseases such as tumor invasion, metastasis, arthritis, vessel wall disease, and neurodegenerative diseases (2, 4 -5, 11, 21-28). Moreover, numerous studies have demonstrated that plasmin can contribute to the proteolytic activation of pro-MMP-1, -2, -3, -9, -10, -12, and -13 in vitro (29). A less clear picture of plasmin as a pro-MMP activator has emerged from cell-based assays. Plasmin has been reported to be both a poor activator and a very efficient activator of MMP-9 (23, 30 -32). Macrophage and endothelial cell-produced uPA contributes to MMP-1 and -13 activation via the activation of Plg (21,32). Plasmin can also activate MMP-13 in fibroblast cultures.
The interstitial, fibrillar type I-III collagens are the most abundant proteins in connective tissues. These collagen monomers consist of three polypeptide chains, which contain a single long uninterrupted section of Gly-X-Y repeats. The three helical chains are intertwined to produce a tight superhelix that buries the peptide bonds within the interior of the helix. The fibrillar collagens spontaneously self-associate to form fibrils that range in diameter from 10 to 300 nm (33,34). This unique supramolecular organization makes fibrillar collagens resistant to degradation by most proteases. However, once cleaved within their triple helical domain, collagens quickly denature and are easily hydrolyzed by a variety of proteases. Several molecular pathways exist for the dissolution of collagen fibrils. One pathway is intracellular and involves the phagocytosis of collagen fibrils via integrin receptors, followed by the degradation of the internalized collagen in the lysosome (35). A second, cathepsin-mediated, pathway for collagen degradation takes place in the acidic microenvironment that underlies osteoclasts during bone resorption (36 -37). However, a subset of the MMPs, the collagenases, are attributed with being primarily responsible for the initial cleavage of interstitial collagen fibrils under normal and pathological conditions (1). They do so in a highly specific manner, cleaving all three chains of the collagen monomer at a specific locus located approximately three-quarters from the N terminus (38 -40). While it was previously believed that only the "genuine" collagenases, MMP-1, -8, and -13, were capable of hydrolyzing fibrillar collagens, other members of the MMP family have now been reported to display "triple-helicase" activity, including MMP-2 and the membrane type MMPs (MT-MMPs) (41)(42)(43)(44).
Previous studies (43,(45)(46)(47) have investigated the dissolution of collagen fibril matrices by mesenchymal cells. Collectively, these studies have demonstrated a pivotal role of MT1-MMP (MMP-14) in mesenchymal cell-mediated collagen turnover. In this paper, we report the identification of an alternative, keratinocyte-specific, pathway for collagenase activation and fibrillar collagen dissolution that requires Plg activation and MMP-13.
Cell Culture-Primary human skin keratinocytes were isolated from human neonatal foreskin (Cooperative Human Tissue Network) as de-scribed previously (51). Briefly, the tissue was cut into 4 -5-mm pieces and floated dermis side down on 0.25% trypsin in Dulbecco's PBS solution overnight at 4°C. The epidermal layer was peeled away from the dermis and further dissociated by pipetting up and down 10 times in Dulbecco's modified Eagle's medium (Invitrogen) containing 20% heat-inactivated fetal bovine serum (FBS), supplemented with 10 ng/ml epidermal growth factor, 1 nM cholera toxin, 0.4 g/ml hydrocortisone, and penicillin/streptomycin (Invitrogen). The cell suspension was seeded into tissue culture flasks. The medium was replaced the following day with Defined Keratinocyte-SFM (Invitrogen).
Primary murine skin keratinocytes were isolated from the epidermis of 1-3-day-old mice essentially as described previously (52). Briefly, the skin of newborn mice was floated overnight at 4°C with the dermis facing down in 0.25% trypsin in Dulbecco's PBS solution (Invitrogen). The epidermal layer was then peeled away from the dermis with sterile forceps. To release the partially dissociated keratinocytes, the epidermis was minced with forceps and pipetted up and down 10 times in minimum Eagle's medium without calcium chloride (BioWhittaker, Inc., Walkersville, MD) supplemented with 8% Chelex-treated heatinactivated FBS (HyClone Laboratories Inc., Logan, UT), 1.3 mM CaCl 2 , and penicillin/streptomycin (Invitrogen). The cell suspension was then filtered through a 70-m cell strainer to remove the stratum corneum, and the filtered suspension was seeded into tissue culture plates for 4 h at 37°C in a humidified 7% CO 2 -buffered tissue culture incubator. The medium was then replaced with minimum Eagle's medium with 8% Chelex-treated, heat-inactivated FBS, containing 0.05 mM CaCl 2 , and penicillin/streptomycin. Primary murine skin fibroblasts were isolated from the dermis as described previously (43). Briefly, the skin of newborn mice was floated dermis side down on 0.25% trypsin in Dulbecco's PBS solution (Invitrogen) overnight at 4°C. The dermis was dissociated by forcing it through a 3-ml syringe. The cell suspension was seeded in Dulbecco's modified Eagle's medium with 50% heat-inactivated FBS and penicillin/streptomycin. The serum concentration was gradually reduced to 10%, and the fibroblasts were maintained in Dulbecco's modified Eagle's medium with 10% heat-inactivated FBS and penicillin/streptomycin.
Collagen Fibril Dissolution Assay-The assay has been described recently and discussed in detail (53). Briefly, 24-well tissue culture plates were coated with type I collagen derived from rat tails (kindly provided by Dr. Jack Windsor, Indianapolis) as described previously (45), and 0.37 ml of 300 g/ml collagen was added to each well for 2 h at 37°C. The collagen film was then dried down, and the salt was removed by distilled water to leave a thin hydrated layer of collagen fibrils. The integrity of the collagen fibrils was verified by incubation of the film with either 25 mg/ml trypsin (Invitrogen), 20 ng/ml MMP-9 (kindly provided by Dr. William Stetler-Stevenson, National Institutes of Health, Bethesda), or 20 ng/ml MMP-1 (kindly provided by Dr. Jack Windsor, University of Indianapolis) at 37°C for 4 h and staining with Coomassie Brilliant Blue as described below. Keratinocytes were grown in Defined Keratinocyte-SFM supplemented with 1.0 or 1.3 mM CaCl 2 , and fibroblasts were grown in Opti-MEM (both from Invitrogen). Thirty thousand cells in 30 l of medium were applied to the center of each well, and the cells were allowed to attach to the collagen layer for 5 h at 37°C. Thereafter, the cells were incubated in 1 ml of medium with or without the following additions: 10% heat-inactivated FBS (HyClone Laboratories), 20 nM Glu-Plg, 50 nM mutant murine PAI-1, 40 nM aprotinin, 200 nM ␣ 2 -AP (all from Calbiochem-Novabiochem), 5 nM tumor necrosis factor-␣ (TNF-␣), 2.5 nM interleukin-1␤ (IL-1␤) (PeproTech, Inc., Rocky Hill, NJ), 160 nM phorbol 12-myristate 13acetate (Sigma), 200 nM TIMP-1 (kindly provided by Dr. Jack Windsor, Indianapolis), or 200 nM TIMP-2 (Fuji Chemical Industries, Toyama, Japan). To visualize zones of cell-mediated lysis, the cells were removed by a trypsin/EDTA solution (Invitrogen) containing 1% (v/v) Triton X-100 (Sigma). The wells were then stained with 0.5% (w/v) Coomassie Brilliant Blue R-250 for 5 min and rinsed with distilled water.
Northern Blot Analysis-Murine keratinocytes were grown in Defined Keratinocyte-SFM supplemented with 1.3 mM CaCl 2 for 18 h in the presence or absence of 5 nM TNF-␣ and 2.5 nM IL-1␤. Murine fibroblasts were grown in Opti-MEM for 15 h in the presence or absence of 5 nM TNF-␣ and 2.5 nM IL-1␤. Total RNA was isolated from the keratinocytes using the RNeasy Mini Kit (Qiagen Inc., Valencia, CA) according to the manufacturer's instructions. RNA (5 g/lane) was separated by electrophoresis on 1% agarose-formaldehyde gels and transferred to Nytran membranes (Schleicher & Schuell) by capillary blotting. The following DNA fragments were used as probes: full-length murine MMP-2, MMP-9, and MMP-13 cDNAs (54, 55); a full-length murine uPAR cDNA (56); a 407-bp murine uPA cDNA fragment containing nucleotides 201-608 (57); and a 302-bp murine MMP-14 cDNA fragment containing nucleotides 2065-2367 (43). The cDNA fragments were radiolabeled with [ 32 P]dCTP by random prime labeling using the Random Primed DNA Labeling Kit (Roche Diagnostics) and purified on ProbeQuant G-50 columns (Amersham Biosciences). Prehybridization was carried out at 65°C in 10 ml of Quickhyb (Stratagene, La Jolla, CA). The labeled DNA probe was boiled in 1 mg of salmon sperm DNA (Research Genetics, Huntsville, AL) and then added to the QuikHyb solution after 30 min of prehybridization. Hybridization was carried out for 2 h at 65°C. The membranes were washed 3 times for 20 min each in 2ϫ SSC with 1% SDS at 65°C and exposed overnight to a Phosphor-Imager screen (Amersham Biosciences).
Western Blot Analysis-Twenty five l of conditioned medium from murine keratinocytes that were grown for 5 days on collagen-coated plates was resolved under reducing conditions by SDS-PAGE on 4 -12% NuPAGE BisTris gels (NOVEX, San Diego, CA). The medium contained equivalent amounts of total protein as estimated by Coomassie Brilliant Blue staining. The separated proteins were electroblotted onto nitrocellulose membranes (Schleicher & Schuell), blocked with 5% nonfat milk and 0.1% Tween 20 (Sigma) in PBS, and probed with an anti-MMP-13 monoclonal antibody, LIPCO IID1 (Labvision Inc., Fremont, CA). Rat pro-MMP-13 (Labvision) and active rat MMP-13 (a kind gift from Dr. Jack Windsor, Indianapolis) were used as controls for the migration of the two forms of MMP-13. Plg/plasmin fragments were detected with a polyclonal rabbit antibody to human plasminogen, A0081 (Dako Inc., Carpinteria, CA). Antibody bound to the membrane was detected with horseradish peroxidase-conjugated goat anti-mouse or rabbit secondary antibody (Dako Inc., Carpinteria, CA), followed by exposure to Renaissance Solutions (PerkinElmer Life Sciences) or a Western Breeze kit (Invitrogen) and visualization by autoradiography.
Casein Zymography-Serum-free conditioned medium (25 l/lane) was resolved by SDS-PAGE under non-reducing conditions on a 12% gel that incorporated 1% boiled, nonfat dry milk, with or without 20 g/ml Plg (58). The gel was washed twice for 30 min each in 2.5% Triton X-100 to remove SDS and renature proteins and developed overnight in 0.1 M glycine, pH 8.0, at 37°C. The gel was stained with 0.1% Amido Black (Bio-Rad) and destained in 30% methanol and 10% acetic acid solution to visualize the zones of lysis.
␣ 2 -Macroglobulin Capture-Conditioned medium from cytokinestimulated keratinocyte cultures was incubated for 1 h with purified ␣ 2 -macroglobulin (55 nM final concentration) at room temperature. The samples were incubated for 3 h in SDS sample buffer with 5% mercaptoethanol and the protein complexes resolved on 4 -12% NuPAGE gels, and MMP-13-␣ 2 -macroglobulin protein complexes were visualized by Western blot analysis with MMP-13 antibodies as described above.

Plg Is Required for Collagen Dissolution by Keratinocytes but
Not by Fibroblasts-We used a previously described cell-mediated collagen fibril film dissolution assay (53), to study the ability of primary keratinocytes and fibroblasts to degrade fibrillar collagen. The film was easily degraded by incubation with purified MMP-1 but resistant to degradation by both purified trypsin and MMP-9, confirming the integrity of the triple helical collagen fibrils (data not shown). Human primary keratinocytes degraded this collagen fibril film when stimulated by phorbol 12-myristate 13-acetate or cytokines to induce MMP expression (Fig. 1, panel A, and data not shown). As shown previously for fibroblasts (43), collagen breakdown took place exclusively under the layer of seeded cells and occurred in the presence of serum. Collagen degradation was prevented by the exogenous addition of both TIMP-1 and TIMP-2 to the medium, demonstrating the critical role of MMPs in the process. Surprisingly, however, collagen dissolution was also completely blocked by the addition of PAI-1, suggesting a specific requirement of Plg activators in the process. To further investigate the possible role of the PA system in collagen breakdown by keratinocytes, Plg was depleted from the culture medium by lysine-Sepharose chromatography (59). Interestingly, the depletion of Plg from the serum also completely prevented collagen dissolution, further implicating the PA system in keratinocyte-mediated collagen breakdown. In accordance with this finding, the keratinocytes dissolved fibrillar collagen under serum-free conditions only when supplemented by exogenous Plg. When aprotinin, which inhibits both cell surface bound plasmin and plasmin in solution, was added at a 2-fold molar excess to Plg, collagen dissolution was completely blocked. In contrast, the serpin ␣ 2 -AP, which is an excellent inhibitor of plasmin in solution but a poor inhibitor of cell surface-bound plasmin, did not prevent collagen breakdown, even when present in a 10-fold molar excess to Plg. Complete inhibition of collagen dissolution occurred only when a greater than 30-fold molar excess of ␣ 2 -AP was added. Taken together, these data demonstrate that human primary keratinocytes require MMP activity and cell surface Plg activation for fibrillar collagen degradation.
The availability of mice with single and combined deficiencies in components of the PA and MMP systems facilitated a more detailed genetic analysis of collagen dissolution by keratinocytes. We first confirmed that murine primary keratinocytes also required both MMP and cell surface Plg activation for fibrillar collagen dissolution by seeding first passage keratinocytes isolated from newborn mice onto collagen-coated wells under serum-free conditions (Fig. 1, panel B). Similar to human keratinocytes, the dissolution of collagen fibrils occurred exclusively under the cell layer and only in the presence of Plg. This breakdown was dramatically enhanced by the addition of TNF-␣ and IL-1␤ (Fig. 1, panel B), as well as by phorbol 12-myristate 13-acetate or TGF-␤1 (data not shown). Furthermore, the addition of TIMP-1, TIMP-2, PAI-1, or aprotinin all completely blocked collagen dissolution, suggesting that human and murine keratinocytes utilize a similar pathway for collagen degradation.
We have shown previously (43) that murine fibroblasts require MMP-14 in order to dissolve a layer of collagen fibrils. Collagen dissolution assays were performed with murine dermal fibroblasts in the presence or absence of Plg to determine whetherfibroblastsweredependentonthePAsystemforMMP-14dependent collagen dissolution (Fig. 1, panel C). Collagen dissolution by fibroblasts occurred in the absence of Plg, compatible with MMP-14 being activated by furin or furin-like proteases. Furthermore, no enhancement of the capacity of fibroblasts to degrade collagen was observed by the addition of Plg to the medium. Consistent with these findings, collagen dissolution was unaffected by the addition of PAI-1. Furthermore, TIMP-1, which efficiently prevented collagen breakdown by both human and murine keratinocytes, did not affect collagen dissolution by murine fibroblasts. In conclusion, keratinocytes and fibroblasts appeared to utilize fundamentally different pathways for the dissolution of fibrillar collagen.

Expression of MMPs and Components of the PA System in Cultured Murine Keratinocytes and Fibroblasts-
The expression of mRNA for collagenases and components of the PA system were analyzed in primary murine keratinocytes and primary dermal fibroblasts, as a first step toward elucidating the novel pathway responsible for keratinocyte-mediated collagen dissolution. Total RNA was isolated from unstimulated cells seeded on collagen or cells that were seeded on collagen and stimulated with TNF-␣ and IL-1␤, and the RNA was subjected to Northern blot analysis (Fig. 2, panel A). Overall, the expression of the components of the PA and MMP systems was quite similar in the two types of cells. MMP-9, MMP-13, MMP-14, uPAR, and uPA mRNA were all expressed by both keratinocytes and fibroblasts. Cytokines enhanced the expression of MMP-9, MMP-13, uPAR, and uPA in both cell types. MMP-2 and MMP-15 mRNA was abundantly expressed by cytokinestimulated fibroblasts but could not be detected in keratinocytes. The low level of mRNA for tPA precluded detection by Northern blot hybridization. However, both murine keratinocytes and fibroblasts expressed tPA as determined by casein-Plg zymography (Fig. 2, panel B).
Murine Keratinocytes Require Either uPA or tPA for Collagen Dissolution-We next sought to identify the specific proteolytic components that were essential for keratinocyte-mediated collagen degradation using keratinocytes derived from mice genetically deficient in either MMP-14, uPA, tPA, uPAR, uPA, and tPA or uPAR and tPA (Fig. 3). In contrast to previous observations (43) with murine fibroblasts, MMP-14 deficiency did not affect the ability of keratinocytes to degrade collagen, underscoring the fundamentally different pathways for collagen dissolution utilized by the two cell types (Fig. 3, column B). Surprisingly, keratinocytes with single deficiencies in either uPA or tPA also retained their ability to dissolve collagen in a Plg-dependent and PAI-1-inhibitable manner (Fig. 3, columns C and D). We therefore generated keratinocytes from mice with a combined deficiency in uPA and tPA to determine whether the lack of effect of single deficiencies in uPA or tPA was a consequence of the existence of a third, functionally important PAI-1-sensitive Plg activator or, rather, was due to a functional redundancy between the two established Plg activators. Combined deficiency in uPA and tPA completely abolished the ability of keratinocytes to dissolve fibrillar collagen, suggesting that uPA and tPA have a redundant function in this process and demonstrating a novel, unanticipated, role of tPA in keratinocyte-mediated collagen dissolution (Fig. 3, column E). Keratinocytes deficient in uPAR retained their full collagenolytic potential (data not shown). This could be due to either the residual tPA expression in these cells, the existence of a uPAmediated, but uPAR-independent, Plg activation pathway, or both. To distinguish between these options, we determined the ability of keratinocytes with combined deficiencies in uPAR and tPA to dissolve collagen. These cells express uPA but are only capable of activating uPA in a uPAR-independent manner. Surprisingly, however, uPAR-and tPA-deficient cells retained their ability to break down collagen fibrils in a Plg-dependent and PAI-1-inhibitable manner, suggesting that uPA-mediated Plg activation associated with collagen dissolution can occur independently of uPAR in this model system (Fig. 3, column F).
In conclusion, collagen dissolution by keratinocytes requires the expression of either uPA or tPA. The function of the two Plg activators in this process is fully redundant, and collagen dissolution can occur independently of the expression of uPAR.
Plg Is Activated by Either uPA or tPA in Murine Keratinocytes-We next determined the specific proteolytic requirements for the activation of Plg during collagen dissolution. Conditioned medium from wild type murine keratinocytes in the process of dissolving collagen was collected, and plasmin was detected by Western blot and casein zymography (Fig. 4). A catalytically active plasmin fragment of ϳ40 kDa could be recovered from the conditioned medium of the culture (Fig. 4,  panels A and E). This fragment corresponds to one of the major plasmin fragments generated by cultured HT1080 fibrosarcoma cells (60). The 40-kDa plasmin fragment was generated both in the absence and in the presence of TNF-␣ and IL-1␤, and the formation of plasmin was not blocked by the addition of TIMP to the culture medium, suggesting that MMP activity is dispensable for the generation of plasmin. In contrast, the addition of PAI-1 to the culture medium completely prevented the generation of this plasmin fragment (Fig. 4, panels A and  E). Plasmin also accumulated in the conditioned medium of keratinocytes with single deficiencies in either tPA or uPA, in accordance with the ability of keratinocytes with single uPA or tPA deficiency to degrade collagen (Fig. 4, panels B and C). However, no catalytically active form of plasmin was present in conditioned medium from keratinocytes deficient in both uPA and tPA (Fig. 4, panel D). Taken together, these data underscore the functional redundancy between uPA and tPA in activating Plg during keratinocyte-mediated collagen dissolution.
Plg and uPA or tPA Is Required for MMP-13 Activation by Murine Keratinocytes-The fibrillar collagenases MMP-2, -8, and -15 were not expressed by murine keratinocytes, and MMP-14 was not required for keratinocyte-mediated dissolution of collagen fibrils (Figs. 2 and 3). In contrast, MMP-13 mRNA was abundant in keratinocytes, and cytokines that stimulated collagen dissolution also enhanced the expression of MMP-13 mRNA ( Fig. 2 and data not shown). Moreover, plasmin has been proposed to be a physiological activator of pro-MMP-13 (see Introduction). We therefore focused on MMP-13 as the potential fibrillar collagenase responsible for collagen dissolution, and we analyzed the role of the PA system in the activation of MMP-13 by murine keratinocytes during collagen dissolution. Aliquots of conditioned medium from wild type keratinocytes were resolved by SDS-PAGE, transferred to nitrocellulose membranes, and probed with an anti-MMP-13 antibody that recognizes both pro-MMP-13 and active MMP-13 (Fig. 5, panel A). Interestingly, only pro-MMP-13 was present in the culture medium in the absence of Plg. When Plg was present, a truncated form of MMP-13 was generated with a corresponding reduction in the amount of pro-MMP-13. This truncated MMP-13 species was shown to be the active form of MMP-13 by its ability to form an SDS-stable complex with ␣ 2 -macroglobulin (data not shown). In further accordance with a critical role of MMP-13 in collagen dissolution, the addition of PAI-1 to the culture medium completely prevented the activation of MMP-13. Interestingly, the presence of either TIMP-1 or TIMP-2 in the culture medium did not prevent the activation of MMP-13, suggesting that the activation of MMP-13 by keratinocytes is independent of the activity of other MMPs. Keratinocytes with single deficiencies in either tPA or uPA displayed an MMP-13 activation pattern that was identical to that of wild type keratinocytes, in accordance with the uncompromised capacity of these cells to degrade collagen (Fig. 5, panels  B and C). In contrast, only pro-MMP-13 was detected in the conditioned medium of keratinocytes that were deficient in both uPA and tPA, demonstrating the fully redundant function of the two Plg activators in the pathway leading to productive activation of pro-MMP-13.
In conclusion, these data suggest that Plg activation by either uPA or tPA is required for keratinocyte-mediated dissolution of fibrillar collagen. Plasmin acts to promote collagen dissolution by directly activating MMP-13 and possibly other fibrillar collagenases. DISCUSSION The data presented in this paper demonstrate that keratinocytes utilize a novel pathway for the dissolution of fibrillar collagen that is different from the collagenolytic pathway employed by mesenchymal cells. The data also provide the first demonstration of a direct requirement of the PA system for cell-mediated collagen degradation. The novel pathway is strictly associated with the keratinocyte cell surface and in- cludes the activation of Plg by either uPA or tPA and the subsequent proteolytic activation of the fibrillar collagenase, MMP-13, by cell surface-associated plasmin. The initial truncation of MMP-13 is mediated directly by plasmin and does not appear to require the participation of other (TIMP-inhibitable) MMPs. Two aspects of the process that lead to the successful activation of MMP-13 merit particular attention. The first is the apparently fully redundant function of uPA and tPA in the activation of Plg. uPA-catalyzed Plg activation has traditionally been associated with extracellular matrix degradation in the context of tissue remodeling, while tPA-catalyzed Plg activation has been traditionally associated with thrombolysis (14,61). Our findings are, nonetheless, in accordance with several recent studies in Plg activator-deficient mice that have demonstrated a remarkable extent of functional redundancy between the two activators in many physiological processes. These include thrombolysis, extravascular fibrin degradation, liver and kidney regeneration, and incisional skin wound healing (48,50,(62)(63)(64). Furthermore, tPA-catalyzed Plg activation has been shown recently (24,63,65) to mediate the degradation of other substrates besides fibrin, including laminin and DSD-1-PG/ phosphacan, and to have physiological and pathological activities not strictly related to fibrinolysis. However, this is the first study that directly links tPA-catalyzed Plg activation to MMP activation and fibrillar collagen dissolution, further extending the range of potential physiological functions of tPA. tPA and Plg bind to fibrin with high affinity, and fibrin strongly potentiates tPA-catalyzed plasmin formation by aligning the proteases in a spatially favorable orientation for activation and by protecting them from specific soluble inhibitors (61). How tPA-catalyzed Plg activation is accomplished on the cell surface is less well understood but seems to involve the specific and saturable binding of tPA to one or more, possibly cell typespecific, tPA receptors (66 -67).
The second unanticipated finding in this study concerning Plg activation by keratinocytes is the lack of a specific requirement of uPAR for uPA-catalyzed, cell surface Plg activation. Previous studies (11,15) with both normal and malignant cells have demonstrated that the concomitant binding of pro-uPA to uPAR and of Plg to specific cell surface receptors strongly potentiates uPA-mediated Plg activation and plasmin-medi-ated activation of pro-uPA, leading to a powerful feedback loop that results in productive plasmin formation. However, mounting evidence now suggests that alternative pathways for uPAcatalyzed, cell surface Plg activation must exist. First, uPARdeficient mice present overall superior health as compared with uPA-deficient mice, and uPAR-deficient mice, unlike uPAdeficient mice, retain the full ability to efficiently repair tissue damage, such as vascular injuries and incisional skin wounds (49 -50, 68). Second, several cell lines that do not express uPAR were reported recently to also strongly potentiate cell surfaceassociated, uPA-catalyzed Plg activation (69). Collectively, the data presented here and in these previous studies imply the existence of alternative cell surface receptors for uPA capable of facilitating uPA-catalyzed, cell surface Plg activation, perhaps acting in parallel to the uPAR-dependent pathway. Indeed, specific binding sites for uPA, different from uPAR, were reported recently (70) to be present on the surface of smooth muscle cells.
The juxtaposition of pro-uPA and Plg on the cell surface strongly potentiates uPA-mediated Plg activation and plasminmediated activation of pro-uPA. How this feedback loop is initiated on the surface of keratinocytes, or other cell types, is unclear but may include the activation of pro-uPA by other proteases besides plasmin, or the activation of Plg by a very low intrinsic activity of pro-uPA. The transmembrane serine protease matriptase/MT-SP1, true tissue kallikrein, hepatocyte growth factor activator, and cathepsin B are all candidate initiators of the pro-uPA cascade (71)(72)(73)(74)(75), and their specific involvement in Plg-catalyzed collagenase activation by keratinocytes remains to be explored. Of the aforementioned pro-uPA activators, the transmembrane serine protease, matriptase/ MT-SP1, is highly expressed by keratinocytes during skin wound healing and represents a plausible candidate pro-uPA activator in this process. 2 The data presented here strongly implicate MMP-13 as the collagenase utilized by murine keratinocytes to dissolve collagen fibrils. However, the possibility of other known MMPs also being involved in Plg-dependent collagenolysis by keratino-2 K. List and T. Bugge, unpublished data. cytes cannot be excluded. MMP-14 could be clearly eliminated as an essential protease in the process, and the participation of other membrane-type MMPs such as MMP-15 and MMP-16 is improbable due to the ability of TIMP-1 to effectively block collagen dissolution by keratinocytes. MMP-2 and -8 were not expressed in detectable amounts by keratinocytes, as assessed by Northern blot hybridization or gelatin zymography (data not shown). The existence of very small, but functionally important, amounts of MMP-2 or -8 nevertheless cannot be unequivocally excluded by the expression studies. However, previous extensive biochemical studies did not render plasmin a likely candidate as an activator of pro-MMP-2, and the participation of MMP-2 and -8 in Plg-dependent collagen dissolution by keratinocytes is altogether improbable. MMP-1 is expressed by human keratinocytes and is involved in the cleavage of fibrillar collagen by these cells (76). The existence of a murine orthologue of MMP-1 remained unclear for a considerable time period, due to the repeated inability to detect a cDNA in murine tissues that hybridized to human MMP-1. Most recently, however, a novel murine MMP gene, Mcol-A, was reported (77) that occupies a position syntenic to the human MMP-1 locus and encodes a functional protease with high homology to human MMP-1. It remains to be shown if Mcol-A is expressed by murine keratinocytes and contributes to plasmin-dependent fibrillar collagen dissolution.
In summary, we have identified a novel pathway for collagen dissolution by keratinocytes that is strictly dependent on fibrillar collagenase activation by the cell surface Plg activation system. The novel pathway may be relevant to physiological remodeling of the skin, oral cavity, and urogenital tract, includ-  ing the healing of excisional, incisional, chemical, freeze, and burn injuries, and pathological tissue remodeling processes such as keloid formation, chronic ulcers, blistering diseases, and squamous cell carcinoma. When taken together with previous studies (43) of MT1-MMP-dependent collagen dissolution by mesenchymal cells, the data underscore the likely existence of multiple, cell type-, and context-specific pathways for collagen dissolution.