Degradation of type I collagen by rat mucosal keratinocytes. Evidence for secretion of a specific epithelial collagenase.

Feeder-cell-independent serially propagating keratinocytes from rat oral mucosa (tongue) dissolved reconstituted type I [3H]collagen fibrils, although rather slowly. Analysis of the conditioned medium from such cultures revealed secretion of a Mr = 65,000 collagenase which remained almost entirely latent in the absence of exogenous protease activity. Addition of trypsin (0.1-1.0 microgram/ml) or plasmin (1.0-4.0 micrograms/ml) resulted in substantial acceleration of the collagenolytic process in stimulated secretion of latent collagenase and, at higher concentrations, in conversion of the latent enzyme to the catalytic form. The keratinocyte collagenase was indistinguishable from interstitial, fibroblast-type collagenases by several criteria including: cleavage of native type I collagen in solution at the characteristic collagenase-sensitive locus at 22 degrees C and dissolution of reconstituted type I collagen fibrils at 35 degrees C; activation by trypsin and by organomercurials and inhibition by Zn2+ and Ca2+ chelators; and cross-reaction with antibody to fibroblast-type procollagenase. Expression of collagenolytic activity in keratinocyte cultures was effectively regulated by cell density. The activity (on a per cell basis) was maximal at 10-20% confluence and was more than 95% "contact-inhibited" at subconfluent and early confluent densities (2-4 X 10(5)/cm2). Our findings show that mucosal keratinocytes possess a potent enzymatic apparatus for degradation of interstitial collagen fibrils which includes a classical vertebrate collagenase.

Feeder-cell-independent serially propagating keratinocytes from rat oral mucosa (tongue) dissolved reconstituted type I [SH]collagen fibrils, although rather slowly. Analysis of the conditioned medium from such cultures revealed secretion of a M, = 65,000 collagenase which remained almost entirely latent in the absence of exogenous protease activity. Addition of trypsin (0.1-1.0 Ng/ml) or plasmin (1.0-4.0 pg/ml) resulted in substantial acceleration of the collagenolytic process in stimulated secretion of latent collagenase and, at higher concentrations, in conversion of the latent enzyme to the catalytic form. The keratinocyte collagenase was indistin~ishable from interstitial, fibroblasttype collagenases by several criteria including: (i) cleavage of native type I collagen in solution at the characteristic collagenase-sensitive locus at 22 "C and dissolution of reconstituted type I collagen fibrils at 35 "C; (ii) activation by trypsin and by organomercurials and inhibition by Zn2+ and Cas+ chelators; and (iii) cross-reaction with antibody to fibroblast-type procollagenase. Expression of collagenolytic activity in keratinocyte cultures was effectively regulated by cell density. The activity (on a per cell basis) was maximal at 10-20% confluence and was more than 95% "contact-inhibited" at subconfluent and early confluent densities (2-4 X 105/cm2), Our findings show that mucosal keratinocytes possess a potent enzymatic apparatus for degradation of interstitial collagen fibrils which includes a classical vertebrate collagenase.
The biologic degradation of interstitial collagen fibrils presumably is initiated by proteolytic attack on the constituent collagen molecules by the enzyme collagenase (1,2). In support of this concept, collagenase and its inactive precursor, procollagenase, have been identified in conditioned culture media or extracts of cells classically associated with connective tissue metabolism in healthy and inflamed tissues (3-lo), Invasion of mesenchymal domains by epithelial cells during fetal development and during carcinoma growth requires dissolution of interstitial (type I and 111) collagen fibrils which form the skeletal framework of the subepithelial connective tissue. It is still unresolved whether invasive epithelial cells degrade stromal connective tissue by secretion of lytic en-zymes or by instruction of surrounding mesenchymal cells (11)(12)(13)(14). Each of these pathways is supported by a body of evidence. Secretion of collagenase by epithelial cells has been reported in explants prepared from metamorphosing tadpole tail skin (15), healing experimental wounds (16,17), and inflamed human oral mucosa (gingiva) (18). Other studies, however, have failed to verify collagenase secretion by epithelial cells in vivo and in vitro and have shown instead that cultured skin and corneal epithelial cells release soluble factors which induce secretion of the enzyme in cocultured fibroblasts (12,19,20). Moreover, immunofluorescent staining of human skin carcinomas (14) and of human gingiva (21) have shown that the enzyme is abundant in mesenchymal cells adjacent to epithelium but absent in epithelial cells. These findings, taken together, suggest that epithelial collagenase may have a narrow ~i n d o w of expression and that its synthesis and secretion may require specific activating stimuli.
We have investigated the degradation of interstitial (type I) collagen by rat mucosal epithelial cells in culture and now provide evidence for density-dependent expression of a collagenolytic apparatus which involves secretion of a true vertebrate, M, = 65,~/60,000, t~sin-activatable collagenase.

MATERIALS AND METHODS'
RESULTS Development and Maintenance of Feeder C~~l -i n d e~n d e n t Clonal Keratinocyte Lines from Rat Oral Mucosa-Rat mucosal keratinocytes established in culture under hypothermal conditions as summarized in Fig. 1, A-H retained distinctive epithelial phenotypic traits at 32 "C as well as 37 "C. These included flattened oval or discoid single cell shape (Fig. 1, E and G), formation of expanding sheets of tightly packed polygonal cells during colonial growth (Fig. 1, F and H), formation of multiple focal cell contacts (Fig. 1H), and eventually, formation of coherent mutilayered sheets with varying degrees of organoid stratification as previously described for a sister line (27). Trypsin-treated cells attached within 1-2 h to pfastic as well as to the collagen fibril coating described below but spread rather slowly and frequently required overnight incubation to spread completely.
Dissolution of Type I Collagen Fibrils by Mucosal Keratinocytes-Keratinocytes seeded at a density of 50,000/cm2 (15-' Portions of this paper (including "Materials and Methods") are presented in miniprint at the end of this paper. Miniprint is easily read with the aid of a standard magnifying glass. Full size photocopies are available from the Journal of Biological Chemistry, 9650 Rockville Pike, Bethesda, MD 20814. Request Document No. 86M-1956, cite the authors, and include a check or money order for $3.20 per set of photocopies. Full size photocopies are also included in the microfilm edition of the Journal that is available from Waverly Press.

6823
FIG. 1. Development of feeder cell-independent keratinocyte lines from rat oral mucosa. A , rat oral mucosa from the ventral surface of the tongue was explanted in a 35-mm dish and incubated for 14 days at 32 "C with MEM, 20% fetal calf serum, 0.5% dimethyl sulfoxide. Note absence of fibroblasts around the continuous epithelial sheet. Stained with Coomassie Blue. B, clonal density subculture (REP) seeded on a bed of human skin fibroblasts and grown for 7 days at 32 "C as described above. Stained with Rhodanile Blue. C, fourth passage culture (REP) seeded at clonal density on a human fibroblast feeder layer and incubated at 37 "C for 14 days. The fibroblasts are largely displaced from the growth surface and appear as dark dotted lines around the epithelial colonies (arrows). Stained with Rhodanile Blue. D, fourth passage colony seeded without feeder cells. Multilayering and stratification (pyknotic nuclei) are evident toward the center (thick arrow). A single fibroblastic cell is seen at the edge of the colony (thin arrow). Phase contrast, bur is 50 pm. E, fifth passage cells (REP) derived by subcultivation of a culture similar to that shown in Fig. 1C. The cells display the discoid or circular shape typical of single epithelial cells. Phase contrast, bur is 100 pm. F, confluent monolayer (REP) showing cobblestone morphology. Phase contrast, bur is 100 pm. G, developing three cell colony (CC1-4). SEM, bur is 10 pm. H, subconfluent to early confluent monolayer. The cells are connected by multiple focal contacts; the intercellular gaps are still open at this time but will close as the culture becomes confluent. Scanning electron micrograph, bur is 10 pm. 20% confluency) dissolved the fibril coating after a character-recovered almost entirely in latent form and by day 8, after istic 3 to 4-day lag period (Fig. 2). At the same time, collagen-the entire substrate layer was dissolved, still only 16% of the olytic activity accumulated in the medium at a rate of 0.08-enzyme had been activated. We considered the possibility that 0.10 unit/ml/day. During the lag period the enzyme was the active enzyme initially bound to the collagen coating, but media from companion cultures seeded on plastic did not contain more active enzyme (date not shown). These findings prompted us to calculate the rate of secretion and activation of collagenase actually required to dissolve the collagen layer. The result showed that generation by the cell layer of as little as 15 milliunits of active collagenase/ml/day or a total of 0.1 unit/ml over an 8-day period, which is close to that observed in Fig. 2, was sufficient to dissolve the 220-225 pg collagen fibrils.
Although the level of fibroblast contamination at this stage was below the detection limit of 0.1%, we proceeded to clone the epithelial cells by limiting dilution to avoid ambiguity as to the origin of the enzyme. Fifteen out of eighteen clones examined secreted clearly detectable levels of latent collagenase activity (0.05-0.35 unit/ml) ( Table I).
Keratinocyte Collagenase-The latent collagenase harvested from unstimulated mucosal keratinocyte cultures was activated to approximately the same level by trypsin and by organomercurials. Optimal activity was achieved at 5-20 pg/ ml of trypsin (10 min, 22 "c) and at 0.5-1.0 mM 4-aminophenylmercuric acetate (present throughout the assay). Trypsin-activated harvest medium protein made a single initial cleavage of rat tail tendon type I molecules as evidenced by generation of 3h and 1 h length fragments (Fig. 3, A and B).
With longer incubation times an additional, shorter fragment accounting for approximately 63% of the a-chain size was formed by excision of 120-130 amino acids from the COOHterminal end of the native 75% fragment. Generation of this 63% fragments (Fig. 3B, arrows), which is characteristic of harvest media from cells and tissues of the rat (45-47), was blocked by addition of the serine-protease inhibitor PMSF2  Rat mucosal keratinocytes (REP) were cloned by limiting dilution in 24-well plates. Eighteen subconfluent to early confluent clones (30-60 X IO3 cells/well) were washed with phosphate-buffered saline and incubated for another 3 days with serum-free MEM. The collagenolytic activity of duplicate samples of 50 pl of culture medium was determined by radiofibril assay after preincubation with 2 pg of trypsin for 10 min at 22 "C followed by addition of 15 molar excess of soybean trypsin inhibitor. Complete lysis of the 30-pg collagen gel was achieved by collagenolytic activities at or above 0.7 unit/ml (16 h, 35 "C). and therefore in all likelihood was catalyzed by a protease distinct from collagenase. The inhibition pattern ( Table 11) was consistent with that of a metallo-protease and closely resembled those previously reported for mesenchymal collagenases (1,2), with the possible exception of a rather surprising 65% inhibition by 1 mM p-nitrophenyl-p'-guanidino benzoate, an active-site titrant for trypsin-like serine proteases. The inhibitory property resided with the intact compound as neither of its two hydrolysis products, p-nitrophenol and guanidinobenzoate, had any effect. No inhibition was observed with diisopropylfluorophosphate. Of the macromolecular protease inhibitors only azmacroglobulin blocked the activity of the enzyme.
The electrophoretic migration of the enzyme was studied by zymography under nondenaturing conditions. Concentrated harvest medium was resolved by SDS-PAGE and the resultant electrophoretogram incubated in contact with a native collagen fibril film after removal of SDS by repeated washing with Triton X-100. The epithelial harvest medium produced a major lytic band at M, = 65,000 daltons which comigrated (within a few kilodaltons) with the human fibroblast procollagenase isozymes (Mr = 65,000, 60,000). In addition, unidentified minor trailing bands were seen at M, = 68,000 and 92,000 (Fig. 4A). The M, of the procollagenase standard, and presumably of epithelial collagenase as well, determined by zymography was 5,000-10,000 higher than that previously determined with the completely unfolded proteins (5). The specificity of the reaction was shown by inclusion of parallel lanes with trypsin (1 pg) and thermolysin (1 pg) which did not produce visible lysis of the substrate under the same conditions ( Fig. 4B, lanes 2 and 3). Substitution of gelatin for native collagen in the zymogram showed that the mass = 65 kDa collagenolytic enzyme had no measurable activity against gelatin and was electrophoretically distinct from the major M , 105,000 gelatin-cleaving protease secreted by rat mucosal keratinocytes (data not shown).
The results summarized above strongly suggested that the collagen-cleaving epithelial protease was a genuine vertebrate collagenase. This was further supported by immunoperoxidase staining of Western blots prepared from epithelial culture media as shown in Fig. 5. A polyclonal antibody raised in rabbits against human fibroblast procollagenase recognized a major double band (60,000, 65,000) which co-migrated with sulfate-polyacrylamide gel electrophoresis; TLCK, tosyl-lysyl-chloromethyl ketone; TPCK, tosyl-phenyl-chloromethyl ketone; t-PA, tissue-type plasminogen activator; u-PA, urokinase-type plasminogen activator; TBS, Tris-buffered saline. Inhibition of keratinocyte collagenase activity in epithelial harvest media Concentrated keratinocyte (CCl-10) harvest medium was dialyzed against assay buffer (50 mM Tris-HC1; 0.2 M NaCl; 5 mM CaCln; pH 7.4) and preincubated with TPCK trypsin (165 pg/unit collagenase) for 10 min at 22 "C. The incubation was terminated by addition of 15-fold molar excess of soybean trypsin inhibitor. Aliquots of 30 milliunits of collagenase activity were mixed with inhibitors dissolved either in assay buffer or in dimethyl sulfoxide (to yield a final solvent concentration of 1%, marked by *), and the resultant collagenase activity was measured by radiofibril assay. NPGB, p-nitropheny1-p'guanidine benzoate; APMA, 4-aminophenylmercuric acetate.

Inhibitor
Concentration  3. Cleavage of type I collagen in solution by keratinocyte collagenase. A, rat tail tendon type I collagen (0.7 mg/ml) dissolved in 50 mM Tris-HC1 buffer, pH 7.5,0.2 M NaCl, 5 mM CaC12, 0.5 M glucose was incubated at 22 "C with concentrated trypsinactivated harvest medium containing 0.8 unit/ml of keratinocyte collagenase. Aliquots of the reaction mixture were withdrawn at various time intervals and mixed with inhibitor (10 mM EDTA for collagenase; 1 mM PMSF for trypsin). Native helical fragments and uncleaved collagen were precipitated by addition of p-dioxane to 50% v/v and harvested by centrifugation. The precipitate was dissolved directly in sample buffer and resolved by electrophoresis in 7% polyacrylamide. Five pg of collagen was applied to each lane. Incubation was with buffer for 16 h ( l a n e I ) or with keratinocyte collagenase for 2 h ( l a n e 2), 4 h ( l a n e 3), 8 h ( l a n e 4), 16 h ( l a n e 5), or with trypsin (100 pg/ml) for 16 h ( l a n e 6). Silver-stained. B, keratinocyte harvest medium (1 unit/ml) was incubated with type I collagen (1 mg/ml) as described in Fig. 3A.
Helical cleavage products were precipitated by p-dioxane and compared with those generated by incubation with 1 unit/ml of human fibroblast collagenase ( l a n e I ) .
Each lane contained 5 pg of type I collagen. Incubation was with lane 2, keratinocyte collagenase; lane 3, keratinocyte collagenase + 1 mM PMSF. Note that PMSF blocked further breakdown of the % fragments (arrows). Silver-stained. human fibroblast procollagenase and two minor bands (45,000, 48,000), presumably activated forms of collagenase (Fig. 5, lane 2). Pre-immune IgG and immune IgG absorbed out by passage over a procollagenase-Sepharose column showed no reactivity (Fig. 5, lanes 3 and 4). On the other hand, each of nine murine monoclonal antibodies raised against human fibroblast procollagenase, including three which strongly inhibit the human e n~y m e ,~ failed to crossreact with the rat enzyme.

Stimulation of Collagen Breakdown by Proteases-
The rather high proportion of latent collagenase activity recovered in medium conditioned by rat mucosal keratinocytes (Fig. 2) suggested that the cells failed to activate the latent enzyme to any significant degree and that addition of proteases capable of converting procollagenase to the catalytic form (trypsin, plasmin) might substantially increase the rate of collagen breakdown. We therefore examined the effect of addition of trypsin (0.1, 0.3, and 1.0 pg/ml) on collagen breakdown by correlating the rate of fibril dissolution with the secretion and activation of collagenase as shown in Fig. 6. Addition of trypsin, as predicted, led to substantial acceleration of the rate of fibril dissolution. Moreover, higher levels of collagenase activity accumulated in the culture medium although not necessarily in an active form (Fig. 6, A uersus B, C and D).
For example, the lowest concentration of trypsin (0.1 pg/ml) gave an almost 5-fold stimulation of collagenase secretion but only a 2.5-fold increase in the rate of fibril dissolution, as most of the collagenase remained latent. Raising the trypsin concentration to 0.3 or 1.0 pg/ml (Fig. 6, C and D)   Each lane was incubated in succession with 1% BSA in borate saline (2 h at 22 "C), first antibody in Tris-buffered saline/Triton X-lOO/BSA (overnight at 4 "C followed by 1 h at 22 "C to allow for temperature equilibration), horseradish peroxidase-conjugated goat-anti-rabbit IgG (1:200) in TBS-Triton X-100 (1 h, 22 "C). Lane I, human fibroblast culture medium (20 pl) stained with rabbit anti-human fibroblast procollagenase IgG (60 pglml). Lane 2, keratinocyte medium (300 pl) stained with rabbit anti-human fibroblast procollagenase IgG (60 pglml). Lane 3, keratinocyte medium (300 pl) stained with rabbit preimmune IgG (60 pglml). Lane 4, keratinocyte medium stained with immune IgG fraction (60 pg/ml) after passage over a procollagenase-Sepharose affinity column. proCOLL, procollagenase; COLL, collagenase. stimulated fibril dissolution 8-10-fold as virtually all of the enzyme was converted to the active form (Fig. 6, B versus C  and D ) . At 1.0 pg/ml of trypsin the cells failed to reattach after dissolution of the collagen coating (arrow) and appar-Collagenase 6827 ently stopped secreting. Taken together, these findings suggest, somewhat surprisingly, that even subactivating concentrations of trypsin stimulated fibril dissolution (Fig. 6, A  versus B ) , although not nearly as much as higher concentrations (0.3 and 1.0 pg/ml) which activated the enzyme (Fig. 6, B versus C). These findings suggest that exogenous protease activity stimulated collagen breakdown in keratinocyte cultures by two independent but synergistic effects, namely by stimulation of secretion of latent collagenase. (Fig. 6, A versus  B-D) and by activation of the latent enzyme (Fig. 6, B versus   C and D ) .
Role of Plasminogen Activation-Based on work by Werb et al. (28) we considered the possibility that plasmin, which mediates tissue proteolysis in a variety of processes (48), might induce collagen breakdown in epithelial cultures. Plasminogen and plasmin (4.0 pg/ml) both stimulated collagen breakdown to the same extent ( Fig. 7) but did not measurably activate the latent collagenase (Fig. 6E). The ability of plasminogen to substitute for plasmin (Fig. 7) suggested that the cells converted the added plasminogen to the catalytic form (plasmin) by secretion of plasminogen activator(s). Analysis of the harvest medium by fibrin zymography revealed two bands of plasminogen-dependent fibrinolytic activity, a major M , = 48,000 band, presumably u-PA, and a minor, barely visible M , = 70,000 band, presumably t-PA (Fig. 7, inset).

Density Dependence of Keratinocyte Collagen Breakdown-Early experiments showed an apparently inverse relationship between cell density and substrate lysis in epithelial cultures.
This was further analyzed using cells of clone CC1-4 seeded at densities varying from 5 to 100% confluence (IO4 to 4 X 105/cm2). As shown in Fig. 8, the rate of collagen breakdown was maximal at 5-10 X 104/cm2 (10-20% confluence) and 15-20-fold lower at confluence (4 x 105/cm2). The rate of secretion of specific collagenase as measured by the radiofibril assay was also inversely related to cell density. Based on the total collagenase activity accumulated in the medium over the duration of the experiment, thinly spread cultures (5 X lo4/ cm2) secreted %fold more collagenase than subconfluent to early confluent (2 X 105/cm2) cultures (0.25 versus 0.03 unit/

DISCUSSION
This study has shown that clonal and parental rat mucosal keratinocytes maintained in culture are capable of dissolving native reconstituted type I collagen fibrils. Moreover, a collagenolytic protease was recovered from the harvest medium which was indistinguishable from genuine vertebrate collagenases by a number of criteria, most notably by its ability to cleave native molecules of type I collagen in solution at 22 "C in a manner characteristic of mammalian collagenases and to completely dissolve reconstituted, trypsin-resistant collagen fibrils at 35 "C. In support of this conclusion it was shown by zymography and immunostaining of Western blots that the collagenolytic activity co-migrated with (human) fibroblast procollagenase and was recognized by antibody to this enzyme. Other properties shared with vertebrate collagenases derived from mesenchymal cells include an inhibition pattern consistent with that of a metallo-protease and a requirement for exposure of the latent form to exogenous protease activity (trypsin) or to organomercurials (4-aminophenylmercuric acetate) for generation of catalytic activity. The epithelial collagenase possessed little or no activity against gelatin and in SDS-PAGE was clearly separated from the major Mr = 105,000 gelatinase also produced by these cells.
It has previously been shown that intact epithelial sheets isolated from the edges of healing wounds in guinea pigs and rabbits, and from thyroxine-induced involuting anuran tail- FIG. 6. Stimulation by trypsin and plasminogen of collagen breakdown by keratinocyte clonal line CCI-4. CC1-4 cells were seeded at a density of 200,000/cm2 in companion wells coated either with labeled or with unlabeled reconstituted collagen fibrils and incubated for 4 days with MEM, with 5 mg/ml BSA in the absence of serum. Trypsin (0.1,0.3, and 1.0 pg/ml) or plasminogen (4.0 pg/ml) was added at the beginning of the incubation period to stimulate collagen breakdown. For each set of companion cultures, the daily release of radioactivity (0, 0 ) (as a measure of fibril dissolution) and the accumulation in the medium of trypsin-activatable and spontaneous collagenase activity (A, A) were monitored. Medium from the latter cultures was divided into two aliquots; one was immediately mixed with 300 pg/ml soybean trypsin inhibitor and assayed for catalytically active enzyme by radiofibril assay; the other was preincubated with trypsin (20 pg/ml; 10 min, 22 "C), then mixed with (300 pglml) soybean trypsin inhibitor, and finally assayed for resultant total collagenolytic activity. 0, 0, radioactivity released from the substrate coating in the presence (0) or absence (0) Fig. 6, B, C, and D. Each point is mean of duplicate determinations. skin, secrete collagenolytic activity in tissue culture (11-13). This study, however, provides the first unequivocal evidence of elaboration of a genuine collagenase by serially propagated parental and clonal epithelial cells of any mammalian species and the first demonstration that keratinocytes are indeed capable of dissolving fibrils of type I collagen. The reactivity of antibody to human fibroblast procollagenase with Western blots of keratinocyte culture media suggests that the epithelial collagenase shares at least some antigenic determinants with the fibroblast enzyme, although further studies are necessary to determine the extent of homology. The maximal production of collagenase (0.25 unit/1O6 cells/day) was several-fold lower than that obtained with dermal fibroblasts in our laboratory (1-2 units/106 cells/day) but, if the smaller cell size of the keratinocyte is taken into account, the destructive potential of a subconfluent epithelial cell layer, as measured by the rate of fibril dissolution, is indeed comparable to that of subconfluent fibroblast cultures.
Previous studies by Johnson-Wint and collaborators (11,12,19,20) have shown that low passage cultures of rabbit epidermal and corneal epithelial cells do not produce collagenase but induce secretion of this enzyme in cocultured fibroblasts. Together with the independent finding by Bauer et at. (14) that stromal fibroblasts in human basal cell carcinomas contain immunoreactive collagenase protein, whereas the epithelial tumor cells do not, these studies have led to the suggestion that epithelial invasion of mesenchymal domains requires induction of degradative enzymes in fibroblasts across the epithelio-mesenchyma1 interphase. Our demonstration of a genuine vertebrate collagenase in media conditioned by keratinocytes provides evidence for existence of an alternate pathway utilizing an endogenous epithelial collagenase. It does not yet prove, however, that this pathway is actually used in vivo nor does it exclude the possibility that epithelial cells enhance their invasive potential by induction of adjacent mesenchymal cells. Still another degradative mechanism deserves further consideration. Studies by Birek et al. (49) have shown that epithelial cells from porcine periodontal ligament are capable of phagocytizing collagen fibrils in culture. This process clearly bears resemblance to the phagocytic uptake of fibril fragments by fibroblasts which has been well documented in vivo and in vitro (50,51). The role of collagenase in this process, if any, remains unclear. It is still uncertain whether collagenase-mediated and phagocytic collagen breakdown are two sequential stages of a single pathway (i.e. initial extracellular fragmentation of fibrils by collagenase followed by phagocytic uptake and further degradation in phagolysosomes) or represent two altogether distinct and alternative degradative mechanisms. The biologic role of either mechanism in keratinocyte-mediated collagen breakdown in vivo remains to be established.
The observation that rat mucosal keratinocytes activated only a small fraction ((20%) of the latent collagenase released to the culture medium raises the question whether exogenous components are required for activation of the enzyme in vivo or, alternatively, whether secretion and activation of procollagenase are independently regulated and therefore not necessarily coordinately expressed in culture. However, this question cannot be answered with any degree of certainty on the basis of current knowledge inasmuch as the biologic pathway for procollagenase activation remains obscure. At least four mechanisms have been proposed involving: (i) an endogenous procollagenase-activating protease (52); (ii) an apparently noncatalytic protein found in the skin and in the ~volutjng uterus (53); (iii) a nonenzymatic thiol-exchange reaction (54); and (iv) an exogenous protease such as plasmin (28). A number of studies have shown that plasmin can activate procollagenase in test tube experiments but it is uncertain to what extent this reaction takes place in vivo or even in vitro in cell culture systems. Studies by Werb et al. (28) have shown that human synovial fibroblasts possess far more activity against collagen substrates when grown in the presence of plasminogen than in its absence, although plasmin by itself has little or no lytic activity against collagen. Further studies (55, 56) have indicated that the mechanism in all likelihood is more complex than merely a conversion of procollagenase to the catalytic form since many proteolytic enzymes, including plasmin, stimulate secretion of procollagenase under conditions which yield little or no activation of the zymogen. In the present study, we have also found elevated secretion, but little or no activation, of latent collagenase in the presence of plasminogen/plasmin. Trypsin had a similar effect at or below 0.1 pg/ml but did activate the latent collagenase at higher concentrations. Although the dissolution of collagen fibrils by the cells was somewhat stimulated even at subactivating protease concentrations, it was apparent that the rate of collagen breakdown was closely linked to the generation of active enzyme.
The observation made in previous studies (16, 17) that epithelium derived from the edges of dermal wounds secretes collagenolytic activity in organ culture, whereas that from uninjured skin does not, suggests that secretion of collagenase by epithelial cells may be a wound healing response. We propose that enzymatic dissociation and plating of epithelial cells in culture at low densities in many ways simulate wound healing conditions and induce a similar response (i.e secretion of collagenase) which is eventually turned off as the cells reach confluence. This offers a rational explanation of the apparent enigma that the enzyme is secreted in culture but not widely expressed in vivo. The inverse relationship between cell density and collagenolytic activity ("contact inhibition") which resulted in a 90-95% reduction of activity per cell in subconfluent and confluent cultures shows that two different phenotypes are expressed in sparsely and densely populated cultures and that secretion of collagenase is predominately associated with the low density phenotype. Johnson-Wint (11) has previously shown that secretion by rabbit corneal epithelial cells of two distinct sets of regulatory factors, one which stimulates and one which inhibits stromal cell collagenase production, is also highly density regulated. The stimulating factors secreted at low cell density and the inhibitory components released at higher densities presumably serve the same overall purpose, namely to regulate the collagenolytic activity in the immediate environment. Although the culture conditions are not quite comparable, it is of note that Johnson-Wint's study (11) as well as ours showed maximal (collagenolytic) activity at 25-50 x lo4 cells/cm* which represents 5-10% of confluent densities. The dependence of collagenolytic activity on cell density represents a radical departure from the behavior of dermal fibroblasts which are rather insensitive to changes of cell density in vitro and continue to secrete collagenase at confluency (57). It will be of considerable interest to determine whether the collagenolytic activity of rat mucosal keratinocytes is regulated by a set of (endogenous) factors similar to those described by Johnson-Wint (11

1662-1671
The development of epithelial cell lines is summarized in Fig. I . Briefly, explants of whole muc06a were attached to culture dishes with a chick plasma Coagulum and incubated at 32% in Eagle's minimum essential medrvm IMEMI with 20% fetal calf serum, 0.5% DMSO. 2 mM glutamine and antibiotics (200 IU/ml penicillin, 20 ug/ml gentamicin. IO0 u/ml fungironel. The ovtgrovth ( S e e Pis. 1 R ) consisted oredominatlv of eoithelial c e l l s since fibroblasts ~~~~ grow poorly, if at all, at this temperature. After one to two weeks the explants were dleaggregated by trypsin at 32OC 10.25U trypsin, 2 mn EDTA In phosphate buffered sallne without Cat+ and Mg") IPBS-I and the resultant single cell suspension was seeded either at subconfluent densit 12-3 x 105/cm21 in ordinary tissue Culture dishes without feeder c e l d t clonal dentity on a feeder layer of human skin fibroblasts IPig. 1s). Feeder cella were established by trypsinization (after removal of fibroblatn by trypsln/EDTAI and seeded a% before.
During the n e x t 2-4 pasrages most cultures (50-70UI Outgrew the dependence on feeder cell6 and farmed seemingly but rose to more than 8OU; at the same time the population doubling tlme lmmortal lines: the plating efficiency was initially low (less than 1%) morphologic alterations. At the end of the transition period the eplthelial fell from 30-40 h to 14-18 h. These Changes were not associated with notable ellmination of fibroblasts was ultimately accomplished elther by (iI numerical cells rapidly outgrew and displaced the fibroblasts [ Fig. 1CI. Complete Overgrowth assisted by washing with trypsin-EDTA or by (iiI cloning by limlting dllutron. l i ) Repeated seeding of passage 3-4 cells at high dilution (10-100/cm2) resulted in less than detectable fibroblast levels I < 0.191 within another 2-3 passages. The efflcacy of removal of fibroblasts by low-temperature cultivation comblned Wlth trypsin/EDTA warhlng was mOnitOIed as followm. Trypsinized single c e l l ~~~p e n s l o n s were Seeded in 10 cm dishes at a density of 10-20 cells per an2, incubated at 37OC for 4-0 days and then stained vlth coomabsie blue.
The fibroblastic and eplthelial lgrossl colonial morphologles were strikingly different and permitted immediate identificaCion of both c e l l types. As many as 1000 colonies per dish could be screened by visual inspection. Uncloned kerarinocyte cultures referred to in the following were free from fibroblasts at the 0.18 level or better.
doobllng times of 1 4 -1 8 h was achievedvith "Supplemented with 20% fetal serial propagation of both cloned and uncloned cells with population calf eerum wlrhour fvrther addition of growth factors or vltamlns.
The cultures were examined for mycoplasma infection at regular Intervals using the fluorescent DNA stain, HoeChSt 133258, and found to be negative. The cells did not propagate in the absence of serum but remained adherent and metabolically active under these conditions for 10-15 days. Repeated attempts to develop tumors in nude mice by injection of aliquats of trypsinized cells unsuccessful and suqqest that development of feeder cell independence was (lo7 c e l l s per sitel at Intraperitoneal and eubcutaneous sites were .