Stromal fibroblasts synthesize collagenase and stromelysin during long-term tissue remodeling

The morphology of a connective tissue can undergo dynamic alterations through progressive synthesis, degradation and resynthesis of its extracellular matrix (ECM) components. This process, called tissue remodeling, is an important mechanism contributing to the morphogenesis of repair tissue deposited after injury (Clark, 1988). Early repair tissue is composed of a haphazard arrangement of collagen fibrils and an abnormal complement of proteoglycan and collagen types. Remodeling transforms the structure and molecular composition of repair tissue. In skin, these changes contribute to an increase in tensile strength, a property which is important to the survival of the organism. However, despite such changes, the structure of repairing skin never returns to normal and its original function is never restored. In contrast, studies using a rabbit model have demonstrated that remodeling of corneal repair tissue can result in functional regeneration (Cintron and Kublin, 1977). The opaque repair tissue deposited in the cornea shortly after injury has a disorganized structure which is similar to the structure of repairing skin. Remodeling, over a period of months to years, gradually reorganizes corneal repair tissue so that its structure comes to approximate the uninjured cornea (Cintron et al., 1973; Cintron and Kublin, 1977; Cintron et al., 1978, 1981, 1988). Changes resulting from remodeling eventually restore the cornea to normal transparency at the site of injury. It is this unique regenerative aspect that makes the cornea particularly interesting for the study of mechanisms controlling remodeling.

The cornea is composed of an epithelium, a stroma and an endothelium. These tissues are arranged in simple layers that can be separated easily from one another. In addition, each of these tissues is remarkably homogeneous in cell type and extracellular matrix composition. These structural features have made possible a precise documentation of the cellular events that occur during corneal repair (Weimar, 1960). Within hours after injury, the corneal epithelium begins to resurface the damaged area by migrating as a sheet from the edge of the wound. Also at this time, the quiescent cells of the underlying stroma transform to a fibroblast phenotype (Weimar, 1957). By 30-36 hours after injury, these fibroblasts begin mitosis and migrate into the damaged area. It is thought that these fibroblasts deposit the ECM comprising early repair tissue (Ross et al., 1970). A few inflammatory cells, including macrophages and polymorphonuclear neutrophils (PMNs), also accumulate in the damaged area, but these cells disappear during the first few weeks after injury. Over time, the cells in repair tissue continue to synthesize matrix molecules and thus mediate the synthetic phase of the remodeling process (Cionni et al., 1986).

Less information is available concerning the mechanisms controlling the degradative phase of repair tissue remodeling. Culturing of tissue fragments on collagen gels has demonstrated that tissue isolated from skin (Grillo and Gross, 1967) or cornea in the early stages of repair (Brown and Weller, 1970) actively mediates collagenolysis. However, collagenolysis during the long-term remodeling of repair tissue has not been examined.

Documentation of the origin of those cells producing collagenase in early repair tissue is still not complete. The fibroblast at the edge of the granulation tissue was recently identified as the predominant collagenase-synthesizing cell in healing skin (Porras-Reyes et al., 1991). Fibroblasts have also been identified as collagenase-producing cells in a pathological skin model (Hembrey and Erlich, 1986). However, the origin of these fibroblasts, dermis or subcutaneous connective tissue, has not been established (Grillo and Gross, 1967). Since the origin of the fibroblasts found in repair tissue of the cornea is known, the corneal model could help to define the origin of collagenase-synthesizing cells found in early repair tissue.

Other important questions also remain unanswered. In vitro experiments have revealed that net collagenolytic activity is determined by many interacting factors, including the level of enzyme synthesis and the degree to which enzyme is processed to its active form. The contribution of these factors to collagenolytic activity during tissue repair has not been examined. To begin defining the mechanisms which regulate the process of remodeling in repair tissue, we have investigated collagenase biosynthesis in a corneal repair model. We have also analyzed biosynthesis of the related metalloproteinase, stromelysin, which is part of the proteolytic cascade that generates activated collagenase from its inactive form.

To make a full-thickness corneal injury, a penetrating keratectomy was performed with a 2.0 mm trephine on one eye of New Zealand White Rabbits (2.5 kg) as previously described (Matsubara et al., 1991). This procedure surgically removes all three layers of the central cornea. It is a simple procedure resulting in reproducible healing with minimal complications and allows easy recovery of repair tissue (Cintron et al., 1973). The contralateral eye was left unwounded to serve as a control. After injury, at times specified in the text, animals were killed by lethal injection of sodium pentobarbital, and 2.0 mm-diameter fragments of corneal epithelium or stroma were isolated with a trephine as described previously (Matsubara et al., 1991). Fragments were isolated from the repair (i.e. site of injury), repair-adjacent (i.e. 1.0 mm from the repair area) and repair-peripheral (i.e. 2.0 mm from the limbus) areas of the cornea. In some cases, epithelium and stroma were isolated as a single unit. Following collection, each tissue fragment was either (1)cultured immediately, (2) disrupted for preparation of primary cell cultures (described below) or (3) placed in an individual tube and frozen immediately in liquid nitrogen.

Tissue fragments were cultured in equal volumes (200 μl) of serum-free medium consisting of Minimal Essential Medium (MEM; Gibco, Grand Island, NY) supplemented with antibiotics/antimycotics (Gibco) in 96-well cluster dishes. Explant cultures were incubated at 37°C in a humidified 5% CO_2-air atmosphere. [³⁵S]methionine (Amersham, Arlington Heights, IL) was included in the medium at a final concentration of 80 μCi/ml. Explant-conditioned medium was collected for analysis after 48 h, centrifuged to remove debris, and frozen at −20°C for later use.

Equal-volume samples (10 μl) of explant-conditioned medium were diluted 2:5 in sample buffer, and proteins were reduced by the addition of β-mercaptoethanol. Electrophoresis of samples was performed on 8% SDS-polyacrylamide gels (Laemmli, 1970) as previously described (Fini and Girard, 1990). Gels were dried and autoradiographed without intensifying screens and over a range of times to ensure the linear response of the film. The relative amounts of individual proteins in each gel lane were quantified by volume densitometry (Molecular Dynamics, Sunnyvale, CA) and compared.

Conditioned media from like tissue explants were pooled, and equal volumes were analyzed by immunoprecipitation using standard techniques (Fini and Girard, 1990). Normal sheep serum or sheep antisera directed against either rabbit collagenase or rabbit stromelysin (gifts from Dr Constance Brinckerhoff, Dartmouth Medical School) were utilized as primary antibodies. Immunoprecipitated proteins were visualized by gel electrophoresis. The relative amounts of specific immunoprecipitated proteins synthesized by tissues derived from different areas of cornea were compared after quantitation by volume densitometry (Molecular Dynamics, Sunnyvale, CA).

Collagenolytic activity in conditioned medium from stromal explants was assayed using the radioactive collagen fibril film method (Johnson-Wint, 1980a). In some cases, medium was treated with 0.01% L-(tosylamino 2-phenyl) ethyl chloromethylketone-trypsin (Worthington Biochemical Corp., Freehold, NJ) for 7 min at 37°C to convert proenzyme to active enzyme. Trypsin was then inactivated by the addition of a five-fold excess of soybean trypsin inhibitor (Worthington, Biochemical Corp.). Activated and unactivated samples were divided into 200 μl samples and placed into wells of a 96-well cluster plates for collagenolytic assay.

Primary cultures of epithelial and stromal cells were prepared from equal-diameter fragments of combined epithelium/stroma, as previously described (Fini and Girard, 1990). To allow cells to attach and spread on the glass slide, released cells were placed in 8chamber, Tissue-Tek slides (VWR Scientific, Boston, MA) and incubated for 18 h in MEM containing antibiotics/antimycotics and 10% calf serum (Hyclone, Logan, UT).

Monensin (1 μM final concentration) was added to cultures for 4 h prior to processing for immunofluorescence. Cells were fixed in 10% sodium phosphate-buffered formalin (pH 7.0) and double, indirect immunofluorescence was performed using standard methods (Harlow and Lane, 1988). The primary antibodies used were a sheep antiserum directed against rabbit collagenase, described above (used at 1:50 final dilution) and a murine monoclonal antibody directed against human vimentin (used at 1:25 final dilution; Sigma, St. Louis, MO). The secondary antibodies were fluorescein isothiocyanate (FITC)-conjugated, donkey antisheep IgG (used at 1:50 final dilution) and rhodamine isothiocyanate (RITC)conjugated, rabbit antimouse IgG (used at 1:50 final dilution). Both secondary antibodies were purchased from Jackson ImmunoResearch Laboratories (West Grove, PA). Cells were viewed and photographed using a Zeiss Axiophot (Atlantex and Zeiler Instrument Corp., Avon, MA) equipped for epiillumination.

Cells to be analyzed for in situ hybridization were fixed in 4% paraformaldehyde and processed further as described by Pardue (1985). A ³⁵S-labeled antisense RNA probe was synthesized by in vitro transcription (Melton et al., 1984) from a rabbit stromelysin cDNA template (Fini et al., 1987), which had been cloned into the vector, pBluescript (Stratagene, La Jolla, CA). The hybridization mixture contained ³⁵S-labeled probe (10⁷ cts per min/ml), 4×SSC (0.6 M NaCl, 0.06 M sodium citrate), 50% deionized formamide, 1 ×Denhardt’s solution and 20 mM dithiothreitol. Hybridization was carried out at 50°C overnight.

Percentages of fluorescent or radiolabeled cells were calculated by dividing the number of positive cells in at least 20, non-overlapping fields by the total number of cells in these fields.

Frozen corneal tissue fragments, prepared as described above, were thawed and immediately extracted with a solution of 2% sodium dodecyl sulfate (SDS) and analyzed by zymography (Heussen and Dowdle, 1980) as previously described (Matsubara et al., 1991). The entire extract prepared from each tissue sample was electrophoresed without reduction on 11% SDS-polyacrylamide gels containing beta-casein (Sigma, St. Louis, MO) at a concentration of 0.05%. After staining, the location of caseinases on the gel could be identified as clear bands, where the casein substrate had been digested, on a blue-stained casein background. Since proenzymes are activated by a conformational change produced by the SDS in the gel without an alteration in molecular size, both procaseinases and activated caseinases can be visualized using this technique. Molecular sizes of caseinases were determined by comparison with reduced size standards run on the same gel.

Isolated epithelium and stroma from control and surgicallyinjured corneas were cultured separately in the presence of [³⁵S]methionine to label newly synthesized proteins. Immunoprecipitation analyses were then performed (Fig. 1) on equal volumes of explant-conditioned medium samples. Collagenase antiserum specifically precipitated a 57 kDa and a 53 kDa protein, of the appropriate sizes to be the glycosylated and unmodified proenzyme forms of collagenase, respectively. Only explants from the stromal layer of repair tissue produced immunoprecipitable collagenase and stromelysin. These enzymes were not synthesized by the epithelial layer of repair tissue or by the stroma or epithelium of control corneas.

Stromelysin antiserum specifically precipitated a 51 kDa protein of the appropriate size to be prostromelysin, as well as a protein doublet migrating at a molecular mass of 40 kDa. This doublet co-electrophoresed with the autoproteolytically-activated form of stromelysin produced in vitro by treating enzyme with the organomercurial, 4aminophenylmercuric acetate (data not shown). Like collagenase, stromelysin was only synthesized by the stromal repair tissue.

To confirm that collagenase synthesized by repairing stromas is in its latent form, we evaluated the collagenolytic activity of conditioned medium from stromal explants harvested one week after injury. We were unable to detect collagenolytic activity in these samples. However after trypsin treatment, which removes the N-terminal peptide from procollagenase, we measured collagenolytic activity in two out of four samples. The level of activity was at the lower limit of the assay. Considering the quantity of procollagenase protein that we had observed by immunoprecipitation in 7day stromal wound samples, this level was much less than we had expected. This result probably indicates that enzyme inhibitors are also present in the medium and contribute to collagenase latency.

The temporal expression of collagenase in repair tissue deposited over nine months after injury was documented in a time course study. A more extensive spatial analysis of enzyme expression was performed in this set of experiments than in the one described above. For analysis of collagenase expression within one week after injury, equaldiameter fragments of tissue were isolated from corneas which had been undergoing repair for one, three, five or seven days. Fragments were harvested from repair, repair-adjacent and peripheral regions of injured corneas and from similar locations in control corneas. Newly-synthesized and secreted proteins from epithelial and stromal explants were analyzed by immunoprecipitation. Relative autoradiographic band densities of immunoprecipitated procollagenase in samples from injured corneas are summarized in Table 1.

As in the previous experiment, collagenase was produced by explants from the stromal layer of repairing corneas. Enzyme expression by explants was easily detectable as soon as one day after injury. In general, collagenase expression increased over the time course, reaching its peak at the seventh (last) day of analysis. Collagenase synthesis was not specifically localized to the repair tissue. A gradient of collagenase expression radiated from the area of repairing stroma. Enzyme expression decreased precipitously with increasing distance from the repair tissue. In fact, collagenase expression in the area most peripheral to the repairing area would have been missed if a long autoradiographic exposure had not been used. This explains why peripheral expression was not noticed in the previous experiment. However, even with the longest autoradiographic exposures, immunoprecipitable collagenase was not detectable in conditioned medium samples from epithelium at any location in repairing corneas, nor was enzyme synthesis detectable in control corneal explants.

In an expanded time course, collagenase expression was analyzed at one, four and eight weeks after injury. To reduce unnecessary manipulation of the corneal tissue, the epithelium was not separated from the stroma for this experiment. Collagenase synthesis was detected in explants from injured corneas over the 8-week time course (Table 2). As in the previous experiment, highest levels were measured in the area of repair; lowest levels in the peripheral cornea. The peak level of collagenase expression occurred at 4 weeks after injury. Again, no collagenase synthesis was detectable in conditioned medium from control tissues.

The final portion of this study determined whether collagenase is still synthesized and secreted during long-term remodeling of corneal repair tissue. Collagenase expression by epithelial/stromal explants from corneas injured 9 months earlier was analyzed. As shown in Fig. 2, collagenase synthesis was still detectable in samples from the repair tissue. However, enzyme synthesis could not be found in samples from either the repair-adjacent or peripheral cornea (data not shown).

Cells which produce collagenase and stromelysin in repaircells from repair-adjacent and peripheral stroma showed that a higher percentage of cells contained intracellular collagenase in the repair-adjacent (172 positive cells out of 216 cells; 80%) than in the peripheral stroma (51 positive cells out of 132 cells; 39%).

Immunoreactive collagenase was not detectable in cells isolated from the repair epithelium (Fig. 3E), nor in epithelia isolated from the repair-adjacent (Fig. 4A) or repairperipheral (Fig. 4C) regions of repairing corneas, nor from control corneas (Fig. 3A). Cells from control corneas (Fig. 3B) or the repair-adjacent and peripheral regions of repairing corneas displayed a vimentin-negative, epithelial phenotype. Two cell phenotypes were apparent in the epithelium isolated from the repair tissue. One (Fig. 3F) was vimentin-negative; the other (data not shown) was vimentin-positive. The significance of these two phenotypes is unknown, but neither phenotype contained collagenase enzyme.

In situ hybridization was used to determine whether stromelysin synthesis in repairing tissue might be controlled by accumulation of new mRNA (Fig. 5). Cultured epithelial and stromal cells, isolated from corneas seven days after injury, were hybridized with a ³⁵S-labeled, rabbiting corneas were characterized by immunofluorescence and in situ hybridization analyses. To increase the sensitivity of our detection methods, cells were first removed from their matrices and allowed to attach and spread on glass slides in tissue culture overnight. For immunofluorescence studies, cells were then treated with monensin to block N-linked glycosylation. This treatment causes collagenase and stromelysin, which are normally secreted immediately after synthesis, to accumulate in the Golgi apparatus. Following this treatment, cells were doublestained with collagenase and vimentin antiserum; vimentin is an intermediate filament which is a marker for corneal fibroblasts, but not for corneal epithelial cells (Risen et al., 1987).

Immunoreactive collagenase was localized by fluorescence microscopy to a specific perinuclear region, corresponding to the location of the Golgi apparatus, in cells isolated from the stromal layer of corneas which had been undergoing repair for one week (Fig. 3G). In contrast, immunoreactive collagenase was not detectable in stromal cells from control corneas (Fig. 3C). Stromal cells from all areas were vimentin-positive with a fibroblastic phenotype (Fig. 3D and H), whether or not they were collagenase-positive. One difference between cells from the repair tissue and those from the undamaged areas of repairing corneas was that collagenase-positive cells from the repair tissue were much larger (compare Fig. 3D and H).

An examination of the percentage of collagenase-positive cells in each region demonstrated that a large number of the cells isolated from the stromal layer of repair tissue contained intracellular collagenase. Although many cells isolated from stroma outside of the repairing area were also collagenase-positive, the percentage of these cells appeared to decrease with increasing distance from the repairing area. We were not able to obtain enough cells from the repair tissue for a quantitative analysis. However, such an analysis was possible for cells from areas of stroma outside of the repair tissue (Fig. 4B and D). A more thorough examination of randomly-selected, non-overlapping fields of stromelysin-specific RNA probe. Labeling above background was observed only over stromal cells from repairing corneas. Labeled cells were observed at a frequency of 64% (472 positive cells out of 740 cells) in the repairing stroma, 82% (508 positive cells out of 618 cells) in the repair-adjacent stroma, and 38% (215 positive cells out of 566 cells) in the peripheral stroma. Positive cells had a fibroblastic phenotype.

To ascertain whether our cell culture studies accurately reflect events occurring in vivo, we utilized the technique of zymography to visualize collagenase and stromelysin in extracts of tissue fragments harvested one week after injury and snap-frozen. Thawed samples were electrophoresed on SDS gels containing casein, since casein is a good substrate for stromelysin (Chin et al., 1985). As shown in Fig. 6, caseinolytic activity was found only in stromal samples from corneas undergoing repair. Two caseinases were observed in the repairing stroma. One had a molecular mass of 88 kDa. The other (at arrow) had an approximate molecular mass of 50 kDa. The size of the smaller caseinolytic enzyme is appropriate for the inactive, proenzyme form of either collagenase or stromelysin; this enzyme coelectrophoresed with prostromelysin produced by cultures of stromal fibroblasts (data not shown). No caseinolytic activity could be detected in the epithelium harvested from repairing corneas. These results are consistent with the timing and localization of stromelysin expression as determined from our explant and cell culture studies, and suggest that our data accurately reflect events occurring in vivo.

Following penetrating keratectomy, the repairing stroma, which consists of a haphazard meshwork of cells and fibrils, is considerably different in structure and composition from that of the normal stroma (Cintron et al., 1978). This lack of matrix order may contribute to the opacity and mechanical weakness of corneal repair tissue. With time, these deficiencies are corrected through a prolonged process of synthesis, degradation and resynthesis. The parallel layers of collagen lamellae, characteristic of the intact cornea, reform across the injured region. Electron microscopic studies reveal that, during the remodeling process, collagen fibril size becomes progressively more regular, and the stromal fibrils attain a more orderly arrangement (Cintron et al., 1978). It is believed that these gradual changes, which can take months to years, contribute to the eventual return of normal corneal transparency marking functional tissue regeneration.

It seems logical to assume that the degradative phase of tissue remodeling requires the participation of matrixdegrading enzymes, called matrix metalloproteinases (MMPs). Together, these enzymes, which are descended from a common ancestral gene (Brinckerhoff and Fini, 1989; Matrisian, 1990), have the capacity to degrade most components of the extracellular matrix. The MMPs are produced by both resident cells in a tissue and invading inflammatory cells and have greatest activity at the neutral pH of the extracellular space. Each MMP is secreted into the extracellular space as an inactive proenzyme which must be converted to an active form. Different MMPs have different substrate specificities. Fibroblast collagenase described in this paper can catalyze degradation of native types I, II or III collagen. In contrast, fibroblast stromelysin can specifically cleave proteoglycans as well as fibronectin and laminin. Stromelysin also appears to play another important role by converting procollagenase to its fully active form (Nagase et al., 1992).

Despite the large body of knowledge that has accumulated about the structure and biochemistry of MMPs in vitro, little is known about how the expression or activity of these enzymes is regulated during corneal remodeling. It has been shown that corneas undergoing wound repair for 1 to 2 weeks elaborate collagenolytic activity which is not produced by explants from uninjured corneas (Brown and Weller, 1970). However, it has not yet been established whether this activity arises from the activation of latent collagenase which was previously synthesized or is the result of new enzyme synthesis. Clarification of this subtle point has not been adequately addressed in the literature. In the present study, we demonstrate that collagenase is synthesized in repairing corneal tissue. In comparison, we found that normal corneas do not synthesize detectable collagenase. Remodeling corneas synthesize and secrete two proteins which can be immunoprecipitated with collagenase antiserum. These proteins, with molecular masses of 57 kDa and 53 kDa, are the appropriate sizes to be the inactive, proenzyme forms of glycosylated and unmodified fibroblast collagenase, respectively (Fini and Girard, 1990; Fini et al., 1987). To our knowledge, these data provide the first direct evidence for the long-held, but previously unproven, view that collagenase biosynthesis is a necessary prerequisite for collagen turnover during early repair process (Cionni et al., 1986; Davison and Galbavy, 1985, 1986). In addition, we demonstrate for the first time, that synthesis of collagenase continues in the repair tissue for as long as nine months after injury. This is the first evidence that this enzyme plays a role in the long-term process of tissue remodelling that leads to stromal regeneration.

In the present study, we also determined which corneal tissue layer produces collagenase following injury. This has been a subject of controversy for many years. When epithelium and stroma from recently injured corneas or skin are separated and cultured individually on collagen gels, epithelial, not stromal, explants lyse the gels (Grillo and Gross, 1967; Slansky et al., 1969; Gnadinger et al., 1969; Brown and Weller, 1970; Slansky and Dohlman, 1970; Pfister et al., 1971; Berman et al., 1971; Brown, 1971). Similar results were obtained in studies on healing skin (Grillo and Gross, 1967). The results of these explant studies have been difficult to reconcile with the findings of more recent cell culture experiments demonstrating that collagenases are generally produced by fibroblasts, PMNs or macrophages (Alexander and Werb, 1991), which are cell types found in the stromal layer of early repair tissue. In the present study, we provide data that might explain the discrepancy between cell culture and explant studies. We show that explants from the stromal layer of corneas undergoing repair do indeed synthesize and secrete immunoprecipitable collagenase. However, the collagenase is in its proenzyme form. Therefore, it would not have been detected by the gel lysis assay which was used in the previously reported explant studies because this assay requires active enzyme.

More difficult to resolve than the lack of collagenolytic activity in stroma is the localization of collagenolytic activity to the epithelium of repairing corneas and skin. These data suggest that epithelial cells also synthesize collagenase in repairing tissue. While epithelial cells are not generally recognized as cells which synthesize collagenase, a few reports have demonstrated some exceptions to the rule. For example, it has been reported that collagenase is synthesized by both skin epithelial cells (keratinocytes) after passage in culture (Petersen et al., 1987) and some epithelial tumor cell lines (Lyons et al., 1989). In these studies, collagenase synthesis could have been induced by cell culture or might reflect the abnormal genetic changes that occur with oncogenic transformation. However, it is also possible that epithelial cells begin to synthesize collagenase during tissue repair and remodeling. If this were the case, we should have been able to detect collagenase synthesis by epithelial cells; the collagenase synthesized by keratinocytes and epithelial tumor cells is the same gene product as the fibroblast collagenase and would have been immunoprecipitated by our antibodies. Since we were unable to detect enzyme synthesis either by immunoprecipitation or immunofluorescence, we must conclude that cells in the repairing corneal epithelium are not an important source of collagenase during tissue remodeling.

What, then, is the biochemical basis for the elaboration of collagenolytic activity by epithelial explants from repairing corneas or skin? A number of reports have demonstrated that, while epithelial cells do not generally produce collagenase enzyme, they can play a regulatory role in collagenolysis. Epithelial cells, including those of cornea and skin, can produce cytokines that positively or negatively regulate collagenase synthesis by fibroblasts (JohnsonWint, 1980b, 1988). In addition, these cells can produce proteinases that participate in converting procollagenase to the active form (He et al., 1989). Therefore, if any procollagenase-producing repair fibroblasts were cultured along with the corneal epithelium, the resulting conversion of collagenase to active collagenase would allow its detection by the gel lysis assay. Importantly, most studies reporting localization of collagenolytic activity to corneal epithelium made no attempt to isolate epithelium as a pure tissue; they simply separated corneas into epithelium/anterior stroma and stroma/endothelium. In only one case was greater care taken to separate the two tissues (Brown and Weller, 1970), but contamination with even a few stromal cells may have been sufficient to provide a source of collagenase. Thus, it seems likely that the basis of the reported epithelial localization of collagenolytic activity in these previous studies actually results from the capacity of the epithelium to regulate collagenolysis, rather than to synthesize collagenase enzyme.

In the present study, we found that expression of stromelysin was, like collagenase, induced de novo in the corneal stroma by injury. This has important implications for the mechanism of procollagenase activation. The conversion of latent collagenase to its active form is thought to occur via an enzymatic cascade which results in successive cleavages from the N terminus (Nagase et al., 1992). The initial cleavages yield collagenase that has only 1425% of full activity, and subsequent cleavage by stromelysin is necessary for conversion of collagenase to its fully active form. The major protein form that we identified had a molecular mass of 51 kDa and is the appropriate size to be the inactive, proenzyme form (Fini et al., 1987; Fini and Girard, 1990). Interestingly, a minor portion of the synthesized enzyme had a size of 40 kDa which is appropriate for a proteolytically activated form. This suggests that, while induction of collagenase and stromelysin synthesis in repair tissue is coordinate, activation of these enzymes is controlled by different factors.

In addition to its role in collagenase activation, stromelysin can also degrade the core protein of proteoglycans. In the cornea, two types of proteoglycans are commonly found: the cornea-specific molecule, keratan sulfate proteoglycan (lumican), and a molecule that is also found in sclera and skin, dermatan sulfate proteoglycan (decorin). Maintenance of an appropriate ratio (3:2) of these specific proteoglycan types is thought to be important for determining the precise degree of corneal hydration needed for tissue transparency (Bettelheim and Plessy, 1975; Hedbys, 1961). In repair tissue, this ratio is altered, with decorin becoming a more prominant component. It is likely that this change contributes to repair tissue opacity (Hassell et al., 1983). Activated stromelysin, via its capacity to catabolize proteoglycans, is likely to help reestablish transparency by returning the proteoglycan ratio to normal during remodeling.

We were able to detect newly synthesized collagenase and stromelysin in explant cultures from repair tissue within 24 h after injury. This is coincident with stromal cell migration into the injured area and the initiation of cell division (Weimar, 1960). Collagenase levels increased (1.4-fold) within the repair tissue between one and four weeks after injury; this increase is considerably smaller than the increase in cell number (4.5-fold) which we reported in an earlier investigation (Cintron and Kublin, 1977). Likewise, collagenase levels decreased (4.2-fold) between four and eight weeks after injury, but cell number decreased only 1.8-fold between four weeks and ten weeks. Thus, changes in the rate of collagenase production by repair tissue over the eight-week time course cannot be explained simply by changes in the number of cells in the repair tissue. This suggests that the rate of collagenase synthesis per cell is a secondary factor controlling the total amount of collagenase produced.

We were surprised to learn that expression of collagenase and stromelysin is not localized specifically to the cells in the repair tissue or to the cells at the wound edge from which wound fibroblasts are derived. Instead, our biosynthetic studies demonstrated that synthesis of collagenase and stromelysin occurred in a gradient across the radius of the corneal stroma, with highest levels in the repair tissue and lowest levels at the corneal periphery. Perhaps collagenase expression peripheral to the repair tissue is required to mediate intercalation of the new collagen fibrils into the existing stromal lamellae (Davison and Galbavy, 1986). In areas peripheral to the repair tissue, the gradient of enzyme expression could be explained, at least in part, by the number of cells engaged in enzyme synthesis across the area. In corneas repairing for one week, a higher percentage of cells containing intracellular collagenase were found in the repair-adjacent stroma (80%) than in the repairperipheral stroma (39%). However in the repair tissue, fewer of the total cells were synthesizing enzyme than in the repair-adjacent tissue. In addition, we previously determined that the total cell number in the repair tissue at the one week time point is less than half that in the undamaged areas (Cintron and Kublin, 1977). Yet despite these factors, the repair tissue produces considerably more enzyme (4-5 times) than the adjacent stroma, demonstrating that the rate of synthesis is also a controlling factor in enzyme production. In situ hybridization revealed that the pattern of mRNA expression paralleled that of enzyme synthesis; this demonstrates regulation of gene expression at the level of mRNA accumulation. The gradient of enzyme expression suggests that factors controlling enzyme expression radiate from the repair tissue. Inductive signals could be passed directly from one cell to another via gap junctions or could be transmitted by diffusible substances originating from the epithelial or stromal layers of the repair tissue.

In summary, we have shown that resident stromal fibroblasts synthesize and secrete the MMPs, collagenase and stromelysin, in the repairing corneal stroma. Our results help to eliminate confusion regarding the tissue source of collagenase in repairing cornea. To our knowledge, this may be the first study which establishes that continued collagenase biosynthesis by resident fibroblasts is a component of long-term tissue remodeling.

Sheep antisera directed against rabbit collagenase and stromelysin were generously donated by Dr Constance Brinckerhoff, Dartmouth Medical School, Hanover, NH. We thank Dr Romaine Bruns for advice and help with photography, and Mr Richard Silverstein for handling and care of the experimental animals. We are also grateful to Dr Jerome Gross for enlightening discussions. This work was supported by NIH grants EY08408 (M.E.F.) and EY01199 (C.C.), and by an agreement between Massachusetts General Hospital and the Shiseido Company.