Matrix metalloproteinase gelatinase B (MMP-9) coordinates and effects epithelial regeneration.

We studied the role of the matrix metalloproteinase gelatinase B (gelB; MMP-9) in epithelial regeneration using the gelB-deficient mouse. We report the novel finding that, in contrast to other MMPs expressed at the front of the advancing epithelial sheet in wounds of cornea, skin, or trachea, gelB acts to inhibit the rate of wound closure. We determined this to be due to control of cell replication, a novel capacity for MMPs not previously described. We also found that gelB delays the inflammatory response. Acceleration of these processes in gelB-deficient mice is correlated with a delay in signal transduction through Smad2, a transcription factor that inhibits cell proliferation, and in accumulation of epithelial-associated interleukin-1alpha, a cytokine that inhibits Smad2 signaling and promotes the inflammatory response. GelB-deficient mice also reveal defects in remodeling of extracellular matrix at the epithelial basement membrane zone, in particular, failure to effectively remove the fibrin(ogen) provisional matrix. We conclude that gelB coordinates and effects multiple events involved in the process of epithelial regeneration.

Epithelial tissues surface all organs of the body exposed to the external environment, acting as a protective barrier. Unlike most organ stromal tissues, epithelia can regenerate completely following injury. Epithelial regeneration involves multiple distinct processes, including the concerted migration of cells as a sheet to resurface the wound bed, mitosis to replace cell numbers, re-stratification of the sheet into a multilayered structure, and restoration of stable adhesive interaction with the underlying tissue stroma (1)(2)(3)(4)(5)(6). These processes must be temporally coordinated with one another, and with the protective inflammatory response, which also occurs in response to injury. This is thought to involve an orderly progression and synergism of cytokines and growth factors within the wound environment (7)(8)(9)(10)(11)(12)(13)(14)(15). Despite many recent advances, mechanisms for integrating and orchestrating the multiple processes involved in epithelial regeneration are still poorly understood.
The matrix metalloproteinases (MMPs) 1 are a family of zinc endopeptidases that act as key effectors and regulators of tissue remodeling in vertebrates (16). Molecular substrates for the MMPs include all classes of extracellular matrix proteins and molecules that organize tissues such as the cadherins (17,18). MMPs are also reactive against signaling molecules such as cytokines and growth factors, controlling their activity and bioavailability (18). While MMP activity is regulated at multiple levels, gene expression constitutes the major control mechanism (19). Induced expression of an array of MMPs occurs as part of the tissue response to injury (20,21). Inappropriate expression or overexpression of MMPs contributes to the pathophysiology of diverse disorders occurring across all organ systems (22), including healing disorders (20,(23)(24)(25).
Gelatinase B (gelB; MMP-9) is an MMP that catalyzes cleavage of denatured collagens of all types and native basement membrane components (26,16). In addition, recent work has identified fibrin(ogen) (27), ␣ 1 -proteinase inhibitor (79), interleukin-1 (IL-1) (28,29), and transforming growth factor-␤ (TGF-␤) as gelB substrates (30). GelB is not expressed in the normal cornea, skin, or trachea; however, expression is induced (along with a number of other MMPs) in cells at the front of the migrating epithelial sheet as it begins to resurface the wound bed following injury (22,(31)(32)(33). Inflammatory cells infiltrating the wound bed also produce gelB (34,35). GelB expression in the epithelium spreads progressively distal to the migrating front once wound resurfacing is complete and persists for several weeks thereafter (32,36). The timing correlates with the period during which provisional wound matrix is resorbed and new structures for epithelial/stromal adhesion are assembled in the epithelial basement membrane zone (37). Overexpres-sion of gelB is associated with defective re-epithelialization and a reduction in adhesion complex integrity (23,31). These data suggest that gelB plays a key role in the process of epithelial regeneration.
The corneal epithelium can be considered a prototype for a stratified epithelium similar to that surfacing skin, lung, and other organs. Moreover, the cornea offers an ideal model for evaluating mechanisms of tissue repair and regeneration because the avascularity of this organ reduces the number of interacting tissues. In this study we have investigated specific roles for gelB in regeneration of the corneal epithelium, utilizing a strain of mice made genetically deficient for gelB.

EXPERIMENTAL PROCEDURES
Mice and Surgical Procedures-GelB-deficient mice (GelB Ϫ/Ϫ ) were on the 129SvEv/CD-1 mixed background. A matched littermate control line (GelB ϩ/ϩ ) was on the same background (38). At intervals throughout this study, genotypes were verified by Southern blotting. Mice were used for experiments between 6 and 12 weeks of age; gelB-deficient mice were age-matched with littermate controls. Surgical procedures were performed on anesthetized mice according to approved institutional protocols. For corneal debridement surgery, the epithelium within an area demarcated by a 1.5-mm diameter trephine was removed to the basement membrane with the aid of an algebrush (Storz Ophthalmic). Photorefractive keratectomy (PRK) was performed on corneas using a Summit Apex Excimer Laser (Summit, Waltham, MA) on phototherapeutic settings. Tissue was ablated to a depth of 40 -44 m within a 2.0-mm diameter spot size using 160 pulses. To create a full thickness skin wound, hair was removed with a depilitory agent. Then the back skin was pulled up to form skin-to-skin double layers and a 6-mm punch biopsy sterilely used to create two identical full thickness wounds. The wounds were covered with transparent Op-Site dressing.
Quantification of Re-epithelialization Rate-Eyes were enucleated between 10 and 20 h after surgery and tethered to paraffin by pinning through the extraocular muscles. The remaining epithelial defects were delineated by application of Richardson's stain to the ocular surface. Images were projected at a fixed distance and the circumference of the area stained by blue dye was traced in pencil. The area was then quantified using NIH image software by an observer unaware of the mouse genotype.
To quantify healing in skin wounds, animals were anesthetized and then the outside perimeter was traced with a fine point marker onto a sterile slide directly applied to the wound site, after removing the old dressing and before applying a new one. The wound shape was traced immediately after surgery and every 2 days from the third day afterward. The image of the slide with the wound tracing was captured through a computerized video camera with Pax-it! Software (version 1.83, Imaging Systems for Science and Industry). The area of wound tracing on slide was quantified using computer image analysis software (Optimus).
Organ Culture-Eyes were enucleated immediately following surgery, tethered to paraffin beds in the individual wells of 24-well plates, and cultured in serum-free medium as described (32). For "rescue" experiments, purified recombinant human gelB proenzyme (Chemicon, Temecula, CA) was added to culture medium at the beginning of an experiment. In other experiments, organ cultures were treated with Ilomastat, a synthetic peptide hydroxamic acid analogue of the transition state of zinc metalloproteinases (39). A peptide analogue lacking the active hydroxamic acid group was used as a control (both agents from AMS Scientific, Pleasant Hill, CA).
Zymography-Epithelium from the repairing areas of corneas was debrided with a surgical scalpel, placed in 200 l of ice-cold 200 Tris-HCl, pH 7.5, and pulverized with a Pellet Pestle (Kimble-Knotes, NJ). Insoluble material was removed by centrifugation at 14,000 ϫ g for 10 min at 4°C. The soluble proteins were precipitated using cold acetone and protein concentration was determined with the Bio-Rad reagent. Gelatin zymography was performed according to our standard methods (31,32).
Light Microscopy-Methacrylate cross-sections (6 m) were stained with hematoxylin and eosin. Relative thickness of the corneal epithelial and stromal layers in gelB-deficient mice and normal mice was compared by examining 10 sections taken through the center of each cornea from 10 different mice of each strain. Measurements were made using a micrometer viewed in the light microscope field alongside the tissue section. Significance of differences between strains was assessed using the Student's t test.
Confocal microscopy was performed on mouse corneas fixed with 1% paraformaldehyde in phosphate-buffered saline. Whole corneas were then stained en bloc with rhodamine-conjugated phalloidin (Molecular Probes, Eugene, OR) to identify actin filaments. Corneas were mounted epithelial side down onto 25-m thick, 60-mm diameter mylar Petri dishes (Bachofer, Hamburg, Germany) and scanned using a confocal microscope (Confocal Laser Scanning Microscopy, Leica, Deerfield, IL) equipped with a Leica Fluorovert microscope, argon/krypton laser and 568/590 excitation/detection filters. Three-dimensional data sets were collected and transferred to a Silicon Graphics work station (Personal Iris 4D-35G), and maximum intensity volume renderings were made using the ANALYZE software program (Mayo Medical Adventure Inc., Rochester, MN).
BrdU Labeling and Detection-Central corneal debridement (1.5mm) wounds were made in the right eyes of the mice and allowed to heal for 16 h in vivo. Two hours before the mice were sacrificed, BrdU (Sigma) dissolved in phosphate-buffered saline was injected subcutaneously in the dorsal subscapular region at 100 g/g body weight. Mice were sacrificed and the eyes removed immediately and frozen for cryostat sectioning. Cross-sections were immunohistochemically stained with a monoclonal antibody against BrdU (Becton Dickinson, San Jose, CA). The primary antibody was detected by avidin-biotin complex immunoperoxidase technique using an ABC Elite Kit (Vector Laboratories, Burlingame, CA) following the manufacturer's directions. Color was developed with 3Ј,3Ј-diaminobenzidine peroxidase substrate (Sigma).
Western Blotting-Eyes were frozen immediately after the mice were sacrificed by immersion in liquid nitrogen to prevent Smad activation during sample preparations. The entire corneal epithelium was then removed by gentle scraping with a scalpel. The pooled tissues from 10 eyes were lysed in a buffer containing 50 mM Tris-HCl, pH 7.4, 1% Nonidet P-40, 0.25% sodium deoxycholate, 150 mM NaCl, 1 mM EGTA, 1 mM phenylmethylsulfonyl fluoride, 1 g/ml each aprotinin, leupeptin, pepstatin, 1 mM Na 3 VO 4 , 10 mM NaF. Equal amounts of protein (10 g/sample) were separated by 10% SDS-PAGE, transferred onto polyvinylidene difluoride membranes and probed with specific primary antibody reagents. Immunodetection of primary antibody binding was accomplished by incubation with horseradish peroxidase-conjugated secondary antibody followed by chemiluminescence (PerkinElmer Life Sciences) and exposure to x-ray film. Rabbit anti-Smad3 and antiphospho-Smad2 were purchased from Upstate Biotechnology (Lake Placid, NY), and monoclonal anti-Smad2 (clone 18) was purchased from Transduction Laboratories (San Diego, CA). A goat polyclonal antibody raised against recombinant mouse IL-1␣ (catalog number AF-400-NA) was purchased from R & D Systems (Minneapolis, MN). Information in the company catalogue states that this antibody was selected for its ability to neutralize the biological activity of recombinant mouse IL-1␣. Based on direct enzyme-linked immunosorbent assay and Western blot results, this antibody showed less than 10% cross-reactivity with recombinant human IL-1␣. Additionally, indirect enzyme-linked immunosorbent assays this antibody showed no cross-reactivity with other cytokines tested.

GelB-deficient Mice Have Normal
Corneas-Mice homozygous for a targeted mutation that inactivates the gene for gelB show delayed growth of long bones due to impaired vascularization of the growth plate, but ultimately they attain normal adulthood (38). Routine gross examination of adult eyes revealed no obvious anatomical differences between gelB-deficient homozygotes and their normal counterparts. Eyes of the gelB-deficient mice were of normal size and lenses were clear. Corneas were clear and avascular, with no evidence of malformation or inflammation by gross or histological analysis (data not shown). No statistically significant difference was observed between strains in stromal or epithelial layer thickness, or in number of epithelial layers (data not shown). These observa-tions indicate that a deficiency of gelB does not affect the formation and maintenance of normal corneal structure.
Increased Rate of Corneal Re-epithelialization in GelB-deficient Mice-To investigate the effects of gelB deficiency on regeneration of the corneal epithelium following injury, we used an epithelial debridement model (42,32) adapted for the mouse in which the epithelium is surgically removed to the level of the basement membrane within a circular region of 1.5-mm diameter. We collected epithelial tissue from the repairing portions of normal and gelB-deficient corneas at a time when wounds were partially closed, and analyzed tissue extracts by gelatin zymography (Fig. 1A). A gelatinase of a size appropriate to be the gelB proenzyme was detectable in extracts derived from littermate control mice, but was absent from extracts derived from gelB-deficient mice. In contrast, a gelatinase of the size appropriate to be the gelatinase A (gelA; MMP-2) proenzyme was detectable in extracts from both gelBdeficient mice and their normal counterparts. The intensity of the gelA band was essentially the same in both mouse lines, indicating that gelB deficiency does not result in compensatory up-regulation of this related enzyme.
A time course experiment was performed over the period of wound closure, and the rate of re-epithelialization was compared between gelB-deficient mice and normal littermate controls. Fig. 1B shows representative images of the remaining epithelial defects at a time when wounds were partially closed. No difference was observed in the quality of re-epithelialization between mouse strains. Corneas from both gelB KO mice and their normal counterparts re-epithelialized without incident, with no gross evidence of inflammation. In contrast, the rate of re-epithelialization was faster in the gelB-deficient mice. This difference was apparent as early as 10 h after surgery, and was highly significant by 20 h (Fig. 1C). All corneas from both mouse strains remained avascular throughout this experiment, indicating that cytokines or other factors contributed by the vasculature could not cause the differing re-epithelialization rates.
The substrate on which the epithelium migrates could influence the rate of wound resurfacing. Therefore, we repeated the experiment described above using a PRK model in which a laser is used to ablate superficial corneal tissue removing the epithelium, the basement membrane, and a portion of the anterior stroma. To resurface this type of wound, the epithelial sheet migrates across stromal collagen fibrils. Again, no difference in the quality of healing was observed between gelBdeficient mice and their normal counterparts; however, reepithelialization was still significantly faster in the gelBdeficient mice (Fig. 1D).
To learn whether faster wound closure generalizes to other organs, we examined the rate of healing of an excisional wound in the skin (Fig. 1E). As in cornea, wound closure in skin was found to occur significantly more rapidly in gelB-deficient mice than in their normal control counterparts. These results indicate that the findings in cornea are not tissue-specific.
The GelB-deficient Re-epithelialization Phenotype Is Due to a Direct Loss-of-function in the Eye and Is Different from the Phenotype Resulting from Broad Spectrum MMP Deficiency-To differentiate between systemic and local factors in the gelB-deficient phenotype, we repeated the surgical debridement experiment, but eyes were enucleated following surgery and placed in organ culture for healing. The difference in the re-epithelialization rate between normal and gelB-deficient mice was less than when re-epithelialization occurred in the mouse in situ, but was still significant at 15 h (p Ͻ 0.002; n ϭ 10; data not shown) and 20 h after surgery ( Fig. 2A). These data indicate that at least some of the effects of gelB deficiency are attributable to local factors.
To provide evidence that the corneal re-epithelialization phenotype in the gelB-deficient mice is due to a direct loss-offunction, we attempted a rescue experiment by adding purified gelB protein to the corneas of gelB-deficient mice undergoing re-epithelialization in organ culture. GelB was added in the latent proenzyme form, because this is the form that predominates in re-epithelializing debridement wounds (see Fig. 1A). Our expectation was that the exogenously added proenzyme would be converted to an active form through the same extracellular mechanisms utilized for activation of endogenously produced enzyme during wound repair. At 1 g/ml of exogenously added gelB proenzyme, the retardation of re-epithelialization rate was significant (Fig. 2B). In other words, exogenous addition of gelB returned the rate of wound reepithelialization in the gelB-deficient mouse back toward normal, consistent with rescue of the deficiency phenotype. These data support the conclusion that the gelB-deficient reepithelialization phenotype results from a direct loss-of-function due to inactivation of the gene for gelB.
Inhibition of broad-spectrum or specific MMP activity has been reported to inhibit epithelial cell migration in culture (21,33,43) and to retard re-epithelialization of skin wounds in organ culture and in vivo (21,44). To learn whether this is also true in our corneal model, we examined the effects of Ilomastat, a broad spectrum MMP inhibitor (39, 45) on re-epithelialization in normal mice. While the K i for gelatinases determined in test tube assay is 0.4 nM, inhibition of MMPs in tissues requires much higher concentrations; we used the dose range determined to be effective by Chin and Werb (46) in their studies with a mandibular organ culture system. At an ilomastat dose of 1 M, the rate of corneal re-epithelialization was significantly retarded in comparison to controls treated with the same dose of an inactive analogue lacking the hydroxamic acid group (Fig.  2C). These data are consistent with a requirement for MMP activity in corneal re-epithelialization, and suggest that gelB must be playing a different role than other MMPs in this process.
Earlier Inflammatory Cell Infiltration, Increased Deposition of Provisional Matrix, and Enhanced Rate of Epithelial Cell Replacement in GelB-deficient Mice-Tangential optical sectioning by confocal microscopy on corneal whole mounts stained with rhodamine-phalloidin revealed no differences between gelB-deficient mice and their normal counterparts in the overall appearance, or in organization of the actin filaments, in cells at the advancing epithelial front (Fig. 3, Confocal, arrows). However, in the course of these experiments we observed the presence of inflammatory cells in the wound bed of gelBdeficient mice (Fig. 3, Confocal, Ϫ/Ϫ, arrowhead), that were not seen in their normal counterparts (Fig. 3, Confocal, ϩ/ϩ, arrowhead). Absence of gelB does not appear to alter the capacity of inflammatory cells to respond to a chemotactic stimulus, extravasate from blood vessels, or accumulate in tissues (47). However, a deficiency in gelB was shown to delay the resolution of the contact hypersensitivity response in skin (48). A systematic investigation of inflammatory cell infiltration by immunofluorescent microscopy on cross-sections revealed consistent staining with the Mac-1 marker in corneas from gelBdeficient mice examined prior to epithelial closure (Fig. 3, Immunofluorescence, Mac-1, Ϫ/Ϫ, arrow), while no staining was observed in normal mice (Fig. 3, Immunofluorescence, Mac-1, ϩ/ϩ, arrow). However, shortly after wound closure (24 h postsurgery), corneas from all normal mice stained for the Mac-1 marker, and the amount of staining appeared only slightly less than in the gelB-deficient mice (data not shown).
In histologically stained cross-sections, the migrating epithelial sheet appeared as a monolayer in normal mice but was multilayered in the gelB-deficient mice (Fig. 3, H & E, arrows). In addition, an eosinophilic deposit was apparent in the wound bed and below the migrating epithelium in some gelB-deficient mice that was not observed in their normal counterparts (Fig.  3, H & E, arrowheads). This seemed likely to represent an accumulation of provisional wound matrix and is discussed further below. Bromodeoxyuridine (BrdU) labeling was per- FIG. 2. The gelB-deficient re-epithelialization phenotype is due to a direct loss-of-function in the eye, and is different from the phenotype resulting from broad-spectrum MMP deficiency. Epithelial debridement surgery was performed on groups of normal wild-type (WT) and gelB-deficient knock-out (KO) mice and then the animals were sacrificed and the eyes were removed to organ culture for re-epithelialization in vitro. The graphs represent areas of remaining epithelial defects at the indicated time point after surgery. Statistical significance was determined using the Student's t test. A, organ-cultured corneas at the 20-h time point; p Ͻ 0.02; n ϭ 8 for each group. B, organ-cultured corneas from gelB-deficient mice were treated with recombinant human pro-gelB for 18 h. p Ͻ 0.003. n ϭ 6 for each group. C, organ-cultured corneas from gelB-deficient mice were treated with 1 M ilomastat (Ilom) or control peptide (CP) for 18 h. p Ͻ 0.017. n ϭ 6 for each group.

FIG. 3. Earlier inflammatory cell infiltration, increased deposition of provisional matrix, and enhanced rate of epithelial cell replacement in gelB-deficient mice.
Epithelial debridement surgery was performed on the corneas of normal (ϩ/ϩ) and gelB-deficient (Ϫ/Ϫ) mice, and corneas were collected for analysis just prior to wound closure (16 -18 h). Confocal, tangential optical sections by confocal microscopy through the epithelium of rhodamine-phalloidin-stained corneal whole mounts, taken 18 h after surgery. formed on corneal cross-sections to identify cells undergoing DNA synthesis (Fig. 3, BrdU). It is known that epithelial injury stimulates proliferation of cells distal to the wound edge, while the rate of proliferation decreases in cells at the migratory front (49,50). Consistent with this, BrdU-labeled cells from corneas of both normal and gelB-deficient mice were concentrated in the peripheral epithelium, with fewer labeled cells localized to the migratory front (central). In both locations, however, the number of labeled cells was clearly greater in the gelB-deficient mice. These data provide evidence that gelB deficiency enhances the rate of epithelial resurfacing of corneal wounds by increasing the pressure exerted as a result of cell proliferation.
GelB Deficiency Is Associated with a Delay in Smad2 Signaling and an Increase in Cell-associated IL-1␣ in the Regenerating Corneal Epithelium-GelB can cleave a variety of molecules involved in cell signaling (18), and thus gelB deficiency might alter the net signaling information received at the cell surface and cause the changes observed above. A deficiency in the transcription factor Smad3 was recently shown to enhance the rate of cutaneous wound re-epithelialization by increasing cell proliferation (51), similar to our findings here on the gelBdeficient phenotype. Both Smad3 and its closely related homologue, Smad2, translocate from the cytoplasm to the nucleus once phosphorylated in response to signals received at the cell surface (52). Immunohistochemical staining of cross-sections through corneas of normal mice revealed that Smad3 undergoes translocation into nuclei throughout the regenerating corneal epithelium following debridement surgery (Fig. 4A). Western blotting revealed that Smad2 and Smad3 activity was clearly increased in the regenerating epithelium of normal mice (Fig. 4B, ϩ/ϩ). For Smad2, increased activity was demonstrated by comparison of the levels of total Smad2 and phosphorylated Smad2. For Smad3, reprobing of the same Western blot revealed a new immunoreactive band of slower electrophoretic mobility in regenerating epithelium, consistent with phosphorylation. These findings are consistent with a role for Smad2 and Smad3 in controlling the rate of epithelial cell proliferation.
Western blot analysis of the regenerating epithelium from gelB-deficient mice showed the same overall changes in Smad2 and Smad3 as the normal mice (Fig. 4B, Ϫ/Ϫ). The amount of the putative active form of Smad3 was similar in the gelBdeficient mice and their normal counterparts. However, activation of Smad2 was clearly delayed in the gelB-deficient mice; an increase was not observed until the 16-h time point, and this increase was less than that seen in normal mice.
Since the cornea is avascular, the corneal wound is devoid of platelets, a major source of TGF-␤ and other cytokine mediators of wound repair. However, the corneal epithelium synthesizes TGF-␤ , IL-1, and other cytokines (53)(54)(55) and is thought to substitute for platelets in controlling the repair process (56). Western blot analysis of cell extracts from regenerating epithelium was performed to compare the levels of cell-associated TGF-␤ and IL-1 at the 8-h time point (Fig. 4C). No difference in the levels of the major TGF-␤ isoform produced by corneal epithelium (54) could be detected between gelB-deficient mice and their normal counterparts (data not shown). However, the amount of the major corneal IL-1 form, IL-1␣ (53), was considerably increased in the gelB-deficient mice (Fig. 4C). Immunofluorescent localization analysis confirmed this increase, with greater IL-1␣ observed within cells throughout the migrating epithelium (data not shown). Since IL-1␣ is an antagonist of TGF-␤ signaling (76,77), these data are consistent with the reduction in Smad-2 signaling in gelB-deficient mice. Since IL-1␣ is an inflammatory cytokine, itself chemotactic for leu-kocytes (82), the increased levels of IL-1␣ are also consistent with the earlier onset of inflammatory cell infiltration.

Impaired Basement Membrane Zone Remodeling and Accumulation of Provisional Matrix After Injury in GelB-deficient
Mice Compromises Corneal Transparency-In normal skin or corneal wounds, a provisional extracellular matrix composed of fibrin(ogen) and fibronectin is deposited in the wound bed from the serum or tear film, and provides a substrate for cell attachment and migration (57)(58)(59)(60)(61)(62). This matrix is resorbed once the wound is resurfaced and is replaced by new epithelial/stromal anchoring complexes (37,64). The eosinophilic deposits apparent in the wound bed of gelB-deficient mice (see Fig. 3, H & E), suggested that remodeling in the basement membrane zone might be defective. We investigated this hypothesis by immunofluorescence analysis (Fig. 5A). In the unwounded areas of both gelB-deficient and normal corneas that had been healing for 6 days following PRK, the anchoring fibril component laminin 5 was present as a thin linear band between the epithelium and the stroma, and there was no staining for fibronectin or fibrinogen. This staining pattern had returned in the wound bed of the normal control mice, consistent with the timing of basement membrane zone remodeling observed in previous studies (32,60). In contrast, the immunoreactive band of laminin-5 was thick and uneven in the wound bed of gelB-deficient corneas at the 6-day time point, and small amounts of fibronec- tin were present. Most strikingly, a large fibrinogen deposit was apparent in the wound bed of the gelB-deficient mice. These results indicate impaired resorption of provisional matrix and re-establishment of epithelial/stromal adhesion complexes in gelB-deficient mice.
Defective remodeling at the epithelial basement membrane zone would have undesirable consequences for long-term health of the epithelium and could compromise organ function. In cornea, defective remodeling might interfere with the transparency necessary for transmittal of light to the retina of the eye. To investigate this hypothesis, we followed wound healing after PRK over a 2-week time course experiment. Representative results are shown in Fig. 5B, and data are summarized in Table I. Two days post-PRK, re-epithelialization was complete, and the corneas of normal mice were completely transparent. A minimal amount of clouding or "haze" was observed in some of the gelB-deficient mice at this time point, at a level too low to be recorded by photography. By 6 days post-PRK, some of the corneas from normal mice exhibited well defined, diffuse haze. However, many of the gelB-deficient mice exhibited haze, of a degree to obstruct iris detail. By day 14, haze was mostly resolved in the normal mice, but overall haze was increased in the gelB-deficient mice. This clouding is very similar to that reported after corneal wound healing in the plasminogen-deficient mouse (41). At no point was there any evidence of neovascularization of the avascular corneas at any time point in any experimental animal; thus the excessive material deposited could not have been due to vascular differences. These results suggest that defective resorption of provisional matrix and remodeling at the basement membrane zone due to the absence of gelB causes corneal clouding. DISCUSSION MMP expression is induced in cells at the migrating epithelial front in healing wounds of cornea, skin, or lung, and the combined evidence to date (including data presented here) has supported the hypothesis that MMPs function to promote epithelial resurfacing by stimulating cell migration (22,43,44). Moreover, there is evidence to support this role specifically for gelB (33,65). MMPs, including gelB, have further been implicated in a larger way in the process of cell migration involving many other tissues. Again, the roles identified for MMPs have been consistently facilitative and include the clearing of extracellular matrix to break down physical barriers, modulation of adhesive interactions with the extracellular matrix to release cells and to provide traction for their movement, and exposure of signals necessary to effect motor function and provide chemoattraction (18). The results reported here reveal that, while gelB may be involved in epithelial sheet migration, it is clearly not essential for this process in the normal in vivo situation. In fact, we show that a deficiency in gelB actually accelerates the rate of normal wound resurfacing. This is the first time that an MMP has been shown to exert negative control over cell migration. We show that gelB does this by inhibiting cell replication in the migrating epithelial sheet. This is a novel role for an MMP, not previously identified.
Peripheral to the main purpose of the current study, we made the new and significant observation that Smad2/Smad3 signaling is activated in the regenerating corneal epithelium. Smad signaling inhibits cell replication (52,66), and thus Smad activation seems at first counterproductive to the requirement for cell replacement in epithelial regeneration. This can be understood, however, when we take into account that epithelial regeneration involves several other distinct processes, including resurfacing of the wound bed, re-stratification into a multilayered structure, and restoration of stable adhesive interaction with the underlying stroma. Each of these processes is associated with a withdrawal from the cell cycle (49,50), and it is here that Smad activation may play a role.
Significantly, we found a specific delay in activation of Smad2 but not Smad3 in the gelB-deficient mice. Smad2 and Smad3 are the major downstream effectors of TGF-␤ signaling; however, there is accumulating evidence to suggest that these proteins are functionally different (52,66). While activity of both are stimulated by TGF-␤ signaling, they can be differentially regulated as a result of cross-talk among signaling pathways activated by other extracellular ligands. The IL-1 signaling pathway interacts with the Smad signaling pathway in this way; therefore the premature increase in IL-1␣ levels observed in this study may translate into a specific delay in Smad-2 signaling (76 -78, 80 -81). GelB is selectively active against the IL-1␤ isoform, although a minor capacity to cleave IL-1␣ could still translate into a major effect in a specific microenvironment in situ. However, GelB could alter the accumulation of IL-1␣ by FIG. 5. Impaired basement membrane zone remodeling and accumulation of provisional matrix in gelB-deficient mice compromises corneal transparency. A, cross-sections from corneas taken from normal (ϩ/ϩ) and gelB-deficient mice (Ϫ/Ϫ) 6 days after PRK. Indirect immunofluorecence microscopy was used to identify laminin-5, fibronectin, and fibrin(ogen). The arrow indicates immunoreactive fibrin(ogen) which accumulates in eyes from gelB-deficient mice. B, eyes of normal mice (ϩ/ϩ) and gelB-deficient mice (Ϫ/Ϫ) were photographed 2 or 6 days after PRK. Arrows indicate corneal opacities. Our findings also indicate that gelB controls the timing of the inflammatory response in the repairing cornea. Again, this may be attributed to premature accumulation of IL-1␣, which is a chemoattractant for leukocytes (63). In a similar vein, an alteration in cytokine profiles was identified as the mechanism for delayed inflammatory cell resolution in a cutaneous chemical hypersensitivity model applied to the gelB-deficient mice (48). The joint regulation of wound closure and inflammation by controlling the levels of a single cytokine, IL-1␣, could serve to coordinate the timing of these two processes.
Our work shows that gelB alters the "instructions" that cells receive from the microenvironment, thus mediating very specific alterations in intracellular signaling pathways, and fine tuning of the regenerative process. Not only must cell replication be temporal-coordinated with the other processes involved in epithelial regeneration, but also spatially coordinated. Thus, the progressive change in expression pattern of gelB during epithelial regeneration may be a key factor in its ability to perform a fine tuning function (31,32,36). Expression at the leading edge of the migrating corneal epithelium may serve to coordinate what is happening at the migrating front with the cell replication that occurs distal to the front. Later, when the wound is closed, expression across the entire regenerating epithelium may be important for the complete resorption of provisional matrix and restoration of normal epithelial/stromal adhesion.
The accumulation of fibrin(ogen) in the wound bed was a striking aspect of the gelB-deficient phenotype found in this study and indicates that gelB acts not only to coordinate but also to effect events involved in epithelial regeneration. Recent work has shown that pericellular fibrinolysis by migrating endothelial cells is MMP-mediated (67). As this article was in preparation, it was reported that fibrin(ogen) accumulation occurs in the gelB-deficient mouse in a kidney disease model and that gelB can cleave purified fibrinogen in a test tube assay (27). Thus gelB may act directly on the temporary fibrin(ogen) scaffolding. GelB may also effect fibrin(ogen) removal indirectly, for example, by its ability to proteolytically activate plasminogen activator, or by its ability to proteolytically inactivate inhibitors of the plasmin-plasminogen activator cascade, such as PAI-1 (26).
Excessive deposition of fibrin(ogen) due to the absence of gelB appeared to be a major reason for corneal clouding. Corneal clouding or haze is an undesirable side effect of the wound healing process after corrective PRK (68) and can be visually debilitating in 1% of human patients (69 -73). Excessive fibrin-(ogen) deposition was also associated with defective restoration of stable adhesive interactions with the corneal stroma. Again, this suggests parallels to human pathology. Chronic wounds of skin and cornea are typically characterized by a hyperplastic epithelium (20,23,31,74), which fails to migrate properly across the wound bed or form proper attachments to the underlying stroma. This leads to persistent or recurrent epithelial defects. While fibrin accumulating in acute wounds is removed within days, it persists in chronic cutaneous wounds (75,74). It has been suggested that excessive fibrin deposition may contribute to pathophysiology, by "trapping" of cytokines controlling epithelial cell dynamics, and by physically interfering with the restoration of stable epithelial/stromal adhesion.
Taking the findings of this study into consideration with the results of previous work, we conclude that a balance of gelB activity must be struck for health of epithelial tissues; too little activity or too much activity can both lead to pathology. However, it appears that timing and location of gelB expression in the microenvironment is also critical to the overall picture of health or pathology. Understanding the mechanisms for controlling the pattern of gelB expression will be a future challenge.