MMP9 cleavage of the β4 integrin ectodomain leads to recurrent epithelial erosions in mice

Healing of the corneal epithelium after injury is rapid and takes place by sheet movement. Hemidesmosomes mediate attachment of the epithelial basal cells to their basement membrane and disassemble upon injury to facilitate the sliding of the sheet across the exposed basement membrane zone (BMZ) (Gipson, 1992; Gipson et al., 1993). Adherens junctions, desmosomes and tight junctions are all still present in the migrating corneal epithelial sheet but are less abundant towards the leading edge (Suzuki et al., 2000; Suzuki et al., 2003; Hutcheon et al., 2007). After migration is complete, cell–cell adhesions reform, first at the sealed wound margins, and later at the wound periphery along with the hemidesmosomes. Hemidesmosomes consist of two integral membrane components: the α6β4 integrin heterodimer and collagen XVII/BPA180 (Stepp et al., 1990; Nishizawa et al., 1993; Borradori and Sonnenberg, 1996). Genetically engineered mice lacking β4 integrin have no hemidesmosomes and lose their epithelium at birth (van der Neut et al., 1996; Dowling et al., 1996). Patients with mutations that reduce the expression or function of α6β4 integrin have subtypes of epidermolysis bullosa (EB) and develop skin blisters because of the reduced structural integrity of their hemidesmosomes (Uitto and Richards, 2004; de Pereda et al., 2009). Patients with mutations in collagen XVII develop a subtype of EB or a related disease called bullous pemphigoid (BP) (Zillikens and Giudice, 1999; Schumann et al., 2000).

Our lab has developed a mouse model to reproducibly induce recurrent epithelial erosions that spontaneously arise within 2 weeks after a single corneal debridement wound (Pal-Ghosh et al., 2004). These erosions occur but with different frequencies in both C57BL6 and BALB/c mice strains (Pal-Ghosh et al., 2008). The current study was initiated to determine the mechanism behind the development of erosions. Studies of patients with recurrent corneal erosions have shown elevated levels of various MMP family members (Garrana et al., 1999; Ramamurthi et al., 2006). Elevated levels of MMPs are also seen in dry eye disease (Chotikavanich et al., 2009). We hypothesized that at sites of erosion, activated MMPs could cleave basement membrane proteins, leading to internalization and degradation of α6β4 integrin, thereby destabilizing adhesion of epithelial cells to the underlying BMZ. Instead, we report that α6β4 integrin itself was targeted for cleavage by activated MMP9 during erosion formation.

The speed of corneal re-epithelialization after wounding is remarkable. At 18 hours after a small wound involving removal of ~35% of corneal epithelium, 90% of the wound area is covered. Large wounds involving removal of 85–90% of the corneal epithelium are over 60% closed within 18 hours (Fig. 1). This rapid migration rate requires an equally rapid rate of disassembly of the hemidesmosomes at the basal surface of the unwounded corneal epithelium. However, rapid re-epithelialization can result in detachment of the epithelial sheet from the underlying surface during or soon after re-epithelialization. This delays wound closure and increases the risk of infection. Unless hemidesmosomes reassemble after migration is complete, resolution of healing will be delayed. Using a mouse cornea debridement model, recurrent epithelial erosions can be induced in the cornea after a single debridement wound (Pal-Ghosh et al., 2004).

If the failure to reassemble hemidesmosomes were the primary cause of the recurrent erosions, we would expect that erosions would form at the center of the cornea where wounds are made and hemidesmosome disassembly is maximal because leading edges merge at the center of the cornea. Also, large wounds, which involve disassembly of hemidesmosomes over a larger surface area, would be expected to show more erosion. Our lab has shown that there are fewer erosions after large wounds, and this difference is significant (Pal-Ghosh et al., 2004). To address the question of where the erosions form, we examined 14 eyes after small wounds and 39 eyes after large wounds 6 weeks after wounding.

Data on the location of the erosions are presented in Fig. 1. To test the hypothesis that erosions form centrally, we used the Chi-square test. The following assumptions were made: expected values for centrally located erosions were 11 of 12, and 29 of 30, for small and large wounds, respectively; 2 of the 14 cornea with small wounds and 9 of the 39 corneas with large wounds were closed at 4 weeks and had no erosions. Of corneas with erosions, 1 of 12 (8%) and 8 of 30 (27%) were in the central region after small and large wounds, respectively, which is significantly fewer than expected (P<0.001) by the Chi-square test. We then tested the assumption that the erosions in the periphery are distributed randomly around the corneal circumference, we found significantly (P<0.001) more erosions in the nasal periphery (66% and 52% for small and large wounds, respectively) compared with the other regions. These results indicate that: (1) erosions are not caused exclusively by delayed reassembly of hemidesmosomes after wound healing, and (2) the anatomy of the mouse eye has a significant role in the development of erosions.

In a TEM study performed to look at the structure of the BMZ after debridement, we showed that the lamina densa remains on the anterior stroma after wounding, but is progressively lost, first at the center, and later at the periphery (Sta Iglesia and Stepp, 2000). Wounds that remain open for more than 24 hours showed a complete loss of the lamina densa at sites where the stroma was exposed. Basal cells that initiate migration at the wound periphery move over a BMZ with an intact lamina densa, whereas cells at the center of the cornea migrate over a BMZ with no lamina densa.

These observations, coupled with data showing more erosions at the corneal periphery, suggest that residual lamina densa has a role in the formation of erosions. We tested this hypothesis by creating superficial keratectomy wounds. Unlike debridement wounds, superficial keratectomy wounds involve removal of the corneal epithelial cells along with their BMZ and a portion of the anterior stroma. As shown in images taken 6 weeks after wounding, there were no erosions seen by slit lamp or by fluorescein staining in any of the mouse corneas subjected to keratectomy injury, although a scar could be seen in the stroma at the periphery of the cornea (Fig. 2A).

Analysis of β4 integrin localization by confocal microscopy using an ectodomain directed antibody and 3D image reconstruction revealed an abrupt transition between the site where the keratectomy wound was made and the unwounded area peripheral to it. Within the keratectomy site, elevated levels of β4 integrin were observed (Fig. 2B). There were more epithelial cell layers within the keratectomy wound zone compared with adjacent sites. More β4 integrin was expressed per basal cell and in more cell layers within keratectomy sites compared with levels in adjacent regions (Fig. 2C). Corneal epithelial cells were extracted 4 weeks after injury by limbal to limbal debridement and used for immunoblot analyses of β4 integrin expression. Immunoblots obtained using a C-terminal tail directed β4 integrin antibody confirmed that production of the ~200 kDa full-length β4 integrin protein was twice as high in keratectomy wounded corneas compared with unwounded corneas (Fig. 2D).

Next, we examined debridement wounds as a function of time after wounding for differences in expression and localization of β4 integrin by immunoblots and confocal imaging (Fig. 3A–D). As expected from our previous studies, β4 integrin expression increased within corneal epithelial cells during the active phase of re-epithelialization and remained above control levels for 1 week after both sizes of wounds. Full-length β4 integrin is ~200 kDa with a ~100 kDa ectodomain and a ~100 kDa cytoplasmic domain. Using an antibody against the C-terminal cytoplasmic tail of β4 integrin, we observed a 200 kDa band corresponding to full-length β4 integrin and a series of lower molecular mass bands. These results show that the β4 integrin N-terminus is subjected to proteolytic cleavage, reducing the size of the protein from 200 kDa to 160 kDa. We also saw 100 and 50 kDa fragments. When these fragments were quantified and presented graphically (supplementary material Fig. S1), there is evidence of increased cleavage of β4 integrin during active re-epithelialization.

To examine the localization of β4 integrin after debridement wounding, we used confocal microscopy and 3D image reconstruction with an antibody against the β4 integrin ectodomain rather than the cytoplasmic domain. Results shown in Fig. 3C indicate a loss of the β4 integrin ectodomain at and around erosion sites. Higher resolution imaging was performed (Fig. 3D) and showed β4 integrin surrounding the basal cells of the corneal epithelium of control corneas, at the leading edge during active re-epithelialization at 18 hours, and adjacent to the erosion site 4 weeks after large wounds. Results show that erosions are accompanied by focal loss of the β4 integrin ectodomain within clusters of epithelial cells.

To determine whether the levels of MMP proteins are elevated during erosion formation, we performed zymography using extracts from control and 4-week wounded corneas. Data shown in Fig. 4A indicate that there is an upregulation of MMPs with molecular mass over 100 kDa in all four corneas with erosions. MMP2 (72 kDa) and MMP3 (54 kDa) were expressed in control and wounded corneas. In two of four corneas assessed, MMP3 was elevated compared with the control. MMP9 is often referred to as the 92 kDa gelatinase, which reflects its size in human cells; in the mouse, it is larger and is over 100 kDa in corneal tissues (Mohan et al., 2002). To determine whether Mmp9 mRNA was also elevated in corneas with erosions, mRNA was isolated from control and wounded corneas for QPCR (Fig. 4B). The level of Mmp9 mRNA was below detectable limits in all control samples (n=6) evaluated. However, 4 weeks after large wounds, three of four samples tested expressed Mmp9 mRNA. These data implicate elevated levels of MMP9 in the etiology of recurrent erosions after debridement wounds.

Based on the zymogram and QPCR results and on studies showing an important role for MMP9 in maintenance of the ocular surface epithelial barrier (Pflugfelder et al., 2005), we proceeded to assess the expression of MMP9 protein as a function of time after injury by immunoblotting (Fig. 4C). We found that MMP9 protein levels were elevated at 18 hours and 2 days after small and large wounds compared with controls, and remained elevated at 1 and 4 weeks after wounding.

Next, we assessed the localization of MMP9 on the ocular surface using whole-mount confocal microscopy. MMP9-positive epithelial cells were seen at and around erosion sites 4 weeks after injury (Fig. 4D). 3D image reconstruction showed no detectable MMP9 in control corneal epithelium. MMP9 is expressed in a few suprabasal cells adjacent to the leading edge 18 hours after debridement wounding and localized within cells at and adjacent to the erosion site 4 weeks after wounding (Fig. 4E).

To determine whether MMP2, MMP3 and/or MMP9 cleave β4 integrin, we extracted proteins from mouse epidermal keratinocytes without protease inhibitors and assayed for the susceptibility of β4, α3 and β1 integrins to cleavage by three activated MMPs (MMP2, MMP3, MMP9) or by TACE (TNF-α converting enzyme/ADAM17). TACE is included in these experiments because it is an MMP family member that has been implicated in the corneal epithelial erosions that develop after exposure of eyes to mustard gas (Mol et al., 2009). After 60 minutes, extracts with no MMPs added showed a time-dependent reduction in the amount of full-length β4 integrin as a result of the action of endogenous proteases (Fig. 5A). MMP2, MMP3 and MMP9 all cleaved β4 integrin and eliminated the ~200 kDa band within 5 minutes. TACE also cleaved β4 integrin, but cleavage was slower. Prolonged exposure of immunoblots similar to those shown in Fig. 5A showed that cleavage by both MMP2 and MMP9 generated a 100 kDa fragment (see supplementary material Fig. S2).

To determine whether other integrins are susceptible to cleavage by MMPs and TACE, we assessed the degradation of α3 and β1 integrins using C-terminal-directed antibodies. α3 integrin showed differential susceptibility for cleavage by MMPs and TACE; it was sensitive to MMP3 and was completely degraded in 30 minutes, but was resistant to cleavage by MMP9 and TACE. MMP2 rapidly degraded a fraction of the α3 integrin. β1 integrin was resistant to cleavage by MMP2 and was sensitive to MMP3, MMP9 and TACE. MMP3 almost completely degraded β1 integrin by 60 minutes in a time-dependent manner. There was a rapid reduction in the level of β1 integrin after MMP9 and TACE were added to the samples, but no further loss of protein over the 60 minute time period. The partial sensitivity to degradation of α3 and β1 suggests that there are MMP-sensitive and -resistant pools of these integrins in cells.

To confirm that endogenous MMPs present within corneal epithelial cells are capable of cleaving integrins, extracts from unwounded and wounded corneal epithelia were incubated without protease inhibitors, with a protease inhibitor cocktail (+PI), or with GM6001, which is a broad-spectrum MMP inhibitor. Because epithelial extracts from animals 18 hours after wounding have 1.8-fold more β4 integrin (Fig. 3A) and 4.7-fold more MMP9 (Fig. 4C) compared with control epithelial extracts, these experiments allow us to determine whether elevated MMP9 levels alter the sensitivity of β4 integrin to degradation.

When no inhibitors were added to unwounded extracts, the amount of full-length β4 integrin decreased, and 80% was degraded by 5 hours. Adding a PI cocktail partially delayed β4 degradation and 60% was degraded by 5 hours (Fig. 5B). When the MMP inhibitor GM6001 was added to control extracts, a doublet appeared over time. When both β4 bands were summed and the percentage degradation calculated, in the presence of GM6001, only 10% of β4 integrin was degraded in extracts from control cornea.

There was more full-length β4 integrin in extracts obtained from corneas 18 hours after wounding. When no inhibitors were added to wounded extracts, β4 integrin was degraded more rapidly over time and intact β4 was below detectable levels by 5 hours. Adding the PI cocktail delayed β4 degradation during the first hour, after which β4 integrin was rapidly degraded and was also below detectable levels by 5 hours. By contrast, when the MMP inhibitor was added, more full-length β4 integrin was observed at time 0 and the rate of degradation over the time course was slowed. Compared with PI-treated samples in which 100% of the β4 integrin degraded by 5 hours, GM6001-treated samples had degraded by 75% at this time point. These data show that β4 integrin is sensitive to proteolytic cleavage by endogenous proteases and MMPs. When MMP levels are low, in unwounded extracts, cleavage can be almost entirely blocked by MMP inhibitors, but when MMP-levels are elevated after wounding, it was not possible to block β4 integrin cleavage over time.

Numerous studies have been performed using the phorbol ester TPA (tissue plasminogen activator) to activate MMPs and induce shedding of various surface proteins, including syndecan-1; these cleavage events are mediated by TACE (Pruessmeyer et al., 2010) and/or MMP7 (Chen et al., 2009) and inhibited by MMP inhibitors (Ludwig et al., 2005). To determine whether α6β4 integrin is shed from keratinocyte surfaces after TPA treatment, we incubated human corneal epithelial cells in medium with 0.02% DMSO or TPA in 0.02% DMSO. Although antibodies directed against α6 and β4 integrin ectodomains work well for staining tissues and cells, they often do not function for immunoblotting. Because shedding of the integrin ectodomains leaves the cytoplasmic tails of these molecules behind on cell surfaces, cell culture medium cannot be used to evaluate shed α6β4 integrin. We extracted cells at 20, 60 and 180 minutes after treatment with 100 mM TPA and performed immunoblots. TPA treatment reduced the levels of all three integrins assessed (β4, α6 and α3) at 20 minutes; α3 integrin levels had partially recovered by 180 minutes (Fig. 5C). To look directly at surface integrin expression, we performed flow cytometry studies using mouse epidermal keratinocytes treated with two different concentrations of TPA (100 mM and 200 mM) for 20 minutes. Similar results were obtained at both TPA concentrations (Fig. 5D). Surface levels of β1 integrin were significantly reduced; however, levels of β4 integrin, although also lower, did not achieve statistical significance.

MMP9 was chronically elevated in debridement-wounded corneas when erosions formed (Fig. 4) and β4 integrin was subject to MMP-mediated cleavage (Fig. 5B), but the experiments shown do not distinguish which MMPs are responsible for β4 cleavage. MMP9 could degrade β4 integrin directly or it could degrade integrin ligands and cause degradation of β4 integrin indirectly. If β4 integrin is a direct MMP9 target, the two proteins should associate with one another.

Coimmunoprecipitation experiments were thus performed using epithelial extracts from unwounded and wounded corneas 4 weeks after injury. We immunoprecipitated with β4 integrin or MMP9 antibodies and immunoblotted with antibodies against the β4 integrin cytoplasmic domain or MMP9 (Fig. 6). When β4 integrin was subjected to immunoprecipitation, the full-length 200 kDa protein was observed, but the majority of β4 integrin immunoprecipitated from both control and wounded samples at 4 weeks was present as a 100 kDa C-terminal fragment, consistent with cleavage of the entire ectodomain; β4 integrin immunoprecipitates from wounded corneas showed the presence of MMP9. When MMP9 was immunoprecipitated, no full-length β4 integrin was seen; however, the 100 kDa C-terminal fragment was observed in both control and wounded samples. As shown schematically in Fig. 6, the association between the β4 integrin ectodomain and MMP9 during immunoprecipitation induced degradation of β4 integrin, leaving a 100 kDa fragment containing the C-terminal intracellular domain, which resisted further MMP-mediated degradation.

The experiments presented above show that β4 integrin and MMP9 associate, and this association induces MMP9 cleavage of the β4 integrin ectodomain. If MMP9 is capable of associating with and/or cleaving β4 integrin in vivo at erosion sites, there should be less β4 integrin ectodomain detected on cells where MMP9 is localized. 3D confocal imaging was performed on whole mounts of corneas with erosions to study the colocalization of MMP9 and the β4 integrin ectodomain. There was an absence of the β4 integrin ectodomain within cells expressing MMP9 (Fig. 7), but we did observe focal but infrequent colocalization of the β4 integrin ectodomain and MMP9. Our in vitro studies show that MMP9 can associate with and cleave β4 integrin, and the in vivo studies confirm that β4 ectodomain cleavage occurs and contributes to the focal development of recurrent erosions in the cornea.

The data presented here, obtained using in vivo and in vitro approaches, show that: (1) MMP9 expression is elevated in the mouse corneal epithelium at erosion sites; (2) the ectodomain of β4 integrin is lost focally within the corneal epithelial sheet and adjacent to erosion sites where MMP9 is also expressed; (3) MMP9 associates with and degrades β4 integrin; (4) β4 integrin cleavage is inhibited by an MMP-inhibitor; and (5) α6β4 and β1 integrin cleavage are induced rapidly in cultured human corneal epithelial cells and mouse epidermal keratinocytes by MMP activation. Our studies also show that MMPs and TACE target other integrins in addition to β4 integrin.

The fact that β4 integrin is easily degraded was appreciated soon after it was identified (Kennel et al., 1989). β4 integrin is degraded by calpain (Giancotti et al., 1992; Potts et al., 1994) and by caspase-3 and caspase-7 (Werner et al., 2007). The intracellular cleavage of β4 integrin by calpain and caspases generates a β4 integrin cytoplasmic domain truncation that cannot mediate cytoskeletal assembly and hemidesmosome formation, induces rapid proteasomal degradation of β4 integrin, and sensitizes cells to apoptosis (Werner et al., 2007). Because β4 integrin is roughly 200 kDa with approximately 100 kDa inside and 100 kDa outside the cell, caspase- and/or calpain-mediated degradation cannot account for the generation of cleavage products with molecular masses of 100 kDa or larger assessed using an antibody against the C-terminal-tail of β4 integrin. Only proteases that act on the ectodomain of β4 integrin generate fragments 100 kDa or larger in size. Although a report describing the ability of matrilysin/MMP7 to cleave β4 integrin derived from human prostate cell line extracts has appeared in the literature (von Bredow et al., 1997), additional investigations of this topic have not been reported. Interestingly, there has been a report that MMP9 can induce the shedding of the β2 family integrins on immune cells (Vaisar et al., 2009). Not only does this shedding inhibit adhesion of leukocytes to inflamed tissues, but also the shed molecules appear to act as dominant-negative inhibitors of CD11 or CD18 function on leukocytes (Gjelstrup et al., 2010).

Our data are the first to show that β4 integrin associates with and is cleaved by MMP9. Other studies have used coimmunoprecipitation to show that MMP9 associates with β1 integrin (Radjabi et al., 2008) and CD11 or CD18 (Vaisar et al., 2009), that MMP2 associates with α5β1 and αv integrins (Menon et al., 2006; Morozevich et al., 2009; Chetty et al., 2010; Choi et al., 2010), and that MT1-MMP associates with both αv- and β1-containing integrins on endothelial cells (Gálvez et al., 2002). The study by Menon and colleagues (Menon et al., 2006) shows that the association between α5β1 integrin and MMP2 induces apoptosis of rat myocytes after stimulation by β-adrenergic receptor and the study by Radjabi and colleagues (Radjabi et al., 2008) shows thrombin induced tumor cell invasion by enhancing the association of MMP9 and β1 integrin. The majority of these studies used immunoprecipitation with an integrin antibody followed by immunoblotting for the MMP so it is not possible to determine whether immunoprecipiation induced integrin cleavage.

One experimental method used to assess MMP-activated shedding of cell surface molecules is to treat cells with TPA (Ludwig et al., 2005). TPA has been shown to induce the activation of several different MMP family members including MT1-MMP, TACE/ADAM17, ADAM10 and MMP7 (Endo et al., 2003; Ludwig et al., 2005; Chen et al., 2009; Pruessmeyer et al., 2010); TPA-induced ectodomain shedding of numerous cell surface proteins can be blocked by specific MMP inhibitors (Ludwig et al., 2005). MMPs are often upregulated in cultured cells in vitro, during cancer development and dissemination, and in numerous pathological conditions in vivo. We show here that TPA induced a rapid shedding of a portion of the α6β4 and α3β1 integrins from the surface of corneal epithelial and epidermal keratinocytes in vitro. Although an exhaustive survey of MMP family members to determine which are capable of cleaving α6β4 and α3β1 integrins is beyond the scope of this study, the association of MMP9 and β4 integrin in corneal epithelial extracts suggests that MMP9-mediated cleavage of α6β4 integrin occurs in vivo and in migrating cells in vitro and in vivo.

Debridement-wounded mouse corneas obtained several weeks after wounding showed focal MMP9 localization and β4 integrin ectodomain loss at erosion sites. The elevated MMP9 protein could derive from immune cells trapped within the epithelial sheet at the erosion site. Regardless of the source, the presence of the elevated levels of MMP9 on the ocular surface for prolonged times leads to pathology. Erosions in the mouse cornea are prevented by removing the basement membrane at the time of injury. This may be due to better epithelial–stromal interaction, elimination of factors released by injured epithelial cells, the presence of BMZ fragments underlying the epithelial sheet, or to a combination of all three processes.

The fact that some of the MMP9 expressed in control epithelial tissues associates with β4 integrin and that they coimmunoprecipitate, indicates that the interaction of these two proteins does not always result in cleavage of the integrin. Our studies of MMP9 expression in control and wounded corneas did not show evidence for conversion of proMMP9 to active MMP in response to wounding, perhaps because differences in the mobility of murine pro- and active-MMP9 are too subtle to be seen in the gel system we used. MMP9 is an extracellular protease that binds to the extracellular domain of β4 integrin and if MMP9 were activated within a MMP9–β4-integrin complex, it would degrade the integrin. MMP9 immunoprecipitates would then not contain β4 integrin because the β4 integrin cytoplasmic domain does not bind to MMP9. The fact that the size of the major β4 integrin fragment detected in these coimmunoprecipitates is 100 kDa, indicates that cleavage was induced during sample processing and the cytoplasmic fragment of β4 integrin was trapped on the Sepharose beads during precipitation.

If inactive proMMP9–β4-integrin forms a complex in vivo under homeostatic conditions, what might the function of this complex be? Studies of the role of MMP9–β2-integrin associations indicate that proMMP9 associates with β2 integrin intracellularly and is expressed at the cell surface upon activation of leukocytes (Stefanidakis et al., 2004). Thus, the interaction of MMP9 with integrins might help to localize MMP9 at the cell surface. Other studies, again on immune cells, suggest that complexes of proMMP9 with integrin function as docking sites for CD44v (Redondo-Muñoz et al., 2008). Additional studies in epithelial tissues are needed to determine the exact role of MMP9 in the regulation of β4 integrin activity and function in health and disease.

All experiments described were conducted in voluntary compliance with the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. All procedures for debridement and keratectomy wounds were approved by the George Washington University Animal Care and Use Committee. BALB/c mice were obtained from NCI-Frederick; C57Bl6j mice were obtained from Jackson Labs (Bar Harbor, ME). For debridement wounding, 8- to 10-week-old mice (22–25 g) were anesthetized with 250 μl of a 1:10 dilution of a 1:1 mixture of ketamine (100 mg/ml; Aveco, Fort Dodge, IA) and xylazine (20 mg/ml; Miles, Shawnee Mission, KS). Once animals were anesthetized, a topical anesthetic (proparacaine ophthalmic solution) was applied to their ocular surfaces until the blink sensation was lost, and their corneas were scraped with a dull scalpel to remove the corneal epithelium as described previously (Pal-Ghosh et al., 2008). Corneas were allowed to heal in vivo for 18 hours, 1 week and 4 weeks for small and 2 days, 1 week and 4 weeks for large wounds. For keratectomy, 1.5 mm circular wounds were created in the central cornea as described previously (Hutcheon et al., 2005). An epithelial-stromal button was removed with forceps. The healing process was monitored under a slit lamp and the corneas were photographed before and after 2% fluorescein staining. The animals were allowed to heal for 4–6 weeks. Tissue extracts were derived from ten mouse corneas for each time point and each wound size and the data presented represent typical data from each study. Every in vivo experiment was conducted at least twice. To reduce the numbers of animals needed, only when discrepancies in experimental results were observed, were the in vivo studies repeated a third time.

After sacrifice, the corneal epithelium was either scraped and frozen in liquid nitrogen for RNA or protein biochemistry, or fixed for whole mounts as described previously (Pajoohesh-Ganji et al., 2004). A minimum of three eyes was used for each time point. For assessment of ocular surface integrity and erosion location, after sacrifice, the corneas of the BALB/c mice were sutured and stained with Richardson stain (1% methylene blue prepared in 1% borax mixed 1:1 with a 1% solution of azure II, filtered and stored at 4°C) to reveal exposed basement membrane and corneas were photographed. For assessment of the ocular surface in C57Bl6j mice after keratectomy wounds, mice were sacrificed at 6 weeks and slit lamp images captured before and after corneas were stained with 2% fluorescein. For the whole mount and biochemical studies of keratectomy wounds, mice were sacrificed at 4 weeks.

The tissues stored in 100% methanol were dissected to remove the lens, iris, and retina and four incisions were made equal distances apart to aid in flattening the corneas. Corneas were then processed for whole mounts as described previously (Pajoohesh-Ganji et al., 2004). The primary antibodies used for these studies are listed in Table 1. Secondary antibodies of the appropriate class were obtained from either Invitrogen or Molecular Probes conjugated with Alexa Fluor 488 or Alexa Fluor 594 (Molecular Probes, Invitrogen, Carlsbad, CA) or conjugated with Dylight 694 (Jackson ImmunoResearch Laboratories, West Grove, PA). Routine protocols included corneas stained with secondary antibodies alone. Corneas were stained with DAPI or propidium iodide before flat mounting to reveal nuclei. To achieve the best flattening, the corneas were placed epithelial side-up, mounting medium was added, followed by coverslips.

Confocal microscopy was performed at the Center for Microscopy and Image Analysis (CMIA) at the George Washington University Medical Center. A confocal laser-scanning microscope (model MRC 1024; Bio-Rad, Hercules, CA) equipped with a krypton-argon laser and an inverted microscope (model IX-70; Olympus) was used to image the localization of Alexa Fluor 488 (488 nm laser line excitation; 522/35 emission filter), Alexa Fluor 594 (568 nm excitation; 605/32 emission filter), and Dylight 694. Optical sections (z=0.5 mm) of confocal epifluorescence images were acquired sequentially with a 63× objective lens (NA 0.63) with image acquisition software (LaserSharp ver. 3.2; Bio-Rad). 63× images for keratectomy were acquired using the Ziess 710 using similar settings as the Bio-Rad confocal microscope. Typically, six to eight optical sections were merged and viewed en face. 3D images were obtained and rotated using Volocity software (Version 5.0; Perkin Elmer)

Immunoblot studies used corneal epithelium removed by debridement after specific time points. Epithelial tissues were extracted with 100 μl of Buffer A (0.1 M Tris-HCl, pH 8.5, 0.15 M NaCl, 0.5 mM MgCl₂ and 0.5% NP-40) with or without various additives to inhibit the activity of generic proteases or MMPs. For samples extracted in Buffer A plus the PI cocktail, to 2 ml Buffer A, 25 μl Pefabloc SC Plus (Roche), 286 μl 7× complete mini protease inhibitor (Roche Complete Mini) and 20 μl HALT phosphatase inhibitor cocktail (Pierce/ThermoFisher) were added. For samples incubated with MMP inhibitors, to 10 ml Buffer A, 4 μl of a 25 mM stock solution of Galardin/GM6001 (Sigma) was added. Protein assays were performed and equal amounts of total protein were loaded onto 4–20% Tris-glycine gels (Invitrogen) and electrophoresis was performed at 140 V. The proteins were transferred to PVDF membrane (Millipore) at 400 mA for 1.5 hours, and the blots blocked in 5% milk in TBS (10×) (Bio-Rad) with 0.1% Tween 20 (TBST) overnight at 4°C. Primary and secondary antibodies were added for 1 hour each at room temperature (RT), with washes in between. Blots were subjected to chemiluminescence using Pierce SuperSignalWestDura (Pierce), and chemiluminescence was detected using x-ray film. When appropriate, data were quantified using NIH ImageJ software. Each immunoblot or immunoprecipitation study used the epithelial tissues obtained from ten corneas for each time point.

For zymography, the tissues were extracted as mentioned above and equal amounts of protein were resolved using ReadyGel Precast Zymography Gels (Bio-Rad), bands were developed as described by the manufacturer and gels were photographed. Each lane shown in the gels used for zymography used the corneal epithelial tissues extracted from one mouse.

Quantitative PCR was performed as described (Stepp et al., 2010). RNA was isolated from debrided corneal epithelia with TRIZOL (Invitrogen) following the manufacturer's protocol. For cDNA synthesis, 2 μg of total RNA was reverse transcribed using SuperScript III Reverse Transcriptase (Invitrogen). PCR amplifications were performed in a volume of 20 μl. For real-time PCR analysis, the expression level of MMP9 was determined using a Bio-Rad MYiQ iCycler and Gene Expression Macro (version 1.1). cDNA (diluted 1:200 in the final reaction) was measured from triplicate samples using iQ SYBR Green Supermix (Bio-Rad). The primer sequences for MMP9 and GAPDH were obtained from Qiagen. Relative standard curves were generated from log input versus the cycle threshold (C_t) and relative levels of cDNAs determined using the comparative C_t method with the cycle threshold difference corrected for GAPDH. Data are presented as fold change in gene expression normalized to GAPDH as described by Onat and colleagues (Onat et al., 2008). All experiments were performed in triplicate and used the epithelial tissues derived from ten corneas for each time point evaluated.

Primary mouse keratinocytes were derived from neonatal Balb/c mouse pups and were cultured as described previously (Stepp et al., 2007) and used after 3 days of culture in low-calcium medium. The human corneal and limbal epithelial (HCLE) cell line was a kind gift from Ilene Gipson (Schepen's Eye Research Institute, Harvard Medical School, Boston, MA). HCLE cells were grown in keratinocyte growth medium (Gibco) and were used at 70% confluence.

Epidermal keratinocytes were cultured for 3 days in low-calcium medium as described (Stepp et al., 2007). Purified recombinant mouse MMP2, MMP3 or MMP9 (all R&D Systems) were used after a 1 hour activation step with 4-aminophenylmercuric acetate (APMA) (Sigma) as described by the manufacturer. 2 μg activated purified MMP2, MMP3, MMP9 or TACE (R&D Systems) were used for each experiment. Cells were extracted in Buffer A without protease inhibitors and divided into five different aliquots and 2 μg of activated MMP2, MMP3, MMP9, TACE or buffer alone was added to each sample. Equal volumes of extract were removed 5, 30 and 60 minutes after incubation at 37°C and placed in 2× sample buffer to stop enzymatic activity. Replicate samples were extracted in Buffer A plus PI cocktail. Samples were run onto gels and processed for immunoblotting. Experiments were repeated three times and the data shown represent typical experimental results.

Unwounded or wounded mouse corneal epithelial tissues were harvested after 18 hours from ten mouse corneas by debridement. The epithelial cells were extracted in 100 μl Buffer A without any inhibitors. 15 μl of the extract was removed and 2× Laemmli sample buffer was added to serve as a control. The remaining extract was divided into three equal volumes and added to tubes containing an equal volume of Buffer A, Buffer A with 2× concentration of PI cocktail, or Buffer A plus a 2× concentration of GM6001. At 1, 3 and 5 hour time points, 15 μl was removed from each tube and added to 2× Laemmli sample buffer. All the samples were boiled for 5 minutes followed by gel electrophoresis and transfer. Each experiment required the use of epithelia from ten control and ten wounded corneas; experiments were repeated twice. The data presented show typical experimental results.

Human corneal epithelial cells were grown and treated with DMSO (ThermoFisher, #191418) only or DMSO plus TPA (Calbiochem, #524400) at a concentration of 100 mM. The final DMSO concentration in the media was maintained at 0.02%. After 20, 60 or 180 minutes, cells were extracted in Buffer A plus PI cocktail and equal amounts of total protein run onto SDS gels and processed for immunoblotting. Control cells had been incubated in DMSO alone for either 20 or 180 minutes. Experiments were repeated twice and the data shown represent typical results. Flow cytometry was performed using primary Balb/c mouse epidermal keratinocytes cultured as described previously (Stepp et al., 2007) treated with either 0.02% DMSO alone for controls, or with 100 mM or 200 mM TPA in 0.02% DMSO for 20 minutes. Experiments were repeated twice at each of the two concentrations of TPA and results were similar for all four experiments. The antibodies used to assess surface β1 and β4 integrins are indicated in supplementary material Table S1.

Samples were extracted in either RIPA buffer (10× RIPA Lysis Buffer, Millipore) or with Buffer A [50 mM Tris-HCl, pH 8.0, 0.15 mM NaCl, 0.5 mM CaC1₂, 0.5% (v/v) Nonidet P-40]; a PI cocktail was added and extracts were pre-cleared with 30 μl of washed and packed Dynabeads (Invitrogen) for 15 minutes at room temperature. 3 μl of the appropriate primary antibody was added to 30 μl of washed and packed Dynabeads and 500 μl of RIPA Buffer and incubated at room temperature for 1 hour. Control beads were incubated with rabbit IgG. Beads were washed three times with RIPA buffer (500 μl). The bead–antibody complex was resuspended in 500 μl of RIPA Buffer with PI cocktail and 15 μg of the cell extract was added and samples incubated for 1 hour at room temperature with rotation. After several washes with 1 ml RIPA Buffer with PI cocktail, 30 μl of 1× Laemmli sample buffer with 20 mM DTT was added to the samples. Samples were heated at 100°C for 5 minutes and resolved on gels, transferred for immunoblotting, and processed as described above.

We used either the Student's t-test (GraphPad Instat) or the Graphpad QuickCalc Chi-Square calculator (http://www.graphpad.com/quickcalcs/chisquared2.cfm) where appropriate, as indicated.

This work was supported by NIH grants EY08512 to M.A.S. and EY005665 to J.D.Z. Zeiss 710 confocal microscopy was made possible by NIH SIG grant F10 RR025565. We would also like to thank Ilene Gipson for the gift of the HCLE cell line; Anastas Popratiloff, director of the GWU Center for Microscopy and Image Analysis for help with imaging; Teresa Hawley, director of the FACs Core Facility at GWU, for help with flow cytometry. Deposited in PMC for release after 12 months.