A poly(oxyethylene)–poly(oxypropylene) copolymer was obtained from BASF (Pluronic F-127 [PF-127]; Poloxamer 407; Ludwigshafen, Germany) and PBS was prepared from PBS tablets (Oxoid Ltd., Basingstoke, UK). Unmodified mouse anti-Cx43 ODN (5′-GTA ATT GCG GCA GGA GGA ATT GTT TCT GTC-3′) and Cy3-tagged mouse anti-Cx43 ODN (5′-Cy3 GTA ATT GCG GCA GGA GGA ATT GTT TCT GTC-3′) were purchased commercially (Sigma-Genosys, Castle Hill NSW, Australia). The Cx43 AsODN was designed to be Cx43 specific, with potential sequence-related side effects such as partial sequence homology to unrelated genes (including other connexin family genes), and GGGG or CpG motifs avoided. This particular antisense sequence has been used in animal skin wound, spinal cord, and optic nerve models with vehicle or sense controls used. ^{18,19,21 –23} In this study, antisense treatment has been compared with standard of care. For immunohistochemical studies, the following primary antibodies were used: monoclonal mouse anti-Cx43 primary antibody (a gift from David Becker ²⁴ ), monoclonal mouse anti–α-SMA antibody (Novocastra Laboratories Ltd, Newcastle upon Tyne, UK) and rabbit polyclonal anti-laminin-1 antibody (Sigma Chemical, St. Louis, MO). Mouse or rabbit secondary antibodies (Alexa Fluor 488 or Alexa Fluor 568 secondary; Invitrogen, Carlsbad, CA) were used and Hoechst 33258 (Sigma) served as a cell nuclei marker. Other reagents for tissue processing included: goat serum (GIBCO/Invitrogen, Auckland, New Zealand), water-soluble resin for cryostat sectioning (Tissue-Tek OCT Compound; Sakura Finetek, Torrance, CA), an antifadent agent (CitiFluor AF1; CitiFluor Ltd., London, UK), and synthetic-resin mounting media (DPX mountant; BDH Chemicals Ltd., Poole Dorset, UK). 

Female Wistar rats (d 33) were raised under conditions specified by the ARVO Resolution on the Use of Animals in Research. Experiments were performed with ethics approval from the University of Auckland Animal Ethics Committee. Animals were anesthetized using an anesthetic mixture (1:1 Hypnorm [Janssen Pharmaceutica, Beerse, Belgium] and Hypnovel [Roche Products Ltd., Auckland, New Zealand]) at a dosage of 0.083 mL/100 g body weight. Initial mechanical scrape wounds were performed under a standard laboratory dissecting microscope (Carl Zeiss MicroImaging, Göttingen, Germany). Using a sterile 2-mm-diameter dermal biopsy punch (Miltex Inc., York, PA), the central part of the corneal epithelium was punched and the epithelium within the demarcated area was scraped off using a scalpel blade (Swann-Morton Ltd., Sheffield, UK). 

For a more detailed analysis a precise wound size (area and depth) excimer laser ablation protocol was used. Animals were divided into three groups, each consisting of 20 eyes: (1) control group with no surgery, (2) untreated laser wounded corneas, and (3) laser wounded corneas treated with Cx43 AsODN in thermoreversible gel (PF-127). Because some animals were euthanized at subsequent time points for histologic analysis the number of eyes per group varied from 20 at 2 hours postwounding down to 10 at 72 hours postwounding (see Table 2). Excimer laser wounding (Technolas 217 Z; Bausch & Lomb Surgical, Rochester, NY) followed a standard algorithm ^25,26 and was performed through the intact epithelium, centered over the pupil, resulting in a 3-mm-diameter and 50-μm-deep ablation. For treatment, Cx43 AsODN (2 μM) in PF-127 gel was freshly prepared on the day of the surgery using a standard protocol. ^27,28 A volume of 50 μL was injected into the subconjunctival space of the lower fornix within 3 minutes after surgery, serving as a drug depot and 20 μL was applied topically, directly in the fornix. 

Mechanical scrape wounding was performed as described earlier and 10 μL of 2 μM Cy3-Cx43 AsODN in PF-127 gel was applied to the wound site immediately after surgery. Three hours after application, rats were euthanized and eyes were enucleated. Eyes were attached to the bottom of a petri dish using a drop of highly durable adhesive (Super Glue; Super Glue Corp., Rancho Cucamonga, CA) and the petri dish was filled with PBS. Drug penetration was assessed performing z-scans on a confocal laser scanning microscope (Leica TCS-SP2; Leica Microsystems, Wetzlar, Germany). Optical slices were taken in 20-μm steps. Three-dimensional (3D) reconstructions of z-stacks were generated using visualizing and analyzing software (AMIRA, version 3.1; Visage Imaging Inc., San Diego, CA). Penetration depths at the wound site and penetration distances along the epithelium at the wound leading edge (n = 3) were measured using ImageJ image-analysis software (developed by Wayne Rasband, National Institutes of Health, Bethesda, MD; available at http://rsb.info.nih.gov/ij/index.html). 

To validate the hypothesis of the antisense treatment, 10 μL of 2 μM Cx43 AsODN in PF-127 gel was applied to the left eye immediately after scrape wounding, whereas the right eye, applying the vehicle alone, served as control (n = 6 per group). Eyes from these animals were removed 12 hours after treatment and 2 drops of concentrated fluorescein solution were applied to the corneal surface to demarcate any wound area remaining. Images were taken with a digital camera (Panasonic Corp., Osaka, Japan) mounted onto a dissection microscope (Takagi OM-5; Takagi Seiko Co., Ltd., Nakano-shi, Japan) and wound areas were measured in ImageJ using a calibration scale. 

In vivo confocal microscopy was performed on anesthetized animals at 0, 2, 8, 24, 48, and 72 hours post-PTK surgery. The apex of the cornea was placed parallel to the in vivo confocal objective lens and the eyelids of the animals were opened and closed a few times to simulate blinking. Using a ×40 nonapplanating lens, the light intensity was decreased to half of the intensity routinely used for human subjects and the in vivo confocal microscope was adjusted at four passes and 400-μm working distance. A drop of eye lubricant gel (Viscotears; CIBA Vision, Castle Hill, Australia) on the objective lens provided immersion, preventing direct contact between the objective lens and the corneal surface. Centration was achieved using the central pupillary zone to maximize reproducibility. Up to 250 sequential images were taken for each cornea and images were analyzed qualitatively by two blinded experienced observers. The epithelium was evaluated for the presence of epithelial cells with classic morphology as well as elongated, so-called sliding cells. The stroma was divided into anterior and posterior regions and three consecutive images for each region were evaluated for hyperreflectivity (Hr, increased brightness or haze), hypercellularity (Hc, increased number of stromal cells), and lost visualization (Lv, loss of nuclear detail or sharpness). Each of these parameters was rated using a 0-to-4 grade scale: absent (0); minor (1); moderate (2); advanced (3); and severe (4) (Fig. 1). The mean score of each subject group for each of the three parameters was taken, and the scores for the three parameters were then summed to give a total score for each group. The endothelium was examined for integrity and visualization. 

At predetermined time points postwounding (8, 24, 48, and 72 hours, n = 4 per group), eyes were enucleated, mounted in water-soluble resin for cryostat sectioning (Tissue-Tek OCT, Sakura Finetek), and frozen in liquid nitrogen. Corneal sections (16 μm) were cut parallel to the optical axis of the eye and mounted on electrostatic slides (Superfrost Plus; Menzel-Gläser, Braunschweig, Germany). Standard hematoxylin and eosin (Gill II H&E) staining was performed to evaluate the stromal thickness. For immunolabeling, nonspecific binding was blocked with 10% normal goat serum in PBS for 1 hour at room temperature. Slides were then incubated with the appropriate primary antibodies (Cx43, α-SMA, and laminin-1) at 4°C overnight. The Cx43 antibody specifically labels gap junction proteins, α-SMA targets myofibroblasts, and laminin-1 labels the basal lamina of the corneal epithelium. After rinsing the slides in PBS three times for 15 minutes, sections were incubated with the appropriate secondary antibody for 2 hours at room temperature. Sections were again washed in PBS three times for 10 minutes, before counterstaining the cell nuclei with Hoechst 33258 for 10 minutes and mounting the sections in antifading medium (CitiFluor Ltd.). Histology slides were viewed under the light microscope (Leica DMRA; Leica Microsystems) and the stromal thickness was determined using ImageJ software. Immunolabeling was analyzed using a confocal laser scanning microscope (Leica TCS-SP2 or Leica TCS-4D; Leica Microsystems) and Cx43 levels and the number of stromal cells were quantified using ImageJ software. 

Figure 2 shows a side-on view following 3D reconstruction from a z-stack image series obtained using confocal laser scanning microscopy and reconstructed with visualizing and analyzing software (AMIRA). The Cy3-tagged AsODN penetrated freely through the hydrophilic stroma in the central cornea (330.3 ± 15.1 μm), where the epithelium had been removed. Further penetration into the anterior chamber seems to be prevented by Descemet's membrane and the endothelium, which appeared to form a barrier to the highly hydrophilic AsODN. In more distal areas where the epithelium was still intact, very little Cy3-tagged AsODN was detectable in the stroma beneath (not shown). Large quantities of AsODN accumulated in the epithelium close to the wound and penetrated centrifugally along the basal layers, especially of the epithelium adjacent to the wound leading edge (287.2 ± 30.2 μm) (Fig. 2). 

A two-way ANOVA revealed no significant difference (P > 0.05) in wound size between the six rats in each group, confirming reproducibility of the mechanical scrape wound model and the associated healing response. Twelve hours after scraping (2-mm diameter, 3.14-mm² lesion), Cx43 AsODN–treated eyes exhibited a significant increase in reepithelialization of the wound bed compared with application of the vehicle alone (1.59 ± 0.37 and 2.29 ± 0.58 mm² remaining open, respectively; P < 0.01). 

Unwounded corneas (Group 1) demonstrated the typical three distinctive cellular layers with in vivo microscopy, the characteristic features of which are presented in Table 1. 

A summary of in vivo confocal microscopy results for excimer laser ablation wounded corneas are shown in Table 2. These corneas had 7.06-mm², 50-μm-deep ablations. Untreated corneas (Group 2) demonstrated severe signs of edema 2 hours after surgery, which made cellular structures difficult to visualize. No endothelial details could be discerned in these corneas. At 24 hours, none of the untreated corneas showed signs of central reepithelialization, although unusual epithelial cells, possibly proliferating cells, were evident at the margin of the photoablation zone from 8 hours onward. By 24 hours the stroma had an increased number of cells that were different in morphology to normal keratocytes, and were assessed as being inflammatory cells. These cells were brighter, smaller, and circular in appearance. They were present in very high quantities in the anterior part of the stroma, gradually decreasing in number toward the endothelium. Reepithelialization was detected after 48 hours in all untreated corneas. The corneal stroma appeared to be hypercellular, but with less haze and reflectivity. Within 72 hours complete reepithelialization was noted in 8 of 10 untreated eyes. 

Corneas treated with Cx43 AsODN (Group 3) also presented with edema 2 hours after ablation. However, this was not as severe since endothelial structure was clearly discernable. Within 24 hours, all corneas examined (n = 16) demonstrated almost complete central reepithelialization. By 24 hours the stroma featured an increased number of cells with a morphology differing from that of the typical keratocyte morphology (identified as inflammatory cells); however, these cells were restricted to the anterior stroma. The posterior stroma had normal keratocyte morphology and only mild edema. The endothelial mosaic was well defined. Within 48 hours, reepithelialization was fully complete in 75% (n = 9) of examined eyes and the stromal architecture was close to normal, with only a few inflammatory cells localized within the anterior stroma. Corneal morphology returned to baseline within 72 hours for all 10 eyes (100%) examined at that time point. 

Table 3 compares changes in stromal thickness measured on H&E-stained corneal sections before and at 24, 48, and 72 hours after laser ablation. It should be noted here that laser ablation was predicted to remove 20 μm of the anterior stroma, which was taken into account when analyzing differences between pre- and postsurgical values. Applying Mann–Whitney nonparametric statistical analysis, the stroma of all Cx43 AsODN–treated corneas was found to be significantly thinner in the periphery as well as the central wound site at all time points when compared with laser ablated but untreated cornea controls (P < 0.05). Most notable was an almost tripling in corneal thickness in the ablated controls arising from edema at the periphery, remaining almost double in thickness at 48 and 72 hours. In contrast, Cx43 AsODN–treated corneas remained close to prewound values. 

Immunohistochemical labeling of laser ablated but untreated control and Cx43 AsODN–treated corneas was carried out at 8, 24, 48, and 72 hours postwounding. Cx43 labeling in unwounded rat cornea is shown in Figure 3A. At 8 hours postwounding Cx43 was found to be downregulated at the wound edge in migrating epithelial cells. This was more pronounced in Cx43 AsODN–treated corneas in which a longer migrating “tongue” was observed correlating with the faster wound closure rates observed with fluorescein staining and in vivo confocal microscopy. In both Cx43AsODN–treated and untreated corneas there was an initial drop in anterior stroma Cx43 levels correlating with loss of stromal keratocytes immediately beneath the lesion (as seen with Hoechst labeling of nuclei). However, by 24 hours, as was seen with in vivo confocal microscopy, an inflammatory response was evident. Cx43 AsODN–treated corneas exhibited lower Cx43 levels and reduced numbers of cells in the central stroma at all time points investigated, correlating with reduced hyperplasia and stromal edema. 

A detailed Cx43 spot count across the stroma was performed at 24 hours postsurgery and demonstrated significantly reduced numbers of Cx43 spots per μm² in treated corneas compared with controls (1.29 × 10⁻³ and 3.29 × 10⁻³, respectively; P < 0.05). In parallel, a detailed cell nuclei count at 24 hours after laser ablation showed stromal cell counts were elevated in the anterior stroma, but to a much lesser extent in the Cx43 AsODN–treated corneas compared with laser ablated but untreated controls (prewound count 36,469 ± 11,122 cells/mm³, n = 17; control 144,643 ± 60,989/mm³, n = 17; Cx43 AsODN–treated 93,468 ± 53,548/mm³, n = 17). In the posterior stroma, laser ablated but untreated controls showed an increase in cell counts compared with baseline levels; however, the Cx43 AsODN–treated corneas showed no increase in cell count (prewound count 33,909 ± 8753/mm³, n = 17; control 46,901 ± 26,964/mm³, n = 17; Cx43 AsODN–treated 33,510 ± 11,350/mm³, n = 14). Both anterior and posterior cell counts for Cx43 AsODN–treated corneas proved to be significantly reduced compared with laser ablated but untreated cornea controls (P < 0.05). At 48 and 72 hours both Cx43 and stromal cell density remained lower and closer to that of normal unwounded tissue in Cx43 AsODN–treated wounds than for wounded controls. 

Normal Cx43 localization in the rat cornea is shown in Figure 3A. At 48 hours after ablation, untreated corneas showed an upregulation of Cx43 into the upper epithelial layers where healing had occurred and in the stroma correlating with hyperplasia of the epithelium as well as stromal edema (Fig. 4A). On the other hand, those treated corneas that had completed reepithelialization showed normal numbers of epithelial layers evenly across the wound site, with Cx43 being restricted to the basal cell layer of the epithelium and gap junction density in the stroma appearing to show little difference from unwounded corneas (compare Fig. 4B with unwounded cornea in Fig. 3A). 

The inflammatory response in the stroma beneath and adjacent to the wound site was evident as increased cellularity associated with hyperreflectivity in the in vivo confocal microscopy analysis (Table 2) and increased cell counts assessed histologically (Hoechst-labeled nuclei) from 24 hours onward. This increase correlated with myofibroblast differentiation as confirmed with α-SMA immunohistochemistry on tissue sections on untreated and treated corneas (Figs. 4C, 4D). At all time points studied (from 24-hour postwounding onward), the laser ablated but untreated corneas showed high levels of myofibroblast labeling at the wound site and in the periphery throughout the entire stromal thickness (Fig. 4C), whereas Cx43 AsODN–treated corneas exhibited only a small number of myofibroblasts concentrated primarily in the anterior part of the stroma (Fig. 4D). 

Laminin-1 labeling was used to evaluate the reestablishment of the basal lamina of the epithelium. Normal basal lamina labeling is shown in Figure 3B. Cx43 AsODN–treated corneas showed a more regular, less undulating lamina being laid down under the new epithelium with a more intense and continuous laminin deposition (Fig. 4F) compared with the laser ablated but untreated controls (Fig. 4E). A Mann–Whitney nonparametric statistical analysis revealed that the amplitude of undulation in the treated corneas was significantly different from controls (P < 0.01). The untreated corneas, showing signs of hyperplasia, did not have laminin deposition at the center of the wound and often showed a discontinuous basal lamina (Fig. 4E). 

Cx43 is generally expressed in the basal layers of the rat corneal epithelium, where it regulates proliferation and differentiation, as well as in the stroma, where it allows for keratocyte coupling, providing a key pathway for nutrient and waste exchange in this avascular tissue and, therefore, coordination of stromal homeostasis. ^15,29 Previous studies have indicated that connexin levels in the epithelium change after corneal wounding. Ratkay-Traub et al. ¹⁷ demonstrated that Cx43 and Cx26 were upregulated in rabbit cornea after excimer corneal wounding, but Cx43 was reduced at the epithelial edge itself that was migrating to close the wound. Similarly, Suzuki et al. ³⁰ showed that actively migrating rat cornea epithelial cells lacked Cx43 gap junctions and desmosomes normally present in the basal cell layer. They proposed that these intercellular junctions reassembled with the reestablishment of the basement membrane. More recently, Morishige et al. ³¹ confirmed Cx43 downregulation in basal cells of the rat cornea at the leading edge of the migrating epithelium, although they suggested that Cx43 levels were stable in the remaining portion of the epithelium. Remarkably, there has been little study of Cx43 between stromal keratocytes after corneal wounding, with Ratkay-Traub et al. ¹⁷ only commenting that Cx43 was always detected in corneal keratocytes and endothelium and noting that “stromal reaction was usually missing in studies presented by others.” 

In the present study we confirmed that Cx43 levels decrease at the epithelial migrating edge, but also showed that further transient downregulation with Cx43-specific AsODN significantly increased the rate of wound closure. Our AsODN delivery studies indicated that Cx43 AsODN accumulated in the stroma beneath the wound and at the wound leading edge of the epithelium, where Cx43 protein knockdown should therefore primarily occur. This profile indeed correlated with the reduction in Cx43 levels at the wound edge compared with untreated controls, as seen by immunohistochemistry, and was consistent with a faster wound closure observed with in vivo confocal microscopy, fluorescein staining, and histologic analysis. In addition, we analyzed Cx43 in the stroma after wounding, where it was initially lost immediately below the injury site, probably as a result of cell dieback, but was subsequently upregulated in the stroma both at and peripheral to the wound site. Cx43 AsODN application reduced stromal levels of Cx43, and in parallel reduced edema and inflammation (both cell counts and myofibroblast differentiation), and promoted improved basal lamina repair. It is not possible to carry out Western blot analysis with sampling limited to the affected areas of tissue, and although we have counted the number of immunolabeled gap junctions, this approach does not take into account possible changes in plaque size. However, our healing results are consistent with reports from skin wound studies, where Cx43 downregulation decreased inflammation and edema, and triggered faster epithelial closure. ^18,19 It is postulated that downregulation of Cx43 with specific AsODN may enhance corneal wound healing by a dual mechanism: increased epithelial cell migration and decreased stromal swelling and inflammation. Reduced Cx43 levels in the epithelium may enable cells to differentiate into a migratory phenotype to close the wound, as occurs with skin epidermal keratinocytes. ¹⁸ Lower Cx43 levels in the stroma lead to reduced edema, which is thought to be associated with Cx43 hemichannel opening, ³² as well as reduced inflammation and bystander-mediated lesion spread. The spreading of proinflammatory responses through Cx43 gap junctions has been reported in lung endothelial cells. ³³ Similarly here, proinflammatory signals appear to be spreading between keratocytes. This would explain why proliferation and myofibroblast differentiation deep beneath and peripheral to the lesion seen in control wounded animals are reduced with Cx43 AsODN treatment. The sensitivity of the fluorescence tag technique means that we cannot entirely exclude the possibility that some AsODN has reached the endothelium. However, the transient knockdown of Cx43 appears to have a limited effect on connexin levels in unwounded tissues, ³⁴ with the major effect being where Cx43 levels are changing in the first hours after injury. The same Cx43 AsODN has been used previously by Nakano et al. ³⁵ to study endothelial wound healing, but we did not observe any effects on the endothelium, which was unwounded in our study. 

Zieske et al. ³⁶ demonstrated that starting 24 hours after epithelial debridement the remaining keratocytes began to undergo proliferation. In their study, keratocytes did not differentiate into myofibroblasts, which was in agreement with Moller-Pedersen et al., ³⁷ who found that scrape wounds produced few myofibroblasts. However, Moller-Pedersen and colleagues ³⁷ also found that PRK wounds can induce many myofibroblasts, and rapid cell proliferation and myofibroblast differentiation also constituted a prominent feature with the laser ablation presented here. ³⁷ Our results showed that the stromal hypercellularity (assessed with in vivo confocal microscopy) correlated with α-SMA labeling for myofibroblast differentiation, and both were reduced after transiently blocking Cx43 gap junction communication. Myofibroblasts are responsible for scar tissue deposition and visual impairment. ^37,38 Although myofibroblasts have also been implicated in proliferation, migration, and differentiation of the overlying epithelial cells, ³⁹ our results with faster wound closure in parallel with reduced myofibroblast activation would suggest this latter role may not be essential in this setting. 

Considering the anatomic characteristics of the eye, the delivery of therapeutic agents can be problematic. We chose PF-127 gel as the delivery vehicle in our experiments because previously published results on skin wounds demonstrated the successful delivery of Cx43 AsODN in living tissues. ^27,28,40 Moreover, as a thermoreversible polymer, PF-127 gel can easily be applied in liquid form at 4°C, thereafter gelling rapidly at body temperature and forming a depot from which the drug is released slowly over time. ^41,42 In terms of the delivery route, we performed a combination of subconjunctival injection, which creates a depot for slow release throughout the entrance site, ^{39 –42} and topical application for immediate action. Although subconjunctival application is widely used in ophthalmology, topical application is more applicable for the clinical setting and provides direct access of the applied drug to the injured tissue. In our scrape wound experiments, the gel was applied topically only, and results would suggest that for this indication, subconjunctival injections would be of questionable value. 

Evaluation of optical properties of the living cornea can be achieved by various instruments. However, in vivo confocal microscopy is a precise technique for detecting information at a microstructural level. ^{43 –45} We have developed an algorithm for dynamic observation and grading of the healing processes in the cornea suitable for use after laser ablation. ^{46 –48} This stable, uniform model provided the opportunity to evaluate the efficacy of Cx43 AsODN to promote corneal wound healing at various time points without having to euthanize the participating animals. In vivo confocal microscopy observations revealed reduced adverse corneal changes and improved structural recovery in the AsODN-treated corneas compared with the laser ablated but untreated controls. The improved corneal clarity in the Cx43 AsODN–treated corneas, and the combined effect of faster reepithelialization and reduced inflammation, would be expected to improve visual outcomes. 

In conclusion, corneal wound healing is of major interest for clinical and experimental ophthalmology, in that minimizing corneal fibrosis may be associated with significant improvement of the optical properties and therefore quality of vision. ^{49 –51} There are a number of aging, degenerative, infective, mechanical, and surgical causative factors that are associated with corneal wound healing aberrations. ⁵⁰ On the basis of our experiments we may conclude that application of Cx43 AsODN reduces edema and inflammation, and thus would be expected to reduce scarring, and promotes uniform repair of the basal lamina as well as faster epithelial recovery. The enhancement of corneal healing and improved optical quality after application of Cx43 AsODN has the potential to be of great benefit for practical ophthalmology after surgical procedures as well as for treating anterior surface diseases such as persistent epithelial defects and trauma.