Connexin43 Knockdown Accelerates Wound Healing but Inhibits Mesenchymal Transition after Corneal Endothelial Injury In Vivo

Abstract

purpose. To explore connexin43 (Cx43) knockdown as an efficient treatment for corneal endothelial injury in an in vivo rat corneal scrape injury model.

methods. Scrape injury was induced in the corneal endothelium, and immunolabeling (ZO-1, α-SMA, Cx43) was performed to analyze changes in Cx43 expression during wound healing. Single injection of Cx43 antisense oligodeoxynucleotide (AS-ODN), small interfering RNA (siRNA), or adenovirus (CMV-Cx43-mRFP1) was applied into the anterior chamber simultaneously with the injury, and wound closure was examined by immunolabeling (ZO-1, Cx43) and propidium iodide staining. Corneal endothelium proliferation on day 1 after injury was studied by Ki67-immunolabeling. Cx43-knockdown treatment was performed also without injury, and its effect on Cx43 expression and Ki67 immunolabeling was examined. The postinjury appearance of myofibroblasts in Cx43 AS-ODN- and sense-ODN–treated corneas was compared by α-SMA-immunolabeling.

results. Complete wound closures were observed in five of six corneas on day 3 after injury with either Cx43 AS-ODN or siRNA treatment, whereas no complete closure was observed on day 3 in the control corneas (S-ODN, zero of six; or nonsense siRNA, zero of six). Consistently, Cx43 overexpression using adenovirus delayed wound closure. Cx43 knockdown increased the number of Ki67-positive proliferating cells on day 1, whereas it decreased the number of α-SMA-positive myofibroblasts on day 5. Cx43 knockdown without injury decreased Cx43 expression and induced endothelial proliferation in vivo.

conclusions. These results show that Cx43 knockdown induces corneal endothelium proliferation but inhibits endothelial–mesenchymal transition/transformation after injury, suggesting that Cx43 knockdown is a new therapeutic approach for acceleration of wound closure and for prevention of retrocorneal fibrous membrane formation.

The corneal endothelium is extremely vulnerable to injury. One reason for this is that corneal endothelial cells in vivo are actively maintained in a nonproliferative, G₁-phase–arrested state and normally lack a robust proliferative response to cell loss. ¹ As a result of excessive cell loss due to accidental or surgical trauma, dystrophy, or disease, the endothelium either does not replicate at all or replicates at a rate too low to close the defect. Several mechanisms that contribute to inhibition of corneal endothelial proliferation can be considered. First, the TGF-β ^{2 3} and cAMP ⁴ present in aqueous humor may inhibit endothelial proliferation. Second, cell–cell contact itself appears to be responsible for inhibiting endothelial proliferation during development and to help maintain the mature endothelial monolayer in a nonproliferative state. ⁵ Furthermore, in both mechanisms, p27kip1, a member of the cyclin-dependent kinase-interacting protein/cyclin-dependent kinase inhibitory protein (Cip/Kip) family, seems to play an important antiproliferative role in the corneal endothelium. ^{5 6 7 8}  

Another reason for the vulnerability of the corneal endothelium to injury is that even if healing occurs after wounding, fibrosis is often induced, which causes loss of corneal clarity, resulting in loss of visual acuity. The most common example of fibrosis observed in vivo is the development of a retrocorneal fibrous membrane. ^{9 10 11 12} Endothelial–mesenchymal transition/transformation (EMT) appears to play an important role in corneal fibrosis after endothelial injury. ¹³ During EMT, corneal endothelial cells lose their characteristic polygonal morphology, and contact-inhibited phenotypes and are converted to multilayered myofibroblasts, leading to an excessive deposition of extracellular matrix components. The key features of EMT include the expression of α-smooth muscle actin (α-SMA), the hallmark of the myofibroblast phenotype. ^{14 15} TGF-β ¹⁶ and FGF-2-tyrosine kinase receptor–associated pathways ¹³ have been shown to be of importance in EMT in the cornea. For the two reasons just mentioned, both circumventing the antiproliferative properties of the corneal endothelium and inhibiting EMT leading to fibrosis would potentially bring about optimal treatment for corneal endothelial injury. However, no appropriate treatment that achieves this purpose has yet been developed. 

Gap junctions are cell–cell communicating channels that consist of multimeric proteins called connexins and mediate the exchange of low-molecular-weight metabolites and ions between contacting cells. ^{17 18} Thus far, 21 human genes and 20 mouse genes for connexins have been identified. ¹⁹ Among their products, connexin43 (Cx43) is the most widely expressed connexin, and its expression is also found in the corneal endothelium. ^{20 21 22 23 24} It has been shown that Cx43 expression pattern alters dynamically during wound healing in the epidermis, ^{25 26 27 28 29 30} the tongue, ³¹ and the rabbit corneal epithelium. ³² Cx43 expression first decreases at the wound edge, but subsequently increases on wound closure, in regions where cells are differentiating. Furthermore, knockdown of Cx43 expression either by Cx43 gene knockout ²⁷ or treatment with Cx43 antisense oligodeoxynucleotide (AS-ODN) ^{26 33} has recently been demonstrated to accelerate the wound closure in the mouse epidermis. To our knowledge, however, so far there have been no reports on the application of Cx43 knockdown in wound healing of the corneal endothelium, not to mention its effects on EMT. 

In this study, we hypothesized that the expression of Cx43 in the corneal endothelium changes drastically after injury, that these changes are crucial for wound healing, and that Cx43-knockdown approach provides an efficient treatment for corneal endothelial injury. To test these hypotheses, we analyzed changes in Cx43 expression in the rat corneal endothelium in vivo after scrape injury and studied the effects of Cx43 AS-ODN and Cx43 siRNA on wound healing. 

Materials and Methods

Rat Models of Corneal Endothelium Wound Healing and Cx43 AS-ODN or siRNA Treatment

Male Wistar rats (aged 8 weeks) were raised under conditions specified by the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. They were anesthetized with an intra-abdominal injection of pentobarbital sodium and topical oxybuprocaine. Under a stereoscopic microscope, a 30-gauge needle was inserted into the anterior chamber of the eye through the limbus, and an endothelial wound was made by gently scraping cells, taking care not to collapse the anterior chamber or to damage Descemet’s membrane. After 20 μL of aqueous humor was aspirated, 20 μL of a 40 μM Cx43 AS-ODN or siRNA solution was injected into the anterior chamber. A single application of Cx43 AS-ODN or siRNA was made to one eye, and an identical application of control sense oligodeoxynucleotide (S-ODN) or nonsense siRNA (control siRNA) was made to the other eye. On day 1 or 3 after injury, the animals were euthanatized, and their eyes were enucleated. The corneas were then carefully excised, leaving a 1-mm ring of sclera. The Cx43 AS-ODN we used is 5′-GTAATTGCGGCAGGAGGAATTGTTTCTGTC-3′, ²⁶ and the control sense oligodeoxynucleotide (S-ODN) is 5′-GACAGAAACAATTCCTCCTGCCGCAATTTAC-3′. ²⁶ The sense sequence of Cx43 siRNA is 5′-CAAUUCCUCGUGCCGCAATT-3′, and the antisense sequence is 5′-UUGCGGCACGAGGAAUUGTT-3′. The sense sequence of control siRNA is 5′-AAUUCUCCGAACGUGUCACGT-3′, and the antisense sequence is 5′-GUGACACGUUCGGAGAAUUTT-3′. 

To confirm penetration of Cx43 AS-ODN or siRNA into the corneal endothelium, we injected 20 μL of a 40 μM FITC-labeled Cx43 AS-ODN- or FITC-labeled Cx43 siRNA into the anterior chamber of the eye without injury and observed its fluorescence 1 day after injection in surface preparations under a confocal microscope. 

Overexpression of Cx43 by Adenovirus Containing Cytomegalovirus Promoter-Cx43-Monomeric Red Fluorescent Protein 1

The expression vector for Cx43-mRFP1 was constructed as previously described, ³⁴ and an adenoviral vector (CMV-Cx43-mRFP1) was constructed by inserting Cx43-mRFP1 into a vector system (AdEasy Adenoviral Vector Systems; Stratagene, La Jolla, CA) according to the manufacturer’s instructions. Just after making an endothelial wound, a 10-μL solution of recombinant adenovirus containing CMV-Cx43-mRFP1 (5 × 10⁸ pfu/mL) was injected into the anterior chamber. To confirm that the adenovirus works in the corneal endothelium, Cx43-mRFP1 was observed under a confocal microscope without fixation or staining on days 2 and 3 after application. 

Cx43-Knockdown Treatment without Injury

A single topical application of Cx43 AS-ODN or siRNA to one eye and an identical application of S-ODN or control siRNA to the other were made without injury in conditions otherwise similar to the injured endothelium. At 6 or 12 hours and 1, 2, 3, or 5 days after application, the animals were euthanatized, and their eyes were enucleated. Corneal buttons with scleral rims were dissected and assayed. 

All animal experiments described in this study were approved by the Committee for Animal Research, Kyoto Prefectural University of Medicine. 

Immunohistochemistry

Two types of sample preparations, surface preparations, and paraffin-embedded corneal cross sections were made for immunohistochemistry. Surface preparations were made according to the method reported by Crewe and Armitage. ³⁵ The corneas were cut into square corneal pieces (4 × 4 mm) with a slit knife. The pieces were incubated with the following primary antibodies, diluted in PBS containing 1% BSA, 1% NaN₃, and 0.1% Triton X-100: a rabbit polyclonal antibody against ZO-1 (1:400 dilution; Zymed Laboratories, South San Francisco, CA), a rabbit polyclonal antibody against Cx43 (1:400 dilution; Zymed Laboratories), a mouse monoclonal antibody against Cx43 (1:400 dilution; Chemicon International, Temecula, CA), a mouse monoclonal antibody against Ki67 (1:20 dilution; Dako, Glostrup, Denmark), and/or a mouse monoclonal antibody against anti-p27kip1 (1:200 dilution; Sigma-Aldrich, St. Louis, MO). They were incubated with the anti-Ki67 antibody for 48 hours and with the other antibodies overnight, at 4°C. After the pieces were rinsed in PBS three times for 5 minutes each, they were incubated for 90 minutes at 37°C with appropriate secondary antibodies, Alexa-488- or Alexa-568-conjugated antibodies to mouse or rabbit IgG (Invitrogen-Molecular Probes, Eugene, OR) at 1:500 dilution. After the final rinses in PBS, some pieces were overlaid with propidium iodide (PI; 1 mg/mL) for 30 minutes at room temperature for nuclear staining. The corneal pieces were mounted on glass slides (Vectashield mounting medium; Vector Laboratories, Burlingame, CA), with the endothelial surface uppermost toward the coverslip. 

For paraffin-embedded corneal cross sections, whole corneoscleral discs were cut into 4 × 3-mm pieces, and then fixed in 1% paraformaldehyde in PBS at room temperature overnight. The pieces were dehydrated in an ethanol series and then treated with xylol and embedded in paraffin, so that the axis of the wound was perpendicular to the bottom of the mold. Corneal cross sections 4 μm thick were cut and then deparaffinized with xylol and an alcohol series. The sections were incubated with a mouse monoclonal antibody against α-SMA (1:5000 dilution; Sigma-Aldrich) overnight at 4°C. After the sections had been rinsed in PBS three times for 5 minutes each, they were incubated with goat anti-rabbit Ig conjugated to a horseradish peroxidase (HRP)-labeled polymer (Envision+ HRP; Dako), for 60 minutes at room temperature. After they were rinsed in Tris-buffered saline (TBS, pH 7.6) three times for 5 minutes each, the sections were colored with 3,3′-diaminobenzidine (DAB) substrate. After a final rinse with distilled water, the sections were counterstained with hematoxylin and mounted. 

Confocal Laser Scanning Microscopy and Image Analysis

These immunolabeled corneas with Alexa-488- or Alexa-568-conjugated secondary antibodies were observed with a confocal laser scanning microscope (Fluoview System; Olympus, Tokyo, Japan). Digital images were collected by adjusting the sensitivity of the photomultiplier according to the fluorescence intensity of the positive control section from nontreated samples in each immunolabeling run, and the fluorescence intensities in the positive control sections were adjusted within the 4096-level scale. Section planes were obtained by using a ×60 objective lens and a zoom setting of 2 (800 × 600 pixels corresponds to 133 × 100 μm) or a ×20 objective lens and a zoom setting of 2 (800 × 600 pixels corresponds to 399 × 300 μm) under identical conditions. Digital images were transferred to a computer equipped with image-analysis software (Photoshop; Adobe Systems Inc., San Jose, CA). 

Complete wound closure of corneas with Cx43-knockdown treatment was evaluated at day 3 after injury using images of the cell–cell boundary labeled with the anti-ZO-1 antibody together with nuclear staining with PI. The composite images 570 × 85 μm in size were used for this evaluation. We deemed that complete wound closure had occurred when the corneal surface was fully covered with endothelial cells without intervening spaces. Altogether six corneas were used for each of Cx43 AS-ODN, Cx43 S-ODN, Cx43 siRNA, and control siRNA. 

In experiments with the recombinant adenovirus containing CMV-Cx43-mRFP, complete wound closure was evaluated on day 5 after injury in images labeled with the anti-Cx43 antibody together with nuclear staining with PI. 

Proliferative activity of the corneal endothelium was evaluated on day 1 after injury in images of double immunolabeling for Ki67 and Cx43. In the 250 × 100-μm areas from the wound edge, the total number of cells was measured with Cx43 immunolabeling, and finally the proliferative activity was expressed as the number of Ki67-positive nuclei per the number of total cells (Ki67 labeling index). 

The effects of Cx43 AS-ODN or siRNA on Cx43 expression and proliferation of normal corneal endothelium without injury were evaluated in images of double immunolabeling for Ki67 and Cx43. The images were converted into 256 grayscale, and for morphometry, the area (280 × 280 μm) was selected from the center of each image. The integrated density of Cx43 immunolabeling in the selected area was quantified on computer (Scion Image; Scion Corp., Frederick, MD). The Ki67 labeling index was calculated as the number of Ki67-positive cells divided by the total number of cells. The percentage of the number of Cx43-negative cells to that of total cells was also quantified. To elucidate the relationship between Ki67-positive cells and Cx43-negative cells, the percentage of Ki67-positive cells among Cx43-negative cells was obtained for each image. 

For image analysis for effects of Cx43 AS-ODN on mesenchymal transition of the corneal endothelium after injury, paraffin-embedded corneal cross sections were used from days 3 and 5 after injury. Determination of sections where the number of myofibroblasts was at its maximum in each cornea and the analysis of their number in such sections were performed as follows: first, serial sections 4 μm thick were cut. Second, every 20th section was stained with hematoxylin and eosin to narrow the choice. Third, sections before and behind the selected (H&E)-stained sections were immunolabeled using the antibody against α-SMA, Envision+ HRP, and DAB. Finally, the number of multilayered α-SMA-positive myofibroblasts in the retrocorneal lesion posterior to Descemet’s membrane were counted in five corneas each with Cx43 AS-ODN and Cx43 S-ODN treatments. The data for morphometric analysis of α-SMA-positive myofibroblasts were expressed as both the number of α-SMA-positive myofibroblasts per retrocorneal lesion and the number of α-SMA-positive myofibroblasts per 100 μm of the linear horizontal distance across the retrocorneal lesion occupied by multilayered α-SMA-positive myofibroblasts. 

Measurement of TGF-β2, cAMP, and FGF-2 in Aqueous Humor after Injury

For Cx43 siRNA and control siRNA, the concentrations of TGF-β2, cAMP, and FGF-2 in aqueous humor on day 1 after injury were measured by using ELISA kits for cAMP (Yamasa Corp., Chiba, Japan), TGF-β2 (R&D Systems, Minneapolis, MN), and FGF-2 (R&D Systems), respectively, according to the manufacturers’ instructions. The results were compared with those of the nontreated control group. Aqueous humor from 18 eyes was used for each group. 

Real-Time RT-PCR Analysis for Cx43 mRNA

The effectiveness of Cx43 knockdown by Cx43 siRNA was quantified with real-time RT-PCR. Briefly, under a stereoscopic microscope, the endothelium together with Descemet’s membrane was carefully peeled off from the cornea and incubated in a solution of RNA stabilization reagent (RNAlater; Qiagen, Tokyo, Japan). Total RNA was then isolated (RNeasy Micro Kit; Qiagen). Forty nanograms of total RNA from the corneal endothelium of each eye was used for quantitative realtime RT-PCR. Real-time RT-PCR experiments were performed with commercial probes (TaqMan probes; Applied Biosystems, Inc. [ABI], Tokyo, Japan) and a PCR system (model 7300; ABI) Total RNAs from 21 corneal endothelial samples (three samples from the nontreated control group, and three samples each from Cx43 siRNA and control siRNA treated for each of 12 hours, 1 day, and 3 days) were analyzed and standardized to values obtained with control 18S rRNA primers. 

Statistical Analysis

Statistical analysis was performed with commercial software (Stat View; SAS Institute Inc., Cary, NC). 

Results

Appearance of Enlarged Endothelial Cells and Multilayered Myofibroblasts after Scrape Injury in the Rat Corneal Endothelium and Drastic Changes in Cx43 Expression in Those Cells

We first asked what kind of cells are involved in the wound-closure process after scrape injury in the rat corneal endothelium in vivo, and what changes including those of Cx43 expression occur in such cells during this process. We found that two different kinds of cell—that is, enlarged corneal endothelial cells and multilayered small, elongated cells, appeared after scrape injury, and that dramatic changes in Cx43 expression occurred in those cells (Fig. 1A) . In the normal, uninjured control corneas, the endothelium showed great uniformity in shape, size, and expression of ZO-1 and Cx43. ZO-1 formed a discontinuous polygonal (mainly hexagonal) pattern, and Cx43 was uniformly expressed as abundant small dots on the plasma membrane between endothelial cells. On the other hand, no α-SMA was found in the endothelium (data not shown). Scrape injury fairly constantly produced a linear wound of approximately 2 × 1 mm. No damage to the peripheral corneal endothelium was apparent after application of this technique. At 3 hours after injury, although the cell shape demonstrated by ZO-1 immunolabeling did not change very much, Cx43 expression was decreased among the endothelial cells around the wound edge (in the third or fourth row from the edge). At 12 hours after injury, small, elongated cells appeared at the border of the wound, whereas enlarged irregular-shaped endothelial cells accompanying the decrease in Cx43 emerged on the periphery of the wound. The increases in the cell size and the extent of reductions in ZO-1 and Cx43, as well as the deformity of cell shapes further proceeded on the periphery of the wound on day 1 after injury, and reached the maximum on day 3. The size of the wound began to decrease on day 1 after injury, and most wounds were closed on day 5. On day 5 after injury, Cx43 expression was found in most of the small, elongated cells at the center of the wound-closure area; some enlarged endothelial cells on the periphery of the wound-closure area were decreased in size, and Cx43 expression reappeared in those endothelial cells. Afterward, Cx43 expression showed gradual recovery, and by day 21 after injury ZO-1 and Cx43 expression among the polygonal cells became comparable to that in the uninjured endothelium, although small, elongated cells were still present at the center of the wound-closure area. Serial confocal images of double immunolabeling for ZO-1 and α-SMA taken at 0.7-μm intervals revealed that multilayered small, elongated cells on day 5 after injury were α-SMA positive (Fig. 1B) , indicating that those cells possess myofibroblast phenotypes. These results suggest that both enlarged endothelial cells and multilayered myofibroblasts derived from endothelial cells via EMT participate in the wound-closure process after scrape injury in the rat corneal endothelium, and that drastic changes in Cx43 expression occur in those cells during wound healing. 

Since EMT-derived myofibroblasts have been suggested to play important roles in formation of retrocorneal fibrous membrane, ¹³ we examined corneal cross-sections after scrape injury by routine hematoxylin and eosin histology and α-SMA immunohistochemistry, to determine whether retrocorneal fibrous membranes are formed in the rat corneal endothelium after scrape injury. We found a retrocorneal fibrous membrane consisting of α-SMA-positive spindle cells arranged in multiple cell layers posterior to Descemet’s membrane at the center of the wound on day 5 after injury (Fig. 1C) . Thus, scrape injury to the rat corneal endothelium induced formation of retrocorneal fibrous membranes via EMT. 

Delivery of Cx43 AS-ODN or siRNA into the Corneal Endothelium In Vivo and Cx43 Protein Knockdown

To confirm penetration of Cx43 AS-ODN or siRNA into the corneal endothelium, we injected FITC-labeled Cx43 AS-ODN or FITC-labeled Cx43 siRNA into the anterior chamber of the eye without injury and observed its fluorescence 1 day after injection. In both cases, clear small dotted labeling could be seen in the endothelium (Fig. 2) , indicating that Cx43 AS-ODN or siRNA can be delivered into the endothelium after injection into the anterior chamber. Cx43 immunolabeling showed that Cx43 expression was downregulated in the endothelium from Cx43 AS-ODN- or siRNA-treated corneas compared with that from nontreated ones (Fig. 2) . 

Acceleration of Wound Closure and Induction of the Proliferative Activity of the Rat Corneal Endothelium after Scrape Injury with Cx43-Knockdown Treatment

We next asked whether Cx43-knockdown treatment can accelerate wound closure in the rat corneal endothelium after scrape injury. We found that a single injection of Cx43 AS-ODN or siRNA accelerated wound closure in the corneal endothelium. ZO-1 immunolabeling and PI nuclear staining demonstrated that complete wound closures occurred in five of six corneas on day 3 after injury, with the injection of either Cx43 AS-ODN or siRNA. In contrast, no complete closure was observed on day 3 in the endothelia of the control corneas (S-ODN; zero of six, or nonsense siRNA, zero of six; Figs. 3A 3B ; Table 1 ). Even in one cornea each from among Cx43 AS-ODN- or siRNA-treated corneas, that we evaluated as showing no complete wound closure, the space without cells was much smaller than that of the control corneas (data not shown). These results show that Cx43-knockdown treatment can accelerate wound closure in the rat corneal endothelium after scrape injury. 

To determine whether Cx43-knockdown treatment induces cell proliferation, we evaluated the proliferative activity of the corneal endothelium on day 1 after injury using immunolabeling for Ki67 as a marker for proliferative cells. ³⁶ We found that Cx43-knockdown treatment increased the proliferative activity of cells after scrape injury. On day 1 after injury, the labeling indexes of Ki67 were significantly higher in the corneas with a single injection of Cx43 AS-ODN (Figs. 4A 4C)or siRNA (Figs. 4B 4D)than those in the control corneas. Morphologic analysis of Ki67-immunolabeled cells on day 1 after injury revealed a clear difference between the phenotypes of Ki67-positive proliferating cells in the Cx43-knocked-down endothelium and in the control corneas. That is, in the Cx43-knocked-down endothelium, the small, elongated cells that appeared at the border of the wound and the enlarged irregular-shaped endothelial cells on the periphery of the wound were Ki67 positive, whereas only small, elongated cells at the border of the wound were Ki67 positive in the control corneas. These results indicate that Cx43-knockdown treatment enhances cell proliferation of enlarged endothelial cells. 

To understand the mechanism by which Cx43 knockdown releases corneal endothelial cells from G₁-phase arrest, we performed immunostaining for p27kip1 in the corneal endothelium on day 1 after injury. We found that Cx43 knockdown markedly decreased the expression of p27kip1 in enlarged irregular-shaped endothelial cells on the periphery of the wound (Fig. 4E) . These results indicate that downregulation of p27kip1 may be important in corneal endothelial cell proliferation induced by Cx43 knockdown. 

To determine whether Cx43 knockdown reduces the release of antiproliferative growth factors into the aqueous humor, which leads to corneal proliferation, we measured the concentrations of TGF-β2 and cAMP in the aqueous humor on day 1 after injury. There were no significant differences among the concentrations of TGF-β2 in the aqueous humor from Cx43 siRNA-treated and control siRNA-treated eyes at day 1 after injury and from nontreated control eyes (Fig. 5A) . However, the concentration of cAMP was increased in the aqueous humor from Cx43 siRNA-treated eyes compared with those from control siRNA-treated and nontreated control eyes (Fig. 5B) . We also measured the concentration of FGF-2 in the aqueous humor on day 1 after injury. There were no significant differences among the concentrations of FGF-2 in the aqueous humor from Cx43 siRNA-treated and control siRNA-treated eyes at day 1 after injury and from nontreated control eyes (Fig. 5C) . 

Effects of Cx43 Overexpression by Adenovirus Containing CMV-Cx43-mRFP1 on Wound Healing of the Corneal Endothelium

To confirm further that Cx43 plays a role in wound healing in the corneal endothelium, we attempted to overexpress Cx43 by using adenovirus containing CMV-Cx43-mRFP1 and determined the effect on wound closure. We found that a single injection of adenovirus containing CMV-Cx43-mRFP1 increased Cx43 expression (Figs. 6A 6B)and delayed wound closure (Fig. 6B) . Complete wound closures occurred only in only two corneas of five on day 5 after injury with the injection of the adenovirus, whereas complete closure was observed on day 5 after injury in all (5/5) of the control corneas (Table 2) . These results indicate that Cx43 plays a role in wound healing in the corneal endothelium. 

Induction of Proliferative Activity of the Rat Corneal Endothelium In Vivo without Injury by Cx43 Knockdown

In wound healing, since the endothelium is already released from contact inhibition by the injury, it is difficult to measure direct effects of Cx43-knockdown treatment on endothelial cell proliferation. We hypothesized that Cx43-knockdown treatment can induce cell proliferation independently from the effect of loss of contact inhibition. To test this hypothesis, we treated the endothelia of rat corneas with Cx43 AS-ODN or siRNA without injury and analyzed the proliferative activity of the endothelium by immunolabeling for Ki67. We found that treatment with Cx43 AS-ODN or siRNA knocked down Cx43 expression and induced the proliferative activity of corneal endothelium, even without injury. Immunolabeling (Figs. 7A 7B)and subsequent morphometry of Cx43 fluorescence (Figs. 7C 7D)demonstrated that treatment with Cx43 AS-ODN or siRNA decreased Cx43 expression starting at 6 hours, with the maximum decrease at 12 hours and continuing until at least day 3. Real-time RT-PCR analysis confirmed that Cx43 mRNA was also knocked down at 12 hours after the beginning of Cx43 siRNA treatment, and this knockdown further continued and proceeded until day 3 of the treatment (Fig. 7G) . In addition to the decrease in Cx43 fluorescence, we found Cx43-negative cells in the corneas on days 2 to 3 after Cx43 AS-ODN or siRNA treatment (Figs. 7A 7B) . Concerning proliferative activity, although Ki67-positive endothelial cells were observed neither in the control corneas nor at 12 hours or at day 1 after Cx43 AS-ODN or siRNA treatment, a small number of Ki67-positive cells appeared on day 2 and the Ki67-labeling indexes were increased on day 3 after the knockdown treatment (Figs. 7A 7B 7E 7F) . A direct comparison between Cx43-negative cells and Ki67-positive cells revealed that all Ki67-positive cells were Cx43-negative and that the percentage of Ki67-positive cells among Cx43-negative cells increased from day 2 to day 3 after Cx43 AS-ODN or siRNA treatment (Figs. 7E 7F) . Thus, Cx43-knockdown treatment can induce proliferation of rat endothelial cells in vivo, even without injury. 

Inhibition of EMT and Formation of Retrocorneal Fibrous Membrane in the Rat Cornea after Injury by Cx43 Knockdown

We finally asked whether Cx43-knockdown treatment affects the formation of retrocorneal fibrous membrane via EMT after scrape injury. We found, on day 3 after injury, more α-SMA-positive myofibroblasts in Cx43 AS-ODN-treated corneas (n = 5) than in S-ODN-treated corneas (n = 5; Figs. 8A 8B 8C ). However, on day 5 after injury, the number of myofibroblasts in Cx43 AS-ODN-treated corneas was less than half of that in S-ODN-treated ones (Figs. 8A 8B 8C) . The number of myofibroblasts in the Cx43 AS-ODN-treated corneas was not significantly different at days 3 and 5 after injury, and the number of myofibroblasts in Cx43 AS-ODN-treated corneas on day 3 after injury was also less than half of that in S-ODN-treated ones on day 5. In other words, although a retrocorneal fibrous membrane consisting of α-SMA-positive myofibroblasts arranged in multiple cell layers was found posterior to Descemet’s membrane on day 5 after injury in all corneas studied (100%) in both the S-ODN- and AS-ODN-treated groups, Cx43 knockdown reduced the number of myofibroblasts and reduced the retrocorneal fibrous membrane. These results indicate that Cx43-knockdown treatment reduces the formation of retrocorneal fibrous membrane via EMT after scrape injury to the corneal endothelium. 

Discussion

Rat corneal endothelial cells in vivo are arrested in the G₁ phase of the cell cycle and are actively maintained in a nonproliferative state. After injury, cells with endothelial phenotypes can proliferate, but their proliferation is not sufficient to close the large wound. Thus, myofibroblasts that are possibly transformed from endothelium via EMT proliferate and form retrocorneal fibrous membrane, leading to loss of corneal clarity. 

This study shows that expression of Cx43 in the rat corneal endothelium in vivo drastically changes after injury and that Cx43-knockdown treatment enhances wound healing but inhibits EMT. In our experiment, a single application of Cx43 AS-ODN or siRNA into the anterior chamber simultaneously with injury accelerated wound closure and increased the number of Ki67-positive proliferating cells in the corneal endothelium. In contrast, we found that Cx43 AS-ODN treatment decreased the number of α-SMA-positive myofibroblasts in the corneal endothelium in vivo on day 5 after injury. Furthermore, even without injury, Cx43-knockdown treatment induced Ki67-positive proliferating endothelial cells in vivo. These results indicate that Cx43 knockdown may be a new therapeutic approach, not only for acceleration of wound closure but also for prevention of fibrosis in the corneal endothelium by inhibiting EMT. 

We knocked down the expression of Cx43 in the rat corneal endothelium in vivo by direct injection of Cx43 AS-ODN or siRNA into the anterior chamber. Using FITC-labeled Cx43 AS-ODN or siRNA, we showed that Cx43 AS-ODN or siRNA can be delivered into the endothelium after the injection. Real-time RT-PCR for Cx43 mRNA and morphometry for Cx43 immunofluorescence demonstrated that Cx43 expression was significantly knocked down by this method. These findings are in disharmony with the general belief that a much more concerted effort, involving methods such as electroporation, hydrophobic transfection agents, or adenoviruses, is needed to transfect cells with AS-ODN or siRNA. One of the reasons that we succeeded in knock down of Cx43 expression may be the unusually high concentration of Cx43 AS-ODN or siRNA that we used. Combined with the markedly short half-life of Cx43 proteins (1.3 hours) in the rat heart, ³⁷ rapid and continuous reduction of Cx43 mRNA by knockdown could explain the very early time points for initial and maximum knockdown of Cx43 protein observed in this study. 

Our present results showing that Cx43-knockdown treatment accelerated wound closure in the rat corneal endothelium in vivo are in accord with the finding that wound closure is accelerated in the mouse epidermis with reduced Cx43 levels. ^{26 27 33} The present study, which confirms the efficacy of Cx43 knockdown for wound treatment, presents the possibility of broad application of Cx43-knockdown treatment for wound healing of the corneal endothelium. Furthermore, our results showing that overexpression of Cx43 by adenovirus containing CMV-Cx43-mRFP1 delayed wound closure indicate that Cx43 plays an important role in wound healing of the corneal endothelium in vivo. 

Several different mechanisms of acceleration of wound closure by Cx43-knockdown treatment could be considered. First, Cx43 knockdown can enhance cell proliferation and subsequently accelerates wound closure. We found that on day 1 after injury, Cx43-knockdown treatments significantly increased the number of Ki67-positive proliferating cells. Similar results have been reported by Coutinho et al. ³³ for mouse epidermal thermal injury (i.e., Cx43 AS-ODN treatment augmented the number of Ki67-positive cells on days 4 and 7 after injury). Our present study further showed that Cx43-knockdown treatment induced proliferation of enlarged irregular-shaped endothelial cells on the periphery of the wound on day 1 after injury in addition to proliferating small, elongated cells at the border of the wound, whereas only small, elongated cells at the border were immunolabeled with Ki67 in the control corneas. 

Enhancing effects of Cx43 knockdown on cell proliferation during wound healing have been further supported by the present results, revealing that even without injury, Cx43-knockdown treatment circumvents the nonproliferative, G₁-phase–arrested state of corneal endothelial cells and induces proliferation of these cells in vivo. Furthermore, a direct comparison between Ki67-positive cells and Cx43-negative cells after Cx43-knockdown treatment without injury indicated that downregulation of Cx43 precedes cell proliferation and that Cx43 downregulation might be a prerequisite for overcoming the contact inhibited, G₁-phase–arrested state of corneal endothelial cells in vivo. 

In contrast to the proliferative effect of Cx43 knockdown on the uninjured corneal endothelium, mechanisms other than release from contact inhibition must be taken into account in regard to the enhancing effects of Cx43 knockdown on cell proliferation after injury. This is because, in wound healing, the endothelium is already released from contact inhibition by injury, and various growth factors, cytokines, and low-molecular-weight compounds, such as TGF-β and FGF, are released from the aqueous humor, cells, and extracellular matrix; some of these substances have inhibitory effects on corneal endothelial cell proliferation during wound healing, whereas the others have the opposite effects. ³⁸ Cx43-knockdown treatment may modulate such inhibitory or enhancing effects of the compounds and stimulate corneal endothelial proliferation, resulting in acceleration of wound closure. However, at least concerning TGF-β2 and cAMP, our results showed that their concentrations in aqueous humor on day 1 after injury were not decreased in Cx43 siRNA-treated eyes compared with control siRNA-treated or nontreated control eyes. Instead, the concentration of cAMP was increased compared with control siRNA-treated or nontreated control eyes. The concentration of FGF-2 in aqueous humor on day 1 after injury was not increased in Cx43 siRNA-treated eyes compared with control siRNA-treated or nontreated control eyes. From these results, it seems reasonable to suppose that Cx43 knockdown induces corneal endothelial proliferation through mechanisms other than reductions in the release of TGF-β2 and cAMP or increase in the release of FGF-2 into aqueous humor. 

Our present results showed that the time courses of appearance of Ki67-positive endothelial cells after Cx43-knockdown treatment were different in corneas with and without injury. Ki67-positive endothelial cells were present on day 1 after the knockdown treatment applied together with injury. In contrast, Ki67-positive cells were barely detected among the endothelial cells on day 1 after the knockdown treatment applied without injury but were present on day 2 after the treatment. Thus, mechanisms other than release from contact inhibition might also work in induction of cell proliferation after injury with Cx43-knockdown treatment. 

Our data demonstrating that Cx43 knockdown markedly decreases the expression of p27kip1 are in good agreement with the important antiproliferative role of p27kip1 in the corneal endothelium reported by different laboratories. ^{5 6 7 8} Our data also coincide with the suppressive role of Cx43 in cell proliferation through upregulation of p27kip1 reported by Zhang et al. ³⁹ Further studies are needed to clarify the detailed mechanisms by which Cx43 knockdown downregulates the expression of p27kip1. 

Our present results showed that α-SMA-positive myofibroblasts are involved in wound healing of the rat corneal endothelium after scrape injury and that retrocorneal fibrous membranes, which cause loss of corneal clarity and visual acuity, are formed after injury. Because the cornea is avascular, EMT appears to play a major role in the emergence of α-SMA-positive myofibroblasts in the corneal endothelium after scrape injury. We found that on day 5 after injury, the number of myofibroblasts in Cx43 AS-ODN-treated eyes was less than half of that in the control eyes. Furthermore, although more α-SMA-positive myofibroblasts were found in Cx43 AS-ODN-treated eyes than in the control eyes on day 3 after injury, the number of myofibroblasts in these eyes on day 3 was still less than half that in the control eyes on day 5, and no increases in their number were observed from day 3 to day 5. These results indicate that Cx43-knockdown treatment reduces the appearance of α-SMA-positive myofibroblasts via EMT and could be expected to have beneficial effects on the corneal endothelium after injury by preventing corneal fibrosis. 

Although we have no detailed experimental data on the mechanism by which Cx43-knockdown treatment inhibits EMT during wound healing of the rat corneal endothelium, several possibilities can be considered. One is that Cx43-knockdown treatment may indirectly reduce the need for myofibroblasts generated via EMT for wound closure by enhancing the proliferation of enlarged endothelial cells. This hypothesis arose because in Cx43-knocked-down corneas, Ki67-positive enlarged endothelial cells appeared on day 1 after injury, most wounds were closed on day 3, and no increase in the number of myofibroblasts was observed between days 3 and 5. The finding on day 3 after injury that more α-SMA-positive myofibroblasts were present in Cx43 AS-ODN-treated eyes than in the control eyes also shows that Cx43-knockdown treatment does not completely inhibit EMT, suggesting that its effect on EMT is indirect. 

Although the indirect effects of Cx43-knockdown treatment on EMT are probably the main mechanism, we cannot absolutely exclude the other possibility—that Cx43-knockdown treatment directly represses EMT after injury, resulting in decreases in the number of myofibroblasts. This view is based on several findings. First, a direct connection to each other through gap junctions is one of the morphologic characteristics of myofibroblasts. ⁴⁰ Gap junctions and Cx43 have been identified between myofibroblasts that are derived from corneal fibroblasts. ⁴¹ Thus, gap-junctional intercellular communication through Cx43 may play an important role in differentiation and function of myofibroblasts, which would be impaired by Cx43-knockdown treatment. Second, it has been shown that gap junctional intercellular communication via Cx43 mediates TGF-β activation and endothelial-induced mural cell differentiation. ⁴² Third, Dai et al. ⁴³ recently reported that Cx43 mediates TGF-β signaling through competitive Smads binding to microtubules. Given the fact that TGF-β signaling pathways play important roles in EMT in the cornea, ¹⁶ Cx43-knockdown treatment possibly has direct inhibitory effects on EMT after injury. 

It might be possible that cell types other than the corneal endothelium are being affected by Cx43 knockdown, because the barrier between the endothelial cells becomes leaky during wound healing and therefore could allow Cx43 AS-ODN and siRNA to make their way into the posterior stroma, affecting resident cells such as keratocytes and lymphocytes, as well as recruited inflammatory cells, including neutrophils, after injury. We cannot completely rule out this possibility, because our imaging method for FITC-labeled Cx43 AS-ODN or siRNA using corneal surface preparations lacks the capacity to image the posterior stroma; therefore, we do not know whether Cx43 AS-ODN or siRNA enters cells residing in the stroma. We also cannot totally exclude the possibility that Cx43 AS-ODN and siRNA may affect other cell types in the anterior chamber, including cells within the iris, ciliary body, and lens. This would disrupt the normal function of these cells, leading to the release of unknown factors into the aqueous humor that might in turn affect endothelial cell growth. 

One concern might be that Cx43 knockdown induces excessive endothelial proliferation, leading to problems, particularly disruption of aqueous humor outflow. However, we observed no excessive endothelial proliferation in Cx43 AS-ODN-treated corneas on day 5 after injury, in the same corneas as those used for analysis of α-SMA-positive myofibroblasts. Yet, the long-term outcome after Cx43-knockdown treatment remains to be clarified. 

In conclusion, in our study, Cx43-knockdown treatment enhanced wound healing but inhibited EMT after injury to the rat corneal endothelium in vivo. The present results indicate that Cx43 knockdown might be a new therapeutic approach not only for acceleration of wound closure but also for prevention of retrocorneal fibrous membrane formation. 

 Supported by grants-in-aid for scientific research from the Ministry of Education, Culture, Sports, Science and Technology, Japan.

 Submitted for publication February 28, 2007; revised July 30 and September 10, 2007; accepted November 14, 2007.

 Disclosure: Y. Nakano, None; M. Oyamada, None; P. Dai, None; T. Nakagami, None; S. Kinoshita, None; and T. Takamats u, None

 The publication costs of this article were defrayed in part by page charge payment. This article must therefore be marked “advertisement” in accordance with 18 U.S.C. §1734 solely to indicate this fact.

 

Figure 1.

 

Changes in the rat corneal endothelium in vivo during wound healing after mechanical scrape injury with a 30-gauge needle. (A) Double immunolabeling for ZO-1 and Cx43 of a surface preparation (wounds on the left) during wound healing (3 hours to 21 days after injury) shows both enlarged corneal endothelial cells ( ) and multilayered small, elongated cells (arrowheads) and drastic changes in Cx43 expression in those cells. (B) Serial confocal images (1–9, from the surface to the deeper region) of double immunolabeling for ZO-1 and α-SMA taken at 0.7-μm intervals on day 5 after injury reveal that multilayered, small, elongated cells were α-SMA-positive myofibroblasts. (C) Immunohistochemistry for α-SMA using a cross section on day 5 after injury shows formation of a retrocorneal fibrous membrane consisting of α-SMA-positive myofibroblasts arranged in multiple cell layers posterior to Descemet’s membrane. Nuclei were stained with hematoxylin. Scale bars: (A, B) 20 μm; (C) 100 μm.

Figure 2.

 

Cx43 AS-ODN or siRNA injected into the anterior chamber of the rat eye was delivered into the corneal endothelium in vivo and downregulated the Cx43 protein. A single injection (40 μM, 20 μL) of FITC-labeled Cx43 AS-ODN or FITC-labeled Cx43 siRNA into the anterior chamber was performed without injury. Cx43 immunolabeling was performed with the anti-Cx43 antibody and Alexa 594-labeled secondary antibody. FITC and Alexa 594 were observed 1 day after injection in a surface preparation by confocal microscope. Clear, small, punctate FITC labeling was seen in the endothelium. Substantial reductions in Cx43 immunolabeling were found in Cx43 AS-ODN- or siRNA-treated corneas compared with the control corneas. Scale bar, 20 μm.

Figure 3.

 

Cx43 AS-ODN or siRNA accelerated wound closure of the rat corneal endothelium after scrape injury. A single injection (40 μM, 20 μL) of Cx43 AS-ODN or siRNA into the anterior chamber was applied simultaneously with injury. (A, B) Immunolabeling for ZO-1 and nuclear staining with PI of a surface preparation on day 3 after injury (wounds at the center) reveal that the corneal surface with the single application of either AS-ODN (A) or siRNA (B) was fully covered with endothelial cells without intervening spaces, indicating complete wound closure. In contrast, spaces without cells ( ) were observed in the control eyes S-ODN or nonsense siRNA (control siRNA), demonstrating that closure was not complete. Scale bar, 20 μm.

Table 1.

 

Acceleration of Postinjury Wound Closure in the Rat Corneal Endothelium by Cx43 AS-ODN or siRNA Treatment

Treatment	Complete Wound Closure on Postinjury Day 3
Cx43 S-ODN	0/6
Cx43 AS-ODN	5/6*
Control siRNA	0/6
Cx43 siRNA	5/6, †

 A single injection (40 μM, 20 μL) of Cx43 AS-ODN or siRNA into the anterior chamber was applied simultaneously with scrape injury to the corneal endothelium with a 30-gauge needle. Complete wound closure was evaluated at postinjury day 3 in composite images (570 × 85 μm, as shown in Fig. 2 ) of the cell-cell boundary labeled with the anti-ZO-1 antibody together with nuclear staining with PI. Complete wound closure was deemed to have occurred when the corneal surface was fully covered with endothelial cells without an intervening space.

*   P < 0.01 versus Cx43 S-ODN.

†   P < 0.01 versus control siRNA (χ² test).

Figure 4.

 

Cx43 AS-ODN or siRNA increased the proliferative activity of the rat corneal endothelium after scrape injury. (A, B) Double immunolabeling for Ki67 and Cx43 in a surface preparation on day 1 after injury (wounds on the left) revealed that not only small, elongated cells at the border of the wound (arrowheads), but also enlarged irregular-shaped endothelial cells ( ) on the periphery of the wound were Ki67-positive in the corneas treated with AS-ODN (A) or siRNA (B). In contrast, only small, elongated cells at the border of the wound were immunolabeled with Ki67 in the control eyes (S-ODN or control siRNA). (C, D) Ki67-labeling indexes in the corneas on day 1 after a single injection of AS-ODN (C) or siRNA (D) simultaneously with injury are shown. Mean ± SD; n = 5. *P < 0.001 versus control eyes (S-ODN or control siRNA); Student’s t-test. (E) Double immunolabeling for p27kip1 and Cx43 using surface preparations on day 1 after injury (wounds on the left) shows that Cx43 siRNA downregulated both p27kip1 and Cx43 expression. Scale bar, 20 μm.

Figure 5.

 

Measurement of the concentrations of TGF-β2 (A), cAMP (B), and FGF-2 (C) in the aqueous humor from Cx43 siRNA-treated and control siRNA-treated eyes at day 1 after injury and from nontreated control eyes. Mean ± SD; n = 3. †P < 0.05 compared with control or control siRNA; NS, not significant compared with control or control siRNA (Scheffé’s test).

Figure 6.

 

Cx43 overexpression by adenovirus containing CMV-Cx43-mRFP1 delayed wound closure of the rat corneal endothelium. (A) Cx43-mRFP1 fluorescence without fixation or any staining on day 3 after injection of 10 μL solution of recombinant adenovirus containing CMV-Cx43-mRFP (5 × 10⁸ pfu/mL) into the anterior chamber without injury. Small, punctate Cx43-mRFP was seen between the endothelial cells. (B) Immunolabeling for Cx43 and nuclear staining with PI of a surface preparation on day 5 after injury (wounds at the center) revealed spaces without cells ( ) in the cornea treated with adenovirus containing CMV-Cx43-mRFP1, demonstrating that wound closure was not complete. In contrast, the control corneal surface was fully covered with endothelial cells without intervening spaces, indicating complete wound closure. Overexpression of Cx43 was seen in the corneal endothelium treated with adenovirus containing CMV-Cx43-mRFP1 compared with the control. Scale bar, 20 μm.

Table 2.

 

Delay in Postinjury Wound Closure in the Rat Corneal Endothelium by Recombinant Adenovirus Containing CMV-Cx43-mRFP1

Treatment	Complete Wound Closure on Postinjury Day 5
Control	5/5
Adenovirus containing CMV-Cx43-mRFP1	2/5*

 A single injection (10 μl) of recombinant adenovirus containing CMV-Cx43-mRFP1 (5 × 10⁸ pfu/mL) into the anterior chamber was applied simultaneously with scrape injury to the corneal endothelium with a 30-gauge needle. Complete wound closure was evaluated at postinjury day 5 using composite images (570 × 85 μm, as shown in Fig. 6 ) of the cell-cell boundary labeled with the anti-Cx43 antibody together with nuclear staining with PI. Complete wound closure was deemed to have occurred when the corneal surface was fully covered with endothelial cells without an interrupting space.

*   P < 0.05 versus control (χ² test).

Figure 7.

 

Cx43 AS-ODN or siRNA knocked down Cx43 expression and induced the proliferative activity of the rat corneal endothelium in vivo without injury. A single injection (40 μM, 20 μL) of Cx43 AS-ODN or siRNA into the anterior chamber was applied without injury. (A, B) Double immunolabeling for Ki67 and Cx43 of a surface preparation revealed that Cx43 AS-ODN (A) or siRNA (B) decreased Cx43 expression from 12 hours until day 3. A few Ki67-positive endothelial cells were present on day 2, and their number increased on day 3 after Cx43 AS-ODN or siRNA treatment. Cx43-negative cells (arrowheads) were found in the cornea on days 2 to 3 after Cx43 AS-ODN or siRNA treatment. Scale bar, 20 μm. (C, D) Morphometry for Cx43 fluorescence showed that Cx43 AS-ODN (C) or siRNA (D) decreased Cx43 expression starting at 6 hours, reaching a maximum decrease at 12 hours, and continuing until at least day 3. Mean ± SD; n = 2 to 16 for various time points. (E, F) Morphometry of Cx43-negative cells and Ki67-positive cells and the ratio of Ki67-positive cells to Cx43-negative cells revealed that no Ki67-positive endothelial cells were present at either 12 hours or day 1 after Cx43 AS-ODN or siRNA treatment, but that the number of Cx43-negative cells and Ki67-positive cells increased from days 2 to 3, that all Ki67-positive cells were Cx43-negative, and that the ratio increased from days 2 to 3. Mean ± SD; n = 4 to 16 for various time points. †P < 0.01; *P < 0.001; Welch’s t-test. (G) Real-time RT-PCR analysis for Cx43 mRNA in the endothelium and Descemet’s membrane from Cx43 siRNA-treated, control siRNA-treated, and nontreated (control) corneas. Mean ± SD; n = 3. †P < 0.01 compared with control or control siRNA; ‡P < 0.05; NS, not significant (Scheffé’s test).

Figure 8.

 

Cx43 AS-ODN treatment inhibits EMT and formation of retrocorneal fibrous membrane in the rat cornea after scrape injury. (A) Immunohistochemistry for α-SMA in cross sections on days 3 and 5 after injury revealed that on day 3, the wound was closed in the AS-ODN-treated cornea but remained open in the S-ODN-treated one. On day 5, the wounds were completely closed in both corneas. No increases in the number of α-SMA-positive myofibroblasts between days 3 and 5 were found in the AS-ODN-treated cornea, whereas the number of myofibroblasts in the S-ODN-treated cornea was markedly increased during the same period. Scale bar, 100 μm. (B, C) Morphometry for α-SMA-positive myofibroblasts. The data are expressed as the number of α-SMA-positive myofibroblasts per retrocorneal lesion (B) and the number of α-SMA-positive myofibroblasts per 100 μm of the linear horizontal distance across the retrocorneal lesion occupied by multilayered α-SMA-positive myofibroblasts (C). On day 5 after injury, the number of myofibroblasts in Cx43 AS-ODN-treated corneas was less than half that in S-ODN-treated corneas. Although more α-SMA-positive myofibroblasts were found in Cx43 AS-ODN-treated corneas than in the S-ODN-treated corneas on day 3 after injury, the number of myofibroblasts in Cx43 AS-ODN-treated corneas on day 3 was still less than half that in the S-ODN-treated corneas on day 5, and no increases in their number were observed from days 3 to 5. Mean ± SD; n = 5. †P < 0.01; ‡P < 0.05 (Scheffé’s test).