Abnormal Connexin Expression Underlies Delayed Wound Healing in Diabetic Skin

All of the rats were supplied from a longstanding breeding colony in University College London's Biological Services Unit. (The original stock came from Harlan [Loughborough, U.K.]). Animals were maintained according to the U.K. Home Office Animals (Scientific Procedures) Act 1986 Code of Practice. They were housed in 12-h light and dark cycles in a climatically controlled room with free access to water and food. In total, 89 rats were used in these studies. Diabetes was induced in adult male Sprague-Dawley rats (350–400 g) by a single intraperitoneal injection containing 65 mg/kg STZ in citrate buffer (n = 6 diabetic and 6 control animals per time point) (15). Diabetes was confirmed 3–5 days after STZ injection using urinary glucose strips (Clinistix; Bayer, Newbury, U.K.). Specific blood glucose readings were also taken at the time of killing, and all STZ-injected rats were found to be severely hyperglycemic (diabetic rat blood glucose level 27.07 ± 1.09 mmol/l). Most diabetic wound-healing studies have been carried out 2 weeks after diabetes induction (16,17); this time point was used for our wound-healing study. However, we also examined Cx expression in the skin of diabetic rats at 8 weeks (n = 6 diabetic and 6 control animals) to confirm that the changes we had detected at 2 weeks remained the same. Normal back skin was excised, cryosectioned, and immunostained for Cxs; imaged by confocal microscopy; and the staining quantified as described previously (18).

Rats were anesthetized with 5% halothane by inhalation and maintained with 1.5% halothane. Their backs were shaved and then wiped with 70% ethanol. Two pairs of 5 × 5-mm full-thickness excisional wounds were made, and 50 μl of 10 μmol/l Cx43-specific antisense ODN in 30% Pluronic F-127 gel (Sigma, Poole, U.K.) was applied to one wound and a matched dose of control (sense) ODN gel to the other (5). Cx43 antisense ODN, sequence GTA ATT GCG GCA GGA GGA ATT GTT TCT GTC and control sense ODN, sequence GAC AGA AAC AAT TCC TCC TGC CGC AAT TAC, were purchased from Sigma-Genosys (Poole, U.K.). Both ODN stocks and 30% Pluronic gels were dissolved in diethylpyrocarbonate-treated water before mixing. The ODNs are very stable inside the gel but have a half-life of ∼30 min when in contact with sera or inside a cell. Slow release of the ODNs from the gel provides a sustained delivery of the antisense over time, which has proved to be a very effective delivery method (5,7). The ODN-gel mixtures and pipette tips were kept on ice until use to keep the gel in liquid form. Tissue was harvested on days 1, 2, 5, 10, and 15 after wounding (n = 6 diabetic and 6 control rats per time point) and cryosectioned at a thickness of 10 μm in preparation for Cx immunohistochemistry or hematoxylin and eosin (H&E) staining (8).

Intercellular communication was assessed by applying 10 μl of a 4% solution of Lucifer Yellow CH (Sigma) in a pledget of gelfoam into a fresh, full-thickness skin incision (4). Dye was allowed to diffuse for 5 min before removal of the gelfoam and fixation of the tissue in 4% fresh paraformaldehyde for 30 min. Several control studies were also included: a 10-kDa fluorescein isothiocyanate dextran, which will enter injured cells but not pass through gap junctions, was applied to the wound as described above and was found only in the injured cells of the wound margin. Application of the gap junction blocker octanol to a wound 5 min before applying Lucifer Yellow resulted in failure of the Lucifer Yellow to spread beyond the wound margin. In addition, a nonfixable form of Lucifer Yellow (Lucifer Yellow methoxycarbonyl, M-1341; Molecular Probes, Eugene, OR) revealed a similar pattern of dye spread to that of Lucifer Yellow CH. Tissues were cryosectioned and imaged on a Leica SP2UV confocal microscope (Leica, Milton Keynes, U.K.).

Dye spread and Cx immunostaining were examined using the confocal microscope. Optimal gain and offset were set in advance and kept constant during image acquisition. A series of single optical section images was taken to generate a montage of the skin from the cut. Digital 8-bit images were analyzed using ImageJ software (National Institutes of Health). To assess dye spread, a 1,500 × 30-pixel region-of-interest box was placed with one short side aligned to the cut edge in the mid-dermis and an image intensity graph across the box was generated. A gray-level intensity drop to <50 was arbitrarily taken as the limit to which Lucifer Yellow had traveled. Similarly, in the epidermis, the distance from the cut to where the Lucifer Yellow signal dropped to <50 was recorded. At least three images were analyzed from each animal. Dye-spread data were tested for significance using the Mann-Whitney U test. All values were expressed as mean ± SE, and significance was taken at P < 0.05.

To compare levels of Cx protein, six single optical section images of dermis or epidermis were taken from different sections for intact tissue and for each wound. All parameters of laser power, pinhole, gain/offset, and objective were kept constant across both control and diabetic groups. Cx expression was quantified, as described previously (18), using ImageJ software. A threshold was set to detect gap-junction plaques with minimal background noise and was then kept constant for all images. The number and size of Cx plaques were recorded for each image and expressed per 100 μm length of epidermis or 10,000 μm² area of dermis. This approach has proved much more accurate than Western blot, as it generates information on protein expression at the cellular level. Western blots would be unable to distinguish between epidermal and dermal cells or detect effects of proximity to the wound edge. Using this approach, we were able to quantify Cx levels in keratinocytes in an adjacent zone at the wound edge and in a zone 500 μm away at 1, 2, and 5 days after wounding. At day 5 after wounding, we also imaged an additional zone of the leading edge of the nascent epidermis.

Images of H&E staining were taken using a Leica DMLFS microscope with a DC300F digital camera. Measurements for reepithelialization rate (total length of nascent epidermis as a percentage of the length of the wound bed) have been described previously (5). Comparisons of Cx expression in intact skin, between control and diabetic rats, were made using the Mann-Whitney U test. The effect of diabetes on the dynamic changes in Cx expression during the wound-healing process was assessed using a two-way ANOVA. One-way ANOVA was used to assess the effect of the antisense treatment on the reepithelialization rate.

STZ-induced diabetes in the rats was confirmed in the following week with a urine glucose test. In addition, the blood glucose level was measured when the diabetic rats were killed (average 27.07 ± 1.09 mmol/l). In diabetic skin, we found typically punctate Cx43 immunostaining in the basal layer of the epidermis and, in dermal fibroblasts, hair follicles, blood vessels, and appendages (Fig. 1). However, quantification of staining revealed that in 8-week diabetic skin, Cx43 staining was significantly reduced in the epidermis, in terms of the number of gap junction plaques (Fig. 1A; P = 0.0039, Fig. 1C), although there was no difference in plaque size distribution. Staining for Cx26 in the upper layers of the epidermis was similarly significantly reduced in diabetic epidermis, with no changes in plaque size distribution (P = 0.0039, Fig. 1C). However, this was not a global downregulation of Cxs in response to diabetes, as a contrasting effect was observed in dermal fibroblasts; in these, we observed a distinct increase in the size and number of Cx43-positive plaques (Fig. 1C). Similar effects on Cx43 expression were seen in skin 2 weeks after STZ induction of diabetes (epidermis P = 0.037, dermis P = 0.025; Fig. 1D).

To assess cell-cell communication in diabetic epidermis and dermis, we examined the extent of spread of the gap junction–permeant dye Lucifer Yellow through the tissue in 5 min (4). This revealed significantly reduced spread in the diabetic epidermis (P = 0.0433) but significantly enhanced spread in the diabetic dermis (P < 0.0472; Fig. 2A), consistent with the changes in Cx protein expression. In control experiments, a nonfixable, methoxycarbonyl version of Lucifer Yellow that is only retained during histological processing when it is inside nonpermeabilized cells showed the same distribution properties as Lucifer Yellow CH (Fig. 2B), whereas a gap junction–impermeant 10-kDa fluorescein isothiocyanate dextran, applied in the same way, was confined to the edge of the wound (data not shown). In addition, application of the gap junction blocker octanol to the wound for 5 min before application of Lucifer Yellow CH effectively prevented dye spread (Fig. 2C). These observations suggest that dye spread occurred through gap junctions rather than by diffusion through the extracellular space. Also, consistent with these findings, elevated expression of Cx43 protein and increased communication have been reported in human diabetic fibroblasts (19), and mixed responses of different Cxs to diabetes have been noted in the renal system (20) and bladder (21).

Quantitative data on Cx43 and Cx26 expression in keratinocytes at the wound edge and in an adjacent zone 500 μm away 1, 2, and 5 days after wounding are shown in Fig. 3, together with data for leading-edge keratinocytes after 5 days. Representative micrographs illustrating Cx expression are shown in Fig. 4. We confirmed that reepithelialization is delayed in diabetic skin by measuring the length of nascent epidermis. After 2 days, there was 15.7% closure of wounds in control rats, whereas diabetic wounds were retarded (6.1% closed); the epidermis had only recently adopted a migratory morphology. The quantitative data were analyzed by two-way ANOVA with diabetes and time after injury as factors. This revealed that a highly significant interaction (P = 0.0001) existed between the effect of diabetes and that of time on Cx43 expression in the wound edge, indicating that diabetes was influencing the dynamic changes in Cx43 expression in the wound-edge region associated with wound healing. No such interaction was demonstrated for Cx43 expression in the intact region adjacent to the wound (Fig. 3C). In contrast, diabetes did not significantly affect the dynamic changes in Cx26 expression in the wound-edge region but did significantly interact (P = 0.016) with these responses in the intact adjacent region (Fig. 3D).

One day after wounding, Cx43 expression was reduced in the wound-edge region relative to the adjacent region (Fig. 4A), as has been reported previously (4,8). This has been associated with the change in the keratinocytes to a migratory phenotype. In marked contrast, and surprisingly in view of the decreased expression of Cx43 in intact skin in diabetes, Cx43 expression was greater in the wound-edge region than in the adjacent region in diabetic rats (Fig. 4A), and Cx43 expression in the wound-edge region from diabetic rats was more than double (P < 0.001) that in the wound-edge region from controls (Fig. 3C). Thus, the initial wound-healing response of Cx43 in diabetic rats was the opposite of that observed in controls. Keratinocytes formed a thickened bulb at the wound edge in diabetic tissue, whereas they thinned out before crawling forward to close the wound in controls. After 2 days, Cx43 expression had fallen in the wound edge of diabetic tissue to levels similar to those of controls (Figs. 3C and 4B). There was an increase in Cx26 expression in the first 2 days following wounding in the wound-edge region in both control and diabetic tissue (Figs. 3D and 4D and E). However, the response to injury appeared to be more widespread in diabetes because Cx26 expression in the adjacent region was significantly (P < 0.05) greater than controls after 2 days (Figs. 3D and 4E).

On day 5, reepithelialization was well underway in both diabetic and control groups, although the diabetic wounds were significantly retarded (27.4% closed compared with 41% in controls, P < 0.01). Cx43 expression increased in the wound edge of control but not diabetic tissue (Figs. 3C and 4C), such that levels were significantly higher (P < 0.01) in control wound edge, a pattern similar to that observed in intact skin. Only low levels of Cx43 expression were observed in keratinocytes in the leading-edge region in both control and diabetic tissue. There were no significant differences in the expression of Cx26 in adjacent, wound edge, or leading edge in control or diabetic tissue after 5 days; the previously elevated Cx26 staining in the diabetic wound-edge and adjacent zones had returned to more normal levels (Figs. 3D and 4F).

Reepithelialization was complete in both control and diabetic animals by day 10, when high levels of Cx43 could be seen in all of the layers of the proliferative, excessively thickened epidermis (Fig. 5). Examination of the dynamic changes in Cx43 and Cx26 protein expression in the newly regenerated epidermis, as it underwent revision, revealed a marked delay in its return to normality for diabetic wounds. A gradient of maturation could be seen from the edge to the center of the wound (marked 3 and 4 in Figs. 5Q and 6Q). At day 10, the epidermis was hyperthickened and expressed Cx43 and Cx26 throughout, with higher levels of Cx43 seen in the diabetic epidermis. The nascent epidermis also lacked the laminar structure and restriction of Cx expression of the normal epidermis seen in zone 1 (Figs. 5 and 6). By day 15, wound revision had progressed and the newly formed epidermis had started to thin down and adopt a more laminated appearance. A marked gradient of maturation from the edge to the center of the wound was still evident, and this reflected the gradual restriction of Cx43 expression to the basal layers and Cx26 to the upper layers of the epidermis (Figs. 5 and 6). Throughout the process of maturation, the diabetic skin appeared to be retarded, lagging behind the normal sequence of healing events by several days.

Having observed these abnormal patterns of Cx immunostaining, gap junctional communication and morphological response in diabetic wounds, we then assessed the effects of the abnormal increase of Cx43 protein in diabetic wound-edge keratinocytes by preventing the increase with a Cx43-specific antisense gel, applied to the wound at the time of injury (5,7). This effectively inhibited the increase in Cx43 that was seen with control (sense) gel (Fig. 7A) and allowed the keratinocytes to adopt a thinner, tongue-like profile (Fig. 7B). More importantly, it also improved wound-healing rates, which more than doubled, reaching untreated control levels or exceeding them (Fig. 7G). One day after injury, diabetic wounds showed virtually no epidermal regrowth compared with controls (Fig. 7C and D), but treatment of diabetic skin with the Cx43 antisense ODN induced a regrowth of epidermis to match that of controls (Fig. 7F). Healing rates were even higher in antisense-treated wounds in control rats (Fig. 7E), as would be expected from our previous findings (5,7).

We have demonstrated that the onset of diabetes has a marked effect on the expression levels of Cxs in the skin and that this effect is even more pronounced when the skin is injured. The mechanism whereby diabetes alters the expression of Cxs, with a differential effect on different Cx proteins, is currently unknown. However, hyperglycemia and/or oxidative stress induced by hyperglycemia have been implicated in the development of other complications of diabetes such as neuropathy and angiopathy (3). Both high glucose and oxidative stress in vitro have been shown to alter Cx expression and/or gap junctional communication in a variety of different cell types (22–24), and Cx43 and Cx26 are known to have different promoters (25,26). It should be emphasized that impaired wound healing has been demonstrated in rats as early as 1 week after induction of diabetes, before either neuropathy or angiopathy develops (16,17). Thus, the changes observed here within 2 weeks are likely to be direct consequences of the diabetic condition rather than secondary consequences of long-term diabetes complications.

The effect of hyperglycemia in inhibiting Cx43 expression has been described in a number of cell types in in vitro models, where it has been suggested that the inhibition is due to increased protein kinase C phosphorylation (27,28). Phosphorylation of Cx43 may lead to its removal from the membrane (29), but this explanation cannot account for our observation of reduced levels of Cx26, which does not have any phosphorylation site on its COOH-terminus. Clearly, the effect of diabetes also is not a uniform downregulation of Cxs, as fibroblasts respond by elevating Cx43. Again, the mechanism of action remains to be determined. High levels of Cx43 in diabetic fibroblasts are likely to have a detrimental effect on their migration to form granulation tissue in the wound-healing process because actively downregulating Cx43 in fibroblasts enhances their migration and proliferation (30).

A significant finding in our study was the abnormal upregulation of Cx43 at the epidermal wound edge in diabetes. This has the potential to affect the process of wound closure in various ways. The formation of communication compartments within the regenerating epidermis has been proposed to play a role in wound healing (1,9,31). Compartmentalization could be effectively brought about in normal conditions by expression of Cx26 and removal of Cx43 in leading-edge cells, as these Cxs do not form junctions with one another. Thus, the delay in wound healing in diabetes could reflect the additional time required for Cx43 expression to downregulate to a point where such a compartmentalization can occur. Alternatively, as the C-tail of Cx43 is known to interact with cytoskeletal components or with P120ctn/Rho GTPase, downregulation of Cx43 could be necessary for changing the motility of keratinocytes at the wound edge, enabling them to migrate and close the wound (32). In this regard, it was notable that, 1 day after wounding, when their Cx43 expression was high, diabetic keratinocytes formed a thickened bulb at the wound edge rather than flattening out and starting to crawl forward across the wound.

Persistence of Cx43 at the wound edge has been reported in chronic diabetic ulcers in humans (10). Thus, our observation that the same Cx is abnormally upregulated in diabetic rats is likely to have implications for the delayed wound healing that occurs in diabetic patients. Clearly, the normal, prompt downregulation of Cx43 after wounding is critical for normal healing, because we have shown here that knocking down Cx43 expression by a single application of the Cx43-specific antisense gel at the time of wounding enables the entire process, from initial migration to complete wound closure, to occur at or above a normal rate, despite the continued presence of the diabetic condition. Thus, antisense treatment has considerable potential as a novel approach for the treatment of chronic wounds in diabetic patients.

We thank the Biotechnology and Biological Sciences Research Council for a studentship for C.M.W.; the National Institute of Diabetes and Digestive and Kidney Diseases, National Institute of Neurological Disorders and Stroke, and Juvenile Diabetes Research Foundation International (DK58010) for support of J.L.; and the Association of Medical Research Charities for a project grant to D.L.B.

An abstract of this work was published at the 46th American Society of Cell Biology Annual Meeting, 9–13 December 2006.