Reepithelialization of Diabetic Skin and Mucosal Wounds Is Rescued by Treatment With Epigenetic Inhibitors

Diabetes is an increasingly difficult health care problem due to its impact on several different organ systems (1). The number of adults with diabetes worldwide has quadrupled in the last 40 years to 8.5% of the global population (2). A serious complication of diabetes is impaired wound healing, which has a high degree of morbidity, increased risk of infection, hospitalization, negatively impacts the quality of life, and can significantly increase health care costs (3–5). Diabetes impairs skin wound healing, which can lead to the formation of chronic wounds and limb amputation (3,4,6,7). In the mucosa, diabetes interferes with corneal wound healing (8), increases the susceptibility and severity of gastric wounds (9), and is a frequent comorbidity factor with ulcerative colitis (10). In the oral cavity, diabetes interferes with healing and increases susceptibility to medication-induced osteonecrosis of the jaws (11–13).

An essential component of diabetic wound healing is keratinocyte migration along the leading edge of the wound. Rapid closure of a wound by reepithelialization is critical to cover the wound and prevent infection. It occurs when basal keratinocytes proliferate and migrate from the wound edge (7). Reepithelialization is facilitated by the expression of matrix metalloproteinases (MMPs), which are tightly regulated to optimize migration (14,15). Both inadequate MMP expression and overexpression interfere with reepithelialization. MMP dysregulation occurs in diabetic wounds in vivo and negatively impacts wound reepithelialization (16–23).

There is substantial interest in the regulation of the transcription factor FOXO1 because of its association with diabetes (24–29). FOXO1 plays a role in diabetes by contributing to hyperglycemia by stimulating glucose production and altering lipid metabolism (24–26), which are improved by specific FOXO1 inhibitors or genetic inhibition of FOXO1 (25,26). Increased FOXO1 activity is detected in diabetic cardiomyopathy and is thought to contribute to the pathologic process (28,29). FOXO1 plays a role in early diabetic retinopathy by increasing apoptosis of vascular cells and contributing to abnormal vascular remodeling in diabetic cardiomyopathy (29,30). Furthermore, FOXO1 promotes diabetes-associated bone loss and impairs long-bone fracture healing in diabetic animal models (31,32). FOXO1 activity is enhanced in keratinocytes of diabetic wounds and plays a significant role in diabetes-impaired skin and mucosal wound healing (33,34).

FOXO1 is regulated by posttranslational modifications that affect its capacity to bind to DNA and induce gene expression (35,36). We examined epigenetic changes that regulate FOXO1-target gene interactions that may be therapeutically important, focusing on matrix metallopeptidase 9 (MMP9), due to the critical need to tightly regulate its expression in wound healing (37). In normal human wounds, epigenetic changes in basal keratinocytes support in vitro studies that MMP9 expression is inversely regulated by the ratio of DNA methylation to demethylation. We show for the first time that diabetes-impaired reepithelialization of skin and mucosal wounds can be completely reversed through epigenetic inhibitors that block DNA demethylation or histone methylation. Mechanistically, high glucose (HG) inhibits keratinocyte migration through increased expression of MMP9 that is due to significantly enhanced FOXO1-MMP9 promoter interaction caused by increased DNA demethylation. This is significant because FOXO1 in an HG environment stabilizes the interaction between FOXO1 and RNA polymerase-II (Pol-II) to increase MMP9 expression. Thus, HG levels found in a diabetic environment significantly alter FOXO1 activity that is induced by epigenetic changes. These results may give insight into other pathologic processes associated with diabetic complications, such as impaired wound healing.

Our study investigated whether epigenetic changes played a key role in FOXO1-mediated gene expression linked to diabetes and examined the therapeutic implications in vivo. This is significant since elevated FOXO1 plays an essential role in hyperglycemia and altered lipid metabolism in individuals with diabetes and has been linked to a number of diabetic complications. To investigate this issue, we examined a model of diabetes-impaired wound healing involving FOXO1-induced MMP9 expression. We examined how HG-induced changes in DNA methylation and demethylation were closely associated with FOXO1 consensus binding sites in the MMP9 promoter. These experiments used bisulfite assays to measure modification of CpG dinucleotides and DNA immunoprecipitation (IP) experiments with antibodies specific for 5-methylcytosine (5mC) and 5-hydroxymethylcytosine (5hmC). The impact of FOXO1 on RNA–Pol-II interactions with the MMP9 promoter was determined by chromatin IP (ChIP) assays, with and without FOXO1 knockdown, as well as IP/immunoblot assays to examine the interaction of FOXO1 with RNA Pol-II.

Histone modification was done by measuring specific methylation of H3K4me3, H3K9me3, and H3K27me3 associated with the MMP9 promoter by ChIP with specific antibodies compared with the control antibody. Cause-and-effect relationships linking HG-induced epigenetic changes or HG-altered keratinocyte migration to methylase or demethylase enzymes were assessed using specific inhibitors and RNA interface (RNAi). To investigate the clinical impact of these epigenetic changes on diabetic wound healing, we used specific inhibitors of DNA demethylation and histone methylation in models of mucosal and dermal wound healing. Wounds were analyzed to assess the degree of reepithelialization. Study sizes and end points were selected on the basis of previous experience with the respective animal models. Animals were randomly assigned to treatment groups. Biological replicates used different animals and were sex-matched compared across groups. Investigators were blinded to treatment when analyzing the data. All animal experiments were approved by the University of Pennsylvania Institutional Animal Care and Use Committee. The detailed methods are shown in Supplementary Methods.

All data are shown as means ± SEM. Statistical analysis between two groups was performed using a two-tailed Student t test. In experiments with multiple treatments, differences were determined by one-way ANOVA with the Tukey post hoc test. P < 0.05 was considered statistically significant.

In human skin, basal keratinocytes have low expression of MMP9 that is significantly increased at the leading edge of the wound (Fig. 1A). This coincides with a high ratio of DNA methylation to demethylation (Fig. 1B and C). Upon wounding, this ratio is reversed due to an increase in demethylation and a reduction in methylated DNA (5mC) and is associated with increased MMP9 expression. To understand mechanistically how this ratio affects MMP9 expression, we used inhibitors and knockdown of key enzymes that regulate DNA methylation in human mucosal keratinocytes. Inhibition of DNA methylation increased MMP9 levels in human mucosal keratinocytes in vitro by 170%, indicating that under normal conditions, MMP9 DNA methylation restrains expression of MMP9 (P < 0.05) (Fig. 1D). Knockdown of DNA methylases, DNA methyltransferases (DNMT) 1, DNMT3a, or DNMT3b increased MMP9 levels by 200–300% (P < 0.05) (Fig. 1E). Since FOXO1 is a key regulator of MMP9 expression (18,38), we examined the interaction of FOXO1 with the MMP9 promoter. Consistent with these results, a DNA methylase inhibitor increased FOXO1 binding to the MMP9 promoter by 200% (P < 0.05) (Fig. 1F). Inhibition of DNA demethylation by use of a specific inhibitor or RNAi, in contrast, reduced MMP9 mRNA by 34%, MMP9 protein levels by 40% (P < 0.05) (Fig. 1G–I), and reduced FOXO1 binding to the MMP9 promoter by 30–50% (P < 0.05) (Fig. 1J and K). The effect of ten-eleven translocation (TET) and DNMT inhibition on global levels of 5hmC and 5mC was examined by immunofluorescence (Supplementary Fig. 1). HG increased global 5hmC levels by 646% (P < 0.05) (Supplementary Fig. 1A and B), while 5hmC levels were inhibited by 53% by TET inhibitors under HG conditions (P < 0.05) (Supplementary Fig. 1A and B). Levels of 5mC were further inhibited by 75% by DNMT inhibitors (P < 0.05) (Supplementary Fig. 1C and D). The knockdown efficiency of small interfering (si)DNMT1, siDNMT3a, siDNMT3b, and siTET3 was ∼80–90% (Supplementary Figs. 2 and 3).

Because diabetes is known to interfere with reepithelialization through excessive MMP9 expression, we examined whether HG modulated MMP9 levels in a FOXO1-dependent manner. HG stimulated an increase in MMP9 levels that was blocked by FOXO1 knockdown (P < 0.05) (Fig. 2A–D). The knockdown efficiency of siFOXO1 was ∼90% (Supplementary Fig. 3). Exposure of keratinocytes to HG increased FOXO1 nuclear translocation in human mucosal keratinocytes (P < 0.05) (Fig. 2E and F) and increased MMP9 mRNA levels several fold in primary cultures of human and murine mucosal cells and in primary dermal keratinocytes (P < 0.05) (Fig. 2G and H).

To better understand how HG modulated MMP9, we determined that HG significantly increased RNA Pol-II binding to the MMP9 promoter (P < 0.05) (Fig. 2I). The interaction of RNA Pol-II with the MMP9 promoter was FOXO1 dependent as knockdown of FOXO1 blocked RNA Pol-II–MMP9 promoter interaction (P < 0.05) (Fig. 2I). IP, followed by immunoblot assay, demonstrated that RNA Pol-II interacts with FOXO1 and that the interaction was enhanced threefold by HG (Fig. 2J and K).

To address whether epigenetic changes could account for altered FOXO1-DNA binding, we examined CpG sites (−1,300 base pairs [bp] to the transcription start site) associated with FOXO1 response elements in the MMP9 promoter using JASPAR software and performed bisulfite analysis for methylation status. There are four FOXO1 binding sites in the MMP9 promoter region (−1,050 to −1040 bp, −763 to −753 bp, −683 to −673 bp, and −239 to –229 bp), and three potential regions were identified (Fig. 3A). ChIP assays determined that HG substantially increased FOXO1 interaction with FOXO1 consensus binding sites 1 and 2 but had minimal effect on site 3 (Fig. 3B). High-throughput DNA bisulfite sequencing and analysis were then performed for sites 1 and 2 to identify 5mC and 5hmC changes induced by HG. As shown in lollipop plots, HG reduced 5mC but increased 5hmC of CpG dinucleotides (Fig. 3C). DNA-IP assays also demonstrated that HG inhibited DNA methylation in the MMP9 promoter by 50% but increased DNA demethylation of the MMP9 promoter by threefold (P < 0.05) (Fig. 3D and E). Thus, HG reduces repressive methylation (5mC) and increases permissive demethylation (5hmC) in sites linked to FOXO1 response elements.

Mechanistic studies were undertaken using specific inhibitors and siRNA to DNA methylases and demethylases to further investigate epigenetic changes. Inhibition of DNA demethylation with a TET1- and TET2-specific inhibitor blocked HG-stimulated MMP9 mRNA and protein levels (P < 0.05) (Fig. 4A–C). Knockdown of TET1 or TET2 alone or together or TET3 alone blocked HG-induced MMP9 mRNA, establishing their role in altering MMP9 expression (P < 0.05) (Fig. 4D). HG stimulated more than a threefold increase in TET activity, which was blocked by a TET1/2 inhibitor or TET1/2 RNAi (P < 0.05) (Fig. 4E). Mannitol did not affect MMP9 levels (Fig. 4D), and the knockdown efficiency of siTET1, siTET2, and siTET3 was ∼80–90% (Supplementary Fig. 3).

The impact of DNA demethylation on FOXO1 binding to the MMP9 promoter was examined by ChIP assay. HG stimulated FOXO1 binding to the MMP9 promoter by twofold (P < 0.05) that was blocked by TET3 knockdown (P < 0.05) (Fig. 4F). In contrast to demethylation, DNA methylation is typically inhibitory. Studies in Fig. 3 indicate that HG reduces MMP9 promoter methylation, which was confirmed by DNA-IP and not affected by TET knockdown (Fig. 4G and H). Although HG reduced DNA methylation of the MMP9 promoter, residual methylation served as a brake to HG-stimulated MMP9 mRNA levels since a DNMT-specific inhibitor and knockdown of DNMT3b led to a small but significant 47.5% and 48.6% increase in MMP9 mRNA (P < 0.05) (Fig. 4I and J) and increased FOXO1-MMP9 promoter interactions (P > 0.05) (Fig. 4K). These results indicate that HG reduces DNA methylation of the MMP9 promoter but that residual methylation acts as a brake to limit further hyperexpression. However, changes in DNA methylation status may affect the expression of other genes and indirectly affect FOXO1-regulated MMP9 expression.

Histone methylation can positively and negatively regulate gene expression by the formation of permissive and repressive modifications (39,40). Histone methylation played a role in regulating MMP9 expression under basal conditions. H3K27me3 demethylase inhibition or siRNA knockdown of the H3K9me3 demethylase KDM4A increased MMP9 mRNA by twofold (P < 0.05) (Fig. 5A and B). The knockdown efficiency of siKDM4A was ∼80% (Supplementary Fig. 4). In contrast, inhibition or knockdown of WDR5, which stimulates permissive histone methylation, significantly reduced basal MMP9 at the mRNA and the protein levels (P < 0.05) (Fig. 5C–F). ChIP assays indicated that inhibition or siRNA knockdown of WDR5 reduced FOXO1 binding to the MMP9 promoter by a small but significant amount (P < 0.05) (Fig. 5G–I), indicating a minor positive role of permissive H3K4me3 histone in FOXO1-MMP9 promoter interactions under basal conditions.

HG increased the formation of the permissive histone modification, H3K4me3, by 100% (P < 0.05) (Fig. 6A). Specific inhibition of WDR5 activity or RNAi reduced expression of WDR5, a key component of the histone-methyltransferase complex that catalyzes histone H3 lysine 4 trimethylation (41,42), blocked HG-stimulated MMP9 mRNA levels (P < 0.05) (Fig. 6B and C) and protein levels (P < 0.05) (Fig. 6D and E). Formation of H3K4me3 associated with the MMP9 promoter was stimulated 2.3-fold by HG and dependent on WDR5 (P < 0.05) (Fig. 6F), and WDR5 inhibition or siRNA knockdown blocked HG-stimulated FOXO1 binding to the MMP9 promoter (P < 0.05) (Fig. 6G and H). The knockdown efficiency of siWDR5 was ∼80% (Supplementary Fig. 4). These results are consistent with HG enhancing permissive H3K4me3 levels linked to increased MMP9 expression. However, the WDR5-MLL complex also leads to the formation of H3K4me2 and H3K4me1 (43), which could also affect FOXO1-mediated MMP9 expression.

Trimethylation of histone-3 at the K27 or K9 residues represents repressive modifications (44). HG reduced H3K27me3 by 55% (P < 0.05) (Fig. 6I). Specific inhibition of H3K27me3 demethylase induced a 50% further increase in HG-induced MMP9 mRNA (P < 0.05) (Fig. 6J). HG also caused a >70% reduction in repressive H3K9me3 formation (P < 0.05) (Fig. 6K) that is functionally significant as knockdown of the H3K9me3 demethylase KDM4A increased HG-induced MMP9 levels by 80% (P < 0.05) (Fig. 6L).

The above studies demonstrate that HG causes epigenetic changes that regulate FOXO1-induced MMP9 expression. To determine whether this relationship has translation potential, specific inhibitors of DNA demethylation or histone methylation were examined as treatments for diabetes-impaired reepithelialization. Immunofluorescence with a kerartin14 (K14)-specific antibody, which labels keratinocytes in mucosal and skin epithelium, demonstrated that diabetes significantly enhanced MMP9 expression at the leading edge of the wound epithelium, as has previously been shown in human studies (19,20). Local injection of a specific TET DNA demethylase inhibitor or a specific inhibitor of WDR5 needed for permissive H3K4me3 methylation blocked diabetes-increased MMP9 expression (Fig. 7A). The number of MMP9-positive cells in the leading edge was 3.8-fold and 2.5-fold higher in diabetic wounds, agreeing well with in vitro data (P < 0.05) (Fig. 7B and C). Diabetes-enhanced MMP9 was completely reversed by treatment with these specific inhibitors (P < 0.05) (Fig. 7B and C). In addition, diabetes increased MMP9 mRNA levels in keratinocytes isolated from the healing diabetic skin wounds 150% compared with control wounds (P < 0.05) (Fig. 8J). The number of H3K4me3-positive and 5hmC-positive cells in the leading edge was 4.7-fold and 2.8-fold higher in diabetic skin wounds, respectively (P < 0.05) (Fig. 7D–G). The diabetes-enhanced H3K4me3 and 5hmC were completely reversed by treatment with WDR5 and TET inhibitors (P < 0.05) (Fig. 7D–G).

We examined the impact of DNA demethylation and histone methylation on reepithelialization using an in vitro scratch-wound assay that replicates the events that occur in vivo (33,45). Reepithelialization under normal conditions occurred in 3 days, whereas HG delayed the reepithelialization at each time point measured (Fig. 7H and I). The addition of an MMP9-specific inhibitor significantly accelerated HG impaired reepithelialization (Fig. 7J). TET and WDR5 inhibitors similarly rescued reepithelialization that was impaired by HG (Fig. 7K and L). Keratinocyte migration was also assessed in a transwell assay with cells cultured under various conditions prior to transfer to transwell chambers. HG reduced migration by 50% (P < 0.05) (Fig. 7M and N), while mannitol had no effect (P > 0.05). In accordance with in vivo data, TET1/2 or WDR5 inhibitors blocked the impact of HG and restored migration to normal levels (P < 0.05) (Fig. 7M and N). The addition of exogenous MMP9 blocked the rescue effect of TET1/2 or WDR5 inhibitors on cell migration (P < 0.05) (Fig. 7M and N), directly linking high levels of MMP9 to the effect of the inhibitors.

The effect of blocking DNA demethylation or histone methylation on reepithelialization was assessed grossly and histologically. Diabetes markedly reduced the closure of mucosal wounds, consistent with results obtained in human wounds (3,4,6,7). Inhibitors of DNA demethylation and histone methylation completely rescued the negative effect of diabetes on reepithelialization (Fig. 8A). The area of the diabetic wound 5 days after wounding was twofold larger than the corresponding wound in matched control mice (P < 0.05) (Fig. 8B). Reepithelialization in diabetic mice was restored by treatment with the inhibitors, measured grossly or histologically (P < 0.05) (Fig. 8B–D). It is striking that the length of the newly formed migrating epithelium was reduced almost 80% in the diabetic wounds, which was entirely rescued by local application of TET and WDR5 inhibitor (P < 0.05) (Fig. 8C and D). The capacity of TET and WDR5 inhibitors to improve reepithelialization of diabetic skin wounds was also examined. Similar to results obtained with mucosal wounds, diabetes caused a 100% increase in the size of the skin wounds measured grossly and a 100% larger gap between the wound edges measured histologically (P < 0.05) (Fig. 8E–G). Both parameters were restored by local application of the TET or WDR5 inhibitors. It is striking that the length of newly formed epithelium in the skin wound was reduced 80% by diabetes and completely rescued by the inhibitors (P < 0.05) (Fig. 8H). To better understand the impact of the inhibitors on keratinocytes, epithelial sheets were removed from healing wounds by enzymatic digestion, and mRNA was assessed. Diabetes reduced urokinase plasminogen activator receptor (uPAR) mRNA levels, a marker of migrating keratinocytes, by 64% (P < 0.05) (Fig. 8I), which was rescued by TET and WDR5 inhibitors (P < 0.05) (Fig. 8I). To establish that the RNA obtained from epithelial sheets reflected mRNA from keratinocytes and not underlying connective tissue, mRNA levels of K14, which is highly expressed in keratinocytes, and Col1α2, which is highly expressed in fibroblasts, was assessed (Fig. 8K). The K14 mRNA levels were several 100-fold higher than Col1α2, consistent with epithelial origin (Fig. 8K).

Optimal levels of MMP expression are needed for normal wound healing. We identified in human wounds at the leading wound edge that there is a shift from DNA methylation to DNA demethylation that corresponds with increased expression of MMP9. In vitro studies demonstrate that MMP9 expression is normally repressed by methylation of the MMP9 promoter, which limits FOXO1-MMP9 promoter interactions. When the MMP9 promoter is demethylated, FOXO1 can interact with the MMP9 promoter and induce MMP9 expression. Thus, an epigenetic switch allows transition from a homeostatic to a wound-healing, migratory phenotype in keratinocytes that is facilitated by FOXO1-mediated MMP9 expression.

Research into the regulation of FOXO1 has focused on the posttranslational modification of FOXO1 (35,36). Here, we took a different approach, investigating the impact of epigenetic changes and used as a model diabetes-impaired reepithelialization that focused on MMP9, based on previous studies that it plays a causative role in delayed healing of human diabetic wounds (15–17). MMP-9 affects keratinocyte migration via degradation of extracellular matrix proteins that are involved in ECM-cell interactions. MMP9 cleaves collagen fibrils, which act as substrates for keratinocytes to promote migration (46). When MMP9 is overexpressed, extracellular matrix is excessively degraded and remodeled. This negatively affects the cell attachment and migration (47). It has been widely reported that delayed wound healing is linked to high levels of MMP9 (16–23,37). To better understand how HG levels affect FOXO1 interaction with the MMP9 promoter, we examined changes in DNA methylation and demethylation. HG reduced DNA methylation (5mC) and increased DNA demethylation (5hmC) of CpG sites near consensus FOXO1 response elements in the MMP9 promoter. This was demonstrated by bisulfite and high-throughput DNA sequencing studies and DNA IP experiments. A DNA methylation inhibitor, decitabine, or knockdown of DNA methylase DNMT3b increased HG-induced FOXO1 binding to the MMP9 promoter and MMP9 mRNA levels. In contrast, a DNA demethylase inhibitor bobcat339 or knockdown of the DNA demethylases enzymes, TET1, -2, or -3 reduced HG-stimulated FOXO1 binding to the MMP9 promoter and MMP9 expression. Our evidence pointed to TETs as important regulators of MMP9, given that knockdown of both TET1, -2, and -3 each have a significant effect. It is noteworthy that the reduction in MMP9 expression by knockdown by TET3 alone or TET1 and TET2 combined was particularly effective. This is the first evidence that modulation of DNA methylation and demethylation is a key factor in FOXO1-regulated gene expression. HG also increased permissive histone-3 H3K4me3 trimethylation and reduced repressive histone-3 H3K9me3 and H3K27me3 trimethylation associated with the MMP9 promoter. Inhibition or siRNA knockdown of the histone demethylase blocked the permissive H3K4me3 trimethylation and reversed HG-induced FOXO1 binding to the MMP9 promoter and MMP9 expression, demonstrating a cause-and-effect relationship. In contrast, HG reduced repressive H3K9me3 and H3K27me3 histone trimethylation and their inhibition. Blocking their formation with a specific inhibitor or knockdown of the H3K9me3 demethylase reduced HG-induced FOXO1-MMP9 promoter interaction and MMP9 mRNA levels. These results suggest that HG reduces the formation of repressive histone markers that are functionally significant in modulating FOXO1-stimulated gene expression.

The importance of HG-stimulated FOXO1-MMP9 interactions was supported by further studies examining RNA Pol-II. RNA Pol-II is required for the expression of protein-coding genes, and its activity is tightly regulated (48). FOXO1 was needed for interaction of RNA Pol-II with the MMP9 promoter since knockdown of FOXO1 by RNAi significantly reduced Pol-II binding to the MMP9 promoter. Moreover, immunoprecipitation pull-down assays indicate that FOXO1 and RNA Pol-II physically interact with each other and that this interaction is enhanced by exposure to HG. This suggests a previously underappreciated function of FOXO1 as a binding partner with RNA Pol-II. This finding may represent a more generalized mechanism for increased expression of FOXO1 target genes in diabetic conditions (i.e., through recruitment of RNA Pol-II to promoter sites). Taken together our data also support the concept that DNMTs, TETs, and histone modifications control FOXO1-MMP9 promoter interaction by increased DNA demethylation, decreased DNA methylation, and reduced permissive histone-3 methylation. However, the pharmacological and mRNA knockdown of DMNTs, TETs, and WDR5 may change the expression of other genes and indirectly regulate FOXO1-MMP9 promoter binding and MMP9 mRNA expression.

In vivo studies investigated the therapeutic value of inhibiting the epigenetic changes described above. We applied by local injection a specific TET DNA demethylase inhibitor, bobcat339, or a specific inhibitor that blocks WDR5, MM-102, which inhibits permissive H3K4 trimethylation. Both treatments blocked diabetes-increased MMP9 expression in keratinocytes in vivo, measured by immunofluorescence and real-time PCR. Diabetes significantly reduced uPAR mRNA levels, a marker of migrating keratinocytes, that was also rescued by local treatment with epigenetic inhibitors. In vitro migration assays showed that bobcat339 and MM-102 rescued HG-impaired keratinocyte migration, consistent with in vivo results measuring uPAR. Moreover, application of MMP9 reversed the rescue of keratinocyte migration by inhibiting DNA demethylation or permissive histone-3 trimethylation. The link to MMP9 was further established by rescue of HG-impaired migration with an MMP9 specific inhibitor, mimicking the effect of bobcat339 and MM-102. Most importantly, the epigenetic inhibitors completely rescued the negative effect of diabetes on reepithelialization in vivo. This is the first evidence that inhibitors of DNA demethylation and histone methylation enhance mucosal and dermal healing in diabetes. It provides an explanation for our previous observations that FOXO1 deletion significantly improves diabetic mucosal and dermal wound healing (33–35). The results may have broader importance and provide a basis to understand how diabetes may alter gene expression through a FOXO1-mediated mechanism that contributes to pathologic changes in diabetic complications such as early diabetic retinopathy (49), cardiomyopathy (28,29), and diabetes-impaired fracture healing (31,32).

Acknowledgments. The authors thank Juan Wang and Subechhya Neupane (Department of Periodontics, School of Dental Medicine, University of Pennsylvania, Philadelphia, PA) for technical assistance with in vitro and in vivo experiments.

Funding. This work was funded by National Institutes of Health National Institute of Dental and Craniofacial Research grants R01DE019108 and R01DE017732 (D.T.G.), R01DE031046 (S.A.), and R01DE030415 (K.I.K.).

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. B.Y., S.A., M.V.G., P.M.T., K.I.K., M.B., and D.T.G. reviewed and edited the manuscript. B.Y., S.A., T.D., M.K., and J.Su. performed in vivo and in vitro experiments. B.Y., S.A., P.M.T., J.Se., M.B., and D.T.G. analyzed data. B.Y. and M.V.G. performed bioinformatic analysis. B.Y., P.M.T., K.I.K., J.Se., M.B., and D.T.G. developed the experimental approach. B.Y. and D.T.G. wrote the original draft. S.A., K.I.K., and D.T.G. acquired funding. D.T.G. conceptualized the study. D.T.G. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.