Insulin and Glucose Alter Death-Associated Protein Kinase 3 (DAPK3) DNA Methylation in Human Skeletal Muscle

Type 2 diabetes is a life-threatening metabolic disease reaching epidemic-like proportions. The pathophysiology of type 2 diabetes is incompletely understood, but both genetic and environmental factors contribute to disease development (1). People with type 2 diabetes are characterized by skeletal muscle insulin resistance arising from impaired insulin signaling and reduced glucose transport activity (2,3). Many of these same steps are impaired in skeletal muscle from relatives of patients with type 2 diabetes who are glucose intolerant (4,5). Thus, skeletal muscle insulin resistance is a characteristic feature of type 2 diabetes that can be noted at all stages of disease development.

Skeletal muscle insulin resistance may arise in part from a dysfunctional metabolic program, either of an epigenetic or genetic origin (6). This notion is partly based on the observation that impairments in glucose and lipid metabolism persist in skeletal muscle cultures derived from people with type 2 diabetes (7–10), as well as in those from relatives of patients with type 2 diabetes who are insulin resistant but do not have diabetes (11,12). Strikingly, these metabolic derangements observed in cultured cells mirror the clinical presentation of type 2 diabetes in the donors, indicating that insulin resistance is an underlying intrinsic defect. However, skeletal muscle insulin resistance can also be acquired by exposure to high levels of insulin, glucose, free fatty acids, or cytokines (7,13–16), systemic factors associated with type 2 diabetes. Although acquired insulin resistance has largely been attributed to defective signal transduction and enzymatic modifications, epigenetic modifications may also play a role (1).

Epigenetic modifications including DNA methylation provide a potential molecular basis for the interaction between genetic and environmental factors on glucose homeostasis. Recent evidence suggests that systematic factors associated with obesity or insulin resistance can modify DNA methylation in somatic tissues (9,13,17–19), indicating that the regulation of the epigenome is a more dynamic process than previously appreciated. Using whole-genome promoter methylation analysis of DNA from skeletal muscle from healthy participants and participants with type 2 diabetes, we screened for differentially methylated genes and identified hypermethylation of promoters of genes involved in glucose metabolism and mitochondrial function (13). We also found that elevated levels of cytokines and free fatty acids altered DNA methylation of genes important for lipid metabolism in primary human muscle cultures (13), suggesting that systemic factors associated with insulin resistance may play a role. Changes in DNA methylation in skeletal muscle are also triggered by exercise (20–22), dietary factors (19,23,24), and gastric bypass–induced weight loss (17). Whether DNA methylation of genes involved in metabolic regulation is altered in response to dynamic shifts in metabolic demand experienced during an acute insulin stimulation or a glucose challenge is unknown.

We have shown that changes in skeletal muscle DNA methylation in type 2 diabetes are inversely correlated with mRNA expression of multiple genes involved in the regulation of insulin sensitivity (13,17), indicating that an epigenetic fingerprint emerges in response to metabolic dysregulation. Insulin treatment acutely alters the gene expression profile of skeletal muscle biopsies of healthy individuals and patients with type 2 diabetes (25). Thus, we hypothesized that insulin directly mediates changes in DNA methylation of skeletal muscle. We found that insulin treatment reduced DNA methylation in the calcium pump ATP2A3 gene and increased DNA methylation in the gene body of death-associated protein kinase 3 (DAPK3), a gene involved in cell proliferation, apoptosis, and autophagy. Short-term fasting/refeeding is also associated with changes in skeletal muscle gene expression profiles concomitant with shifts in substrate utilization between carbohydrates and fats as main fuel sources (26). Thus, we determined whether a glucose challenge alters DNA methylation of ATP2A3 and DAPK3 in skeletal muscle. Establishing the epigenomic response to these physiological challenges in metabolically healthy people with normal glucose tolerance will build profiles to better predict the development of skeletal muscle insulin resistance.

Two separate cohorts were studied. Ten healthy sedentary men with normal glucose tolerance were recruited for the open muscle biopsy study. Clinical characteristics of these participants are presented in Table 1A. Twenty-four men with either normal glucose tolerance (n = 12) or type 2 diabetes (n = 12) were recruited for the glucose tolerance study. Clinical characteristics of these participants are presented in Table 1B. Study protocols were approved by the local ethics committee in Stockholm and performed according to the Declaration of Helsinki. Informed written consent was obtained from all participants. All subjects included were instructed to not participate in any physical exercise 2 days prior to the visit to the clinic. Subjects reported to the clinic at 0800 h after an overnight fast; they had been asked not to eat or drink anything but water after 2000 h the night before.

Open muscle biopsies were performed as previously described (27). A skeletal muscle biopsy (∼2 g) was excised and placed in oxygenated Krebs-Henseleit buffer (KHB) containing 5 mmol/L HEPES and 0.1% BSA (RIA Grade; Sigma-Aldrich, St. Louis, MO). Smaller muscle strips were dissected from the biopsy specimen and mounted on Plexiglas clamps; placed in oxygenated KHB containing 5 mmol/L HEPES 0.1% BSA, 15 mmol/L mannitol, and 5 mmol/L glucose; and transported to the laboratory. Thereafter, muscle strips were incubated in the absence (basal) or presence of insulin (120 nmol/L) for 60 min in individual glass vials containing oxygenated KHB plus the addition of 5 mmol/L HEPES 0.1% BSA, 15 mmol/L mannitol, and 5 mmol/L glucose. Muscle strips were frozen in liquid nitrogen for subsequent methylation analysis.

Fasting blood samples were taken for clinical chemistry (Table 1B), and a skeletal muscle biopsy (40–100 mg) was obtained from the vastus lateralis with a Weil-Blakesley conchotome tong instrument (Agnthos, Sweden) under local anesthesia (mepivacaine hydrochloride 10 mg/mL (AstraZeneca). Biopsies were frozen and stored in liquid nitrogen. Thereafter, the subjects ingested a standardized solution containing 75 g of glucose. Two hours after the glucose ingestion, a blood sample and skeletal muscle biopsy were collected as described above.

Skeletal muscle biopsies were obtained from healthy male volunteers as previously described (28). Cell subculture was performed as described (28), with the following adaptations. Primary myoblasts were grown in DMEM/F-12 with 20% FBS, 1% penicillin-streptomycin (10,000 units/mL), and 1% Fungizone on uncoated culture flasks. At 80% confluence, the medium was changed to a fusion medium of 76% DMEM with 25 mmol/L glucose, 20% M199 (5.5 mmol/L), 2% HEPES, 1% penicillin-streptomycin (10,000 units/mL), and 1% Fungizone (all from Thermo Fisher Scientific, Waltman, MA, except the FBS from Sigma-Aldrich) with 0.03 µg/mL zinc sulfate, 1.4 mg/mL vitamin B12 (both from Sigma-Aldrich), 100 µg/mL apo-transferrin (from BBI solutions, Cardiff, U.K.), and 0.286 IU/mL insulin (Novo Nordisk, Bagsværd, Denmark) for 4–5 days. Following fusion, the medium was switched to the same medium but without apo-transferrin and insulin for an additional 4–5 days (postfusion media) with 2% FBS, except for the insulin-treatment study, where no FBS was used. In the For the low/high glucose experiment, DMEM with 5.5 mmol/L glucose was used in the postfusion medium. The medium was changed every 2–3 days and prior to each experiment. Myotubes were incubated for 1 h in the absence or presence of insulin (120 nmol/L) and either harvested immediately or washed with fresh postfusion media and harvested after an additional 4 h incubation in the absence of insulin.

Myotubes were transfected (Lipofectamine RNAiMAX Transfection Reagent; Thermo Fisher Scientific) with 10 nmol/L of a nontargeting negative control small interfering RNA (siRNA) (Silencer Select Negative Control No. 2, catalog no. 4390847; Thermo Fisher Scientific) or siRNA directed against ATP2A3 or DAPK3 (validated Silencer Select siRNA s1749 or s557, respectively; Thermo Fisher Scientific). Cells were transfected twice, separated by a period of 48 h, and the final transfection was initiated 48 h prior to further experiments.

Lipid oxidation and glucose incorporation into glycogen was determined as described (28). For lipid oxidation, myotubes were exposed to serum-free postdifferentiation media containing 0.025 mmol/L palmitate and incubated in the absence or presence of 2 mmol/L 5-aminoimadazole-4-carboxamide ribonucleotide (AICAR; Toronto Research Chemicals, Toronto, CA) for 6 h. For glycogen synthesis, cells were incubated for 4 h in serum-free postfusion media prior to a 2-h incubation in fresh serum-free postfusion media in the presence or absence of 120 nmol/L insulin.

To assess signal transduction, cells were incubated in the absence or presence of 120 nmol/L insulin or 2 mmol/L AICAR for 1 h. Cells were rinsed twice in ice-cold PBS and then frozen at −20°C. mRNA analysis and description of primer oligonucleotides are detailed in Supplementary Table 1. Western blot analysis and description of antibodies are detailed in Supplementary Table 2.

Genomic DNA and mRNA was extracted from skeletal muscle and cultured myotubes using AllPrep DNA/RNA Protein Kit or DNeasy and RNeasy kits (Qiagen, Hilden, Germany). DNA and RNA concentration was measured with NanoDrop 1000 (Thermo Fisher Scientific). Genomic DNA was bisulfite treated using EpiTect Fast Bisulfite Conversion Kit (Qiagen).

Genome-wide methylation was run on the Infinium HumanMethylation450 BeadChip arrays (Illumina, San Diego, CA). Analysis was performed at the Bioinformatics and Expression Analysis core facility at Karolinska Institutet. Beta-values were converted into M-values in R, using Bioconductor packages lumi and methylumi (29,30). Paired data were analyzed as Beta-values and then converted to M-values. Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis was performed using WebGestalt as updated on 30 January 2013 (31,32). DNA pyrosequencing analysis and primer oligonucleotides are detailed in Supplementary Table 4.

RNA was reverse transcribed using the High-Capacity cDNA Reverse Transcription Kit (Thermo Fisher Scientific). Real-time PCR was performed using SYBR Green–based primers. HPRT1 was used as the reference gene for skeletal muscle biopsies. The geometric mean of reference genes (B2M, GAPDH, HPRT, and RPLP0) was used to compare results of experiments performed using cultured cells, except for gene-silencing experiments, where the geometric mean of PPIB and TBP were used. Primer oligonucleotides for the quantitative PCR analysis is shown in Supplementary Table 1.

Parametric tests or nonparametric tests were used as appropriate followed by either Bonferroni post hoc correction or Benjamini-Hochberg correction (KEGG pathway analysis). Correlation analysis was determined using Spearman correlation coefficient. Analysis was performed using GraphPad Prism version 6.0 (GraphPad Software, Inc., La Jolla, CA) or R base version 3.3.1 and open-source packages. Comparisons were considered to be statistically significant at P < 0.05.

Genome-wide methylation was determined in isolated skeletal muscle strips incubated in the absence (basal) or presence of insulin using an Illumina 450k array. The methylation array data are deposited in the Gene Expression Omnibus (GEO accession: GSE87655). Genes showing differentially methylated CpGs were selected (P < 0.05). Genes showing CpGs with differences in Beta-values and M-values above 0.03 between basal and insulin-stimulated muscle were considered further (Supplementary Table 3). Insulin treatment altered methylation of over 1,600 genes on at least one CpG site. The list was further interrogated using WebGestalt (31,32). If several CpGs relating to same gene were identified, the list was curated, such that any gene associated with a significant change in CpG methylation between basal and insulin-stimulated skeletal muscle occurred only once in the list. KEGG pathway analysis revealed that insulin stimulation significantly altered DNA methylation of genes related to over 15 pathways, including genes involved in insulin signaling, metabolic pathways, and type 2 diabetes (Fig. 1).

To validate the results from the 450k array, we studied four specific regions/genes of interest: DAPK3 (cg18748062), ATP2A3 (cg08498987), HK1 (cg17575314), and HDAC5 (cg16217242) (Table 2). We purchased or designed primers for pyrosequencing analysis (Supplementary Table 4). HK1 was selected as part of the insulin signaling pathway as revealed by the KEGG analysis. The other three sites were selected based on either the high M-value or Beta-value difference, as well as the role of each gene in metabolism. Pyrosequencing of DNA extracted from isolated skeletal muscle strips incubated in the absence or presence of insulin was performed. Consistent with results from the 450k array, methylation of one CpG site (cg08498987) in the gene body of DAPK3 was increased in response to insulin. As pyrosequencing allows for the sequencing of regions over 100 base pairs in length, we investigated the adjacent CpGs and found that methylation differed between CpGs and that insulin impacted methylation in the whole region (Fig. 2A). We next analyzed five different CpGs in the 3′ untranslated region of the ATP2A3 gene. CpG5 corresponded to the probe on the array (cg08498987). Whereas methylation was unchanged in CpG5 in response to insulin, methylation of CpG2 and CpG3 was altered (Fig. 2B). Moreover, the methylation pattern was unaltered over five different CpGs in HK1 in response to insulin (Fig. 2C). The level of methylation in the HDAC5 region probed was close to the unmethylated negative control (<1% of the CpGs methylated) and therefore considered below background (data not shown). Given the change in methylation of DAPK3 and ATP2A3, we next determined mRNA expression of each respective gene. Despite the changes in DNA methylation, mRNA of DAPK3 or ATP2A3 was unaltered in response to insulin (Fig. 2D).

To determine whether an insulin-induced methylation change occurred specifically at the studied CpG site or whether it was a more general effect over a stretch of the chromosome, we interrogated the methylation array by focusing on the genes directly adjacent to DAPK3 (Supplementary Table 5).

DNA methylation of DAPK3 and ATP2A3 was determined in myotubes incubated for 1 h in the absence or presence of insulin (120 nmol/L) (Fig. 3). Myotubes were studied immediately after insulin exposure or allowed to recover in standard media for an additional 4 h. Insulin treatment did not significantly increase DAPK3 DNA methylation in myotubes acutely (P < 0.09; Fig. 3A). Although ATP2A3 methylation was consistently lower after insulin stimulation, no significant change was noted between time points (Fig. 3B). The level of methylation in basal and insulin-stimulated myotubes was variable between donors (Fig. 3C), whereas the insulin-stimulated change in percent DNA copies methylated was more stable (Fig. 3D). mRNA of DAPK3 and ATP2A3 was unaltered between basal and insulin-stimulated conditions (data not shown).

We next measured DNA methylation of DAPK3 and ATP2A3 in skeletal muscle obtained from a separate cohort of men with normal glucose tolerance or type 2 diabetes before and 2 h after a glucose challenge. ATP2A3 methylation was not altered between patients with type 2 diabetes and healthy men (data not shown). Moreover, a glucose challenge did not alter ATP2A3 methylation (data not shown). We assessed methylation of DAPK3 at two CpG sites (CpG1; Fig. 4A and CpG2; Fig. 4B). Methylation of DAPK3 at CpG1 but not CpG2 was lower in men with type 2 diabetes compared with healthy control men. Moreover, a glucose challenge was associated with lower CpG1 and CpG2 methylation in DAPK3 in both men with type 2 diabetes and healthy men. To determine whether glucose altered DNA methylation in vitro, differentiated human myotubes were incubated in the presence of 25 mmol/L glucose for 2 h and CpG1 and CpG2 methylation in DAPK3 was assessed. Whereas methylation at the CpG1 site (Fig. 4C) was unaltered, a trend for reduced methylation was observed at the CpG2 site (Fig. 4D).

As plasma glucose was significantly different between the groups before and after the glucose tolerance test, as well as between men with type 2 diabetes and healthy men (Table 1B), we correlated percentage of DAPK3 methylation with glucose concentration. DAPK3 methylation at CpG1 and CpG2 was inversely correlated with the 2-h glucose concentration. However, the relationship between fasting glucose concentration and DAPK3 methylation did not reach statistical significance (Fig. 5). BMI did not associate with DAPK3 methylation, while waist-to-hip ratio correlated with DAPK3 CpG1 methylation in skeletal muscle obtained 2 h after a glucose challenge (P = 0.004) (data not shown).

siRNA was used to knock down the expression of ATP2A3 or DAPK3 in human skeletal muscle. Subsequently, glucose incorporation into glycogen, palmitate oxidation, and signal transduction was assessed. ATP2A3 and DAPK3 mRNA was reduced after gene silencing in differentiated primary human skeletal muscle cells (Fig. 6A and B). Silencing of ATP2A3 (Fig. 6C) or DAPK3 (Fig. 6D) did not alter basal or insulin-stimulated glucose incorporation into glycogen. Silencing ATP2A3 (Fig. 6E) did not alter basal or AICAR-stimulated palmitate oxidation. Conversely, DAPK3 silencing increased basal and AICAR-stimulated palmitate oxidation (Fig. 6F).

To further explore the effects of ATP2A3 or DAPK3 on myotube function, we determined insulin signaling and autophagic signaling after silencing these genes. Silencing ATP2A3 or DAPK3 did not impact phosphorylation of protein kinase B (PKB; also known as Akt) or TBC1 domain family member 4 (TBC1D4; also known as Akt substrate of 160 kDa, AS160) in basal or insulin-stimulated myotubes (Fig. 7A–D). As DAPK3 is involved in autophagy, we measured p62 and LC3 protein abundance. DAPK3 silencing did not alter the protein abundance of p62 and LC3 in cultured myotubes (Fig. 7E–H). GAPDH protein abundance was not altered by gene silencing (data not shown).

Epigenetic modifications of the genome, including DNA methylation, provide a molecular basis for the interaction between genetic and environmental factors on glucose homeostasis (1). While dynamic changes in DNA methylation have been mainly studied during the narrow window of early development, evidence is emerging that diabetes, obesity, and weight loss have a dynamic effect on the epigenome (1). Changes in skeletal muscle DNA methylation in type 2 diabetes are inversely correlated with mRNA expression of multiple genes involved in the regulation of insulin sensitivity (13,17), indicating that an epigenetic fingerprint emerges in response to metabolic dysregulation. Careful mapping of epigenetic signatures in healthy people and patients with type 2 diabetes at rest and in response to various metabolic challenges may uncover novel biomarkers and targets that can be leveraged to predict and treat insulin resistance. Here we show that insulin stimulation or oral glucose ingestion rapidly alters the DNA methylation in human skeletal muscle, supporting the notion that systemic factors can modify the epigenome in type 2 diabetes.

Type 2 diabetes and obesity are associated with genome-wide changes in DNA methylation in skeletal muscle (13,17). Interestingly, therapeutic interventions to improve insulin sensitivity including weight loss (17) and exercise training (20–22) remodel DNA methylation patterns in skeletal muscle. Moreover, exposure of cultured muscle cells to systemic factors that induce insulin resistance such as elevated lipids or cytokines or to agents that mimic the effects of exercise on glucose and lipid metabolism alter DNA methylation of metabolic genes (9,33). Thus, alterations in the extracellular milieu, including hyperglycemia, hyperinsulinemia, and elevated free fatty acids and cytokines may reprogram skeletal muscle insulin sensitivity through changes in DNA methylation and gene expression profiles. Here we determined the direct effects of an acute insulin stimulation on genome-wide DNA methylation profiles in intact skeletal muscle from healthy donors incubated in vitro. We studied the epigenomic response to insulin in metabolically healthy people with normal glucose tolerance to gain insight into the development of insulin resistance. Our bioinformatic analysis of genome-wide DNA methylation revealed rapid and broad changes in CpG methylation patterns in skeletal muscle from healthy men in response to an acute insulin exposure. Furthermore, insulin-induced changes in CpG sites were identified in genes related to the insulin signaling pathway. Moreover, several pathways related to insulin action were enriched including those annotated as “metabolic pathways,” “inositol phosphate metabolism,” and “type 2 diabetes.” Thus, we reveal an insulin-induced epigenetic fingerprint in skeletal muscle on multiple pathways involved in glucose and lipid metabolism.

Our motivation for taking a genome-wide approach to assess DNA methylation was fueled in part by our desire to identify previously unrecognized insulin-responsive targets. Confirming the array results on a base-pair level, we found that insulin stimulation increased DNA methylation in the DAPK3 gene. DAPK3 (ZIPK/DLK) is a Ca²⁺/calmodulin-regulated serine/threonine kinase that shares a high level of homology within its kinase domain with the other two DAPK family members (DAPk), DAPK1 and DAPK2. DAPK3 plays a role in autophagy (34) and apoptosis (35). To further validate the effects of insulin on DAPK3 methylation, we exposed primary human myotubes to insulin and assessed methylation at the identified CpG sites. We observed a trend for insulin-induced DAPK3 methylation in primary human skeletal muscle that persisted over 4 h Although not statistically significant, this effect mirrored the insulin-induced DNA methylation of DAPK3 in incubated human muscle strips. Interestingly, DAPK3 is a tumor suppressor that is frequently methylated or mutated in many cancer types (36) and is negatively regulated by Akt signaling (37). DAPK3 is important for myosin light chain phosphorylation, leading to muscle contraction or cell mobility (34). While the role of DAPk family members in metabolic disease is unknown, acute exercise alters DAPK2 phosphorylation in skeletal muscle from healthy untrained men (38).

One of the most dramatic shifts in fuel utilization throughout the day occurs during the transition between the use of lipids in the fasting state and the use of carbohydrates in the insulin-stimulated fed state. In this regard, skeletal muscle is a highly malleable tissue and can transition between these states with a high degree of metabolic flexibility (39). Moreover, activation of energy/nutrient-sensing pathways in the fasted versus fed state subsequently modifies gene expression (40). Thus, we determined whether an oral glucose challenge altered DNA methylation of CpG sites in DAPK3. Contrasting the insulin-stimulated condition, we found that DAPK3 DNA methylation was reduced in skeletal muscle from healthy men after the glucose tolerance test. The difference in DAPK3 DNA methylation between the insulin- and glucose-stimulated conditions is difficult to reconcile, given that plasma insulin levels rise during a glucose challenge. However, metabolomic studies have revealed that thousands of metabolites from classes that include free fatty acids, acylcarnitines, bile acids, lysophosphatidylcholines, and amino acids are altered during an oral glucose tolerance test (41,42). Thus, the systemic milieu between the in vitro muscle incubation and the oral glucose tolerance test may account for the differences. During the glucose tolerance test, the rise in blood glucose, as well as any one of a number of metabolites, may act as a nutrient signal to suppress processes controlling DAPK3 DNA methylation. However, we did not observe similar reductions in DAPK3 DNA methylation in human muscle cells exposed to high glucose in vitro. Thus, other metabolites in addition to glucose may influence DAPK3 DNA methylation. In mouse knockout models, DAPK3 has an independent effect from ERK or Akt activation to negatively regulate mTOR-S6K-S6 signaling (43), implicating DAPK3 in nutrient signaling pathways. We observed an inverse correlation between DNA methylation and the 2-h blood glucose concentration, suggesting dynamic regulation of the DAPK3 gene with refeeding. Furthermore, methylation of the DAPK3 gene at CpG1 but not CpG2 was lower in patients with type 2 diabetes. Even though the patients with type 2 diabetes are fasted, they are in a state of metabolic inflexibility, with peripheral tissues facing nutrient overload (39), which may account for the differences in DNA methylation.

We used siRNA against DAPK3 or ATP2A3 and assessed glucose or lipid metabolism. Silencing of either DAPK3 or ATP2A3 did not alter basal or insulin-stimulated glucose incorporation into glycogen in cultured myotubes. However, DAPK3 silencing increased basal and AICAR-stimulated palmitate oxidation. Since carbohydrate ingestion acutely reduces lipid oxidation, our data suggests that glucose/carbohydrate feeding may lead to a long-lasting inhibition of lipid oxidation via reduced methylation of DAPK3, a gene which may normally modulate lipid oxidation, as evidenced by our siRNA experiments. Although we did not observe any changes in DAPK3 mRNA at 2 h following the glucose challenge, changes in DAPK3 expression may occur at a later time point. DAPK3 has been implicated in the regulation of autophagy (34), and therefore we determined an abundance of the autophagy markers p62 and LC3. Silencing of DAPK3 did not alter the abundance of these targets. We cannot exclude the possibility that DAPK3 or ATP2A3 may be involved in other aspects of metabolism and skeletal muscle physiology. For example, ATP2A3 encodes a SERCA3 Ca²⁺-ATPase, an intracellular pump located in the sarcoplasmic and endoplasmic reticulum that plays a role in excitation and contraction, and thus ATP2A3 silencing may affect skeletal muscle contraction.

Pyrosequencing studies confirmed that insulin altered DNA methylation of ATP2A3 in incubated human skeletal muscle from men with normal glucose tolerance. ATP2A3 is implicated in glucose-activated β-cell regulation of calcium homeostasis in type 2 diabetes (44) and endoplasmic reticulum stress (45). Rare missense mutations in the ATP2A3 gene are associated with type 2 diabetes (46). In cultured human muscle, we observed that insulin acutely decreased ATP2A3 methylation, with levels normalized within 4 h after the stimulation, indicating that systemic factors play a role. Chronic hyperglycemia in rats subjected to partial pancreatectomy reduces ATP2A3/SERCA3 mRNA, implicating that glucose may have a direct effect on gene expression (47). However, we found that ATP2A3 DNA methylation was unaltered by a glucose challenge, and similar levels were noted between patients with type 2 diabetes and healthy subjects. Thus, insulin rather than glucose or type 2 diabetes appears to drive ATP2A3 methylation changes.

DNA methylation can affect gene expression by altering histone interactions, influencing transcription factor binding, and/or influencing recruitment of methyl-binding proteins (48). Thus, a simple filter for analyzing genome-wide methylation and transcriptomic data is to search for inverse relationships between DNA methylation and mRNA expression of the corresponding gene. In the current study, changes in DNA methylation did not correlate with changes in mRNA levels of DAPK3 or ATP2A3 in skeletal muscle at the time points investigated. However, DNA methylation may not always correlate with mRNA expression (17,18,49,50). Although epigenetic processes are one part of the complex system of regulation of gene transcription and a localized CpG methylation might be insufficient to control gene expression, changes in mRNA could occur at later time points than investigated in the current study.

In conclusion, insulin acutely alters the DNA methylation profile of human skeletal muscle. These findings advance the notion that DNA methylation is a rapidly adaptive epigenetic mark in somatic cells. Acute insulin exposure altered DNA methylation of DAPK3 and ATP2A3, genes involved in apoptosis and endoplasmic reticulum stress, respectively. ATP2A3 was unaffected by glucose challenge or type 2 diabetes, indicating that insulin has a direct role. In contrast, DAPK3 DNA methylation was altered in type 2 diabetes and after an oral glucose challenge, with levels of DNA methylation inversely correlated with glucose levels. Moreover, insulin and glucose modulate skeletal muscle DAPK3 DNA methylation in a reciprocal manner, suggesting a feedback control on gene expression. Continued efforts to establish the epigenomic response to systemic factors associated with type 2 diabetes in metabolically healthy people with normal glucose tolerance will be useful to better predict genomic landscape involved in the development of skeletal muscle insulin resistance.

Funding. The Strategic Diabetes Program at Karolinska Institutet, European Research Council Ideas Program (ICEBERG, ERC-2008-AdG23285), Swedish Research Council (Vetenskapsrådet) (2011-3550, 2012-1760, 2015-165), Swedish Diabetes Foundation (DIA2012-082, DIA2012-047), Torsten Söderbergs Foundation (M71/15), Swedish Foundation for Strategic Research (SRL10-0027), Diabetes Wellness Sweden, Novo Nordisk Foundation (NNF14OC0009941), and Stockholm County Council (20150326) supported this research. The Novo Nordisk Foundation Center for Basic Metabolic Research is an independent research center at the University of Copenhagen partially funded by an unrestricted donation from the Novo Nordisk Foundation.

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

Author Contributions. J.M.M. helped design the study, collected and analyzed data, and drafted and helped revise the manuscript. D.G.L. helped design the study, collected and analyzed data, and helped revise the manuscript. C.N. and S.G.-C. collected and analyzed data. E.N. helped design the study and collected and analyzed data. A.K. and J.R.Z. designed the study, interpreted data, and drafted and revised the manuscript. J.R.Z. 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.