Loss of Mbd2 Protects Mice Against High-Fat Diet–Induced Obesity and Insulin Resistance by Regulating the Homeostasis of Energy Storage and Expenditure

The ongoing obesity epidemic and its associated complications of chronic diseases, such as type 2 diabetes, dyslipidemia, nonalcoholic fatty liver disease, and hypertension, exert formidable challenges and burden to human health (1). Obesity is caused by a complex interplay between genetic and environmental factors. It is believed that environmental factors interact with susceptible genes to modulate the risk for obesity, which may also happen through direct chemical modifications of the genome by so-called DNA methylation. Although high-fat diet (HFD), as a highly influential environmental factor, has long been used to induce obesity and insulin resistance, its impact on epigenetic modifications of the genome, especially in the adipose genome, is yet to be fully elucidated. Evidence shows that short-term high-fat overfeeding impacts genome-wide DNA methylation patterns in human skeletal muscle (2). A global study of DNA methylation in human adipose tissue also characterized changes for the epigenetic patterns in response to long-term exercise, potentially affecting adipocyte metabolism (3). Given that DNA methylation acts as a “footprint” to record gene–environment interactions or accumulated environmental exposures during the course of daily life processes (4,5), we thus hypothesize that HFD induces adipose tissues to undergo a DNA methylation turnover, which would be in favor of developing obesity and insulin resistance.

Previous studies including ours demonstrated that information encoded by the DNA methylome is read by a family of methyl-CpG–binding domain (MBD) proteins, including MBD1, MBD2, MBD3, MBD4, and the founding member, MeCP2 (6–8). It has been recognized that MBD1, MBD2, and MeCP2 selectively bind to methylated CpGs, by which they prevent transcription factors binding to gene promoters and/or recruit histone deacetyltransferases and corepressors (9–11). In contrast, binding affinity for MBD3 is not dependent on DNA methylation, and MBD4 has been primarily regarded as a thymine DNA glycosylase with little role in transcriptional repression (12,13). Mice lacking MeCP2 is associated with specific neurological defects that mimic the human neurological disorder Rett syndrome (14). Lack of Mbd1 develops deficits in adult neurogenesis and hippocampal function (15), whereas loss of Mbd4 suppresses CpG mutability and tumorigenesis (13,16). In sharp contrast, mice deficient in Mbd2 are generally normal except for a minor phenotype in maternal behavior (12), which endowed MBD2 to be an ideal target for the study of DNA methylation on disease pathoetiology. In the current study, we used Mbd2^−/− mice as a model to investigate the effect of DNA methylation on HFD-induced obesity. HFD induced genomic DNA of adipose tissue to undergo a DNA methylation turnover, which altered the homeostasis of energy expenditure and storage in favor of obesity development, and, as a result, loss of Mbd2 provided protection for mice against HFD-induced obesity and insulin resistance.

Mbd2 knockout (KO; Mbd2^−/−) mice in the C57BL/6 background were provided by Dr. Adrian Bird (Edinburgh University, Edinburgh, U.K.) (12). Both male Mbd2^−/− mice and their wild-type (WT) littermates (8 weeks old) were housed individually and either maintained on normal diet (ND) (MD12031; 10% of kcal from fat; Medicience Ltd., Jiang Su, China) or switch to HFD (MD12033; 60% of kcal from fat; Medicience Ltd.) for 16 consecutive weeks. Mean daily food consumption was determined in each mouse by calculating the amount of accumulatively consumed food for 1 month (8–12 weeks old) and then normalized with body weight. All protocols for animal studies were approved by the Tongji Hospital Animal Care and Use Committee in accordance with the National Institutes of Health guidelines.

After 16 weeks of feeding, glucose and insulin tolerance tests were conducted by the intraperitoneal injection of glucose (Sigma-Aldrich, St. Louis, MO) and insulin (Novolin R; Novo Nordisk, Bagsvaerd, Denmark) as described previously (17). HOMA of insulin resistance (HOMA-IR) was calculated according to the formula: fasting insulin (μU/L) × fasting glucose (nmol/L)/22.5 (18).

Hematoxylin and eosin (H&E) and Oil Red O staining of epididymal adipose tissue and liver was performed as reported (19,20). The adipocyte area was estimated by measuring the area of >120 cells per ×100 original magnification tissue sections using an image quantitative digital analysis system (Image-Pro plus 6.0; Media Cybernetics) (21,22).

Western blotting and real-time PCR analysis were performed as reported (6). Antibodies against Mbd2 (sc-10752), Raptor (sc-27744), adipose triglyceride lipase (ATGL; sc-50223), carnitine palmitoyltransferase (Cpt1; sc-20514), and phosphorylated insulin receptor substrate (p-IRS; sc-17200) were purchased from Santa Cruz Biotechnology (Santa Cruz, CA), whereas antibodies against AKT (4685s), phosphorylated AKT (p-AKT Thr³⁰⁸) (5056s), and phosphorylated hormone-sensitive lipase (p-HSL; 4126s) were ordered from Cell Signaling Technology (Danvers, MA). The primer sequences for real-time PCR are listed in Supplementary Table 1.

For analysis of metabolic index, the mice were placed in metabolic cages individually connected with a comprehensive laboratory animal-monitoring system (Columbus Instruments, Columbus, OH). The mice were acclimatized to respiratory chambers for 48 h, followed by recording in real time for the data of VO₂, VCO₂, and respiratory exchange ratio (RER).

Paired samples of visceral tissues were obtained from 47 females with Chinese Han origin during abdominal surgery. Among those 47 females, 24 subjects were scheduled for bariatric surgery due to morbid obesity, whereas the other 23 patients were subjected to exploratory surgeries to exclude cancer or inflammatory diseases. The subjects were assigned to two cohorts (lean, BMI <25; and obese, BMI ≥30). All subjects had a stable weight with fluctuations <2% of the body weight for at least 3 months before surgery. This study was approved by the Institutional Review Board of Tongji Hospital, and informed consent was obtained from each subject. Biopsy samples from omental adipose tissue were collected during surgery, frozen immediately in liquid nitrogen, and stored at −80°C before use. The clinical characteristics for subjects included in the study are shown in Supplementary Table 7.

Global DNA methylation was determined using a MethylFlash Methylated DNA Quantification Kit (Epigentek, Farmingdale, NY) as instructed. Bisulfite DNA sequencing was conducted as previously described (23). Primer sequences for bisulfate DNA PCR are listed in Supplementary Table 2.

Fresh epididymal adipose tissues from HFD-induced obese mice were ground into small pieces (1 to 2 mm³) in liquid nitrogen and then crosslinked for 15 min by 1% formaldehyde in PBS. Chromatin immunoprecipitation (ChIP) assays were performed using a ChIP Assay Kit (Beyotime Biotechnology, Jiang Su, China) as detailed in the previous studies (24). Polyclonal antibodies against C/EBPα (sc-61X) and ATF6 (sc-22799X) were used for the assays, and a normal rabbit IgG (sc-2027X; Santa Cruz Biotechnology, Santa Cruz, CA) was used for negative controls. Primers used for ChIP assays are listed in Supplementary Table 3.

MBD2 ChIP sequencing (ChIP-seq) was performed following a previously reported protocol (25). Briefly, genomic DNA was isolated from epididymal adipose tissues of WT mice following 16 weeks of HFD or ND induction. Genomic DNA originated from six mouse of each group was pooled and then fragmented into 100–350 bp in length using a M220TM Focused-ultrasonicator (Covaris). The fragmented DNA from each pool was then subject to enrichment of methylated CpG DNA using a MethylMiner methylated DNA enrichment kit (Invitrogen, Carlsbad, CA) as instructed. The captured DNA was next subjected to high-throughput DNA sequencing using a HiSeq 2500 platform in BGI-Shenzhen (Shenzhen, China).

The peak regions for Raptor (−4,128 to −3,802) and Ucp1 (−3,467 to −3,220) (start codon as +1) were amplified from mouse genomic DNA. The mutated peak region for Raptor and Ucp1 were directly synthesized by the Tsingke Biological Technology (Beijing, China), in which cytosines located in all CpG sites were mutated into adenosine. The core promoter regions for Raptor (−3,801 to +82) and Ucp1 (−3,221 to −1) were PCR amplified and then PCR ligated to the peak region prepared above as previously reported (26). The resulting products were subcloned into a pGL-3 vector (pGL3-Raptor wt, pGL3-Raptor mut, pGL3-Ucp1 wt, and pGL3-Ucp1 mut) and then confirmed by DNA sequencing, respectively. In vitro methylation was carried out by incubating all constructs with SssI CpG Methyltransferase (Thermo Fisher, Beverly, MA) at 37°C for 15 min, which was then confirmed by bisulfite DNA sequencing as above. For Ucp1 promoter reporter assays, C3H10T 1/2 cells were transduced with Mbd2 adenovirus (043798A; Abcam), followed by transfecting with a mixture of plasmids for pGL3 luciferase reporter and pRL-TK luciferase (20:1). The cells were treated with 10 μmol forskolin for 4 h following 24 h of transfection and then harvested for analysis of luciferase activities using the dual-luciferase reporter assay system (Promega, Madison, WI). For analysis of Raptor promoter reporter activity, the plasmids were transfected into 3T3-L1 cells as above, and the cells were harvested 30 h after transfection without forskolin stimulation.

All data were expressed as mean ± SEM. Prism 5.0 software (GraphPad, La Jolla, CA) was used for statistical analysis of all data using the Student t test or one-way or two-way ANOVA where appropriate. In all cases, P < 0.05 was considered with statistical significance. Detailed approaches for bioinformatic analysis of ChIP-seq data were described in the Supplementary Data.

We first sought to address the impact of DNA methylation and Mbd2 deficiency on the development of obesity and type 2 diabetes. For this purpose, 8-week-old Mbd2^−/− and control littermates were fed with HFD for 16 weeks. It was interestingly noted that Mbd2^−/− mice were significantly protected from HFD-induced obesity (Fig. 1A and B), and importantly, the lower body weight in Mbd2^−/− mice was predominantly featured by the reduction of white adipose tissue (WAT) mass such as the epididymal adipose tissues (Fig. 1C and D). Of note, no significant difference in terms of tibia length, a marker for mouse development, was characterized between Mbd2^−/− mice and control littermates (Fig. 1E), indicating that the lower body weight identified in Mbd2^−/− mice was not caused by the delay of body development (growth retardation). Furthermore, measurement of food intake revealed that Mbd2^−/− mice actually consumed more food than their counterparts (Fig. 1F). Together, these data suggest that loss of Mbd2 provides protection for mice against HFD-induced obesity.

To elucidate the impact of Mbd2 deficiency on type 2 diabetes, we examined the development of insulin resistance and glucose intolerance in Mbd2^−/− and WT littermates after 16 weeks of HFD feeding. Consistent with the above observations, Mbd2^−/− mice showed significantly lower levels of blood glucose both in the fasted and fed state than that of control littermates (Fig. 2A). Similarly, Mbd2^−/− mice displayed significantly decreased serum insulin levels in the fasted state (Fig. 2B) along with a significant reduction for the HOMA-IR index (Fig. 2C). Unlike their WT littermates, Mbd2^−/− mice were characterized by the significantly improved glucose tolerance (Fig. 2D) and insulin sensitivity (Fig. 2E and Supplementary Fig. 1). Indeed, Western blot analysis of AKT and IRS-1, the two critical insulin-signaling molecules, revealed that epididymal adipose tissue originated from Mbd2^−/− mice was featured by the significantly increased phosphorylated AKT (p-AKT Thr³⁰⁸) and IRS-1 (p–IRS-1 Tyr⁹⁸⁹) as compared with their control counterparts following 16 weeks of HFD induction, whereas no perceptible difference was noted between Mbd2^−/− mice and littermate controls in terms of their expressions (Fig. 2F). Collectively, our results support that loss of Mbd2 improves glucose tolerance and insulin sensitivity following HFD induction.

Next, we examined the effect of Mbd2 deficiency on HFD-induced hyperlipemia and hepatic steatosis. Remarkably, mice deficient in Mbd2 were manifested by the significantly lower levels of plasma triglycerides (TG) and total cholesterol (TC) than their littermate controls after 16 weeks of HFD induction, and particularly, significantly lower TC levels were even noted in Mbd2^−/− mice under ND (Fig. 3A). In line with this observation, hepatic steatosis was detected in HFD-induced WT controls as featured by the significant change of liver color (Fig. 3B, left panel) and weight (Fig. 3B, right panel) as compared with that of Mbd2^−/− mice. Indeed, a marked intrahepatic lipid accumulation was noted in HFD-induced littermate controls, whereas it was significantly attenuated in Mbd2^−/− mice, as determined by H&E and Oil Red O staining (Fig. 3C). Moreover, the control littermates displayed a 3.6-fold higher level of hepatic TG than that of Mbd2^−/− mice (Fig. 3D). Taken together, these data provided evidence suggesting that loss of Mbd2 also protects mice against HFD-induced dyslipidemia and hepatic steatosis.

To further assess the impact of Mbd2 deficiency on the development of obesity, we measured metabolic index by assessing RER of mice housed in metabolic cages. Remarkably, HFD-induced Mbd2^−/− mice showed significantly higher RER than that of control mice (Fig. 4A), whereas they manifested comparable RER as control mice once they were under ND condition (Fig. 4B and Supplementary Fig. 2), indicating that WT mice displayed abnormal fat metabolism upon HFD induction as indicated by the decreased RER, whereas loss of Mbd2 attenuated HFD-induced RER decrease. Consistently, the size of epididymal adipocytes was significantly larger in littermate controls (Fig. 4C) as manifested by the 1.4-fold higher mean adipocyte area than that of Mbd2^−/− mice (Fig. 4C) (9,087 ± 240 vs. 3,810 ± 355 μm²) following HFD induction, which was consistent with the aforementioned reduced epididymal adipose mass in Mbd2^−/− mice (Fig. 1D). We then checked plasma leptin, an adipocyte-derived hormone. HFD-induced littermate controls displayed a 4.5-fold increase for plasma leptin, whereas no perceptible change in terms of plasma leptin in Mbd2^−/− mice was noted following HFD induction (Fig. 4D).

Next, we examined the expression of genes associated with energy expenditure by real-time PCR. Similarly, no discernible difference for the expression of Ucp1, Pgc1-a, and Ppar-a was noted between Mbd2^−/− and control littermates under ND, whereas significantly higher levels of Ucp1, Pgc1-a, and Ppar-a mRNA were detected in Mbd2^−/− mice following HFD induction, and particularly, Mbd2^−/− mice showed significant upregulation of Ucp1 mRNA once fed with HFD (Fig. 4E). Indeed, Western blot analysis of epididymal adipose tissues revealed that Mbd2^−/− mice were featured by the enhanced lipolysis and β-oxidation as evidenced by the higher levels of p-HSL, ATGL, and CPT1 than that of littermate controls (Fig. 4F). Comparable results were obtained for analysis of tissue lysates derived from skeletal muscle and liver (Supplementary Figs. 3 and 4).

Given that obesity is generally associated with inflammation in the visceral adipose tissues, we thus next selectively examined several inflammatory markers. It was interestingly noted that mRNA levels for F4/80, Mcp-1, and Tnf-α were significantly lower in epididymal WAT (eWAT) originated from Mbd2^−/− mice as compared with that of control littermates (Fig. 4G). Together, our data support that loss of Mbd2 protects mice against HFD-induced alteration for energy homeostasis and inflammatory response in adipose tissues.

To confirm the above observations, we bred the Mbd2 null allele into ob/ob mice to generate Mbd2 and leptin double-KO (Mbd2-ob/ob) mice. Remarkably, Mbd2-ob/ob mice gained substantially less weight than that of ob/ob mice (Fig. 5A). Consistent with this observation, Mbd2-ob/ob mice displayed significantly smaller size for adipocytes in epididymal adipose tissues than that of ob/ob mice (Fig. 5B) (9,722 ± 309.2 vs. 11,720 ± 504.4 m²). Similarly, lipid deposition in the liver was significantly attenuated in Mbd2-ob/ob mice (Fig. 5C), along with a marked reduction for the levels of hepatic TG (Fig. 5D). Consistently, Mbd2-ob/ob mice were characterized by the significantly improved glucose tolerance (Fig. 5E) and insulin sensitivity (Fig. 5F) as compared with that of ob/ob mice. Collectively, these results suggest that loss of Mbd2 also provides protection for ob/ob mice against obesity and insulin resistance.

To dissect the mechanisms by which Mbd2 deficiency prevents obesity and insulin resistance, we first compared global DNA methylation levels in epididymal adipose tissues between normal and HFD-induced mice. Interestingly, HFD induced a global DNA hypomethylation as manifested by the reduction of total 5-methylcytosine (5-mC) levels as compared with that of mice with ND (Fig. 6A). Importantly, a similar trend was noted by the analysis of genomic DNA originated from omental adipose tissues of obese patients and normal subjects (Fig. 6B), indicating that the development of obesity is associated with changes of DNA methylation levels and/or patterns (methylation turnover). To exclude that the above global methylation changes were not a result from shift of infiltrated inflammatory cells such as macrophages, we isolated mature adipocytes from epididymal adipose tissue of obese mice and normal mice, followed by analysis of global DNA methylation as above, and comparable results were obtained (Supplementary Fig. 5). In line with these observations, HFD attenuated the expression of MBD2, a reader for DNA methylome-encoded information (12,27), in the eWAT, liver, and skeletal muscle (Fig. 6C). Of note, we also compared global DNA methylation levels between WT and Mbd2^−/− mice fed with normal or HFD and failed to detect a perceptible difference (data not shown). Given that MBD2 itself does not affect the methylation of DNA (7,12), our results indicate that MBD2 modulates HFD-induced obesity and insulin resistance by deciphering the information resulted from DNA methylation changes.

To demonstrate the detailed information of above HFD-induced DNA methylation turnover, eWAT MBD2–Methyl–CpG DNA complexes were first precipitated from mice fed with HFD and ND, followed by high-throughput DNA sequencing (ChIP-seq). Interestingly, 11,218 peak-related genes were actually shared by both normal and HFD-induced mice, whereas 7,439 peak-related genes were noted to be HFD specific, and 1,069 genes were associated with ND (Fig. 6D). Methylation-associated representative genes relevant to energy metabolism are listed in Supplementary Tables 5 and 6. Indeed, the color of heat map and curve associated with DNA methylation levels and/or patterns generated by bioinformatics analysis indicated a typical DNA methylation turnover following HFD induction (Fig. 6E).

Mbd2 ChIP-PCR for selective analysis of promoter regions of above-identified genes demonstrated positive results, confirming that these methylated peaks are the endogenous binding sites for Mbd2 (data not shown). To further dissect the mechanisms by which HFD-induced DNA methylation turnover predisposes to the development of obesity, we defined gene panels for energy storage and expenditure as described, which allowed us to characterize 159 genes associated with energy expenditure (Supplementary Table 10) and 216 genes relevant to energy storage (Supplementary Table 9). Analysis of total signals of ChIP-seq data in the promoter region (5K up transcription start site) of two panels failed to detect a significant difference in terms of methylation state for most of the genes between HFD- and ND-induced mice. However, once we analyzed the data again with a 1.5-fold cutoff, a significant reduction for methylation levels of genes associated with energy storage was noted after HFD induction as evidenced by the box-plot analysis of energy storage and expenditure genes (Fig. 7A). Although HFD induced a trend for decreased methylation levels of energy expenditure genes, it was without a statistical significance (Fig. 7A). Collectively, these data support that HFD induces DNA demethylation of genes both for energy storage and expenditure, but its potency to induce DNA demethylation of genes associated with energy storage is much higher. To confirm this notion, we compared the enrichment of peaks between HFD- and ND-induced mice and found that HFD-induced mice manifested 2,836 decreased peaks and 2,103 increased peaks as compared with ND-induced mice. Gene ontology (GO) analysis for differential peaks of related genes demonstrated mitogen-activated protein kinase (MAPK) signaling pathway was significant enriched in the decreased peaks (Fig. 7B). Given the role of MAPK signaling pathway played in obesity and insulin resistance (28–30), the ChIP-seq data after GO analysis also support that HFD-induced DNA hypomethylation of genes associated with energy storage.

We then sought to confirm the methylation state of above-identified genes by selectively analyzing the promoters of representative genes. Raptor, a typical gene involved in energy storage, and Ucp1, the most critical gene associated with energy expenditure, were selected for this analysis. Indeed, HFD induced a DNA hypomethylation for Raptor and Ucp1 promoters as evidenced by the lower total methylation levels in mice following HFD induction (Fig. 7C and Supplementary Fig. 6). Bioinformatic analysis revealed that CpG site at position −3,881 of Raptor promoter (start codon as +1) contains a potential Cebp-α binding site (Supplementary Fig. 7), whereas CpG site at position −3,302 of Ucp1 promoter (Start codon as +1) contains a potential ATF6 binding site (Supplementary Fig. 8). ChIP assays confirmed that Cebp-α selectively bound to Raptor promoter at CpG site of −3,881 (Fig. 7D, top panel), and ATF6 selectively bound to Ucp1 promoter at CpG site of −3,302 (Fig. 7D, bottom panel).

The above results prompted us to examine with focus for the methylation levels of those two CpG sites following HFD induction. It was interestingly noted that CpG site at position −3,881 of Raptor promoter was highly demethylated in HFD-induced mice (methylation levels: 55 vs. 90%) (Fig. 7E, left panel). In sharp contrast, analysis of CpG site at position −3,302 of Ucp1 promoter failed to detect such a great difference in terms of DNA demethylation (methylation levels: 80 vs. 95%) (Fig. 7E, right panel), which in fact was similar to the methylation levels from analysis of Ucp1 promoter (Fig. 7C, right panel). Similar as above, analysis of Mbd2^−/− mice revealed similar methylation rates for the above-indicated CpG sites following HFD induction (data not shown). In line with this observation, Western blot analysis of eWAT lysates detected significantly higher levels of Raptor in HFD-induced mice (Fig. 7F), and PCR analysis of Raptor mRNA obtained similar results (Supplementary Fig. 9). In contrast, Western blotting still failed to detect Ucp1 in eWAT of HFD-induced mice (data not shown). Together, those data support the notion that HFD is more potent to induce DNA hypomethylation of genes associated with energy storage.

Finally, we have cloned Raptor promoter into a pGL-3 vector (pGL-Raptor wt) as described to confirm that MBD2 modulation of obesity is associated with HFD-induced DNA methylation turnover. A mutated Raptor promoter reporter vector was also constructed, in which cytosines in all CpG sites at the ChIP-seq peak region between −4,089 and −3,874 were mutated into adenosine (pGL-Raptor mut). CpG sites within the reporter vectors were methylated by SssI as described. 3T3-L1 cells were first transduced with Mbd2 adenoviruses and then transfected with above reporters, respectively. As expected, pGL-Raptor mut showed lower reporter activities than pGL-Raptor wt under unmethylated condition. However, treatment of reporters with SssI resulted in a significant reduction for the reporter activities, but SssI treated pGL-Raptor mut displayed similar reporter activities as that of its pGL-Raptor wt counterpart (Fig. 7G), indicating that methylation of CpG sites attenuated reporter activities. We next conducted reporter assays for Ucp1 promoter, and consistent results were obtained (Fig. 7H). Taken together, our results suggest that HFD induces a DNA methylation turnover, and MBD2 modulates HFD-induced obesity by reading the information encoded by this methylation turnover.

Previous studies including ours demonstrated evidence indicating that environmental insult-induced alterations of DNA methylome are implicated in the pathoetiology of complex diseases (3,7,31). Particularly, we found that ischemic insult induced endothelial cells to undergo a DNA methylation turnover as manifested by the changes of methylation levels and patterns, and MBD2 regulates angiogenesis by deciphering the information resulted from this DNA methylation turnover (7). Obesity is a multifactorial chronic disease, which involves altered homeostasis of energy storage and expenditure via physiological processes (32). Specifically, alterations in energy homeostasis could be caused by the changes of epigenetic factors such as DNA methylation, which can be induced by inadequate dietary habits, diminished physical exercise, and genetic background (33). Despite past extensive studies, the exact impact of dietary style on DNA methylome relevant to the pathogenesis of obesity, however, is yet to be fully addressed. Given that consumption of unhealthy foods such as high carbohydrates and high fat with low fiber is considered the leading cause for developing metabolic disorders, we thus used Mbd2^−/− mice to dissect the impact of DNA methylation on the pathoetiology of HFD-induced obesity and insulin resistance.

It was interestingly noted that mice deficient in Mbd2 were protected from HFD-induced obesity, which was not caused by the reduction of less food intake, and in fact, food consumption was found to be slightly higher in Mbd2^−/− mice than that of control mice. We also excluded the possibility that the phenotype was caused by retarded growth, as there was no perceptible difference in terms of the length for tibias between Mbd2^−/− mice and control littermates. In fact, previous studies have already suggested that loss of Mbd2 does not affect murine development (12). Because obesity is generally associated with an array of metabolic abnormalities termed “metabolic syndrome” such as insulin resistance and hyperlipidemia, we also noted that Mbd2^−/− mice were protected from HFD-induced insulin resistance, hyperlipidemia, and hepatosteatosis. Particularly, the reduced adipocyte size characterized in Mbd2^−/− mice could be the main cause of attenuated adipose tissue accumulation following HFD induction. Indeed, it has been widely accepted that adipocyte size is associated with TG accumulation (34). To confirm the above phenotype, we then bred Mbd2 null allele into ob/ob (Mbd2-ob/ob) mice. Consistently, Mbd2-ob/ob mice displayed significantly less amount of weight gain along with improved glucose homeostasis and lipid deposition in the liver. Collectively, we demonstrated convincing evidence that MBD2 regulates glucose homeostasis and lipid metabolism implicated in the pathogenesis of HFD-induced obesity and insulin resistance.

To dissect the mechanisms by which loss of Mbd2 represses HFD-induced obesity, we first compared total DNA methylation state in epididymal adipose tissues between control and HFD-induced mice. Interestingly, HFD induced a global DNA hypomethylation in epididymal adipose tissues, and a similar trend was also noted in omental adipose tissues originated from obese patients. These results are actually consistent with previous studies, in which oxidative stress leads to a global DNA hypomethylation by interfering with the ability of DNA to function as a substrate for DNA methyltransferases (35,36). Surprisingly, attenuated MBD2 expression was characterized in the epididymal adipose tissues, skeletal muscle, and liver following HFD induction. The cause by which HFD attenuates MBD2 expression is currently unknown. Given that loss of Mbd2 prevents HFD-induced obesity, the attenuated MBD2 expression could be a compensated response due to HFD-induced imbalance of energy homeostasis. Nevertheless, additional studies would be necessary to address the exact mechanism underlying this phenomenon.

MBD2-based ChIP-seq assays were next conducted to characterize target genes with manifestation of DNA methylation changes following HFD induction. Indeed, the assays characterized a large number of genes relevant to energy storage and expenditure. Importantly, those genes exhibited unique changes either in DNA methylation levels or patterns, indicating that HFD rendered epididymal adipose tissues to undergo a DNA methylation turnover. Bioinformatic analysis of ChIP-seq data characterized 216 genes associated with energy storage and 159 genes relevant to energy expenditure, in which a significant decrease of DNA methylation levels for those genes associated with energy storage was noted. Importantly, selective analysis of Raptor, a specific and essential component of mammalian target of rapamycin complex 1 that positively regulates adipogenesis, lipogenesis, and glucose uptake (37), and Ucp1, an essential gene for energy expenditure (38), by bisulfite DNA sequencing confirmed our ChIP-seq data. It was interestingly noted that although HFD-induced DNA hypomethylation for both genes involved in energy storage and expenditure, the extent of its impact on the induction of DNA hypomethylation in genes relevant to energy storage was much more potent. Indeed, GO pathway analysis revealed that MAPK signaling pathway was significantly hypomethylated. Given the role the MAPK signaling pathway played in adipogenesis, obesity, and insulin resistance, those data further support that HFD is more potent to induce DNA hypomethylation of energy storage genes in favor of obesity development.

It is worthy of note that the discrepancy in terms of HFD-induced DNA hypomethylation between genes responsible for energy storage and expenditure is likely associated with the binding sites for transcription factors. For example, the extent of DNA hypomethylation for CpG site at position −3,881 of Raptor promoter was significantly higher than that of CpG site at position −3,302 of Ucp1 promoter as compared with their corresponding counterpart. Bioinformatic analysis and ChIP assays revealed that CpG site at position −3,881 of Raptor promoter contains a binding site for Cebp-α, whereas CpG site at position −3,302 of Ucp1 promoter contains a binding site for ATF6. Importantly, it seems that this methylation discrepancy contributed to the enhanced Raptor expression following HFD induction, as determined by the methylation-dependent promoter reporter assays and Western blot analysis of epididymal adipose tissues. In line with our assumption, the DNA methylome settings have also been noted to be the ultimate integration sites of both environmental and differentiative inputs, which preferentially determine the transcription of nuclear factor-κB–dependent genes such as interleukin-1β and tumor necrosis factor-α (39–42). Taken together, our studies indicate that HFD induced a global DNA hypomethylation, but its impact on the induction of DNA hypomethylation in genes responsible for energy storage is much more potent than that in genes associated with energy expenditure. MBD2 implicates in the pathogenesis of HFD-induced obesity and insulin resistance by reading the information resulted from those DNA methylation changes. Therefore, once MBD2 is depleted, the effect of this methylation turnover cannot be implemented, which then promotes energy expenditure to prevent HFD-induced obesity. It is noteworthy that the changes of DNA methylation profiles identified in eWAT are unlikely caused by the shift of infiltrated immune cells such as macrophages, because the isolated mature adipocytes from eWAT of HFD induced obese mice and normal mice displayed comparable results as the ChIP-seq data.

In summary, we demonstrated evidence that HFD renders epididymal adipose tissues to undergo a DNA methylation turnover as manifested by the changes of methylation levels and/or patterns. MBD2 interprets the information encoded by this DNA methylation turnover for regulation of an array of genes to alter the homeostasis of energy storage and expenditure in favor of obesity development. Therefore, loss of Mbd2 provides protection for mice against HFD-induced obesity and insulin resistance. Given that MBD2 itself does not affect DNA methylation and is dispensable for daily life processes, our results suggest that MBD2 could be a viable epigenetic target for developing more efficacious and cost-effective therapies for prevention and treatment of obesity in clinical settings.

Acknowledgments. The authors thank Dr. Adrian Bird (Wellcome Trust Centre for Cell Biology, University of Edinburgh, Edinburgh, U.K.) for providing the Mbd2 KO mice and Dr. Qi-Qun Tang (Department of Biochemistry and Molecular Biology, Fudan University Shanghai Medical College, Shanghai, China) for providing the C3H10T 1/2 cell line. The authors also thank Dr. Wenye Mo (Center for Biomedical Research, Tongji Hospital, Tongji Medical College, Huazhong University of Science and Technology) for animal studies.

Funding. This study was supported by the National Natural Science Foundation of China (grants 81130014, 81428001, and 81530024), the European Foundation for the Study of Diabetes/Chinese Diabetes Society/Lilly Program for Collaborative Diabetes Research between China and Europe, the Program for Changjiang Scholars and Innovative Research Team in University (IRT_14R20), and Innovative Funding for Translational Research from Tongji Hospital.

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

Author Contributions. J.C. and J.S. were responsible for conducting all experiments and data analysis and wrote the manuscript. X.H., M.Z., S.H., and P.Y. were responsible for conducting all experiments and data analysis. S.Z. was responsible for conducting all experiments and data analysis and contributed to discussion and review of the manuscript. Q.Y., F.X., and D.W.W. were involved in review of the manuscript. J.Z., Q.N., Z.C., and Z.Z. contributed to discussion and review of the manuscript. D.L.E. contributed to discussion and review of the manuscript, study design, and manuscript preparation. C.Z. was responsible for conducting all experiments and data analysis and contributed to discussion and review of the manuscript, study design, and manuscript preparation. C.-Y.W. contributed to the study design and manuscript preparation. C.-Y.W. 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.