Hepatic Gadd45β promotes hyperglycemia and glucose intolerance through DNA demethylation of PGC-1α

The liver is a dynamic organ that plays a key role in the regulation of systemic glucose metabolism. During periods of nutrient deprivation, the liver is the main source responsible for maintaining euglycemia through hepatic glucose production (HGP). HGP consists of glycogenolysis and gluconeogenesis, which breaks down glycogen and de novo synthesizes glucose from noncarbohydrate precursors, respectively (Petersen et al., 2017). In the fasted state, hepatic gluconeogenesis is activated to maintain blood glucose levels within a normal range, ensuring enough energy supply for glucose-dependent tissues like the brain. However, uncontrolled gluconeogenesis is believed to be one of the major factors that contribute to hyperglycemia and glucose intolerance in obesity and type 2 diabetes mellitus (T2DM). Therefore, inhibiting HGP, especially by decreasing gluconeogenesis, is a potentially appealing avenue for the treatment of diabetes (Rines et al., 2016).

The rate of hepatic gluconeogenesis is mainly controlled by the pancreatic hormone glucagon and adrenal hormone glucocorticoids (GCs). Glucagon activates the cAMP/protein kinase A/cAMP-responsive element binding protein (CREB) signaling pathway, while GCs activate the nuclear receptor glucocorticoid receptor (GR; Goldstein and Hager, 2015). These two signals act either independently or synergistically to activate many transcription factors (e.g., hepatocyte nuclear factor 4α [HNF4α], FoxO1, KLF15, and CREB protein H) that stimulate the expression of two key gluconeogenic enzymes: phosphoenolpyruvate carboxykinase (PEPCK) and glucose-6-phosphatase (G6Pase; Goldstein and Hager, 2015). Of interest, proper metabolic gene regulation and the transactivation of transcription factors require the recruitment of coregulatory proteins (Mouchiroud et al., 2014), such as peroxisome proliferator–activated receptor γ coactivator-1α (PGC-1α). The gluconeogenic effect of PGC-1α is achieved by directly coactivating HNF4α and FoxO1, leading to an increase in the transcription of gluconeogenic genes and HGP (Lin et al., 2005; Piccinin et al., 2019). In contrast, both PGC-1α–KO mice and acute liver-specific PGC-1α–knockdown mice displayed hypoglycemia under a fasted state (Lin et al., 2004; Koo et al., 2004). Moreover, hepatic PGC-1α is markedly up-regulated in multiple models of diabetes (Yoon et al., 2001), and selective chemical inhibition of PGC-1α suppressed hepatic gluconeogenesis and alleviated hyperglycemia in diabetic mice (Sharabi et al., 2017), highlighting its importance in systemic glucose homeostasis. In addition to the transcriptional regulation, epigenetic changes, including DNA methylation and histone modifications, emerge as an important signature involved in reprogramming of metabolic pathways (Ling and Rönn, 2019). For instance, alterations in DNA methylation of multiple gene loci have been linked to the development of obesity and T2DM by epigenome-wide association studies (Wahl et al., 2017). Besides, specific expression and methylation differences were observed for several genes coding metabolic enzymes between distinct liver phenotypes (Ahrens et al., 2013). Nevertheless, our understanding of the importance of epigenetic function in the hepatic gluconeogenesis and HGP is relatively limited.

In the present study, we sought to explore the role of the epigenome in HGP using two mouse models of HGP: dexamethasone (Dex)-treated mice and fasted mice. Dex and other synthetic GC analogues have been used clinically for decades to treat many types of autoimmune diseases as anti-inflammatory drugs and also employed to prevent organ transplant rejection as immunosuppressants. Unfortunately, elevated GC levels are associated with hyperglycemia and glucose intolerance, which becomes one of the key limitations of the long-term use of GC drugs (Rose and Herzig, 2013). As a result, we found that both Dex treatment and fasting induce the expression of growth arrest and DNA damage-inducible β (Gadd45β), which has been implicated in the active DNA demethylation of specific loci in vertebrates (Ma et al., 2009; Schüle et al., 2019). Using liver-specific Gadd45β overexpression and deficient mice, we further demonstrate the physiological and pathological significance of Gadd45β-mediated gluconeogenesis. Our results illustrate an example of how hepatic epigenetic changes coordinate HGP and systemic glucose homeostasis, which may have important implications for the treatment of diabetes, including steroid-associated diabetes.

To identify potential genes that may contribute to the development of GC-associated hyperglycemia, we performed RNA sequencing (RNA-seq) analysis of the livers collected from two mouse models (Fig. S1, A and B). In the first model, C57BL/6 mice were treated with Dex (1.0 mg/kg) daily for 14 d (Fig. S1 A). This dose of Dex has been shown to cause hyperglycemia in mice by our previous work (Lu et al., 2012). In the second model, C57BL/6 mice were fed ad libitum or were fasted for 16 h (Fig. S1 B). Differentially expressed genes were identified using a q value <0.05 and fold change >1.5. The q value is an adjusted P value, taking into account the false discovery rate. In the livers of Dex-treated mice, 1,107 genes were significantly differentially expressed compared with control mice. Of these genes, 129 showed increased expression, and 978 showed decreased expression. As expected, Dex treatment induced the expression of gluconeogenic genes, such as Ppargc1a (coding PGC-1α) and Pck1 (coding PEPCK; Table S1). Besides, the most highly up-regulated genes included a range of novel genes but also those known to be regulated by GCs, such as Cyp2b10 (Jarukamjorn et al., 2001), Mt2 (Wray et al., 2019), and Lcn2 (Owen et al., 2008; Table S1). In fasted livers, 612 genes were significantly differentially regulated compared with the livers of ad libitum-fed mice. Of these genes, 207 were up-regulated, and 405 were down-regulated. When comparing both mouse models, we identified 21 genes that were commonly up-regulated and 56 genes that were commonly down-regulated (Fig. S1, C and D; and Table S2 and Table S3).

Of note, our screen revealed a pronounced up-regulation of Gadd45β in the livers of Dex-treated and fasted mice compared with the corresponding controls (Table S2). Increased Gadd45β expression in both mouse models was further confirmed by quantitative real-time PCR (qRT-PCR), which showed a 21.76-fold up-regulation of Gadd45β expression in the livers of Dex-treated mice and an 8.98-fold up-regulation of Gadd45β expression in the livers of fasted mice (Fig. 1 A and Fig. S1 E). Western blotting analysis confirmed that hepatic Gadd45β protein was up-regulated in both models (Fig. 1 B and Fig. S1 F). In contrast, expression levels of Gadd45α and Gadd45γ, two other members of the Gadd45 family, were not affected by Dex or fasting (Fig. S1, G and H). Moreover, qRT-PCR and Western blotting analysis showed that the mRNA and protein levels of Gadd45β were increased in mouse primary hepatocytes (MPHs) following exposure to Dex (Fig. 1, C and D). Furthermore, we observed a dose-dependent induction in Gadd45β mRNA expression following Dex treatment (Fig. S1 I). Time-course experiments revealed that Gadd45β mRNA expression was rapidly induced in MPHs: in as little as 30 min, expression was significantly increased over baseline levels (Fig. S1 J).

Gluconeogenesis can also be induced by glucagon, chronic inflammation, and persistent ER stress. Thus, we examined the ability of forskolin (which mimics glucagon), TNFα (which mimics chronic inflammation), and tunicamycin (which mimics ER stress) to activate Gadd45β expression. As a result, these factors failed to up-regulate Gadd45β expression in MPHs (Fig. S1, K–M). Therefore, Gadd45β may specifically respond to steroid hormone signaling in the process of hepatic gluconeogenesis.

The effects of GCs are mediated through activation of the GR, a member of the nuclear hormone receptor superfamily. Treatment with low-dose RU486, a dual antagonist of GRs and progesterone receptors (PRs), decreased Dex-induced Gadd45β expression in MPHs, and high-dose RU486 almost completely blocked it (Fig. 1 E). Given that PR expression is absent in hepatocytes (Lu et al., 2020; Yang et al., 2006), the effects of RU486 suggest that the transactivation of Gadd45β by Dex is mediated by the GR. These results were confirmed by adenoviral shRNA targeting of the GR. Knockdown of endogenous GR reduced basal Gadd45β expression and attenuated Dex-mediated induction of Gadd45β in MPHs (Fig. 1 F).

To further demonstrate the role of the GC–GR axis in the regulation of Gadd45β, C57BL/6 mice were administered GR shRNA or control shRNA, respectively (Fig. S1 N). Then, we measured the rhythm of plasma GC levels and compared hepatic Gadd45β expression in two groups of mice. Consistent with a previous report (Yang et al., 2006), hepatic GR expression was constant throughout the day and night (Fig. S1 N). A significant diurnal pattern for plasma corticosterone concentrations was observed in two groups of mice (Fig. S1 O). As expected, hepatic Gadd45β mRNA showed a circadian expression similar to that of plasma corticosterone in control shRNA-infected mice (Fig. S1 P, left). However, the basal expression of Gadd45β mRNA was significantly decreased in mice infected with GR shRNA, and the circadian pattern of Gadd45β expression ceased (Fig. S1 P, right).

We further explored the molecular basis of Gadd45β up-regulation by the GC–GR signaling pathway using an online transcription factor scanning system (PROMO software; Messeguer et al., 2002). Bioinformatic analysis showed that the promoter region of the Gadd45β gene, spanning from −2,000 to +1 bp (transcription start site), contains two potential half–glucocorticoid receptor response element (half-GRE) binding sites (Fig. 1 G). When this region was cloned and inserted into a luciferase reporter gene, transfection of a GR expression plasmid activated the luciferase, and this activation was robustly enhanced by Dex treatment (Fig. 1 H). Mutation of the reporter constructs clearly showed that each GRE is functional; mutation of any site decreased, while mutation of both sites blocked the transactivation role of GCs (Fig. 1 H). A chromatin immunoprecipitation (ChIP) assay further showed that the GR could bind to both sites in MPHs in the presence of Dex (Fig. 1 I). We also confirmed the GC-mediated recruitment of endogenous GR proteins onto the Gadd45β promoter region in the livers of Dex-treated mice and fasted mice (Fig. 1, J and K). Together, these results indicate that GCs up-regulate Gadd45β expression through the GR.

Next, we studied the functions of hepatic Gadd45β in the regulation of glucose metabolism. An adenovirus (Ad) containing Gadd45β (Ad-Gadd45β) was prepared and administered into C57BL/6 mice via tail vein injection (Fig. 2 A). An equal amount of Ad-GFP was used as a control. As a result, blood glucose levels were significantly elevated in fasted mice transfected with Ad-Gadd45β compared with mice transfected with Ad-GFP (Fig. 2 B). The hepatic glycogenolysis and gluconeogenesis pathways are the main sources of endogenous glucose production to maintain blood glucose levels in a fasted state (Rines et al., 2016). Thus, we first examined the glycogenolysis pathway. No significant difference was detected in the hepatic glycogen content of mice transfected with Ad-Gadd45β versus those transfected with Ad-GFP (Fig. S2 A). In addition, genes related to glycogenolysis were not altered (Fig. S2 B). In contrast, the pyruvate tolerance test (PTT) showed that hepatic gluconeogenesis was enhanced in mice transfected with Ad-Gadd45β (Fig. 2 C). A glucose tolerance test (GTT) indicated that Ad-Gadd45β–infected mice developed impaired glucose tolerance (Fig. 2 D). In agreement, mRNA levels of the gluconeogenic enzymes PEPCK and G6Pase were markedly increased in the livers of Gadd45β-expressing mice (Fig. 2 E).

To further confirm the gluconeogenic effects of Gadd45β in vitro, MPHs were transfected with Ad-Gadd45β or Ad-GFP (Fig. S2 C). Overexpression of Gadd45β promoted cellular glucose production and increased the mRNA levels of gluconeogenic enzymes (Fig. S2, D and E). Thus, Gadd45β appears to function as an important regulator of gluconeogenesis to promote HGP.

Gadd45β has been shown to activate gene expression by reducing site-specific DNA methylation (Ma et al., 2009; Schüle et al., 2019). Therefore, to explore the molecular mechanism involved in Gadd45β regulation of hepatic gluconeogenesis, we performed DNA methylation profiling to analyze the methylomes of livers treated with Dex or vehicle control by methylated DNA immunoprecipitation sequencing (MeDIP-seq). The results showing significant differences in DNA methylation levels are presented as a heatmap (Fig. S2 F). We identified 751 differentially methylated regions (DMRs) between Dex- and vehicle control–treated livers (Fig. S2 G). Strikingly, 394 of these 751 DMRs (52.46%) showed increased methylation levels, while 357 (47.54%) were hypomethylated (Fig. S2 H). To determine whether DMRs correlated with changes in gene expression, we compared MeDIP-seq with RNA-seq, as depicted in Fig. S1 A. This comparison revealed 20 DMR-associated genes that were also differentially expressed (Fig. S2 H). Notably, Ppargc1a, located at chromosome 5, is one of the DMR-associated hypomethylated genes that were up-regulated (Fig. S2, H and I). Ppargc1a encodes a protein called PGC-1α, a well-established transcriptional coactivator in hepatic gluconeogenesis (Yoon et al., 2001). Therefore, we asked whether PGC-1α is a downstream target of Gadd45β in the liver.

Our qRT-PCR analysis confirmed that mRNA levels of PGC-1α were markedly elevated in the livers of mice transfected with Ad-Gadd45β (Fig. 2 F), while the expression of other transcriptional coactivators or transcription factors involved in gluconeogenesis, such as CRTC2, steroid receptor coactivators, HNF4α, forkhead box Os, and Krüppel-like factors, remained unchanged (Fig. 2, F and G). The induction of PGC-1α expression was further confirmed by Western blotting analysis (Fig. 2 H). Overexpression of Gadd45β also increased PGC-1α mRNA and protein expression in MPHs (Fig. S2, J and K). To determine whether the gluconeogenic effects of Gadd45β were PGC-1α dependent, an Ad encoding an shRNA targeting PGC-1α was generated and transfected into MPHs (Fig. S2 L). Knockdown of endogenous PGC-1α significantly attenuated the effects of Gadd45β to induce the expression of PEPCK and G6Pase (Fig. S2). Consistently, enhanced cellular glucose production by Gadd45β was also largely blocked by depletion of PGC-1α (Fig. S2 M). Thus, we can conclude that PGC-1α is essential for the gluconeogenic functions of Gadd45β in hepatocytes.

Next, we examined whether the induction of PGC-1α expression by Gadd45β can be attributed to changes in DNA methylation. Previous studies have demonstrated that cytosine hypermethylation levels of the PGC-1α promoter were negatively correlated with the mRNA expression and biological functions of PGC-1α (Barrès et al., 2009; Barrès et al., 2012; Laker et al., 2014). This promoter region spans −360 to −60 bp relative to the transcription start site and contains several methylation sites; one major locus is located within the CREB binding site (Fig. S3 A), the most important cis-regulatory element for PGC-1α transcription (Herzig et al., 2001). Our ChIP assays showed that Gadd45β can occupy the PGC-1α promoter region in MPHs, which was enhanced by Dex treatment (Fig. S3 B). Through methylated DNA immunoprecipitation (MeDIP) followed by qRT-PCR (MeDIP-qPCR) and pyrosequencing analysis, we found that the methylation status of the PGC-1α promoter was significantly reduced in the livers of mice infected with Ad-Gadd45β (Fig. 2, I and J). This reduction was also observed in MPHs overexpressing Gadd45β (Fig. S3, C and D). However, global DNA methylation levels were not altered by Gadd45β expression (Fig. S3, E and F). In addition, the methylation status at the promoter region of Dhx15, a gene adjacent to PGC-1α on chromosome 5, remained unaffected in the livers of mice or MPHs overexpressing Gadd45β (Fig. S3, G and H). These results confirm that Gadd45β regulates DNA methylation patterns at specific gene loci. Moreover, Gadd45β did not increase the cytosine methylation of the PEPCK or G6Pase promoter regions in mouse livers or MPHs (Fig. S3, I–L), indicating that Gadd45β indirectly up-regulates their gene expression. In agreement with the gluconeogenic role of Gadd45β, the reduced DNA methylation pattern in the PGC-1α promoter was also found in the livers of Dex-treated or fasted mice (Fig. S3, M–P).

Changes in DNA methylation status usually affect the binding affinity of specific transcription factors and histones, leading to altered gene transcription (Martínez et al., 2014). Using ChIP assays, we found that the interaction of phosphorylated CREB with the PGC-1α promoter, which was activated by cAMP signaling, was dramatically increased when Gadd45β was overexpressed in MPHs (Fig. S3 Q). In addition, we observed a reduction in repressive histones (histone H3 lysine 9 trimethylation [H3K9me3] and H3K27me3) and an enrichment in active histones (H3K4me3) in the PGC-1α promoter in MPHs overexpressing Gadd45β (Fig. S3 R). Together, these data indicate that a reduced DNA methylation status caused by Gadd45β at the PGC-1α promoter region is likely associated with enhanced CREB binding and active histone modifications, resulting in the up-regulation of PGC-1α expression. Therefore, our results also well explained why GCs had a synergistic effect with cAMP signaling to activate PGC-1α expression (Yoon et al., 2001).

Previous studies have shown that TET methylcytosine dioxygenase is responsible for converting 5-methylcytosine (5mC) sequentially to 5-hydroxymethylcytosine (5hmC), 5-formylcytosine, 5-carboxylcytosine, and unmethylated cytosine (Wu and Zhang, 2017). Therefore, we determined the molecular basis in Gadd45β-mediated active demethylation and activation of the PGC-1α gene. Using ChIP assays, we observed a considerable occupancy of TET1, but not TET2 or TET3, at the PGC-1α promoter in the MPHs overexpressing Gadd45β (Fig. 3 A), which was associated with an increment of DNA 5hmC (Fig. 3 B). Consistently, Dex stimulation dramatically increased TET1 binding to the PGC-1α promoter, resulting in increased 5hmC at its promoter (Fig. 3, C and D). Coimmunoprecipitation experiments further showed that there is a direct protein–protein interaction between Gadd45β and TET1 in MPHs (Fig. 3 E). Besides, compared with Gadd45β alone, cotransfection of TET1 with Gadd45β further reduced DNA methylation status at the PGC-1α promoter (Fig. 3, F and G), leading to enhanced expression of PGC-1α and gluconeogenic enzymes (Fig. 3 H). Cellular glucose production was also increased by combination of TET1 with Gadd45β (Fig. 3 I), suggesting a functional cooperation of Gadd45β and TET1. On the other hand, knockdown of endogenous TET1 expression impaired Gadd45β-induced DNA demethylation (Fig. 3, J and K) and reversed Gadd45β-mediated gluconeogenic effects (Fig. 3, L and M). These results indicate that Gadd45β acts as a bridging protein between TET1 and methylated DNA at the PGC-1α promoter region, while TET1-associated DNA demethylation might be crucial for the transcriptional activation of the PGC-1α gene.

To further examine the metabolic function of Gadd45β in vivo, two loxP sites were inserted into the Gadd45β gene flanking exons 1–4 to generate Gadd45β flox/flox (F/F) mice (Fig. S4 A). We then obtained liver-specific Gadd45β-knockout mice (LKO mice) by injection of adeno-associated virus (AAV)–expressing Cre or GFP recombinase into Gadd45β F/F mice through the tail vein. The AAV approach was chosen in our study because (1) AAV-mediated gene transduction did not induce immune responses and toxicity in mice; (2) the Cre or GFP recombinase expression was driven by the liver-specific thyroxine binding globulin promoter; (3) it can persistently be expressed in the liver for >20 wk (Liu et al., 2016; Yazdanyar et al., 2013); and (4) we could carry out this experiment in adult mice to avoid perturbing the proliferation or differentiation of hepatocytes at the developmental stage. Our qRT-PCR and Western blotting analyses confirmed that Gadd45β expression was substantially and specifically reduced in the livers of LKO mice, while its expression was unaffected in other tissues examined (Fig. S4, B and C). No differences were found in body weight, body length, food intake, and percentage of fat mass between Gadd45β LKO and F/F mice (data not shown).

Notably, the fasting glucose levels were significantly decreased in Gadd45β LKO mice (Fig. 4 A). We also found that Gadd45β LKO mice had a greater glucose tolerance compared with F/F mice (Fig. 4 B). HGP, determined by PTT, was lower in LKO than in F/F mice (Fig. 4 C). The lower blood glucose levels in LKO animals are due to impaired gluconeogenesis, as reflected in the reduced expression of PGC-1α and gluconeogenic enzymes (Fig. 4 D). Levels of plasma insulin, glucagon, corticosterone, and liver injury markers (alanine aminotransferase and aspartate aminotransferase) were unchanged (Fig. S4 D).

We then fed two groups of mice a high-fat diet (HFD) for 12 wk, which can induce hyperglycemia and glucose intolerance. Depletion of hepatic Gadd45β in mice did not alter body weight gain; however, LKO mice had a significantly improved glucose metabolism, as shown by reduced fasting blood glucose concentrations (Fig. 4 E). Mice without hepatic Gadd45β also had lower basal blood insulin levels (Fig. 4 F), suggesting insulin hypersensitivity in these mice. Consistently, LKO mice showed markedly improved glucose and insulin tolerance (Fig. 4, G and H). HGP was also lower in the LKO mice than in the F/F mice (Fig. 4 I). In addition, expression levels of gluconeogenic genes were lower in the livers of LKO mice (Fig. 4 J). Thus, these data indicate that liver-specific suppression of Gadd45β could prevent dietary obesity-induced metabolic impairment independently of changes in body weight.

Based on the above results, we speculate that hepatic Gadd45β might also be required for steroid-induced hyperglycemia. To test this hypothesis in vivo, we treated Gadd45β LKO and F/F mice with Dex or vehicle control for 14 d. Body weights and food intake were comparable between each subgroup (Fig. S4, E and F). Dex treatment led to hyperglycemia in Gadd45β F/F mice (Fig. 5 A, column 2 versus column 1), but not in Gadd45β LKO mice (Fig. 5 A, column 4 versus column 3). Moreover, Dex-induced HGP and glucose intolerance were largely abrogated by hepatic Gadd45β deficiency (Fig. 5, B–E). Thus, ablation of Gadd45β in the liver can protect against Dex-associated hyperglycemia. Consistent with this notion, we found that Dex treatment can induce the expression of PGC-1α due to decreased DNA methylation at its promoter region in Gadd45β F/F mice, but not in LKO mice (Fig. 5, F and G). Additionally, induction of PEPCK and G6Pase by Dex was also defective in Gadd45β LKO mice (Fig. 5 H).

To confirm these findings in an independent setting, MPHs were isolated from the livers of Gadd45β F/F and LKO mice, followed by incubation with Dex or vehicle control. In agreement with the in vivo observations, Dex treatment of MPHs from Gadd45β F/F mice resulted in a significant elevation in cellular glucose production and gluconeogenic gene expression (Fig. 5, I and J, column 2 versus column 1). However, these effects were blocked in MPHs from Gadd45β LKO mice (Fig. 5, I and J, column 4 versus column 3). Collectively, our data clearly demonstrate the essential role of Gadd45β in mediating steroid-induced hyperglycemia and hepatic gluconeogenesis.

It has been established that enhanced hepatic gluconeogenesis is a major contributor to hyperglycemia in diabetic patients (Petersen et al., 2017; Rines et al., 2016). Therefore, we speculate that Gadd45β expression would be dysregulated in the livers of diabetic mice. To test it, leptin receptor–deficient db/db mice, leptin-deficient ob/ob mice, and C57BL/6 mice fed an HFD for 12 wk were used. Plasma corticosterone levels are usually higher in db/db and ob/ob mice, compared with normal lean mice (Liu et al., 2003; Liu et al., 2005; Liu et al., 2006). Besides, while no significant differences in circulating corticosterone concentration between mice fed an HFD and normal chow diet, hepatic GR signaling or sensitivity was enhanced in the livers of HFD-induced obese mice (Quagliarini et al., 2019; Dassonvalle et al., 2020). As expected, these mice displayed hyperglycemia and insulin resistance, which were described in our previous study (Lu et al., 2014). The expression of several gluconeogenic genes, including PGC-1α, PEPCK, and G6Pase, was significantly increased in the livers of these mice compared with their corresponding control mice (Fig. S4, G–I). Furthermore, qRT-PCR and Western blot analysis showed that hepatic Gadd45β expression levels were significantly up-regulated in db/db mice compared with lean controls (Fig. 6, A and B). Similar results were observed in ob/ob mice and HFD mice (Fig. 6, C–F). Besides, either RU486 treatment or GR shRNA infection led to a decreased expression of Gadd45β in the livers of HFD-induced obese mice, indicative of the important role of GC–GR signaling in the regulation of HFD-associated hepatic Gadd45β expression (Fig. S4, J–M). Consistent with these findings, we found that the DNA methylation status was reduced at the PGC-1α promoter in all three types of obese mice (Fig. S4, N–P).

In addition, we obtained human liver samples from patients with T2DM and from nondiabetic normal individuals through ultrasonography-guided percutaneous needle liver biopsies. qRT-PCR analysis showed that expression levels of Gadd45β, PGC-1α, and gluconeogenic enzymes (PEPCK and G6Pase) were up-regulated in diabetic patients compared with normal subjects (Fig. 6 G). Moreover, there was a positive correlation between Gadd45β mRNA levels and both fasting plasma glucose and hemoglobin A1c (HbA1c) levels (Fig. 6, H and I). A significant correlation between mRNA levels of Gadd45β and gluconeogenic genes was also observed (Fig. S4, Q–S). Thus, these results indicate that elevated hepatic Gadd45β expression is associated with the development of hyperglycemia in T2DM.

The dysregulation of hepatic Gadd45β in diabetic mice and human subjects prompted us to investigate whether its deficiency could improve glucose homeostasis. We transfected db/db mice with Gadd45β-specific adenoviral shRNA via the tail vein, which inhibited endogenous Gadd45β expression in the liver (Fig. 7 A). The reduction of Gadd45β caused a marked increase in DNA methylation status of the PGC-1α promoter (Fig. 7, B and C), subsequently leading to its down-regulation (Fig. 7, D and E). In agreement, expression levels of PEPCK and G6Pase were reduced (Fig. 7 F). Compared with negative controls, knockdown of Gadd45β alleviated hyperglycemia and hyperinsulinemia in db/db mice (Fig. 7, G and H), although body weight and food intake remained unchanged (data not shown). In db/db mice with Gadd45β deficiency, PTT and GTT results showed a decrease in gluconeogenesis and an improvement in glucose tolerance, respectively (Fig. 7, I and J). Moreover, insulin sensitivity was enhanced, as evidenced by the results of insulin tolerance tests (ITTs; Fig. 7 K). The beneficial effects on glucose metabolism were also observed in HFD-induced obese mice. Knockdown of Gadd45β in the livers of HFD mice also reduced the expression of PGC-1α and gluconeogenic enzymes due to hypermethylation at the PGC-1α promoter (data not shown). Furthermore, diabetes-associated metabolic syndromes, including hyperglycemia, hyperinsulinemia, glucose intolerance, and insulin resistance, were alleviated (data not shown).

In summary, our results demonstrated that, upon binding of GCs such as cortisol and Dex, GR up-regulates the expression of Gadd45β, which in turn activates PGC-1α gene transcription through DNA demethylation at its promoter region. PGC-1α can further induce the expression of gluconeogenic genes to enhance hepatic gluconeogenesis and glucose production (Fig. 8), leading to hyperglycemia, glucose intolerance, and diabetes.

In the present study, through RNA-seq analysis of two mouse models, we identified Gadd45β as an important potentiating epigenetic factor that activates hepatic gluconeogenesis to enhance HGP. Ad-mediated overexpression of Gadd45β enhanced gluconeogenesis and glucose production in vivo and in primary hepatocytes. In contrast, Gadd45β LKO mice not only showed hypoglycemia in a fasted state, but they were also resistant to HFD-induced and Dex-associated glucose intolerance. Therefore, our findings might be of both physiological and pathophysiological interest, as Gadd45β expression can be induced by starvation, steroid use, and obesity.

Methylation of the PGC-1α promoter can be regulated in skeletal muscle in response to proinflammatory cytokines or free fatty acids through the actions of DNA methyltransferase 3B (Barrès et al., 2009). However, whether the methylation status of PGC-1α promoter is dynamically regulated in the liver remains elusive. Here, we show that fasting, steroid use, and obesity can reduce the methylation at the PGC-1α promoter through the actions of Gadd45β, a member of a small gene family that also includes Gadd45α and Gadd45γ. This family was originally identified as a group of growth-suppressing genes that induce cell cycle arrest and apoptosis (Schäfer, 2013). However, alterations of hepatic Gadd45β activity caused little change in the expression levels of cell cycle–related genes (data not shown). Besides, there were no significant differences in terminal deoxynucleotidyl transferase dUTP nick end labeling incorporation assays in the livers of mice with hepatic Gadd45β overexpression or depletion (data not shown). Of note, all of the in vivo experiments in this study, including overexpression, depletion, and knockdown of Gadd45β, were performed in the livers of adult mice. Therefore, our findings suggest that the metabolic role of Gadd45β in quiescent tissues, such as the liver, might be independent of cell proliferation, apoptosis, or differentiation. Consistent with this notion, recent studies demonstrated that components of the cell cycle machinery like Cyclin D1-CDK4 could control hepatic glucose homeostasis independently of cell division (Lee et al., 2014; Bhalla et al., 2014). Indeed, emerging studies have revealed the diverse functions of the Gadd45 family in various biological systems via active DNA demethylation of specific gene loci (Schäfer, 2013). Gadd45-mediated DNA demethylation is usually initiated by cellular stimuli and is accompanied by simultaneous Gadd45 family gene induction. For instance, Gadd45α was selectively up-regulated during the osteogenic differentiation of mesenchymal stem cells to demethylate key osteogenic genes, like Runx2 and Dlx5 (Zhang et al., 2011). In addition, Gadd45β expression was rapidly induced by neuronal activity, which is essential for the DNA demethylation of specific promoters and the expression of genes for adult neurogenesis, like BDNF and FGF-1 (Ma et al., 2009). These findings indicate that Gadd45 proteins may represent an important nexus linking physiological or pathological stimuli to epigenetic DNA modification and gene expression. While Gadd45 proteins lack enzymatic activity, it is nowadays believed that Gadd45α and Gadd45β contribute to demethylation in conjunction with other modifiers, such as TET or thymine DNA glycosylase (TDG; Schüle, et al., 2019; Bayraktar and Kreutz, 2018; Li et al., 2015). This interaction enhances TET or TDG activity, possibly by tethering them to specific sites of DNA demethylation. Therefore, together with these studies, our results support the notion that Gadd45β could be an adaptor protein that enhances turnover of DNA methylation. Nevertheless, the precise mechanism(s) contributing to demethylation at specific genomic sites warrants more detailed investigation. Questions remain, such as how is the Gadd45β specifically targeted to the PGC-1α promoter? Are other DNA demethylation machineries, such as TDG protein, also involved in hepatic gluconeogenesis and hyperglycemia?

In addition to the regulation of hepatic gluconeogenesis, PGC-1α acts as a critical switch for multiple biological functions in the liver, including insulin sensitivity and fatty acid oxidation (Lin et al., 2005; Piccinin et al., 2019). PGC-1α has been shown to promote insulin resistance through PPARα-dependent induction of tribbles homolog 3 (TRB3), which inactivates AKT to disrupt insulin signaling (Koo et al., 2004). Here, we showed that Gadd45β is up-regulated in the liver of diabetic mice and patients. Besides, liver-specific Gadd45β-deficient mice are resistant to HFD-induced insulin resistance and hyperinsulinemia. Moreover, knockdown of hepatic Gadd45β increased insulin sensitivity and reduced plasma insulin concentrations in db/db mice and HFD-induced obese mice. Consistently, we found that TRB3 expression was up-regulated, while insulin-stimulated phosphorylated AKT and GSK3β were reduced, in the livers of mice with hepatic Gadd45β overexpression (Fig. S5, A and B). Therefore, we speculate that Gadd45β might impair insulin sensitivity through a PGC-1α/TRB3/AKT-dependent pathway.

In addition, PGC-1α was shown to coactivate PPARα to enhance the expression of genes involved in fatty acid oxidation and ketogenesis. As a result, mice lacking PGC-1α displayed fasting-induced liver steatosis (Leone et al., 2005). Similarly, increased triglyceride (TG) contents were observed in our Gadd45β LKO mice in the fasted state (Fig. S5 C), at least in part due to reduced expression of PPARα target genes and impaired fatty acid oxidation (Fig. S5, D and E). In agreement, expression of these genes was induced in the livers of mice overexpressing Gadd45β, resulting in a lower hepatic TG content (Fig. S5, F–H). To evaluate the possibility that TET1 might be also involved in the Gadd45β-associated hepatic TG metabolism, we introduced Gadd45β in the liver while simultaneously depleting TET1 with an Ad that expresses TET1-targeting shRNA in C57BL/6 mice (Fig. S5 I). As expected, knockdown of TET1 largely abolished the effects of Gadd45β in lowering TG contents and induction of fatty acid oxidation–related genes (Fig. S5, J–L). Besides, to confirm that the decrease in PGC-1α contributes to hepatic TG accumulation observed in Gadd45β LKO mice, we performed PGC-1α rescue in livers of Gadd45β LKO mice (Fig. S5 M). As a result, restoration of PGC-1α could alleviate the fasting-associated lipid accumulation by hepatic Gadd45β deficiency (Fig. S5 N). PGC-1α overexpression also reversed the decreased expression of fatty acid oxidation–related genes and plasma β-hydroxybutyrate concentrations (Fig. S5, O and P). On the other hand, while TRB3 is down-regulated in the livers of Gadd45β LKO mice, its restoration cannot rescue the phenotypes of lipid metabolism in Gadd45β LKO mice (Fig. S5, Q–T). Collectively, our results showed that the role of Gadd45β in the regulation of hepatic lipid metabolism could be, at least in part, through TET1 and PGC-1α, suggesting that the functions of Gadd45β and PGC-1α in hepatic glucose and lipid metabolism may partially overlap (Fig. S5 U).

In prolonged fasting, elevated GCs have been shown to inhibit fatty acid synthesis through repression of SREBP-1c (Roqueta-Rivera et al., 2016) while promoting fatty acid oxidation through up-regulation of PGC-1α, PPARα, and FGF21 (Goldstein and Hager, 2015). This might be important for fuel production during fasting. Especially, activation of fatty acid oxidation is required for the endergonic steps of gluconeogenesis for the supply of ATP and nicotinamide adenine dinucleotide plus hydrogen, as both pathways are partially localized in liver mitochondria (Goldstein and Hager, 2015). For instance, PGC-1α–, PPARα-, or FGF21-null mice subjected to prolonged fasting develop fatty liver accompanied with severely hypoglycemia, at least in part due to defects in fatty acid oxidation (Leone et al., 2005; Kersten et al., 1999; Liang et al., 2014), which is similar to our observations in Gadd45β LKO mice. On the other hand, human studies showed that there was a strong correlation between TCA cycle flux and PEPCK (Sunny et al., 2011), consistent with the induction of gluconeogenesis by fat oxidation (Chen et al., 1999; Roden et al., 2000; Staehr et al., 2003). Therefore, we speculate that up-regulation of Gadd45β by fasting GC response may mediate several metabolic adaptation responses to starvation, including hepatic gluconeogenesis and fatty acid oxidation. Moreover, our findings also support the notion that hepatic steatosis under certain conditions can be dissociated from glucose intolerance and insulin resistance, which have been reported in recent genetic mouse models and clinical studies (Sun and Lazar, 2013).

Additionally, global Gadd45β-KO mice generated by Fuhrmeister et al. (2016) showed no obvious changes in blood glucose levels under normal diet but impaired glucose homeostasis under an HFD. Although the reasons behind these conflicting data remain unclear, one possibility is that Gadd45β is also highly expressed in other metabolic organs/tissues beyond the liver, including adipose tissue, skeletal muscle, and the pancreatic islet (Solomon et al., 2010; Chadwick et al., 2015; Larsen et al., 2006). These observations suggest that the global Gadd45β deficiency might be insufficient to explore the hepatic role of Gadd45β in glucose homeostasis, as distinct phenotypes between global KOs and LKOs have been shown in many reports (Koo et al., 2004; Yu et al., 2015; Okamoto et al., 2007). Therefore, the liver-specific role of Gadd45β in the regulation of glucose metabolism is a question of interest and needs to be elucidated. To address this issue, multiple strategies were employed in our current study, including LKO mice, Ad-mediated hepatic overexpression and knockdown, and in vitro studies performed in MPHs, which, in our opinion, might be sufficient to explore the liver-specific functions of Gadd45β in glucose homeostasis. Indeed, all of our data consistently demonstrated that Gadd45β in the liver acts as an important checkpoint to respond to hormonal stimuli, which in turn activate a gluconeogenic program to induce hyperglycemia and glucose intolerance.

In conclusion, we propose that activation of the Gadd45β-mediated epigenetic effects represent a novel mechanism underlying hyperglycemia and diabetes. Inhibition of gluconeogenesis and HGP has been used as an effective strategy to screen novel antidiabetic drugs (Rines et al., 2016). Especially, compounds that selectively inhibit the hepatic gluconeogenic role of PGC-1α have been shown to possess a glucose-lowering effect in a preclinical diabetic mouse model (Sharabi et al., 2017). In addition, circulating GC concentrations are elevated in genetically diabetic mice (e.g., db/db, ob/ob mice), and elevated plasma cortisol levels have also been observed in diabetic patients or subjects with glucose intolerance by some clinical reports (Lee et al., 1999; Reynolds et al., 2001). Therefore, our findings suggest that inactivation of the Gadd45β DNA methylation machinery may serve as a potential avenue for treating T2DM, including long-term GC exposure–associated diabetes.

8-wk-old male db/db and ob/ob mice were purchased from the Nanjing Biomedical Research Institute of Nanjing University. 8–10-wk-old male C57BL/6J mice were purchased from the Shanghai Laboratory Animal Company. HFD-induced obese mice were generated by feeding C57BL/6J mice an HFD (D12492; Research Diets) for 12 wk. The metabolic features of HFD, ob/ob, and db/db mice have been described in our previous study (Lu et al., 2014). Gadd45β F/F mice obtained from GemPharmatech Co., Ltd. were generated by CRISPR/Cas9 technology (Fig. S4 A). Exons 1–4 of the Gadd45β transcript are recommended as the KO region. Briefly, small guide RNA was transcribed in vitro, and a donor vector was constructed. Cas9, small guide RNA, and donor were microinjected into the fertilized eggs of C57BL/6J mice. Fertilized eggs were transplanted to obtain positive F0 mice, which were confirmed by PCR and sequencing. A stable F1 generation mouse model was obtained by mating positive F0 generation mice with C57BL/6J mice. To generate Gadd45β LKO mice, 8-wk-old male Gadd45β F/F mice were injected with AAV9-expressing Cre recombinase through the tail vein. The Gadd45β F/F littermates injected with AAV9-expressing GFP recombinase were used as controls. All mice were housed at 21°C ± 1°C with 55% ± 10% humidity and a 12-h/12-h light/dark cycle. Dex (Sigma-Aldrich) or vehicle control was administered daily for 14 d at a dose of 1.0 mg/kg body weight by i.p. injection. Plasma insulin, glucagon, corticosterone, and β-hydroxybutyrate levels were measured using kits from Crystal Chem (90080), Crystal Chem (81518), Labor Diagnostika Nord (ARE-8100), and BioVision (K632), respectively. For analyzing hepatic TG content, liver tissues (∼100 mg) were homogenized in 1 ml of 5% NP-40 solution and heated to 100°C and then cooled to room temperature. The tissue homogenates were centrifuged for 2 min, and the supernatants were processed to measure TG content using the Triglyceride Quantification Kit (K622; BioVision). The animal protocol was reviewed and approved by the animal care committee of Shanghai Jiao Tong University School of Medicine.

The human subjects were recruited from the fatty liver clinic of Zhongshan Hospital, Fudan University. Liver samples were collected from normal participants (n = 13, male) and patients with type 2 diabetes (n = 16, male) through ultrasonography-guided percutaneous needle liver biopsies. Subjects who had type 1 diabetes, chronic viral hepatitis, hypothyroidism, excessive use of alcohol, use of drugs that may affect glucose metabolism, abnormal renal function, or infectious diseases were excluded. Blood samples were collected after a 12-h overnight fast, and plasma was stored at −80°C for biochemical assays. The clinical characteristics of human subjects are shown in Table S4. The human study was approved by the human research ethics committee of Zhongshan Hospital, Fudan University. Written informed consent was obtained from each subject.

Briefly, full-length mouse Gadd45β cDNA or the GFP gene was cloned into GV314 adenoviral vector (CMV-MCS-3FLAG-SV40-EGFP) from GeneChem. 2 × 10⁹ PFU per mouse were delivered into C57BL/6 mice for 15 d. Ad-shRNA particles targeting GR, Gadd45β, and PGC-1α were generated using pAD_BLOCK_IT_DEST vectors (Invitrogen). The shRNAs had the following sequences: GR (5′-AGA​AAT​GAC​TGC​CTT​ACT​A-3′), Gadd45β (5′-GCG​ACA​ATG​ACA​TTG​ACA​TCG-3′), PGC-1α (5′-GGT​GGA​TTG​AAG​TGG​TGT​AGA-3′), or TET1 (5′-GCC​TCT​TCT​ACG​GGA​ACA​TTC-3′). To knock down hepatic GR expression, 2 × 10⁹ PFU per mouse were delivered into C57BL/6 or HFD mice for 7 d. To knock down hepatic Gadd45β expression, 4 × 10⁹ PFU per mouse were delivered into db/db mice for 14 d. All viruses were purified by the cesium chloride method, dialyzed in PBS containing 10% glycerol, and administered to mice through tail vein injection.

For the PTTs and GTTs, mice were fasted for 16 h and injected i.p. with pyruvate (1.5 g/kg body weight; Sigma-Aldrich) or D-glucose (1.5 g/kg body weight; Sigma-Aldrich). For the ITTs, mice were injected with regular human insulin (0.75 U/kg body weight; Eli Lilly) after a 6-h fast. Blood glucose was collected from tail veins after injection and measured by a portable blood glucose meter (Johnson & Johnson).

High-throughput mRNA sequencing service was provided by CloudSeq Biotech. Total RNA (1 µg) was used for removing the ribosomal RNAs (rRNAs) using Ribo-Zero rRNA removal kits (Illumina) following the manufacturer’s instructions. RNA libraries were constructed using rRNA-depleted RNAs with the TruSeq Stranded Total RNA Library Prep Kit (Illumina) according to the manufacturer’s instructions. Libraries were controlled for quality and quantified using the BioAnalyzer 2100 system (Agilent Technologies, Inc.). Libraries were denatured as single-stranded DNA molecules, captured on Illumina flow cells, amplified in situ as clusters, and finally sequenced for 150 cycles on an Illumina HiSeq Sequencer according to the manufacturer’s instructions. Paired-end reads were harvested from the Illumina HiSeq 4000 sequencer and were quality controlled by Q30. After 3′ adaptor trimming and low-quality read removal by Cutadapt software (version 1.9.3), the high-quality clean reads were aligned to the reference genome (University of California, Santa Cruz [UCSC] Genome Browser hg19) with HISAT2 software (version 2.0.4). Then, guided by the Ensembl GTF gene annotation file, Cuffdiff software (part of cufflinks) was used to get the gene-level fragments per kilobase million as the expression profiles of mRNA. Fold change and P value were calculated based on fragments per kilobase million, and differentially expressed mRNAs were identified. The RNA-seq data have been deposited in the Gene Expression Omnibus database (GSE159038 and GSE159084).

HepG2 cells were obtained from the Cell Bank of Shanghai Institute of Biochemistry and Cell Biology (Chinese Academy of Sciences). MPHs were isolated from the livers of 10-wk-old male C57BL/6 mice by collagenase perfusion through the portal vein and purified by centrifugation. Freshly prepared MPHs were resuspended in attachment media (Science Cell) and seeded in 6-well plates at a final density of 5 × 10⁵ cells per well. 12 h later, unattached cells were washed away, and DMEM (Gibco-BRL) was added for an overnight incubation.

MPHs were infected with Ads as indicated for 36 h. Then, the medium was replaced with glucose-free DMEM supplemented with 2 mM sodium pyruvate and 20 mM sodium lactate. 4 h later, media were collected, and the glucose concentrations were measured with a colorimetric glucose assay kit (MAK263; Sigma-Aldrich). The readings were normalized to total protein contents determined from the whole-cell extracts.

HA-tagged full-length mouse GR and Gadd45β expression plasmids were cloned by PCR amplification and inserted into porcine cytomegalovirus vector with HindIII/BamHI restriction sites. The promoter region of the mouse Gadd45β gene extending from position −2,000 bp (relative to the mRNA transcription start site) to +1 was cloned into the PGL4.12 luciferase reporter plasmid with the XhoI/HindIII restriction sites. Mutant GRE motif was generated using the QuikChange II XL Site-Directed Mutagenesis Kit (Agilent Technologies Inc.) with the following primers (mutation sites underlined): mutant 1: 5′-CAT​TCC​AAG​TGGGCTC​TCC​CCC​TCC​CTT​TCC​GGA​AG-3′; mutant 2: 5′-AAG​GCC​CGG​CAGGGCAC​TTA​GTA​AGC​CTC​CTG​GCG​CAT-3′. For luciferase assays, HepG2 cells were placed in 24-well plates and cotransfected with GR expression plasmids and luciferase reporter plasmids using Lipofectamine 3000 (Thermo Fisher Scientific) according to the manufacturer’s instructions. The plasmid expressing Renilla luciferase with thymidine kinase promoter (Promega) was used to normalize the luciferase activity. 24 h after transfection, cells were treated with Dex (10 nM) or vehicle for 12 h before harvest. Luciferase activities were measured using the Dual-Luciferase Reporter Assay System (Promega).

ChIP assays were performed using the EZ ChIP kit (Merck Millipore). MPHs or minced liver tissues were fixed in 1% formaldehyde (Sigma-Aldrich) for 10 min. Cross-linking was stopped by addition of glycine at a concentration of 125 mM for 5 min, followed by centrifugation and washing in SDS lysis buffer at room temperature. DNA was sheared to fragments of 200–1,000 bp by sonication. Lysates containing soluble chromatin were incubated and precipitated overnight with 5 µg anti-GR antibody (ab3579; Abcam), anti-TET1 (ab191698; Abcam), anti-TET2 (ab94580; Cell Signaling Technology), anti-TET3 (ab139311; Abcam), or IgG (ab172730; Abcam). DNA–protein immune complexes were removed with protein G agarose and then washed and eluted. Protein–DNA cross-links were reversed by treatment with proteinase K for 2 h at 45°C. The DNA was subsequently purified, diluted, and subjected to qRT-PCR. The mouse Gadd45β gene promoter fragments containing the GRE motif were amplified using the following primers: site 1: 5′-TCT​CAC​TTT​GGG​GAG​AGG​ATC-3′ (forward), 5′-GAG​TCC​CTA​GCC​TTT​CCC​AA-3′ (reverse); site 2: 5′-TTC​GAG​GCC​AGC​CTG​GTC​TA-3′ (forward), 5′-CCT​CAT​GAG​GTC​TCT​GCT​CA-3′ (reverse). The promoter region of the GAPDH proximal promoter was set as the negative control. Primers for the GAPDH promoter were: 5′-CTA​TCC​TGG​GAA​CCA​TCA-3′ (forward) and 5′-AAG​CGT​GTG​GGC​TCC​GAA-3′ (reverse).

Genomic DNA from mouse livers was isolated using phenol–chloroform, precipitated with ethanol, and sonicated to 100–500 bp using Bioruptor (Diagenode). Sonicated DNA was end repaired, A tailed, and ligated to adaptors by using the NEBNext Ultra DNA Library Prep Kit (New England Biolabs). Then, MeDIP was performed with an mAb against 5mC by following the standard manufacturer’s protocol (MilliporeSigma). MeDIP DNA libraries were quantified using Quant-iT PicoGreen double-strand DNA kits (Life Technologies) and subjected to high-throughput 150–base paired end sequencing on an Illumina HiSeq sequencer according to the manufacturer’s recommended protocol. High-quality reads (raw data) were generated after sequencing, image analysis, base calling, and quality filtering on an Illumina sequencer. Cutadapt software was used to for adaptor trimming. After adaptor trimming, the trimmed reads were aligned to the mouse genome (UCSC Genome Browser, mm10) using BOWTIE software (version 2.1.0) with default parameters. Peak calling was performed by MACS1.4 software. DMRs were identified by DiffReps software. The enriched peaks and DMRs were then annotated with the UCSC RefSeq database to connect the peak information with the gene annotation. The enriched peaks were visualized on the UCSC Genome Browser.

Purified genomic DNA (4 µg) was fragmented to a mean size of 300 bp using a Covaris machine, denatured, and immunoprecipitated with antibodies against 5mC (MilliporeSigma) or 5hmC (MilliporeSigma). Immunoprecipitated DNA was recovered with proteinase K digestion followed by column-based purification (DNA Wizard; Promega). Recovered DNA fractions were diluted 1:50 and measured using RT-PCR with an ABI PRISM 7000 sequence detector system and fluorescence-based SYBR Green technology (Applied Biosystems).

DNA methylation levels were determined by pyrosequencing after bisulfite conversion using the EpiTect bisulfite kit (Qiagen; Wu et al., 2016). Quantitative PCR was performed on a PSQ 96MD system with the PyroGold sequence analysis reagent kit (Qiagen), and results were analyzed using the PyroMark CpG software (version 1.0.11.14; Qiagen). The quantitative accuracy and sensitivity of our pyrosequencing assays have been validated using mixtures of fully methylated and unmethylated mouse genomic DNA at different ratios (A&D Technology Corporation).

Total RNAs from mouse livers or MPHs were extracted using TRIzol reagent (Invitrogen). RNA purity and concentrations were measured using the NanoDrop ONE spectrophotometer (Thermo Fisher Scientific). The mRNAs were reverse transcribed into cDNA using the Promega Reverse Transcription System. Oligo(dT) was used to prime cDNA synthesis. cDNA was stored at −20°C before use. qRT-PCR was performed using SYBR Green Premix Ex Taq (Takara Bio) on a Light Cycler 480 (Roche), and data were analyzed by the comparative cycle threshold method. PCR conditions included an initial holding period at 95°C for 5 min, followed by a two-step PCR program consisting of 95°C for 5 s and 60°C for 30 s for 50 cycles. Relative abundance of mRNA was normalized to ribosomal protein 36B4. The primer sequences are available upon request.

To prepare total protein extracts, liver tissues and primary hepatocytes were lysed in RIPA buffer containing 50 mM Tris-HCl, 150 mM NaCl, 5 mM MgCl₂, 2 mM EDTA, 1 mM NaF, 1% NP-40, and 0.1% SDS. Equivalent amounts of protein samples were denatured in loading buffer and resolved by 10–12% SDS-PAGE and transferred onto polyvinylidene fluoride membranes. Membranes were blocked in 5% nonfat milk for 1 h before incubation with primary antibody overnight at 4°C. Membranes were washed with PBS with Tween-20 five times and incubated with secondary antibody for 1 h. The signals of the proteins were then visualized by an electrochemiluminescence system. The following primary antibodies were used: anti-Gadd45β at 1:2,000 (sc-377311; Santa Cruz Biotechnology), anti–PGC-1α at 1:2,000 (AB3242; EMD Millipore), anti-TET1 at 1:1,000 (61741; Active Motif), anti-TRB3 at 1:1,000 (sc-365842; Santa Cruz Biotechnology), anti–p-AKT at 1:1,000 (4060; Cell Signaling Technology), anti–t-AKT at 1:2,000 (9272; Cell Signaling Technology), and anti-GAPDH antibody at 1:6,000 (KC-5G5; Kangchen Technology).

Data are presented as mean ± SEM. For animal and cellular experiments, a two-tailed unpaired Student’s t test was performed to compare two groups. One-way ANOVA followed by the Student-Newman-Keuls test was used to compare more than two groups. The correlation between mRNA expression levels of target genes and plasma glucose levels was analyzed by Pearson correlation. P values <0.05 were considered statistically significant. *, **, and *** represent P < 0.05, P < 0.01, and P < 0.001, respectively.

Fig. S1 shows the up-regulation of hepatic Gadd45β by GCs. Fig. S2 shows the identification of PGC-1α as a target of Gadd45β. Fig. S3 shows the methylation status of gluconeogenic genes. Fig. S4 shows the generation of Gadd45β LKO mice and the methylation status and expression of hepatic gluconeogenic genes in diabetes. Fig. S5 shows the hepatic insulin signaling and lipid metabolism in mice. Table S1 shows the up-regulated genes by RNA-seq analysis. Table S2 shows the overlap of the up-regulated expressed genes. Table S3 shows the overlap of the down-regulated expressed genes. Table S4 shows the clinical characteristics of human subjects.

We thank Professor Peng Li (Institute of Metabolism and Integrative Biology, Fudan University, Shanghai, China) and Professor Guoliang Xu (Shanghai Institute of Biochemistry and Cell Biology, Shanghai, China) for critical discussion of the data.

This work is supported by the National Natural Science Foundation of China (81722013 to Y. Lu, 81771574 to L. Wu, 81900727 to Y. Jiao, 82070887 to Y. Li), the National Key Research and Development Program of China (2018YFA0800402 to Y. Lu), the Shanghai Municipal Commission of Health and Family Planning Foundation (20174Y0036 to L. Wu), the Shanghai Sailing Program (19YF1406900 to Y. Jiao), and the Shanghai Science Foundation (18ZR1437800 to Y. Li).

Author contributions: Y. Lu designed and directed experiments; L. Wu, Y. Jiao, and Y. Li performed animal and cellular experiments; M. Li, B. Li, and Z. Yan contributed to animal and cellular experiments; J. Jiang and L. Zhao performed human experiments; X. Chen and X. Li contributed to overall experimental analysis and discussion. L. Wu, Y. Jiao, and Y. Lu wrote the manuscript. Y. Lu supervised the entire project and edited the manuscript. All authors reviewed the manuscript and provided final approval for submission.