O-GlcNAcylation of Orphan Nuclear Receptor Estrogen-Related Receptor γ Promotes Hepatic Gluconeogenesis

O-GlcNAcylation works as a nutrient sensor in the liver to maintain energy homeostasis in response to varying nutrient flux (1). O-GlcNAcylation is extensively linked with glucose metabolism in liver. L-Glutamine fructose-6-phosphate amidotransferase (GFAT) overexpression leads to peripheral insulin resistance (2,3). Transgenic mice overexpressing O-GlcNAc transferase (OGT) in skeletal muscle and fat exhibit elevated circulating insulin levels and insulin resistance (4). Insulin receptor substrate 1/2 of insulin signaling is O-GlcNAcylated (5), and O-GlcNAcylation has been shown to be a negative regulator of insulin signaling (6). O-GlcNAcylation of forkhead box class O1 (FOXO1), CRTC2, and peroxisome proliferator–activated receptor γ coactivator 1α (PGC-1α) modulates expression of gluconeogenic genes (7–10). Chronic increase in O-GlcNAcylation levels of PDX1 and NeuroD1 may contribute to hyperinsulinemia in type 2 diabetes (11,12). Thus, by being intimately intertwined with metabolism, the hexosamine biosynthetic pathway (HBP) and its end product O-GlcNAc link transcriptional processes to cellular glucose metabolism and insulin resistance.

Estrogen-related receptors (ERRs) are members of the NR3B subfamily of nuclear receptors, which include ERRα, ERRβ, and ERRγ. ERRγ is primarily expressed in heart, brain, kidney, pancreas, and liver tissues and is induced during fasting in murine liver (13–15). ERRγ plays an important role in the regulation of glucose, lipid, alcohol and iron metabolism in mouse liver (16,17). Hepatic ERRγ expression is induced in fasting and the diabetic state and causes insulin resistance and glucose intolerance (18). Induction of hepatic ERRγ impairs insulin signaling through diacylglycerol-mediated protein kinase ε activation (19), suggesting that ERRγ transcriptional activity could be involved in insulin action to maintain glucose homeostasis. Recently, our laboratory reported that insulin-dependent phosphorylation of ERRγ alters its transcriptional activity to suppress hepatic gluconeogenesis in the fed state (20).

PGC-1α is a transcriptional coactivator involved in hepatic glucose metabolism. Fasting induces hepatic PGC-1α expression that directly interacts with transcription factors, including hepatocyte nuclear factor 4α, FOXO1, and glucocorticoid receptor, to increase the expression of gluconeogenic genes (21,22). PGC-1α overexpression leads to increased expression of G6Pase and PEPCK, key enzymes in the hepatic gluconeogenesis. Conversely, knockdown or knockout of PGC-1α results in lower blood glucose levels as a result of reduced gluconeogenesis.

As insulin-dependent posttranslational modification regulates the transcriptional activity of ERRγ in the fed state, in this study, we investigated whether fasting-dependent activation of the transcriptional activity of ERRγ involves O-GlcNAcylation. We demonstrate that the fasting condition triggers O-GlcNAcylation of ERRγ that results in protein stabilization. O-GlcNAcylation of ERRγ by OGT decreases its ubiquitination and cooperatively upregulates gluconeogenesis. In contrast, the fed condition decreases O-GlcNAcylation of ERRγ, resulting in ubiquitin-mediated protein degradation. Overall, our study describes how O-GlcNAcylation modulates gluconeogenesis via ERRγ.

Male 8-week-old C57BL/6J mice, maintained at the Korea Research Institute of Bioscience and Biotechnology (KRIBB) were fed either a high-fat diet (HFD) (D12492; Research Diets, New Brunswick, NJ) or a normal chow diet for 12 weeks. At the end of 12 weeks, mice were sacrificed, and liver tissue was used for identification of O-GlcNAcylation of ERRγ. ob/ob and db/db mice (7–12 weeks old; Charles River Laboratories) were maintained at KRIBB in an animal facility with ad libitum access to water and a standard laboratory diet. Liver tissue was used for identification of O-GlcNAcylation of ERRγ. Male 7-week-old C57BL/6J mice (The Jackson Laboratory, Bar Harbor, ME) were obtained from Ochang Branch Institute, KRIBB. After 2 weeks, adenoviruses (Ad-green fluorescent protein [GFP], Ad–wild-type [wt] ERRγ, and Ad-S317A+S319A ERRγ; 5.9 × 10⁹ plaque-forming units/mouse) were delivered by tail-vein injection into mice. Glucose tolerance test was performed at day 5 after a tail-vein injection of adenoviruses. Briefly, mice fasted 16 h were injected intraperitoneally with 1 g/kg glucose, and blood glucose was measured in tail-vein blood using a blood glucose meter and test trips (Accu-Chek Aviva meter system; Roche Diagnostics, Indianapolis, IN). All mice were acclimatized to a 12-h light/dark cycle at 22 ± 2°C with free access to food and water in a specific pathogen-free facility. All animal experiments were approved and performed by the Institutional Animal Care and Use Committee of KRIBB.

Glucose production from primary mouse hepatocytes was measured using a colorimetric glucose oxidase assay kit according to the manufacturer’s protocol. Briefly, after the experimental time period as indicated, the cells were washed three times with PBS. Then the cells were incubated for 3 h at 37°C, 5% CO₂, in glucose production buffer (glucose-free DMEM [pH 7.4], containing 20 mmol/L sodium lactate, 1 mmol/L sodium pyruvate, and 15 mmol/L HEPES, without phenol red), and the glucose assays were performed.

Blood glucagon levels were measured using the mouse glucagon EIA kit (Ray Biotech) following the manufacturer’s protocol.

Primary hepatocytes were isolated from C57BL/6J mice (male, 20–30 g) by collagenase perfusion (23) and seeded with Medium 199 (Cellgro). After 3–6 h of attachment, cells were infected with the indicated adenoviruses for overexpression or treated with various chemicals as indicated. HEK 293T and AML12 cells were maintained as described previously (24). Transient transfection was performed using Lipofectamine 2000 (Invitrogen) or SuperFect (Qiagen, Hilden, Germany) according to the manufacturers’ instructions. β-N-Acetyl-d-glucosamine (GlcN), 6-diazo-5-oxo-L-norleucine (DON), glucagon, insulin, cycloheximide, ANTI-FLAG M2 affinity gel, glucose oxidase assay kit, and streptozotocin (STZ) were purchased from Sigma-Aldrich. MG-132 protease inhibitors were purchased from Calbiochem. Express Protein Labeling Mix [³⁵S] (NEG072002MC) was purchased from PerkinElmer. The Mouse Glucagon EIA kit (EIAM-GLU) was purchased from Ray Biotech. Antibodies were purchased as follows: O-GlcNAc from Covance, α-tubulin from Abfrontier, ERRγ from Perseus Proteomix, OGT from Abcam, FLAG, Anti-FLAG M2, and hemagglutinin (HA) from Cell Signaling Technology, and G6Pase, Mdm2, ubiquitin, Gal4, PGC-1α, and PEPCK from Santa Cruz Biotechnology.

Expression vectors for HA-ERRγ, FLAG-ERRγ, HA-PGC-1α, and Sft4-luc containing three copies of the ERRγ binding site were described previously (25). HA-ERRα was described previously (26). FLAG-human OGT was constructed by inserting the full PCR fragment of the open reading frame into the Not1/Sal1 sites of the p3XFLAG-CMV-7.1 vector. FLAG-human O-GlcNAcase (OGA) was constructed by inserting the full PCR fragment of the open reading frame into the Bgl2/Sal1 sites of the pFLAG-CMV-7.1 vector. FLAG-mutant ERRγ’s (S317A, S319A, and S317A+S319A) were constructed using wild-type ERRγ as a template by the Quick Change Lightning Site-Directed Mutagenesis kit from Agilent Technologies. Gal4–DNA binding domain (DBD) and Gal4-tk-Luc were described previously (27). Briefly, Gal4–ERRγ–ligand-binding domain (LBD) is a fusion protein consisting of the Gal4–DBD (amino acids 1–147) and ERRγ–LBD (amino acids 189–458). This fusion protein activates transcription of a reporter construct (Gal4–tk-luc) containing five GAL4 binding sites (upstream activator sequence) upstream of the firefly luciferase gene in the pGL2-promoter. Gal4-DBD-S317A ERRγ, Gal4-DBD-S319A ERRγ, and Gal4-DBD-S317A+S319A ERRγ were constructed using Gal4-DBD-wild-type ERRγ as a template by the Quick Change Lightning Site-Directed Mutagenesis kit from Agilent Technologies. wtPEPCK-luc, and ERRγ response element (ERRE) mutant PEPCK-luc were described previously (28). Adenoviruses expressing unspecific short hairpin (sh)RNA, shERRγ, control GFP, and ERRγ were described previously (28). Ad-OGT encoding the human OGT gene, adenovirus OGA (Ad-OGA) encoding the human OGA gene, and adenovirus S317A+S319A ERRγ (Ad-S317A+S319A ERRγ) were generated with the pAd-easy system as described previously (29). All viruses were purified by using CsCl gradient protocol.

Overexpressed wild‐type FLAG-ERRγ proteins from mouse liver were purified using FLAG‐M2 agarose and subjected to SDS-PAGE. Purified protein was digested with trypsin (Promega, Madison, WI) (25 ng/μL) for 16 h at 37°C. After in-gel digestion, tryptic peptides were separated by online reversed-phase chromatography using a Thermo Scientific Eazy nano LC II autosampler (Thermo Scientific) with a reversed-phase peptide trap EASY-Column (100-μm inner diameter, 2-cm length) and a reversed-phase analytical EASY-Column (75-μm inner diameter, 10-cm length, 3-μm particle size; both from Thermo Scientific). Electrospray ionization was performed using a 30-μm (inside diameter) nano-bore stainless steel online emitter (Thermo Scientific) and a voltage set at 2.6 V, at a flow rate of 300 nL/min. The chromatography system was coupled on-line with an LTQ Velos Orbitrap mass spectrometer (Thermo Scientific). Protein identification was accomplished using the Proteome Discoverer v1.3 database search engine (Thermo Scientific), and searches were performed against IPI.Humanv3.87 FASTA database or ERRγ FASTA database. A fragment mass tolerance of 1.2 Da, peptide mass tolerance of 25 ppm, and maximum missed cleavage of 2 were set.

Total RNA was isolated using TRIzol reagent (Invitrogen), cDNA was synthesized using a reverse transcriptase kit (Intron Biotechnology), and real-time quantitative RT-PCR (qPCR) was performed with the SYBR green PCR kit (Enzynomics). The amount of mRNA for each gene was normalized to that of actin mRNA.

Nuclear isolation of primary hepatocytes and cross-linking of protein to DNA were performed as described previously (28). After sonication, soluble chromatin was subjected to immunoprecipitation using anti-ERRγ antibody. DNA was recovered by phenol/chloroform extraction and analyzed by PCR using primers against relevant promoters. Primer sequences were mouse PEPCK1 promoter forward, 5′-CTAGCCAGCTTTGCCTGACT-3′ and reverse, 5′-GGGTCCCCACGACCTTCCAA-3′.

Whole-cell extracts were prepared using radioimmunoprecipitation assay (RIPA) buffer (50 mmol/L Tris [pH 7.5], 150 mmol/L NaCl, 1% Nonidet P-40, and 5 mmol EDTA). Proteins from whole-cell lysates were separated by 10% SDS-PAGE and then transferred to nitrocellulose membranes. The membranes were probed with different antibodies. Immunoreactive proteins were visualized using an Amersham Biosciences ECL kit (GE Healthcare) according to the manufacturer's instructions.

C57BL/6J mice were infected with respective viruses via tail-vein injections. Four days postinjection, mice were fasted for 16 h and imaged using an IVIS Lumina II imaging system (Caliper Life Sciences, Hopkinton, MA) as described previously (28).

At 24 h after transfection, the cells were fixed with 2% formaldehyde, immunostained, and subjected to observation by confocal microscopy using a laser-scanning confocal microscope (Olympus, Lake Success, NY).

AML12 cells were transfected with FLAG-wt ERRγ, FLAG-S317A ERRγ, FLAG-S319A ERRγ, and FLAG-S317A+S319A ERRγ and incubated in methionine and cystine-free medium for 2 h. Translabel mixture (PerkinElmer) containing ³⁵S-methionine was added for 30 min, and then cells were cultured in normal medium up to 3 h. FLAG-ERRγ was immunoprecipitated with M2 antibody in RIPA buffer, and radioactive FLAG-ERRγ was detected by autoradiography.

All values are expressed as means ± SEM. The significance between mean values was evaluated by two-tailed Student t test.

ERRγ is a key positive regulator of hepatic gluconeogenesis (28). ERRγ phosphorylation by protein kinase B/Akt also contributes to insulin-mediated inhibition of hepatic gluconeogenesis (20). Many factors that regulate hepatic gluconeogenesis are O-GlcNAcylated in various conditions (7,30,31). Hence, we sought to determine whether ERRγ is modified by O-GlcNAc. HBP intermediate GlcN treatment significantly increased ERRγ O-GlcNAcylation levels as well as ERRγ protein content in a dose-dependent manner (Fig. 1A). Consistent with a rise in ERRγ protein levels, GlcN treatment significantly reduced ERRγ ubiquitination levels and increased ERRγ transcriptional activity (Supplementary Fig. 1A and B). Glucose (Glc) flux through HBP regulates O-GlcNAcylation (1). Therefore, to determine whether glucose could directly affect O-GlcNAcylation of ERRγ, mouse primary hepatocytes (MPH) were treated with 5 and 25 mmol/L glucose. High glucose significantly increased O-GlcNAcylation of ERRγ along with ERRγ protein levels (Fig. 1B). This was further confirmed when OGT cotransfection markedly enhanced O-GlcNAcylation of ERRγ (Fig. 1C). Because ERRα and ERRγ are both associated with hepatic glucose metabolism, we also examined whether OGT overexpression could lead to ERRα O-GlcNAcylation. Unlike ERRγ, O-GlcNAcylation was not detected for ERRα (Fig. 1D).

O-GlcNAcylation is linked with protein stability (30–33). From our results (Fig. 1A and B), we speculated that O-GlcNAcylation could stabilize ERRγ by decreasing protein degradation. To test this, HEK 293T cells were cotransfected with ERRγ and OGA or OGT expression vectors followed by MG-132 treatment to inhibit proteasome-mediated protein degradation. ERRγ ubiquitination levels were significantly raised and O-GlcNAcylation levels were decreased in presence of OGA, whereas OGT had an entirely opposite effect, suggesting that OGT triggers ERRγ O-GlcNAcylation that inhibits ERRγ ubiquitination and stabilizes it (Fig. 1E). Next, to examine the functional implications of O-GlcNAcylation, reporter gene assay with transient transfection was carried out in the 293T cell line. ERRγ significantly enhanced the Sft4-luc reporter activity, which was further augmented in presence of OGT. OGA had an inverse effect to that of OGT and significantly reduced ERRγ transcriptional activity. However, ERRα could not markedly activate the reporter gene even in presence of OGT (Fig. 1F). Activation of the Gal4-tk-luc reporter gene by Gal4–ERRγ–LBD was significantly augmented by OGT, whereas OGA significantly repressed it, indicating that O-GlcNAcylation of ERRγ might occur in its LBD (Fig. 1G). Overall, these results suggest that ERRγ is subject to O-GlcNAcylation, which in turn increases protein stability by decreasing ubiquitin mediated protein degradation and also enhances ERRγ transcriptional activity.

Hepatic ERRγ expression is increased by fasting-dependent activation of the CREB-CRTC2 pathway (18,28). Hence, we sought to determine whether fasting-dependent increase in ERRγ expression is associated with O-GlcNAcylation. Fasting significantly enhanced O-GlcNAcylation of ERRγ compared with the fed state. This stands in contrast to ERRγ ubiquitination levels that were higher in the fed condition than the fasting condition, suggesting that O-GlcNAcylation mediates ERRγ stability in the fasting condition (Fig. 2A). To confirm that O-GlcNAcylation mediates ERRγ stability in the fasting condition, we overexpressed OGA in this condition. Overexpression of OGA significantly reduced O-GlcNAcylation of ERRγ and simultaneously increased ubiquitination of ERRγ, resulting in lowering of ERRγ abundance, unambiguously establishing that ERRγ stability is governed by O-GlcNAcylation in the fasting condition (Fig. 2B). Increased protein stability is associated with decreased interaction between protein and E3 ubiquitin-ligases such as Mdm2 (32). We observed that fasting increased OGT–ERRγ interactions in a time-dependent manner, resulting in enhanced O-GlcNAcylation and reduced Mdm2–ERRγ interactions in mice (Fig. 2C). The binding of OGT and ERRγ was reduced, but the levels of ERRγ O-GlcNAcylation were still intensified after 3 h of fasting (Fig. 2C). To clarify this, we tested whether the interactions between ERRγ and OGT or OGA might be dynamically altered during the fasting period. The interaction between ERRγ and OGT increased in a fasting time manner, reaching a peak between 3 and 6 h and then declining, whereas the interaction between ERRγ and OGA decreased in a fasting time manner (Supplementary Fig. 1C), suggesting that in the fed state, OGA interacts with ERRγ and destabilizes it, whereas in the fasting condition OGT interacts with ERRγ and stabilizes it. Interestingly, protein levels of both OGT and OGA were increased in the fasting (Supplementary Fig. 1C), which was supported by elevated OGT and OGA mRNA levels in fasting (Supplementary Fig. 1D). In spite of low circulating glucose, fasting promoted O-GlcNAcylation of ERRγ. Hence, we speculate that glucagon could be more important than blood glucose for O-GlcNAcylation of ERRγ during fasting. To test this, we measured blood glucose and glucagon levels in a fasting time-course experiment (Supplementary Fig. 1E and F). Blood glucose levels in fasting mice decreased from 0 to 12 h, whereas serum glucagon levels steadily increased from 0 h, reaching a peak at 6 h, followed by a decrease at 12 h, though it remained significantly higher at 12 h than at 0 h. ERRγ O-GlcNAcylation levels gradually increased from 0 h, reaching a peak at 6 h and then remaining almost the same until 12 h, suggesting that circulating glucagon levels are more important than circulating glucose levels for O-GlcNAcylation of ERRγ during fasting.

Glucagon increases hepatic ERRγ transcriptional activity during fasting (28). Thus, we sought to ascertain whether glucagon increases ERRγ transcriptional activity by affecting O-GlcNAcylation. Glucagon treatment significantly increased ERRγ O-GlcNAcylation levels, resulting in higher protein stability through reduced ERRγ ubiquitination (Fig. 2D). As both high glucose (Fig. 1B) and glucagon (Fig. 2D) induced O-GlcNAcylation, we tested which is more important for O-GlcNAcylation of ERRγ. Both high levels of glucose and glucagon increased ERRγ O-GlcNAcylation levels and total protein levels. They have a cumulative effect when applied together (Supplementary Fig. 1G), but unlike glucagon, glucose did not increase ERRγ mRNA levels (Supplementary Fig. 1H). The notion that O-GlcNAcylation mediates the effect of glucagon on ERRγ protein stability was further corroborated when overexpression of OGA significantly reduced both O-GlcNAcylation and protein levels of ERRγ (Fig. 2E). Insulin induced by feeding suppresses ERRγ transcriptional activity as well as gene expression (20,28). Thus, we speculate that insulin may inhibit O-GlcNAcylation of ERRγ, leading to decreased protein stability. As we expected, glucagon-induced ERRγ O-GlcNAcylation was significantly suppressed by insulin treatment in AML12 cells. ERRγ ubiquitination was markedly enhanced in presence of insulin compared with glucagon treatment, resulting in lower protein stability (Fig. 2F). Taken together, these results indicate that fasting increases O-GlcNAcylation of ERRγ, leading to protein stability, whereas feeding acts reciprocally to degrade ERRγ by suppressing its O-GlcNAcylation.

In order to identify the O-GlcNAc site of ERRγ, we infected mice with Ad-FLAG-ERRγ and extracted the ERRγ protein from liver tissue through immunoprecipitation (Fig. 3A and B). The mass spectrometry results show that S317 and S319 of ERRγ are O-GlcNAcylated (Supplementary Figs. 2 and 3). The primary structure of ERRγ revealed that S317 and S319 were present in the LBD of ERRγ (Fig. 3C), which was consistent with the previous result that suggested O-GlcNAcylation site is in the LBD (Fig. 1G). However, only S317 was conserved in ERRα, ERRβ, and ERRγ (Fig. 3C). The three-dimensional structure of ERRγ–LBD provides a comprehensive view of the O-GlcNAcylation site. The two O-GlcNAcylated serine residues are located at the end of helix5 of the reported ERRγ–LBD structures, which is locally stabilized with helix 6, helix 7, and a couple of strands. Although S319 is completely exposed to the solvent-accessible surface, S317 is partially hidden (Fig. 3D). To verify the specific O‐GlcNAcylation sites indicated by mass analysis, we constructed three site‐specific point mutant cDNAs of ERRγ, S317A ERRγ, S319A ERRγ, and S317A+S319A ERRγ. The single mutants (S317A ERRγ and S319A ERRγ) showed less O-GlcNAcylation compared with wild-type ERRγ, whereas the double mutant (S317A+S319A ERRγ) showed complete absence of an O-GlcNAc signal in the presence of GlcN (Fig. 3E). Interestingly, even though the single mutants showed higher O-GlcNAcylation than the double mutant, all three mutants were highly ubiquitinated compared with wild-type in presence of GlcN, unambiguously suggesting that O-GlcNAcylation of both S317 and S319 is required for ERRγ protein stability (Fig. 3E). Consistent with these findings, we noticed that all three mutants showed significantly lower stability compared with wild-type in a pulse-chase experiment, demonstrating the importance of O-GlcNAcylation in ERRγ protein stability (Fig. 3F). This was further confirmed when all three mutants showed significantly reduced protein levels in the presence of cycloheximide, a protein synthesis inhibitor (Supplementary Fig. 4A). Next, reporter gene assay reveled that unlike wild-type, the mutant ERRγ had no significant transcriptional activity, suggesting that O-GlcNAcylation is required for ERRγ transcriptional activity as well (Fig. 3G). All three mutants were unable to bind to the PEPCK promoter revealing why they had no transcriptional activity (Fig. 3H). Collectively, these results illustrate that O-GlcNAcylation stabilizes ERRγ and increases its transcriptional activity.

Increase in transcriptional activity of transcription factors in response to O-GlcNAcylation was previously linked to their nuclear transport (7,34). Because the O-GlcNAcylation mutant ERRγ had practically no transcriptional activity compared with wild-type (Fig. 3G and H), we speculate that the O-GlcNAcylation mutant may not translocate into the nucleus. Unexpectedly, the S317A+S319A ERRγ was located in the nucleus, indicating that subcellular localization of ERRγ was not governed by O-GlcNAcylation (Fig. 4A). Transcriptional coactivator PGC-1α interacts with ERRγ and is critical for ERRγ transcriptional activity (28). Hence, we examined whether PGC-1α could interact with S317A+S319A ERRγ. In AML12 cells, GlcN treatment significantly augmented O-GlcNAcylation of ERRγ as well as ERRγ–PGC-1α interaction, whereas inhibiting HBP by treating with DON, an inhibitor of GFAT, the rate-limiting enzyme of HBP, significantly reduced O-GlcNAcylation of ERRγ as well as ERRγ–PGC-1α interaction. We could not detect any interaction between S317A+S319A ERRγ and PGC-1α, supporting the idea that the HBP mediates ERRγ–PGC-1α interaction (Fig. 4B). We also observed that GlcN treatment increased PGC-1α protein levels (Fig. 4B), but not mRNA levels (Supplementary Fig. 4B). The rise in PGC-1α protein levels could be because O-GlcNAcylation also stabilizes PGC-1α protein (30). Next, STZ treatment, an inhibitor of OGA and promoter of O-GlcNAcylation (7,35), significantly augmented PGC-1α–ERRγ interaction, demonstrating that OGA inhibition enhances O-GlcNAcylation of ERRγ that results in substantial increase in PGC-1α–ERRγ interaction (Fig. 4C). Along with ERRγ (FLAG), PGC-1α (HA) protein levels were also elevated in response to STZ treatment (Fig. 4C), which is consistent with Fig. 4B, in which GlcN treatment elevated PGC-1α protein levels. Next, we performed in vitro interaction study between PGC-1α and Gal4 construct containing either wild-type or S317A+S319A ERRγ-LBD as the LBD contains the O-GlcNAcylation site (Fig. 3C). Similar to Gal4–wild-type ERRγ–LBD, Gal4-S317A+S319A ERRγ-LBD protein was as stable as the wild-type one, but we could not detect any interaction between PGC-1α and Gal4-S317A+S319A ERRγ-LBD, suggesting that O-GlcNAcylation in the LBD regulates ERRγ–PGC-1α interaction (Fig. 4D). Similar to the double mutant, the two single mutants also could not interact with PGC-1α (Supplementary Fig. 4C). Gal4–tk-Luc reporter gene assay revealed that Gal4-S317A+S319A ERRγ-LBD was incapable of activating the reporter gene in the presence of PGC-1α (Fig. 4E). Moreover, Gal4-S317A+S319A ERRγ-LBD was also unable to activate the reporter gene even in presence of GlcN or OGT (Supplementary Fig. 4D). Taken together, these results demonstrate that O-GlcNAcylation regulates ERRγ–PGC-1α interaction critical to ERRγ transcriptional activity.

HBP induces gluconeogenic enzymes gene expression through O-GlcNAcylation (7), (30). ERRγ is a key positive regulator of gluconeogenic enzymes gene expression (28), (18). Our previous results described that ERRγ was O-GlcNAcylated through HBP (Figs. 1A, 4B and C). Therefore, to determine the contribution of O-GlcNAcylated ERRγ in HBP-induced gluconeogenesis in MPH, we knocked down endogenous ERRγ. ERRγ knockdown markedly reduced GlcN-induced PEPCK and G6Pase protein levels (Fig. 5A). OGT overexpression leads to induction of gluconeogenesis and OGT knockdown improves glucose homeostasis in diabetic mice (7,30). OGT overexpression significantly increased PEPCK and G6Pase mRNA levels, and this increase in mRNA levels was greatly suppressed by ERRγ knockdown in MPH (Fig. 5B, from left, first and second panel). In line with PEPCK and G6Pase mRNA levels results, glucose production was also significantly reduced in response to ERRγ knockdown (Fig. 5B, rightmost panel). Effectiveness of OGT overexpression was confirmed by Western blot analyses (Supplementary Fig. 4E). Next, to examine the effect of glucose or GlcN on gluconeogenic gene promoter activity, we used wild-type and ERRE mutant PEPCK promoter, which is devoid of ERRγ binding site. Exposure to glucose or GlcN increased wild-type promoter activity, but this increase was greatly reduced with the ERRE mutant PEPCK promoter (Fig. 5C). In a parallel approach, 293T cells were transfected with wild-type PEPCK promoter along with wild-type and S317A+S319A ERRγ. Wild-type ERRγ considerably increased the promoter activity that was further augmented in presence of OGT, whereas OGA coexpression greatly impaired ERRγ effect. At the same time, S317A+S319A ERRγ could not greatly activate the promoter (Fig. 5D). Next, chromatin immunoprecipitation (ChIP) assay was performed in MPH to monitor the effect of HBP inhibition on ERRγ recruitment to the endogenous PEPCK gene promoter. Under basal conditions, ERRγ occupied the PEPCK promoter. However, GlcN treatment significantly augmented ERRγ occupancy on the PEPCK promoter, whereas HBP inhibition by DON treatment markedly diminished the occupancy. We also observed a similar binding pattern of PGC-1α (Fig. 5E). Together, these results demonstrate that O-GlcNAcylation by HBP governs gluconeogenic activity of ERRγ.

Diabetic conditions induce ERRγ gene expression and promote gluconeogenesis (18). Hence, we speculated that ERRγ could be O-GlcNAcylated under diabetic conditions. As we expected, O-GlcNAcylation of ERRγ was greatly increased in HFD-fed, ob/ob, and db/db mice (Fig. 6A and B). Previously, it was reported that diabetic conditions promoted OGT gene expression (7). We also noticed a significant increase in OGT mRNA levels in HFD-induced diabetic mouse, although OGA mRNA levels were also elevated (Supplementary Fig. 4F). Overexpression of ERRγ promotes hepatic gluconeogenesis and elevates blood glucose levels (18,28). Therefore, we compared the effect of wild-type and S317A+S319A ERRγ overexpression in mouse liver. In accordance with previous results, glucose excursion during intraperitoneal glucose tolerance test was significantly higher in Ad–wild-type ERRγ-injected mice compared with control mice, but Ad-S317A+S319A ERRγ–injected mice showed normal blood glucose levels (Fig. 6C). Fasting blood glucose levels and PEPCK and G6Pase mRNA levels were significantly higher for Ad–wild-type ERRγ infection compared with Ad-S317A+S319A ERRγ (Fig. 6D). Effectiveness of ERRγ overexpression was confirmed by Western blot analyses (Supplementary Fig. 4G). Next, to negate the effect of O-GlcNAcylation, we overexpressed OGA in mouse liver. Disrupting O-GlcNAcylation of ERRγ in mice infected with Ad–wild-type ERRγ through overexpression of hepatic OGA by Ad-OGA greatly lowered the gluconeogenic profile (Fig. 6E). Effectiveness of ERRγ overexpression was confirmed by Western blot analyses (Supplementary Fig. 4H). Finally, based on the previous result (Fig. 6C and D), we performed in vivo imaging analysis to verify the effect of O-GlcNAcylation of ERRγ on hepatic gluconeogenesis at the transcriptional levels. Ad–wild-type ERRγ-dependent induction of PEPCK promoter activity was significantly reduced in mice injected with Ad-S317A+S319A ERRγ (Fig. 6F). Overall, these results suggest that O-GlcNAcylation is prerequisite for ERRγ to trigger hepatic gluconeogenesis.

Several transcription factors promote gluconeogenesis in the fasted state and type 2 diabetes. We hypothesize that in the fasting and diabetic states, gluconeogenesis generates fructose-6-phosphate, which is used by HBP to O-GlcNAcylate transcription factors and coactivators to further promote gluconeogenesis. In the current study, we show that ERRγ is stabilized by O-GlcNAcylation in the fasted and diabetic states to promote gluconeogenic gene induction. Hence, we suggest a positive feed-forward loop in which glucose entry into the HBP promotes gluconeogenesis in the fasting and diabetic conditions in the liver.

Glucagon–insulin crosstalk regulates ERRγ protein stability and transcriptional activity (20,28). In this study, we investigated whether ERRγ protein stability and transcriptional activity is influenced by O-GlcNAcylation. In fact, glucagon stabilized ERRγ by promoting its O-GlcNAcylation (Fig. 2D–F). O-GlcNAcylation can increase protein stability and transcriptional activity by inhibiting ubiquitination or promoting deubiquitination (30–33). It can also reduce protein stability and transcriptional activity by increasing ubiquitination (36,37). However, we observed that O-GlcNAcylation stabilized ERRγ protein by inhibiting its ubiquitination (Fig. 2D–F). Our results clearly demonstrated that incremental O-GlcNAcylation–mediated reduction in ubiquitination was a result of greater inhibition of the interaction between ERRγ and E3 ubiquitin ligase Mdm2 that resulted in increased protein stability (Fig. 2C). Phosphorylation of GFAT1 by cAMP-dependent protein kinase blocks its enzyme activity (38), whereas phosphorylation of GFAT2 by cAMP-dependent protein kinase increases its enzyme activity (39). Our observation that glucagon robustly enhances O-GlcNAcylation of ERRγ and insulin inhibits it (Fig. 2F) could be due to GFAT2 activation. O-GlcNAcylation can affect transcription factors by modifying key residues involved in their interaction with coactivators (40). It can also induce important conformational changes within transcription factors, which might have a direct impact on their activity, as demonstrated for the estrogen receptor (41). Three-dimensional structural features of ERRγ also suggest that O-GlcNAcylation may trigger a conformational change of ERRγ–LBD that contains the AF-2 domain (Fig. 3D). This probable conformational change may be crucial for inhibition of ubiquitination and ERRγ–PGC-1α interaction. Perhaps, S317A ERRγ and S319A ERRγ, in spite of being partially O-GlcNAcylated, were heavily ubiquitinated (Fig. 3E) and unable to interact with PGC-1α (Supplementary Fig. 4C). The two single mutants may not attain the desired conformational change required to inhibit ubiquitination and interact with coactivator PGC-1α due to incomplete O-GlcNAcylation. Further investigation is required to confirm whether O-GlcNAcylation influences ERRγ–LBD structure. Previously, Yang et al. (42) reported that the transcripts of ERRγ followed a cyclic pattern in its diurnal rhythmicity in the liver. ERRγ transcripts reach maximum levels in the liver during the daytime. The observed result of O-GlcNAcylation–mediated stability of ERRγ may have been influenced by the diurnal rhythmicity in the liver.

O-GlcNAcylation takes place in response to high glucose or insulin (31,43,44). PGC-1α and CRTC2 undergo O-GlcNAcylation in hyperglycemic and hyperinsulinemic diabetic mouse (7,30). Interestingly, low-glucose conditions also promote O-GlcNAcylation (7,9,30). Glucose deprivation induces protein O-GlcNAcylation and OGT expression as well (45). All of these reports suggest that both high- and low-glucose conditions promote O-GlcNAcylation in vitro and in vivo. Glucose in the form of fructose-6-phosphate is used by hexosamine biosynthetic pathway to O-GlcNAcylate proteins under different conditions. In the fed conditions, high circulating glucose could be converted to fructose-6-phosphate and used in O-GlcNAcylation. Although the circulating glucose concentration is high in the diabetic condition, the diabetic condition differs from the fed condition in that the activity of glucokinase, which converts glucose to glucose-6-phosphate, is low in the liver (46), (47,48). Therefore, circulating glucose is less readily converted to fructose-6-phosphate in the diabetic condition. In the diabetic conditions hepatic gluconeogenesis is significantly upregulated (49), producing fructose-6-phosphate, which could be used in O-GlcNAcylation. Furthermore, in low circulating glucose (fasting) conditions, glycogenolysis and gluconeogenesis are both called into play to maintain blood glucose levels. During early fasting, glycogenolysis, stimulated by glucagon, produces glucose-6-phosphate, which is in equilibrium with fructose-6-phosphate (50). Therefore, glycogenolysis, especially during early fasting, is surely a major source of fructose-6-phosphate. During short-term fasting, glucagon also triggers the initial induction of hepatic gluconeogenesis through activation of CREB-CRTC2 (7). In prolonged fasting, PGC-1α, FOXO1, and ERRγ promote hepatic gluconeogenesis (28,30,51). During that initial induction period (short-term fasting), hepatic glycogenolysis and gluconeogenesis produce fructose-6-phosphate, which might be used to O-GlcNAcylate ERRγ to further promote gluconeogenesis during prolonged fasting. We observed that blood glucose levels were decreased during fasting in a time-dependent manner, whereas serum glucagon levels and ERRγ O-GlcNAcylation levels were steadily increased (Supplementary Fig. 1E and F), indicating that circulating glucagon levels are more important than circulating glucose levels for O-GlcNAcylation of ERRγ in fasting. Moreover, OGT mRNA levels were significantly enhanced in response to fasting, even though OGA mRNA levels were also enhanced, but it was less significant compared with OGT (Supplementary Fig. 1D). Enhanced OGT mRNA levels could also be responsible for fasting-dependent O-GlcNAcylation of ERRγ.

Conserved ERRE on the PEPCK promoter is required for the transcription of that gene in response to fasting and diabetes-mediated gluconeogenesis (18,28); hence, we explored the role of ERRγ in HBP mediated gluconeogenesis. Loss of endogenous ERRγ in primary hepatocytes led to a significant decrease in HBP-induced gluconeogenic profile (Fig. 5), implying the importance of O-GlcNAcylation of ERRγ in the context of gluconeogenesis. Diabetic conditions promote O-GlcNAcylation mediated gluconeogenesis in the diabetic mice (7,30). As a matter of fact, ERRγ was highly O-GlcNAcylated in diabetic mice (Fig. 6A and B). Hepatic overexpression of wild-type ERRγ caused glucose intolerance with hyperglycemia, whereas O-GlcNAcylation mutant ERRγ overexpression showed glucose tolerance with euglycemia in mice (Fig. 6C and D). O-GlcNAcylation was blocked by using either OGT or GFAT inhibitor or by enzymatically modulating OGA or OGT expression to investigate the effect of their inhibition on glycemia in diabetic mice. The OGT inhibitor alloxan was used in many studies (52–54), but it has wide off-target effects and general cellular toxicity (55). Another OGT inhibitor, Ac₄-5S-GlcNAc, was used in wide range of studies (56–58), but it affects other glycosyltransferase and impairs N-glycosylation and extracellular glycan synthesis in cultured cell lines (59). Moreover, the enzymatic approach modulating OGA or OGT expression was successfully used to restore glucose homeostasis. Overexpression of hepatic OGA or knockdown of hepatic OGT inhibited aberrant gluconeogenesis and significantly improved glycemic conditions in diabetic mice (7,30,31). We observed that disruption of O-GlcNAcylation by hepatic OGA overexpression reduced gluconeogenic profile in normal mice expressing Ad-ERRγ (Fig. 6E), unambiguously illustrating that O-GlcNAcylation is prerequisite for gluconeogenic function of ERRγ.

We conclude that the fasting and diabetic conditions promote O-GlcNAcylation of ERRγ. O-GlcNAcylation is imperative for ERRγ protein stability and enhances gluconeogenic activity of ERRγ. The fed state, however, reduces O-GlcNAcylation of ERRγ, resulting in ubiquitin-mediated degradation of ERRγ (Fig. 6G). Our results indicate a vital role for O-GlcNAcylation of ERRγ in maintaining normal glucose levels during fasting and also in mediating the elevated blood glucose levels in type 2 diabetes. Hence, pharmacological inhibition of O-GlcNAcylation–mediated hyperactivation of ERRγ might provide a pathway for preventing hyperglycemia and treating type 2 diabetes.

Acknowledgments. The authors thank David Moore (Baylor College of Medicine, Houston, TX) and Seok-Yong Choi (Department of Biomedical Sciences, Chonnam National University Medical School, Gwangju, Republic of Korea) for helpful discussion and Ji Min Lee, Yong Soo Lee, Ki Sun Kim, Soon-Young Na, and Yaochen Zhang (Chonnam National University, Gwangju, Republic of Korea) for technical assistance.

Funding. This work was supported by National Creative Research Initiatives Grant 20110018305 through the National Research Foundation of Korea (NRF) funded by the Korean government (Ministry of Science, ICT & Future Planning) (to H.-S.C.). This research was supported by Grant HI16C1501 of the Korea Health Technology R&D Project through the Korea Health Industry Development Institute (KHIDI) and funded by the Ministry of Health & Welfare, Republic of Korea (to I.-K.L.). This research was also supported by the NRF funded by the Ministry of Education, Science, and Technology (Grants NRF-2013R1A2A1A01008067 and NRF-2015M3A9B6073840 to J.W.C.).

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

Author Contributions. J.M. conceived and designed research, performed experiments, analyzed the data, and wrote the paper. D.-K.K., Y.S.J., E.-K.Y. provided reagents/materials/analysis tools. H.B.K. and B.G.K. performed liquid chromatography-tandem mass spectrometry collision-induced dissociation site mapping analysis for identification of O-GlcNAcylation sites. Y.-H.K. performed mouse in vivo imaging. S.K. performed liquid chromatography-tandem mass spectrometry collision-induced dissociation site mapping analysis for identification of O-GlcNAcylation sites and analyzed the data. I.-K.L., R.A.H., J.-S.K., and C.-H.L. analyzed the data. J.W.C. analyzed the data and provided reagents/materials/analysis tools. H.-S.C. conceived and designed research, analyzed the data, and wrote the paper. H.-S.C. 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.