Novel Mechanism of Foxo1 Phosphorylation in Glucagon Signaling in Control of Glucose Homeostasis

Glucagon and insulin are the most important pancreatic hormones in target tissues, such as liver, in control of glucose homeostasis in the body in response to food intake (1,2). During the fasting state, glucagon is secreted from the pancreatic α-cells to elevate blood glucose through stimulation of gluconeogenesis and glycogenolysis, protecting the body from hypoglycemia (3). However, an excess of blood glucagon level is present in animals and patients with diabetes, stimulating excessive hepatic gluconeogenesis and contributing to diabetic hyperglycemia (1,4–7).

Glucagon exerts its function through binding to the glucagon receptor (GCGR), a G-protein–coupled receptor. Upon binding to glucagon, GCGR in the cell membrane activates adenylate cyclase to elevate intracellular cAMP and subsequently activates protein kinase A (PKA) signaling (1,2). PKA regulates the metabolic enzymes or gene expression for glycogenolysis and gluconeogenesis, including glycogen phosphorylase, glucose-6-phosphatase catalytic subunit (G6pc), and phosphoenolpyruvate carboxykinase-1 (Pck1) (8), increasing blood glucose. Importantly, recent studies demonstrated that inhibition of GCGR by a genetic approach or chemical administration in mice significantly decreases the blood glucose and completely prevents streptozotocin-induced hyperglycemia in type 1 diabetes (1), suggesting that glucagon-mediated hepatic glucose production (HGP) or gluconeogenesis is a potential target for therapeutic intervention of diabetes (9,10).

Gluconeogenesis is suppressed by insulin involved in the gene transcriptional regulation in the nucleus. Foxo1, a member of the Foxo family, is an important component of insulin-signaling cascades in regulating cellular growth, differentiation, and metabolism (11). Our and other studies demonstrated that Foxo1 enhances gene transcription of the rate-limiting enzymes responsible for gluconeogenesis, including G6pc and Pck1, through interaction with a conserved insulin-responsive element on the promoter region of the target genes (12–14), and increases HGP (11,15–17). However, insulin suppresses Foxo1 by the insulin receptor substrate 1 and 2 (IRS1 and 2)–dependent activation of Akt that phosphorylates Foxo1 at Ser²⁵⁶ in human hepatocytes, which is equivalent to mouse Foxo-S253, promoting Foxo1 nuclear export and/or degradation and then decreasing HGP and blood glucose (18–21).

Foxo1 protein is stabilized in the liver of fasting mice when insulin decreases and glucagon increases in the blood circulation. Our early studies indicate that Foxo1 may partially mediate the effect of cAMP, a second messenger of glucagon, or glucocorticoid-induced expression of genes encoding gluconeogenic enzymes, such as G6pc and Pck1 (22,23). Recently, a calcium-sensing enzyme, Ca²⁺/calmodulin-dependent protein kinase II (CaMKII), is shown to increase seven non-Akt phosphorylation sites of Foxo1, mediating the action of glucagon and cAMP signaling and enhancing Foxo1 nuclear localization and HGP, but detailed phosphorylation sites involved in glucagon or cAMP signaling remain elusive in vivo (24). In this study, we hypothesize that Foxo1 may play a role in the action of glucagon in control of HGP or gluconeogenesis and investigated the molecular and physiological mechanism of Foxo1 regulation by glucagon via PKA with the aim of better understanding the fundamentals of blood glucose homeostasis and pathogenesis of diabetes.

To determine whether Foxo1 is directly phosphorylated by PKA, we performed in vitro kinase assay, in which 20 µg of recombinant Foxo1–glutathione-S-transferase (GST) fusion protein was purchased from Millipore Inc. as previously described (18). The purified Foxo1-GST recombinant protein was incubated with 5 units of recombinant catalytic subunit of PKA (PKAc; Upstate Biotechnology) for 1 h, at 30°C, in 50 µL reaction buffer (25 mmol/L Tris-HCl, pH 7, 1 mmol/L EGTA, 0.5 mmol/L EDTA, 0.5 mmol/L β-mercaptoethanol, and 3 mmol/L magnesium acetate) with or without 0.1 mmol/L ATP. The protein bands of interest in the SDS-PAGE gel were incised and prepared for liquid chromatography–mass spectrometry (LC-MS) analysis, as described in the Supplementary Data.

Plasmid DNA expressing human Foxo1-WT (wild-type) was previously described (12). Oligos specific for human Foxo1-S153A, Foxo1-S153D, Foxo1-S276A, or Foxo1-S276D were designed, the QuickChange Site-Directed Mutagenesis kit was purchased (Stratagene, La Jolla, CA), and mutant sites were confirmed by DNA sequencing.

All animal experiments were performed according to procedures approved by the Texas A&M University Institutional Animal Care and Use Committee. The floxed Foxo1 mice (Foxo1^L/L) (25) and albumin-Cre mice were previously described (26). All of the mice on a C57BL/6 and 129 Sv mixed background were maintained on regular chow (Isopro 5P76; Prolab). Foxo1-S273A and Foxo1-S273D knock-in (KI) mice were generated using the CRISPR/Cas9 approaches (27) and the microinjection and mouse colonies production performed by Cyagen Biosciences Inc. (Santa Clara, CA; contract number: NTMCN-150310-AML-01). The founder mice were backcrossed to C57/BL6 mice, and F3 offspring male mice were used for analysis. The tail DNA of pups was genotyped by PCR and then confirmed by DNA sequencing analysis. The db/db mice were purchased from The Jackson Laboratory (Bar Harbor, ME). If not specified elsewhere, all of the mice were male at the ages of 8–12 weeks. For the high-fat diet (HFD)–induced insulin-resistant mouse model, 8–10-week-old male mice were fed with HFD (42% kcal from fat, TD.88137; Envigo) or control low-fat diet (LFD; 13% kcal from fat, TD08485; Envigo) for 6 months. For glucagon-tolerant tests, mice after an 18-h fast were i.p. injected with glucagon of 16 µg/kg body weight, and blood glucose was measured with a glucometer (Elite XL; Bayer). The dose of glucagon used was based on a recent report (28).

We synthesized two peptides of human Foxo1 corresponding to residues 143–162 (GPLAGQPRKSpSSSRRNAWGN) and 265–285 (KSRSRAAKKKApSLQSGQEGAG), where “p” indicates the site of phosphorylation (S153 and S276). The peptides were conjugated to both keyhole limpet hemocyanin and BSA and used to immunize rabbits at Covance (Denver, PA; Custom Immunology Service, CRPQ20383-01P), as previously described (18).

Serum was analyzed for insulin (Crystal Chem) and glucagon (Alpco) using commercial protocol and reagents. Blood glucose was measured using a glucometer (Bayer). For glucagon stimulation experiments, 8–12-week-old mice were subjected to an 18-h fasting, injected i.p. by glucagon (pharmaceutical grade, 0.25 mg/kg body weight), and blood glucose monitored. For GCGR antagonist injection, 6-h fasted control mice were i.v. injected with GCGR antagonist ([des-His¹, Glu⁹]-glucagon amide; 1 mg/kg body weight).

Proteins were prepared from cells or livers, resolved by SDS-PAGE, and transferred to nitrocellulose membrane for immunoblotting analysis using specific antibodies, as previously described (29). The signal intensity was measured and analyzed by ImageJ software (National Institutes of Health).

Nuclear and cytoplasmic proteins from HepG2 or liver tissue were extracted with NE-PER nuclear (NER) and cytoplasmic (CER-I and CER-II) extraction reagent (Pierce), as previously described (29).

The primary mouse hepatocytes were isolated from 8–12-week-old mice via 0.05% collagenase (type 1; Worthington Biochemical) and cultured as previously described (16). For HGP assays, freshly isolated hepatocytes were resuspended in DMEM with 2% FBS for 4 h, then rinsed with PBS, and cultured in HGP buffer (118 mmol/L NaCl, 2.5 mmol/L CaCl₂, 4.8 mmol/L KCl, 25 mmol/L NaHCO₃, 1.1 mmol/L KH₂PO₄, 1.2 mmol/L MgSO₄, 10 μmol/L ZnSO₄, 0.6% BSA, 10 mmol/L HEPES, 10 mmol/L sodium DL-lactate, and 5 mmol/L pyruvate, pH 7.4) in the presence or absence of 100 nmol/L glucagon or 8-Br–cAMP. Cell culture medium was collected at 3 and 6 h and glucose in the medium measured according to the manufacturer’s protocol using Amplex Red Glucose Assay (Invitrogen) (30,31). For glycogenolysis assay, lactate and pyruvate in the HGP buffer were removed and glucose release into the medium measured after treatments.

RNA was extracted with TRIzol reagent (Invitrogen), cDNA synthesis used the Super-Script first-strand synthesis system (Bio-Rad), and gene expression was measured with the SYBR Green Supermix system (Bio-Rad), as previously described (29).

HepG2 cells were cultured with DMEM with 10% FBS for 6 h. Cell cultures were then subjected to Lipofectamine 3000 (Life Technologies) with plasmids according to the manufacturer’s instruction. The dual-luciferase reporter assay system from Promega was purchased, luciferase activity measured with an Optocomp-I luminometer (MGM Instruments), and cells expressing eGFP and eGFP-Foxo1 visualized with the confocal microscope (Leica Biosystems) and calculated as previously described (32).

The comparison of differences between two groups was performed using a Student two-tailed t test to determine the significance (26). ANOVA tests were used when comparing two groups among multiple groups. All data are presented as mean ± SEM. A P value <0.05 was considered statistically significant.

Previous studies from our group and others have shown that liver-specific Foxo1 knockout mice (L-F1KO) or hepatic Foxo1 suppression exhibited a 15–20% reduction in blood glucose as compared with control mice (16,17,33). To determine whether Foxo1 is involved in the action of glucagon-mediated blood glucose homeostasis, we administered glucagon i.p. in L-F1KO and control-fed mice. Glucagon significantly increased blood glucose by 32% in control mice 60 min after glucagon injection, but the increased potential in L-F1KO mice was 21%, with a significant reduction compared with control mice (Fig. 1A). We next performed the glucagon tolerance test in the mice in the fasting state. Both control and L-F1KO mice exhibited an ∼1.7-fold elevation in blood glucose 15 min after glucagon injection. Importantly, the ability of blood glucose increase was maintained in 60 min after glucagon injection in control mice; however, L-F1KO mice exhibited a 50% reduction in blood glucose after glucagon stimulation compared with control mice (P < 0.05) (Fig. 1B). Gene expression analysis indicated that glucagon stimulated hepatic gene transcriptional levels of Igfbp-1, G6pc, and Pck1 in control mice by 4.4-, 2.6-, and 2.0-fold, respectively. However, glucagon stimulation on Igfbp-1, G6pc, and Pck1 was 3.2-, 2.0-, and 1.8-fold, respectively, in L-F1KO liver, exhibiting significant reductions compared with control (Fig. 1C–E).

To determine whether Foxo1 deficiency impairs glucagon-mediated glucose production via glycogenolysis and gluconeogenesis in an autonomous manner in hepatocytes, we measured HGP with glucagon treatment in primary hepatocytes isolated from L-F1KO and control mice. Glucagon significantly increased HGP by 42% in control hepatocytes and by 35% in L-F1KO hepatocytes, exhibiting a 28% reduction in L-F1KO cells (Fig. 1F). Both glycogenolysis and gluconeogenesis contribute to HGP; L-F1KO cells exhibited a 50% reduction in glucagon-stimulated gluconeogenesis compared with control (46% stimulation in control vs. 23% in L-F1KO; P < 0.01). Glucagon significantly increased glycogenolysis by 21% in control cells; L-F1KO hepatocytes had no blockage for glucagon, even though the basal level of glycogenolysis was 29% lower in L-F1KO than control cells (71 ± 4.2 µmol/h/g protein vs. 50 ± 4.9 µmol/h/g protein; P < 0.01) (Fig. 1F). Taken together, these data indicate that Foxo1 deficiency impaired glucagon-induced liver gene expression, gluconeogenesis, and blood glucose.

We next examined the effect of insulin versus glucagon on Foxo1 regulation in hepatocytes. HepG2 cells were cultured and treated the hormones over a 6-h time course. Insulin treatment for 3 h significantly reduced total Foxo1 protein by 20%. However, glucagon treatment for 3 h significantly increased total Foxo1 protein by 1.7-fold (Fig. 2A and B). The phosphorylation level of Foxo1 at S256 was increased 2.5-fold by insulin for 3-h treatment and persisted up to 6 h, whereas glucagon barely stimulated Foxo1-S256 phosphorylation even that the phosphorylation was slightly reduced at 3 h of treatment (Fig. 2A and C). Insulin treatment for 15 min increased by 23-fold on Akt phosphorylation at T308, which was increased 5-fold by glucagon when compared with the basal level of Akt phosphorylation in cells (Fig. 2A and D). By contrast, insulin barely increased phosphorylation of CREB at S133, a target of PKA, but glucagon treatment for 15 min significantly increased CREB-S133 phosphorylation by 1.5-fold without changing the total CREB protein level (Fig. 2A, B, and E). Moreover, Foxo1 mRNA levels were unaffected by glucagon in HepG2 cells (Fig. 2F). These results prompted us to hypothesize that glucagon may promote Foxo1 stability through activation of PKA and control Foxo1 target gene expression, impelling us to assess the impact of Foxo1 in glucagon-regulated HGP in vivo.

PKA and Akt serve as important downstream effectors in glucagon and insulin signaling, respectively, in control of glucose homeostasis (15,34). In order to determine whether PKA affects Foxo1-mediated gene transcriptional activity in hepatocytes, we transfected HepG2 cells with the Igfbp-1–luciferase reporter gene construct with plasmid DNA expressing control GFP, PKAc, or constitutively active Akt (Myr-Akt), and then luciferase activities were measured. Overexpression of PKAc alone increased the basal promoter activity by at least twofold, whereas overexpression of the active form of Akt (MyrAkt) inhibited the promoter activity by 55% (Fig. 3A). Moreover, expression of Foxo1 alone increased Igfbp-1 promoter activity by 2.9-fold, and coexpressing PKAc and Foxo1 resulted in a robust increase of Igfbp-1 promoter activity, at least by >10-fold. By contrast, co-overexpression of Myr-Akt and Foxo1 almost completely diminished the Foxo1-stimulated Igfbp-1 promoter activity (Fig. 3A). Together, these data suggest that PKA had an additive effect on promoting Foxo1-stimulated Igfbp-1 promoter activity.

Analysis of Foxo1 phosphorylation sites (https://scansite4.mit.edu/4.0/#home) revealed that human Foxo1 contains three PKA consensus sequences (R/KXRT/S, R is arginine, K is lysine, and X is any amino acid), located at Ser¹⁵³, Thr¹⁸², and Ser²⁷⁶, which are evolutionarily conserved across species (Fig. 3B). We hypothesize that glucagon-activated PKA phosphorylates Foxo1 via these putative PKA phosphorylation sites and enhances Foxo1 stability and transcriptional activity.

Next, we performed an in vitro kinase assay to determine whether PKA can directly phosphorylate Foxo1. GST–fused Foxo1 (Foxo1-GST) and active PKAc were incubated in a protein kinase assay buffer system and then resolved by SDS-PAGE followed by the Coomassie Brilliant Blue staining. Phosphorylated Foxo1 fragment (60 kDa) by PKAc slightly upshifted in mobility (Fig. 3C). The Foxo1-GST protein (105 kDa) and Foxo1 fragment (60 kDa) were incised from the gel and subjected to LC–tandem MS (MS/MS) analyses. The results from Mascot analysis of MS of PKAc with or without ATP-treated samples indicated a presence and high score for Foxo1-S276 phosphorylation, a low score for Foxo1-S153 phosphorylation, and no presence and score for T182 phosphorylation (Fig. 3D). Based on these results, we generated the phosphospecific antibody against human Foxo1-S276 or Foxo1-S153, which is equivalent to mouse Foxo1-S273 or S150, respectively. In HepG2 cells, glucagon significantly stimulated the phosphorylation of Foxo1-S273 and S153 by 2.5- and 1.5-fold, respectively (Fig. 3E). The phosphorylation of S276 and S153 by glucagon was blocked when PKA catalytic subunit C was knocked down by small interfering RNA (siRNA) (Fig. 3E). The knockdown of PKA markedly decreased glucagon-stimulated CREB phosphorylation and Foxo1 protein level, suggesting that PKA is essential for maintaining Foxo1 protein stability (Fig. 3E). We further confirmed these findings in mouse primary hepatocytes; glucagon treatment enhanced the phosphorylation of Foxo1 at S273 and S150, as well as total Foxo1 protein levels, whereas these effects of glucagon were abolished in the cells when treated with H89, a PKA chemical inhibitor (Fig. 3F). In mouse liver, glucagon injection resulted in a 3.8-fold increase in Foxo1-S273 phosphorylation and a 1.6-fold increase in Foxo1-S150 phosphorylation (Fig. 4A). Moreover, we administered GCGR antagonist into WT mice and examined whether endogenous Foxo1 protein level is regulated by the glucagon receptor in the liver. Intravenous injection of GCGR antagonist ([des-His¹, Glu⁹]-Glucagon amide) resulted in a 20% decrease in blood glucose (Fig. 4B and C), as well as a 60% decrease of CREB phosphorylation, 55% decrease in Foxo1-S273 phosphorylation, and 35% decrease in total Foxo1 protein level, whereas Foxo1-S150 phosphorylation was barely affected compared with vehicle treatment (Fig. 4D). These results suggest that glucagon signaling via its receptor and PKA stimulates hepatic Foxo1 phosphorylation at S273, enhancing Foxo1 stability.

We next examined whether the newly identified Foxo1 phosphorylation sites play a role in Foxo1 nuclear/cytoplasmic trafficking and stability in cells. We generated plasmid DNA expressing GFP-Foxo1-WT, GFP-Foxo1-S153A, or GFP-Foxo1-S276A, in which serine residue was substituted with an alanine (A) at S153 or S276 to block phosphorylation, whereas GFP-Foxo1-S153D or GFP-Foxo1-S276D containing an aspartate (D) substitution to mimic phosphorylation by introduction of a negative charge. HepG2 cells were transfected with one of these plasmids and then stimulated with 8-Br–cAMP for 15 min prior to confocal imaging. In cells that overexpressed Foxo1-WT, Foxo1 localized in both cytoplasm and nucleus at the basal level, but cAMP stimulation promoted Foxo1 nuclear localization. However, cAMP-induced Foxo1 nuclear localization was blocked by S276A mutation, in which overexpressed Foxo1-S276A dispersed primarily within the cytoplasm. Conversely, Foxo1-S276D primarily localized in the nucleus (Fig. 5). Compared with Foxo1-WT, Foxo1-S153A or S153D did not exhibit distinct differences in Foxo1 subcellular localization in cells, and overexpressed Foxo1-S153A and 153D were primarily located to the nucleus in responses to cAMP, similar to Foxo1-WT (Fig. 5). Thus, these results suggest that Foxo1-S276 phosphorylation is crucial for mediating cAMP-stimulated Foxo1 nuclear localization.

To confirm the Foxo1 subcellular localization upon glucagon or cAMP stimulation, we extracted nuclear and cytoplasmic protein from cells for Foxo1 analysis. Glucagon or cAMP stimulated endogenous Foxo1 nuclear localization by nearly threefold compared with vehicle control group (Fig. 6A) (P < 0.05). By contrast, insulin reduced nuclear total Foxo1 by 25% (Fig. 6A). These data indicate that either glucagon or cAMP induced Foxo1 nuclear localization, which is distinct from insulin.

We next further focused on Foxo1-S276 phosphorylation in control of Foxo1 stability in cells. HepG2 cells were transfected with plasmid DNA that expresses Foxo1-WT, Foxo1-S276A, or Foxo1-S276D and then stimulated with glucagon for 30 min followed by nuclear and cytoplasmic protein extraction. In Foxo1-WT–transfected cells, glucagon increased total Foxo1 protein in the nucleus by 1.5-fold (Fig. 6B and C). Importantly, Foxo1-S276D protein, which mimics S276 phosphorylation with introduction of a negative charge, increased total Foxo1 protein abundance in both nucleus and cytoplasm, mimicking the effect of glucagon on Foxo1 nuclear localization (Fig. 6B). By contrast, Foxo1-S276A, which contains a substitution of alanine, largely prevented glucagon-stimulated Foxo1 nuclear localization (Fig. 6C).

To further examine the mechanism of Foxo1-S276 phosphorylation in control of Foxo1 stability, we transfected with an equal amount of Foxo1 gene construct with HepG2 cells and then treated cells for 8 h with MG132, a 26S proteasome complex inhibitor, preventing Foxo1 protein ubiquitination or cycloheximide (CHX), a protein synthesis inhibitor, inhibiting Foxo1 protein synthesis. When Foxo1-WT, Foxo1-S276A, or Foxo1-S276D plasmid DNA was transfected, the total Foxo1 protein in Foxo1-S276A group was reduced by 50% compared with overexpressed Foxo1-WT (P < 0.05) (Fig. 6D). However, in the MG132-treated group, the reduction of Foxo1-S276A protein was totally prevented, Foxo1-S276A protein was restored and stabilized to a normal level when compared with Foxo1-WT, and there was no significant difference in total Foxo1 protein levels observed in all of the cells transfected with Foxo1-WT, S276A, or S276D (Fig. 6D). Moreover, when Foxo1 protein synthesis was inhibited by CHX in cells, the remaining Foxo1 protein level in the Foxo1-S276D group increased by 40%, whereas it decreased by 44% in the Foxo1-S276A group compared with the Foxo1-WT group (Fig. 6D). These data suggest that Foxo1-S276 phosphorylation promoted Foxo1 stability, whereas nonphosphorylation of S276 enhanced 26S proteasome-dependent degradation of Foxo1.

We next determined whether Foxo1-S276 phosphorylation or Foxo1-S276D affects insulin-induced Foxo1 ubiquitination in cells. HepG2 cells were transfected with an equal amount of Foxo1-WT or S276D plasmid DNA together with ubiquitin for 10 h, then stimulated with insulin for an additional 12 h, and then total cellular proteins were extracted for analysis. Western blotting analysis revealed that insulin reduced total Foxo1 protein by 50% in Foxo1-WT– or S276A-expressed cells, whereas the effect of insulin was abolished in cells expressing Foxo1-S276D mutant (Fig. 6E).

We further assessed whether Foxo1-S276 phosphorylation or Foxo1-S276D affects insulin-induced Foxo1 nuclear translocation in cells. HepG2 cells were transfected with an equal amount of Foxo1-WT or S276D plasmid DNA for 18 h and then stimulated with insulin over a 120-min time course. Upon insulin stimulation for either 90 or 120 min, Foxo1-WT protein was primarily located in the cytoplasm (Fig. 7A–C). Foxo1-S276D protein primarily localized at both cytoplasm and nucleus with increased levels compared with Foxo1-WT. However, Foxo1-S276D largely blocked insulin-stimulated nuclear export, particularly at 120-min treatment of insulin (Fig. 7A), which was confirmed by Western blotting. Moreover, the blotting result showed that the cytoplasmic Foxo1-S256 phosphorylation by insulin for a 90- or 120-min treatment was significantly impaired by S276D mutation because the basal level of Foxo1-S256 phosphorylation was higher in S276D than WT in cells (Fig. 7B and C). These results reveal that Foxo1-S276 phosphorylation or aspartate mutation promoted Foxo1 nuclear localization and stability and impaired the ability of insulin to induce Foxo1 nuclear export and/or degradation, which may serve as a mechanism for insulin resistance at the Foxo1 level.

To further explore the physiological function of equivalent human Foxo1-S276 in glucose metabolism in mice, we used the CRISPR/Cas9 approach to generate Foxo1-S273A and Foxo1-S273D KI mice, in which the endogenous Foxo1-S273 alleles were replaced by alanine (A) or aspartate (D) mutations (Supplementary Fig. 1). All of the mice survived with no obvious physiological deficiency. The feeding blood glucose increased by 20% in Foxo1-S276^D/D mice, and there was no change in Foxo1-S276^A/A mice compared with control. The fasting blood glucose increased by 31% in Foxo1-S276^D/D mice but decreased significantly by 10% in Foxo1-S276^A/A mice (Fig. 8A). Hepatic Igfbp-1 and G6pc mRNA expression significantly increased by 4.5- and 2.1-fold in the liver of Foxo1-S273^D/D compared with control liver, respectively. However, there was a 25% reduction for both gene expression in S273^A/A liver compared with control liver (P < 0.05) (Fig. 8B). The total Foxo1 protein level in the liver was increased by 2.6-fold in Foxo1-S273^D/D mice, but reduced by 40% in Foxo1-S273^A/A mice when compared with control (P < 0.05) (Fig. 8C).

We further assessed the direct role of Foxo1-S273 phosphorylation in control of HGP using the primary hepatocytes isolated from Foxo1-S273^A/A, S273^D/D, and WT mice. The basal level of HGP was increased by 46% in Foxo1-S273^D/D hepatocytes, but the basal potential was reduced by 40% in Foxo1-S273^A/A hepatocytes, compared with control hepatocytes (P < 0.01) (Fig. 8D). Upon glucagon stimulation, HGP was increased by 53% in control hepatocytes; however, this stimulatory effect was significantly impaired or attenuated by either Foxo1-S273D or S273A mutation, which exhibited an increase by 41% or 20%, respectively (P < 0.01) (Fig. 8D). Glucagon-stimulated glycogenolysis barely impaired by Foxo1 mutation in the cells.

We next performed glucagon tolerance tests in Foxo1-S273^A/A, S273^D/D, and WT mice in the fasting state. Both WT and Foxo1-S273^D/D mice exhibited a 1.8-fold increase, and Foxo1-S273^A/A mice had a 1.6-fold increase in blood glucose after 15 min of glucagon injection, indicating that Foxo1-S273^A/A mutant impaired glucagon stimulation of blood glucose. Importantly, in WT and Foxo1-S273^A/A mice, blood glucose returned to initial level 60 min after glucagon injection; however, high blood glucose level was maintained in Foxo1-S273^D/D mice, indicating that Foxo1-S273^D/D mutant prolonged glucagon stimulation of blood glucose (P < 0.05) (Fig. 8E and F). The impaired effects of glucagon stimulation on HGP and blood glucose by Foxo1 mutations at S273 suggest the important role of Foxo1-S273 phosphorylation plays in glucagon signaling in control of hepatic gluconeogenesis and blood glucose.

Finally, we examined Foxo1-S273 phosphorylation in the liver of db/db diabetic mice, as well as HFD-induced insulin-resistant mice. The db/db mice exhibited elevations in blood glucagon, glucose, and serum insulin by 3-, 3-, and 10-fold, respectively, compared with control mice (Fig. 8G). Moreover, total Foxo1 protein and S273 phosphorylation significantly increased in the liver of db/db mice and accumulated in the nucleus (Fig. 8H). In parallel with increased Foxo1-S273 phosphorylation, PKA-T197 phosphorylation indicative of PKA activity was also markedly enhanced in both nucleus and cytoplasm in the liver of db/db mice (Fig. 8I). The HFD-fed mice exhibited increased blood glucagon, glucose, and serum insulin by 1.8-, 1.4-, and 5-fold, respectively, compared with LFD-fed mice (Fig. 8J). Moreover, total Foxo1 protein and Foxo1-S273 phosphorylation significantly increased in the liver of HFD-fed mice when compared with the LFD-fed liver (Fig. 8K and L). These data suggest that Foxo1-S273 phosphorylation and its nuclear retention, along with excess glucagon and PKA activity, coexist in the liver of animals with type 2 diabetes or insulin resistance.

We and other groups have previously established that insulin phosphorylates human Foxo1-S256 by activation of Akt that promotes Foxo1 cytoplasmic sequestration or nuclear export and then suppresses expression of genes responsible for HGP (12,18,20,21,35). In this study, using in vitro kinase assay–coupled LC-MS, phosphospecific antibodies, and CRISPR/Cas9-based Foxo1 KI mutant mice, we were able to identify a novel molecular, cellular, and physiological mechanism by which Foxo1 mediates glucagon signaling via phosphorylation at Ser²⁷⁶ (human) or Ser²⁷³ (mouse) in control of hepatic gluconeogenesis and blood glucose.

Glucagon has been implicated in the pathogenesis of diabetic hyperglycemia, largely by enhancing HGP, which is believed to be a key mechanism for pathogenesis of diabetes (36). In mice with type 2 diabetes, we recently demonstrated that hepatic Foxo1 deletion reduced HGP and blood glucose in db/db mice (16). Given that a high glucagon level is present in both types of diabetes (6), hyperactivation of Foxo1 by activated PKA and inactivated Akt upon insulin resistance or deficiency may serve as fundamental mechanisms for excess liver gluconeogenesis resulting in diabetic hyperglycemia.

The mechanism by which Foxo1-S276 phosphorylation promotes Foxo1 nuclear retention and stability is involved in the prevention of proteasome-mediated Foxo1 ubiquitination. We expect that Foxo1-S276 phosphorylation might enhance its recruitment with a number of cofactors, such as importin or CREB, to participate in Foxo1 subcellular trafficking and gene transcription, but detailed molecular aspects warrant further investigation. By activating heterologous signaling cascades, glucagon primarily activates PKA and downstream effectors, including glycogen phosphorylase for glycogenolysis and CREB for gluconeogenesis (9,37). Glucagon also minimally activates Akt phosphorylation, but its duration and intensity seemed insufficient for induction of Foxo1-S256 phosphorylation and degradation (Fig. 2). Thus, insulin is a major hormone for the endogenous Akt activation downstream of the IRS1- and 2-associated phosphoinositide 3-kinase, as we observed in the liver or heart of male mice (15,26,38). In this study, our in vitro data established that Foxo1-S276 is a target of PKA in glucagon signaling and promotes Foxo1 nuclear retention. Foxo1-S276D mutation mimics Foxo1-S276 phosphorylation and promotes Foxo1 nuclear localization and stability. Moreover, Foxo1-S273D mutation impairs the ability of insulin-stimulated Foxo1 nuclear export or cytoplasmic sequestration. This finding suggests a novel mechanism for insulin resistance when Foxo1-S273 is phosphorylated. Conversely, Foxo1-S276A mutation, which blocks phosphorylation, enhances Foxo1 ubiquitination and/or degradation, a process reversed by the 28S proteasome inhibitor MG132. This study reveals a unique mechanism of Foxo1 regulation by which glucagon stimulates HGP via Foxo1 at the posttranslational level.

Foxo1 can be regulated by multiple protein kinases (15). It has been shown that mitogen-activated protein kinase phosphorylates Foxo1 at S246, S284, S295, S326, S413, S415, S429, S467, S475, and T557, and p38α phosphorylates 5 of the 10 sites: S284, S295, S326, S467, and S475 (39). Moreover, these phosphorylation sites are shared by CaMKII, expression of a Foxo1-S7A mutation (S246A/S284A/S295A/S413A/S415A/S429A/S475A) attenuated glucagon-induced Foxo1 nuclear localization, and it is suggested that CaMKII promotes Foxo1 nuclear localization through p38α activation that stimulates hepatic gluconeogenesis (24,40). The Foxo1-S273 phosphorylation identified as a target of PKA is distinct from these sites, but whether it is involved by other protein kinases, such as p38α, remains to be determined.

HGP is generally regulated by direct mechanisms of insulin action in hepatocytes autonomously (41). Our early studies suggest that direct hepatocytes signaling via IRS1- and 2-mediated Akt activation or hepatic Foxo1 suppression is required for hepatic insulin suppression on HGP and blood glucose (26). However, a recent report (42) suggests that an indirect insulin that suppresses free fatty acid influx to hepatocytes from lipolysis of adipose tissue is critical for suppression of HGP so that hepatocyte Akt signaling is dispensable for HGP suppression. Indeed, HGP can be regulated by indirect insulin action in extrahepatic tissues nonautonomously (41). In addition, to suppress glucagon secretion from pancreatic α-cells or antagonize the effect of cAMP, insulin also suppresses HGP by the central nervous system in hypothalamic neurons via a vagal efferent (43). Thus, we do not rule out the possibility of other tissues including the central nervous system, adipose tissue, and pancreas on HGP control in the L-F1KO and KI mouse models, but our data from mouse primary hepatocytes do support that the cellular autonomous effect of glucagon in hepatocytes has critical roles in controlling Foxo1, gluconeogenesis, and HGP.

Foxo1 has a variety of target genes responsible for multiple physiological functions in the body (11,15,44,45). We have recently identified two other Foxo1 target genes in the liver, such as angiotensinogen, a precursor of angiotensin II that regulates blood pressure (44). Foxo1 also promotes expression of heme oxygenase-1, reducing synthesis of heme and mitochondria (46) and driving bodily metabolic inflammation and insulin resistance in mice and humans (47). In this study, we demonstrated a novel regulatory mechanism of Foxo1 by glucagon-cAMP–PKA signaling in hepatocytes, but we do not rule out the possibility of a similar Foxo1 regulation by other hormones via activation of PKA in tissues, including but not limited to glucagon (48). Indeed, hormones such as catecholamines or glucocorticoids also increase hepatic gluconeogenesis or promote insulin resistance under certain conditions involving activation of PKA, such as stress (49). Our studies demonstrate that Foxo1 is a mediator of multiple signaling cascades (e.g., PKA and Akt); it integrates different hormones and intracellular protein kinases into the nuclear gene transcriptional programming in control of insulin sensitivity, HGP, and blood glucose. Thus, Foxo1-S276 phosphorylation may not only serve as a novel and important biomarker for Foxo1 stability and bioactivity contributing to hyperglycemia upon development of diabetes, but also provides a potential therapeutic target in control of Foxo1 stability and activity to prevent diabetes and associated diseases in the future.

Acknowledgments. The authors thank Mike Honig, who provided English editing for the manuscript. S.G. also thanks the scientists at Covance (Denver, PA) and Cyagen Biosciences (Santa Clara, CA) for help generating Foxo1-phosphospecific antibodies and KI mice, respectively.

Funding. This study was supported by grants from the National Institutes of Health (RO1-DK-095118), American Diabetes Association Career Development Award (1-15-CD-09) and Minority Undergraduate Internship Award (1-17-MUI-008), the American Heart Association (BGIA-7880040), a Faculty Startup from the Texas A&M University Health Science Center and AgriLife Research, and the U.S. Department of Agriculture National Institute of Food and Agriculture (Hatch 1010958) to S.G. S.G. is the recipient of the 2015 American Diabetes Association Research Excellence Thomas R. Lee Award.

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

Author Contributions. Y.W., Q.P., H.Y., K.Z., X.G., Z.X., W.Y.,Y.Q., C.A.G., C.H., L.Z., A.Z., L.L., and Y.C. performed experiments. Y.W., Q.P., H.Y., and S.G. wrote the manuscript. L.Z. and A.Z. performed LC-MS and analyzed the data. W.Z., Y.S., H.Z., F.W., and L.H. analyzed and interpreted data. S.G. conceived of the hypothesis and designed the experiments. S.G. is the guarantor of this work and as such, had full access to all of the data in the study and takes responsibility for the integrity of the data and accuracy of the data analysis.

Prior Presentation. Parts of this study were presented at the 75th Scientific Sessions of the American Diabetes Association, Boston, MA, 5–9 June 2015 and the 78th Scientific Sessions of the American Diabetes Association, Orlando, FL, 22–26 June 2018.