Estrogen Improves Insulin Sensitivity and Suppresses Gluconeogenesis via the Transcription Factor Foxo1

The prevalence of type 2 diabetes (T2D) has shown sex disparities, with a reduced incidence in women (1). Both clinical and animal studies show a strong correlation between estrogen deficiency and metabolic dysfunction (2,3). The reduction of estrogen in postmenopausal women accelerates the development of insulin resistance and T2D (4). Clinical trials of estrogen replacement therapy in postmenopausal women demonstrated an amelioration of insulin resistance and reductions of plasma glucose level and incidence of T2D (5,6). However, the potential risk of breast cancer and stroke upon estrogen therapy underscores 17β-estradiol (E₂) as a therapeutic agent (7,8). Thus, understanding of tissue-specific E₂ action and its molecular mechanism in metabolic regulation is required for the development of targeted estrogen mimics conveying metabolic benefit without side effects.

The maintenance of glucose homeostasis depends on glucose uptake in muscle and adipose tissue and glucose production through gluconeogenesis and glycogenolysis in liver (9). Estrogen reduces glucose level, but it is thought to be derived from the promotion of glucose uptake in muscle (10). However, some studies have also indicated that estrogen suppresses glucose production in the liver (11–13) and that impairing the estrogen receptor (ER)-mediated signaling results in hepatic insulin resistance and hyperglycemia with elevated hepatic glucose production (HGP) (14,15). These lines of evidence underscore the importance of estrogen signaling in the liver in control of glucose homeostasis; yet, the liver-specific action of estrogen on HGP and its mechanism remain unclear.

Foxo1, a member of O-class forkhead transcription factor, is a predominant regulator for HGP, promoting gene transcription of the rate-limiting gluconeogenic enzyme glucose-6-phosphatase (G6pc) (16,17). Foxo1 is suppressed by insulin through the activation of Akt via the insulin receptor substrate (Irs)1 and Irs2-associated phosphoinositide 3-kinase (PI3K) (18,19). In a T2D mouse model, insulin fails to activate Akt and suppress Foxo1, resulting in enhanced HGP and hyperglycemia; this can be prevented by deletion of hepatic Foxo1 (20). In this study, we investigated the effect of E₂ on the maintenance of glucose homeostasis and the involvement of hepatic Foxo1 in mice of both sexes. We further examined E₂ regulatory mechanisms for HGP and gluconeogenesis in primary hepatocytes. In particular, we tested the hypothesis that hepatic Foxo1 plays important roles in the E₂ action of suppressing hepatic gluconeogenesis through ER-associated Akt signaling.

Animal protocols were approved by the institutional animal care and use committee at Texas A&M University. Liver-specific Foxo1 knockout (L-F1KO) mice were generated by breeding floxed Foxo1^L/L mice with albumin-Cre mice as previously described (21). The Cre recombinase induced by the mouse albumin enhancer/promoter shows no influence on animal performance, as previously reported, and either the floxed-Foxo1^L/L or albumin-Cre littermates were used as control mice in our studies (22,23). Eight- to 12-week-old control (Foxo1^L/L) and L-F1KO (albumin-Cre^+/−::Foxo1^L/L) mice on a mixed genetic background of C57BL/6J and 129/Sv were fed a standard chow diet (54% calories from carbohydrate, 14% from fat, and 32% from protein and 3.0 kcal/g) (Envigo Teklad Diet) ad libitum.

Control and L-F1KO mice were randomly assigned to experimental groups (n = 4–7). All mice were subjected to the subcutaneous implantation of placebo or E₂ pellet (0.05 mg/pellet, 60-day release) (Innovative Research of America, Sarasota, FL). Female mice underwent a bilateral ovariectomy (OVX) surgery (except for intact control mice) at the time of pellet implantation.

Blood glucose levels were measured using a glucometer (Bayer, Whippany, NJ). Glucose tolerance tests (GTTs) and pyruvate tolerance tests (PTTs) were performed in mice fasted for 16 h. Insulin tolerance tests (ITTs) were conducted in mice fasted for 5 h. Serum samples were collected from mice fasted for 16 h by cardiac puncture for determination of serum insulin, glucagon, and other hormone levels by Bio-Plex mouse diabetes immunoassay (Bio-Rad, Hercules, CA) and serum E₂ concentration by an ELISA kit (Cayman, Ann Arbor, MI). Liver glycogen content was measured from mice fasted 16 h as previously described (24).

Primary hepatocytes were isolated from 8- to 16-week-old mice and cultured in DMEM with 10% FBS as previously described (21). Cells were serum starved for 6 h before treatment with 100 nmol/L E₂ and a variety of signal transduction inhibitors, which were applied 30 min prior to E₂. For HGP assay, freshly isolated hepatocytes were cultured in DMEM with 2% FBS. After 3 h of attachment, cells was cultured in HGP buffer (120 mmol/L NaCl, 5.0 mmol/L KCl, 2.0 mmol/L CaCl₂, 25 mmol/L NaHCO₃, 2.5 mmol/L KH₂PO₄, 2.5 mmol/L MgSO₄, 10 mmol/L HEPES, 0.5% BSA, 10 mmol/L sodium dl-lactate, and 5 mmol/L pyruvate, pH 7.4) and treated with chemicals. Culture medium was collected to determine glucose production using an Amplex Red Glucose Assay kit (Invitrogen, Carlsbad, CA). Total HGP, glycogenolysis, and gluconeogenesis were determined and calculated as previously described (25).

RNA was extracted with Trizol reagent (Invitrogen) and used for cDNA synthesis via an iScript cDNA Synthesis kit (Bio-Rad). Gene expression was measured with SYBR Green Supermix in real-time PCR (Bio-Rad) using primers as previously described and cyclophilin gene as an internal control (21).

An equal amount of protein was resolved in SDS-PAGE and transferred to nitrocellulose membrane for Western blot. Antibodies for Foxo1, phosphorylated Foxo1 at Ser²⁵³, Akt, phosphorylated Akt at Ser⁴⁷³, Pck1, and GAPDH were purchased from Cell Signaling Technology. Protein densitometry was performed and analyzed using ImageJ as previously described (21).

Data were analyzed by one-way or two-way ANOVA to determine the significance of the model as appropriate. The interaction between E₂ and L-F1KO was analyzed in two-way ANOVA to determine the dependency. Differences between groups were determined by Tukey post hoc test or Student t test as appropriate. P < 0.05 was considered indicative of statistical significance. Results are presented as means ± SEM.

We determined the role of estrogen in control of glucose homeostasis and its dependence on hepatic Foxo1. Generally, intact female mice exhibited a 16% lower fasting glucose level compared with male mice (intact female 51 ± 2.8 mg/dL vs. male: 61 ± 2.3 mg/dL, P = 0.03) (Fig. 1A and C). In female mice, OVX mice had a 22% increase in fasting glucose compared with intact females (intact 51 ± 2.8 mg/dL vs. OVX 62.4 ± 2.2 mg/dL, P < 0.01) (Fig. 1A). E₂ reduced glucose levels by 21% in OVX control females (OVX + placebo 62.4 ± 2.2 mg/dL vs. OVX + E₂ 49.4 ± 1.2 mg/dL, P < 0.001) and almost restored glucose levels to those of intact mice (Fig. 1A and B). In control male mice, E₂ decreased glucose levels by 16% (placebo 61.0 ± 2.3 mg/dL vs. E₂ 51.0 ± 2.0 mg/dL, P = 0.03) (Fig. 1C). Deficiency of hepatic Foxo1 decreased fasting glucose by 19% in OVX females and by 23% in males (Fig. 1B and C). In L-F1KO mice, E₂ led to a further 17% reduction of glucose levels in OVX females but failed to reduce glucose levels in males (Fig. 1B and C). These results indicate that hepatic Foxo1 is required for E₂ to regulate glucose homeostasis in males rather than females.

To explain the sex disparity in glucose levels of L-F1KO mice in response to E₂, we measured fasting glycogen content in the liver. Intact female mice showed a sevenfold higher liver glycogen than that of males, indicating a higher capacity of females to store glycogen and/or prevent glycogenolysis in the liver than males (Fig. 1D and F). In control females, OVX reduced liver glycogen by 80% (P < 0.01), but this effect of OVX was reversed by E₂ (P < 0.01) (Fig. 1D and E). L-F1KO increased liver glycogen by 2.2-fold in OVX females, and E₂ further increased liver glycogen by 2.5-fold in OVX L-F1KO females (Fig. 1E); however, the effect of L-F1KO is masked by E₂ when E₂ is abundantly available. Collectively, the sex differences in liver glycogen may help explain the sex disparity in glucose homeostasis in response to E₂.

Compared with intact female mice, OVX increased body weight by 40% (Fig. 1G), while E₂ reduced body weight of OVX females by 14% in both control and L-F1KO mice (Fig. 1H). Body weight was unaffected by E₂ in male mice (Fig. 1I). Deletion of hepatic Foxo1 had no effect on body weight in both sexes (Fig. 1G–I). OVX mice exhibited a 44% lower serum E₂ level compared with intact females owing to absence of ovarian estrogen production (Fig. 1J). E₂ implant significantly increased serum E₂ levels of control and L-F1KO mice in both sexes (P < 0.05) (Fig. 1K and L). Interestingly, in males, serum E₂ concentration was undetectable in control mice, but L-F1KO males exhibited a detectable serum E₂ value at 1.56 ± 0.53 pg/mL (Fig. 1L).

We next examined the role of hepatic Foxo1 in E₂ action on insulin sensitivity and gluconeogenesis in mice. Surgical removal of the ovaries impaired glucose clearance after intraperitoneal injection of glucose (Fig. 2A and D). Either E₂ implant or deletion of liver Foxo1 resulted in an improved glucose tolerance in both sexes, but there were no additive effects of these two in L-F1KO mice (Fig. 2B–F). We examined whether estrogen-lowered fasting glucose is derived from suppression of gluconeogenesis, as determined by PTT. OVX had impaired pyruvate tolerance compared with control intact females (P < 0.05) (Fig. 2G and J), while E₂ suppressed gluconeogenesis in both sexes of control mice (P < 0.05) (Fig. 2H–L). L-F1KO mice displayed a decrease in gluconeogenesis in both sexes (P < 0.05); however, the effect of E₂ was abolished in L-F1KO mice (Fig. 2H–L). OVX impaired insulin sensitivity as determined by ITT (Fig. 2M and P). E₂ and L-F1KO improved ITT in both sexes but had no additive effect (Fig. 2N–R). Taken together, E₂ improves insulin sensitivity and suppresses gluconeogenesis in a Foxo1-dependent manner.

We measured mRNA levels of genes responsible for insulin signaling and gluconeogenesis. The OVX surgery suppressed hepatic Irs2 mRNA expression by 40% and induced G6pc and phosphoenolpyruvate carboxykinase 1 (Pck1) expressions by 22% and 42%, respectively, in control females (P < 0.05); however, E₂ induced Irs2 by 35% and suppressed G6pc and Pck1 by 22% and 30% in OVX control females (Fig. 3A and B). E₂ also induced Irs2 by 50% and suppressed G6pc by 24% in male mice. The liver of L-F1KO mice displayed a reduction of Irs1 expression by 45% in OVX females and 20% in males independent of E₂ treatment (Fig. 3B and C). Hepatic expression of G6pc was reduced by 46–48% in both sexes upon deletion of liver Foxo1 (P < 0.05) (Fig. 3B and C). However, E₂ failed to have the inhibitory effect on G6pc and Pck1 expressions in both sexes of L-F1KO mice (Fig. 3B and C).

Levels of metabolic hormones were determined in the serum from mice fasted for 16 h. OVX mice exhibited increases in leptin and plasminogen activator inhibitor 1 levels by 3.0- and 1.3-fold, respectively, compared with intact female mice, while E₂ implant decreased levels of leptin and plasminogen activator inhibitor 1 by 56% and 30% in OVX mice (P < 0.05) (Fig. 3D and E). In male mice, deletion of liver Foxo1 increased glucagon and ghrelin levels by 5.1- and 2.6-fold, respectively (P < 0.05) (Fig. 3F). E₂ increased ghrelin level by 1.8-fold in control males (P < 0.05) (Fig. 3F). Although sex differences exist in the release of metabolic hormones, the increased glucagon and ghrelin levels, as well as decreased leptin level, may be a result of lower blood glucose upon E₂ stimulation and ablation of Foxo1, which suggests a feedback regulation in glucose homeostasis.

We next assessed E₂ action on hepatic glucose metabolism in primary hepatocytes. L-F1KO hepatocytes exhibited a 32% lower HGP after 1 h culture in assay buffer compared with control cells (P < 0.05) (Fig. 4A). E₂ decreased HGP by 14% and 20% after 3- and 6-h culture in control cells, respectively (P < 0.05), but failed to suppress HGP in L-F1KO cells (Fig. 4A). We further divided HGP into gluconeogenesis and glycogenolysis in response to E₂. Although E₂ decreased glycogenolysis by 25% in control hepatocytes (P < 0.05), the contribution of glycogen breakdown to total HGP was much smaller than that of gluconeogenesis (Fig. 4B). Thus, 18% reduction of gluconeogenesis by E₂ mainly contributed to suppression of HGP in control cells (P < 0.05) (Fig. 4B). Consistently, E₂ reduced expressions of G6pc by 25% and Pck1 by 30% in control hepatocytes (P < 0.05) (Fig. 4C). L-F1KO hepatocytes showed reductions of G6pc and Pck1 expressions by 55% compared with control cells (P < 0.05); however, E₂ failed to further suppress these genes’ expression in L-F1KO cells (Fig. 4C).

Akt phosphorylates Foxo1 at Ser²⁵³, thus promoting Foxo1 nuclear export and/or degradation (26). E₂ increased phosphorylation of Akt and Foxo1 by 1.7- and 1.5-fold, respectively. The total Foxo1 degradation was unnoticeable by E₂ treatment alone but evident by E₂ and insulin cotreatments (Fig. 4D and E). To confirm the importance of phosphorylation of Foxo1 at Ser²⁵³ for E₂ action, we generated Foxo1 Ser²⁵³ knock-in mice (S253A), in which serine 253 was replaced with alanine to prevent Akt phosphorylation (K. Zhang, X. Guo, H. Yan, Y. Wu, , Q. Pan, Z. Shen, X. Li, Y. Chen, L. Li, Y. Qi, Z. Xu, W. Xie, W. Zhang, D. Threadgill, L. He, Y. Sun, M. F. White, H. Zheng, and S. Guo, unpublished observations). Hepatocytes from S253A mice exhibited increases in HGP by 31% and gluconeogenesis by 36% owing to the blockage of Foxo1 phosphorylation at Ser²⁵³ (P < 0.05); meanwhile, E₂ failed to reduce HGP in S253A hepatocytes (Fig. 4F). These data suggested that phosphorylation of Foxo1 Ser²⁵³ by Akt is required for E₂ in the regulation of Foxo1 activity and hepatic gluconeogenesis.

We next determined whether activation of Akt signaling downstream from PI3K is necessary for E₂ in suppression of Foxo1 and gluconeogenesis in hepatocytes using Akt1/2 kinase inhibitor (Akti) or PI3K inhibitor wortmannin (Wort). Both Akti and Wort completely prevented E₂ effects on HGP, glycogenolysis, and gluconeogenesis (Fig. 5A), as well as E₂-suppressed G6pc and Pck1 mRNA expression and Pck1 protein levels (Fig. 5B and C). Meanwhile, E₂-induced reduction of Pck1 protein level and phosphorylation of both Akt and Foxo1 were also blocked by these two inhibitors (Fig. 5C and D). Additionally, Akti alone showed increases in basal levels of G6pc and Pck1 but no significant effect on HGP or gluconeogenesis in the primary hepatocytes (Fig. 5A–D).

Moreover, we further determined whether Irs1 and Irs2 are required for E₂-mediated glucose homeostasis using liver-specific Irs1 and Irs2 double knockout (L-DKO) mice as we previously characterized (27). In a random-fed state, L-DKO male mice displayed hyperglycemia with 158% increased glucose level (control 124.7 ± 4.38 mg/dL vs. L-DKO 321.50 ± 51.73 mg/dL, P = 0.01); however, L-DKO females exhibited euglycemia (control 119.9 ± 4.37 mg/dL vs. L-DKO 135.0 ± 6.93 mg/dL, P = 0.2) (Fig. 5E). In the fasting state, L-DKO males had increased glucose level by 62% (control 54.2 ± 1.9 mg/dL vs. L-DKO 87.6 ± 5.5 mg/dL, P < 0.01), while L-DKO female mice only had increased glucose level by 29% (control 44.8 ± 1.65 mg/dL vs. L-DKO 57.8 ± 2.63 mg/dL, P < 0.01) (Fig. 5E). Significantly, L-DKO females developed hyperglycemia 2 weeks after OVX, with 56% increased glucose, in the random-fed state (intact L-DKO 135.0 ± 6.9 mg/dL vs. OVX L-DKO 210.3 ± 13.5 mg/dL, P < 0.01), while OVX only increased the glucose level by 21% in control females (intact control 119.9 ± 4.4 mg/dL vs. OVX control 136.2 ± 5.9 mg/dL, P = 0.04) (Fig. 5E). Collectively, for OVX females, random-fed L-DKO mice had a 54% increase in glucose level (control 136.2 ± 5.91 mg/dL vs. L-DKO 210.3 ± 13.5 mg/dL, P = 0.04) and a 43% increase in fasting glucose (control 60.4 ± 5.36 mg/dL vs. L-DKO 86.0 ± 4.89 mg/dL, P = 0.03) compared with control mice (Fig. 5E). Moreover, subcutaneous implantation of E₂ significantly decreased blood glucose of both control and L-DKO OVX mice in both fed and fasting states and totally abolished the effect of L-DKO on glucose level (Fig. 5E). Additionally, in in vitro HGP assay, L-DKO primary hepatocytes exhibited increases in HGP, glycogenolysis, and gluconeogenesis by 65%, 27%, and 71%, respectively (P < 0.05). Moreover, E₂ inhibited HGP and gluconeogenesis by 17% and 20%, respectively, in L-DKO cells (Fig. 5F). These results suggested that ovarian hormone protects against hyperglycemia induced by loss of hepatic insulin function and that E₂-mediated suppression of hepatic gluconeogenesis requires the activation of PI3K-Akt-Foxo1 signaling but in an Irs1- and Irs2-independent manner.

ERα is the major form of ERs in the liver (28). We next investigated whether ERα modulates E₂ effect on suppression of HGP in hepatocytes using its specific agonist and antagonist. ERα-selective antagonist methylpiperidinopyrazole (MPP) completely blocked E₂-reduced HGP, gluconeogenesis (Fig. 6A), and expression of G6pc and Pck1 (Fig. 6B), as well as phosphorylation of Akt and Foxo1 (Fig. 6C). Moreover, MPP alone reduced basal phosphorylation of Akt and Foxo1 by 30% and 55%, respectively (P < 0.05) (Fig. 6C); this resulted in increases in G6pc and Pck1 levels by 68% and 25% (P < 0.05) (Fig. 6B). ERα agonist propylpyrazoletriol (PPT) showed an effect similar to that of E₂ to reduce HGP and gluconeogenesis in both control (Fig. 6A) and L-DKO (Fig. 5E) hepatocytes. Meanwhile, PPT enhanced phosphorylation levels of Akt and Foxo1 by 2.7-fold and 1.8-fold, respectively (P < 0.05) and suppressed expression of G6pc by 30% in control hepatocytes (P < 0.05) (Fig. 6B and C).

We finally determined whether overexpression of ERα enhances insulin- or E₂-induced Akt and Foxo1 phosphorylation in HepG2 cells, human hepatoma cells. Insulin and E₂ increased Akt phosphorylation by 10-fold and 5.4-fold, respectively, and induced Foxo1 phosphorylation by 2.9-fold and 2.8-fold (P < 0.05). Overexpression of ERα enhanced basal phosphorylation of Akt and Foxo1 by 3.4-fold and 2.0-fold. Strikingly, ERα overexpression further enhanced insulin- and E₂-stimulated Akt phosphorylation by 67% and 42% and Foxo1 phosphorylation by 1.3- and 1.2-fold (P < 0.05) (Fig. 6D).

Taken together, these results indicate that ERα is necessary and sufficient for E₂ to activate Akt-Foxo1 signaling and suppress gluconeogenesis.

Estrogen has been shown to be involved in both central and peripheral regulation in glucose homeostasis (29,30). Estrogen deficiency or impaired estrogen signaling is associated with insulin resistance and dysregulation of metabolic homeostasis, thus contributing to the development of T2D and obesity in both human and animal models (31–33). However, the contribution of tissue-specific action of estrogen to metabolic changes and underlying mechanisms have yet to be elucidated. In this study, we have established that activation of ERα-Akt-Foxo1 signaling is an important mechanism for estrogen in the liver in maintaining glucose homeostasis.

In general, glucose homeostasis is maintained by glucose uptake in muscle and adipose tissue and endogenous glucose production in the liver. Previous studies reported that estrogen lowers glucose level, which is associated with enhanced glucose uptake in muscle through activation of Akt and induction of GLUT4 expression (10,34). However, blockage of E₂ signaling in global or liver ERα knockout mice causes hepatic insulin resistance and enhanced HGP, resulting in hyperglycemia (14,35). In this study, E₂ improved glucose tolerance in both sexes; however, such an effect was blocked upon deletion of hepatic Foxo1. Our results supported that improvement of glucose homeostasis by E₂ is regulated by hepatic Foxo1-mediated gluconeogenesis rather than by promoting muscle glucose uptake (12). In liver, Foxo1 promotes gluconeogenesis through activating transcription of G6pc and Pck1 by directly binding to insulin response element CAAAACAA on their promoter regions (26,36). Liver-specific ablation of Foxo1 impairs gluconeogenesis and reduces fasting glucose level (37). Deletion of hepatic Foxo1 protects diabetic mice, such as L-DKO mice or db/db mice, from hepatic insulin resistance (23,27). In this study, suppression of gluconeogenesis by E₂ was abolished upon deletion of Foxo1; this supported a notion that hepatic Foxo1 is required for E₂ action in the modulation of gluconeogenesis.

This study illustrated that E₂ inhibits Foxo1 transactivation via the interaction with ERα and activation of PI3K-Akt signaling, thus serving as a molecular basis for E₂ action on suppression of hepatic gluconeogenesis. ERα is involved in glucose homeostasis, and impaired ERα function is associated with insulin resistance, T2D, and metabolic syndrome in both human and animal models (14,15,35). Our results demonstrated that activation of ERα is required for E₂ to suppress HGP and that overexpression of ERα enhances insulin- and E₂-induced phosphorylation of Foxo1. Ambiguity about how ERα modulates insulin signaling and glucose metabolism still remains. A recent study indicated that an estrogen response element half-site (AGGTCA) presents in promoter regions of G6pc and Pck1 (38). Recruitment of ERα to those promoters upon E₂ activation inhibits transcriptions of G6pc and Pck1 (35), but Yasrebi et al. (39) reported that E₂ still decreased fasting glucose and improved glucose tolerance in the ERα knock-in mice, in which the knock-in mutation blocks the functional estrogen response element binding activity for DNA, ruling out this genomic signaling of ERα. An E₂-dependent binding of ERα with Foxo1 was reported to suppress Foxo1 transactivation in the MCF7 cell line (40). Inhibition of Akt signaling by inhibitors or by S253A mutant of the Akt phosphorylation site in Foxo1 totally blocked effects of estrogen on HGP; this suggested the importance of indirect E₂ action on Foxo1 Ser²⁵³ phosphorylation via ERα-PI3K-Akt rather than a direct interaction between ERα and Foxo1. Direct interactions of ERα with several proteins in the insulin signaling cascade have been reported. ERα binds to Irs1 and Irs2 and promotes stability of the ERα/Irs complex in breast cancer lines (41,42); it also binds to the p85α catalytic subunit of PI3K in human embryonic kidney (HEK)293T, MCF7, and aortic endothelial cells (43,44). A direct interaction between p85α and ERα in proopiomelanocortin (POMC) progenitor neurons has been reported, and deletion of PI3K in POMC neurons attenuates E₂-reduced glucose level in OVX females (45). However, the question of whether these protein interactions contribute to hepatic glucose metabolism remains. In this study, L-DKO intact females were protected from hyperglycemia and deletion of Irs1 and Irs2 did not disrupt E₂-suppressed fasting glucose level in vivo or hepatic gluconeogenesis in vitro, whereas PI3K inhibitor totally blocked the E₂ action; this suggested that the interaction of ERα with PI3K may contribute to activation of Akt-Foxo1 and regulation of glucose homeostasis independent of Irs1 and Irs2. The phosphorylation sites of Foxo1 by Akt are also identified in other Foxo members, including Foxo3 and Foxo4. Our previous study indicated that Foxo1, rather than Foxo3 and -4, significantly influences glucose homeostasis (21). Even if E₂ modifies the activity of other Foxo isoforms via Akt, the contribution of Foxo3 and Foxo4 by E₂ to glucose metabolism is limited.

E₂ exhibits a Foxo1-independent mechanism to reduce glucose level only in females. OVX decreased hepatic glycogen (46), which was restored by E₂, as previously reported (11,12,47,48). E₂ increased glycogen independent of hepatic Foxo1, which may account for the reduced glucose level in L-F1KO females. Hepatic glycogen is regulated by glycogen synthase kinase 3α (GSK3α), which phosphorylates glycogen synthase and inhibits glycogen synthesis (49). GSK3α is inactivated by Akt-mediated phosphorylation, and loss of function of GSK3α is associated with increased hepatic glycogen and lowered glucose levels (50). Thus, we speculate that activation of Akt and inhibition of GSK3α to promote glycogen synthesis by E₂ may be involved in a Foxo1-independent mechanism for E₂ to reduce glucose level in females. In fact, a study reported that hepatic Foxo1 ablation is associated with an increase in liver glycogen and inhibition of glycogenolysis in hepatocytes (37); another study indicated that Foxo1 has limited effects on hepatic glycogen (20). Nevertheless, we found that deficiency of hepatic Foxo1 increased liver glycogen only in OVX females, which is masked by E₂ implant. That deletion of hepatic Foxo1 increased liver glycogen might be the result of inhibition of gluconeogenesis and accumulation of glucose-6-phosphate, a substrate for glycogen synthesis.

Sex differences exist in liver glycogen levels and responses to E₂. Male mice exhibited an 85% reduction in liver glycogen compared with intact females, and E₂-increased hepatic glycogen levels as observed in OVX females disappeared in male mice. Several studies have indicated that long-term exercise training increases glycogen storage in both muscle and liver with enhanced glycogen synthase activity (51). Female mice naturally exhibited a higher level of physical activity than male mice (52); we suspect that such a higher physical activity may contribute to a greater capacity to store glycogen. This also implies a sexual dimorphism in the contribution of glycogen metabolism to glucose homeostasis.

E₂ suppressed body weight in OVX females independent of hepatic Foxo1 but reduced leptin level dependent on Foxo1, whereas E₂ was less effective in reduction of body weight and leptin in males. Our previous studies showed that estrogen increases energy expenditure and decreases body weight through binding to its receptors in the central nervous system (CNS) (53). In the OVX female mice, obesity developed and E₂ administration reduced the body weight and serum leptin level, suggesting E₂ increases the leptin sensitivity, which is likely via increasing expression of Irs2, a modulator of leptin action in the CNS (54). Leptin level is highly associated with the amount of fat mass. E₂-decreased fat mass accumulation through CNS may explain the decreased leptin level. However, E₂ administration reduced the body weight in L-F1KO mice; the effect can be independent of leptin level but dependent on leptin sensitivity with an increase of Irs2 expression and energy expenditure (54). The sex differences may result from differences in expression of ER isoforms in various tissues (55), absorbance rates of exogenous E₂ and local E₂ concentrations, and capacity of E₂ to cross the blood-brain barrier to CNS.

Estrogen is critical to maintaining metabolic homeostasis in both sexes (31). In this study, E₂ implant improved insulin sensitivity and lowered fasting glucose in both sexes. In premenopausal women, E₂ is converted from androstenedione by aromatization in the ovaries and is released as circulating hormone acting on distant tissues. In men, E₂ is produced locally in extragonadal tissues by aromatization from testosterone (31,56). Patients lacking aromatase function and aromatase-deficient mice (ArKO mice), with endogenous estradiol synthesis diminished in both cases, exhibit impaired glucose homeostasis (3,57). Men lacking aromatase also exhibit hyperinsulinemia (58), and male ArKO mice develop insulin resistance (57); this suggests that E₂ signaling is similarly crucial in metabolic homeostasis in both sexes. Of note, L-F1KO males had a detectable serum E₂ level compared with the undetectable serum E₂ in control males. Given that deletion of hepatic Foxo1 increases cholesterol levels (59) and cholesterol is a precursor for steroid hormone biosynthesis, we speculate that hepatic Foxo1 may also be involved in estrogen biosynthesis indirectly through the regulation of cholesterol production in the liver.

In summary, our study demonstrated that E₂ improves insulin sensitivity and suppresses hepatic gluconeogenesis through inhibition of Foxo1 via activation of ERα-PI3K-Akt signaling (Fig. 7). PI3K is required for E₂ action, but comprehensive studies into the direct interaction of ERα with PI3K regulating hepatic glucose metabolism have not been conducted. Since estrogen promotes the development of the reproductive system, the metabolism and reproductive biology can be regulated by different sets of target genes. Therefore, the identification of tissue-specific actions of E₂ and direct targets of ERs will facilitate the development of novel selective ligands that prevent T2D, cardiovascular disease, and obesity without promoting abnormal sex characteristics or breast cancer.

Acknowledgments. The authors thank Jennifer DeLuca (Department of Nutrition and Food Science, College of Agriculture and Life Sciences, Texas A&M University) for technical assistance for the OVX surgery.

Funding. This work was supported by National Institutes of Health, National Institute of Diabetes and Digestive and Kidney Diseases, grants (RO1-DK-095118 and R56-DK-118334-01), American Diabetes Association Career Development Award 1-15-CD-09, an American Heart Association grant (BGIA-7880040), Faculty Start-up funds from Texas A&M University Health Science Center and AgriLife Research, and a U.S. Department of Agriculture National Institute of Food and Agriculture grant (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. H.Y., W.Y., and F.Z. designed and conducted experiments and performed data analyses. H.Y. and S.G. wrote the manuscript. X.L., Q.P., Z.S., and K.A. conducted experiments. G.H., A.N.-F., Y.T., R.M., W.L., Y.X., C.W., C.A., and Y.S. reviewed and edited the manuscript. S.G. supervised the project, conceived of the hypothesis, and designed experiments. S.G. 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.

Prior Presentation. Parts of this study were presented in abstract form at the 78th Scientific Sessions of the American Diabetes Association, Orlando, FL, 22–26 June 2018.