Nonsteroidal anti-inflammatory drugs (NSAIDs), including acetylsalicylic acid (ASA), improve glucose metabolism in diabetic subjects, although the underlying mechanisms remain unclear. In this study, we observed dysregulated expression of cyclooxygenase-2, prostacyclin biosynthesis, and the I prostanoid receptor (IP) in the liver’s response to diabetic stresses. High doses of ASA reduced hepatic prostaglandin generation and suppressed hepatic gluconeogenesis in mice during fasting, and the hypoglycemic effect of ASA could be restored by IP agonist treatment. IP deficiency inhibited starvation-induced hepatic gluconeogenesis, thus inhibiting the progression of diabetes, whereas hepatic overexpression of IP increased gluconeogenesis. IP deletion depressed cAMP-dependent CREB phosphorylation and elevated AKT phosphorylation by suppressing PI3K-γ/PKC-ζ–mediated TRB3 expression, which subsequently downregulated the gluconeogenic genes for glucose-6-phosphatase (G6Pase) and phosphoenol pyruvate carboxykinase 1 in hepatocytes. We therefore conclude that suppression of IP modulation of hepatic gluconeogenesis through the PKA/CREB and PI3K-γ/PKC-ζ/TRB3/AKT pathways contributes to the effects of NSAIDs in diabetes.
Introduction
Glucose is the major source of energy required by most mammalian cells to maintain normal physiological functions. Glucose homeostasis is tightly regulated within a relatively narrow range by hormones such as glucagon and insulin and through balancing of glucose output by the liver and its utilization by peripheral tissues such as skeletal muscle, heart, and adipocytes. Liver is the dominant organ in the maintenance of glucose homeostasis, which is regulated by way of glucose production through glycogenolysis and gluconeogenesis, glucose uptake by glycogenesis, and glycolytic conversion to pyruvate (1). Circulating insulin increases in response to feeding, leading to glycogenesis and lipogenesis and the suppression of hepatic glucose production (HGP). Conversely, during fasting conditions or in the case of untreated type 1 diabetes, insulin secretion drops and glucagon secretion rises, prompting hepatic glycogenolysis and gluconeogenesis. The key regulatory enzymes for hepatic gluconeogenesis include glucose 6 phosphatase (G6Pase), fructose-1, 6-bisphosphatase, and phosphoenolpyruvate carboxykinase 1 (PEPCK), also known as PCK1. However, in patients with type 2 diabetes, the rate of hepatic gluconeogenesis is considerably elevated, contributing to fasting hyperglycemia and to exaggerated postprandial hyperglycemia.
Prostaglandins (PGs) play important roles in inflammation-mediated diseases, including diabetes (2). Elevated PGs have been observed in type 1 and type 2 diabetes (3,4). Epidemiological studies have indicated that the use of nonselective nonsteroidal anti-inflammatory drugs (NSAIDs), including acetylsalicylic acid (ASA), is associated with a significant reduction in the risk of diabetes in healthy populations (5). Selective cyclooxygenase (COX)-2 inhibitors have also been reported to increase insulin sensitivity in healthy individuals (6) and to ameliorate diabetes in experimental animals (7). These observations strongly suggest that COX-derived PGs are involved in the pathogenesis of diabetes. Moreover, clinical trials have revealed that treatment with ASA (8) results in the reduction of fasting plasma glucose and improves insulin sensitivity, whereas high doses of selective COX-2 inhibitors have been reported to cause hypoglycemia (9) and increase the hypoglycemic effect of oral antidiabetic drugs (10). These observations raise the possibility that PGs play a role in carbohydrate metabolism, especially in hepatic gluconeogenesis, the predominant source of increased hepatic glucose in type 2 diabetes.
Prostanoids in liver are produced by parenchymal hepatocytes (11) and nonhepatocyte cells, such as Kupffer cells (12), and their biosynthesis and release can be regulated in response to a range of (patho) physiological stimuli to modulate hepatocyte function. In isolated rodent livers, infusion of PGF2α but not thromboxane A2 stimulates gluconeogenesis and glycogenolysis, and PGD2 induces hepatic glycogenolysis (13), whereas PGE2 inhibits glucagon-mediated gluconeogenesis from lactate (14). However, the relevance of these infusion experiments to the autocoidal role and concentrations of endogenous eicosanoids is unclear, and the potential importance of PGI2 in regulating glucose metabolism in liver is unknown.
In this study, we observed upregulation of the COX-2/PGI2/I prostanoid receptor (IP) axis in the livers of fasted mice or mice treated with a high-fat diet (HFD), or mice in which diabetes had been induced genetically or pharmacologically. Deletion of the IP conferred protection against diabetes in mice due to the suppression of hepatic gluconeogenesis. Conversely, reexpression of IP in the liver augmented hepatic glucose output and led to insulin resistance by enhancing intracellular adenylate cyclase activity, upregulating Tribble 3 (TRB3)-dependent AKT phosphorylation, and subsequently promoting transcription of the key hepatic gluconeogenic enzymes PCK1 and G6Pase. These data suggest that PGI2 is involved in modulation of hepatic gluconeogenesis through the IP.
Research Design and Methods
Mice
The mice used in this study were maintained at a C57BL/6 background. For the diet-induced obese model, mice were fed a regular chow diet (SLRC, Shanghai, People’s Republic of China) or an HFD (60% fat, D12492; Research Diets, New Brunswick, NJ) ad libitum for 16 weeks. For ASA treatment, 7-week-old ob/ob mice received normal drinking water or drinking water containing 600 mg/L (high-dose) ASA, respectively, which was replaced every other day. All procedures were approved by the Institution for Nutritional Sciences Institutional Animal Care and Use Committee, Shanghai Institutes for Biological Sciences, Chinese Academy of Sciences, China.
Metabolic Studies
Insulin, glucose, and pyruvate tolerance tests were performed by an intraperitoneal injection of 0.8 units/kg insulin after 6-h fasting, 1.5 g/kg glucose after overnight fasting, or 2 g/kg sodium pyruvate after overnight fasting, respectively. Glucose clearance was evaluated by calculating the areas under curve, as previously described (15). A glucagon challenge test was performed by an intraperitoneal injection of glucagon (150 μg/kg) after a 15-h fast (16). Measurements of Vo2 and the respiration exchange ratio (Vco2/Vo2) were acquired using indirect calorimetry, as previously described (17). Insulin and glucagon quantitation of murine plasma samples collected at 8–9 a.m. was done using the Mercodia Ultrasensitive Rat Insulin ELISA kit and the Glucagon ELISA kit (ALPCO Diagnostics, Salem, NH), respectively. Serum triglyceride (TG), total cholesterol, HDL cholesterol, and LDL cholesterol were measured using assay kits (BHKT, Beijing, People’s Republic of China). Tissue TG and total cholesterol were measured using assay kits (BHKT), as previously described (18).
Western Blotting
Proteins from total cell lysates were separated by SDS-PAGE and probed with different primary antibodies against AKT (ser473), AKT (ser308), AKT, insulin receptor (IR) substrate (IRS; Ser307), IRS, FoxO1 (Ser253), FoxO1, glycogen synthase kinase 3β (GSK3β) (ser9), GSK3β, phospho-p44/42 mitogen-activated protein kinase (Thr202/Tyr204), p44/42 mitogen-activated protein kinase, PDK1, and carboxy-terminal modulator protein (CTMP; Cell Signaling Technology, Danvers, MA); COX-1 and COX-2 (Cayman Chemical Company, Ann Arbor, MI); G6Pase and TRB3 (Santa Cruz Biotechnology, Santa Cruz, CA); PCK1 (Abcam); β-actin (Sigma-Aldrich, St. Louis, MO); and HK2 and peroxisome proliferator–activated receptor (PPAR)-α (Proteintech, Wuhan, People’s Republic of China).
Primary HGP
Primary hepatocytes were washed three times with PBS and then changed to glucose- and phenol-free DMEM with 20 mmol/L sodium lactate and 1 mmol/L sodium pyruvate for 6 h. Glucose levels in the culture were determined using a Glucose Assay kit (Sigma-Aldrich). Total protein was used for normalization. For adenovirus (Ad) experiments, cells were infected with various Ad, and HGP was quantitated after 48 h.
Cell Culture and Treatments
Primary hepatocytes were cultured in DMEM with 25 mmol/L glucose, 10% FBS, and 50 μg/mL penicillin and streptomycin at 37°C in 5% CO2 and 95% air. Insulin (100 nmol/L) from Sigma-Aldrich; glucagon (100 nmol/L), cicaprost (1 μmol/L), LY294002 (25 μmol/L), RO32-0432 (10 μmol/L), and CAY-10441 (1 μmol/L) from Cayman Chemical; and Akt inhibitor IV (1 μmol/L) from Calbiochem were used to treat hepatocytes as indicated.
Determination of Cellular cAMP Levels
Cellular cAMP levels were measured as previously described (19).
In Vivo Insulin Signaling Assay
For measurement of insulin signaling in liver, mice maintained on different diets were fasted 6 h before receiving an insulin injection. Mice were anesthetized, and a piece of liver was excised through an incision and snap frozen in liquid nitrogen as the untreated control. Within 4–5 min after injection via the portal vein with 10 units/kg human insulin (Eli Lilly, Indianapolis, IN), another piece of liver was snap frozen for subsequent protein extraction and Western blot analysis.
Construction of Adenoviral Vector Encoding IP
To construct an IP overexpressing adenoviral vector, mouse IP full-length cDNA was inserted into the pAdTrack-cytomegalovirus construct, which was then subcloned into pAdEasy-1 adenoviral backbone vector through homologous recombination in BJ5183. To package the Ad, the adenoviral DNA was linearized by PacI restriction enzyme and transfected into HEK293 cells using Lipofectamine 2000 (Invitrogen). After several rounds of propagation, recombinant Ad was purified by ultracentrifugation in cesium chloride gradient.
Isolation of Liver Kupffer Cells
PG Extraction
PG extraction from liver tissue or culture medium was routinely performed in the laboratory (22).
PCK1 Activity
PCK1 activity was examined as previously described (23).
Generation of Streptozotocin-Induced Diabetic Mice
Streptozotocin (STZ)-induced diabetic mice were produced by using previously described methods (24)
mRNA Quantification by RT-PCR
Total RNA from cultured hepatocytes or tissues was extracted using Trizol reagent (Invitrogen, Carlsbad, CA) and then treated with RNase-free DNase (Takara, Dalian, People’s Republic of China) at 37°C for 2 min to remove genomic DNA. The primers used for each target gene are summarized in Supplementary Table 1.
Statistical Analysis
The Student t test was used for all statistical analysis. Time course studies were analyzed by two-way ANOVA, followed by the Bonferroni posttest (GraphPad Prism 5 software). P < 0.05 was considered statistically significant. Data are represented as mean ± SEM.
Results
COX-2/PGI2/IP Axis Is Upregulated in Liver in Response to Fasting and Diabetic Stress
We first examined the expression of all prostanoid receptors in metabolic tissues from chow-fed mice (Supplementary Fig. 1). All of the prostanoid receptors were variously expressed in liver except FP. Of note, thromboxane receptor (TP) and IP were abundantly detected in liver, white adipose tissue (WAT), and skeletal muscle. In contrast to the dominant constitutive expression of COX-1, only COX-2 was upregulated by fasting or administration of the HFD to wild-type (WT) mice, and a similar induction was evident in ob/ob mice (Fig. 1A and B), with a corresponding induction of prostanoid formation (Fig. 1C). The most abundant product formed in the livers was PGI2 under physiological conditions. Further investigation of hepatocytes and Kupffer cell–enriched nonparenchymal cells (>80% by immunostaining) revealed that COX-2 expression was upregulated in response to COX-1 deficiency (Fig. 1D and F), suggesting that products of COX-1 contribute to the expression of COX-2, which is consistent with previous observations (25). However, the deletion of COX-1 or COX-2 reduced all prostanoids compared with WT controls (Fig. 1E and G). Interestingly, COX-2 expression and prostanoids production induced by high glucose concentrations appeared more robust in Kupffer cells than in hepatocytes (Fig. 1D–G). IP expression in liver was also elevated significantly in fasted, HFD-fed, and ob/ob mice (Fig. 1H and I); however, we did not observe any alterations of TP expression in metabolic organs in response to fasting and the HFD challenge (data not shown).
IP Deficiency and High Doses of ASA Reduced Fasting Blood Glucose Levels in Mice
Given that the hepatic PGI2/IP axis was activated in response to fasting and diabetic stresses, we first examined the effect of IP ablation on glucose metabolism in mice. IP deletion had no detectable effect on blood glucose levels in chow-fed mice but significantly decreased blood glucose during fasting as measured at 4 h and 8 h (Fig. 2A), and the hyperglycemic response to glucagon was markedly attenuated in IP-deficient mice (Fig. 2B). Interestingly, mRNA and protein expression of key gluconeogenic genes in liver, G6Pase and PCK1, were consistently reduced in IP knock-out (KO) mice compared with WT littermates, before and after fasting (Fig. 2C–E). PCK1 activity in liver was also impaired in IP KO mice in response to fasting (Fig. 2F). Moreover, we did not observe alterations of the key hepatic glycogenolytic gene for glycogen phosphorylase (Pygl) in IP KO mice (Fig. 2G), suggesting that the fasting hypoglycemia in the mutants resulted from a defect of hepatic gluconeogenesis, not glycogenolysis.
To further investigate whether IP disruption interferes with lipid metabolism and glycolysis in mice, we analyzed the blood lipid profiles and expression of enzymes involved in fatty acid synthesis, fatty acid oxidation, and glycolysis. No differences in blood cholesterol or TG were detected between IP KO and WT in fed or fasted states (Supplementary Fig. 2). In addition, the expression of key enzymes for de novo fatty acid synthesis, including ATP citrate lyase (ACL), acetyl-CoA carboxylase 1 (ACC1), and fatty acid synthase in livers, gastrocnemii, and WAT, were unaltered in IP KO mice (Supplementary Figs. 3A–C and 4A–C). Moreover, IP deletion did not influence hepatic fatty acid oxidation, muscular glycolysis, or glucose uptake by skeletal muscle or WAT in mice (Supplementary Figs. 5A–C and 6A–C). Taken together, these results suggest that IP is involved in hepatic gluconeogenesis in response to fasting.
High doses of ASA reportedly lower fasting blood glucose in type 2 diabetes (8,26). As expected, a marked reduction of PG formation in liver, including PGI2 (Supplementary Fig. 7A) and a significant depression in fasting blood glucose were observed in ob/ob mice treated with 600 mg/L ASA in drinking water (Supplementary Fig. 7B). Likewise, PCK1 and G6Pase expression in liver was also dramatically reduced by high doses of ASA, further suggesting that the hypoglycemic effect of ASA is due to suppression of hepatic gluconeogenesis (Supplementary Fig. 7C). Interestingly, the decline of blood glucose in ob/ob mice by ASA was restored completely by administration of the IP agonist cicaprost (Supplementary Fig. 7D), suggesting that the hypoglycemic effect of ASA might be at least partly mediated through inhibition of PGI2/IP signaling.
To directly determine the effect of IP deficiency on gluconeogenesis in vivo, pyruvate tolerance tests were performed on age- and weight-matched lean male mice. IP KO mice have a decreased gluconeogenic capacity (Fig. 3A). In addition, hypoglycemic response to insulin was more pronounced in IP KO mice compared with WT littermates (Fig. 3B and C). There were no marked improvements in glucose tolerance and normal glucose clearance in IP KO mice (Fig. 3D). However, we did not detect any overt differences in blood insulin and glucagon levels between IP KO and WT mice in fed or fasted states, despite a sharp decrease in insulin and a marked elevation of glucagon in response to fasting (Fig. 3E and F). No significant differences were detected in O2 consumption, and CO2 and heat production, the respiratory exchange ratio, and activity between IP KOs and WTs (Supplementary Fig. 8A–E).
IP Deficiency Protected Against STZ- and HFD-Induced Diabetes
Multiple doses of STZ induced severe degeneration and necrosis of pancreatic β-cells (Fig. 4A) and islet shrinkage (Fig. 4B–C) in IP KO and WT mice, as reflected by a striking reduction of insulin secretion (Fig. 4D). The elevation in blood glucose evoked by STZ in WT mice fed regular chow was significantly blunted by IP deletion (Fig. 4E). No differences in insulin and glucagon levels were detected between IP KO and WT mice (Fig. 4D). After the HFD challenge, IP KO mice gained approximately the same body weight as their WT littermates (Fig. 4F) and exhibited normal body composition (Fig. 4G). There was no significant difference of blood glucose levels between HFD-challenged IP KO and WT mice (Fig. 4H). In response to insulin and glucose tolerance tests after 16 weeks of HFD feeding, however, the plasma glucose levels were lower in IP KO mice than in WT controls (Fig. 4I and J). Moreover, hepatic TG and cholesterol accumulation (Fig. 4K) and hepatocellular vacuolation (Fig. 4L) induced by the HFD were also attenuated in IP KOs compared with littermate WT controls, whereas plasma levels of insulin and glucagon were unaltered (Fig. 4M). The increased expression of the proinflammatory genes, tumor necrosis factor-α, monocyte chemotactic protein-1, and interleukin 6, evoked in peritoneal macrophages and epididymal fat by the HFD, was restrained by IP deletion (Fig. 4N and O). Thus, IP deletion attenuated the inflammatory response and appeared to confer protection against STZ- and HFD-induced diabetes in mice through improvement of glucose metabolism.
Hepatic Reexpression of the IP Augmented Gluconeogenesis and Reduced Insulin Sensitivity in IP KO Mice
We sought to determine whether reexpression/overexpression of the IP could rescue hepatic gluconeogenesis and insulin tolerance in IP KO mice. Ads encoding the IP were introduced into IP KO mice by tail vein injection. Adenoviral delivery of genes resulted in their hepatic expression, as indicated by fluorescence intensity (Fig. 5A) and quantitative RT-PCR of IP expression (Fig. 5B). In the fasted state, Ad-mediated overexpression of IP in liver significantly improved hypoglycemia (Fig. 5C) and elevated gluconeogenesis, as reflected by pyruvate challenge in IP overexpression (Fig. 5D). Thus, reexpression of the IP in liver alone substantially rescued the metabolic phenotype exhibited in mice globally deficient in the receptor (Fig. 5E).
Disruption of the PKA/CREB Pathway Contributed to Impaired Gluconeogenesis Consequent to IP Deficiency
To explore the molecular mechanisms underlying IP-mediated regulation of gluconeogenesis in the liver, we first determined whether cAMP/PKA activity was regulated by PGI2 in hepatocytes (27). In cultured primary hepatocytes, IP disruption dramatically reduced glucose production at different doses of insulin, resulting in suppression of glucose output (Fig. 6A). Hepatocyte intracellular cAMP was suppressed by IP deletion under basal and glucagon-evoked conditions (Fig. 6B). Similarly, glucagon-evoked phosphorylation of CREB, PCK1, and G6Pase (Fig. 6C and D) was also attenuated by IP deletion. In contrast, reexpression of the IP in livers of IP KOs (Fig. 6E) substantially rescued these phenotypes (Fig. 6F–I). The PKA inhibitor, H-89, lacked the augmentation of CREB phosphorylation induced by Ad-IP and partially inhibited the induction of PCK1 and G6Pase (Fig. 6G and H). In addition, reexpression of IP boosted HGP (from 54 ± 3.9 μg ⋅ mg−1 ⋅ h−1 to 87 ± 4.8 μg ⋅ mg−1 ⋅ h−1), which could be depressed only 29.8% by H-89 treatment (from 87 ± 4.8 μg ⋅ mg−1 ⋅ h−1 to 61 ± 3.2 μg ⋅ mg−1 ⋅ h−1) in Ad-IP–infected hepatocytes. However, this remained higher than that in GFP-expressing hepatocytes (Fig. 6I), indicating that an additional IP-mediated mechanism, independent of the cAMP/PKA pathway (Fig. 6J), may influence hepatic gluconeogenesis.
IP Deficiency Increases AKT Activation in Liver
In addition to glucagon/CREB signaling, HGP metabolism is also regulated by the insulin/AKT/FoxO1 axis. In the fed state, increased insulin secretion activates the AKT pathway in hepatocytes, which in turn, phosphorylates and inhibits FoxO1, which is subjected to ubiquitination and degradation in the cytoplasm, resulting in reduced expression of gluconeogenic genes, PCK1 and G6Pase (28). Under basal conditions, the level of phosphorylation of AKT (Ser473 and Thr308) is quite low in primary hepatocytes from WT mice. However, deletion of the IP in primary hepatocytes markedly increased phosphorylation of AKT, FoxO1 (Ser253), and GSK3β (Ser9, another AKT downstream substrate), even in the absence of insulin stimulation (Fig. 7A and B). Similar results were observed in livers obtained from mice fed a chow diet or HFD (Supplementary Fig. 9A and B). Again, the expression levels of PCK1 and G6Pase decreased in IP KO hepatocytes (Fig. 7A). However, we did not detect alterations of IRS-1 protein expression and its tyrosine phosphorylation in IP-deficient hepatocytes (Fig. 7A and B), consistent with unchanged plasma insulin levels in the IP KO mice (Fig. 3E). Thus, IP deletion appeared to modulate AKT activity independent of an effect on the IR. The AKT inhibitor IV, which efficiently restrained phosphorylation of AKT at Ser-473 and Thr-308 (Supplementary Fig. 10A and B), blocked the augmented phosphorylation of AKT and FoxO1 in IP-deficient hepatocytes (Fig. 7C and D) but did not completely abolish the differences in expression of PCK1 and G6Pase between the two genotypes (Fig. 7C and D). Knockdown of AKT in primary hepatocytes further confirmed the findings described above (Supplementary Fig. 11). In addition, IP disruption in cultured hepatocytes reduced HGP approximately by half (Fig. 7E). This corresponds to a reduction of approximately one-fifth after treatment with AKT inhibitor IV (215 ± 10.83 μg ⋅ mg−1 ⋅ h−1 vs. 175 ± 6.7 μg ⋅ mg−1 ⋅ h−1), which was consistent with the premise that elevated activation of AKT contributed to suppression of hepatic gluconeogenesis in IP KO mice (Fig. 7F).
PGI2 Modulated Hepatic Gluconeogenesis via the PI3K/PKC-ζ/TRB3 Pathway to Depress AKT
The activity of AKT can be modulated through its interaction with various binding partners (29). For example, AKT activity can be downregulated by CTMP (30) and pseudokinase TRB3, an endogenous AKT inhibitor that binds to AKT and prevents insulin-mediated AKT phosphorylation (31). We observed that TRB3 expression at mRNA (Supplementary Fig. 12A and B) and protein (Fig. 8A and Supplementary Fig. 13A) levels was markedly suppressed in livers from IP KO mice under fed and fasted conditions and that this expression could be rescued by hepatic reexpression of the IP in the KO mice (Fig. 8B and Supplementary Fig. 13B). Deletion or reexpression of IP in hepatocytes failed to influence expression of PDK1 and CTMP (data not shown). Insulin also induces the expression of TRB3, which, in turn, modulates AKT activity to maintain normal glucose metabolism (31). Insulin treatment led to rapid AKT phosphorylation (peaking at 2 h), followed by subsequent deactivation, whereas TRB3 induction lagged behind AKT phosphorylation (Fig. 8C and Supplementary Fig. 13C and D), which was consistent with the notion that TRB3 negatively regulates AKT activation (32). In addition to augmenting activation of AKT, the dynamic expression of TRB3 was reduced in IP-deficient hepatocytes (Fig. 8C).
TRB3 expression depends on the activity of phosphatidyl inositol 3 kinase (PI3K) (33) and its downstream atypical PKC, PKC-ζ (32). As with other G-protein–coupled receptors, stimulation of the IP can activate class 1B PI3K, (PI3K-γ, catalytic unit p110-γ), through heterotrimeric G proteins Gα and Gβγ that bind to the pleckstrin homology domain in the NH2-terminal region of PI3K-γ (34). Phosphorylation of PKC-ζ was decreased in IP-deficient hepatocytes (Fig. 8D and Supplementary Fig. 13E). Reexpression of the IP significantly elevated phosphorylation of PKC-ζ (Fig. 8E and Supplementary Fig. 13F), whereas the IP-specific antagonist CAY-10441, PI3K inhibitor LY294002, and PKC inhibitor RO32-0432 all suppressed the induction of PKC-ζ phosphorylation (Fig. 8F). Similarly, differences in hepatic TRB3 expression resulting from deficiency or reexpression of IP were blunted by the PI3K inhibitor LY294002 and the PKC-ζ inhibitor RO32-0432 (Supplementary Fig. 14A and B), indicating that regulation of TRB3 by IP is mediated by the PI3K/PKC-ζ pathway. Knockdown of PKC-ζ in primary hepatocytes further confirmed our findings described above (Supplementary Fig. 15) Silencing of TRB3 (Ad-shTRB3) suppressed IP-mediated HGP by ∼40% (from 99 ± 4.1 μg ⋅ mg−1 ⋅ h−1 to 56 ± 4.9 μg ⋅ mg−1 ⋅ h−1), whereas the combination of Ad-shTRB3 and the PKA inhibitor H-89 completely abolished HGP induced by IP reexpression in hepatocytes (Fig. 8G). Likewise, we also observed the suppression of PI3K-γ signaling and decreased phosphorylation of CREB in mice treated with high doses of ASA, and these trends were reversed by activation of IP (Supplementary Fig. 16). These findings implicated PI3K-γ/PKC-ζ/TRB3/AKT signaling in IP modulation of hepatic gluconeogenesis (Fig. 8H).
Discussion
In this study, we found that hepatic expression of COX-2, PGI2 production, and IP expression were enhanced under conditions associated with augmented hepatic gluconeogenesis (e.g., during fasting in WT mice and in mice predisposed to diabetes due to genetic mutations or pharmacological treatment). Disruption in this pathway, specifically by deleting the IP, inhibited hepatic gluconeogenesis and ameliorated diabetes for the latter conditions, whereas this phenotype could be rescued in large part by hepatic reexpression of the IP. These findings establish a novel connection between PGI2, a cardioprotective (35) and proinflammatory eicosanoid (27), and carbohydrate metabolism, which may at least partly explain the reported metabolic effects of NSAIDs in humans (8).
Inhibition of COX-2, the dominant source of PGI2 biosynthesis in humans (36), has been reported to increase insulin sensitivity in healthy (6) and in overweight individuals (7). Pharmacological inhibition and genetic ablation of COX-2 in mice reduces fasting glucose and protects against STZ-induced diabetes (37). Treatment with high doses of ASA, which would be expected to inhibit COX-1 and COX-2, ameliorates insulin resistance in patients with type 2 diabetes by reducing gluconeogenesis and stimulating peripheral glucose uptake (8), likely by inhibiting I-κB kinase/nuclear factor-κB activity (26), which influences COX-2 transcription and activity. These observations are consistent with the notion that COX-2–dependent products play a functionally important role in the regulation of gluconeogenesis in humans.
We also found dysregulated expression of COX-2, PGI2 biosynthesis, and IP expression in the livers of fasted, diet-induced obese, and ob/ob mice. IP deficiency depressed fasting-induced hepatic gluconeogenesis and enhanced the hypoglycemic effect of insulin in mice, probably due to impaired counterregulatory response (38). Moreover, deletion of IP in mice slowed the progression of diabetes induced by STZ or the HFD. Conversely, hepatic overexpression of IP in the liver increased gluconeogenesis, resulting in insulin resistance. IP deletion in primary hepatocytes resulted in the diminished glucagon-mediated phosphorylation of CREB and diminished transcription of gluconeogenic genes, and this phenotype was rescued by reexpression of the IP in liver. These results suggest that the COX-2/PGI2/IP axis influences gluconeogenesis, at least in part, by enhancing the glucagon-mediated pathway. Conversely, insulin has been shown to depress hepatic gluconeogenesis through the IR/IRS-mediated PI3K/AKT/FoxO1 pathway (39). Indeed, IP ablation also increased AKT/FoxO1 signaling in response to insulin in cultured hepatocytes and in livers from starved and HFD-challenged mice. IP activation also perturbs hepatic insulin signaling to regulate gluconeogenesis through modulation of AKT activity. Class 1A PI3Ks (including PI3K-α and PI3K-β) are activated by receptor tyrosine kinase (such as IR), whereas class 1B PI3K (i.e., PI3K-γ) is activated by the binding of p110-γ to the Gβγ unit of the G-protein–coupled receptor (34). Meanwhile, p110-γ interacts physically with PKC-ζ (40). We did observe that the elevation of TRB3 expression induced by activation of IP (a Gαs–coupled receptor) could be abrogated by both PI3K and PKC-ζ inhibitors, consistent with previous studies, indicating that hepatic TRB3 could be regulated by PI3K and PKC-ζ (32,41). Moreover, we did not observe any effects of cAMP/CREB on TRB3 expression by chromatin immunoprecipitation assay (Supplementary Fig. 17), which is consistent with PKA inhibitor (H89) treatment and was unable to entirely block HGP induced by IP reexpression. Collectively, these results indicated that activation of the IP regulated hepatic gluconeogenesis via both cAMP/PKA/CREB and PI3K-γ/PKC-ζ/TRB3/AKT pathways, and a combination of H89 treatment and TRB3 silencing could entirely restrain IP activation–induced HGP.
PGI2 displays proinflammatory (e.g., rheumatoid arthritis) and anti-inflammatory (e.g., atherosclerosis) properties, depending on the inflamed organs and the pathological models (27). Our results indicated that PGI2/IP may modulate the pathological process of excessive hepatic gluconeogenesis in diabetes. As one of the effective drugs for the management of pulmonary arterial hypertension, synthetic PGI2-epoprostenol has been reported to elevate serum glucose in humans and animals (42), whereas treatment with another PGI2 analog, iloprost, significantly reduced lactate and slightly increased glucose in patients with critical limb ischemia (43), thereby indirectly implicating IP signaling in gluconeogenesis. However, PGI2 in fat tissue also promotes adipose cell differentiation (44) and de novo brown adipocyte tissue recruitment in WAT (45) through another natural nuclear receptor, PPAR-γ. Interestingly, treatment with beraprost, a PGI2 analog, ameliorated complications of diabetes, such as nephropathy, in severely diabetic rodents by reduction of inflammation in peripheral tissues through upregulation of PPAR-γ (46) and perhaps PPAR-δ (47). Subsequent activation of PPAR-γ, which reduces a flux of free fatty acid and cytokines from adipose tissue to the liver, results in increased insulin sensitivity in diabetes (48). However, a slight increase of hepatic PPAR-γ expression was observed in IP KO mice (data not shown), suggesting that beraprost might regulate PPAR-γ directly (49) rather than through IP.
NSAIDs, particularly those specific for inhibition of COX-2, can influence carbohydrate metabolism, specifically gluconeogenesis, in humans. Among the prostanoids, COX-2 is the dominant contributor to biosynthesis of PGI2, and PGI2 is the major product of mouse hepatocytes under physiological conditions, Here, we show that the IP for PGI2 modulates hepatic gluconeogenesis through both the Gαs/PKA/CREB and Gβγ/PI3K-γ/PKC-ζ/TRB3/AKT pathways (Fig. 8H) and that disruption of this receptor confers protection against the progression of diabetes by inhibition of hepatic gluconeogenesis. These observations provide a mechanistic rationale for clinical observations suggesting a role for COX-2 inhibition in the regulation of carbohydrate metabolism.
Article Information
Acknowledgments. The authors thank Dr. Shengzhong Duan (Institute for Nutritional Sciences Shanghai Institutes for Biological Sciences, Chinese Academy of Sciences) for technical assistance.
Funding. This work was supported by grants from the Ministry of Science and Technology of China (2012CB945100, 2011CB503906, 2011ZX09307-302-01, and 2012BAK01B00), the National Natural Science Foundation of China (NFSC 81030004), NSFC-Canadian Institutes of Health Research (CIHR) joint grant (NSFC 81161120538 and CIHR-CCI117951), the Knowledge Innovation Program of the Chinese Academy of Sciences (KSCX2-EW-R-09), and the Clinical Research Center at the Institute for Nutritional Sciences, Shanghai Institutes for Biological Sciences (CRC2010007). Y.Y. was supported by the One Hundred Talents Program of the Chinese Academy of Sciences (2010OHTP10) and Pujiang Talent Program of Shanghai Municipality (11PJ1411100). Y.Y. is a Fellow at the Jiangsu Collaborative Innovation Center for Cardiovascular Disease Translational Medicine.
Duality of Interest. No potential conflicts of interest relevant to this article were reported.
Author Contributions. S.Y., Q.Z., and Y.Y. designed the research associated with the project. S.Y., Q.Z., X.Z., J.T., Y.W., Y.Z., and J.Z. performed experiments. J.Y., F.G., Y.L., and G.A.F. provided important reagents. S.Y., G.A.F., and Y.Y. wrote the manuscript. Y.Y. 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.