The Farnesoid X Receptor Modulates Hepatic Carbohydrate Metabolism during the Fasting-Refeeding Transition*

The liver plays a central role in the control of blood glucose homeostasis by maintaining a balance between glucose production and utilization. The farnesoid X receptor (FXR) is a bile acid-activated nuclear receptor. Hepatic FXR expression is regulated by glucose and insulin. Here we identify a role for FXR in the control of hepatic carbohydrate metabolism. When submitted to a controlled fasting-refeeding schedule, FXR-/- mice displayed an accelerated response to high carbohydrate refeeding with an accelerated induction of glycolytic and lipogenic genes and a more pronounced repression of gluconeogenic genes. Plasma insulin and glucose levels were lower in FXR-/- mice upon refeeding the high-carbohydrate diet. These alterations were paralleled by decreased hepatic glycogen content. Hepatic insulin sensitivity was unchanged in FXR-/- mice. Treatment of isolated primary hepatocytes with a synthetic FXR agonist attenuated glucose-induced mRNA expression as well as promoter activity of L-type pyruvate kinase, acetyl-CoA carboxylase 1, and Spot14. Moreover, activated FXR interfered negatively with the carbohydrate response elements regions. These results identify a novel role for FXR as a modulator of hepatic carbohydrate metabolism.

The liver plays a major role in maintaining plasma glucose homeostasis by controlling a delicate balance between hepatic glucose uptake/utilization and hepatic glucose production. In the fed state, the liver stores energy from glucose by synthesizing glycogen and fat. Conversely, when plasma glucose concentrations decrease during fasting, the liver produces glucose by the glycogenolytic and gluconeogenic pathways. This fasting-refeeding transition involves a highly coordinated adaptation of the expression of genes encoding key metabolic enzymes that is orchestrated by hormones and nutrients.
In the fed state, insulin and glucose act in concert to promote the expression of genes controlling glucose utilization and fatty acid (FA) 1 synthesis, including glucokinase (hexokinase type IV), L-type pyruvate kinase (Lpk), acetyl-coenzyme A carboxylase-1 (Acc-1), and fatty acid synthase (Fas) (1). A primary action of insulin appears to be the induction of glucokinase expression, which catalyzes the first step of intracellular hepatic glucose metabolism and signaling (2). The sterol regulatory element-binding protein-1c (Srebp-1c) has emerged as a major mediator of this insulin action (3). On the other hand, glucose also directly affects gene transcription via carbohydrate-response elements (ChoREs) that have been identified in both glycolytic (Lpk) and lipogenic (Acc, Fas, and Spot14) genes (4,5). Recently, the transcription factor carbohydrate response element-binding protein (Chrebp) was identified based on its ability to bind the ChoRE of the Lpk promoter (6,7). Additionally, hepatocyte nuclear factor 4␣ (HNF4␣) acts in concert with proteins that bind to ChoREs to elicit maximal Lpk promoter activity (8,9). In the fasting state, the production of glucose by the liver is crucial for tissues unable to use FA as a source of energy, such as the brain and red blood cells. Phosphoenolpyruvate carboxykinase (Pepck) is considered to be a rate-controlling enzyme of gluconeogenesis (10). Inhibition of Pepck activity by insulin is considered an important physiological mechanism to repress hepatic glucose production in the early post-prandial state (11).
The farnesoid X receptor (FXR) is a nuclear receptor that is activated by bile acids (BAs) (12)(13)(14). FXR regulates BA synthesis from cholesterol by controlling the expression of several key enzymes (15,16). FXR also influences blood lipids because FXR-deficient (FXR Ϫ/Ϫ ) mice display elevated serum levels of triglycerides and cholesterol (17). Recently, we found that FXR expression in rat liver is regulated by glucose and insulin (18), suggesting that this nuclear receptor may also play a role in carbohydrate metabolism. In accordance with this hypothesis, FXR has been shown to regulate hepatic Pepck expression (19 -21). Nevertheless, the physiological relevance of this regulation rests to be determined. * This work was supported in part by the FEDER-Conseil Régional Nord-Pas de Calais Génopole 01360124 and the Fondation Leducq. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
§ Both authors contributed equally to this manuscript. To address this issue, we performed fasting-refeeding experiments in FXR wild-type (FXR ϩ/ϩ ) and FXR Ϫ/Ϫ mice. Our results indicate that FXR interferes with glycolytic and lipogenic pathways in liver by controlling, among other genes, Lpk, Acc-1, and Spot14 transcription. These data suggest that FXR also functions as a molecular modulator of hepatic carbohydrate and lipid metabolism, which uncovers a novel physiological response to the enterohepatic circulation of bile acids.
Animals and Diets-All studies were approved by the institutional review boards for the care and use of experimental animals. Homozygous FXR-deficient (FXR Ϫ/Ϫ ) mice (17) and sex-and age-matched wildtype mice (FXR ϩ/ϩ ) bred on the C57BL/6N genetic background, housed in a pathogen-free barrier facility with a 12-h light/12-h dark cycle, were maintained on a standard laboratory chow diet (UAR AO3, Villemoison/Orge, France). For the fasting-refeeding experiments, 8 -12-week-old female mice were divided into three groups: non-fasted (NF), fasted (F), and refed (RF). The non-fasted group was fed standard chow diet ad libitum and killed at 9.00 a.m. The fasted group was fasted during 24 h (from 9:00 a.m to 9:00 a.m) before sacrifice, whereas the refed group was subjected to an additional refeeding with a high carbohydrate/low fat diet (Harlan Teklad TD88122, Madison, WI) for 6 or 24 h. The high carbohydrate diet contained 48.6% sucrose, 16.6% corn starch, 22.2% casein, 5.5% cellulose, 1% corn oil, 3.9% mineral, 1% vitamins.
Hyperinsulinemic Euglycemic Clamp-Male FXR Ϫ/Ϫ and agematched wild-type mice (n ϭ 5-6 mice/group) were equipped with a permanent catheter in the right atrium via the jugular vein and fasted during 9 h before the start of the experiment, as previously described (22,23).
In Vivo Insulin Stimulation and Analysis of Insulin Signaling-Mice were starved during 24 h, anesthetized with pentobarbital, and injected with 1 IU/kg of human insulin (Actrapid, Novo Nordisk) into the portal vein. Livers were removed 5 min after injection and frozen in liquid nitrogen. Liver proteins were extracted in lysis buffer (20 mmol/liter Tris-HCl, pH 7.5, 150 mmol/liter NaCl, 5 mmol/liter EDTA, 30 mmol/ liter sodium pyrophosphate, 50 mmol/liter NaF, 1% Triton X-100, 1 g/ml pepstatin, 2 g/ml leupeptin, 5 g/ml aprotinin, 1 mM phenylmethylsulfonyl fluoride, 1 mmol/liter sodium orthovanadate). For coimmunoprecipitation experiments, proteins were incubated overnight at 4°C with the indicated antibodies in the presence of protein A-agarose. The immunoprecipitates were washed in lysis buffer, subjected to SDS-PAGE, and transferred to nitrocellulose. Membranes were subjected to immunodetection with the indicated antibodies. Immunoreactive bands were revealed using an ECL detection kit (Amersham Biosciences).
Primary Hepatocytes-Rat primary hepatocytes were isolated by collagenase perfusion from livers of Wistar rats and cultured as previously described (18). Mouse primary hepatocytes were isolated from the livers of fed mice by a modification of the collagenase method (24). Briefly, livers from control and FXR Ϫ/Ϫ mice were perfused with Hanks' balanced salt solution (KCl, 5.4 mM; KH 2 PO 4 , 0.45 mM; NaCl, 138 mM; NaHCO 3 , 4.2 mM; Na 2 HPO 4 , 0.34 mM; glucose, 5.5 mM; HEPES, 1 M; EGTA, 50 mM; CaCl 2 , 50 mM; pH 7.4) at a rate of 5 ml/min via portal vein before addition of collagenase (Sigma) (0.025%) was added. Cell viability was assessed by the trypan blue exclusion test and was always higher than 60%. Hepatocytes were cultured as a monolayer in serumfree William's E medium (Invitrogen, France) supplemented with 2 mmol/liter glutamine, 25 g/ml gentamicin, 100 nmol/liter dexamethasone (Soludecadron, MSD), 0.1% fatty acid-free bovine serum albumin, 2% (v/v) Ultrocer G (Invitrogen) at 37°C in a humidified atmosphere of 5% CO 2 , 95% air. After overnight culture, cells were incubated for the indicated time in fresh culture medium supplemented with different glucose or insulin concentrations as indicated. When indicated, GW4064 (5 mol/liter) or Me 2 SO were added 12 h before the culture medium change.
Statistical Analysis-Statistical significance was assessed using analysis of variance followed by unpaired t test. Values of p Ͻ 0.05 were considered significant.

Fxr Deficiency Accelerates the Expression Response of Genes Involved in Hepatic Glycolysis and de Novo Lipogenesis upon
Refeeding-To determine whether Fxr deficiency interferes with the physiological control of hepatic carbohydrate metabolism in vivo, FXR ϩ/ϩ and FXR Ϫ/Ϫ mice were subjected to a 24-h fasting period followed by refeeding with a high carbohydrate diet (HCHO). There was no difference in food intake between genotypes (data not shown). Whereas plasma glucose levels were comparable between both strains during the non-fasted and fasted states, plasma glucose concentrations were significantly lower at 6 h after refeeding in FXR Ϫ/Ϫ mice (Table I). These differences were not associated to altered intestinal glu-cose absorption in FXR Ϫ/Ϫ mice, because an early blood glucose excursion curve after an oral glucose tolerance test was similar between FXR Ϫ/Ϫ and FXR ϩ/ϩ (data not shown). Insulin concentrations in non-fasted and fasted states were not different between the two genotypes. However, plasma insulin levels were significantly lower in FXR Ϫ/Ϫ mice upon refeeding, reflecting either a pancreatic adaptative response to the relative hypoglycemia or an increased insulin sensitivity. Lactate concentrations were higher in non-fasted and refed FXR Ϫ/Ϫ mice, suggesting an increase of glucose utilization in these animals. In contrast, plasma concentrations of ␤-hydroxybutyrate were not statistically different between both groups. Whereas plasma FFA concentrations were slightly higher in FXR Ϫ/Ϫ during the non-fasted state, FFAs did not differ in both fasted and refed states between both genotypes.
To evaluate whether FXR plays a role in hepatic glucose metabolism, mRNA levels of several key enzymes of glycolysis, lipogenesis, and gluconeogenesis were measured. Hepatic mRNA levels of Lpk, a major enzyme of the glycolytic pathway, increased more rapidly in FXR Ϫ/Ϫ than in FXR ϩ/ϩ mice in response to refeeding (Fig. 1A), being approximately 3-fold higher in FXR Ϫ/Ϫ after 6 h refeeding. However, the maximal increase of Lpk mRNA obtained after 24 h of refeeding did not differ between the two genotypes. On the other hand, the expression of glucokinase was not significantly different between both strains (data not shown). In the absence of Fxr, increased glycolysis may direct substrate flow toward hepatic de novo lipogenesis, because both Acc-1 and Fas mRNAs were induced more rapidly upon refeeding (Fig. 1, B and C). Concerning the gluconeogenesis pathway, the expression of Pepck appeared to be more rapidly down-regulated upon refeeding in FXR Ϫ/Ϫ mice, being already returned to its basal value after 6 h of refeeding (Fig. 1D). In contrast, the expression of the catalytic subunit of glucose-6-phosphatase (Glc-6-Pase) did not differ between both strains (data not shown).
To determine the metabolic consequences of Fxr deficiency in the liver in vivo, glycogen and TG contents were measured after fasting-refeeding. Liver glycogen content was significantly reduced in both fed and refed FXR Ϫ/Ϫ mice ( Fig. 2A). Hepatic expression of glycogen synthase and glycogen phosphorylase, two key enzymes of glycogen metabolism, were similar in both groups (data not shown). As previously described, FXR Ϫ/Ϫ mice displayed a higher liver TG content in the nonfasted state. Nevertheless, hepatic TG content did not differ after 6 ( Fig. 2B) and 24 h (data not shown) of refeeding in both strains. Under fasting conditions, however, hepatic TG concentrations rose to a similar extent in FXR Ϫ/Ϫ and FXR ϩ/ϩ mice. Therefore, hepatic TG concentrations followed the plasma FFA increase during fasting (Table I), likely reflecting the changes in adipose tissue lipolysis. As expected (17), plasma TG concentrations were increased in FXR Ϫ/Ϫ mice in the non-fasted state (Fig. 2C). Whereas plasma TG concentrations decreased to a similar level in the two groups during fasting, they were significantly higher in FXR Ϫ/Ϫ mice upon refeeding. Plasma TG concentrations in FXR Ϫ/Ϫ mice upon refeeding is associated with increased export, because hepatic very low density lipoprotein production was significantly increased in FXR Ϫ/Ϫ mice after 6 h of refeeding (Fig. 2D). Taken together, these results suggest that increased mRNA expression of lipogenic enzymes observed in FXR Ϫ/Ϫ mice during refeeding leads to increased triglyceride synthesis.
Influence of Fxr Deficiency on the Expression of Hepatic Transcription Factors Controlling Fasting-Refeeding Transition-To investigate the molecular mechanism of Fxr deficiency on hepatic carbohydrate metabolism, mRNA levels of transcription factors controlling hepatic gene expression in response to nutritional changes were measured in FXR ϩ/ϩ and FXR Ϫ/Ϫ mice. FXR mRNA was induced during fasting as reported previously (26), and its expression rapidly decreased upon refeeding (Fig. 3A). Surprisingly, the expression of small heterodimer partner (Shp), a well characterized FXR target gene (16), did not change upon fasting-refeeding in wild-type mice. As expected (17), Shp was expressed at very low levels in FXR Ϫ/Ϫ mice (Fig. 3B). As previously reported (27), Srebp-1c mRNA expression decreased upon fasting and was induced FIG. 1. FXR ؊/؊ mice display an altered hepatic gene expression profile of glycolytic and lipogenic enzymes during the fastingrefeeding schedule. Non-fasted (NF) FXR Ϫ/Ϫ and age-matched FXR ϩ/ϩ mice were subjected to a 24-h fasting (F), and then refed (RF) for 6 or 24 h with a high carbohydrate/low fat diet, as described under "Experimental Procedures." Total RNAs from livers of FXR Ϫ/Ϫ (filled bars) and FXR ϩ/ϩ (empty bars) mice were subjected to real-time PCR quantification. Values are normalized relative to 36B4 mRNA and are expressed (mean Ϯ S.E.) relative to those of ad libitum non-fasted FXR ϩ/ϩ mice, which are arbitrarily set at 1. Statistical differences were indicated: *, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001 compared with FXR ϩ/ϩ mice (n ϭ 5-7 mice/group). upon refeeding (Fig. 3C). A similar, albeit less pronounced, response was observed for Chrebp (Fig. 3D). However, both genes responded similarly in FXR ϩ/ϩ and FXR Ϫ/Ϫ mice. Taken together, these results suggest that the observed differences in glycolytic (LPK) and lipogenic (Fas and Acc-1) gene expression in FXR Ϫ/Ϫ mice are not mediated via alterations in the transcriptional regulation of Srebp-1c and Chrebp. Conversely, peroxisome proliferator-activated receptor ␥-coactivator 1␣ (PGC-1␣) expression increased upon fasting and decreased upon refeeding, in line with previous reports (28). Pgc-1␣ mRNA levels were significantly lower in the basal (nonfasted) state, slightly increased during fasting, and more rapidly decreased after 6 h of refeeding in FXR Ϫ/Ϫ mice (Fig. 3E). Hnf4␣ mRNA also decreased more rapidly after 6 h of refeeding in FXR Ϫ/Ϫ mice (Fig. 3G). Finally, Foxo1 mRNA levels displayed a similar expression profile as Pgc-1␣ and Hnf4␣, with a faster decrease upon refeeding in FXR Ϫ/Ϫ mice (Fig. 3F). However, the level of serine 256 phosphorylation of endogenous Foxo1 protein did not differ in liver of both strains after 6 h of refeeding (data not shown).
Fxr Deficiency Does Not Alter Hepatic Insulin Signaling-Because insulin and glucose act synergistically to induce the transcription of lipogenic genes, we examined whether insulin or glucose signaling or both were altered in the liver of FXR Ϫ/Ϫ mice. Basal hepatic glucose production, assessed during an euglycemic, hyperinsulinemic clamp, after a moderate fasting period of 9 h, was lower in FXR Ϫ/Ϫ compared with FXR ϩ/ϩ mice (136 versus 103 mol/kg/min; p Ͻ 0.05). In contrast, in the steady state condition, clamped hepatic glucose production was suppressed to a similar extent by insulin in both groups (FXR ϩ/ϩ , 54 Ϯ 7 mol/kg/min; FXR Ϫ/Ϫ , 48 Ϯ 7 mol/kg/min), indicating that hepatic insulin sensitivity was not altered in FXR Ϫ/Ϫ mice. To further investigate insulin receptor (IR)-mediated signaling in the liver, mice were fasted overnight and injected with either recombinant human insulin (1 IU/kg) or saline via the portal vein. Insulinstimulated tyrosine phosphorylation of IR and insulin receptor substrate-2 (IRS-2), the main IR-docking substrate in liver, were similar in both genotypes (Fig. 4, A and B). In addition, Fxr deficiency did not alter the binding of the p85␣ regulatory subunit of phosphatidylinositol 3-kinase to IRS-2 nor the level of insulin-stimulated phosphorylation of Akt (Fig. 4, B and C). These data confirm that the changes in hepatic glucose metabolism observed in FXR Ϫ/Ϫ mice are not because of altered hepatic insulin signaling.
Ligand-activated Fxr Modulates Glucose-induced, but Not Insulin-regulated Gene Expression in Primary Hepatocytes-To determine whether Fxr activation interferes with glucose signaling directly in the hepatocyte, the effect of Fxr activation on the regulation of gene expression by glucose and insulin was studied in vitro using the primary rat hepatocyte model.
First, the effect of the specific Fxr ligand GW4064 on the transcriptional repression of Pepck by insulin was investigated (Fig. 5A). As expected, the PEPCK mRNA level was increased Ͼ20-fold by a combination of dexamethasone and the cAMP Values are normalized relative to 36B4 mRNA and are expressed (mean Ϯ S.E.) relative to those of ad libitum non-fasted FXR ϩ/ϩ mice, which are arbitrarily set at 1. Statistical differences were indicated: §, p Ͻ 0.05; § §, p Ͻ 0.01 compared with fed mice. *, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001 compared with FXR ϩ/ϩ mice (n ϭ 5-7 mice/group). analog 8-CPT-cAMP, and dropped significantly upon incubation with insulin. Pre-treatment of the cells with GW4064 did not alter the insulin-induced repression of Pepck mRNA expression (Fig. 5A), whereas it efficiently increased the expression of the Fxr-target gene Shp under the same conditions (data not shown). Because glucose is able to inhibit Pepck expression in the absence of insulin (29,30), it was then investigated whether Fxr activation by GW4064 could alter this effect. Pepck expression was decreased by ϳ50% in the presence of a high (25 mmol/liter) glucose concentration (Fig. 5B). Interestingly, pre-treatment with GW4064 abolished the inhibitory action of glucose on Pepck expression.
To determine whether activated Fxr also affects the activation of other glucose-responsive genes, primary rat hepatocytes were pre-treated with GW4064 (5 mol/liter) and subsequently cultured in low or high glucose concentrations. Lpk expression was strongly induced by high glucose (Fig.  5C). Glucose-dependent induction of LPK expression was reduced upon treatment with GW4064 (Fig. 5C). Moreover, the glucose-mediated induction of lipogenic genes such as Spot14 (S 14 ) and Acc-1 was also reduced by GW4064 (Fig. 5, D and  E). Glucose induction of Glc-6-Pase, another gene positively regulated by glucose (31), was also significantly blunted by Fxr activation with GW4064 (Fig. 5F). This suggests that Fxr modulates a broad panel of glucose-responsive genes, including glycolytic, lipogenic, and gluconeogenic enzymes. To verify the efficiency of Fxr activation, expression of the Fxrtarget gene Shp was measured (Fig. 5G). As expected, GW4064 strongly induced the expression of Shp, independent of the prevailing glucose concentration. These results support the concept that activated Fxr interferes with glucose signaling independent of changes in Shp expression.
To further confirm the role of Fxr in the regulation of Lpk by glucose, hepatocytes were isolated from FXR Ϫ/Ϫ and FXR ϩ/ϩ mice. A 2-fold induction of Lpk mRNA levels was observed in wild-type mice hepatocytes in response to high glucose concentrations, similarly as previously reported (2,32). Treatment with GW4064 abolished almost completely the glucose induction of Lpk mRNA (Fig. 6A, left panel). In hepatocytes isolated from FXR Ϫ/Ϫ mice, Lpk mRNA levels were drastically in- creased, already in the presence of low glucose concentrations, and incubation in high glucose concentration resulted in an additional slight increase of LPK mRNA expression (Fig. 6A,  right panel). In contrast to FXR ϩ/ϩ hepatocytes, GW4064 treatment did not alter Lpk gene expression in FXR Ϫ/Ϫ hepatocytes cultured both in low and high glucose concentrations (Fig. 6A,  right panel). As expected, Shp mRNA expression was increased in FXR ϩ/ϩ hepatocytes in response to GW4064 treatment (Fig.  6B, left panel). In contrast, this effect was lost in hepatocytes isolated from FXR Ϫ/Ϫ mice (Fig. 6B, right panel). Taken together, these data strongly suggest that FXR directly regulates the induction of Lpk by glucose.
Fxr Inhibits the Activation of Glucose-responsive Promoters at the Transcriptional Level-To determine the mechanism by which Fxr regulates glucose-induced gene expression, primary rat hepatocytes were co-transfected with the Lpk and Acc-1 promoters and subsequently incubated in either high or low glucose concentrations and treated with GW4064. High glucose strongly increased Lpk promoter activity, an effect that was prevented by GW4064 treatment (Fig. 7A, left panel). A similar effect, albeit in a lesser extent than for Lpk, was observed on the Acc-1 promoter (Fig. 7A, right panel). Moreover, a heterologous promoter driven by the L3L4-LPK glucose response element was also repressed by activated Fxr and this effect was enhanced by co-transfected FXR (Fig. 7B). In a similar way, glucose induction of an heterologous promoter containing the ChoRE-Spot14 region was also repressed by activated Fxr (Fig.  7C). Under similar transfection conditions, GW4064 and FXR induced the expression of a heterologous promoter driven by a consensus FXRE site (Fig. 7D). These results suggest that activated Fxr modulates the glucose response via negative interference with the ChoRE.
In light of the strong effect of activated Fxr on Lpk promoter activity, it was determined whether FXR binds physically to the L3L4 site. As previously described (25), FXR can bind to consensus IR-1 FXRE either as a monomer or as a heterodimer in the presence of RXR (Fig. 7E, lanes 2 and 4). FXR bound to the L3-Lpk site as monomer (Fig. 7E, lane 7), because the presence of RXR did not modify FXR complex mobility (Fig. 7E,  lane 8). The complex was specific, because the addition of an antibody against FXR was able to eliminate FXR binding (Fig.  7E, lane 9). No binding was observed on the L4-Lpk site (Fig.  7E, lanes 11 and 12). These results suggest that FXR also bind to the ChoRE region of the Lpk promoter, an effect that may contribute to the modulation of its glucose response. DISCUSSION FXR regulates genes controlling biological pathways such as the synthesis and transport of BAs as well as lipid and lipoprotein metabolism. Here, we demonstrate that FXR also plays a role in regulating hepatic carbohydrate metabolism. In vivo analysis of the phenotype of FXR Ϫ/Ϫ mice revealed an exaggerated response to HCHO refeeding, resulting in an earlier, more pronounced induction of expression of glycolytic and lipogenic genes. Conversely, ligand-activated Fxr decreased the expression of the same genes in response to high glucose concentrations in primary rodent hepatocytes. In addition, LpK mRNA expression was up-regulated in hepatocytes isolated from FXR Ϫ/Ϫ mice. At the molecular level, FXR decreased the transcription of the Lpk promoter induced by glucose acting via its ChoRE. Absence of Fxr therefore leads to an enhanced glycolytic flux, providing substrates for lipogenesis that may contribute to the hypertriglyceridemia observed in FXR Ϫ/Ϫ mice.
The liver is a major site of carbohydrate metabolism and de novo lipogenesis. Under HCHO refeeding conditions, glucose provides the main source of substrates for FA synthesis via production of acetyl coenzyme A. The observation that hepatic Fxr expression varies during nutritional changes (18,26) led us to hypothesize that Fxr may play a role in the control of carbohydrate metabolism. Several lines of evidence suggest that glucose utilization is increased in FXR Ϫ/Ϫ mice: (i) plasma glucose and insulin levels were significantly lower upon refeeding in FXR Ϫ/Ϫ mice even though food consumption and glucose absorption were similar in FXR Ϫ/Ϫ and FXR ϩ/ϩ mice, (ii) hepatic expression of Lpk, Fas, and Acc-1 was more rapidly induced upon refeeding in FXR Ϫ/Ϫ mice, (iii) the changes in gene expression of these glycolytic and lipogenic enzymes correlated with increased hepatic de novo triglyceride synthesis.
FA from the lipogenic pathway are partly used to synthesize TG, which can either be stored in the liver or exported as very low density lipoprotein particles. Whereas hepatic TG content was significantly higher in non-fasted FXR Ϫ/Ϫ mice, it rose to a similar extent under HCHO refeeding in both FXR Ϫ/Ϫ and FXR ϩ/ϩ mice. Furthermore, hepatic TG secretion was increased in FXR Ϫ/Ϫ mice upon refeeding, which may contribute to the elevated plasma TG levels observed under these conditions. FA from de novo lipogenesis efficiently mobilize apolipoprotein B and induce very low density lipoprotein particle assembly (33,34). Moreover, FXR Ϫ/Ϫ mice produce larger very low density lipoprotein particles than wild-type mice (35).
Because both insulin and glucose act synergistically to regulate glycolytic and lipogenic gene expression (1), the exaggerated response to refeeding in FXR Ϫ/Ϫ mice may reflect either an enhanced hepatic insulin sensitivity, a potentiation of carbohydrate-induced signaling, or a combination of both. Whereas a recent study reported that BAs may interfere with insulin signaling in hepatocytes (36), we were unable to detect any modification of hepatic insulin sensitivity in FXR Ϫ/Ϫ mice. This finding is in accordance with the similar mRNA levels of glucokinase in livers of FXR Ϫ/Ϫ and wild-type mice. Therefore, we determined whether changes in carbohydrate metabolism could explain the phenotype of FXR Ϫ/Ϫ mice. Primary hepatocytes were used to differentiate between systemic and hepatocyte-specific effects on glycolytic and lipogenic gene transcription. Interestingly, activation of Fxr with its specific ligand GW4064 significantly reduced glucose-induced expression of Lpk, Acc-1, and S 14 . This effect was glucose-dependent, because insulin concentrations were kept constant throughout the experiments. The genes encoding these enzymes all contain ChoREs in their promoters. In the case of the Lpk promoter, four nuclear protein-binding regions (L1 to L4) have been described (8). The L4 site contains the Lpk ChoRE (9) and binds the transcription factor Chrebp, which is thought to translate glucose signaling in transcriptional responses (6). The contiguous L3 site binds Hnf4␣ and several studies suggest that the L3L4 site is required for full promoter activity (9,37). In addition, the Acc-1 and S 14 promoters also contain ChoRE and are regulated by ChREBP (7,32). Results from transfection assays demonstrated that FXR inhibits glucose-induced Lpk and Acc-1 promoter transcription. Moreover, our results indicate that FXR negatively interferes with the isolated ChoRE from the Lpk and S 14 genes and decreases the glucose response. In addition, gel shift assays indicated that FXR physically interacts with the L3 region of the Lpk promoter, an effect that may contribute to the strong effect of Fxr on its transcriptional activity induced by glucose.
Our in vivo results indicate that Fxr regulates glucose-induced Lpk gene expression. Moreover, in vitro in primary mouse hepatocytes, Fxr deficiency results in a derepression of basal Lpk expression. Carbohydrate responsiveness is mainly mediated by the L4 region to which, among other transcription factors, Chrebp binds (6, 7). However, Chrebp expression was not different in the FXR Ϫ/Ϫ mice. Thus, additional molecular mechanisms for Fxr-mediated inhibition of glucose-regulated genes may exist. Considering that Srebp-1c may be negatively controlled by BAs (26,38), its expression levels were also measured to determine its potential involvement in the regulation of glycolytic and lipogenic gene expression in FXR Ϫ/Ϫ mice. However, Fxr deficiency was not associated with an increased Srebp-1c expression in vivo. Therefore, it is unlikely that the observed effects in FXR Ϫ/Ϫ mice occur via an indirect pathway implicating alterations in expression of Srebp-1c or Chrebp. Because both Chrebp and Srebp-1c are mainly regulated at the post-translational level, it cannot be excluded that Fxr interferes with the activity of any or more of these transcription factors. Taken together, our observations suggest that at least one of the molecular mechanisms of Fxr action on carbohydrate metabolism implicate a direct effect on the expression of glucose-regulated genes, such as Lpk.
Fxr modulates the kinetics rather than the amplitude of the response to dietary carbohydrate intake, because a strong in- FIG. 7. FXR inhibits the induction of ChORE-containing promoters. Primary rat hepatocytes were transfected with: A, Lpk or Acc-1 promoter-driven luciferase reporter plasmids, incubated in low (5 mmol/liter) or high (25 mmol/liter) glucose, insulin (1 nmol/liter), and GW4064 (5 mol/liter) as indicated; B, the heterologous L3L4 containing thymidine kinase (Tk) promoter; or C, the S 14 -ChoRE promoter driven luciferase reporter plasmids in the presence of cotransfected FXR expression vector and incubated in low (5 mmol/liter) or high (25 mmol/liter) glucose, insulin (1 nmol/liter), and GW4064 (5 mol/liter). D, primary rat hepatocytes were transfected with a heterologous luciferase reporter plasmid driven by a consensus FXRE-Tk promoter in the presence of a FXR expression vector and GW4064 (5 mol/liter) as indicated. Transfections were carried out in triplicate, and each experiment was repeated at least three times. E, the gel retardation assay was performed with labeled doublestranded consensus FXRE (IR-1, lanes 1-5), L3 site (L3-LPK, lanes 6 -9), or L4 site (L4-LPK, lanes 10 -12) oligonucleotides in the presence of in vitro translated FXR, RXR (retinoid X receptor), or HNF4␣, and of an anti-FXR antibody as indicated.
duction of Lpk, Fas, and Acc-1 mRNAs levels occurred as early as 6 h after refeeding in FXR Ϫ/Ϫ mice. Thus, Fxr appears to act as a molecular modulator of carbohydrate metabolism during the fasting-refeeding transition. The nutritional regulation of Fxr expression correlates well with the changes in hepatic glucose metabolism in wild-type mice. During fasting, Fxr is induced by Pgc-1␣ (26). However, the increased expression and activity of Cyp7a1, a negative regulated Fxr target gene (15,16), during fasting (20) suggests that the transcriptional activity of Fxr is decreased. From a physiological point of view, this would prepare BAs for the digestion and absorption of fats after a subsequent meal. Upon meal ingestion, the increased enterohepatic circulation of BA negatively feedbacks BA synthesis. In addition, our results show that the induction of Fxr may not only result in suppression of lipogenesis, but also in simultaneous inhibition of the glycolytic pathway to promote glycogen storage (Fig. 8). The rise in plasma insulin levels upon refeeding induces a decrease of Fxr expression (18), which may contribute to a redirection of the glucose flux to glycolysis in the inter-prandial state.
Fxr may also modulate gluconeogenesis. Previous studies have demonstrated that bile acids could regulate the gluconeogenic pathway by regulating Pepck gene expression, but the molecular mechanism remained unclear (20,21,39). Additionally, a recent study suggests that Fxr positively regulates Pepck expression and increases glucose output in rat primary hepatocytes (19). In the present study, basal hepatic glucose production was significantly decreased after a short-term fasting in FXR Ϫ/Ϫ mice, reinforcing the hypothesis of a role of Fxr in gluconeogenesis. In addition, the decrease in Pepck expression following refeeding was significantly more pronounced in FXR Ϫ/Ϫ mice. However, the molecular mechanism of this effect is unclear. Insulin-induced inhibition of Pepck gene expression was not altered in response to FXR activation in vitro, confirming that Fxr does not regulate hepatic insulin sensitivity. On the other hand, both Pgc-1␣ and Foxo1 mRNA expression levels paralleled the kinetics of Pepck expression during refeeding, and may thus mediate the observed genotypic differences. Whereas Pgc-1␣ gene expression correlates well with its transcriptional activity, Foxo1 is mostly regulated at a post-translational level, being phosphorylated by Akt in response to insulin stimulation (40). In accordance with unaltered hepatic insulin signaling in FXR Ϫ/Ϫ mice, the phosphorylation level of Foxo1 was not different in livers of refed FXR ϩ/ϩ and FXR Ϫ/Ϫ mice, suggesting that the observed differences in Foxo1 mRNA levels likely have a minor physiological impact. Another hypothesis may be that increased hepatic carbohydrate utilization in FXR Ϫ/Ϫ mice may interfere with PEPCK expression. Indeed, the glucose-dependent inhibition of Pepck gene expression (29) was decreased in response to Fxr activation in primary hepatocytes. Alternatively, an increased glycolytic flux per se may be involved because it decreases gluconeogenic substrates and directs pyruvate to the lipogenic pathway (41). Clearly, additional studies are needed to unravel the role of Fxr in gluconeogenesis.
In summary, we have demonstrated a new role for Fxr as a modulator of carbohydrate metabolism in the liver. Fxr appears to contribute to the coordinated regulation of the shift from hepatic glucose production to hepatic glucose utilization during the fasted to fed transition by interfering with carbohydrate-induced changes in gene expression. Because ingestion of a meal stimulates enterohepatic cycling of BAs, our findings suggest a novel physiological role of bile acids in the control of post-prandial carbohydrate metabolism via Fxr activation. In the post-prandial state, BAs secreted into bile and discharged into the intestine stimulate absorption of lipids that form chylomicrons in lymph. Carbohydrates, like glucose, are absorbed and reach the liver via the portal vein. BAs undergo enterohepatic recirculation and return to the hepatocyte where they activate, among other actions, Fxr. Activated Fxr subsequently interferes with glycolysis, via inhibition of LPK expression, to stimulate glycogen storage, and inhibits de novo lipogenesis via inhibition of Fas, Acc-1, and Spot14 gene expression. Finally, Fxr inhibits BA synthesis by inhibition of Cyp7a1 expression. Thus, Fxr participates in the metabolic adaptation of the hepatocyte in the postprandial state. MIT, mitochondria; Ac-CoA, acetyl coenzyme A; VLDL, very low density lipoprotein.