Nuclear bile acid signaling through the farnesoid X receptor

Abstract

Bile acids (BAs) are amphipathic molecules produced from cholesterol by the liver. Expelled from the gallbladder upon meal ingestion, BAs serve as fat solubilizers in the intestine. BAs are reabsorbed in the ileum and return via the portal vein to the liver where, together with nutrients, they provide signals to coordinate metabolic responses. BAs act on energy and metabolic homeostasis through the activation of membrane and nuclear receptors, among which the nuclear receptor farnesoid X receptor (FXR) is an important regulator of several metabolic pathways. Highly expressed in the liver and the small intestine, FXR contributes to BA effects on metabolism, inflammation and cell cycle control. The pharmacological modulation of its activity has emerged as a potential therapeutic strategy for liver and metabolic diseases. This review highlights recent advances regarding the mechanisms by which the BA sensor FXR contributes to global signaling effects of BAs, and how FXR activity may be regulated by nutrient-sensitive signaling pathways.

Introduction

Bile acid (BA) synthesis and the enterohepatic BA cycle

Bile acids are steroid molecules synthesized in hepatocytes from cholesterol. The BA biosynthetic pathway includes a number of successive enzymatic conversions involving multiple hydroxylations of cholesterol in the endoplasmic reticulum, and the mitochondrial shortening of its 17β-side chain from a C27 to a C24 molecule [1]. Primary products, resulting directly from this multi-enzymatic pathway and dependent on the key enzymes cytochrome P450 family members Cyp7A1, Cyp8B1 and Cyp27A1 [2], are the so-called primary BA cholic acid (CA) and chenodeoxycholic acid (CDCA) in humans, whereas an additional step converts CDCA into muricholic acid (α- and β-MCA) in mice. It is worth noting that ursodeoxycholic acid (UDCA), first identified in Chinese black bear bile, is also a primary BA in rodents [1, 3]. Prior to their secretion by hepatocytes into the canalicular lumen, BAs are conjugated to taurine or glycine at the C₂₄ position by bile acid coenzyme A: amino-acid N-acyl transferase (BAAT) to decrease their hydrophobicity and allow their secretion into the bile [4–6]. In humans, BAs can also be conjugated to sulfate mostly at C₃ by sulfotransferase 2A1 (SULT2A1) (Sult2A9 in mice), or glucuronidated by UDP glucuronosyltransferase family 2 (UGT2B4 and UGT2B7), a modification resulting in reduced toxic properties and increased urinary or fecal excretion [7–9]. Conjugated BAs are actively secreted into the canalicular lumen through the bile salt export pump (BSEP), stored and concentrated into the gallbladder. BA export promotes the simultaneous excretion of phospholipids and cholesterol into bile, through the ATP-binding cassette subfamily B member 4 (MDR3/ABCB4) and subfamily G member 5 and 8 (ABCG5/8) transporters, thus favoring cholesterol solubilization [10]. The human plasma bile acid pool is mainly composed of CDCA derivatives, especially glycochenodeoxycholic acid (GCDCA, 24 %), deoxycholic acid (DCA, 19 %), CA (16 %) and CDCA (15 %), whereas the mouse plasma bile acid pool is constituted of CA (24 %), ω-MCA (24 %), β-MCA (19 %) and DCA (7 %), thus revealing major differences between species. Glycine-conjugated BAs are the most abundant conjugated BAs in humans, whereas tauro-conjugated BAs are most abundant in rodents [11]. Modification of the BA pool composition may result in the accumulation of more hydrophobic BA species, causing severe hepatic alterations that may lead to liver failure and eventually extrahepatic tissue injury, as observed in cholestatic liver diseases [12].

Upon ingestion of a meal, gallbladder contraction, induced by cholecystokinin (CCK) released from the gut, discharges bile into the small intestine where it participates in lipid and lipid-soluble vitamin absorption by the intestine. BAs are also actively reabsorbed in the terminal ileum by the enterocyte transporter system composed of apical sodium-dependent bile acid transporter (ASBT/SLC10A2) and the heterodimeric organic solute transporter (OSTα/β). BAs escaping from reabsorption (~5 %) are converted into the secondary BAs DCA, lithocholic acid (LCA) and UDCA in humans and DCA, LCA, ωMCA, hyodeoxycholic acid (HDCA) and murideoxycholic acid (MDCA) in mouse. This conversion is catalyzed by the gut microbiota which may generate, according to its composition, dozens of secondary and tertiary BA metabolites [13, 14]. Secondary BAs can further be passively reabsorbed by the colonocytes or lost into feces. Reabsorbed primary and secondary BAs are then transported by the portal venous system and taken up by hepatocytes via the Na⁺-taurocholic acid cotransporting polypeptide (NTCP/SLC10A1), specifically involved in conjugated BA reabsorption, or by members of the organic anion transporting polypeptide family (OATPs/SLCOs [15]). In hepatocytes, BAs are then reconjugated and secreted back by the liver together with newly synthesized BAs which compensate for fecal loss. The BA pool, defined as the total amount of BAs circulating within this enterohepatic cycle, cycles between two and six times per day depending on the dietary regimen in humans.

Gut microbiota as a driver of BA pool composition

Recent reports highlighted the importance of gut microbiota as a driver of the BA pool composition. Germ-free mice display a decrease in BAs diversity, with an absence of secondary BAs. In addition, germ-free mice have an increased BA pool size, concentrated in the gallbladder and small intestine, but a decreased BA content in the cecum, colon, feces and serum [3]. This increased BA pool size results from a specific increase in βMCA conjugated and unconjugated metabolites rather than CA, and may be caused by enhanced intestinal BA reabsorption in the absence of gut microbiota [3, 16, 17]. Changes in microbiota due to probiotic, antioxidant or antibiotic treatments result in the modification of the BA pool composition [3, 18, 19]. In addition, dysbiosis has been described in diabetes and obesity [20], pathological conditions in which BA metabolism is altered [21–23]. Joyce and colleagues recently emphasized the importance of bacterial bile salt hydrolase (BSH) enzymes in determining BA pool composition. BSH enzymes are widespread enzymes in normal microbiota and are implicated in the deconjugation of glyco- and tauro-conjugated BAs. Monocolonization of germ-free mice with an Escherichia coli strain expressing a single bacterial BSH enzyme (BSH1) led to decreased plasma BAs and to a specific reduction of tauro-conjugated BAs associated with increased free CA, and α- and β-MCA content [24]. Thus, microbiotas control BA pool size and composition. Conversely, as a high concentration of DCA (0.5 mM) may inhibit bacterial overgrowth [25], BAs are also known to influence the structure and composition of gut microbiome, underlining the complex and reciprocal interaction between BA metabolism and microbiota composition [26, 27].

Roles of BAs in metabolic regulation

The role of BAs in metabolic regulation has been highlighted by the clinical use of BAs and oral non-absorbed BA sequestrants. First, BAs regulate their own synthesis and transport. UDCA, used as treatment for cholestatic manifestations such as primary biliary cirrhosis (PBC) or intrahepatic cholestasis of pregnancy (ICP) [28–33] and delays liver transplantation by increasing the rate of excretion of intracellular BA across liver cells and into the canaliculus [34], thus reducing toxic hydrophobic BA accumulation within liver cells [35]. In addition, the use of CDCA and UDCA in dissolving radiolucent cholesterol gallstone in gallstone disease patients emphasizes the role of BAs in cholesterol metabolism [36]. Lipid metabolism is impacted upon administration of BA sequestrants such as colesevelam, used to reduce cardiovascular disease (CVD) risk in prediabetic hypercholesterolemic patients [37], by increasing cholesterol efflux and lowering low-density lipoprotein cholesterol (LDL) plasma levels [38]. Finally, BA sequestration affects glucose metabolism, as pointed out by decreased blood glucose and glycated hemoglobin levels (HbA1c) in type 2 diabetic patients treated with colesevelam [39–41]. The use of BA sequestrants and BAs as treatment for various human metabolic dysfunctions emphasizes the pleiotropic role of BAs in the control of diverse metabolic pathways.

Part I: FXR, a BA receptor with pleiotropic biological functions

In 1999, the BA field witnessed a major leap forward with the discovery of a BA receptor: the nuclear receptor farnesoid X receptor (FXR) [42, 43]. Further studies revealed that BAs are also endogenous ligands for other nuclear receptors (NRs) including pregnane X receptor (PXR) [44], vitamin D receptor (VDR) [45] and the bile acid membrane receptor (TGR5/GPBAR1) [46]. The role of these NRs in metabolic regulation has been discussed previously [47, 48].

FXR structure

Farnesoid X receptor shares the common nuclear receptor structure composed of a central DNA-binding domain (DBD) close to the N-terminal ligand-independent activation function 1 domain (AF1) and a C-terminal ligand-binding domain (LBD) encompassing the ligand-dependent activation function 2 domain (AF2). The LBD and DBD are separated by a hinge region, which is considered as a flexible link between these two regions [49]. The FXR gene (also called NR1H4) is evolutionarily conserved showing high sequence similarities between several species such as Syrian hamster, mouse, rat and humans [50]. The murine and human FXR gene contains 10 introns and 11 exons [50, 51].

Using alternative promoters and splicing, four FXR isoforms have been identified (FXRα1, α2, α3 and α4 [50]) which display tissue- and species-specific expression. In humans, FXRα1 and FXRα2 are mainly expressed in the liver and in adrenals, whereas the four isoforms are present in the small intestine. The FXRα3 and α4 isoforms predominate in the colon and kidney. Contrary to humans, mouse liver and small intestine express the four isoforms, whereas mouse kidney and stomach express only the FXRα3 and α4 isoforms [50, 52]. FXR mRNA has been detected in human and mouse pancreas and the FXR protein also in reproductive tissues and human vascular smooth muscle cells [53–55]. Interestingly, FXR isoforms show distinct responsiveness to agonist-mediated activation. In transient-transfection experiments, FXRα2 is the most potent isoform to promote ligand-induced FXR target-gene expression [51, 52, 56–58]. This differential isoform efficiency is moreover gene dependent: the OSTβ gene is similarly activated by exogenous human FXR isoforms in the presence of an agonist, contrary to the small heterodimer partner (SHP) and BSEP. In addition, the isoform specificity is variable according to species, as only the hBSEP promoter is differentially transactivated by FXRα2 and FXRα1 isoforms, in contrast to rat and mouse Bsep [59].

The phenotype of FXR-deficient mice

Studies with Fxr-null (Fxr^−/−) mice have convincingly demonstrated the physiological role of this nuclear receptor. Three different models of Fxr-null mice have been used in the literature. The first and most used model results from a deletion of a genomic fragment encoding a large portion of the LBD and the 3′UTR region of the Fxr mRNA [60]. The phenotype of these mice has been extensively studied and characterized by defective BA transport identified by elevated plasma BA concentrations, a decreased BA fecal loss and a high susceptibility to cholestasis upon CA feeding [60]. Fxr ^−/− mice have a reduced body and adipose tissue mass [61]. In addition, Fxr ^−/− mice fed a 1 % cholesterol-enriched diet showed an alteration of cholesterol and lipid metabolism, with increased hepatic triglycerides and cholesterol content and a pro-atherogenic serum lipoprotein profile [60, 62]. In line with this altered lipoprotein profile, Fxr ^−/− ApoE ^−/− mice develop more extensive atherosclerotic aortic plaques and display more profound lipid abnormalities than ApoE ^−/− mice [63]. However, a study combining Fxr and Ldlr deficiency surprisingly detected less atherosclerotic lesions in Fxr ^−/− Ldlr ^−/− compared to Ldlr ^−/− mice in male, but not in female mice, calling into question the beneficial effect of FXR activity on atherosclerosis [64]. In addition, high fat diet-fed Fxr ^−/− Ldlr ^−/− male mice develop hepatic lesions mimicking human NASH [65].

Fxr ^−/− mice display an altered glucose homeostasis. A transient hypoglycemia upon short-term fasting is associated with reduced hepatic glycogen content and decreased hepatic glucose production [66], whereas fed mice are hyperglycemic and display impaired insulin signaling in skeletal muscle and liver [67, 68]. However, a beneficial effect of Fxr deficiency on glucose parameters in leptin-deficient (ob/ob) mice has been reported [61, 69], suggesting that metabolic conditions might affect FXR activities.

Male and female Fxr ^−/− mice spontaneously develop liver tumors at 14 months of age [69–72] and show defective liver regeneration after partial hepatectomy [73] or upon CCl₄-induced liver injury [74]. Fxr deficiency in male and female APC^min mice leads to intestinal tumor development [75], hinting at a role for FXR in cellular proliferation.

The second model of Fxr ^−/− mice has been generated by Deltagen Inc. and presents a deletion of a part of Fxr exon 2, thus affecting the DNA-binding domain integrity [76]. This model has been less extensively studied, but displays an increased body mass, increased liver mass, increased plasma and biliary bile salt concentration, increased BA fecal loss and increased bile acid pool size, contrary to the first model described above [76]. In agreement with the first Fxr ^−/− model, this model shows reduced endogenous glucose production, delayed intestinal glucose absorption and reduced hepatic and peripheral insulin sensitivity [77]. No liver function alterations as measured by ALT, AST or bilirubin plasma levels have been reported. Recently, a third model of Fxr ^−/− mice has been generated by deletion of Fxr exon 9, thereby removing part of the FXR LBD [78]. This mouse model has a decreased body mass associated with increased energy expenditure at the age of 21 weeks. At 30 weeks, these mice show decreased body fat mass and altered bone mineral density when compared with their wild-type littermates. These aging Fxr ^−/− mice spontaneously develop cholestasis and exhibit liver damage resembling NASH. In addition, and contradictory to previous models of Fxr deficiency, they display improved glucose tolerance and lower HOMA-IR index [78]. To sum up, these different models present clear and unexplained discrepancies in their phenotype which might be due to KO-induced artefacts, such as the possible occurence of a truncated protein and removal of non-coding RNAs, but also to experimental conditions of the study (age, diet, treatment, etc.). A formal comparative evaluation of the phenotype of these three different Fxr ^−/− strains would be of particular interest to identify true phenotypic differences. Furthermore, the use of tissue-specific knockout animal models is mandatory to dissect the actual implication of FXR in different tissues. Despite these discrepancies, FXR has undoubtedly a broad role in whole body homeostasis, emphasizing the importance of understanding its molecular and mechanistic properties.

FXR function in bile acid homeostasis

Bile acid synthesis and transport need to be tightly regulated because of their high cellular toxicity and FXR plays a critical role in this process. In the liver, FXR activation increases Shp gene expression [79], a known repressor of Cyp7A1 expression. SHP represses liver receptor homolog 1 (LRH-1) transcriptional activity through direct interaction with LRH-1 and the recruitment of corepressors [80–82]. CYP8B1 gene regulation by FXR stems from two opposite mechanisms in humans, but only one in rats. The common mechanism in both species involves an SHP-dependent inhibition of LRH1 activity and the subsequent decrease of CYP8B1 expression. In human hepatocytes, a second FXR-dependent mechanism coexists, leading to the FXR-mediated transactivation of the CYP8B1 promoter through increased hepatocyte nuclear factor 4α (HNF4α) transcriptional activity. These two opposite FXR-dependent regulations of CYP8B1 counteract each other, explaining probably why CDCA treatment in humans strongly represses CYP7A1 expression but has little effect on CYP8B1 expression [83–85], whereas in the mouse CA treatment represses Cyp8B1 and Cyp7A1 expression to a similar extent [60]. The differential activation of the human CYP7A1 and CYP8B1 genes may modulate BA pool hydrophobicity, as CYP8B1 is required for CA synthesis [86]. FXR also represses BA uptake by the liver by repressing Ntcp through an SHP-dependent mechanism [87, 88]. FXR increases the efflux of BAs from the liver to the canalicular lumen through upregulation of ABCB11/BSEP and ABCB4/MDR3 gene expression and BA efflux from the liver to the portal vein by increasing OSTα and OSTβ expression [85, 89]. Finally, FXR also regulates key enzymes involved in BA conjugation and detoxification, such as UGTs, SULTs and glutathione S-transferase [90].

In the intestine, FXR induces the expression of a fibroblast growth factor family member, FGF15 (in mouse) and FGF19 (in humans) in vitro and in vivo [91]. FGF15/19 expression and secretion by enterocytes contributes to BA synthesis regulation, through the activation of the hepatic fibroblast growth factor receptor 4 (FGFR4) membrane receptor. FGF19 is induced upon FXR activation in primary human hepatocytes revealing, in addition to endocrine signaling from the intestine, a possible hepatic autocrine or paracrine FGF19 signaling pathway [92]. Activation of FGFR4 and of its immediate target, fibroblast growth factor receptor substrate 2α (FRS2α), leads to the activation of the JNK and ERK pathways, resulting in decreased Cyp7A1 gene expression [93]. Intestinal deletion of the Fxr gene prevents Cyp7A1 repression caused by an FXR agonist in normal physiological conditions [94–96]. These observations therefore highlight the major regulatory role of intestinal FXR (iFXR) in regulating hepatic BA synthesis. Interestingly, numerous links have been made between the iFXR/FGF15 pathway and hepatic FXR activity (lFXR) on BA synthesis. Both in Fgf15 and Fgfr4 KO mice, Cyp7a1 expression is not inhibited after oral administration of an FXR agonist [94, 97]. Furthermore, a component of hepatic FGFR4 signaling, cytoplasmic tyrosine phosphatase SHP2, is necessary for the induction of Shp expression after FXR agonist treatment [98] and hepatic SHP is required for FGF15-mediated repression of Cyp7a1 [97, 99]. This suggests that FGFR4 signaling is necessary for lFXR-mediated repression of BA synthesis, making iFXR a strong driver of liver BA synthesis. In addition to BA synthesis, FXR also regulates BA transepithelial transport in the intestine, by regulating BA absorption and export to the portal system. Indeed, FXR inhibits intestinal Asbt expression responsible for BA reabsorption in mice, but not in rats [100], and increases fatty acid binding protein 6 (Fabp6/Ibabp) transporter expression, which shuttles BAs within the enterocytes [101]. Expression of the heterodimeric transporter OSTα/β, the major player of BA transport from enterocyte to the portal vein, is also regulated by FXR [102]. Thus, by reducing BA synthesis, favoring BA reabsorption in the intestine and promoting export and biotransformation in hepatocytes, FXR reduces metabolic requirements for BA synthesis and protects against BA toxicity (Fig. 1). In line with this protective effect, FXR activation leads to improvement of cholestasis in models of extra- and intrahepatic cholestasis by decreasing liver injury [103–106]. Paradoxically, Fxr ^−/− mice have less liver injury after bile duct ligation, an aggressive surgical model of cholestasis, probably due to the downregulation of the canalicular transporter Bsep expression, thereby reducing biliary pressure. By generating a more hydrophilic bile acid pool and enhancing phase I detoxification of BAs through polyhydroxylation by cytochrome P450 member 3A11 (CYP3A11), Fxr^−/− mice display increased urinary BA excretion [107–109]. A decreased FXR activity can also be generated upon intestinal Sirtuin 1 (Sirt1) knockout which, as a result, inhibits HNF1α binding to the Fxr promoter [110]. This suggests that aging, which is associated with decreased Sirt1 expression, may also affect the FXR signaling pathway [111]. Cholestatic conditions such as progressive familial intrahepatic cholestasis type 1 (PFIC1) and intrahepatic cholestasis of pregnancy (ICP) have been also associated with decreased FXR activity and/or expression [112–114], suggesting that FXR activation may be a promising therapeutic option in the treatment of intrahepatic cholestasis.

Fig. 1.

FXR controls the enterohepatic cycle of bile acids. FXR activation by BAs protects the liver against BAs’ accumulation through a multistep mechanism involving a communication between the liver and intestine. In the intestine, FXR activation by BAs reduces bile acid uptake from the intestinal lumen and increases BAs’ cellular trafficking and export into the portal circulation. In addition, FXR activation in the intestine (distal ileum) increases FGF15 synthesis and export into the portal system. FGF15 through its membrane receptor FGFR4 decreases hepatic bile acid biosynthesis through SHP-dependent and SHP-independent mechanisms, mostly affecting Cyp7a1 activity, and hence CDCA synthesis. Hepatic FXR activation increases bile acid efflux from hepatocytes through the regulation of transporters’ expression (OSTα/β, BSEP and MDR3). FXR represses bile acid uptake into the hepatocyte through the repression of the NTCP transporter. This figure has been made using Servier Medical Art according to the license Creative Commons Attribution 3.0 France

Role of FXR in lipoprotein metabolism

In addition to its role in BA metabolism, FXR also regulates lipoprotein metabolism. As mentioned above with respect to the Fxr-deficient mice phenotype, the implication of FXR in HDL/LDL metabolism is controversial depending on the genotype background. In vitro, treatment of immortalized human hepatocytes with CDCA or the synthetic FXR agonist GW4064 reduces proprotein convertase subtilisin/kexin type 9 (PCSK9) expression, a natural inhibitor of LDL-R activity [115]. Although suggestive of a plasma LDL-C(holesterol) lowering capacity of activated FXR, in vivo treatment with CDCA did not alter or even increase LDL-C plasma concentrations upon short or prolonged treatment, respectively [116, 117]. A recent clinical trial in NAFLD diabetic patients reported increased LDL-C plasma levels after administration of the semi-synthetic FXR agonist INT747 during 6 weeks [118]. FXR also acts on reverse-cholesterol transport and has a profound effect on HDL metabolism, by repressing apolipoprotein A1 (ApoAI) expression and increasing cholesteryl ester transfer protein (CETP) expression [119]. Moreover, FXR regulates hepatic HDL-C uptake and HDL remodeling, by modulating phospholipid transfer protein (PLTP) and hepatic lipase (HL) [120, 121]. In a clinical study, CDCA decreased HDL-cholesterol levels [122]. Administration of the FXR agonist PX20606 decreases plasma cholesterol and reduces aortic plaque formation in CETP-transgenic LDLR^−/− mice and in cynomolgus monkey, probably through increased HDL-C subclass 2 uptake by the liver resulting from increased hepatic SR-BI expression [123]. This counterintuitive observation in light of the plasma lipoprotein profile, but concordant with the pro-atherogenic profile of Fxr-deficient mice, makes FXR a potential anti-atherogenic target. More studies are needed to determine whether prolonged FXR activation limits atherosclerotic risk in controlled clinical studies.

FXR function in lipid metabolism and inflammation

The farnesoid X receptor regulates lipogenesis and triglyceride synthesis by repressing sterol regulatory binding protein 1c (Srebp1c) gene transcription, a key transcription factor stimulating the transcription of lipogenic genes, fatty acid desaturases and elongases, through an SHP-mediated inhibition of LRH-1/LXR transactivation [124]. A similar mechanism operates in the repression of microsomal transfer protein (Mtp) expression, which controls triglyceride assembly with apolipoprotein B in very low-density lipoproteins (VLDL) [125]. In agreement with its role in modulating hepatic triglyceride content, FXR is implicated in the progression of non-alcoholic fatty liver disease (NAFLD), a spectrum of liver disorders ranging from benign hepatic steatosis to steatohepatitis. FXR activation indeed protects against steatosis in rat and mice [126, 127]. FXR activation by the synthetic agonist WAY-362450 attenuates liver inflammation and fibrosis in mice fed a methionine and choline-deficient diet [128]. Moreover, FXR also regulates orosomucoid-1 (AGP) acute-phase protein expression, which is secreted in plasma upon liver inflammation, and hepatic FXR may modulate adipose tissue inflammation in mice [129].

FXR function in glucose homeostasis

Fxr-deficient mice display mild systemic glucose intolerance with hepatic and peripheral insulin resistance [66–68]. Moreover, Fxr expression is reduced in the liver of diabetic animal models [130] and its expression can be increased by insulin [130]. FXR activation by intraperitoneal injection of GW4064 promotes insulin sensitivity in the liver and skeletal muscle in ob/ob mice [66, 67, 130]. By contrast, long-term high fat diet feeding and concomitant treatment with GW4064 reduced insulin sensitivity, decreased energy expenditure and promoted weight gain [131]. These contradictory results may result from differences in experimental design such as the use of two different administration modes for GW4064. Given the poor bioavailability of this compound, the activation of peripheral FXR by orally administrated GW4064 is questionable. Indeed, Prawitt et al. [61] demonstrated that hepatic FXR deficiency does not protect against diet-induced obesity and insulin resistance, suggesting the involvement of peripheral FXR in metabolic homeostasis. Another point of divergence is the use of different mouse models of obesity. The importance of the genetic background in assessing FXR functions is emphasized by two different studies. Fxr deficiency in the obesogen ob/ob background improves glucose homeostasis and prevents weight gain [61, 69], whereas in the pro-atherogenic Ldlr ^−/− background, Fxr deficiency alters neither insulin sensitivity nor glucose tolerance [69]. Further studies with tissue-specific FXR knockouts are needed to decipher FXR-dependent mechanisms involved in peripheral metabolic effects.

FXR regulates different pathways of hepatic glucose metabolism by increasing glycogen synthesis [68, 132] and decreasing glycolysis by downregulation of L-type pyruvate kinase gene expression (LPK) [133]. Interestingly, the effect of FXR activation on gluconeogenesis changes according to the physiological context. In fed mice, FXR activation represses the expression of phosphoenolpyruvate carboxykinase 1 (Pepck) and glucose-6-phosphatase (G6pase) genes, whereas in fasted mice, FXR activation increased their expression via a mechanism involving GR [134]. The evaluation of FXR differential activity in fasted versus fed mice needs to be mechanistically studied by transcriptomic and genomic analysis. In addition to its hepatotropic effects, FXR has a general protective role in β-cell function and notably glucose-induced insulin secretion, thereby affecting glucose homeostasis [53, 135–137].

FXR functions in cellular proliferation

In addition to its metabolic phenotype, aging Fxr ^−/− mice spontaneously develop hepatocellular carcinoma (HCC), characterized by the upregulation of genes involved in inflammation and cell cycle control [70, 138]. The mechanism underlying the oncosuppressive property of FXR in the liver remains ill defined, but could be partially explained by an SHP-mediated increase of apoptosis, repression of BA synthesis and signal transducer and activator of transcription 3 (STAT3) inactivation through increased suppressor of cytokine signaling 3 (SOCS3) expression [139–141]. In addition, Deuschle et al. [142] identified the tumor suppressor N-myc downstream-regulated gene 2 (NDRG2) as a novel FXR target gene implicated in cell cycle control. In the liver, the tumorigenic effect of Fxr deficiency seems to stem from increased inflammation and fibrosis, and the Fxr ^−/− phenotype recapitulates the progression of human HCC [72]. Furthermore, treatment of human HCC cell lines by a combination of an acyclic retinoid [targeting retinoid X receptor α (RXRα)] and GW4064 inhibits growth and causes apoptosis [143]. Moreover, lentiviral-mediated FXR overexpression represses liver cancer cell proliferation and tumor growth in nude mice [144]. In agreement with these in vivo results, FXR expression is decreased and inversely correlated to disease progression in human HCC [72]. Interestingly, Vaquero et al. [145, 146] reported that FXR activation stimulates the chemoprotection of hepatocytes against genotoxic compounds, thus reducing the response of tumor cells to pharmacologically induced cellular toxicity in vitro. This mechanism possibly involves increased expression of genes implicated in drug efflux, DNA repair and cell survival. Recently, Degirolamo et al. [147] showed that intestinal Fxr expression protects Fxr ^−/− mice from spontaneous HCC development by decreasing BA overload, hepatic inflammation and lesions, emphasizing the importance of the FGF15 signaling pathway in HCC progression. FGF19-based therapy has been proposed as a potential therapeutic option in hepatocellular carcinoma treatment. However, FGF19 also appears to play a role in tumorigenesis by favoring hepatocyte proliferation, raising a potential safety concern. Recently, a specific variant of FGF19 displaying a selective activity on the regulation of bile acid metabolism without exhibiting adverse effects on tumorigenesis has been described, pointing to the therapeutic potential of FGF19 as a hepatic oncosuppressor [148, 149].

FXR expression is reduced in human colorectal cancers [150, 151], and accumulating evidence reveals a link between intestinal inflammation and tumorigenesis. The role of FXR in intestinal inflammation is underlined by the ability of activated FXR to decrease the production of proinflammatory cytokines such as IL1β, IL2, IL6, TNFα and IFNγ [152]. In addition, iFXR activation induces the transcription of genes involved in enteroprotection and prevention of bacterial translocation in the intestinal tract [153]. In addition, FXR deficiency increased the number and size of tumors in experimental models of intestinal cancer [75]. In line with the oncosuppressive activity of FXR, a chimeric, constitutively active FXR (VP16 AD-FXR) led to a markedly increased apoptosis and reduction of tumor growth in a xenograft mouse model [154]. FXR depletion in 8-week-old APC^+/− mice increases adenoma-like lesions and aberrant crypt foci displaying extensive intestinal inflammation, suggesting the implication of FXR in all cancer progression stages [154]. In contrast, FXR upregulation is correlated to the induction of carcinogenesis and metastasis in the stomach and esophagus, via a mechanism involving probably the activation of the caudal-related homeobox 2 (Cdx2) gene, a transcription factor implicated in cancer development [155]. Consequently, increasing FXR signaling may be an interesting strategy in the management of inflammation and possibly carcinogenesis in the liver or the intestine, whereas such beneficial effects in other cancers remain to be firmly established.

FXR function in liver regeneration

Farnesoid X receptor deficiency strongly inhibits liver growth and increases mortality after a 70 % partial hepatectomy (PH) [73]. Fxr-deficient mice exhibit increased hepatocyte apoptosis after CCl₄ treatment, a chemical compound promoting hepatic fibrosis, by delaying activation of the STAT3 signaling pathway. Mirroring these effects, FXR expression promotes liver regeneration [73] and stimulates liver repair after CCl₄-induced liver injury [74]. These data suggest a dual role of FXR in promoting liver regeneration and protecting from hepatocyte cell death. The liver regenerative action of FXR stems from its ability to activate the forkhead box M1 transcription factor (FOXM1b), a key cell cycle regulator controlling the G1/S and G2/M transitions [73, 156]. Interestingly, overexpression of SIRT1 in mice and the subsequent reduction in FXR activity resulted in impaired hepatocyte proliferation after PH [157]. Strikingly, intestinal-specific FXR deficiency results in an impaired liver regeneration process, indicating that iFXR participates in hepatic regeneration. Exogenous delivery of FGF15 in iFXR-deficient mice restores liver repair capacity, showing that the iFXR effect is mediated through the endocrine FGF15 pathway independently of FoxM1b activation and may involve decreased BA synthesis and FGF15 mitogenic activity [158].

Part II: mechanism of transcriptional regulation by FXR

To date, the vast majority of FXR biological activities has been ascribed to its ability to interact with cis-acting DNA sequences. The molecular basis for these activities are described below.

DNA-binding properties of FXR

Farnesoid X receptor regulates gene expression by binding directly to DNA in vitro as a heterodimer with RXR or as a monomer [159]. The FXR DNA-binding domain confers specific recognition of different DNA motifs called FXR response element (FXRE), for which genome-wide studies have revealed that the most prominently recognized DNA sequence is an inverted repeat of the consensus sequence AGGTCA separated by one nucleotide (IR1) in mouse liver and intestine and in the human hepatoma HepG2 cell line [160–162]. In addition, FXR has been reported to bind other response elements with various geometries [161, 163–166] such as IR0, everted repeats separated by two or eight nucleotides (ER2–8) or direct repeats separated by one, four or five nucleotides (DR1–4–5). Interestingly, only 11 % of FXR binding sites are shared between mouse liver and intestine, revealing a clear tissue specificity of the FXR cistrome, which emphasizes its tissue-specific functions [161]. Indeed, gene ontology analysis of genes associated with FXR binding sites in the liver and intestine revealed tissue-specific pathway/function enrichment, according to which iFXR binding sites are strikingly associated with genes involved in catalytic reactions, oxidoreductase activity, monooxygenase activity, cofactor binding and the complement and coagulation cascade. In contrast, lFXR binding sites are in close vicinity of genes involved in cholesterol, lipid and glucose homeostasis which coincides with FXR biological activities deduced from studies in Fxr ^−/− mice [161]. In addition, motif analysis reveals that FXR preferentially binds in intestinal cells to ER2 motifs in addition to the IR1, underlining possible tissue-specific motif recognition by FXR. How this specificity is achieved is not yet known [161]. FXR thus demonstrates a broad diversity in bound DNA motifs, which could emphasize the high variety of mechanisms by which it regulates the expression of its target genes.

Ligand-binding properties of FXR

As other NR LBDs, the FXR LBD is constituted of 12 α-helices, forming a hydrophobic pocket allowing binding of ligands such as endogenous BAs and synthetic BA analogs. The orientation of the BA steroid scaffold in the ligand-binding pocket (LBP) is opposite to that found observed in other steroid NRs, with the carboxylic moiety being buried inside the LBP [167]. As mentioned above, BAs exhibit distinct potencies to activate FXR, with CDCA being the most active natural human FXR activator (EC₅₀ = 20 µM in a mammalian 2-hybrid system) followed by DCA and LCA [42, 43, 52]. BAs have an “α” side on which hydroxyls are exposed at the 3α, 7α and 12α positions. The hydrophobic “β” side is similar between primary and secondary BAs. Relative affinities of BAs for FXR are dictated by their specific pattern of axial hydroxyl groups at the 7 and 12 positions [167]. Intriguingly, the potency of BAs to activate FXR is species specific and may involve cell-specific expression of membrane transporters and/or of transcriptional modulators. In HepG2 cells, CA treatment fails to activate FXR [52]. A possible explanation is the absence of the NTCP transporter in those cell lines [168]. In addition, human intestinal cell lines are sensitive to CDCA, but not to its conjugated derivatives GCDCA and TCDCA at similar intracellular concentrations, whereas FXR-target genes are activated by these CDCA conjugates in human hepatocyte cell lines, suggesting cell type-specific FXR activation by conjugated BAs not related to BA uptake properties [52].

Due to the poor selectivity of endogenous BAs hampering the dissection of FXR-dependent from FXR-independent pathways, synthetic and semi-synthetic ligands have been developed with improved selectivity for FXR. The semi-synthetic ligand INT747 (6-ECDCA) is derived from the structure of CDCA and is the most advanced in clinical trials (phase III trial, Primary Biliary Cirrhosis, Intercept Pharmaceuticals, NCT01473524). The non-steroidal synthetic agonist GW4064 is extensively used for in vitro and in vivo evaluation of FXR activity; it activates intestinal FXR, but displays poor bioavailability. As mentioned above, GW4064 has opposite metabolic effects in short- versus long-term treatment, which might stem from the different administration routes and preferential activation of liver (intraperitoneal) or intestinal FXR (oral), thereby inducing bias and discrepancies between studies [66, 68, 131]. Novel, less characterized agonists such as Px-102, WAY-362450 and fexaramine have been reported [169]. Surprisingly, fexaramine and GW4064 induce different gene expression profiles in primary mouse hepatocytes, suggesting that, in line with other NRs, FXR has distinct transcriptional properties when bound to structurally different ligands [170]. However, these differences can also be partly explained by the relaxed specificity of GW4064 which activates the estrogen-related receptor (ERR) isoforms, calling for a careful pharmacological evaluation of newly synthesized ligands [171].

In addition to natural or synthetic FXR agonists, FXR antagonists have been discovered. The secondary murine BA tauro-β muricholic acid (T-βMCA) has antagonistic properties [3]. Interestingly, the gut microbiotas modulate the FXR signaling pathway through modulation of T-βMCA conversion [18, 19]. Germ-free mice show increased T-βMCA concentrations, leading to the inhibition of iFXR. This in turn leads to increased hepatic BA synthesis and a larger total BA pool size as a result of decreased FGF15 signaling. Despite the increase in T-βMCA concentrations in the liver of germ-free mice, lFXR activity is not regulated as a function of gut microbiotas, suggesting a lack of sensitivity of lFXR to T-βMCA [3]. T-βMCA is however not produced in humans and further studies are needed to evaluate the presence of a “T-βMCA-like” antagonist activity in the human BA pool. Guggulsterone, a natural extract from the guggul tree, has FXR antagonistic activity on several genes, but surprisingly enhances FXR-induced Bsep gene expression [172]. This selective antagonism may be explained by the poor selectivity of guggulsterone which interacts also with other NRs [173]. Other antagonists such as 15d-PGJ2, suvanine or theonellasterol have been described in structural and screening studies [174–177], but remain poorly characterized with respect to their in vivo and in vitro activities.

FXR transactivation mechanisms

Nuclear receptors activate gene transcription through three currently described mechanisms: simple transactivation, composite transactivation and tethering transactivation, in analogy to other NR such as glucocorticoid receptor (GR) [178]. Simple transactivation is the “classical” way by which an NR activates transcription. Monomeric, homodimeric or RXR heterodimeric NRs directly bind to nuclear receptor response elements (NRE), while ligand (agonist) binding triggers a multistep process initiated by a conformational change of the LBD. This allows the dismissal of a corepressor complex and the subsequent recruitment of coactivating complexes. These are essentially, but not exclusively, made of histone writers which alter chromatin structure through post-translational histone modifications, promoting access of the general transcription machinery to proximal promoters or enhancers. Composite transactivation is defined as the direct binding of the NR to an NRE and its synergy with other DNA-bound NRs or other transcription factors (TF) present in close proximity. The tethering transactivation mechanism requires indirect binding of the NR to DNA through protein–protein interaction and has been described for several NRs, including GR which synergizes with DNA-bound signal transducer and activator of transcription 5 (STAT5) [179] or COUP-transcription factor 2 (COUP-TFII) [180]. RXR/FXR heterodimers have often been shown to act through a simple transactivation mechanism by the use of reporter-gene assays [89]. However, monomeric transactivation has been suggested for UGT2B4 and the glucose transporter (GLUT4) [181, 182]. Composite transactivation has equally been proposed for FXR. Genome-wide analyses indeed showed co-binding of FXR and of the orphan nuclear receptor LRH-1, both cooperating in activating hepatic lipid metabolism [160, 183]. Moreover, the existence of a synergy between FXR and HNF4α has been suggested, as FXR activation by the synthetic agonist GW4064 in vivo increases HNF4α binding in the vicinity of an FXRE [184]. This synergy is restricted to a particular set of regions and associated genes implicated in the complement and coagulation cascade. These two transcription factors were suggested to have cooperative, compensatory or independent effects on transcriptional regulations depending of the target, revealing a complex relationship between these two factors [184].

FXR regulates transcription through the recruitment of coregulators able to modulate chromatin properties. This occurs through the recognition of a hydrophobic groove at the surface of the NR generated upon agonist binding and ensuing LBD structural transitions. This hydrophobic groove accommodates LXXLL motifs found in single or multiple copies in coactivating proteins. Unexpectedly, the FXR LBD harbors two distinct docking sites for LXXLL motifs, the first one being responsible for coactivator recruitment and the second generating a binding interface for coactivators with more than one LXXLL motif, thereby increasing their relative affinity for FXR [167]. Steroid receptor coactivator 1 (SRC1) [185, 186], peroxisome proliferator-activated receptor gamma coactivator 1α (PGC1α) [187], thyroid hormone receptor-associated protein (TRAP) [188], SET domain-containing lysine transferase (SET7/9) [189] as well as histone methyltransferases such as PRMT1 [190], CARM1 [191], ASCOM (MLL3/MLL4) [192, 193], histone acetylases such as P300/CBP [186, 194] and chromatin remodelers of the Swi/Snf ATPases family (BRG1, [195]) have been identified as FXR coactivators (Table 1). Corepressors have also been identified, including the silencing mediator for retinoid and thyroid hormone receptors (SMRT) and the nuclear receptor corepressor 1 (NCoR1) complexes [196], Ku autoantigen proteins [197], the nuclear receptor 0B1 (DAX-1/NR0B1) [198] and the histone deacetylase SIRT1 [199] (Table 2). The composition of the FXR-tethered bound cofactor complex appears gene specific. SHP activation by FXR is highly dependent on PGC1α and SIRT1, whereas BSEP and FGF19 are activated independent of PGC1α and SIRT1 in HepG2 cells [200]. FXR can also regulate genes from more distal regulatory regions or enhancers. A genome-wide study revealed that FXR binds often to intergenic and intronic regions, as 41 % of FXR binding sites in mouse liver are located at more than 10 kb upstream of a Refseq gene [161]) and can possibly induce chromatin looping to connect distant binding sites and the promoter-bound transcriptional machinery. A head-to-tail chromatin looping between an FXRE located in the Shp/Nr0b2 promoter and another found in a downstream enhancer has been demonstrated [79], and similar looping processes have been proposed for the Cyp8b1 and Gr gene regulation by FXR [82, 201]. Emerging Hi-C technologies (three-dimensional analysis of genome architecture) might elucidate the functions of FXR in distal chromatin looping and whether gene activation by enhancer-bound FXR is also mediated through a cohesin-dependent mechanism as described for the estrogen receptor [202]. Undoubtedly, these various mechanisms implicating different FXR partners establish another layer of complexity in FXR transcriptional regulation.

Table 1.

List of FXR coactivators, their functions and the context (cell type, gene) of their association with FXR

Coactivator	Function	Substrate	Model	Genes	Reference
P300	Histone acetyl transferase (HAT)	H3K9/K14	HepG2; IHH; mouse liver	SHP TXNIP	[133, 194, 227]
CARM1	Arginine methyltransferase	H3R17	HepG2	BSEP	[191]
PRMT1	Arginine methyltransferase	H4R3	HepG2	SHP BSEP	[190]
MLL3	Lysine methyltransferase	H3K4	HepG2; mouse liver	SHP BSEP Mrp2	[193]
MLL4	Lysine methyltransferase	H3K4	HepG2	SHP BSEP	[193]
ASC2 (NCOA6)	Recruitment of MML3/4		HepG2; mouse liver	SHP BSEP KNG1 Mrp2	[192, 193, 200]
Brg1	ATPase chromatin remodeler		HepG2; mouse liver	SHP	[195]
TRRAP	Recruitment of HAT		HepG2	Reporter assay with FXRE	[188]
Set7/9	Non-histone methyltransferase	FXR	Huh7	SHP BSEP	[189]
SRC1			HepG2	Two-hybrid assay SHP BSEP OSTα	[198, 228, 229]
SRC2			HepG2	KNG1	[200]
PGC1α	Recruitment of other cofactors		HepG2	IBABP BSEP	[187]
SIRT1	Deacetylase	PGC1α, FXR	Mouse liver and intestine	Shp Ostβ	[220]

Table 2.

List of FXR corepressors, their functions and the context (cell type, gene) of their association with FXR

Corepressor	Function	Substrate	Model	Genes	Reference
SIRT1	Deacetylase	H3K9/14, FXR	HepG2; mouse liver	SHP	[199]
DAX1	NCOR recruitment		HepG2	SHP BSEP	[198]
KU80 and 70	DNA-dependent protein kinases	unknown	HepG2	BSEP	[197]
SMRT			HEK; IHH	Two-hybrid assay TXNIP	[133, 177]
NCOR			HEK	Two-hybrid assay	[177]

FXR transrepression mechanisms

Similarly to transactivation processes, several mechanisms have been described for transrepression of target genes by NRs, including simple transrepression, competitive transrepression, tethering transrepression and coactivator squelching. Like simple transactivation, simple transrepression involves direct binding of NR as a monomer, homo- or heterodimer to a negative nuclear receptor response element (nNRE). Tethering transrepression requires indirect binding of the NR to DNA via interaction with another TF, thereby preventing coactivator recruitment or stabilizing corepressor binding to this TF. This scenario has been abundantly documented to explain the inhibitory effect of several NRs such as GR, peroxisome proliferator-activated receptor γ (PPARγ or retinoid receptors on the activator protein 1 (AP-1) and nuclear factor of kappa light polypeptide gene enhancer in B-cells (NF-kappaB) signaling pathways. Coactivator squelching implies the competition for cofactors between the ligand-activated NRs and other TF.

Transrepression by FXR has been reported for the regulation of human APOAI in human ApoA1 transgenic mice treated with taurocholic acid (TCA), which occurs independently of RXR upon binding to a negative FXRE [203]. These results, raising questions about the potential atherogenic properties of FXR agonists as they lower HDL-C, have however been recently challenged by Gardès et al. [204], who suggested that the observed decrease is unlikely to be an FXR-dependent process, as synthetic FXR agonists do not mimic the effect of TCA in vivo. Competitive transrepression has been described at the APOA/Lp(A) promoter, as ligand-activated FXR competes for binding to a HNF4α-bound DR1 site in its promoter. In addition, an increased hepatic BA concentration caused by bile duct ligation in transgenic hAPOA mice decreased APOA expression, suggesting that BAs repress hAPOA expression in vivo in an FXR-dependent manner [205]. Transrepression has also been described for the ApoCIII gene which encodes for an inhibitor of lipoprotein lipase, providing a molecular basis for the triglyceride-lowering effect of FXR agonists. BA treatment induces the binding of FXR and RXR heterodimers and displaces HNF4α from a DR-1 site. Since this repressive effect is also observed in hepatic HNF4a-deficient mice, a direct repressive role of FXR has been proposed [166]. In the colonic Caco2 cell line, LCA treatment decreased UGT2B7 gene expression through a mechanism dependent on RXR and involving a negative FXRE composed of the sequence “GATCCTTGATATTA” [206]. A similar mechanism has been identified for the negative regulation of SULT2A1 by FXR in HepG2 cells and in vivo [207]. FXR represses liver pyruvate kinase (LPK) gene expression through tethering transrepression. At high glucose concentrations, HNF4α and carbohydrate response element-binding protein (ChREBP) bind to the LPK promoter and activate LPK expression. Agonist-activated FXR interacts with CHREBP and HNF4α, leading to the release of the transcriptionally active CHREBP and stabilizing the SMRT corepressor onto the FXR/HNF4α-containing complex, hence decreasing LPK transcription [133]. In addition, FXR repressed autophagic genes in the liver through the transrepression of cAMP response element-binding protein (CREB). Activated FXR interacts with CREB bound at autophagic gene promoters and disrupts the transcriptionally active CREB/CRTC2 complex [208]. FXR-mediated repression may also operate through competition with PPARα at autophagic gene promoters [209]. Furthermore, FXR cross talks with the NF-kappaB pathway. FXR activation inhibits the NF-kappaB-mediated hepatic inflammatory response without, surprisingly, affecting NF-kappaB-activated antiapoptotic genes [210]. This observation has been replicated in murine macrophages and data were suggestive of a model by which FXR activation led to the stabilization of the NcoR complex on the NF-kappaB-responsive element of the Il-1β promoter [211]. Conversely, NF-kappaB represses FXR activity in liver and intestinal cells [210, 212]. More studies are needed to dissect the mechanism of transrepression of NF-kappaB by FXR, and in particular whether it occurs independently of direct FXR DNA binding as shown for PPARα [213]. Genome-wide studies, together with comprehensive transcriptomic analyses, are required to address the gene selectivity of such a mechanism, its functional significance in vivo and the relative importance of the transrepression mechanism in the FXR activity spectrum.

Part III: mechanism of modulation of FXR activity

As emphasized above, questions remain about the role of endogenous BAs in activating FXR in different organs, as normal plasma BA concentrations are in the low micrometer range and EC50s in the high micrometer range. In vivo imaging of FXR activity in mice underlines the poor sensibility of lFXR to a CDCA-enriched diet, which only leads to increased iFXR activity. The FXR agonist GW4064 dramatically increases FXR activity in ileum, kidney, adrenals and liver [214], albeit that this efficiency was not replicated in another study [200]. The mechanism by which lFXR is activated in physiological conditions thus remains enigmatic, as no significant changes in hepatic BA concentrations were reported during a 24-h cycle in mice fed a chow diet [215]. Fxr gene expression oscillates weakly during a 24-h cycle in the mouse [216]. In contrast, the expression of FXR-target genes such as Shp, Cyp7A1, Ntcp and Bsep in the liver and Fgf15 and Ostα in the intestine clearly show circadian rhythmicity, suggesting a possible regulation of FXR activity through ligand-independent mechanisms [215, 217]. These observations suggest that in physiological conditions, lFXR activity may not be modulated by varying BA concentrations, in contrast to ileal FXR. In addition, Renga et al. [134] demonstrated that FXR activity and expression are dependent on the nutritional status, as activation of FXR exerts opposite effects on gluconeogenic gene expression in fasting versus fed conditions. This study raised the possibility of a differential FXR transcriptional activity as a function of the pathophysiological conditions, which was further strengthened by a genome-wide comparison of FXR binding sites in chow versus high fat diet-fed mice, which evidenced distinct FXR cistromes in both conditions [218].

These different results suggest that FXR activity is modulated by nutrient fluxes through a mechanism which may be, at least in part, independent of BA composition and concentration in the liver. A recent set of studies established that FXR-dependent transcriptional responses can also be modulated by nutrients essentially through post-translational modifications (PTMs). Glucose favors FXR expression and transcriptional activity [132] through the hexosamine biosynthetic pathway (HBP) and O-GlcNac transferase (OGT), which catalyzes FXR O-GlcNacylation [162]. Another signaling pathway regulated by the cellular energetic status is the AMP-activated protein kinase (AMPK) pathway. In low-energy conditions, AMPK is activated and targets a range of physiological processes to block anabolic reactions and to favor ATP synthesis [219]. Phosphorylation of FXR on serine 250 by AMPK decreases its transcriptional activity via the inhibition of coactivator recruitment [200]. In addition to AMPK, SIRT1 is activated in low-energy conditions by increased NAD⁺ concentrations. By counteracting the effect of p300-mediated FXR acetylation, which decreases its interaction with RXR thereby preventing its interaction with DNA, SIRT1 favors FXR recycling to regulated promoters. In line, deletion of hepatic SIRT1 negatively affects the FXR signaling pathway and triggers negative metabolic outcomes [111, 194, 199, 220]. However, SIRT1 is also a regulator of HNF1α, a known inducer of FXR expression in both liver and intestine, whose hepatic or intestinal deletion decreases hepatic and intestinal Fxr expression [110, 220]. SIRT1 transgenic mice display FXR persistent deacetylation correlated with decreased FXR protein content, leading to decreased FXR activity [157]. These paradoxical results highlighted a more complex regulation of FXR signaling by SIRT1, which acts at different levels, including direct deacetylation of FXR and also of its transcriptional partners and histones. Other FXR PTMs have been described, including phosphorylation by PKCα [221] and PKCζ [222], ubiquitination and sumoylation [223]. Further studies are warranted to understand the contribution of each PTMs to FXR activation in distinct physiological contexts.

Concluding remarks

Given the multiple roles of FXR in whole body metabolic regulation, FXR activation may be generally beneficial to normalize metabolic parameters in metabolically disturbed individuals. Recently, FXR signaling has been shown to be essential for the positive effect of vertical sleeve gastrectomy on weight loss and metabolic improvement [224], although the phenotype of Fxr ^−/− mice, which display a decreased body mass, calls for a careful evaluation of this study [225]. Further studies using tissue-specific Fxr deletion are needed to unravel FXR functions. Nevertheless, the potential benefit of FXR activation in many diseases including NASH and cholestasis has been highlighted in several clinical studies using 6α-ethyl CDCA (INT747, obeticholic acid). A phase II clinical trial in type 2 diabetes and NAFLD patients demonstrated that INT747 treatment increased insulin sensitivity, as evidenced by an increased glucose infusion rate in a hyperinsulinemic–euglycemic clamp study. There was also a drop in plasma γ-glutamyltransferase (GGT) and alanine aminotransferase (ALT), suggesting reduced liver injury. Liver fibrosis, which occurred in 81 % of the patients, was also improved significantly after INT747 treatment by reducing hyaluronic acid, procollagen III amino terminal peptide and tissue inhibitor of metalloproteinase 1 [118, 226]. A phase IIb clinical study with NASH patients has been stopped prematurely in January 2014 because of early achievement of the pre-defined efficacy criterion (Intercept Pharmaceuticals, FLINT trial NCT01999101). In addition, INT747 has also been tested in primary biliary cirrhotic (PBC) patients, an autoimmune disease which causes progressive cholestasis, fibrosis and ultimately cirrhosis. Clinical trials have shown beneficial effects of INT747 treatment in PBC patients. Short- and long-term treatment of INT747 together with UDCA reduced alanine aminotransferase (ALT) and alkaline phosphatase (ALP) levels in PBC patients which did not respond to UDCA alone (Intercept Pharmaceutical NCT00550862). A phase III clinical trial is currently ongoing (Intercept Pharmaceuticals NCT01473524). The more commonly observed adverse effect observed with INT747 treatment is pruritus, but HDL-C and LDL-C levels need to be monitored in long-term treatments. Other phase II clinical trials are planned on different hepatic pathologies such as primary sclerosing cholangitis (PSC) (NCT02177136) and alcoholic steatohepatitis (ASH) (NCT02039219). Another non-steroidal FXR agonist Px104 is being evaluated in a clinical phase II trial in NAFLD patients (Phenex Pharmaceuticals AG NCT01999101).

Furthermore, the comparative evaluation of several FXR agonists in rodent models of atherosclerosis and obesity and normolipemic cynomolgus monkeys evidenced distinct effects on lipoprotein and cholesterol plasma levels, while equally reducing aortic atherogenic plaque size [123]. These observations demonstrate that development of selective bile acid receptor modulators (SBARM) is warranted, allowing the partial and selective activation of the FXR transcriptome in a tissue-specific manner.