Recent advances on FXR-targeting therapeutics

The bile acid receptor FXR has emerged as a bona fide drug target for chronic cholestatic and metabolic liver diseases, ahead of all non-alcoholic fatty liver disease (NAFLD). FXR is highly expressed in the liver and intestine and activation at both sites differentially contributes to its desired metabolic effects. Unrestricted FXR activation, however, also comes along with undesired effects such as a pro-atherogenic lipid profile, pruritus and hepatocellular toxicity under certain conditions. Several pre-clinical studies have confirmed the potency of FXR activation for cholestatic and metabolic liver diseases, but overall it remains still open whether selective activation of intestinal FXR is advantageous over pan-FXR activation and whether restricted or modulated FXR activation can limit some of the side effects. Even more, FXR antagonist also bear the potential as intestinal-selective drugs in NAFLD models. In this review we will discuss the molecular prerequisites for FXR activation, pan-FXR activation and intestinal FXR in/activation from a therapeutic point of view, different steroidal and non-steroidal FXR agonists, ways to restrict FXR activation and finally what we have learned from pre-clinical models and clinical trials with different FXR therapeutics.


Introduction
Treatment with bile and bile constituents has a longstanding tradition. Traditional Chinese Medicine has been using dried bear bile for thousands of years to treat fever, inflammation, jaundice, for detoxification purposes and many more (Feng et al., 2009). A major constituent of bile are bile acids. In bear bile ursodeoxycholic acid (UDCA) is the predominating bile acid and constitutes up to 38% of total bile acids (at least in the black bear bile) (Hagey et al., 1993). For the sake of animal welfare, UDCA has been chemically synthesized for the past 60 years and it has subsequently been approved to treat gallstone disease as a single compound in 1977 and for the treatment of primary biliary cholangitis (PBC) in 1997. A mayor molecular step forward in bile therapeutics came in 1999 with the identification that bile acids can signal via a specific hormone receptor, the farnesoid X receptor (FXR) (Hofmann and Hagey, 2014). A few years later, in 2002, the first semisynthetic FXR ligand, 6alpha-ethyl-chenodeoxycholic acid (better known today as obeticholic acid (OCA)) has been generated (Pellicciari et al., 2002). Since then, subsequent work from researchers over the world has further characterized a bile acid-orchestrated metabolic network that is not only regulated by FXR but, from a therapeutic point of view, is also druggable by FXR ligands (for reviews e.g. (Fuchs and Trauner, 2022;Panzitt and Wagner, 2021;Perino et al., 2020;Fiorucci et al., 2021;). In 2017, finally, the European Association for the Studies of the Liver (EASL) recommended the use of the potent FXR ligand OCA for patients with PBC, who do not respond to standard UDCA treatment (European Association for the Study of the Liver, 2017). Currently, OCA and several other "next generation" FXR ligands are in the pipeline not only for the treatment of various cholestatic liver diseases but also for complex metabolic disorders such as non-alcoholic fatty liver disease (NAFLD). This review will give an overview on different molecular aspects of FXR ligands, on translational biases from mice to humans and on what disorders may also targeted by FXR ligands in the future.

Ligand binding -natural ligands
Bile acids are the natural FXR ligands, but the affinity of bile acids to the ligand pocket differs among different bile acids in the rank order of potency: chenodeoxycholic acid (CDCA) > deoxycholic acid (DCA) > lithocholic acid (LCA) ≫ cholic acid (CA) (Parks et al., 1999;Wang et al., 1999;Makishima et al., 1999). UDCA and its conjugated form glycine-UDCA, which appear in up to 5% in human bile, are antagonists to FXR (Mueller et al., 2015;Sun et al., 2018). In murine liver the enzyme Cyp2c70 converts CDCA into α-muricholic acid (αMCA) and further to βMCA (Takahashi et al., 2016;de Boer et al., 2020). These bile acids and their conjugates, which are primary bile acids in mice, are also antagonistic to FXR (Sayin et al., 2013;Jiang et al., 2015a). Since MCAs dominate in mice, the mouse bile acid pool is rather FXR antagonistic compared to the human bile acid pool. Conjugated bile acids, which account for 98% of all bile acids, require active cellular uptake by a bile acid transporter. Conjugated bile acids, therefore, activate FXR primarily in tissues that express bile acid transporters (this is predominantly the liver, the intestine and the kidney). Unconjugated bile acids may activate FXR in tissues that do not actively transport bile acids. Thus, potent high affinity FXR targeting drugs, which do not require straight active bile acid transport, may also show FXR effects outside bile acid transporting tissues.

DNA-binding
FXR binds to DNA hormone response elements as a monomer, a homodimer oras in most scenariosas a heterodimer with RXR. The classical FXR-RXR binding motif consists of two hexa-half sites arranged as an inverted tandem repeat separated by one nucleotide (IR-1). However, chromatin-immunoprecipitation sequencing (ChIP-seq) analysis during the past decade revealed some unexpected findings in respect to DNA binding of FXR, which needs to be considered when using FXR therapeutics. First (i), comparing FXR binding between different tissues indicates that FXR binding is largely tissue specific (Thomas et al., 2010). Hepatic binding sites overlap to only 11% with the binding events in the intestine. In the intestine an additional everted repeat (ER-2) binding motif for FXR is present, implying that tissue-specificity may also come along with motif variations (Thomas et al., 2010). Of note, the FXR ER-2 motif exists also in the liver (Jungwirth et al., 2021) and appears to be targeted by specific FXR isoforms, which predominantly regulate metabolic gene expression (Ramos Pittol et al., 2020). Second (ii), also the metabolic background dictates FXR binding. A comparison of hepatic FXR binding between normal and obese mice revealed that FXR binding was significantly reduced and divergent in obese mice (Lee et al., 2012). This is of utmost importance, when considering that, therapeutically, FXR agonists are used in diseased conditions. A potential explanation for different binding between obese and non-obese mice is post-translational acetylation of FXR in obese conditions, which reduces heterodimerization with RXR, DNA binding and transactivation activity (Kemper et al., 2009). Interestingly, several FXR associated genes were not inducible by FXR agonists in the obese condition indicating a higher number of functionally inactive FXR binding sites in obesity. Even more, agonistic FXR treatment often resulted in repression of gene expression (Lee et al., 2012). In addition to obese conditions, also a cholestatic background appears to result in different FXR binding patterns in human patients (Panzitt et al., 2020) and in rodents (Sutherland et al., 2018), which show robust and time dependent changes in FXR binding. Finally (iii), we performed a comprehensive comparison of FXR binding sites and associated genes between different species (Jungwirth et al., 2021) and revealed that the overlap of FXR binding sites between different species is modest and matches to only approx. 50% between humans and rodents. In addition, also invitro conditions appear to significantly impact on FXR binding (Jungwirth et al., 2021). Other interesting aspects revealed by cistromic studies included that -based on gene and functional pathway analysis -FXR may overall have a much wider role in cellular metabolism than previously appreciated (Thomas et al., 2010;Chong et al., 2010).

FXR tissue distribution
FXR is enriched in tissues that face high bile acid concentrations along the enterohepatic circulation of bile acids: liver ≫ small intestine > duodenum > gall bladder and to low extents also colon > rectum (Uhlen et al., 2015)(www.proteinatlas.org). In more detail, in hepatic tissue FXR is expressed in hepatocytes, liver resident macrophages (i.e. Kupffer-cells) (Jin et al., 2020), bile duct epithelial cells (Jung et al., 2014), liver sinusoidal endothelial cells (Tegge et al., 2018) and to a questionable extent in hepatic stellate cells (HSCs) (Fickert et al., 2009;Fiorucci et al., 2004;Garrido et al., 2021). The expression of FXR in HSCs, however, appears to be important for the effects of FXR on fibrosis. A recent study came up with an interesting explanation for diverging results of FXR in HSCs: activated HSCs show limited response to FXR agonists due to enhanced FXR sumoylation. Sumoylation inhibitors rescue FXR signaling and thereby increase the efficacy of FXR agonists against HSC activation and fibrosis (Zhou et al., 2020). In addition to FXR expression in hepatobiliary tissues, FXR is also highly present in tissues which are not primarily regarded as bile acid sensing and metabolizing tissues. These tissues include the kidney Herman-Edelstein et al., 2018;Jiang et al., 2007) and to a low extent the urinary bladder, steroidogenic tissues such as the adrenal glands (Baptissart et al., 2013;Liu et al., 2019;Theiler-Schwetz et al., 2019) and the ovary (Takae et al., 2019), but also the endocrine and exocrine parts of the pancreas (Popescu et al., 2010;Renga et al., 2010;Zhou et al., 2017a), immune cells (Schote et al., 2007), the brain (i.e. neurons (Huang et al., 2016) and astrocytes (He et al., 2021)), various cells of the cardiovascular system Pu et al., 2013;Bishop-Bailey et al., 2004) and in modest amounts also in adipose tissue (van Zutphen et al., 2019;Cariou et al., 2006)(for further review on non-gastrointestinal FXR functions see e.g. (Yan et al., 2021)). From a pharmacological viewpoint it may be relevant to consider that steroidal FXR ligands such as natural bile acids or OCA are conjugated during their repeated hepatic passage and therefore require active transport via the selective bile acid transporters NTCP or ASBT for cellular uptake (Boyer, 2013;Dawson, 2011). NTCP and ASBT are only expressed in primary bile acid target tissues, namely the liver (NTCP expression) and intestine (ASBT expression) -of note, kidney and bile duct epithelial cells also express ASBT to some degrees (Geier et al., 2007). Thus, steroidal/natural FXR ligands primarily target the liver and intestine and may target extra hepato-intestinal tissues to only minor extents. In theory, unconjugated synthetic FXR ligands, which may enter cells independently of the specific aforementioned high-capacity bile acid transporters, may therefore have more prominent effects on tissues other than the above mentioned primary bile acid target tissues.

Principle physiological effects of FXR regulated pathways
Due to the purpose of this review only the main FXR regulated physiological pathways, which are relevant for the therapeutic understanding of effects and side effects will be mentioned here without further detailing molecular mechanisms or conflicting results (Fig. 1). For details the reader is referred to other articles of this special issue or other recent in-depth review articles on FXR (Fuchs and Trauner, 2022;Panzitt and Wagner, 2021;Perino et al., 2020;Fiorucci et al., 2021;Sun et al., 2021).
Bile acid metabolism: FXR is the main sensor and regulator of bile acid metabolism. Bile acid pool size is regulated in a negative feedback loop via orchestrated action of intestinal and hepatic FXR signaling. In brief, FXR activation reduces bile acid production and favors bile acid excretion from the liver resulting in a quantitative decrease and a qualitative change of the bile acid pool. In the liver activated FXR suppresses hepatocellular bile acid uptake via downregulation of bile acid uptake transporters (NTCP) while it favors bile acid export via stimulation of sinusoidal (OSTα/β) and canalicular (MRP2, BSEP) bile acid transporters. Bile acid production in hepatocytes is suppressed by direct hepatic FXR effects on CYP7A1 via the FXR-SHP axis and by indirect extra-hepatic hormonal signaling of FXR induced FGF19 from the intestine. In addition to quantitative effects on bile acid pool size via CYP7A1, FXR also qualitatively modulates the bile acid pool composition. FXR regulates CYP8B1, the enzyme responsible for the hydroxylation of CDCA into CA, and thereby overall hydrophobicity of the bile acid pool is modulated. FXR activation also detoxifies hydrophobic bile acids via stimulation of phase 1 hydroxylating and phase 2 conjugating enzymes Perino et al., 2020). The qualitatively modified bile acid pool reciprocally affects hepatic FXR signaling and even more importantly impacts on extra-hepatic bile acid signaling molecules such as TGR5 with profound effects on overall energy homeostasis (Watanabe et al., 2006). FXR also stimulates canalicular phospholipid secretion via MDR3 into bile  and overall induces bile flow (Baghdasaryan et al., 2011).
Glucose metabolism: The effects of FXR on hepatic glucose homeostasis depend in part on the physiological context. In the fed state FXR activation lowers plasma glucose levels via repression of gluconeogenesis and favors glucose disposition and glycogen storage (Duran-Sandoval et al., 2005;Ma et al., 2006). The repressive effects on gluconeogenesis and the shunt from glycolysis towards glycogen storage are supported by FXR induced FGF19 release from the intestine (Potthoff et al., 2011;Kir et al., 2011). On the contrary, in the fasted state FXR activation stimulates gluconeogenesis (Renga et al., 2012a).
Cholesterol metabolism: FXR considerably determines atherosclerotic risk factors. The impact of FXR on cholesterol metabolism is established by regulating cholesterol breakdown and cholesterol distribution via apolipoproteins and significantly varies between mice and humans. In brief, FXR inhibits CYP7A1, the rate limiting enzyme in the conversion of cholesterol into bile acids. For sufficient reduction of hepatic CYP7A1 crosstalk between hepatic FXR and intestinal FXR-FGF19 release is required (Inagaki et al., 2005). Experiments in Cyp7a1 Ko mice show, that simply inhibition of Cyp7a1 on its own does not result in hypercholesterolemia, because cholesterol is rerouted for canalicular excretion into bile via the cholesterol transporter ABCG5/8 (Schwarz et al., 1998;Zhang et al., 2010). In addition, the lack of bile acids in the intestine significantly reduces intestinal cholesterol absorption and increases neutral sterol losses in the feces (Schwarz et al., 1998). Yet, another FXR pathway to reduce cholesterol is elimination of neutral cholesterol from the intestine via active transport mechanism, called transintestinal cholesterol excretion (TICE) (de Boer et al., 2018).
The effects of FXR on lipoprotein metabolism are complex and in part controversial. There exist ligand, sex and species specific differences . Overall, in mice FXR ligand activation is anti-atherogenic by enhancing reverse cholesterol transport and limiting intestinal cholesterol absorption, which leads to reduced atherosclerotic plaque formation (Xu et al., 2016;Hartman et al., 2009). However, there exist significant differences in HDL metabolism between mice and man and taking the plasma lipoprotein levels into account it appears that FXR activation in human rather shows a pro-atherogenic profile. An important contributor to the observed differences in cholesterol metabolism between mice and man are the aforementioned differences in bile composition mediated by the mouse specific enzymes Cyp2c70 (and Cyp2a12 (Honda et al., 2020)). In mice FXR antagonistic muricholic acids (MCAs) dominate the murine bile acid pool whereas in human the FXR agonistic CDCA and DCA are dominating (de Boer et al., 2020;Honda et al., 2020;Straniero et al., 2020). Moreover, in human hepatocytes APOA1, which is the main lipoprotein of HDL is under direct negative regulation of FXR, and FXR stimulation reduces HDL levels (Claudel et al., 2002;Neuschwander-Tetri et al., 2015). The pro-atherogenic impact of FXR on human HDL metabolism is extended by the stimulating effects on cholesterol ester transfer protein (CETP), which promotes the exchange of cholesteryl esters and triglycerides between HDL and APOB-containing lipoproteins, thereby increasing VLDL and LDL cholesterol levels and decreasing HDL cholesterol (Gautier et al., 2013). Albeit VLDL loading in hepatocytes is reduced by FXR activation in human hepatocytes (Hirokane et al., 2004), subsequent LDL formation is increased directly by FXR. FXR regulates the expression of the secreted cofactors required for lipoprotein lipase (LPL) activity. Activation of FXR induces APOC2 (Kast et al., 2001), which activates LPL and reduces the expression of APOC3, which inhibits LPL . Thus, FXR activation results in increased LDL cholesterol (Neuschwander-Tetri et al., 2015;Perez-Aguilar et al., 1985) by the transformation of TG-rich VLDL into TG-poor but cholesterol-rich LDL particles. In addition, hepatic LDL-receptor mRNA and its modulator PCSK9 are decreased upon bile acid treatment and thus also contribute to an LDL increasing profile (Nilsson et al., 2007;Ghosh Laskar et al., 2017;Langhi et al., 2008).
Lipid metabolism: FXR regulates lipogenesis and TG metabolism and thus contributes to the overall "fat" content of the liver. Overall, FXR activation reduces fatty liver in mice. FXR reduces the major transcription factor for lipogenic pathways, SREBP-1c and thereby reduces key enzymes of de novo lipogenesis (Watanabe et al., 2004). In addition, intestinal FGF19 supports the repressive effects on SREBP-1c and lipogenic enzymes (Bhatnagar et al., 2009). The role of FXR for fatty acid oxidation is less clear and species specific differences may exist. FXR activation induces human but not murine PPARα expression and its target genes (Pineda Torra et al., 2003), suggesting synergistic effects of FXR on PPARα-mediated β-oxidation.
Anti-inflammatory properties: Anti-inflammatory properties of FXR can mainly be attributed to transrepressive effects on the inflammation induced transcription factor NFκB, but also AP-1 and STAT3, which results in reduction of classical pro-inflammatory cytokines such as TNFα IL1β, IL6 and iNOS (Wang et al., 2008;Gadaleta et al., 2011;Bijsmans et al., 2015;Vavassori et al., 2009;Xu et al., 2012). These transrepressive effects are functional in hepatocytes as well as macrophages. Reciprocally, NFκB activation (and also inflammation activated STAT1 (Renga et al., 2009a)) antagonizes FXR activity and target gene expression (Wang et al., 2008), resulting in reduced FXR expression levels and function in inflamed tissues (Renga et al., 2009a). Reduced FXR expression in inflamed conditions may be a confounding aspect, when FXR ligands are therapeutically considered in such conditions. Another level of FXR-dependent control of inflammation represents the inflammasome. Inflammasomes are cytoplasmic protein complexes, which sense inflammatory stimuli and induce an inflammatory response by mediating cleavage of pro-inflammatory cytokines (Martinon et al., 2002). FXR was found to be an important negative regulator of the NLRP3 inflammasome either by direct physical interaction with NLRP3 and caspase 1 (Hao et al., 2017) or indirectly via ameliorating ER stress inducing inflammasome activation (Han et al., 2018). Of note, the role of bile acids on the inflammasome is conflicting and bile acids have been shown to activate (Hao et al., 2017) or inhibit the inflammasome via TGR5 (Guo et al., 2016) in monocytes. For further immunomodulatory functions of FXR the reader is referred to another review within this special issue or recently published articles on this topic (Fiorucci et al., 2018).
From a clinical perspective it may be favorable to identify or develop FXR modulators which selectively repress only inflammatory pathways without interfering with the metabolic effects of FXR. In such an effort mometasone furoate was identified as a compound that selectively reduced NFκB reporter activity in an FXR-dependent manner (Bijsmans et al., 2015).
Effects on fibrosis and vascular remodeling: FXR activation modulates hepatic stellate cell activation and results in decreased fibrosis in several animal models (Fiorucci et al., 2004(Fiorucci et al., , 2005a(Fiorucci et al., , 2005bSchwabl et al., 2017Schwabl et al., , 2021Zhang et al., 2009). However, the direct impact of FXR in HSC is not that clear, since conflicting results on FXR expression and function in HSC and myofibroblasts exist (Fickert et al., 2009;Fiorucci et al., 2004;Garrido et al., 2021). FXR also inhibits transdifferentiation and contractility of HSC, which comes along with reduced endothelin-1 expression and is predicted to lower intrahepatic resistance (Li et al., 2010). In addition to the potential anti-fibrotic properties, FXR activation also improves vascular inflammation and remodeling (Li et al., 2007). FXR affects vascular nitride oxide signaling and sinusoidal vasodilation by upregulating eNOS in vascular endothelial cells  and restores eNOS levels in cirrhotic animals (Verbeke et al., 2014;Mookerjee et al., 2015). These vascular effects of FXR contribute to reduction of portal hypertension in respective animal models (Schwabl et al., 2017(Schwabl et al., , 2021Verbeke et al., 2014).

FXR isoforms
FXR exists in 4 different isoforms (FXRα1-4), which arise from differential promotor usage and alternative splicing of the same gene (Huber et al., 2002;Zhang et al., 2003). Importantly, the isoforms are species-specifically expressed in different organs and isoform-specific activity depends on bile acid composition (Vaquero et al., 2013). In human liver FXRα1/2 dominates, with FXRα2 being the transcriptionally most active form. A recent study shows that the isoform FXRα2 is the dominating isoform in response to FXR agonism and binds to almost 90% of binding sites in an isoform selective fashion via an ER-2 motif, in addition to the canonical IR-1 FXR binding motif (Ramos Pittol et al., 2020). Feeding and fasting cycles (at least in mice) dynamically regulate FXR splicing (Correia et al., 2015). Fasting, exercise and a healthier liver increase FXRα2 expression and the switch in the FXRα2/FXRα1 ratio is sufficient to reprogram the transcriptional output. The FXRα2 isoform in the liver more effectively promotes fatty acid α-oxidation, reduces hepatic lipogenesis, enhances glycerate metabolism and improves ammonia clearance (Ramos Pittol et al., 2020). These effects, which are opposing the action of FXRα1 (in the fed state), canat least in micepartly be attributed to induction of PPARα mRNA. FXRα1 is rapidly induced in the refeeding state to become the predominant isoform and starts a transcriptional program that primarily leads to reduced de-novo lipogenesis (Correia et al., 2015). From a clinical perspective, FXRα2 specific FXR agonists would be favorable over currently available pan-FXR agonists in diseased conditions such as NAFLD since the major source of triglycerides in the liver is derived from free fatty acids and to a lesser extent from de-novo lipogenesis (Donnelly et al., 2005). However, to date, there exist no isoform selective FXR agonists and generating isoform selective ligands will be challenging, since the ligand binding domain is identical among all FXR isoforms (Mukha et al., 2021). Moreover, expression of FXR and in particular FXRα2, is reduced in NASH livers (Aguilar-Olivos et al., 2015), cirrhosis  and HCC tumors (Chen et al., 2013) compared to healthy livers, and it is expected that ER-2 targets are poorly activated by FXR ligands in these conditions (Ramos Pittol et al., 2020). Thus, strategies aimed at increasing FXRα2 expression would increase the therapeutic efficacy of FXR agonists against metabolic diseases (Ramos Pittol et al., 2020). Therefore, one future option could be interfering with the splicing process to induce one particular isoform (Mukha et al., 2021). From the perspective of a biomarker, the expression level of FXRα2 in liver might be used to predict responses of patients to the treatment with FXR agonists (Ramos Pittol et al., 2020). Of note, FXR dependent regulation of bile acid metabolism and consequently also application of FXR agonists for cholestatic liver disease is independent of the isoform (Ramos Pittol et al., 2020).

Therapeutic potential of steroidal and non-steroidal FXR agonists
Various FXR agonists with different physicochemical properties exist, which can broadly be sub-classified into ones with a steroidal or non-steroidal scaffold (Gege et al., 2019). A major limitation of the best studied and first-in-class steroidal FXR ligand OCA is the poor atherogenic profile of OCA with decreased HDL and increased LDL cholesterol levels, albeit cardiovascular fatalities have not been reported on the short-term. Another major side-effect is the pruritogenic potency of OCA, which has been-at least in partrelated to a TGR5 activating potential of steroidal FXR agonists. Since overall it is not entirely clear how much of the side effects are attributable to the steroidal properties or to intrinsic FXR agonistic effects, other potent and selective non-steroidal FXR agonists have widely been developed and tested (Mookerjee et al., 2015). A hope for these new FXR agonists is to limit pruritogenic and atherogenic side effects despite of still having pronounced effects on inhibiting bile acid synthesis, gluconeogenesis, lipogenesis and inflammatory pathways. These second generation (semi)synthetic pan-FXR agonists are designed to enhance the risk-benefit ratio by structural optimization. Due to their different structure and physicochemical properties these agonists differ in various aspects: the potency to activate FXR as full or partial agonist, the specificity to other receptors in particular TGR5, the bioavailability which determines their capacity to act on intestinal > hepatic FXR, the enterohepatic cycling and the affinity for different cellular uptake transporters. Of note, plasma concentrations of FXR ligands are not predictive of the potency of FXR agonists, and liver concentrations of FXR ligands do not necessarily correlate directly with FXR activity (Hambruch et al., 2012). Moreover, liver residency time may be far longer than plasma lifetime (Gege et al., 2019). The best method to test the potential of the novel compound for hepatic and or intestinal FXR activity should be by addressing direct gene expression analysis in vivo (Hambruch et al., 2016). Although many of the various structurally leading compounds have been tested in in vivo models or even clinical trials, a thorough head-to-head comparison in a comparable fashion is often lacking. Most compounds are tested for their effects on serum lipid parameters and their ability to induce intestinal FGF19. However, critical information on hepatobiliary enrichment versus plasma enrichment and their genome-wide transcriptomic profile is largely missing. This is important information because very likely transcriptomic profiles vary among different ligands. In fact, very early comparative experiments indicate that with the exception of consensus FXR target genes, broader gene expression patterns among different FXR agonistsin that particular study CDCA, fexaramine and GW4064 -are strikingly different (Downes et al., 2003). More recently, a comparable transcriptomic analysis among two structurally different FXR agonists showed a clear separation in principal component analysis (Harrison et al., 2021a). Since FXR ligands likely do not alter DNA binding on cistromic levels (Jungwirth et al., 2021) but mainly transcriptional events, differences in gene transcription may be primarily caused by conformational changes after binding of different ligands (Downes et al., 2003) and subsequently different co-regulator recruitment.
The first-in-class steroidal FXR agonist OCA has a potency (EC50) of approximately 100 nM (for comparison the EC50 of the most potent natural FXR agonist CDCA is 90-fold weaker at approximately 8.7 μM) in cell based assays (Pellicciari et al., 2002). OCA is efficiently cycling enterohepatically and its conjugated product (90-95%) is highly enriched in hepatocytes compared to plasma (Roda et al., 2014). A second FXR agonist with a steroidal scaffold, named EDP-305, has been established only recently and has a 16-fold more potent EC50 than OCA. In contrast to OCA, EDP-305 does not show activity against TGR5. However, pruritus was still an important clinical issue in the recently conducted phase II clinical trial with EDP-305 in NASH (Ratziu et al., 2022). An interesting semi-synthetic compound with a steroid scaffold that has not yet been tested in clinical trials should also be mentioned here: TC-100, which is a C11β-hydroxylated OCA derivate (Pellicciari et al., 2016). It is a fully selective bile acid derivate of FXR with a slightly higher potency than OCA and no activity against TGR5. The additional hydroxyl group renders TC-100 highly water-soluble (16-fold higher than OCA and comparable to CA) and therefore would limit the risk of bile acid induced toxicity. TC-100 is rapidly released from hepatocytes and shows an intestinal tropism (Pellicciari et al., 2016).
The best studied class of non-steroidal FXR agonists belong to isoxazole core compounds and are commonly named as the "hammerhead class" (Gege et al., 2014). The first synthetic FXR ligand in this class, GW4064, was already identified in the year 2000 -that was before the discovery of OCA (Maloney et al., 2000), and is widely used in animal experiments. GW4064 has an EC50 of approx. 40 nM but a poor bioavailability, a short half-live and it has estrogen-like activity and cross-reacts with the estrogen related receptor alpha (Dwivedi et al., 2011). Improved structural derivates of GW4064 are the full FXR agonists tropifexor (LJN452, EC50 < 10 nM (Tully et al., 2017),) and cilofexor (GS9674 or Px201, EC50 approx. 50 nM (Jiang et al., 2022)), which both have already undergone phase 2 trials in NASH (Kremoser, 2021;Patel et al., 2020) or cholestatic liver disease (Trauner et al., 2019a). Cilofexor is the follow-up compound of Px-102/PX20606 and Px-104, which have been tested in small human volunteer phase 1 and phase 2 studies, respectively and then abandoned (Gege et al., 2019;Traussnigg et al., 2021;Al-Khaifi et al., 2018). Another compound in the "hammerhead class" is TERN-101 (LY2562175A), which is a partial FXR agonist with higher EC50 potency compared to the mother compound GW4064 (Genin et al., 2015) and has passed phase 1 ). An interesting pharmacokinetic aspect of the "hammerhead" core compounds is their transport capacities by bile acid transporters. These compounds can be taurine conjugated to variable degrees by the bile acid conjugating enzymes BAAT and BACS just like the natural unconjugated bile acids (Gege et al., 2019). However, unlike their conjugated bile acid counterparts these synthetic taurine conjugated "hammerhead" FXR agonists may not be taken up by the intestinal bile acid uptake transporter ASBT anymore and therefore conjugated forms may not undergo enterohepatic cycling and have only limited bioavailability (<10%) (Gege et al., 2019). Still, their transcriptomic liver response is very efficient (Hambruch et al., 2012). It is not clear, if this can be attributed to bioavailability of more abundant unconjugated over conjugated forms and thus more potent stimulation of hepatic FXR by unconjugated compounds (Hambruch et al., 2016).
A completely different structure owes nidufexor (LMB763), which is a partial FXR agonist with a medium potency (Chianelli et al., 2020) and has now entered clinical phase 2 studies in NASH. Interestingly, its transcriptomic profile in rodents with NASH shows way more reversal of the NASH gene signature than OCA does (Chianelli et al., 2020). Yet another different compound, which is speculated to be a derivate of fexaramine (Gege et al., 2019), is called MET409, and has shown promising results in a recent NASH trial (Harrison et al., 2021a). All the compounds listed above have undergone at least phase 1 testing and have entered clinical trials. The only compound, which has already been approved for clinical usage is OCA in primary biliary cholangitis (PBC) (European Association for the Study of the Liver, 2017; Nevens et al., 2016).
There exist a few more compounds, all non-steroidal from their structure, which need to be mentioned here. WAY-450 (FXR-450) is a potent (EC50 < 10 nM) but insufficiently soluble full FXR agonist which has been used in several mouse experiments (Gege et al., 2019;Flatt et al., 2009;Evans et al., 2009). No follow-up compounds and clinical trials with these compounds have yet been released. Fexaramine has been discovered already back in 2003 with a decent EC50 of 25 nM. However, the hydrophobicity and insolubility of fexaramine turned this compound into the prototype of intestinally restricted FXR agonists with remarkable metabolic effects without potential liver related side effects (Gege et al., 2019;Downes et al., 2003;Fang et al., 2015). There are yet no reports on follow-up compounds of fexaramine with clinical data which could confirm the promising rodent data. Yet, another compound which is interesting to mention is Merck FXR agonist #1 (MFA-1), which has a decent potency and is by structure a steroidal FXR agonist but the steroid ring of MFA-1 binds to the FXR ligand binding domain in flipped orientation compared to the natural bile acid ligands (Soisson et al., 2008). Unfortunately, it is not known if this opposed ligand binding changes transcriptional output since no further transcriptomic data on this interesting compound is available.
Are non-steroidal FXR agonists superior? There are only a few headto-head comparisons in rodents. In mouse models of cirrhosis and portal hypertension the effects of PX-102/PX20606 were only marginally more pronounced compared to OCA in respect to vascular remodeling, effects on fibrosis and angiogenesis (Schwabl et al., 2017). In a chimeric mouse model with a humanized liver the effects of OCA, cilofexor and tropifexor were comparable with respect to effects on cholesterol, HDL and LDL levels and enzymes involved in cholesterol metabolism (Papazyan et al., 2018). One study again compared effects of OCA and PX-102/PX20606 and found a much higher liver exposure and enhanced FXR liver transcriptional response in PX-102/PX20606 compared to OCA and GW4064 (Hambruch et al., 2012). Pruritus was a significant side-effect also for the non-steroidal FXR agonists, which in general lack TGR5 agonistic activity and thus, pruritus cannot be attributed to the steroidal properties (Kremoser, 2021). Since human comparative studies do not exist and the clinically most important questions of pruritus and effects on the atherogenic profile cannot be answered sufficiently in rodents, neither a preference for steroidal nor non-steroidal FXR agonists may yet be made. Fig. 2 summarizes the effects of steroidal FXR agonists in the intestine and the liver.

Partial agonists/selective agonists for FXR
Full FXR agonism comes along with side effects, the most prominent in human being its pruritogenic potential as well as the HDL cholesterol lowering and LDL cholesterol increasing capacities. Another yet unpredictable risk is the induction of native FGF19 (mainly from the intestine), which in contrast to clinically tested FGF19 mimetics (Harrison et al., 2021b), may be mitogenic (Krones and Wagner, 2016). Moreover, from what we know of cistromics and transcriptomics studies, FXR is regulating a much broader spectrum of genes than initially anticipated (Thomas et al., 2010;Chong et al., 2010), thus, "unselective" targeting might bear a risk for yet unpredictable effects. Development or identification of partial agonists or selective bile acid receptor modulators (SBARMs) could represent a way, which specifically modulates FXR action to enhance desired pathways and minimize unwanted side effects. Partial agonists bind to the LBD, and the resulting conformational change provides only a partial activation of transcription, which may be due to less proficient recruitment of co-activators or competition with full agonists (Burris et al., 2013).
Selective regulation can be broadly seen as gene-selective and/or tissue selective FXR effects. An excellent review on this topic can be found elsewhere . The main mechanisms for gene-selective FXR modulation include controlling of co-factor binding and subsequent post-translational modifications to FXR as well as differential DNA binding. Consequently, controlling post-translational modifications of FXR represents an opportunity to target a specific subset of FXR-regulated genes Kim et al., 2015) and, activation of specific FXR isoforms would also represent a theoretic opportunity to modulate gene-selective programs via distinct motif binding (Ramos Pittol et al., 2020). Moreover, structure analyses suggest that different ligands may induce different receptor conformations, which then drive co-regulator, tissue-and gene-selectivity (Burris et al., 2013). Binding to allosteric pockets within the ligand binding domain may lead to conformational modulation with different biological outcomes Meijer et al., 2019;Pellicciari et al., 2006). One example in respect to FXR is the selective agonist/antagonist guggulsterone (Urizar et al., 2002;Wu et al., 2002), which binds to an allosteric pocket in the ligand binding domain of FXR (Meyer et al., 2005). Guggulsterone, which is an herbal extract from the guggul tree, has initially been discovered as an FXR antagonist and predictably reduces LDL in human (Urizar et al., 2002;Wu et al., 2002). Guggulsterone is an FXR antagonist in co-activator association assays, but it enhances FXR agonist-induced transcription of Bsep and Shp without affecting Cyp7a1 and Cyp8b1. Thus, guggulsterone rather is a selective bile acid receptor modulator that regulates expression of a subset of FXR targets (Cui et al., 2003). Of note, a clinical trial with guggulipid, failed to show benefits on LDL cholesterol (Szapary et al., 2003). Identification of FXR ligands which interact with allosteric/non-canonical sites may represent yet another option to modulate FXR transactivation properties Pellicciari et al., 2006).
Selective FXR agonists may also be designed to target FXR in a tissue specific manner. Restriction of FXR to specific target tissues may increase pharmacologically wanted effects and limit side effects. The main tissue sites which are especially targeted are either the intestine or the liver. However, it has to be kept in mind that intestinal FXR communicates with the liver e.g. via FGF15/19 signaling and thus, intestinal specific FXR agonism or antagonism also affects liver diseases. Tissue specific effects are dictated by the prevalence of tissue dominating binding motifs and binding patterns (Thomas et al., 2010). Only a minority of FXR binding sites are overlapping between the liver and the intestine (Thomas et al., 2010). In addition, variation in the co-activator to co-repressor levels in a particular cell type determine the agonistic and antagonistic properties of selective modulators as best exemplified for the estrogen receptor (Burris et al., 2013). Moreover, FXR isoforms are differentially prevalent in various tissues (e.g. liver, intestine, kidney), which then also show different responses to FXR agonistic treatment (Vaquero et al., 2013). Pharmacologically, tissue specificity can be achieved by modulating the physicochemical properties of the FXR ligands, which critically depends on the uptake by bile specific transporter systems in the liver or intestine (see above) and the residence time of a ligand in a specific tissue (Pellicciari et al., 2016). Examples of FXR ligands with intestine specificity include fexaramine (Fang et al., 2015), TC-100 (Pellicciari et al., 2016) and ivermectin Jin et al., 2013). So far, there is no exclusive liver specific FXR agonist known. In general, the non-steroidal FXR agonists show an intestinal preponderance. The molecular and functional effects of selective intestinal and hepatic FXR modulation are discussed below.

FXR antagonists
Teleologically, FXR agonism is the promising molecular therapy for cholestatic and metabolic liver diseases with an inflammatory and fibrotic component (Fig. 3). The rationales for such FXR agonistic mechanisms have been confirmed in several clinical trials, and the firstin-class FXR agonist, OCA, became approved for PBC in 2017 and is expecting approval for NAFLD. Even more astonishing is the fact that also FXR antagonists may havein addition to the lack of aforementioned typical side effects of the classical first-in-class FXR agonistsmolecular and clinical benefits in cholestatic (Wagner et al., 2003;Stedman et al., 2006;Renga et al., 2012b) and metabolic liver diseases (Jiang et al., 2015a(Jiang et al., , 2015b. The best known example of an FXR antagonistically acting drug is UDCA, which is the mainstay in the treatment of chronic cholestatic liver diseases in human patients, ahead of all PBC. UDCA itself does not show activity against FXR (Lew et al., 2004). However, upon treatment UDCA-conjugates, especially the glycine-conjugate of UDCA, become the prominent bile acid species in the bile acid pool (Mueller et al., 2015;Marschall et al., 2005) and thereby reduce overall FXR activity by substituting for FXR agonistic bile acids such as CDCA (Sun et al., 2018). Glycine-UDCA can also be endogenously generated by metformin treatment and specific reduction of bile salt hydrolase (BSH) expressing bacterial strains (Sun et al., 2018). Mechanistically, glycine-UDCA and tauro-UDCA reduce FXR activation by CDCA-dilution in several binding assays and are potent antagonists particularly against intestinal FXR (Sun et al., 2018). In addition to reduced FXR activation by UDCA, also reduced overall FXR binding upon UDCA treatment has been shown in DNA binding assays in obese patients, which comes along with reduction of the intestinal FXR target FGF19 and increased hepatic CYP7A1 (Mueller et al., 2015). However, it appears that under certain conditions some FXR agonistic activity for BSEP remains (Mueller et al., 2015;Lew et al., 2004;Marschall et al., 2005), placing UDCA in the category of a partial antagonist-agonist. There do not exist any genome wide FXR binding assays under UDCA treated conditions, which would greatly contribute to the understanding how UDCA is modulating FXR binding on a gene by gene level. Why then is UDCA beneficial in chronic cholestatic liver disease despite being FXR antagonistic? Several modes of action in combination may explain this and should be listed here (Beuers, 2006): rendering a hydrophobic biliary bile acid pool into a more hydrophilic and less toxic one; stimulation of bile-acid dependent and independent bile flow; increasing biliary protection by a bicarbonate umbrella (Beuers et al., 2012), anti-apoptotic effects; autophagy stimulating effects (Panzitt et al., 2020); induction of the alternative bile acid export pump MRP4 (Marschall et al., 2005;Renga et al., 2011) and induction of canalicular BSEP (Marschall et al., 2005). Similar to the FXR agonistic glycine-UDCA, which is a low abundant natural bile acid in human, the murine primary bile acids tauro-αMCA and tauro-βMCA are FXR antagonists (Sayin et al., 2013) (Fig. 3). Endogenous conjugated bile acids are sensitive to deconjugation by bacterial BSH. Thus strategies that reduce BSH producing bacteria (e.g. tempol, metformin or germ-free conditions) can increase natural FXR antagonistic bile acids (Sun et al., 2018(Sun et al., , 2021Sayin et al., 2013;Li et al., 2013). Synthetic glycine-βMCA is more resistant to BSH deconjugation than the endogenously occurring tauro-βMCA in mice and is a strong intestinal FXR antagonist in rodents (Jiang et al., 2015a). In contrast to intestinal specific FXR agonists such as fexaramine the metabolically beneficial effects of glycine-βMCA (reduction of obesity, improvement of insulin resistance, reduction of fatty liver in high/fat diet treated mice) are independent of TGR5 modulation and strictly dependent on FXR antagonism (Jiang et al., 2015a). Mechanistically, the metabolic improvements with glycine-βMCA depend on reduced biosynthesis of intestinal-derived ceramides, which directly compromise thermogenic function of beige fat cells (Jiang et al., 2015a). The FXR agonist GW4064 (and ceramides as well) block the effects of glycine-βMCA (Jiang et al., 2015a). Theonellasterol is a marine sponge sterol and a selective FXR antagonist that protects against cholestatic liver injury at least in mice (Renga et al., 2012b). Mechanistically, theonellasterol recruits the co-repressor NCOR1 to FXR and prevents inhibition of FXR on Mrp4. Increased Mrp4 expression may then be the central beneficial part to reduce injury after bile duct ligation in theonellasterol treated animals (Renga et al., 2012b). Sulfated progesterone metabolites are elevated in the serum and urine of women with intrahepatic cholestasis of pregnancy and epiallopregnanolone sulfate (PM5S) acts partially FXR-antagonistic. PM5S competes with FXR agonists for the ligand binding pocket of FXR leading to reduced bile acid efflux and FGF19 secretion (Abu-Hayyeh et al., 2013).

FXR as a drug target in liver, intestine or both?
In vivo imaging of FXR activity in a luciferase reporter mouse revealed that under physiological conditions only FXR in the terminal ileum has basal activity, which follows a diurnal rhythm (Houten et al., 2007). However, after pharmacological stimulation (in this particular study with GW4064) FXR activity becomes boosted in the intestine, but also in the liver, kidney and adrenal glands. Similarly, bile duct ligation recruits hepatic FXR activation (Houten et al., 2007). This suggests that the diurnal physiological homeostasis is maintained predominantly via intestinal FXR, but under pharmacological and stress conditions, additional sites of FXR tissue expression become activated (Houten et al., 2007). This overall implies that FXR may have tissue specific basal functions and adapt in its activity to pathophysiological requirements that may be exploited pharmacologically. It is important to note, that under basal physiological conditions, neither the liver specific nor the intestine specific FXR knockout mouse show an apparent phenotype, except for an increased bile acid pool size (Kim et al., 2007a). Tissue-selective functions of FXR are supported by comparison of intestinal and liver specific FXR binding profiles in ChIP-seq analyses which exhibit a high-degree of tissue specific binding sites and tissue selective gene regulation (Thomas et al., 2010). To name just two prominent gene specific examples: Shp, which is a major gene repressor activated by FXR, has higher basal expression and FXR promotor binding in the liver, but can be fully pharmacological activated under pharmacological conditions in both, liver and intestine (Thomas et al., 2010;Kim et al., 2007a). On the other hand, FXR induced FGF19 is fully expressed in the terminal ileum and absent in the liver but can be induced in the liver upon pharmacological stimulation and in cholestatic conditions in human (Al-Dury et al., 2019; Schaap et al., 2009).
Activation of intestinal FXR is sufficient to repress hepatic bile acid synthesis via Cyp7a1 through the FGF19 pathway but also to modulate Fig. 3. FXR antagonists. FXR antagonists such as glycine-UDCA or taurin a/ bMCA deactivate FXR in the intestine whereas liver FXR remains unaffected. FXR antagonism in the intestine improves hepatic and extrahepatic metabolic signaling by decreasing ceramide production. the bile acid pool hydrophobicity index via suppression of Cyp8b1. The effects on Cyp8b1, however, are mainly regulated via hepatic FXR (Kim et al., 2007a;Kong et al., 2012). Thus liver FXR is the main factor that lowers bile acid hydrophilicity in humans (i.e. makes the bile acid pool more hydrophobic). A hydrophilic bile acid pool is a major determinant of transintestinal cholesterol excretion (TICE), which contributes to at least 30% of neutral sterol excretion (de Boer et al., 2017(de Boer et al., , 2018. Liver FXR is also the major factor for reducing HDL cholesterol (Papazyan et al., 2018;Lambert et al., 2003). From this point of view, targeting intestinal FXR appears favorable over hepatic FXR activation for both pathological conditions, i.e. bile acid overload conditions as in cholestasis and metabolic overload conditions as coming along with the metabolic syndrome. In fact, experiments with FXR agonists that tissue selectively activate primarily intestinal FXR (i.e. fexaramine) (Fang et al., 2015) or mice selectively overexpressing intestinal FXR (Modica et al., 2012) support that intestinal FXR activation may be sufficient to alleviate cholestatic and metabolic liver diseases, while on the contrary side-effects and toxicity issues arising from hepatic FXR activation may be reduced (Fig. 4). Many of the metabolic functions of intestinal FXR activation are mediated via secretion of FGF15 (mouse ortholog)/FGF19 (human ortholog) (Inagaki et al., 2005). FGF19 is an essential repressor of hepatic bile acid synthesis which determines its anti-cholestatic properties. FGF19 also promotes energy expenditure, reduces gluconeogenesis and lipogenesis and thereby prevents fatty liver and improves insulin resistance (Kir et al., 2011). In line with these metabolic effects, FGF19 analogues have proved their efficiency in animal models of cholestasis (Modica et al., 2012;Luo et al., 2014;Zhou et al., 2016) and NAFLD (Zhou et al., 2017b) as well as in human clinical trials (Harrison et al., 2021b;Hirschfield et al., 2019;Mayo et al., 2018). Of note, part of the effects of intestinal FXR activation and FGF19 induction on energy expenditure appear to depend on a shift in bile acid composition and subsequent TGR5 activation (Fang et al., 2015;Pathak et al., 2018). The metabolic effects on browning of white adipose tissue and energy expenditure induced by the intestinal FXR agonist fexaramine can be blocked in TGR5 knockout mice, but surprisingly also in antibiotics treated mice, suggesting that the gut microbiome plays a critical role in the intestinal FXR effects (Pathak et al., 2018). For further information the reader is referred to a review article on FGF19 action in this special issue series as well as excellent recent reviews (Gadaleta and Moschetta, 2019;Owen et al., 2015).
The view of intestinal FXR activation to treat metabolic liver diseases such as NAFLD (not cholestatic liver diseases) has been challenged by a set of elegant studies in mice which overall show that also intestinal FXR antagonism has beneficial metabolic effects on NAFLD and insulin resistance. These studies show that intestinal FXR antagonism reduces local ceramide production, which can result in detrimental hepatic effects due to increased gluconeogenesis. FXR antagonists can be internally generated by changing the gut microbiome or via synthetically generated compounds (Sun et al., 2018;Jiang et al., 2015aJiang et al., , 2015bLi et al., 2013;Pathak et al., 2018;Xie et al., 2017).

FXR-targeting therapeutics in chronic liver diseases and beyond
In the following section we will only briefly introduce the concept for targeting FXR in cholestatic liver diseases and NAFLD and discuss the preclinical and clinical effects of FXR agonism with particular emphasis on different FXR agonists, if data are available. Detailed reviews on FXR and its role in bile acid metabolism and the metabolic syndrome are reviewed in separate articles of this special issue. At the end of this chapter we will give an outlook on what we think could become future relevant disease targets for FXR modulating compounds.

Rational to target FXR in cholestasis
Cholestatic liver diseases comprise a spectrum of acute and chronic liver diseases, which are characterized by impairment in regular bile formation and bile flow (Fickert and Wagner, 2017). In adults, PBC and primary sclerosing cholangitis (PSC) represent the two most prevalent chronic cholestatic liver diseases. Re-establishment of regular bile formation and/or increasing adaptive mechanisms to handle cholestatic bile acid overload are central parts in the therapeutic concepts of chronic cholestasis. Bile acid metabolism can be targeted at different levels (Wagner and Fickert, 2020): strategies to increase choleresis via stimulation of bicarbonate-rich bile flow, strategies to reduce bile acid pool size, strategies to change bile acid composition and strategies to increase adaptive transporter and detoxification mechanisms for toxic accumulating bile acids. Depending on the disease stage additional properties such as anti-inflammatory or anti-fibrotic effects later in the disease course may be required to support the bile-modulating effects. FXR targeting drugs act on all levels of bile formation, are anti-inflammatory and also inherit anti-fibrotic effects. FXR agonists therefore represent ideal anti-cholestatic drugs.

What we have learned from pre-clinical models
In general, the most common animal models of cholestasis either use bile acid feeding to mimic bile acid overload, specific bile acid transporter deletions to mimic human inborn errors of cholestasis, bile duct ligation to imitate biliary obstruction or chemical exposition to cause damage of intrahepatic bile ducts. Mdr2 transport knockout mice (Mdr2 Ko) mimic PSC and secondary sclerosing cholangitis, show many features of chronic cholestasis, are widely available and represent a very well comparable mouse model to study cholestasis and effects of anticholestatic drugs (Fickert et al., 2014). In Mdr2 Ko mice the canalicular phospholipid floppase Mdr2/Abcb4 has been genetically deleted (Smit et al., 1993). The knockout results in a lack of phospholipids in bile, which are required to pack bile acid in mixed micelles with cholesterol. Non-micellar free bile acids represent a toxic hit to the bile duct epithelium which triggers a cascade of inflammatory events, finally leading to sclero-fibrotic destruction of bile ducts (Fickert et al., 2004).

Fig. 4. Intestinal specific FXR agonists.
Intestine specific FXR agonists such as fexaramine specifically activate only intestinal FXR but can affect the liver by FGF19. Changes in the bile acid composition additionally activate TGR5, which adds to the beneficial effects outside the liver.
Activation of intestinal FXR is sufficient to alleviate cholestatic liver disease in this model. Genetic overexpression of intestinal FXR significantly reduced hepatic Cyp7a1 via Fgf15 signaling and consequently robustly decreased bile acid pool size by 75%. This is accompanied by decreased serum parameters of liver injury, improved liver histology and reduced pathological bile duct proliferates. In contrast, genetic silencing of intestinal FXR increased hepatic Cyp7a1 and significantly worsened cholestatic liver injury in Mdr2 Ko mice (Modica et al., 2012). These experiments suggest that reduction of the bile pool size and selective stimulation of intestinal FXR may be sufficient to alleviate cholestatic liver injury. Surprisingly, pharmacological treatment of Mdr2 Ko mice with OCA even aggravated injury (Baghdasaryan et al., 2011). OCA in this model robustly increases Fgf15 and represses Cyp7a1 but it does not induce bile acid independent bile flow and biliary bicarbonate secretion and it does not reduce biliary bile acids which are a main culprit in disease pathogenesis in the Mdr2 Ko model of sclerosing cholangitis. It rather appeared that hydrophobic OCA accumulated in the liver without stimulation of hepato-protective effects which resulted in aggravated injury (Baghdasaryan et al., 2011). Interestingly, the 10-fold more (FXR-)potent but hydrophilic steroidal FXR/TGR5 dual agonist INT-767 reduced injury in an FXR dependent manner in the Mdr2 Ko model by inducing bile flow and bicarbonate secretion (Baghdasaryan et al., 2011). These experiments suggest that FXR-induced choleresis, in particular bicarbonate-rich choleresis along with reduced biliary bile acids are important to alleviate cholestatic injury. Of note, also UDCA treatment resulted in aggravated injury in the Mdr2 Ko model, which is most likely not attributed to the fact that UDCA is an FXR antagonist but due to increased choleresis in the presence of significant bile duct obstruction (Fickert et al., 2002(Fickert et al., , 2006. Therefore, from animal experiments it is not entirely conclusive which FXR -ileal, hepatocyte or bothshould be targeted and if the effects of FXR activation are more dependent on stimulation/restoration of (bile acid independent) bile flow or repression of bile acid synthesis and pool size or both. Another conclusion derived from these experiments is that hydrophobic steroidal FXR agonists trapped in hepatocytes during cholestasis represent a potential risk for toxicity issues. Indeed, all too frequent dosing of OCA in cholestasis with already advanced cirrhotic injury resulted in severe liver damage and even death (John et al., 2021;Eaton et al., 2020). In other models of cholestasis such as LCA treatment and ethinylestradiol-induced cholestasis as models for intrahepatic cholestasis, OCA restored bile flow and improved cholestasis (Pellicciari et al., 2002;Fiorucci et al., 2005c). The non-steroidal FXR agonist GW4064 improved markers of cholestasis along with the reduction of hepatic bile acid accumulation in bile duct ligated and α-naphthylisothiocyanate treated rats, too (Liu et al., 2003). Some of the inconclusive effects in particular with the treatment of OCA may also be related to the cholestatic model and/or species. In this respect, it is noteworthy to again point to the fact that the bile acid composition between rodents and human significantly differs in its hydrophobicity and FXR activity and, therefore the translation of results from mouse to human is limited.

What we have learned from clinics
It is important to understand that for none of the FXR agonists hard clinical endpoints -that is survival or transplant free survival -have been demonstrated. All clinical trials in cholestatic liver disease use reduction of alkaline phosphatase (AP), which is a serum cholestasis biomarker, as primary readout. Only for UDCA treatment in patients with PBC survival benefits have been shown and calculated (Table 1).
Ursodeoxycholic acid is the first-line therapy in PBC and all novel therapies are primarily intended as add-on therapy to UDCA (European Association for the Study of the Liver, 2017). The 30-40% of PBC patients who do not respond adequately to UDCA (i.e. AP levels >1.67 upper limit of normal (ULN)) are at increased risk for progressive disease, particularly when diagnosed at later stages (Corpechot et al., 2005). Since non-responders to UDCA therapy nonetheless profit from therapy compared to untreated patients, they should also receive lifelong UDCA treatment (Harms et al., 2018).
Obeticholic acid is approved as second line treatment in PBC for UDCA non-responders or patients who do not tolerate UDCA. The POISE trial, which led to the approval of OCA in PBC, showed response rates to OCA in up to 47% after 12 month of treatment, but only 7% normalized their AP levels (Nevens et al., 2016). OCA treatment resulted in a dose-dependent increase in the FGF-19 level and a significant decrease in serum bile acid levels. In addition, inflammatory markers such as high-sensitivity CRP and TNF-α levels decreased significantly in OCA treated patients. Response rates remained stable after 3 years of open label extension (Trauner et al., 2019b). The results have recently been confirmed with real-life data, which show an overall response rate of 43%. Not surprisingly, patients with advanced liver disease and cirrhosis responded with lower efficacy due to less tolerability and discontinuation (D'Amato et al., 2021). The most prominent side effect was dose-dependent pruritus in more than 50% of the patients. Another concern with OCA is the reduction of HDL and increase of LDL cholesterol levels. This is less problematic in PBC patients but may be of greater concern in the ongoing OCA trials in patients with NAFLD who are at increased risk for cardiovascular diseases (Neuschwander-Tetri et al., 2015). Yet another concern are potential hepatotoxic issues with OCA in patients with severely impaired liver function (John et al., 2021;Eaton et al., 2020). In PSC patients with baseline UDCA medication, OCA significantly decreased serum AP levels by about 15% after 24 weeks. The most significant side effect of OCA was again dose-dependent pruritus (Kowdley et al., 2020). Whether OCA will have any positive or negative effects on the risk of cholangiocellular or colorectal cancer in humans is open to question but will need special attention and careful surveillance (Fu et al., 2019).
Cilofexor, a synthetic non-steroidal FXR agonist, led to dosedependent significant reductions in serum AP of up to 21% in noncirrhotic PSC patients (Trauner et al., 2019a). Adverse events were similar between cilofexor and placebo-treated patients. Interestingly, pruritus was more frequent in the placebo group and the rate of pruritus in the cilofexor groups was overall low. It is not clear if less pruritus is due to the non-steroidal character of cilofexor or probably due to other reasons. In NASH trials a minor dose dependent increase of pruritus has been observed (Kremoser, 2021).
There are several clinical trials with other FXR agonists in cholestatic patients ongoing (including cilofexor, tropifexor and EDP-305), which have not been published as full papers. The reader is kindly referred to other reviews in this field, which cover the clinical trials in more depth (Simbrunner et al., 2021).

Rational to target FXR in NAFLD
Besides regulating bile acid metabolism, FXR also regulates glucose and lipid metabolism. In general, FXR activation physiologically reduces postprandial plasma glucose levels by limiting gluconeogenesis, glycolysis and favors the shuttling of glucose into glycogen production. FXR activation lowers lipid contents by reducing fatty acid uptake and de novo lipogenesis and increasing β-oxidation. In addition to the metabolic effects, FXR also has major effects on the liver stress response: FXR activation alleviates ER stress, oxidative stress, apoptotic cell death and inflammation via NFκB signaling (reviewed in (Yan et al., 2021)). Moreover, it is increasingly appreciated that NAFLD and in particular non-alcoholic steatohepatitis (NASH) comes along with impaired bile acid metabolism, which has been termed micro-cholestasis (Radun and Trauner, 2021). These changes consist of reduced FXR mRNA in NAFLD, increased bile acid concentrations, a shift in primary and secondary as well as conjugated and unconjugated bile acids, overall changes which suggest that less FXR activity or "FXR resistance" in NAFLD/NASH conditions prevail (Radun and Trauner, 2021;Yang et al., 2010;Jiao et al., 2018;Puri et al., 2018). Reconstitution of micro-cholestasis may therefore be yet another event that can be targeted by FXR. Overall, FXR activation appears promising to combat diseases of the metabolic syndrome, ahead of all NAFLD which often comes along with obesity and diabetes. The effects of FXR appear, however, to differ between physiological regulation and pathophysiological effects. In addition, FXR activation increases the conversion from VLDL into LDL and reduces LDL receptor mRNA and its modulator PCSK9, which overall results in increased LDL levels. HDL on the contrary is reduced by enhancing reverse cholesterol transport (for more detailed reviews e.g. Perino et al., 2020)). These are major limitations for the use of FXR agonists in a patient population with increased risk for cardiovascular events.

What we have learned from pre-clinical models
Up to 80% of patients with NAFLD are obese and/or have diabetes type 2 with insulin resistance. Therefore, treatment should ideally not only reduce fat in the liver but should also reset underlying insulin resistance and/or obesity. Deletion of global FXR in regular lean mice results in fatty liver and impairs glucose homeostasis and insulin resistance (Cariou et al., 2006;Ma et al., 2006;Zhang et al., 2006;Sinal et al., 2000). However, deletion of global FXR in obese and insulin resistant mice results in improved insulin resistance and reduced weight gain (Prawitt et al., 2011;Zhang et al., 2012). Importantly and in contrast to the beneficial effects on weight and glucose, liver steatosis in global FXR deleted obese mice aggravated as a result of repressed β-oxidation (Prawitt et al., 2011). Interestingly, liver-specific FXR deficiency did not protect from diet-induced obesity, insulin resistance and fatty liver suggesting that intestinal FXR antagonistic effects may control critical metabolic outputs in obesity (Prawitt et al., 2011;Schmitt et al., 2015). In line with antagonistic effects on intestinal FXR either genetic FXR deletion in the intestine or pharmacological FXR antagonism with synthetic glycine-βMCA, natural tauro-βMCA or glycine-UDCA protects from obesity and improves insulin sensitivity in high-fat diet treated mice. These models also show significant reduction of fatty liver which was attributed to reduced circulating ceramides and reduced de novo lipogenesis (Jiang et al., 2015a(Jiang et al., , 2015bLi et al., 2013;Xie et al., 2017). However, also treatment with endogenous FXR ligands and GW4064 improves glucose homeostasis in lean mice and genetically obese insulin resistant mice (Cariou et al., 2006;Ma et al., 2006;Zhang et al., 2006). FXR agonistic treatment with OCA, GW4064, WAY-450 and tropifexor also improves fatty liver disease (Zhang et al., , 2009Tully et al., 2017;Cipriani et al., 2010;Hernandez et al., 2019). Moreover, constitutive hepatic FXR activation showed similar beneficial metabolic effects on insulin sensitivity and fatty liver . These studies are backed up by experiments in liver and intestine specific FXR knockout mice, which show that only liver specific knockout mice gain fatty livers, suggesting hepatic FXR is the site for liver fat reduction (Schmitt et al., 2015). In contrast to the above referenced work, intestinal specific FXR activation with fexaramine also shows improved insulin sensitivity and alleviated fatty liver, effects which however depend on additional TGR5 signaling and also on an intact gut microbiome (Fang et al., 2015;Pathak et al., 2018). In striking contrast, prolonged treatment with GW4064 has also been found to induce obesity and diabetes through reduced energy expenditure by a constantly repressed bile acid pool (Watanabe et al., 2011). Other relevant studies showed, that disruption of the entire FXR-Shp signaling axis in the liver leads to improved insulin resistance and alleviated fatty liver under high-fat diet conditions (Akinrotimi et al., 2017;Kim et al., 2017).
Overall, treatment with FXR modulating drugs shows effects for pan-FXR agonists, intestine selective FXR agonists as well as intestine specific FXR antagonist. In the most relevant setting of NASH with a marked inflammatory component or advanced NAFLD with a fibrotic component, direct hepatic FXR activation might be a logical approach. However, also FXR-induced FGF15/19 from the intestine and engineered FGF19 blunts hepatic inflammation and fibrosis in NASH (Zhou et al., 2017b), favoring intestinal FXR activation. From mouse studies it will not be possible to conclude which of the FXR modulating strategies will translate effectively into human conditions. The limitation with mouse models for NAFLD is strikingly exemplified with a chimeric model that harbors both, human and murine hepatocytes. Feeding Western diet results in pronounced steatosis in human hepatocytes only, whereas murine hepatocytes remained normal. Transcriptomic and metabolomic analyses showed that molecular responses diverged sharply between murine and human hepatocytes, demonstrating robust species differences in liver function (Bissig-Choisat et al., 2021). Therefore, further clinical studies in patients have to pin-point the above raised very relevant questions.

What we have learned from clinics
In clinical studies treatment success in NAFLD and in particular NASH is assessed by various non-invasive parameters (e.g. serum transaminases, MR liver fat imaging and non-invasive fibrosis scores including fibroelastometry) andas gold standard -by histological improvements. A typical study goal would test the effect of the drug on liver histology, defined as improvement of fibrosis with no worsening of NASH or resolution of NASH with no worsening of fibrosis. UDCA has been tested in the treatment of NASH but did not show any improvement of histology after 2 years (Lindor et al., 2004). A higher dose of UDCA had some, but marginal effects (Ratziu et al., 2011;Haedrich and Dufour, 2011). Of note, most UDCA in humans is conjugated to glycine-UDCA which is regarded as an FXR antagonist and has shown promising effects in mouse models (see above). OCA is the most advanced FXR agonist for NASH. In a small phase 2 study OCA resulted in improvement of insulin resistance in patients with NAFLD and type 2 diabetes, which has however, not been observed in subsequent studies (Mudaliar et al., 2013). In the FLINT trial OCA improved several histological NASH features, including reduction of liver steatosis, inflammation and fibrosis, but no improvement in insulin sensitivity (Neuschwander-Tetri et al., 2015). As main side effects dose dependent pruritus occurred and a pro-atherogenic lipid profile consisting of increased LDL and decreased HDL developed (Neuschwander-Tetri et al., 2015). Currently ongoing long term studies with OCA investigate in detail the effects on the lipid profile and safety issues in phase 3 studies (REGENERATE trial), which also include cirrhotic patients (REVERSE trial). Interim results from the REGENERATE trial after 18 month showed fibrotic improvement of 23% compared to 12% placebo treated patients in the higher OCA dose group but no resolution of NASH was accomplished (Younossi et al., 2019). Again LDL was increased, HDL decreased and pruritus occurred dose-dependently.
As non-steroidal FXR agonists cilofexor has been tested in NASH patients in a phase 2 study for 6 months, which showed an improvement of liver fat in MR imaging in approx. 40% of patients along with significant improvements of other non-invasive biochemical serum tests and markers of insulin resistance (Patel et al., 2020). However, non-invasive markers for fibrosis were not changed. The serum lipid profile did not change significantly, but dose dependent pruritus occurred. Recently, also another non-steroidal FXR agonist, MET409, showed a robust decrease of liver fat content along with only moderate changes in the lipid profile. Pruritus still was an issue (Harrison et al., 2021a). Several more clinical trials in NASH with other non-steroidal (tropifexor, nidufexor) and steroidal FXR (EDP-305) agonists are ongoing and only abstracts have been reported. A recent editorial by Claus Kremoser comprehensively summarizes all the relevant information (i.e. fat content, fibrosis, LDL, HDL and pruritus) of ongoing trials with FXR agonists in NASH and the reader is referred for further insights into ongoing clinical trials to this excellent overview (Kremoser, 2021).

Portal hypertension
Chronic liver diseases can progress and the further prognosis is significantly affected by the degree of fibrosis and portal hypertension. Portal hypertension determines the risk for complications such as ascites and variceal bleeding. Therefore, a treatment goal in advanced chronic liver disease is to lower portal hypertension, which is so far being achieved with non-selective β-blockers. FXR has been shown to lower vascular resistance by affecting several pathways required for endothelial remodeling and vasodilatation, including nitrite oxide generation via eNOS , inhibition of endothelin-1 generation (Li et al., 2010) and stimulation of the microcirculation via hydrogen sulfide production (Renga et al., 2009b). These effects are predicted -along with the antifibrotic properties of FXR activation -to lower portal hypertension. In several cirrhotic models OCA has been shown to reduce portal hypertension via eNOS restoration and enhanced nitrite oxide production (Verbeke et al., 2014;Mookerjee et al., 2015). In non-cirrhotic and cirrhotic rat models the non-steroidal FXR agonist PX20606 ameliorates portal hypertension by reducing liver fibrosis, vascular remodeling and sinusoidal dysfunction (Schwabl et al., 2017). In a recent study in NASH rats also the non-steroidal FXR agonist cilofexor significantly decreased portal hypertension and reduced liver fibrosis. While cilofexor seems to primarily decrease sinusoidal resistance in cirrhotic portal hypertension, the combination with propranolol additionally reduced mesenteric hyperperfusion (Schwabl et al., 2021).

Alcoholic steatohepatitis
Alcoholic liver diseases (ALD) come along with disrupted bile acid metabolism (Li and Chiang, 2020). The degree of bile acid levels is positively correlated with histological scores of ALD (Trinchet et al., 1994). Patients with ALD show elevated bile acid levels, but repressed de-novo bile acid synthesis, which correlates with disease severity. In line, on a molecular level CYP7A1 and the bile acid precursor 7α-hydroxycholestene-4-one (C4) are reduced and FGF19 is increased (Brandl et al., 2018). In mice, alcoholic diet also reduces Cyp7a1 and expands the bile acid pool. Repression of Cyp7a1 exacerbates, whereas overexpression of Cyp7a1 ameliorates alcoholic liver injury in mouse models (Donepudi et al., 2018). The expansion of the bile acid pool appears to be unique to ALD as it does not occur in NAFLD models (Li and Chiang, 2020). Whole body FXR knockout mice are more susceptible to ALD (Kong et al., 2019). However, this depends on ileal FXR and/or increased bile acid levels, since aggravated injury is not observed in hepato-specific FXR Ko mice (Zhang et al., 2018). In line with a role of bile acids and FXR for the severity of ALD, FXR activation with OCA or WAY-450 in mouse models of ALD reduces injury (Iracheta-Vellve et al., 2018;Livero et al., 2014;Wu et al., 2014). Noteworthy, also selective activation of intestinal FXR with fexaramine ameliorates injury . A clinical trial with OCA in patients with ALD has been started and prematurely halted (ClinicalTrials.gov Identifier: NCT02039219). Results have not been published, yet.

Hepatic encephalopathy
Hepatic encephalopathy is a severe neurological dysfunction in patients with advanced liver diseases characterized by neuropsychiatric and motor dysfunction. The pathogenesis is multifactorial and the gut microbiome and urea cycle are critically involved (Bloom et al., 2021). FXR may be involved at several steps. First, FXR is directly involved in amino acid catabolism and the urea cycle, which transforms ammonia from amino acid catabolism into urea for renal elimination (Massafra et al., 2017). Functionally, FXR activation promotes protein degradation and ammonium clearance in mice through direct binding to genes of amino acid degradation, ureagenesis and glutamine synthesis. FXR Ko mice have reduced plasma urea concentration after a protein rich diet, but accumulated precursors of ureagenesis, which is due to reduced hepatic expression of enzymes that regulate ammonium detoxification. This mechanism is important, since hyperammonia is a typical problem of patients with acute or chronic liver diseases and FXR activation might promote ammonium clearance in these patients (Massafra et al., 2017). Second, FXR critically determines the gut microbiome and changing the gut microbiome is a strategy to combat hepatic encephalopathy (Bloom et al., 2021;Wahlstrom et al., 2017). Third, only recently, a direct role for increased bile acid levels and hepatic encephalopathy has been acknowledged (Williams et al., 2022). In a mouse model of acute liver injury with a neurological decline, hepatic encephalopathy was delayed after intracerebroventricular infusion of FXR vivo-morpholinos. Bile acid administration increased neurological symptoms whereas cholestyramine feeding reduced it (McMillin et al., 2016). The exact downstream mechanisms of intracerebral FXR signaling remain, however, unknown.

Diarrhea due to primary bile acid malabsorption
Primary, or idiopathic, bile acid malabsorption (BAM) is characterized by increased fecal BAs and watery diarrhea. It responds well to bile acid sequestrants in the absence of secondary causes such as ileal resection (Vijayvargiya and Camilleri, 2018). The most likely reason for BAM is reduced levels of FGF19, which result in increased hepatic bile acid synthesis (Walters et al., 2009;Wong et al., 2012). Other factors involved in insufficient FGF19 action are modest variations in FGFR4, β-Klotho and ASBT (Johnston et al., 2016). Treatment with OCA increases circulating FGF19 levels and reduces C4 and bile acid levels. This treatment has been associated with improved stool frequency and consistency in patients with primary BAM (Walters et al., 2015). Another FXR agonist, GW4064, attenuated Cl− secretory responses in the colonic epithelium and thereby regulated fluid and electrolyte transport upon various secretory stimuli. These anti-secretory effects may also support anti-diarrheal action (Mroz et al., 2014). In contrast to the bile acid sequestrants, FXR agonists do not show the typical side effects of nausea and only need to be administered once daily. Therefore, FXR agonists represent a likely future treatment option for these patients.

Kidney injury
FXR is highly expressed in the kidneys, but its functional role there is by far less well understood than its hepatic and intestinal roles (Libby et al., 2021). Within the kidneys FXR is ubiquitously distributed with the highest expression in the cortex . Functionally, FXR is involved in the regulation of urine volume as its activation increases urinary concentrating capacity mainly via up-regulating AQP2 expression in the collecting ducts . In the kidney FXR plays a protective role against fibrotic events and may thus have potential as drug target for diabetic nephropathy and nephrosclerosis (Herman-Edelstein et al., 2018;Jiang et al., 2007). In mouse models of diet induced obesity and diabetic nephropathy FXR agonism with GW4064 and CA reduces renal Srebp1 expression and thereby shifts lipogenesis towards fatty acid oxidation and lipid catabolism (Jiang et al., 2007). FXR agonism further reduced renal expression of pro-fibrotic growth factors, pro-inflammatory cytokines, and oxidative stress enzymes and decrease glomerulosclerosis, tubulointerstitial fibrosis, and proteinuria (Jiang et al., 2007). These effects were not observed in FXR Ko mice and renal injury was even aggravated in FXR Ko mice (Wang et al., 2010). Pretreatment with OCA also protected from renal ischemia-reperfusion injury via reduced oxidative stress. This protective effects required FXR dependent translocation of Nrf2 into the nucleus (Gai et al., 2017). FXR agonism has also been shown to protect from kidney fibrosis in various renal disease models (Libby et al., 2021). FXR mediated inhibition of yes-associated protein (YAP) signaling appears to be a central renal anti-fibrotic mechanism Li et al., 2019).

Primary cancers of the liver
FXR is an established regulator of liver growth and proliferation (Huang et al., 2006). It is also well noted that FXR deficiency correlates with the development of primary cancers of the liver, that is both hepatocellular carcinoma (HCC) and cholangiocellular carcinoma (CCC) and that FXR agonism counteracts carcinogenesis (Yang et al., 2007;Kim et al., 2007b;Girisa et al., 2021;Attia et al., 2017;Di Matteo et al., 2019;Erice et al., 2018). FXR expression levels are reduced in both types of cancer (Su et al., 2012;Obama et al., 2005), which may lead to a (micro-)cholestatic environment with exposure to toxic bile acids and subsequent establishment of a chronic inflammatory condition. In CCC the downregulation of FXR comes along with the upregulation TGR5, which is highly expressed in cholangiocytes but absent in hepatocytes (Dai et al., 2011;Reich et al., 2016). Activation of FXR with different agonists such as CDCA, OCA, GW4064 and precursors of cilofexor had different molecular effects in cancerogenic cell lines and animals but all of them resulted in a tumor suppressive gene signature and a partly restoration of FXR expression. FXR agonists reduce the time hepatocytes are exposed to toxic bile acids and thereby limit liver injury (Kjaergaard et al., 2021) and also improve liver fibrosis, often the preliminary stage to liver cirrhosis and cancer. On a molecular level FXR agonism interferes with the NFκb inflammatory pathway and downregulates oncogenes such as different cyclins, STAT3, gankyrin and disrupts the β-catenin-TCF4 complex, subsequently reducing transcription, proliferation, migration and invasion of tumor cells. In addition, malignant cells do not enter S-phase as frequently and instead rest in a pre-apoptotic phase (Attia et al., 2017;Erice et al., 2018). Moreover, FXR balances the ratio of tumorigenic and tumor suppressive miRNAs (Girisa et al., 2021). Along this line, the FXR antagonist NorCholic acid promotes tumor progression (Gong et al., 2021). FXR also plays a role in many other cancer types e.g. breast, lung, pancreatic or esophageal cancer, where FXR is usually not expressed. FXR was found to be upregulated in these cancers as opposed to liver cancer (Girisa et al., 2021). If FXR agonists hold future promise as drugs with anti-cancer properties remains to be determined. It should be kept in mind that FXR activation in the intestine promotes FGF19 secretion. FGF19 along with the FGFR4 receptor have a clear mitogenic potential and are implicated in HCC development (Krones and Wagner, 2016).

Summary and conclusion
From a molecular viewpoint FXR modulation represents an ideal strategy for the treatment of chronic cholestatic and metabolic liver diseases such as NAFLD. FXR activation not only targets imbalances of "macro-cholestasis" in chronic cholestasis and supposedly "microcholestasis" in NAFLD, but also has anti-inflammatory and anti-fibrotic function, which are both pathophysiological mechanisms to perpetuate a chronic disease course. However, unrestricted FXR activation may come along with some side effects, which may limit its usability. Therefore, more restricted FXR activation appears to be an opportunity to combine required beneficial FXR effects with limited side effects. Preclinical models point out, that tissue selective FXR activation in the intestine may be sufficient for the treatment of metabolic liver diseases and maybe also cholestasis. Even more, intestinal FXR antagonism appears as a strategy for treating NAFLD. However, comparative studies are missing and many open questions remain, the most important ones being, whether activation of intestinal FXR might be in fact sufficient for treatment and to limit side effects, or if rather a concerted intestinal and hepatic action is required? If intestinal FXR activation is sufficient, how much of the effect is contributed by FGF19 and may then treatment with non-mitogenic FGF19 mimetics be an even more preferred alternative treatment option? How should the side effects be weighted in light of cholestasis with low atherogenic risk and NAFLD with a higher proatherogenic risk profile, i.e. do we need separate FXR agonists for cholestasis and other FXR agonists or even FXR antagonists for NAFLD? Do we need a second treatment for the side effects already combined with FXR agonists, e.g. anti-pruritogenic medication such as fibrates, or LDL lowering drugs such as statins? Moreover, are FXR agonists acting differently depending on the metabolic background (e.g. fed versus fasted, lean versus obese, cholestatic versus non-cholestatic conditions)?
Another interesting question is, if biomarkers can help us in future to predict the response to FXR modulation and make FXR treatment personalized? C4 levels may be used as predictor of CYP7A1 activity and may be a relevant estimate for an already repressed endogenous bile acid synthesis, thus helping to stratify treatment with FXR activating compounds (Molinaro and Marschall, 2022). FGF19 levels are a marker for intestinal FXR activity. The FXRα2 isoform predicts a metabolically favorable response to pan-FXR agonistic treatment (Ramos Pittol et al., 2020). The bile acid composition changes with disease stages and thereby modulates endogenous FXR activity and activity to other bile acid sensing receptors such as TGR5. Moreover, the gut microbiome and BSH producing strains significantly impact on FXR agonistic or antagonistic properties. Finally, FXR binding is strikingly different among different metabolic conditions and very likely installs FXR dependent pathways which then might produce measurable biomarkers. Measurement of these parameters before, during and after FXR treatment may help in future to better stratify therapeutic approaches.
Pre-clinical animal models have paved the way, but these final questions should be answered in clinical settings.

Financial Support
The current work was supported by the Austrian Science Fund (FWF) #P30482 to MW.

Declaration of competing interest
KP, GZ, HUM and MW have nothing to disclose of relevance for this manuscript.