Abstract

The farnesoid X receptor (FXR) is a nuclear receptor that controls bile acid, lipid, and cholesterol metabolism. FXR-targeted drugs have shown promise in late-stage clinical trials for non-alcoholic steatohepatitis. Herein, we used clinical results from our first non-steroidal FXR agonist, 4-[2-[2-chloro-4-[[5-cyclopropyl-3-(2,6-dichlorophenyl)-4-isoxazolyl]methoxy]phenyl]cyclopropyl] benzoic acid (Px-102), to develop cilofexor, a potent, non-steroidal FXR agonist with a more manageable safety profile.

Px-102 demonstrated the anticipated pharmacodynamic (PD) effects in healthy volunteers but caused a 2-fold increase in alanine aminotransferase (ALT) activity and changes in cholesterol levels. These data guided development of a high fat diet mouse model to screen FXR agonists based on ALT and cholesterol changes. Cilofexor was identified to elicit only minor changes in these parameters. The differing effects of cilofexor and Px-102 on ALT/cholesterol in the model could not be explained by potency or specificity, and we hypothesized that the relative contribution of intestinal and liver FXR activation may be responsible. Gene expression analysis from rodent studies revealed that cilofexor, but not Px-102, had a bias for FXR transcriptional activity in the intestine compared with the liver. Fluorescent imaging in hepatoma cells demonstrated similar subcellular localization for cilofexor and Px-102, but cilofexor was more rapidly washed out, consistent with a lower membrane residence time contributing to reduced hepatic transcriptional effects. Cilofexor demonstrated antisteatotic and antifibrotic efficacy in rodent models and antisteatotic efficacy in a monkey model, with the anticipated PD and a manageable safety profile in human phase I studies.

SIGNIFICANCE STATEMENT Farnesoid X receptor (FXR) agonists have shown promise in treating non-alcoholic steatohepatitis and other liver diseases in the clinic, but balancing efficacy with undesired side effects has been difficult. Here, we examined the preclinical and clinical effects of the first-generation FXR agonist, 4-[2-[2-chloro-4-[[5-cyclopropyl-3-(2,6-dichlorophenyl)-4-isoxazolyl]methoxy]phenyl]cyclopropyl] benzoic acid, to enable the selection of an analog, cilofexor, with unique properties that reduced side effects yet maintained efficacy. Cilofexor is one of the few remaining FXR agonists in clinical development.

Introduction

Farnesoid X receptor (FXR; NR1H4) is a nuclear hormone receptor expressed in the gastrointestinal tract and liver that functions as the main regulator of bile acid (BA) homeostasis through the transcription of genes involved in BA, triglyceride, cholesterol, and lipoprotein metabolism. FXR is also expressed in the kidneys, adrenal glands, and ovaries, but its function in these organs is less well defined (Forman et al., 1995; Makishima et al., 1999; Parks et al., 1999; Lee et al., 2006; Calkin and Tontonoz, 2012).

In the intestine, the hormone FGF19 is the main target gene of FXR, and BAs such as chenodeoxycholic acid (CDCA) act as endogenous FXR agonists, causing the release of FGF19 into portal circulation. On hepatocytes, FGF19 binds to heterodimeric cell surface receptors comprising FGF receptor 4 and β-klotho, which stabilizes and increases the activity of the transcriptional repressor small heterodimer partner (SHP). The repression of cytochrome P450 (CYP) 7A1 by SHP reduces the production of 7α-hydroxy-4-cholesten-3-one (C4), which is the rate-limiting step for BA synthesis (Goodwin et al., 2000; Holt et al., 2003; Inagaki et al., 2005; Kliewer and Mangelsdorf, 2015), and thereby generates a negative feedback loop to control BA levels. In the liver, direct FXR agonism and FGF19 signaling have distinct but overlapping functions in the suppression of BA synthesis. The former also increases the transcription of SHP (Lu et al., 2000; Kliewer and Mangelsdorf, 2015), along with many hepatoprotective effects (Modica et al., 2009; Jiang et al., 2013; Prakash et al., 2015).

Modulation of the FXR–FGF19 pathway using FXR agonists [or conceivably antagonists, as discussed in (Wang et al., 2018; Panzitt et al., 2022; Ma et al., 2023)] has emerged as a therapeutic strategy for cholestatic liver diseases (in which increased BA levels are implicated in promoting tissue damage, hepatic inflammation, and fibrosis) (Woolbright and Jaeschke, 2012; Carey et al., 2015; Kowdley et al., 2020), and for non-alcoholic fatty liver disease (NAFLD) and non-alcoholic steatohepatitis (NASH). Treatment of cholestatic diseases is limited and there is no approved treatment of NAFLD/NASH. The steroidal FXR agonist obeticholic acid, is approved for the treatment of primary biliary cholangitis and, in clinical trials of NAFLD and NASH, decreased liver fat and liver fibrosis (Neuschwander-Tetri et al., 2015; Younossi et al., 2019). Decreasing liver fibrosis in NASH is important as it is the only histologic parameter associated with positive outcomes (Angulo et al., 2015; Dulai et al., 2017). However, decreased liver fat is often accompanied by increased liver enzymes, decreased high-density lipoprotein-cholesterol (HDL-C), increased low-density lipoprotein-cholesterol (LDL-C) and/or pruritus in clinical studies with other FXR agonists. These class effects have resulted in the discontinuation of treatment and led to a debate on the ideal pharmacological profile for agents targeting this pathway (Kremoser, 2021). A more nuanced agonism of FXR is likely required, as expression of constitutively active FXR in mouse liver and intestine showed spontaneous liver toxicity and increases in alanine aminotransferase (ALT) activity (Cheng et al., 2015).

An important consideration in this debate is the relative contribution of intestinal and liver FXR activation. Obeticholic acid, for example, activates both and, as a bile acid analog, undergoes extensive enterohepatic recirculation. Intestinally selective FXR agonists have suppressed BA synthesis and lowered triglycerides in plasma and the liver (Zhou et al., 2017; Harrison et al., 2018; Pathak et al., 2018), leading to benefits on liver steatosis, fibrosis, and the organism’s metabolic state through induction of intestinal fibroblast growth factor (FGF) 19 (or its rodent homolog FGF15) (Fang et al., 2015). There is less certainty regarding the benefit of liver FXR activation (Stroeve et al., 2010; Kong et al., 2012; Schmitt et al., 2015; Massafra et al., 2018). It is also unclear which structural class of FXR agonists might be best suited for the optimal pharmacological balance (Verbeke et al., 2017).

Herein, we first describe 4-[2-[2-chloro-4-[[5-cyclopropyl-3-(2,6-dichlorophenyl)-4-isoxazolyl]methoxy]phenyl]cyclopropyl]benzoic acid (Px-102) which, to our understanding, was the first non-steroidal FXR agonist candidate tested in humans. The safety, pharmacokinetics and pharmacodynamics (PD) of Px-102 were evaluated in a first-in-human dose ranging study in healthy volunteers. Px-102 induced dose-dependent increases in plasma FGF19 and reductions in C4, demonstrating FXR target engagement. However, clinical development was halted as a result of significant increases in ALT, aspartate aminotransferase (AST), and alkaline phosphatase in plasma. We subsequently devised a preclinical model to screen for second generation agonists that would harness the beneficial effects of FXR induction but avoid unwanted effects on liver enzymes. This led to the identification of cilofexor, an analog that confers biased agonism of intestinal FXR versus liver FXR. Cilofexor reduced hepatic fibrosis and steatosis in rodent and primate models, respectively, and has since advanced into a phase IIb trial in NASH (NCT04971785).

Materials and MethodsFluorescent Resonance Energy Transfer Assay

A biochemical ligand-dependent nuclear receptor-cofactor peptide interaction assay was conducted as described (Hambruch et al., 2012), with biotinylated SRC-1 peptide b-CPSSHSSLTERHKILHRLLQEGSPS-COOH (0.4 µM) and purified FXRaa187–472-LBD fused to glutathione S-transferase (2.5 ng) together with 200 ng streptavidine-allophycocyanin and 6 ng europium labeled-anti-glutathione S-transferase as reagents in 25 µl assay buffer (20 mM Tris/HCl at pH 7.5; 5 mM MgCl2; 60 mM KCl; 1 mM dithiothreitol; 0.9 g/L bovine serum albumin). Fluorescent resonance energy transfer values are given in nM and Efficacy (%) is the maximum efficacy of the compound relative to 3-[2-[2-chloro-4-[[3-(2,6-dichlorophenyl)-5-(1-methylethyl)-4-isoxazolyl]methoxy]phenyl]ethenyl]benzoic acid (means of at least two assays).

Mammalian One Hybrid Reporter Assay

The cellular FXR assay was performed as described (Hambruch et al., 2012), with human FXRaa187–472-LBD C-terminally fused to a Gal4 DNA-binding domain under transcriptional control of the cytomegalovirus promoter in pCMV-BD (Stratagene). This chimeric plasmid construct was transiently co-transfected into HEK293 cells together with pFRluc (contains a synthetic promoter with five tandem repeats of the yeast GAL4 binding sites that control expression of the Photinus pyralis luciferase gene, Stratagene) encoding a Gal4 promoter-driven firefly luciferase. Cells were treated with serial dilutions of the test compound. EC50 values were calculated from at least three experiments. Percentage efficacy is the extrapolated maximum signal generated by the test compound in a dose-response dilution. Efficacy of 3-[2-[2-chloro-4-[[3-(2,6-dichlorophenyl)-5-(1-methylethyl)-4-isoxazolyl]methoxy]phenyl]ethenyl]benzoic acid in the respective assay is set as 100%.

Cell-Based Subcellular Localization Assays

All compounds were dissolved in 100% dimethylsulfoxide and diluted in growth medium to a final concentration containing 0.2% dimethylsulfoxide. For organelle localization and compound uptake, cells were imaged using a laser scanning confocal microscope (Leica TSP). Fluorescence signals were detected at 10–20% excitation power to prevent bleaching using laser lines 488 nm/510 nm and 561 nm/595 nm for excitation and emission, respectively.

The human hepatoma cell line Huh7 cells were incubated overnight with fluorescently labeled cilofexor (cilofexor-F) or Px-102 (Px-102-F) and organelle markers. Endoplasmic reticulum (ER) and Golgi apparatus were labeled using ER Tracker Red and Bodipy TR Ceramide (Thermo Fisher Scientific), respectively. Alternatively, CellLight reagents ER-red fluorescent protein (RFP) and Golgi-RFP (Thermo Fisher Scientific) were used following manufacturers’ protocols. We chose Huh-7 cells as opposed to primary human hepatocytes or enterocytes as Huh-7 cells are large with good resolution of subcellular structures, making them advantageous for fluorescence microscopy.

For agonist uptake, cells were incubated with organelle marker-supplemented growth medium for 30 minutes, then the medium was replaced by growth medium supplemented with the cilofexor-F or Px-102-F and organelle marker microscopic images were recorded at different intervals over 60 minutes.

To compare uptake and localization between both labeled FXR agonists, cells had to express ectopic and transient sodium taurocholate cotransporting polypeptide. Sodium taurocholate cotransporting polypeptide was fused C-terminally to detection marker mRuby (a monomeric RFP) with and without a linker containing a P2A cleavage site. Cells expressing the fusion protein for 48 hour after transfection were split into microscope chambers and grown in medium supplemented with cilofexor-F. Thereafter, cells were observed under the microscope.

Cells in washout experiments were observed using a Zeiss Cellobserver Epifluorescence microscope equipped with green fluorescent protein and RFP filters. Images were recorded applying the Zeiss ZEN software. Here, cells were allowed to grow in medium supplemented with fluorescent FXR agonists for 8 hours. Cells were then washed with five volumes of medium and images taken immediately. Cells were returned into the incubator overnight and imaged again the next morning and subsequent afternoon.

Nuclear Receptor and TGR5 Assays

To check for possible cross reactivity on other nuclear receptors, Px-102 and cilofexor were tested for activity against 18 human nuclear receptors. ERalpha, ERbeta, AR, GR, MR, PR, PPARalpha, PPARbeta/delta, PPARgamma, RXRalpha, RARalpha, LXRalpha, LXRbeta, TRalpha, VDR, ERRgamma, CAR, and PXR receptor activity in comparison with the respective control ligands was monitored in a cellular GAL4-driven reporter gene assay system. All assays were done in HEK293 cells by transfecting the reporter plasmid pFRluc and pCMV-BD (for fusions of nuclear receptor ligand-binding domains to the DNA-binding domain of the yeast protein GAL4). The mammalian one hybrid assay for AR was modified by the addition of a plasmid constitutively expressing a PGC1alpha fragment (aa1 to aa677). The assays were run in agonist mode and were done in 11 pt dose-response experiments in biological triplicates. EC50 values were calculated from at least three experiments. Percentage efficacy is the extrapolated maximum signal generated by the test compound in a dose-response dilution. Efficacies of known nuclear receptor agonists as positive controls were set as 100% in the respective assays.

The cell-based assay to evaluate agonism of TGR5 (also known as G protein-coupled bile acid receptor 1) was based on the Promega Glosensor cAMP Assay (cAMP is the key second messenger involved in signal transduction mediated by TGR5). The assay uses a genetically modified form of firefly luciferase into which a cAMP-binding protein moiety has been inserted, and binding of cAMP results in increased light output. cAMP Hunter CHO-K1 G protein-coupled bile acid receptor 1 (95-20049C2, DiscoveRx) cells were co-transfected with the plasmid pGloSensorTM-22F cAMP Plasmid (E2301, Promega). The resulting clones were screened for stable genomic integration of the vector. Following two selection rounds, one stable clone was isolated. Cilofexor was assayed over a concentration range of 0.0005 to 10 µM; lithocholic acid served as the positive control.

Intestinal and Hepatic Gene Expression

For Fig. 4, gene expression of FXR targets was performed as described (Schwabl et al., 2021). For Fig. 6, ileum SHP and FGF15 gene expression was measured by reverse transcriptase polymerase chain reaction using Taqman probes SHP: Mm00442278_m1 (Cat #4331182) and FGF15: Mm00433278_m1 (Cat #4331182) and normalized to housekeeping gene GAPDH: Mm99999915_g1 (Cat #4331182). Liver SHP gene expression was measured by Nanostring all nCounter panel following the manufacturer’s standard procedure (NanoString Technologies, Inc., Seattle, WA).

Clinical Chemistry

For the clinical chemistry analysis from plasma, ALT, cholesterol and triglycerides were determined using a fully automated bench top analyzer (Respons910, DiaSys Greiner GmbH, Flacht, Germany) with system kits provided by the manufacturer or equivalent. BA in plasma were mixed with three volumes of cold acetonitrile, the precipitate removed by centrifugation, and the supernatant was transferred to a glass HPLC injection vial. BA were separated on a Reprosil-pur-phenyl analytical column (Dr Maisch, Germany) and determined by negative ion reverse phase high-performance liquid chromatography tandem mass spectrometry. For the representation of major changes in BA composition in mice over time, the sum of concentrations of all CDCA-derived BA species (CDCA, muricholic acid, and their taurine conjugates) were divided by the sum of concentrations of all cholic acid (CA)-derived species (CA, DCA, and their taurine conjugates). BA in feces were determined as described (Grundy et al., 1965; Miettinen et al., 1965).

Other Assays

FGF19 plasma levels were quantitated by ELISA (BioVendor, Heidelberg, Germany, cat.# RD191107200R) according to the manufacturer’s instructions. Formalin-fixed paraffin-embedded slides were stained with Picrosirius red (PSR) with quantification using the Visiopharm platform.

Animal Models

All animal experiments were conducted at Association of Assessment and Accreditation of Laboratory Animal Care International-accredited sites, were approved by the respective site’s Institutional Animal Care and Use Committee, and performed according to a prewritten protocol. Animals were housed in rooms with a 12:12 hour light/dark cycle and full access to food and water. Sample sizes were prospectively determined by power analysis to ensure the minimum number of animals needed to achieve statistical significance.

For mouse pharmacology studies, the streptozotocin high-fat diet mouse model was used as described (Fujii et al., 2013). FXR−/− male mice were kindly provided by Barry Forman (City of Hope, California, USA). For the high fat diet (HFD) model, male, ∼8-week-old C57BL/6J, Balb/c, FVB, NOD, and FXR−/− mice or Sprague-Dawley rats were maintained on an HFD with 60% of kcal from fat and 1% (w/w) from cholesterol (Surwit, Ssniff, Soest, Germany) for up to 24 days. Oral administration of vehicle or 3, 10, or 30 mg/kg compound, as indicated, was initiated at the time of HFD induction. Food consumption was not affected by FXR agonist administration (data not shown). This study design was selected as it provided a strong signal for ALT, AST, cholesterol, and total BA within a time frame that was short enough to be amenable to serve as a compound screen. For the HFD plus carbon tetrachloride “two-hit” model, C57BL/6J male mice were maintained on an HFD (Research Diet D12492) for 16 weeks and administered 0.5 μl/g CCl4 in olive oil intraperitoneally twice weekly and cilofexor at 0 (vehicle: 0.5% carboxymethylcelluose and 1% ethanol in 50 mM Tris buffer, pH 8.0), 10. or 30 mg/kg q.d. cilofexor orally for the final 4 weeks. For rat bile duct cannulation studies, male Sprague-Dawley rats (n = 3) with bile duct and jugular vein cannulae were administered cilofexor and Px-104 (the (-)-enantiomer of the racemic Px-102) at 1 mg/kg i.v. Bile was collected at indicated times, and cilofexor, Px-104, and the glyco- and tauro-conjugates of Px-104 were quantified by liquid chromatography with tandem mass spectrometry (LC/MS-MS). For the monkey PD study, 3–5 kg male cynomolgus monkeys were individually housed and administered a single dose of cilofexor orally at 5 mg/kg or i.v. at 1 mg/kg. For monkey pharmacology studies, 3–5 kg male cynomolgus monkeys were comingled and provided an HFD (Research Diets #C16062304) for 16 weeks and orally administered vehicle, 30 mg/kg cilofexor, or 5 mg/kg Px-102 for the final 12 weeks.

Participants and Study Design

From healthy human volunteers, informed consent in writing was obtained and the study protocol conformed to the ethical guidelines of the 1975 Declaration of Helsinki.

Clinical Phase I, Randomized, Placebo-Controlled Safety Study for Px-102.

The protocol was approved by the Ethics Committee of the Aerztekammer Nordrhein, Germany. The phase I studies are registered on ClinicalTrials.gov: identifier NCT01998659 for the single-ascending dose (SAD) study and NCT01998672 for the multiple-ascending dose (MAD) study. Px‐102 was suspended in 0.1% polysorbate 80, 0.5% polyvinyl‐pyrrolidone, 1.5% benzyl alcohol, and 0.3 ppm denatonium benzoate anhydrous (Bitrex) in purified water. A liquid suspension of Px-102 or placebo was administered orally to fasted, male, healthy patients.

Clinical Phase I, Randomized, Double-Blind, Placebo-Controlled Safety Study for Cilofexor.

The phase I study at Seaview Research, Miami, Florida, USA, is registered on ClinicalTrials.gov, identifier: NCT02654002 (Supplemental Fig. 1). Healthy male and non-pregnant, non-lactating female patients received placebo on Day −1 and 10 mg (n = 12), 30 mg (n = 12), 100 mg (n = 12), or 300 mg (n = 12) cilofexor on Day 1. After a 5-day washout, volunteers received the same doses of cilofexor q.d. for 14 consecutive days. Blood draws were taken for FGF19, C4, lipid, cholesterol and liver enzyme (ALT, AST, or gamma-glutamyl transferase) analysis.

Statistical Analysis

Data are expressed as the means ± S.D. Differences between specific groups were determined using an unpaired Student’s t test and analyzed by the Instat software package (GraphPad Software, San Diego, California, USA). Data without error bars represent pooled samples. Significance was indicated with *** for p < 0.001, ** for p < 0.01, and * for p > 0.01.

ResultsPx-102 is a Potent, Selective, Non-Steroidal FXR Agonist

Using 3-[2-[2-chloro-4-[[3-(2,6-dichlorophenyl)-5-(1-methylethyl)-4-isoxazolyl]methoxy]phenyl]ethenyl]benzoic acid (Maloney et al., 2000) as a template, we synthesized derivatives of the substituted isoxazole core to yield a potent FXR agonist with improved oral bioavailability and selected Px-102 (Fig. 1A), which had sub-micromolar potencies in biochemical and cellular assays (Fig. 1, B and C) for further evaluation. Px-102 proved to be selective for FXR in a panel of 18 human nuclear receptors (Supplemental Table 1) and did not activate TGR5 (data not shown). The oral bioavailability of Px-102 was approximately 50% in most species tested, with a 6-fold liver-to-plasma ratio in mice. Px-102 dose-dependently reduced the NAFLD activity score (NAS) and hepatic fibrosis as measured by percent fibrotic area in PSR-stained liver sections (Supplemental Fig. 2) in the streptozotocin high-fat diet mouse model of liver fibrosis (Fujii et al., 2013). The efficacy of Px-102 in this model compared favorably to the effects of the antidiabetic agent pioglitazone and obeticholic acid; both have shown some clinical benefit in certain NASH patient populations. In a CCl4 rat model, Px-102 yielded potent reductions in measures of hepatic fibrosis and improved portal hypertension (Schwabl et al., 2017).

Fig. 1.

Px-102 is a potent, selective, non-steroidal agonist of FXR. (A) Structure of Px-104 (Px-102 is the racemate) and (B and C) Px-102 and Px-104 potencies in the fluorescence resonance energy transfer biochemical and mammalian one hybrid cellular assays.

Px-102 Potently and Dose-Dependently Induced FGF19 in Humans, but also Increased Markers of Liver Injury

The safety, pharmacokinetic, and PD of Px-102 were evaluated in a first-in-human SAD study. Plasma FGF19 and C4 were selected as PD markers. In a SAD design (Fig. 2A), seven doses of Px-102 (0.15–4.5 mg/kg) were administered as a liquid suspension. Px-102 plasma exposure was approximately dose linear (Fig. 2, B and C) and dose-dependent increases in FGF19 (Fig. 2D) and reductions in C4 were observed.

Fig. 2.

Px-102 causes dose-dependent increases in serum FGF19 but also in ALT and AST. (A) Clinical study design, (B and C) Px-102 exposures, (D and E) plasma FGF19 levels in SAD and MAD studies, (F) serum ALT and AST activities, (G) plasma and fecal BA, and (H) ratio of CDCA/CA-derived BA species. Data are presented as mean ± S.D.; *p < 0.05, **p < 0.01, ***p < 0.001 versus placebo by unpaired t test.

In a 7-day MAD study, volunteers were administered 0.5, 1.0, or 1.5 mg/kg Px-102 q.d., and FGF19 levels were analyzed over 24 hours (Fig. 2E). The daily amplitude of the FGF19 response was maintained for the 0.5 and 1.0 mg/kg Px-102 doses, while 1.5 mg/kg Px-102 resulted in an increase of effect over time, indicating possible accumulation of parent drug or active metabolites. Px-102 at doses of 1.5 mg/kg also resulted in significant increases in plasma ALT, AST, and alkaline phosphatase (Fig. 2F). Px-102 treatment resulted in a reduction of BAs in plasma and feces (Fig. 2G). The reduction in plasma BAs was less prominent than in feces but the qualitative changes in BA composition were far more pronounced, i.e., an increase in the ratio of CDCA/CA-derived BA species (Fig. 2H), which accords with FXR’s function as master regulator of BA synthesis. Changes in total cholesterol, LDL-C, and HDL-C were recorded but showed no clear dose-dependency over time (Supplemental Fig. 3). Furthermore, two of 12 volunteers in the 1.0 mg/kg group and three of 12 in the 1.5 mg/kg group experienced mild-to-moderate pruritus. These were deemed treatment related but resolved within 7 days after the final dose. However, two isolated incidences of ventricular extrasystoles in the SAD and one volunteer experiencing multiple extrasystoles over 24 hours in the 1.5 mg/kg arm in the MAD were regarded as potentially drug-related cardiac safety signals. Therefore, clinical development of Px-102 was halted.

Identification of a Preclinical Model that Recapitulates the Px-102-Induced Increases in ALT and Shifts in BA Composition Observed in Humans

None of the standard toxicology studies predicted the increase in liver enzymes seen in humans (data not shown); therefore, we evaluated this effect in rodent models. We discovered that C57BL/6J mice maintained on an HFD recapitulated the increase in ALT when administered 10 mg/kg Px-102 for 17 days (Fig. 3A). Other mouse strains or Sprague-Dawley rats did not show this effect. Furthermore, studies with C57BL/6 mice on standard chow and FXR−/− mice on an HFD revealed that the ALT increase was dependent on both the HFD and functional FXR (Fig. 3B). We then measured FXR-induced changes in plasma BAs and compared them with increases in plasma ALT and decreases in plasma cholesterol over 24 days. Treatment with 30 mg/kg Px-102 q.d. induced pronounced changes, most notably an increase in the ratio of CDCA/CA-derived BA species and conjugated/unconjugated BAs (Fig. 3C and see Supplemental Fig. 4 for individual BA species), recapitulating what had been seen in the MAD study. We further observed that the temporal onset of the BA ratio changes coincided with ALT increase and cholesterol lowering (Fig. 3D), suggesting that these parameters are all linked and affected by FXR.

Fig. 3.

Identification of rodent screening model. (A) ALT activities in different mouse and rat strains maintained on HFD and administered vehicle or 10 mg/kg Px-102 for 17 days. (B) Effect of 17 days of 10 mg/kg Px-102 dosing on ALT activities in C57BL/6J or FXR−/− mice maintained on standard chow or HFD, (C) ratio of plasma CDCA/CA-derived BA species in mice on HFD treated at 30 mg/kg Px-102 for 24 days, and (D) ALT, ratio of plasma CDCA/CA-derived BA species, and cholesterol in C57BL/6J mice on HFD administered 30 mg/kg Px-102 for 24 days. Data are presented as mean ± S.D.; *p < 0.05, **p < 0.01, ***p < 0.001 versus vehicle by unpaired t test.

HFD Mouse Model Identified Cilofexor as an FXR Agonist with Differential Effects in Liver and Intestine

We then used the HFD model to identify FXR agonists that would dissociate the beneficial induction of intestinal FGF15 from changes in BA composition, plasma cholesterol, and liver enzymes. Agonists with different biochemical and cellular potencies (Supplemental Fig. 5) were screened at 10 mg/kg q.d. for 10 days. Marked phenotypic differences emerged: cilofexor showed no ALT increases, while GS-713514 displayed the most pronounced effect. The same rank order of compounds was observed for the reduction of plasma cholesterol (Fig. 4A). We next determined the mRNA expression levels of FXR target genes in the liver (Ostβ) and ileum (FGF15) by quantitative reverse transcriptase polymerase chain reaction. There were significant differences among the agonists, with the rank order of changes in mRNA expression identical to ALT or cholesterol changes (Fig. 4B). Furthermore, cilofexor showed the lowest ratio of OSTβ liver expression to FGF15 ileal expression (Fig. 4C). We extended the quantitative reverse transcriptase polymerase chain reaction analysis to other known FXR target genes such as Cyp7A1 and Cyp8B1 in the liver (Fig. 4D). The rank order of magnitude of effects on these genes remained the same as for FGF15 and Ostβ, with cilofexor representing the weakest and GS-713514 the strongest FXR-dependent transcriptional regulator. The sequence of the strength of the transcriptional effects was mirrored by the induction of plasma BA composition. The increase in the ratio of CDCA/CA-derived BA species was most pronounced for GS-713514 and least distinct for cilofexor (Fig. 4E). Interestingly, the in vivo transcriptional effects were more closely predicted by the cellular potencies than the biochemical potencies (Supplemental Fig. 5).

Fig. 4.

Rodent screening model identifies cilofexor as an FXR agonist with differential effects in liver and intestine. (A) Plasma ALT activities and serum cholesterol levels in C57BL/6J mice on HFD after 10 days of oral treatment with FXR agonists (n = 8, 10 mg/kg). (B) Expression of the FXR targets Ostβ in the liver and FGF15 in the ileum at 4 hours post dose on Day 10. (C) Ratios of liver over intestinal target gene activation. (D) Expression of the liver FXR related genes Cyp8B1 and Cyp7A1 at 4 hour post dose on Day 10. (E) Ratio of CDCA- to CA-derived BA species in plasma on Day 10. The individual bile salts were analyzed and the ratio of plasma CDCA/CA-derived bile acids was calculated and graphed. (F) Correlation of tissue exposure and gene regulation (ileum/FGF15 [right]), (liver/Ostβ [left]) of 3 mg/kg and 10 mg/kg Px-104 or cilofexor treatment over 10 days. White circles denote Px-104 and black triangles denote cilofexor treatment. Bar graphs are means ± S.D. Significance was calculated using Student’s unpaired t test (*p ≤ 0.05; **p ≤ 0.01; ***p ≤ 0.001).

Px-104 and Cilofexor Display Substantial Differences in Transcriptional Efficiency

Cilofexor was selected as the most promising candidate from the FXR agonists screened. Cilofexor proved to be selective for FXR in a panel of 18 human nuclear receptors (Supplemental Table 1) and did not activate TGR5 (Supplemental Fig. 6). In the HFD mouse model at 3 and 10 mg/kg over 10 days, Px-104 showed dose-dependent differences in compound levels in each tissue that corresponded well to the changes in gene expression (Fig. 4F). Cilofexor levels in the ileum were more variable, but even exposures exceeding those of Px-104 did not lead to similar increases in FGF15 expression. In general, cilofexor levels in the liver were lower than those of Px-104, but even at comparable liver exposures, the magnitude of Ostβ induction was less. Thus, two compounds with similar potency, but different amphiphilicity, demonstrated major differences in the relationship between tissue level and FXR transcriptional activity. A further distinction between the compounds that likely contributes to in vivo activity is their metabolite profile: the taurine and glycine conjugates of Px-104, which are also potent FXR agonists, were found in abundance, but analogous conjugation of cilofexor was not observed (data not shown).

Fluorescent Derivatives of Px-104 and Cilofexor Reveal a Distinct Subcellular Localization Pattern with Different Affinities to ER Membranes

We previously observed (Schwabl et al., 2017) that amphiphilicity, i.e., the asymmetry of polarity distribution within a molecule with a hydrophobic core and a hydrophilic, charged, terminal part, affected liver compound levels and in vivo potencies of FXR agonists. Thus, we speculated that differences in transcriptional efficiency could be due to concentration differences in subcellular location.

Huh7 cells were used to assess the subcellular distribution of fluorescently labeled (nitrobenzoxadiazolyl) derivatives of Px-104 (“Px-104-F”) and cilofexor (“cilofexor-F”; Supp. Fig. 4). Both compounds localized to ER membranes (Fig. 5A) and showed little fluorescence in the cytoplasm. The nuclei, where FXR is expected to reside, showed only a sparse, punctuate fluorescence in the presence of either compound, and overall, there was no apparent difference in the staining pattern of the two compounds (Fig. 5B). A kinetic analysis of the fluorescent FXR-agonist Px-104-F (Fig. 5C) showed that the compound appeared first in the plasma membrane and within 10–20 minutes in the ER membranes, reflecting a direct primary partitioning into these membranes.

Fig. 5.

Cilofexor and Px-104 have similar subcellular localization but different retention times. (A) Huh7 cells, transiently transfected with either a KDEL ER-retention signal or a Venus RFP-tagged Golgi marker, were incubated at 100 nM with Px-104-F for 4 hours prior to fixation. (B) Sodium taurocholate cotransporting polypeptide -RFP-expressing Huh7 cells were incubated with 1 µM cilofexor-F (marked as GS-9674-F), Px-104-F (marked as PX-50026), or nitrobenzoxadiazolyl-CDCA for 4 hours. The left column is the RFP channel, the middle column is the fluorescent compound channel, and the right column is the overlay. (C) Huh7 cells were incubated with 100 nM Px-104-F and single-shot time lapse analysis was performed. Intensity profiles of Px-104-F in green and the ER marker protein in magenta are plotted over time. The µm line scan represents the longitudinal cut axis through a representative Huh7 cell; the plasma membrane at x = 28 µm and ER at x = 15 µm. (D) Huh7 cells were incubated with 100 nM Px-104-F (white circles) or cilofexor-F (black circles) for 30 minutes, washed by medium exchange, and the fluorescence was measured at t = 0 (green) or 15 (black) h post medium change. (E) Huh7-derived membrane vesicles were incubated with 5 µM Px-104, Px-104-F, cilofexor, or cilofexor-F. The vesicles were washed three times by centrifugation and the ratio of compounds in the vesicles was determined by LC/MS-MS. (F) Cilofexor and Px-104 were administered at 1 mg/kg i.v. to bile duct-cannulated rats, bile was collected over 8 hours, and the amounts of cilofexor, Px-104, glycol-Px-104, and tauro-Px-104 were quantified.

Given the higher polarity of cilofexor, we expected a reduced affinity for ER membranes compared with Px-104. Thus, we performed kinetic washout experiments which showed that cilofexor-F largely disappeared from the ER after 15 minutes, whereas Px-104-F persisted for several hours (Fig. 5D). These results were corroborated in washout studies with giant plasma membrane vesicles (GPMVs) prepared from cells that were pre-labeled with the indicated combinations of cilofexor, Px-102 and their fluorescent derivatives (Fig. 5E). The isolated GPMVs were subsequently washed and the amount of compound retained in the membranes was quantitated by LC/MS-MS. Both Px-104 and Px-104-F were retained in the membranes to a greater extent than cilofexor or cilofexor-F, consistent with their respective logD values (Supplemental Fig. 4). Importantly, there was little difference in membrane partitioning between each respective parent and fluorescent derivative. Deducing that the faster washout in cells may translate to a faster transit time through the liver, bile duct-cannulated rats were administered Px-104 and cilofexor at 1 mg/kg i.v. and bile was collected over 8 hours; for both compounds, approximately 8% was recovered in the bile. However, cilofexor bile levels peaked at 15–60 minutes post dose and could not be detected in bile past 2 hours post-dose, while Px-104 and its active glycol- and tauro-conjugates, of which only about 0.5% was parent compound, peaked at 30–120 minutes (Fig. 5F) and could still be detected 4–8 hour post-dose.

Cilofexor has Antifibrotic and Antisteatotic Efficacy in a Mouse Model of NASH

We conducted further pharmacology studies using an HFD-CCl4 mouse model to determine the therapeutic potential of cilofexor on liver steatosis and fibrosis. C57BL/6J HFD-CCl4 is a well-established mouse model to anticipate drug effects on advanced human NASH with regard to degree of steatosis, inflammation, ballooning, and fibrosis (Tsuchida et al., 2018). Mice were maintained on an HFD for 16 weeks and administered CCl4 intraperitoneally from Weeks 12 to 16. Animals were also treated with vehicle or cilofexor at 10 or 30 mg/kg q.d. orally from Weeks 12 to 16 (Fig. 6A). At study end, the ileum and liver were analyzed for effects on gene regulation, liver histopathology, steatosis, and fibrosis.

Fig. 6.

Cilofexor significantly reduces fibrosis in a 2-hit mouse model. (A) study design. Mice were orally administered cilofexor at 0 (vehicle: 0.5% carboxymethylcelluose and 1% ethanol in 50 mM Tris buffer, pH 8.0), 10 or 30 mg/kg q.d. from weeks 12 to 16. (B) NAS, ballooning, steatosis, and inflammation scores from hematoxylin and eosin-stained liver sections, (C) representative microphotographs of PSR staining (pink area) of liver sections; collagen deposition was assessed in PSR-stained sections using quantitative morphometry and in liver samples by hydroxyproline measurement, (D and E) gene expression in liver and ileum, respectively. Data are presented as mean ± S.D.; *p < 0.05, **p < 0.01, ***p < 0.001 versus vehicle and †††p < 0.001 versus CCl4 control by unpaired t test, n = 7–12 per group.

Cilofexor dose-dependently reduced the NAS and the individual components of steatosis, ballooning, and inflammation (Fig. 6B). Cilofexor also reduced the fibrosis levels at both doses as shown by the reduction in PSR-positive area and in hydroxyproline content (Fig. 6C). Treatment with cilofexor at both dose levels significantly induced SHP gene expression 3- to 4-fold in the liver (Fig. 6D) and 200- to 400-fold in the ileum (Fig. 6E), demonstrating a strong intestinal bias. Cilofexor also significantly induced FGF15 gene expression in the ileum (Fig. 6E). Cilofexor significantly decreased serum BA at 30 mg/kg, which was reflected in the liver by decreased CYP7A1 and increased bile salt export pump expression at both dose levels. Furthermore, although liver CYP8B1 expression was decreased in mice on HFD and CCl4, cilofexor further reduced its expression (Supplemental Fig. 7).

Cilofexor and Px-104 Induced FGF19 in HFD-Fed Monkeys

Mice produce FGF15, which is not a direct ortholog of human FGF19 (Wright et al., 2004), and their lipid particle and BA compositions differ significantly from those of humans. Therefore, a monkey model was used to evaluate the effects of cilofexor on FGF19 production. As cilofexor was expected to activate intestinal FXR to produce FGF19, we first tested whether this activation was via systemic circulation. Monkeys were administered cilofexor at 5 mg/kg orally or 1 mg/kg i.v. Oral administration of cilofexor generated a pulse of systemic FGF19 with Tmax at 4 hours. In contrast, intravenous administration produced no such p