High-resolution ESI-MS spectra were performed with a LTQ-orbitrap spectrometer (Thermo Fisher Scientific, Waltham, MA, United States). NMR spectra were obtained on Bruker 400 spectrometer (¹H at 400, ¹³C at 100 MHz) equipped with a Sun hardware and recorded in CDCl₃ (δ _H = 7.26 and δ _C = 77.0 ppm) and CD₃OD (δ _H = 3.30 and δ _C = 49.0 ppm). J are in hertz and chemical shifts (δ) are reported in ppm and referred to as CHCl₃ and CHD₂OD as internal standards. HPLC was performed using an Agilent 1260 Infinity II equipped with a Quaternary Pump VL model G7111A, a manual Rheodyne injector, and a Diode Array Detector WR model G7115A (Agilent Technologies, Santa Clara, California, United States). Reaction progress was monitored via thin-layer chromatography (TLC) on Alugram^® silica gel G/UV254 plates. Silica gel MN Kieselgel 60 (70–230 mesh) from Macherey-Nagel Company was used for column chromatography. All chemicals were obtained from Zentek S. r.l. (Milano, Italy) or Sigma Aldrich (St. Louis, Missouri, United States), and reagents were used as supplied from commercial sources with the following exceptions. Tetrahydrofuran was distilled from calcium hydride immediately prior to use. All reactions were carried out under argon atmosphere using flame-dried glassware. The purity of all the intermediates, checked by ¹H NMR, was greater than 95%. The purity of tested compounds was determined to be always greater than 95% by analytical HPLC analysis as reported for each compound.

Williamson reaction. The 2-(cloromethyl)quinoline was added to a stirred mixture of the phenol (1.0 eq), K₂CO₃ (2.5 eq), and dry DMF. The reaction mixture was stirred at 100°C for 12 h. The mixture was diluted with H₂O and extracted with EtOAc (3 × 30 mL). The combined organics were washed with brine, dried (Na₂SO₄), concentrated, and purified by flash chromatography.

Basic hydrolysis. An aliquot of esters was dissolved in MeOH/H₂O (1:1 v/v) and treated with NaOH (5 mol eq.) at 0°C. The resulting mixture was stirred under reflux for 8 h. The mixture was treated with 6 N HCl, until neutrality, and then was extracted thrice with EtOAc. The combined organic extracts were dried over Na₂SO₄ and the solution was concentrated in vacuum. The residue was purified on the silica column to give corresponding carboxylic acids.

DIBAL-H reduction. A solution of DIBAL-H (2.0 eq, 1.0 M in THF) was added dropwise to a solution of esters in dry THF (25 mL) at 0 C. The resulting mixture was stirred at room temperature. When the TLC shows the end of the substrate, the reaction was quenched by slow addition of a solution of saturated sodium potassium tartrate and, after dilution with CH₂Cl₂ stirred for 2 h. The mixture was partitioned thrice with CH₂Cl₂, and the combined organic extracts dried over Na₂SO₄.

GPBAR1. To investigate the GPBAR1 activation, HEK-293 T cells were transiently transfected with Fugene HD reagent (Promega, Madison WI) using the following vectors: pCMVSPORT6-human GPBAR1, pGL4.29 (Promega, Madison WI), a reporter vector containing a cAMP response element (CRE) cloned upstream to the luciferase reporter gene luc2P and pGL4.70 (Promega, Madison WI), a vector encoding the human Renilla gene. At 24 h post transfection, the cells were stimulated with specific receptor agonist TLCA (10 μM) or compounds 1–12 at 10 μM. For the dose–response curve the cells transfected for GPBAR1 were stimulated with increasing concentrations (0.1–100 μM) of compound 2. At 18 h post stimulations, cellular lysates were assayed for luciferase and Renilla activities using the Dual-Luciferase Reporter assay system (E1980, Promega Madison WI). Luminescence was measured using Glomax 20/20 luminometer (Promega, Madison WI). Luciferase activities were normalized with Renilla activities.

Human CysLT₁ (LTD4) (h) (antagonist effect) Cellular Functional Assay. These assays were performed at Eurofins Cerep-Panlabs (France). The cells were suspended in DMEM buffer (Invitrogen) and then distributed in microplates at a density of 3.104 cells/well. The fluorescent probe (Fluo4 Direct, Invitrogen) mixed with probenecid in Hank’s balanced salt solution (HBSS) buffer (Invitrogen) complemented with 20 mM Hepes (Invitrogen) (pH 7.4) was then added into each well and equilibrated with the cells for 60 min at 37°C then 15 min at 22°C. Thereafter, the assay plates are positioned in a microplate reader (CellLux, PerkinElmer), which was used for the addition of the test compound or HBSS buffer, then 5 min later, 0.1 nM LTD4 or HBSS buffer (basal control), and measured the changes in fluorescence intensity, which varies proportionally to the free cytosolic Ca²⁺-ion concentration. The results are expressed as a percent inhibition of the control response to 0.1 nM LTD4. The standard reference antagonist is MK 571. (Sarau et al., 1999; Fiorillo et al., 2021).

CysLT₁R. The crystal structure of the Homo sapiens Cysteinyl leukotriene receptor 1 (PDB ID 6rz4) (Luginina et al., 2019) was downloaded from the Protein Data Bank website. The soluble cytochrome b562 fragment and the co-crystallized ligand were removed and the residue Gln274 was reconstructed, whereas crystallographic water molecules within 5Å of the co-crystallized ligand were included. Residues protonation states were assigned in accordance with the most populated ones predicted by the H++ webserver (Anandakrishnan et al., 2012) at pH 7.4.

GPBAR1. The GPBAR1 homology model reported in D’Amore et al. (D’Amore et al., 2014) was employed for docking calculations. The receptor was prepared as in Biagioli et al. (Biagioli et al., 2020b).

Both receptors were treated with the Protein Preparation Wizard (Sastry et al., 2013) tool implemented in Maestro ver. 11.8.

Ligands. The 3D structure of 1–12 was built using the Graphical User Interface (GUI) of Maestro ver. 11.8 (Sepe et al., 2014). The protonation state of 1–12 at pH 7.4 in water has been calculated using the Epik module (Shelley et al., 2007). Finally, 1–12 were then minimized using the OPLS 2005 force field through 2500 iteration steps of the Polak-Ribiere Conjugate Gradient (PRCG) (Grippo and Lucidi 1997) algorithm.

Docking calculations of 1–12 were performed with the Glide software package, using the Standard Precision (SP) algorithm of the GlideScore function (Halgren et al., 2004) and the OPLS 2005 force field (Banks et al., 2005). A grid box of 25 × 16 × 17 Å for GPBAR1 receptor and one of 16 × 20 × 18 Å for CysLT₁R centered on the ligand binding cavity were created. A total amount of 100 poses was generated and the conformational sampling of the ligand was enhanced by two times, as reported by the default setting of Glide. Docked conformations of 1–12 were then clustered based on their atomic RMSD. Globally, seven clusters were obtained and, among them, only the conformation included in the most populated cluster with both the Glide Emodel and GlideScore lowest-energy value was considered (Figure 3). All figures were rendered by UCSF Chimera (Pettersen et al., 2004).

Solubility. Ten microliters of a 10 mM solution of tested compounds were diluted either in 490 µL of PBS pH 7.4 or MeOH and agitated for 24 h at 250 rpm. The obtained solutions were first centrifuged for 5 min at 4,000 rpm and then further diluted adding 10 μL of each sample to 490 µL of MeOH. Samples were analyzed by LC-MS/MS and the ratio of mass signal area obtained in PBS and in organic solvent was then measured and used to determine solubility.

LogD. Measurements 40 µL of selected compounds were dissolved in 1960 µL of PBS pH 7.4/Octanol. Samples were shacked for 2 h at rt, 10 µL of each phase were withdrawn, diluted in 490 µL of MeOH, and analyzed by LC-MSMS. The ratio of mass signal area obtained in octanol and PBS at pH 7.4 and in organic solvent was employed to calculate LogD.

Metabolic Stability. All compounds were tested at the final concentration of 1 µM in a 50 mM potassium phosphate buffer (pH 7.4) containing 1% DMSO as vehicle. For microsomes assay, the incubation mixtures contained 0.15 mg of Human liver microsomes (Sigma-Aldrich, St. Louis, MO, United States), 5 mM MgCl₂, 1 mM NADPH, 5 mM glucose 6-phosphate, 0.4 U mL⁻¹ glucose 6-phosphate dehydrogenase. For S9 fraction analysis, the buffer contained 0.15 mg of S9 proteins (Sigma-Aldrich, St. Louis, MO, United States), 0.3 mM NADPH, 5.6 mM glucose-6-phosphate, 0.6 units/mL glucose-6-phosphate dehydrogenase, 5.8 mM UDP-glucuronic acid, 0.05 mM acetyl-CoA, 0.5 mM dithiothreitol, 0.5 mM 3′-phosphoadenosine 5′-phosphosulfate, 1 mM glutathione, 0.2 mM acetyl carnitine, 4 units/mL carnitine acetyl transferase, 0.5 mM glycine, and 0.5 mM taurine.

Samples were kept at 37°C and aliquots were removed at 0, 5, 15, 30, 45, 60, 90, 120, 150 min after protein additions. The reaction was stopped by adding 200 µL of ice-cold acetonitrile to withdrawn aliquots. After 2 h, samples were centrifuged for 10 min at 10,000 rpm, and supernatants were subjected to LC-MS/MS analysis.

The slope of the linear regression of the curve obtained reporting the natural logarithm of compound area versus incubation time (−k) was used in the conversion to in vitro t1/2 values by t1/2 = −ln (2)/k.

In vitro intrinsic clearance (Clint expressed as µL/min/mg) was calculated according to the following formula: Clint = Vx 0.693/t1/2 were V = volume of reaction (µL)/protein in the incubation (mg). Testosterone was used as a positive control for microsome and phase I enzymes, and 7-hydroxycoumarin was used as positive control for phase II enzymes.

The amount of compound 2 in plasma samples of animals treated with compound 2 was evaluated measuring the peak area of the compound at 1, 6, and 24 h after administration. 50 μL of plasma samples were mixed with 200 μL of acetonitrile, vortexed and incubated for 1 h on ice. Samples were then centrifuged for 10 min at 10,000 rpm and supernatants were evaporated to dryness. For LC/MS-MS analysis a QTRAP 6500 System (AB Sciex) equipped with Shimadzu LC-20A LC and AutoSampler was employed. Samples were dissolved in 100 μL of H₂O, 30% MeOH, 0.2% FA, and the chromatographic separation was performed using a Luna Omega 3 µm Polar C18 100 × 2.1 mm (Phenomenex) and the best chromatographic results were achieved using the following gradient: 20% B from 0 to 1 min, 20–80% B from 1 to 5.0 min, 80–95% B from 5.0 to 5.1 min, held at 95% B for 3 min and then to 20% B (Buffer A: H₂O 0.2% FA; Buffer B: MeOH 0.1% FA). The transition 356 →143 was employed for quantification. The following parameters were set: positive mode, DP 120 eV, EP 12 eV, CE 40 eV, CXP 23 V, CUR 30 psi, CAD Medium, IS 5500 V, TEM 350°C, GS1 and GS2 50 psi.

The GSE135251 series includes gene expression profiles (RNA-seq analysis, Illumina NextSeq 500 system) of 216 snap frozen liver biopsies, comprising 206 NAFLD cases with different fibrosis stages and 10 controls (Govaere et al., 2020; Pfister et al., 2021).

C57BL/6J male mice were fed a high fat diet (HFD) containing 59 KJ% fat plus 1% cholesterol, w/o sugar (ssniff ^® EF R/M acc. D12330 mod. 22,7 ME/kg) and fructose (HFD-F) in drinking water (42 g/L), or normal chow diet for 61 days. Food intake was estimated as the difference of weight between the offered and the remnant amount of food at 3-days intervals. The food was provided as pressed pellets and the residual spillage was not considered. After 7 days, HFD-F mice were randomized to receive HFD-F alone or in combination with compound 2 (30 mg/kg/day) by gavage until the end of the experiment. Doses of each agent were chosen according to previously published data. Mice were housed under controlled temperature (22°C) and photoperiods (12:12-h light/dark cycle), allowed unrestricted access to standard mouse chow and tap water. The experimental protocol was approved by the Animal Care and Use Committee of the University of Perugia and by the Italian Minister of Health and Istituto Superiore di Sanità (Italy) and was in agreement with the European guidelines for use of experimental animals (permission n. 583/2017-PR). The general health of the animals was monitored daily by the Veterinarian in the animal facility. At the day of sacrifice, fed mice were deeply anesthetized with a mixture of tiletamine hypochloride and zolazepam hypochloride/xylazine at a dose of 50/5 mg/kg and sacrificed before 12 a.m.

Body weight and body length (nose-to-anus or nose–anus length) were measured in anesthetized mice at the time of sacrifice and were used to calculate the body mass index (BMI) (=body weight (g)/length² (cm²)) and the Lee index (=cube root of body weight (g)/nose-to-anus length (cm)).

AST, ALT, total- and HDL-cholesterol, and triglycerides plasmatic levels were quantified using an automated clinical chemistry analyzer (Cobas, Roche).

After 8 weeks of HFD-F mice were fasted overnight and orally administered glucose (1.5 g/kg body weight) for OGTT. The blood glucose concentrations were measured at 0, 15, 30, 60, 90, and 120 min after feeding or injection using a portable glucose meter (Accu-Check Go, Roche).

Temperature of brown adipose tissue (BAT) was recorded through the study using a noninvasive technology. Briefly, mice were maintained at 25°C and the thermal images were taken by a FLIR E6 thermal imaging camera (FLIR System, Wilsonville, Oregon) and the surface temperature quantified by the FLIR Tools.

For histological examination, portions of the right and left liver lobes were fixed in 10% formalin, embedded in paraffin, sectioned and stained with Sirius red and/or Hematoxylin/Eosin (H&E), for morphometric analysis (Fiorucci et al., 2004). NASH severity (steatosis, hepatocytes ballooning, lobular inflammation, and portal inflammation) was scored in H&E-stained cross sections using an adapted grading system of human NASH as described previously (Carino et al., 2017b).

RNA was extracted from eWAT and colon with Direct-zol™ RNA MiniPrep w/Zymo-Spin™ IIC Columns (Zymo Research, Irvine, CA) and it was subjected to reverse transcription with FastGene Scriptase Basic Kit (Nippon Genetics Europe) in a 20 μL reaction volume.

For real time PCR, 50 ng cDNA were amplified in a 20 μL solution containing 200 nM of each primer and 10 μL of SYBR Select Master Mix (Thermo Fisher Scientific, Waltham, MA, United States). All reactions were performed in triplicate, and the thermal cycling conditions were as follows: 10 min at 95°C, followed by 40 cycles of 95°C for 10 s and 60°C for 45 s with QuantStudio™ 3 Real-Time PCR System (Applied Biosystems). The relative mRNA expression was calculated and expressed as 2−(ΔΔCt). Expression of the respective gene was normalized against Gapdh mRNA as an internal control. The following primers for Real-Time PCR were used: mouse-Gapdh: ctg​agt​atg​tcg​tgg​agt​cta​c and gtt​ggt​ggt​gca​gga​tgc​att​g; mouse-Srebf1: gat​caa​aga​gga​gcc​agt​gc and tag​atg​gtg​gct​gct​gag​tg; mouse-Fasn: tca​aga​tga​agg​tgg​cag​agg​tgc​t and ttg​agc​agt​gcc​ggg​att​cgg; mouse-Pgc1α: ctt​agc​act​cag​aac​cat​gca​g and aat​gct​ctt​cgc​ttt​att​gct​c; mouse-Fxr: act​gga​cca​cga​aga​tca​gat​t and gag​cgt​act​cct​cct​gag​tca​t; mouse-adiponectin: tga​cag​atc​agc​tcg​agt​gg and cag​tgc​cgt​cag​ttc​ttg​tg; mouse-Tnf-α: cca​cca​cgc​tct​tct​gtc​ta and agg​gtc​tgg​gcc​ata​gaa​ct; mouse-ucp2: ttg​ccc​gta​atg​cca​ttg​tc and gca​agg​gag​gtc​atc​tgt​ca; mouse-cd11b gtc​aga​gtc​tgc​ctc​cgt​gt and cag​ggt​cta​aag​cca​ggt​ca; mouse-pparγ: gcc​agt​ttc​gat​ccg​tag​aa and aat​cct​tgg​ccc​tct​gag​at; mouse-Il-1β: gct​gaa​agc​tct​cca​cct​ca and agg​cca​cag​gta​ttt​tgt​cg; mouse-Gcg: cca​aga​ttt​tgt​gca​gtg​gtt and cct​tca​gca​tgc​ctc​tca​aa.

High-quality RNA was extracted from mice livers, using the Direct-zol™ RNA MiniPrep w/Zymo-Spin™ IIC Columns (Zymo Research, Irvine, CA) according to the manufacturer’s instructions. RNA quality and quantity were assessed with the Qubit^® RNA HS Assay Kit and a Qubit 3.0 fluorometer (Invitrogen, Carlsbad, CA) followed by agarose gel electrophoresis. Libraries were generated using the Ion AmpliSeq™ Transcriptome Mouse Gene Expression Core Panel and Chef-Ready Kit (Comprehensive evaluation of AmpliSeq transcriptome, a whole transcriptome RNA sequencing methodology) (Thermo Fisher Scientific, Waltham, MA). Briefly, 10 ng of RNA was reverse transcribed with SuperScript™ Vilo™ cDNA Synthesis Kit (Thermo Fisher Scientific, Waltham, MA) before library preparation on the Ion Chef™ instrument (Thermo Fisher Scientific, Waltham, MA). The resulting cDNA was amplified to prepare barcoded libraries using the Ion Code™ PCR Plate, and the Ion AmpliSeq™ Transcriptome Mouse Gene Expression Core Panel (Thermo Fisher Scientific, Waltham, MA), Chef-Ready Kit, according to the manufacturer’s instructions. Barcoded libraries were combined to a final concentration of 100 p.m., and used to prepare Template-Positive Ion Sphere™ (Thermo Fisher Scientific, Waltham, MA) Particles to load on Ion 540™ Chips, using the Ion 540™ Kit-Chef (Thermo Fisher Scientific, Waltham, MA). Sequencing was performed on an Ion S5™ Sequencer with Torrent Suite™ Software v6 (Thermo Fisher Scientific, Waltham, MA). The analyses were performed with a range of fold <−2 and >+2 and a p value <0.05, using Transcriptome Analysis Console Software (version 4.0.1), certified for AmpliSeq analysis (Thermo-Fisher). The transcriptomic data have been deposited as dataset on Mendeley data repository (reference number linking to the repository: Fiorucci, Stefano; Biagioli, Michele; Marchianò, Silvia; Di Giorgio, Cristina (2021), “Discovery of potent dual GPBAR1/CysLT₁R modulator for the treatment of metabolic fatty liver disease,” Mendeley Data, V1, doi: 10.17632/6dnrk9fc72.1).

The ANOVA followed by nonparametric Mann-Whitney U test or a one-tailed unpaired Student t test were used for statistical comparisons (*p < 0.05) using the Prism 6.0 software (GraphPad).

Figure 2 highlights the synthetic routes in obtaining derivatives 1–12, using few reaction steps and simple and cheap intermediates (i.e., chloromethylquinoline and resorcinol), with a view to future large-scale preparation, e.g., in a kilo lab, for future industrial applications.

Williamson reaction on the 2-chloromethylquinoline with methyl 4′-hydroxy- [1,1′-biphenyl]-3-carboxylate (13) or methyl 4′-hydroxy- [1,1′-biphenyl]-4-carboxylate (14) gave the methyl esters 1 (quantitative yield) and 4 (87% yield), respectively. Alkaline hydrolysis and DIBAL-H reduction furnished carboxylic acids 2 and 5 in quantitative yields and alcohols 3 and 6 (quantitative yield and 92%, respectively) (Figure 2A). For the preparation of compounds 7–12 (Figure 2B), first resorcinol was functionalized with the bromo alkyl methyl esters 15 and 16 by Williamson reaction to afford methyl esters 17 and 18 in fair yields (64 and 43%, respectively). Second, Williamson reaction between 17 or 18 with the commercially available 2-(chloromethyl)quinoline allowed the formation of the ether bridge in compounds 7 and 10 (75 and 72%, respectively). Basic hydrolysis and LiBH₄ reduction gave the carboxylic acids (compounds 8 and 11, 68 and 97% respectively) or alcohols (compounds 9 and 12, 80% for both reactions).

Derivatives 1–12 were tested for GPBAR1 activity, in a luciferase reporter assay with HEK-293T cells transfected with GPBAR1. Agonistic activities of compounds on GPBAR1 were compared to TLCA, which was set as 100%. All new compounds were screened in vitro for their CysLT₁R antagonistic activity, as previously described (Sarau et al., 1999; Fiorillo et al., 2021). Compounds 2 and 7–12 were demonstrated antagonists of the CysLT₁R, with efficacies in the range 76–109% and as shown in Table 1, the best results in terms of efficacy on both receptors were found in compound 2.

The binding modes of compounds 1–12 to GPBAR1 and CysLT₁R were investigated by means of molecular docking calculations. This computational approach is commonly employed to elucidate the ligand/protein binding interaction and it has been successfully utilized by our group to characterize the binding mode of new compounds to several targets (Anzini et al., 2008; Nuti et al., 2010; Anzini et al., 2011; Troussicot et al., 2015; Limongelli 2020). In particular, we performed docking calculations using the Glide software package (see Methods for details) (Friesner et al., 2004; Halgren et al., 2004).

As regards CysLT₁R, the crystallographic structure with PDB ID 6rz4 (Di Leva et al., 2019a) has been employed, while for GPBAR1, we used the 3D structure that has already been successfully employed in numerous drug design studies (D’Amore et al., 2014; Sepe et al., 2014; De Marino et al., 2017; Di Leva et al., 2019b) (see Methods for details).

Docking calculations of compounds 1–12 in GPBAR1 led to convergent results showing the quinoline group placed in the amphipathic pocket between transmembrane helices (TM) TM3 and TM5, making contacts with key residues known to participate in the GPBAR1 activation like Tyr89^3.29, Asn93^3.33, Phe96^3.36, and Trp237^6.48 (Macchiarulo et al., 2013; D’Amore et al., 2014; Di Leva et al., 2015; De Marino et al., 2017). On the other hand, the binding mode elucidated by docking calculations performed in CysLT₁R showed the quinoline portion of compounds 1–12 positioned in the pocket formed by TM3, TM4, and TM5. In both cases, the ligand binding mode agrees with that we have recently reported for a series of alpha-pentyl-3-[2-quinolinylmethoxy] benzyl alcohol - REV5901–derivatives (Fiorillo et al., 2021). In the following section, we discuss in detail the binding mode of 2 that is the only dual-activity compound of the series and represents an interesting lead compound to achieve potent CysLT₁R/GPBAR1 dual ligands.

Compound 2 was evaluated for physicochemical properties by LC-MS and compared with those of reference compound REV5901 (Fiorillo et al., 2021) (Figure 3D). Compound 2 showed a good aqueous solubility and a proper LOG_D at pH 7.4. To further assess the drug-likeness of selected compounds, the liability to hepatic metabolizing enzymes contained in microsomes and S9 fractions was evaluated in vitro, measuring by high-performance liquid chromatography-MS/MS the disappearance of unmodified compounds. Compound 2 revealed to be highly stable to Phase I enzymes contained in microsomal fraction. Therefore, employing liver S9 fraction, the metabolic stability of compound 2 was investigated also in the presence of enzymes responsible for Phase II reactions. Even in the presence of S9 contained enzymes derivative 2 was poorly modified and showed a t_1/2 of 385 min (CL_int = 6). All these data highlight the pharmacological potential of compound 2, which associates the potency and the efficacy in dual modulation of GPBAR1/CysLT₁R (Table 1) with excellent metabolic stability (Figure 3D).

Given the promising results obtained during in vitro pharmacokinetics studies, the profile of disappearance of compound 2 was also evaluated in vivo. Mice were treated with compound 2 and plasma samples withdrawn 1, 6, and 24 h after administration were analyzed by LC-MSMS to evaluate the concentration of compound 2. As shown in Figure 3E considerable percentage of compound 2 is still present 6 h after administration whereas approximately 20% is remaining after 24 h.

Because compound 2 is a dual CysLT₁R antagonist/GPBAR1 agonist and NAFLD/NASH is an orphan indication (Fiorucci et al., 2018), we have decided to investigate whether this novel chemical entity is effective in protecting against the development of steatosis and fibrosis in a rodent model of NASH. To validate the relevance of the CysLT₁R/GPBAR1 in this model, we have first investigated whether the cystenyl-leukotriene pathway (Figure 4A) and that of GPBAR1 are modulated in the liver of NAFLD/NASH patients. To explore this concept, we accessed a GSE135251 repository (Govaere et al., 2020; Pfister et al., 2021) and assessed the expression of key genes in the two pathways in a large cohort of patients categorized as early and moderate NAFLD. As shown in Figures 4B–E, while the expression of CysLT₁R mRNA was unchanged in the liver of these patients, the expression of GPBAR1 was robustly increased in both subsets of patients (Figures 4B,E). Furthermore, analysis of key genes in leukotrienes biosynthetic pathways (Martínez-Clemente et al., 2011) demonstrated that either the 5-lipoxygenase (ALOX5) or FLAP (also known as ALOX5AP) were robustly upregulated, and upregulation of FLAP followed the same trend of disease progression (Figures 4C,D).

Since these human data demonstrated that either the leukotrienes pathway or GPBAR1 were upregulated in patients with NAFLD, we have decided to explore whether their modulation is beneficial in a model of NAFLD/NASH. We and others have previously reported that feeding mice a diet enriched in lipid and cholesterol and fructose in the drinking water (HFD-F) to simulate the high caloric intake caused by Western diet, leads to the development of liver injury showing the prototypical changes, steatosis, steatohepatitis, and fibrosis, detected at histopathology analysis in NASH patients (Carino et al., 2017a; Carino et al., 2019a; Carino et al., 2019b; Marchianò et al., 2022). In this experimental setting, HFD-F feed mice were treated orally with compound 2, 30 mg/kg for 54 days. As shown in Figure 5, feeding mice with HFD-F promoted the development of an obesogenic phenotype, with a 32% of weight gain in 61 days, which was partially reversed by compound 2. Importantly, treating mice with compound 2 not only reduced the body weight gain and body mass index (BMI) (Figures 5A–C), but also increased insulin sensitivity as demonstrated by the result obtained at the oral glucose tolerance test (OGTT) shown in Figures 5D,E), reduced AST and ALT plasma levels (Figures 6A,B), and partially reversed the proatherogenic lipid profile caused by feeding mice with an HFD-F (Figures 6C–F).

The analysis of the macroscopic characteristics of the liver (Figures 6G,H) at the end of the experiment showed an increase in liver weight in the animals treated with HFD-F compared to the control group. The administration of compound 2 reversed the effect exerted by the diet. The histopathology analysis of livers collected at the end of the study shown in Figures 6I,J demonstrated that while feeding mice an HFD-F promotes liver steatosis and hepatocytes ballooning (death), compound 2 was highly effective in reversing these features (Figures 6I,J). Additionally, Sirius red stained of liver sections revealed that feeding an HFD-F caused the development of a mild liver fibrosis and increased liver collagen deposition (Figure 6K), we found that these histopathology findings were robustly attenuated by feeding HFD-F mice with compound 2, which also reversed the liver fibrosis at molecular level. Thus, while HFD-F mice showed a robust increase in the liver expression of several pro-fibrogenic genes, αSma and Col1α1, this pattern was reversed by treating mice with compound 2 (Figures 6L,M).

The beneficial effects exerted by compound 2 on liver steatosis were manifested also at molecular levels, since treating HDF-F mice with the dual CysLT₁R/GPBAR1 modulator reshaped the liver transcriptome, as shown in Figure 7. Specifically, treating mice with compound 2 resulted in robust remodeling of the expression of genes encoding for proteins involved in regulating cholesterol and triacylglycerol metabolism. The HFD-F diet compared to the untreated group modulated 647 genes of which 216 upregulated and 431 downregulated (Figure 7B and Supplementary Table S1). Among the genes most upregulated by HFD-F were many genes involved in the lipid metabolism such as Cfd (it is the most upregulated gene with a fold change of 105.43), Mogat1, Me1, Cidec, and Cidea (fold change of 5.67, 2.37, 23.38, and 7.63, respectively) and many genes of the Cyp family (Supplementary Table S1). The administration of compound 2 modulated 154 compared to the group subjected only to HFD-F (Figure 7B and Supplementary Table S2). As shown in Figures 7C–N, compound 2 exerted no effect on Srepb1f but reduced significantly Fasn, along with other genes, such as Elov5, Mogat1, fabp2, Cd36, Me1, and Cidea. Among the 154 genes modulated by compound 2 treatment compared to the group of mice subjected to HFD-F alone, the most downregulated gene is the chemokine Ccl3 with a fold change of–45.68 (Figure 7O and Supplementary Table S2). This finding is very interesting because elevated levels of Ccl3 have been documented in the serum and liver of patients with NASH and moreover the deletion of this gene protects mice from diet-induced steatohepatitis, hepatic fibrosis, and insulin resistance (Xu et al., 2021). Furthermore, among the genes most downregulated by compound 2 we also find the chemokines Cxcl2 and Ccl2 with a fold of–12.23 and–9.45, respectively (Figures 7P,Q) and Serpine 1 (Figure 7R). Importantly, compound 2 increased the expression of Cyp7a1, thus increasing bile acid synthesis and cholesterol disposal (Figure 7T), and decreased Fgf21, responsible for fatty acids and lipoproteins hepatic uptake (Figure 7S). Additionally, as shown in Figure 8, we found that compound 2 increased the expression of Pparα, a gene involved in fatty acid oxidation and Pparγ, in comparison to untreated mice, a validated target in NAFLD/NASH management. No differences are shown instead in the modulation of GPBAR1, FXR, and its target genes (Figures 8A–D). The cystenyl leukotriene pathway was also not modulated by HFD-F (Figures 8G–I). On the other hand, compound 2 statistically modulates only the expression of the Alox5 gene that was downregulated (Figure 8H).

Because adipose tissues are a major therapeutic target in NASH, and both CysLT₁R and GPBAR1 are expressed in this tissue, we have then assessed how compound 2 modulates the functionality of brown and white adipose tissue (BAT and WAT) (Scheja and Heeren 2016). The results shown in Figures 9A–C demonstrate that treating mice with compound 2 reduced the volume of BAT as well as the ratio of BAT weight/body weight. Because expansion of BAT is usually considered a protective event in the setting of high caloric intake, our results on BAT volume are likely due to reduction of body weight and BMI exerted by compound 2.

Because the WAT is the major storage tissue for fat, we have then examined whether treatment with compound 2 modulates WAT metabolism in the model. Exposure to the HFD-F significantly increased the ratio of epididymal white fat (eWAT) weight/body weight (Figures 9D,E) while the treatment with compound 2 reduced these changes. Feeding mice with the HFD-F increased the expression of GPBAR1 and CysLT₁R (Figures 9F,G) and promoted inflammation as shown by increased levels of Tnf-α and Cd11b (Figures 9M–N). No statistically differences are shown in modulation of lipogenic genes (Figures 9H,I,L); feeding an HFD-F diet only shows a tendency to downregulate the expression of Pparγ and Pgc1α (Figures 9H-P). Treating mice with compound 2 reshaped the eWAT transcriptome and upregulated the expression of Ucp2, a marker of brown/beige trans-differentiation (Figure 9J).

Among the effects of GPBAR1 activation in the colon is the increase in GLP-1 secretion (Harach et al., 2012). Therefore, to confirm that the administration of compound 2 induced activation of GPBAR1, we measured the expression of the Gcg, gene encoding GLP-1, in the colon of mice. The data showed that compound 2 upregulated Gcg (Supplementary Figure S1B). The data also confirm the anti-inflammatory activity exerted by compound 2 also in the colon where it strongly reduced the expression of the pro-inflammatory cytokine Il-1β, whose expression was strongly increased by the HFD-F (Supplementary Figure S1A).