Gα_q activation of free fatty acid receptor 4 suppresses metabolic dysfunction by disrupting Nr1h3-PPARγ axis

Abstract

Maintenance of glucose and lipid homeostasis is essential for metabolic health, and its dysregulation, driven by complex gene-environment interactions-underlies various metabolic disorders. Free fatty acid receptor 4 (FFAR4) has been proposed to link dietary signals with genetic metabolic predisposition, yet the precise mechanisms underlying the pathophysiological function of FFAR4 remain elusive and lack of the highly selective FFAR4 agonists. Our study shed light on the pivotal role of FFAR4 in metabolic homeostasis within metabolic organs. Hepatic FFAR4 deficiency in mice exacerbates lipid accumulation and promoted severe steatosis, whereas its overexpression ameliorates diet-induced metabolic dysfunction. Mechanistically, suppression of hepatic FFAR4 promotes lipogenesis by enhancing co-activation of the nuclear receptor Nr1h3 and PPARγ. Furthermore, we identify bavachalcone, a non-carboxylated compound isolated from the traditional Chinese medicine Psoralea corylifolia L., as a functionally effective FFAR4 agonist, which could robustly attenuate metabolic dysfunction. Structural analysis using cryo-electron microscopy reveals the binding mode of bavachalcone within the FFAR4-Gα_iq complex and illuminated the underlying mechanisms. In conclusion, our findings highlight an indispensable role of hepatic FFAR4 in counteracting metabolic dysregulation and identify bavachalcone as a selective and translatable FFAR4 agonist worthy of further clinical evaluation.

FFAR4 safeguards hepatic metabolic homeostasis by restraining Nr1h3-PPARγ-driven lipogenesis. Loss of hepatic FFAR4 aggravates steatosis, whereas its restoration improves metabolic dysfunction. Here the authors identify bavachalcone as a selective FFAR4 agonist and reveal its binding mechanism by cryo-EM.

Introduction

Metabolic disorders, including obesity, diabetes, and metabolic dysfunction-associated fatty liver disease (MAFLD), have emerged as a worldwide health crisis due to exponential prevalence growth¹. Despite recent advancements for the treatment, a deeper mechanistic understanding is urgently needed to uncover novel targets and drive therapeutic innovation^2–5. Although lifestyle intervention remains a cornerstone of management, poor long-standing adherence underscores the compelling need for innovative therapies targeting hubs integrating genetic and dietary factors.

G protein-coupled receptors (GPCRs) regulate diverse physiological processes and constitute the largest class of druggable targets^6,7. Among them, free fatty acid receptors (FFARs), belonging to class A GPCRs, are widely distributed in various tissues and serve as nutrient sensors coordinating metabolic homeostasis^8–10. FFAR2 and FFAR3 preferentially bind gut microbiota-derived short-chain fatty acids^11–13, while FFAR1 and FFAR4 respond to dietary medium- and long-chain fatty acids (LCFAs)¹⁴.

FFAR4 has been implicated in a number of physiological processes, ranging from insulin sensitization^15–17 and reducing sugar preference¹⁸ and to confer renal¹⁹ and neurocognitive protections²⁰ and has generated immense interest and hopes for the treatment of metabolic diseases. Despite these promising observations, the precise mechanisms by which FFAR4 exerts its effects on multiple organs are still not fully understood. Further development is hindered by limited insight into tissue-specific signaling differences and pronounced interspecies variation in FFAR4 expression.

Although preclinical studies strongly support FFAR4 agonism as a strategy to improve glucose and lipid metabolism, the development of selective agonists has faced considerable challenges²¹. Notably, the recent phase IIb ICONA trial of icosabutate, a dual FFAR1/FFAR4 agonist, failed to meet its primary endpoints, despite showing significant improvement in liver fibrosis²². This clinical shortfall may stem from icosabutate’s limited selectivity for FFAR4-driven pathways and its inability to engage specific signaling required for full therapeutic efficacy²³.

In this work, we compare metabolic responses to high-fat diet (HFD)-induced stress across three tissue-specific FFAR4 knockout mouse models to delineate potential differences among the major metabolic organs. We discover that hepatic FFAR4 plays a dominant role in maintaining systemic metabolic homeostasis and negatively regulates metabolic disorders by modulating the nuclear receptor subfamily 1 group H member 3 (Nr1h3) and peroxisome proliferator activated receptor gamma (PPARγ) pathway. Moreover, we identify bavachalcone, a non-carboxylated compound derived from the traditional Chinese medicine Psoralea corylifolia L., with potent therapeutic potential for restoring glucose tolerance and attenuating dysfunction. Collectively, these findings provide mechanistic insights into hepatic FFAR4 signaling and highlight a promising targeted therapeutic strategy for metabolic diseases.

Results

FFAR4 expression is suppressed in metabolic dysfunction and genetically linked to MAFLD susceptibility

To investigate whether hepatic FFAR4 exhibits differential expression in metabolic dysfunction, we evaluated its expression in two classic metabolic syndrome models (HFD-induced obesity and db/db mice). Hepatic FFAR4 expression was significantly decreased in both models (Fig. 1a, b). Furthermore, FFAR4 levels inversely correlated with key indicators of hepatic injury, namely hepatic alanine aminotransferase (ALT) and aspartate aminotransferase (AST), as well as parameters related to dyslipidemia, including hepatic triglycerides (TG), total cholesterol (TC), and low-density lipoprotein cholesterol (LDL-C) (Fig. 1c, d). To further explore a potential causal role of FFAR4 in human MAFLD, we conducted two-sample Mendelian randomization (MR) analysis using summary-level data from publicly available genome-wide association studies (GWAS). Expression quantitative trait locus (eQTL) data for FFAR4 were sourced from the IEU OpenGWAS database (ID: ebi-a-GCST90091033; https://gwas.mrcieu.ac.uk/; Fig. 1e). Both MR-Egger and inverse-variance weighted (IVW) methods consistently indicated that genetically predicted reduced FFAR4 expression is associated with an elevated risk of MAFLD (Fig. 1f, g). These results suggest that hepatic FFAR4 may function as a protective regulator in systemic metabolic homeostasis.

Fig. 1. Decreased hepatic FFAR4 is associated with the advance of MAFLD.

a,b Hepatic FFAR4 mRNA and protein expression in C57BL/6 J mice maintained on normal chow diet or high-fat diet (HFD) for 16 weeks, and db/m and db/db mice (n = 4). mRNA levels were quantified by qPCR with β-actin normalization (a); protein expression was measured by immunoblotting, with β-actin serving as the loading control (b) (n = 4). c, d Correlation analyses of hepatic FFAR4 mRNA levels with metabolic parameters (TG, TC, ALT, AST, LDL) in dietary (c) and genetic (d) models of metabolic syndrome (n = 8). e Schematic of the Mendelian randomization (MR) framework used to assess causality between FFAR4 expression and MAFLD risk. Created in BioRender. Kong, Y. (2026) https://BioRender.com/w19a426. f Scatter plots of SNP effect estimated for FFAR4 expression against MAFLD association from MR analyses. Regression slopes reflect causal estimates for each method. g Forest plot showing the effect of individual SNPs on MAFLD risk. Combined estimates (red lines) derived from random-effects IVW analysis. The bolded and highlighted parts represent key SNPs related to obesity. Data are displayed as mean ± SEM. Significance was determined by an unpaired two-tailed Student’s t test (a, b). Pearson correlation coefficients were calculated between continuous variables (c, d). Source data are provided as a Source Data file.

Hepatic FFAR4 deficiency exacerbates diet-induced metabolic disorders

To delineate the tissue-specific contribution of FFAR4 to metabolic regulation, we generated conditional knockout mice targeting adipose (FFAR4^AKO, Supplementary Fig. 1a), liver (FFAR4^LKO, Fig. 2a, b), and muscle (FFAR4^MKO, Supplementary Fig. 1h) tissues-key organs involved in systemic lipid homeostasis^24,25. Among these models, only hepatocyte-specific FFAR4 knockout mice (FFAR4^LKO) exhibited the most pronounced metabolic perturbations on an HFD. Neither FFAR4^AKO nor FFAR4^MKO mice showed significant phenotypic alterations under normal chow diet (NCD) or HFD conditions compared to controls (Supplementary Fig. 1b–g and Supplementary Fig. 1i–n). FFAR4^LKO mice exhibited elevated body weight, liver weight, and adipose tissue deposition (white adipose tissue [WAT], inguinal WAT [iWAT], and brown adipose tissue [BAT]) compared to FFAR4^flox/flox controls (Fig. 2c–h). Hepatic steatosis was markedly exacerbated in FFAR4^LKO-HFD mice, as evidenced by elevated hepatic TG and TC content, increased hepatic ALT and AST levels, enhanced Oil Red O-stained lipid droplet accumulation, and histopathological signs of steatosis (Fig. 2k, l). Notably, hepatocyte-specific FFAR4 ablation did not significantly modulate systemic insulin sensitivity, as determined by glucose tolerance test (GTT) and insulin tolerance test (ITT) (Fig. 2i, j). No significant phenotypic differences were observable between genotypes under NCD conditions (Supplementary Fig. 2a–j). Collectively, these findings establish a critical and liver-specific role for FFAR4 in protecting against diet-induced metabolic dysfunction.

Fig. 2. Hepatocyte-specific FFAR4 deletion exacerbates HFD-induced metabolic dysfunction.

a Strategy for generating hepatocyte-specific FFAR4-knockout mice (Alb-cre; FFAR4Loxp/Loxp, abbreviated FFAR4^LKO) and littermate controls (FFAR4Loxp/Loxp without Alb-cre, abbreviated as FFAR4^flox/flox). b Left: representative genotyping PCR. Wild-type (WT) alleles yielded 234-bp (FFAR4) and 351-bp (Cre) bands; floxed alleles produced 273-bp (FFAR4) and 390-bp (Cre) products. Middle: Hepatic FFAR4 mRNA levels in FFAR4^flox/flox and FFAR4^LKO mice, quantified by qPCR normalized to β-actin (n = 3). Right: Hepatic FFAR4 protein levels in FFAR4^flox/flox and FFAR4^LKO mice, assessed by western blot analysis using β-actin as loading control (n = 3). c Diagram of experimental design. FFAR4^flox/flox and FFAR4^LKO were fed HFD (60 kcal%) for 16 weeks. Created in BioRender. Kong, Y. (2026) https://BioRender.com/w19a426. d–h Metabolic phenotypes: body weight (d), terminal liver weight (e), epididymal white adipose tissue (eWAT) (f), inguinal WAT (iWAT) (g), and brown adipose tissue (BAT) mass (h) (n = 5). i, j Metabolic assessments: glucose tolerance test (GTT) (i) and insulin tolerance test (ITT) (j) with corresponding AUC values (n = 5). k Hepatic lipid profiling: triglyceride (TG), total cholesterol (TC), alanine aminotransferase (ALT), and aspartate aminotransferase (AST) levels (n = 5). l Histopathological analysis: representative image of liver, liver Oil red O and H&E staining (scale bar, 50 μm). Data are displayed as mean ± SEM. Statistical significance was determined by an unpaired two-tailed Student’s t test. Source data are provided as a Source Data file.

Hepatic FFAR4 overexpression confers resistance to diet-induced metabolic dysfunction

To further validate the role of hepatic FFAR4 in metabolic homeostasis, a hepatocyte-specific FFAR4 overexpression mice model was established (FFAR4^LOE) (Fig. 3a, b). FFAR4^LOE mice challenged with an HFD exhibited significant protection against metabolic dysfunction, including reduced body weight, liver weight, and adipose tissue mass (Fig. 3c–h), improved glucose tolerance and insulin sensitivity (Fig. 3i, j), ameliorated dyslipidemia, and attenuated hepatic steatosis with decreased lipid droplet accumulation (Fig. 3k, l). Under normal chow diet conditions, FFAR4^LOE mice showed no significant differences in most metabolic parameters compared to controls (Supplementary Fig. 3a, b, d, e, f, i, j), except for reduced liver weight and modestly improved glucose metabolism (Supplementary Fig. 3c, g, h). These results demonstrate that specific hepatic overexpression of FFAR4 confers resistance to diet-induced metabolic disorders.

Fig. 3. Hepatic FFAR4 overexpression protects against HFD-induced metabolic disorders.

a Strategy for generating hepatocyte-specific FFAR4-overexpressing mice (FFAR4^LOE) via Cre-mediated excision of a STOP cassette from a knock-in allele at the Hipp11 locus. b Left panel displays a representative genotyping PCR. Wild-type (WT) alleles generated 412-bp (FFAR4-cag) and 351-bp (Cre) fragments; homozygous cag/cag alleles yielded 1051-bp (FFAR4-cag) and 390-bp (Cre) amplicons. The middle panel displays hepatic FFAR4 mRNA levels in control and FFAR4^LOE mice, quantified by qPCR normalized to β-actin (n = 3). The right panel displays hepatic FFAR4 protein levels in control and FFAR4^LOE mice, assessed by western blot analysis using β-actin as loading control (n = 3). c Diagram of experimental design. FFAR4^LOE mice and littermate controls were fed an HFD (60 kcal%) for 16 weeks. Created in BioRender. Kong, Y. (2026) https://BioRender.com/w19a426. d–h Metabolic parameters: body weight (d), liver weight (e), eWAT weight (f), iWAT weight (g), and BAT weight (h) (n = 5). i, j Metabolic assessments: GTT (i) and ITT (j) with corresponding AUC values (n = 5). k Hepatic TG, TC, ALT and AST levels (n = 5). l Histopathological analysis: representative image of liver, liver Oil red O and H&E staining (scale bar, 50 μm). Data are displayed as mean ± SEM. Statistical significance was determined by an unpaired two-tailed Student’s t test. Source data are provided as a Source Data file.

Hepatic FFAR4 deficiency activates PPAR signaling pathway and facilitates Nr1h3 expression

Hepatocytes constitute approximately 80% of liver mass and orchestrate over 90% of hepatic lipogenic regulation, coupled with single-cell RNA-seq analysis from the DISCO database, revealing that predominant FFAR4 localization to hepatocytes with minimal detection in Kupffer cells (Supplementary Fig. 4a). These observations underscore the importance of investigating the functional roles and regulatory mechanisms of FFAR4 specifically in hepatocytes. To elucidate the molecular mechanism by which hepatic FFAR4 modulates lipid metabolism, we utilized two established hepatic steatosis cell models (AML12 and BNL CL.2) that recapitulate key features of hepatocyte lipid accumulation^26–28. FFAR4 knockdown significantly exacerbated TG accumulation and lipid droplet formation in both cell lines (Fig. 4a–c). RNA sequencing of FFAR4-deficient cells identified 709 and 240 upregulated genes and 743 and 412 downregulated genes in AML12 and BNL CL.2 cells, respectively (Fig. 4d). Among the two upregulated gene sets, 53 genes overlapped between the two models (Fig. 4e). KEGG pathway revealed significant enrichment in the PPAR signaling pathway (Fig. 4f), with core regulators including Dbi, Cp1ta, Nr1h3 and Plin4 (Fig. 4g). qPCR validation confirmed these RNA-seq findings, with Nr1h3 demonstrating a particularly significant upregulation (Supplementary Fig. 4b, c). In addition, qPCR and immunoblotting analysis demonstrated that Nr1h3 was markedly increased due to FFAR4 deficiency during hepatic steatosis in vitro and in vivo (FFAR4^LKO-HFD mice) (Fig. 4h–j and Supplementary Fig. 4d), whereas hepatic FFAR4 overexpression mice exhibited significant attenuation of Nr1h3 expression (Fig. 4k and Supplementary Fig. 4e). Taken together, these observations suggested that Nr1h3 might participate in the FFAR4-mediated lipid metabolism.

Fig. 4. PPAR pathway and Nr1h3 expression are enhanced in the absence of hepatocyte FFAR4.

a Knockdown efficiency of FFAR4 siRNA in AML12 validated through qPCR and western blot (n = 3). b Cells were transfected with siRNA, treated after 24 h with palmitic acid/oleic acid (PA/OA) for 48 h. Cellular TG content in siRNA-transfected AML12 and BNL CL.2 cells (n = 3). c Oil Red O staining of lipid droplet accumulation in AML12 and BNL CL.2 cells (scale bar: 250 μm). d Volcano plots displaying differentially expressed genes following FFAR4 silencing. Red and blue dots indicate up- and downregulated genes, respectively. e Venn diagram of upregulated genes shared between AML12 and BNL CL.2 cells after FFAR4 knockdown. f KEGG pathway enrichment analysis of upregulated genes (red) in both cell lines. Enrichment P- valuses were assessed using a one-sided Fisher’s exact test with Benjamini-Hochberg correction (adjusted P < 0.05 were considered significant). g Heatmap of upregulated PPAR signaling pathway genes. h, i Nr1h3 protein expression in AML12 (h) and BNL CL.2 (i) cells transfected with FFAR4 siRNA or control siRNA (n = 3). j, k Nr1h3 expression in liver tissues from FFAR4^LKO (j) and FFAR4^LOE (k) mice under HFD. β-actin served as a loading control (n = 4 mice). Data are displayed as mean ± SEM. Statistical significance was determined by one-way ANOVA with Tukey’s multiple comparisons test (a) or unpaired two-tailed Student’s t test (b, h–k). Source data are provided as a Source Data file.

Pharmacological and genetic inhibition of Nr1h3 rescues metabolic dysfunction in FFAR4-deficient mice

To determine whether Nr1h3 is the critical target for accelerated metabolic disruption by FFAR4 deficiency, we administered the selective Nr1h3 inhibitor GSK2033 by oral gavage to HFD-fed FFAR4^LKO and FFAR4^flox/flox mice for 4 weeks (Fig. 5a). Pharmacological inhibition of Nr1h3 markedly attenuated the metabolic perturbations in FFAR4^LKO mice, as shown by reduced body weight, liver weight, and adipose tissue mass (Fig. 5b–f), improved hepatic lipid profiles (TC, TG) and hepatic function (ALT, AST) (Fig. 5i), and restored glucose tolerance and insulin sensitivity (Fig. 5g, h). GSK2033 treatment also ameliorated hepatic steatosis and lipid deposition (Fig. 5j). Consistent with these in vivo findings, Nr1h3 antagonism with GSK2033 reversed lipid accumulation in FFAR4 knockdown hepatocytes (Supplementary Fig. 5a–c). Moreover, siRNA-mediated knockdown of Nr1h3 in FFAR4-deficient hepatocytes effectively rescued the lipogenic phenotype, reducing triglyceride content and lipid droplet formation (Supplementary Fig. 6a–d). These results collectively demonstrate that hepatic FFAR4 regulates metabolic homeostasis by negatively modulating Nr1h3.

Fig. 5. Pharmacological inhibition of Nr1h3 ameliorates metabolic dysfunction in FFAR4-deficient mice.

a Schematic diagram of GSK2033 administration in HFD-fed mice. FFAR4^LKO and FFAR4^flox/flox mice under HFD conditions were treated daily with GSK2033 (10 mg/kg) or vehicle for 4 weeks (n = 5). Created in BioRender. Kong, Y. (2026) https://BioRender.com/w19a426. b–f Metabolic parameters: body weight (b), liver weight (c), eWAT mass (d), iWAT mass (e), and BAT mass (f) (n = 5). g, h Metabolic assessments: GTT (g) and ITT (h) with corresponding AUC values (n = 5). i Hepatic TG, TC, ALT and AST levels (n = 5). j Representative image of liver Oil red O and H&E staining (scale bar, 50 μm). Data are displayed as mean ± SEM. One-way ANOVA with Tukey’s multiple comparisons test. Source data are provided as a Source Data file.

FFAR4 orchestrates lipogenesis via Nr1h3-PPARγ axes in hepatocytes

FFAR4 activation is known to enhance PPARγ activity, with their combined agonism synergistically improving glucose tolerance and insulin sensitivity in obese models^29,30. Previous mechanistic investigations have elucidated that Nr1h3 can form functional heterodimeric complexes with PPARγ isoforms, which coordinately regulate transcriptional programs governing lipid homeostasis^31,32. To determine whether hepatic FFAR4 regulates lipid metabolism dependent on the Nr1h3-PPARγ axis, key lipogenic regulators in FFAR4^LKO mice liver were determined. Expression of PPARγ, sterol regulatory element-binding protein 1 (SREBP1), and fatty acid synthase (FASN) were significantly increased compared to FFAR4^flox/flox controls (Fig. 6a). Conversely, hepatic FFAR4 overexpression suppressed these proteins (Supplementary Fig. 7a). Consistent with these findings, FFAR4 knockdown in hepatocytes upregulated these lipogenic factors (Supplementary Fig. 7b, c). Pharmacological inhibition or siRNA-mediated knockdown of Nr1h3 attenuated this lipogenic program, significantly reducing PPARγ, SREBP1, and FASN levels (Fig. 6b and Supplementary Figs. 5d, e,6e, f). In addition, immunofluorescence and immunoprecipitation assays showed that FFAR4 deficiency enhanced the interaction between Nr1h3 and PPARγ (Fig. 6c–e). To further validate that FFAR4 deficiency promotes lipogenesis by enhancing the Nr1h3-PPARγ interaction, we treated FFAR4-silenced hepatocytes with the PPARγ inhibitor GW9662, the Nr1h3 inhibitor GSK2033, or their combination. Both individual and combined inhibitor treatments significantly attenuated lipid accumulation (Fig. 6f and Supplementary Fig. 7d) and downregulated key lipogenic proteins (Fig. 6g and Supplementary Fig. 7e). However, no additive or synergistic effect was observed upon co-treatment, further supporting that Nr1h3 and PPARγ function within a convergent pathway driving steatosis in response to FFAR4 loss.

Fig. 6. FFAR4 restrains hepatic lipogenesis by suppressing the Nr1h3-PPARγ axis.

a Protein levels of PPARγ, SREBP1, and FASN in FFAR4^LKO and FFAR4^flox/flox mice fed with HFD, using β-actin as a loading control (n = 4). b Hepatic expression of PPARγ, Nr1h3, SREBP1, and FASN in FFAR4^LKO and FFAR4^flox/flox mice, with or without GSK2033 administration, maintained on a HFD (n = 3). β-actin was used as a loading control. c, d Immunofluorescence analysis of PPARγ (red) and Nr1h3 (green) colocalization in FFAR4^LKO and FFAR4^flox/flox mice fed with HFD (scale bar, 5 μm, c) and in AML12 cells transfected with control or FFAR4 siRNA and treated with PA/OA (scale bar, 20 μm, d). Nuclei were stained with Hoechst (blue). Images were acquired by confocal microscopy. e Representative immunoblot of co-immunoprecipitation of PPARγ and Nr1h3 in AML12 cell lysates using specific antibodies (representative of three independent experiments with similar results). f, g AML12 cells were transfected with control or FFAR4 siRNA, followed by treatment with GSK2033 (Nr1h3 inhibitor), GW9662 (PPARγ inhibitor), or both, in the presence of PAOA for 48 h. Cellular TG level (f). Protein levels of PPARγ, Nr1h3, SREBP1, and FASN (g). Data are displayed as mean ± SEM. Significance was assessed by unpaired two-tailed Student’s t test (a) or one-way ANOVA with Tukey’s multiple comparisons test (b, f). Source data are provided as a Source Data file.

Bavachalcone evokes FFAR4-Gα_q signaling to suppress hepatic lipogenesis

Current development of FFAR4 agonists has largely relied on carboxylate-containing analogs of fatty acids, which restricts the chemical diversity and precision of pharmacological modulation³³. There is a pressing need to explore structurally distinct chemotypes that enable more precise targeting and functional selectivity. Natural products, with their vast and evolutionarily refined chemical space, represent a promising source for novel scaffolds^34,35. To identify therapeutic agents targeting hepatic FFAR4 for metabolic diseases, we screened a library of 145 natural compounds using dynamic mass redistribution (DMR) assays followed by validation of TG accumulation inhibition in vitro (Fig. 7a). Similarity analysis of DMR profiles in a heat map identified six hits, namely C54 (bavachalcone), C98 (oleanonic acid), C92 (hederagenin), C132 (emodin), C96 (corosolic acid), and C76 (punicalagin), which induced significant DMR signal and attenuated the probe TUG891-induced DMR signal (Fig. 7b and Supplementary Fig. 8a). Among them, bavachalcone triggered the most similar DMR real-time signal pattern to that of the FFAR4 agonist probe TUG891 and completely desensitized TUG891-induced DMR signal, suggesting that bavachalcone is a potential FFAR4 agonist (Supplementary Fig. 8a). To verify its agonistic activity against FFAR4, we conducted concentration-dependent DMR assays. Bavachalcone elicited a DMR signal in a concentration-dependent manner with a pEC₅₀ ± SEM value of 6.01 ± 0.03 (Fig. 7c, d). And its DMR signal was absolutely inhibited by FFAR4 antagonist AH-7614, leading to a pIC₅₀ ± SEM value of 7.19 ± 0.06 (Fig. 7e), suggesting that bavachalcone is an FFAR4 agonist. Meanwhile, bavachalcone triggered concentration-dependent Ca²⁺ mobilization, and the Ca²⁺ signal was blocked by AH-7614 in fluorometric imaging plate reader (FLIPR) assays, confirming its FFAR4 agonistic activity (Supplementary Fig. 8b). Receptor selectivity assays showed that bavachalcone had no activity against FFAR1, FFAR2, or FFAR3 (Fig. 7f–h). Thus, bavachalcone is a selective FFAR4 agonist with a distinct scaffold compared to TUG891. Furthermore, bavachalcone markedly inhibited intracellular TG accumulation in AML12 and BNL CL.2 hepatocytes (Supplementary Fig. 8g) and reduced lipid droplet deposition (Supplementary Fig. 8h), accompanied by downregulation of the Nr1h3-PPARγ pathway (Fig. 7n, o and Supplementary Fig. 9a, b).

Fig. 7. FFAR4 agonist bavachalcone attenuates lipid accumulation via Gα_q signaling.

a Schematic workflow of the screening platform for identifying FFAR4 agonists. Created in BioRender. Kong, Y. (2026) https://BioRender.com/w19a426. b Label-free cell phenotypic heatmap of 145 compounds based on DMR profile similarity in CHO-K1-FFAR4 cells. This heat map was obtained using similarity analysis of the compound-induced DMR signals and the FFAR4 probe TUG891-induced DMR signals in cells pretreated with compounds. For each DMR profile, the real amplitudes at 5, 7, 12, 20 and 60 min were used and color-mapped (red, positive amplitude; black, zero; green, negative amplitude). The assayed concentrations of compounds and the probe TUG891 were 40 μM and 2 μM, respectively. c Real-time DMR responses of bavachalcone at varied concentrations in CHO-K1-FFAR4 cells (n = 3). d Concentration-dependent DMR responses of bavachalcone and TUG891 (n = 3). e Concentration-dependent DMR responses of bavachalcone and TUG891 in CHO-K1-FFAR4 cells pretreated by FFAR4 antagonist AH-7614 at varied concentrations (n = 3). f–h Concentration-dependent DMR responses of bavachalcone and the agonist probe-TAK875 (FFAR1, f), 4-CMTB (FFAR2, g), and AR420626 (FFAR3, h) in corresponding receptor-expressing cells (n = 3). i–k Concentration-dependent response curves of FFAR4 in response to bavachalcone and reference agonist TUG891, measured via Gɑ_q-Gγ (i), Gɑ_i-Gγ (j), and Gɑ_s-Gγ (k) dissociation assay (n = 3). l, m Cellular TG levels of AML12 cells treated with bavachalcone (10 μM) combined with cholera toxin (Gα_s inhibitor, 100 ng/mL) or YM-254890 (Gα_q inhibitor, 300 nM) (l) or siRNA knockdown (m) for 48 h (n = 3). n, o Protein levels of PPARγ, SREBP1, FASN, and Nr1h3 in PAOA-treated AML12 cells with bavachalcone (10 μM) and YM-254890 (Gα_q inhibitor, 300 nM) (n), or Gα_q siRNA knockdown (o). β-actin was used as the loading control (n = 3). Data are displayed as mean ± SEM. Significance was assessed by one-way ANOVA with Tukey’s multiple comparisons test (l, m). Source data are provided as a Source Data file.

FFAR4 is known to exhibit ligand-directed signaling bias, wherein different agonists preferentially engage distinct intracellular pathways³⁶. To determine the signaling response triggered by the binding of bavachalcone to FFAR4, we performed G protein dissociation assays using TUG891 as a reference agonist. Notably, bavachalcone exhibited preferential activation of Gα_q and Gα_s signaling (pEC₅₀± SEM = 5.42 ± 0.09 and 5.69 ± 0.18, respectively), whereas TUG891 preferentially activated Gα_q and Gα_i (pEC₅₀± SEM = 6.08 ± 0.33 and 5.47 ± 0.20, respectively; Fig. 7i–k). The cell pretreatment with the FFAR4 antagonist AH-7614 was found to block the BRET signals arising from the bavachalcone-induced Gα_q and Gα_s dissociation, suggesting the specificity to FFAR4 activation (Supplementary Fig. 8c, d). Furthermore, β-arrestin recruitment assays showed that bavachalcone concentration-dependently recruited β-arrestin 1 and β-arrestin 2, similar to TUG891 (Supplementary Fig. 8e, f). Mechanistic investigation into the contribution of individual signaling pathways to lipid regulation revealed that pharmacological inhibition of Gα_q (with YM-254890) markedly attenuated the anti-lipogenic effects of bavachalcone, whereas blockade of Gα_s (with cholera toxin) showed minimal impact (Fig. 7l and Supplementary Fig. 8i). Consistent with these pharmacological findings, siRNA-mediated silencing of Gα_q, but not Gα_s or β-arrestins, blunted the compound’s lipid-lowering benefits, establishing Gα_q as the principal mediator (Fig. 7m, Supplementary Figs. 8j,9c, d, f, g). Furthermore, only Gα_q inactivation abrogated bavachalcone’s suppression of the Nr1h3-PPARγ pathway and lipogenic effectors (Fig. 7n, o and Supplementary Fig. 9a, b, e, h). These results highlight that bavachalcone elicits lipid-lowering effects via FFAR4-Gα_q activation, providing a tool to understand the druggable molecular landscape of FFAR4.

Bavachalcone recognition and activation of FFAR4

To elucidate the structural basis of bavachalcone-mediated FFAR4 activation, we identified the cryo-EM structure of the bavachalcone-bound FFAR4-Gα_iq complex at 2.99 Å resolution (Fig. 8a, Supplementary Figs. 10b–f and Supplementary Table 1). In the presence of bavachalcone, FFAR4 and the heterotrimeric G proteins formed a stable complex (bavachalcone-FFAR4-Gα_iq-Gβ₁-Gγ₂) (Supplementary Fig. 10a). Gα_iq is an engineered chimera in which residues 329–354 of the Gα_i1 α5-helix (αH5) were replaced by the corresponding Gα_q segment (residues from 334 to 359); this construct has previously enabled structure determination of the Gα_q-coupled FFAR4, revealing the engagement between Gα_q and FFAR4. A focused refinement improved local map quality for the FFAR4-Gα_iq-bavachalcone region, despite a slight resolution drop to 3.25 Å (Supplementary Fig. 10c, g–j).

Fig. 8. Structural basis of bavachalcone recognition and FFAR4 activation.

a Cryo-EM density map (left panel) and structural mode (right panel) of the bavachalcone-bound to FFAR4 in complex with Gα_iq, Gβ₁, and Gγ₂.Bavachalcone is shown in blue, FFAR4 in purple, Gα_iq in salmon, Gβ₁ in green, and Gγ₂ in yellow. The middle panel shows the density for bavachalcone. b Surface representation of FFAR4, visualized as a cross-section and colored by hydrophobicity, depicting the ligand-binding pocket. c Molecular interactions between bavachalcone and FFAR4 residues. Hydrogen bonds are indicated by black dashed lines. d, e Superimposition of the active Bav-FFAR4 complex (purple) onto the Alphafold-predicted inactive structure (AF-FFAR4, gray). Views from the extracellular (d) and intracellular sides (e). f–i Conformational changes in key activation motifs between active (purple) and inactive (gray) FFAR4. j Interactions between FFAR4 and Gα_iq. Hydrogen bonds are shown as black dashed lines.

Mimicking the structural topology of endogenous lipids^23,37,38,bavachalcone adopts a L-shaped pose within the orthosteric pocket (Fig. 8b). Its 4-hydroxyphenyl, enone bridge and 3-prenyl-resorcinol moieties descend sequentially toward the pocket floor, engaging almost exclusively hydrophobic contacts (Fig. 8b, c). Specifically, the 4-hydroxyphenyl packs against F27^N-term, L196^ECL2, W207^5.38, D208^5.39, I284^6.55, and I287^6.58; the enone bridge complements the hydrophobic surface formed by F27^N-term, F115^3.29, and L173^4.61; the 3-prenyl-resorcinol occupies the bottom cavity lined by M118^3.32, V307^7.39, F211^5.42, I280^6.51, W277^6.48, and I126^3.40. An additional hydroxyl-carboxylate hydrogen-bond network (phenolic-OH to E204^5.35 and D208^5.39) locks the ligand orientation (Fig. 8c).

Structural alignment between bavachalcone-bound FFAR4 and an inactive FFAR4 model reveals how bavachalcone binding initiates a cascade of conformational changes associated with receptor activation. The AlphaFold-predicted apo-FFAR4 structure closely resembles the structure of inhibitor-bound CB16 (PDB: 5TGZ)^39–41, with an RMSD of ~ 1.05 Å, confirming its representation of an inactive receptor state (Supplementary Fig. 11a). Compared to the apo model, the extracellular side of bavachalcone-bound FFAR4 shows minimal conformational change, while the cytoplasmic end of TM6 moves outward significantly, and that of TM7 shifts inward (Fig. 8d, e), which are hallmarks of class A GPCR activation⁴². The toggle residue W277^6.48, located beneath the 3-prenylresorcinol group of bavachalcone, swings downward to avoid steric clash with the ligand (Fig. 8f). This movement induces conformational rearrangements in I126^3.40 and F273^6.44 of the PIF motif and R136^3.50 of the DRY motif (Fig. 8g, h). The displacement of the R136^3.50 sidechain disrupts its electrostatic interaction with D259^6.30 present in inactive FFAR4. Together with the rotation of the F273^6.44sidechain, this results in a large outward movement of the cytoplasmic end of TM6 by up to 10 Å (Fig. 8e, g, h). Concurrently, the upward shift of the R136^3.50 sidechain attracts Y321^7.53 in the NPxxY motif to swing toward TM3, leading to an inward movement of the cytoplasmic end of TM7 by approximately 2.5 Å (Fig. 8e, i). These rearrangements create the cytosolic cavity required for G-protein engagement.

The N- and C-terminal segments of αH5 anchor Gα_iq to FFAR4 mainly through hydrophobic contacts (Fig. 8j). At the C-terminus, V354^G.H5.26, L353^G.H5.25, Y351 ^G.H5.23, L348^G.H5.20, and L344^G.H5.16 pack against a surface comprising L262^6.34, L266^6.38, I140^3.54, R136^3.50, and I233^5.64 of FFAR4, while the Y351^G.H5.23-R136^3.50 cation-π interaction provides additional stability. At the N-terminus, A337^G.H5.9, A338 ^G.H5.10, and F334^G.H5.6 form a hydrophobic patch complementary to A246^ICL3, Y247^ICL3, and L241^5.72 of FFAR4, reinforced by a F336^G.H5.9-R240^5.71 cation-π pair. Mid-segment polar contacts further tether αH5: Q345^G.H5.17 and D341^G.H5.13 hydrogen-bond to R238^5.69 and S237^5.68, respectively, and the D341^G.H5.13-R238^5.69 salt bridge completes the interface (Fig. 8j).

TUG891, a known agonist of FFAR4, activates both Gα_i and Gα_q signaling pathways^23,37,43. The overall structure of the bavachalcone-FFAR4-Gα_iq complex is highly similar to that of the TUG891-FFAR4-Gα_iq complex, with an RMSD of ~ 0.9 Å (Supplementary Fig. 11b). However, differences exist in the ligand-binding pose and receptor conformation. Compared to the corresponding group in TUG891, the enone bridge of bavachalcone is shifted by about 2.3 Å toward TM3 of FFAR4. To avoid steric clash, F115^3.29 in the bavachalcone-bound structure is displaced by 1.6 Å relative to its position in the TUG891-bound structure (Supplementary Fig. 11c). In addition, the extracellular ends of TM5 and TM6 shift toward the ligand-binding pocket (Supplementary Fig. 11c). Accompanying these extracellular changes, the intracellular ends of TM5 and TM6 move toward the G protein-binding pocket by 1.1 Å and 2.5 Å, respectively (Supplementary Fig. 11d). In addition, in the bavachalcone-bound FFAR4, R240^5.71 forms a hydrogen bond with Q144^ICL2, stabilizing the cytoplasmic end of TM3 and the TM3-TM4 loop more effectively than in the TUG891-bound complex (Supplementary Fig. 11d). These structural insights define the atomic-level mechanism of bavachalcone binding, receptor activation, and G-protein engagement, providing a foundation for designing FFAR4 agonists.

In vivo pharmacology studies of bavachalcone

To systematically evaluate the therapeutic potential of bavachalcone, an FFAR4 agonist, we first investigated its efficacy in two established models of metabolic dysfunction (Supplementary Fig. 12a, h). In the HFD-induced model, bavachalcone treatment (particularly at 10 mg/kg) produced dose-dependent reductions in body and liver weight (Supplementary Fig. 12b, c), lowered hepatic lipid profiles (TC, TG) and improved markers of liver function (ALT, AST) (Supplementary Fig. 12e). Consistently, bavachalcone treatment also ameliorated hepatic steatosis and reduced lipid deposition (Supplementary Fig. 12d). In addition, bavachalcone administration significantly improved glucose tolerance and insulin sensitivity (Supplementary Fig. 12f, g). Consistent beneficial effects were observed in mice fed a methionine-choline-deficient diet (MCD), in which bavachalcone markedly decreased hepatic lipid content (TC, TG) and restored liver function (ALT, AST) (Supplementary Fig. 12i). Histological analysis confirmed marked attenuation of steatosis, inflammation, and fibrosis in bavachalcone-treated mice (Supplementary Fig. 12j). Notably, bavachalcone demonstrated potent suppressive effects through FFAR4-Nr1h3 pathway, with quantitative immunoblot analysis showing substantial decreases in Nr1h3, PPARγ, SREBP1 and FASN protein expressions compared with control group (Supplementary Fig. 12k).

Given the FFAR4 expression in hepatocytes, we performed FLIPR‑based real‑time calcium mobilization assays in the AML12 hepatocyte cell line. Bavachalcone induced concentration‑dependent Ca²⁺ influx, mirroring the signaling profile of the canonical FFAR4 agonist TUG891 (Supplementary Fig. 13a), with both compounds eliciting comparable intracellular Ca²⁺ dynamics (Supplementary Fig. 13b). Moreover, siRNA‑mediated FFAR4 knockdown or pharmacological inhibition by the selective antagonist AH‑7614 abrogated Ca²⁺ influx and blunted bavachalcone’s lipid‑lowering activity (Supplementary Fig. 13c–h). Furthermore, to delineate the target specificity of bavachalcone in vivo, we employed FFAR4^LKO mice (Fig. 9a). In the FFAR4^LKO-HFD model, bavachalcone administration failed to mitigate body weight and liver mass gain, alongside persistent hepatic lipid accumulation (TC, TG) and liver dysfunction (elevated ALT, AST), impaired glucose homeostasis, sustained insulin resistance and unresolved hepatic lipid deposition (Fig. 9b–g). This stood in marked contrast to the robust lipid‑lowering efficacy observed in FFAR4^flox/flox‑HFD littermates, where bavachalcone significantly suppressed the protein expression of Nr1h3, PPARγ, SREBP1, and FASN. These effects were largely abrogated in FFAR4^LKO‑HFD mice (Fig. 9h). Collectively, these findings demonstrate bavachalcone alleviates hepatic steatosis by modulating FFAR4-Nr1h3 in a receptor-dependent manner.

Fig. 9. Bavachalcone ameliorates diet-induced hepatic steatosis via FFAR4-dependent modulation of the Nr1h3-PPARγ signaling axis.

a Experimental design schematic: FFAR4^flox/flox and FFAR4^LKO mice were fed HFD for 16 weeks, with daily oral gavage of vehicle (0.5% CMC-Na) or bavachalcone (10 mg/kg) administered during the final 4 weeks. Created in BioRender. Kong, Y. (2026) https://BioRender.com/w19a426. b Body weight. c Liver weight. d Hepatic TG, TC, ALT and AST levels (n = 5). e, f GTT (e) and ITT (f) tests presented as area under the curve (AUC) (n = 5). g Representative macroscopic liver morphology with corresponding Oil Red O and H&E staining (scale bar, 50 μm). h Immunoblot analysis of hepatic PPARγ, SREBP1, FASN, and Nr1h3 protein expression (n = 3). β-actin served as a loading control. Data are displayed as mean ± SEM. Significance was assessed by one-way ANOVA with Tukey’s multiple comparisons test. Source data are provided as a Source Data file.

Discussion

FFAR4, a key metabolic regulator, has been extensively documented to orchestrate glucolipid metabolic homeostasis in metabolic syndrome^8,44–46. However, with the evolving concept of precision medicine, increasing attention has been directed toward the tissue-specific contributions to disease onset and progression. In this context, the functional role of FFAR4 across different tissues has come under renewed scrutiny. FFAR4 is expressed in various tissues, and its function may vary, or even oppose, depending on the tissue microenvironment, ligand availability, and downstream signaling cascades^23,47–49. To systematically evaluate the tissue-specific regulatory influence of FFAR4 on metabolic homeostasis, we generated three distinct tissue-specific knockout mouse models. Strikingly, only hepatic deletion of FFAR4 significantly altered the progression of metabolic syndrome, while loss of FFAR4 in muscle or adipose tissue had minimal impact, underscoring the liver as a critical site of FFAR4 action in metabolic pathophysiology. Previous studies proposed that PUFAs activate FFAR4 in adipose tissue to suppress inflammation, attenuate lipid accumulation, and improve insulin sensitivity^50,51. However, these findings remain equivocal, as PUFAs are rapidly metabolized into compounds that do not engage FFAR4, complicating the attribution of observed effects to FFAR4 activation per se^15,52. Thus, PUFA-based models often lack target specificity and causal clarity. Our results unequivocally demonstrate that FFAR4 function in the liver, but not in adipose or muscle, plays a determinant role in metabolic syndrome, thereby resolving lingering uncertainties about tissue-specific causality and offering a clearer mechanistic framework for FFAR4 agonism.

We further delineate a regulatory cascade in which hepatic FFAR4 deficiency promotes heterodimerization between Nr1h3 (LXRα) and PPARγ, leading to the upregulation of lipogenic genes and metabolic dysfunction. While Nr1h3 is known for its roles in metabolism and inflammation^53,54, our study redefined Nr1h3 as a co-regulator of PPARγ in a non-classical pathway, emphasizing the complexity of nuclear receptor networks. Prior evidence indicates that LXRs can interact with PPAR subtypes with different affinities, and these interactions are regulated by ligand binding^55,56. Based on the ability of PPARs to dimerize with LXRs, we observed that Nr1h3 acts not merely as a nuclear receptor but also as a chaperone facilitating the nuclear translocation of PPARγ. Our results further indicate that different ligands (TUG891 vs bavachalcone) engaging FFAR4 can produce distinct signalings, which in turn differentially modulate the Nr1h3-PPARγ pathway. This regulatory cascade may therefore underlie the functional selectivity often observed with different FFAR4 agonizts. The identification of this pathway not only clarifies how hepatic FFAR4 maintains metabolic homeostasis but also provides a mechanistic basis for understanding how ligand-specific signaling preferences translate into divergent physiological outcomes.

Current development of FFAR4 agonists has largely relied on carboxylate-containing analogs of fatty acids, which restricts the chemical diversity and precision of pharmacological modulation. Recent clinical outcome of icosabutate underscores the significance of receptor selectivity and ligand bias, as well as clear pharmacodynamics, in shaping the clinical efficacy of FFAR4-targeting therapies^22,57. There is a pressing need to explore structurally distinct chemotypes that enable more precise targeting and functional selectivity. Natural products, with their vast and evolutionarily refined chemical space, represent a promising source for novel scaffolds. By integrating DMR, BRET, and FLIPR assays, coupled with functional validation in vitro and in vivo, we identified bavachalcone, a natural compound with a distinct chemical backbone, as a functionally selective FFAR4 agonist. In addition, we delineate a distinct signaling profile of bavachalcone at FFAR4. Unlike the reference agonist TUG891, bavachalcone concurrently activates Gα_q, Gα_s, and β-arrestin signaling pathways, representing a divergent functional profile. Mechanistically, FFAR4-mediated suppression of the Nr1h3-PPARγ-regulated lipogenic axis was dependent on Gα_q coupling, with minimal contributions from Gα_s or β-arrestins. Notably, although the Gα_s and β-arrestin signaling elicited by bavachalcone did not modulate hepatic lipid metabolism in our model, these branches may, analogous to EPA, mediate other processes such as anti-inflammatory^46,58,59. Further structural investigations will be essential to characterize bavachalcone-FFAR4 complexes with Gα_s, β-arrestins, and even downstream effectors, including TAB1 and NLRP3²³, thereby providing a framework for the rational design of FFAR4-targeted therapeutics. Bavachalcone is a natural constituent of the traditional medicine Psoralea corylifolia L. The use of Psoralea corylifolia L. in traditional medicine has a long history, and it remains clinically applied for the treatment of conditions such as vitiligo and osteoporosis^60,61. This established and enduring clinical use of the source plant provides supportive evidence for the favorable safety profile of bavachalcone as a potential therapeutic FFAR4 agonist.

While our findings elucidate key mechanisms underlying bavachalcone’s mode of action, several limitations should be acknowledged. First, the study relies exclusively on preclinical models, which limits direct extrapolation to human pathophysiology. In the absence of clinical validation, it remains uncertain whether the FFAR4-mediated effects observed in mice translate to patients with metabolic dysfunction, particularly individuals carrying certain FFAR4 polymorphisms. Second, although the study focused on lipid metabolic endpoints, it did not systematically assess other relevant physiological consequences such as insulin sensitivity or fibrosis progression, potentially overlooking broader implications of FFAR4 modulation. Thirdly, the potential off-target covalent reactivity of the electrophilic C = C motif was not fully assessed. Although our data support an FFAR4-associated mechanism, this motif may act as a Michael acceptor-like structure and react with cellular nucleophiles, such as cysteine, lysine, or glutathione, leading to FFAR4-independent effects^62–64. Future studies using glutathione or cysteine-trapping assays and activity-based protein profiling are needed to define its covalent reactivity and target selectivity. Finally, constrained by the current experimental conditions and our ongoing structural optimization efforts, the key pharmacokinetic and pharmacodynamic profiles of bavachalcone remain uncharacterized. These unresolved gaps hinder its clinical translation, and we intend to address in our follow-up structural modification and pharmacological characterization studies. Addressing these gaps will be essential for advancing bavachalcone as a viable therapeutic candidate.

In summary, this work pinpoints hepatic FFAR4 as a therapeutic target for metabolic disease, revealing a mechanism centered on the Nr1h3-PPARγ pathway. Furthermore, we identify the natural compound bavachalcone as an FFAR4 agonist with a distinct signaling profile, highlighting its therapeutic potential for targeted intervention.

Methods

Experimental animals

All animal procedures strictly adhered to the care guidelines and Use of Laboratory Animals of the School of Medicine, Jiangnan University, and were approved by the Animal Ethics Committee of Jiangnan University (JNU20210310IRB01 and JN No. 20250315c0480930[120]). Mice were housed under specific pathogen-free (SPF) conditions, maintained at a temperature range of 20–26 °C, 50% relative humidity and a 12 h light/dark cycle. Male C57BL/6 J mice (6 weeks old, weighing 20 ± 2 g) were obtained from GemPharmatech (Nanjing, China). FFAR4^flox/flox mice (fl/fl, RRID: MGI: 7256541), engineered with exons 1 and 2 of the FFAR4 gene flanked by loxp sites using the CRISPR-Cas9 technique, were constructed commercially by Shanghai Biomodel Organism. Hepatocyte-specific FFAR4 knockout (FFAR4^LKO) mice were generated by crossing FFAR4^flox/flox mice with Alb-Cre mice were purchased from Shanghai Biomodel Organism. FFAR4 transgenic (cag/w, RRID: MGI: 7256542) mice were constructed commercially by Nanjing Biomedical Research Institute of Nanjing University. Hepatocyte-specific FFAR4 overexpression mice (FFAR4^LOE) mice were generated by crossing cag/w mice with Alb-Cre mice.

Pharmacological treatment in vivo

Diet intervention started at 8-week-old. Normal chow diet (1010083) was supplied by Jiangsu Xietong Pharmaceutical Bio-Engineering Co., Ltd. For diet-induced MAFLD, mice were fed with a high-fat diet (HFD, 60 kcal%; TP23300; Nantong Trofi Feed Technology Co., Ltd) for 16 weeks. For diet-induced MASH, mice were fed with a methionine-choline-deficient diet (MCD; TP3005G; Nantong Trofi Feed Technology Co., Ltd) for 4 weeks. For Nr1h3 inhibition in vivo, Nr1h3 inhibitor GSK2033 (10 mg/kg body weight, in 0.5% CMC-Na in saline) was intragastric administered into FFAR4^flox/flox and FFAR4^LKO mice with HFD feeding. For bavachalcone treatment in vivo, bavachalcone (5,10 mg/kg body weight) or TUG891 (5 mg/kg body weight), both dissolved in 0.5% CMC-Na/saline, were administered via oral gavage under the two dietary conditions of HFD and MCD. The dose and route of GSK2033, bavachalcone and TUG891 administration were conducted based on previously reported^19,65–69.

Metabolic analysis

Total cholesterol (TC), triglyceride (TG), alanine aminotransferase (ALT) and aspartate aminotransferase (AST) levels were measured using commercial kits (Jiancheng, China). Glucose tolerance was assessed in fasted mice following an intraperitoneal injection of 2 g/kg glucose, with blood glucose levels measured at predefined time points (0, 15, 30, 60, 90, and 120 min). Insulin sensitivity was evaluated via an insulin tolerance test (ITT) by intraperitoneally administering 0.75 U/kg insulin in non-fasted mice. The areas under the curves (AUC) were calculated to quantitatively evaluate glucose and insulin dynamics.

Oil Red O staining

Cells or liver sections processed conventionally are first fixed in 4% paraformaldehyde (PFA), then gently washed with phosphate-buffered saline (PBS). Samples were subsequently stained with 60% saturated Oil Red O working solution for 15 min to visualize lipid droplets. Stained sections were then briefly washed with 60% isopropanol and counterstained with hematoxylin.

Tissues staining

A portion of liver tissue was placed in a 4% PFA solution for fixation to enable subsequent histopathological sectioning, including Oil Red O, H&E and Masson staining. These staining procedures were performed in accordance with previous studies⁷⁰. Histological images were subsequently captured using a light microscope (3D Histech, Hungary).

Cell culture

AML12 and BNL CL.2 cells were purchased from the National Collection of Authenticated Cell Cultures, with catalog numbers GNM42 and GNM22, respectively. AML12 cells were cultured in DMEM/F12 medium (BasalMedia Technologies, L310KJ) supplemented with 1% ITS liquid media (Beyotime, C0341) and BNL CL.2 were maintained in DMEM (Gibco, C11995500), both of which were supplemented with 5% fetal bovine serum (FBS, VivaCell Biotechnology, C04001) and 1% penicillin–streptomycin at 37 °C and 5% CO₂. As previously reported⁷¹, AML12 and BNL CL.2 cells were stimulated with a mixture of palmitic acid (PA; 0.125 mM) and oleic acid (OA; 0.25 mM) at the indicated concentrations for 48 h in order to establish a hepatic steatosis model.

Small interfering RNA (siRNA) transfection

AML12 and BNL CL.2 cells were transfected at 50–70% confluence using 50 pM target-specific siRNA or negative control (NC) siRNA with jetPRIME (Polyplus Transfection, 101000046) according to the manufacturer’s instructions. The siRNAs were designed and synthesized by Gene Pharma, and their sequences used in this study are provided in Supplementary Table 2.

RNA sequencing

RNA sequencing was performed as previously described⁷². Total RNA extraction was conducted using the FastPure Cell/Tissue Total RNA Isolation Kit V2 (Vazyme, RC112-01), followed by first-strand cDNA synthesis with the HiScript III 1st Strand cDNA Synthesis Kit (Vazyme, R312-01/02). Subsequently, the Second Strand cDNA Synthesis Kit (Beyotime Biotechnology, D7172) was employed. The second-strand DNA samples were digested and labeled by using Tn5 transposase. Amplification of DNA libraries was performed with HiFi PCR Mix (CWBIO, CW2648). Libraries were quality-checked and sequenced reads on Illumina NovaSeq (completed by GENEWIZ Biotech). Reads were aligned to the mouse genome using STAR. Differentially expressed genes (FC ≥ 1.5, adjusted P < 0.05) underwent KEGG pathway analysis via Metascape (http://metascape.org).

Western blot

Protein extraction from tissues and cells was performed using RIPA lysis buffer (Solarbio, BC3710) followed by protein quantification with a bicinchoninic acid (BCA) assay kit (Novoprotein, PA002). After homogenization, lysates were centrifuged at 12,000 × g for 15 min at 4 °C, separated by SDS-polyacrylamide gels, and then transferred to a 0.45 mm polyvinylidene difluoride (PVDF) membrane (Millipore, IPVH00010). After blocking with 5% skim milk at room temperature for 2 h, the membrane was incubated with primary antibodies: Nr1h3 (Proteintech, 14351-1-AP, RRID: AB_10640525, 1:1000), FFAR4 (ABclonal, A18689, RRID: AB_2862424, 1:1000), FASN (Cell Signaling Technology, 3180, RRID: AB_2862424, 1:1000), PPARγ (Cell Signaling Technology, 2435, RRID: AB_2166051, 1:1000), SREBP1 (Proteintech, 14088-1-AP, RRID: AB_2255217, 1:1000), β-arrestin1 (ABclonal, A0998, RRID: AB_2255217, 1:1000), β-arrestin 2 (Proteintech, 10171-1-AP, RRID: AB_2255217, 1:1000) and β-actin (ABclonal, AC026, RRID: AB_2255217, 1:1000) overnight at 4 °C, and subsequently mixed with HRP-conjugated goat anti-rabbit IgG H&L (Biodragon, BF03008, RRID: AB_3073988, 1:4000) at room temperature for 2 h. Membrane visualization was achieved by employing an enhanced chemiluminescent (ECL) reagent (Millipore, WBKLS0500).

Quantitative real-time PCR (qPCR)

qPCR analysis was conducted using Hieff UNICON qPCR SYBR Green Master Mix (Shanghai YEASEN Biotech, 11198ES). Relative quantification was determined using the 2^−ΔΔCt method, with β-actin serving as the internal control. All qPCR experiments were manipulated in triplicate. Primer sequences are presented in Supplementary Table 3.

Co-immunoprecipitation (Co-IP)

AML12 cells were lysed with 500 μL IP buffer, which was pre-chilled in ice water, followed by the addition of protease inhibitor. To eliminate nonspecific binding, lysates underwent preclearing through sequential incubation with Protein A/G Magnetic Beads (MedChemExpress, HY-K0202) and normal immunoglobulin G (IgG). After centrifugation, the clarified supernatants were subjected to immunoprecipitation overnight at 4 °C using either anti-Nr1h3 or anti-PPARγ antibody conjugated with fresh Protein A/G Magnetic Beads. The antibody-bound complexes were subsequently washed with IP lysis buffer, then eluted and boiled in 2 × SDS loading buffer prior to probed with immunoblotting.

Immunofluorescent staining

Following deparaffinization and rehydration, liver sections underwent heat-induced epitope retrieval in 10 mM sodium citrate buffer (pH 6.0) maintained at 95 °C for 30 min, subsequently blocked with 5% BSA for 1 h. AML12 cells cultured on poly-L-lysine-coated coverslips (Corning) were fixed with 4% PFA for 20 min. Cell membranes were permeabilized using 0.1% Triton X-100 (Sigma-Aldrich) in PBS for 10 min, followed by identical blocking procedures as tissue sections. Both tissue and cellular specimens were incubated with primary antibodies against Nr1h3 (rabbit polyclonal, Proteintech14351-1-AP, 1:200) and PPARγ (mouse monoclonal, Proteintech 66936-1-Ig, RRID: AB_2882260, 1:350) in antibody dilution buffer (1% BSA/PBS) overnight at 4 °C. Following wash with PBS, specimens were incubated with secondary antibodies: Goat Anti-Rabbit IgG H&L (Alexa Fluor 488, Abcam, ab150077, RRID: AB_2630356, 1:500) and Goat Anti-Mouse IgG H&L (Alexa Fluor 594, Abcam, ab150116, RRID: AB_2650601, 1:500) for 1 h at RT protected from light, mounted with DAPI-containing vectashield mounting medium (Solarbio, C0065), and visualized under a confocal laser scanning microscope (Zeiss, LSM 880).

Dynamic mass redistribution (DMR) assay

All DMR assays were systematically conducted on Epic Imager and Epic Filtration (Corning, NY, USA) as previously described⁷³. Chinese hamster ovary (CHO)-K1 cells (GNHa7), obtained from the National Collection of Authenticated Cell Cultures, were stably transfected to express FFAR1-4. Cells were seeded at 15,000 cells/welll in Epic 384-well biosensor microplate and cultured under standard conditions for 20 h to establish confluent monolayers (95% confluence). Prior to DMR assay, cellular microenvironments were washed and meticulously standardized in 30 μL of assay buffer (1 x HBSS, 20 mM HEPES, pH 7.4) for 60 min. Stock compounds were dissolved in pure DMSO, stored at − 20 °C, and diluted prior to use. For the DMR agonist assay, a 2 min baseline was recorded, then compounds were added, and the DMR signal was monitored continuously for 1 h.

Fuorometric imaging plate reader (FLIPR) assay

Calcium release was detected by the FLIPR Tetra instrument (Molecular Devices). Stable CHO-K1-FFAR4 cells were seeded into 384-well black-walled plates (2.5 × 10⁴ cells/well) and cultured overnight. Cells were then incubated with Calcium-6 dye (Molecular Devices, USA) for 2 h at 37 °C in HBSS containing 20 mM HEPES (pH 7.4). For agonist tests, TUG891 and bavachalcone were diluted with assay buffer and then automatically dispensed into wells. For antagonist tests, cells were pretreated with AH-7614 at varied concentrations, followed by the addition of TUG891 (6.25 μM) and bavachalcone (20 μM). Emission at 520 nm (excitation 488 nm) was monitored for 5 min. Data were presented as fluorescence (arbitrary units) versus time.

Measurement of intracellular Ca²⁺ concentration

Intracellular Ca²⁺ concentration was measured using the Ca²⁺-sensitive dye Fluo‑4 AM (Beyotime, S1061S). Briefly, cells were seeded and subsequently incubated with Fluo‑4 AM at 37 °C for 30 min. Following incubation, fluorescence signals were detected using a microplate reader at excitation/emission wavelengths of 490/525 nm, and data were expressed as relative fluorescence units (RFU).

Gα-Gγ dissociation assay

Activation of G proteins was tracked using the BRET2-based TRUPATH biosensor system (Addgene, kit #1000000163, USA)⁷⁴. Briefly, HEK293 cells (CRL-1573), purchased from ATCC, were co-transfected with the human FFAR4-short (pcDNA3.1), along with the following TRUPATH plasmids at a 1:1:1:1 DNA ratio: Gα-RLuc8 (specifically, Gα_q-RLuc8, Gα_i-RLuc8 or Gα_S-RLuc8), Gβ₃ and Gγ₉-GFP2. Transfection was performed using PEI40000 (Shanghai YEASEN Biotech, 40816ES). After 24 h, cells were reseeded into white 96-well plates (50,000 cells/well) coated with PEI10000 (Aladdin) and cultured for an additional day. Prior to BRET2, the growth medium was carefully exchanged for 60 µL assay buffer (1 x HBSS, 20 mM HEPES, pH 7.4), and 10 µL freshly prepared coelenterazine 400a (50 μM) (GLPBIO, Montclair, CA, USA) was added. After 5 min equilibration, 30 µL of test compounds were employed. Luminescence was read on a BioTek Cytation™ 5 at 410 nm (RLuc8-coelenterazine 400a) and 515 nm (GFP2), and the BRET2 ratio was calculated (GFP2/RLuc8). For experiments involving AH-7614, other procedures followed the aforementioned steps, followed by the addition of 10 µL of AH-7614 or vehicle for 5 min. Signal acquisition and BRET ratio calculation were performed as described above.

β-arrestin recruitment assay

β‑arrestin recruitment to human FFAR4-short was assessed using a BRET‑based assay. Briefly, HEK293 cells were co‑transfected with EYFP‑tagged FFAR4-short, β‑arrestin 1/β‑arrestin 2–Renilla luciferase, and GRK2 plasmid at a 3:1:1 ratio using PEI40000. After transfection for 24 h, the cells were seeded at a density of 50,000 cells per well into white 96‑well plates pre‑coated with PEI10000. Following an additional 24 h incubation, cells were treated with 6 μM coelenterazine (TargetMol, China) for 10 min, followed by sample stimulation for 15 min. BRET signals were then measured as the ratio of luminescence at 540 ± 25 nm to that at 460 ± 40 nm using a Cytation 5 microplate reader (BioTek, USA).

Plasmid construction

For structural determination, the coding sequence of human FFAR4 (UniProt ID: Q5NUL3, residues 1–361) was cloned into a modified pFastBac vector (ThermoFisher) carrying an N-terminal HA signal sequence. To facilitate expression and purification, an N-terminal cassette comprising a FLAG tag, cytochrome b562RIL (BRIL)⁷⁵, and a Gly-Ser-Ala linker was fused to FFAR4. Previous work confirmed that the BRIL insertion does not impair ligand-induced FFAR4 signaling²³. For the FFAR4-Gα_q complex, an engineered Gα_iq chimera was generated from Gα_i1 by replacing residues 329–354 of its α5 helix with residues 334–359 from Gα_q. Human Gβ₁ and Gγ₂ were separately cloned into pFastBac.

Preparation of scFv16 antibody

The scFv16 gene was modified to carry an N-terminal GP67 signal peptide and a C-terminal TEV site followed by an 8 × His tag, then subcloned into pFastBac. scFv16 was expressed and purified as reported⁷⁶, snap-frozen in liquid nitrogen, and preserved at − 80 °C for utilization in FFAR4-Gα_iq complex formation.

Expression and purification of FFAR4-Gα_iq complex

The Spodoptera frugiperda (Sf9) insect cells (2.5 × 10⁶ cells/mL; Invitrogen, 11496015) were co-infected with four baculoviruses (FFAR4, Gα_iq, Gβ₁, Gγ₂) at a 1:1:1:1 ratio⁷⁷. After 48 h at 27 °C, cells were harvested (3000 × g, 10 min), flash-frozen, and stored at − 80 °C. All subsequent stages were carried out under 4 °C conditions.

Cell pellets were resuspended in lysis buffer (20 mM HEPES pH 7.5, 75 mM NaCl, 5 mM MgCl₂, 5% glycerol and 0.3 mM TCEP) containing protease inhibitors and lysed by high-pressure homogenization. FFAR4-Gα_iq complex formation was initiated by adding 10 μM bavachalcone plus apyrase (25 mU/ml, NEB) for the FFAR4-Gα_iq complex. After 1 h, membranes were solubilized with 0.5% lauryl maltose neopentyl glycol (LMNG, Anatrace) and 0.1% cholesteryl hemisuccinate (CHS, Anatrace) for 2 h. Lysates were clarified (100,000 × g, 1 h) and incubated with FLAG affinity resin (GenScript) for 2 h. The resin was washed successively with: (i) 5 column volumes of 20 mM HEPES pH 7.5, 100 mM NaCl, 10 μM bavachalcone, 0.05% LMNG and 0.01% CHS and 0.3 mM TCEP; (ii) 5 column volumes buffer containing 20 mM HEPES pH 7.5, 100 mM NaCl, 10 μM bavachalcone, 0.0075% LMNG, 0.0015% CHS, 0.0025% GDN, and 0.3 mM TCEP. Complexes were eluted with 200 μg/mL FLAG peptide in buffer comprising 20 mM HEPES pH 7.5, 100 mM NaCl, 10 μM bavachalcone, 0.0075% LMNG, 0.0015% CHS, and 0.0025% GDN. Eluted fraction was incubated with scFv16 for 1 h, followed by further purification via SEC (Superose™ 6 Increase 10/300 GL) under conditions of 20 mM HEPES pH 7.5, 100 mM NaCl, 10 μM bavachalcone, 0.3 mM TCEP, 0.00075% LMNG, 0.00015% CHS, and 0.00025% GDN. The peak fractions were concentrated to 5–8 mg/ml (100 kDa Amicon Ultra, Millipore) for cryo-grid preparation.

Cryo-EM grid preparation and data collection

Aliquots (3 μL) of FFAR4-Gα_iq complex (concentration 5–8 mg/ml) were applied onto Quantifoil R1.2/1.3 300-mesh gold grids treated with glow discharge, blotted at 100% humidity and 4 °C, and plunge-frozen in liquid ethane using a Vitrobot Mark IV (Thermo Fisher). Grids were preserved in liquid nitrogen. Movies were recorded on a 300 kV Titan Krios (FEI) equipped with a Gatan K3 Summit detector and 15 eV slit energy filter using SerialEM⁷⁸. Acquisition parameters: pixel size 1.061 Å, defocus range − 1.5 to − 2 μm, total dose 60 e-/Å2 over 4 s (34 frames), 6378 movies collected.

Image processing and 3D reconstruction

Processing was carried out in CryoSPARC 4.6.2⁷⁹. Frames were motion-corrected (Patch Motion) and CTF parameters estimated (Patch CTF). Micrographs with resolution < 5 Å and defocus 7000-30,000 Å were selected. Particles were picked (blob picker, 60–180 Å), 2 × binned, and subjected to iterative 2D classification (400 classes). Cleaned particles were used for ab initio reconstruction (5 classes) followed by heterogeneous refinement. After multiple rounds, the best class was re-extracted without binning and refined (non-uniform and local refinement), yielding a 2.99 Å global map. Local reconstruction focusing on the FFAR4-Gα_iq core improved receptor density, producing a 3.25 Å map lacking Gβ₁ and Gγ₂.

Model building and structure refinement

Initial models were generated in ModelAngelo⁸⁰ and manually adjusted in Coot⁸¹. Ligand restraints were prepared with PHENIX.elbow⁸². Iterative manual rebuilding (Coot) and real-space refinement (phenix.real_space_refine⁸³) were performed with secondary-structure and geometry restraints. Validation (clash score, MolProbity, Ramachandran statistics) used phenix.comprehensive_validation⁸⁴. Figures were prepared in ChimeraX⁸⁵.

Statistical analysis

The data are depicted as means ± standard error of the mean (SEM). GraphPad Prism 8.0, SPSS 21.0, and R 3.6.0 were performed for data analysis. Unpaired Student’s t test was used for two‑group comparisons, and one‑way analysis of variance (ANOVA) with Tukey’s multiple comparisons test was applied for multiple‑group comparisons. Statistical significance was considered as P < 0.05.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

Y.L.K. performed most experiments, data analysis, and drafted the initial manuscript. J.X.W. designed drug screening protocols and helped in the manuscript writing. Z.W. and S.T.Y. contributed to the development of the strategy for bavachalcone-FFAR4-Gα_iq complex formation, and acquired the accession to the cryo-EM facility for collecting cryo-EM data. X.L.Y. contributed to the complex formation and purification. W.W., H.W., and W.J.D. took part in the animal experiments. Y.F.L., F.F.X., T.H., and Y.P.Z. assisted in screening and identification of FFAR4 agonizts from natural compound libraries. B.Y.Z contributed to mutant construction during the manuscript revision. X.K.Y. provided design ideas and writing during the revision. Y.Q.C. participated in the study design. X.M.L. helped in the study design and discussion. S.L.Z. organized the studies, supervised the project, and revised the manuscript.

Peer review

Peer review information

Nature Communications thanks Graeme Milligan and the other anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Funding

S.Z. discloses support for the research of this work from the National Natural Science Foundation of China (grant number 32471225-3247090211). Z.W. discloses support for the research of this work from the National Natural Science Foundation of China (grant number 32000896) and the Natural Science Foundation of Shanghai (grant number 24ZR1477600). X.Y. discloses support for the research of this work from the National Key Research and Development Program of China (grant number 2022YFC3400500).

Data availability

The data that support the findings of this study are available in the main text, the supplementary information, and the Source Data file. The RNA-Seq data generated have been deposited in the Genome Sequence Archive (GSA) at the National Genomics Data Center (China) under the accession code PRJCA048977. Atomic coordinates for the Bavachalcone bound FFAR4-G_iq complex (local refinement) and the Bavachalcone bound FFAR4-G_iq complex reported in this study are deposited in the RCSB with accession codes 9XFJ and 9XFI, respectively. The corresponding cryo-EM density maps have been deposited into the Electron Microscopy Data Bank with accession numbers EMD-66824 and EMD-66823. Source data are provided in this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-026-75589-2.