Animal health and behaviour were monitored once every 2 days. The mice with methionine-choline deficient (MCD) diet-induced NASH, mice with NASH and intestinal decontamination, and Hnf1α knockout (Hnf1α^H-KO) mice with NASH were sacrificed at the end of 6, 4 and 24 weeks, respectively, and blood (from cardiac puncture), liver, terminal ileum and caecal content samples were collected. The mice were sacrificed by cervical dislocation following the intraperitoneal injection of pentobarbital sodium (>120 mg/kg). Animal welfare was considered, including minimizing suffering and distress, and using the most appropriate dose of anaesthetic. In addition, death was verified by respiratory and cardiac arrest. All animal experiments were approved by the Naval Medical University (approval no. SYXK2021-0075) and Nanchang University (approval no. LL-202303280001).

A total of 24 male C57BL/6 mice (age, 6 weeks; weight, ~20 g) were purchased from the Shanghai Experimental Center of the Chinese Academy of Sciences and were housed in the Experimental Animal Center of the Second Military Medical University (also known as Naval Medical University, Shanghai, China) or Nanchang University (Nanchang, China) in a specific pathogen-free environment at 24°C and 50% humidity under a 12-h light/dark cycle. The mice were randomly divided into two groups: One group was fed a methionine-choline sufficient (MCS) diet ad libitum (n=8), and the other group was fed an MCD diet (Trophic Animal Feed High-tech Co.) ad libitum (n=16). After 2 weeks, the mice fed the MCD diet were randomly separated into two groups: The MCD group (n=8) and the MCD + rifaximin group (n=8). The mice in the MCD + rifaximin group were treated with rifaximin (MedChemExpress) by oral gavage (100 mg/kg/day) for 4 weeks (Fig. S1A). The rifaximin dose was chosen based on a previous report (22). The mice in the MCS and MCD groups were fed water according to their body weight for 4 weeks.

A total of 32 male C57BL/6 mice (age, 6 weeks; weight, ~20 g) were purchased from the Shanghai Experimental Center of the Chinese Academy of Sciences and were housed in the Experimental Animal Center of the Second Military Medical University or Nanchang University in a specific pathogen-free environment at 24°C and 50% humidity under a 12-h light/dark cycle. The mice were randomly divided into the following four groups: MCD group (n=8), MCD + rifaximin group (n=8), MCD + Abx group (n=8) and MCD + Abx + rifaximin group (n=8). To determine whether rifaximin relies on the intestinal microflora to serve a biological role in mice with NASH, after 1 week of treatment with broad-spectrum antibiotics (ampicillin 1 g/l, neomycin sulfate 1 g/l, metronidazole 1 g/l and vancomycin 0.5 g/l; MedChemExpress), which were added to the drinking water as described previously (23), 6-week-old male C57BL/6 mice in the MCD + Abx + rifaximin group were administered rifaximin (100 mg/kg/day) by oral gavage for 3 weeks. The mice in the MCD + Abx group were administered broad-spectrum antibiotics in their drinking water for 1 week and were then treated with water by oral gavage according to their body weight for 3 weeks. The mice in the MCD group were administered water for 1 week and were then treated with water by oral gavage according to their body weight for 3 weeks. The mice in the MCD + rifaximin group were administered water for 1 week and were then treated with rifaximin (100 mg/kg/day) by oral gavage for 3 weeks. All mice were fed an MCD diet for 4 weeks. Animal health and behaviour were monitored once every 2 days. All mice were sacrificed after being fed an MCD diet for 4 weeks, and blood (from cardiac puncture), liver and terminal ileum samples were collected.

Hnf1α^H-KO mice were generated by Shanghai Model Organisms Centre, Inc. by crossing mice homozygous for floxed Hnf1α (Hnf1α^f/f) with Alb-Cre transgenic mice as described in our previous research (24). Male Hnf1α^H-KO mice (age, 6 weeks; weight, ~20 g, n=10) were fed a normal chow diet ad libitum in a specific pathogen-free environment at 24°C under a 12-h light/dark cycle for 20 weeks to induce NASH, as described in our previous research (24). In addition, the male mice in the Hnf1α^f/f group (age, 6 weeks; weight, ~20 g, n=5) were fed a normal chow diet ad libitum for 20 weeks. Subsequently, Hnf1α^H-KO mice were randomly separated into two groups: The Hnf1α^H-KO group (n=5) and the Hnf1α^H-KO + rifaximin group (n=5). The mice in Hnf1α^H-KO + rifaximin group were administered rifaximin (100 mg/kg/day) by oral gavage for 4 weeks. The mice in the Hnf1α^f/f group and Hnf1α^H-KO group were treated with water according to their body weight for 4 weeks. Animal health and behaviour were monitored once every 2 days. All mice were sacrificed at the end of 24 weeks, and blood (from cardiac puncture), liver and terminal ileum samples were collected.

Mouse livers that were fixed in 4% paraformaldehyde at 4°C for 24 h were used for frozen sectioning for Oil Red O staining, and those fixed in 10% formalin at room temperature for 24 h were used for paraffin sectioning according to standard procedures. Paraffin-embedded liver sections (4 µm) were stained with haematoxylin and eosin (H&E; haematoxylin staining for 5 min at room temperature and eosin staining for 15 sec at room temperature) for histopathological examination. Oil Red O (Sigma-Aldrich; Merck KGaA) staining was performed for 20 min at room temperature to evaluate the degree of hepatocyte steatosis. Sirius red (Leagene; Beijing Regen Biotechnology Co., Ltd.) staining was performed for 1 h at room temperature for collagen detection. Images were captured using a light microscope (magnifications, ×10, ×20 and ×40). Liver histology was estimated according to the steatosis score, ballooning score, inflammation score and NAFLD activity score (NAS) designed and validated by the NASH-Clinical Research Network (25). Image analysis software Image-Pro Plus 6.0 (Media Cybernetics, Inc.) was used to semi-quantify the intensity of steatosis or collagen deposition according to the percentage of the positive area of Oil Red O staining or Sirius red staining in the corresponding field of liver tissue.

The paraffin-embedded liver sections underwent immunohistochemical staining according to standard procedures. Briefly, liver sections were deparaffinized in xylene and rehydrated in a series of alcohol concentrations. Subsequently, liver sections underwent antigen retrieval (alkali repair; Tris base 0.3 g, EDTA 0.1 g, H₂O 250 ml) at 100°C for 20 min. They were then soaked in 3% H₂O₂ for 10 min at room temperature to remove endogenous peroxidase. Liver sections were blocked with 10% goat serum (cat. no. c0265; Beyotime Institute of Biotechnology) at room temperature for 1 h. The slides were incubated overnight at 4°C with primary antibodies against α-smooth muscle actin (α-SMA; 1:1,000; cat. no. ab5694; Abcam) and collagen type 1 α1 (Col1A1; 1:1,000; cat. no. BA0325; Wuhan Boster Biological Technology, Ltd.), followed by incubation with a horseradish peroxidase-linked immunoglobulin G secondary antibody [cat. no. GK500710; reagent A; 1:500; Gene Technology (Shanghai) Co., Ltd.] at room temperature for 1 h. Finally, an GTVision^™ III Detection Rabbit/Mouse Kit [cat. no. GK500710; reagent B1; Gene Technology (Shanghai) Co., Ltd.] was used for staining for 3 min at room temperature. Images were captured using a light microscope (magnifications, ×10, ×20 and ×40).

Total RNA was extracted from the liver and distal ileum tissues following the standard TRIzol^® (Invitrogen; Thermo Fisher Scientific, Inc.) method. Subsequently, 1 µg total RNA was used as a template for cDNA synthesis using PrimeScript RT Master Mix (Takara Bio, Inc.) at 37°C for 15 min and 85°C for 30 sec. The transcript levels were detected by qPCR with a SYBR Green PCR Kit (Takara Bio, Inc.). qPCR conditions were as follows: 95°C for 5 min, followed by 40 cycles at 95°C for 30 sec and 60°C for 30 sec. mRNA expression levels were calculated using the 2^−ΔΔCq method and Gapdh was used as the internal reference (26). The primers used are listed in Table I. All reactions were repeated three times.

Total protein was extracted from the liver tissues using lysis buffer [Tris-HCl (pH 6.8), 5 ml; 20% SDS, 10 ml; glycerol, 9.9 ml] supplemented with phenylmethanesulfonyl fluoride (Beyotime Institute of Biotechnology). The protein concentration was determined using a BCA protein assay kit (Beyotime Institute of Biotechnology). Protein samples were electrophoresed on 10% SDS-PAGE gels and were then electrotransferred onto nitrocellulose membranes (cat. no. HATF29325; MilliporeSigma). After blocking with 5% skim milk in PBS-0.1% Tween, the membranes were incubated with specific primary antibodies overnight at 4°C, followed by incubation with donkey-anti-mouse or donkey-anti-rabbit secondary antibodies (1:3,000; cat. nos. 926-32212 and 926-32213; IRDye 700- or 800-conjugated; LI-COR Biosciences) for 1 h at room temperature. Finally, the signals were visualized using an Odyssey infrared imaging system (LI-COR Biosciences) at a wavelength of 700 or 800 nm. The densities of the protein bands were semi-quantified using ImageJ 1.8.0 software (National Institutes of Health). Gapdh was used as the internal control. The primary antibodies used were as follows: Pparγ (cat. no. sc-7273; Santa Cruz Biotechnology, Inc.), sterol regulatory element-binding protein 1 (Srebp1; cat. no. ab28481; Abcam), α-Sma (cat. no. ab5694; Abcam), Hnf1α (cat. no. ab272693; Abcam) and Gapdh (cat. no. BSAP0063; Bioworld Technology, Inc.).

The liver hydroxyproline content was measured according to the protocol of the Hydroxyproline Detection Kit (cat. no. A030-2; Nanjing Jiancheng Bioengineering Institute) as described in our previous study (27).

Caecal contents were collected from the mice and immediately frozen at −80°C. Microbial DNA was extracted from the caecal contents using the PF Mag-Bind Stool DNA Kit (cat. no. M9016-02, Omega Bio-Tek, Inc.) according to the manufacturer's instructions. Briefly, 150 mg caecal contents were added to microcentrifuge tubes containing lysis buffer, and 200 µl Buffer AL was added to the sample and mixed. The 1.5-ml microcentrifuge tube was used for collecting extracted DNA (centrifugation, 4°C, 13,000 × g, 5 min). The DNA quality was checked by 1% agarose gel electrophoresis. The V3-V4 region of the 16S rRNA gene sequence was PCR-amplified with primers (338 forward, 5′-ACTCCTACGGGAGGCAGCAG-3′; 806 reverse, 5′-GGACTACHVGGGTWTCTAAT-3′). The PCR conditions were as follows: 95°C for 3 min, followed by 27 cycles at 95°C for 30 sec and 55°C for 30 sec. DNA (100 ng) underwent paired end sequencing (sequencing kit: cat. no. NOVA-5144; NEXTflex Rapid DNA-Seq kit; Bioo Scientific Corporation) on the Illumina MiSeq PE300 platform (Illumina, Inc.) using standard protocols by Shanghai Majorbio Bio-pharm Technology Co., Ltd.

The initial raw sequences were clustered into operational taxonomic units (OTUs) of a 97% identity threshold using Usearch (version 7.1) (28). In addition, α diversity analysis was calculated with mothur index analysis (version v.1.30.2; http://mothur.org/wiki/calculators). β diversity analysis was performed using R software (version 3.3.1) (29) for mapping, using FastTree (version 2.1.3 http://www.microbesonline.org/fasttree/) for constructing the evolutionary tree, and using FastUniFrac (http://github.com/biocore/unifrac) for analysing the distance matrix between samples. α diversity analysis and β diversity analysis were used to assess the relative abundance of intestinal microflora among the MCS group, MCD group and MCD + rifaximin group. Non-metric multidimensional scaling (NMDS) analysis, Adonis analysis and partial least squares discriminant analysis (PLS-DA) based on amplicon sequence variant were used to show that the overall composition of the intestinal microbiome was significantly altered by different diets. Linear discriminant analysis effect size (LEfSe) analysis was used to estimate the impact of each microbiota species on the difference between the MCD group and the MCD + rifaximin group via linear discriminant analysis. Kruskal-Wallis test and Dunn's test were performed to detect the significant differences in abundance, and the groups with significant differences in abundance were identified.

Terminal ileum samples (0.1 g) were mixed with NaOH and acetonitrile and centrifuged (4°C, 5,698 × g, 5 min). After centrifugation, the supernatants were placed in a chromatographic bottle to detect bile acid levels (performed by Shanghai Majorbio Bio-pharm Technology Co., Ltd.). Chlorpropamide was used as an internal standard for bile acid levels. The bile acid concentrations of the terminal ileum samples were qualitatively and quantitatively determined by liquid chromatography-electrospray ionization-tandem mass spectrometry (LC-ESI-MS/MS) analysis method [LC-ESI-MS/MS (UHPLC-Qtrap); Waters Corporation]. The mass spectrometry system consisted of air curtain gas 40, ion spray voltage of −4,500 V, temperature 550°C, ion source Gas1 50, and ion source Gas2 50. Chromatographic separation was performed using the BEH C18 liquid chromatography column (100×2.1 mm, 1.8 µm; Waters Corporation). The sample size was 5 µl (flow rate, 0.4 ml/min), and the mobile phases were phase A (0.1% formic acid in water) and phase B (0.1% formic acid in acetonitrile). The bile acid standard was used to identify different bile acid metabolites detected by LC-ESI-MS/MS. Finally, the peak mass spectrum area of the analytical sample was substituted into a linear equation to calculate the concentration of bile acid.

Mouse serum was extracted from whole blood via centrifugation at 3,000 rpm for 15 min at 4°C. Alanine transaminase (ALT) and aspartate transaminase (AST) levels in mouse serum were measured using an automated analyser at Shanghai Sipur-Bika Experimental Animal Co., Ltd.

Data are presented as the mean ± SEM. GraphPad Prism 7.0 software (Dotmatics) was used to analyse the experimental data through one-way ANOVA or Kruskal-Wallis test among multiple groups, and two-tailed unpaired Student's t-test or Mann-Whitney test between two groups. The correlation between the gut microbiota and bile acid levels in the terminal ileum was investigated using a nonparametric Spearman's test. Gut microbiota clustering analysis was performed using Pearson's correlation coefficient. P<0.05 was considered to indicate a statistically significant difference.

After 4 weeks of rifaximin treatment, the livers of the MCD + rifaximin group were smoother and moister than those of the MCD group (Fig. 1A). Furthermore, MCD-fed mice had a significant decrease in body weight and increase in liver-to-body weight ratio compared with in the MCS-fed mice, while rifaximin treatment significantly reduced the liver-to-body weight ratio (Fig. 1B). As expected, the MCD diet caused hepatic steatosis and inflammation, eventually leading to steatohepatitis in MCD group mice compared with the MCS group. As shown in Fig. S1B, the body weight of mice in the MCD + rifaximin group did not significantly differ from that of mice in the MCD group. However, H&E staining of liver sections showed decreased lipid deposition, ballooning and interlobular inflammation in the MCD + rifaximin group compared with that in the MCD group, revealing a notable decrease in the severity of hepatic steatosis after rifaximin treatment (Fig. 1C). Additionally, rifaximin treatment significantly decreased the ballooning score, interlobular inflammation score and total NAS (Fig. 1D). Oil red O staining also showed reduced lipid accumulation in the liver after rifaximin treatment (Fig. 1E). Consistent with these findings, the mRNA expression levels of Srebp1, Pparγ and CD36, which are involved in liver lipid synthesis (30), were significantly decreased in the MCD + rifaximin group compared with in the MCD group (Fig. 1F). Moreover, compared with in the MCD group, plasma ALT in the MCD + rifaximin group was significantly lower, and plasma AST exhibited a downward trend, further indicating improvements in liver injury after rifaximin treatment (Fig. 1G). These results indicated that rifaximin treatment ameliorated NASH in MCD diet-fed mice.

With the development of NASH, hepatic stellate cells are activated, and hepatic collagen deposition increases, leading to liver fibrosis (31). Consistent with the findings of previous studies (32,33), the present study revealed that collagen deposition was increased in the livers of mice fed an MCD diet, while rifaximin treatment significantly reduced collagen deposition (Fig. 2A). Moreover, the α-Sma and Col1a1 mRNA and protein expression levels were significantly decreased after rifaximin treatment, as confirmed by immunohistochemistry and RT-qPCR (Fig. 2B-D). Additionally, compared with that in the MCD group, the liver hydroxyproline content in the MCD + rifaximin group was lower, which also indicated that rifaximin treatment alleviated liver fibrosis in mice with MCD diet-induced NASH (Fig. 2E).

To explore the molecular mechanism by which rifaximin improves NASH, 16S rRNA pyrosequencing was performed on the caecal contents of mice with NASH. It was observed that rifaximin treatment significantly decreased the faecal OTU, Chao1 estimator and Simpson index, but markedly increased the Shannon index, indicating that rifaximin could markedly reduce the total intestinal microflora and increase bacterial richness (Fig. 3A). As confirmed by plots from NMDS analysis, Adonis analysis and PLS-DA, the gut microbiota composition was substantially reshaped after rifaximin treatment (Figs. 3B, and S2A and B). Notably, LEfSe bar analysis, which estimated the magnitude of the influence of the abundance of each component on the differential effect, indicated that the phylum Epsilonbacteraeota was the most abundant gut microbiota associated with group separation between the MCD group and the MCD + rifaximin group (Fig. 3C). Notably, it was observed that, at the phylum level, MCD markedly increased the abundance of Epsilonbacteraeota, the most significantly changed bacteria, while rifaximin treatment significantly decreased the increase in Epsilonbacteraeota abundance in MCD diet-induced NASH mice (Fig. 3D and E). To further determine alterations in the gut microbiota, the intestinal microflora was analysed at the genus level. It was revealed that MCD significantly increased the abundance of Helicobacter hepaticus, a member of the Epsilonbacteraeota phylum, while rifaximin treatment markedly decreased the increase in the abundance of Helicobacter hepaticus (P=0.0000314; Figs. 3F, 3G and S2C). These results suggested that rifaximin treatment modulated the gut microbiota community, especially Helicobacter hepaticus, in the caecal contents of mice with MCD diet-induced NASH.

A close interaction exists between the intestinal microflora and intestinal bile acids; in particular, bile acids directly inhibit the growth of the gut microbiota. In addition, bile saline hydrolase (BSH), encoded by some intestinal microflora, can hydrolyse bile acids to resist the antibacterial effect of bile acids (34). To evaluate the potential link between rifaximin-induced changes in the intestinal microflora composition and bile acid levels in the terminal ileum, Spearman's correlation analysis was performed. As shown in Fig. 4A, the abundance of Helicobacter hepaticus, which was the most significantly altered bacteria in mice with MCD diet-induced NASH, was positively correlated with DCA. Therefore, it was speculated that DCA in the terminal ileum may decrease with decreasing Helicobacter hepaticus abundance after rifaximin treatment. As expected, a bile acid assay of the terminal ileum demonstrated that DCA was significantly elevated in mice with MCD diet-induced NASH, whereas rifaximin treatment markedly decreased the increase in DCA (Fig. 4B). In addition, the total bile acid content, conjugated bile acid/unconjugated bile acid ratio, and primary bile acid/second bile acid ratio did not significantly change after rifaximin treatment (Fig. 4C). DCA has been reported to be an intestinal Fxr agonist involved in activating the intestinal Fxr-fibroblast growth factor 15 (Fgf15) signalling pathway to affect liver cholesterol metabolism and bile acid synthesis (35). Therefore, the present study further detected the intestinal Fxr-Fgf15-Cyp7a1 signalling pathway. It was observed that rifaximin treatment significantly downregulated the expression of Fxr and Fgf15 in the distal ileum, but markedly increased Cyp7a1 and Cyp7b1 expression in the liver (Fig. 4D and E), compared with that in the MCD group. These results indicated that rifaximin may inhibit the intestinal Helicobacter-DCA-Fxr signalling pathway.

To determine whether the anti-NASH effects of rifaximin are mediated through the gut microbiota, the effects of rifaximin on hepatic steatosis, inflammation and fibrosis were assessed in mice with NASH and intestinal decontamination. After 1 week of treatment with broad-spectrum antibiotics (ampicillin 1 g/l, neomycin sulphate 1 g/l, metronidazole 1 g/l and vancomycin 0.5 g/l) in the drinking water, as previously described (23), the mice in the MCD + Abx + rifaximin group were administered rifaximin 100 mg/kg/day by oral gavage for 3 weeks, and the mice in the MCD + Abx group were treated with water by oral gavage according to their body weight for 3 weeks (Fig. 5A). Notably, no further improvement was observed in liver steatosis, hepatocyte ballooning or lobular inflammation after rifaximin treatment in mice with NASH and intestinal decontamination (Fig. 5B and C). In addition, Oil Red O staining showed no further reduction in lipid accumulation in the livers of the MCD + Abx + rifaximin group compared with those of the MCD + Abx group (Fig. 5D). Moreover, the protein expression levels of Pparγ and Srebp1 in the MCD + Abx + rifaximin group were not significantly different from those in the MCD + Abx group (Fig. 5E). Additionally, α-Sma protein expression was not significantly reduced after rifaximin treatment in mice with NASH and intestinal decontamination (Fig. 5F). These findings indicated that intestinal decontamination could impair the ability of rifaximin to ameliorate hepatic steatosis, inflammation and fibrosis, suggesting that the intestinal microbiota is required for the ability of rifaximin to ameliorate MCD diet-induced NASH in mice.

The present study revealed that rifaximin could improve bile acid and cholesterol metabolism by regulating the intestinal Helicobacter-DCA-Fxr pathway. Therefore, the study aimed to determine how rifaximin specifically affects liver cell function. Our previous study indicated that Hnf1α^H-KO mice spontaneously develop NASH (24). Notably, the present study revealed that rifaximin enhanced liver Hnf1α and Fxr mRNA expression compared with that in the MCD group (Fig. S1C). Therefore, it was hypothesized that the activation of hepatic Hnf1α may be required for rifaximin to ameliorate NASH in mice. To confirm this hypothesis, Hnf1α^H-KO mice were generated by crossing mice homozygous for Hnf1α^f/f with Alb-Cre transgenic mice (Fig. 6A). Male Hnf1α^H-KO mice were fed a normal chow diet for 20 weeks to induce NASH, then treated with rifaximin by oral gavage for 4 weeks (Fig. 6B). As shown in Fig. 6C, the protein expression levels of Hnf1α in the Hnf1α^H-KO group and Hnf1α^H-KO + rifaximin groups were lower than those in Hnf1α^f/f group. Notably, it was observed that liver steatosis, hepatocyte ballooning, lobular inflammation and NAS in the Hnf1α^H-KO + rifaximin group were not significantly different from those in the Hnf1α^H-KO group (Fig. 6D and E). In addition, no further improvement in lipid accumulation was observed after rifaximin treatment in mice with Hnf1α knockout (Fig. 6F). Moreover, the protein expression levels of Pparγ and Srebp1 in the Hnf1α^H-KO + rifaximin group were not significantly different from those in the Hnf1α^H-KO group (Fig. 6G). Additionally, α-Sma protein expression was not significantly reduced after rifaximin treatment in Hnf1α^H-KO mice with NASH (Fig. 6H). Based on these results, it was indicated that the anti-NASH biological effects of rifaximin depend on the activation of liver Hnf1α.

On the basis of these results, it could be concluded that the anti-NASH biological effects of rifaximin depend on modulation of the intestinal Helicobacter-DCA-Fxr-Hnf1α signalling pathway.