Butyrate Protects Mice Against Methionine–Choline-Deficient Diet-Induced Non-alcoholic Steatohepatitis by Improving Gut Barrier Function, Attenuating Inflammation and Reducing Endotoxin Levels

1The prevalence of non-alcoholic fatty liver disease (NAFLD), a major public health concern, has increased worldwide, and this disease is commonly associated with obesity, insulin resistance, diabetes, and the metabolic syndrome. NAFLD encompasses a pathological spectrum ranging from isolated steatosis to non-alcoholic steatohepatitis (NASH), which can eventually progress to fibrosis, cirrhosis or hepatocellular carcinoma.

2A combination of genetic, metabolic, lifestyle, and environmental factors substantially contribute to the pathogenesis of NAFLD. However, despite extensive studies conducted in recent decades, the underlying molecular mechanisms of NAFLD progression remain mostly unclear; common treatment strategies, other than modifiable lifestyle factors such as weight loss diets and/or intensive physical activity are not yet available, and no effective drug for NASH treatment has been approved by the Food and Drug Administration.

3Increasingly, studies have confirmed the important role of the gut microbiota in NAFLD development and progression, but the underlying mechanisms remain unclear. Endogenous short-chain fatty acids (SCFAs) are mainly produced from dietary carbohydrates by the gut microbiota, composed of many intestinal commensal bacteria and/or probiotic bacteria. Interestingly, dietary supplementation with SCFAs, specifically butyrate, has been shown to protect against high-fat diet (HFD)-induced obesity, insulin resistance and hepatic steatosis in animal models. However, the ability of butyrate to protect mice from methionine–choline-deficient (MCD) diet-induced NASH development and to modulate the gut microbiota and gut luminal metabolism has not yet been determined. We hypothesized that butyrate can attenuate liver steatosis and liver injury and improve gut microbiota and gut luminal metabolism in a mouse model of NASH induced by an MCD diet.

4The animal protocol was approved by the Animal Care Committee of Zhejiang University School of Medicine (permit number: 2017-591) and performed in accordance with the “Guidelines for Experimental Animals” of the Ministry of Science and Technology (Beijing, China).

5Eight-week-old male C57BL/6J mice were housed in a laminar flow, specific pathogen-free facility. After their arrival, the mice were acclimatized for 1 week with free access to a standard chow diet and water. Then, the mice were randomly divided into the following four groups (n = 15–17/group) for a 6-week-long feeding period: (1) One group was fed a methionine–choline-sufficient (MCS) diet (Research Diet, New Brunswick, NJ, United States) and was treated by gavage with 0.12 g/ml of sodium butyrate (SoB) (Sigma-Aldrich, St. Louis, MO, United States) (Control + SoB group, 0.6 g/kg body weight). (2) One group was fed an MCD diet (Research Diet, New Brunswick, NJ, United States) and was treated by gavage with 0.12 g/ml of SoB (MCD + SoB group, 0.6 g/kg body weight). (3) The Control group received the MCS diet and vehicle administered by gavage. (4) The MCD group received the MCD diet and vehicle administered by gavage. The specifications of the MCD and MCS diets are summarized in Supplementary Table S1. After 6 weeks, the mice were euthanized by intraperitoneal injection of 4% chloral hydrate (with 1 mg/100 ml atropine to inhibit respiratory secretions) for tissue collection.

6Fecal samples were collected from all mice upon defecation at the end point (week 6) and were stored in a -80°C freezer for the subsequent analyses. Blood samples were collected immediately after the mice were anesthetized and were centrifuged at 3,000 rpm for 15 min; all serum aliquots were kept at -80°C. Liver and colon samples were either fixed in 4% neutral-buffered formaldehyde and embedded in paraffin or in Allprotect Tissue Reagent (Qiagen, Valencia, CA, United States) or snap-frozen in liquid nitrogen and stored at -80°C.

7Paraffin-embedded liver sections were stained with hematoxylin–eosin (HE) or Masson’s trichrome to detect liver injury and fibrosis and were scanned using a NanoZoomer Digital Pathology system (Hamamatsu Photonics, K.K., Japan), which digitally scans the sections into a specific image format for further evaluation. The HE-stained sections were scored using the NAFLD activity score (NAS) system. Fibrosis was analyzed and quantified using Image-Pro Plus software (version 6.0, Media Cybernetics, Rockville, MD, United States). For each section, the blue regions (collagen) were normalized to the red regions (hepatocytes). The fibrosis index (%) was calculated as a percentage of the total tissue area and represented the average of six randomly selected fields (20× magnification) from each section. Liver triglycerides (TG) levels were determined using a TG assay kit (Applygen Technologies Inc., Beijing, China) according to manufacturer’s protocols. Final concentrations were normalized to the corresponding protein content.

8Alanine aminotransferase (ALT) and aspartate aminotransferase (AST) levels were measured using a dry chemistry analyzer (FUJI DRI-CHEM 7000V, FUJIFILM, Tokyo, Japan). To indirectly quantify endotoxin levels, lipopolysaccharide (LPS)-binding protein (LBP) concentrations were measured by ELISA according to the manufacturer’s guidelines (Guduo, Shanghai, China) because of the well-documented inaccuracies inherent to direct endotoxin measurements. Serum cytokine levels were measured using a Bio-Plex Mouse Cytokine 23-Plex Panel kit (Bio-Rad, Hercules, CA, United States) according to the manufacturer’s instructions. The following mouse cytokine levels were analyzed for each sample: interleukin (IL)-1α, IL-1β, IL-2, IL-3, IL-4, IL-5, IL-6, IL-9, IL-10, IL-12 p40, IL-12 p70, IL-13, IL-17α, eotaxin, granulocyte colony-stimulating factor (G-CSF), granulocyte-macrophage colony-stimulating factor (GM-CSF), gamma interferon (IFN-γ), keratinocyte-derived chemokine (KDC), macrophage chemoattractant protein 1 (MCP)-1, macrophage inflammatory protein-1α (MIP)-1α, MIP-1β, regulated upon activation normal T-cell expressed and secreted (RANTES), and tumor necrosis factor (TNF)-α.

9Paraffin-embedded liver sections were stained for F4/80 (a marker of active macrophages) and alpha smooth muscle actin (α-SMA) (a hallmark of fibrosis) with standard immunohistochemical (IHC) staining procedures as previously detailed. Briefly, liver sections were incubated with a specific primary antibody (Abcam PLC, Cambridge, United Kingdom), followed by incubation with an HRP-linked secondary antibody (Dako, Glostrup, Denmark) and 3,3′-diaminobenzidine (Dako, Glostrup, Denmark) and were scanned using the NanoZoomer Digital Pathology system. F4/80+ cells were counted using Image-Pro Plus. Quantitative analysis of α-SMA was performed using Image-Pro Plus software as previously described. Six fields (20× magnification) of view were randomly selected on each section. The mean optical density of the images was considered the representative staining intensity of the target protein. Likewise, paraffin-embedded colon sections were stained for ZO-1 (a marker of the intestinal barrier) with standard immunofluorescence staining procedures as previously described. In brief, colon sections were incubated with rabbit polyclonal ZO-1 antibody (Proteintech, Rosemont, IL, United States), followed by incubation with Texas Red-conjugated goat anti-rabbit antibody (Jackson, PA, United States) and 4′,6-diamino-2-phenyl indole (DAPI) (Sigma-Aldrich, St. Louis, MO, United States) and were visualized with a Zeiss LSM T-PMT confocal microscope (Zeiss, Jena, Germany).

10Total liver and colon tissue RNA was extracted using an RNeasy Plus Mini kit (Qiagen, Valencia, CA, United States) according to the manufacturer’s protocols. Real-time PCR was performed with an Applied Biosystems ViiA7 Real-time PCR system using a one-step SYBR PrimeScript plus RT-PCR kit (Takara Biomedicals, Kusatsu, Japan). The PCR primer sequences are listed in Supplementary Table S2. Samples were tested in duplicate in a 96-well plate. The expression data were normalized to the expression of the internal control GAPDH and were calculated according to the ΔΔCT method.

11Fecal SCFA concentration was detected by ELISA according to the manufacturer’s guidelines (Jianglai, Shanghai, China). The data were normalized to the total weight of feces.

12DNA extraction from fecal samples was conducted using a QIAamp Fast DNA Stool Mini Kit (Qiagen, Valencia, CA, United States) following the manufacturer’s protocols. DNA concentration and integrity were measured by a NanoDrop 2000 spectrophotometer (Thermo Fisher Scientific, Hudson, NH, United States) and agarose gel electrophoresis, respectively. The library for 16S rDNA amplicon sequencing was constructed based on the PCR-amplified V3–V4 variable regions. The amplicon was purified using Agencourt AMPure XP beads (Beckman Coulter, Inc., Brea, CA, United States). The qualified libraries were paired-end sequenced with an Illumina MiSeq platform according to manufacturer’s protocols. Raw sequencing data were subjected to filtration using Trimmomatic, FLASH, and QIIME software. Then, clean reads were clustered into operational taxonomic units (OTUs) using UPARSE software with a 97% threshold. A representative read from each OTU was selected with the QIIME package. Representative OTU sequences were annotated and taxonomically classified using the Ribosomal Database Project (RDP) Classifier v.2.2, which had been aligned with the Silva database version 123. The linear discriminant analysis (LDA) effect size (LEfSe) method1 was used to identify taxa with statistical significance and biological relevance.

13Metabolomic profiling was performed as previously described with some modifications. Briefly, fecal metabolites were extracted by mixing 15 mg of feces with 800 μl of ice-cold methanol (Sigma-Aldrich, St. Louis, MO, United States). After homogenization and centrifugation, the supernatant was transferred into an Eppendorf tube containing 20 μl of 1 mg/ml heptadecanoic acid (Sigma-Aldrich, St. Louis, MO, United States) as the internal standard; then, the sample was dried using a nitrogen stream (Aosheng, Hangzhou, China). The residue was reconstituted in 50 μl of 15 mg/ml methoxylamine hydrochloride (Sigma-Aldrich, St. Louis, MO, United States) in anhydrous pyridine (Sigma-Aldrich, St. Louis, MO, United States) and was incubated at 37°C for 24 h. Then, 50 μl of N,O-bistrifluoroacetamide (BSTFA) [with 1% trimethylsilyl chloride (TMCS)] (Sigma-Aldrich, St. Louis, MO, United States) was added to the mixture, and the sample was incubated at 70°C for 120 min. Metabolomic analysis was performed with gas chromatography-mass spectrometry (GC-MS) on an Agilent 7890A GC system coupled to an Agilent 5975C inert mass selective detector (MSD) system (Agilent Technologies, Santa Clara, CA, United States). For data analysis, ChemStation software (version E.02.02.1431, Agilent, Santa Clara, CA, United States) and ChromaTOF software (version 4.34, LECO, St Joseph, MI, United States) were used. Metabolites were identified by the NIST and Fiehn databases. Principal component analysis (PCA) and orthogonal partial least squares-discriminant analysis (OPLS-DA) were performed to visualize the metabolic differences among the experimental groups. Differential metabolites were selected according to the statistically significant variable importance in the projection (VIP) values obtained from the OPLS-DA model and P values from two-tailed Student’s t-tests on the normalized peak areas; metabolites with VIP values >1 and P values <0.05 were included.

14The data are presented as the means ± SEM, and the Kolmogorov–Smirnov test was used to check for normality. For most data, one-way ANOVA with Tukey’s post hoc test was used to determine the significance between the groups. The Wilcoxon rank sum test was performed to evaluate alpha diversity and principal coordinates between the different cohorts in the 16S sequencing analysis. PERMANOVA (Adonis) was used to test for microbial community clustering using weighted and unweighted UniFrac distance matrices and Bray–Curtis distance matrices. Correlations between variables were computed using Spearman’s rank correlation. P-values < 0.05 were considered significant. The data were analyzed using SPSS version 17.0 for Windows (SPSS Inc., Chicago, IL, United States).

15As expected, the liver sections from mice fed the MCD diet showed extensive macrovesicular fat accumulation, which is associated with mixed inflammatory infiltration (Figure 1A). Similarly, the evaluation of liver damage using the NAS system demonstrated major hepatic steatosis with lobular inflammation and ballooning hepatocytes in MCD diet-fed mice (Table 1). In contrast, the mice fed the MCD diet supplemented with SoB (the MCD + SoB group) predominantly developed microvesicular steatosis and had a significantly lower number of inflammatory focus in the liver (Figure 1A). These effects were also associated with a reduction in the NAS score (Table 1), but fat accumulation in the MCD + SoB group was significantly higher than that in the Control group or the Control + SoB group (Figure 1A and Table 1). After 6 weeks, compared with the Control group, the MCD group displayed significantly increased serum levels of ALT (P < 0.01), AST (P < 0.01), and intrahepatic TG (P < 0.001) (Table 1), but compared to those in the MCD group, these levels were significantly reduced in the MCD + SoB group (P < 0.05, P < 0.01, and P < 0.05, respectively; Table 1).

16The mRNA expression levels of IL-1β and TNF-α were significantly higher in the livers of mice in the MCD group than those in the livers of mice in the Control group (P < 0.05 and P < 0.05, respectively) or the Control + SoB group (P < 0.05 and P < 0.05, respectively) but were normalized by butyrate treatments (Figure 1D). Similarly, the mRNA expression of F4/80, a specific marker of macrophage infiltration, was significantly increased in the MCD diet-fed mice compared to that in the control mice (P < 0.05) and was reduced by butyrate supplementation (P < 0.001; Figure 1D). Consistent with this finding, IHC analysis indicated that the infiltration of F4/80+ cells was significantly reduced in the livers of mice fed the MCD diet supplemented with SoB compared with that in the livers of mice fed the MCD diet alone (P < 0.01; Figures 1B,C). Moreover, the expression of IL-4 and IL-10, two well-known anti-inflammatory cytokines, were downregulated in mice fed the MCD diet but were significantly up-regulated in mice concomitantly fed butyrate with the MCD diet (P < 0.01 and P < 0.05, respectively; Figure 1D). Consistent with these inflammatory modifications, lipid metabolism was impaired in the MCD diet-fed mice. The expression of peroxisome proliferator activated-receptor (PPAR-γ), which promotes fatty acid uptake and increases insulin sensitivity, was decreased in the livers of the mice fed the MCD diet (P < 0.01), and this reduction was significantly prevented by butyrate supplementation (P < 0.05; Figure 1E). Similarly, the expression of fibroblast growth factor (FGF)-21, which regulates glucose and lipid metabolism via pleiotropic actions, was significantly downregulated by the MCD diet (P < 0.001) and was restored by butyrate supplementation (P < 0.05; Figure 1E). The mRNA expression of the PPAR-γ, the coactivator PGC1-α, was also significantly reduced in the MCD group (P < 0.001), and butyrate treatment prevented this change (P < 0.05; Figure 1E). Free fatty acid receptor (FFAR2) was also involved in the observed metabolic impairment. As shown in Figure 1E, the MCD diet induced a significant decrease in hepatic FFAR2 mRNA expression (P < 0.05), and SoB supplementation significantly increased FFAR2 transcription (P < 0.01).

17To validate whether butyrate administration affected the physiological relevance of the model, we measured total SCFA concentrations. As shown in Figure 1F, compared to those in the MCD and Control groups, SCFA concentrations were significantly increased in the MCD + SoB and Control + SoB groups.

18Fibrosis progression in the liver was evaluated using Masson’s trichrome staining, which stains collagen fibers blue. According to Figure 2A, mice in the Control group displayed with a normal lobular structure with veins and radial hepatic cords in the liver. However, mice in the MCD group displayed with signs of fibrosis including periportal and interstitial collagen deposition. The MCD + SoB group displayed marked improvements in liver fibrosis was markedly improved, as evidenced by a decrease in the collagen-stained area in the liver sections. Quantitative analysis indicated that the MCD diet induced a fibrosis index of 17.66%, but concomitantly treatment with butyrate resulted in a 9.54% reduction in fibrosis (P < 0.001; Figure 2C).

19Additionally, we examined the mRNA levels of early pro-fibrogenic markers in the liver. Compared to the MCS diet, the MCD diet significantly increased the expression of TGF-β1 (P < 0.05; Figure 2E), which is involved in the early stage of fibrosis. Interestingly, butyrate treatment significantly downregulated TGF-β1 expression (P < 0.05; Figure 2E). Similarly, the MCD diet induced the expression of Acta2, a gene encoding alpha smooth muscle actin (α-SMA), but the Acta2 expression level was significantly reduced in mice concomitantly fed butyrate (P < 0.001; Figure 2E). IHC analysis further confirmed our results: compared to that in the Control group, the protein expression of α-SMA was significantly upregulated in the MCD group (P < 0.001; Figures 2B,D). Butyrate supplementation significantly reduced the level of α-SMA-positive staining in cells (P < 0.01), which was consistent with the Masson’s trichrome staining results (Figures 2B,D).

20Toll-like receptor (TLR)-4 and its co-receptor CD14 are involved in LPS recognition. In this study, MCD induced the mRNA expression of endotoxin-associated TLR4 and CD14 in the liver (P < 0.01 and P < 0.001, respectively), and butyrate treatment significantly reversed these effects (P < 0.01 and P < 0.001, respectively; Figure 2F).

21It has been demonstrated that cytokines, especially pro-inflammatory cytokines, play important roles in the pathological progression of NASH, which was evident in the MCD-induced NASH mouse model in this study because of the significant increase in the levels of systemic inflammatory cytokines, including the pro-inflammatory cytokines IL-1α, IL-1β, IL-2, IL-3, IL-6, IL-12 p70, IL-17α, and TNF-α, anti-inflammatory cytokines IL-4 and IL-10, and the chemokine Eotaxin (Figures 3A,B). Interestingly, butyrate treatment not only significantly down-regulated the production of pro-inflammatory cytokines (IL-1α, IL-1β, IL-2, IL-3, IL-6, IL-12 p70, IL-17α, TNF-α, and Eotaxin) but also slightly upregulated the production of the anti-inflammatory cytokines IL-4 and IL-10 in the MCD + SoB group relative to that in the MCD group (Figures 3A,B). No significant differences were found among the groups in the other cytokines studied.

22Lipopolysaccharide acts as a ligand for activating specific TLRs. Therefore, we measured the serum LBP levels to indirectly determine the effect of butyrate treatment on endotoxemia (Table 1). Mice in the MCD group displayed significantly higher LBP levels than mice in the Control group (P < 0.01). However, endotoxemia was significantly improved in the MCD + SoB group relative to that in the MCD group (P < 0.01).

23The impairment of intestinal tight junctions (TJs) has been shown to be associated with NASH. Therefore, we studied the expression and distribution of TJs proteins in the colon, where the most abundant gut microbiota resides and where butyrate is mainly produced and absorbed. As depicted in Figure 4B, the MCD diet downregulated claudin-1 and ZO-1 mRNA levels (P < 0.01 and P < 0.001, respectively), whereas butyrate treatment reversed this effect and significantly increased claudin-1 and ZO-1 expression (P < 0.05 and P < 0.05, respectively). ZO-1 immunostaining further revealed that the colon tissues in the MCD group had increased disruption and disorganization on the apical surface and in the crypts, but butyrate treatment stabilized TJs structures at the protein level, as evidenced by smooth and organized localization of ZO-1 (Figure 4A).

24Furthermore, the colon tissue adopted an expression pattern similar to that of the liver, with the MCD diet significantly increasing TLR2 and TLR4 mRNA expression (P < 0.01 and P < 0.01, respectively); however, these changes in expression levels were strongly suppressed by butyrate supplementation (P < 0.001 and P < 0.001, respectively; Figure 4C).

25To obtain further insights into the protective effects of butyrate, we investigated the impact of butyrate on the gut microbiome using metagenomic sequencing of the 16S rRNA gene. The Simpson index, an estimated species diversity value, was significantly higher in the Control + SoB group than in the Control group (P < 0.05; Supplementary Table S3). No other significant differences were observed in overall microbial richness and species diversity among the four groups, as estimated by the Chao 1 metric and the Shannon index, respectively (Supplementary Table S3).

26The unweighted (Figure 5A) and weighted (Figure 5B) UniFrac principal coordinate analysis (PCoA), which evaluates phylogenetic similarities between microbial communities, was used to calculate the beta-diversity values. The Control group and the MCD group were clearly separated into different clusters (PERMANOVA, unweighted P < 0.01, r = 0.391; weighted P < 0.01, r = 0.341). Likewise, the microbiota of the MCD + SoB group was clustered separately from that of the MCD group (PERMANOVA, unweighted P < 0.01, r = 0.298; weighted P < 0.05, r = 0.285). Similar results were obtained using the Bray–Curtis distance matrices (Supplementary Figure S1).

27Despite significant interindividual differences and diverse bacterial communities, the MCD diet with SoB supplementation clearly affected the gut microbial configuration. The top most abundant taxa at the phylum, family, and genus levels are presented in Figures 5C–E. Tenericutes was significantly more abundant in the fecal microbiota of the MCD group than in that of the Control group (P < 0.01), whereas the Actinobacteria population was significantly reduced in the fecal microbiota of the MCD group (P < 0.01). Compared with that of the MCD group, the fecal microbiota of the MCD + SoB group showed a marked decreases in the abundance of Firmicutes and Tenericutes (P < 0.01 and P < 0.05, respectively), and a significant increase in the abundance of Verrucomicrobia and Proteobacteria were significantly increased (P < 0.05 and P < 0.05, respectively). At the family level, the relative abundance of Lachnospiraceae, Bacteroidaceae, and Rikenellaceae were significantly higher in the MCD group than in the Control group (P < 0.05, P < 0.01, and P < 0.05, respectively), while the relative abundance of Bifidobacteriaceae and Desulfovibrionaceae were significantly lower in the MCD group (P < 0.05 and P < 0.05, respectively). Interestingly, compared with the MCD group, butyrate treatment in the MCD + SoB group significantly restored the abundance of Rikenellaceae (P < 0.05) and increased the abundance of Bacteroidales_S24_7_group, Alcaligenaceae, and Verrucomicrobiaceae (P < 0.01, P < 0.05, and P < 0.05, respectively). At the genus level, the fecal microbiota of the MCD group had a significantly higher abundance of Bacteroides, Blautia, and Rikenellaceae_RC9_gut_group (P < 0.01, P < 0.05, and P < 0.01, respectively), and a significantly reduced abundance of Bifidobacterium, Desulfovibrio, and Enterorhabdus (P < 0.05, P < 0.05, and P < 0.05, respectively) than the fecal microbiota of the Control group. Compared with the MCD group, butyrate supplementation significantly reduced the abundance of Rikenellaceae_RC9_gut_group and Lachnospiraceae_NK4A136_group levels in the MCD + SoB group (P < 0.05 and P < 0.05, respectively) and markedly promoted the abundance of Parasutterella and Akkermansia (P < 0.05 and P < 0.05, respectively). To further identify the distinguishing phylotypes in the gut microbiota of the different groups, we performed LEfSe analysis based on the RDP taxonomy data. Compared with the MCD group, butyrate supplementation in the MCD + SoB group partially restored the microbial structure, as evidenced by preponderant Akkermansia, Roseburia, Coprococcus, Coprobacillus, Delftia, Corynebacterium, Sutterella, Bacteroides, Clostridium, and Coriobacteriaceae populations (LDA score (log10) > 4.8), and reduced Bilophila and Rikenellaceae abundance (LDA score (-log10) > 4.8) (Figures 5F,G).

28With an untargeted strategy, we studied the fecal metabolome associated with each functional state of the gut microbiota. A total of 322 metabolites were identified and quantified in the non-targeted metabolomic analysis, covering the following KEGG metabolic pathways: carbohydrates, amino acids, peptides, lipids, nucleotides, energy, xenobiotics, cofactors and vitamins. The following non-metabolic KEGG pathways were also identified: environmental information processing, human diseases, cellular processes, and organismal systems.

29A PCA was performed (R2X = 0.550, Q2 = 0.222) using a 4-state model corresponding to the four groups (Control, MCD, Control + SoB, and MCD + SoB), and the score plot (Figure 6A) showed a clustering tendency in the direction of the X-axis (the first predictive principal component). The MCD + SoB group adopted a configuration closer to that of the Control group. An OPLS-DA using a 4-state model was then performed. As depicted in Figure 6B, the score plot of the first predictive principal component (X-axis) and first orthogonal principal component (Y-axis) showed a significant separation in the metabolomic datasets among the four groups. The cross-validated predictive ability Q2 was 0.909, indicating that a random fecal GC-MS spectrum discriminates among the four groups at 90.9% of the time. The explained variance R2 was 0.963. Thus, these results suggest distinct clustering of fecal metabolomic profiles among the groups.

30The main focus of the present analysis was to identify differential metabolites that potentially contributed to the separation between the MCD and MCD + SoB groups. Therefore, an OPLS-DA model was established for the two groups (R2Y = 0.990, Q2 = 0.950) (Figure 6C). The characteristic metabolites that were extensively altered following butyrate treatment were selected according to the OPLS-DA model, with VIP values > 1 and P-values < 0.05 (Figure 6D and Supplementary Table S4). Ultimately, 98 metabolites were selected and were mainly associated with pathways involved in the metabolism of lipids, amino acids, carbohydrates, cofactors, vitamins, and nucleotides. Specifically, the abundance of several metabolites involved in lipid metabolic pathways were significantly higher or lower in the MCD + SoB group than those in the MCD group (Figure 6D). For example, the levels of saturated fatty acids, such as stearic acid and behenic acid, unsaturated fatty acids, such as oleic acid and linoleic acid, and squalene, a natural 30-carbon triterpene, were significantly increased in the feces of mice in the MCD + SoB group, but the level of arachidonic acid, an unsaturated fatty acid, was significantly decreased by butyrate in this study.

31To further elucidate the beneficial role of the altered microbial community, we conducted a correlation analysis of the interaction matrix. The analysis was restricted to the following: representative microbial genera with significantly altered relative abundance between the MCD group and the MCD + SoB group, metabolites with a high discriminative power, and other representative injury parameters (Figure 7 and Supplementary Table S5). Bilophila and Rikenellaceae, which were clearly increased in the MCD group, may be involved in the inflammatory responses and NASH progression. For example, both Bilophila and Rikenellaceae were positively correlated with the levels of ALT, IL-17α, LBP, TLR2, TLR4, and TNF-α. Additionally, the abundance of Bilophila abundance was highly associated with that of Rikenellaceae. In contrast, the relative abundance of several butyrate-promoted probiotic strains may play a role in alleviating liver and intestinal damage. Specifically, Akkermansia was significantly negatively correlated with the levels of ALT, NAS, TG, F4/80, IL-12 p70, IL-17α, IL-1β, IL-6, TNF-α, TGF-β1, TLR2, TLR4, and LBP and was positively associated with IL-4 and ZO-1 levels. The enrichment of promising probiotics after butyrate treatment had strong correlations with fecal metabolites. The under-represented arachidonic acid levels in the MCD + SoB group were negatively associated with Akkermansia, while over-represented lactic acid levels were positively associated with Coprobacillus; squalene and stearic acid levels were positively associated with Sutterella. Squalene and stearic acid may improve liver damage, as indicated by the ALT level. Therefore, these data provide evidence that the modified gut microbiota and fecal metabolites induced by butyrate supplementation have important roles in ameliorating NASH.

32The major finding of the present study was that supplementation with SoB substantially ameliorated MCD-induced steatohepatitis in mice. Additionally, several complementary mechanisms were identified, including the following: (i) alleviation of microbial dysbiosis; (ii) protection of the intestinal barrier; (iii) improvement of the metabolomic profile; (iv) regulation of lipid metabolism; and (v) attenuation of the inflammatory response.

33The liver–gut axis plays a pivotal role in NAFLD pathogenesis and progression. TLR signaling in hepatic immune cells induces the initial inflammatory response by producing a broad array of cytokines, which may mediate the destruction of the intestinal barrier and subsequent bacterial translocation, and these events can induce secondary TLR signaling activation. We found that butyrate treatment downregulated the expression of TLR4, the TLR4 co-receptor CD14 in the liver, and TLR2 in the colon. It has been demonstrated that TLR4 signaling may promote NAFLD progression, and NASH is characterized by an increased sensitivity to bacterial LPS, a TLR4 ligand. LPS derived from the gut microbiota may be transferred to the liver through the portal circulation, consequently activating TLR4 on Kupffer cells (KCs) and hepatic stellate cells (HSCs) and triggering a cascade of inflammatory signaling pathways, which in turn can induce liver injury and fibrosis.

34Kupffer cells have been implicated in NASH development, and in the release of various inflammatory cytokines such as TNF-α, IL-6, and IL-1β upon activation. These cytokines are known to be involved in liver function impairment, liver fibrogenesis, and inflammation-driven dyslipidemia. In this study, hepatic and systemic inflammatory responses were improved in the MCD + SoB group, as evidenced by the decreased expression of pro-inflammatory cytokine genes, such as IL-1β, F4/80, and TNF-α, and increased expression of anti-inflammatory cytokine genes, such as IL-4, and IL-10, in the liver. Similarly, significantly decreased levels of pro-inflammatory cytokines (IL-1α, IL-1β, IL-2, IL-3, IL-6, IL-12 p70, IL-17α, TNF-α, and eotaxin) and increased production of anti-inflammatory cytokines (IL-4 and IL-10) were observed in the serum. As shown by IHC staining, mice in the MCD + SoB group also displayed with a significant reduction in the number of F4/80+ infiltrating macrophages in the liver, which indicated a decreased activation of KCs. Improved liver function and lipid metabolism are indicated by the ALT, AST, and TG levels, which were reduced after butyrate treatment, resulting in the alleviation of both hepatocyte injury and inflammation-driven dyslipidemia.

35Hepatic stellate cells serve as another mediator of the liver–gut axis due to their high TLRs expression. HSCs are key cells that express contractile α-SMA and produce excessive extracellular matrix (ECM) components during liver fibrosis. Activation of HSCs during liver fibrosis mainly depends on KC secretion of TGF-β1, the most powerful modulator, and, to a lesser extent, IL-1β. The increased expression of TGF-β1 and α-SMA induced by the MCD diet was significantly reduced by butyrate, and the alleviation of fibrosis was further verified by Masson’s trichrome and α-SMA IHC staining.

36The ability of butyrate to modulate inflammatory and immune responses, including the suppression of NF-κB activation by inhibiting histone deacetylase (HDAC) and to act as a signaling molecule via targeting FFAR-1, 2, and 3 are the most frequently investigated mechanisms related to this SCFA. In energy metabolism, butyrate regulates energy expenditure and mitochondrial function through the promotion of PPAR-γ coactivator PGC-1α, via AMP kinase (AMPK) activation and HDAC inhibition. Furthermore, FGF-21, which is involved in the stimulation of fatty acid β-oxidation in the liver, is also induced by butyrate. We found that butyrate could not only increased the expression of FFAR2, PPAR-γ, FGF21, and PGC-1α, which is consistent with previous studies, but may also alter the intestinal microbiota and barrier function to protect the liver from NASH.

37The elevated LPS level in NASH patients is mainly due to the impaired intestinal barrier and dysbiosis of the gut microbiota. Consistently, we found that the MCD diet disrupted barrier homeostasis, as evidenced by mRNA expression and immunostaining analyses. The MCD diet favored Anaeroplasma, Bilophila, Anaerotruncus, Turicibacter, SMB53 and Flexispira, some species of which are pro-inflammatory bacteria or opportunistic pathogens. Notably, butyrate shifted the gut microbiota composition by reducing Bilophila and Rikenellaceae abundance and strongly promoting Akkermansia, Roseburia, Coprococcus, Coprobacillus, Delftia, Sutterella, and Coriobacteriaceae genera. Some species of Bilophila were identified as LPS-producing bacteria and mucosa-damaging bacteria. Our study consistently found that the abundance of Bilophila was positively correlated with LBP levels. Some species in Rikenellaceae family have been demonstrated to be pathogenic by augmenting inflammation and the production of mutagenic toxins, which is consistent with the correlation analysis showing that Rikenellaceae abundance was positively associated with F4/80, IL-17α, IL-1β, TGF-β1, and TNF-α levels. Accumulating evidence has confirmed the beneficial roles of the butyrate-producing genera Akkermansia and Roseburia in alleviating endotoxemia and improving immunological disorders and gut barrier function. We consistently found that both Akkermansia and Roseburia were inversely correlated with several pro-inflammatory cytokine levels and positively correlated with the gut barrier marker ZO-1. Additionally, Akkermansia had a negative association with LBP levels. Our previous study found that Akkermansia muciniphila protects against immune-mediated liver injury in a mouse model by alleviating inflammation and hepatocellular death. The genus Coprobacillus and another butyrate-producing genus, Coprococcus, have been found to play beneficial roles in maintaining intestinal stability. The probiotic properties of Delftia are due to the ability of this genus to transform or degrade multiple organic and inorganic toxins. Furthermore, Sutterella, an organism that promotes a protective immunoregulatory profile in vitro, and Coriobacteriaceae, some species of which have been suggested to improve host lipid metabolism, were also increased following butyrate treatment. Therefore, the depletion of opportunistic pathogens and promotion of promising probiotics provide evidence for the beneficial role of butyrate.

38To link microbial community structure with metabolic functions, we integrated metabolomic approaches into this study. The over-represented metabolites including stearic acid, behenic acid, oleic acid, linoleic acid, and squalene in the feces of the MCD + SoB group are involved in lipid metabolism pathways for the biosynthesis of unsaturated fatty acids, steroids, and fatty acids, and in the metabolism of linoleic acid. Studies have reported that stearic acid, a potent anti-inflammatory lipid, can exert a protective effect by promoting the recovery of hepatic dysfunction. Our study also found that stearic acid was negatively correlated with several pro-inflammatory cytokines, ALT, and the NAS. suggested that behenic acid results from reduced TG absorption. Although a high concentration of oleic acid is traditionally used to induce steatosis in vitro, dietary supplementation with oleic acid has been shown to exert several protective effects associated with energy homeostasis in different experimental models of NAFLD, especially a hypocholesterolemic effect. Surprisingly, supplementation with an oleic acid-derived compound (S1) restores high fat diet-induced gut microbiota dysbiosis. A negative correlation between oleic acid and Rikenellaceae was found in our study. Recently, a combination of fatty acids including linoleic acid was found to be associated with a lower incidence of type 2 diabetes (T2D). Squalene is capable of reducing cholesterolemia through the activation of liver X receptor (LXR) α and β without promoting lipogenesis in hepatocytes. Arachidonic acid, an unsaturated fatty acid that was significantly decreased after butyrate treatment in this study, was positively associated with pro-inflammatory F4/80, IL-12 p70, and IL-6 and inversely correlated with the probiotic Akkermansia. Studies have also found that arachidonic acid exacerbates NASH through favoring pro-inflammatory microbiota, reducing butyrate production, amplifying inflammation via the TLR4-NF-κB pathway and inducing insulin resistance. Interestingly, we found that lactic acid was significantly increased in the MCD + SoB group, which indicated cross-feeding between lactate-producing bacteria and lactate-converting bacteria. Butyrate increased the abundance of lactate-producing bacteria, such as Coprobacillus, in our study. Lactate favors lactate-converting microbiota, such as Coprococcus and Roseburia, which ultimately produce butyrate, constituting a positive feedback loop in the microbiota ecosystem, which substantially amplifies the protective effects.

39Nevertheless, our study has some limitations. First, although an MCD dietary model was used as a NASH animal model in this study because morphological changes in the mouse liver resemble human NASH, characterized by hepatic steatosis and inflammatory infiltration, this model is not entirely representative of human NASH because of the absence of features such as obesity and insulin resistance. Additionally, the metabolic profile of the MCD diet-induced NASH animal model is distinct from that of NASH in humans. Therefore, further clinical studies are needed. Second, our study identified a clear correlation between the butyrate-induced shifts in the gut microbiome and metabolome and alleviated hepatic and intestinal impairment. However, whether the alterations in the gut microbiome and metabolome are the cause or the effect of the hepatic and intestinal improvement requires further investigation.

40Our findings demonstrated that butyrate induced a protective shift in the gut microbiome and metabolome, thereby effectively preventing the MCD diet-induced liver and gut impairments associated with NASH. Thus, the current study indicates that oral butyrate administration is a promising strategy for treating NASH.

41JY, LXL, WW, YL, and LJL designed the study. JY, WW, and YL collected the data for the study and wrote and edited the manuscript. DS, DF, FG, HJ, RY, and WY contributed to discussions and reviewed the manuscript. All authors read and approved the final manuscript.

42The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.