Ling-Gui-Zhu-Gan decoction ameliorates nonalcoholic fatty liver disease via modulating the gut microbiota

As part of a standardization process, we carried out a high-performance liquid chromatography (HPLC) fingerprint analysis of LG. In this HPLC fingerprint, LG and the reference standards were compared to identify the major peaks. There are seven compounds in LG, viz. (i) liquiritin apioside; (ii) liquiritin; (iii) isoliquiritin; (iv) cinnamic acid; (v) cinnamal dehyde; (vi) licoricesaponin G2; and (vii) ammonium glycyrrhizinate were properly identified (). Fig. S1

Our results showed that the body weight gain rate of the mice was significantly higher in the groups fed with HD than in the group fed with a normal chow diet (NC). After 4 weeks of administration, the body weight gain rate of the LG group was significantly lower than that of the HD group (Fig. 1a). The body fat ratio and liver index of the HD group were significantly higher than those of the NC group, which decreased significantly with LG treatment (Fig. 1b and c). According to these results, LG can reduce HD-induced NAFLD mice’s body weight gain rate, body fat ratio, and liver index. Compared to the NC group, the hepatic total cholesterol (TC) and total triglycerides (TG) levels increased significantly in HD, but the hepatic TC decreased significantly in LG (Fig. 1e). The hepatic TG and the serum free fatty acid (FFA) had a decreasing trend with LG administration (Fig. 1d and f). The hepatic HE staining sections of HD demonstrated severe micro/macrovesicular steatosis and frequent ballooning of hepatocytes, which were significantly ameliorated (Fig. 1g and h). HD liver tissue was stained with oil red O to reveal a significant reduction in liver lipid deposition after LG treatment (Fig. 1i and j). Compared to the NC group, the fasting blood glucose, fasting serum insulin, and homeostasis model assessment of insulin resistance (HOMA-IR) levels all increased significantly in HD, and the fasting blood glucose showed a downward trend in LG (Fig. 1k), while LG decreased the fasting serum insulin and HOMA-IR levels significantly (Fig. 1l and m). The levels of serum alanine aminotransferase (ALT), aspartate aminotransferase (AST), and hepatic malondialdehyde (MDA) in HD all increased significantly versus the NC group. Serum ALT and hepatic MDA were reduced by LG (Fig. 1n and p), and serum AST decreased in LG with near marginal significance (Fig. 1m). The above results showed that LG could ameliorate liver function and liver lipid deposition in mice fed HD.

To examine whether the gut microbiota is involved in the improvement of NAFLD by LG, we sequenced the 16S rRNA gene. At the phylum level, the gut microbiota is dominated by Firmicutes, Bacteroides, and Proteobacteria (Fig. 2a). Proteobacteria are composed of numerous opportunistic pathogens, whose abundance increases significantly in the HD group and decreases significantly in the LG group. At the family level (Fig. 2b), LG showed a tendency to reverse HD-induced changes in bacterial composition, which were characterized by higher abundances of S24_7 and Lachnospiraceae with lower abundances of Desulfovibrionaceae and Ruminococcaceae. Moreover, the Firmicutes/Bacteroidetes (F/B) ratio increased significantly in HD but decreased in LG with a certain trend (P = 0.0833) (Fig. 2c).

Compared to the NC group, the Simpson index of HD reflected a trend toward less diversity, while there was a recovery tendency in the diversity of LG close to statistical significance (P = 0.064) (Fig. 2d). The gut microbiota structure was significantly different among three groups (analysis of one-way similarity (ANOSIM), r = 0.7841, P = 0.001). Furthermore, β-diversity based on weighted UniFrac distances showed differences in microbiome composition between LG and HD, while LG had a similar diversity with NC (Fig. 2e). To identify key bacteria, we conducted a differential analysis using random forest analysis and the Wilcoxon rank-sum test. In total, 20 families were identified as discriminating families in the three groups via random forest analysis, and S24-7, Ruminococcaceae, and Desulfovibrionaceae were ranked as discriminative families (Fig. 2f). After analyzing the network of high-abundance families, we investigated how the microbial communities interacted. The NC group showed a cluster of strongly related bacteria Bifidobacteraceae, Lactobacilaceae, Alcaligenaceae, and Turicibacteraceae, while LG could recuperate the portion of them under the adverse impact of HD. The enriched bacteria, such as Desulfovibrionaceae, Ruminococcaceae, and Dehalobacteriaceae were associated, whereas LG reduced the enrichment of these bacteria (Fig. 2g). Spearman’s correlation heatmap was used to indicate the relations between major bacteria and significant metabolic biomarkers related to NAFLD (Fig. 2h). These bacteria enriched in HD, such as Ruminococcaceae, Desulfovibrionaceae, and Christensenellaceae, were positively correlated with metabolic biomarkers related to NAFLD. However, these bacteria enriched in NC and LG, such as S24_7, Bifidobacteriaceae, and Lactobacillaceae, were negatively correlated with metabolic biomarkers related to NAFLD, including body weight, insulin, hepatic TC, TG, and FFA, and serum levels of ALT and AST.

Phylogenetic Investigation of Communities by Reconstruction of Unobserved States (PICRUSt) analysis was performed to predict the gut microbiota’s function. Through statistical algorithms, 30 kyoto encyclopedia of genes and genomes (KEGG) pathways showed significant differences (Fig. 3a). Analysis of the clustering of the 30 KEGG pathways indicated that the LG group clustered with the NC group, which was analogous to the results of PCoA (Fig. 2e). We found that the pathways of steroid hormone biosynthesis and primary and secondary BAs biosynthesis were higher in the NC and LG groups, whereas the pathways of butanoate metabolism, synthesis, and degradation of ketone bodies were lower in NC and LG. These results suggested that LG could restore the functional profile of the HD-induced gut microbiota to the NC.

To confirm the results of PICRUSt, we examined bacteria-associated metabolites. Compared to the NC group, the concentration of lipopolysaccharide-binding protein (LBP) in serum was increased significantly in HD and decreased in LG at the margin of statistical significance (P = 0.0692) (Fig. 3b). The concentration of total BAs in serum increased significantly in HD and decreased significantly in LG (Fig. 3c).

Subsequently, we analyzed the specific components of the BAs in serum and found that the relative abundances of deoxycholic acid (DCA) and hyodeoxycholic acid (HDCA) were increased in HD compared to NC and were decreased in LG (Fig. 3d). Nevertheless, the ratio of primary/secondary BAs in serum decreased significantly in HD and increased significantly in LG (Fig. 3e).

We compared the levels of different BA components in serum between groups. Compared to the NC group, HDCA, DCA, ursodeoxycholic acid (UDCA), tauro-α-muricholic acid (T-α-MCA), ω-muricholic acid (ω-MCA), glycohyodeoxycholic acid (GHDCA), glycoursodeoxycholic acid (GUDCA), tauro ursodesoxy cholic acid (TUDCA), and α-muricholic acid (α-MCA) showed a significant increase in HD. The chenodeoxycholic acid (CDCA), cholic acid (CA), HDCA, DCA, UDCA, T-α-MCA, tauro-β-muricholic acid (T-β-MCA), ω-MCA, taurohyodeoxycholic acid, GUDCA, α-MCA, and ω-MCA in LG showed a significant decrease compared to HD (Fig. 3f). Based on the above, we found that total serum BAs can be increased by HD, but the majority of secondary BAs components, which are derivatives of primary BAs, were reduced after LG treatment.

Compared to the HD group, the concentrations of acetic acid, propionic acid, and butyric acid in feces all increased significantly in LG (Fig. 3g through i). Compared to the NC group, the concentrations of acetic acid and propionic acid in feces decreased significantly in HD, while the concentration of butyric decreased slightly in HD (Fig. 3i). Spearman’s correlation heatmap showed the relations between major bacteria and their significant metabolites related to NAFLD (Fig. 3j). These bacteria enriched in the HD, such as Peptococcaceae, Dehalobacteriaceae, Odoribacteraceae, Ruminococcus, and Paraprevotellaceae had a positive relationship with total BAs but a negative relationship with SCFAs, such as acetic acid, propionic acid, and butyric acid. However, bacteria enriched in NC and LG, such as S24_7 and Mogibacteriaceae, had a positive relationship with propionic acid and primary/secondary BAs but a negative relationship with BAs.

Based on the annotated genomes and genes differentially expressed between NC versus HD and HD versus LG, KEGG functional enrichment analysis was conducted (Fig. 4a and b).

Compared to the NC group, HD is related to the peroxisome proliferator-activated receptor (PPAR) signaling pathway, primary BA biosynthesis, and the butanoate metabolism signaling pathway. Compared to the HD group, LG could regulate the PPAR signaling pathway, cytokine‒cytokine receptor interaction, and primary BA biosynthesis signaling pathways. Reverse transcription PCR (RT-PCR) analysis was performed to verify the transcriptome results on the metabolism of lipids, BAs, SCFAs, and inflammatory signaling pathways. The expression of genes related to hepatic lipid synthesis (PPARγ, CD36, scd1, fas, Srebp1c), genes related to lipolysis (Lipe, Pnpla2, Fabp5), genes related to lipid oxidation (Cpt2, Acox1, Ppargc1a, PPARα), genes related to BAs biosynthesis (CYP7A1, CYP27A1, FgFr4, FXR, shp), and genes related to hepatic inflammation (Mcp-1, CD14, Nlrc4, Tlr2, Tlr4, IL-10) in liver tissues was measured by RT-PCR. The expression of genes related to BAs biosynthesis (Fgf15, FXR), a gene related to SCFAs (SLC5A8), and genes related to intestinal permeability (Occludin, ZO-1, Muc5) in ileal tissue was also measured by RT-PCR (Fig. 4c through g and i through k). Compared to the NC group, the expression of hepatic lipogenesis genes PPARγ, hepatic bile acid biosynthesis genes CYP7A1, hepatic inflammation genes Mcp-1, Tlr2, Tlr4, IL-10 upregulated significantly, while the expression of hepatic lipid oxidation genes Cpt2, Acox1, Ppargc1a, hepatic lipolysis gene Lipe, Pnpla2 and Fabp5, ileum SCFAs receptors SLC5A8, and intestinal permeability gene Occludin and ZO-1 downregulated significantly in HD.

However, compared to the HD group, the expression of hepatic BAs biosynthesis genes CYP27A1, hepatic inflammation genes CD14, Nlrc4, and IL-10 was downregulated significantly, while the expression of hepatic lipolysis gene Fabp5, ileum BAs biosynthesis gene Fgf15, intestinal permeability gene ZO-1, and ileum SCFAs receptors SLC5A8 was upregulated significantly in LG. The higher gene expression of PPARγ and CYP7A1 in the livers of HD and the higher gene expression of Fgf15 in the ileum of LG were verified by Western blot (Fig. 4h through j, n and o). The higher gene expression of Occludin in the ileum of LG was verified by immunohistochemistry (Fig. 4p).

The components of LG are listed in Table 1. Sichuan Neo-Green Pharmaceutical Technology Development Co. Ltd. supplied all herbs. LG was administrated at the standard dose (715 mg/kg/d) in mice, which is equivalent to that given to human patients in clinical studies (9, 42). HPLC was applied to analyze the ingredients of LG according to methods developed by the Hong Kong Chinese Materia Medica Standards (43) (Fig. S1).

The Shanghai Research Center of Southern Model Organisms (Shanghai, China) provided 6- to 8-week-old male C57BL/6 mice that were group-housed (3–4 per cage) in specific pathogen-free conditions. The Animal Ethics Committee at Shanghai University of Traditional Chinese Medicine authorized all animal experiments in this work (approval number PZSHUTCM210512413, 12 December 2019), which were carried out according to the “Guide for the Care and Use of Laboratory Animals” approved by the US National Institutes of Health.

We randomly divided mice into two groups after 1 week of acclimatization: one group was given a normal chow diet (NC, n = 10) and the other group was given a high-fat diet (HD, n = 18). Two mice per group were dissected randomly after 16 weeks on an HD diet to determine the success of the NAFLD mouse model. The HD group was then separated into two groups, each with eight mice: one group received HD and 0.5% sodium carboxymethyl cellulose intragastrically once daily; the other group received HD and LG intragastrically once daily. All animals were euthanized under anesthesia with Pentobarbita (100 mg/kg, intraperitoneal injection) after 8 weeks of treatment, and blood, liver, and intestinal tissues were collected according to strict protocols.

The Catalyst One Chemistry Analyser (IDEXX Laboratories) was used to measure the levels of serum ALT and AST. Commercial kits were used to measure total TG, TC, FFAs, and MDA in extracted hepatic samples (Nanjing Jiancheng Bioengineering Institute, A110-1-1, A111-1-1, A042-2-1, A003-1-2↗, respectively). The serum insulin and LBP levels were measured with ELISA kits (Crystalchem, 90080; Abcam, ab269542, respectively).

Mouse liver tissues were extracted and preserved for 48 hours in a 10% formalin solution. Haematoxylin and eosin staining was performed after the tissues were dehydrated, cleaned, waxed, embedded, sliced, and embedded in plastic. Oil red O staining of frozen liver slices confirmed the diagnosis of steatosis. Under a light microscope, the morphological differences of the tissues were observed, and all quantification was completed by Image-Pro Plus software 6.0.

Using Illumina high-throughput sequencing, we evaluated the composition and functional profile of the gut microbiota by sequencing the 16S rRNA gene V3–V4 region in mouse feces. Using the DNeasy PowerSoil Kit, total bacterial DNA was isolated from mouse feces (QIAGEN, Inc., the Netherlands). The forward primer 338F (5′-ACTCCTACGGGAGGCAGCA-3′) and the reverse primer 806R (5′GGACTACHVGGGTWTCTAAT-3′) were used to generate an amplicon library for double-ended (2 × 300 bp) sequencing with the Illumina MiSeq platform. The PicoGreen dsDNA Assay Kit was used to quantify PCR amplicons purified using Agencourt AMPure Beads (Beckman Coulter, Indianapolis, IN; Invitrogen, Carlsbad, CA, USA). Amplicons were pooled in equal numbers once they had been quantified. The raw readings were combined after they were quality-filtered. FLASH was used to construct paired-end readings. After chimera detection, the remaining sequences were used. UCLUST grouped the remaining high-quality sequences into operational taxonomic units (OTUs) at 97% sequence identity after chimera identification. Using default settings, a typical sequence was chosen from each OTU. The representative sequences were BLAST searched against the Greengenes database, and the best hit was used for OTU taxonomic categorization. An averaged, rounded rarefied OTU table was constructed by averaging 100 uniformly resampled OTU subsets under 90% of the minimum sequencing depth for subsequent analysis to reduce the variation in sequencing depth among samples. Overall samples, OTUs with less than 0.001% of the sequences were filtered. PICRUSt2 software (https://github.com/gavinmdouglas/q2-picrust2/releases/tag/2021.11_0↗) (44) was used to predict the functional potential of the microbial community. Metabolic pathways were annotated using the KEGG database (https://www.kegg.jp/↗) (45).

Gas chromatography/mass spectrometry (GC/MS) was used to determine the concentration of SCFAs in the feces from the caecum as previously reported (46). In summary, 50 mg of fecal material was homogenized in 0.8 mL ddH₂O, followed by 0.01% H₂SO₄ and 1 mL diethyl ether. After vortexing for 30 seconds, the mixture was centrifuged for another 5 minutes at 4°C and 10,000 rpm. Two hundred microliters of supernatant were collected using an Agilent 7000B system (Agilent Technologies, CA, USA), equipped with flame ionization and thermal conductivity detectors, capillary columns, and GC ChemStation software. Pure standards diluted in diethyl ether were used to measure acetate, butyrate, propionate, valerate, isobutyrate, and isovalerate.

Liquid chromatography/mass spectrometry (LC/MS) was used to determine the concentration of BAs in the serum as previously reported (47). The separation of BAs was conducted using an Acquity UPLC BEH C18 column (2.1 × 100 mm, 1.7 µm) (Waters, Milford, MA, USA). The sample injection volume was 5 µL, with a column temperature of 45°C. The mobile phase consisted of formic acid (0.01% in water, solvent A) and acetonitrile (solvent B), at a flow rate of 0.3 mL/minute. The gradient elution program was as follows: 0–1 minute, 55% B; 1–9 minutes, 55%–80% B; 9–11.4 minutes, 80%–90% B; 11.4–14 minutes, 90%–90% B; 14–14.1 minutes, 90%–55% B. The samples were analyzed by a mass spectrometer in negative electrospray ionization under a multiple-reaction mode. Operating parameters were ion source temperature: 120°C; desolventizing temperature: 350°C; desolventizing gas (N2) flow rate: 600.0 L/hour; cone gas (N2) flow rate: 5.0 L/hour; capillary voltage: 2.8 kV; cone voltage: 55 V. Data for the determination of BAs were analyzed and processed using Analyst Software 1.6 (Agilent).

Total RNA was selected from three randomly chosen NC, HD, and LG livers (n = 3). The RNA Seq library was constructed using the Illumina TruSeq RNA Sample Preparation Kit according to the manufacturer’s instructions. The library was then enriched for segments 300–400 bp in length using PCR amplification. The library quality was then verified using an Agilent 2100 Bioanalyzer. These libraries were then sequenced at paired ends using Next-Generation Sequencing on the Illumina HiSeq sequencing platform (Personalbio, Shanghai, China). We then utilized HTSeq to compare each gene’s read count value to the original expression of that gene using fragments per kilobase million (FPKM) to optimize the expression.

Total RNA was extracted from liver and colon tissue and reverse transcribed into cDNA utilizing the SuperScriptTM First-Strand Synthesis System (Invitrogen). A LightCycler96 (Hangzhou Bioer Technology, Share-Holding Co.), and iQ SYBR Green Supermix (BIO-RAD, USA) were used for RT-PCR. Genes related to hepatic lipid synthesis (Scd1, fas, PPARγ, CD36), genes related to lipolysis (Lipe, Pnpla2, Fabp5), genes related to lipid oxidation (Cpt2, Acox1, Ppargc1a, PPARα), genes related to bile acid transporters (shp, CYP7A1, CYP27A1, FgFr4), genes related to hepatic BAs biosynthesis (FXR), genes related to ileal bile acid transporters (Fgf15), genes related to liver inflammation (CD14, TLR2, TLR4, NLRC4, MCP-1, IL-10), genes related to intestinal permeability (Occludin, ZO-1, Muc5), and genes related to short-chain fatty acids (SLC5A8) were all adjusted with glycerol triphosphate dehydrogenase (GAPDH) as a housekeeping gene. Relative quantification was calculated using the −∆∆Ct method. The primer sequences for RT-PCR can be found in. Table S1

To extract total protein, mouse liver tissues were lysed in RIPA buffer with one protease and phosphatase inhibitor cocktail. The BCA technique was used to determine the total protein content in the supernatant. After denaturation for 5 minutes at 100°C, equivalent quantities of proteins were run on a 10% SDS-PAGE gel and transferred to a polyvinylidene fluoride (PVDF) membrane. The membrane was blocked for 2 hours at room temperature with 5% nonfat milk in 0.1% tris-buffered saline with tween-20 (TBST) before being incubated with a primary antibody overnight at 4°C. After thorough washing, the membranes were incubated for 2 hours at room temperature with the appropriate HRP-conjugated secondary antibody. Finally, an ECL reagent was used to develop the blots. The primary antibodies listed below were used: GAPDH [monoclonal antibody (2B8), YM3029, 1:5,000], PPARγ (polyclonal antibody, YM33072, 1:1,000), CYP7A1 (polyclonal antibody, NBP-04836, 1:500), and Fgf15 (polyclonal antibody, ab229630, 1:500). The quantitative image analysis was performed using Image J (2.0.0 version, National Institutes of Health, Bethesda, MD, USA).

Intestinal Occludin expression was evaluated using paraffin-embedded liver and ileum tissues. After deparaffinization, the slides were heated in an autoclave with sodium citrate for antigen repair, followed by 1% hydrogen peroxide to abolish 10 endogenous peroxidase activity, and blocked with 2% goat serum. Slides were then incubated with primary antibodies, including Occludin (Servicebio, GB111401↗, 1:500), at 4°C overnight. HRP-conjugated secondary antibodies (1:200) were incubated for 50 minutes at room temperature. After washing with PBS, the peroxidase substrate DAB (Dako, K5007) was used for color development.

The results are displayed as the mean ± standard deviation. The graphs depict the data distribution in the form of box and violin graphs. The Mann‒Whitney U test was used to examine significant differences in biochemical markers, F/B, and α-diversity of gut microbiota, across groups in GraphPad Prism 8.4 software. Significant differences in the microbiota structure among the three groups were evaluated by nonparametric analysis of similarity (ANOSIM, permutations = 9,999) using the vegan package in R (version 3.6.1). Permutational multivariate analysis of variance permutational multivariate analysis of variance (PERMANOVA; permutations = 9,999) was used to assess the structural difference between the two groups (vegan package in R). Statistical significance was defined as P < 0.05. Differential abundance of pathways was identified by using the Kruskal–Wallis test. Benjamini–Hochberg method was used to adjust P values generated by multiple comparisons from PERMANOVA.

The original sequences obtained in this investigation have been deposited in the NCBI sequence read archive (SRA) with the accession number PRJNA757473↗.