Er-Chen Decoction improves fatty liver disease linked to metabolism by changing gut bacteria and liver metabolism through the gut barrier

The raw herbal materials comprising ECD were purchased from Kang Mei Pharmaceutical Co. Ltd. (Guangdong, China), and the detailed composition is presented in Table 1. To prepare the decoction, the mixed crude herbs were soaked in purified water at a ratio of 1:10 (w/v) for 30 min. The mixture was boiled for 30 min and subsequently maintained at a gentle simmer (approximately 100 °C) for another 30 min. After filtration, the residues were subjected to a second extraction under identical conditions. The filtrates from both extractions were combined and concentrated to a final density of 2.639 g crude drug/mL (Deng et al., 2024). The extract was prepared weekly, stored at 4 °C, and reheated to ambient temperature prior to administration. To ensure quality consistency, the chemical profile of ECD was characterized using UPLC–MS/MS analysis.

The chemical profile of ECD was analyzed using an Agilent 1290 Infinity II UPLC system coupled with an Agilent 6546 Quadrupole Time-of-Flight (Q-TOF) mass spectrometer (Agilent Technologies, CA, United States). Separation was achieved using an Agilent InfinityLab Poroshell 120 EC-C18 column (2.1 × 150 mm, 2.7 μm) maintained at a constant temperature of 40 °C. The mobile phase consisted of 0.1% formic acid in water (A) and (B) acetonitrile. The gradient elution program was set as follows: 0–2 min, 5% B; 2–8 min, 5–50% B; 8–25 min, 50–95% B; 25–30 min, 95% B; 30–30.1 min, 95–5% B; and 30.1–35 min, 5% B for equilibration. The flow rate was maintained at 0.2 mL/min. Mass spectrometric detection was conducted utilizing a Dual Agilent Jet Stream (AJS) ESI source operating in Auto MS/MS mode. The key source parameters were as follows: gas temperature, 320 °C; drying gas flow, 8 L/min; nebulizer pressure, 35 psig; sheath gas temperature, 350 °C; sheath gas flow, 11 L/min; capillary voltage, 3,500 V; nozzle voltage, 1,000 V; fragmentor voltage, 175 V; and skimmer, 65 V. Data were acquired in the mass range of 100–1,100 m/z. Compound identification was performed using the Agilent MassHunter Workstation Software by matching the spectra against the Agilent Natural Standard NPs PCDL with a mass error tolerance of 20 ppm.

The study utilized male C57BL/6J mice (6–8 weeks old) sourced from Guangdong Medical Laboratory Animal Center (Guangzhou, China; License No. SCXK (Yue) 2022-0002). All animal experiments were approved by the Animal Ethics Committee of Guangzhou University of Chinese Medicine (Approval No.20240612003) and were conducted in strict accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals. The animals were housed in a specific-pathogen-free (SPF) facility at the Research Center of Basic Integrated Medicine, Guangzhou University of Chinese Medicine (License No. SYXK [Yue] 2023-0182), under controlled environmental conditions (22 ± 2 °C; 12-h light/dark cycle) with ad libitum access to food and water.

Following a one-week acclimatization period, the mice were randomly allocated into four groups (n = 9 per group) based on body weight to ensure uniformity in baseline conditions: Control, Model, Er-Chen Decoction (ECD), and semaglutide (SE). The Control group was fed a standard chow diet, whereas the Model, ECD, and SE groups were fed a high-fat diet (HFD, 60% energy from fat, 5.24 kcal/g; Cat. No. GD60, Guangdong Medical Laboratory Animal Center) formulated based on Research Diets D12492 to induce MASLD. HFD feeding lasted for 12 weeks to establish the MASLD model (Zhao et al., 2025). From weeks 13 to 20, the mice received specific treatments while maintaining their respective diets. ECD via intragastric gavage at a daily dose of 26.39 g crude drug/kg (calculated based on the weight of raw herbal materials). This dosage corresponds to 4 times the clinical human equivalent dose, calculated based on the body surface area (BSA) normalization method (human-to-mouse conversion factor of 9.1) (Deng et al., 2024). The SE group received a daily subcutaneous injection of semaglutide (Novo Nordisk A/S, Denmark) at a dose of 30 nmol/kg/day (Yue et al., 2022). The Control and Model groups were administered an equivalent volume (10 mL/kg) of saline intragastrically.

Weekly body weight was recorded. An oral glucose tolerance test (OGTT) was performed in the 19th week. At the end of the 20-week experimental period, mice were euthanized for serum and tissue collection. The terminal procedure was conducted under deep anesthesia induced by inhalation of 3–4% isoflurane. Upon reaching a surgical plane of anesthesia (loss of righting reflex and absence of response to toe pinch), blood was collected from the retro-orbital sinus using a capillary tube. Subsequently, euthanasia was completed by cervical dislocation, which was performed only by personnel who had demonstrated proficiency in the technique. Death was confirmed by the absence of a heartbeat, respiration, and corneal reflex. Following this, liver and colon tissues were rapidly dissected, weighed, and either snap-frozen in liquid nitrogen or fixed in 4% paraformaldehyde for subsequent analyses.

Fresh liver and colon tissues were fixed in 4% paraformaldehyde for 24 h, dehydrated, embedded in paraffin, and sectioned into 4-μm sections. The sections were stained using a Hematoxylin and Eosin (H&E) Staining Kit (Cat. No. C0105M, Beyotime, Shanghai, China) to assess general tissue morphology and inflammatory infiltration. For the evaluation of hepatic lipid accumulation, fresh liver tissues were embedded in OCT compound and cryosectioned at 8 μm thickness using a cryostat. The frozen sections were then stained with Oil Red O working solution (Cat. No. O0625, Sigma-Aldrich, MO, United States) for 20 min at room temperature in the dark, followed by differentiation in 60% isopropanol and counterstaining with hematoxylin. Representative images of each group were acquired using a light microscope (Olympus BX46F, Olympus, Tokyo, Japan). Scale bars (200 μm for colon tissues and 50 μm for liver tissues) were integrated into the images for dimensional reference.

At week 19, an oral glucose tolerance test (OGTT) was performed following a 6-h fast with ad libitum access to water. Mice were administered glucose (2.0 g/kg) via oral gavage, and blood glucose levels were measured in tail vein blood at 0, 30, 60, 90, and 120 min post-administration using an Accu-Chek Active meter (Roche Diagnostics, Germany).

At the experimental endpoint, blood samples were collected and centrifuged (3,500 rpm, 10 min, 4 °C) to obtain serum. Serum levels of alanine aminotransferase (ALT), aspartate aminotransferase (AST), total cholesterol (TC), and triglycerides (TG) were determined using a Mindray BS-460 Automatic Biochemical Analyzer (Shenzhen Mindray Bio-Medical Electronics Co., Ltd., Shenzhen, China) according to the manufacturer’s instructions. Hepatic TG concentrations were quantified using a commercial assay kit (Cat. No. A110-1-1, Nanjing Jiancheng Bioengineering Institute, Nanjing, China) following the standard protocol.

Fresh fecal samples were collected, snap-frozen in liquid nitrogen, and stored at −80 °C. Genomic DNA was extracted from approximately 0.2 g of fecal samples utilizing the FastPure Stool DNA Isolation Kit (Magnetic Bead) (MJYH, Shanghai, China), adhering strictly to the manufacturer’s protocol. DNA concentration and purity were assessed using a TBS-380 fluorometer and NanoDrop 2000 spectrophotometer, respectively, while DNA integrity was confirmed via 1% agarose gel electrophoresis. For library construction, genomic DNA was fragmented to an average size of approximately 350 bp using a Covaris M220 instrument (Gene Company Limited, China). Sequencing libraries were prepared using the NEXTFLEX® Rapid DNA-Seq Kit (Bioo Scientific, Austin, TX, United States). Subsequently, the libraries were sequenced on the Illumina NovaSeq X Plus platform (Illumina, San Diego, CA, United States) at Majorbio Bio-Pharm Technology Co., Ltd. (Shanghai, China) to generate paired-end reads.

The raw sequencing data were quality-filtered using fastp (Chen et al., 2018) to remove adapters and low-quality reads. To remove host contamination, clean reads were aligned to the Mus musculus genome (Vertebrates: mus_musculus_c57bl6nj) using BWA (Li and Durbin, 2009). The remaining high-quality reads were assembled into contigs utilizing MEGAHIT (Li et al., 2015). Contigs with a length of ≥ 300 base pairs were selected as the final assembly result. Subsequently, open reading frames (ORFs) were predicted from each assembled contig using Prodigal (Hyatt et al., 2010), and ORFs with a length ≥ 100 bp were retrieved. Gene abundance was calculated using SOAPaligner (Li et al., 2008). Taxonomic annotation was performed by aligning the non-redundant gene catalog against the NCBI Non-Redundant (NR) database using DIAMOND (Buchfink et al., 2015) with an e-value threshold of 1 × 10⁻⁵. The relative abundance at each taxonomic level was determined based on gene abundance. Differentially abundant bacterial taxa among the groups were determined using Analysis Effect Size (LEfSe) analysis, with a significance threshold set at an LDA score of > 3.0. All bioinformatics analyses were conducted on the Majorbio Cloud Platform1 (Ren et al., 2022).

Serum metabolites were extracted using a mixture of methanol and acetonitrile (1:1, v/v) spiked with internal standards. Following sonication, protein precipitation was carried out at −20 °C, followed by centrifugation. The resulting supernatants were evaporated to dryness and reconstituted in acetonitrile:water (1:1, v/v) for analysis. LC–MS/MS analysis was conducted using a Thermo Scientific Vanquish UHPLC system coupled to an Orbitrap Exploris 240 mass spectrometer. Chromatographic separation was performed on an ACQUITY UPLC HSS T3 column (100 mm × 2.1 mm, 1.8 μm; Waters, Milford, MA, United States) maintained at 40 °C. The mobile phases consisted of 95% water and 5% acetonitrile containing 0.1% formic acid (A) and 47.5% acetonitrile, 47.5% isopropanol, and 5% water containing 0.1% formic acid (B). The injection volume was set at 3 μL. Raw data processing, including peak alignment and identification against the HMDB and METLIN databases, was performed using Progenesis QI software (Waters). Differential metabolites were identified on the Majorbio Cloud Platform based on the criteria of Variable Importance in Projection (VIP) > 1 and p < 0.05.

Colon and liver tissues were accurately weighed and homogenized in ice-cold RIPA Lysis Buffer (Medium) (Cat. No. FD008, Fude Biological Technology, Hangzhou, China) at a ratio of 1:10 (w/v, 30 mg tissue in 300 μL buffer). The buffer was supplemented with protease inhibitor (Cat. No. FD1001) and phosphatase inhibitor (Cat. No. FD1002) cocktails at a 1:100 dilution. The lysates were centrifuged at 12,000 × g for 15 min at 4 °C to collect the supernatant. Protein concentration was determined using a BCA Protein Assay Kit. Equal amounts of protein (30 μg) were separated by 10% SDS-PAGE and electrotransferred onto 0.22 μm PVDF membranes (Millipore, MA, United States). The membranes were blocked with Fast Blocking Buffer (Cat. No. G2052, Servicebio, Wuhan, China) at room temperature for 15 min and subsequently incubated overnight at 4 °C with primary antibodies.

The primary antibodies used in this study included: anti-AHR (1:1000, Cat. No. 67785-1-lg, Proteintech, Wuhan, China), anti-Occludin (1:1000, Cat. No. GB111401-50, Servicebio, Wuhan, China), anti-TNF-α (1:5000, Cat. No. YM8306, ImmunoWay, TX, United States), anti-IL-1β (1:1000, Cat. No. 12242S, Cell Signaling Technology, MA, United States), anti-PPARγ (1:5000, Cat. No. YM8211, ImmunoWay), anti-CIDEA (1:1000, Cat. No. A7655, ABclonal, Wuhan, China), and anti-DNMT3B (1:1000, Cat. No. A11079, ABclonal). Anti-GAPDH (1:5000, Cat. No. AF7021, Affinity Biosciences, Jiangsu, China) and anti-β-actin (1:20000, Cat. No. 66009-1-Ig, Proteintech) were used as loading controls.

After three washing steps with TBST (10 min each), the membranes were incubated with HRP-conjugated anti-rabbit (Cat. No. 7074S) or anti-mouse (Cat. No. 7076S) secondary antibodies (1:2000, Cell Signaling Technology) for 1.5 h at room temperature. Following three additional washing steps with TBST, protein bands were visualized using an ECL chemiluminescence kit and imaged on a Tanon 5200ce Chemiluminescence Imaging System. Band intensity was quantified using ImageJ software (NIH, Bethesda, MD, United States) and normalized to GAPDH or β-actin levels.

For endotoxin analysis, serum LPS levels were measured using a Mouse Lipopolysaccharides ELISA Kit (Cat. No. JL20691, Shanghai Jianglai Biotechnology Co., Ltd., Shanghai, China) strictly according to the manufacturer’s protocol. Absorbance was measured at 450 nm using a microplate reader.

Paraffin-embedded colon tissues were sectioned at 4 μm thickness, deparaffinized in xylene, and rehydrated through a graded ethanol series. Antigen retrieval was performed by 95 °C heating in TRIS-EDTA Antigen Retrieval Solution (Cat. No. BL617A, Biosharp, Hefei, China) for 15 min to expose the specific binding sites. After cooling to room temperature and washing with PBS, the sections were blocked with 10% Goat serum for 1 h at room temperature to prevent non-specific binding. Subsequently, the sections were incubated overnight at 4 °C with the primary antibody against ZO-1 (1:200, Cat. No. YM8448, ImmunoWay, TX, United States).

Following three washes with PBS, the sections were incubated with Alexa Fluor 594-conjugated goat anti-rabbit IgG (1:1000, Cat. No. A11012, Thermo Fisher Scientific, MA, United States) for 1 h at room temperature in the dark. Finally, the sections were mounted and nuclei counterstained using Antifade Mounting Medium with DAPI (Cat. No. P0131, Beyotime, Shanghai, China). Fluorescence images were acquired using a Zeiss LSM800 confocal laser scanning microscope (Carl Zeiss, Oberkochen, Germany). The acquisition parameters were standardized across all samples to ensure comparable signal intensity.

Total RNA was isolated from liver tissues utilizing the SteadyPure Universal RNA Extraction Kit (Accurate Biology, Hunan, China), adhering strictly to the manufacturer’s protocol. RNA quality and integrity were determined by 1% agarose gel electrophoresis and the Agilent 5300 Fragment Analyzer (Agilent Technologies, CA, United States). Only high-quality RNA samples (total amount ≥ 0.5 μg, concentration ≥ 20 ng/μL, RQN > 4.5) were used for the library construction. Sequencing libraries were prepared utilizing the Illumina® Stranded mRNA Prep, Ligation Kit (Illumina, San Diego, CA, United States) in accordance with the manufacturer’s protocol. The libraries were then sequenced on the Illumina NovaSeq X Plus platform (Illumina, San Diego, CA, United States) to generate paired-end reads. Raw reads were filtered using fastp to remove adapters and low-quality reads (Chen et al., 2018). The clean reads were aligned to the Mus musculus reference genome (GRCm39) using HISAT2 (Kim et al., 2015). Gene expression levels were quantified using RSEM software (Li and Dewey, 2011), and the expression abundance was normalized to Transcripts Per Million (TPM). Differential expression analysis between groups was performed using DESeq2 (Love et al., 2014). Differentially expressed genes (DEGs) were identified using the criteria of a false discovery rate (FDR) less than 0.05 and an absolute log2 fold change (|log₂FC|) of at least 1.

Total RNA was isolated from liver and colon tissues utilizing the SteadyPure Universal RNA Extraction Kit (Accurate Biology, Hunan, China), adhering strictly to the manufacturer’s protocol. Briefly, approximately 20–30 mg of fresh tissue was homogenized in 600 μL of lysis buffer provided in the kit. RNA concentration and purity were measured using a NanoDrop One spectrophotometer (Thermo Fisher Scientific, WI, United States). Subsequently, 1 μg of total RNA was reverse-transcribed into cDNA using the 4× EZscript Reverse Transcription Mix II (with gDNA Remover) (Cat. No. EZB-RT2GQ, EZBioscience, Roseville, MN, United States). RT-qPCR was conducted on an Applied Biosystems QuantStudio 5 Real-Time PCR System using the 2× Color SYBR Green qPCR Master Mix (ROX2) (Cat. No. A0012-R2, EZBioscience). The cycling conditions were as follows: initial denaturation at 95 °C for 5 min, followed by 40 cycles of 95 °C for 10 s and 60 °C for 30 s. Melting curve analysis was performed to verify the specificity of PCR products. The primers used for the target genes (Dnmt3b, Cidec, Cidea, Pparg, Tnf, Il1b, Cyp1a1, Cyp1b1) and internal controls (Actb and 18S) are listed in Supplementary Table S1. The comparative 2^ ^-ΔΔCt was utilized to analyze relative mRNA abundance, with normalization to the geometric mean of the reference genes.

GraphPad Prism 10.0 (GraphPad Software, San Diego, CA, United States) was used for both statistical evaluation and graphical representation. All data are expressed as the mean ± standard deviation (SD). Data normality was verified using the Shapiro–Wilk test, and differences among groups were analyzed using One-way Analysis of Variance (ANOVA) followed by Dunnett’s multiple comparisons test to compare the treatment groups against the Model group. Bioinformatics analyses were performed utilizing the Majorbio Cloud Platform (see Footnote 1). Beta-diversity and metabolic profile differences were assessed using ANOSIM, PCA, and PLS-DA. Differentially abundant taxa were identified by LEfSe (LDA > 3.0), differential metabolites were screened based on VIP > 1 and p < 0.05, and differentially expressed genes were identified using DESeq2 (adjusted p < 0.05). For the comparison of specific omics features (e.g., bacterial abundance and metabolite intensity), the non-parametric Kruskal-Wallis H test was used, followed by Tukey–Kramer post-hoc tests with False Discovery Rate (FDR) correction. Correlations were evaluated using Spearman’s rank analysis. All results were considered statistically significant at p < 0.05.

To elucidate the chemical components of ECD, the chemical fingerprint of the ECD extract was comprehensively characterized using UPLC-MS/MS. The total ion chromatograms (TIC) revealed a complex chemical profile (Figure 1). By matching the accurate mass and MS/MS fragmentation patterns against the Agilent Natural Standard NPs PCDL database, we screened and tentatively identified a series of constituents. The complete list of compounds identified in this study is provided in Supplementary Table S2. Based on this profile, nine representative compounds with potential biological significance were selected for further investigation (Table 2). The retention times of these key constituents are indicated by arrows in the TIC (Figure 1), and their corresponding extracted ion chromatograms (EIC) and mass spectra are presented in Supplementary Figure S1. These compounds included L-tryptophan, hesperetin 7-O-glucoside, naringin, glycyrrhizic acid, 6-gingerol, narirutin, aromadendrin, liquiritin, and limonin. Notably, the presence of L-tryptophan, a biosynthetic precursor of indole derivatives, together with anti-inflammatory agents such as naringin and glycyrrhizic acid, suggests that ECD may provide potential substrates for the tryptophan-indole metabolic pathway, while collectively contributing to the regulation of hepatic inflammatory processes.

To assess the therapeutic efficacy of ECD, we developed a MASLD mouse model utilizing an HFD. The experimental design was shown in Figure 2A. The successful induction of obesity by HFD was confirmed by the significant increase in body weight observed in the Model group compared to the Control group (Figure 2B). However, ECD treatment effectively suppressed body weight gain throughout the experiment and significantly reduced the final body weight at week 20 (Figure 2C). Systemic metabolic disorders were also assessed. The Model group presented with severe dyslipidemia and liver injury, characterized by markedly elevated serum TG, TC, ALT, and AST levels. ECD intervention significantly reversed these pathological changes (Figures 2D–G). Furthermore, glucose metabolism was assessed through the administration of an oral glucose tolerance test (OGTT). The Model group showed impaired glucose tolerance, whereas ECD treatment significantly lowered blood glucose levels at various time points (Figure 2H) and reduced the area under the curve (AUC) (Figure 2I), indicating an improvement in systemic insulin sensitivity.

We further examined the protective effects of ECD on liver pathologies. Gross examination showed that the livers from the Model group were enlarged and pale yellow, typical of fatty liver, whereas ECD treatment restored a reddish-brown appearance similar to that of the Control group (Figure 3A). Compared to the Model group, which exhibited a substantial increase in liver mass at the 20-week endpoint, ECD administration resulted in a notable reduction in liver weight (Figure 3B). Histologically, H&E staining revealed severe pathological alterations in the Model group compared to the Control group. The liver tissue exhibited diffuse macrovesicular steatosis (characterized by large cytoplasmic vacuoles displacing the nucleus, indicated by black arrows) and widespread hepatocyte ballooning (swollen cells with rarefied cytoplasm). Additionally, the normal radial arrangement of hepatic cords was disordered, accompanied by scattered inflammatory cell infiltration. Consistent with these findings, Oil Red O staining visualized substantial intracellular lipid droplet accumulation (stained brilliant red) in the Model group. In contrast, ECD treatment markedly attenuated these pathological changes, restoring the hepatic architecture and significantly decreasing both the quantity and size of lipid droplets (Figure 3C). Consistent with the histological findings, quantitative analysis confirmed that the hepatic TG content was significantly reduced in the ECD group compared to that in the Model group (Figure 3D).

To investigate the regulatory effect of ECD on the gut microbiome, we performed a metagenomic sequencing analysis. Venn diagram analysis revealed the extent of species intersection among the groups, reflecting specific alterations in the microbial repertoire (Figure 4A). We first examined the α-diversity to evaluate microbial richness and evenness. HFD feeding significantly decreased the Chao1 and Simpson indices compared with the Control group. However, ECD treatment ameliorated this trend, showing a tendency to improve microbial diversity compared to the Model group (Figures 4B,C). We then assessed the overall microbial community structure using β-diversity analyses. As shown in the PCoA plot (Figure 4D), a distinct separation was observed between the Model and Control groups (R=0.272, p=0.001), confirming the profound impact of HFD on the gut microbiome. Importantly, the overall microbial composition of ECD-treated mice was clearly separated from that of the Model group, showing a trend of recovery toward the Control phenotype. Further examination at the phylum (Figure 4E), genus (Figure 4F), and species (Figure 4G) levels confirmed that ECD modulated the relative abundance of dominant taxa, effectively correcting the HFD-induced deviations. Collectively, these findings indicate that ECD effectively ameliorates HFD-induced gut microbiota dysbiosis.

To identify specific bacterial taxa driving the structural changes, we performed LEfSe analysis (Figure 5A). As shown in the LDA score plot (Figure 5B), specific biomarkers were identified for each experimental group. SE treatment significantly increased the abundance of Akkermansia (Figure 5F). ECD treatment resulted in a distinct modulation pattern. In contrast to the SE, the ECD did not significantly alter the Akkermansia levels. In contrast, ECD uniquely and significantly enriched the abundance of Bifidobacterium (Figure 5C) and Lactobacillus (Figure 5D) in the gut. Additionally, both ECD and SE treatments exhibited a trend of suppressed HFD-induced opportunistic pathogens, such as Acetatifactor (Figure 5E). These results indicate that while both ECD and SE improve MASLD, ECD operates through a distinct microbiome-modulating mechanism characterized by specific enrichment of Lactobacillus and Bifidobacterium. Given that these genera are known to participate in tryptophan metabolism, their enrichment suggests that ECD may exert its hepatoprotective effects by modulating downstream tryptophan metabolites.

To systematically evaluate the metabolic impact of ECD, we performed an untargeted serum metabolomics analysis. The distribution of common and distinct metabolites among the four groups is shown in the Venn diagram (Figure 6A). Principal Component Analysis (PCA) revealed a clear separation in the metabolic profiles between the Control and Model groups, indicating significant metabolic dysregulation in MASLD. Notably, the metabolic profile of the ECD group shifted distinctly away from the Model group and clustered toward the Control group (R = 0.7211, p = 0.001), indicating a global reversal of HFD-induced metabolic deviations (Figure 6B). A detailed list of these significantly different metabolites is provided in Supplementary Table S3. Based on our metagenomic findings that ECD specifically enriched tryptophan-metabolizing bacteria (Lactobacillus and Bifidobacterium), we focused our analysis on tryptophan-derived metabolites. Heatmap and VIP analysis revealed that the levels of indole derivatives were generally suppressed in the Model group. However, ECD treatment uniquely and significantly upregulated a cluster of key metabolites in this pathway, including 3-methyloxindole, indole-3-acetylglycine, indoleacrylic acid, and indole-3-acetic acid (Figure 6C). In contrast, the SE group did not exhibit similar activation.

To further verify whether this metabolic shift was driven by the altered gut microbiota, we performed a Spearman correlation analysis (Figure 6D). Notably, the levels of key indole derivatives were positively correlated with the abundance of Lactobacillus and Bifidobacterium (p < 0.05). In contrast, the opportunistic pathogen Acetatifactor exhibited an inverse relationship with these metabolites. These statistical associations support the notion that ECD-induced gut microbiota remodeling, particularly the enrichment of indole-producing bacteria, is a key factor in the activation of the tryptophan-indole metabolic axis. Furthermore, we evaluated the associations between the altered gut microbiota/metabolites and host lipid phenotypes. As shown in Supplementary Figure S2, key indole derivatives exhibited negative correlations with serum TG and TC levels, supporting a potential link between these metabolic changes and lipid regulation.

Given that the indole derivatives, such as indole-3-acetic acid, identified above function as natural ligands for the aryl hydrocarbon receptor (AHR), we examined whether their systemic accumulation leads to the activation of the intestinal AHR pathway.

First, we evaluated the structural integrity of the colonic barrier. H&E staining showed that the colonic mucosa in the Model group exhibited disturbed crypt architecture and reduced mucosal thickness compared to the Control group. Notably, ECD treatment improved these morphological defects, preserving the overall mucosal structure (Figure 7A).

To further visualize the tight junction assembly, we performed immunofluorescence staining for ZO-1. As shown in Figure 7B, the Control group displayed a continuous and distinct linear signal along the epithelial border. In contrast, the Model group showed a fragmented and discontinuous distribution of ZO-1. Importantly, ECD treatment restored the continuity and intensity of the ZO-1 signal. Consistent with these morphological findings, Western blot analysis confirmed that the protein expression of the tight junction protein Occludin was significantly upregulated in the ECD group compared to the Model group (Figures 7C,E).

To determine the molecular mechanism underlying this barrier restoration, we examined the AHR signaling pathway. Western blot analysis demonstrated that ECD treatment markedly increased AHR protein expression in the colon relative to the Model group (Figures 7C,D). Furthermore, to confirm the transcriptional activation of AHR, we quantified the mRNA levels of its canonical target genes. RT-qPCR results showed a significant induction of Cyp1b1 (p < 0.05) and an increasing trend of Cyp1a1 in the ECD group (Figures 7F,G). These data provide direct evidence that the elevated indoles effectively triggered AHR signaling.

Functionally, this reinforcement of the intestinal barrier effectively blocked the translocation of gut-derived endotoxins, as evidenced by the significant reduction in serum LPS levels in the ECD group (Figure 7H). These findings suggest that ECD remodels the gut microbiota to increase indole production, which subsequently activates AHR signaling to restore barrier integrity and mitigate metabolic endotoxemia, thereby reducing the influx of LPS into the liver.

Following the observation that ECD restored the intestinal barrier and reduced serum LPS levels, we investigated its downstream effects on the liver. To comprehensively explore the molecular mechanisms, we performed RNA-seq on the liver tissues. PCA analysis revealed a distinct separation among the Control, Model, and ECD groups, indicating that ECD treatment effectively shifted the hepatic transcriptomic profile away from the disease state (Figure 8A). We then examined the specific changes in gene expression that drive these profiles. As illustrated in the volcano plot comparing the Model and Control groups (Figure 8B), HFD feeding induced extensive transcriptomic alterations. Notably, the lipid droplet fusion gene, Cidec, was significantly upregulated. However, ECD treatment successfully reversed this pathological pattern. Comparison between the ECD and Model groups showed that Cidea and its homolog Cidec were significantly downregulated, indicating targeted inhibition of lipid droplet expansion (Figure 8C). To elucidate the biological functions of these changes, we performed a Reactome pathway enrichment analysis. Pathway enrichment analysis highlighted that the genes modulated by ECD were predominantly involved in metabolic processes, with a specific emphasis on lipid metabolism and biological oxidations (Figure 8D).

Furthermore, we examined the expression patterns of key genes implicated in LPS-induced inflammatory responses and lipid metabolic processes, identifying them as integral components of gut-liver crosstalk. As shown in the heatmap (Figure 8E), HFD feeding induced a broad upregulation of genes involved in LPS sensing and signaling, including Tlr1, Tlr2, Tlr4, Myd88, and Tollip expression. This was accompanied by elevated expression of downstream inflammatory mediators such as Rela, Nfkb1, Tnf, Il1b, and Il6, confirming the hepatic response to gut-derived endotoxemia. Consistent with the blockade of LPS translocation observed in Figure 7, ECD treatment systematically reversed the pathological inflammatory profile. Concurrently, genes critical for lipid droplet formation (Cidec, Cidea, Plin1, Plin2, and Plin4) and fatty acid uptake/synthesis (Pparg, Cd36, and Fabp4) were significantly downregulated. Most importantly, ECD uniquely restored the expression of Dnmt3b, a key epigenetic regulator suppressed by LPS.

To rigorously validate these transcriptomic findings and confirm the functional execution of these gene expression changes, we performed both RT-qPCR and Western blot analyses. At the transcriptional level, RT-qPCR results showed high consistency with the RNA-seq data: the mRNA levels of Dnmt3b were markedly upregulated, whereas Cidec, Cidea, Pparg, Tnf, and Il1b were significantly downregulated in the ECD group relative to the Model group (Figures 9A–F).

Crucially, to verify these changes at the translational level, we assessed the protein abundance of key nodes in the proposed “Gut-Liver” axis. As shown in Figures 9G–L, Western blot analysis confirmed that ECD treatment significantly restored the protein expression of DNMT3B, a critical maintenance DNA methyltransferase. This restoration was accompanied by the significant suppression of the lipid droplet fusion protein CIDEA and the key lipogenic transcription factor PPARγ. Furthermore, consistent with the alleviated hepatic inflammation, the protein levels of pro-inflammatory cytokines TNF-α were reduced, while IL-1β exhibited a consistent downward trend. Collectively, these multi-level validations demonstrate that ECD ameliorates MASLD by attenuating LPS-driven hepatic inflammatory and metabolic responses, potentially mediated by the DNMT3B-dependent epigenetic remodeling of the hepatic landscape.

The raw sequencing data generated in this study have been deposited in the NCBI Sequence Read Archive (SRA). The fecal metagenomic data are available under BioProject accession number PRJNA1380942, and the liver transcriptomic data are available under BioProject accession number PRJNA1380962.