We procured male 8 weeks BALB/c mice (body weight: 18–22 g,) from Hunan SJA Laboratory Animal Co., Ltd. (LL20240312001). All the animals were raised in Specific Pathogen Free level barrier facilities at the Animal Experiment Center of Hunan University of Traditional Chinese Medicine (located in Changsha, China) which maintain a ventilated living environment, suitable humidity and temperature, sufficient food and drinking water. All the experimental procedures were approved by the Animal Ethics Committee.

We randomly divided 30 mice into the following five groups: control, model, high-dose FAE, medium-dose FAE, and a low-dose FAE groups, with six mice in each group. Except for the blank group, all other groups were injected with 0.2 mL of 20% CCl₄ (Fuchen Chemical Reagent Co., Ltd., Tian-Jing, China) (CCl₄ and olive oil in a 1:4 ratio) twice a week for a total of 4 weeks. The blank group was administered with the same amount of ordinary olive oil.

FAE (0.2 mL; Shanghai Aladdin Chemical Co., Ltd., Shanghai, China) was administered to mice through gavage (at concentrations ranging from 0.125 to 1.25 g/mL) once a day for 4 weeks. According to the Chinese Pharmacopoeia, the clinically used dose of FAE is 3–9 grams(g). According to the Chinese Pharmacopoeia, the clinically used dose of FAE is 3–9 grams (17). Therefore, a pre-test was conducted with reference to this dose range, and it was finally decided to use 0.125 g/mL as the low dose group, 0.5 g/mL as the medium dose group, and 1.25 g/mL as the high dose group. The mice in both the blank and model groups were administered normal saline for 4 weeks following the same administration regimen as for the FAE groups. The drug intervention period for all groups commenced from the first week of modeling.

At the end of the experiment, all animals underwent fasting for 24 h. Blood, liver, and intestinal contents were then sampled. Blood (0.5 mL) was collected from the orbital vein. The serum was obtained through centrifugation; the remaining serum and intestinal contents were stored in a refrigerator at −80°C for further testing. Liver specimens were weighed and stored in paraformaldehyde.

The mouse liver was removed, rinsed with phosphate-buffered saline, and the water was eliminated. The wet weight was measured as follows: Liver index = liver weight (g)/mouse body mass (g) × 100%.

Liver tissues were fixed in 4% paraformaldehyde and embedded in paraffin. Liver tissue was collected from 3 μm-thick sections. Hematoxylin and eosin (H&E) staining (Wuhan Weier Biotechnology Co., Ltd., Wuhan, China) was performed for pathological evaluation. Masson’s trichrome staining was performed to assess collagen content.

Blood was allowed to stand at room temperature for 2 h and then centrifuged at 3,000 rpm for 10 min to obtain the serum. The serum levels of alanine aminotransferase (ALT), aspartate aminotransferase (AST), triglyceride (TG), and total bilirubin (T-Bil) were determined using kits from Shenzhen Mindray Biomedical Electronics Co., Ltd., Shen-Zhen, China, following the manufacturer’s instructions.

Paraffin-embedded liver tissues were cut into 3 cm-thick sections. After deparaffinization, antigen recovery was performed, followed by an overnight incubation at 4°C with rabbit anti-α-SMA (1:600), anti-collagen 1 (1:600), anti-TLR4 (1:500), anti-NF-κ B (1:500), anti-JUN (1:1000), anti-HSP90A11 (1:500), anti-PTGS2 (1:500), and anti-CASP3 (1:600) antibodies obtained from Anjie High Tech Wuhan Weier Biotechnology Co., Ltd. Hematoxylin was used for nuclear counterstaining, and a light microscope was used to capture TIFF images. They were quantitatively analyzed by Case Viewer (2.4_RTM_v2.4.0.)

The Ultimate 3,000 ultra high-performance liquid chromatograph and Thermo Q-Exactive Orbitrap mass spectrometer were used for ultrapure LC–MS/MS analysis. Chromatographic separations were carried out on a Waters ACQUITY UPLC HSS T3 column (100 mm × 2.1 mm × 1.8 μ m). Subsequently, all raw data were analyzed using Progenesis QI and SIMCA-P14.0 software. Partial least-squares discriminant analysis (PLS-DA) and orthogonal partial least-squares discriminant analysis (OPLS-DA) were performed to identify overall metabolic differences among the three groups, with variable importance in the projection (VIP) used to identify characteristic metabolites. Distinct metabolites were analyzed based on VIP > 1.0 and p < 0.05. Additional information is provided in Supplementary materials.

16S rDNA gene sequencing of fecal samples was conducted. The samples were subjected to polymerase chain reaction (PCR) amplification, followed by amplicon purification and paired-end sequencing to evaluate the raw data.

Differential metabolites involved in important metabolic pathways in the control, model, and FAE groups were identified using correlation analysis; 45 most abundant liver species across all the three groups were identified. Spearman correlation analysis and heatmaps were used to assess the correlation between the gut microbiota and metabolite content.

FAE targets were predicted using TCMSP,¹ and disease targets were obtained from the Gene Cards database. Using “Hepatic fibrosis” as the search term and limiting the species to “Homo sapiens,” predict the relevant targets of HF in the Gene Card database, and obtain targets with AS related scores greater than 1.0 as the relevant targets of HF. On the Venny 2.1.0 drawing tool platform² take the intersection of precognitive sub targets and HF related targets, obtain the intersection targets of precognitive active ingredients and HF, and draw a Venny plot. A protein–protein interaction (PPI) network was assembled using the STRING database version,³ with the organism set to Homo sapiens; only those interactions with a minimum score of >0.4 were considered significant. Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analyses were performed to further explore the biological functions of the targets and related pathways. To validate the binding of core components in FAE to the predicted core targets, the 3D molecular structures of the compounds were retrieved from the PubChem database, and the structural files of the target proteins were obtained from the RCSB Protein Data Bank.⁴ Molecular docking calculations were performed using SYBYL-X 2.0.

All data are presented as the mean ± SD of values from at least three independent experiments. Statistical analysis and comparisons between groups were performed using Prism 9.0 software (GraphPad Software, United States) and a one-way ANOVA. Statistical significance was set at p < 0.05.

We examined the in vivo effects of FAE in male mice with CCl₄-induced HF (Figure 1A). H&E staining results showed that the hepatic lobule structure of the mice in the blank group was complete, the hepatocytes were neatly arranged, the central vein was radial, and no infiltration of inflammatory cells into the portal area had occurred. Inflammatory cell infiltration was observed in the portal areas in the model group. A small number of inflammatory cells infiltrated the portal area of the high-dose FAE group, the fibrous septum was reduced, and the pathological changes were less severe than those in the model group. Masson staining showed that the blank group had intact liver tissue, no collagen deposition, and hyperfibrous tissue; in the model group, a large amount of collagen deposition occurred in the central vein and portal area, the hyperfibrous tissue formed fibrous cables connecting the adjacent collecting areas to form false lobules, and the degree of collagen deposition in the FAE high-dose group was significantly reduced. In summary, FAE treatment reduced HF (Figure 1C). These results were based on the liver index (Figure 1B) and serum AST, ALT, TG, and T-Bil levels (Figure 1D).

Given that α-SMA and collagen 1 expression by activated type HSC specifically in the liver can play a crucial role in HF progression. NF-κ B and TLR4 can upregulate the transcription and translation of pro-inflammatory genes and further promoting HF (18), inhibiting their expression could potentially mitigate fibrosis. Immunohistochemistry experiments revealed that FAE treatment suppressed the expression of α-SMA, collagen 1, NF-κ B and TLR4 (Figure 1E).

The principal component analysis (PCA) results highlighted a clear separation between the model and blank groups, indicating distinct differences between the metabolites of mice with HF and those of the control group (Figures 2A,B). PLS-DA further revealed significant variations in metabolites between the model and blank groups, with notable disparities in both the positive- and negative-ion PLS-DA models (R2 = 0.996, Q2 = 0.934; R2 = 0.995, Q2 = 0.933, respectively) (Figures 2C,D). Based on the importance of VIP values, we used multivariate analysis along with the p-values and multiple differences for screening. Metabolites used for differential expression (Figure 2E). p-value <0.05, VIP ≥ 1, and fold change ≤0.833 or ≥ 1.2 (Tables 1, 2) were considered as the screening criteria. We integrated differential metabolites identified from mouse liver tissue, used pathway analysis modules to enrich the metabolic pathways involved in the aforementioned differential metabolites, and finally determined the metabolic pathway with the highest correlation with FAE intervention in HF based on impact >0.1 or p < 0.05 (Figures 2F,G). The most influential pathways were starch and sucrose metabolism, fructose and mannose metabolism, galactose metabolism, amino and nucleotide sugar metabolism, lysosomes, and taste transduction (Table 3).

IlluminaPE250 sequencing of 16S rDNA genes from 24 fecal samples and subsequent clustering into operational taxonomic units (OTUs) identified 1903 OTUs. The sequencing coverage indices (> 0.99) indicated reliable data quality for all the three sample groups (Figure 3A). The dilution curves demonstrated that the sequencing data for the gut flora across the three sample groups were reasonable (Figure 3B).

Alpha diversity analysis of the gut microbiota revealed that the Chao1 index and observed species index were positively correlated with species richness, indicating increased diversity post-CCl₄ induction but decreased diversity following FAE intervention (Figures 3B,C). The results of PLS-DA (Figure 3D) displayed a distinct separation among the three sample groups, emphasizing the significant differences in gut microbiota composition.

At the phylum level (Figure 3E), distinct dominant species, including Firmicutes, Bacteroidetes, and Verrucomicrobiota, were identified in the intestinal colonies of mice within each group. The relative abundance of Firmicutes and Bacteroidetes was the highest, with variations among the groups. Cyanobacteria, Proteobacteria, and Verrucomicrobiota were significantly enriched in the model group compared to those in the blank group. Moreover, the relative abundance of these phyla in the FAE group indicated their potential therapeutic effects on the modulation of gut microbiota composition.

At the genus level (Figure 3F), the relatively high abundance in each group of mice belonged to Muribaculaceae (Muri), Lactobacillus, and Alistipes (Alis), with Muri and Lactobacillus exhibiting the highest proportions. The relative abundance of Muri in each group was as follows: blank group: 29.96%; model group: 42.27%; FAE group: 42.23%. The relative abundance of Lactobacillus spp. was as follows: blank group: 17.34%; model group: 17.44%; FAE group: 15.74%. Akkermansia, Colidextribacter, and Eubacterium were significantly enriched in the model group compared to those in the control group (p < 0.05). Additionally, the relative abundance of Akkermansia, Colidextribacter, and Eubacterium was higher in the FAE group than in the model group (p < 0.05).

Upon setting the linear discriminant analysis threshold at 3, the linear discriminant analysis effect size analysis (Figure 3G) revealed significant differences in species abundance in the gut microbiome of the mice across all groups. In the blank group, the notable differential flora included Rikenellaceae, Ruminococcaceae, and Saccharimonadaceae, whereas the model group showed Intestinimonas as the differential flora. The FAE group was enriched in Verrucomicrobiae and Akkermansiaceae. Notably, the model group displayed an increase in the relative abundance of Trichosporonaceae (Lachnospiraceae) and Verrucomicrobiota and a decrease in that of Cyanobacteria and Proteobacteria when compared with the blank group. The FAE group showed increased relative abundance of Verrucomicrobiota and Akkermansia. They exert an inhibitory effect on the progression of HF in mice.

To further explore the relationship between the intestinal flora and metabolites, Spearman correlation analysis was utilized to correlate the mouse liver differential metabolites with the flora at the portal and genus levels and to generate correlation heatmaps (Figure 4). Positive and negative correlations are depicted in red and blue, respectively. Color intensity indicates the strength of the correlation coefficient.

The differential metabolites Notoperol and Sinefungin were downregulated in both the blank and model groups, with stronger downregulation observed in the model group. They exhibited negative correlations with the differential genera Verrucomicrobiota, Akkermansia, and Muribaculaceae. Additionally, Sinefungin negatively correlated with the differential genus Lactobacillus and positively correlated with the genus Firmicutes.

Similarly, the differential metabolites D-turanose, Maltotetraose, and Stachyose were downregulated in both the model and FAE groups. They showed negative correlations with the genera Verrucomicrobiota and Akkermansia but positive correlations with the genera Firmicutes, Lactobacillus, and Colidextribacter. These findings provide insights into the intricate relationship between the intestinal flora and metabolites.

NP analysis was conducted to explore the anti-HF mechanism of the FAE. Initially, 83 drug targets and 4,738 HF-related targets were identified using the Uni-Prot prediction database. The intersection of these sets was visualized using a Venn diagram (Figure 5A). Utilizing the network analyzer tool in Cytoscape 3.7.2 software, a network topology analysis was performed on the interaction network diagram of “traditional Chinese medicine-active ingredient-intersecting target.” The mean values of the meso-, proximity centrality, and degree of network topology parameters were 0.057, 0.39, and 6.23, respectively, and the active ingredients with median, proximity centrality, and degree values exceeding the average were identified as FAE core active ingredients (Table 4).

A protein–protein interaction (PPI) network, consisting of 52 nodes representing functional proteins and 189 edges representing the interactions between functional and other proteins, was constructed using Cytoscape (Figure 5B). Within the network, 11 key targets, including JUN, CASP3, NOS3, PTGS2, HSP90AA1, NR3C1, PGR, NOS2, HMOX1, PRKACA, and SLC6A4, which were determined based on node degree, were identified (Table 5). These key targets may play crucial roles as target genes in the treatment of HF. Furthermore, to analyze the mechanism of action of FAE in HF treatment, 55 intersecting targets were subjected to GO function enrichment analysis in the DAVID database, with a significance threshold set at p < 0.05. A total of 207 biological process (BP) pathways, 36 cellular component (CC) pathways, and 55 molecular functions (MF) were enriched in the intersecting targets. The top ten pathways for BP, CC, and MF were selected and presented in a drift diagram (Figure 5C).

The 55 intersecting targets were also analyzed for KEGG pathway enrichment in the DAVID database, resulting in a total of 85 enriched KEGG pathways. The top 15 KEGG pathways with the smallest p-values were selected to create a KEGG pathway bubble diagram (Figure 5D). The targets involved in anti-HF prediction were associated with various signaling pathways, including pathways involved in cancer, lipid and atherosclerosis, serotonergic synapses, chemical carcinogenesis-receptor activation, AGE-RAGE signaling pathway in diabetic complications, and the estrogen signaling pathway.

Overall, these findings shed light on potential mechanisms underlying HF treatment and highlight the significance of key target genes in HF therapy.

Six key targets, JUN, CASP3, NOS3, PTGS2, HSP90AA1, and NR3C1, were selected for molecular docking with β-sitosterol, stigmasterol, ferulic acid, carotenoid, anthocyanin, and 5-hydroxymethylfurfural, which are the core active components in the treatment of HF. Among these key targets, the highest binding activities were observed for HSP90AA1, PTGS2, NOS3, HSP90AA1, JUN, NOS3. The core active components HSP90AA1, PTGS2 and NOS3 exhibited favorable binding affinities for the aforementioned six targets (Figure 5E). JUN, CASP3, NOS3, PTGS2, HSP90AA1, and NR3C1 have good binding activity with the core components of treating HF, suggesting that these may be the core targets for intervening in HF. Through immunohistochemical analysis of the levels of JUN, HSP90AA, PTGS2, and CASP3, it was found that compared to the blank group, the JUN, PTGS2, and HSP90AA levels in the model group were significantly increased, while the FAE intervention group showed a significant decrease, and the CASP3 content was reduced in the model group and significantly increased in the FAE intervention group (Figure 6).