The drugs of QJ were purchased from Sichuan New Green Pharmaceutical Technology Development Co., Ltd. The preparation method was described in detail in previous study (Chen, et al., 2022) and is briefly described as follows: according to the national pharmacopoeia standard, the drugs were soaked, and 1.5 L of pure water was added and the mixture brought to a boil. The drug solution was then filtered, and the above steps were repeated. Finally, the two drug solutions were mixed, concentrated, and stored in the refrigerator for further use. A patient needs 156 g/d crude drugs per 70 kg body weight. The rat needs 14 g/kg/d crude drugs following the method from Experimental Course of Pharmacology of Traditional Chinese Medicine (Xinguo, 2017). The high dose (QJ) is twice the medium dose. In this research, a high dose of QJ (28 g/kg) was chosen as the optimal dose in previous study (Chen, et al., 2022) with no significant side effects.

Twenty-four male Sprague-Dawley rats (RRID: MGI: 5651135) at 6–8 weeks, weighing approximately 160–180 g, were purchased from Beijing SPF Biotechnology Co., Ltd. with License No. SCXK (Beijing) 2019-0008. Rats were maintained in a room on a 12-h light/dark cycle with 50%–60% humidity and a temperature of 20°C–24°C. Free access to both water and food was provided. Animal experiments were approved by the Ethics Committee for Animal Experiments of Chengdu University of Traditional Chinese Medicine (No. 2021-66) and performed in accordance with the National Institutes of Health guidelines. The rats were randomly divided into three groups (n = 8 per group): the control (C) group—an equal amount of olive oil solution and saline during modeling and gavage; the model (M) group—CCl₄ olive oil solution (40%, 3 mL/kg, s.c.) twice a week for 8 weeks, then once per week for 6 weeks and an equal amount of saline daily for the same 6 weeks; the QJ group—CCl₄ olive oil solution twice a week for 8 weeks, then once per week for 6 weeks and QJ daily for the same 6 weeks. Rats were anesthetized with 3% sodium pentobarbital (50 mg/kg, i.p.). The contents of the large intestine were collected under sterile conditions in a clean bench, and then samples of abdominal aortic blood, liver, and ileum were collected. Rats were finally euthanized by an overdose of sodium pentobarbital.

Blood samples were centrifuged at 3500 rpm for 10 min. Serum was collected and analyzed for rat aspartate aminotransferase (AST), alanine aminotransferase (ALT), and alkaline phosphatase (ALP) using a fully automated blood biochemistry analyzer (Mindray BS-240 VET, China).

The liver and ileum were fixed with 4% paraformaldehyde for 24 h and then trimmed, dehydrated, embedded, and sectioned (4 μm). Hematoxylin-eosin (HE) staining was performed on both liver and ileum tissues, whereas only liver tissues underwent Masson staining. Stained sections were transformed into images through the use of a pathology scanner (HS6, Sunny Optical Technology, China). The study quantitatively measured the positive areas of Masson staining utilizing Image-Pro Plus 6.0 (RRID:SCR 007369). Each sample was measured by randomly selecting four fields of view and calculating the mean integrated optical density (IOD) values.

Immunohistochemistry (IHC) was performed according to Qu et al. (2023). Paraffin-embedded tissue sections were deparaffinized with xylene and rehydrated in an ethanol gradient. Sodium citrate solution was used for antigen retrieval in microwave oven, and 3% H₂O₂ solution was utilized for endogenous peroxidase blocking. Nonspecific antigens were blocked with goat serum. Afterwards, the slices were incubated overnight at 4°C with Collagen I (1:500, Abcam Cat# ab270993, RRID: AB 2927551), α-SMA (1:100, Cell Signaling Technology Cat# 19245, RRID: AB 2734735), Claudin-1 (1:150, Thermo Fisher Scientific Cat# 51-9000, RRID:AB 2533916), Occludin (1:150, Thermo Fisher Scientific Cat# 40-4700, RRID:AB 2533468) and ZO-1 (1:150, Thermo Fisher Scientific Cat# 61-7300, RRID:AB 2533938).Secondary antibodies were incubated for 1 h, followed by DAB chromogenization and hematoxylin staining. The sections were subsequently dehydrated and fixed. Semi-quantitative analysis of the images was performed utilizing Image-Pro Plus, and mean IOD values were calculated by randomly selecting four fields of view for each section.

Total RNA was isolated from liver and ileum tissues via FastPure Tissue Total RNA Isolation Kit (RC112, Vazyme, China), and then cDNA was synthesized by reverse transcription using HiScript III All-in-one RT SuperMix (R333, Vazyme, China). Finally, qPCR measurements were accomplished with a fluorescence quantitative PCR instrument (qTOWER3G, Jena, Germany) using Taq Pro Universal SYBR qPCR Master Mix (Q712, Vazyme, China), and relative gene expression was calculated by 2^−ΔΔCT method. The gene primer sequences are detailed in Table 1.

TGuide S96 Magnetic Soil/Stool DNA Kit (Tiangen Biotech Co., Ltd., China) was used to extract genomic DNA from intestinal contents following the manufacturer’s instructions. The V1-V9 hypervariable regions of the 16S rRNA gene were amplified using primers (27F: 5′-AGRGTTTGATYNTGGCTCAG -3′; 1492R: 5′-TASGGHTACCTTGTTASGACTT -3′). Amplicon concentrations were quantified and equimolarly normalized before pooling for sequencing on the PacBio Sequel II platform (Beijing Biomarker Technologies Co., Ltd., China).

USEARCH (version 10.0) was used to assign sequences that exceeded the 97% similarity threshold to an operational taxonomic unit (OTU). Taxonomy annotation of the OTUs was performed based on the Naive Bayes classifier in QIIME2 (RRID:SCR 021258) using the SILVA database (RRID:SCR 006423) with a confidence threshold of 70%. Alpha analysis was conducted using QIIME2 to characterize the biodiversity complexity of each sample. Beta diversity calculations were analyzed by principal coordinate analysis (PCoA) to assess the diversity in samples for species complexity. Linear discriminant analysis (LDA) coupled with effect size (LEfSe) was applied to evaluate the differentially abundant taxa.

The LC/MS system consisted of a UPLC (Acquity I-Class PLUS, Waters, United States) in tandem with a high-resolution MS (Xevo G2-XS QTOF, Waters, United States) on an Acquity UPLC HSS T3 column (1.8 μm 2.1*100 mm, Waters, United States) operated in positive/negative ion mode with mobile phase A: 0.1% formic acid aqueous solution and mobile phase B: 0.1% formic acid acetonitrile. MSe mode acquisition of primary and secondary mass spectral data were under MassLynx V4.2 (RRID:SCR 014271) control. ESI ion source parameters were capillary voltage set to 2000 V for positive ion mode or −1500 V for negative ion mode, cone voltage at 30 V, ion source temperature maintained at 150°C, desolvent gas temperature kept at 500°C, backflush gas flow rate of 50 L/h, and desolventizing gas flow rate of 800 L/h.

The raw data were processed using Progenesis QI software to extract and align peaks, in addition to conducting various data processing operations. This was achieved by using the METLIN database (RRID:SCR 010500) and Biomark’s self-created library for identification. Furthermore, theoretical fragment identification and mass deviation were both limited to 100 ppm. Following the normalization of the original peak area information with the total peak area, subsequent analysis is carried out. The identified metabolites were searched for classification and pathway information in KEGG (RRID:SCR 012773), HMDB (RRID:SCR 007712), and LIPID MAPS (RRID:SCR 006579) databases. Based on the grouping, fold change (FC) was calculated, and a t-test was used to calculate the P value for each metabolite. The orthogonal partial least squares discriminant analysis (OPLS-DA) model was utilized to screen the differential metabolites. The screening criteria were FC >1, P-value < 0.05 and VIP >1.5. Differential metabolites of KEGG pathway enrichment significance were calculated using the hypergeometric distribution test.

SPSS 26.0 (RRID:SCR 002865) was utilized for statistical analysis, and most bar charts were created using GraphPad Prism (RRID:SCR 002798). Experimental data were presented as mean ± standard deviation (SD). Group comparisons were executed through either t-test or one-way ANOVA, and multiple comparisons were performed using the Bonferroni method (equal variances) or Tamhane T2 (unequal variances). We conducted Spearman correlation analysis using the R language (RRID:SCR 001905) and plotted heat maps. Statistical significance was set at p < 0.05.

Liver function tests displayed a significant increase in serum levels of AST, ALT and ALP in the M group compared with the C group. After QJ treatment, these levels were significantly reduced (Figure 1B). QJ also significantly downregulated the liver index in HF (Figure 1C). As illustrated in Figure 1A, HE staining revealed that hepatocytes in group C were normal in shape and tightly arranged, without any discernible pathological or inflammatory cellular infiltration in the liver parenchyma. After CCl₄ treatment, hepatocytes exhibited obvious signs of steatosis, inflammation, and necrosis, all of which were alleviated by QJ. The levels of hepatic inflammatory indices in each group were detected by qPCR. The results showed that IL-1β, IL-6, and TNF-α levels were observably increased in the M group, and QJ treatment significantly downregulated this change, which further confirmed that QJ inhibited CCl₄-induced liver inflammation (Figure 1D).

Masson staining and its semi-quantitative analysis (Figures 2A, B) showed that massive fibrous streaks and even pseudolobule formation were visible in group M with respect to group C, while QJ treatment significantly attenuated CCl₄-induced fibrosis. Histologically, HF was characterized by disruption of hepatic architecture and excessive deposition of ECM, with type I collagen being a predominant constituent of ECM. HSC activation is the major cellular source of ECM production, and α-SMA is a marker of HSC activation. Collagen I and α-SMA expression in liver of each group was determined using IHC and qPCR. The results indicate that the mean IOD values and mRNA expression of Collagen I and α-SMA significantly increased in the M group but were significantly inhibited after treatment with QJ (Figures 2A, C, D). The above results demonstrated the inhibitory impact of QJ on CCl₄-induced HF.

HE staining of the ileum showed clear intestinal tissue structure, normal cell morphology, and no obvious inflammatory changes in the C group (Figure 3A). In contrast, CCl₄ modeling resulted in sparsely arranged connective tissue, reduced intestinal villus height and crypt depth, and the infiltration of inflammatory cells. However, QJ administration restored intestinal villus height and crypt depth without observable inflammatory changes.

Tight junctions are vital for maintaining the structural integrity and normal function of the enteric epithelial barrier by serving as the primary connection between intestinal mucosal epithelial cells. Disruption of tight junctions can damage the intestinal epithelial barrier, which may allow intestinal microbes and their metabolites to translocate and exacerbate pre-existing liver disease (Wang and Liu, 2021). Research has shown that the expression of Occludin in the small intestine of mice decreased significantly after the first injection of CCl₄, and the bacterial translocation (CCl₄, 4 times) preceded the imbalance of gut microbes (CCl₄, 24 times; bridging fibrosis) (Fouts et al., 2012). The increased permeability caused by the change of intestinal tight junction may explain the early bacterial translocation in the CCl₄-induced liver injury model. Claudin-1, Occludin and ZO-1 are important components of intestinal tight junction proteins (TJPs). Their mean IOD values and mRNA expression were measured by IHC and qPCR. IHC analysis of the ileum, depicted in Figures 3A, B, displayed a significant reduction in the mean IOD values for Claudin-1, Occludin, and ZO-1 after CCl₄ modeling. Nevertheless, their values were restored post QJ administration. Moreover, the mRNA expression measured by qPCR was consistent with the IHC results (Figure 3C). The findings indicated that CCl₄ modeling causes downregulation of TJPs, which may lead to harmful gut microbes and their metabolites affecting the liver through the gut-liver axis. However, QJ has a beneficial effect on CCl₄-induced intestinal epithelial barrier damage.

To investigate the impact of QJ on the composition of gut microbiota in HF rats, we collected intestinal contents and subjected them to high-throughput 16S rRNA gene sequencing. Venn diagrams demonstrated the number of shared and unique OTUs between groups (Supplementary Figure S1A). The flattening rarefaction curve indicates sufficient sequencing depth across samples (Supplementary Figure S1B). The flattening of rank abundance curve reveals that the samples contained a rich and homogeneous species composition (Supplementary Figure S1C).

The richness and diversity of gut microbiota were assessed using alpha diversity. The indices of Ace, Chao 1, and Shannon demonstrated that the richness and diversity of the gut microbiota increased following CCl₄ induction. However, after the QJ administration, Ace and Chao 1 indices decreased (Figure 4A). The results indicated that QJ treatment counteracted the effects of HF on the richness of gut microbiota.

Then, weighted unifrac-based PCoA and ANOSIM were used to analyze the changes in the structure of the gut microbial community among the groups. ANOSIM (R=0.613, p = 0.016, Supplementary Figure S1D) showed that there were significant differences in gut microbial structure among the three groups. PCoA (PC1 = 62.62%, PCo2 = 13.39%) showed that the C group was completely separated from the M group. Moreover, the gut microbial composition of the QJ group was closer to the C group than to the M group (Figure 4B). The above results indicate that QJ regulated the community structure of the rat’s gut microbiota.

Bugbase is used to predict the potential phenotypes in each group in which the abundances of gram-negative and positive bacteria changed significantly (Figure 4D). Compared with group C, group M significantly raised the abundance of Gram-negative bacteria and significantly reduced that of Gram-positive bacteria, while QJ restored it. At the phylum level, we identified 10 gut microbes in total, of which Firmicutes and Bacteroidota were the two most dominant phyla, making up over 90% of the gut microbial composition (Figure 4C). As shown in Figures 4C–E, the M group exhibited reduced levels of beneficial bacteria Firmicutes and a higher presence of Bacteroidota than the C group. Following QJ treatment, however, the relative abundances of both Firmicutes and Bacteroidetes were reversed (Figures 4C, F).

LEfSe analysis was conducted to identify statistically significant biomarkers and the dominant gut microbes in each group (Figures 4E, F; Supplementary Table S1, S2). The larger the logarithmic LDA score, the more significant differences of that gut microbe between groups. Based on an LDA score >3, the M group displayed an enrichment of Muribaculaceae and Prevotellaceae in Bacteroidota, as well as Erysipelotrichaceae, Clostridiaceae, and Peptococcaceae in Firmicutes, Enterobacteriaceae in Proteobacteria, and Bifidobacteriaceae in Actinobacteriota. Both the C and QJ groups demonstrated enrichment of Lactobacillaceae, uncultured rumen bacterium (belonging to Clostridia UCG 014), Rikenellaceae, etc. Research results indicated that gut microbial structure underwent considerable alterations after induction by CCl₄, and QJ was capable of adjusting or improving these alterations. As illustrated in Figures 4E, F, marked alterations in the abundance of 19 shared genera were recognized at the genus level. Furthermore, at the species level, QJ treatment resulted in downregulation of Escherichia coli and upregulation of Lactobacillus johnsonii and Limosilactobacillus reuteri, i.e., Lactobacillus reuteri. These gut microbes may potentially have an essential role in QJ inhibition of HF.

To investigate the metabolic characteristics of each group, non-targeted metabolomics analysis was conducted using intestinal content samples. OPLS-DA showed that the C group was separated from the M group, and the M group was separated from the QJ group (Supplementary Figure S2A, D). The corresponding permutation tests established good validity and reliability, showing no "overfitting” (Supplementary Figure S2B, E). Volcano plots display the differential metabolites (Figures 5A, B). The metabolic pathways of QJ in inhibiting HF were analyzed based on the differential metabolites between the M and QJ groups. Relevant KEGG pathways included Drug metabolism—cytochrome P450, Fatty acid degradation, Taurine and hypotaurine metabolism, and Lysine degradation (Figure 5D).

Between C vs. M and M vs. QJ, there were 82 common differential metabolites (co-differential metabolites), all of which were inversely regulated by QJ (Figure 5C). After excluding metabolites that lacked annotation in KEGG or LIPID MAPS or HMDB databases, as well as metabolites that were not categorized in HMDB, there were 61 co-differential metabolites (Supplementary Table S3), which may be important metabolites for QJ inhibition of HF.

Spearman correlation analysis was utilized to calculate the associations between 19 shared genera or 61 co-differential metabolites and indices related to HF (AST, ALT, ALP, liver index, and mRNA expression of IL-1β, IL-6, TNF-α, COL1a1, and α-SMA), as well as mRNA expression of TJPs (Claudin-1, Occludin, and ZO-1).

Figure 6A indicates that 10 of the 19 shared genera display a strong association with both HF-related indices and TJPs. Specifically, uncultured rumen bacterium, Rikenellaceae RC9 gut group, and Lactobacillus exhibit a significant negative correlation with HF-related indices, while demonstrating a significant positive correlation with TJPs. Conversely, unclassified Peptococcaceae, Prevotellaceae UCG 001, Turicibacter, Dubosiella, Fournierella, Faecalibaculum, and Oscillibacter exhibit the opposite relationship.

There were 37 metabolites out of 61 co-differential metabolites intimately correlated with HF-related indices and TJPs and reversed by QJ (Figure 6B), which may be the key metabolites for QJ to regulate metabolic changes. Among these 37 metabolites, 7-Hydroxyenterolactone, 2,3-Dihydroxypropyl octanoate, 5″-Phosphoribostamycin, Adhumulinic acid, Talaromycin A, 2′-Hydroxyenterolactone, 2-Phenylethyl 3-phenyl-2-propenoate, 3-(2-Hydroxyphenyl) propionic acid, Cinnamyl phenylacetate, (±)-Enterolactone, Atropaldehyde, Felbamate, Antibiotic X 14889A, and Piquindone showed significant negative correlations with HF-related indices and significant positive correlations with TJPs, while the other metabolites showed positive correlations with HF-related indices and negative correlations with TJPs. In addition, Figure 6C shows that these metabolites are involved in lignans, bile acids, Lipids, Drug metabolism - cytochrome P450, Phenylalanine metabolism, and Taurine and hypotaurine metabolism, etc.

Figure 6C illustrates the correlation between the aforementioned 10 genera and 37 differential metabolites. Piquindone, Talaromycin A, Adhumulinic acid, Cinnamyl phenylacetate, 7-Hydroxyenterolactone, 2′- Hydroxyenterolactone, (±)-Enterolactone, Atropaldehyde, Felbamate, 5″-Phosphoribostamycin, 3-(2- Hydroxyphenyl)propionic acid, 2-Phenylethyl 3-phenyl-2-propenoate, 2,3 Dihydroxypropyl octanoate, and Antibiotic X 14889A were positively correlated with Rikenellaceae RC9 gut group, uncultured rumen bacterium, and Lactobacillus, and adversely related to Prevotellaceae UCG 001, Turicibacter, unclassified Peptococcaceae, Faecalibaculum, Oscillibacter, Fournierella, and Dubosiella; the opposite was true for the other metabolites. Turicibacter, Faecalibaculum, Prevotellaceae UCG 001, and unclassified Peptococcaceae were the most highly correlated with 37 differential metabolites among the 10 genera. Notably, both Faecalibaculum and Turicibacter were significantly correlated with all 37 differential metabolites, while Prevotellaceae UCG 001 and unclassified Peptococcaceae were significantly correlated with 36 metabolites. These four genera might serve as the core gut microbes that inhibit HF by QJ.