YE (MF-4, purity≥98%, nutrient composition are shown in Supplementary Table S1)was supplied from Mufan Biotechnology Co., Ltd. (Henan, China). Silymarin (H20181067, 140 mg/capsule) was obtained from Plantextrakt GmbH & Co. KG. The liquid diet was purchased from Trophic Animal Feed High-Tech Co., Ltd. (Nantong, China). Detection kits for alanine aminotransferase (ALT), aspartate aminotransferase (AST), total antioxidant capacity (T-AOC), glutathione peroxidase (GPX), malondialdehyde (MDA), superoxide dismutase (SOD), lactic dehydrogenase (LDH), and superoxide Anion (O2-) were purchased from Beijing Boxbio Science & Technology Co., Ltd. Detection kits for alkaline phosphatase (AKP), triglyceride (TG), total cholesterol (T-CHO), lipid peroxidation (LPO), alcohol dehydrogenase (ADH), Hydroxyl Radical Scavenging Ability (OH Scavenging Ability), total nitric oxide synthase (TNOS), and inducible nitric oxide synthase (iNOS) were purchased from Jiancheng Corp (Nanjing, China). ELISA kits for leukocytes-1β (IL-1β), tumor necrosis factor-α (TNF-α), interferon-γ (INF-γ) were supplied from Elabscience Biotechnology Co., Ltd. Fish oil was purchased from Shanghai Yien Chemical Technology Co., Ltd. (Shanghai, China). Leukocytes-1β (IL-1β), tumor necrosis factor-α (TNF-α) and interferon-γ (INF-γ) were supplied from Elabscience Biotechnology Co., Ltd. Fish oil was purchased from Shanghai Yien Chemical Technology Co., Ltd. (Shanghai, China).

Four-week-old male Wistar rats were purchased from Lanzhou University Animal Experimental Center (Lanzhou, China). The rats were housed in a clean environment with a 12-h light–dark cycle at 20–22°C and 45 ± 5% humidity. All the rats were free to access food and water ad libitum. All experiments involving animals were approved by the Ethics Committee of School of Public Health, Lanzhou University (IRB22030601). After one-week of acclimatization, the rats were randomly divided into the following five groups (11 rats per group): liquid diet control group (LC), alcohol liquid diet group (AL), low-dosage yeast extract intervention group (YEL, 60 mg/kg/day), high-dosage yeast extract intervention group (YEH, 120 mg/kg/day), and silymarin intervention group (PC, 120 mg/kg/day). AL, PC, YEL, and YEH groups were all given liquid diet with an alcohol concentration of 8% (v/v). Both types of liquid diet (with and without alcohol) had calorie densities of 1 kcaL/mL. The source of fatty acids in the diet was fish oil. During the first 3 days of adaptive feeding, all groups were given a control liquid diet. From the fourth to the nineth day, all groups (except for the LC group) entered the adaptive feeding stage of an alcohol liquid diet. The ratio of liquid control feed to liquid alcohol feed was 2:1, 1:1, and 1:2, and each ratio was fed for 2 days. After the adaptive feeding period, the rats were provided with a complete liquid alcohol diet. At the start of the alcohol feeding procedure, YE and silymarin were well-dissolved in ultrapure water prior to daily administration to the animals via gavage, over a period of 6 weeks. Food intake and body weight were monitored daily and weekly, respectively. At the end of the experiments, the rats were gavaged with 40% alcohol (15 mL/kg). After 12 h, the rats were anesthetized with an intraperitoneal injection of 3% sodium pentobarbital and then sacrificed. The liver of each rat was weighed to calculate the liver organ coefficient (wet liver weight, mg/ body weight, g). The livers were removed and rinsed with cold PBS. Portions of the liver were used for histological evaluation, and others were stored at −80°C for further analysis. Fecal samples were immediately frozen in liquid nitrogen and stored at −80°C until analysis was done.

Samples of the right liver lobe were fixed in 4% paraformaldehyde for 24 h, embedded in paraffin wax, and then cut into 5 μm-thick sections. The sections were blurred by Hematoxylin–eosin (H&E) staining and Masson’s trichrome. The pathological changes on the liver were observed using the BX53 system microscope with an attached camera (Olympus, Japan) at 400× magnification.

The liver samples (about 100–150 mg) were mixed with 0.9% normal saline at a ratio of 1:9 to make homogenates. Centrifugation was then done at 3000 rpm for 15 min, at 4°C. The supernatant was collected for subsequent analyzes. The levels of ALT, AST, AKP, ADH, and LDH were assessed according to the instructions on the commercial assay kits.

To evaluate the oxidative stress indicators, homogenates were prepared using liver samples weighing 200 mg combined with normal saline, at a ratio of 1:9. The levels of MDA, T-AOC, LPO, SOD activity, and GPX activity were detected, following the manufacturer’s instructions.

The liver tissues (about 50–100 mg) were mixed with normal saline at a ratio of 1:9 to prepare homogenates. Supernatants were collected after centrifugation. The levels of TNOS and iNOS in the liver tissues were analyzed by assay kits (Nanjing, China), according to the manufacturers’ instructions. The levels of IL-1β, TNF-α, and INF-γ in liver tissue were measured using ELISA kits (Elabscience Biotechnology, China) followed by the method described in the instructions.

The analysis was conducted at Metware Biotechnology Co., Ltd. (Wuhan, China). Genomic DNA from the fecal samples (six samples were randomly selected from each group) was extracted by the Magnetic Soil and Stool DNA kit (TianGen Biotech Co., Ltd., Beijing, China). The bacterial sample was used to amplify the V4 hypervariable region of the 16S rRNA gene with primers 515F(GTGCCAGCMGCCGCGGTAA) and 806R(GGACTACHVGGGTWTCTAAT) (Itoh et al., 2014). Purified PCR products were quantified, using enzyme-labeled quantification, and the sequencing library preparation was performed based on the manufacturer’s instructions (Illumina, San Diego, CA, United States). Quantification was performed by Qubit and Q-PCR;sequencing was done using NovaSeq6000 (Illumina, San Diego, CA, United States). For these data, the reads of each sample were assembled using FLASH (v1.2.11, http://ccb.jhu.edu/software/FLASH/) after truncating the barcode and primer sequences, resulting in the Raw Tags. The Raw Tags were then truncated at the first low-quality base site, with consecutive low-quality values (default quality threshold: ≤ 19) and a base number reaching the set length (default length value is 3). Subsequently, the truncated Tags were filtered to remove those with a continuous high-quality length below 75% of the Tags length. The effective Tags were sequenced using Novaseq 6,000 (Illumina, San Diego, CA, United States), and the Uparse algorithm (v7.0.1001, http://www.drive5.com/uparse/) was used to cluster all sequences into OTUs (Operational Taxonomic Units). For the species annotation analysis of OTUs sequences, the Mothur method and the SSUrRNA database of SILVA138.1¹ were used, setting the threshold between 0.8 ~ 1.

Furthermore, the Qiime software (version 1.9.1) was used to the “observed number,” as well as the Shannon and Simpson indices. R software (version 4.1.2) was employed in creating the Rank abundance curve. The nonparametric test of the difference between groups in Alpha diversity index was done using the Kruskal-Wallis test. LEfSe analysis was performed using the LEfSe software (version 1.0, https://huttenhower.sph.harvard.edu/lefse/), and the default LDA score filtering value was set to 4. The Mothur software (University of Michigan, Michigan, United States) was used in Metastats analysis at each classification level to obtain the p value (p < 0.05). The Benjamini and Hochberg False Discovery Rate method was then used to correct the p value to obtain the q value.

In this research, the metabolites were analyzed at Metware Biotechnology Co., Ltd. (Wuhan, China). Fecal samples (six samples were randomly selected from each group) stored at −80°C refrigerator were thawed. A 400 μL solution (methanol: water = 7:3, v/v) containing an internal standard was added into each sample (20 mg) and vortexed for 3 min. All samples were sonicated in an ice bath for 10 min and vortexed for 1 min before being placed in −20°C for 30 min. These samples were then centrifuged at 12000 rpm for 10 min at 4°C. Each supernatant was recentrifuged at 12000 rpm for 3 min at 4°C. 200 μL aliquots of each supernatant were transferred for ultra-performance liquid chromatography tandem mass spectrometry (UPLC-MS/MS, Sciex, United States) analysis, which included both non-targeted and broad-spectrum targeted detection. The non-targeted metabolites were separated on a pre-column (ACQUITY HSS T3 C18,1.8 μm, 2.1 mm*100 mm), with the column temperature being set at 40°C. mobile phase A consisted of water with 0.1% formic acid, and mobile phase B was comprised of acetonitrile with 0.1% formic acid. The gradient program was as follows: 95:5 v/v at 0 min; 10:90 v/v at 10.0 and 11.0 min; 95:5 v/v at 11.1 min; 95:5 v/v at 14.0 min. The flow rate was 0.4 mL/min, with an injection volume of 5 μL. The Quadrupole-Time of Flight mass spectrometry analysis was performed with an electrospray ionization (ESI) temperature of 500°C, voltage of 5,500 V (positive) and − 4,500 V (negative), ion source gas I (GS I) at 50 psi, gas II (GS II) at 50 psi, and Curtain gas (CUR) at 25 psi. The delustering potential collision energy was set at 80 psi, and the collision energy spread was at 15 psi. For Broad-spectrum targeted detection, a waters ACQUITY UPLC HSS T3 C18 chromatographic column (1.8 μm, 2.1 mm*100 mm) was employed, with an injection volume of 2 μL. Tandem mass spectrometry (QTRAP^®) analysis was carried out using the same ESI temperature, voltage, ion source gas, and curtain gas condition as those in the Quadrupole-Time of Flight mass spectrometry.

All sample extracts were mixed in equal amounts to form a QC sample. Based on the database MWDB (including secondary spectrum, retention time RT), DB-all public database (including HMDB, KEGG and other databases), AI prediction library, and MetDNA, metabolites identification and extraction of multiple ion pairs, as well as retention time RT information were performed The MRM accurate quantification of the metabolites in the new library was carried out using the Q-Trap instrument platform.

The OPLS-DA function of the MetaboAnalyst R package in the R software was used for analysis. Based on the OPLS-DA results, the Variable Importance in Projection (VIP) of the OPLS-DA model was determined from the obtained multivariate analysis, and the differential metabolites were further screened by combining with the fold change. The selected data with different amounts of the metabolites were treated with UV (unit variance scaling). The R software Complex Heatmap package was used to draw the cluster heat map. Pearson correlation analysis was used to analyze the association between different metabolites that were identified according to the screening criteria.

Finally, the KEGG (Kanehisa and Goto, 2000)² and HMDB³ database were used to annotate metabolites and identify the pathways of differential metabolite enrichment.

All statistical analysis was carried out using the SPSS statistical software (version 22.0, IBM, White Plains, New York, United States), and the measurement data were expressed as mean ± standard error of mean deviation. One-way ANOVA was used for comparison among multiple groups and the significance of difference between the two groups was determined by LSD test. When the variances are not equal, Dunnett’s T3 was used to test the differences between the two groups. Values of p < 0.05 were considered significant. All images were generated by OriginPro (2019, OriginLab, Massachusetts, United States) and GraphPad Prism (version 9.5.1, Graphpad Software, California, United States).

To explore the effects of YE on ALD, H&E and Masson staining were performed to check the histomorphological changes that were induced by alcohol. The hepatocytes arranged neatly without obvious pathological changes in the LC group, but there was irregular arrangement of hepatocytes, diffuse steatosis, and a large number of inflammatory cell infiltration in AL group (Figure 1B). According to Masson staining, there were more parts that were dyed blue in the AL group, suggesting a more serious case of fibrosis (Figure 1A). Compared with the AL group, the pathological changes observed in the liver were alleviated in the PC, YEL, and YEH groups, with YEH exhibiting the most significant improvement (Figures 1A,B). In comparison to the AL group, a remarkable decrease in the liver organ coefficient was observed in the YEH group (p < 0.05, Figure 1C), whereas no significant reduction was noted in the PC and YEL groups. Biochemical analysis demonstrated that the levels of hepatic ALT, AST, and AKP, which are diagnostic markers of liver injury, remarkably increased in the AL group compared to the LC group. However, the administration of YE and Silymarin effectively inhibited the increase (Figures 1D–F). The administration of YE also reduced the accumulation of TC and TG (p < 0.01, Figures 1G,H). Additionally, high doses of YE significantly decreased LDH levels (p < 0.01, Figure 1J). Furthermore, YE administration effectively prevented the decrease in ADH levels (p < 0.001, Figure 1I).

The levels of T-AOC, GPX, SOD and·OH scavenging ability remarkably decreased in the AL group than in the LC group (Figures 2A,B,D,F). On the other hand, the levels of MDA and O₂⁻ exhibited opposing trends (p < 0.05, Figures 2C,E). It is noteworthy that the groups with the YE intervention exhibited strong antioxidant activity by resisting alcohol-induced liver damage, which showed good protective effects (p < 0.05, Figures 2A–F). However, there were no significant changes in the levels of T-AOC and O₂⁻ after treatment with Silymarin (p > 0.05, Figures 2D,E). The highest level of ·OH scavenging ability was observed in the YEL group (p < 0.05, Figure 2F).

Some studies have reported that excessive or chronic alcohol use can trigger an inflammatory response in the liver (Xiao et al., 2020; Hyun et al., 2021). As shown in Figure 3, alcohol ingestion significantly increased the levels of IL-1β, TNF-α, IFN-γ, iNOS, and TNOS in the liver of rats (p < 0.05, Figures 3A–E). Compared with the AL group, administration with high-dosage YE inhibited alcohol-induced inflammatory response, and this was consistent with LC group (p < 0.05, Figures 3A–E). However, the levels of IL-1β, TNF-α, and IFN-γ in the YEL group and those of IFN-γ in the Silymarin group did not exhibit obvious changes (p > 0.05, Figures 3A–C). These findings suggested that YE intervention can reduce the levels of inflammatory factors and alleviate alcohol-induced liver damage. High dosage of YE appear to provide superior protective effects.

ALD is closely associated with the gut microbiota. Long-term consumption of alcohol can substantially alter the structure of the gut microbiota in ALD patients (Bishehsari et al., 2017; Wang et al., 2021).To characterize alterations in gut microbiota, fecal samples were subjected to 16S rRNA gene sequencing. The analysis revealed specific and shared OTUs among the five groups of mice (Supplementary Figure S1A). In comparison to the LC group, all experimental groups exhibited an increase in both the Shannon and Simpson diversity index (Supplementary Figures S1B,C). The Weighted Unifrac analysis demonstrated that the β-diversity of gut microbiota in the PC, YEL, and YEH groups was significantly distinct from that of the AL group, but closer to that of the LC group (Supplementary Figures S1D,E). These results suggest that YE intervention could reverse alcohol-induced dysbiosis of gut microbiota in rats.

The structure of the gut microbiome was investigated at both phylum and genus levels (Figures 4A,B). Bacteroidetes and Firmicute were the most abundant at the phylum level in all groups. However, when compared with the AL group, the ratio of Firmicutes to Bacteroidetes (F/B) was reduced in the other groups, though the change was not significant (Figure 4C). On the other hand, the population of Proteobacteria was significantly elevated in both the YEH and PC groups (p < 0.05), compared to the AL group. At the genus level, intervention with YEL and YEH resulted in a decrease in the relative abundance of Peptococcus and Tyzzerella, while an increase in the relative abundance of Gemella (Figures 4D–F), compared to the AL group was observed. In the YEL group, the relative abundance of Ruminococcus and Faecalibacterium was significantly decreased (Figures 4G,H). It was also noted that the administration of YEH resulted in a significantly reduced relative abundance of Faecalibaculum, Helicobacter, Parasutterella and Romboutsia compared to the AL group, while the Corynebacterium remarkably increased (Figures 4I–M).

Linear discriminant analysis (LDA) and LEfSe analysis were used to identify biomarkers with significant differences. Based on LDA scores, the relative abundance of Ruminococcus was significantly higher in the AL group compared to the YEL group (Figure 5A). In addition, Romboutsia, Faecalibaculum, and Parasutterella were found to be more abundant in the AL group than in the YEH group (Figure 5B). The alterations in gut microbiota have a certain degree of correlation with some physicochemical indicators that are related to liver injury. As shown in Figure 5C, the genera such as Tyzzerella and Peptococcus decreased after YE intervention, and these positively correlated with ALT, AST, and MDA. In addition, positive correlation was observed between Peptococcus and AKP. However, negative correlation was noted between Peptococcus and ADH, SOD, GSH, and T-AOC. After YE intervention, Gemella increased, while Corynebacterium levels only increased in the YEH group. Both genera showed significant positive correlations with hydroxyl radical scavenging ability, but significant negative correlations with TNOS were observed (Figure 5C). In the YEL group, Faecalibacterium and Ruminococcus significantly decreased. Faecalibacterium showed a positive correlation with MDA and TC, but a negatively correlation with SOD. On the other hand, the abundance of Ruminococcus positively correlated with ALT, AST, AKP, TC, TG, and MDA, while it was negatively associated with levels of ADH, SOD, GPX, and hydroxyl radical scavenging ability (Figure 5C). The genera Romboutsia and Faecalibaculum decreased in the YEH intervention group. Romboutsia exhibited a significant positive correlation with TNOS and iNOS, while Faecalibaculum showed a positive correlation with TC and iNOS, but a negative association with GPX (Figure 5C).

To further explore and understand the effects of altered gut microbiota on metabolite changes, a total of 2029 metabolites were detected through the extensive targeted metabolomics analysis (Figure 6A). According to the orthogonal partial least-squares discriminant analysis (OPLS-DA), metabolites in the YEH group could be well separated from those in the AL group (Q² > 0.5) (Figure 6C). Based on the threshold of fold change ≥2 or fold change ≤0.5 and VIP > 1, 201 differential metabolites were identified between the AL and YEL groups, while 470 differential metabolites were found between AL and YEH groups. By combining pathways with significant enrichment in KEGG and HMDB databases, it was discovered that among the samples of AL and YEH, 14 metabolites were significantly different (two up-regulated metabolites and 12 down-regulated metabolites). Additionally, there were six remarkedly different metabolites (five up-regulated metabolites and one down-regulated metabolite) among the samples of AL and YEL groups. The heatmaps of specific metabolic difference substances are shown (Figures 7A,B).

To explore the relationship between the co-regulation of gut microbes and metabolites, the Spearman correlation analysis was performed (Figure 7C). Gemella levels were found to exhibit notable negative correlation with 17β-estradiol 3-(β-D-glucuronide) concentrations. Parasutterella positively correlated with histamine and adenosine, while Corynebacterium inversely correlated with histamine.

According to the KEGG database, the YEL notably affected insulin resistance (Figure 8A), while the HMDB database indicated that the intervention of the YEL significantly affects two major metabolic pathways: taurine and hypotaurine metabolism, as well as phospholipid biosynthesis. According to the KEGG database, there are significant differences in steroid hormone biosynthesis, morphine addiction, cGMP-PKG signaling pathway, and gastric acid secretion between the AL and YEH groups (Figure 8C). Based on the HMDB database, YEH significantly impacted three metabolic pathways: intracellular signaling via adenosine receptor A2a and adenosine, intracellular signaling through adenosine receptor A2b and adenosine, and intracellular signaling mediated histamine receptor H2 and histamine (Figure 8D).