Experimental materials were obtained from ‘XJ4’ apple fruits grown in the Modern Agricultural Science and Technology Demonstration Park of Qingdao Agricultural University, Jiaozhou, China (36°16′49′′N, 120°00′36′′E).

The preparation method of the RAFE was a slightly modified version of the method of Xiang (39). The apple fruits were cleaned, and 0.1% phosphoric acid/ethanol was added as the extracting solution in liquid nitrogen grinding, with a material-to-liquid ratio of 1:30 (g ml^–1). The mixture was then sonicated (Ultrasonic cleaner, Baoding Shengfeng Instrument Technology Co., LTD., HP) for 30 min and macerated for 12 h with protection from light, then finally filtered to obtain a crude extract solution of flavonoids. Concentration was conducted with a rotary evaporator (Rotary evaporators, Shanghai Yarong Biochemical Instrument Factory, RE-6000A) at 35°C and low speed to remove the organic solvent from the extract until the crude extract became viscous. Finally, the crude extract was diluted by adding an appropriate amount of deionized water (100 ml).

The flavonoid extract was added to NKA-9 macroporous sorbent resin at pH7 and the flow rate of the upper sample was controlled at 1 ml min^–1. After reaching equilibrium of resin adsorption, the impurities were eluted with deionized water, the macroporous resin was eluted with 100% (V/V) ethanol eluate at the same flow rate, and all eluates were collected. The eluate was concentrated and dried using a nitrogen blowing machine and vacuum freeze dryer sequentially, and the volume was fixed at a final concentration of 30 mg ml^–1. Figure 1 shows the process flow for the preparation and purification of RAFE.

Protocatechin, catechin, epicatechin, rhizobioside, quercetin, centaureidin-3-O-galactoside, and centaureidin-3-O-glucoside standards were purchased from Sigma-Aldrich (St. Louis, MO, USA). Chromatography-grade methanol was purchased from Tedia (USA). All flavonoid standard stock solutions (1 mg ml^–1) were stored at a low temperature and protected from light, and the standard curve was constructed using the external standard method. High-performance liquid chromatography (HPLC, USA, Agilent, 1260) was performed under the following conditions: the chromatographic column was an XDB-C18 column (150 × 4.6 mm, 5 μm); the mobile phase A was 1% aqueous acetic acid solution and the mobile phase B was methanol solution; the elution program was 0–10 min, 5–30%, 10–25 min, 30–50%, 25–35 min, 50–70%, and 35–40 min, 70–5% (all mobile phase B); the flow rate was 0.8 ml/min; the sample volume was 20 μl; the column temperature was 30°C; and the wavelength was 280 nm. Qualitative and quantitative analyses of flavonoids in ‘XJ4’ RAFE were performed according to the above methods and conditions.

All animal experimental procedures were conducted according to the protocol of the Animal Care and Use Committee of Qingdao Agricultural University (license number: SYXK [SD] 20170005). Specific Pathogen Free (SPF) grade male ICR mice (3 weeks old, 20.0 ± 2.0 g) were housed in an SPF chamber at 22 ± 2°C with a 12 h light/dark cycle and 50–70% humidity, and all mice were fed and watered ad libitum. After 3 days of adaptation to the living environment, the mice were randomly divided into the following six groups (10 mice per group): normal control (NC), pathological control (MC), positive control (PC), low RAFE concentration (L-RAFE), medium RAFE concentration (M-RAFE), and high RAFE concentration (H-RAFE). The prepared RAFE was diluted into different concentrations and fed to mice in the RAFE intervention groups by gavage (0.2 mg kg^–1, L-RAFE; 1 mg kg^–1, M-RAFE; and 5 mg kg^–1, H-RAFE). The NC group was administered saline gavage, the MC group was administered saline and CCl₄ as a model (0.1% CCl₄ olive oil solution 10 ml kg^–1, intraperitoneal injection, twice a week, 4 weeks in total), and the PC group was treated with bifenthrin drops in aqueous solution (1 mg kg^–1, gavage). CCl₄ was injected intraperitoneally at the same time as the gavage in all groups except the NC group, and the injection volume was the same as that of the MC group. The gavage volume was 0.1 ml mouse^–1, which was administered for 4 weeks. The establishment of the CCL₄ liver injury model was established with reference to Keshavarz et al. and slightly modified (40, 41).

After 35 days, the mice were deprived of food for 12 h then euthanized by cervical dislocation under ether anesthesia. After euthanasia, heart blood and liver tissues were taken from the mice, weighed, and stored at −80°C for further analysis.

Superoxide dismutase, T-AOC, and MDA diagnostic kits were purchased from Solebro Biotechnology (Beijing, China). AST, ALT, alkaline phosphatase (ALP), and total glutathione (T-GSH) diagnostic kits were purchased from the Nanjing Jiancheng Institute of Biological Engineering, Nanjing, China. Tests were performed according to the manufacturer’s instructions.

We processed the mice liver tissue samples fixed in 4% paraformaldehyde. The samples were paraffin-embedded and cut into sections 7–8 μm thick, which were stained with hematoxylin-eosin (H&E) dye, then observed and imaged using an optical microscope (Leica fluorescence microscope, Germany Leica Microsystems, DM2500) with a 40× objective to assess pathological changes in the liver tissue.

The samples used for 16S rRNA gene sequencing were fecal samples from mice fed for 5 weeks. Genomic DNA was extracted using the CTAB method (Nobleryder, China), followed by agarose gel electrophoresis to determine DNA purity and concentration. Finally, the DNA sample was diluted with sterile water to 1 ng μl^–1.

The template was diluted genomic DNA, and the variable region V4 of the 16S rRNA gene consisted of forward primers 515 F (5′-GTGCCAGCMGCCGCGGTAA) and 806 R (3′- GGACTACHVGGGTWTCTAAT). PCR uses specific primers with barcodes and a highly efficient High-Fidelity DNA polymerase [Phusion^® High-Fidelity DNA polymerase (M0530S, New England Biolabs, USA)] to ensure amplification efficiency and accuracy. The PCR system comprised Phusion Master Mix (2×) 15 μl, PrimerF (1 μM) 1 μl (1 μM), PrimerR (1 μM) 1 μl (1 μM), gDNA (1 ng/μl) 10 μl (5–10 ng), and ddH₂O to make up 30 μl of the system. The PCR reaction program was: 98°C pre-denaturation for 1 min; 30 cycles of 98°C for 10 s, 50°C for 30 s, 72°C for 30 s, and 72°C for 5 min.

PCR products were detected by electrophoresis on 2% agarose gel. The PCR products that passed the test were purified using magnetic beads and quantified using enzyme labeling, then detected by 2% agarose gel electrophoresis after mixing the product in equal amounts according to the concentration of the product. Target bands were recovered using a gel recovery kit (Qiagen). Library construction was performed using a library construction kit (TruSeq DNA PCR-Free Sample Preparation Kit). The constructed libraries were then quantified by Qubit and Q-PCR. If the library qualified, sequencing was performed using NovaSeq6000 (Illumina Novaseq6000, Illumina Corporation, Illumina, San Diego, CA, USA).

Data analysis of the gut microbes was performed using the cloud platform.¹ If there were only two groups of samples, the t-test and Wilcoxon test were used; if there were more than two groups of samples, the Tukey test and Kruskal–Wallis test were used. The remaining graphs were produced using GraphPad Prism software (version 8.0.2; GraphPad Software Inc., San Diego, CA, USA). The data were tested for multiple comparisons using Dunnett’s method in SPSS (version 24.0; SPSS Inc., Chicago, IL, USA) and ordinary one-way ANOVA. P < 0.05 was set as the threshold to indicate a significant difference.

The microbiota raw sequencing data generated in this study have been uploaded to the NCBI SRA database with accession numbers: RJNA901622.²

To clarify the composition and content of flavonoids in ‘XJ4’ RAFE, we used HPLC to draw standard curves for eight standards, such as catechins, and delineated the linear range of each standard (Table 1). Flavonoids were identified by comparing the retention times of sample peaks and standard peaks, and the contents of various flavonoids in RAFE were calculated using the standard curves (Figure 2). In the chromatogram of ‘XJ4’ RAFE (Figure 2), we detected eight flavonoid compounds, including one flavonol, three flavanols, two anthocyanins, and two dihydrochalcone. The data showed that ‘XJ4’ RAFE had a high content of catechin (49.346 mg⋅100 g^–1), followed by quercetin (23.087 mg⋅100 g^–1), and the least abundant substance was rhizobioside (0.827 mg⋅100 g^–1). Interestingly, the levels of centaureidin-3-O-galactoside and centaureidin-3-O-glucoside reached 13.306 mg⋅100 g^–1 and 4.426 mg⋅100 g^–1, respectively (Table 1).

Studies have shown that flavonoids have protective effects against CCl₄-induced liver injury in mice. An elevated liver index is an important indicator of liver injury. During rearing, body weight, fluid consumption, and food consumption were measured every 2 days in each group. Based on the weekly data, we found no significant changes (P > 0.05) in the above three basic life indicators (Figures 3A–C). CCl₄-treated mice showed a significant increase in liver size and liver index compared to the normal controls, but RAFE treatment produced a certain recovery effect, especially in the H-RAFE group, where the liver coefficient was significantly lower than that of the MC group (P < 0.01) and almost returned to the level of the NC group (Figures 3D, E). The PC group showed a significant effect on the recovery of liver coefficients (Figure 3E). According to the liver images, the livers of the MC group exhibited obvious edema expansion and a yellowish color, as well as many white dotted lesions on the surface of the liver. Conversely, treatment with three different concentrations of RAFE and bifendatatum restored the color and edema of the liver, especially in the M-RAFE, H-RAFE, and PC groups, which were almost restored to normal levels (Figure 3D).

Liver injury adversely affects the transport function and membrane permeability of hepatocytes and causes enzymes to leak out of cells. The most important enzymes that reflect liver injury are ALT, AST, and ALP. In this study, we determined the enzymatic activity levels of ALT, AST, and ALP in mouse plasma (Figure 4). We found that ALT, AST, and ALP enzyme activities increased sharply after CCl₄ intervention in plasma, to 3.86, 12.93, and 82.16 U/L, respectively, whereas the enzyme activity of NC group was only 0.46, 7.38, and 46.56 U/L (Figure 4). This indicated that the liver tissue of mice was severely damaged after CCl₄ treatment. After RAFE intervention, the levels of ALT, AST, and ALP were all significantly reduced, especially in H-RAFE and M-RAFE groups, where the enzymatic activity of AST was restored to almost normal levels (P < 0.0001) (Figure 4B). Simultaneously, ALP and ALT levels were significantly decreased in the H-RAFE and M-RAFE groups. Interestingly, RAFE reduced ALT activity in a dose-dependent manner (Figure 4A). Bifendate, a clinical drug used for the treatment of hepatitis, also significantly reduced the plasma levels of ALT, AST, and ALP (Figure 4). These results suggest that RAFE can effectively relieve liver injury caused by CCl₄.

Carbon tetrachloride-induced liver injury is associated with lipid peroxidation in mice, and elevated MDA levels are a key manifestation of liver injury caused by lipid peroxidation. In this study, the level of MDA increased from 17.36 nmol/g quality in the MC group to 27.65 nmol/g quality in the NC group (Figure 5D), which indicated that CCl₄ treatment caused significant peroxidative injury in the mouse liver. However, the level of MDA decreased after RAFE treatment, and the most pronounced effect was in the H-RAFE group, with a reduction of 18.60 nmol/g quality (Figure 5D). This suggests that RAFE alleviates CCl₄-induced liver peroxidation. SOD is the most important antioxidant enzyme in biological systems. T-GSH is the main non-enzymatic antioxidant and regulator of intracellular redox homeostasis. Normally, there is a physiological balance between the production of reactive oxygen species, free radicals, and antioxidant enzymes. Peroxidation of liver tissues usually results in high consumption of antioxidant enzymes/antioxidants, such as SOD and T-GSH. In this study, SOD, T-GSH, and T-AOC decreased to 44.80 U/g quality, 157.59 μmol/L, and 29.76 μmol/g quality, respectively, in the MC group (Figures 5C, E, F). However, RAFE intervention significantly increased the levels of SOD, T-GSH, and T-AOC (Figures 5C, E, F), which followed the same trend as that of enzyme activity in the mouse plasma (Figure 4). Interestingly, the best recovery of T-GSH was observed in the L-RAFE group, which returned to normal levels of 237.67 μmol/L (Figure 5C). The PC group also showed better recovery of biochemical parameters in the liver (Figures 5C–F). Recovery of these liver indices showed that RAFE can improve CCl₄-induced oxidative stress and help reduce oxidative injury in the mouse liver.

Histopathological observations provide visual data on the recovery effect of RAFE in mouse liver injury. H&E staining of liver sections showed that the NC group had well-preserved cytoplasm and typical hepatocytes with clearly visible central veins. The hepatocytes were tightly bound to each other, and the liver porosity was only 2.34% (Figures 5A, B). In contrast, significant pathological changes were observed in the hepatocytes around the central vein of the liver in the MC group, and the cells were loosely bound with a liver porosity of 24.05%. However, after RAFE intervention, hepatic tissue cells normalized, liver porosity significantly decreased, hepatocellular lesions around the central vein were reduced, and the H-RAFE group showed the best recovery, with a 7.89% reduction in porosity, indicating a recovery effect that was second only to that of bifendate. The results of the histological observation analysis were consistent with the trend of serum and liver recovery, which suggested that RAFE intervention significantly ameliorates severe histological lesions in mice with CCl₄-induced liver injury.

It has been shown that there is a bidirectional relationship between the liver and gut microbes, with the portal vein enabling direct transport of gut-derived products to the liver, as well as a hepatic feedback pathway of bile and antibody secretion to the intestine (42). Here, we investigated whether flavonoids extracted from red-fleshed apples can restore liver injury by regulating intestinal microorganisms. We determined the effect of RAFE on the overall structure and abundance of mouse intestinal microorganisms by analyzing the gene sequence of the 16S rRNA (V4 region) in mouse feces. The smoothed dilution curve indicated the reliability of the sequencing results (Figure 6A). The petal diagram showed that CCl₄ reduced the number of operational taxonomic units (OTUs) in the pathological group from that in the normal control group, whereas the bifendate and RAFE intervention groups both restored the number of similar microorganisms to a certain extent, especially in the H-RAFE group. This suggests that RAFE has the potential to restore intestinal microbes (Figure 6B). The statistical results of the alpha index showed that the diversity of NC and MC groups was negatively correlated, and intervention with RAFE partially restored the diversity of intestinal microbes in mice with CCl₄-induced liver injury. Overall, H-RAFE intervention showed the best effect; however, drug treatment also restored intestinal microbe levels (Figure 6C).

Beta diversity analysis was used to assess the similarity between groups. OTU-based non-metric multidimensional scaling analysis showed that the MC group was clearly distinct from the other treatment groups, whereas all other treatment groups had certain similarities with the NC group, which suggested that RAFE could restore the composition of intestinal microbes (Figure 7A). We then performed principal co-ordinates analysis based on the unweighted UniFrac distance. The results showed that the MC group had lower community similarity with the NC group, whereas the distance to the NC group after RAFE treatment was relatively close and the species composition structure was similar, illustrating the success of CCl₄ modeling and the therapeutic effect of RAFE (Figure 7B). The unweighted pair-group method with arithmetic mean dendrogram based on the unweighted UniFrac distance showed that the NC and MC groups had different gut microbial compositions because the two groups were located on different branches, and the microbial composition after RAFE and bifendate intervention was closer to that of the NC group, especially for the M-RAFE group (Figure 7C).

To identify the specific taxa associated with RAFE, the relative abundance of intestinal microbes was determined at the phylum and genus levels (Figures 8A, B). At the phylum level, the microbial community was mainly composed of Firmicutes, Proteobacteria, and Bacteroidetes. The abundance of Firmicutes was higher in the MC group, whereas the abundance of Bacteroidetes was lower, resulting in an increase in the ratio of Firmicutes to Bacteroidetes (F/B) compared to the NC group. RAFE treatment affected the F/B ratio in two ways, increasing the relative abundance of Bacteroidetes and decreasing the relative abundance of Firmicutes, thereby reducing the F/B ratio and returning F/B to normal levels. Interestingly, the interventions of both RAFE and bifendate increased the relative abundance of Proteobacteria and Actinobacteria compared to that in the MC group (Figure 8A). At the genus level, the intervention of RAFE reduced the relative abundance of Staphylococcus, increased the relative abundance of probiotics such as Lactobacillus, and restored the intestinal microbial level almost back to normal (Figure 8B). To further validate the differences in intestinal microbes across all groups, the abundance of different species was analyzed at the phylum and genus levels. On the phylum-level abundance heat map, the top 35 dominant bacterial groups in the NC group were similar to those in the bifendate group and RAFE intervention groups (Figure 8A). Interestingly, on the genus-level heatmap, the dominant flora of the NC group was only similar to that of M-RAFE (Figure 8B).

Linear discriminant analysis effect size indicated that the dominant microbes in the MC group were Firmicutes, and intervention with M-RAFE mainly increased nine microbiota induced by CCl₄ in mice, including Lachnospiraceae, Muribaculaceae, Clostridia, and other probiotics (Figure 8C). Taken together, these data suggest that the gut microbiota of mice induced by CCl₄ are imbalanced, and that RAFE can modulate gut disturbance in mice with liver injury by increasing the microbial diversity and relative abundance of probiotics.