Twelve male 8-week-old C57BL/6J mice and 12 female 8-week-old FXR^−/− mice were randomly divided into four groups of six mice: Wild-type control group (WT), FXR control group (FXR^−/−), wild-type model group (WT + ANIT) and FXR model group (FXR^−/− + ANIT). The mice in WT and FXR^−/− groups were treated with vehicle (olive oil), and the mice in the WT + ANIT and FXR^−/− + ANIT groups were treated with ANIT (75 mg/kg) by oral gavage. After 48 h, the mice in each group were fasted for 12 h on the night prior to experimental sample collection. Blood samples (∼0.6 mL for each mouse) were collected from the infraorbital venous plexus of mice and were separated by centrifugation at 3,000 rpm for 10 min for untargeted metabolomics analysis. The ileum and the part of the large hepatic lobe were fixed with 10% formaldehyde for histopathological examination. The intestinal contents were collected and immediately snap-frozen in liquid nitrogen for 16s rRNA gene sequencing analysis. All animal experiments performed in this study were approved and supervised by the ethics committee of the fifth medical center of PLA general hospital (Beijing, China).

The serum concentrations of aspartate transaminase (AST), alanine transaminase (ALT), total bile acid (TBA), and direct bilirubin (DBIL) were measured according to the manufacturer’s protocols.

The liver and ileum tissues of mice obtained from each group were fixed in 4% paraformaldehyde, embedded in paraffin, and stained with hematoxylin and eosin (H&E). Pathological photographs were obtained using a microscope (magnification ×200 and 400×). The hepatic HE scores were assessed according to previous research (Gong et al., 2021). Briefly, the hepatic histological parameters of inflammation, necrosis, and hepatocyte vacuolization were all assessed on a scale of 0–3 (0 defined as absent; 3 defined as severe). Chiu’s score was used to evaluate the level of ileum injury (Li et al., 2020).

To prepare samples for the untargeted metabolomics analysis, frozen serum samples were allowed to thaw at room temperature. 200 μL of each serum sample was mixed with 600 μL of 100% methanol and was then centrifuged for 10 min at 48,000 rpm. A total of 500 μL supernatant was transferred to a new 1.5 mL tube and dried with a nitrogen blower. The dried samples were reconstituted with 500 μL of 50% acetonitrile methanol. Dissolved samples were filtered into the sample bottle with a 0.22-μm membrane for Q-TOF/MS analysis as described previously (Wei et al., 2018).

Data analysis was performed within Mass Hunter Profinder software (Agilent) for post-acquisition data processing. Then, the MetaboAnalyst database was used to normalize the original data. Multivariate data analysis was performed with SIMCA 14.0 (Umetrics, Sweden). OPLS-DA analysis based on non-targeted metabolomics was conducted in the serums between WT and WT + ANIT group, and WT and FXR-KO group (n: WT = 6, WT + ANIT = 6, FXR^−/− = 6). The endogenous metabolites with VIP >1.5 and |p(corr)| ≥ 0.58 were selected as the differential biomarker for further pathway enrichment analysis. MetaboAnalyst and DAVID database were used to analyze the metabolic pathway and KEGG pathway. Metabolite targets were collected through the MBrole database (http://csbg.cnb.csic.es/mbrole2/) and applied to the Sangerbox database (http://sangerbox.com/) for GO enrichment analysis and the Metascape database (https://metascape.org/gp/) for protein interaction analysis.

The total DNA of intestinal bacteria was extracted from intestinal contents using HiPure Stool DNA Kit B according to the manufacturer’s instructions. The DNA extractions were quantified by ultraviolet spectroscopy and amplified using universal primers of 341F: 5′-CCTACGGGNGGCWGCAG-3′ and 806R: 5′-GGACTACHVGGGTATCTAAT-3′ to target the V3–V4 region of bacterial 16s rRNA. The raw data obtained were then analyzed as described previously (Guo et al., 2017).

Statistical analyses were performed with SPSS 19.0, All experimental data were expressed as the mean ± standard deviation. The differences between the group means were calculated by one-way analysis. p < 0.05 were considered statistically significant.

As shown in Figure 1, ANIT caused significant increases in serum ALT, AST, TBIL, and TBA levels in WT and FXR^−/− mice. Compared with WT + ANIT mice, no significant change in serum ALT, AST, TBIL, and TBA levels in FXR^−/− + ANIT mice.

Histological analysis of the liver and ileum provided direct evidence of FXR function in cholestatic liver injury. As shown in Figures 2A, B, compared with WT mice, the WT + ANIT and FXR^−/− + ANIT mice showed enlarged gallbladders. The liver tissue of mice in WT + ANIT, FXR^−/−, and FXR^−/− + ANIT groups showed marked acute infiltration, hepatic necrosis, and hepatocyte vacuolization. Pathological changes in the mouse ileum suggested that the ileum in the WT + ANIT, FXR^−/−, and FXR^−/− + ANIT groups had a more intense yellow color compared with the mice in the WT group. And compared with the WT group, the mice in the WT + ANIT, FXR^−/− and FXR^−/− + ANIT groups had a shed, sparse, and low ileum villi combined with obvious inflammatory cell infiltration (Figures 2C, D).

To discover the regulatory mechanism of FXR on the pathogenesis of intrahepatic cholestasis, the effect of FXR on serum metabolic microenvironment in vivo was studied by metabolomics, and the signal network regulated by FXR was further tracked by FXR-related specific biomarkers. The PCA results showed that four clusters could be better distinguished (Figures 3A, B). OPLS-DA analysis of metabolite profiles also showed a global metabolic difference between WT and WT + ANIT groups (R2X(cum) = 0.431, R2Y(cum) = 0.998, Q2(cum) = 0.956), WT and FXR^−/− groups (R2X(cum) = 0.658, R2Y(cum) = 0.833, Q2(cum) = 0.334), FXR^−/− and FXR^−/− + ANIT groups (R2X(cum) = 0.621, R2Y(cum) = 1, Q2(cum) = 0.825), WT + ANIT and FXR^−/− + ANIT groups (R2X(cum) = 0.505, R2Y(cum) = 0.999, Q2(cum) = 0.775) in positive-ion (ESI+) mode (Figures 3C–F). Meanwhile, the OPLS-DA analysis had similar results in negative-ion (ESI-) mode (Supplementary Figure S1).

To further analyze the disturbance of the serum biomarkers in mice with intrahepatic cholestasis caused by FXR knockout, we conducted an in-depth analysis of the OPLS-DA models of serum samples between the WT and WT + ANIT groups, and between the WT and FXR^−/− groups. A scatter analysis diagram (Figures 4A, B and Supplementary Figure S2) was established using p and p (corr) parameters. The metabolites with VIP>1 and |p(corr)|>0.58 were regarded as the differential biomarkers of the WT vs. WT + ANIT and WT vs. FXR^−/− respectively. 35 differential serum metabolites were found in both the positive-ion mode and the negative-ion mode. Compared with the WT group, the expression of the above 35 biomarkers in the WT + ANIT model was almost significant (Figure 4C). In the scatter analysis data of WT vs. FXR^−/−, 31 differential serum biomarkers were found in both the positive-ion and the negative ion modes (Figure 4D).

Several differential biomarkers triggered by FXR deficiency and intrahepatic cholestasis induced by ANIT in WT mice were obtained. Considering the function of FXR in many diseases, the differential biomarkers shared by WT vs. WT + ANIT and WT vs. FXR^−/− were further screened by using a Venn plotter to explore the specific biomarkers associated with the pathogenesis of cholestasis caused by FXR knockout (Figure 5A). To this end, 8 differential biomarkers were further screened in serum. (Figure 5B; Table 1).

The results showed that 8 serum differential biomarkers related to FXR deficiency-induced intrahepatic cholestasis were enriched to 6 metabolic pathways (Figure 6A; Table 2), mainly glycophoric metabolism and primary bile acid biosynthesis. This result is consistent with the known function of FXR. To further reveal the upstream signals of the differential biomarkers related to FXR deficiency-induced intrahepatic cholestasis, the relevant targets of the above 8 differential metabolic markers were collected. As shown in Figure 6B, 131 targets have been collected and most targets can regulate each other.

To investigate the signal pathway regulated by FXR in causing intrahepatic cholestasis, KEGG and GO enrichment analyses were performed. As we expected, the KEGG enrichment analysis results showed that FXR knockout caused bile acid metabolisms and its upstream lipid metabolism-related signal pathways, such as bile secret, primary bile acid biosynthesis, ABC transporters, glycerophospholipid metabolism, and steroid hormone biosynthesis. In addition, other signals were also enriched, such as the GnRH signaling pathway (p = 1.36 × 10⁻⁴) and Fc gamma R-mediated phase (p = 0.02) signals (Figure 7A and Supplementary Table S1). The GO enrichment analysis results showed that the targets were mainly enriched in lipid metabolism (Figure 7B).

Alpha diversity analysis reflects the species abundance and species diversity. Alpha diversity was presented as the index of Chao and Sobs. The results showed that the Sobs and Chao values of the WT + ANIT group and FXR^−/− group (p < 0.05) were both downregulated compared with WT mice (Figures 8A–D). The results of PCoA, NMDSD, and PCA showed that the microbial community structure was significantly different between the WT and WT + ANIT (Figures 8E–H), as well as WT and FXR^−/− (Figures 8I–L). Box diagram analysis and Anosim test results showed a significant difference between the WT and WT + ANIT groups (Figure 8H), as well as WT and FXR^−/− groups (Figure 8L).

The relative abundance of microbial species between the WT and WT + ANIT groups, and between the WT and FXR^−/− groups at genus, phylum, and species levels was calculated and depicted by stacked bar plots (Figure 10). At the genus level, compared with the WT group, the relative abundance of Escherichia-Shigella in the WT + ANIT group (Supplementary Figures S3A, B) and Akkermansia in the FXR^−/− group increased significantly, while the relative abundance of Lactobacillus in the WT + ANIT and FXR^−/− groups was significantly lower (Supplementary Figures S3C, D). At the phylum level, the relative abundance of Proteobacteria in the WT + ANIT group (Supplementary Figures S3E, F) and Verrucomicrobia in the FXR^−/− group (Supplementary Figures S3G, H) increased significantly compared with the WT group, while the relative abundance of Firmicutes in the WT + ANIT and FXR^−/− groups was significantly lower. At the species level, the relative abundance of Parabacteroides_goldsteinii, Parabacteroides_distasonis, Enterococcus_faecalis, Bacteroides_vulgatus, Bacteroides_acidifaciens and Alistipes_finegoldii in the WT + ANIT group (Figures 9A, B) increased significantly compared with WT group, while the relative abundance of Lactobacillus_reuteri and Lactobacillus_johnsonii_FI9785 in the WT + ANIT and Lactobacillus_johnsonii_FI9785 and Lactobacillus_gasseri in the FXR^−/− group was significantly lower (Figures 9A–D).

Compared with the WT group, the main signals enriched by different bacterial flora in the WT + ANIT group were inflammatory signals, bile acid metabolism signals, steroid metabolic signals, and fatty acid metabolism signals (Supplementary Figure S4). While, in the FXR^−/− group, the main signals enriched by different bacterial flora were bile acid metabolism, steroid metabolism, sphingomyelin metabolism, and amino acid metabolism, such as steroid hormone biosynthesis, sphingolipid, secondary bile acid biosynthesis, protein digestion and absorption, phenylalanine metabolism, NOD-like receptor signaling pathway, and beta-Alanine metabolism (Supplementary Figure S5).

A correlation heatmap was constructed to evaluate the covariation between altered gut microbiota species and differential biomarkers. As shown in Figure 10C, the differential gut microbiota and biomarkers were significantly correlated. The result demonstrated that the changes in these differential biomarkers may be associated with gut microbiota disruption. According to the above research, we found that the Lactobacillus_johnsonii_FI9785 and Lactobacillus_gasseri were significantly lower in FXR^−/− group. Correlation analysis results also found that Lactobacillus_johnsonii_FI9785 and Lactobacillus_gasseri were negatively correlated with the differential biomarkers (HMDB0000067, HMDB0007851, HMDB0002639, and HMDB0006893). This result proves that HMDB0000067, HMDB0007851, HMDB0002639, and HMDB0006893 play prominent roles in cholestasis caused by FXR knockout which is associated with Lactobacillus_johnsonii_FI9785 and Lactobacillus_gasseri.