In this study, the experimental design and procedures in this study were reviewed and approved by the Committee of the Institute of Subtropical Agriculture at the Chinese Academy of Sciences. All the experiment operations were conducted following the guidelines of the institute of Subtropical Agriculture on Animal Care, Chinese Academy of Sciences (No. ISA-2020-18).

Eucommia ulmoides leaves (EL) were grinded and filtered with 2 mm screener. The EL powder (500 g) was extracted by soaking twice in 5 L 95% alcohol for 7 days at room temperature, and the collected supernatant was filtered and concentrated using vacuum evaporator. Subsequently, the EL concentrate was decolorized with petroleum ether and then rotated again, then the extract of ELE was obtained by vacuum freeze-drying.

The content of total flavonoid and total phenolic acids in ELE was determined by using the plant flavonoids test kit (A142-1-1, Nanjing Jiancheng Bioengineering Institute, Nanjing, China) and plant total phenols test kit (A143-1-1, Nanjing Jiancheng Bioengineering Institute). The content of sugar was determined by using the plant total sugar and reducing sugar test kit (Shanghai Yuanye Bio-Technology Co., Ltd, Shanghai). All operations are in accordance with the relevant instructions. The content of total flavonoid, total phenolic acids, total soluble sugar, and reducing sugar in ELE was 44.59, 89.27, 104.23, and 25.81 mg/g, respectively.

The small molecule metabolites in ELE were detected by using liquid chromatography tandem secondary mass spectrometry (LC-MS/MS). The instrument parameters are as follows: LC was conducted by using UltiMate 3000 (Thermofisher, Waltham, MA, United States) and equipped with C18 chromatographic column (the column temperature is 18°C during operation). Mobile phase was A: 100% acetonitrile + 0.1% formic acid, B: H₂O + 0.1% formic acid. Elution gradient: A: 0−7 min, 5−50%; 7−8 min, 50−75%; 8−9 min, 75−80%; 9−11 min, 80−90%; 11−15 min, 90−95%; 15−20 min, 95%. Velocity of flow: 0.2 ml/min. MS/MS was conducted by using Q-Exactive Focus (Thermofisher, Waltham, MA, United States). Ion source: ESI source, atomization temperature: 300°C, atomization gas (sheath gas) pressure: 40 ARB, auxiliary gas pressure: 10 ARB; transmission capillary temperature: 3p 20°C, scanning mode: (1) full scan, resolution 35,000, in source CID: 0 ev; ddms2, resolution 17,500, HCD-stepped NCE 10, 30, and 50. All the reagents for LC-MS were purchased from Sigma-Aldrich; the mass spectrum data were compared with the mzVault and MassList database. The relative content of metabolites was reflected by peak area. The relative content of metabolites was reflected by peak area. The relative abundance of the small molecular components is presented in peak area relative to the total peak area. Geniposide and chlorogenic acid accounted for 12% and 10%, respectively. The top 30 small molecular metabolites are shown in Supplementary Table 1.

C57BL/6 mice (11 weeks of age) were purchased from STJ Laboratory Animal Co., LTD (Hunan, China). The mice were housed in four per cage under the same condition (temperature, 25 ± 1°C; lighting cycle, 12 h:12 h light/dark; 8:00−20:00 for light) and had free access to food and drinking water. Dextran sulfate sodium (36,000–5,000 MW) was purchased from MP Biomedicals (Santa Ana, CA, United States) and dissolved in drinking water to 2% (w/v) and given ad libitum to mice beginning on day 0 for 7 days (Tanaka et al., 2003). EL extract were dissolved in distilled water. The mice were then gavage with distilled water (CON and DSS group), 200 and 400 mg/kg (according to the body weight) ELE for 11 days (D5–D8). All the mice were evaluated in terms of body weight change and diarrhea index every day.

At the end of this experiment, all the mice were fasted for 6 h, and anesthesia was induced by intraperitoneal injection of 2% pentobarbital sodium (45 mg/kg body weight). The blood samples were collected in 1.5 ml centrifuge tubes (germ-RNase- and DNase-free) after enucleation of the eyeball. The distal ileum (near the cecum to 2 cm) and colon were collected and stored in 4% neutral polyformaldehyde fixative or frozen in liquid nitrogen for real-time quantitative PCR (RT-qPCR) or western blotting. The cecum contents were collected in sterile enzyme free centrifuge tube and frozen in liquid nitrogen, and then stored at −80°C until total genomic DNA extraction. The genomic DNA was used for subsequent 16S rRNA gene sequencing analysis.

The colonic tissue was embedded in paraffin, cut into 5 μm sections, and stained with hematoxylin and eosin (H&E). The colon crypt depth was observed by microscope (Olympus, Japan), and the depth of crypt was measured using VistarImage (Olympus, Japan) software, matching with the Olympus microscope (Zhai Z. et al., 2019).

To evaluate the inflammatory and colon epithelium integrity, we assessed the expression level of genes related to inflammation, tight junction, and bile acid receptor. RNA was extracted from colon tissue using column RNA extraction kit (R4121, Magen, Guangzhou, China). The total RNA concentration was measured using a NanoDrop 2000C spectrophotometer (Thermo Fisher Scientific, Waltham, MA, United States). The total RNA was reverse transcribed into cDNA by using cDNA synthesis kit (CW2582M, CWBIO, Jiangsu, China). The RT-qPCR primers were designed and synthesized by a commercial service of Sangon Biotech (Shanghai, China); the primer sequences are shown in Supplementary Table 2. The RT-qPCR was then performed according to the direction of FastSYBR Mixtrue (CW0955M, CWBIO) and run in the ABI 7500 FAST system (Applied Biosystems Instruments, Thermo Fisher Scientific, United States). The values of the target genes were normalized by the housekeeping gene.

The process of genomic DNA extraction, quality detection and sequencing operation of intestinal microbiota were performed by the methods described in our previous study (Zhai Z. et al., 2019). Briefly, the sequencing raw data was obtained the Ion S5^TM XL platform. The raw data was measured by Cutadapt (version 1.9.1) to obtain the clean data. Operational taxonomy units (OTUs) were clustered with 97% identity by using Uparse (v7.0.1001). The non-metric multidimensional scaling (NMDS) were conducted based on OTU by the using R software. OTU is annotated and divided into phylum, class, order, family, genus, and species. The linear discriminant analysis (LDA) effect size (LEfSe) was used to elucidate the differences among bacterial taxa. An LDA score ≥4 was considered to be an important contributor to the model. The Spearman’s analysis was used to measure the correlation between gut microbiota and bile acids.

The BA concentration in serum was measured by LC-MS. The standard curves of 25 BAs were established (Supplementary Table 3). Internal standards (Supplementary Table 4) were added into 50 μl serum samples and concentrated in a concentrator. After concentration, samples were redissolved with 100 μl 50% methanol-H₂O solution (v:v). LC was conducted by using LC30AD (Shimadzu, Japan) and equipped with C18 chromatographic column. Mobile phase was composed of solution A: acetonitrile (containing 0.01% acetic acid) and solution B: H₂O (0.01% acetic acid and 5 mmol/L ammonium acetate). Elution gradient is shown in Supplementary Table 5. MS was conducted by using QTRAP 6500 (ABSCIEX, MA, United States). Ion source: ESI source, scanning mode: negative ion scanning, Sperry voltage: 4,500 V (negative), capillary temperature: 550°C, curtain gas: 35 psi. The content of bile acid in blood was calculated by standard curve.

Human colon cell line Caco-2 was kindly provided by Professor Baichuan Deng (College of Animal Science, South China agricultural University). The cell was cultured in DMEM/F12 (1:1) medium supplemented with 10% fetal bovine serum (FBS) (Gibco, Thermo Fisher, Waltham, MA, United States) and then incubated at 37°C in a humidified atmosphere containing 5% CO₂ in air. For transwell experiment, 1 × 10⁶ cells were seeded in 0.32 cm² (0.4 μm) transwell plate and detect the TEER until differentiation is complete (Xiaojun et al., 2020). The cells were then treated with 40 μg/ml LPS (derived from Escherichia coli, O111: B4, Beyotime Biotechnology, Shanghai, China) for and treated with 12.5 μM TGR5 agonist INT-777 (I884046, ≥ 99%, MedChemExpress, Shanghai, China) at the same time, then the TEER was detected at 24h and 48h. For TGR5 RNA interference assay, the operation was carried out according to the kit instructions. In brief, negative control (NC) and TGR5-siRNA were mixed with serum-free DMEM/F12 medium and Lipo8000^TM (Beyotime Biotechnology) and incubated at room temperature for 20 min. The mix was then added to each well and treated for 24 and 48 h. The NC and TGR5-siRNA sequences were designed and synthesized by (Sangon Biotech) and the sequence were shown in Supplementary Table 6. The FITC-dextran (M.w.10000, MACKLIN, Shanghai, China) permeability of monolayer cell detection has been described in a previous study (Zhai et al., 2018). Briefly, 200 μl FITC-dextran solution (dissolved in HBSS, 5 mg/ml) was added to the apical compartment of transwell plate and 500 μl HBSS were added to basal compartment and cultured for 2 h. The absorbance in the basal compartment was measured using microplate reader (Spark10M, TECAN, Männedorf, Switzerland), and the content of FITC-dextran was calculated relative to a standard curve.

For the experiment, 1 × 10⁶ cells were seed in cell culture dish (6 cm diameter, JET, Guangzhou, China). After 24 h of being seeded, the cells were treated with LPS, LPS + INT-777, and INT-777 for 24 h. Caco-2 samples were lysed in Radio Immunoprecipitation Assay Lysis buffer (RIPA) lysate buffer (containing 1 mM PMSF) on ice for 15 min and then collected. RIPA buffer and PMSF are purchased from Beyotime Technology (Shanghai, China). The total protein concentration was detected by using BCA protein assay kit (Pierce BCA Protein Assay Kit, Beyotime technology) and mixed with loading buffer, and the protein was denatured in a metal bath (95°C, 5 min). In electrophoresis, 60 μg protein samples were loaded in each sampling well. After electrophoresis and membrane transfer, the primary and secondary antibodies were incubated, and electrochemiluminescence (ECL) was detected. The primary and secondary antibodies were shown as follows: claudin-1 (51-9000) was purchased from Thermo Fisher Scientific. TGR5 (NBP2-23669) were purchased from NOVUS (CO, United States). Goat antimouse (bs-40296G-HRP) and goat anti-rabbit (bs-0295G-HRP) antibodies were purchased from Bioss (Beijing, China). In order to detect proteins with similar molecular weight, we used striping buffer to remove antibodies, and then performed blocking, first antibody and second antibody incubation, and ECL chemiluminescence detection. The gray value of protein bands was measured by ImageJ (National Institutes of Health, Germany). The relative protein expression was expressed by the ratio of target protein to β-actin.

Data are expressed as the mean ± SEM. One-way ANOVA and the least significance difference (LSD) method were used to determine the differences among the groups by using SPSS 20.0 (IBM, SPSS, United States) with significant criteria set to ^∗p < 0.05, ^∗∗p < 0.01, and ^∗∗∗p < 0.001. GraphPad Prism 7 (GraphPad Software Inc., San Diego, CA, United States) was used to generate statistical plots.

To evaluate the therapeutic effect of ELE on IBD, we constructed IBD mouse model using 2% DSS in drinking water for 1 week and administered different doses of ELE by gavage for 10 days, as shown in Figure 1A. Compared with the DSS group, ELE did not promote the recovery of body weight loss (Figure 1B) but remarkably attenuated the diarrhea status (Figure 1C, p < 0.05). In addition, oral administration alleviated DSS-induced splenic hypertrophy (Figure 1D).

The effects of oral ELE administration on colon length and the epithelium morphology were evaluated. ELE supplementation alleviated the DSS-induced shortening of colon length (p < 0.05, Figures 2A,B) and crypt depth (p < 0.05, Figures 2C,D). Furthermore, ELE administration downregulated inflammation-related gene expression in the colon, including TLR4 and IL-6, which were more highly expressed in the DSS-treated group (Figure 2E).

Oral administration of 200 and 400 mg/kg ELE showed similar effects in the preliminary results, thus 200 mg/kg was chosen for assessment in the following experiments. The supplementary effect of 200 mg/kg ELE on the gut microbiota in DSS-induced mice was further investigated. The rank abundance and rarefaction curve showed that the selected sequences were sufficient to determine most bacterial diversity (Supplementary Figure 1). The NMDS plot showed a clear separation of the gut microbiota among the control, ELE, and DSS groups (Figure 3A). The Venn plot showed that there were 1,862, 2,023, and 1,807 OTUs in control, ELE, and DSS groups, respectively (Figure 3B). The observed species (OS) and Shannon, Simpson, Chao1, and ACE indexes showed that ELE supplementation recovered the DSS-induced decline in OS and Chao1 index-dependent richness and Simpson and Shannon index-dependent diversity of the microbiota (Figure 3C). The phyla Firmicutes, Proteobacteria, Bacteroidetes, Verrucomicrobia, and unidentified_Bacteria were the microbiota with the top 5 relative abundances (Figure 3D). Compared with the DSS group, ELE increased the relative abundance of Verrucomicrobia. At the family level, DSS significantly reduced the relative abundance of Lactobacillaceae and increased the levels of Bacteroides and Marinifilaceae compared with the control group. After ELE treatment, the relative abundances of Burkhoderiaceae, Akkermansiaceae, and Ruminococcaceae were increased, while those of Erysipelotrichaceae and Bacteroidaceae were decreased compared with those in the DSS group (Figure 3E). LEfSe analysis demonstrated that Bacteroidetes, Marinifilaceae, and Helicobacteraceae were the main features in IBD mice. Akkermansia belonging to the Verrucomicrobia phylum was a feature in ELE-administrated mice. However, compared with the control group, the relative abundances of Lactobacillus, Dubosiella, and unidentified_Ruminococcaceae belonging to Firmicutes were significantly decreased in IBD- and ELE-treated mice (Figure 3F).

The content of bile acids in the serum of mice was analyzed by LC-MS. ELE administration reduced the DSS-induced changes higher in total bile acids (p < 0.05) and tended to reduce the DSS-induced changes higher in total primary bile acids (0.05 < p < 0.1) in the serum (Figures 4A,B). Compared with the control group, ELE administration did not recover the DSS-induced changes of total secondary bile acids (Figure 4C) or the P/S ratio (Figure 4D).

Furthermore, we analyzed the composition of bile acids in the serum of mice in each group. Compared with the control group mice, DSS significantly increased the content of CA and β-MCA and decreased the DCA and ω-MCA content (p < 0.05, Figures 4E,F). EL extract treatment restored CA, β-MCA, and DCA to normal levels. Additionally, compared with the DSS and control groups, ELE significantly increased the contents of TCA and TUDCA (p < 0.05). The intestinal microbiota can regulate the process of bile acid synthesis, uncoupling, and secondary bile acid production in mice. We analyzed the correlation between bile acid content and the relative abundance of intestinal microbiota (Figure 4G). Consistent with the above results, in normal mice, unidentified_Ruminococcaceae, Lactobacillus showed a negative correlation with primary bile acids, such as β-MCA, Tβ-MCA TCDCA, CA (−0.55 < R² < −0.40). Bacteroides showed positive correlation with primary bile acids such as Tβ-MCA and CDCA (0.4 < R² < 0.5),while with negative correlation with DCA and TDCA (0.4 < R² < 0.5). In addition, Dubosiella showed strong positive correlation with α-MCA, UDCA and ω-MCA (R² = 0.54, 0.62, and 0.51). The feature taxon Akkermansia in ELE treated mice showed strong positive correlation with TUDCA (R² = 0.53).

Based on the change in bile acid content in serum, we further detected the mRNA expression levels of TGR5, a bile acid receptor and tight junction protein in the mouse colons. The data showed that, compared with DSS treatment, ELE treatment significantly increased the decreased mRNA levels of TGR5, claudin-1, and occludin (p < 0.05, Figures 5A–C) while had no significant effect on ZO-1 (Figure 5D). Immunofluorescence results showed that claudin-1 was expressed from the top to the bottom of the crypt in the control group. Dextran sulfate sodium administration decreased this expression due to the destruction of intestinal epithelial structure. After ELE treatment, the expression of claudin-1 gradually increased from the bottom to the top of the crypt, suggesting that the expression and distribution of claudin-1 are involved in the process of recovery (Figure 5E).

In this study, we used Caco-2, a human colon epithelium cell line, to confirm the role of bile acid receptor TGR5 in alleviating the process of colon epithelium injury. To activate TGR5, a selective agonist INT-777 was used. The results showed that INT-777 prevented the LPS-induced decrease in the TEER value in the Caco-2 monolayer cell barrier and increase in FITC-dextran permeability (Figures 6A,B). Western blotting data showed that INT-777 significantly increased LPS-induced reductions in claudin-1 protein level (Figures 6C,D), suggesting that INT-777 can promote the barrier function of colon cells. TGR5-siRNA was then used for TGR5 mRNA interference and TEER and FITC-dextran permeability were also evaluated after treatment for 24 and 48 h. The results showed that interference with TGR5 expression could significantly decrease TEER and increase the permeability of FITC-dextran in monolayer cells (p < 0.05, Figures 7A,B). Western blotting results showed that, TGR5 interference significantly decreased claudin-1 expression (p < 0.05, Figures 7C,D).