Commercial DSS (molecular weight 36–50 kDa) used for animal model was obtained from MP Biomedicals (Irvine, CA, United States). HPLC grade methanol was bought from Fisher Scientific (Fair Lawn, NJ, United States). The sartorius water purification system (arium®mini, Gottingen, Germany) was applied to gain ultrapure water. Besides, pyridine and chloroform were GC grade from China National Pharmaceutical Group Corporation (Shanghai, China). The preparation of S. boulardii (strain number is CICC 1903) was provided by lyophilization from Angel Nutritech Co., Ltd. (Hubei, China. 2,017,081,701), which was suspended in double-distilled water and administered orally by intragastric.

All procedures were strictly executed in the light of the international guidelines for the ethical use of laboratory animals and the guidelines of Animal Management Rules of the Ministry of Health of the People’s Republic of China (documentation Number 55, 2001, Ministry of Health of PR China). Moreover, all animal care and experimental procedure were approved by the Animal Care and Use Committee (ACUC) at Tongji Medical College, Huazhong University of Science and Technology, China (S2135).

It is reported that UC commonly affects young females (Ooi et al., 2010). Thus, female C57BL/6 mice, weighing 20 to 24 g (8 weeks old), purchased from Beijing Hua Fukang Laboratory Animal Technology Co., Ltd. (Beijing, China), were maintained in specific pathogen-free (SPF) condition with ambient temperature (25 ± 2°C) and 55% humidity, under a 12:12 h light:dark cycle, Animals were divided into three groups of equal size (n = 18) in a randomized allocation after one-week acclimatization: Control (non-colitis) group, DSS (colitis) group and (Sb + DSS) group. Distilled water or S. boulardii suspension [the dose set at 10⁶ Colony-Forming Units (CFU)/kg/day] was administered orally by intragastric for 21 consecutive days. Experimental colitis was duplicated in mice by adding 2.5% (wt/vol) DSS to drinking water for 7 consecutive days from the 22nd day (Supplementary Figure S1). Mice were housed in a plastic cage with wood shaving, with standard chow and distilled water freely available. We monitored each mouse daily for these indicators such as body weight, presence of hematochezia and the stool consistency. The estimate of weight loss is the variance between the original weight (day 0) and the weight on any given day. The disease activity index (DAI) was calculated in conjunction with scores of body weight loss, diarrhea and blood from the 21th to the 30th day to evaluate the severity of colitis, according to the Cooper method with slightly modified (Cooper et al., 1993). Once sacrificed, blood was gathered directly by cardiac puncture from the inferior vena cava with a heparinized tube. All measurements were performed blindly. The serum, colonic tissue and fecal samples were collected and stored at −80°C for further detection.

After fixation in 4% paraformaldehyde solution, the distal segments of colon (1–2 cm from anal margin) were prepared and taken for hematoxylin and eosin (H&E) staining once embedded in paraffin wax. Colonic histological damage was assessed based on tissue damage and cell infiltration. Each section was randomly selected with an inverted microscope in six fields (magnification ×200) and blindly evaluated under a light microscope (Olympus, Japan) by two pathologists. The Standard for pathological scoring was shown in Supplementary Table S1.

Serum contents of high mobility group box 1 protein (HMGB1), tumor necrosis factor α (TNF-α), interleukin-1β (IL-1β), IL-6, and monocyte chemotactic protein-1 (MCP-1) were thoroughly detected with commercial enzyme-linked immunosorbent assay kits (ELISA) (Neobioscience, China) against the manufacturer’s specifications. All measurements were performed in duplicate.

The colon tissues of mice (6 mice each group) were fixed and permeabilized in methanol at-20°C. Subsequently, the sections were stained at 4°C with primary antibodies ZO-1 (1:100) and Occludin (1:100), next to the incubation with secondary antibody for fluorescent labelling. Nuclei were stained for 10 min with mounting medium containing 4,6-diamidino-2-phenylindole (DAPI) (Roche, Switzerland). Images selected from the same area of tissue section were observed by fluorescence microscopy (Olympus, Japan).

The genomic DNAs were extracted from fecal samples by a commercial DNA stool kit (D4015, Omega, Inc., United States) and quantified by NanoDrop 3,300 fluorospectrometer (Thermo Scientific, Wilmington, DE) as directed. Total DNAs were detected by PCR (LC-Bio Technology Co., Ltd., China) after eluted in 50 μL Elution buffer. For PCR amplification, each amplicon of the 16S rRNA genes V3-V4 region set through the forward primer 338Forward (5’-ACTCCTACGGGAGGCAGCAG-3′) and reverse primer 806Reverse (5’-GGACTACHVGGGTWTCTAAT-3′) with slight modification. Successful amplicons were purified using AMPure XT beads (Beckman Coulter Genomics, Danvers, MA, United States) and quantified by Qubit (Invitrogen, United States). Moreover, assessments of the amplicon library size and quantity were performed on Agilent 2,100 Bioanalyzer (Agilent, United States) using the Library Quantification Kit for Illumina (Kapa Biosciences, Woburn, MA, United States). In addition, PhiX Control library (V3) (Illumina) was combined with the amplicon library (expected at 30%), which was sequenced on 300PE MiSeq. The sequenced reads and operational taxonomic units (OTUs) were applied to assign bacterial taxonomic classification, with a 97% threshold of pairwise identity. The α-diversity was assessed by Mothur to discriminate significantly different species between groups. As to evaluate the β-diversity, several unsupervised multivariate statistical methods including principal coordinate analysis (PCoA), non-metric multidimensional scaling (NMDS) analysis and principal components analysis (PCA) were operated by QIIME2. We deposited the original sequencing data in a NCBI BioProject with the accession number PRJNA639324. To identified differences in biologically relevant taxonomic biomarkers among groups, linear discriminant analysis effect size (LEfSe) was carried out. We further generated abundance-based correlation networks through enriching different group bacterial genus to detect their correlation. Pearson correlation coefficients were calculated to identify bacteria-metabolite correlations. Important relationships were identified as |Rho| ≥ 0.5 and p-values <0.05, between the centered log-ratio-transformed bacterial genus and correlative metabolites. Differences in metabolite concentrations between 2 phenotypes were detected by Student t-tests, where the cut-off value of VIP > 1.0 screened as statistically significant.

After precooled with 50% methanol and incubation, approximate 20 μL extraction mixture was gained and preserved overnight at −20°C. All metabolites were collected by the liquid chromatography–tandem mass spectrometry (LC–MS/MS) system followed machine orders. In brief, an ultra-performance liquid chromatography (UPLC) system (SCIEX, United Kingdom) was used to separate all chromatographicions, followed by an ACQUITY UPLC T3 column (100 mm × 2.1 mm, 1.8 μm, Waters, United Kingdom) used for the reversed phase separation. The collected MS data preprocessing were analyzed by a software of XCMS, which identified every ion according to integrating retention time (RT) with m/z data. As a result, outlier detection and batch effects evaluation were conducted through PCA in view of the pre-processed dataset.

Data are expressed as means ± SEM. One-way analysis of variance (ANOVA) followed by Dunnett’t test were performed to reveal the differences among groups. Correlation analysis was conducted to achieved the pairwise relationships among the secondary metabolites obtained by metabolomics, the significantly different genera acquired by 16S sequencing analysis, and inflammatory parameters. Statistical analyses were carried out by SPSS 25.0 software (SPSS Inc., Chicago, IL, United States) with p < 0.05 served as significance. Difference between metabolites and OTU were conducted in R version 3.5.2 using QIIME2.

Compared to the control, DSS administration induced remarkable reductions in body weight (p < 0.05, Figure 1A) and colon length (p < 0.05, Figure 1B), caused increased colon weight/length ratio (p < 0.05, Figure 1C) and led to higher DAI scores in mice from the seventh day (p < 0.05, Figure 1D). Treatment with S. boulardii alleviated the colon shortening and body weight loss, lowered the ratio of colon weight/length, and gradually attenuated the elevated DAI scores in DSS-induced colitis mice (all p < 0.05). Histological examinations further revealed that in contrast to the healthy colon tissues of control mice, DSS administration caused obvious distortion of crypt epithelium, severe mucosal damage, colonic epithelial cell apoptosis and inflammatory cells infiltration into the mucosa and sub mucosa. However, these were relieved along with S. boulardii treatment (p < 0.05, Figure 1E). Moreover, the TUNEL assay revealed that S. boulardii treatment effectively inhibited cell apoptosis triggered by DSS in colonic tissue, indicating a strong prevention of S. boulardii against cell death in colitis mice (p < 0.05, Figure 1E).

Given the above histopathological changes, we next evaluated the impact on the intestinal barrier integrity. Compared to control, tight junction components Occludin and ZO-1 were markable reduced in DSS-fed mice revealed by immunofluorescence analysis (both p < 0.05, Figure 1F). Whereas S. boulardii treatment improved the epithelial integrity by enhancing tight junction proteins in colitis mice induced by DSS.

For the sake of exploring the potential pharmacological effects of S. boulardii on inflammation in colitis, MPO, which is a typical inflammatory indicator of colitis reflecting the neutrophil counts, was measured. As depicted in Figure 1G, MPO activity was more obviously in mice administered DSS than the control (p < 0.05). Moreover, the serum pro-inflammatory factors including HMGB1, TNF-α, IL-6, IL-1β and MCP-1 were dramatically increased after DSS administration (all p < 0.05), while significantly reduced as S. boulardii treatment (all p < 0.05). Altogether, these results reveal that S. boulardii can mitigate the pathogenic symptoms of DSS-induced colitis, suggesting the preventive effects of S. boulardii on experimental colitis in mice.

Overlapping OTU data from the Venn diagram displayed that the control mice existed the highest amount of microbes (2846), whereas this number in DSS-fed mice was minimum (1901). Moreover, S. boulardii-treated mice shared more overlapping microbes with the control than mice fed with DSS (Figure 2A). The α-diversity indices of Chao1, ACE, Shannon and Simpson were analyzed utilizing OTU species and abundance to dissect species diversity in samples. In general, Chao1 and ACE proxied for species abundance, meanwhile Shannon and Simpson indicated for community diversity, respectively. As shown in Figure 2B, mice administered DSS displayed a significant reduction in all α-diversity indexes compared to the control animals, while S. boulardii treatment increased the community richness but not community diversity in DSS-induced colitis mice. Overall species richness was evaluated through the rarefaction curve of the observed OTUs. Richness in S. boulardii-treated colitis mice was close to that in the controls, but higher than that in mice fed with DSS (Figure 2C). The dominant bacterial communities at the phylum level were Bacteroidetes, Firmicutes and Proteobacteria in control mice. However, the gut microbial profile was drastically changed by DSS feeding, with an obvious improvement in the relative abundance of Firmicutes, Verrucomicrobia and a dramatical diminish in Bacteroidetes, resulting in a diminished Firmicutes/Bacteroidetes ratio. S. boulardii administration prevented the DSS-driven reduction in microbial richness and restored their levels similar to those in control mice (Figures 2D,E).

Taking all available sequences into account, comparative analysis revealed a prominently high level in β-diversity (heterogeneity) among groups, suggesting the different variations in response to DSS feeding and S. boulardii treatment (Figures 3A–C). Relative abundance performed by Metastats analysis exhibited that enrichment of the Verrucomicrobia phylum, and the Turicibacter, Alistipes, Clostridium and Bacteroides genera in mice after DSS treatment was the major differences as compared to controls. While S. boulardii treatment fully prevented a tendency towards expansion of the microbial communities and, to some extent, restored to the control levels (Figure 3D). Moreover, the LEfSe approach identified Bacteroides, Akkermansia, Verrucomicrobia, Erysipeiotrichaceae, Alistipes, Turicibacter, along with a major decrease in family Porphyromonadaceae as the major feature discriminating fecal bacterial communities of DSS-induced colitis mice from that of control animals (Figures 3E,F). Importantly, higher abundance of Porphyromonadaceae was observed in S. boulardii-treated colitis mice when compared with colitis mice induced by DSS (Figures 3G,H), which in turn, supporting the restorative effect of S. boulardii on microbiota dysbiosis.

The representative ion current chromatograms of LC–MS disclosed high reproducibility and small retention time drift. Based on chemical ontologies and structure similarity, 436 and 186 peaks were distinguished and assigned to corresponding metabolites in positive and negative ion mode, respectively (Figures 4A,B). A schematic overview of the identified metabolites was classified into nine major groups: amino acid, carbohydrate, xenobiotic, vitamin and cofactor, terpenoid and polyketide, lipid, nucleotide, energy and membrane transport related metabolites (Figures 4C,D).

To describe the dispersion tendency among groups, the PCA was performed and the score plot described an obvious distinction between the colitis mice and the control, similarly with the S. boulardii supplementation groups (Figure 4E), indicating striking differences in endogenous metabolites within DSS group and other two groups. It is worth noting that the control and S. boulardii group were not fully separated from each other. Subsequently, we employed PLS-DA to further investigate metabolic variations and differential metabolites among groups. Compared to the control, the DSS mice plot showed a markedly variation in the PLS-DA score, characterized by R² = 0.9997 and Q² = 0.9923 (Figure 4F). Likewise, a distinctly difference was also observed between the S. boulardii supplementation and the DSS groups (R² = 0.9988 and Q² = 0.9811, Figure 4G). These plots supplied additional evidence of distinguish in metabolic profiles of DSS-induce colitis and S. boulardii treatment.

Specifically, comparison between the DSS-fed mice and the control mice revealed that 58 metabolites were identified as discriminant biomarkers, with 27 enriched and 31 depleted (all p < 0.05, Table 1). Moreover, 40 metabolites were significantly changed by S. boulardii treatment in contrast to those in the DSS group, with 21 enriched and 19 depleted (all p < 0.05). Representative metabolites are visualized in Figure 4H. DSS treatment altered the contents of metabolites involved in the tricarboxylic acid (TCA) cycle and protein metabolism, including _L-glutamic acid, _L-tyrosine, citric acid, hydroxypyruvic, tryptophan, arginine and histidine. A hierarchical clustering analysis with selected variant metabolites was run, and the resultant heatmaps presented the relative abundance of discrepant metabolites and relevant metabolic profiles (Figures 4I,J).

Apart from taxonomic composition, differential functional pathways among groups were outlined by MetaboAnaylst based on enrichment and topology analysis (Figures 5A,B). DSS administration affected six pathways, while S. boulardii supplementation influenced seven pathways (Supplementary Table S2). Of which, the relevant pathways of synthesis and metabolism, involving arachidonic acid metabolism, linoleic acid metabolism, tyrosine metabolism, phenylalanine, tyrosine and tryptophan biosynthesis, turned out highly consistent (Figure 5C). Based on the calculation results, the final interplay network diagrams among the common metabolite pathways that links with several highly significant biological pathways were also drew up (Figure 5D).

To better understand the relationships among the inflammatory parameters, gut microbiota and corresponding metabolites in DSS-induced colitis, spearman correlation analysis was carried out. Metabolites were selected according to the intensities within the top 30 expression levels. As shown in Figure 6A, it was found that these 30 gut flora genera appeared either negative or positive association with at least one parameter of inflammation. Thereinto, Bacteroides, Akkermansia, Clostridium, Alistipes, Turicibacter were positively correlated with parameters that promote colitis, while Lactobacillus showed negative correlations. Accordingly, fifteen metabolites (Fentanyl, Hypoxanthine, Xanthine, Deoxycholic acid, 7-Ketodeoxycholic acid, Glutamic acid, Nutriacholic acid, Oxypurinol, Tyrosine, Nicotinic acid, Allopurinol, Mangiferic acid, Ursocholic acid, Cavipetin C, and Valyl-Leucine) were negatively, while seven metabolites (Myricolal, Soyasapogenol E, Isoferulic acid, Linoleoyl ethanol amide, linoleic acid, 5B-Cyprinol sulfate, and Ecabte) were positively correlated with these inflammatory cytokines. In addition, correlations between intestinal flora genera and metabolites were also observed. Wherein, strong correlations were shown in Supplementary Table S3 with |rho| > 0.8 and p < 0.05. It’s worth noting that the associations of citric acid with Turicibacter and Clostridium were negative, but positive with Lactobacillus and Lactococcus. Tryptophan was negatively associated with Lactobacillus, Lactococcus and positively correlated with Turicibacter and Clostridium. Moreover, Riboflavin was negatively correlated with Bacteroides, Turicibacter, Clostridium and Alistipes (Figure 6B).