A CW–component–target network was constructed by active components with a cut-off of drug-likeness (0.18) and oral bioavailability (20%). All the plant names have been checked with http://www.theplantlist.org. Based on the database of Laboratory of Systems Pharmacology (Ru et al., 2014), the active components of SR and DC were collected. The active components of OS, Cb, GB, and FS were acquired, according to the published articles (Lee et al., 2010; Pan et al., 2016; Scandiffio et al., 2018; Guo et al., 2020; Kintsu et al., 2020). According to these active components, the targets were obtained from PubChem, and the genes of ulcerative colitis (UC) were collected from GeneCards (Stelzer et al., 2016), Online Mendelian Inheritance in Man (OMIM) (Amberger and Hamosh, 2017), DrugBank (Wishart et al., 2018a), PharmGkb (Barbarino et al., 2018), and Statistics of Therapeutic Target Database (TTD). Based on the commonly shared genes, Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) Pathway Enrichment analyses were conducted (The Gene Ontology Consortium, 2019). The herb–ingredient–target (HIT) interaction network and a protein–protein interaction (PPI) network with a confidence score ≥ 0.7 based on the Search Tool for Retrieval of Interacting Genes/Proteins database were computed by CytoNCA (Tang et al., 2015), which calculates the median values of betweenness centrality (BC), closeness centrality (CC), degree centrality (DC), local average connectivity (LAC), eigenvector centrality (EC), and network centrality (NC).

The CW formula was prepared by mixing the powder of SR, OS, GB, Cb, FS, and DC at a ratio of 7:4:3:3:2:1, respectively. Briefly, the FS powder was diluted in water and mixed with other herbs. The dried drug was smashed and given to mice or subjected to mass spectrometry (MS) detection. To validate the components in the formula, a connected system of LC-30 (Shimadzu)-Hybrid Quadrupole Time-of-Flight MS (TOF MS) with an electrospray ionization source (ESI) was used. The mobile phase system had solution A (acetonitrile) and equate B (0.1% HCOOH-H₂O): 25 min (A:20%: B:80%), 50 min (A:35%: B:65%), 9 min (A:90%: B:10%), and 7 min (A:20%: B:90%). Data were processed in information-dependent acquisition (IDA) with a high-sensitivity mode.

Experiments were approved by the guidelines of the Institutional Animal Care and Use Committee of Jining Medical University in China (SYXK-Shandong province-2018-0002).

Mice (Pengyue Animal Center of Shandong province, China) (male, about 22 g, 2 months old) were given 2% dextran sodium sulfate (DSS, MP Biomedicals) for 1 week, and then, the CW formula was administered for 21 days. The dose of the human being was 120 mg/kg per day based on the Chinese Pharmacopoeia 2020 version, and correspondingly, murine was 1.44 g/kg calculated by the Meeh–Rubner formula: S k i n   s u r f a c e   a r e a = M a s s   C o e f f i c i e n t s × ( W e i g h t × 2 3 10000 ) . An amount of 36 mg (equal to 1.44 g/kg) (the high dose) or 18 mg (equal to 0.7 g/kg) (the low dose) CW diluted in 200 μl water was administered each day. The positive control of the study was mesalazine (MedChemExpress, China) (Veloso et al., 2021). Every 15 mice were grouped, and five groups were set as follows: 0.9% saline, 2% DSS, the low-CW-DSS group, high-CW-DSS group, and mesalazine group (200 mg/kg). Isoflurane (RWD Life Science, Shenzhen City, China) was used for gas anesthetization. The histopathological score was determined: weight loss (normal: 0; <5%: 1; 5%–10%: 2; 10%–20%: 3; >20%: 4); feces (normal: 0; soft: 1; very soft: 2; liquid: 3); blood test (no blood within 2 min: 0; positive in 10 s: 1; light purple within 10 s: 2; heavy purple within 10 s: 3) (Leagene, China); and histological indices (destruction of the epithelium, immune cell infiltrates, edema, and crypt loss, each for 1).

The colon samples were soaked in paraformaldehyde (PFA) for 72 h, embedded in paraffin and then cut into 3-μm thick slices, and subjected to hematoxylin and eosin (H&E) staining.

After the high-dose treatment of CW, we collected the serum from colitis mice in 2 hours for the following experiments. To examine the toxicity, NCM460 cells, an epithelial cell line derived from the normal colon of a 68-year-old Hispanic male, were treated with serum and subjected to an MTT assay (3- [4,5-dimethylthiazol-2-yl]-2,5-diphenyl tetrazolium bromide) and cellular immunofluorescence assay with PI staining (Abcam, United States).

Enzyme-linked immunosorbent assay (ELISA) kits were bought from Abcam and were utilized to examine the cytokine concentrations in the serum: TNFα (sensitivity: 0.1 pg/ml; range: 47–3,000 pg/ml) and IL-1b ELISA kits (sensitivity: 15.2 pg/ml; range: 28.1–1,800 pg/ml).

The feces and mucus within the colon were collected and mixed and processed by 16S rRNA gene sequencing. The raw data were filtered by Trimmomatic (Bolger et al., 2014), and paired-end reads were produced after chimera removal by UCHIME. Corrected paired-end reads generated circular consensus sequencing (CCS) reads, which were distinguished based on the barcode sequences. After removing chimeras, an OTU (operational taxonomic unit) analysis was performed with a similarity >97%. Species annotation and taxonomic analysis, diversity analysis including alpha and beta diversity, significant difference analysis, and functional prediction were performed. Differential analysis among groups was calculated by the t-test. The rarefaction curve and Shannon curve were plotted using Mothur and R packages. The rank abundance curve (Guo et al., 2020) was generated by ranking the features in each sample based on abundance and plotting them in abundance rank (X-axis) and relative abundance (Y-axis) (Salari et al., 2016). Beta diversity analysis was processed by QIIME software to compare species diversity between different samples utilizing binary jaccard, which was an OTU-based algorithm. Principal coordinates analysis (PCoA) was drawn by the R language tool, and PERMANOVA (Adonis) created by the vegan package was used to analyze whether there was a significant difference in beta diversity between samples from different groups (Xie et al., 2020). LEfSe [Linear Discriminant analysis (LDA) Effect Size] was conducted to identify biomarkers with statistical difference between different groups (Zhou et al., 2020). Metastatsconducted a t-test on species richness data between groups at the taxonomic level of the genus (Hu et al., 2018). Krona, as a powerful metagenomic visualization tool, demonstrates species annotation. Each fan means the corresponding annotation’s proportion (Ondov et al., 2011). PICRUSt (Phylogenetic Investigation of Communities by Reconstruction of Unobserved States) analysis was conducted to predict the alterations in KEGG pathways. The raw data were available with the accession number in SRA (PRJNA827781).

Fecal and mucus samples were treated with methanol and standard internal substances. After the ultrasound and frozen treatment, the samples were centrifuged. The supernatant was subjected to ultra-high-performance liquid chromatography (UHPLC) coupled with TOF-MS (Garcia and Barbas, 2011). Acquisition software (Analyst TF 1.7, AB Sciex) was utilized to assess the full scan survey MS data, which were then converted by ProteoWizard and generated the retention time (RT), mass-to-charge ratio (m/z) values, and peak intensity. According to the in-house MS2 database, substances were identified and determined by orthogonal projections to latent structures-discriminant analysis (OPLS-DA). The cut-off was p value < 0.05, fold change (FC) > 1, and Variable Importance in the Projection (VIP) > 1. The HMDB (Human Metabolome Database) (Wishart et al., 2018b) and KEGG database were utilized to annotate metabolites.

Mice were anesthetized, and sterilized PBS was injected into the abdominal cavity. After massage, the liquids were isolated and centrifuged. The pellet was diluted in RPMI-1640 complete medium.

Diluted microparticles (Thermo Fisher, United States) [fluorescein isothiocyanate (FITC) wavelength-488 nm] were incubated with peritoneal Mφs collected from anesthetized mice for 2 h at 37°C, were centrifuged and washed, and subjected to flow cytometry analysis.

A density of 10⁵ NCM460 cells was seeded in the 24-well cell culture plate at a 37°C 5% CO₂ incubator. When reaching 100% confluency, a scratch was made, and a transwell insert (Corning, 0.3 μm diameter, United States) carrying 10⁴ peritoneal Mφs was placed. After 24 h, the cells in five random fields were examined.

Total protein (40 µg) was subjected to SDS-PAGE (sodium dodecyl sulfate–polyacrylamide gel electrophoresis) and transferred. After blockage, antibodies [CASP3, vascular endothelial growth factor A (VEGFa), IL-1b, IL-10, and arginase 1 (ARG1)] were incubated with the membrane for 24 h. After washing with PBS, secondary antibodies conjugated with horseradish peroxidase (HRP) were utilized, and an enhanced chemiluminescence (ECL) substrate was added to visualize the protein bands. All chemicals were bought from Invitrogen, United States.

The colon samples were subjected to the Caspase-3 activity assay kit (Abcam, United States), and the absorbance was determined at an optical density at 405 nm (BioTek Instruments, Inc.).

The small intestine was cut into small pieces and washed with cold PBS until the solution became clear. After 35 min of digestion by the digestive solution (Yu et al., 2020), the filtered medium was centrifuged and cultured in a growth medium. After 9 days, tumor necrosis factor-α (TNFα) was added with or without the CW serum. After 24 h, IOs were incubated at 37°C with propidium iodide (PI)/Hoechst33342 (480 nm/630 nm) for 20 min. To measure mitochondrial stress, MitoSOX™ Red mitochondrial superoxide indicator (590 nm) was added. After 10 min at 37°C, cells were examined using a fluorescent microscope. Two dyes and culturing reagents were bought from Thermo Fisher and STEMCELL Technologies, respectively.

Data are means ± SEM. *p < 0.05, **p < 0.01, and ***p < 0.001. Every experiment was repeated at least five times. Comparisons between two groups were analyzed by unpaired Student’s t-test. GraphPad 8.0 and R software were utilized.

We determined the chemical profile of the CW formula by UPLC-MS/MS and identified the presence of the ingredients (Supplementary Figure S1, Supplementary Table S1). DSS exacerbated weight loss, caused rectal bleeding, shortened the colon length, and induced edema. These symptoms were markedly alleviated by the CW therapy at a low and high dose (1.44 g/kg). As compared with the mesalazine treatment, the high dose of CW produced a longer colon (Figure 1A). H&E staining showed that the high dose of CW successfully rescued colitis-caused crypt loss and alleviated epithelial breach, as well as immune cell infiltrates, showing its superiority in mucosal healing and resolution of inflammation compared with mesalazine (Figure 1B). Of note, both doses of CW rescued weight loss (Figure 1C) and suppressed the DSS-evoked IL-1b and TNFα levels (Figures 1D,E).

The CW formula encompassed six herbs (Supplementary Table S2) and shared 85 putative targets with the acquired 5811 colitis-relevant genes (Figure 2A). A HIT network representing the interactions between active components and targeted genes was established: pink rectangle nodes represented colitis-associated genes, and nodes with different colors inside were active components (Figure 2C). Based on the database and published articles, 12 active components of SR, 30 of DC, and 2 of GB were identified, while OS, FS, and Cb had three components. Among the 115 targets of these components, 85 genes were associated with the progression of colitis.

The KEGG analysis demonstrated that these shared genes were enriched in carcinogenesis, infections, neuroactive ligand–receptor interaction, and VEGF and HIF-1 signaling pathways (Figure 2B). As illustrated in the pie charts, and weight or target proportion, SR occupied much of the formula (Figure 3A). All the ingredients are associated with infection and CRC, as well as inflammatory responses (Figure 3B).

Based on the shared genes between colitis and the CW formula, a PPI network with a PPI enrichment p-value (<1.0e-16) was established utilizing the STRING database, and a sub-network composed of 24 genes is shown in Figure 4A with the median values (BC: 16.157288675, CC: 0.151452282, DC: 5, EC: 0.0531741455, LAC: 1.625, and NC: 2.5833333335). To predict the major active components in the formula, the calculated 24 hub genes and corresponding components were utilized to construct a component–hub gene network (Figure 4B). Western blotting confirmed that the high dose of CW treatment significantly suppressed the protein abundance of Casp3 and IL-1b, and increased the Casp3 activity (Figures 4C,D), indicating an alleviated inflammation by the CW formula.

To determine the effect of the CW formula on microbiota composition, 16S rRNA gene sequencing was conducted utilizing the feces and mucus. All sample libraries covered >99%, suggesting a sufficient library size to represent most microbes (Supplementary Table S3). DSS treatment led to a marked decrease in the alpha-diversity of microbiota including Chao and Ace, and Shannon indexes, as well as the Simpson index, all of which were reversed by the CW therapy (1.44 g/kg) (Figure 5A). β-diversity was calculated by the binary Jaccard method. Principal Co-ordinates Analysis (PCoA) combined with PERMANOVA showed that CW treatment significantly altered the β-diversity of commensal microorganisms (Figure 5B). Non-metric multidimensional scaling analysis (NMDS) indicated the CW-DSS group clustered distinctly from the DSS group with a stress value of 0.02 (Figure 5C). At the phylum level, we observed an apparent reduction in Bacteroidota and Firmicutes in the DSS group compared with other groups, with an increase in Proteobacteria (Figure 5D). LEFSe showed that the CW treatment increased the relative abundance of the phyla Bacteroidota and Firmicutes, whereas decreased the prevalence of Proteobacteria at the phylum level. The administration of CW suppressed the abundance of species Romboutsia ilealis, Bacteroides thetaiotaomicron, Paeniclostridium sordellii, and Escherichia, and Shigella, while increased Lachnospiraceae NK4A136, Phascolarctobacterium faecium, Mucispirillum schaedleri, and Helicobacter winghamensis (Figure 5E). BugBase feature prediction calculated by Mann–Whitney–Wilcoxon analysis showed that the CW treatment significantly reversed the DSS-induced abundance of potentially pathogenic phenotype microorganisms (Figure 5F). PICRUSt analysis demonstrated that the administration of XQ inhibited infectious diseases and cancers, and lipid metabolism, controlled the digestive system, and improved the immune system (Figure 5G).

To examine the alterations in the metabolic profile, untargeted metabolomics was performed. OPLS-DAOPLA-DA showed that the metabolites between the saline and the DSS group were clustered distinctly between the two groups with Q2Y (0.894, 0.782) and R2Y (0.953, 0.953) (Supplementary Figure S2A). A Q2 value of more than 0.5 indicates the precise predictivity of the model. A permutation test was performed to validate the model, and R2 and Q2 should be higher than the permutated models with a vertical axis intersection of Q2 that was lower than zero. As illustrated, the intersection of Q2 was (0.0, −1.35) and (0.0,−1.35), which confirmed the validity of the OPLS-DA model (Supplementary Figure S2B). According to VIP, the top five regulated outlier metabolites in the DSS group are shown in (Supplementary Figure S2C, Supplementary Table S4). The KEGG analysis showed that DSS suppressed linoleic acid metabolism, whereas it increased the cytochrome P450 activity (Supplementary Figure S2D).

OPLS-DAOPLA-DA combined with a permutation test also validated the differential metabolites between the DSS and DSS-CW (1.44 g/kg) groups with Q2Y (0.943, 0.844) and R2Y (0.997, 0.94) (Figure 6A). The top five regulated metabolites are listed in Figure 6B (Supplementary Table S5). The differential metabolites between the CW and DSS groups were associated with carboxylic acids and derivatives, glycerophospholipids, fatty acyls, and steroids and derivatives (Figure 6C). The administration of the CW abolished the DSS-induced cytochrome P450 activity and DSS-impaired linoleic acid metabolism, and inhibited serotonin levels in mice subjected to colitis (Figure 6D).

Peritoneal Mφs were collected from UC murine models. The CW formula increased the protein abundance of Vegfa and Arg1 (Figure 7A). As shown in microparticle experiments, the CW treatment increased the proportion of P2–P4 from 9.28% to 25.6% (Figure 7B), suggesting an enhanced M2 transition and alleviated inflammatory reactions. Moreover, NCM460 cells were co-cultured with the CW peritoneal Mφs for 24 h. The wound healing experiment showed an increased migration in the CW group, confirming the favored wound healing capacity of M2 Mφs (Figure 7C).

IOs are composed of nearly all the intestinal cell types, including epithelial cells, endocrine cells, and paneth cells, hence being considered an ideal tool to reflect the status of epithelial recovery. We treated IOs with TNFα (20 ng/ml, 24 h) to construct an ex vivo UC cellular model. The serum was isolated from murine colitis models subjected to the CW treatment and healthy blank controls, and both doses of CW serum did not induce apoptosis and impaired the viability of NCM460 cells (Supplementary Figure S3). TNFα treatment increased the ROS level, and this elevation was diminished by the CW serum (Figure 8A). Moreover, TNFα-induced IO death was also rescued after the administration of the CW serum (Figure 8B).