Dahuang Gancao Dandelion Decoction (DGD-D) was prepared using 40 g rhubarb (Rheum palmatum L.), 10 g licorice (Glycyrrhiza uralensis Fisch.), and 10 g dandelion (Taraxacum mongolicum Hand.-Mazz.) sourced from Ailikan Herbal Store (Urumqi, Xinjiang, China). Herbs were authenticated by botanical experts at Xinjiang Agricultural University and voucher specimens (No. XJAU2022-001–003) were deposited at the university’s herbarium. The dried herbs were decocted in 20 volumes of distilled water for1.5 h, after which a second extraction in 15 volumes of water for 1 h. The mixed filtrates were concentrated to 0.15 g/mL. The UHPLC-QE-MS was used to measure the chemical composition of them.

Bioactive constituents of DGD-D were identified through UHPLC-MS/MS analysis. The compounds were retrieved from the TCM System databases such as TCMSP and HERB based on the following criteria: bioavailability (OB) ≥ 30%, drug-likeness (DL) ≥ 0.18, and relevance to inflammation treatment. SwissTargetPrediction (http://swisstargetprediction.ch/, probability > 0.01), BATMAN-TCM (score ≥ 80), and TargetNet (http://targetnet.scbdd.com, probability ≥ 70%) were used to predict potential targets. Targets related to UC were obtained from GeneCards, DrugBank, OpenTargets, and the Comparative Toxicogenomics Database. Overlapping targets were illustrated through Venn diagrams generated using Microbioinformatics tools (https://www.bioinformatics.com.cn/). Compound-target-disease interaction networks were constructed with Cytoscape v3.9.1 (https://cytoscape.org/), whilst analysing protein-protein interactions (PPIs) using String (https://stringdb.org/). Finally, the relevant targets of DGD - D and UC were subjected to KEGG pathway enrichment analysis and GO functional enrichment analysis.

The Animal Welfare and Ethics Committee examined and permitted each experimental protocol (No. 2022016). The mice (C57BL/6) (Wuhan, China) were 6–8 weeks old and weighed 20–24 g.

The experiment was divided into two parts. In Part I, following a week of acclimatization, 32 mice (half male and half female) were divided into four groups at random: control (CK), LOW (950mg/kg), Mediue (1450mg/kg), High (1950mg/kg). During the 0–7 days of the experiment, Mice were euthanized 24 hours after the last administration under anesthesia induced by intraperitoneal injection of 1.25% avertin (20 mL/kg; Nanjing Aibei Biotechnology Co., Ltd). Then, samples of liver, spleen, kidney and small intestine tissues were collected for drug safety experiments.

In Part II, as depicted in the diagram of experimental design (Figure 1A), 24 mice (male) were divided into four groups at random: control (CK), DSS model (DSS), DGD-D group (1950 mg/kg, based on TCM literature (15, 16)and pilot studies), and 5-ASA group (400 mg/kg; Shanghai Aidefa Pharmaceutical Co., Ltd, Shanghai, China) via gavage. During the 0–14 days of the experiment, all groups received daily gavage as follows: CK and DSS groups, saline; DGD-D group, extrac; 5-ASA group, mesalazine. In order to replicate chronic UC progression in the DSS model, starting from day 8, the CK group was maintained on double-distilled water, whereas the remaining groups were administered a 3% DSS solution (molecular weight: 36,000–50,000 Da; Shanghai Macklin Biochemical Technology Co., Ltd.) for 7 day. Mice were euthanized 24 hours after the last administration under anesthesia induced by intraperitoneal injection of 1.25% avertin (20 mL/kg; Nanjing Aibei Biotechnology Co., Ltd). Serum, colon tissues, and spleen samples were then collected for subsequent analysis.

The DAI was determined as the average of the weight change score (WCS), fecal blood score (FBS), and fecal consistency score (FCS), calculated using the formula: DAI = (WCS + FBS + FCS)/3 (17). The spleen index was computed as follows: Spleen index = [spleen mass (mg)/body weight (g)] × 10 (17). Following measurement of length, colon tissue samples were divided into portions: one subset was fixed in 4% paraformaldehyde solution for histological analysis, while the remaining portions were snap-frozen in liquid nitrogen and stored at -80°C for molecular and biochemical assessments.

Total RNA from colonic tissue was quantified and assessed by Nanodrop spectrophotometer (Thermo, USA). The complementary DNA was synthesized from total RNA using the 2× RT OR-Easy™ Mix and Real-Time PCR Easy™ kits (Fujij Biotech Co., Ltd.) from all RNA for qRT-PCR. GAPDH was the internal control, while target genes were IL-1β, IL-6, IL-17, IFN-γ, TNF-α, MPO, ZO-1, Occludin, MUC2, TLR4, MyD88, NF-κB p65 (Rela), NLRP3, and 5-LOX. Relative gene expression was measured by 2^−ΔΔCt method and following primers were used (Table 1).

Colon tissues were first homogenized in PBS (1:9, w/v) supplemented with protease inhibitors (Roche, Switzerland). Then the supernatant was collected from it after centrifugation (12,000 × g, 10 min, 4°C). Cytokine levels were then quantified by commercial kits (Shanghai Kexing Trading Co., Ltd) according to the provided protocols.

Following the histopathological protocol, specimens of colonic tissue underwent fixation in neutral-buffered formalin and subsequent embedding within paraffin wax blocks; thereafter, consecutive sections measuring 3 μm in thickness were microtome-sectioned, affixed to charged glass slides, and subjected to conventional H&E staining to enable detailed microscopic study, with resultant tissue alterations being systematically scored via the semiquantitative rubric specified in Table 2 (18). Goblet cells and glycoproteins were further visualized using periodic acid–Schiff (PAS) and alcian blue–periodic acid–Schiff (AB–PAS) staining. Quantitative assessments were performed using Image-Pro Plus 6.0 software.

Colon sections that had been dewaxed were blocked with 3% BSA for 30 minutes after undergoing antigen retrieval in EDTA buffer (pH 8.0; Servicebio, Wuhan, China). Primary antibodies targeting ZO-1 (catalog no. 21773-1-AP), occludin (catalog no. 27260-1-AP), and MUC2 (catalog no. 27675-1-AP) sourced from Proteintech Group (Wuhan, China) were incubated with the samples at 4°C for 12–16 h. Following PBS washes, slices were counterstained with DAPI (Servicebio) for 10 minutes, treated for 1 hour in the dark with a fluorescein-conjugated secondary antibody (Cat. No. SA00001-1), and examined under the microscope.

Paraffin-dewaxed sections of colonic tissue were first treated with 10% normal goat serum (Servicebio) for 10 minutes to facilitate permeabilization at 4°C using primary antibodies to ZO-1 (catalog no. GB15195), occludin (catalog no. GB111401), MUC2 (catalog no. 27675-1-AP), TLR4 (catalog no. GB15186), MyD88 (catalog no. GB111554), NF-κB p65 (RELA) (catalog no. GB11997), NLRP3 (catalog no. GB114320), and 5-LOX (catalog no. GB111330). Subsequent rinsing in phosphate-buffered saline preceded a 2 h incubation at ambient temperature with a fluorophore-conjugated goat anti-rabbit secondary antibody (catalog no. G1213), after which nuclear counterstaining with DAPI was applied for 10 minutes prior to fluorescence microscopy acquisition. Digital image analysis for quantitative assessment was conducted via Image-Pro Plus 6.0 software.

The CK, DSS, and DGD-D groups’ cecal contents were taken on day 15 and kept at -80°C. Using the Illumina MiSeq platform, Suzhou PANOMIX Biomedical Tech Co., Ltd. (Suzhou, China) carried out 16S rRNA gene sequencing. Microbial α-diversity (Chao1, Pielou_e, Shannon, Simpson) and β-diversity (PCoA) were analyzed in QIIME2. Taxonomic differences at phylum and family levels were assessed, and differential taxa were identified using LEfSe (LDA > 4, P < 0.05).

Colon tissues underwent non-targeted metabolomics profiling using LC-MS at Suzhou Panomics Biomedical Technology Co., Ltd. (Suzhou, China). Chromatographic separation of metabolites was accomplished employing a UPLC HSS T3 column (2.1 × 100 mm, 1.8 μm) under a binary gradient of 0.1% formic acid in water (solvent A) and acetonitrile (solvent B) (0–3 min, 5% B; 3–10 min, 5-95% B) delivered at a flow rate of 0.4 mL/min. Mass spectrometry analysis involved electrospray ionization (ESI) in both positive and negative polarities for ionization and detection, covering a mass-to-charge (m/z) spectrum from 70 to 1000 Da. To explore data patterns, principal component analysis (PCA) was performed via the ropls package within the R environment. Differentially abundant metabolites were selected based on variable importance in projection (VIP) values exceeding 1, combined with Benjamini-Hochberg adjusted P-values below 0.05, enabling robust identification of key biomarkers. These findings were visually represented through volcano plots to highlight fold changes and significance, Venn diagrams to show overlaps across groups, and pie charts to depict proportional distributions of metabolite classes (19). Additionally, Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway mapping was applied to reveal enriched biological pathways, with statistical significance set at P < 0.05 to prioritize relevant metabolic networks (20).

The data is expressed as the mean ± standard deviation (SD). Statistical analyses were conducted using GraphPad Prism (version 8.0). Data conforming to a normal distribution were analyzed using one-way ANOVA, followed by Tukey’s multiple comparisons post hoc test to identify pairwise differences. P-value adjustment was performed using the Bonferroni correction for multiple tests. Differences were deemed statistically significant when p values were P < 0.05, ensuring reliable detection of treatment effects while controlling for type I errors.

The 418 components were found in Dahuang Gancao Dandelion Decoction (DGD-D), with the following primary classes: flavonoids (25.12%), phenolic acids and derivatives (9.33%), terpenoids (8.13%), coumarins and derivatives (6.22%), amino acids and their derivatives (4.78%) and lipids (3.83%), (Figure 2B). Key compounds, consistent with the Pharmacopoeia of the People’s Republic of China (2020), included aloe-emodin, rhein, emodin, chrysophanol, and physcion from rhubarb; glycyrrhizin and glycyrrhizic acid from licorice; and chicoric acid from dandelion (Figures 2C, D). These findings confirm the presence and chemical stability of DGD-D’s active components.

Based on the detection and identification results of UHPLC-MS, combined with the screening of TCMSP and HERB databases (OB ≥ 30%, DL ≥ 0.18), 28 major active ingredients in DGD-D were identified, targeting 1,024 potential sites in DGD-D, while 1,156 targets associated with UC were recovered. A Venn analysis showed that 305 targets overlapped (Figure 3A). According to a compound-target-disease network, the top targets were TP53, STAT3, TNF, AKT1, JUN, IL-6, EP300, HSP90AA1, CTNNB1, and ESR1 (Figures 3B, C). Analysis of protein-protein interactions (PPI) revealed that eight of these targets were linked to NF-κB (Figure 3D). Among the 20 KEGG pathways (P < 0.05), nine were inflammation-related, with NF-κB signaling as the primary focus (Figure 3E). GO analysis revealed enrichment in the control of apoptosis, inflammatory responses, and lipopolysaccharide-mediated signaling (Figure 3F).

CK group showed no pathological alterations in the liver, spleen, kidneys, and small intestine, with normal tissue structure and healthy status. Compared to the CK group, the Low, Medium, and high groups exhibited no significant differences in any of the aforementioned tissues. (Supplementary Figure S1).

DSS induced diarrhea, rectal bleeding, weight loss, colon shortening, and spleen enlargement in mice, confirming successful UC modeling (Figure 1B). Compared to the DSS group, DGD-D treatment significantly reduced diarrhea and rectal bleeding, attenuated weight loss (P < 0.01 or P < 0.001), lowered Disease Activity Index (DAI) scores (P < 0.001), restored colon length (P < 0.001), and decreased spleen index (P < 0.01) (Figures 1C, G). These effects were comparable to mesalazine (5-ASA; P > 0.05).

ELISA showed DSS significantly elevated IL-17, TNF-α, and IFN-γ levels (P < 0.001) and reduced IL-4 (P < 0.001), with no change in IL-10 (Figures 4A–E). DGD-D and 5-ASA treatments decreased IL-1β, IL-17, TNF-α, and IFN-γ while increasing IL-4 and IL-10 (P < 0.001). qRT-PCR confirmed DSS-induced upregulation of IL-1β, IL-6, IL-17, TNF-α, IFN-γ, and myeloperoxidase (MPO) mRNA (P < 0.01 or P < 0.001), which DGD-D and 5-ASA significantly downregulated (P < 0.01 or P < 0.001) (Figures 4F–K).

In mice treated with DSS, hematoxylin-eosin (HE) staining showed increased histological scores (P < 0.001) along with significant epithelial damage, crypt loss, and inflammatory cell infiltration. Histological scores were lowered (P < 0.001) by DGD-D and 5-ASA treatments, which also decreased crypt damage, submucosal edema, and immune cell infiltration (Figures 5A, B). PAS and AB-PAS staining revealed reduced content goblet cell numbers and glycoprotein in the DSS group, which 5-ASA and DGD-D restored (P < 0.001) (Figures 5C–F).

The findings of IF, IHC, and qRT-PCR analysis showed DSS significantly reduced ZO-1, Occludin and MUC2 expression levels (P < 0.001). DGD-D treatment increased these barrier proteins’ expression (P < 0.01 or P < 0.001), comparable to 5-ASA (P > 0.05) (Figures 6A, K).

DSS decreased microbial α-diversity, according to 16S rRNA sequencing (Chao1, Pielou_e, Shannon, Simpson indices). These parameters were improved by DGD-D (P < 0.05 or P < 0.001) (Figure 7A). Different microbial profiles were shown by Principal Coordinate Analysis (PCoA) for the CK, DSS, and DGD-D groups (Figure 7B). DSS reduced the Firmicutes/Bacteroidetes ratio at the phylum level by increasing Bacteroidetes and Proteobacteria and decreasing Firmicutes. These patterns were reversed by DGD-D (P < 0.001) (Figures 7C–F). At the family level, DSS increased Alcaligenaceae, Moraxellaceae, and Xanthomonadaceae, while decreasing Lachnospiraceae; DGD-D restored these abundances and increased Ruminococcaceae (P < 0.05) (Figures 7G, H),. LEfSe confirmed enrichment of Ruminococcaceae in theDGD-D groups, while Alcaligenaceae Moraxellaceae, and Xanthomonadaceae, was significantly enriched in the DSS group (Figures 7I, G). The results of KEGG pathway prediction showed that there were significant differences in the enrichment levels of pathways such as Cyanoamino acid metabolism, Butirosin and neomycin biosynthesis, and Apoptosis between the DSS Vs DGD-D group (P < 0.001) (Figure 7K).

Non-targeted metabolomics via LC-MS showed distinct metabolic profiles among CK, DSS, and DGD-D groups (Figures 8A, B). Of 1,232 identified metabolites, 535 (258 upregulated, 277 downregulated) differed between DSS and CK (VIP > 1, P < 0.05), and 555 (264 upregulated, 291 downregulated) differed between DGD-D and DSS (VIP > 1, P < 0.05) (Figures 8C, D). Venn analysis identified 339 shared differential metabolites, with lipids and lipid-like molecules comprising 23.89% (Figures 8E, F). KEGG enrichment revealed ARA, prolactin, and FoxO signaling pathways as significantly affected (P < 0.05) in both the DSS vs CK and DGDG-D vs DSS comparisons. Additionally, in the ARA metabolic pathway, DGD-D downregulated ARA, leukotriene A4 (LTA4), leukotriene B4 (LTB4), and leukotriene D4 (LTD4) levels (P < 0.001) (Figures 8G–J). As shown in Supplementary Figure S2, Lachnospiraceae and Ruminococcaceae were significantly negatively correlated with ARA, LTA4, LTB4, and LTD4 (P < 0.001). In contrast, Alcaligenaceae and Moraxellaceae showed significantly positive correlations with ARA, LTA4, LTB4, and LTD4 (P < 0.001). Additionally, Xanthomonadaceae was also significantly positively correlated with LTA4 and LTB4 (P < 0.001) (Supplementary Figure S2).

IHC and qRT-PCR analyses showed DSS upregulated TLR4, MyD88, NF-κB p65, NLRP3, and 5-LOX expression in colon tissues (P < 0.001). DGD-D treatment downregulated these markers (P < 0.01 or P < 0.001) (Figures 9A–E). These findings indicate DGD-D suppresses NF-κB/ARA signaling activation.