The details regarding the reagents utilized in this investigation can be found in Supplementary Table S1.

The fresh bark of Phellodendron chinense var. glabriusculum C.K.Schneid. and Acalypha australis L. (Dechangxiang, Guizhou) was crushed and blended in a 1:1 ratio. Subsequently, 1000 g of ZLD was soaked in 10 L of water for 1 h, refluxed at 100°C for 2 h, filtered and extracted with 8000 mL of water at the same temperature for another 1.5 h. The resulting filtrates were combined, condensed, and subjected to vacuum freeze-drying to yield a powdered extract. Finally, a 58.2g ZLD was obtained for the follow-up experiment.

0.5g ZLD powder was dissolved with 10 mL distilled water and extracted with 50% methanol. After vibration and centrifugation at 12,000 rpm for 30 min, the supernatant was transferred and evaporated to dryness at a temperature under vacuum. The residue was dissolved in 1 mL of ddH2O and centrifugated at 12,000 rpm for 15 min twice prior to analysis. The standard substances were dissolved in methanol. The chemical analyses were performed using Thermo Fisher Q Exactive coupled with a Vanquish UPLC system. The mobile phase in positive mode included H2O with 0.1% FA (A) and methanol (B). The gradient conditions included: 0–2 min at 10% B; 2–3 min at 10%–35% B; 3–10 min at 35%–45% B; 10–11 min at 45%–90% B; 11–12 min at 90% B; and 12–12.5 min at 90%–10% B. The mobile phase also consisted of 0.1% FA (C) and methanol (D) under negative mode. The negative mode gradient conditions are: 0–2 min at 10% D; 2–3 min at 10%–35% D; 3–5 min at 35%–90% D; 5–7 min at 90%–95% D; 7–9 min at 95% D; 9–11 min at 95%–10% D. The experimental conditions included a flow rate of 200 μL/min and a constant temperature of 40°C. The mass conditions were: spray voltage: +3.4KV/-3 KV; sheath gas pressure: 40 arb; aux gas pressure: 10 arb; capillary temp: 300°C; heater temp: 320°C.

Eight-week-old specific pathogen-free (SPF) C57BL/6J mice, weighing 20 ± 2 g, were obtained from Beijing Huafukang Animal Breeding Center (License No. SCXK-[jing]-2019–0008). Mice were adapted and fed for 1 week, and experimental phase keeping was performed under standard laboratory conditions. The experimental procedures received approval from Guizhou University of Traditional Chinese Medicine Animal Research Institute (approval number 20230047).

UC induction with dextran sulfate sodium (DSS) was performed as previously reported. Mice were kept in a controlled space under SPF conditions and received sterile diet and water. Experimental mice were divided randomly into control, DSS, sulfasalazine (SASP), ZLD low-dose (ZLD-L), and ZLD high-dose (ZLD-H) groups (n = 15 per group). UC was induced in groups DSS, SASP, ZLD-L, and ZLD-H by administering 3% DSS (w/v) over a period of 7 consecutive days. Mice in the control group were provided with deionized drinking water for a period of 14 days. Following UC induction, the reference group mice were administered 125 mg/kg/d SASP. Mice in the ZLD-L and ZLD-H groups were fed 291.25 mg/kg/d and 873.75 mg/kg/d ZLD, respectively. SASP and ZLD were administered orally once daily for 7 days. All treated mice were euthanized on the 14th day.

The severity of the disease was assessed daily basis by monitoring body weight, dietary intake, fecal occult blood, and consistency of stool. The Disease Activity Index (DAI) was computed based on established criteria (Li et al., 2021) (Supplementary Table S2). DAI scores were calculated by summing the weight change, defecation, and occult blood scores, and then dividing it by three. On day 13, fresh feces were harvested from each mouse and frozen at −80°C in sterile EP tubes. The following day, colon tissues were obtained also stored at −80°C.

Colon tissues from the mice were fixed in a 4% paraformaldehyde solution, then embedded in paraffin and subsequently sectioned at a thickness of 5 μM. The sections were processed for hematoxylin and eosin (HE) staining, followed by visualization and photography under a light microscope (Nikon Fi3, Japan). Pathological scoring of colonic injury was conducted using the Geboes score (Supplementary Table S3) (Shi et al., 2017).

Colon tissues (<1 mm³) were fixed in 2.5% glutaraldehyde and phosphate buffer solution. Following fixation, gradual dehydration was performed using progressive ethanol concentrations. After embedding and solidification, sections of 70 nm were cut using an ultramicrotome (Leica UC7). These sections were stained with electron microscopy dyes to enhance contrast for cellular organelles and structures. Subsequently, the samples were observed for their ultrastructure using 80 kV transmission electron microscopy (HT7800/HT7700), and images were captured.

Add 10 mg of tissue to 500 μL of extract (containing protease inhibitor and Ripa lysate) and homogenize using a homogenizer. The homogenate was subjected to centrifugation at a speed of 12,000 revolutions per minute for a duration of 10 min at a temperature of 4°C. Following this, the supernatant was carefully extracted for use in the ELISA assay. Pro-inflammatory cytokine levels, including TNF-α, IL-6, IL-1β, and IL-18, were assessed using an ELISA kit (Sevier, China) as per the manufacturer’s guidelines.

Fecal samples were selected and subjected to three freeze-thaw cycles at −20°C. They were then filtered and extracted using a butanol: methanol: water (5:25:70, v/v/v) solution. The specific metabolites were quantified following the instructions provided with the ELISA kits.

Paraffin sections of colon tissue were stained using rabbit polyclonal antibodies against Toll-like receptor 4 (TLR4, dilution 1:400) and NLR family pyrin domain containing 3 (NLRP3, dilution 1:50), along with polyclonal anti-rat IgG (1:200) antibodies. Section images were taken with NIS Elements Imaging Software Version 4.0.

Total RNA was extracted using the Tissue RNA Purification kit (EZBioscience, China), following the manufacturer’s guidelines. Quantify RNA using the kit (refer to Supplementary Table S1 for details). RT-qPCR analysis was conducted using the CFX96 Real-Time System (BIO-RAD, United States). Supplementary Table S4 contains the primer sequences used in this study.

FastPrep-24TM5G (MP Biomedicals, United States) for extraction of total proteins from PBS-washed colon tissues. The specific experimental procedures of the Western blot have been cited in previous studies (Yang et al., 2024). Used the following antibodies at the indicated dilutions: β-actin (dilution 1:100,000), NF-κB (dilution 1:1,000), Phospho-inhibitor of NF-κB alpha (p-IKBα, dilution 1:1,000), cleaved caspase-1 (dilution 1:3,000), apoptosis-associated speck-like protein containing a CARD (ASC; dilution 1:1,000), NLRP3 (dilution 1:500), and HRP-conjugated (dilution 1:10,000). Bands were visualized and quantified utilizing the ChemiDocTM Imaging System (BIO-RAD, United States) and Image Lab software.

Genomic DNA of feces collected from control, DSS, ZLD-L, and ZLD-H groups was extracted using the DNA kit (Omega Bio-Tek, China) according to the instructions. The V3-V4 hypervariable regions of the bacterial 16S rRNA gene were amplified from genomic DNA using specific primers (338F 5′-ACT​CCT​ACG​GGA​GGC​AGC​A-3′ and 806R 5′-GGACTACHVGGGTWTCTAAT-3′). Samples were individually labeled with unique barcodes during PCR amplification, utilizing Tks Gflex DNA Polymerase. The quantity and quality of the amplified products were assessed with the gel electrophoresis. Following this, a purification of the PCR product was conducted using VAHTSTM DNA Clean Beads and AMPure XP Beads, respectively. The purified PCR products were characterized with the Quant-iT PicoGreen dsDNA Assay Kit. Libraries were constructed and sequenced using the MiSeq Reagent Kit v3 (Illumina, United States). Raw data were generated in FASTQ format. Microbiome biological information was analyzed using QIIME2 to obtain Operational Taxonomic Units (OTU). The index of Chao1, Shannon, and Simpson, as well as principal coordinates analysis (PCoA) and non-metric multidimensional scaling (NMDS) plots, were generated using R software and QIIME2 software. To detect taxonomic units that differ significantly between groups, we used the Least Discriminant Analysis Effect Size (LEfSe) method.

Referring to previous studies described (Lu et al., 2021), to implement the operation, metabolite extraction involved adding 50 mg of fecal sample to 400 μL of extraction solution (methanol: water = 4:1) containing 0.02 mg/mL of the internal standard (L-2-chlorophenylalanine) in an EP tube. The supernatant obtained after sample grinding was pipetted into an analytical vial. The UPLC-Q Exactive system (Thermo Fisher Scientific) with an HSS T3 column (100 mm × 2.1 mm, 1.8 μm) and Q Exactive mass spectrometry were used for detection. The mobile phases were composed of a solution of 0.1% formic acid in water and acetonitrile (95:5, v/v) (referred to as solvent A), as well as a solution of 0.1% formic acid in acetonitrile, isopropanol, and water (47.5:47.5:5, v/v) (referred to as solvent B) (Zhang et al., 2023). The automatic injector was set to 4°C and programmed to dispense a volume of 3 μL. Information Dependent Acquisition (IDA) controlled by Xcalibur software (Thermo-Fisher) was used to collect mass spectrometry data. The mass spectra were continuously assessed in positive and negative status for further analysis in the subsequent stage.

Data were subjected to analysis using GraphPad Prism 9.0 software. The student’s test and one-way ANOVA were used, and all data were expressed as mean ± standard error of the mean (SEM). The co-occurrence among intestinal flora and metabolites was calculated based on the relative abundance by Spearman’s rank correlation coefficient (P< 0.05) using R package Hmisc. Visualization of results was achieved using R 4.03. Differences were considered statistically significant for P-values < 0.05.

The chemical analysis showed that 108 compounds were contained in ZLD (Supplementary Table S5). Among them, hellodendrine and berberine were identified under the positive mode, according to the standard substances (Figure 1A). The three chemicals in the ZLD extract were identified as gallic acid, quercetin, and galanin (Figure 1B).

Protocol design is illustrated in Figure 2A. In contrast to the control mice, the DSS group exhibited a reduction in body weight (Figure 2B), elevated DAI score (Figure 2C), and shortened colon length (Figures 2D, E). Histopathological examination revealed extensive inflammatory cell infiltration and ulceration in the colon mucosa of the DSS group compared with the other experimental groups, with significantly higher Geboes values (Figures 2F, G). Treatment with SASP, ZLD-L, and ZLD-H significantly ameliorated the pathological alterations of the colon tissue and reduced the Geboes values (Supplementary Table S3). In addition, TEM results showed that the intestinal barrier and tight junctions were disrupted in the DSS group (Figure 2H). Nevertheless, the ZLD-L and ZLD-H groups exhibited significantly improved intestinal barrier structures as compared to the model group, while the SASP group did not show a significant improvement.

Mouse fecal samples were subjected to high-throughput 16S rRNA sequencing in order to evaluate the impact of ZLD on gut flora. We utilized sparse curves, as well as α and β diversity indices, to assess the microbial species diversity present in the samples. These curves were relatively flat for each group (Supplementary Figure S1). A Venn diagram unveiled a combined total of 442 OTUs across the four groups, as illustrated in Figure 3D. To assess alpha diversity in the gut microbiota, Chao1, Simpson, and Shannon indices were utilized at the OTU level. Compared to other groups, the DSS group exhibited reduced.

Chao1, Simpson, and Shannon indices, indicating decreased microbiota richness and diversity (Figures 3A–C). Treatment with ZLD-L and ZLD-H significantly raised Chao1, Simpson, and Shannon indices values (Chao1 and Shannon: P< 0.05) versus the DSS group. However, differences in those three indices for the ZLD-L group and in the Simpson index for the ZLD-H were not significant. PCoA and NMDS were conducted to evaluate the β diversity across groups. Microbial community analysis using PCoA and NMDS revealed distinct clustering patterns between groups. The DSS group displayed a broader distribution than the control group, while the ZLD-L and ZLD-H groups demonstrated distributions similar to the control group (Figures 3E, F). Thus, ZLD could comprehensively elevate the α and β diversity of microbiota in mice with UC.

The gut microbiota in the DSS group was predominantly predominantly composed of the phylum level Bacteroidota, Firmicutes, and Proteobacteria (see Supplementary Figure S2). The relative abundance of Bacteroidota decreased, while that of Firmicutes, Proteobacteria increased in DSS group (Figure 4B). Additionally, the DSS group exhibited a higher Firmicutes/Bacteroidota (F/B) ratio was elevated in the DSS group than in the control group (Supplementary Figure S3). Bacteroidota, Firmicutes, Proteobacteria, and Verrucomicrobia were the main components of the intestinal flora in the ZLD-L and ZLD-H groups (Supplementary Figure S2).

Compared to the DSS group, the relative abundance of Firmicutes and Bacteroidetes increased in the ZLD-L and ZLD-H groups (Figure 4B). The F/B ratio decreased in the ZLD-H group and increased in the ZLD-L group compared to the DSS group (Supplementary Figure S3). However, neither difference was statistically significant. Additionally, ZLD treatment elevated the abundance of Verrucomicrobia and depressed Proteobacteria. The taxonomic composition by group at various levels is illustrated in Supplementary Figures S4–7 to provide further elucidation.

Lactobacillus, Akkermansia, and Prevotella were the predominant genera in all four groups, as illustrated in Figure 4A. LEfSe analysis was conducted to determine distinctive bacteria across the groups. In contrast to the DSS group, at the genus level, the ZLD-H exhibited an enhanced relative abundance of Coprococcus, [Prevotella], Clostridium, Christensenella, Lactobacillus, Candidatus_Arthromitus, Turicibacter, and degraded relative abundance of 13 other microorganisms, including Prevotella, Enterococcus, Bacteroides, Parabacteroides, Ruminococcus, Enterococcus, Streptococcus, Staphylococcus, Proteus, Enterobacter, Shigella, Rikenella, Escherichia, and Mucispirillum. The relative abundance of Turicibacter and Anaerofustis was elevated, yet that of Streptococcus, Staphylococcus, Enterococcus, Proteus, Enterobacter, Parabacteroides, Rikenella, Escherichia, Desulfovibrio, and Butyricicoccus was reduced for the ZLD-L group than for the DSS group. The findings suggest that ZLD treatment induces changes in the composition and diversity of gut microbiota, as depicted in Figure 4D.

We identified 50 critical microbial species with linear discriminant analysis (LDA) scores >4 (Figure 4C). 21 key differential gut microbes species were identified utilizing LDA >4 as a criterion, with 7 signature differential gut microbes at the genus level: [Prevotella], Clostridium, Christensenella, Lactobacillus, Candidatus_Arthromitus, Turicibacter, Enterococcus.

Our previous results indicated that, among the tested compounds, ZLD-H treatment was the most effective in managing UC. Furthermore, it significantly promoted specific beneficial flora. Subsequently, we analyzed the effect of ZLD-H on fecal metabolism using liquid chromatography-tandem mass spectrometry. Mice metabolic profiling carried out using Partial Least Squares Discriminant Analysis (PLS-DA) and Orthogonal Partial Least Squares Discriminant Analysis (OPLS-DA) (Figures 5, 6), revealed significant differences in metabolite profiles between the DSS and ZLD-H groups (Supplementary Figures S8, S9), with samples in the ZLD-H group resembling those from the control group. The DSS group formed distinct metabolic clusters compared to the ZLD-H group (cationic model: R²Y = 0.9728, Q² = −0.1439; anionic model: R²Y = 0.91555, Q² = −0.2754).

The top 30 metabolites, identified in both positive and negative ion modes, that exhibited differentially expressed between the DSS and ZLD-H groups, as well as between the control and DSS groups are shown in Figure 7A, Supplementary Figures S10–S12. Metabolites with VIP (Variable importance for the projection) scores> 1 and P-values < 0.05 were selected. This metabolite screen yielded 96 metabolites, as illustrated in Supplementary Table S6 and Figure 7B. The DSS and ZLD-H groups exhibited differential metabolites totaling 58. ZLD-H significantly enhanced the levels of 29 metabolites, including butyrate acid (BA), zeranol, macromomycin B, and taurocholic acid 3-sulfate, while significantly reducing the levels of other 29 metabolites including 5(S)-HETE, arachidonic acid (AA), tetrahydrodeoxycorticosterone, docosahexaenoic acid, estradiol, and 5b-dihydrotestosterone.

Metabolic pathway enrichment analysis using MetaboAnalyst 6.0 indicated significant alterations in five major pathways (Figure 7C). Four pathways, including steroid hormone biosynthesis, AA metabolism, unsaturated fatty acid biosynthesis, and BA metabolism, were the main enriched in these differential metabolites. Thus, ZLD treatment significantly reduced the abundance of AA, 5(S)-HETE, while increasing that of BA. Furthermore, AA metabolism was shown to be hyperactivated in UC (Wu et al., 2022), whereas the butanoate metabolism pathway was shown to be inhibited (Jia et al., 2023).

Quantitative analysis by ELISA further revealed that ZLD significantly increased the levels of BA while reducing the levels of AA (Figure 7D).

To delve deeper into the interplay between specific microbes and differential metabolites, we performed correlation analyses to reveal the coordinated changes between ZLD treatment metabolites and 7 specific microbiomes.

Based on the criteria of |r|>0.57 and P< 0.05, the correlation identified 111 pairs of significantly correlated differential flora-metabolites (P< 0.05), with 19 pairs of correlations being more significant (P< 0.01). Christensenella, Clostridium, Candidatus_Arthromitus, Lactobacillus abundance was negatively correlated with most of the elevated differential metabolites and positively correlated with most of the reduced differential metabolites in the DSS group. The ZLD-induced specific enterobacterial genera Christensenella, Candidatus_Arthromitus with AA, and Christensenella were negatively correlated with 5(S)-HETE; whereas Clostridium, Candidatus_Arthromitus, Lactobacillus, and Turicibacter were all positively correlated with BA (Figure 8). In accordance with these results, we hypothesized that the modulation of key metabolite levels by specific gut flora influences corresponding metabolic pathways, thereby potentially explaining the therapeutic mechanism of ZLD for UC. This involves downregulation of the levels of AA, 5(S)-HETE to inhibit the over-activity of the AA metabolic pathway in its participation, and upregulation by BA to promote the BA metabolic pathway.

In Figure 9B, levels of IL-6, IL-1β, IL-18, and TNF-α were markedly increased in the DSS group compared to the control group. However, the levels of these pro-inflammatory factors were reduced in the SASP, ZLD-L, and ZLD-H groups than the DSS group. This demonstrated a significant mitigates the inflammatory response by ZLD. Previous studies have highlighted a positive feedback loop between NF-κB and NLRP3 in gut inflammation (Zhang et al., 2022). IHC staining showed than the control group, that the expression of TLR4 and NLRP3 was significantly increased in the UC group (Figure 9A). However, their expression was notably reduced by SASP and ZLD-H treatment. To further elucidate the involvement of the TLR4/NF-κB/NLRP3 pathway in UC, analysis of mRNA and protein expression levels of key targets in the colons of experimental animals. RT-qPCR analysis revealed significant upregulation of genes encoding tlr4, nf-κb, caspase-1, nlrp3 and asc in the DSS group, as illustrated in Figure 10A. Treatments with SASP and ZLD mitigated this upregulation. WB analysis (Figure 10B) confirmed these findings, showing significantly increased protein concentrations of Cleaved caspase-1, NF-κB p65, P-p65, p-IKBα, NLRP3, and ASC by DSS. These changes markedly were inhibited by the SASP, ZLD-L, and ZLD-H groups.

As Figure 11A showed, TNF-α was significantly negatively correlated with Clostridium, Lactobacillus, Candidatus_Arthromitus, IL-6 was significantly negatively correlated with Clostridium, Christensenella, Lactobacillus, Candidatus_Arthromitus. As well as IL-18 was significantly negatively correlated with Lactobacillus and IL-1β with Clostridium, Turicibacter. Changes in the levels of inflammatory factors were closely related to metabolites in addition to gut microbes as well. TNF-α, IL-1β, and IL-18 exhibited a negative correlation with BA, and IL-1β was positively correlated with 5(S)-HETE and AA. And TNF-α, IL-1β, IL-6 and IL-18 were associated with 22, 16, 16 and 37 metabolites, respectively (Figure 11B).