A total of 42 4-week-old C57BL/6 male mice (18 ± 2 g) were obtained from Chengdu Dashuo Laboratory Animal Co., Ltd. [Chengdu, China; Certificate No SCXK (Chuan) 2020-0030]. Mice were housed in an animal laboratory with a temperature and humidity of 22 °C ± 2 °C and 50% ± 5%, respectively, and a 12 h light/dark cycle (Andersson et al., 2024). The mice received standard water and food ad libitum. All experimental procedures were approved by the Animal Ethics Committee of Dali University (Approval No. 2023-SL-274, Dali, China), and conducted in accordance with the ARRIVE guidelines.

PA (Lot. 20200521) was provided by Yunnan Jingxin Biotechnology Co., Ltd. (Yunnan, China) and authenticated by Yang Zizhong. Olsalazine sodium capsules (Lot. 210206) were procured from Zhejiang Zhongyi Pharmaceutical Co., Ltd. (Zhejiang, China). DSS (Lot. 1132-222) was obtained from Shenzhen Lijing Biochemistry Technology Co., Ltd. (Shenzhen, China). Anhydrous ethanol (Lot. 1703113601) and formic acid (FA, Lot. 20171122) were acquired from Fuchen Chemical Reagent Co., Ltd. (Tianjin, China) and Tianjin Windship Chemical Reagent Technology Co., Ltd. (Tianjin, China), respectively. Acetonitrile (ACN, HPLC grade, Lot. 204127) was sourced from Fisher Scientific (Fair Lawn, NJ, United States). A fecal occult blood test kit (Lot. 20230520) was purchased from Nanjing Jiancheng Bioengineering Institute (Nanjing, China). Commercial ELISA kits for myeloperoxidase (MPO), inducible nitric oxide synthase (iNOS), tumor necrosis factor-alpha (TNF-α), interleukin-6 (IL-6), IL-17A, IL-4, IL-10, and interferon-gamma (IFN-γ) (Lots. 20220101-20852A, 20220101-21486A, 20220101-20852A, 20220101-20188A, 20220101-20171A, 20220101-20186A, 20220101-20162A, 20220101-20140A, respectively) were supplied by Shanghai Enzymes Biotechnology Co., Ltd. (Shanghai, China).

Dried PA powder (15 g) was extracted with ultrapure water (1:10, w/v) at 90 °C for 2 h. This extraction procedure was repeated twice. The combined aqueous extracts were filtered, and the filtrate was concentrated under reduced pressure using a rotary evaporator to a final concentration of 1.0 g/mL (equivalent to crude extract). Subsequently, anhydrous ethanol was added to the concentrate and the mixture was stored at 4 °C for 24 h to facilitate precipitation. Following centrifugation (4,000 × g, 20 min, 4 °C), the supernatant (PAW) was collected. This supernatant was sequentially fractionated by ultrafiltration using membranes with nominal molecular weight cut-offs of 10,000 Da and 3,000 Da. The resulting fractions were collected as follows: PAW1, Filtrate < 3 kDa; PAW2, Retentate 3–10 kDa (filtrate from 10 kDa membrane retained by 3 kDa membrane); PAW3, Retentate > 10 kDa. The crude PAW and the three ultrafiltered fractions (PAW1, PAW2, PAW3) were lyophilized (Freeze-dried) and stored at −20 °C until further use.

The freeze-dried PA was redissolved in ultrapure water and filtered to obtain the PA test solution. The PA test solution was reduced with 10 mM of DL-dithiothreitol (DTT) at 56 °C for 1 h, followed by alkylation with 50 mM of iodoacetamide (IAA) for 40 min in the dark at room temperature to obtain the PA peptide substance. Before LC-MS/MS analysis, PA peptides were resuspended in 0.1% FA. The analysis was performed using an Ultimate 3000 System Nanoflow UPLC system (Thermo Fisher Scientific, United States) coupled to a Q Exactive™ Hybrid Quadrupole-Orbitrap™ Mass Spectrometer with an ESI nanospray source (Thermo Fisher Scientific, United States), equipped with ReproSil-Pur C18-AQ resin (150 μm × 15 cm, 1.9 μm, Dr. Maisch GmbH, Germany). Elution was performed using solvent A (2% CAN with 0.1% FA) and solvent B (80% CAN with 0.1% FA) at a 600 nL/min flow rate and an injection volume of 5 μL. The gradient elution procedure was as follows: B from 4% to 8%, 2 min; B from 8% to 28%, 43 min; B from 28% to 40%, 10 min; B from 40% to 95%, 1 min; and B from 95% to 95%, 10 min. Additionally, the samples were analyzed using a Q Exactive™ hybrid quadrupole-orbitrap™ mass spectrometer (Thermo Fisher Scientific, United States) with a spray voltage of 2.2 KV and a capillary temperature of 270 °C (Wang et al., 2023).

The mass spectrometry parameters were configured to a mass resolution of 70,000 at 400 m/z, an automatic gain control target of 3e6, a maximum fill time of 40 ms, and a mass spectrometry precursor scanning range of 300.0–1,800.0 m/z. The 20 most intense peptide ions were scanned using MS/MS starting at 100 m/z with an isolation width of 3.00 and a normalized collision energy of 40.0. Raw MS files were analyzed using Byonic and searched in the target protein database according to the sample type (Cadang et al., 2023). The parameters were set as follows: carbamoylmethylation (C) was the fixed protein modification, oxidation (M) was the variable, enzyme specificity was set to non-specific, the maximum number of missed cuts was set to 3, the precursor ion mass tolerance was set to 20 ppm, and the MS/MS tolerance was 0.02 Da (Carnevale Neto et al., 2022).

The freeze-dried PA were dissolved in ultrapure water and filtered through a 0.22 μm microporous membrane. The target compounds were separated using an EXION LC System (SCIEX) ultra-high performance liquid chromatograph (Sciex, ExionLC AD) and passed through a Waters UPLC liquid chromatography column (ACQUITY UPLC HSS T3, 1.8 μm, 2.1 × 100 mm) at 400 μL/min (Fu et al., 2024). The mobile phases were 0.1% FA aqueous solution (A) and CAN (B), the column oven temperature was 40 °C, the autosampler temperature was 4 °C, and the injection volume was 2 μL. The gradient elution procedure was as follows: 2% B in 0–0.5 min, 2%–50% B in 0.5–10 min, 50%–95% B in 10–11 min, 95% B in 11–13 min, 95%–2% B in 13–13.1 min, and 2% B in 13.1–15 min. A SCIEX 6500 QTRAP+ triple quadrupole mass spectrometer (Sciex, Qtrap 6500 +) with an IonDrive Turbo V ESI ion source was used to perform mass spectrometry analysis in multiple reaction monitoring mode (You et al., 2024). The specific ion source parameters were as follows: ion spray voltage of +5,500/−4,500 V, curtain gas of 35 psi, temperature of 400 °C, ion source gas of 1:60 psi, ion source gas of 2:60 psi, and DP of ± 100 V (Dong et al., 2023).

The mice were randomly divided into seven groups (n = 6 per group): control, UC model, Olsalazine (200 mg/kg), PA (100 mg/kg), PA_3KD (100 mg/kg), PA_3-10KD (100 mg/kg), and PA_10KD (100 mg/kg) group. The dosages administered were selected based on prior research findings and have been proven effective in the UC model (Tao et al., 2020; Zhang et al., 2022). Following a 1-week acclimatization period, UC was induced in all groups except the control group by administration of dextran sulfate sodium (DSS) in drinking water. Specifically, a 3% DSS solution was administered for four consecutive days, followed by a 2% DSS solution for six consecutive days (Liu et al., 2023; Zhao et al., 2024). Commencing on the eighth day of modeling, therapeutic intervention was administered for 7 days. During this phase, the control and UC model groups were treated with enemas of 0.9% NaCl solution, while the remaining groups were administered enemas containing their respective suspensions. The comprehensive experimental procedure is depicted in Figure 1A.

Daily observations of body weight, fecal consistency, and rectal bleeding were recorded throughout the experiment. The DAI score was assigned to UC mice based on a composite of three scores: body weight loss (0, <1%; 1, 1%–5%; 2, 5%–10%; 3, 10%–20%; and 4, >20%), fecal consistency (0, normal feces; 1, well-formed pellets; 2, paste and semi-formed stools that do not adhere to the anus; 4, diarrhea, liquid stools that adhere to the anus), and rectal bleeding (0, bloodless; 1, red blood; 2, dark red blood; 4, rectal prolapse) (Niu et al., 2013). Following the completion of the treatment, the mice were humanely euthanized through cervical dislocation. Fecal samples were collected in sterile bottles on a clean bench and stored at −80 °C. The intestinal, thymus, and spleen tissues were promptly collected. The colon tissues were washed with ice saline, photographed, measured, and divided into two portions; one portion was fixed in a 4% paraformaldehyde solution, and the other was placed in an enzyme-free cryopreservation tube and stored at −80 °C. The weights of the thymus and spleen tissues were recorded. The thymus index was calculated by dividing the weight of the thymus by the weight of the body, while the spleen index was calculated by dividing the weight of the spleen by the weight of the body, expressed in grams per kilogram (Huang et al., 2024).

The homogenized colon tissues were centrifuged at 8,000 rpm for 20 min at 4 °C, and the supernatants were analyzed for TNF-α, IL-6, IL-17A, INF-γ, IL-4, IL-10, INOS, and MPO using an ELISA kit according to the instructions of the manufacturers.

The colon of the end tissue was fixed in 4% paraformaldehyde for 24 h, embedded in paraffin, and sectioned into 4-μm-thick slices (Nie et al., 2022). These sections were stained with hematoxylin-eosin (H&E) and alcian blue/periodic acid Schiff (AB-PAS) and examined under light microscopy to evaluate the secretory function of colonic mucus in mice and observe morphological alterations in histological sections. Histological scores (HS) were determined based on the degree of histopathologic alterations observed in colon tissue stained with H&E, which included epithelial cell changes assessment (0, normal morphology; 1, absence of cupped cells; 2, extensive loss of cupped cells; 3, deletion of crypt cells; 4, significant absence or polypoid regeneration of the crypt fossa) and cellular inflammatory infiltrates (0, none; 1, infiltration around crypts; 2, infiltration extending into the muscular layer of the mucosa; 3, widespread infiltration of the muscular layer of the mucosa with mucosal swelling; 4, infiltration into the submucosa). The overall HS was calculated by adding the scores for epithelial cell presence and cellular inflammatory infiltrate.

Samples were immersed in a sodium citrate buffer solution and heated in a microwave oven for antigenic repair. Subsequently, the sections were placed in a closed solution at room temperature for 30 min before being incubated with primary antibodies (NF-κB-p65, TLR4, MyD88, occludin-1, and ZO-1) at 4 °C overnight. All sections were treated with secondary antibodies, and the color was developed using an IHC kit (DAB). Positive DAB-stained regions were quantified using ImageJ software (version 1.48) according to the methodology reported for previous IHC analysis (Zhao et al., 2024).

Total proteins were extracted from murine colon tissues utilizing a tissue protein extraction kit (Lot. 061623231130, Beyotime, China). Protein concentrations were determined using the bicinchoninic acid (BCA) protein assay kit (Lot. B500, LABLEAD, China). The extracted protein samples were denatured by boiling in sample buffer at 100 °C for 10 min and subsequently resolved via electrophoresis on a 10% sodium dodecyl sulfate-polyacrylamide gel. Proteins were then transferred onto a polyvinylidene fluoride (PVDF) membrane (Lot. ISEQ00010, Millipore, United States). Subsequently, the membranes were blocked with 5% milk for 2 h at room temperature and then incubated overnight at 4 °C with primary antibodies targeting NF-κB p65 (1:1,000, Lot 8242S, Cell Signaling Technology, United States), β-actin (1:20,000, Lot AC026, ABclonal, China), and Lamin-B (1:500, Lot WL01775, Wanleibio, China). Following incubation, the membranes underwent three washes with TBST, each lasting approximately 15 min. Thereafter, the membranes were incubated with a horseradish peroxidase (HRP)-conjugated secondary antibody (1:1,000, Lot AS014, ABclonal, China) at room temperature for 1 h. Protein expression was subsequently detected using an enhanced chemiluminescence system (Advansta, United States), and imaging was conducted with an Azure 280 imaging system (Azure Biosystems, United States). Anti-β-actin served as an internal loading control. Analysis was carried out in ImageJ and signal intensities were normalized to loading controls, where applicable.

Fecal samples stored at −80 °C were removed from the refrigerator and thawed at 4 °C. Genomic DNA was extracted from feces using the aMG-Soil kit (Omega BioTek) according to the instructions of the manufacturers (Wu et al., 2022b). A 1% agarose gel was used to test the integrity of the extracted DNA, and a NanoDrop 2000 (Thermo Scientific, United States) was used to determine the concentration and purity (Li et al., 2023b). The highly variable region V3-4 of the 16S rRNA gene was PCR-amplified using an ABI GeneSide ^® 9700 PCR (ABI, United States) thermal cycler with the forward primer 338 F (5′- ACT​CCT​ACG​GGA​GGC​AGC​AG-3′) and reverse primer 806R (5′- GGACTACHVGGGTWTCTAAT-3′). The PCR products from the same samples were purified using 2% agarose gel electrophoresis to determine the size of the band fragments and recovered with the AxyPrepDNA Gel Kit (AXYGEN, United States). Subsequently, the purified PCR products were prepared with the NEXTFLEX Rapid DNA-Seq Kit (Bioo Scientific, Austin, United States) and sequenced on the Illumina MiSeq PE 300 platform (Illumina, San Diego, United States) following the standard protocol of Majorbio Bio-Pharm Technology Co., Ltd. The raw FASTQ files were multiplexed and quality filtered using QIIME software and subsequently aligned with high-quality 16S rRNA sequences sourced from the GreenGenes (http://greengenes.secondgenome.com/) database. The UPARSE (version 11) algorithm was employed to generate operational taxonomic units (OTUs) with a 97% similarity threshold and identify and remove chimeric sequences using UCHIME. The alpha diversity indices of the samples were assessed using Mothur software (version 1.30.2), and group differences were tested with Wilcoxon rank-sum or Kruskal–Wallis tests (with post hoc Dunn test). The QIIME software was utilized to conduct principal coordinate analysis (PCoA) and non-metric multidimensional scaling (NMDS) based on the Bray-Curtis distance metric to examine the resemblance of microbial community structures across the samples. Differences in relative abundances of flora at the phylum, family and genus levels were tested using Wilcoxon rank-sum and Kruskal–Wallis tests (with post hoc Dunn test). The LEfSe analysis software (version 1.2.11) (LDA > 2, P < 0.05) was used to identify bacterial taxa with significant differences in abundance from phylum to genus level among different groups. The functional gene composition of the samples was inferred using PICRUST2 software (version 2.2.0) to analyze different samples or groups for functional differences by comparing the species composition information obtained from 16S rRNA sequencing data (Douglas et al., 2020). Distance-based redundancy analysis (db-RDA) was used to investigate the effect of clinical indicators on the structure of the gut bacterial community.

All data are presented as the mean ± standard deviation (SD). Statistical analysis was performed using SPSS software (version 22.0). The normality of the data distribution was assessed using the Shapiro-Wilk test, and the homogeneity of variances was verified using Levene’s test. For comparisons among multiple groups (e.g., Control, UC, PA, PAW1, PAW2, PAW3), a one-way analysis of variance (ANOVA) was applied if the data met the assumptions of normality and homogeneity of variances, followed by Tukey’s post hoc test for pairwise comparisons. If the assumptions were violated, the non-parametric Kruskal–Wallis H test was used, followed by Dunn’s post hoc test with Benjamini–Hochberg correction for multiple comparisons. P < 0.05 was considered statistically significant. Graphs were generated using GraphPad Prism software (version 8.0.2).

The active constituents of the unsegmented PA extracted from P. americana L. were meticulously identified and characterized using Nano LC-MS/MS analysis. The comprehensive total ion chromatogram of these active compounds is depicted in Supplementary Figure S1, which recognizes a total of 438 distinct sequences (Supplementary Table S1). From this dataset, sequences that presented with a sequence score exceeding 200 and a mass accuracy deviation beyond ± 5.0 ppm were rigorously excluded. This filtration process culminated in the identification of the top 30 active component sequences, selected based on their peak area intensity, and these are systematically cataloged in Supplementary Table S2.

A total of 1,326 molecular compounds were identified by LC-MS, with detailed annotations provided in Supplementary Table S3. This chemically diverse profile encompassed phenolic compounds, amino acids and derivatives, flavonoids, fatty acyls, purine nucleosides, carboxylic acids and derivatives, alkaloids, aromatic compounds, sugars, coumarins, and plant-derived metabolites, among other classes. Notably, the following courses exhibited significant relative abundance: phenolics (25.45%), amino acids and derivatives (18.80%), nucleotides and derivatives (8.78%), carboxylic acids and derivatives (8.27%), and alkaloids (7.92%). The 30 compounds displaying the highest peak areas were rigorously selected, and their chromatographic data are presented in Supplementary Table S4. Additionally, Supplementary Figure S2 presents the total ion chromatograms (TIC) of PA extract constituents, visually illustrating the compound distribution profile.

Compared with the control mice, the UC mice exhibited a significant reduction in body weight and an increase in DAI scores (Figures 1B,C), which was reversed by administration of olsalazine, PA, and its molecular weight fractions (PA_3KD, PA_3-10KD, and PA_10KD). Remarkably, the group treated with the PA exhibited the most pronounced improvement. The indices of the thymus and spleen were also evaluated, with the UC group showing a significantly diminished thymus index and an elevated spleen index. In contrast, the treatment groups showed the opposite trend (Figures 1D,E), with the PA once again showing the most substantial therapeutic impact. Compared with the control group, the colon in the UC group exhibited a significantly shortened colon length, attributed to congestion and edema, which are key indicators of UC severity, as shown in Figures 1F,G. The PA and its fractionated components (<3, 3–10, and > 10 kDa) effectively mitigated DSS-induced colonic congestion and atrophy, thereby substantiating the therapeutic potential of PA in the treatment of UC. These findings strongly support the potential application of PA in the management and treatment of ulcerative colitis.

H&E staining revealed the extent of colonic tissue damage and the potential protective role of PA components. In the control group, the structural integrity of the colonic tissues, including the mucosal, submucosal, and muscular layers of the mice, was intact. In contrast, mice in the UC group exhibited extensive crypt and goblet cell loss, significant inflammatory cell infiltration, and damage to the submucosal structures (Figures 2A,B). Furthermore, mice in the UC group exhibited significantly higher epithelial cell integrity and inflammatory cell scores than those in the control group. PA treatment reduced inflammatory cell infiltration, protected crypt and epithelial cell structures, and significantly decreased the colonic histology score (Figure 2D).

AB-PAS staining, which targets mucopolysaccharides and goblet cells, was an essential morphological marker for evaluating intestinal function and structure (Cheng et al., 2023a). The goblet cells of mice in the control group exhibited robust morphology and were densely arrayed flanking the crypts (Figure 2C). Conversely, those of the mice in the UC group were significantly diminished, with significantly lower relative expression rates of positive cells (Figure 2E). A significant increase in goblet cell expression was observed in the mice in the PA and PA_3KD groups than in those in the UC group following the therapeutic agent administration. This suggests that the PA components alleviated the morphological damage and restored the functional integrity of the colonic epithelium.

Immunohistochemical staining revealed significant insights into the barrier function of the colonic mucosa (Figure 3). NF-κB-p65, TLR4, and MyD88 are essential in the inflammatory signaling cascade implicated in UC pathogenesis. They modulate the inflammatory response and promote pro-inflammatory cytokine synthesis, exacerbating colonic epithelial barrier dysfunction (Subramaniam et al., 2016). We observed increased NF-κB-p65, TLR4, and MyD88 expression levels among mice in the UC group, indicating an increased inflammatory response and intestinal barrier function deterioration. However, drug treatment significantly reduced the expression of these proteins in colonic tissue, with the PA blend yielding the most significant reduction (Figures 3A–D).

To investigate whether the anti-inflammatory effects of PA are associated with the NF-κB p65 signaling pathway, the expression levels of NF-κB p65 in both the nucleus and cytoplasm were assessed. As illustrated in Figures 3E,F, there was an overexpression of NF-κB p65 in the nucleus and a concomitant reduction in the cytoplasm within the colon tissues of the UC group. This observation suggests that UC facilitates the nuclear translocation of NF-κB p65, thereby activating the NF-κB p65 pathway (P < 0.05). Conversely, PA treatment markedly inhibited the nuclear translocation of NF-κB p65 in mice with colitis. These findings indicate that PA treatment mitigates the inflammatory response by suppressing the activation of the NF-κB p65 signaling pathway.

Occludin-1 and ZO-1 were identified as key components of the tight junctions that constitute the mechanical barrier between intestinal epithelial cells. The mice in the UC group demonstrated significantly reduced occludin-1 and ZO-1 expression levels than those in the control group, indicating a breakdown of intestinal barrier integrity (Figures 4A–C). However, PA treatment significantly increased occludin-1 and ZO-1 expression, indicating intestinal barrier restoration. The PA significantly increased occludin-1 levels. These findings revealed that PA components are efficacious in mitigating the DSS-induced intestinal barrier dysfunction in UC, highlighting their potential therapeutic value.

The concentration of inflammatory cytokines within the colonic mucosa is a strong indicator of UC disease severity (Cheng et al., 2023b). Figure 5 indicates that UC-affected mice have significantly higher levels of pro-inflammatory mediators, including TNF-α, IL-6, IL-17A, and INF-γ, than those in the control group (Figures 5A–D). Simultaneously, increased MPO and INOS expressions were observed in the UC cohort (Figures 5E,F). Overproduction of these pro-inflammatory cytokines was considered a central factor in UC development and exacerbation. This study demonstrates that the administration of PA and its various molecular weight fractions significantly reduced the levels of these inflammatory mediators in mice in the PA groups than in those in the untreated UC group. The complete PA extract was more effective than its components in diminishing the levels of pro-inflammatory cytokines, MPO, and INOS, suggesting a synergistic effect of the combined ingredients in PA anti-inflammatory action.

In contrast to pro-inflammatory mediators, we found that post-treatment levels of anti-inflammatory cytokines (IL-4 and IL-10) were significantly higher in the mice in the UC model group than in those in the control group (Figures 5G,H). PA administration was significantly effective in boosting these anti-inflammatory cytokine levels. These findings indicate that the PA extract reduces inflammation in the UC mouse model by suppressing pro-inflammatory factors expression and increasing anti-inflammatory mediators’ expression. Accordingly, the extract may play a dual modulatory role in the inflammatory response, potentially contributing to its therapeutic efficacy in UC management.

The utilization of 16S rRNA gene sequencing to examine the effect of PA on intestinal microflora diversity and composition in a DSS-induced UC mouse model highlights the role of the gut microbiome in UC and the potential therapeutic effects of PA. These findings demonstrated that the sequencing effort was sufficient to capture a representative snapshot of microbial richness across the samples, as indicated by the plateauing of the rarefaction curves (Figures 6A,B). Microbiome studies need this validation step to ensure that further analysis is based on a comprehensive dataset.

The PCoA and NMDS analyses (Figures 6C,D) are multivariate statistical techniques used for reducing the dimensionality of complex data sets. The significant spatial separation between mice in the UC model group and those in the control group in these analyses indicates that UC causes a distinct shift in the gut microbiota composition. The fact that the gut microbiota composition of mice in the groups treated with various molecular weight fractions of PA is closer to that of mice in the control group suggests that these treatments can alter the gut microbiota to mimic that of the healthy controls.

The identification of 263 OTUs across all samples, with 183 OTUs shared by all groups, indicates that a core microbiome exists under all conditions. The unique OTUs found in each treatment group (Figure 6E) may be of particular interest because they may represent specific bacterial taxa associated with the therapeutic effects of PA and its fractions.

The alpha diversity analysis (Figures 6F–I), utilizing the Kruskal–Wallis test, indicated no significant differences in the Ace (H = 2.412, df = 6, P = 0.878), Chao1 (H = 2.230, df = 6, P = 0.897), Shannon (H = 4.777, df = 6, P = 0.573), and Simpson (H = 8.726, df = 6, P = 0.190) indices across groups. This lack of significant variation was corroborated by Dunn’s test with Benjamini–Hochberg correction, which also revealed no significant differences between groups (all adjusted P > 0.05). These findings suggest that DSS-induced colitis and PA treatment did not significantly impact overall microbial abundance and homogeneity. However, the moderate effect sizes (ε ² = 0.12–0.21) indicate that further validation with an expanded sample size is necessary to explore potential underlying biological trends. Detailed results are presented in Tables 1 and 2.

This study identified species from 11 phyla, 17 classes, 45 orders, 78 families, 170 genera, and 654 OTUs.Through a comprehensive multilevel taxonomic analysis encompassing the phylum, family, and genus levels, distinct variations in the floral structure among the groups were identified (Figures 7A–C). At the phylum level, the Control group was predominantly composed of Firmicutes (63.2%) and Proteobacteria (23.5%). In contrast, the UC group exhibited a marked increase in Desulfobacterota (17.4% vs. 2.03%) and Actinobacteriota (14.08% vs. 5.60%), alongside a significant reduction in Proteobacteria (0.06% vs. 23.5%). The Kruskal–Wallis test indicated significant intergroup differences across five phyla, notably Proteobacteria (H = 16.133, P = 0.013, ε ² = 0.393) and Verrucomicrobiota (H = 25.689, P < 0.001, ε ² = 0.627), as presented in Table 3. Notably, the PA intervention significantly reinstated the abundance of Proteobacteria (compared to UC: H = 19.083, Z = 2.695, P = 0.007). Furthermore, both PA and PA_10KD interventions effectively mitigated the abnormal increases in Deferribacterota (H = 17.750/14.667, Z = 3.145/2.598, P ≤ 0.009) and Desulfobacterota (H = 16.667/19.333, Z = 2.353/2.730, P ≤ 0.019), as detailed in Table 4. At the family level (Table 5), the UC group exhibited a significant enrichment in Akkermansiaceae (H = 23.500, Z = 3.321, P = 0.001), Bifidobacteriaceae (H = 23.083, Z = 3.326, P = 0.001), and Erysipelotrichaceae (H = 20.917, Z = 2.953, P = 0.003), while Moraxellaceae (H = 22.500, Z = 3.206, P = 0.001) was significantly reduced. This microbial imbalance was effectively reversed by PA and PA_3-10KD treatments. At the genus level (Table 6), there was a significant elevation in Akkermansia, Bifidobacterium, Dubosiella and Faecalibaculum within the UC group (P < 0.01). However, PA treatment significantly reduced the abundance of Bifidobacterium (H = 26,000, Z = 3,746, P < 0.001), Dubosiella (H = 24,750, Z = 3,496, P < 0.001), and Faecalibaculum (H = 22,500, Z = 3,188, P < 0.001). These findings suggest that PA may ameliorate UC pathology by modulating specific microbial communities.

A comprehensive analysis was performed on all samples at the phylum to genus level, utilizing the LefSe method with an LDA threshold of >3.0 to distinguish distinct bacterial communities under varying conditions. The results were presented using branching plots and histograms to illustrate the distribution of LDA values (Figures 7D,E). Branching plots revealed significant differences in gut microbial taxa between groups, while histograms of LDA scores revealed the degree of influence of significantly different species, resulting in the identification of 62 microbial markers. The number of microorganisms that differed between the control, UC, oxalazine, PA, PA_3KD, PA_3-10KD, and PA_10KD groups were 8, 12, 12, 11, 6, 9, and 4. Among them, four key genera, Psychrobacter, Limosilactobacillus, unclassified, and Jeotgalicoccus, were found in the mice in the control group, while those found in the mice in the UC group included Dubosiella, Desulfovibrionia, Bifidobacterium, and Clostridium sensu stricto 1.

The above findings suggest that PA effectively reduces intestinal inflammation and modulates intestinal flora composition in UC mice. However, the potential impact of PA on the relationship or coherence between these two factors remains unclear. The effects of cytokines on gut microbiota composition were assessed using canonical correlation analysis (RDA/CCA), db-RDA, and Spearman correlation heatmaps (Figures 8A–C). In RDA/CCA and db-RDA analyses microbes in the gut exhibited significant positive correlations with anti-inflammatory factors (IL-4 and IL-10) and negative correlations with pro-inflammatory cytokines (TNF-α, IL-6, IL-17A, INF-γ, INOS, and MPO). Among them, IL-10, IL-17A, and MPO exhibited the most significant influence on the changes in intestinal flora in UC mice. The Spearman correlation heatmaps depicted in Figure 8C indicate a positive correlation between Psychrobacter and anti-inflammatory cytokines (IL-4 and IL-10), suggesting that Psychrobacter may play a role in the regulation of inflammation and intestinal flora composition. Conversely, Dubosiella, Bifidobacterium, Akkermansia, and Desulfovibrio exhibited contrasting associations. These findings imply that modulating the abundance and composition of Psychrobacter, Dubosiella, Bifidobacterium, Akkermansia, and Desulfovibrio flora in UC mice may provide a therapeutic approach to alleviate DSS-induced intestinal inflammation.

The functions of microbial communities were predicted using PICRUSt2 software combined with EggNOG (evolutionary genealogy of genes: non-supervised orthologous groups) and Kyoto encyclopedia of genes and genomes (KEGG) databases based on the 16S rRNA amplicon sequencing results (Figures 8D–F). Microorganisms predicted from the eggNOG database have known functions primarily related to biosynthesis and metabolism, with translation, ribosomal structure, and biogenesis accounting for the highest abundance. Similarly, KEGG function predictions indicate that microbial community function is primarily associated with carbohydrate, amino acid, energy, and nucleotide metabolic pathways. When combined with the informative expression of KEGG data at a deeper (tertiary) level, the microbial association with secondary metabolites and amino acid biosynthesis, carbon metabolism, purine metabolism, and glycolysis was confirmed.