Female C57BL/6J mice and female Pdcd1lg1 knockout C57BL/6J mice (Pdcd1lg1 -/- mice), aged 6–8 weeks, were obtained from Cyagen Biosciences, Suzhou, China. All experimental procedures were performed according to the guidelines of the Animal Care and Use Committee of Fudan University (Shanghai, China). Animals were housed at the Experimental Animal Center of the Eye and ENT Hospital of Fudan University with a temperature of 22-26°C, 12-hour light/dark cycles, and relative humidity of 40-70%.

Then, animals were randomly divided into four groups (n=3/group) including wt_control (wild-type without DSS), ko_control (Pdcd1lg1 -/- without DSS), wt_DSS (wild-type with DSS-induced colitis), and ko_DSS (Pdcd1lg1 -/- with DSS-induced colitis) (Dextran Sulfate Sodium, Yeasen biotech Co., Ltd., Shanghai, China). Colitis was induced in mice by daily administration of 3% DSS for 7 days, while the control group received equivalent volumes of sterile distilled water ( Figure 1A ). Body weights were recorded at baseline (day 0) and on day1, 3, 5, and 7 post-modeling. On day 7, mice were sacrificed via cervical dislocation under anesthesia (1.25% bromethol, 0.2 mL/10g body weight administered intraperitoneally; Aibei Biotechnology Co., Ltd., Nanjing, China). Collect fecal specimens and obtain colonic samples extending from the cecum junction to the anal verge. Immediately after collection, the fecal specimens were flash-frozen in liquid nitrogen and then stored at -80°C to maintain microbial viability and DNA integrity for subsequent sequencing analysis. Colon length was measured along the mesenteric border before fixation in 4% paraformaldehyde, followed by paraffin embedding and hematoxylin & eosin (H&E) staining. The disease activity index (DAI) was calculated to assess the severity of UC in animal models in light of the study (22). DAI = (Weight loss score + Stool score + Bleeding score)/3.

Total microbial genomic DNA was extracted from fecal samples using the OMEGA Mag-Bind Soil DNA Kit (M5635-02; Omega Bio-Tek, GA, USA) according to the manufacturer’s protocol, and stored at -20°C until further assessment. DNA concentration and purity were quantified using a Qubit™ 4 Fluorometer (Thermo Fisher Scientific Inc., MA, USA), while integrity was verified by agarose gel electrophoresis. The extracted microbial DNA was processed to construct metagenome shotgun sequencing libraries with insert sizes of ~400 bp by using Illumina TruSeq Nano DNA LT Library Preparation Kit (Illumina, CA, USA). Each library was sequenced by Illumina NovaSeq platform (Illumina, CA, USA) with PE150 strategy at Personal Biotechnology Co., Ltd. (Shanghai, China).

Raw sequencing reads were processed through adapter removal with Cutadapt (v1.2.1) (23), sliding-window quality trimming using fastp (v0.23.2) (24), and host DNA removal by alignment to the mouse genome via Minimap2 (v2.24-r1122) to generate quality-filtered data (25). Once quality-filtered reads were obtained, taxonomical classifications of metagenomics sequencing reads from each sample were performed using Kraken2 (v2.0.8 beta) (26) against databases including NCBI-nt, GTDB and RVDB. Reads assigned to metazoans or viridiplantae were removed for downstream analysis. Megahit (v1.1.2) (27) was used for assembly of reads in each sample using the meta-large presetted parameters. The generated contigs (longer than 300bp) were then pooled together and clustered using MMseqs2 (version 13.45111) (28) with “easy-linclust” mode, setting sequence identity threshold to 0.95 and covered residues of the shorter contig to 90%. The lowest common ancestor taxonomy of the non-redundant contigs was obtained by aligning them against the NCBI-nt database by MMseqs2 with “taxonomy” mode, and contigs assigned to Viridiplantae or Metazoa were dropped in the following analysis. Prodigal (V2.6.3) (29) was used to predict the genes in the contigs. CDS sequences of all samples were clustered by MMseqs2 with “easy-cluster” mode, setting protein sequence identity threshold to 0.95 and covered residues of the shorter contig to 90%. To assess the abundances of these genes, the high-quality reads from each sample were mapped onto the predicted gene sequences using Minimap2 with “-ax sr –sam-hit-only” and using featureCounts to count the number of reads aligned to gene sequences, i.e., the Reads Count (RC) (30) for each gene. The functionality of the non-redundant genes was obtained by annotated using MMseqs2 with the “search” mode against the protein databases of KEGG.

In this study, statistical tests (Kruskal-Wallis test, ANOVA) were conducted using R software (version 3.6.1). Based on the taxonomic and functional profiles of non-redundant genes, LEfSe (Linear discriminant analysis effect size) was performed to detect differentially abundant taxa and functions across groups using the default parameters (31). Beta diversity analysis was performed to investigate the compositional and functional variation of microbial communities across samples using Bray-Curtis distance metrics (32) and visualized via principal coordinate analysis (PCoA), nonmetric multidimensional scaling (NMDS) and unweighted pair-group method with arithmetic means (UPGMA) hierarchical clustering (33). P < 0.05, P < 0.01 and P < 0.001 were considered significant. GraphPad Prism 9.0 software (GraphPad Software Inc, La Jolla, CA, USA) was used to draw images.

All mice exhibited normal activity, dietary intake, stool consistency, and weight gain before modelling. In comparison with those in the control group, the mice in the model group showed not only significantly decreased body weight ( Figure 1B ) and colon length ( Figures 1D, E ), but also observably increased DAI scores ( Figure 1C ). Both wt_DSS and ko_DSS groups experienced significant weight loss compared to their respective controls (p<0.05). Notably, the ko_DSS group showed more pronounced weight loss than the wt_DSS group (p<0.05), indicating exacerbated colitis in Pdcd1lg1 -/- mice ( Figure 1B ). The wt_DSS group had shorter colons than the wt control (p<0.05), reflecting colitis-induced shortening. The ko_DSS group exhibited the shortest colons (p<0.05 vs. wt_DSS), suggesting more severe colitis in Pdcd1lg1 -/- mice ( Figures 1D, E ). Both wt_DSS and ko_DSS groups had higher DAI scores than wt group (p<0.05). The ko_DSS group had the highest DAI scores (p<0.05 vs. wt_DSS), indicating the most severe colitis. In addition, the macroscopic morphology of colonic tissues showed colon edema, a reduced caecal volume and an enteric cavity with bloody stools. The ko_DSS group mice exhibited severed colonic shortening and colonic edema than both wt group mice and wt_DSS group mice ( Figure 1F ).

We performed metagenomic sequencing on fecal samples from four groups of mice: wt, wt_DSS, ko, and ko_DSS. Our analysis revealed distinct gut microbiota compositions among the groups. We focused on microbes with an abundance of 30. In the colitis model groups (wt_DSS and ko_DSS), the gut microbiota shifted notably. Specifically, Escherichia coli and Bacteroides increased significantly in the ko_DSS group, while Tritrichomonas was more abundant in the wt_DSS group. In the control groups, the gut microbiota also differed between the wt and ko groups. The ko group had higher abundances of Limosilactobacillus and CAG-485 sp002361215, whereas the wt group showed higher levels of Dubosiella, Akkermansia, and CAG-485 sp002491945 ( Figure 2A ). A PCoA plot ( Figure 2B ) and a clustering heatmap ( Figure 2C ) both highlighted the distinct compositional differences between groups. Alpha diversity analysis ( Figure 2D ) indicated no significant differences in some indices, but notable changes in others, such as the trend in Chao1 diversity. Species differential analysis revealed that Escherichia levels varied significantly across groups (p < 0.05). While Odoribacter sp910589025, Limosilactobacillus reuteri, Akkermansia muciniphila A, Ligilactobacillus murinus, and Tritrichomonas foetus did not show statistical differences, their abundances changed markedly ( Figure 2E ). Finally, LEfSe analysis identified differentially abundant gut microbial taxa at various taxonomic levels, pinpointing specific bacterial genera or families that changed significantly in each group ( Figure 2F ).

This study primarily examines the alterations in gut microbiota composition between the wt_DSS and ko_DSS groups.

Evaluation of gut microbiota in the DSS-induced colitis model revealed alterations in both the wt_DSS and ko_DSS groups ( Figures 3A, B ). The two groups shared 8,049 bacterial genera, while the wt_DSS and ko_DSS groups exhibited 1,509 and 4,258 unique genera, respectively ( Figure 3D ). Alpha diversity indices were used to assess the gut microbiota, showing significant differences in richness (Chao1, Observed species, and ACE) between the two groups (p = 0.05). However, no significant differences were observed in diversity (Simpson index), evenness (Pielou’s evenness), or coverage (Good’s coverage) ( Figure 3C ). The wt_DSS group showed higher enrichment of Schaedlerella sp009911175, Acetatifactor sp910589655, Kineothrix sp000403275, Bacteroides sp00249163, UBA3263 sp001689615, Bacteroides acidifaciens, CAG-485sp002361215, Acetatifactor sp003612485, Bacteroides thetaiotaomicron, Bacteroides caecimuris, Escherichia coli, CAG-873 sp009775265, CAG-510 sp003979355, CAG-95 sp910587295, Acetatifactor sp910585015, and UBA3282 sp009774585. In contrast, Ligilactobacillus murinus, Paraprevotella sp910586505, Mucispirillum schaedleri, Eubacterium J sp910577225, Prevotella sp002298815, CAG-115 sp910587465, Tritrichomonas foetus, CAG-485 sp002493045, Odoribacter sp910589025, CAG-485 sp009775375, Cryptobacteroides sp910585445, Sporofaciens sp910574715, COE1 sp009774375, and CAG-557 sp000435275 were less enriched. The ko_DSS group exhibited the opposite trend ( Figure 3E ). Further differential species analysis identified significantly higher enrichment of Escherichia coli (p = 0.0058), CAG-510 sp003979355 (p = 0.010), UBA3263 sp001689615 (p 0.027), and Bacteroides thetaiotaomicron (p = 0.015). Conversely, Tritrichomonas foetus (p < 0.001), CAG-557 sp000435275 (p < 0.001), Paraprevotella sp910586505 (p = 0.003), Cryptobacteroides sp910585445 (p = 0.0044), Odoribacter sp910589025 (p = 0.0069), and Prevotella sp002298815 (p = 0.008) were significantly less enriched ( Figure 3F ).

We next performed functional analysis of the gut microbiota in colitis model groups. The intergroup differences in microbial composition significantly altered gene functional profiles. Statistical analysis of feature tables using the KEGG database enabled visualization of compositional distribution at functional classification levels (top 30 categories), presented as a bar chart ( Figure 4A ). Comparative analysis revealed 4,898 shared functional genes between groups, with 394 and 1,937 genes uniquely identified in the ko_DSS and wt_DSS groups, respectively ( Figure 4B ). Subsequently, LEfSe (LDA Effect Size) analysis was employed to rank differentially enriched functional pathways between ko_DSS and wt_DSS groups. Results demonstrated that the ko_DSS group exhibited significant enrichment of pathways associated with pro-inflammatory and pathogenic functions, including: Bacterial secretion systems (Type III/IV), linked to virulence traits in Escherichia coli and CAG taxa. ABC transporters (e.g., K02003/K02004) and drug resistance pathways, suggesting enhanced antibiotic resistance in ko_DSS. Immune dysregulation-related pathways, such as the lysosome pathway (associated with heightened phagocytic activity due to inflammation) and toxification/detoxification (reflecting oxidative stress in chronic colitis). In contrast, the wt_DSS group showed enrichment of pathways supporting anti-inflammatory and metabolic homeostasis, including: Short-chain fatty acid (SCFA) metabolism (e.g., butyrate synthesis via K01990), which promotes Treg differentiation and epithelial barrier integrity. Carbohydrate metabolism (e.g., glycosyl hydrolases), driven by dietary fiber fermentation from Ligilactobacillus and Bacteroides. Ribosome biosynthesis, indicative of stable microbiome-protein synthesis. Commensal-mucosal crosstalk pathways (e.g., arginine/proline metabolism) modulating nitric oxide (NO) production and immune signaling ( Figures 4C, D ). Further normalization of functional unit abundances revealed statistically significant enrichment of pro-inflammatory/virulence-associated genes in ko_DSS (e.g., K07133 [Type III secretion system, p = 0.038], K02529 [uncharacterized protein, p = 0.030], and K01190 [glycoside hydrolase, p = 0.046]). The wt_DSS group harbored higher abundances of anti-inflammatory/commensal genes (e.g., K01990 [acetyl-CoA synthetase] and K01153 [glycosyl hydrolase]), though these lacked statistical significance ( Figure 4F ). Finally, linear regression analysis of functional and taxonomic correlations identified a significant negative association (p = 0.003) between an immune variable (x; e.g., inflammatory cytokine levels) and microbial/metabolic responses (y; e.g., beneficial taxon abundance), underscoring PD-L1’s role in microbiota-immune crosstalk during colitis ( Figure 4E ).

Finally, we performed gene-level differential analysis of fecal samples from the mouse model. Compared to the wt_DSS group, ko_DSS group exhibited 222,335 genes upregulated and 59,504 genes downregulated genes ( Figure 5A ). KEGG enrichment analysis of these differentially expressed genes (DEGs) revealed significant associations with immune/inflammatory pathways in ko_DSS, including:IL-17 signaling, Lipopolysaccharide biosynthesis, Endocrine resistance. Additionally, infection/disease-related pathways were enriched, such as COVID-19, HIV infection and Salmonella infection. Notably, metabolic and cellular dysregulation pathways, including thermogenesis and lysosome, were also altered ( Figures 5B, C ), suggesting shifts in response to DSS-induced inflammation in Pdcd1lg1 deficient mice.