All bioinformatics analyses in the present study were performed using R software (version 4.3.3; R Foundation for Statistical Computing). Furthermore, all bioinformatics packages described here and below (including 'harmony', 'edgeR', 'limma', 'DESeq2', 'WGCNA' and 'clusterProfiler') are R-based packages. scRNA-seq data from three human UC samples were downloaded from the Gene Expression Omnibus (GEO) database (https://www.ncbi.nlm.nih.gov/geo/; GSE116222) (22) and processed using the 'Seurat' package (v4.3.0) within the R environment. Data quality was assessed, normalized using the 'NormalizeData' function, and the top 2,000 highly variable genes were identified with 'FindVariableFeatures'. Batch effects were corrected using the 'harmony' package (v1.2.0) (23). Cell clustering was performed with 'FindNeighbors' and 'FindClusters' (resolution= 0.8) and visualized via Uniform Manifold Approximation and Projection. Cell types were annotated based on marker genes from the CellMarker database (http://xteam.xbio.top/CellMarker/). Furthermore, the AddModuleScore function implemented in the 'Seurat' package (v4.3.0) was utilized to evaluate the activity of specific signaling pathways (such as the NF-κB pathway) at the single-cell level. This function calculates the average expression levels of a predefined target gene set for each cell, calibrated against the aggregated expression of randomly selected control gene sets.

Bulk RNA-seq datasets from control individuals and patients with IBD, including GSE75214 (97 UC, 8 CD and 11 controls) (24), GSE13367 (16 inflamed UC, 18 non-inflamed UC and 20 controls) (25) and GSE87466 (87 UC and 21 controls) (26), were obtained from GEO. Differentially expressed genes (DEGs) in these datasets were identified using the 'edgeR' (v3.40.0), 'limma' (v3.54.0) and 'DESeq2' (v1.38.0) packages in R, with screening criteria of |log2FC|>1 and P<0.05 (27-30). The intersection of the results from these methods was calculated using the 'intersect()' function in R, and the overlapping genes were defined as the final DEG set.

Weighted Gene Co-expression Network Analysis (WGCNA) offers a comprehensive analytical framework for clustering genes sharing analogous expression patterns (31). In the present study, the 'WGCNA' package in R was employed to construct a co-expression network based on the GSE116222 dataset, specifically exploring the association between FXR and CD133 in intestinal cells. Subsequently, a correlation matrix was generated using the WGCNA package to analyze the association between FXR and distinct modules derived from GSE87466. This specific dataset was selected for WGCNA and subsequent Gene Set Enrichment Analysis (GSEA) because it provides a large, focused cohort strictly comparing UC and healthy controls (n=108). This avoids the confounding disease heterogeneity (such as the presence of CD) or varying inflammatory states present in the other cohorts, thereby ensuring optimal statistical power and biological clarity for network construction and pathway analysis. After selecting an optimal soft threshold (power=8) based on scale-free topology criteria (R²>0.85), the correlation matrix was transformed into an adjacency matrix, and a topological overlap matrix (TOM) was derived. Genes demonstrating similar expression profiles were clustered into cohesive modules through average linkage hierarchical clustering, employing the TOM-based dissimilarity measure. Modules with a dissimilarity index <0.25 were merged. The modules exhibiting the strongest positive or negative correlations with FXR expression, characterized by the highest absolute Pearson correlation coefficients and showing statistical significance (P<0.05), were selected as the focal points for deeper investigation.

The GSE87466 cohort was stratified into high and low expression groups based on median FXR expression. Differential expression analysis was performed using 'DESeq2'. GSEA was then conducted on the entire transcriptome using the 'ClusterProfiler' package, with a significance threshold of P<0.05.

Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analyses were performed on the key module genes identified via WGCNA. The analyses were conducted using the 'clusterProfiler' package (v4.6.0) in R. The GO database (http://geneontology.org/) was used for annotating biological processes, cellular components and molecular functions, while the KEGG database (https://www.genome.jp/kegg/) was utilized to identify significantly enriched signaling pathways. The threshold for identifying statistically significant GO terms and KEGG pathways was set at an adjusted P<0.05.

DSS (molecular weight, 36-50 kDa) was purchased from MP Biomedicals, LLC. Anti-FXR (cat. no. 25055-1-AP), anti-leucine-rich repeat-containing G-protein coupled receptor 5 (LGR5) antibody (cat. no. 30007-1-AP, for western blotting), anti-CD133 (cat. no. 18470-1-AP), anti-SALL4 (cat. no. 24500-1-AP), anti-TNF-α (cat. no. 17590-1-AP), anti-IL-1β (cat. no. 16806-1-AP) and anti-NF-κB p65 (cat. no. 66535-1-Ig) antibodies were purchased from Proteintech Group, Inc. Anti-LGR5 [cat. no. YM9275, for immunohistochemistry (IHC) and immunofluorescence (IF)], anti-OCT3/4 (cat. no. YM0485), anti-NANOG (cat. no. YM8110) and anti-SOX2 (cat. no. YM9292) antibodies were purchased from ImmunoWay Biotechnology Company. GAPDH antibody (cat. no. AF0006), α-tubulin antibody (cat. no. AF2831) and RIPA lysis buffer (cat. no. P0013B) were purchased from Beyotime Biotechnology. The mouse/rabbit triple-target four-color multiplex immunofluorescence kit (cat. no. RS0035) was purchased from ImmunoWay Biotechnology Company. The kit includes the HRP-polymer anti-rabbit/mouse secondary antibody, 488-labeled Tyramide (green fluorescence), 594-labeled Tyramide (red fluorescence), anti-fade mounting medium with DAPI and high-temperature antibody stripping buffer. The ready-to-use streptavidin-biotin complex-peroxidase Mouse/Rabbit IgG immunohistochemistry staining kit (cat. no. SA1020) (containing a ready-to-use HRP-conjugated universal anti-mouse/anti-rabbit secondary antibody) was purchased from Boster Biological Technology. DMEM high-glucose (cat. no. KGL1211) was purchased from Nanjing KeyGen Biotech Co., Ltd., and penicillin-streptomycin solution (cat. no. VC2003) was purchased from VICMED. The Tanon High-sig ECL Western Blotting Substrate (cat. no. 180-5001) was purchased from Tanon Science & Technology Co., Ltd. A chemiluminescence instrument (Nanjing Zhiheng Intelligent Technology Co., Ltd.) was utilized for gel imaging.

FXR global knockout (KO; FXR^−/−) mice (C57BL/6 background) were obtained from Jackson Laboratory. Wild-type (WT) C57BL/6 mice were purchased from GemPharmatech Co. Ltd. All mice were housed in a specific pathogen free-grade animal facility under standard controlled conditions (temperature, 22±2°C; humidity, 50±10%; 12/12-h light/dark cycle) with ad libitum access to standard rodent chow and water. After 1 week of acclimatization, a total of 24 male mice aged 8-10 weeks, weighing 20-22 g, were selected for experiments (n=6 per group). All animal experimental procedures were approved by the Medical Ethics Committee of Anhui Medical University (approval no. LLSC-20232140; Hefei, China).

FXR^−/− and WT mice were randomly assigned to control or model groups. Model group mice received 3% DSS in sterile drinking water ad libitum for 7 days, followed by normal water for 3 days. Control mice received normal water throughout. Body weight, stool consistency and rectal bleeding were recorded daily. To minimize animal suffering, humane endpoints were established in accordance with ethical guidelines. Criteria for immediate euthanasia included rapid weight loss >20% of the initial body weight, severe lethargy, inability to access food or water or persistent severe rectal prolapse. During the experimental period, no mice met these predefined humane endpoints prior to the scheduled experimental endpoint.

On day 11, mice were euthanized by cervical dislocation. Colon length (cecum to anus) was measured. Mid-colon segments were harvested and briefly washed with cold DMEM (high glucose) supplemented with 1% penicillin-streptomycin to remove fecal contents and prevent bacterial contamination. One part was fixed in 4% paraformaldehyde at room temperature for 24 h, paraffin-embedded and sectioned (4 µm) for H&E staining, IHC and IF, and the other part was snap-frozen in liquid nitrogen for western blotting.

Human colon tissue samples were prospectively collected from 30 patients at Xuzhou Rehabilitation Hospital (Xuzhou, China) between June 2023 and June 2025. The cohort included 20 patients with pathologically confirmed active UC and 10 healthy controls. The healthy control samples were obtained from macroscopically normal tissues of patients undergoing screening colonoscopy or surgery for non-inflammatory, non-malignant conditions. The UC group comprised 13 men and 7 women (mean age, 45.6±8.4 years), while the control group consisted of 8 men and 2 men (mean age, 46.2±7.5 years).

The inclusion criteria for patients with UC were confirmed clinical, endoscopic and histological diagnosis of UC. The exclusion criteria included: i) CD or indeterminate colitis; ii) colorectal cancer or other intestinal malignancies; iii) severe systemic infections; and iv) prior treatment with targeted biologics. All procedures were approved by the Medical Ethics Committee of Xuzhou Rehabilitation Hospital (approval no. XK-LW-20230529-014).

Paraffin sections were deparaffinized in xylene, rehydrated in descending ethanol series and subjected to antigen retrieval in 10 mM citrate buffer (pH 6.0) at 100°C. Endogenous peroxidase was blocked with 3% H₂O₂. Sections were blocked with 5% BSA (Sigma-Aldrich; Merck KGaA) at room temperature for 30 min. Subsequently, multiplex immunofluorescence was performed using the mouse/rabbit triple-target four-color multiplex immunofluorescence kit according to the manufacturer's instructions. Briefly, sections were subjected to sequential staining cycles. In each cycle, tissue sections were incubated overnight at 4°C with the primary antibody (rabbit anti-FXR at 1:200, rabbit anti-CD133 at 1:150 or rabbit anti-LGR5 at 1:1,000). After washing with PBS containing 0.1% Tween-20, sections were incubated with the kit-provided HRP-polymer anti-rabbit/mouse secondary antibody. The fluorescent signals were then developed using specific fluorophore-conjugated Tyramide substrates (488-labeled Tyramide for green fluorescence or 594-labeled Tyramide for red fluorescence). After signal visualization, the slides were subjected to microwave heating in the high-temperature antibody stripping buffer to remove the bound antibodies before initiating the next staining cycle for the subsequent target. Following the completion of all staining cycles, sections were mounted and nuclei were counterstained simultaneously using the ready-to-use anti-fade mounting medium with DAPI at room temperature. Images were subsequently acquired using a confocal microscope and analyzed with ImageJ software (version 1.54; National Institutes of Health).

Frozen colon tissue was homogenized in RIPA lysis buffer (containing 1% protease inhibitor and 1% phosphatase inhibitor) on ice, followed by centrifugation (12,000 × g, 4°C, 15 min) and collection of the supernatant. Protein concentration was determined using the BCA method. After adjusting the protein concentration in each sample, 5X loading buffer was added and samples were boiled at 95°C for 5 nim. A total of 30 µg of protein per sample was subjected to SDS-PAGE and then transferred to a PVDF membrane. Membranes were blocked with 5% skimmed milk at room temperature for 2 h, followed by incubation with primary antibodies (FXR, OCT3/4, NANOG, SOX2, SALL4, TNF-α, IL-1β and NF-κB p65, all at 1:1,000; GAPDH and α-tubulin at 1:1,000) overnight at 4°C. After washing with TBS containing 0.1% Tween-20 buffer, an appropriate HRP-conjugated secondary antibody (1:5,000; anti-rabbit or anti-mouse depending on the host species of the primary antibody) was applied and incubated at room temperature for 1 h. Signals were visualized using an ECL chemiluminescence kit, and images were captured using a gel imaging system. The grayscale value ratio of target proteins to the internal reference was quantified using ImageJ software (version 1.54).

For histological analysis, paraffinembedded colon sections (4 µm) were subjected to H&E and periodic acid-Schiff (PAS) staining. For H&E staining, sections were deparaffinized, rehydrated, and stained with hematoxylin for 5 min at room temperature, followed by eosin staining for 2 min at room temperature. For PAS staining, sections were incubated with periodic acid solution for 10 min at room temperature, rinsed with deionized water, and then stained with Schiff reagent for 15 min at room temperature in the dark.

For IHC, paraffin sections were deparaffinized in xylene and rehydrated in a descending ethanol series. Antigen retrieval was performed in 10 mM citrate buffer (pH 6.0) at 100°C for 15 min, followed by treatment with 3% H₂O₂ for 10 min to block endogenous peroxidase activity. Sections were then blocked with 5% BSA at room temperature for 30 min. Subsequently, sections were incubated overnight at 4°C with primary antibodies against OCT3/4, SOX2, LGR5 and FXR at 1:200. The kit-provided ready-to-use HRP-conjugated universal anti-mouse/rabbit secondary antibody was applied and incubated for 30 min at room temperature, followed by 3,3′-diaminobenzidine development and counterstaining with hematoxylin for 5 min at room temperature. Representative images were captured using a light microscope (Olympus Corporation).

Statistical analysis was performed using SPSS 26.0 software (IBM Corp.), and graphs were plotted using GraphPad Prism 8.0 (Dotmatics). Data are presented as the mean ± standard deviation. Comparisons between two groups were performed using an independent samples Student's t-test. Comparisons among multiple groups were performed using one-way ANOVA followed by Tukey's post hoc test. Correlation analyses between gene expression levels or module eigengenes were evaluated using Pearson's correlation coefficient. P<0.05 was considered to indicate a statistically significant difference.

Integrative analysis of bulk transcriptomic datasets (GSE13367, GSE75214 and GSE87466) revealed consistent and significant downregulation of FXR in human UC tissues compared with controls (Fig. 1A-C). Analysis of single-cell data (GSE116222) confirmed this finding at a cellular resolution, showing a gradient of FXR loss from normal to inflamed tissue (Fig. 1D). This cross-omics validation highlighted FXR downregulation as a robust feature of the colitis molecular landscape. To investigate FXR expression in vivo, a mouse model of DSS-induced IBD was established. The results showed that the length and weight of the large intestine in the DSS group were both significantly shorter than those of the control group (Fig. 1E-G). The spleen weight in the DSS group increased significantly (Fig. 1H). Concurrently, mouse body weight progressively decreased over time (Fig. 1I). IHC was further used to detect the expression and tissue localization of FXR in human and mouse colon tissue sections. The results revealed that the number of FXR⁺ cells and intensity of FXR staining decreased in the UC and DSS groups compared with those in the control groups, indicating downregulation of expression (Fig. 1J and K). Western blot analysis consistently showed downregulation of the FXR protein in DSS-treated mouse colon compared with the control group (Fig. 1L and M). These results demonstrated that FXR expression is downregulated in colitis.

While the multi-layer transcriptomic analysis revealed reduced FXR expression in colitis, the precise cellular populations affected remained unclear. Therefore, scRNA-seq data analysis was performed to localize FXR within the intestinal epithelium and assess its co-expression with the stem cell marker CD133 (32). Analysis of the scRNA-seq dataset GSE116222 revealed overlapping expression patterns of FXR with the canonical stem cell marker CD133 (PROM1) (Fig. 2A). Based on the CellMarker database, the single-cell data were annotated and categorized into five major cell types (Fig. 2B and C). Within the epithelial compartment, a CD133-high population was defined within the Marker-defined Stem Cell cluster (MSC_like cluster) and FXR expression was found to be specifically co-localized with this population (Fig. 2D). Simultaneously, immunofluorescence staining of mouse colon tissue revealed notable yellow overlapping signals between FXR (green fluorescence) and CD133 (red fluorescence) in colonic epithelial cells, indicating co-localization between FXR and the ISC marker CD133 at the cellular level (Fig. 2E). To definitively confirm the presence of FXR in the crypt base columnar (CBC) stem cell niche, additional in vivo validation was performed using the gold-standard ISC marker LGR5 (33). Immunofluorescence staining demonstrated that LGR5⁺ cells were strictly localized to the base of the intestinal crypts (Fig. 2F). Upon DSS-induced injury, a compensatory expansion of LGR5⁺ cells was observed in WT mice, representing a classical mucosal regenerative response. FXR expression exhibited marked co-localization with LGR5 within these CBC cells in DSS-induced mice, supporting its direct regulatory role within the active stem cell niche.

The FXR^−/− mouse model was used to functionally test the hypotheses generated from human omics data. The experimental design for the FXR^−/− mouse colitis model is shown in Fig. 3A. Tissue index measurements showed that compared with the WT-DSS group, the KO-DSS group had a markedly decreased colon weight and shortened colon length (Fig. 3B-D). The spleen weight in the KO-DSS group exhibited a substantial increase compared with that in the WT-DSS group (Fig. 3E). Furthermore, the body weight of KO-DSS mice progressively declined over time relative to that of the WT-DSS group (Fig. 3F). H&E and PAS staining results showed that the WT model group exhibited partial epithelial damage and inflammatory cell infiltration in the colonic mucosa. By contrast, the FXR^−/− model group displayed severe disruption of colonic mucosal structure, loss of epithelial layer integrity, massive inflammatory cell infiltration in the lamina propria, disorganized or even absent glands and notably aggravated pathological damage (Fig. 3G and H). These results confirmed the central pathogenic role of FXR loss inferred from transcriptional data.

To functionally determine if FXR is required for the maintenance of this stem cell pool during severe inflammation, protein expression levels were assessed in the FXR^−/− DSS model. Western blot analysis revealed a notable downregulation of LGR5 protein levels in the colon tissues of FXR^−/− DSS mice compared with those of WT DSS controls (Fig. 4A), indicating a failure to maintain the stem cell niche. Consistent with this depletion of the primary CBC stem cell population, notably reduced protein levels of core stemness-associated transcription factors were also observed, including SALL4, OCT3/4 and NANOG in FXR^−/− DSS mice (Fig. 4A and B). Consistent with the protein expression data, IHC demonstrated a marked decrease in the positive staining area and intensity for SOX2 and OCT3/4 in the epithelial and crypt regions of the FXR^−/− DSS colon (Fig. 4C). Furthermore, high-magnification IHC images confirmed the spatial depletion of the LGR5⁺ population. Specifically, LGR5⁺ signals, which are tightly related to the CBC cells, were diminished in both intensity and cell number compared with those in WT DSS controls (Fig. 4C). This localized loss of the LGR5⁺ pool consistently aligns with the western blot findings, further validating the collapse of the functional stem cell niche upon FXR deletion. Together, these data demonstrated that FXR deficiency leads to the collapse of the ISC pool and its associated regenerative protein network, thereby impairing mucosal repair.

To explore the broader regulatory network underlying this phenotype, WGCNA was performed on the GSE87466 transcriptome. While multiple modules displayed statistical significance, the MEbisque4 module was specifically selected for downstream analysis because it exhibited the strongest positive correlation with FXR expression (highest correlation coefficient and lowest P-value) (Fig. 4D). This strong positive co-expression indicates that genes within the MEbisque4 module are most tightly co-regulated with FXR, making it the optimal candidate for representing the homeostatic biological functions lost during FXR downregulation in colitis. By applying strict filtering criteria (a Pearson correlation >0.4 with FXR and >0.8 with the MEbisque4 module), 140 core genes were identified (Fig. 4E). Subsequent GO and KEGG enrichment analysis revealed that these 140 genes are enriched in lipid and glucose metabolism pathways (Fig. 4F and G). This transcriptomic profile suggested that FXR may support epithelial stemness and mucosal repair, at least in part, by modulating local cellular metabolism.

WGCNA of the transcriptome dataset GSE87466 revealed a strong negative correlation between FXR and the MEdarkmagenta module. This suggested that diminished FXR expression may activate specific genes, triggering downstream biological effects. Applying filtering criteria, an absolute Pearson correlation of >0.4 with FXR and >0.8 with the MEdarkmagenta module, 165 key genes were identified (Fig. 5A). Subsequent GO and KEGG enrichment analysis were performed on these 165 genes. GO enrichment analysis revealed that the DEGs were significantly enriched in critical biological processes, including 'leukocyte cell-cell adhesion', 'leukocyte-mediated immunity', 'regulation of immune effector process' and 'regulation of leukocyte cell-cell adhesion' (Fig. 5B). KEGG pathway analysis further demonstrated significant enrichment of the DEGs in pivotal signaling pathways: 'TNF signaling pathway', 'B cell receptor signaling pathway' and 'NF-κB signaling pathway' (Fig. 5C). Collectively, these results underscored that FXR is intimately linked to inflammatory processes.

To investigate the relationship between FXR and inflammatory signaling, the GSE87466 cohort was divided into FXR high- and low-expression groups. Differential expression analysis uncovered distinct transcriptional profiles, as illustrated in the resulting volcano plot (Fig. 6A). Subsequently, GSEA was performed, demonstrating marked activation of the NF-κB signaling pathway in the FXR low-expression group (Fig. 6B). Consistently, FXR expression showed negative Pearson correlations with key NF-κB-related genes, including TNF, IL-1β, RELA and TUBB (34,35) (Fig. 6C-F).

To determine the cellular context of this association, sub-clustering analysis was performed on epithelial-related stem cell populations from the single-cell dataset GSE116222 (Fig. 6G and H). Stratifying these cells into FXR⁺ and FXR⁻ groups and scoring NF-κB pathway activity using AddModuleScore revealed specific pathway activation in FXR⁻ cells (Fig. 6I and J). This activation was progressively elevated across healthy tissues (lowest activation), through non-inflamed tissues (intermediate activation) to inflamed tissues (highest activation) (Fig. 6K), with concomitant upregulation of NF-κB-related inflammatory genes in the FXR⁻ population (Fig. 6L).

Finally, these findings were validated in vivo using the FXR^−/− DSS colitis model. Western blot analysis showed notably increased protein levels of TNF-α, IL-1β and NF-κB p65 in FXR^−/− colon tissues compared with those in WT controls (Fig. 6M and N), indicating that FXR loss is associated with the hyperactivation of NF-κB signaling and promotes inflammatory responses at the functional level.