IL1B-high macrophages promote ferroptosis-like lipid peroxidation injury in Crohn's disease intestinal epithelium

Abstract

Crohn's disease (CD)-associated mucosal injury involves immune-cell infiltration, epithelial cell death, and junctional disruption, but how specific myeloid states contribute to epithelial lipid peroxidation remains unclear. We integrated longitudinal single-cell transcriptomics, spatial transcriptomics, macrophage–epithelial co-culture, epithelial RNA-seq, exploratory candidate-gene prioritization, GPX4 functional validation, and a TNBS-induced chronic colitis model to investigate IL1B-high macrophage-associated epithelial injury. In paired pretreatment and post-treatment biopsies, persistent active disease (POST-ACT) showed a higher relative representation of myeloid cells, increased epithelial programmed-cell-death-related transcriptional activity, and stronger predicted myeloid–epithelial communication than remission samples. Myeloid subclustering identified a pro-inflammatory IL1B-high macrophage state relatively enriched in pooled POST-ACT cells, while RCTD-based spatial mapping supported non-random local co-occurrence of inferred IL1B-high macrophage and injury-associated epithelial signals. In vitro, IL1B-high macrophage co-culture increased epithelial cell death, reduced GSH, and increased MDA and lipid ROS in FHC and NCM460 cells; these changes were partially attenuated by Ferrostatin-1. Integration of ferroptosis-related genes, differential expression, LASSO, and random forest prioritized GPX4, PGD, and LPCAT3. GPX4 was consistently downregulated and was selected for functional validation; GPX4 overexpression partially restored ZO-1, E-cadherin, and Occludin expression. Blocking IL-1β/IL1R1 signaling with an anti-IL-1β neutralizing antibody or the IL1R1 antagonist AF12198 attenuated lipid peroxidation, GPX4 downregulation, and loss of junction-associated molecules in the shared co-culture system. In TNBS colitis, Ferrostatin-1 or anti-IL-1β treatment alleviated selected mucosal injury phenotypes, and adding IL-1β blockade to anti-TNF produced incremental improvement over anti-TNF alone. Collectively, these findings support an IL1B-high macrophage-associated epithelial injury axis involving IL-1β–IL1R1 signaling, reduced GPX4-related lipid-peroxide defense, ferroptosis-like lipid peroxidation injury, and junction-associated molecular alterations in CD.

1. Introduction

Crohn's disease is one of the major subtypes of inflammatory bowel disease, and its pathological process is typically characterized by chronic, recurrent, and segmental intestinal inflammation, often accompanied by mucosal ulceration, immune-cell infiltration, epithelial barrier disruption, and tissue remodeling [1,2]. The intestinal epithelial barrier plays a key role in maintaining the spatial separation between luminal microorganisms, dietary antigens, and the mucosal immune system [3,4]. Tight junction- and adherens junction-related molecules maintain epithelial structural integrity, whereas increased epithelial cell death or loss of barrier molecules can promote the entry of luminal antigens into the lamina propria and further amplify local immune responses [5,6]. Therefore, clarifying the cellular sources and molecular mechanisms that drive epithelial death and barrier disruption within the inflammatory microenvironment is important for understanding the progression of Crohn's disease-associated mucosal injury.

Anti-TNF agents are an important therapeutic strategy for moderate-to-severe Crohn's disease, as they reduce inflammatory activity and promote mucosal healing in some patients. However, primary non-response, secondary loss of response, and persistent inflammatory activity after treatment remain clinical challenges [7]. Previous studies suggest that insufficient treatment response is not necessarily explained solely by the TNF pathway itself; local tissues may harbor TNF-independent inflammatory networks or networks that remain active after TNF blockade [8]. For example, IL-1-related tissue inflammatory programs have been associated with certain treatment-nonresponsive phenotypes [9]. These findings indicate that identifying pathways related to residual mucosal injury in the context of anti-TNF therapy may help in understanding persistent disease activity and developing combination intervention strategies.

Macrophages are an important component of the intestinal mucosal innate immune system. Under homeostatic conditions, they participate in phagocytic clearance, immune tolerance, and tissue repair; under inflammatory conditions, macrophages can acquire a continuum of context-dependent pro-inflammatory transcriptional states and produce cytokines such as IL-1β, TNF, and IL-6 [10]. IL-1β is an important mediator in the inflammatory cascade and can activate downstream inflammatory signaling through IL1R1 and its accessory receptor IL1RAP, thereby affecting epithelial function, myeloid cell recruitment, and tissue injury [11]. In recent years, single-cell transcriptomic studies have further shown that multiple myeloid cell states exist in the mucosa of inflammatory bowel disease, and some subsets are associated with inflammatory amplification, tissue destruction, and treatment response [7,12]. However, whether a transcript-defined IL1B-high macrophage state is associated with Crohn's disease-related epithelial lipid peroxidation and loss of junction-associated molecules remains to be further validated.

Ferroptosis is a regulated form of cell death characterized by iron-dependent lipid peroxidation accumulation. Unlike apoptosis, necroptosis, and pyroptosis, the key events of ferroptosis include polyunsaturated lipid peroxidation, impairment of antioxidant systems, and oxidative damage to the cell membrane [13]. GPX4 is a core protective molecule that limits the propagation of membrane lipid peroxidation and uses glutathione to remove lipid peroxides, thereby maintaining membrane redox homeostasis [14]. Increasing evidence suggests that ferroptosis or ferroptosis-related lipid peroxidation injury in intestinal epithelial cells may participate in epithelial barrier disruption and mucosal inflammation in inflammatory bowel disease [15]. However, in the inflammatory microenvironment of Crohn's disease, whether GPX4 downregulation is driven by a specific immune-cell subset and whether this process contributes to the loss of barrier molecules remain insufficiently supported by systematic evidence.

Recent studies have also suggested that IL-1β-producing macrophages can affect susceptibility to intestinal inflammation by regulating intestinal epithelial lipid metabolism, indicating that macrophage-derived IL-1β is not only an inflammatory cytokine but may also participate in mucosal injury by influencing epithelial metabolic states [16]. In addition, host immunometabolic status and microbial metabolic pathways may affect responses to anti-TNF therapy, providing new directions for exploring mechanisms of residual inflammation and mucosal injury in the setting of anti-TNF treatment [17]. Based on these advances, we hypothesized that, within the inflammatory microenvironment of Crohn's disease, IL1B-high macrophages may contribute to epithelial lipid peroxidation injury through IL-1β–IL1R1 signaling, reduced GPX4-dependent lipid-peroxide defense, and accompanying loss of junction-associated molecules.

This study was designed around the following logical framework: myeloid cell remodeling in an aggravated inflammatory state, enrichment of IL1B-high macrophages, enhanced IL-1β–IL1R1 communication, GPX4 downregulation, ferroptosis-like injury of intestinal epithelial cells, and loss of junction-associated molecules. By integrating single-cell transcriptomics, spatial transcriptomics, in vitro co-culture experiments, exploratory candidate-gene prioritization, GPX4 functional validation, and a TNBS-induced mouse model of chronic colitis, this study aimed to examine the association of IL1B-high macrophages with Crohn's disease-associated epithelial injury and junction-associatedmolecular alterations and to determine whether blockade of IL-1β signaling or Fer-1 treatment can attenuate the related pathological changes.

2. Results

2.1. Single-cell analysis suggests a higher relative representation of myeloid cells and increased epithelial PCD-related transcriptional activity in POST-ACT samples

To characterize the cellular composition and injury-related features of the intestinal mucosa in different Crohn's disease states, we analyzed 30 biopsies from 15 patients, each of whom contributed one pretreatment biopsy(PRE; n = 15) and one longitudinally paired post-treatment biopsy classified as remission (POST-REM; n = 9) or persistent active disease (POST-ACT; n = 6). After quality control and integration, a single-cell atlas was generated (Fig. S1A–H). Comparison of UMAP structures before and after Harmony integration showed reduced residual sample-associated segregation while retaining biological group-associated heterogeneity (Fig. S2A–D). Representative marker genes showed clear cell type-specific expression patterns, supporting the reliability of the annotation results (Fig. 1D; Fig. S2H and I). Based on these annotations, major cell types including T cells, plasma cells, B cells, myeloid cells, epithelial cells, stromal cells, fibroblasts, smooth muscle cells, enteric neurons, and mast cells were identified in PRE, POST-REM, and POST-ACT samples (Fig. 1A–C, Fig. S2E and F). Further comparison of cellular composition across disease states revealed differences among PRE, POST-REM, and POST-ACT samples. Notably, myeloid cells showed a higher relative representation within pooled POST-ACT cells (Fig. 1F and G). This pattern suggests that altered myeloid composition may be a cellular feature of persistently active mucosal inflammation. To further assess injury-related states across cell types during inflammatory progression, we calculated a composite PCD-related transcriptional score from the union of genes in 19 programmed cell death-related categories (Fig. 1E). Different cell types showed distinct distributions of PCD-related transcriptional activity (Figs. S2G, J, K). In POST-ACT samples, PCD-related transcriptional activity was relatively elevated in myeloid and epithelial cells (Fig. S2J and K). Further between-group comparisons showed differences in PCD scores across multiple cell types between POST-ACT and POST-REM samples, with more pronounced changes in myeloid and epithelial cells (Fig. 1H and I). Because increased PCD-related activity in epithelial cells may be associated with interactions with immune cells in the inflammatory microenvironment, we further analyzed the cell–cell communication features of epithelial cells with different PCD activity levels. In the POST-ACT state, CellChat predicted a greater number and aggregate strength of interactions between high-PCD-activity epithelial cells and myeloid cells (Fig. 1J–L). By contrast, communication strength was relatively lower in low-PCD-activity epithelial cells or epithelial cells in the POST-REM state (Fig. S3A–I). Together, these cell-level findings associate POST-ACT samples with a higher relative representation of myeloid cells, increased epithelial PCD-related transcriptional activity, and stronger predicted myeloid-epithelial communication.

Fig. 1.

Single-cell analysis suggests a higher relative representation of myeloid cells and increased epithelial PCD-related transcriptional activity in POST-ACT samples. (A-C) UMAP distributions of POST-ACT, POST-REM, and PRE samples showing 10 major cell types. (D) Expression characteristics of representative marker genes in each cell type. (E) Distribution of 19 programmed cell death-related gene sets. (F, G) Changes in the proportions of major cell types across disease states. (H, I) Comparison of programmed cell death (PCD) activity scores across cell types, showing higher relative PCD-related transcriptional activity in myeloid cells and epithelial cells in the POST-ACT group. (J-L) Cell–cell communication analysis showing greater predicted numbers and aggregate strength of interactions between high-PCD-activity epithelial cells and myeloid cells. PRE, pretreatment biopsies before anti-TNF therapy (n = 15); POST-REM, longitudinally paired post-treatment remission biopsies (n = 9); POST-ACT, longitudinally paired post-treatment biopsies with persistent active disease (n = 6). Data are presented as mean ± SEM. ns, not significant; *P < 0.05, **P < 0.01, ***P < 0.001.

2.2. Single-cell and tissue analyses identify an IL1B-high inflammatory macrophage state enriched in POST-ACT samples

Given the increased proportion of myeloid cells in the POST-ACT state and the enhanced communication between myeloid cells and high-PCD-activity epithelial cells, we further performed subcluster analysis of myeloid cells. UMAP analysis showed that myeloid cells could be further divided into monocyte, macrophage, cDC1, cDC2, cDC3, and pDC subsets (Fig. 2A; Fig. S4A). Differential expression analysis of myeloid cells between the POST-ACT and POST-REM groups revealed a set of inflammation-related genes upregulated in the POST-ACT group (Fig. 2B). Protein-protein interaction network analysis showed that IL1B was located at the center of this pro-inflammatory molecular network, suggesting that IL-1β may be closely associated with altered myeloid cell function under aggravated inflammatory conditions (Fig. 2C). Further analysis of the distribution of IL1B transcript expression across myeloid cell subsets showed that IL1B was mainly enriched in a specific macrophage subset, Mac_3 (Fig. 2D). This subset also highly expressed inflammation-related genes including SLAMF1, HILPDA, SPP1, PTGS2, IL6, and TNFAIP6, indicating a marked pro-inflammatory transcriptional profile (Fig. S4B). We therefore designated Mac_3 as an IL1B-high inflammatory macrophage state. Analysis of subset proportions across disease states showed that Mac_3 was relatively enriched in the POST-ACT group, suggesting that IL1B-high macrophages may represent an important component of myeloid cell remodeling under aggravated inflammatory conditions in Crohn's disease (Fig. 2E; Fig. S4C). To further evaluate the functional features of IL1B-high macrophages, we performed enrichment analysis of their associated genes. GO analysis showed that genes associated with IL1B-high macrophages were mainly enriched in terms related to the cell surface, extracellular matrix, secretory granules, cytokine activity, receptor ligand activity, and chemokine activity (Fig. 2F; Fig. S4D and E). KEGG analysis showed that these genes were mainly involved in inflammation- and immunity-related pathways, including cytokine-cytokine receptor interaction, NOD-like receptor signaling pathway, Toll-like receptor signaling pathway, and HIF-1 signaling pathway (Fig. S4F). These results further support the pro-inflammatory and cytokine communication-related functional features of IL1B-high macrophages. We then combined spatial transcriptomics and tissue-based experiments to examine, respectively, the distribution of the transcript-defined IL1B-high macrophage state and directly detected IL-1β+ macrophages in inflamed tissues. In one 10x Visium section, RCTD-based mapping showed a positive correlation between inferred IL1B-high macrophage abundance and neighboring high-injury epithelial abundance(Spearman ρ = 0.443; empirical P = 0.006), supporting non-random local co-occurrence at spot-level resolution (Fig. 2G; Fig. S4G and H). In an independent validation cohort of 12 infliximab-treated patients (POST-REM, n = 6; POST-ACT, n = 6), immunofluorescence showed more CD68+IL-1β+ cells in POST-ACT tissues (Fig. 2H). Flow cytometry likewise showed a higher proportion of IL-1β+ cells within the macrophage-enriched myeloid population (Fig. 2I and J). ELISA and qRT-PCR showed higher IL-1β protein and IL1B transcript levels, respectively (Fig. 2K and L). Together, these findings identify an IL1B-high inflammatory macrophage state that was relatively enriched in pooled POST-ACT cells and locally associated with high-injury epithelial signals.

Fig. 2.

Single-cell transcriptomics identified an IL1B-high inflammatory macrophage state and revealed its pro-inflammatory features. (A) UMAP distribution of myeloid cell subsets in POST-ACT, POST-REM, and PRE samples. (B) Volcano plot of differentially expressed genes in myeloid cells between POST-ACT and POST-REM groups. (C) Protein-protein interaction network of differentially expressed genes showing IL1B at the center of the pro-inflammatory molecular network. (D) IL1B transcript-expression density plot showing concentration of IL1B transcripts in Mac_3. (E) Proportional distribution of myeloid cell subsets across PRE, POST-REM, and POST-ACT groups. (F) GO enrichment analysis of IL1B-high macrophage-associated genes. (G) RCTD-based mapping in one 10x Visium section showing non-random spot-level local co-occurrence of inferred IL1B-high macrophage and high-injury epithelial signals. (H) Immunofluorescence staining of CD68 and IL-1β (n = 3). (I, J) Flow cytometry of CD45+CD14+HLA-DR + CD64+ macrophage-enriched myeloid cells and intracellular IL-1β (n = 3). (K, L) ELISA of IL-1β protein and qRT-PCR of IL1B transcripts (n = 6). Data are presented as mean ± SEM. ns, not significant; *P < 0.05, **P < 0.01, ***P < 0.001.

2.3. Co-culture with IL1B-high macrophages is associated with ferroptosis-like lipid peroxidation injury in intestinal epithelial cells

To further explore the cell death mode potentially involved in IL1B-high macrophage-associated intestinal epithelial injury, we first scored multiple programmed cell death-related gene sets based on the single-cell transcriptomic data. The results showed that the ferroptosis-related score was relatively high, suggesting that ferroptosis-related transcriptional features may participate in inflammation-associated intestinal epithelial injury (Fig. 3A). We next compared inflammatory IL1B-high macrophages with matched IL1B-knockdown macrophages, designated the IL1B-low state. Two IL1B-targeting shRNAs reduced IL1B mRNA by approximately 50% and 70%, respectively; the more efficient construct was selected for subsequent experiments. Reduced IL-1β protein and comparable macrophage viability through 72 h were confirmed (Fig. S5A–D). On this basis, IL1B-low macrophages or IL1B-high macrophages were co-cultured with FHC and NCM460 intestinal epithelial cells, and epithelial cell death was detected by PI flow cytometry. Compared with the Ctrl group and the IL1B-low macrophage co-culture group, co-culture with IL1B-high macrophages increased the proportion of PI-positive FHC and NCM460 cells, indicating greater epithelial cell death under IL1B-high macrophage co-culture (Fig. 3B). To further characterize the molecular features of IL1B-high macrophage-induced epithelial injury, we performed RNA-seq analysis of FHC and NCM460 cells co-cultured with IL1B-low macrophages or IL1B-high macrophages. Heatmaps and volcano plots of differentially expressed genes showed marked transcriptomic changes in both intestinal epithelial cell lines after co-culture with IL1B-high macrophages (Fig. 3C). GO enrichment analysis showed that differentially expressed genes were mainly involved in biological processes such as redox processes, mitochondrial function, ribosome biogenesis, RNA metabolism, and substance transport (Fig. 3D; Fig. S5G). KEGG enrichment analysis showed that these genes were involved in cell cycle, HIF-1 signaling pathway, p53 signaling pathway, glycolysis/gluconeogenesis, PI3K-Akt signaling pathway, focal adhesion, oxidative phosphorylation, autophagy, peroxisome, lysosome biogenesis, and ferroptosis-related pathways (Fig. 3E; Fig. S5H). These results suggest that IL1B-high macrophage-induced epithelial injury is accompanied by metabolic stress, redox imbalance, and changes in ferroptosis-related pathways. We next compared pharmacologic probes of apoptosis, necroptosis, lysosomal/autophagic flux, and lipid peroxidation. At the tested concentrations, inhibitor-only controls had little effect on epithelial viability, although Bafilomycin A1 caused a modest decline in NCM460 cells (Fig. S5E and F). Under IL1B-high macrophage co-culture conditions, Fer-1 showed the largest rescue pattern in FHC and NCM460 viability among the tested interventions (Fig. 3F and G). LDH release assays further showed that co-culture with IL1B-high macrophages increased membrane injury in FHC and NCM460 cells, whereas Fer-1 treatment reduced LDH release (Fig. 3H and I). These results suggest that ferroptosis-related processes may participate in IL1B-high macrophage-induced epithelial cell injury. Further detection of ferroptosis-like injury-related indicators showed that co-culture with IL1B-high macrophages reduced intracellular GSH levels in FHC and NCM460 cells, whereas Fer-1 treatment partially restored GSH content (Fig. 3J and K). Meanwhile, IL1B-high macrophage co-culture increased intracellular MDA levels in FHC and NCM460 cells, and Fer-1 treatment reduced MDA accumulation (Fig. 3L and M). C11-BODIPY staining combined with confocal microscopy showed that IL1B-high macrophage co-culture enhanced membrane lipid ROS signals, whereas Fer-1 treatment decreased lipid peroxidation (Fig. 3N). Representative transmission electron microscopy images showed relatively small, electron-dense mitochondria with reduced or disorganized cristae following IL1B-high macrophage co-culture, whereas Fer-1 treatment partially preserved mitochondrial morphology and cristae architecture in FHC and NCM460 cells (Fig. S5I). Together, the partial Fer-1-sensitive changes in viability, membrane injury, GSH, MDA, and lipid ROS support a ferroptosis-like lipid peroxidation injury phenotype during IL1B-high macrophage co-culture.

Fig. 3.

Transcriptomic and functional experiments support an association between IL1B-high macrophages and ferroptosis-like lipid peroxidation injury in intestinal epithelial cells. (A) Scoring analysis of multiple programmed cell death modes based on single-cell transcriptomic data. (B) IL1B-low macrophages and IL1B-high macrophages were co-cultured with FHC and NCM460 intestinal epithelial cells, and cell death was detected by PI flow cytometry (n = 3). (C) RNA-seq analysis of FHC and NCM460 cells co-cultured with IL1B-low macrophages or IL1B-high macrophages (n = 3). (D, E) GO and KEGG enrichment analyses of differentially expressed genes. (F, G) CCK-8 assays assessing the effects of different cell death inhibitors on the viability of FHC and NCM460 cells at 0, 24, 48, and 72 h (n = 3). (H, I) LDH assays over the same time course showing that IL1B-high macrophages enhanced membrane injury in FHC and NCM460 cells, whereas Fer-1 treatment reduced LDH release (n = 3). (J, K) GSH assays showing that IL1B-high macrophages reduced intracellular GSH levels in FHC and NCM460 cells, whereas Fer-1 treatment partially restored GSH content (n = 6). (L, M) MDA assays showing that IL1B-high macrophages increased intracellular MDA levels in FHC and NCM460 cells, whereas Fer-1 treatment markedly reduced MDA accumulation (n = 6). (N) C11-BODIPY probe combined with confocal microscopy was used to specifically detect membrane lipid ROS levels (n = 3). Data are presented as mean ± SEM. In panels F-I, Normal denotes epithelial cells cultured without macrophages; DMSO denotes IL1B-high macrophage co-culture with vehicle; and the Z-VAD, Necrostatin-1, Bafilomycin A1, and Fer-1 groups denote co-culture with the indicated compound. Colored significance symbols beside each curve indicate comparison of that group with Fer-1. ns, not significant; *P < 0.05, **P < 0.01, ***P < 0.001.

2.4. Candidate prioritization and functional validation support GPX4 as a component of epithelial lipid-peroxide defense in IL1B-high macrophage-associated injury

To further screen for key molecules potentially involved in IL1B-high macrophage-induced ferroptosis-like injury in intestinal epithelial cells, we first intersected ferroptosis-related genes with differentially expressed genes in FHC and NCM460 cells co-cultured with IL1B-high macrophages, obtaining a set of candidate ferroptosis-related differentially expressed genes (Fig. 4A). LASSO regression and random forest analyses were then used for exploratory candidate-gene prioritization (Fig. 4B and C). Intersection analysis of the two algorithms finally identified three shared candidate genes: GPX4, PGD, and LPCAT3 (Fig. 4D). At the transcriptomic level, RNA-seq results showed that GPX4, PGD, and LPCAT3 all tended to be downregulated in FHC and NCM460 cells after co-culture with IL1B-high macrophages (Fig. 4E). qRT-PCR further validated the expression changes of these three candidate genes in both intestinal epithelial cell lines, among which GPX4 showed a relatively stable downward trend in both FHC and NCM460 cells (Fig. 4F). Considering its established role in lipid-peroxide detoxification and its consistent reduction in both epithelial cell lines, we selected GPX4 for subsequent validation. In human tissue samples, qRT-PCR showed that GPX4 mRNA expression was lower in POST-ACT samples than in POST-REM samples (Fig. 4G). Immunofluorescence staining showed weakened GPX4 signals in POST-ACT tissues (Fig. 4H and I). Western blotting also showed, at the protein level, that GPX4 expression was lower in POST-ACT samples than in POST-REM samples (Fig. 4J). In the in vitro co-culture system, Western blotting showed that GPX4 protein expression was reduced in FHC and NCM460 cells after co-culture with IL1B-high macrophages (Fig. 4K). These results suggest that GPX4 downregulation may be associated with IL1B-high macrophage-related epithelial ferroptosis-like injury. To further evaluate the protective role of GPX4 in IL1B-high macrophage-associated epithelial injury, we established a GPX4 overexpression system. Western blotting confirmed clear GPX4 overexpression efficiency in FHC and NCM460 cells (Fig. 4L). To determine whether the protective effect of GPX4 extended to junction-associated molecular alterations, we next examined epithelial junction-related molecules. Notably, the preceding epithelial RNA-seq enrichment analysis indicated that transcriptional changes induced by IL1B-high macrophage co-culture were not restricted to redox- and ferroptosis-related processes, but also involved adhesion-associated pathways, including focal adhesion. We therefore examined the tight junction-associated genes TJP1 (encoding ZO-1) and OCLN (Occludin), together with the adherens junction-associated gene CDH1 (E-cadherin). IL1B-high macrophage co-culture decreased the mRNA expression of these junction-associated genes, whereas GPX4 overexpression partially restored their expression levels (Fig. 4M). Western blotting further showed that GPX4 overexpression partially ameliorated the decreases in ZO-1, E-cadherin, and Occludin at the protein level (Fig. 4N). In summary, GPX4 downregulation was associated with IL1B-high macrophage-related epithelial ferroptosis-like injury and loss of junction-associated molecules, whereas GPX4 overexpression partially restored their expression, supporting a protective contribution of GPX4 to the measured molecular phenotype.

Fig. 4.

Exploratory candidate-gene prioritization and functional validation support GPX4 as a component of epithelial lipid-peroxide defense in IL1B-high macrophage-associated injury. (A) Intersection analysis between ferroptosis-related genes and differentially expressed genes in FHC and NCM460 cells after co-culture with IL1B-high macrophages. (B, C) LASSO regression and random forest analyses used for exploratory candidate-gene prioritization. (D) Intersection of the two algorithms identified three prioritized candidate genes: GPX4, PGD, and LPCAT3. (E) RNA-seq results showing that GPX4, PGD, and LPCAT3 were downregulated in FHC and NCM460 cells after co-culture with IL1B-high macrophages (n = 3). (F) qRT-PCR validation of GPX4, PGD, and LPCAT3 expression changes in FHC and NCM460 cells (n = 6). (G-J) qRT-PCR (n = 6), immunofluorescence (n = 3), and Western blot (n = 3) validation of GPX4 expression changes in POST-REM and POST-ACT samples at mRNA and protein levels. (K) Western blot analysis of GPX4 protein expression in FHC and NCM460 cells after co-culture with IL1B-high macrophages (n = 3). (L) Western blot validation of GPX4 overexpression efficiency in FHC and NCM460 cells (n = 3). (M) qRT-PCR analysis of the effect of GPX4 overexpression on TJP1 (ZO-1), CDH1 (E-cadherin), and OCLN (Occludin) transcript levels (n = 6) and (N) Western blot analysis of ZO-1, E-cadherin, and Occludin protein expression (n = 3). Data are presented as mean ± SEM. ns, not significant; *P < 0.05, **P < 0.01, ***P < 0.001.

2.5. IL-1β/IL1R1 blockade partially attenuates epithelial lipid peroxidation and loss of junction-associated molecules in co-culture

To further dissect the potential communication mechanism between IL1B-high macrophages and highly injured epithelial cells, we performed cell–cell communication and ligand-receptor analyses. The CellChat analysis predicted interactions between multiple immune cell subsets and high-PCD-activity epithelial cells, among which the communication link between IL1B-high macrophages and highly injured epithelial cells was prominent (Fig. 5A; Fig. S6A). Further ligand-receptor analysis predicted IL1B–IL1R1/IL1RAP signaling between IL1B-high macrophages and highly injured epithelial cells, suggesting that the IL-1β–IL1R1 axis may participate in inflammatory crosstalk between macrophages and epithelial cells (Fig. 5B). To assess the potential role of IL1R1 in the epithelial injury state, we further performed scTenifoldKnk-based in silico gene-regulatory-network perturbation. The results suggested that IL1R1 perturbation was associated with changes in multiple epithelial injury- and inflammation-related genes (Fig. 5C; Fig. S6B–D). KEGG enrichment analysis of IL1R1-associated genes showed that they were mainly involved in the PPAR signaling pathway, peroxisome, cell adhesion, arachidonic acid metabolism, ferroptosis, and lipid metabolism-related pathways (Fig. 5D). These results suggest that IL1R1 may be closely associated with epithelial lipid metabolism, oxidative stress, and injury states. To further evaluate the role of the IL-1β–IL1R1 axis, we applied the anti-IL-1β neutralizing antibody CDP-484 or the IL1R1 antagonist AF12198 in the co-culture system. CCK-8 assays showed that blocking IL-1β signaling or antagonizing IL1R1 partially alleviated the decrease in FHC and NCM460 cell viability induced by IL1B-high macrophages (Fig. 5E and F). LDH release assays further showed that treatment with the anti-IL-1β neutralizing antibody or the IL1R1 antagonist AF12198 reduced IL1B-high macrophage-induced membrane injury (Fig. 5G and H). In parallel, GSH and MDA assays showed that blockade of IL-1β–IL1R1 signaling partially restored GSH levels and reduced MDA accumulation, indicating attenuation of IL1B-high macrophage-associated oxidative lipid injury (Fig. 5I–L). At the molecular level, qRT-PCR showed that treatment with the anti-IL-1β neutralizing antibody or the IL1R1 antagonist AF12198 partially restored the transcript levels of GPX4 and the epithelial junction-associated genes TJP1 (ZO-1), CDH1 (E-cadherin), and OCLN (Occludin) (Fig. 5M and N). Western blotting further showed that GPX4, ZO-1, E-cadherin, and Occludin protein levels were partially restored after blockade of IL-1β–IL1R1 signaling (Fig. 5O). In addition, C11-BODIPY staining showed that treatment with the anti-IL-1β neutralizing antibody or the IL1R1 antagonist AF12198 reduced IL1B-high macrophage-induced accumulation of membrane lipid ROS (Fig. 5P). Representative transmission electron microscopy images further showed that anti-IL-1β antibody or IL1R1 antagonist treatment partially preserved mitochondrial morphology and cristae architecture in FHC and NCM460 cells exposed to IL1B-high macrophages (Fig. S6E). Together, these results support the involvement of IL-1β/IL1R1 signaling in the epithelial lipid-peroxidation and junction-associated molecular phenotype observed in the co-culture system.

Fig. 5.

IL-1β/IL1R1 signaling is associated with IL1B-high macrophage-related epithelial lipid peroxidation injury. (A) Cell–cell communication analysis showing broad interactions between different immune cell subsets and highly injured epithelial cells. (B) Ligand-receptor analysis showing significant IL1B–IL1R1/IL1RAP signaling between IL1B-high macrophages and highly injured epithelial cells. (C) scTenifoldKnk in silico IL1R1 perturbation analysis suggesting that IL1R1 may be an important receptor molecule affecting the epithelial injury state. (D) KEGG enrichment analysis of IL1R1-associated genes. (E-H) CCK-8 and LDH assays showing that an anti-IL-1β neutralizing antibody or AF12198 treatment attenuated IL1B-high macrophage-induced decreases in FHC and NCM460 cell viability and membrane injury (n = 3). (I-L) GSH and MDA assays showing that blockade of IL-1β–IL1R1 signaling alleviated IL1B-high macrophage-induced oxidative stress and lipid peroxidation (n = 6). (M, N) qRT-PCR showing that an anti-IL-1β neutralizing antibody or AF12198 treatment partially restored the expression of GPX4 and the epithelial junction-associated genes TJP1 (ZO-1), CDH1 (E-cadherin), and OCLN (Occludin) (n = 6). (O) Western blotting further validated that blockade of IL-1β–IL1R1 signaling partially restored GPX4, ZO-1, E-cadherin, and Occludin protein expression (n = 3). (P) C11-BODIPY probe combined with confocal microscopy was used to specifically detect membrane lipid ROS levels (n = 3). Data are presented as mean ± SEM. ns, not significant; *P < 0.05, **P < 0.01, ***P < 0.001.

2.6. Fer-1 or IL-1β blockade attenuates selected injury phenotypes in TNBS-induced chronic colitis

In the TNBS-induced chronic colitis model (Fig. 6A), we compared Ctrl, TNBS, TNBS + Fer-1, and TNBS + anti-IL-1β antibody groups. We first evaluated the effects of different treatments on disease phenotypes. Gross colon observation showed that colons were shortened in the TNBS group, whereas Fer-1 or anti-IL-1β antibody treatment alleviated colon shortening (Fig. 6B, Figure S7C). Dynamic monitoring showed more pronounced body weight loss and higher DAI scores in the TNBS group; by contrast, Fer-1 or anti-IL-1β antibody intervention improved the trend of body weight loss and reduced DAI scores (Fig. 6C and D). Statistical analysis of colon length further supported these observations, suggesting that Fer-1 or anti-IL-1β antibody treatment attenuated TNBS-induced colon injury phenotypes to some extent (Fig. 6E). We then measured inflammatory and lipid peroxidation-related indicators in colon tissues. Detectable colonic IL-1β signal was higher in the TNBS group and lower after anti-IL-1β antibody treatment, as assessed by ELISA, Western blotting, and immunohistochemistry (Fig. 6F; Figure S7A, B). Meanwhile, MDA levels were increased and GSH levels were decreased in the TNBS group, suggesting that TNBS-induced mucosal injury was accompanied by enhanced lipid peroxidation and reduced antioxidant capacity. Fer-1 or anti-IL-1β antibody treatment reduced MDA levels and partially restored GSH content (Fig. 6G and H). These results indicate that ferroptosis-like lipid peroxidation changes may participate in TNBS-induced intestinal mucosal injury and that IL-1β blockade can attenuate this process to some extent. We further examined the expression changes of GPX4 and epithelial junction-associated molecules at the molecular level. qRT-PCR showed that transcript levels of GPX4, TJP1 (ZO-1), CDH1 (E-cadherin), and OCLN (Occludin) were decreased in colon tissues from the TNBS group, whereas Fer-1 or anti-IL-1β antibody treatment restored the expression of these molecules to varying degrees (Fig. 6I). This result suggests that TNBS-induced intestinal mucosal injury was accompanied by GPX4 downregulation and loss of epithelial junction-associated molecules, whereas Fer-1 treatment or IL-1β blockade partially improved these molecular alterations. Histological analysis further showed that colon mucosal architecture was disrupted, inflammatory cell infiltration was increased, tissue injury was aggravated, and histological scores were elevated in the TNBS group; after Fer-1 or anti-IL-1β antibody treatment, colonic mucosal pathological injury was attenuated and histological scores were reduced (Fig. 6J). Immunohistochemical staining showed that GPX4, ZO-1, E-cadherin, and Occludin protein expression was decreased in colon tissues of the TNBS group, whereas these proteins showed partial recovery after Fer-1 or anti-IL-1β antibody intervention (Fig. 6K). In summary, Fer-1 and anti-IL-1β antibody attenuated selected TNBS-induced mucosal injury phenotypes and partially reversed lipid peroxidation, GPX4 downregulation, and loss of epithelial junction-associated molecules.

Fig. 6.

Fer-1 or IL-1β blockade attenuates selected injury phenotypes in TNBS-induced chronic colitis. (A) Schematic workflow of TNBS-induced chronic colitis and intervention in mice. (B) Gross morphology of the colon in each group. (C, D) Body weight changes and DAI scores in each group. (E) Quantification of colon length in each group. (F-H) Detection of IL-1β signal, MDA, and GSH levels in colon tissues. (I) qRT-PCR analysis of GPX4, TJP1 (ZO-1), CDH1 (E-cadherin), and OCLN (Occludin) transcript levels in colon tissues from each group. (J) H&E staining assessing pathological injury in colon tissues from each group. (K) Immunohistochemical staining detecting protein expression changes of GPX4, ZO-1, E-cadherin, and Occludin in colon tissues from each group. Six mice were initially assigned to each group; one TNBS mouse died, yielding endpoint n = 5 for TNBS and n = 6 for all other groups. Data are presented as mean ± SEM. For endpoint comparisons, * indicates TNBS versus TNBS + Fer-1, and # indicates TNBS versus TNBS + anti-IL-1β Ab; ns, not significant; *P < 0.05, **P < 0.01, ***P < 0.001; #P < 0.05, ##P < 0.01, ###P < 0.001.

2.7. Addition of IL-1β blockade to anti-TNF yields incremental improvement in selected TNBS-associated phenotypes

Because POST-ACT samples retained active disease after anti-TNF therapy, we next examined whether adding IL-1β blockade to anti-TNF altered selected TNBS-associated phenotypes. Compared with the TNBS + anti-TNF group, mice in the TNBS + anti-TNF + anti-IL-1β antibody group showed lower DAI scores and a more favorable body-weight recovery trajectory (Fig. 7A and B). Gross colon morphology showed less injury after anti-IL-1β antibody was added to anti-TNF (Fig. 7C, Fig. S7D). Statistical analysis of colon length further showed that colon length was greater in the TNBS + anti-TNF + anti-IL-1β antibody group than in the TNBS + anti-TNF group, consistent with incremental improvement when IL-1β blockade was added to anti-TNF (Fig. 7D). We then measured oxidative lipid injury-related indicators. Compared with anti-TNF treatment alone, the addition of anti-IL-1β antibody treatment increased GSH levels and decreased MDA levels in colon tissues, suggesting that IL-1β blockade may further reduce lipid peroxidation-related injury that persists after anti-TNF antibody treatment (Fig. 7E and F). H&E staining showed that the combination group had relatively better-preserved colonic mucosal architecture, reduced inflammatory infiltration and tissue injury, and lower histological scores than the anti-TNF treatment-alone group (Fig. 7G). Further immunohistochemical staining showed that, compared with the TNBS + anti-TNF group, protein expression of GPX4, ZO-1, E-cadherin, and Occludin was partially restored in colon tissues of the TNBS + anti-TNF + anti-IL-1β antibody group (Fig. 7H). Together, these findings indicate incremental improvement in the evaluated disease-activity, histologic, redox, and junction-associated molecular phenotypes after IL-1β blockade was added to anti-TNF. Fig. 7I summarizes the proposed relationship among the IL1B-high macrophage state, IL-1β–IL1R1 signaling, GPX4 downregulation, epithelial lipid peroxidation, junction-associated molecular alterations, and mucosal injury.

Fig. 7.

Addition of IL-1β blockade to anti-TNF provides incremental improvement in selected TNBS-associated mucosal injury phenotypes. (A, B) DAI scores and body weight changes in the TNBS + anti-TNF group and the TNBS + anti-TNF + anti-IL-1β antibody group. (C, D) Gross colon morphology and colon length quantification in each group. (E, F) Detection of GSH and MDA levels in colon tissues. (G) H&E staining assessing pathological injury in colon tissues from each group. (H) Immunohistochemical staining detecting protein expression changes of GPX4, ZO-1, E-cadherin, and Occludin in colon tissues from each group. (I) Mechanistic schematic summarizing the proposed model: IL1B-high macrophage-associated IL-1β–IL1R1 signaling may contribute to epithelial GPX4 downregulation, lipid peroxidation, ferroptosis-like injury, junction-associated molecular alterations, and mucosal injury; IL-1β blockade or Fer-1 treatment attenuated selected phenotypes in this experimental system. The comparison used n = 6 mice per group. Data are presented as mean ± SEM. ns, not significant; *P < 0.05, **P < 0.01, ***P < 0.001.

3. Discussion

This study supports an immune-epithelial injury model in which an IL1B-high macrophage state is linked to reduced GPX4-related lipid-peroxide defense, ferroptosis-like epithelial injury, and loss of junction-associated molecules through IL-1β–IL1R1 signaling. Recent single-cell studies have shown that IBD mucosal inflammation is not driven by a single cell type or a single inflammatory cytokine, but is shaped by multiple cellular states involving epithelial cells, myeloid cells, stromal cells, and lymphocytes [[18], [19], [20]]. In CD samples, we observed myeloid cell remodeling, enrichment of IL1B-high macrophages, and stronger predicted communication between these cells and highly injured epithelial cells, consistent with previous studies on CD mucosal cellular-state heterogeneity and cellular remodeling after anti-TNF therapy [7]. Combined with spot-level spatial co-occurrence and tissue-based validation, our findings suggest that IL1B-high macrophages may represent a myeloid cell state of interest in the CD mucosal injury microenvironment.

Inflammatory macrophages have recently been considered an important cellular module in IBD tissue inflammation [21]. Previous studies have shown that IBD-associated genetic regulatory regions can drive inflammatory transcriptional programs in human macrophages, and IL-1-related tissue inflammatory programs are also associated with certain treatment-nonresponsive or persistently active mucosal pathological states [9,21,22]. Therefore, the enrichment of IL1B-high macrophages observed in this study should not be interpreted as an isolated cellular phenomenon, but rather should be understood within the broader context of inflammatory myeloid cell remodeling in IBD. The additional significance of this study lies in linking IL1B-high macrophages not only to a pro-inflammatory phenotype but also to epithelial GPX4 downregulation, enhanced lipid peroxidation, and loss of junction-associated molecules, thereby proposing a more specific IL1B-high macrophage–epithelial injury pathway.

The effects of IL-1β on intestinal inflammation and epithelial barrier function have been well studied [23]. Previous studies have shown that IL-1β can affect intestinal epithelial tight junctions and barrier permeability [24,25]; recent studies also suggest that IL-1β-producing intestinal macrophages can affect intestinal epithelial lipid metabolism and mucosal inflammation [16]. Against this background, our findings extend the established effects of IL-1β by suggesting that, in CD mucosal injury, IL1B-high macrophages may be associated with altered epithelial oxidative lipid homeostasis through IL-1β–IL1R1 signaling, making epithelial cells more susceptible to GPX4 downregulation, lipid ROS accumulation, and loss of junction-associated molecules. In this context, IL1B-high macrophages may not only serve as a source of pro-inflammatory cytokines but may also act as an upstream cellular source of immunometabolic stress signals that help shape the injury outcome of intestinal epithelial cells.

GPX4-related lipid peroxidation defense may represent an important mechanistic link between IL1B-high macrophages and epithelial injury. After co-culture with IL1B-high macrophages, epithelial cells showed GPX4 downregulation, decreased GSH, increased MDA, and lipid ROS accumulation, whereas Fer-1 or GPX4 overexpression partially attenuated injury-associated molecular changes and loss of junction-associated molecules. Complementary representative transmission electron microscopy images showed mitochondrial ultrastructural alterations after IL1B-high macrophage co-culture and partial preservation of mitochondrial morphology following Fer-1 or IL-1β/IL1R1 pathway blockade. These findings are consistent with previous studies showing that GPX4 limits intestinal epithelial lipid peroxidation and maintains mucosal homeostasis [14,26], suggesting that IL1B-high macrophages may promote CD-associated ferroptosis-like epithelial injury by affecting epithelial redox defense status. Although GPX4 was selected for functional validation, the computational analyses also prioritized PGD and LPCAT3. PGD may influence pentose-phosphate-pathway-derived NADPH and redox buffering, whereas LPCAT3 may affect phospholipid remodeling and the availability of oxidizable polyunsaturated phospholipids; their parallel, cooperative, or upstream/downstream contributions remain to be tested.

Mechanistically, IL-1β–IL1R1 signaling may serve as an important bridge connecting IL1B-high macrophages to epithelial GPX4 downregulation. CellChat and scTenifoldKnk in silico IL1R1 perturbation analyses suggested that IL1R1-associated genes are related to lipid metabolism, peroxisome function, arachidonic acid metabolism, cell adhesion, and ferroptosis-related pathways; blockade of IL-1β or antagonism of IL1R1 partially restored GPX4 and junction-associated molecule expression and attenuated lipid peroxidation injury. These results suggest that IL-1β–IL1R1 signaling may not only participate in inflammatory amplification but may also influence epithelial redox and membrane lipid homeostasis. Similarly, recent work on the METTL1-m7G-SLC7A11 axis emphasized that epithelial cell fate is jointly regulated by inflammatory intensity, nutritional environment, and metabolic stress [27]. From this context-dependent perspective, our results also suggest that the effects of IL1B-high macrophages on epithelial injury may depend on the local inflammatory background, epithelial antioxidant capacity, and GPX4-related lipid peroxidation defense state. Because the IL1R1 antagonist was present in the shared co-culture system, these data do not establish epithelial-cell-autonomous IL1R1 signaling.

In vivo experiments further support the biological relevance of this mechanism in inflammatory mucosal injury. In the TNBS model, either Fer-1 or anti-IL-1β antibody attenuated selected mucosal injury phenotypes and partially reversed lipid peroxidation, GPX4 downregulation, and loss of junction-associated molecules. This is consistent with previous studies implicating Fer-1 treatment or IL-1-related inflammatory programs in IBD-like inflammation [15,26]. Nevertheless, the TNBS model can only recapitulate some CD-like enteritis features and cannot fully represent the genetic background, immune heterogeneity, or chronic relapsing course of human CD. Adding anti-IL-1β antibody to anti-TNF treatment was associated with incremental improvement in several injury phenotypes, suggesting that IL-1β-related injury signals may remain pathologically relevant in the context of TNF blockade. Further validation in patient samples, primary cells, organoid models, and multiple animal models is still needed. A formal anti-TNF × anti-IL-1β factorial interaction was not tested, so pharmacologic synergy cannot be inferred.

This study has several limitations. First, although we integrated multi-level evidence from single-cell transcriptomics, spatial transcriptomics, in vitro co-culture experiments, and animal models, biological background differences remain among these experimental systems; therefore, the relevant conclusions require further validation in additional independent cohorts and experimental models [12,28]. Second, THP-1-derived macrophages and intestinal epithelial cell lines provide a relatively controllable system for dissecting specific cellular interactions, but they cannot fully recapitulate the complex interactions among primary immune cells, epithelial cells, stromal cells, and microenvironmental components in the human intestinal mucosa. Therefore, future studies may combine patient-derived primary cells, intestinal organoids, and co-culture systems that better approximate tissue architecture to further evaluate the effects of IL1B-high macrophages on epithelial injury and recovery of junction-associated molecular expression [29,30]. In addition, the TNBS model reflects some CD-like enteritis features, but human CD is characterized by marked clinical heterogeneity, a long-term chronic relapsing course, and variable treatment responses; thus, animal intervention results should primarily be interpreted as mechanistic support rather than directly extrapolated to clinical therapeutic conclusions [31,32]. Based on these limitations, future studies should combine larger clinical sample sets, longitudinal follow-up cohorts, and patient-derived models to further define the boundaries of the IL-1β–IL1R1-GPX4 axis across disease stages, treatment contexts, and patient subgroups [[33], [34], [35], [36]]. Additional limitations include cell-level rather than patient-level single-cell comparisons and the absence of direct barrier-function assays. PI/LDH readouts and partial Fer-1 rescue do not establish ferroptosis as the exclusive death mechanism. Only female mice were studied, and potential antibody interference with post-treatment IL-1β detection was not excluded.

Overall, this study supports a contributory IL1B-high macrophage-associated immune-epithelial injury axis. This pathway links IL-1β–IL1R1 signaling to reduced GPX4-related lipid-peroxide defense, ferroptosis-like epithelial injury, and loss of junction-associated molecules. IL-1β blockade or Fer-1 treatment partially attenuated these changes in the experimental systems used. These findings provide preclinical evidence linking an inflammatory myeloid state to epithelial oxidative injury in Crohn's disease.

4. Conclusion

This study supports an immune-epithelial interaction mechanism associated with CD mucosal epithelial injury. IL1B-high macrophages were relatively enriched in CD-associated inflamed tissues and showed stronger predicted communication with highly injured intestinal epithelial cells. The combined data implicate IL-1β–IL1R1 signaling and reduced GPX4-related lipid-peroxide defense in ferroptosis-like epithelial injury accompanied by decreased junction-associated molecules, including ZO-1, E-cadherin, and Occludin. In vivo experiments further suggest that IL-1β signaling blockade or Fer-1 treatment attenuated selected injury phenotypes in TNBS-induced chronic colitis; in the context of anti-TNF treatment, additional IL-1β blockade was associated with incremental improvement in several mucosal injury phenotypes. These results link IL1B-high macrophages, GPX4 downregulation, and epithelial lipid peroxidation injury, providing preclinical evidence for a myeloid contribution to epithelial oxidative injury in CD.

5. Materials and methods

5.1. Single-cell data source, processing, and annotation

This study used the public single-cell transcriptomic dataset GSE282122 to analyze intestinal mucosal cell composition and injury-related transcriptional features across different treatment states of Crohn's disease. This dataset was derived from a longitudinal single-cell atlas study of anti-TNF therapy, in which the original study enrolled biologic-naive patients with inflammatory bowel disease and collected intestinal biopsy samples before and after adalimumab treatment [7]. In the original study, remission status in Crohn's disease was mainly determined by integrated follow-up clinical, endoscopic, and histological indicators, including the Harvey-Bradshaw Index, macroscopic ulceration, and Nancy histological score. According to the original sample annotation information, this study extracted samples annotated as Crohn's disease (CD) and compiled clinical annotations using information on patient, treatment, inflammation, age, and sex; detailed sample information is provided in Supplementary Table S1. According to treatment time points and post-treatment inflammation/remission status, samples were divided into a pretreatment group (PRE), an anti-TNF post-treatment remission group (POST-REM), and an anti-TNF post-treatment active group (POST-ACT). PRE represents intestinal mucosal samples collected before adalimumab treatment; POST-REM represents samples that achieved remission after treatment; and POST-ACT represents samples with persistent inflammatory activity or failure to achieve remission after treatment. The dataset comprised 30 biopsies from 15 patients. Each participant contributed one PRE biopsy and one corresponding post-treatment biopsy; the post-treatment biopsy was classified as POST-REM (n = 9) or POST-ACT (n = 6), yielding a longitudinally paired structure rather than 30 independent patients. Based on these original grouping annotations, we further compared the proportions of major intestinal mucosal cell types, programmed cell death-related transcriptional activity, and cell–cell communication features across treatment states. Cells originating from the same biopsy or patient were not considered independent biological replicates. Individual cells were used for clustering, dimensionality reduction, annotation, state characterization, module-score calculation, and exploratory differential-expression analyses; cell-level analyses were used for candidate identification and transcriptional characterization rather than patient-level population inference.

Single-cell data were processed in R 4.4.3, mainly using Seurat v5.5.0, SeuratObject v5.4.0, and Harmony v2.0.3. Expression matrices from each sample were first imported, and quality control was performed based on the number of detected genes, UMI counts, the proportion of mitochondrial genes, the proportion of ribosomal genes, and the proportion of hemoglobin-related genes, together with DoubletFinder to remove potential doublets. After quality control, high-quality cells were retained for downstream analysis. Quality-controlled data were normalized using the LogNormalize method, highly variable genes were identified, and data scaling and PCA were performed. Harmony was then used for cross-sample integration with individual sample identity (orig.ident) as the integration variable, followed by graph construction, unsupervised clustering, and UMAP visualization using the first 15 Harmony dimensions. Clustering resolution was determined by integrating multi-resolution comparison, clustree analysis, and marker gene expression patterns. Cell type annotation was performed based on differentially expressed genes and classical lineage marker genes for each cell cluster. Specific marker genes were identified using the FindAllMarkers function. Major cell types, including T cells, plasma cells, and B cells, were ultimately identified. The cell annotation results were validated by UMAP distribution, marker gene expression plots, and bubble plots and were used for subsequent analyses. Integration was assessed by comparing UMAP representations before and after Harmony correction with cells colored by sample and by PRE/POST-REM/POST-ACT group. For the POST-ACT versus POST-REM myeloid comparison, genes detected in at least 10% of cells in either group were tested; displayed differentially expressed genes met |average log2 fold change| ≥ 0.25 and Benjamini–Hochberg-adjusted P < 0.05. Complete output is provided in Supplementary Table S9.

5.2. Programmed cell death-related transcriptional scoring

To evaluate programmed cell death (PCD)-related gene expression features across different cell types, we compiled 19 PCD-related gene sets, including ferroptosis, apoptosis, necroptosis, pyroptosis, autophagy-related cell death, and other PCD-related categories [[37], [38], [39], [40], [41], [42], [43], [44]]. The union of genes from the 19 categories was used with the AddModuleScore function in Seurat to calculate one composite PCD-related transcriptional score for each cell, reflecting the relative enrichment of PCD-related gene sets at the single-cell transcriptomic level. We then compared the distribution of PCD scores among different cell types and across disease states to assess changes in PCD-related transcriptional activity under aggravated inflammatory conditions. For exploratory communication analysis, a single global median across all epithelial cells was used: cells above the median were assigned to the high-PCD state and cells at or below it to the low-PCD state. The same cutoff was applied to PRE, POST-REM, and POST-ACT cells, and the resulting states were used for subsequent cell–cell communication analysis to explore potential interactions between epithelial cells in distinct injury-related states and immune cells. The sources and specific gene lists of the PCD-related gene sets are provided in Supplementary Table S2. This split was an operational classification of a continuous transcriptional score rather than a biologically absolute boundary.

5.3. Cell–cell communication analysis

Cell–cell communication was inferred using CellChat version 1.6.1 and CellChatDB.human. Normalized expression and cell annotations were supplied to identifyOverExpressedGenes() and identifyOverExpressedInteractions(), followed by projectData(). Communication probabilities were estimated with computeCommunProb() using the trimean expression summary, raw.use = TRUE, population.size = FALSE, nboot = 100, and seed.use = 1. Cell groups containing fewer than five cells were excluded with filterCommunication(min.cells = 5), and predicted interactions with P < 0.05 were retained. Pooled cells within each clinical state were used to generate exploratory group-level ligand–receptor networks, with focused assessment of the IL1B–IL1R1/IL1RAP axis.

5.4. Myeloid cell subcluster analysis, differential expression, and functional enrichment

To further characterize the heterogeneity of myeloid cells under aggravated inflammatory conditions, myeloid cells were extracted from the integrated single-cell object for secondary analysis. This process included re-normalization, highly variable gene selection, data scaling, PCA, nearest-neighbor graph construction, clustering, and UMAP visualization. Based on subcluster marker gene expression features, myeloid cells were further annotated as monocytes, macrophages, cDC1, cDC2, cDC3, and pDCs. Differentially expressed genes between POST-ACT and POST-REM myeloid cells were then compared, and their distribution was displayed using volcano plots. To identify core inflammation-related molecules, differentially expressed genes were input into a protein-protein interaction network analysis. Mac_3 emerged from unsupervised clustering and showed relatively high IL1B expression together with SLAMF1, HILPDA, SPP1, PTGS2, IL6, and TNFAIP6; it was therefore termed the IL1B-high macrophage state, and this state was not considered a fixed lineage. This subset was further used for subsequent functional enrichment and cell–cell communication analyses. GO and KEGG enrichment analyses were used to evaluate the functional features of IL1B-high macrophage-associated genes and other differentially expressed genes. GO analysis included biological process, cellular component, and molecular function categories; KEGG analysis was used to identify relevant signaling pathways. The complete cell-level differential-expression output is provided in Supplementary Table S9.

5.5. Spatial transcriptomic analysis

Spatial transcriptomic analysis used 10x Genomics Visium from GSE228360. Visium capture spots have a nominal diameter of 55 μm and approximately 100 μm center-to-center spacing. RCTD, implemented in spacexr, was run in doublet mode using the scRNA-seq raw-count matrix and detailed cell-state annotations as the reference and the spatial count matrix and coordinates as the query; the complete gene list is provided in Supplementary Table S8. For each central spot, mean inferred high-injury epithelial abundance among first-order neighboring spots was calculated and correlated with central-spot inferred IL1B-high macrophage abundance using Spearman correlation. Significance was evaluated with 5000 within-section label permutations.

5.6. Clinical tissue sample collection

The independent human-tissue validation cohort included 12 patients with Crohn's disease, with one intestinal mucosal specimen per participant. All patients received infliximab induction at 5 mg/kg intravenously at weeks 0, 2, and 6, and post-induction status was assessed approximately 12 weeks after treatment initiation. Six independent patients were classified as POST-REM and six as POST-ACT. Remission was defined as meeting at least two of three criteria: Harvey–Bradshaw Index <5, absence of macroscopic ulceration, and Nancy histological index ≤1; participants not meeting this definition were classified as POST-ACT. The groups were independent rather than longitudinally paired. Demographic characteristics, disease duration and location, treatment, and classification variables are provided in Supplementary Table S6. The protocol was approved by the Ethics Committee of The Second Xiangya Hospital, Central South University (Z0919-01).

5.7. Immunofluorescence staining

Paraffin-embedded intestinal mucosal tissue sections were deparaffinized, rehydrated, and subjected to antigen retrieval, followed by blocking of nonspecific binding sites with blocking solution. Primary antibodies were then added for incubation to detect the expression of proteins including CD68, IL-1β, and GPX4. Primary antibody information was as follows: IL-1β (P50520-1R1, Abmart), CD68 (MN50019, Abmart), and GPX4 (67763-1-Ig, Proteintech). After primary antibody incubation, sections were washed with PBS and incubated with corresponding fluorescent secondary antibodies in the dark, followed by nuclear counterstaining with DAPI. Images were acquired under identical settings within an experiment. For each independent patient, one section and three non-overlapping fields were analyzed in ImageJ using one uniform threshold across groups. Target fluorescence intensity was measured in each field, the three field values were averaged to generate one value per patient, and analysis was performed with group allocation concealed. Fields were not treated as independent biological replicates.

5.8. Flow cytometric detection of IL-1β+ cells within a macrophage-enriched myeloid population

Fresh intestinal mucosal tissues were washed with PBS, obvious necrotic tissue and blood clots were removed, and single-cell suspensions were prepared by mechanical mincing combined with enzymatic digestion. The digested cell suspension was filtered through a cell strainer, cells were collected by centrifugation, and red blood cell lysis was performed when necessary according to sample condition. Cells were then washed with staining buffer and counted, and an appropriate number of cells was used for flow cytometry. To evaluate the proportions of myeloid cells and IL-1β-related cell populations in intestinal mucosal tissues, single-cell suspensions were first stained with a live-cell dye, followed by surface staining with antibodies against CD45, CD14, HLA-DR, and CD64. After surface staining, cells were fixed and permeabilized, and IL-1β antibody was added for intracellular staining. Flow cytometry data were analyzed using FlowJo software. The gating strategy was as follows: cell debris was first excluded and the main cell population was gated based on FSC-A and SSC-A; doublets were then excluded using FSC-A/FSC-H to obtain single cells; viable cells were further selected using the live-cell dye; CD45⁺ immune cells were selected among viable cells, and CD14+HLA-DR + myeloid cells were identified within the CD45⁺ population. CD14, HLA-DR, and CD64 were used in combination to enrich monocyte/macrophage-lineage cells and reduce contamination from other CD14+ myeloid populations. The CD45+CD14+HLA-DR + CD64+ population was therefore operationally defined as a macrophage-enriched myeloid population, and intracellular IL-1β positivity was determined within this population using the corresponding negative control and one consistent threshold across groups. The final population was defined as singlet CD45+CD14+HLA-DR + CD64+IL-1β+ cells.

5.9. ELISA detection of IL-1β levels

Tissue samples were homogenized and centrifuged to collect supernatants, and culture supernatants were collected from cell experiments for detection. Human IL-1β protein levels in cell culture supernatants were measured using the Human IL-1 beta ELISA Kit (KE00021, Proteintech), whereas mouse IL-1β levels in colon tissue homogenates were measured using the Mouse IL-1 beta ELISA Kit (KE10003, Proteintech), according to the manufacturer's instructions. Standards and test samples were included, and IL-1β concentrations were calculated using the standard curve. For tissue homogenate samples, IL-1β levels were normalized to total protein concentration to reduce the influence of differences in starting sample amounts.

5.10. Cell culture, macrophage treatment, GPX4 overexpression, and co-culture system

Human normal colonic epithelial cells FHC and NCM460 were obtained from the Cell Bank of the Chinese Academy of Sciences (Shanghai, China; GNHu74) and Otwobiotech (HTX1841C), respectively. The human monocytic cell line THP-1 was purchased from Shanghai Jinyuan Biotechnology (JY097). Cell identities were confirmed according to authentication information supplied by the corresponding cell banks or commercial suppliers. Mycoplasma contamination was routinely assessed using a PCR-based Mycoplasma Detection Kit (Servicebio, G1900-50T), with positive and negative controls; only mycoplasma-negative cultures were used. Cells were cultured in DMEM (Gibco) supplemented with 10% fetal bovine serum (FBS; ExCell, FSD500) and 1% penicillin-streptomycin (P/S; NCM Biotech, C100C5) in a humidified incubator at 37°C with 5% CO2. Cells with good growth status, appropriate confluence, and logarithmic growth phase were selected for subsequent experiments.

THP-1 cells were treated with 100 nM phorbol 12-myristate 13-acetate (PMA) for 48 h, washed, and allowed to recover for 24 h in PMA-free medium. They were then stimulated with lipopolysaccharide (100 ng/mL) plus interferon-γ (20 ng/mL) for 24 h. The same differentiation and stimulation schedule was used for IL1B-knockdown and non-targeting-control cells. Plasmid construction and lentiviral packaging for IL1B-targeting shRNAs were provided by GeneChem (Shanghai, China). Two human IL1B-targeting shRNAs and one non-targeting control were used; target and hairpin sequences are provided in Supplementary Table S7. Knockdown was confirmed by qRT-PCR and Western blotting, and viability was assessed by CCK-8 through 72 h. Inflammatory control cells were designated IL1B-high macrophages, whereas IL1B-knockdown cells under the same stimulation were designated IL1B-low macrophages; these labels denote relative expression states rather than absolute positive/negative populations.

To evaluate the role of GPX4 in intestinal epithelial cell injury, a GPX4 overexpression vector was constructed and transfected into FHC and NCM460 cells. The experimental groups included an empty vector control group (Vector) and a GPX4 overexpression group (GPX4-OE), and GPX4 protein expression was detected by Western blotting to verify overexpression efficiency.

Subsequently, THP-1-derived macrophages under the indicated IL1B-high or IL1B-low conditions were co-cultured with FHC or NCM460 intestinal epithelial cells in a Transwell system. After co-culture, intestinal epithelial cells and culture supernatants were collected for subsequent functional validation experiments.

5.11. Detection of cell death, cell viability, and lipid peroxidation-related indicators

PI flow cytometric staining was used to detect the proportion of dead intestinal epithelial cells after co-culture. After co-culture, FHC or NCM460 cells in the lower chamber were collected, washed with PBS, incubated with PI staining solution in the dark, and analyzed by flow cytometry to determine the proportion of PI-positive cells.

To compare pharmacologic probes of cell-death-related pathways, FHC or NCM460 cells were pretreated for 2 h before co-culture with Z-VAD-FMK (30 μM; Y062550, Beyotime), Necrostatin-1 (30 μM; HY-15760, MedChemExpress), Bafilomycin A1 (50 nM; Y262115, Beyotime), or Ferrostatin-1 (Fer-1; 1 μM; HY-100579, MedChemExpress). The corresponding compound was maintained at the same concentration throughout the subsequent 72-h co-culture; DMSO served as the vehicle control. Inhibitor-only CCK-8 assays under the same concentration and exposure conditions assessed intrinsic cytotoxicity. Bafilomycin A1 was used as an exploratory inhibitor of vacuolar H + -ATPase-dependent lysosomal acidification and autophagic flux, not as a specific inhibitor of autophagic cell death. Independent experiments, rather than technical wells, were treated as biological units. CCK-8 and LDH assays were performed according to the manufacturers' instructions.

After lysis of cell or tissue samples, total protein concentration was measured using the BCA assay. Glutathione (GSH) levels were measured using a Total Glutathione Assay Kit (S0052, Beyotime), and malondialdehyde (MDA) levels were measured using a Lipid Peroxidation MDA Assay Kit (S0131S, Beyotime). GSH and MDA results were normalized to total protein concentration. Intracellular lipid peroxidation levels were measured using a Lipid Peroxidation Assay Kit (S0043, Beyotime). After treatment, cells were incubated with BODIPY 581/591 C11 staining working solution at 37°C in the dark for 10-30 min, washed with PBS, and observed by fluorescence microscopy.

For transmission electron microscopy, the culture medium was removed after co-culture and the indicated interventions, and electron microscopy fixative was added directly to the culture dish for approximately 5 min at room temperature in the dark. Cells were gently scraped in one direction with a cell scraper, and the resulting cell suspension was transferred to a centrifuge tube using a Pasteur pipette and subjected to low-speed centrifugation for 3–5 min. After the supernatant was discarded, the cell pellet was gently pipetted and resuspended in fresh electron microscopy fixative. Samples were fixed for an additional 30 min at room temperature in the dark, stored at 4°C, and transported at 4°C with ice packs for subsequent transmission electron microscopy processing and imaging.

5.12. RNA-seq sequencing and differential expression analysis

To analyze transcriptomic changes in intestinal epithelial cells after co-culture with IL1B-high macrophages, FHC and NCM460 cells co-cultured with IL1B-low macrophages or IL1B-high macrophages were collected for RNA-seq analysis, with three biological replicates per group. Total RNA was extracted using TRIzol reagent (Invitrogen). After quality assessment, RNA-seq libraries were constructed using the Hieff NGS® EvoMax RNA Library Prep Kit (dUTP, 12340ES97, Yeasen Biotechnology), and 150-bp paired-end sequencing was performed on the DNBSEQ-T7 platform. Raw sequencing data were subjected to adapter removal and low-quality read filtering, aligned to the human reference genome GRCh38 (Ensembl 104) using STAR, and gene counts were generated using HTSeq. All raw sequencing data have been deposited in the GEO database under accession number GSE338571. Differential expression analysis was performed using the limma package, with screening criteria of |log2FC| ≥ 0.5 and P value < 0.05. Heatmaps and volcano plots were used to visualize differentially expressed gene patterns. Differentially expressed genes were further used for GO and KEGG enrichment analyses. Complete gene-level outputs for both cell lines are provided in Supplementary Table S10.

5.13. Screening of ferroptosis-related candidate genes

Ferroptosis-related gene sets were intersected with differentially expressed genes in FHC and NCM460 cells after co-culture with IL1B-high macrophages. Each independent RNA-seq biological replicate was treated as one observation. The binary outcome represented co-culture with IL1B-low versus IL1B-high macrophages, whereas expression values of the ferroptosis-related differentially expressed genes were used as input features. LASSO regression with cross-validation and random forest variable-importance analysis were applied for exploratory candidate-gene prioritization rather than clinical prediction. Their shared results prioritized GPX4, PGD, and LPCAT3. Ferroptosis-related gene sets are listed in Supplementary Table S2.

5.14. qRT-PCR detection

Total RNA was extracted from cell and tissue samples using TRIzol reagent (NCM Biotech, China) according to the manufacturer's instructions. RNA concentration and purity were measured using a NanoDrop spectrophotometer (Thermo Fisher Scientific, USA). RNA was then reverse-transcribed into cDNA using the Hifair III 1st Strand cDNA Synthesis SuperMix for qPCR Kit (Yeasen, China), and real-time quantitative PCR was performed using the Hieff qPCR SYBR Green Master Mix Kit (Yeasen, China). The detected genes included IL1B, GPX4, PGD, LPCAT3, TJP1 (ZO-1), CDH1 (E-cadherin), and OCLN (Occludin). Relative expression levels of target genes were calculated using the 2^-ΔΔCt method and normalized to an internal reference gene. Primer sequences are listed in Supplementary Table S3.

5.15. Western blotting

Total protein was extracted from cells or tissues and quantified using the BCA assay. Equal amounts of protein were separated by SDS-PAGE and transferred to PVDF membranes. A prestained protein ladder was used as a molecular weight reference during electrophoresis, including Servicebio (G2087-250UL) or Thermo Fisher Scientific (26619). After transfer, PVDF membranes were blocked and incubated with the corresponding primary antibodies. The primary antibodies used included ZO-1 (Abmart, TA5145), IL-1β (Abmart, P50520-1R1), GPX4 (Proteintech, 67763-1-Ig), Occludin (Servicebio, GB15149), E-cadherin (Servicebio, GB11082-100), and GAPDH (Servicebio, GB15002-100). After primary antibody incubation, membranes were washed and incubated with HRP-conjugated secondary antibodies, and signals were detected by chemiluminescence. GAPDH was used as the internal control. Band intensities were quantified using ImageJ software and normalized to the internal control protein.

5.16. Immunohistochemical staining

Colon tissues were fixed, paraffin-embedded, and sectioned. Sections were deparaffinized, rehydrated, subjected to antigen retrieval, endogenous peroxidase blocking, and blocking, and then incubated with the corresponding primary antibodies. The primary antibodies used included GPX4 (Proteintech, 67763-1-Ig), ZO-1 (Abmart, TA5145), E-cadherin (Servicebio, GB11082-100), and Occludin (Servicebio, GB15149). Sections were then incubated with the corresponding secondary antibodies, visualized with DAB, and counterstained with hematoxylin. IHC staining was quantified in ImageJ as the DAB-positive area percentage (PA, %) from one section and three non-overlapping fields per biological sample using a uniform threshold across groups. The three field values were averaged to one value per patient or mouse, and analysis was performed blinded to group allocation.

5.17. IL-1β/IL1R1 blockade experiment

For pharmacological blockade of IL-1β/IL1R1 signaling, the anti-IL-1β neutralizing antibody CDP-484 (5 μg/mL; HY-P990563, MedChemExpress) or the IL1R1 antagonist AF12198 (hereafter referred to as IL-1Ra, 10 nM; HY-P1110, MedChemExpress) was used in the macrophage–epithelial Transwell system. CDP-484 was added directly to the shared co-culture medium at co-culture initiation and maintained throughout the subsequent 72 h; a concentration-matched human IgG1κ isotype antibody (5 μg/mL) served as the antibody control. For IL1R1 blockade, epithelial cells were pretreated with AF12198 for 2 h, after which AF12198 was maintained at 10 nM in the shared medium throughout the 72-h co-culture; an equivalent volume of DMSO served as vehicle control.

5.18. ScTenifoldKnk in silico IL1R1 perturbation analysis

The potential transcriptional consequences of IL1R1 perturbation were explored using scTenifoldKnk, an in silico gene-regulatory-network perturbation framework based on scRNA-seq data. After filtering lowly expressed genes, a single-cell gene-regulatory network was reconstructed from the epithelial expression matrix. IL1R1 was then computationally perturbed by removing its outgoing regulatory connections to generate a pseudo-perturbed network. The original and perturbed networks were compared to identify differentially regulated genes potentially affected by IL1R1 perturbation, followed by KEGG enrichment analysis. This procedure provided computational support for IL1R1-associated network effects.

5.19. TNBS-induced mouse model of chronic colitis

All experimental mice were housed under specific pathogen-free (SPF) barrier conditions. All animal procedures were reviewed and approved by the Institutional Animal Care and Use Committee (IACUC), The Second Xiangya Hospital, Central South University, China (approval number: 20261133). Eight-week-old female SPF C57BL/6 mice were used to maintain consistency with a chronic TNBS protocol previously optimized in our laboratory. Six mice were initially included per group and were allocated before model induction using a computer-generated randomization sequence. Before model induction, the hair on the backs of the mice was shaved, and 150 μL of 5% TNBS in acetone was evenly applied to the dorsal skin for presensitization. Subsequently, mice were anesthetized with isoflurane and placed in the prone position. A flexible catheter was slowly inserted approximately 3–4 cm into the rectum through the anus, followed by TNBS enema. After enema, mice were kept inverted for approximately 30 s to reduce leakage. This study used a six-cycle TNBS enema protocol with gradually increasing doses. TNBS was dissolved in 45% ethanol and administered as 1.0% TNBS on days 0 and 7, 1.5% TNBS on days 14 and 21, and 2.0% TNBS on days 28 and 35 [[45], [46], [47]]. According to the experimental objectives, the TNBS model group and corresponding drug intervention groups were established, including the Ferrostatin-1 (Fer-1) group, anti-IL-1β antibody group, anti-TNF antibody group, and anti-TNF plus anti-IL-1β antibody group. Fer-1 was purchased from MedChemExpress (HY-100579), the anti-IL-1β antibody was an anti-mouse IL-1β antibody (Clone B122, I-437, Leinco Technologies), and the anti-TNF antibody was an anti-mouse TNF-α antibody (Clone XT3.11, BE0058, Bio X Cell). Treatments began at week 4 and continued to sample collection at week 6. Fer-1 was administered intraperitoneally at 5 mg/kg every other day; anti-mouse IL-1β antibody was administered intraperitoneally at 200 μg/mouse every 3 days; and anti-mouse TNF-α antibody was administered intraperitoneally at 10 mg/kg twice weekly. During the experiment, body weight, stool consistency, and fecal bleeding were recorded regularly, and DAI scores were assigned according to weight loss, stool consistency, and occult/gross blood. The final DAI score was calculated as the sum of each component, and the detailed scoring criteria are provided in Supplementary Table S4. At the experimental endpoint, mice were euthanized, and the colon was completely isolated, cleared of contents, and gently rinsed. The colon length from the anus to the cecal end was then measured, and gross colon morphology was photographed under standardized background and scale conditions. Prespecified exclusions were severe procedure-related injury unrelated to TNBS colitis, unsuccessful intracolonic administration or substantial leakage, technical failure to obtain evaluable endpoint tissue, and attainment of humane endpoints; animals were not excluded on the basis of treatment response. One TNBS-group mouse died, so endpoint analyses included n = 5 in that group and n = 6 in all other groups; no other animals were excluded. Investigators assessing body weight, DAI, colon length, histopathology, and tissue-image quantification were blinded to allocation after samples were coded.

5.20. H&E staining

Freshly collected colon tissues were gently rinsed with PBS to remove residual contents and immediately fixed in tissue fixative (Servicebio, G1101). Fixed tissue samples were dehydrated, paraffin-embedded, sectioned, and subjected to routine H&E staining. Briefly, paraffin sections were deparaffinized, rehydrated, stained sequentially with hematoxylin and eosin, dehydrated, cleared, and mounted. Stained sections were examined under a light microscope, and representative images were selected after coded quantitative assessment. Histopathological assessment mainly included epithelial injury, inflammatory cell infiltration, crypt architectural destruction, and mucosal edema. Detailed scoring criteria are provided in Supplementary Table S5. One section per animal was independently scored by two investigators blinded to treatment using the prespecified criteria. Their mean was used as the final score; substantial discrepancies were re-evaluated by consensus while allocation remained concealed.

5.21. Statistical analysis

Quantitative data are presented as mean ± SEM. The biological/statistical unit was the patient, mouse, independent RNA-seq culture, or independent cell-culture experiment, as specified in each legend. Technical wells and microscopic fields were averaged within the biological unit and were not analyzed as independent replicates. Distributional assumptions were assessed using residual and Q–Q plots and Shapiro–Wilk testing, and variance homogeneity using Brown–Forsythe testing when applicable. Two-group comparisons used two-sided Student's t-test, Welch's t-test, or Mann–Whitney U test according to assumptions. Multiple groups were analyzed by one-way ANOVA with Tukey-adjusted comparisons or Kruskal–Wallis with Dunn-adjusted comparisons. Single-cell score comparisons used Wilcoxon tests with Benjamini–Hochberg correction when reported but were interpreted at the cell level. Longitudinal body-weight and DAI data were analyzed with REML mixed-effects models including treatment group, time, and treatment × time, with mouse as the repeated subject and Šídák-adjusted prespecified comparisons. Pre-death observations from the TNBS-group mouse were retained; subsequent values were missing and were not imputed. Analyses were performed in R 4.5.1 and GraphPad Prism 10.0. Tests were two-sided and P < 0.05 was considered significant. Exact P values are reported where retained in the archived output; otherwise figures use ns, *P < 0.05, **P < 0.01, ***P < 0.001; #P < 0.05, ##P < 0.01, ###P < 0.001.

CRediT authorship contribution statement

Qi Sun: Conceptualization, Formal analysis, Writing – original draft, Validation, Methodology, Software, Visualization, Writing – review & editing. Zhixian Jiang: Validation, Writing – original draft, Methodology, Visualization, Writing – review & editing. Xueliang Chen: Writing – original draft, Validation, Visualization, Writing – review & editing. Ganglei Liu: Writing – original draft, Validation, Writing – review & editing, Visualization. Mengyao Song: Validation, Writing – original draft. Lianwen Yuan: Writing – original draft, Writing – review & editing, Funding acquisition, Conceptualization, Project administration. Lichao Yang: Conceptualization, Methodology, Supervision, Writing – original draft, Writing – review & editing, Formal analysis.

Data availability statement

The RNA sequencing data generated in this study have been deposited in the Gene Expression Omnibus (GEO) database under accession number GSE338571. The public datasets used for analysis are also available from GEO. All other relevant raw data are available from the corresponding author upon reasonable request (yuanlianwen@csu.edu.cn, ylc0312@163.com).

Ethics statement

Human tissue samples used in this study were obtained from patients undergoing surgical treatment at The Second Xiangya Hospital of Central South University. All participants provided written informed consent before sample collection. The study protocol was reviewed and approved by the Institutional Ethics Committee of The Second Xiangya Hospital of Central South University (approval number: Z0919-01), and all procedures were conducted in accordance with the Declaration of Helsinki. All animal procedures were reviewed and approved by the Institutional Animal Care and Use Committee (IACUC), The Second Xiangya Hospital, Central South University, China (approval number: 20261133) and were performed in accordance with institutional guidelines for the care and use of laboratory animals. All efforts were made to minimize animal suffering.

Funding

This work was supported by the National Natural Science Foundation of China (82270590), the Hunan Provincial Natural Science Foundation (2025JJ50584), and the Postgraduate Research Innovative Project of Central South University (2026ZZTS0517).

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary data

The following are the Supplementary data to this article: