Fig 2: The interaction between senescent T cells and epithelial cells in vitro.a A schematic image illustrating how GzmA affects colonic epithelial cells through PARs signaling. b Overall scheme of the T cell isolation and senescence process (upper panel). Isolated pan-T cells after stimulation with anti-CD3/28 cocktail and rhIL-2 were assessed for cumulative population doubling level (PDL) (lower panel). Young cells were defined as 2 < cPDL < 5 and senescent cells as cPDL > 9. c The C12FDG staining analysis using flow cytometry (upper panels). The mRNA expression level of representative T cell senescence markers (p21Waf1, p16INK4A, CD28, and CD57) are shown (lower panel). ND not detected. d The mRNA (upper panel) and protein (lower panel) levels of GzmA in young and senescent T cells are performed by real-time PCR and ELISA, respectively. e The IHC analysis of cleaved caspase-3 in normal colon tissues from young and elderly individuals (upper panel) and 4-month and 18-month-old mouse colon tissues (lower panel). The right panels display the quantification data, respectively. f,g Human primary colon epithelial cells were co-cultured with young or senescence T cells (f) or treated with CM of young or senescence T cells (g) for 3 days. The luminescence-based assay and ICC to determine cleaved caspase-3/7 activity and cleaved caspase-3 expression level were performed, respectively. h IHC analysis for IL8 was performed in normal colon from young and elderly tissues (upper panel). The IHC analysis results are categorized into “Low,” “Moderate,” and “High” based on the intensity of the immunostaining, with representative images provided for each category. The left lower panel shows the quantification data for IHC analysis. The right lower panel shows a dot plot illustrating the expression of CXCL8 (IL8) in colonic epithelial cells from young and elderly individuals in the GSE178341 data set. i The CXCL8 (IL8) mRNA expression was analyzed by real-time PCR in human colon epithelial cells, which were co-cultured with young or senescent T cells (upper panel) and treated with CM of young or senescent T cells (lower panel) for 3 days. A p-value in (h) is calculated using the Chi-square test. The rest of the p-values are calculated using the Mann–Whitney U test. The graph in (b) is shown as mean ± standard deviation, while the rest of the bar graphs are shown as mean + standard deviation. “Young” and “Sen” in (c,d,f,g,i) indicate the young and senescent T cells, respectively. “Young” and “Old” in (e,h) indicate the young and the elderly individuals, respectively.

Fig 3: Epothilone A maintains intestinal barrier integrity by inhibiting GEF‐H1 to regulate the RhoA/ROCK pathway. Under physiological conditions (homeostasis), GEF‐H1 is maintained in an inactive, phosphorylated state, while RhoA remains bound to GDP. This arrangement keeps downstream effectors, including ROCK, LIMK, MLC2, and Cofilin, in an inactive form; thus, supporting a stable intestinal epithelial cytoskeleton and normal barrier function. However, during sepsis, immune‐cell‐derived Gzma induces the dephosphorylation of GEF‐H1 at Ser886. This activation promotes the formation of RhoA‐GTP, which subsequently activates ROCK. Activated ROCK then phosphorylates LIMK, MLC2, and Cofilin—resulting in cytoskeletal contraction and disruption of tight junctions that ultimately lead to barrier dysfunction. Treatment with Epothilone A stabilizes microtubules and inhibits GEF‐H1 activation; thereby suppressing the RhoA/ROCK pathway and preventing cytoskeletal hypercontraction while restoring epithelial barrier integrity during sepsis.

Fig 4: Activation of the GEF‐H1/RhoA signalling axis by Gzma under sepsis conditions. (A) Heatmap of transcriptome sequencing analysis showing differential expression profiles of small GTPases and cytoskeleton‐related genes in the intestinal tissues of WT‐Sham and the CLP sepsis model group (rows: genes; columns: experimental groups; colour scale: relative expression level). (B) Immunofluorescence staining of phalloidin (green, labelling F‐actin) in human colonic epithelial NCM460 cells after co‐culture with LPS‐pretreated NK92MI cells (DAPI staining for nuclei, blue; top row: control group; bottom row: co‐culture group; scale bar = 20 µm). (C) Coomassie blue‐stained SDS‐PAGE gel following Gzma immunoprecipitation (IP); the arrow indicates the GEF‐H1 band. (D) Interaction network between Gzma and candidate binding proteins predicted by the STRING database (Nodes: proteins). (E) Mass spectrum (MS/MS) of high‐affinity peptides binding Gzma to GEF‐H1, showing peptide sequences (EVEGLKDLLVGPGVELLLTPR, LVNLYGLLHGLQAAVAQQDTLMEAR, VGLFAEMTHFQAEEDGGSGMALPTLPR), Xcorr (cross‐correlation coefficient), and charge state. (F) Left: The three‐dimensional structure of the molecular docking of Gzma (yellow) and GEF‐H1 (blue) simulated by the GRAMM software; Right: A close‐up view highlighting the key interacting residues at the interface. (G) SPR analysis of GEF‐H1 binding to Gzma. Sensorgrams show concentration‐dependent binding of Gzma (.125–1 µM). The measured K D is 4.27⨯10−7 M, indicating a medium‐high affinity interaction. (H) Immunofluorescence co‐localization analysis of GEF‐H1 (red) and Gzma (green) in NCM460 cells after co‐culture with LPS‐pretreated NK cells (DAPI staining for nuclei, blue; top row: control group; bottom row: co‐culture group; scale bar = 20 µm). (I) RhoA G‐LISA activity assay in NCM460 (left) and Caco2 (right) cells after co‐culture with LPS‐stimulated NK cells (mean ± SEM, n = 3; ****p < .0001, one‐way ANOVA). (J) Western blot analysis of GEF‐H1, p‐GEF‐H1 (Ser886), MLC2, p‐MLC2 (Thr18/Ser19), LIMK, p‐LIMK (Thr508), Cofilin, p‐Cofilin (Ser3), and β‐actin (used as a loading control) was performed in NCM460 and Caco2 cells following co‐culture with LPS‐stimulated NK cells.

Fig 5: Elevated expression of Gzma in sepsis and its mediation of intestinal epithelial barrier dysfunction. (A) Western blot analysis of Gzma and β‐actin in PBMCs from healthy controls and septic patients. Data: mean ± SEM; n = 3 per group. Student's t‐test, **p < .01. (B) RT‐qPCR analysis of Gzma mRNA in PBMCs from both groups. Data: mean ± SEM; n = 3. *p < .05 by Student's t‐test. (C) Violin plots showing Gzma expression in NK and CD8+ T cells from WT‐Sham and WT‐CLP mice, based on scRNA‐seq of intestinal tissues. (D) Cell–cell interaction network in intestinal tissue under sepsis. Line thickness indicates interaction strength. (E) RT‐qPCR analysis of pro‐inflammatory cytokines (IL‐6, IL‐1β, TNF‐α, IFN‐β) in intestinal tissues from WT‐Sham and WT‐CLP mice. Data: mean ± SEM; n = 3. One‐way ANOVA with Tukey's test. (F) IHC staining showing Gzma localization in intestinal tissues. Scale bar: 100 µm. (G) Western blot analysis of Gzma, tight junction proteins (Claudin1, ZO‐1, Occludin), and E‐cadherin in intestinal tissues. Right panel: densitometry normalized to β‐actin. Data: mean ± SEM; n = 3. ***p < .001, **p < .01, *p < .05 by Student's t‐test. (H) Correlation between Gzma mRNA in septic patient PBMCs and SOFA scores: Pearson r = .6234, p < .0001. (I) RT‐qPCR analysis of tight junction proteins (Occludin, Claudin1, ZO‐1) and E‐cadherin in NCM460 cells co‐cultured with NC. Data: mean ± SEM; n = 3. ****p < .0001, ***p < .001, **p < .01 by one‐way ANOVA. (J) Protein levels of intestinal epithelial barrier‐related molecules (Occludin, Claudin1, ZO‐1, E‐cadherin) in NCM460/Caco2 cells decreased after co‐culture; Gzma secretion by LPS‐stimulated NK92MI cells increased. (K) TEER of NCM460/Caco2 cells was significantly reduced in the co‐culture system. Data: mean ± SEM; n = 3. **p < .01, *p < .05 by one‐way ANOVA. (L) FITC‐dextran permeability of NCM460/Caco2 cells was markedly elevated in the co‐culture system. Data: mean ± SEM; n = 3. ***p < .001, **p < .01 by one‐way ANOVA. (M) Immunofluorescence showed downregulated ZO‐1 in Caco2 cells (left, red) and increased Gzma secretion by LPS‐stimulated NK92MI cells (right, orange), confirming impaired barrier integrity. Scale bar: 20 µm.