KDM4A-induced tumor senescence enhances the efficacy of immunotherapy by inhibiting AGT-PHB1 axis-mediated mitophagy in colorectal cancer

ABSTRACT

Immune checkpoint inhibitors (ICIs) can re-active the immune response and induce a complete response in mismatch repair-deficient and microsatellite instability-high (dMMR/MSI-H) colorectal cancer (CRC). However, most CRCs exhibit proficient mismatch repair and microsatellite stable (pMMR/MSS) phenotypes with limited immunotherapy response because of sparse intratumoral CD8⁺ T-lymphocyte infiltration. Cellular senescence has been reported to involve immune cell infiltration through a senescence-associated secretory phenotype (SASP). However, the relationship between CRC cellular senescence and CD8⁺ T-lymphocyte infiltration remains unclear. Through integrated analysis of clinical cohorts and transcriptomic data across mismatch repair (MMR) subtypes, we identified cellular senescence as a hallmark of dMMR tumors, accompanied by elevated expression of KDM4A (lysine demethylase 4A). Clinically, KDM4A^high CDKN2A/p16^high expression correlated with improved CRC patient prognosis. Mechanistically, KDM4A upregulated AGT (angiotensinogen) expression through H3K9me3 demethylation and promoted CRC cellular senescence. Meanwhile, KDM4A-driven senescence suppressed tumor growth and enhanced intratumoral CD8⁺ T-lymphocyte infiltration via enhancing SASP-associated secretion. Furthermore, AGT disrupted PHB1 (prohibitin 1)-mediated basal mitophagy, triggering cytoplasmic mitochondrial DNA (mtDNA) accumulation that activated CGAS-STING1 signaling and enhanced SASP secretion. Crucially, KDM4A overexpression potentiated anti-PDCD1/PD1 efficacy in MSI-H CRC and reversed therapy resistance in MSS CRC. Conclusively, we established a KDM4A-AGT-PHB1 (KAP) grade system that robustly predicts immunotherapy responsiveness in pMMR CRC patients.

Abbreviation: AGT: angiotensinogen; BafA: bafilomycin A₁; CCCP: carbonyl cyanide 3-chlorophenylhydrazone; CRC: colorectal cancer; CDKN1A/p21: cyclin dependent kinase inhibitor 1A; CDKN2A/p16: cyclin dependent kinase inhibitor 2A; CHX: cycloheximide; Co-IP: co-immunoprecipitation; dMMR: deficient mismatch repair; EdU: 5-ethynyl-2’-deoxyuridine; GAPDH: glyceraldehyde-3-phosphate dehydrogenase; IL6: interleukin 6; IL8: interleukin 8; IHC: immunohistochemical; KDM4A: lysine demethylase 4A; mtDNA: mitochondrial DNA; MS: mass spectrometry; NFKB/NF-κB: nuclear factor kappa B; PHB1: prohibitin 1; PHB2: prohibitin 2; PINK1: PTEN induced kinase 1; pMMR: proficient mismatch repair; PRKN/parkin: parkin RBR E3 ubiquitin protein ligase; SASP: senescence-associated secretory phenotype; SA-GLB1/β-gal: senescence-associated galactosidase beta 1; TIMM23: translocase of inner mitochondrial membrane 23; TOMM20: translocase of outer mitochondrial membrane 20; TRIM21: tripartite motif containing 21; TUBB/beta-tubulin: tubulin beta class I.

Introduction

Colorectal cancer (CRC) persists as a major global health burden, with 2022 GLOBOCAN estimates documenting 1.93 million new cases (third highest incidence) and 904,000 deaths (second leading cancer mortality) worldwide [1]. DNA mismatch repair (MMR) is a conserved system that recognizes and repairs DNA replication and damage errors. Deficient mismatch repair (dMMR) tumors caused by MMR genes deficiency or MLH1 promoter hypermethylation, have a higher neoantigen and dense immune cell infiltration than predominant proficient MMR (pMMR), which is the basic of immunotherapy response [2]. While immune checkpoint inhibitors targeting PD-1 or PD-L1 have transformed treatment paradigms for 12–15% of dMMR CRC [3], their clinical impact remains limited in pMMR subtype (85% of cases) that exhibits immunologically “cold” microenvironments [4]. Several studies have shown that the density of CD8⁺ T-lymphocytes within the tumor is the main cause of the tumor immune response and is associated with a better prognosis in pMMR CRC [5–7]. Therefore, enhancement of intratumoral CD8⁺ T-lymphocytes infiltration is the key point to improve the response of immunotherapy in pMMR CRC patients.

Emerging evidence implicates cellular senescence, a stable proliferation arrest coupled with senescence-associated secretory phenotype (SASP), as a dual regulator of tumor immunity [8]. While SASP components can recruit CD8⁺ T-lymphocytes in some contexts [9,10], paradoxically, senescent pMMR CRC cells inhibit cytotoxic T-lymphocytes infiltration through secreting a high concentration of CXCL12 [11]. This mechanistic contradiction indicated whether senescence differentially modulates T-lymphocyte infiltration across MMR subtypes remains unexplored. Therefore, we speculate that senescent tumor cells may provide a new angle for enhancing intratumoral T-lymphocytes infiltration of CRC patients.

Here, we identified histone demethylase KDM4A as a molecular switch governing senescence-mediated immune remodeling by integrating clinical specimens and RNA-seq analysis across different MMR CRC. KDM4A^high CDKN2A/p16^high tumors exhibited increased CD8⁺ T-lymphocyte infiltration and superior 3-year survival in CRC cohorts. Through chromatin immunoprecipitation sequencing (ChIP-seq), RNA-seq, mass spectrometry (MS) and CRC cohort, the KDM4A-AGT-PHB1 axis was identified as the mechanism by which, KDM4A induced CRC senescence and facilitated intratumoral CD8⁺ T-lymphocyte infiltration. Furthermore, we demonstrated that KDM4A-induced CRC senescence enhanced the efficacy of immunotherapy in both MSI-H and MSS models. More importantly, our findings establish the KDM4A-AGT-HB1 axis as both a predictive biomarker for anti-PDCD1/PD1 therapy response and a therapeutic target for reprogramming immunologically “cold” pMMR CRC into immune-hot tumors.

Results

KDM4A^high CDKN2A^high expression indicated favorable clinical outcomes in CRC

To investigate the interplay between tumor cellular senescence and immune microenvironment remodeling, we performed spatial analysis of senescence markers and CD8⁺ T-lymphocyte infiltration in dMMR and pMMR CRC specimens using frozen and paraffin-embedded tissues. The dMMR CRC tumors exhibited substantially increased SA-GLB1/β-gal-positive cells accompanied by concomitant upregulation of both CDKN1A and CDKN2A compared to pMMR counterparts (Figure 1A,B). This senescence signature was further validated by increased levels of CDKN1A and CDKN2A (Figure 1C, Figure S1A), confirming cell cycle arrest activation in dMMR tumors. Immunohistochemical (IHC) quantification of CDKN2A also supported this result (Figure 1D, Figure S1B). Notably, the enhanced senescence phenotype spatially correlated with adaptive immune responses. IHC quantification demonstrated significantly (p < 0.05) elevated CD3⁺ and CD8⁺ T-lymphocyte infiltration within dMMR tumor (Figure 1A,B). To establish clinical relevance, we performed correlation analysis in an independent CRC cohort [12]. Intriguingly, CDKN2A expression showed significant (p < 0.001) positive correlations with both CD3⁺ (r = 0.328) and CD8⁺ (r = 0.281) cell densities in CRC tumors (Figure 1E,F). Subgroup analysis revealed enhanced correlations in dMMR tumors (CD3⁺: r = 0.330; CD8⁺: r = 0.283) compared to pMMR subgroup (CD3⁺: r = 0.200; CD8⁺: r = 0.244) (Figure S1C-D). To reveal the effect of CDKN2A expression on CRC prognosis, we performed Kaplan-Meier analysis based on CRC cohort (n = 375). These results revealed improved 3-year overall survival in patients with high CDKN2A expression (p = 0.029), particularly pronounced in dMMR subgroup (p = 0.003) compared to pMMR cases (p > 0.05) (Figure 1G). These findings collectively suggest that dMMR CRC tumors developed obvious tumor senescence that may facilitate CD8⁺ T-lymphocyte recruitment, potentially explaining the improved clinical outcomes observed in this molecular subtype.

Figure 1.

KDM4A^high CDKN2A/p16^high expression indicated favorable clinical outcomes in colorectal cancer. (A) Representative IHC images of SA-GLB1/β-gal, CDKN2A/p16, CDKN1A/p21, CD8 and CD3 within dMMR (top) and pMMR (bottom) CRC tumor tissues. Scale bars: 50 μm. (B) Quantification of SA-GLB1/β-gal, CDKN2A/p16, CDKN1A/p21, CD8 and CD3 within dMMR (n = 7) and pMMR (n = 7) CRC tumor tissues. (C) Western blot detected CDKN2A/p16 and CDKN1A/p21 expression (normalized to GAPDH expression) in dMMR and pMMR CRC tumors. Relative quantification is shown on the right. n = 5 per group. (D) Quantification of CDKN2A/p16 expression by IHC score in all 375 human CRC samples. (E) Pearson’s correlation analysis between CDKN2A/p16 expression levels and CD3⁺ cell density in CRC tissues. (F) Pearson’s correlation analysis between CDKN2A/p16 expression levels and CD8⁺ cell density in CRC tissues. (G) Kaplan – Meier survival curve of patients with colorectal cancer layered by the CDKN2A/p16 expression. (H) KEGG analysis of pathways enriched between dMMR CRC tumors and pMMR CRC tumors. (I) Volcano plot showing histone methylation genes expression between dMMR CRC tumors and pMMR CRC tumors. (J) Quantification of histone methylation genes by qRT-PCR in 16 (each) dMMR and pMMR tumors. (K) Western blot detected KDM4A expression (normalized to GAPDH expression) in 10 dMMR and pMMR CRC tumors. Relative quantification is shown on the right. (L) Representative immunostaining images of KDM4A expression within dMMR (top) and pMMR (bottom) CRC tumor tissues; scale bars: 100 μm (left), 20 μm (right). Quantification data is shown on the right. (M) Representative immunostaining images showing the expression correlation of KDM4A and CDKN2A/p16 within dMMR and pMMR CRC tumor tissues; scale bars: 50 μm. Pearson’s correlation results is shown on the right. (N) Pearson’s correlation analysis between KDM4A expression levels and CD3⁺ cell density in CRC tissues. (O) Pearson’s correlation analysis between KDM4A expression levels and CD8⁺ cell density in CRC tissues. (P) Kaplan – Meier survival curve of patients with CRC layered by the KDM4A expression in tumor tissue sections. (Q) Kaplan – Meier survival curve of patients with CRC layered by the KDM4A and CDKN2A/p16 expression in tumor tissue sections. (R-S) the density of CD8+ T-lymphocyte of patients with CRC layered by the KDM4A and CDKN2A/p16 expression in tumor tissue sections. Values are represented as mean ± SD. *, p < 0.05; **, p < 0.01; ***, p < 0.001; ns (no significance), by two-tailed Student t test (B-D and J-L), one-way ANOVA (R-S), Person’s correlation analysis (E-F and M-O) or log-rank test (G, P and Q).

To investigate the mechanistic connection between cellular senescence and CD8⁺ T-lymphocyte infiltration in dMMR tumors, we performed a transcriptome analysis with three dMMR tumors and three pMMR tumors. Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis revealed significant enrichment of cell cycle regulation and cellular senescence pathways in dMMR tumors (Figure 1H). Complementary Gene Ontology (GO) analysis demonstrated prominent involvement of chromatin remodeling-related epigenetic modifications (Figure S1E), consistent with established roles of epigenetic regulation in senescence processes [13]. Given the critical role of lysine methylation in chromatin dynamics, we performed RNA-seq to identify differentially expressed histone methyltransferases and demethylases. Heatmap and volcano plot analyses identified four candidate regulators (KDM4A, SETDB1, KDM1B, and EZH1) with significant expression variations (Figure 1I, Figure S1F). Subsequent qRT-PCR assays confirmed significant (p < 0.05) upregulation of KDM4A and KDM1B in dMMR specimens, with KDM4A showing more pronounced differential expression. This finding aligns with recent reports linking KDM4A overexpression to senescence-associated phenotypes [14], prompting focused investigation of this demethylase. Immunoblot analysis revealed elevated KDM4A protein levels in dMMR compared to pMMR specimens, and IHC assays confirmed predominant KDM4A expression within dMMR tumors (Figure 1K,L). Expanding to our CRC cohort, we identified significant positive correlations between KDM4A expression, senescence marker CDKN2A, and intratumoral CD8⁺ T-lymphocyte density – particularly pronounced in dMMR subgroups (Figure 1M-O, Figure S1G-H). Clinically, patients with high KDM4A expression demonstrated improved overall survival (Figure 1P), a trend maintained in MMR-stratified analyses. Besides, Tumor specimens with strong KDM4A immunoreactivity were predominantly found in patients with limited nodal dissemination (Table S1). Notably, two-gene prognostic and immune infiltration analysis revealed that KDM4A^high CDKN2A^high patients exhibited the most favorable prognosis with the highest intratumoral CD8⁺ T-lymphocyte infiltration (Figure 1Q-S). These findings suggest synergistic protective effects of KDM4A-mediated senescence and enhanced anti-tumor immunity in CRC.

KDM4A suppressed tumor growth through inducing CRC cellular senescence

Prior to investigating the role of KDM4A in CRC senescence, we first evaluated its expression level in normal human intestinal epithelial cell and CRC cell lines. Notably, KDM4A exhibited marked upregulation in CRC cells, with particularly elevated expression in dMMR CRC lines (LoVo, HCT116, DLD1, HCT8) compared to pMMR counterparts (HT29, SW480, SW620) (Figure S2A). To establish a senescence model, HCT116 and DLD1 cells were treated with H₂O₂. Increased SA-GLB1/β-gal-positive cells and enhanced expression of CDKN1A and CDKN2A indicated models successful constructed (Figure S2B-C). Intriguingly, H₂O₂-treated cells concurrently displayed elevated KDM4A levels, suggesting that high expression of KDM4A may be crucial for CRC cellular senescence (Figure S2C). To test this hypothesis, we generated stable KDM4A-overexpressing CRC cell lines (HCT8 and SW480) (Figure 2A, Figure S2D). KDM4A overexpression significantly increased SA-GLB1/β-gal-positive cells (Figure 2B), reduced proliferation-phase cell populations (Figure 2C,D) and upregulated CDKN1A and CDKN2A expression (Figure 2E). Cell cycle analysis revealed G₀/G₁ phase arrest (Figure 2F, Figure S2E-F), consistent with senescence-associated proliferation arrest. Functional assays confirmed suppressed colony formation capacity (Figure 2G) and reduced cell viability (Figure S2G), corroborating the senescence phenotype. These data indicated that KDM4A overexpression induced cellular senescence in CRC cells across different MMR status. To further validate these observations, we silenced KDM4A in H₂O₂-induced senescent cells (HCT116 and DLD1) using two efficient shRNAs (Figure 2H, Figure S2H). Knockdown KDM4A in senescent CRC cells decreased SA-GLB1/β-gal-positive cells, increased the number of proliferative cells and downregulated the expression of CDKN1A and CDKN2A (Figure 2I-L), which indicated that cellular senescence was ameliorated. Further cell cycle analysis also revealed an increase in S-phase cells and a decrease in G₀/G₁-phase cells (Figure 2M, Figure S2I-J). Moreover, colony formation assays and CCK8 assays confirmed that downregulation of KMD4A improved the capacity of cell proliferation in senescent cells (Figure 2N, Figure S2K). These data indicated that knockdown KDM4A impaired H₂O₂-induced cellular senescence in CRC. Given the critical role of SASP in tumor microenvironment remodeling, we analyzed IL6 and IL8 secretion using cell supernatants. As expected, KDM4A-overexpressing cells exhibited elevated cytokine levels, whereas KDM4A knockdown in senescent cells reduced their secretion (Figure 2O,P). Collectively, these results concluded that KDM4A overexpression was involved in CRC cellular senescence.

Figure 2.

KDM4A suppressed tumor growth through inducing CRC senescence. (A) Western blot analysis of KDM4A protein expression in HCT8 and SW480 cells with KDM4A overexpression. (B) Percentage of SA-GLB1/β-gal-positive cells (blue) in HCT8 and SW480 cells with KDM4A overexpression. Scale bars: 50 μm. Relative quantification is shown on the right (n = 3). (C-D) Edu assays detected the proliferation of KDM4A-overexpressing HCT8 and SW480 cells. Scale bars: 50 μm. Relative quantification is shown on the right (n = 3). (E) Western blot analysis of CDKN1A/p21 and CDKN2A/p16 expression in HCT8 and SW480 cells with KDM4A overexpression. (F) Relative quantification of cell cycle analysis in HCT8 and SW480 cells with KDM4A overexpression. (G) Colony formation assays of HCT8 and SW480 cells. Relative quantification is shown on the right (n = 3). (H) Western blot analysis of KDM4A protein expression in HCT116 and DLD1 cells (H₂O₂-induced senescence cells) with KDM4A knockdown. (I) Percentage of SA-GLB1/β-gal-positive cells (blue) in HCT116 and DLD1 cells (H₂O₂-induced senescence cells) with KDM4A knockdown. Scale bars: 50 μm. Relative quantification is shown on the right (n = 3). (J-K) Edu assays detected the proliferation of KDM4A-knockdown HCT116 and DLD1 cells (H₂O₂-induced senescence cells). Scale bars: 50 μm. Relative quantification is shown on the right (n = 3). (L) Western blot analysis of CDKN1A/p21 and CDKN2A/p16 in HCT116 and DLD1 cells (H₂O₂-induced senescence cells) with KDM4A-knockdown. (M) Relative quantification of cell cycle analysis in HCT116 and DLD1 cells (H₂O₂-induced senescence cells) with KDM4A-knockdown. (n = 3). (N) Colony formation assays of HCT116 and DLD1 cells (H₂O₂-induced senescence cells). Relative quantification is shown on the right (n = 3). (O-P) The level of IL6 and IL8 of cell supernatant were detected by Elisa assays in KDM4A overexpression cells (HCT8 and SW480, n = 5) and KDM4A-knockdown cells (HCT116 and DLD1, n = 5). (Q) Average growth curves of subcutaneous xenograft tumors in immune-competent mice after inoculation of MC38 or CT26 cells transfected with vector or with Kdm4a plasmid (n = 5). (R) tumors weight of xenograft tumors in immune-competent mice after inoculation of MC38 or CT26 cells transfected with vector or with a Kdm4a-expressing plasmid (n = 5). (S) Quantification of SA-GLB1/β-gal, KDM4A, CDKN2A/p16 and CDKN1A/p21 in mouse tumor tissues (n = 5). (T) Quantification of CD3⁺ and CD8⁺ T lymphocytes in mouse tumor tissues (n = 5). Values are represented as mean ± SD. *, p < 0.05; **, p < 0.01; ***, p < 0.001; ns (no significance), by two-tailed Student t test (B-D, F-G, O and R-T), one-way ANOVA (I-K, M-N and P-Q).

Building on the clinical correlation between KDM4A expression, tumor senescence, and CD8⁺ T-lymphocyte infiltration in CRC patients, we established syngeneic mouse models using CT26 (MSS/pMMR) and MC38 (MSI-H/dMMR) cell lines to investigate the role of KDM4A in tumor senescence and immune modulation (Figure S2L-M). Stable KDM4A-overexpressing CRC cells or vector controls were subcutaneously inoculated into immunocompetent mice. KDM4A overexpression significantly attenuated tumor growth in both MSS and MSI-H models, as evidenced by reduced tumor volumes and weights compared to vector controls (Figure 2Q-R, Figure S2M). To examine whether KDM4A suppressed tumor growth through inducing tumor senescence, we accessed tumor senescence with established senescence markers. KDM4A-overexpressing tumors exhibited significantly increased SA-GLB1/β-gal, CDKN2A and CDKN1A staining (Figure 2S, Figure S2N), confirming that KDM4A overexpression induced CRC tumor senescence. Furthermore, we assessed the density of CD8⁺ T-lymphocyte in xenograft tumor models. Consistent with clinical observations, KDM4A-overexpressing tumors displayed a marked increase in CD3⁺ and CD8⁺ T-lymphocyte infiltration compared to the control group (Figure 2T, Figure S2O). Notably, this immunomodulatory effect was observed in both MSS/pMMR and MSI-H/dMMR models, indicating that the immune modulation function of KDM4A across MMR subtype. Taken together, these results demonstrated that KDM4A inhibited tumor growth through inducing CRC cellular senescence and enhancing CD8⁺ T-lymphocyte infiltration.

KDM4A upregulated AGT expression by demethylating H3K9me3

Previous studies established the role of KDM4A in cellular senescence through epigenetic regulation [15]. To determine whether its histone demethylase activity drives CRC cellular senescence, we generated a catalytically inactive KDM4A mutant (KDM4A-Mut) using established mutant structure [16]. Transfection of KDM4A-Mut into HCT8 and SW480 cells failed to alter senescence-associated phenotypes, as evidenced by unchanged CDKN1A and CDKN2A protein levels, SA-GLB1/β-gal-positive cells and proliferation phase cells (Figure S3A-D). These findings confirm that KDM4A-mediated CRC cellular senescence requires its demethylation enzymatic activity.

We subsequently investigated the epigenetic landscape underlying this phenomenon. Immunoblot analysis showed that KDM4A overexpression markedly diminished H3K9me3 levels, with no significant impact on H3K36me3 levels (Figure 3A). Conversely, KDM4A knockdown elevated H3K9me3 modification (Figure 3B), indicating that H3K9me3 was the primary epigenetic target of KDM4A in CRC cells. Given that KDM4A regulates transcription of target genes by demethylation, H3K9me3 ChIP-seq and RNA-seq were hence selected for filtering target gene in HCT8 cells. ChIP-seq results revealed that trimethylation peaks of H3K9me3 around transcription start site were significantly decreased in KDM4A overexpression group compared with the Vector group (Figure 3C). RNA-seq analysis also demonstrated that cellular senescence pathways enriched in KDM4A-overexpressing CRC cells and senescence-associated transcriptional programs were activated (Figure 3D,E, Figure S3E). To identify candidate genes mediating the observed phenotype, we performed integrative analysis of transcriptional profiles (RNA-seq: p < 0.05&log2FC > 1) and epigenetic modifications (H3K9me3 ChIP-seq) using KDM4A-overexpressing HCT8 cells. Subsequent KEGG pathway analysis revealed significant enrichment of cardiovascular-associated signaling pathways (Figure S3F). Through cross-comparison of these datasets, AGT emerged as a prime candidate gene, showing concurrent transcriptional activation and H3K9me3 demethylation at its transcriptional start site (Figure 3F,G). Moreover, AGT has been reported to have a promoting role in cardiovascular aging disease [17]. Besides, AGT was recognized as key factor in K-Ras driven tumor senescence [18]. Given the unreported connection between AGT and CRC cellular senescence, we prioritized this target for mechanistic investigation.

Figure 3.

KDM4A upregulated AGT expression by demethylating H3K9me3. Western blot analysis of H3K9me3 and H3K36me3 protein expression (A) in KDM4A-overexpressing cells (HCT8 and SW480) and (B) KDM4A knockdown H₂O₂-induced senescence cells (HCT116 and DLD1). (C) Heatmaps showing H3K9me3 ChIP-seq enrichment of peaks near the accessible promoters (3.0kb upstream and downstream TSS per gene) of HCT8 cells transfected with vector or KDM4A plasmid. (D) GSEA profiling of gene expression with significant enrichment scores exhibiting a cellular senescence signature in KDM4A-overexpressing HCT8 cells compared with vector HCT8 cells. (E) Heatmaps showing the differential expression genes of SenMayo. Genes displayed are upregulated upon KDM4A-overexpressing HCT8 cells. (F) Venn diagram depicting the intersection gene set of ChIP-seq and RNA-seq in KDM4A-overexpressing HCT8 cells. (G) Genome browser tracks showing H3K9me3 occupancy near the promoters of AGT (angiotensinogen) from ChIP-seq analysis of HCT8 cells. (H-I) Relative quantification of AGT mRNA expression in KDM4A overexpression cells (HCT8 and SW480) and KDM4A knockdown cells (HCT116 and DLD1). (J) Western blot analysis of AGT protein expression in KDM4A-overexpressing cells (HCT8 and SW480) and (K) KDM4A knockdown H₂O₂-induced senescence cells (HCT116 and DLD1). (L-M) Independent ChIP-qPCR was conducted to evaluate the enrichment of H3K9me3 in KDM4A overexpression cells (HCT8 and SW480) and KDM4A knockdown cells (HCT116 and DLD1). (N) the relative activities of the AGT promoter in KDM4A overexpression cells (HCT8 and SW480). (O) the relative activities of the AGT promoter in KDM4A knockdown cells (HCT116 and DLD1). (P) Representative immunostaining images of AGT expression within dMMR (left) and pMMR (right) CRC tumor tissues (n = 375); scale bars: 50 μm. Quantification data is shown on the right. (Q) Representative immunostaining images showing the expression correlation of KDM4A and AGT within dMMR and pMMR CRC tumor tissues; scale bars: 50 μm. Correlation analysis result is shown on the right. (R) Kaplan – Meier survival curve of patients with CRC layered by the AGT expression in tumor tissue sections (n = 375). Values are represented as mean ± SD. *, p < 0.05; **, p < 0.01; ***, p < 0.001; ns (no significance), by two-tailed Student t test (H, L, N and P), one-way ANOVA (I, M and O), Person’s correlation analysis(Q), or log-rank test (R).

We then validated AGT expression and H3K9me3 modification patterns in H₂O₂-induced senescence models. Senescent CRC cells exhibited markedly elevated AGT levels accompanied by significant reduction of H3K9me3 modification compared to non-senescent controls (Figure S3G). Functional experiments demonstrated that KDM4A knockdown substantially decreased AGT mRNA and protein expression in HCT116 and DLD1 cells, whereas KDM4A overexpression enhanced AGT expression in HCT8 and SW480 cells (Figure 3H-K). Crucially, catalytically inactive KDM4A-Mut failed to modulate AGT expression (Figure S3A, S3H), confirming the requirement for its demethylase activity. To delineate the regulatory mechanism, ChIP-qPCR revealed increased H3K9me3 occupancy at the AGT promoter following KDM4A knockdown, while KDM4A overexpression reduced this repressive histone mark (Figure 3L,M), which was consistent with H3K9me3-mediated transcriptional silencing. Complementary luciferase reporter assays demonstrated that KDM4A overexpression enhanced AGT promoter activity, whereas its suppression diminished transcriptional output (Figure 3N,O), confirming that AGT transcriptional expression was regulated by KDM4A-mediated H3K9me3 modification.

To establish the clinical relevance of the KDM4A-AGT axis, we investigated its association with tumor immunology and patient prognosis in CRC. Immunohistochemical analysis of CRC specimens revealed predominant AGT expression in tumor, with significantly elevated levels in dMMR tumors compared to pMMR counterparts (Figure 3P). Importantly, AGT expression showed a significant (Pearson’s r = 0.522, p < 0.001; Figure 3Q) positive correlation with both KDM4A levels and CD8⁺ T-lymphocyte infiltration density (Figure S3I-K). Survival analysis demonstrated that high AGT expression correlated with improved overall survival (Figure 3R), a prognostic significance further validated in subsequent stratified analysis (Figure S3L).

Collectively, these findings revealed a mechanistically coherent pathway wherein KDM4A-mediated H3K9me3 demethylation at the AGT promoter drives its transcriptional activation in senescent CRC cells. The resultant AGT upregulation associates with enhanced anti-tumor immunity through CD8⁺ T-lymphocyte recruitment and confers favorable clinical outcomes, positioning AGT as both a functional mediator of senescence-associated immune modulation and a potential prognostic biomarker in CRC.

AGT is required for KDM4A-mediated cellular senescence and immune remodeling

Based on these findings, we investigated whether KDM4A-induced cellular senescence requires AGT expression. To examine the functional role of AGT in CRC senescence, we first generated stable AGT-overexpressing HCT8 and SW480 cell lines (Figure 4A). Compared to vector control, AGT expression significantly (p < 0.001) upregulated the protein levels of CDKN2A and CDKN1A, elevated the number of SA-GLB1/β-gal-positive cells and reduced cell proliferation (Figure 4A-D). Cell cycle profiling demonstrated that AGT overexpression induced G₀/G₁ phase arrest with corresponding S phase reduction (Figure 4E, Figure S4A). Furthermore, AGT-overexpressing cells exhibited higher secretion of senescence-associated cytokines IL6 and IL8 (Figure 4F). To confirm the necessity of AGT in senescence progression, we employed H₂O₂-induced senescence models in HCT116 and DLD1 cells with AGT knockdown (Figure 4G). AGT knockdown significantly (p < 0.001) downregulated protein expression of CDKN1A and CDKN2A, reduced the number of SA-GLB1/β-gal-positive cells and restored proliferative capacity (Figure 4G-J). Cell cycle analysis showed decreased G₀/G₁ populations and increased S phase cells following AGT knockdown (Figure 4K, Figure S4B-C). Notably, AGT silencing attenuated H₂O₂-induced IL6 and IL8 secretion (Figure S4L). These results collectively demonstrated that AGT expression was required for inducing CRC cellular senescence.

Figure 4.

AGT was required for KDM4A induced CRC senescence. (A) Western blot analysis of AGT, CDKN1A/p21 and CDKN2A/p16 protein expression in HCT8 and SW480 cells with AGT overexpression. (B) Percentage of SA-GLB1/β-gal-positive cells (blue) in HCT8 and SW480 cells with AGT overexpression. Scale bars: 50 μm. Relative quantification is shown on the right (n = 3). (C-D) Edu assays detected the proliferation of AGT-overexpressing HCT8 and SW480 cells. Scale bars: 50 μm. Relative quantification is shown on the right (n = 3). (E) Relative quantification of cell cycle analysis in HCT8 and SW480 cells with AGT overexpression (n = 3). (F) the levels of IL6 and IL8 of cell supernatant were measured by Elisa assay in AGT overexpressing cells (HCT8 and SW480). (G) Western blot analysis of AGT, CDKN1A/p21 and CDKN2A/p16 protein expression in HCT116 and DLD1 cells (H₂O₂-induced senescence cells) with AGT knockdown. (H) Percentage of SA-GLB1/β-gal-positive cells (blue) in HCT116 and DLD1 cells (H₂O₂-induced senescence cells) with AGT knockdown. Scale bars: 50 μm. Relative quantification is shown on the right (n = 3). (I-J) Edu assays detected the proliferation of AGT-knockdown HCT116 and DLD1 cells. Scale bars: 50 μm. Relative quantification is shown on the right (n = 3). (K) Relative quantification of cell cycle analysis in HCT116 and DLD1 cells (H₂O₂-induced senescence cells) with AGT knockdown. Scale bars: 50 μm (n = 3). (L) the levels of IL6 and IL8 of cell supernatant were measured by Elisa assay in AGT knockdown cells (HCT116 and DLD1). (M) HCT8 and SW480 cells with KDM4A overexpression were transfected with AGT knockdown. The protein expression of AGT, CDKN1A/p21 and CDKN2A/p16 were measured by western blot. (N) Percentage of SA-GLB1/β-gal-positive cells (blue) in HCT8 and SW480 cells. Scale bars: 50 μm. Relative quantification is shown on the right (n = 3). (O-P) Edu assays, (Q) colony formation assays, (R) CCK8 assays and (S) cell cycle analysis and (T) Elisa assays evaluated the cellular senescence in HCT8 and SW480 cells. Scale bars: 50 μm. Relative quantification is shown on the right (n = 3). (U-V) Quantification of tumor volume and tumor weight in a xenograft mouse model (n = 5). (W-X) Quantification of SA-GLB1/β-gal, AGT, CDKN2A/p16, CDKN1A/p21, CD3 and CD8 in mouse tumor tissues (n = 5). Values are represented as mean ± SD. *, p < 0.05; **, p < 0.01; ***, p < 0.001; ns (no significance), by two-tailed Student t test (B-F), one-way ANOVA (H-L and N-X).

To validate the essential role of AGT in mediating KDM4A-induced CRC senescence, functional rescue assays were conducted in HCT8 and SW480 cells. AGT knockdown significantly (p < 0.01) reduced the expression of CDKN1A and CDKN2A, decreased the number of SA-GLB1/β-gal-positive cells and restored the inhibitory effect of KDM4A overexpression on cell proliferation (Figure 4M-R). In addition, cell cycle analysis revealed that AGT knockdown rescued the KDM4A-induced S phase reduction and G₀/G₁ phase accumulation (Figure 4S, Figure S4D). Consistently, silencing AGT antagonized the increased secretion of IL6 and IL8 induced by KDM4A overexpression (Figure 4T). Collectively, these data demonstrated that AGT expression was required for KDM4A-induced cellular senescence in CRC cells.

To validate these findings in vivo, we established KDM4A-overexpressing syngeneic transplant models with Agt knockdown (Figure S4E). Agt knockdown markedly rescued the suppressive effects of KDM4A overexpression on tumor growth and reduced the staining of senescence markers (SA-GLB1/β-gal, CDKN2A and CDKN1A) (Figure 4U-W, Figure S4F-H). Strikingly, Agt knockdown diminished intratumoral CD8⁺ T-lymphocyte infiltration, suggesting the involvement of KDM4A-AGT axis immune microenvironment remodeling through the KDM4A-AGT axis (Figure 4X, Figure S4H). Chemokine signals, such as CCL5 and CXCL10, are the major factors in the migration of T-lymphocytes from the peripheral blood to the tumor region [19]. We hypothesized that KDM4A-AGT axis mediated intratumoral CD8⁺ T-lymphocyte infiltration through T lymphocyte-related chemokines. Therefore, peripheral blood samples were collected from the MC38 and CT26 mouse models to examine the concentration of chemokines. We found that KDM4A overexpression elevated serum levels of chemokines CCL5 and CXCL10 in mice, which was reversed by Agt knockdown (Figure S4I). In vitro T-lymphocyte migration assays confirmed that conditioned media from KDM4A-overexpressing cells enhanced T cell migration, which was reversed by Agt knockdown (Figure S4J-K). These data illustrated that AGT serves as the critical executor of KDM4A-mediated CRC cellular senescence and establishes an immunostimulatory tumor microenvironment through chemokine-dependent CD8⁺ T-lymphocyte recruitment.

AGT promoted degradation of PHB1 in ubiquitination manner and impaired mitophagy in senescence cells

To elucidate the molecular mechanism of AGT-driven cellular senescence, we performed anti-AGT co-immunoprecipitation (co-IP) combined with mass spectrometry (MS) in HCT8 cells. MS analysis identified PHB1 (prohibitin 1), a mitochondrial inner membrane protein implicated in cellular senescence [20], as the forefront ranking interactor (Figure S5A-B, Table S2). Reciprocal co-IP assays validated endogenous AGT-PHB1 interaction in HCT8 and SW480 cells, and immunofluorescence staining showed their cytoplasmic colocalization (Figures 5A,B, Figure S5C).

Figure 5.

AGT promoted degradation of PHB1 in ubiquitination manner and impaired mitophagy in senescence cells. (A) Co-IP assays confirmed the interaction between AGT and PHB1 using anti-AGT antibody or anti-PHB1 antibody in HCT8 and SW480 cells. IgG was used as a negative control, and protein extracts were used as the positive control (“Input”). (B) Quantification of colocalization immunofluorescence images show AGT (red) and PHB1 (green) in HCT8 and SW480 cells. (C) Relative quantification of PHB1 mRNA in AGT-overexpressing cells (HCT8 and SW480). (D) Immunoblotting analysis of protein expression of PRKN, PINK1, PHB2 and PHB1 in the total, cytosolic (cyto) and mitochondrial (mito) fractions of AGT overexpressing cells (HCT8 and SW480) (n = 3). The mitochondria protein was quantified and normalized to TOMM20; others were quantified and normalized to TUBB. (E) Quantification of colocalization immunofluorescence images show TOMM20 (red) and PHB1 (green) in AGT-overexpressing cells (HCT8 and SW480 cells). (F) AGT overexpressing cells (HCT8 and SW480) treated with 10 nM BafA for 1.5 h. Immunoblotting analysis of protein expression of mitophagy flux (TIMM23 and TOMM20), senescence marker (CDKN1A/p21 and CDKN2A/p16), PHB1 and PHB2 in HCT8 and SW480 cells. (G) Colocalization of mitochondria and autolysosomes analyzed by staining HCT8 and SW480 cells with MitoTracker (green) and LysoTracker (red). Yellow dots indicated ongoing mitophagy. White arrows indicate colocalization points. Scale bar: 5 μm. Quantification is shown on the right (n = 3). (H) AGT overexpressing cells (HCT8 and SW480) were treated with or without 10 μM CCCP for 2 h. Fluorescent dots in HCT8 and SW480 cells transfected with mito-Keima plasmid were observed by confocal microscopy. Scale bar: 5 μm. The relative ratio of red dots to green dots per image was quantified (n = 3). (I-J) Quantification of colocalization immunofluorescence images show TOMM20 (red) and PHB1 (green) in AGT overexpressing cells (HCT8 and SW480) treated with or without 10 μM CCCP for 2 h. (K) Detection of mitochondrial ultrastructure using TEM. Black triangle: mitophagosome. The number of mitophagosome was quantified in per cell. Scale bar: 5 μm (top), 1 μm (bottom). (L) Western blot analysis of AGT and PHB1 protein levels in HCT8 or SW480 cells transfected with vector or AGT with cycloheximide (CHX, 50 μg/mL) treatment at the indicated time. The quantification of PHB1 expression is shown in the right. (M) Western blot analysis of PHB1 expression in AGT-overexpressing HCT8 and SW480 cells treated with or without MG132 (10 μM, 12 h). (N) Ubiquitination amounts of PHB1 in the co-IP products of in AGT-overexpressing (HCT8 and SW480) cells treated with MG132 (left) and the expression of AGT and PHB1 in total protein lysates (right) are shown. (O) Co-IP assays confirmed the interaction between PHB1 and TRIM21 using anti-PHB1 antibody in HCT8 and SW480 cells. IgG was used as a negative control, and protein extracts were used as the positive control (“Input”). (P) Western blot analysis of IP using anti-PHB1 antibody in HCT8 and SW480 cells transfected with vector or AGT plasmids. Values are represented as mean ± SD. *, p < 0.05; **, p < 0.01; ***, p < 0.001; ns (no significance), by two-tailed Student t test (C, G, and K), one-way ANOVA (H and L).

PHB1 typically forms a complex with PHB2 on the inner mitochondrial membrane to regulate mitophagy [21]. Meanwhile, mitophagy dysfunction is linked to senescence [22]. To explore the effect of AGT on PHB1, we first assessed PHB1 expression in AGT-overexpressing cells. Although PHB1 mRNA levels remained unchanged (Figure 5C), its protein expression was significantly reduced in the mitochondria of AGT-overexpressing cells (Figures 5D,E, Figure S5D-E). This post-transcriptional suppression suggested AGT might impair PHB1-mediated mitophagy to induce senescence. To test this hypothesis, we analyzed mitophagy flux using bafilomycin A₁ (BafA), a lysosomal inhibitor. AGT overexpression elevated mitochondrial membrane proteins TOMM20 and TIMM23 to levels comparable to BafA treatment alone (Figures 5F). Notably, BafA failed to further increase TOMM20 and TIMM23 in AGT-overexpressing cells, indicating AGT-induced mitophagy blockade. Consistently, AGT overexpression reduced mitochondria-lysosome colocalization and blunted CCCP-induced mitophagy activation (Figure 5G-K). Importantly, BafA synergized with AGT to increase the expression of senescence markers CDKN1A and CDKN2A (Figure 5F), suggesting lysosomal dysfunction exacerbates AGT-driven CRC cellular senescence.

Given the centrality of PINK1-PRKN/Parkin signaling in mitophagy [23], we examined the impact of AGT on their expression. AGT overexpression downregulated mitochondrial PINK1 and PRKN protein levels (Figure 5D). Notably, CCCP-induced PRKN mitochondrial translocation was attenuated in AGT-overexpressing cells (Figure 5I,J, Figure S5F-G), further supporting AGT-mediated PINK1-PRKN pathway inhibition. It has been reported that PHB2 is required for PRKN-mediated mitophagy and binds directly to LC3, while PHB1 binds indirectly to LC3 [24]. Notably, knockdown of PHB1 also downregulated PHB2 expression, and PHB1 expression is necessary for the maintenance of PHB2 protein [25]. Therefore, we hypothesized that PHB1 mediates PINK1-PRKN-dependent mitophagy by regulating PHB2 protein expression in AGT-induced senescent CRC cells. Co-IP assays confirmed the interaction of AGT with PHB2 in HCT8 and SW480 cells (Figure S5H). In addition, PHB1 knockdown reduced PHB2 protein without affecting its mRNA (Figure S5I-J). Moreover, we examined the protein expression of PHB2 in AGT overexpression cells. As anticipated, AGT overexpression decreased the protein expression of PHB2 in mitochondrial (Figure 5D). These data implied that AGT destabilized the PHB1-PHB2 complex and impaired PINK1-PRKN signaling. Collectively, AGT promoted CRC cellular senescence by suppressing PHB1-PHB2-PINK1-PRKN mediated mitophagy.

To determine how AGT reduces PHB1 protein levels without affecting its mRNA, we first assessed PHB1 protein stability using cycloheximide (CHX) chase assays. AGT-overexpressing cells exhibited accelerated PHB1 degradation compared to the controls (Figure 5L), suggesting AGT enhances PHB1 protein degradation. Given the two major protein degradation pathways in eukaryotes [26], we systematically tested their contributions. Proteasome inhibition with MG132 significantly restored PHB1 levels in AGT-overexpressing cells (Figure 5M), whereas lysosomal inhibition by BafA showed no rescuing effect (Figure 5F). Notably, AGT overexpression markedly increased PHB1 ubiquitination levels (Figure 5N), indicating that AGT-induced PHB1 degradation was proteasome-dependent degradation. Since TRIM21 is a known E3 ligase for PHB1 ubiquitination in nasopharyngeal carcinoma [27], we investigated its role in CRC cells. Co-IP revealed endogenous PHB1-TRIM21 complexes in both HCT8 and SW480 cells (Figure 5O). Importantly, AGT overexpression strengthened this interaction (Figure 5P), suggesting TRIM21 recruitment mediated AGT-induced PHB1 protein degradation. Collectively, these data demonstrate that AGT coordinates with TRIM21 to mediated ubiquitination of PHB1, thereby promoting its proteasome-dependent degradation in CRC cells.

PHB1 expression restored basal mitophagy and restrained mtDNA release

To further demonstrate the important role of PHB1 mediated mitophagy in cancer cellular senescence. We knocked down PHB1 in different cancer cells. In HCT8 and SW480 cells, PHB1 knockdown upregulated CDKN1A and CDKN2A expression, increased the number of SA-GLB1/β-gal-positive cells, reduced cell proliferation and activated the secretion of IL6 and IL8 (Figure S6A-E). Besides, knockdown of PHB1 downregulated the total and mitochondrial levels of PHB2, PINK1 and PRKN proteins and decreased the co-location dots of lysosome and mitochondrial, indicating PHB1 expression was required for PHB2-PINK1-PRKN mediated mitophagy (Figure S6F-G). To investigate whether loss of PHB1-PINK1-PRKN axis is a feature of senescence in other cell types, we knocked down PHB1 in MCF-7 and PANC-1 cells. We also found that knockdown of PHB1 promoted cellular senescence and inhibited PHB2-PINK1-PRKN mediated mitophagy (Figure S6H-N). These findings demonstrated that the mechanistic link between PHB1-mediated mitophagy and cellular senescence may be conserved across diverse cancer models.

To explore the functional link between PHB1-mediated mitophagy and AGT-driven senescence, we examined phenotype of mitophagy and cellular senescence. In AGT-overexpressing cells, PHB1 overexpression reversed the accumulation of protein expression of TIMM23 and TOMM20 proteins, restored mitophagy flux, and enhanced mitochondria-lysosome colocalization (Figure 6B-D). To examine whether PHB1 rescued mitophagy by restoring PHB2-PINK1-PRKN pathway, we assessed the expression of PHB2-PINK1-PRKN pathway and analyzed the colocalization of PRKN and TOMM20. PHB1 overexpression rescued the protein levels of mitochondrial PHB2, PINK1, and PRKN and rescued PRKN mitochondrial translocation under AGT overexpression (Figure 6E-G, Figure S7A-B). Concomitantly, PHB1 overexpression decreased the protein expression of CDKN1A and CDKN2A (Figure 6B), reduced the number of SA-GLB1/β-gal-positive cells (Figure 6H), increased cell proliferation (Figure 6I,J), and reduced SASP factors (IL6 and IL8) (Figure 6K,L). These data confirm that PHB1 rescues mitophagy and reverses senescence via PHB2-PINK1-PRKN axis restoration in AGT-induced CRC cellular senescence.

Figure 6.

PHB1 expression restored basal mitophagy and restrained mtDNA release. (A) HCT8 and SW480 cells with AGT overexpression were transfected with control or PHB1 plasmids. The protein expression of AGT, PHB1, TIMM23, TOMM20, CDKN1A/p21 and CDKN2A/p16 were measured by western blot. (B) Fluorescent dots in HCT8 and SW480 cells transfected with mito-Keima plasmid were observed by confocal microscopy. Scale bar: 5 μm. (C) the relative ratio of red dots to green dots per image was quantified and shown on the right (n = 3). (D) Colocalization of mitochondria and autolysosomes analyzed by staining HCT8 and SW480 cells with MitoTracker (green) and LysoTracker (red). Yellow dots indicated ongoing mitophagy. White arrows indicate colocalization points. Scale bar: 5 μm. Quantification is shown on the right (n = 3). (E) Immunoblotting analysis of protein expression of PRKN, PINK1, PHB2 and PHB1 in the total and mitochondrial (mito) fractions of HCT8 and SW480 cells (n = 3). (F-G) Quantification of colocalization immunofluorescence images show TOMM20 (red) and PRKN (green) in KDM4A overexpressing cells (HCT8 and SW480 cells). (H) Percentage of SA-GLB1/β-gal-positive cells (blue) and (I-J) Edu assays in HCT8 and SW480 cells. Scale bars: 50 μm. Relative quantification is shown on the right (n = 3). (K-L) the levels of IL6 and IL8 in cell supernatant were detected by Elisa assays in HCT8 and SW480 cells (n = 5). (M) the DNA of AGT-overexpressing cells (HCT8 and SW480) with or without PHB1 overexpression was extracted. qRT-PCR assays were used to detect the mtDNA levels (D-loop) in the cytosol relative to nuclear DNA levels (18S) in the whole cell lysates (n = 5). (N) Quantitative analysis of the colocalization between cytoplasmic DNA (Pico green) and mitochondria (TOMM20). (O) the protein expression of CGAS, STING1 and p-NFKB were measured by western blot. (P) Quantification of immunostaining results of PHB1 expression in CRC tumor tissues (n = 375). (Q) Kaplan-Meier survival curve of patients with CRC layered by the PHB1 expression in tumor tissue sections (n = 375). Values are represented as mean ± SD. *, p < 0.05; **, p < 0.01; ***, p < 0.001; ns (no significance), by two-tailed Student t test (P), one-way ANOVA (C-D and H-N), or log-rank test (Q).

Mitochondria is a major regulator of SASP [28]. A recent study has revealed that PHB1-knockdown enhanced cytoplasmic release of mitochondrial DNA (mtDNA) [29]. Besides, conditional knockout PHB1 in mouse intestinal epithelial cells significantly increased the secretion of inflammatory factors [30]. Therefore, we focused on PHB1-mtDNA signaling in AGT-induced senescent CRC cells. AGT overexpression increased the ratio of cytoplasmic mtDNA to nuclear DNA and enhanced cytoplasmic mtDNA accumulation, both of which could be reversed by PHB1 overexpression (Figure 6M,N, Figure S7C). Consistent with the results of mtDNA-CGAS-STING1 axis driven SASP [31], AGT upregulated CGAS-STING1 signal, which was abolished by PHB1 overexpression (Figure 6O). These results imply a PHB1-mtDNA-CGAS-STING1 axis in AGT-mediated SASP regulation.

Besides, we found that PHB1 was highly expressed in pMMR tumors, and low PHB1 expression was correlated with improved CRC prognosis and enhanced CD8⁺ T-lymphocyte infiltration (Figure 6P,Q, Figure S7D-G). All these findings confirm that the AGT-PHB1 axis modulates SASP activation via the mtDNA-CGAS-STING1 signaling pathway, thereby orchestrating intratumoral CD8⁺ T-lymphocyte infiltration.

KDM4A enhanced the anti-PD1 therapy in CRC

Given the established role of CD8⁺ T-lymphocyte infiltration in potentiating immunotherapy for colorectal cancer, we investigated whether KDM4A-driven tumor senescence could enhance anti-PDCD1/PD1 therapeutic efficacy (Figure 7A). In MC38 (MSI-H) tumor-bearing mice, KDM4A overexpression significantly potentiated anti-PDCD1 responses, as evidenced by reduced tumor growth and prolonged survival compared to control groups (Figure 7B-E). Notably, this combinatorial effect extended to immunotherapy-resistant CT26 (MSS) models, where KDM4A overexpression converted anti-PDCD1-refractory tumors into responsive ones, achieving significant tumor regression and markedly improved survival (Figure 7B,F-H).

Figure 7.

KDM4A enhanced the anti-PDCD1/PD1 therapy in both MSI-H and MSS CRC. (A) Schematic of anti-PDCD1 treatment model: anti-PDCD1 or IgG antibodies were given every 3 days for a total of 5 doses starting on day 6 after tumor inoculation. Mouse cells (MC38 or CT26) with KDM4A or vector overexpression were implanted into syngeneic mouse (C57BL/6 or BALB/c) and treated with anti-PDCD1 or IgG. (B-H) Tumor wight, representative pictures, tumor growth and survival of MSI-H and MSS CRC tumors were presented (each group, n = 7). (I) We retrospectively collected the clinical data and pretreatment pathological slides of 112 patients with pMMR CRC who had received anti-PDCD1 treatment (anti-PDCD1 CRC cohort). The assessment of KAP grade. KAP grade was designed to evaluate KDM4A-AGT-PHB1 axis in predicting the response time to anti-PDCD1 therapy in pMMR CRC. (J) The prognosis differences of pMMR CRC treated with anti-PDCD1 therapy between low KAP grade and high KAP degrade. (K) Representative pictures of KDM4A, AGT, PHB1, CD3 and CD8 of pMMR CRC before anti-PDCD1 therapy in indicated KAP grade. Scale bars: 50 μm. (L) Quantification of CD3⁺ and CD8⁺ T lymphocytes in 112 pMMR colorectal cancer samples. (M) Pearson’s correlation analysis between KAP grade and disease-free survival in anti-PDCD1 CRC cohort. Values are represented as mean ± SD. *, p < 0.05; **, p < 0.01; ***, p < 0.001; ns (no significance), by two-tailed Student t test (L), one-way ANOVA (B, D, G), Pearson’s correlation analysis (M), or log-rank test (E, H, J).

To evaluate clinical relevance, we developed a KAP grading system (integrating KDM4A, AGT, and PHB1 immunohistochemical scores) in a retrospective cohort of anti-PDCD1-treated pMMR CRC patients (Figure 7I). High KAP grades were correlated with superior clinical outcomes, demonstrating significant (p < 0.01) increases in disease-free survival and overall survival (Figure 7J). Pre-treatment immunohistochemical analysis revealed that tumors with high KAP grades exhibited a significant (p < 0.001) increase in CD8⁺ T-lymphocyte infiltration density compared to tumors with low-KAP degrades (Figure 7K,L). More importantly, further analysis revealed a significant positive correlation between KAP scores and disease-free survival time in patients received immunotherapy, implying the association between KDM4A-AGT-PHB1 axis and persistent pathological response to anti-PDCD1 therapy (Figure 7M). Collectively, these results indicated that KDM4A enhanced the anti-PDCD1 therapy in CRC and high KAP grade was associated with better immune response in patients with pMMR CRC.

Discussion

Our study established KDM4A as a critical epigenetic regulator of CRC cellular senescence and immunomodulation. Key findings demonstrated that KDM4A-driven cellular senescence programs orchestrated CD8⁺ T-lymphocyte infiltration through the novel KDM4A-AGT-PHB1 axis (Figure 8). Notably, KDM4A overexpression synergized with immune checkpoint inhibitors in MSI-H tumors and reversed immunotherapy resistance in MSS CRC models. Clinical translation potential was evidenced by the ability of KDM4A-AGT-PHB1 grade system to identify pMMR patients likely to benefit from anti-PDCD1 therapy.

Figure 8.

A schematic model illustrating the mechanisms of KDM4A in CRC cellular senescence. Overexpression of KDM4A induces cellular senescence in colorectal cancer cells, suppresses tumor growth, and recruits CD8⁺ T lymphocytes. Mechanistically, KDM4A demethylates H3K9me3 modifications at the AGT promoter region, thereby upregulating AGT expression. Subsequently, AGT enhances ubiquitination-mediated protein degradation of PHB1, which disrupts the stability of the PHB protein complex and inhibits PINK1-PRKN signaling-mediated mitophagy initiation, ultimately leading to a cellular senescence phenotype. Concurrently, PHB1 downregulation promotes mitochondrial DNA (mtDNA) release into the cytoplasm, activating the intracellular CGAS-STING1-SASP signaling axis.

Cellular senescence constitutes a dual-edged tumor suppression mechanism involving cell cycle arrest and immune activation [32]. While therapeutic senescence induction shows clinical promise [33–36], the differential immune consequences between MMR subtypes remain poorly understood. Building on our prior work comparing pMMR and dMMR microenvironments [37], we revealed that dMMR tumors exhibit enhanced senescence markers (CDKN1A and CDKN2A) and CD8⁺ T-lymphocyte infiltration compared to pMMR counterparts. Given that senescent tumors are defined by growth arrest and SASP-mediated immune activation [38,39], we systematically investigated the functional interplay between KDM4A overexpression and CRC senescence. Current understanding of KDM4A in cellular senescence remains paradoxical. In pancreatic cancer models, miR-137-mediated KDM4A downregulation activated p53 to drive cellular senescence [40], suggesting tumor-suppressive senescence requires KDM4A inhibition. Conversely, emerging evidences position KDM4A as a senescence promoter: It facilitated nucleus pulposus cell senescence through H3K9me3-dependent ALKBH5 upregulation [14], and critically maintained CDKN2A expression and SASP production during stable senescence [15]. Our study identified a novel pro-senescence role of KDM4A in CRC pathogenesis, where its overexpression triggered classic hallmarks of cellular senescence through CDKN1A and CDKN2A upregulation, activated secretion of senescence-associated proinflammatory mediators, and enhanced infiltration of CD8⁺ T-lymphocytes.

Building on the epigenetic regulatory function of KDM4A, we first revealed that its demethylase activity is essential for initiating CRC cellular senescence. Through systematic screening, we identified AGT as the critical downstream effector mediating this process. AGT, as a precursor of angiotensin, is an important component of the renin-angiotensin system, which is classically known as a circulating or hormonal system that regulates blood pressure, electrolytes, and fluid homeostasis [41]. A myriad of studies demonstrated that the downstream active substance of AGT, Ang II, induced vascular senescence [42]. Notably, transcriptional upregulation of AGT via KLF6 was reported in K-Ras-driven senescence models, and AGT promoted cancer cells senescence through oxidative stress potentiation [18]. However, the epigenetic regulation of AGT and its mechanistic role in tumor senescence remain unexplored. Our study bridged this knowledge gap by demonstrating that KDM4A directly upregulates AGT expression through H3K9me3 demethylation at its promoter region. Functional validation revealed that AGT overexpression recapitulated senescence phenotypes, whereas AGT knockdown not only attenuated H₂O₂-induced senescence, but also abrogated KDM4A-mediated senescence initiation. Crucially, we established the clinical relevance of this axis: the expression of AGT was correlated positively with both senescence markers and CD8⁺ T-lymphocyte infiltration in CRC specimens, with significantly higher levels observed in dMMR tumors compared to pMMR counterparts. These findings revealed AGT as a molecular linchpin connecting KDM4A-mediated epigenetic remodeling to senescence-driven immune activation.

While extensive research has focused on the paracrine effects of angiotensin II in age-related pathologies [43] the intracellular mechanisms of its precursor AGT in senescence regulation remain enigmatic. To delineate the cell-autonomous functions of AGT in CRC cellular senescence, we employed co-IP coupled with proteomic profiling, identifying PHB1 as a key downstream effector. PHB1, an inner mitochondrial membrane protein and is required for maintenance of mitochondrial function [21], has been shown to regulate basal mitophagy and redox homeostasis, with its depletion accelerating senescence via mitochondrial dysfunction [20]. Similarity, our investigation revealed that AGT overexpression promoted the protein degradation of PHB1 in ubiquitination manner, thereby impairing the PHB1-mediated basal mitophagy. Under homeostatic conditions, PHB1 and PHB2 assemble into a complex predominantly localized within the mitochondrial membrane structure. Besides, PHB2 is required for PINK1-PRKN signal mediated mitophagy [44]. In addition, PHB1 Knockdown impaired the expression of PHB2, implying the importance of PHB1 expression for the stability of the PHB complex [25]. In this regard, we investigated the effect of PHB1 expression on PHB2 expression. Notably, PHB1 knockdown not only reduced mitochondrial PHB2 protein levels without affecting its transcription, but also accelerated PHB2 degradation in cycloheximide chase assays (Figure S8A-B). Conversely, PHB1 overexpression rescued PHB2 expression in AGT-overexpressing cells (Figure 6D,E), confirming the role of PHB1 as a post-translational stabilizer of PHB2. Our study pioneers the connection between AGT-mediated PHB1 degradation and mitophagy impairment in senescence pathogenesis.

Emerging evidence has highlighted mitochondria as central regulators of SASP through their involvement in mtDNA-mediated signaling pathways [45]. Specifically, PHB1 has been identified as a critical gatekeeper of mtDNA release through its regulation of mitochondrial inner membrane permeability [29]. Notably, genetic ablation of PHB1 in paneth cells induced mitochondrial dysfunction and subsequently enhanced inflammatory factor secretion in the intestinal microenvironment [30], suggesting its pivotal role in modulating intestinal inflammatory responses. Building on these mechanistic insights, our study extended the current understanding by elucidating the functional hierarchy of the KDM4A-AGT-PHB1 axis in SASP regulation. Through systematic investigation, we demonstrated that PHB1 degradation triggered cytoplasmic mtDNA accumulation in senescent CRC cells, which subsequently activated the CGAS-STING1-mediated innate immune signaling cascade. This mechanistic link between cytosolic mtDNA accumulation and SASP secretion in KDM4A-AGT-PHB1 axis-driven CRC senescence aligns with established links of mtDNA-mediated SASP activation [45].

While this study has provided novel insights into the epigenetic regulation of tumor senescence, several limitations warrant cautious interpretation of our findings. First, although we have demonstrated that KDM4A transcriptionally upregulates AGT expression via H3K9me3 demethylation, the pleiotropic nature of this histone demethylase merits further exploration. Given that KDM4A also modulated H3K27me3 and H3K36me3 modifications [46,47], future studies should systematically dissect whether these alternative epigenetic mechanisms contribute to its senescence-regulatory functions. Second, the reliance on subcutaneous tumor models constrains our ability to delineate the spatiotemporal dynamics of CD8⁺ T-lymphocyte infiltration in KDM4A-induced senescent tumors. Establishing orthotopic colorectal cancer xenografts hat preserve tumor-stroma interactions would better recapitulate the immunosuppressive microenvironment and enable longitudinal tracking of immune cell trafficking using advanced imaging techniques. Third, the translational implications of our findings require rigorous validation. While KDM4A-induced senescence shows therapeutic promise, its clinical applicability necessitates: (1) comprehensive biomarker development using patient-derived organoid models, (2) pharmacological evaluation of senescence-inducing compounds in immunocompetent systems, and (3) longitudinal assessment of therapy-induced senescence in clinical cohorts through multiplex immunohistochemistry.

Collectively, our study delineated a novel epigenetic axis governing tumor senescence and anti-tumor immunity in CRC. Mechanistically, KDM4A overexpression induced CRC cellular senescence through the AGT-PHB1 signaling cascade by triggering mitochondrial genome instability and subsequent CGAS-STING1 pathway activation, which ultimately enhanced intratumoral CD8⁺ T-lymphocyte infiltration and mediated tumor growth suppression. Therapeutically, we provided compelling preclinical evidence that pharmacological induction of KDM4A expression synergizes with anti-PDCD1/PD1 therapy, effectively overcoming adaptive resistance in immunocompetent models. Clinically, the development of KDM4A-AGT-PHB1 systematic grade demonstrates significant predictive value for immunotherapy response in pMMR CRC patients, a population currently lacking reliable biomarkers for immune checkpoint inhibitor therapy. These findings established tumor senescence induction as a viable immunomodulatory strategy and propose a dual-target therapeutic paradigm combining epigenetic reprogramming with immunotherapy.

Materials and methods

Patients and tissue samples

Human CRC samples were collected from patients received surgery in the department of colorectal surgery at the Sixth Affiliated Hospital of Sun Yat-sen University (Guangzhou, China). Clinical information was collected from the patient electronic medical records of our hospital. All clinical samples and data were carried out in accordance with the principles of the Declaration of Helsinki and were approved by the Institutional Review Committee of the Sixth Affiliated Hospital of Sun Yat-sen University (2025ZSLYEC–166).

Cell lines and cell cultures

The HEK293T, HCT8, SW480, Lovo, HT29, SW620, HCT116, DLD1, MCF-7 and PANC-1 were all obtained from the American Type Culture Collection (CRL-3216, CCL-244, CCL-228, CCL-229, HTB-38, CCL-227, CCL-247, CCL-221, HTB-22, CRL-1469). The mouse colon cancer cell line CT26 and MC38 were obtained from the American Type Culture Collection (CRL-2638) and the National Infrastructure of Cell-Line Resource (3101MOUSCSP5431), respectively. DMEM (Gibco, Thermo Fisher Scientific 11,960,077) supplemented with 10% FBS (Gibco, Thermo Fisher Scientific 10,082,147), 100 units mL⁻¹ penicillin (GIBCO), and 100 µg mL⁻¹ streptomycin (GIBCO) was used to culture all cells in a 5% CO₂ atmosphere. All cell lines were proven to be free of mycoplasma prior to the experiments.

Animal assays

For the xenograft tumor model, 1 × 10⁶ CT26 or MC38 cells transfected with indicated plasmids were inoculated into BALB‐C or C57BL‐6J mice (VitalRiver Laboratory Animal Technology), respectively. For the anti-PD1 treatment research, Mice received intraperitoneal injections of anti-PD1 (Bio X Cell, BE0146) or IgG (Bio X Cell, BE0083) every three days starting on day 6 of subcutaneous tumor transplantation. When subcutaneous tumors grow to a visible size, tumor volume was measured every two days and calculated according to the formula: V = L × S ² ×0.5 (V, volume; L, long diameter; S, short diameter). When tumor reached 2000 mm³ or there was a significant difference in tumor volume among groups, mice were sacrificed for next analysis of tumor volume and tumor weight. Tumor tissues were collected for further IHC analysis. For mouse survival analysis, the mouse was sacrificed when its own tumor volume reached 2,000 mm³ and survival time was calculated from the time after tumor inoculation. All of the animal experiments were approved by the Institutional Animal Care and Use Committee of Sun Yat-sen University and conformed to the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health (National Academies Press, 2011) in China.

Senescence CRC cells induction

For senescence CRC cell construction, 200 µM H₂O₂ was co-cultured with CRC cells (HCT116 and DLD1) for 2 h. Cell culture of senescent cells was performed according to the recommendations [48].

Senescence associated β-galactosidase (SA-GLB1/β-gal) assay

For tissue SA-GLB1/β-gal assay, fresh frozen sections were washed with PBS for 5 min, fixed with 1X fixative solution for at least 25 min and washed with PBS with three times. Then, Frozen sections were stained with the β‐galactosidase staining solution at pH 6.0 at 37°C overnight. For cells SA-GLB1/β-gal assay, CRC cells were grown on 12‐well plates were fixed with 1X fixative solution for 10 min and then washed with PBS three times. Subsequently, cells were stained with the β‐galactosidase staining solution at pH 6.0 at 37°C overnight. Images were obtained from Olympus microscope.

Immunohistochemistry

Formalin-fixed, paraffin-embedded (FFPE) tissue sections were used for Immunohistochemistry. FFPE tissue slides were dewaxed, rehydrated, and boiled in pH 9.0 EDTA antigen retrieval solution (Servicebio, G1218). Then, tissue sections were incubated at 4°C overnight with primary antibodies against CDKN1A/p21/Waf1/Cip1, CDKN2A/p16/INK4A, CD3, CD8, KDM4A, AGT and PHB1, then exposed to secondary antibody (ZS-bio, SAP-9100) for 20 min at room temperature. Diaminobenzidine solution (ZS-bio, ZLI-9018) was used for further DAB stain. Detailed staining scores with calculation methods have been described in previous study [12].

RNA extraction and qRT-PCR

Total RNA was extracted using TRizol Reagent (Invitrogen, 15596026CN), and cDNA was synthesized using Evo M-MLV RT Mix Kit (Accurate Biotechnology, AG11728). qPCR was performed using SYBR Green Pro Taq HS qPCR Kit (Accurate Biology, AG11759) with a LightCycler 96 (Roche, AXYPCR96LC480CNF). The primers used in this study are presented in Table S3. The 2^−∆∆CT method was applied to calculate relative expression.

RNA sequencing (RNA-seq) and data analysis

Total RNA was extracted from three dMMR and three pMMR colorectal carcer tissues of patients to analyze the transcriptomic profile changes. Besides, total RNA was obtained from KDM4A-overexpressing HCT8 cells and their corresponding control cells to evaluate the effects of KDM4A on the gene expression profile of CRC cells.

Above RNA was collected by Trizol reagent separately. The RNA quality was checked by Agilent 2200 and kept at − 80°C. The RNA with RIN (RNA integrity number) > 7.0 is acceptable for cDNA library construction. The cDNA libraries were constructed for each RNA sample using the VAHTS Universal V6 RNA-seq Library Prep Kit for Illumina (vazyme, Inc) according to the manufacturer’s instructions. The libraries were quality controlled with Agilent 2200 and sequenced by DNBSEQ-T7 on a 150 bp paired-end run. The clean reads were then aligned to human genome (GRCh38_Ensembl104) using the Hisat2. HTseq was used to get gene counts and RPKM method was used to determine the gene expression. DESeq2 algorithm was applied to filter the differentially expressed genes following these criteria: Fold change > 2 or < 0.5; p value < 0.05.

Western blot

Proteins were extracted with RIPA buffer containing protease and phosphatase inhibitor (Thermo Fisher Scientific 78,440). concentration of protein was detected with a bicinchoninic acid (BCA) kit (Beyotime Biotechnolog, P0012). protein was separated by SDS – polyacrylamide gels and transferred to PVDF membranes (Millipore, IPVH85R). specific primary antibodies against CDKN1A/p21/Waf1/Cip1, CDKN2A/p16/INK4A, KDM4A, H3K9me3, H3K36me3, Histone H3, AGT, PHB1, PHB2, PINK1, PRKN/Parkin, CGAS, STING1, NFKB, GAPDH, TUBB, TIMM23 and TOMM20 were diluted in TBST containing 5% skim milk powder and incubated with corresponding primary antibodies at 4°C overnight. After washed with PBST three times, Membranes incubated with horseradish peroxidase (HRP)-conjugated anti-rabbit or anti-mouse secondary antibody (Proteintech, SA00001–1) for 1 h at room temperature. Signals were detected and visualized using ECL (Meilun, MA0186). Details of antibodies used here are provided in Table 1.

Table 1.

The catalog numbers and dilutions of antibodies used in this study.

Antibody	Manufacturer	Catalog #(RRID)	Application
Anti-CDKN1A/p21/Waf1/Cip1	Cell Signaling Technology (CST), USA	2947S	WB (1:1000) IHC (1:50)
Anti-CDKN2A/p16/INK4A	Proteintech, China	10883–1-AP	WB (1:2000) IHC (1:200)
Anti-KDM4A	Abclonal, China	A7953	WB (1:1000) IHC (1:250)
Anti-H3K9me3	Abclonal, China	A22295	WB (1:500)
Anti-H3K9me3	Abcam, USA	Ab8898	ChIP (4 µg for 25 µg of chromatin)
Anti-H3K36me3	Abclonal, China	A20379	WB (1:500)
Anti-Histone H3	Abcam, USA	Ab1791	WB (1:2000)
Anti-AGT	Abcam, USA	Ab213705	WB (1:1000) IP (1:40)
Anti-AGT	Proteintech, China	11992–1-AP	IHC (1:200) IF (1:200)
Anti-PHB1	Abcam, USA	Ab75766	WB (1:10000) IP (1:40) IHC (1:250)
Anti-PHB2	Selleck, USA	F1345	WB (1:2000)
Anti-PINK1	Proteintech, China	23274–1-AP	WB (1:600)
Anti-PRKN/Parkin	Proteintech, China	66674–1-Ig	WB (1:4000) IF (1:400)
Anti-ubiqutin	Cell Signaling Technology (CST), USA	3936s	WB (1:1000)
Anti-GAPDH	Proteintech, China	10494–1-AP	WB (1:20000)
Anti-TUBB/β-tubulin	Proteintech, China	66240–1-Ig	WB (1:40000)
Anti-TIMM23	Proteintech, China	67535–1-Ig	WB (1:5000)
Anti-TOMM20	Abcam, USA	Ab186735	WB (1:2000) IF (1:250)
Anti-TRIM21	Proteintech, China	12108–1-AP	WB (1:6000)
Anti-CGAS	Proteintech, China	26416–1-AP	WB (1:4000)
Anti-STING1	Proteintech, China	19851–1-AP	WB (1:5000)
Anti-phospho-NFKB	CST, USA	3033S	WB (1:1000)

Plasmid construction and viral infection

For KDM4A, AGT, PHB1, and Kdm4a overexpression plasmids, the full-length ORF sequences of these genes were subcloned into the pSin-EF2-Pur vector (Addgene 116,877; deposited by Marcos Alcocer). For construct the short hairpin RNA (shRNA)-expressing plasmids targeting KDM4A, AGT, PHB1 and Agt, the target shRNA sequences were individually cloned into pLKO.1 vector. All constructs were verified by sequencing.

For stable lentivirus infection, HEK293T cells were transfected with the above plasmids, psPAX2 (Addgene 12,260; deposited by Didier Trono) and pMD2.G (Addgene 12,259; deposited by Didier Trono). The supernatant containing viral particles were collected and filtered through Millex‐GP Filter Unit (0.45-µm pore size; Millipore, SLHA025NB). Then, colorectal cancer cells were infected with lentivirus for 48 h, together with 10 mg mL⁻¹ polybrene (Bioship, BL628A) at 37°C. Infected cells were under puromycin (5 µg mL⁻¹) or G418 (100 µg mL⁻¹) selection for 5–7 d. The shRNA sequences are listed in Table S4.

Edu cell proliferation assay

BeyoClick™ EdU Cell Proliferation Kit with Alexa Fluor 488 (Beyotime Biotechnolog, C0071S) was used for EdU cell proliferation assay. About 2.5 × 10⁴ cells were seeded in 96-well plates and cultured for 24 h. Then, cells were incubated with EdU for 2 h before being fixed with 4% paraformaldehyde for 15 min. after three times wash with PBS, cells were permeabilized using 0.1% Triton X-100 (Servicebio, G1204) for 15 min, and then incubated with the click-reaction reagent for 30 min at room temperature in the dark. The Hoechst 33,342 reagent (1x) was used to mark cell nuclear. Images were acquired by an Olympus microscope.

Colony formation and cell viability assays

For colony formation assays, 500 cells were seeded in 6-well plates in triplicate and cultured for 10–14 days. The colonies were fixed and stained with 4% paraformaldehyde (Servicebio, G1101) and 0.1% crystal violet (Servicebio, G1014). The number of colonies were counted and analyzed.

Cell viability assay was performed following the Cell Counting Kit‐8 (CCK8) assay (Dojindo Laboratories, CK04) as described in the manufacturer’s manual. Briefly, stable transfected cells were seeded into 96-well plates (3 × 10³ cells/well) and incubated for 96 h. Then, 10 μL of CCK-8 reagent was added and incubated at 37°C for 2 h. The absorbance of the cells at 450 nm was then measured using a spectrophotometer (Molecular Devices).

Enzyme-linked immunosorbent assay (ELISA)

For human IL6 and IL8 levels, CRC cellular supernatant was collected for concentrations detection. For mouse CCL5 and CXCL10 levels, xenograft tumor mouse serum was collected for concentrations detection. Elisa assays were performed according to the manufacturer’s instructions (elabscience).

Chromatin immunoprecipitation (ChIP) assay

The ChIP assay was performed using a SimpleChIP® Plus Sonication Chromatin IP Kit (Cell Signaling Technology, 56383S) according to the manufacturer’s instructions. Briefly, after cross-linking and chromatin sonication from HCT8 cells, sonicated chromatin was incubated with ChIP-Grade Protein G Magnetic Beads and with anti-H3K9me3 at 4°C overnight. Enrichment of target DNA fragments was evaluated by qRT-PCR using the specific primers listed in table S3.

Dual luciferase reporter assay

Dual luciferase reporter assay was performed in HCT8, SW480, HCT116 and DLD1 cells with established AGT promoter reporter (MiaoLingBio, P64494). After 48 h, Firefly and Renilla luciferase activities were measured by the Dual Luciferase Reporter Gene Assay Kit (Abbkine, KTA8010). The relative firefly luciferase activities were measured using Renilla luciferase activities as an internal control.

Co-immunoprecipitation (CO-IP) and mass spectrometry (MS) analysis

Briefly, cells were lysed with IP lysis buffer (Beyotime Biotechnolog, P0013) with protease inhibitor cocktail for 30 min at 4°C. Then the cell lysates were centrifuged at 13000rpm for 20 min at 4°C. The supernatants were collected and incubated with antibody with specific primary antibodies against AGT and PHB1 followed by the addition of Protein A/G magnetic beads (MedChemExpress, HY-K0202). After incubation at 4°C for 12 h, the beads were washed three times with IP lysis buffer and boiled for 10 min in 2×sample loading buffer (Beyotime Biotechnolog, P0015L). The eluted samples were used for MS identification and western blot analysis. Details of antibodies used here were provided in Table 1.

For MS analysis, the gels from the co-IP assays were digested with trypsin. Peptides were dissolved in 0.1% FA and 2% ACN, directly loaded onto a reversed-phase analytical column. The gradient was comprised of an increase from 5% to 50% solvent B (0.1% FA in 80% ACN) over 40 min, and climbing to 90% in 5 min, then holding at 90% for the 5 min. MASCOT (Matrix Science) software was used for protein identification through searching Uniprot Aedis aegypti. The overall protein score was calculated by MASCOT for each protein match, with higher scores indicating more confident match.

Mouse T cells isolations and activation

T cells were isolated from spleens of C57BL/6 mouse. Briefly, lymphocyte suspensions were obtained by grinding mouse spleens with lymphocyte separation solution. The lymphocyte suspension was collected, mixed with serum-free DMEM and centrifuged further at 4°C for 30 min at 300 g. The tunica albuginea layers were collected, washed with precooled serum-free DMEM, and centrifuged at 100 g for 10 min at 4°C. Finally, the supernatant was discarded and the naive T cells were located at the bottom of the tube. For T-cell activation, 50 U/mL IL2 (MedChemExpress, HY-P7077), and 200 mL mouse CD3/CD28 dynbeads (Invitrogen, 11452D) were added for stimulation, and functional experiments were performed 3 days later.

T cell migration assays in vitro

T cell vitro migration assays were performed in a transwell system with a polycarbonate membrane with an 8-mm pore size. Activated CD8⁺ T cells were washed twice and seeded in the top of the chamber with serum-free medium. Then conditioned media from MC38 or CT26 cells transfected with indicated plasmids was added to the bottom of the chamber. After the coculture for 24 hours, the cells in the bottom media were collected and counted.

Mitochondria isolation assay

The mitochondria were extracted from CRC cells (HCT8 and SW480) following the protocol of the mitochondria isolation kit (Beyotime Biotechnolog, C3601). First, CRC cells were washed three times with PBS, detached from petri dish with trypsin digestion, and centrifuged at 600 g for 5 min at 4°C. Secondly, after removing supernatant, the collected cells were resuspended with 1 ml of cell lysis reagent (Beyotime Biotechnolog, C3601–1) and then subjected to an ice bath for 15 min. Third, lysed cells were homogenized using a glass homogenizer for 25 cycles. Fourth, the homogenized cells were centrifuged at 600 ×g for 10 min at 4°C. Then, the supernatant was transferred to a centrifuge tube and centrifuged at 11,000 g for 10 min at 4°C. Last, the cell pellet collected was used for mitochondria protein analysis. The supernatant was collected after another centrifugation at 12,000 g for 10 min at 4°C. The supernatant was used for analysis of cytosolic proteins without mitochondria.

Immunofluorescence (IF)

Cells were fixed in 4% paraformaldehyde, permeabilized with 0.5% Triton X-100 (Servicebio, G1204), and then blocked with 5% BSA (Biosharp, BL2182A) for 1 h at room temperature. Next, the cells were incubated with specific primary antibodies against AGT, PHB1, TOMM20 and PRKN/Parkin overnight at 4°C, followed by incubation with Alexa Fluor 488-(Abcam, ab150077) or 647-conjugated secondary antibodies (Abcam, ab150115) at room temperature for 1 h. DAPI was used to stain the nuclei, and images were obtained using a confocal microscope (Zeiss). Quantification of colocalization immunofluorescence was analyzed by ImageJ. For dsDNA staining, cells were fixed, permeabilized, and blocked, incubated with TOMM20 overnight at 4°C, washed three times with PBS, and followed by incubation with Alexa Fluor 647-conjugated secondary antibodies at room temperature for 1 h. After washing three times, cells were stained with DNA dye Draq5 (1/200) for 5 min to label dsDNA. images were obtained using a confocal microscope (Zeiss). Details of antibodies used here were provided in Table 1.

Detection of mt-Keima

HCT8 and SW480 cells were seeded in 6-well plates and transfected at 70% confluency with 3 μg per well mitochondria-targeted Keima (mt-Keima) plasmid (MiaoLingBio, P41642) using Lipofectamine 3000 (Invitrogen, L3000015). Cells were maintained in antibiotic-free medium for 48 h post-transfection.

mKeima fluorescence density was observed using a Zeiss LSM780 confocal microscope. The mKeima protein can be used to determine the binding of mitochondria to lysosomes as an indicator to evaluate mitophagic flux. The sensitivity of mKeima to pH can be examined whether the mitochondria is in acidic lysosome (excitation: 561 nm, red) or neutral compartments (excitation: 430 nm, green). The relative activity of mKeima was calculated as the ratio of red dots (561 nm) to green dots (430 nm).

Co-localization of mitochondria and lysosomes

Cells were stained with MitoTracker (Beyotime Biotechnolog, C1048) and LysoTracker (Beyotime Biotechnolog, C1046) according to the manufacturer’s instructions, and then pictures were taken using the confocal microscope (Zeiss). Mitophagy activity was quantification by the number of co-localization dots (yellow dots).

Transmission electron microscopy

Cells transfected with indicated plasmids were co-cultured with 10 μM CCCP for 2 h. Then, cells were harvested, washed twice with PBS and fixed with electron microscope fixative (Servicebio, G1102) for 30 min. The cells were then embedded in epoxy resin and dehydrated using a series of ethanol concentrations. Finally, the images were observed using a transmission electron microscope (HT7700, Tokyo, Japan). The number of mitophagosome was evaluate for mitophagy activity.

Cycloheximide (CHX) chase assays

AGT overexpression or PHB1 knockdown cells were used to perform cycloheximide (CHX) chase assays. Briefly, cells were plated in 6-well plates, followed by treatment with 50 μg/mL CHX for indicated time. Cells were harvested at 0, 2, 4, 6, 8 hours for protein extraction, and the protein expression of PHB1 or PHB2 was analyzed by western blot.

MG132 treatment

Cells were treated with 10 μM MG132 (Sigma-Aldrich, M8699) for 12 h, and the total protein was extracted. Western blot assays evaluated the ubiquitination amounts of PHB1 protein.

Detection of mitochondrial DNA versus nuclear DNA

DNeasy Blood and Tissue Kit (Qiagen 69,504) was used for isolation of total DNA (genomic and mitochondrial) in HCT8 and SW480 cells transfected with indicated plasmids. The amount of mitochondrial DNA and nuclear DNA was measured by qRT-PCR assays with D-Loop and RNA18S primers, respectively. The primers used in this study are presented in Table S3.

Statistical analysis

Data are presented as mean ± SD from at least three independent experiments. The statistical significance was determined by two-tailed Student’s t test for two groups and one- way analysis of variance (ANOVA; followed by Tukey’s post hoc test) for multiple groups. The Kaplan-Meier method and the log- rank test were used for Survival analysis. Spearman’s correlation analysis was used to evaluate the correlation among the expression of KDM4A, CDKN2A/p16, AGT, PHB1 and the density of CD8+ T-lymphocyte. Statistical analyses were conducted using Statistical Product and Service Solutions (SPSS) v.22.0 (International Business Machines Corporation) and GraphPad Prism 8.0 (GraphPad) software. A p value < 0.05 was considered statistically significant.

Correction Statement

This article was originally published with errors, which have now been corrected in the online version. Please see Correction (https://doi.org/10.1080/15548627.2025.2575534)

Funding Statement

This work was supported by national natural Science Foundation of China [82472930 to L.K, 82303060 to Z.L. and 82300908 to L.X.], Guangdong Basic and applied Basic research Foundation [2022A1515110785 to Z.L., 2023A1515010525 to Z.L., 2023A1515012455 to L.X. and 2025A1515012589 to L.X.], China Postdoctoral Science Foundation [2023M734027 to Z.L. and 2022M723647 to L.X.], Fundamental research Funds for the central Universities, Sun Yat-sen University [23qnpy149 to Z.L.], Science and Technology Projects in Guangzhou [202206010062 to L.K.], Shenzhen “Sanming Project” research [lc202002 to L.K.], and national Key clinical discipline.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data and materials availability

All data associated with this study are present in the paper or the Supplementary Materials. RNA-seq data and ChIP-seq data have been deposited in the Gene expression ombibus (GEO) database (www.ncbi.nlm.nih.gov/sra), with accession number GSE290475 and GSE290175. MS data are available at ProteomeXchange (www. proteomexchange.org/) with identifier PXd047640.

Supplementary material

Supplemental data for this article can be accessed online at https://doi.org/10.1080/15548627.2025.2551680