3.1 ICI analysis

Senescent cells are characterized by a unique gene set that also predicts senescence-related processes across various tissues [27]. Subsequently, 328 ECM-associated genes were identified based on gene ontology terms and are listed in Table S3. Lastly, we identified 19 ECM-SRGs (MMP2, SERPINE1, ICAM3, TNF, MMP13, PECAM1, MMP3, LCP1, SPP1, MMP14, TIMP2, TNFRSF11B, FGF2, ITGA2, ICAM1, MMP10, MMP1, MMP12, and MMP9) (Fig. S1). We established a significant correlation between GSVA enrichment scores (ES) of ECM-SRGs and ICI in 33 different cancers (P-value and FDR ≤ 0.05, Fig. 1). We also demonstrated a significant correlation between 19 ECM-SRGs and ICI in 33 cancers (Table S4). Moreover, the results showed a positive correlation between GSVA-ES of ECM-SRGs and Macrophage, DCs, iTreg, cytotoxic and exhausted T cell, Th1, Tfh, and NK cells. Additionally, we observed a negative correlation between GSVA-ES of ECM-SRGs and naive CD8 and CD4 T cell, neutrophil, and B cells in pan-cancer (P-value and FDR ≤ 0.05, Fig. 1). Conversely, in pan-cancer, GSVA-ES of ECM-SRGs were positively correlated with the InfiltrationScore of patients (P-value and FDR ≤ 0.05, Fig. 1). Thus, aberrant ECM-SRGs expression regulates ICI in patients, thereby indicating a significant role of ECM-SRGs in cancer progression.

Fig. 1

Correlation between gene set expression (GSVA) enrichment scores of extracellular-matrix-senescence-related genes and immune cell infiltration in 33 cancers. *P value ≤ 0.05; # FDR ≤ 0.05

Alterations in the immune microenvironment could be involved in tumorigenesis, cancer progression, diagnosis, prognosis, and therapeutic of patients. Further, a significant difference in ICI in patients harboring SNVs in ECM-SRGs in UCS, UCEC, THYM, THCA, STAD, SKCM, SARC, READ, PRAD, PCPG, PAAD, OV, LUAD, LIHC, LGG, KIRP, KIRC, HNSC, GBM, DLBC, COAD, CESC, BRCA, and BLCA (FDR ≤ 0.05, Fig. 2A and Table S5). A significant correlation between CNV amplification of ECM-SRGs and ICI in UVM, UCEC, THYM, THCA, STAD, SARC, READ, PCPG, PAAD, MESO, LUSC, LIHC, LGG, KIRP, KICH, HNSC, GBM, ESCA, COAD, CESC, and BRCA was observed. Moreover, the results revealed a significant correlation between CNV deletions of ECM-SRGs and ICI in THCA, SARC, PRAD, LUSC, LUAD, LIHC, LGG, KIRC, KICH, HNSC, ESCA, DLBC, CHOL, BRCA, and BLCA (FDR ≤ 0.05, Fig. 2B and Table S6). Moreover, a significant correlation between the methylation of ECM-SRGs and ICI in ECM-SRGs (Table S7).

Fig. 2

Difference of immune cell infiltration in patients with (A) single nucleotide variation and (B) copy number variation (CNV) amplification in extracellular-matrix-senescence-related genes. WT wild-type

These findings suggest that abnormalities in ECM-SRGs within the immune microenvironment play a role in cancer initiation, progression, diagnosis, prognosis, and therapeutic outcomes.

3.2 Analysis of ECM-SRGs expression, cancer subtypes, and stages

We assessed differences in ECM-SRGs expression among cancer patients based on the GSVA score. Compared to normal tissue, patients with 13 solid tumors displayed significant differences in the expression of ECM-SRGs (FDR ≤ 0.05, Fig. 3A and Table S8). However, there were no significant differences in the expression of ECM-SRGs among patients with ESCA.

Fig. 3

Gene set expression analysis and gene-set enrichment analysis (GSEA) of extracellular-matrix-senescence-related genes (ECM-SRGs). A The mRNA differences between normal samples and tumor samples. B Enrichment score of ECM-SRGs. C Enrichment plot of ECM-SRGs in ESCA. D Enrichment plot of ECM -SRGs in HNSC

The bar plot displays the GSVA scores of ECM-SRGs compared to all 20,000 + genes (background) in various cancers (Fig. 3B and Table S9). The results indicated elevated expression of most ECM-SRGs in patients with ESCA (Fig. 3C), HNSC (Fig. 3D), COAD, STAD, THCA, BRCA, and KIRP (P < 0.05).

We conducted a screening for clinically relevant genes influencing cancer subtypes. The results revealed significant differences in the expression of ECM-SRGs in patients with LUAD, BRCA, LUSC, GBM, KIRC, STAD, HNSC, and BLCA compared to other genes (FDR ≤ 0.05, Fig. 4A and Table S10). Additionally, we noted variations in ICAM3 expression among patients with KIRC subtypes (Fig. 4B) and FGF2 expression among patients with BRCA subtypes (Fig. 4C).

Fig. 4

Genes set expression and subtype and stage analysis of extracellular-matrix-senescence-related genes (ECM-SRGs). A Subtype difference between high and low gene expression of ECM-SRGs in cancers. B ICAM3 mRNA expression in subtype of KIRC. C FGF2 mRNA expression in subtype of BRCA. D Expression difference between stages. E MMP13 mRNA expression in pathologic stage of THCA. F TIMP2 mRNA expression in pathologic stage of BLCA

We conducted a screening for clinically relevant genes influencing cancer stages. The results indicated significant differences in the expression of ECM-SRGs among patients in different stages (from stage I to IV) of THCA, BLCA, KIRC, KIRP, SKCM, and TGCT (FDR ≤ 0.05, Fig. 4D and Table S11). Such as MMP13 expression in pathologic stages of THCA (Fig. 4E) and TIMP2 expression in pathologic stages of BLCA (Fig. 4F).

Furthermore, survival outcomes varied between groups with high and low ECM-SRG expression. Significant correlations were observed between ECM-SRGs and the survival (DFI, DSS, OS, and PFS) of patients (Cox P < 0.05, Fig. 5 and Table S12). Then, a significant correlation between ECM-SRGs and the survival (OS and DFS) of patients was also identified in GEPIA2 (P < 0.05, Fig. S2). These findings suggest that cancer may be affected by abnormal ECM-SRG expression.

Fig. 5

Survival difference between high and low gene expression in 33 cancers of extracellular-matrix-senescence-related genes. Red points represents worse survival of the high expression group, light blue points represent worse survival of the low expression group. The size of the point represents the statistical significance, where the larger the dot size, the higher the statistical significance

3.3 CNVs in ECM-SRGs

To identify CNVs in ECM-SRGs, we analyzed patient CNV data from TCGA. The CNV distribution pie chart indicated that the most commonly observed CNVs in patients were heterozygous amplifications and deletions (Fig. 6A and Table S13). Furthermore, genomic alterations were predominantly amplifications and deep deletions in cBioPortal (Fig. S3). The bubble plot will be filled with bubbles because heterozygous amplifications and deletions of ECM-SRGs were frequently observed in various cancers, as indicated by CNV percentage analyses (P < 0.05, Fig. 6B). In most cancers, homozygous evaluation of ECM-SRGs showed both amplifications and deletions (P < 0.05, Fig. 6C). Additionally, there was a positive association between ECM-SRG expression and CNV in patients with OV, HNSC, SKCM, BRCA, LUAD, SARC, LUSC, LGG, CESC, and others. However, the results revealed a negative correlation between MMP2 expression and CNV in KIRP patients and between ICAM1 expression and CNV in ACC patients (P < 0.0001, Fig. 7A and Table S14). These results suggested that abnormal ECM-SRGs expression resulting from CNVs may be involved in the initiation and progression of cancer. Survival analysis (DFI, DSS, OS, and PFS) revealed a connection between high CNVs in ECM-SRGs and poor patient survival in various malignancies (P < 0.05, Fig. 7B and Table S15).

Fig. 6

copy number variation (CNV) distribution in 33 cancers of extracellular-matrix-senescence-related genes. A CNV distribution. CNV pie chart showing the combined heterozygous/homozygous CNV of each gene in each cancer. A pie chart representing the proportion of different types of CNV of one gene in one cancer, and different colors represent different types of CNV. Hete Amp heterozygous amplification; Hete Del heterozygous deletion; Homo Amp homozygous amplification; Homo Del homozygous deletion; None no CNV. CNV profile showing the percentage of heterozygous CNVs (B) and Homozygous CNVs (C), including the percentage of amplification and deletion for inflammatory aging clock-related genes in 33 cancers. Only genes with > 1% CNV in a given cancer are shown as a point on the Figure

Fig. 7

Copy number variation (CNV) correlation with mRNA expression and survival difference between CNV groups in 33 cancers of extracellular-matrix-senescence-related genes. A CNV correlation with mRNA expression. The association between paired mRNA expression and CNV percentage in samples was based on a Spearman’s product moment correlation coefficient. The size of the point represents the statistical significance, where the bigger the dot size, the higher the statistical significance. FDR, false discovery rate. B Survival difference between CNV groups. Red points represents worse survival of the hyper-CNV group, light blue points represent worse survival of the hypo-CNV group. The size of the point represents the statistical significance, where the larger the dot size, the higher the statistical significance

3.4 Analysis of methylation levels of ECM-SRGs

We investigated epigenetic regulation by examining the methylation status of ECM-SRGs. The methylation status of ECM-SRGs exhibited significant heterogeneity among patients (Fig. 8A). The findings indicated a prevalence of hypomethylation in ECM-SRGs among patients with various cancers (KIRC, COAD, LIHC, UCEC, THCA, BRCA, etc.), and hypermethylation of TNFRSF11B among patients with PRAD, ESCA, LUSC, HNSC, BRCA, and THCA (FDR ≤ 0.05, Fig. 8A and Table S16). The relationship between gene expression and methylation status was subsequently investigated. The results revealed a negative correlation between methylation and the expression of most ECM-SRGs, especially LCP1 expression, in patients with various malignancies (SKCM, THYM, THCA, KIRP, LUAD, BRCA, LIHC, LGG, etc.). However, a positive correlation was observed between MMP2 methylation and expression in patients with COAD, READ, BLCA, and ITGA2 methylation and expression in patients with TGCT (FDR ≤ 0.05, Fig. 8B and Table S17). Survival analysis (DFI, DSS, OS, and PFS) revealed a correlation between ECM-SRGs hypomethylation and poor patient survival in various malignancies (Cox P < 0.05, Fig. 9 and Table S18).

Fig. 8

Methylation of extracellular-matrix-senescence-related genes (ECM-SRGs). A Differential methylation in ECM-SRGs between tumor and normal samples in each cancer. Blue points represent decreased methylation in tumors and red points represent increased methylation in tumors, where the darker the color, the larger the difference of methylation level. B Correlation between methylation and mRNA expression. Blue points represent a negative correlation and red points represent a positive correlation, where the darker of color, the higher the correlation. All the FDR of gene and cancer types were less than 0.05 in the Fig. FDR, false discovery rate

Fig. 9

Survival difference between samples with extracellular-matrix-senescence-related genes with high and low methylation. Red points represents worse survival of the hypermethylation group, light blue points represent worse survival of the hypomethylation group. The size of the point represents the statistical significance, where the larger the dot size, the higher the statistical significance

3.5 ECM-SRG somatic mutations

To assess the frequency of gene variants in each cancer subtype, we investigated SNPs in ECM-SRGs. LUAD, UCEC, COAD, STAD, SKCM, and LUSC patients exhibited an SNV frequency in ECM-SRGs ranging from 1 to 39%, as illustrated in Fig. 10A and Table S19. The frequency of SNVs in the regulatory genes was 80.28% (867 out of 1080 patients, Fig. 10B). Furthermore, missense mutations were the predominant type of SNPs in patients. The proportion of SNVs in the top 10 genes with mutations, ITGA2, MMP9, MMP2, MMP13, MMP10, MMP3, LCP1, FRSF11B, MMP1, and MMP14, was 18, 14, 13, 12, 11, 11, 10, 10, 8, and 8%, respectively. Patients with LUAD, UCEC, SKCM, and LUSC exhibited a higher frequency of SNVs in the regulatory genes (Fig. 10B). For patients with certain malignancies, survival analysis (DFI, DSS, OS, and PFS) demonstrated a significant difference in SNVs between mutant and WT ECM-SRGs (Cox P < 0.05, Fig. S4 and Table S20).

Fig. 10

Single nucleotide variation (SNV) frequency and variant types of extracellular-matrix-senescence-related genes (ECM-SRGs). A Mutation frequency of ECM-SRGs. Numbers represent the number of samples that have the corresponding mutated gene for a given cancer. ‘0’ indicates that there was no mutation in the gene coding region, and no number indicates there was no mutation in any region of the gene. B SNV oncoplot. An oncoplot showing the mutation distribution of ECM-SRGs and a classification of SNV types

3.6 Pathway activity analysis

The pathway activity analysis revealed a notable contribution of ECM-SRGs to cancer-related pathways, encompassing the cell cycle, apoptosis, PI3K/AKT, RAS/MAPK, RTK, and TSC/mTOR signaling pathways, EMT, hormone AR, and ER, as well as the response to DNA damage (Fig. 11A). These ECM-SRGs primarily played roles in activating EMT, apoptosis, and the RAS/MAPK signaling pathways, while inhibiting the cell cycle, hormone AR, and the response to DNA damage signaling pathways (P < 0.05, Fig. 11A and Table S21). Subsequently, we assessed the pathway activity of ECM-SRGs based on the GSVA score, and the findings indicated the participation of ECM-SRGs in the activation of the EMT, apoptosis, and RAS/MAPK signaling pathways, as well as the suppression of the cell cycle, hormone AR, PI3K/AKT, and response to DNA damage signaling pathways (P < 0.05, Fig. 11B). Furthermore, we conducted an analysis of the correlation between ECM-SRGs scores and 14 functional states in various tumors using CancerSEA. The results demonstrated the engagement of ECM-SRGs in activating EMT, apoptosis, angiogenesis, hypoxia, inflammation, and metastasis signaling pathways, while suppressing cell cycle, DNA repair, and DNA damage signaling pathways (P < 0.05, Fig. S5). Thus, ECM-SRGs could modulate pathways related to cancers.

Fig. 11

The cancer related pathway activity between extracellular-matrix-senescence-related genes (ECM-SRGs). A The combined percentage of the effect of ECM-SRGs on pathway activity. B Pathway activity of ECM-SRGs based on the GSVA score

3.7 Analysis of drug sensitivity

Patient sensitivity to chemotherapy and targeted therapy can be influenced by genomic abnormalities. Thus, we explored the role of ECM-SRGs in mediating patient responses to chemotherapy and targeted therapy. Initially, we integrated data on gene expression in cancer cells and drug sensitivity from the GDSC. Subsequently, through Spearman's correlation analysis, we identified that the expression of TNF, LCP1, and ICAM3 was negatively correlated with drug sensitivity to compounds such as Navitoclax, AR-42, CAY10603, CP466722, I-BET-762, KIN001-102, Tubastatin A, GSK1070916, GSK690693, KIN001-260, NG-25, NPK76-II-72-1, PIK-93, TPCA-1, Vorinostat, 5-Fluorouracil, BX-912, WZ3105, XMD13-2, BMS345541, CUDC-101, Methotrexate, PHA-793887, TAK-715, THZ-2-102-1, ZSTK474, and AT-7519 (with a negative correlation with IC50 values). Conversely, resistance to Navitoclax, AR-42, CAY10603, CP466722, I-BET-762, KIN001-102, Tubastatin A, GSK1070916, GSK690693, KIN001-260, NG-25, NPK76-II-72-1, PIK-93, TPCA-1, Vorinostat, 5-Fluorouracil, BX-912, WZ3105, XMD13-2, BMS345541, CUDC-101, Methotrexate, PHA-793887, TAK-715, THZ-2-102-1, ZSTK474, and AT-7519 was associated with MMP14, ITGA2, FGF2, MMP3, MMP1, and MMP2 expression (positively correlated with IC50 values) (FDR ≤ 0.05, Fig. 12A and Table S22).

Fig. 12

Correlation between (A) GDSC and (B) CTRP drug sensitivity and mRNA expression in pan-cancer. Spearman’s correlation represents how the gene expression correlates with a drug. A positive correlation means that a gene with high expression was resistant to a drug, and a negative correlation means that a gene with high expression was sensitive to a drug. FDR false discovery rate, GDSC Genomics of Drug Sensitivity in Cancer, CTRP Cancer Therapeutics Response Portal

Additionally, we integrated gene expression data from CTRP for cancer cell lines and their drug sensitivity. Through Spearman's correlation analysis, it was observed that drug sensitivity to Vorinostat, necrosulfonamide, BI-2536, GSK461364, GW-405833, ML311, PRIMA-1, belinostat, panobinostat, piperlongumine, CR-1-31B, Compound 23 citrate, I-BET151, ISOX, apicidin, 138A-II-SR, 743921-SB, 12-PX, DI-PL, 1-IN-LRRK2, tacedinaline, triazolothiadiazine, vincristine, BRD-K66453893, COL-3, ciclopirox, cytarabine hydrochloride, decitabine, narciclasine, and parbendazole was negatively correlated with LCP1, PECAM1, TNF, MMP9, and ICAM3 expression (with a negative correlation with IC50 values). In contrast, resistance to drugs such as Vorinostat, necrosulfonamide, BI-2536, GSK461364, GW-405833, ML311, PRIMA-1, belinostat, panobinostat, piperlongumine, CR-1-31B, Compound 23 citrate, I-BET151, ISOX, LRRK2-IN-1, apicidin, 138A-II-SR, 743921-SB, 12-PX, DI-PL, tacedinaline, triazolothiadiazine, vincristine, BRD-K66453893, COL-3, narciclasine, decitabine, cytarabine hydrochloride, ciclopirox, and parbendazole was positively correlated with TNFRSF11B, TIMP2, SPP1, SERPINE1, MMP14, MMP1, ITGA2, FGF2, and MMP2 expression (with a positive correlation with IC50 values) (FDR ≤ 0.05, Fig. 12B and Table S23).

These findings suggest that aberrant expression of ECM-SRGs may serve as a mediator of resistance to both chemotherapy and targeted therapy.