Skip to main content
NCBI home page
Discover Oncology logoLink to Discover Oncology
. 2025 Mar 20;16:363. doi: 10.1007/s12672-025-02127-9

Classification of lung adenocarcinoma based on senescence-related genes identifies a cluster with immunotherapy resistance and poor prognosis

Xinrui Gao 1,#, Xiang Shen 2,#, Shasha Huang 3,✉, Shangke Huang 4,✉
PMCID: PMC11925839  PMID: 40111675

Abstract

Lung adenocarcinoma is one of the major contributors to cancer-related mortality, with immunotherapy emerging as a key treatment. However, many patients exhibit resistance to immune checkpoint inhibitors. Cellular senescence has been linked to tumor progression and drug resistance, influencing the tumor microenvironment. This study applied consensus clustering to classify lung adenocarcinoma patients into two clusters based on senescence-related gene expression, revealing differing immune characteristics. One of the identified clusters exhibited immunosuppressive characteristics and showed resistance to immunotherapy. A senescence-related risk score was developed using machine learning to predict immunotherapy response and prognosis. High senescence-related risk score correlated with poorer survival and increased immunotherapy resistance across multiple cancer types. The senescence-related risk score model showed robust predictive ability in both the training and validation cohorts. These findings suggest a link between senescence and immunotherapy resistance, and further investigation into their relationship could reveal new perspectives for cancer treatment.

Supplementary Information

The online version contains supplementary material available at 10.1007/s12672-025-02127-9.

Keywords: Senescence, Lung adenocarcinoma, Immunotherapy

Introduction

Lung cancer is the most frequently diagnosed malignant tumor, with over 2 million new cases worldwide, accounting for about 12% of all cancer diagnoses globally. It remains the leading cause of cancer-related deaths, with an estimated 1.8 million deaths attributed to lung cancer, representing 18.7% of all cancer deaths [1]. Lung adenocarcinoma (LUAD) represents the most prevalent subtype of lung cancer [2]. Current treatment options for LUAD include surgical resection, chemotherapy, radiotherapy, targeted therapy, and immunotherapy [3]. Immunotherapy, while considered a breakthrough in cancer treatment, does not work for all patients, with a substantial portion experiencing resistance. Based on several cohorts using immune checkpoint inhibitor (ICI) monotherapy for lung cancer, more than 70% of lung cancer patients exhibit resistance to ICI monotherapy [4–8]. Tumor immunotherapy resistance arises from both intrinsic and extrinsic mechanisms. Intrinsic mechanisms enable tumor cells to escape detection by T cells, which can result from insufficient tumor antigen expression, defective antigen presentation processes, or a lack of human leukocyte antigen (HLA). Additionally, these cells may constitutively express immune checkpoint ligands like PD-L1 or undergo alterations in immune related signaling pathways. Extrinsic resistance arises from the formation of an immunosuppressive tumor microenvironment, marked by the buildup of cells such as regulatory T cells (Tregs), tumor-associated macrophages (TAMs), and myeloid-derived suppressor cells (MDSCs) [9–11]. Overcoming immunotherapy resistance remains a significant challenge in cancer treatment.

Cellular senescence is defined by four interrelated characteristics: permanent cell-cycle arrest, the formation of a senescence-associated secretory phenotype (SASP), macromolecular damage, and metabolic changes [12, 13]. Cellular senescence can be triggered by oncogene activation, as well as by cancer treatments like chemotherapy and radiotherapy [14, 15]. As senescent cells accumulate, they undergo changes in morphology and metabolism, adopting a SASP, which is characterized by the release of a wide range of molecules that can remodel the microenvironment [16]. There is mounting evidence that cellular senescence promotes tumor progression, increases invasiveness and metastasis, and contributes to drug resistance [17–20].

Cellular senescence plays a crucial role in shaping and influencing the tumor microenvironment, with significant impacts on tumor immune responses. Senescent cells can secrete CCL2, which attracts CCR2+ myeloid cells, fostering the growth of hepatocellular carcinoma and inhibiting the tumor-killing activity of natural killer (NK) cells [21]. In thyroid tumors, senescent cells induced by HrasG12V promote the production of PGE2, which induces macrophages to polarize toward the M2 phenotype, leading to immunosuppression [22]. In a similar manner, p53-deficient stellate cells produce a SASP that promotes M2 macrophage polarization, which in turn accelerates tumor growth [23]. In colorectal cancer, senescence driven by the METTL3/CDKN2B axis also promotes M2 polarization and correlates with both senescence and cancer progression [24]. Moreover, senescent fibroblasts induced by p27 contribute to tumor progression by recruiting MDSCs and Tregs through IL-6 signaling, which suppresses cytotoxic T cell responses [25]. Tumor cells are also capable of triggering senescence in normal T cells, thereby facilitating immune evasion by hindering antitumor immune responses and promoting the development of Tregs [26, 27]. These senescent T cells accumulate in the immunosuppressive tumor microenvironment, where they undermine immune responses by inhibiting effector T cells or inducing additional Tregs. Research has demonstrated that the presence of senescent cells can lead to the failure of immunotherapy, and that the use of senolytics, which eliminate senescent cells, can enhance the efficacy of immunotherapy [28].

Currently, the role of cellular senescence in lung cancer immunotherapy remains incompletely understood. Exploring the relationship between senescence and cancer, and constructing senescence-related clinical prediction models, may help identify populations that benefit from immunotherapy and uncover potential mechanisms underlying resistance. Several studies have used senescence-related gene signatures to explore their relationships with immune-related molecules and the immune microenvironment, as well as to develop predictive models for patient prognosis. However, these studies typically focus on analyzing the immune microenvironment without assessing immunotherapy response, or when they do analyze immunotherapy response, they fail to validate their findings using datasets from patients who have received ICIs [29–32]. Whether senescence-related gene signatures can predict immunotherapy efficacy in clinical models remains unknown.

In our study, we applied consensus clustering to classify LUAD tumors into two distinct clusters based on the expression of senescence-related genes. These clusters exhibited different immune characteristics within the tumor microenvironment, with one showing an immunosuppressive environment and resistance to immunotherapy. Through machine learning, we developed of a senescence related risk score (SRRS). This SRRS effectively predicts both immunotherapy response and prognosis in LUAD, providing valuable insights for clinical decision-making.

Materials and methods

Data sets collection

Information about the datasets used in the study is presented in Supplementary Table 1. RNA-seq, genomic mutation, and clinical data for LUAD samples were obtained from The Cancer Genome Atlas (TCGA) (https://portal.gdc.cancer.gov/). Microarray data for LUAD were downloaded from the Gene Expression Omnibus (GEO) under accession number GSE72094 [33]. These two cohorts were used to construct senescence-related clusters.

Transcriptomic and clinical data for urothelial cancer patients treated with anti-PD-L1 were accessed via the ‘IMvigor210CoreBiologies’ R package [34]. RNA-seq data for NSCLC patients receiving anti-PD-1 or anti-PD-L1 therapy were sourced from the GEO database (accession number GSE135222) [35]. Data for melanoma patients treated with anti-PD-1 were obtained from a large melanoma genome sequencing project (MGSP) [36]; we used the preprocessed data provided in the published article rather than the raw data. These three cohorts were used to validate the predictive efficacy of the SRRS for immunotherapy response.

Senescence-related and hallmark gene sets were sourced from the Molecular Signatures Database (MSigDB) (https://www.gsea-msigdb.org/gsea/msigdb/) or from previously published articles [37–40].

Gene set variation analysis (GSVA)

The enrichment scores for the gene sets of each sample were calculated using the ‘GSVA’ R package [41], with the analysis method set to ‘ssgsea’. The other parameters were configured as follows: mx.diff = TRUE, min.sz = 5, max.sz = 500, and abs.ranking = FALSE. Differences in GSVA scores were assessed using the ‘limma’ R package [42], and a threshold of |t|> 4 was used to indicate significant differences.

Consensus clustering

To identify distinct LUAD clusters, we performed consensus clustering using gene expression data from the SenMayo gene set with the ‘ConsensusClusterPlus’ R package [43]. The clustering parameters were as follows: maxK = 10, reps = 100, pItem = 0.8, pFeature = 1, clusterAlg = ”km”, and distance = ”euclidean”. The optimal number of clusters was determined based on the cumulative distribution function (CDF) curves. The clustering results were validated through principal component analysis (PCA) and the SubMap method [44].

Mutational landscape analysis

The R package ‘maftools’ was utilized for analyzing and visualizing the mutational landscape, identifying differential mutations between groups, and examining co-occurring and mutually exclusive events [45].

Tumor microenvironment analysis

The ESTIMATE method was employed to evaluate the immune, stromal, and tumor score of each sample. For RNA-seq data, the platform parameter was set to “illumina”, while for microarray data, the “platform” parameter was set to “affymetrix” [46]. Tumor contexture was deconvoluted using the quanTIseq method [39]. Since tumor tissues were analyzed, the parameter “tumor = TRUE” was specified. For RNA-seq data, the “arrays” parameter was set to “FALSE”, whereas for microarray data, it was set to “TRUE” [47].

Immunotherapy response prediction

The Tumor Immune Dysfunction and Exclusion (TIDE) module was employed to predict patients’ responses to immunotherapy [40]. A TIDE score greater than 0 indicates resistance to immunotherapy.

Identification of differentially expressed genes (DEGs)

The ‘DESeq2’ R package was employed to detect differentially expressed genes (DEGs) between groups. Genes with an absolute log fold change of 1 or greater and an adjusted p-value below 0.05 were classified as DEGs.

Development of the SRRS

A univariate Cox regression analysis was conducted on the genes upregulated in the C1 clusters to pinpoint those linked to prognosis. To determine the most relevant features for overall survival in LUAD patients, we applied LASSO (Least Absolute Shrinkage and Selection Operator) regression, utilizing the ‘glmnet’ R package. Based on the multivariate LASSO model, the SRRS for each patient was calculated using the following formula:

SRRS=∑LASSO regression coefficient×normalized expression level of each gene

Development of the predictive model

To further validate the SRRS as an independent prognostic factor for LUAD, multivariate Cox regression analyses were performed. Prognostic model were then developed for the LUAD cohort using the R package ‘rms’. A nomogram was created to predict the 1-, 3-, and 5 year survival rates of LUAD patients. The calibration plots were generated to compare actual outcomes with predicted results. The accuracy and sensitivity of the model were evaluated through receiver operating characteristic (ROC) curves.

Statistical analysis

All data processing, statistical analysis, and visualizations were performed using R software version 4.3.1. Categorical variables were compared using either the chi-squared test or Fisher’s exact test, while continuous variables were analyzed using the Wilcoxon rank-sum test or the t-test. The ‘survminer’ R package was employed to determine the optimal cut-off values. Cox regression and Kaplan–Meier analyses were conducted using the ‘survival’ R package. The time-dependent area under the ROC curve (AUC) for survival variables was calculated with the ‘timeROC’ R package. A P value of less than 0.05 was considered statistically significant.

Result

Development of molecular clusters based on senescence related genes

A gene set called SenMayo, which identifies senescent cells, was used as the senescence-related gene set. The GSVA method was used to calculate the SenMayo score for each sample, which was then utilized as the SenMayo-enrichment score. Analyses were conducted on two LUAD datasets. Survival analyses illustrate that patients with elevated SenMayo-enrichment score had worse prognoses, indicating a possible association between cellular senescence and LUAD outcomes (Fig. 1 A, B). To classify TCGA patients, consensus clustering was conducted using senescence-related genes. The analysis, guided by the CDF curves, revealed that the optimal clustering result was achieved by dividing the patients into two distinct clusters (Supplementary Fig. 1A, B). PCA results further confirmed that LUAD samples were clearly separated into two clusters (Fig. 1C, D). This classification was replicated in another lung cancer dataset, similarly identifying two distinct clusters of LUAD patients (Supplementary Fig. 1C–D). The SubMap method was employed to verify the consistency of the clustering results across both datasets (Fig. 1E). Next, we examined the survival differences between the two clusters and found that patients in cluster C1 had poorer outcomes (Fig. 1F). Moreover, the C1 cluster exhibited a markedly higher SenMayo-enrichment score compared to the C2 cluster (Fig. 1G). We validated these findings in an independent dataset, which yielded consistent results (Fig. 1H, I).

Fig. 1.

Fig. 1

Open in a new tab

Development of molecular clusters based on senescence-related genes in LUAD. A Kaplan–Meier survival curves of TCGA cohort. B Kaplan–Meier survival curves of GSE72094 cohort. C PCA confirming the separation of TCGA samples into two distinct clusters. D PCA confirming the separation of GSE72094 samples into two distinct clusters. E SubMap analysis confirming consistency of clustering across datasets. F Kaplan–Meier survival curves of TCGA cohort. G SenMayo-enrichment scores of C1 and C2 clusters in TCGA cohorts. H Kaplan–Meier survival curves of the GSE72094 cohort. I SenMayo-enrichment scores of C1 and C2 clusters in GSE72094 cohort. ****P < 0.0001

Differences in pathways, clinical features, and mutation landscape between clusters

Using GSVA, scores for senescence-related gene sets were calculated. In the C1 clusters, multiple senescence-related pathways were upregulated (Fig. 2A, B), suggesting that C1 cluster may represent a senescence-associated cluster. The clinical features of patients from each cluster are presented (Fig. 2C, D). There were no significant differences in the clinical information of patients between the two clusters (Supplementary Table 2, Supplementary Table 3). Additionally, we analyzed the expression of several senescence markers between the two clusters. In both TCGA-LUAD and GSE72094 cohorts, CDKN1A, IL6, and SERPINE1 were upregulated in the C1 cluster, which is consistent with the senescence phenotype. However, LMNB1, a molecule downregulated in the senescence phenotype, was upregulated in the C1 cluster of both cohorts (Supplementary Fig. 2A, B). The expression of CDKN2B was highly expressed only in the C1 cluster of the TCGA-LUAD cohort, while TP53 was expressed at low levels in the TCGA-LUAD cohort (Supplementary Fig. 2A, B). The mutation landscapes of the two clusters showed slight differences. In the C1 cluster, TP53 was the gene with the highest mutation frequency, whereas in the C2 cluster, TTN mutations were the most common. Additionally, some mutations, such as STK11 and KEAP1, ranked among the top 10 most frequent mutations in C2 cluster but did not make it into the top 20 in C1 clusters (Supplementary Fig. 3A, B). The mutation frequencies of specific genes differed between the C1 and C2 patient groups. Mutations in STK11 and KEAP1 were more prevalent in the C2 group, while TP53 mutations were significantly more common in the C1 group (Supplementary Fig. 3C). The majority of the frequently mutated genes in both clusters showed patterns of co-mutation (Supplementary Fig. 3D, E). Hallmark pathway scores were also calculated using GSVA. When comparing the C1 and C2 clusters, pathways such as epithelial-mesenchymal transition (EMT), hypoxia, p53 signaling, and TGF-β signaling were upregulated in the C1 cluster (Fig. 3A, B).

Fig. 2.

Fig. 2

Open in a new tab

Differences in pathways, clinical features, and mutation landscapes between LUAD clusters. A GSVA analysis of senescence-related gene sets in TCGA cohort. B GSVA analysis of senescence-related gene sets in GSE72094 cohort. C Clinical characteristics of patients in TCGA cohort. D Clinical characteristics of patients in GSE72094 cohort

Fig. 3.

Fig. 3

Open in a new tab

Hallmark pathway differences between LUAD clusters. A GSVA analysis of hallmark pathways comparing C1 and C2 clusters in TCGA cohort. B GSVA analysis of hallmark pathways comparing C1 and C2 clusters in GSE72094 cohort

Immune characteristics between LUAD clusters

Substantial differences in immune regulators were evident between the two clusters. Immune checkpoints such as PD-L1 (CD274), PD-1 (PDCD1), TIM-3 (HAVCR2), CTLA4, TIGIT, and LAG3, were all elevated in the C1 cluster (Fig. 4A, C). Furthermore, the C1 cluster exhibited elevated immune and stromal scores relative to the C2 cluster, whereas the C2 cluster demonstrated greater tumor purity (Fig. 4B, D). To gain deeper insights into the immune microenvironment of the clusters, the quanTIseq method was used for deconvolution, assessing the immune cell composition in the TCGA cohort. It was found that immunosuppressive cells, such as M2 macrophages and Tregs, were found in considerably higher quantities in the C1 cluster compared to the C2 cluster (Fig. 5A, B). Additionally, GSVA was employed to assess the scores for different immune signatures, revealing that signatures related to Tregs, MDSCs, T cell exhaustion, CAFs, and TAMs were higher in the C1 cluster relative to the C2 cluster (Fig. 5C). These findings were validated in the GSE72094 dataset (Fig. 5D–F). These findings suggest that the C1 cluster is characterized by an immunosuppressive microenvironment.

Fig. 4.

Fig. 4

Open in a new tab

Immune characteristics between LUAD clusters. A Differential expression of immune regulators between the two clusters in TCGA cohort. B Immune score, stromal score, estimate score and tumor purity of C1 and C2 clusters in TCGA cohort. C Differential expression of immune regulators between the two clusters in GSE72094 cohort. D Immune score, stromal score, estimate score and tumor purity of C1 and C2 clusters in GSE72094 cohort. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001

Fig. 5.

Fig. 5

Open in a new tab

Immune cell composition and immunosuppressive signatures in LUAD clusters. A The proportion of each cell type in the TCGA cohort. B The differences in cell composition between C1 and C2 clusters in the TCGA cohort. C GSVA scores for immune signatures between C1 and C2 clusters in the TCGA cohort. D The proportion of each cell type in the GSE72094 cohort. E The differences in cell composition between C1 and C2 clusters in the GSE72094 cohort. F GSVA scores for immune signatures between C1 and C2 clusters in the GSE72094 cohort. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001

Prediction of immunotherapy response between clusters

The TIDE algorithm predicts a patient’s response to immunotherapy by evaluating various immune-related transcriptomic features. Using the TIDE algorithm, The TIDE score was computed for every patient in the TCGA cohort. (Fig. 6A). Based on the predictions from the TIDE algorithm, the proportion of patients in the C1 subgroup who are likely to be resistant to immunotherapy may be higher than that in the C2 subgroup. (Fig. 6B). Additionally, The TIDE score for the C1 cluster was markedly elevated compared to that of the C2 cluster (Fig. 6C). To validate these findings, we applied the same analysis to the GSE72094 cohort, where similar results were observed. Based on the TIDE algorithm’s prediction, the proportion of immunotherapy-resistant patients was estimated to be higher in the C1 cluster compared to the C2 cluster, and the TIDE score for C1 cluster was consistently elevated compared to C2 cluster (Fig. 6D–F). Based on the aggregated findings, it is likely that the C1 cluster exhibits a higher resistance to immunotherapy.

Fig. 6.

Fig. 6

Open in a new tab

Prediction of immunotherapy response between LUAD clusters. A TIDE score calculation for each patient in the TCGA cohort. B Difference in response to immunotherapy between different clusters in TCGA cohort. C The differences in TIDE score between the C1 and C2 clusters in the TCGA cohort. D TIDE score calculation for each patient in the TCGA cohort. E Difference in response to immunotherapy between different clusters in TCGA cohort. F The differences in TIDE score between the C1 and C2 clusters in the TCGA cohort. ****P < 0.0001

Construction and validation of SRRS via the machine learning

Given the poorer prognosis of patients in the C1 cluster, an analysis was conducted to compare gene expression between the C1 and C2 clusters. The genes upregulated in the C1 cluster were identified as marker genes for C1 clusters (Fig. 7A). Through univariate Cox regression analysis, 139 marker genes were found to have a significant association with patient prognosis. These 139 prognostic marker genes were then further refined using LASSO regression, resulting in a set of 15 genes with non-zero coefficients (Fig. 7B, C). Based on the identified genes, a risk assessment model was developed, and each sample’s SRRS was computed. Patients were classified into high-risk and low-risk groups according to the median SRRS. This stratification revealed that individuals with a high SRRS had significantly poorer overall survival compared to those with a low SRRS in both the TCGA cohort and the GSE72094 validation cohort (Fig. 8A–D). To evaluate the SRRS’s predictive ability for survival, ROC curves were generated for 1 year, 3 year, and 5 year survival outcomes. In the TCGA cohort, the AUC values were 0.699 for 1 year, 0.679 for 3 year, and 0.635 for 5 year survival, reflecting strong predictive capability (Fig. 8E). Likewise, in the GSE72094 validation cohort, the AUC values were 0.693 for 1 year, 0.713 for 3 year, and 0.748 for 5 year survival, indicating consistent and robust predictive performance (Fig. 8F).

Fig. 7.

Fig. 7

Open in a new tab

Construction of SRRS via machine learning. A Identification of marker genes upregulated in the C1 cluster through differential gene expression analysis. B LASSO regression results selecting 15 prognostic genes with non-zero coefficients. C The coefficients of selected genes

Fig. 8.

Fig. 8

Open in a new tab

The performance of SRRS for predicting prognosis. A SRRS and patient survival status in TCGA cohort. B SRRS and patient survival status in GSE72094 cohort. C Kaplan–Meier survival in patients with different SRRS in TCGA cohort. D Kaplan–Meier survival in patients with different SRRS in GSE72094 cohort. E ROC curves assessing the predictive performance of SRRS for 1-, 3-, and 5 year survival in the TCGA cohort. F ROC curves assessing the predictive performance of SRRS for 1-, 3-, and 5 year survival in GSE72094 cohort

Development and evaluation of the prognostic model

To investigate how SRRS correlates with patient outcomes, a multivariate Cox regression analysis was conducted, incorporating SRRS alongside other relevant prognostic variables, including clinical stage and age. The analysis revealed that both SRRS and clinical stage were significant determinants of patient prognosis (Fig. 9A, B). Subsequently, a nomogram was developed to predict survival probabilities at 1 year, 3 years, and 5 years, with SRRS identified as a pivotal prognostic indicator (Fig. 9C). The calibration curves confirmed a strong agreement between the predicted survival probabilities and actual observed outcomes (Fig. 9D). Patients were categorized into high-risk and low-risk groups based on their linear predictor scores, with those in the high-risk category experiencing significantly worse prognoses (Fig. 9E, F). The model’s efficacy in predicting survival outcomes was evidenced by high AUC values for 1 year, 3 year, and 5 year survival, demonstrating robust predictive performance and consistency across both training and validation cohorts (Fig. 9G, H).

Fig. 9.

Fig. 9

Open in a new tab

Development and evaluation of the nomogram survival model. A Multivariate analysis of the clinicopathologic characteristics and SRRS in TCGA cohort. B Multivariate analysis of the clinicopathologic characteristics and SRRS in GSE72094 cohort. C Nomogram predicting 1-, 3-, and 5 year survival probabilities. D Calibration curves showing consistency between predicted and observed survival probabilities. E Kaplan–Meier survival curves for two risk groups of TCGA cohort. F Kaplan–Meier survival curves for two risk groups of GSE72094 cohort. G ROC curves of risk model in TCGA cohort. H ROC curves of risk model in GSE72094 cohort

Association between SRRS and immunotherapy response

Patients in the TCGA cohort were categorized into high-risk and low-risk groups according to their SRRS values. Consistent with the clustering results mentioned earlier, individuals with elevated SRRS levels showed a greater tendency toward immunotherapy resistance, as indicated by higher TIDE scores (Fig. 10A–C). This pattern was also evident in the GSE72094 cohort (Fig. 10D–F). To further corroborate the link between SRRS and immunotherapy response, we performed validation across three independent cohorts representing various cancer types: NSCLC, UC, and melanoma. All participants in these cohorts had received ICI treatment. Across these validation groups, patients with high SRRS consistently experienced shorter overall survival compared to those with low SRRS. Additionally, a higher proportion of patients with elevated SRRS exhibited resistance to ICI therapy in all cohorts, suggesting that SRRS might be a valuable predictive biomarker for guiding clinical decisions regarding ICI treatment (Fig. 11A–F).

Fig. 10.

Fig. 10

Open in a new tab

Association between SRRS and immunotherapy response in LUAD. A SRRS calculation for patients in different groups in the TCGA cohort. B Difference in response to immunotherapy between different groups in TCGA cohort. C The differences in TIDE score between high SRRS and low SRRS groups in the TCGA cohort. D SRRS calculation for patients in different groups in the GSE72094 cohort. E Difference in response to immunotherapy between different groups in GSE72094 cohort. F The differences in TIDE score between high SRRS and low SRRS groups in the GSE72094 cohort. ****P < 0.0001

Fig. 11.

Fig. 11

Open in a new tab

Validation of SRRS as a predictor of immunotherapy response in independent cohorts. A Kaplan–Meier survival curves of high SRRS and low SRRS patients in GSE135222 cohort. B Difference in response to immunotherapy between different groups in GSE135222 cohort. C Kaplan–Meier survival curves of high SRRS and low SRRS patients in IMvigor210 cohort. D Difference in response to immunotherapy between different groups in IMvigor210 cohort. E Kaplan–Meier survival curves of high SRRS and low SRRS patients in MGSP cohort. F Difference in response to immunotherapy between different groups in MGSP cohort

Discussion

Lung cancer is one of the most lethal cancers worldwide. Immunotherapy targeting immune checkpoints such as PD-1, PD-L1, and CTLA-4 holds great promise for lung cancer treatment. However, the low response rate remains a significant challenge for immunotherapy. Currently available clinical biomarkers, including PD-L1 expression, tumor mutational burden (TMB), and microsatellite instability-high (MSI-H), offer some predictive value but have limitations in accuracy and applicability across different cancer types [48]. Therefore, overcoming immunotherapy resistance and identifying patients most likely to benefit from these therapies are therefore critical for improving clinical outcomes. Cellular senescence plays a dual and complex role in cancer. Initially recognized as a protective mechanism, emerging evidence has revealed its potential to promote tumor progression and suppress immune surveillance, making it as one of the hallmarks of cancer [49, 50].

Several studies have explored the relationship between the senescence phenotype and lung cancer prognosis, with most findings indicating that tumors exhibiting a senescence signature are associated with unfavorable outcomes. Shao et al. utilized immunohistochemistry (IHC) to detect SMP30, a senescence-associated protein, and analyzed its correlation with prognosis, finding that higher SMP30 expression was associated with worse outcomes [51]. Similarly, Giatromanolaki et al. employed IHC to identify lipofuscin, a marker of senescence, and observed that its accumulation was linked to poor prognosis [52]. Domen et al. extended this approach by utilizing four markers—lipofuscin, p16INK4a, p21WAF1/Cip1, and Ki67—to define senescent tumors, further corroborating the association between senescence signatures and adverse outcomes [53]. In addition to IHC-based approaches, researchers have developed prognostic models using gene expression data to predict clinical outcomes. Liu et al., for instance, developed a 12-gene cellular senescence score from the CellAge database. This model effectively predicted prognosis in LUAD patients [29]. Similarly, Ru et al. devised a senescence-related gene risk score that stratified patients into high- and low-senescence score groups, with poorer outcomes observed in the high-score group [54]. Furthermore, Khadirnaikar et al. categorized patients into two subtypes, C1 and C2, based on senescence-related genes linked to ionizing radiation– and doxorubicin-induced senescence [55]. The C1 subtype, representing the high-senescence group, was characterized by high expression of senescence-inducing genes and low expression of senescence-suppressing genes. Notably, contrary to previous studies, patients in the high-senescence group showed a better prognosis despite resistance to HDAC and CDK9 inhibitors. This finding highlights the complexity of cellular senescence. The heterogeneity of senescence presents a major challenge in identifying senescent cells in vivo, as relying on a single marker for detection often lacks precision and fails to capture the full spectrum of senescence phenotypes [56]. Moreover, senescence-related markers, including the SASP, exhibit considerable variability across different cancer types [57]. Our study also observed this phenomenon, where in the C1 cluster enriched for senescence pathways, the expression of senescence markers exhibited opposite trends (Supplemental Fig. 2). This variability can lead to discrepancies in findings and affect the clinical applicability of these transcriptomic signatures.

SenMayo, a gene set derived from aging-associated senescence, has been validated across bulk and single-cell RNA-seq datasets in human and murine tissues. Compared to other senescence-related gene sets, SenMayo demonstrates superior performance in identifying senescent cells and has been successfully applied in oncology. For example, SenMayo has been successfully utilized to predict prognosis in colorectal cancer [58]. Building on its demonstrated efficacy, we selected SenMayo as the senescence-related gene set for this study.

Using SenMayo, LUAD patients were classified into two distinct clusters with differing molecular characteristics, mutation landscapes, and immune profiles. Patients in the high SenMayo-enrichment cluster (C1) showed significant overlap with high-senescence phenotypes, poorer prognoses, and enriched senescence-related pathways. Multiple senescence-related pathways, including the SenMayo pathway, were enriched in the C1 cluster, indicating that it represents a senescence-associated cluster. Furthermore, patients with high senescence scores, as well as those in the C1 cluster, exhibited worse prognoses compared to those with lower senescence scores or those in the C2 cluster, consistent with previous finding [29, 51–54]. This underscores a strong connection between cellular senescence and poor prognosis in lung cancer patients. To improve the prediction of patient prognosis, differential gene analysis was performed between the C1 and C2 clusters to identify genes highly expressed in the C1 cluster. Cox regression and LASSO regression were then used to select 15 prognostic genes to construct a risk model. This model demonstrated strong predictive performance for LUAD patient outcomes. Additionally, a multivariate Cox regression model that integrated SRRS with clinical characteristics provided even greater accuracy in predicting patient prognosis.

TIDE analysis revealed that the C1 cluster had higher TIDE scores, suggesting increased resistance to immune checkpoint blockade therapy compared to the C2 cluster. PD-L1 (CD274) is commonly regarded as a key biomarker for predicting response to αPD-1/αPD-L1 therapies [59]. Interestingly, despite the higher expression of PD-L1 in the C1 cluster, patients in this group still exhibited resistance to immunotherapy. Several reasons for this discrepancy are hypothesized. First, mRNA levels do not always correlate with protein levels, as PD-L1 protein is subject to post-translational regulation. For example, SPOP can ubiquitinate PD-L1, leading to its degradation [60], while GLT1D1 promotes the N-glycosylation of PD-L1, enhancing its stability and immunosuppressive function [61]. Additionally, USP5 deubiquitinates PD-1, preventing its degradation and influencing immune therapy responses [62]. Second, beyond the PD-1/PD-L1 axis, other immune checkpoints such as CTLA-4, LAG3, TIGIT, BTLA, and TIM-3 (HAVCR2) were upregulated in the C1 cluster, potentially contributing to reduced immunotherapy efficacy [63, 64]. Additionally, immunosuppressive cells such as Tregs and M2 macrophages were more prevalent in the C1 cluster, and signatures associated with MDSCs, T cell exhaustion, CAFs, and TAMs were enriched. This immunosuppressive microenvironment likely contributes to the observed resistance to immunotherapy in the C1 cluster. These findings suggest that additional biomarkers beyond PD-L1 are necessary for better predicting immunotherapy response. Currently available clinical biomarkers, such as tumor mutational burden (TMB) and microsatellite instability-high (MSI-H), provide some predictive value but still have limitations in terms of accuracy and applicability across different cancer types.

Upon further investigation, we found that the mutation frequency of TP53 was higher in the C1 group compared to the C2 group, while the mutation frequencies of STK11 and KEAP1 were lower in the C1 group. TP53 is one of the markers of senescence [65]. Our analysis revealed that in the C1 subgroup, which has a high senescence enrichment score, the frequency of TP53 mutations was even higher, reflecting the heterogeneity of senescence. Furthermore, TP53 mutations are considered a favorable factor for immunotherapy [66], while STK11 and KEAP1 mutations are regarded as unfavorable factors [67–69]. However, our predictive analysis indicated relative resistance to immunotherapy in the C1 cluster, characterized by frequent TP53 mutations and infrequent STK11 and KEAP1 mutations. This highlights that there are many factors influencing immunotherapy responses, and relying solely on the mutation frequencies of TP53, STK11, and KEAP1 may not fully predict the efficacy of immunotherapy.

Some studies have explored the use of gene set expression to predict immunotherapy efficacy [70–72]. However, research specifically leveraging senescence-related gene sets for predicting immunotherapy outcomes in lung cancer remains rare. This study explored the relationship between SRRS and immunotherapy efficacy, revealing that SRRS can serve as a valuable predictor of immunotherapy response. Using the TIDE method, it was found that the C1 cluster exhibited resistance to ICIs. Similarly, patients with higher SRRS had elevated TIDE scores, indicating a greater likelihood of non-response to immunotherapy. Further validation using datasets from NSCLC, UC, and melanoma revealed that high SRRS were associated with poorer immunotherapy efficacy in ICI-treated patients. Additionally, patients with higher SRRS had poorer survival outcomes following immunotherapy, further demonstrating the robustness and accuracy of the model.

While our findings demonstrate a strong association between cellular senescence and both poor prognosis and resistance to ICIs in lung cancer patients, it is essential to consider the potential confounding effects of therapy-induced senescence (TIS). Chemotherapeutic agents and radiotherapy are well-established inducers of senescence in tumor cells, which can, in turn, remodel the tumor microenvironment, alter immune cell infiltration dynamics, and modulate transcriptomic profiles [73–75]. These TIS-driven alterations could influence clinical outcomes, therapeutic responses, and the results of clustering based on senescence-related gene expression, potentially confounding our findings. However, the absence of detailed treatment data in our study cohorts—such as specific therapies and their timing—restricted our ability to fully assess the impact of TIS on the predictive model. This underscores the importance of including comprehensive treatment metadata, such as therapy histories and longitudinal biomarker assessments, in future studies. Moreover, integrating multi-omics approaches, including single-cell transcriptomics paired with spatial profiling, may further refine the accuracy of predictive models for prognosis and immunotherapy responses in lung adenocarcinoma.

These findings highlight the clinical potential of SRRS in guiding personalized treatment and improving patient outcomes. As sequencing costs continue to decrease, detailed transcriptomic analysis of LUAD tumors has become increasingly feasible. Utilizing SRRS for patient subtyping could help identify those most likely to benefit from immunotherapy, thereby enabling more targeted and individualized treatment strategies.

Limitation

This study has a few limitations. Firstly, the clustering analysis relied exclusively on transcriptomic data and did not incorporate proteomic information. Secondly, the research depends on previously published datasets and lacks validation through direct in vivo or in vitro experimentation. Third, the absence of treatment regimen information in the analyzed datasets represents a notable constraint. Since clinical interventions such as chemotherapy are known to induce senescence, this unaccounted variable may have affected the accuracy of our results.

Conclusion

In summary, this research has revealed two distinct clusters of LUAD, with the senescence-associated cluster exhibiting poor prognosis and resistance to immunotherapy. The senescence-related SRRS was shown to predict patient response to immunotherapy. Collectively, our results offer new perspectives on the connection between senescence and immunotherapy resistance, setting the stage for future investigations.

Supplementary Information

Additional file1 (DOCX 7924 KB) (7.7MB, docx)

Acknowledgements

We are grateful to the researchers and institutions who provided the data from publicly accessible sources used in this study, as well as the participants involved in the original studies.

Author contributions

All authors contributed to designing the study and writing the manuscript. X.R.G and X.S conducted data collection and analysis. S.S.H and S.K.H revised the manuscript.

Funding

This project is supported by the Doctoral Startup Fund of Affiliated Hospital of Southwest Medical University (No. 19025).

Data availability

All of the data used in this study have been obtained from publicly available sources. The transcriptomic data were retrieved from TCGA (https://portal.gdc.cancer.gov/) or GEO (https://www.ncbi.nlm.nih.gov/geo/) under accession number GSE72094, GSE135222. The IMvigor210 data were accessed via the ‘IMvigor210CoreBiologies’ R package (http://research-pub.gene.com/IMvigor210CoreBiologies/). The MGSP data were accessed from published article ().

Code availability

The code utilized in this study can be obtained from the corresponding author upon reasonable request.

Declarations

Ethics approval and consent to participate

Not applicable.

Competing interests

The authors declare no competing interests.

Footnotes

Publisher's Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Xinrui Gao and Xiang Shen contributed equally.

Contributor Information

Shasha Huang, Email: hsalsa@qq.com.

Shangke Huang, Email: huangshangke001@swmu.edu.cn.

References

  • 1.Bray F, Laversanne M, Sung H, et al. Global cancer statistics 2022: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J Clin. 2024;74(3):229–63. 10.3322/caac.21834. [DOI] [PubMed] [Google Scholar]
  • 2.Herbst RS, Morgensztern D, Boshoff C. The biology and management of non-small cell lung cancer. Nature. 2018;553(7689):446–54. 10.1038/nature25183. [DOI] [PubMed] [Google Scholar]
  • 3.Thai AA, Solomon BJ, Sequist LV, et al. Lung cancer. Lancet. 2021;398(10299):535–54. 10.1016/s0140-6736(21)00312-3. [DOI] [PubMed] [Google Scholar]
  • 4.Herbst RS, Baas P, Kim DW, et al. Pembrolizumab versus docetaxel for previously treated, PD-L1-positive, advanced non-small-cell lung cancer (KEYNOTE-010): a randomised controlled trial. Lancet. 2016;387(10027):1540–50. 10.1016/s0140-6736(15)01281-7. [DOI] [PubMed] [Google Scholar]
  • 5.Reck M, Rodríguez-Abreu D, Robinson AG, et al. Pembrolizumab versus chemotherapy for PD-L1-positive non-small-cell lung cancer. N Engl J Med. 2016;375(19):1823–33. 10.1056/NEJMoa1606774. [DOI] [PubMed] [Google Scholar]
  • 6.Mok TSK, Wu YL, Kudaba I, et al. Pembrolizumab versus chemotherapy for previously untreated, PD-L1-expressing, locally advanced or metastatic non-small-cell lung cancer (KEYNOTE-042): a randomised, open-label, controlled, phase 3 trial. Lancet. 2019;393(10183):1819–30. 10.1016/s0140-6736(18)32409-7. [DOI] [PubMed] [Google Scholar]
  • 7.Carbone DP, Reck M, Paz-Ares L, et al. First-line nivolumab in stage IV or recurrent non-small-cell lung cancer. N Engl J Med. 2017;376(25):2415–26. 10.1056/NEJMoa1613493. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 8.Herbst RS, Giaccone G, de Marinis F, et al. Atezolizumab for first-line treatment of PD-L1-selected patients with NSCLC. N Engl J Med. 2020;383(14):1328–39. 10.1056/NEJMoa1917346. [DOI] [PubMed] [Google Scholar]
  • 9.Bashash D, Zandi Z, Kashani B, et al. Resistance to immunotherapy in human malignancies: mechanisms, research progresses, challenges, and opportunities. J Cell Physiol. 2022;237(1):346–72. 10.1002/jcp.30575. [DOI] [PubMed] [Google Scholar]
  • 10.Bejarano L, Jordāo MJC, Joyce JA. Therapeutic targeting of the tumor microenvironment. Cancer Discov. 2021;11(4):933–59. 10.1158/2159-8290.Cd-20-1808. [DOI] [PubMed] [Google Scholar]
  • 11.Lefler DS, Manobianco SA, Bashir B. Immunotherapy resistance in solid tumors: mechanisms and potential solutions. Cancer Biol Ther. 2024;25(1):2315655. 10.1080/15384047.2024.2315655. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 12.Gorgoulis V, Adams PD, Alimonti A, et al. Cellular senescence: defining a path forward. Cell. 2019;179(4):813–27. 10.1016/j.cell.2019.10.005. [DOI] [PubMed] [Google Scholar]
  • 13.Jha SK, De Rubis G, Devkota SR, et al. Cellular senescence in lung cancer: molecular mechanisms and therapeutic interventions. Ageing Res Rev. 2024;97: 102315. 10.1016/j.arr.2024.102315. [DOI] [PubMed] [Google Scholar]
  • 14.Le ON, Rodier F, Fontaine F, et al. Ionizing radiation-induced long-term expression of senescence markers in mice is independent of p53 and immune status. Aging Cell. 2010;9(3):398–409. 10.1111/j.1474-9726.2010.00567.x. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 15.Liu Y, Lomeli I, Kron SJ. Therapy-induced cellular senescence: potentiating tumor elimination or driving cancer resistance and recurrence. Cells. 2024. 10.3390/cells13151281. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 16.Faget DV, Ren Q, Stewart SA. Unmasking senescence: context-dependent effects of SASP in cancer. Nat Rev Cancer. 2019;19(8):439–53. 10.1038/s41568-019-0156-2. [DOI] [PubMed] [Google Scholar]
  • 17.Krtolica A, Parrinello S, Lockett S, et al. Senescent fibroblasts promote epithelial cell growth and tumorigenesis: a link between cancer and aging. Proc Natl Acad Sci U S A. 2001;98(21):12072–7. 10.1073/pnas.211053698. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 18.Canino C, Mori F, Cambria A, et al. SASP mediates chemoresistance and tumor-initiating-activity of mesothelioma cells. Oncogene. 2012;31(26):3148–63. 10.1038/onc.2011.485. [DOI] [PubMed] [Google Scholar]
  • 19.Demaria M, O’Leary MN, Chang J, et al. Cellular senescence promotes adverse effects of chemotherapy and cancer relapse. Cancer Discov. 2017;7(2):165–76. 10.1158/2159-8290.Cd-16-0241. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 20.Liu H, Xu Q, Wufuer H, et al. Rutin is a potent senomorphic agent to target senescent cells and can improve chemotherapeutic efficacy. Aging Cell. 2024;23(1):e13921. 10.1111/acel.13921. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 21.Eggert T, Wolter K, Ji J, et al. Distinct functions of senescence-associated immune responses in liver tumor surveillance and tumor progression. Cancer Cell. 2016;30(4):533–47. 10.1016/j.ccell.2016.09.003. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 22.Mazzoni M, Mauro G, Erreni M, et al. Senescent thyrocytes and thyroid tumor cells induce M2-like macrophage polarization of human monocytes via a PGE2-dependent mechanism. J Exp Clin Cancer Res. 2019;38(1):208. 10.1186/s13046-019-1198-8. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 23.Lujambio A, Akkari L, Simon J, et al. Non-cell-autonomous tumor suppression by p53. Cell. 2013;153(2):449–60. 10.1016/j.cell.2013.03.020. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 24.Chen Z, Zhou J, Wu Y, et al. METTL3 promotes cellular senescence of colorectal cancer via modulation of CDKN2B transcription and mRNA stability. Oncogene. 2024;43(13):976–91. 10.1038/s41388-024-02956-y. [DOI] [PubMed] [Google Scholar]
  • 25.Ruhland MK, Loza AJ, Capietto AH, et al. Stromal senescence establishes an immunosuppressive microenvironment that drives tumorigenesis. Nat Commun. 2016;7:11762. 10.1038/ncomms11762. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 26.Montes CL, Chapoval AI, Nelson J, et al. Tumor-induced senescent T cells with suppressor function: a potential form of tumor immune evasion. Cancer Res. 2008;68(3):870–9. 10.1158/0008-5472.Can-07-2282. [DOI] [PubMed] [Google Scholar]
  • 27.Ye J, Ma C, Hsueh EC, et al. TLR8 signaling enhances tumor immunity by preventing tumor-induced T-cell senescence. EMBO Mol Med. 2014;6(10):1294–311. 10.15252/emmm.201403918. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 28.Maggiorani D, Le O, Lisi V, et al. Senescence drives immunotherapy resistance by inducing an immunosuppressive tumor microenvironment. Nat Commun. 2024;15(1):2435. 10.1038/s41467-024-46769-9. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 29.Liu X, Lin L, Cai Q, et al. Construction and validation of a prognostic model based on novel senescence-related genes in non-small cell lung cancer patients with drug sensitivity and tumor microenvironment. Adv Biol (Weinh). 2023;7(12):e2300190. 10.1002/adbi.202300190. [DOI] [PubMed] [Google Scholar]
  • 30.Lv Y, Wu L, Jian H, et al. Identification and characterization of aging/senescence-induced genes in osteosarcoma and predicting clinical prognosis. Front Immunol. 2022;13: 997765. 10.3389/fimmu.2022.997765. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 31.Zhang X, Huang Y, Li Q, et al. Senescence risk score: a multifaceted prognostic tool predicting outcomes, stemness, and immune responses in colorectal cancer. Front Immunol. 2023;14:1265911. 10.3389/fimmu.2023.1265911. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 32.Yang W, An L, Li Y, et al. A cellular senescence-related genes model allows for prognosis and treatment stratification of cervical cancer: a bioinformatics analysis and external verification. Aging (Albany NY). 2023;15(18):9408–25. 10.18632/aging.204981. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 33.Schabath MB, Welsh EA, Fulp WJ, et al. Differential association of STK11 and TP53 with KRAS mutation-associated gene expression, proliferation and immune surveillance in lung adenocarcinoma. Oncogene. 2016;35(24):3209–16. 10.1038/onc.2015.375. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 34.Necchi A, Joseph RW, Loriot Y, et al. Atezolizumab in platinum-treated locally advanced or metastatic urothelial carcinoma: post-progression outcomes from the phase II IMvigor210 study. Ann Oncol. 2017;28(12):3044–50. 10.1093/annonc/mdx518. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 35.Jung H, Kim HS, Kim JY, et al. DNA methylation loss promotes immune evasion of tumours with high mutation and copy number load. Nat Commun. 2019;10(1):4278. 10.1038/s41467-019-12159-9. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 36.Liu D, Schilling B, Liu D, et al. Integrative molecular and clinical modeling of clinical outcomes to PD1 blockade in patients with metastatic melanoma. Nat Med. 2019;25(12):1916–27. 10.1038/s41591-019-0654-5. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 37.Saul D, Kosinsky RL, Atkinson EJ, et al. A new gene set identifies senescent cells and predicts senescence-associated pathways across tissues. Nat Commun. 2022;13(1):4827. 10.1038/s41467-022-32552-1. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 38.Rooney MS, Shukla SA, Wu CJ, et al. Molecular and genetic properties of tumors associated with local immune cytolytic activity. Cell. 2015;160(1–2):48–61. 10.1016/j.cell.2014.12.033. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 39.Wang G, Lu X, Dey P, et al. Targeting YAP-dependent MDSC infiltration impairs tumor progression. Cancer Discov. 2016;6(1):80–95. 10.1158/2159-8290.Cd-15-0224. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 40.Jiang P, Gu S, Pan D, et al. Signatures of T cell dysfunction and exclusion predict cancer immunotherapy response. Nat Med. 2018;24(10):1550–8. 10.1038/s41591-018-0136-1. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 41.Hänzelmann S, Castelo R, Guinney J. GSVA: gene set variation analysis for microarray and RNA-seq data. BMC Bioinf. 2013;14:7. 10.1186/1471-2105-14-7. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 42.Ritchie ME, Phipson B, Wu D, et al. limma powers differential expression analyses for RNA-sequencing and microarray studies. Nucleic Acids Res. 2015;43(7):e47. 10.1093/nar/gkv007. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 43.Wilkerson MD, Hayes DN. ConsensusClusterPlus: a class discovery tool with confidence assessments and item tracking. Bioinformatics. 2010;26(12):1572–3. 10.1093/bioinformatics/btq170. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 44.Hoshida Y, Brunet JP, Tamayo P, et al. Subclass mapping: identifying common subtypes in independent disease data sets. PLoS ONE. 2007;2(11):e1195. 10.1371/journal.pone.0001195. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 45.Mayakonda A, Lin DC, Assenov Y, et al. Maftools: efficient and comprehensive analysis of somatic variants in cancer. Genome Res. 2018;28(11):1747–56. 10.1101/gr.239244.118. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 46.Yoshihara K, Shahmoradgoli M, Martínez E, et al. Inferring tumour purity and stromal and immune cell admixture from expression data. Nat Commun. 2013;4:2612. 10.1038/ncomms3612. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 47.Finotello F, Mayer C, Plattner C, et al. Molecular and pharmacological modulators of the tumor immune contexture revealed by deconvolution of RNA-seq data. Genome Med. 2019;11(1):34. 10.1186/s13073-019-0638-6. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 48.Wang Y, Tong Z, Zhang W, et al. FDA-approved and emerging next generation predictive biomarkers for immune checkpoint inhibitors in cancer patients. Front Oncol. 2021;11: 683419. 10.3389/fonc.2021.683419. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 49.Hanahan D. Hallmarks of cancer: new dimensions. Cancer Discov. 2022;12(1):31–46. 10.1158/2159-8290.Cd-21-1059. [DOI] [PubMed] [Google Scholar]
  • 50.Lau L, David G. Pro- and anti-tumorigenic functions of the senescence-associated secretory phenotype. Expert Opin Ther Targets. 2019;23(12):1041–51. 10.1080/14728222.2019.1565658. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 51.Shao C, Guo K, Xu L, et al. Senescence marker protein 30 inhibits tumor growth by reducing HDAC4 expression in non-small cell lung cancer. Transl Lung Cancer Res. 2021;10(12):4558–73. 10.21037/tlcr-21-982. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 52.Giatromanolaki A, Kouroupi M, Balaska K, et al. A novel lipofuscin-detecting marker of senescence relates with hypoxia, dysregulated autophagy and with poor prognosis in non-small-cell-lung cancer. In Vivo. 2020;34(6):3187–93. 10.21873/invivo.12154. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 53.Domen A, Deben C, De Pauw I, et al. Prognostic implications of cellular senescence in resected non-small cell lung cancer. Transl Lung Cancer Res. 2022;11(8):1526–39. 10.21037/tlcr-22-192. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 54.Ru K, Cui L, Wu C, et al. Exploring the molecular and immune landscape of cellular senescence in lung adenocarcinoma. Front Immunol. 2024;15:1347770. 10.3389/fimmu.2024.1347770. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 55.Khadirnaikar S, Chatterjee A, Shukla S. Identification and characterization of senescence phenotype in lung adenocarcinoma with high drug sensitivity. Am J Pathol. 2021;191(11):1966–73. 10.1016/j.ajpath.2021.07.005. [DOI] [PubMed] [Google Scholar]
  • 56.Kirschner K, Rattanavirotkul N, Quince MF, et al. Functional heterogeneity in senescence. Biochem Soc Trans. 2020;48(3):765–73. 10.1042/bst20190109. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 57.Jochems F, Thijssen B, De Conti G, et al. The cancer SENESCopedia: a delineation of cancer cell senescence. Cell Rep. 2021;36(4):109441. 10.1016/j.celrep.2021.109441. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 58.Ungvari Z, Ungvari A, Bianchini G, et al. Prognostic significance of a signature based on senescence-related genes in colorectal cancer. Geroscience. 2024;46(5):4495–504. 10.1007/s11357-024-01164-6. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 59.Brody R, Zhang Y, Ballas M, et al. PD-L1 expression in advanced NSCLC: insights into risk stratification and treatment selection from a systematic literature review. Lung Cancer. 2017;112:200–15. 10.1016/j.lungcan.2017.08.005. [DOI] [PubMed] [Google Scholar]
  • 60.Zhang J, Bu X, Wang H, et al. Cyclin D-CDK4 kinase destabilizes PD-L1 via cullin 3-SPOP to control cancer immune surveillance. Nature. 2018;553(7686):91–5. 10.1038/nature25015. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 61.Liu X, Zhang Y, Han Y, et al. Overexpression of GLT1D1 induces immunosuppression through glycosylation of PD-L1 and predicts poor prognosis in B-cell lymphoma. Mol Oncol. 2020;14(5):1028–44. 10.1002/1878-0261.12664. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 62.Xiao X, Shi J, He C, et al. ERK and USP5 govern PD-1 homeostasis via deubiquitination to modulate tumor immunotherapy. Nat Commun. 2023;14(1):2859. 10.1038/s41467-023-38605-3. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 63.Borgeaud M, Sandoval J, Obeid M, et al. Novel targets for immune-checkpoint inhibition in cancer. Cancer Treat Rev. 2023;120: 102614. 10.1016/j.ctrv.2023.102614. [DOI] [PubMed] [Google Scholar]
  • 64.Kong X, Zhang J, Chen S, et al. Immune checkpoint inhibitors: breakthroughs in cancer treatment. Cancer Biol Med. 2024;21(6):451–72. 10.20892/j.issn.2095-3941.2024.0055. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 65.Tesei A, Arienti C, Bossi G, et al. TP53 drives abscopal effect by secretion of senescence-associated molecular signals in non-small cell lung cancer. J Exp Clin Cancer Res. 2021;40(1):89. 10.1186/s13046-021-01883-0. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 66.Dong ZY, Zhong WZ, Zhang XC, et al. Potential predictive value of TP53 and KRAS mutation status for response to PD-1 blockade immunotherapy in lung adenocarcinoma. Clin Cancer Res. 2017;23(12):3012–24. 10.1158/1078-0432.Ccr-16-2554. [DOI] [PubMed] [Google Scholar]
  • 67.Ricciuti B, Arbour KC, Lin JJ, et al. Diminished efficacy of programmed death-(Ligand)1 inhibition in STK11- and KEAP1-mutant lung adenocarcinoma is affected by KRAS mutation status. J Thorac Oncol. 2022;17(3):399–410. 10.1016/j.jtho.2021.10.013. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 68.Skoulidis F, Goldberg ME, Greenawalt DM, et al. STK11/LKB1 mutations and PD-1 inhibitor resistance in KRAS-mutant lung adenocarcinoma. Cancer Discov. 2018;8(7):822–35. 10.1158/2159-8290.Cd-18-0099. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 69.Marinelli D, Mazzotta M, Scalera S, et al. KEAP1-driven co-mutations in lung adenocarcinoma unresponsive to immunotherapy despite high tumor mutational burden. Ann Oncol. 2020;31(12):1746–54. 10.1016/j.annonc.2020.08.2105. [DOI] [PubMed] [Google Scholar]
  • 70.Li J, Cao Y, Liu Y, et al. Multiomics profiling reveals the benefits of gamma-delta (γδ) T lymphocytes for improving the tumor microenvironment, immunotherapy efficacy and prognosis in cervical cancer. J Immunother Cancer. 2024. 10.1136/jitc-2023-008355. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 71.Chen D, Liu P, Lu X, et al. Pan-cancer analysis implicates novel insights of lactate metabolism into immunotherapy response prediction and survival prognostication. J Exp Clin Cancer Res. 2024;43(1):125. 10.1186/s13046-024-03042-7. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 72.Zhao Q, Yu H, Shi M, et al. Tumor microenvironment characteristics of lipid metabolism reprogramming related to ferroptosis and EndMT influencing prognosis in gastric cancer. Int Immunopharmacol. 2024;137: 112433. 10.1016/j.intimp.2024.112433. [DOI] [PubMed] [Google Scholar]
  • 73.Saleh T, Bloukh S, Hasan M, et al. Therapy-induced senescence as a component of tumor biology: evidence from clinical cancer. Biochim Biophys Acta Rev Cancer. 2023;1878(6):188994. 10.1016/j.bbcan.2023.188994. [DOI] [PubMed] [Google Scholar]
  • 74.Schmitt CA, Wang B, Demaria M. Senescence and cancer—role and therapeutic opportunities. Nat Rev Clin Oncol. 2022;19(10):619–36. 10.1038/s41571-022-00668-4. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 75.Domen A, Deben C, Verswyvel J, et al. Cellular senescence in cancer: clinical detection and prognostic implications. J Exp Clin Cancer Res. 2022;41(1):360. 10.1186/s13046-022-02555-3. [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Additional file1 (DOCX 7924 KB) (7.7MB, docx)

Data Availability Statement

All of the data used in this study have been obtained from publicly available sources. The transcriptomic data were retrieved from TCGA (https://portal.gdc.cancer.gov/) or GEO (https://www.ncbi.nlm.nih.gov/geo/) under accession number GSE72094, GSE135222. The IMvigor210 data were accessed via the ‘IMvigor210CoreBiologies’ R package (http://research-pub.gene.com/IMvigor210CoreBiologies/). The MGSP data were accessed from published article ().

The code utilized in this study can be obtained from the corresponding author upon reasonable request.

Articles from Discover Oncology are provided here courtesy of Springer

RESOURCES