Development of a Cellular Senescence-Related Risk Model as a Prognostic Biomarker and Immunotherapy Response in Lung Squamous Cell Carcinoma

By Kai Kang¹, Sheng Wang¹, Xinjun Liang², Yajun Mao³

Affiliations

1. Department of Thoracic Surgery, Hubei Cancer Hospital, Tongji Medical College, Huazhong University of Science and Technology, Hubei, China
2. Department of Abdominal Oncology, Hubei Cancer Hospital, Tongji Medical College, Huazhong University of Science and Technology, Hubei, China
3. Operating Room, Shanghai Pulmonary Hospital, Tongji University School of Medicine, Shanghai, China

doi: 10.29271/jcpsp.2025.12.1546

ABSTRACT
Objective: To identify cellular senescence-associated genes in lung squamous cell carcinoma (LUSC) and explore their relationship with tumour microenvironment, immunotherapy response, and prognosis.
Study Design: An observational study.
Place and Duration of the Study:  Department of Thoracic Surgery, Hubei Cancer Hospital, Tongji Medical College, Huazhong University of Science and Technology, Hubei, China, from September to October 2023.
Methodology: Prognostic genes and clinical information of LUSC were derived from the Cancer Genome Atlas (TCGA). Subsequently, R packages were used to obtain differentially expressed genes (DEGs) associated with cellular senescence prognosis. In addition, univariate and least absolute shrinkage and selection operator (LASSO) Cox regression analyses were used to construct cellular senescence-related gene signatures to determine the degree of risk. Sensitivity to immune checkpoint inhibitors (ICI) was evaluated according to the median risk score, which served as an independent prognostic factor.
Results: By univariate Cox regression analysis of gene expression data and overall survival data in the TCGA-LUSC cohort, 11 DEGs associated with prognostic cellular senescence were identified. The ageing prognostic models include genes of IGFBP1, SERPINE1, SIX1, and TRPM8. Tumour immune dysfunction and exclusion (TIDE) score and tumour mutation burden (TMB) score were better in the low-risk group, with advantages in terms of prognostic and immunotherapy response rate.
Conclusion: The results suggest that senescence-related determinants are involved in the tumour microenvironment through senescence-related secretory phenotypes and can be used as reliable biomarkers for LUSC immunotherapy and prognosis.

Key Words: Cellular senescence, Lung squamous cell carcinoma, Tumour microenvironment, Immunotherapy, Prognosis.

INTRODUCTION

Cellular senescence is a conserved stress response characte- rised by irreversible cell cycle arrest, triggered by diverse stimuli such as oncogene activation, DNA damage, and oxidative stress.¹

This biological process exerts dual roles in tumourigenesis: on one hand, it acts as a tumour-suppressive barrier by halting the proliferation of genetically damaged cells;² on the other hand, senescent cells secrete a complex cocktail of factors, termed as senescence-associated secretory phenotype (SASP), which pro- motes chronic inflammation, angiogenesis, and immunosuppression,  thereby  facilitating tumour  progression  and  metastasis.³

Recent advances have enhanced the understanding of how senescence modulates cancer immunity. Senescent cells release pro-inflammatory cytokines (e.g., Interleukin-6 [IL-6], C-X-C motif chemokine ligand 8 [CXCL8]), growth factors (e.g., vascular endothelial growth factor [VEGF]), and matrix metalloproteinases (MMPs) via the SASP, which orchestrate tumour immune escape through multiple pathways.^4,5 For instance, IL-6 secreted by senescent cells activates STAT3 signalling in tumour-associated macrophages (TAMs), promo-ting their polarisation into an immunosuppressive M2 phenotype that inhibits cytotoxic T-cell function.⁶ Additionally, SASP-induced upregulation of PD-L1 on tumour cells and stromal cells directly impairs T-cell recognition.⁷ Moreover, SASP factors recruit regulatory T cells (Tregs) and myeloid-derived sup- pressor cells (MDSCs) into the tumour microenvironment (TME), further blunting antitumour immunity.⁸ These findings highlight senescence as a critical modulator of immune evasion, making it a potential target for improving immunotherapy efficacy.

This study aimed to identify differentially expressed genes (DEGs) associated with cell senescence prognosis in lung squamous cell carcinoma (LUSC), to construct a senescence-related signature (SRS) model, and to evaluate correlations between different patient risk groups and the TME, as well as the differences in immunotherapy response and prognosis. The goal was to guide personalised treatment strategies and improve prognostic prediction in LUSC. By elucidating the basic senescence biology, this study sought to provide actionable clinical insights for the management of LUSC patients and identify novel biomarkers related to immunotherapy and prognosis in this patient population.

METHODOLOGY

Data integration and bioinformatic analyses were conducted in the Department of Thoracic Surgery, Hubei Cancer Hospital, affiliated to Tongji Medical College, Huazhong University of Science and Technology, Hubei, China, from September to October 2023. Ethical approval was obtained from the Institutional Ethics Committee (Approval No. LLHBCH2024YN-108).

The transcriptomic data and related clinical parameters of LUSC patients were selected and downloaded (database is from the Cancer Genome Atlas [TCGA]; httpss://portal.gdc.cancer. gov/). The selected data category was bronchus and lung, the data type was TCGA, and the project was TCGA-LUSC. Useful information, such as fragments per kilo base of transcript per million mapped reads (FPKM) format and complete pathological information of each clinical sample was obtained using Strawberry Perl. Samples were included if they met four key conditions: histologically confirmed primary LUSC; complete transcriptomic data available in FPKM format; access to vital clinical metadata, including age, gender, clinical stage, and survival time; and a survival duration of 30 days or longer. Conversely, samples were excluded if they lacked critical clinical para- meters—such as unknown TNM stage or ambiguous survival status—to ensure data integrity for subsequent analyses.

Systematic pre-processing of transcriptomic data was conducted to standardise expression values and eliminate technical bias. Firstly, FPKM values were converted to transcripts per kilobase of the exon model per million mapped reads (TPM) for normalisation. Secondly, batch effects were removed using the sva package in R, ensuring consistency across the dataset. Finally, genes with an average TPM below 1 across all samples were filtered out to exclude low-expression genes. The expression, clinical parameters, as well as survival rates of the immuno-therapy cohort were obtained using the bioinformatics tool IMvigor210 (httpss://research-pub.gene.com/IMvigor210Core Biologies/).

Bioinformatics analysis was first conducted on the expression of senescence-related genes in tumour and normal specimens. The senescence gene expression data were merged with LUSC survival data from the TCGA database using the limma package in R software. Univariate Cox proportional hazard + vb analysis was applied to identify prognostic genes (p <0.05). Subsequently, R packages were used to investigate DEGs and prognostic genes. The DEGs and the prognostic genes were intersected, and then the interrelationships were analysed and presented with the correlation circle diagram.

Patients were randomly classified into a low-risk group and a high-risk group (50% each). The training group is characterised by the construction of senescence-associated gene signatures, while the testing group is applied for validation of the signatures. Using the glmnet R-package, Cox regression analysis was performed on random seeds, and the SRS-based risk scoring model was successfully established, and the risk was obtained through the model formula. Based on the median risk score, patients in the training and testing groups could be classified into a high-risk group and a low-risk group, followed by the evaluation of the SRS performance. With R packages, survival receiver operating characteristic (ROC) curves were created to assess the predictive value and accuracy of the risk model. Genes in the risk assessment model were independently tested using ROC to obtain the area under the curve (AUC) value. To confirm whether the prognostic model is an independent predictor of clinical outcomes, univariate and multivariate Cox regression analyses were performed, adjusting for age, gender, and clinical stage, with p <0.05 considered statistically significant.

The mutation spectrum of the TCGA-LUSC cohort was downloaded from the GDC data portal, and tumour mutation burden (TMB) level of each patient was calculated as the total number of non-synonymous somatic mutations per megabase. Patients were grouped by TMB level (high/low; using the median as the cut-off) for survival analysis. Subsequently, SRS was combined with TMB to compare survival differences among the sub- groups. The biological functions of selected genes were identified by Gene Set Enrichment Analysis (GSEA) using the clusterProfiler package, with adjusted p <0.05 (FDR correction) and |NES|>1 as significant enrichment thresholds.

Several algorithms (TIMER, CIBERSORT, and quanTIseq) were used to evaluate immune cell infiltration in the TCGA RNA-seq cohort based on gene expression profiles. Pearson’s correlation analysis was conducted to determine the relationship between SRS and immune cell infiltration, with p <0.05 considered statistically significant.

To predict the efficacy of immunotherapy and chemotherapy, the tumour immune dysfunction and exclusion (TIDE) algorithm was used for the prediction of the clinical response to immune checkpoint inhibitor (ICI).⁹ When the TIDE score was high, it indicated a high possibility of immune surveillance escape, and a relatively low success rate of blocking therapy at immune checkpoints occurred. The anti-CD274 (PD-L1) immunotherapy cohort (IMvigor 210) for LUSC was included in the analysis.

RESULTS

First, cellular ageing-associated DEGs and their prognostic role were identified. According to the gene expression and overall survival (OS) data in TCGA-LUSC analysed by univariate Cox regression, 19 prognostic DEGs in tumour samples (p <0.05; Figure 1A) and 126 DEGs between tumour samples and normal samples (Figure 1B) were identified. There were eight intersecting genes: cellular communication network factor 1 (CCN1), malate dehydrogenase 1 (MDH1), programmed cell death 10 (PDCD10), SIX1, Snail family transcriptional repressor 1 (SNAI1), SRY-box transcription factor 5 (SOX5), cyclin-dependent kinase inhibitor 1A (CDKN1A), and CCAAT enhancer-binding protein beta (CEBPB)). The intersection analysis showed that these genes are related to senescence (Figure 1B). Correlation analysis revealed tight interactions among these eight genes (Figure 1C). To avoid overfitting, the above 19 prognostic DEGs were incorporated into the least absolute shrinkage and selection operator (LASSO) regression analysis, which identified 11 candidate cellular senescence-related genes: B-cell lymphoma 6 (BCL6), selenoprotein H (SELENOH), enhancer of zeste homolog 2 (EZH2), G-protein subunit gamma 11 (GNG11), Holliday junction recognition protein (HJURP), IGFBP1, inositol-trisphosphate 3-kinase B (ITPKB), Klotho (KL), SERPINE1, SIX1, and TRPM8 (Figure 1D-F).

Multivariate Cox regression of the eleven prognostic genes identified four genes with strong associations with LUSC prognosis: IGFBP1, SERPINE1, SIX1, and TRPM8. The corresponding risk scoring formula was derived as follows: risk score = (0.3869 × IGFBP1 expression value) + (0.8635 × SERPINE1 expression value) + (-0.4248 × SIX1 expression value) + (-0.6037 × TRPM8 expression value). Patients were stratified into high- and low-risk groups using the median risk score. The high-risk group consisted of 124, 120, and 244 cases in the training, testing, and the entire TCGA-LUSC sets, respectively. The low-risk group included 124, 127, and 251 cases (Figure 1G-I). Kaplan-Meier analysis demonstrated significantly better OS in the low-risk group (p <0.001; Figure 1G-I), with risk scores positively correlated with mortality (Figure 1J-Q). The expression levels of the four genes in the indicated groups were shown in Figure 1P-R.

Univariate and multivariate Cox regression analysis revealed the SRS risk model as an independent predictor for clinical outcome (Figure 2A, B). Its AUC was 0.619, which also proved that it could better predict the prognosis of LUSC compared with other clinical parameters (Figure 2C). The 1-, 3-, and 5-year ROC AUCs were 0.619, 0.632, and 0.610, which implied that SRS still had a good prognosis with the prolongation of time (Figure 2D). Notably, the SRS model effectively discriminated survival in the early-stage (I-II) patients (p <0.001) but not in the advanced- stage (III-IV) cases (p = 0.582; Figure 2E, F), reflecting increased tumour heterogeneity or confounding from aggressive late- stage treatments.

GSEA analysis showed a high-risk group enriched in immune- related pathways, while a low-risk group was associated with metabolic processes (e.g., butanoate metabolism, DNA replication, homologous recombination, nucleotide excision repair, and spliceosome; Figure 3A). The SASP refers to the enhancement of the secretory function of senescent cells. A variety of cytokines can be secreted by senescent cells and thus play an important regulatory role in vivo. These cytokines may also cause alterations in TME and thus contribute to the relapse and development of tumours.¹⁰ The results showed that different types of SASP were overexpressed in the high- risk group (Figure 3B). High levels of interleukin, growth factors, regulatory factors, as well as soluble or ablative receptors or ligands further confirm the presence of the obvious SASP in the high-risk group (Figure 3B).

It is noteworthy that upregulated SASP factors such as IL-6, CXCL8, and VEGF can contribute to immunosuppressive pro- perties.¹¹ For example, IL-6 can regulate the presence of immunosuppressive tumour-infiltrating lymphocyte populations with the tumour immune microenvironment (TIME).¹¹ Immune infiltration analysis (TIMER2.0 and TISIDB) revealed distinct patterns by the risk group: lower SRS scores correlated with higher CD4+T cell, CD8+T cell, and uncharacterised cell infiltration, while cancer-associated fibroblasts, macrophages, Tregs, and neutrophils showed positive correlations with SRS scores (Figure 3C). These data support the point that heightened cellular ageing reshapes an immunosuppressive TIME via the SASP.

Twenty-seven immune checkpoint genes showed differential expression between the high- and low-risk groups (Figure 4A), with significantly elevated CTLA-4 and PDCD-1 (clinically relevant immune checkpoint) in high-risk patients (Figure 4A, B). The survival analysis of four subgroups (stratified by the SRS and CTLA-4) showed longer OS in low-risk/low-CTLA-4 patients, with low-risk status conferring better outcomes even in the high-CTLA-4 group. TIDE software predicted higher immune surveillance escape probability and lower immunotherapy response in high-risk patients (Figure 4C),¹² indicating low SRS scores may correlate with improved immunotherapy responses.

High-risk patients had significantly higher TMB levels than low-risk counterparts (Figure 4D). In addition, after classifying patients based on SRS and TMB levels, their survival rates were compared. Comprehensive immune score (IC) and TMB score reclassified patients into four groups according to high or low IC and TMB scores: IC (↑)+TMB (↑); IC (↑)+TMB (↓); IC (↓)+TMB (↑); IC (↓)+TMB (↓). Figure 4E revealed that there were obvious differences in the survival among the above-mentioned four groups. To sum up, the combination of SRS and immune checkpoint expression or TMB could well predict the clinical results of patients, and a patient with a low SRS score might take advantage of the immunotherapy.

The Kaplan-Meier analysis of the IMvigor210 immunotherapy cohort showed improved OS in patients with low-SRS (Figure 5A), with further survival benefits in those with high CD274 (PD-L1), CTLA-4, LAG3, PDCD1, or TIGIT expression (Figure 5B-F). These data confirmed SRS combined with these immune checkpoints as promising predictors of ICI response in LUSC.

Figure 1: Identification of prognostic cellular senescence-associated DEGs in TCGA and internal validation of the SRS model for OS (including the three groups: training, testing, and TCGA-LUSC). (A) A total of 19 prognostic genes correlated with OS in LUSC patients were identified using univariate Cox regression. (B) Eight intersecting genes were identified compared with 126 DEGs and 19 prognostic genes. (C) Correlation network of the eight intersecting genes. (D-F) Cellular senescence-associated genes were identified at the minimum cross-validation point using LASSO regression. (G-I) Kaplan-Meier survival analysis with log-rank tests comparing OS between high- or low-risk groups in the three groups. (J-L) Risk score for all samples in the three groups. (M-O) Scatter plots for showing the survival status of all samples across the three groups. (P-R): Heatmaps detailing the expression levels of the four SRS genes in each group. Figure 2: Evaluation of the predictive accuracy of the SRS model using the entire TCGA-LUSC group. (A, B) Univariate (A) and multivariate (B) Cox regression analyses were performed to confirm the independent prognostic value of the SRS risk score. (C) AUC curves for the SRS risk score as well as other clinical parameters. (D) Time-dependent receiver operating characteristic (ROC) curves for 1-, 3-, and 5-year OS based on the SRS signature. (E, F) The OS of LUSC patients in stage I-II and stage III-IV according to the risk score. Figure 3: SRS-associated SASP and immune cell infiltration. (A) Pathways enriched in high- or low-risk gene groups. (B) Expression levels of SASP factors in high- or low-risk groups. (C) Correlation analyses between different immune cells infiltration and SRS risk scores estimated using CIBERSORT-ABS, EPIC, MCPCOUNTER, QUANTISEQ, TIMER, and XCELL. The box in the figure represents the p-value. Figure 4: SRS and immunotherapy in LUSC. (A) Expression levels of 28 immune checkpoint genes in the high- or low-risk groups. The number on top of each box in the boxplot is the p-value. (B) The OS of four sets of patients classified by SRS and CTLA-4. (C) Prediction of patients’ responses to either anti-CD274 or anti-CTLA4 therapy using TIDE online software. (D) Tumour mutational burden (TMB) survival analysis in LUSC. (E) Combined analysis of SRS and TMB survival analysis of LUSC. Figure 5: SRS showed predictive potential for immunotherapy response. (A) Survival analysis of patients with high or low SRS scores using the IMvigor210. (B-F). Survival analysis of patients classified into four groups based on SRS and the expression of CD274 (PD-L1): CTLA4, LAG3, PDCD1, and TIGIT (p <0.05).

DISCUSSION

Cellular ageing/senescence refers to the stagnation of cell cycle progression. While cellular ageing can promote tissue remodelling, regeneration, and functions, it may lead to inflammation and tumourigenesis in ageing organisms. Cellular senescence plays an essential role in various diseases, including cancer and age-related disorders, by altering biological functions.¹ Increasing evidence suggests that the clearance of senescent cells is caused by innate and adaptive immunity through SASP-induced immune responses.⁴ Nevertheless, it remains unclear how immune cell infiltration is affected by senescent/aged cells, and what role senescent cells play in assessing clinical efficacy.

In this study, the expression patterns, prognostic value, and the TIME of cellular ageing-associated genes in LUSC were analysed in detail. A novel survival prediction model, SRS, was established in the training cohort according to the expression of four ageing-associated genes from the TCGA database and verified in the test cohort. The characteristics of the TIME, such as immune cell distribution and inflammatory activity, were also explored across different SRS score groups. Importantly, TIME remodelling via SASP was regarded as a potential mechanism of immune escape, tumour development, and metastasis. In addition, the SRS score was found to be an independent predictor in patients with LUSC and could be combined with specific IC or TMB to predict immunotherapy response to ICI.

Notably, the SRS model differs from and improves upon previously developed senescence-related prognostic models for LUSC. For example, the model constructed by Hu et al. includes cell cycle-related genes (e.g., CDKN1A) but lacks genes directly regulating the SASP (e.g., SERPINE1) and ion channels (e.g., TRPM8) involved in senescence-immune crosstalk.⁵ In contrast, the SRS model integrates two pro- senescence genes (IGFBP1, SERPINE1) and two anti-senescence genes (SIX1, TRPM8), reflecting the dual regulatory role of senescence in LUSC. In terms of predictive performance, the 5-year AUC of SRS (0.610) is higher than that of Hu et al.’s model (0.57), indicating higher long-term prognostic accuracy. Furthermore, Hu et al. do not explore SASP- mediated TIME remodelling, whereas the present study confirms that high SRS scores are associated with elevated SASP factors (IL-6, CXCL8) and increased infiltration of immunosuppressive cells (Tregs, M2 macrophages; Figure 3B, C), providing a clear mechanism for senescence-driven immune escape. This study also further incorporates the TIDE score, a validated predictor of immunotherapy response. The results show that low-risk patients have lower TIDE scores, which are directly associated with better responses to ICI. The model developed by Li et al., based on DNA methylation (epigenetic level), can only predict prognosis, lacking analyses related to immunotherapy.¹³ Whereas the SRS model, based on the senescence-associated gene expression (transcriptional level), directly reflects the functional status of senescent cells. And by integrating immune checkpoint expressions (e.g., CTLA-4, PD-L1) and TMB, it can guide ICI selection, addressing the unmet need for biomarkers in immunotherapy for LUSC. Huang et al. established a glycolysis-immune score (GIS) model with 2 glycolysis- related genes (GRGs: PYGB, MDH1) and 3 immune-related genes (IRGs: TSLP, SERPIND1, GDF2), focusing on metabolic- immune crosstalk to predict LUSC prognosis and therapy sensitivity.¹⁴ Whereas the SRS model centres on senescence- SASP-immune interactions. Additionally, the SRS model clari- fies SASP-mediated immune remodelling, an aspect not addressed in Huang et al.'s study.¹⁴

The risk model in the present study consists of four senescence-related genes, including IGFBP1, SERPINE1, SIX1, and TRPM8. As a downstream protein of Jagged1, IGFBP1 (a 30 kDa protein) is related to the severity of coronary atherosclerosis in ageing patients and serves as a promising biomarker for lung cancer diagnosis.¹⁵ Previous studies demonstrated that IGFBP1 protects endothelial cells against passage-induced senescence via Akt signalling in cell culture studies.¹⁶ SERPINE1 is highly expressed in senescent cells and induces the senescence of lung epithelial cells via elevation of p53 and p21 levels.¹⁷ Silencing SERPINE1 reverses senescent phenotypes, protects organ structure and functions, and even lengthens the lifespan of mice.¹⁸ SIX1 was reported to influence cellular ageing via the modulation of p16INK4A, as well as differentiation-associated genes. TRPM8 belongs to the TRP superfamily of ion channels, which usually plays a role in various physiological reactions.

The TME refers to the surrounding environment during the process of tumour formation and growth. Normally, immune cells in the TME can identify and eliminate tumour cells in time and play a role in promoting tumour or anti-tumour immunity during tumour growth. However, tumour cells can also escape the recognition and attack of body immunity through a variety of mechanisms to produce immunosuppressed TME, thus achieving immune escape. Immunosuppression is closely related to immunosuppressive cells, such as regulatory T-cells (Tregs), tumour-associated macro-phages, and neutrophils, as well as inhibitory cytokines.¹⁹ Studies have shown that Tregs can inhibit CD80 and CD86 co-stimulatory signals through CTLA4, secrete inhibitory cytokines, and kill effector T-cells.²⁰ By secreting chemo-kines, e.g. CCL3 and CCL20, tumour-associated macro-phages enrol Tregs and thus constitute immunosuppressive TME.²¹ This study shows that immunosuppressive cells, including tumour-associated neutrophils, macrophages, and regulatory T-cells, have higher levels of infiltration, which also indicates that the immunosuppressive microenvironment is more active. Spearman’s correlation coefficient showed the positive association between risk scores and neutrophils, cancer-associated fibroblasts, endothelial cells, macrophages, Treg, and NK cells, suggesting that a patient with high SRS score might have an immunosuppressive TME that prevents immune cells from clearing tumour cells.

GSEA also identified the most significant enrichment path-way between different risk groups, with significant changes in the immunological pathway of high-risk group patients. In addition, the combination of SRS and immune checkpoint expression or TMB can well predict the clinical results, and a patient with a low SRS score may profit from the immuno-therapy. Moreover, this study further explores the principle of immune remodelling caused by senescent cells in cancer development. The results showed the acceleration of SASP on tumour growth by promoting immune escape, which was formed by affecting the establishment of TIME. Pro-inflammatory cytokines were increased by high SRS, including IL-1β, IL-6, IL-1α, IL-7, and IL-16. Growth factors, such as amphiregulin (AREG), epiregulin (EREG), and insulin-like growth factor binding protein (IGFBP), along with matrix metalloproteinase, may regulate the recruitment of immune cells within the TME, thereby transforming the local tissue microenvironment into one that promotes tumour growth.²² Furthermore, as mentioned above, immune remodelling related to cellular senescence may lead to poor efficacy in blocking immune checkpoints.

The present study shows that patients with a low SRS score have a survival advantage. Therefore, it is necessary to strengthen clinical surveillance of high-risk patients to control tumour recurrence and prevent tumour progression. To date, microsatellite status, TIDE, and TMB scores have been recognised as key factors impacting the treatment response to ICI therapy.²³ This study reveals that the TMB level in the low-risk set was higher than that in the high-risk set, which also implies that better response to ICI treatment exists for patients in the low-risk set. Consistent with this, a clinical study on locally advanced NSCLC showed that combined therapy with intensity-modulated radiotherapy (IMRT) and Camrelizumab (a PD-1 inhibitor) significantly improved the levels of CD3+ and CD4+ T cells, enhanced objective response rate (70% vs. 47.5%), and improved physical status without increasing adverse reactions. This suggests that ICI-based combination therapy can effectively restore anti-tumour immune function, especially in patients with a favourable immune microenvironment, such as those with low SRS score.²⁴ The TIDE score, a new biomarker of immunotherapy, serves as a predictor of response to ICI. A higher TIDE score indicates a greater likelihood of tumour immune escape and reduced response rate to immuno-therapy.^12,25 In this study, the TIDE score of LUSC patients with a low-risk score is decreased, which indicates that patients with low-risk score have a high immunogenicity, and that means these patients can benefit more from the ICI treatment.

However, this research has several limitations. Firstly, the TCGA-LUSC cohort primarily includes Western populations, and the generalisability of the SRS to Asian or other ethnic groups remains untested. Therefore, validation in multi- ethnic cohorts—such as Chinese LUSC databases—is warranted. Secondly, the analysis relies on bulk RNA-seq, which masks cell- type-specific senescence signatures. Single-cell RNA sequen-cing would better resolve the interactions between senescent cells and immune components in the TME. Thirdly, treatment history, such as prior chemotherapy or radio- therapy, was not considered, despite its potential to induce senescence and alter the SASP profiles, which may confound risk stratification. Finally, while correlations between the SRS and TME features were observed, the causal mechanisms—such as whether SERPINE1 directly promotes Treg recruitment—require validation through in vitro and in vivo functional experiments. Furthermore, the prognostic efficacy and underlying mechanisms of this model also require further investigation using real-world clinical data and basic experiments.

CONCLUSION

In this study, a risk model related to cellular senescence was established for predicting the prognosis of patients with LUSC. The model can be used as an evaluation index of TIME immune cell infiltration and has independent prognostic significance for patients. Moreover, it was found that the SRS was correlated with the infiltration of immune cells and the regulation of SASP in the immune microenvironment. Additionally, this study shows that the SRS scores have the possibility to combine with ICI, which has important clinical implications for patients receiving checkpoint inhibitor-based immuno-therapy. Therefore, the identification of cellular senescence prognostic genes, the establishment of prognostic models, and the prediction of immunotherapy response are significant for LUSC diagnosis and clinical treatment in the future.

ETHICAL APPROVAL:
Ethical approval was obtained from the Institutional Ethics Committee of Hubei Cancer Hospital affiliated to Tongji Medical College, Huazhong University of Science and Technology, Hubei, China (Approval No. LLHBCH2024YN-108).

PATIENTS’ CONSENT:
The data for this study were all sourced from the TCGA database, and no real-world patient data were collected.

COMPETING INTEREST:
The authors declared no conflict of interest.

AUTHORS’ CONTRIBUTION:
KK: Wrote the initial draft of the manuscript.
YM: Collected and organised data and collected materials.
SW: Conducted data analysis and wrote the manuscript.
KK, XL: Designed the flowchart of the study.
KK, SW: Contributed equally to this work.
All authors approved the final version of the manuscript to be published.

REFERENCES

1. Hernandez-Segura A, Nehme J, Demaria M. Hallmarks of cellular senescence. Trends Cell Biol 2018; 28(6):436-53. doi: 10.1016/j.tcb.2018.02.001.
2. Braig M, Schmitt CA. Oncogene-induced senescence: putting the brakes on tumour development. Cancer Res 2006; 66(6): 2881-4. doi: 10.1158/0008-5472.CAN-05-4006.
3. Coppe JP, Desprez PY, Krtolica A, Campisi J. The senescence-associated secretory phenotype: The dark side of tumour suppression. Annu Rev Pathol 2010; 5:99-118. doi: 10.1146/annurev-pathol-121808-102144.
4. Schneider JL, Rowe JH, Garcia-de-Alba C, Kim CF, Sharpe AH, Haigis MC. The aging lung: Physiology, disease, and immunity. Cell 2021; 184(8):1990-2019. doi: 10.1016/j.cell. 2021.03.005.
5. Hu X, Guo L, Liu G, Dai Z, Wang L, Zhang J, et al. Novel cellular senescence-related risk model identified as the prognostic biomarkers for lung squamous cell carcinoma. Front Oncol 2022; 12:997702. doi: 10.3389/fonc.2022.997702.
6. Fu XL, Duan W, Su CY, Mao FY, Lv YP, Teng YS, et al. Interleukin 6 induces M2 macrophage differentiation by STAT3 activation that correlates with gastric cancer progression. Cancer Immunol Immunother 2017; 66(12):1597-608. doi: 10.1007/ s00262-017-2052-5.
7. Wang TW, Johmura Y, Suzuki N, Omori S, Migita T, Yamaguchi K, et al. Blocking PD-L1-PD-1 improves senescence surveillance and ageing phenotypes. Nature 2022; 611(7935): 358-64. doi: 10.1038/s41586-022-05388-4.
8. Siegel RL, Miller KD, Wagle NS, Jemal A. Cancer statistics, 2023. CA Cancer J Clin 2023; 73(1):17-48. doi: 10.3322/caac. 21763.
9. Fu J, Li K, Zhang W, Wan C, Zhang J, Jiang P, et al. Large-scale public data reuse to model immunotherapy response and resistance. Genome Med 2020; 12(1):21. doi: 10.1186/s13073- 020-0721-z.
10. Acosta JC, O'Loghlen A, Banito A, Guijarro MV, Augert A, Raguz S, et al. Chemokine signaling via the CXCR2 receptor reinforces senescence. Cell 2008; 133(6):1006-18. doi: 10.1016/j. cell.2008.03.038.
11. Kato T, Noma K, Ohara T, Kashima H, Katsura Y, Sato H, et al. Cancer-associated fibroblasts affect intratumoural CD8+ and FoxP3+ T cells via IL6 in the tumour microenvironment. Clin Cancer Res 2018; 24(19):4820-33. doi: 10.1158/1078-0432. CCR-18-0205.
12. Jiang P, Gu S, Pan D, Fu J, Sahu A, Hu X, et al. Signatures of T cell dysfunction and exclusion predict cancer immunotherapy response. Nat Med 2018; 24(10):1550-8. doi: 10.1038/s41 591-018-0136-1.
13. Li XS, Nie KC, Zheng ZH, Zhou RS, Huang YS, Ye ZJ, et al. Molecular subtypes based on DNA methylation predict prognosis in lung squamous cell carcinoma. BMC Cancer 2021; 21(1):96. doi: 10.1186/s12885-021-07807-7.
14. Huang Q, Yang S, Yan H, Chen H, Wang Y, Wang Y. Development and validation of a combined glycolysis and immune prognostic signature for lung squamous cell carcinoma. Front Genet 2022;13:907058. doi: 10.3389/fgene.2022. 907058.
15. Li F, Cao Y, Li J, Gao C, Dong X, Ren P, et al. The clinical significance of serum adipocytokines level in patients with lung cancer. J Thorac Dis 2019; 11(8):3547-55. doi: 10. 21037/jtd.2019.07.66.
16. Wu X, Zheng W, Jin P, Hu J, Zhou Q. Role of IGFBP1 in the senescence of vascular endothelial cells and severity of aging-related coronary atherosclerosis. Int J Mol Med 2019; 44(5):1921-31. doi: 10.3892/ijmm.2019.4338.
17. Jiang C, Liu G, Luckhardt T, Antony V, Zhou Y, Carter AB. Serpine 1 induces alveolar type II cell senescence through activating p53-p21-Rb pathway in fibrotic lung disease. Aging Cell 2017; 16(5):1114-24. doi: 10.1111/acel.12643.
18. Eren M, Boe AE, Murphy SB, Place AT, Nagpal V, Morales-Nebreda L, et al. PAI-1-regulated extracellular proteolysis governs senescence and survival in Klotho mice. Proc Natl Acad Sci USA 2014; 111(19):7090-5. doi: 10.1073/pnas. 1321942111.
19. Li L, Yu R, Cai T, Chen Z, Lan M, Zou T, et al. Effects of immune cells and cytokines on inflammation and immuno- suppression in the tumour microenvironment. Int Immunopharmacol 2020; 88:106939. doi: 10.1016/j.intimp  .2020.106939.
20. Tekguc M, Wing JB, Osaki M, Long J, Sakaguchi S. Treg- expressed CTLA-4 depletes CD80/CD86 by trogocytosis, releasing free PD-L1 on antigen-presenting cells. Proc Natl Acad Sci USA 2021; 118(30):e2023739118. doi: 10.1073/ pnas.2023739118.
21. Cassetta L, Pollard JW. Tumour-associated macrophages. Curr Biol 2020; 30(6):R246-8. doi: 10.1016/j.cub.2020. 01.031.
22. Cuollo L, Antonangeli F, Santoni A, Soriani A. The senescence-associated secretory phenotype (SASP) in the challenging future of cancer therapy and age-related diseases. Biology (Basel) 2020; 9(12):485. doi: 10.3390/biology 9120485.
23. Zeng C, He R, Dai Y, Lu X, Deng L, Zhu Q, et al. Identification of TGF-β signaling-related molecular patterns, construction of a prognostic model, and prediction of immunotherapy response in gastric cancer. Front Pharmacol 2022; 13: 1069204. doi: 10.3389/fphar.2022.1069204.
24. Xu JK. Effect of intensity modulated radiotherapy (IMRT) on the immunity, physical status and clinical effect of locally advanced NSCLC patients. Pak J Med Sci 2021; 37(5): 1480-5. doi: 10.12669/pjms.37.5.4188.
25. Xia Y, Zhang R, Wang M, Li J, Dong J, He K, et al. Development and validation of a necroptosis-related gene prognostic score to predict prognosis and efficiency of immuno-therapy in gastric cancer. Front Immunol 2022; 13:977338. doi: 10.3389/fimmu.2022.977338.