Implications of different cell death patterns for prognosis and immunity in lung adenocarcinoma

Abstract

In recent years, lung adenocarcinoma (LUAD) has become a focus of attention due to its low response to treatment, poor prognosis, and lack of reliable indicators to predict the progression or therapeutic effect of LUAD. Different cell death patterns play a crucial role in tumor development and are promising for predicting LUAD prognosis. From the TCGA and GEO databases, we obtained bulk transcriptomes, single-cell transcriptomes, and clinical information. Genes in 15 types of cell death were analyzed for cell death index (CDI) signature establishment. The CDI signature using necroptosis + immunologic cell death-related genes was established in the TCGA cohort with the 1-, 2-, 3-, 4- and 5-year AUC values were 0.772, 0.736, 0.723, 0.795, and 0.743, respectively. The prognosis was significantly better in the low CDI group than in the high CDI group. We also investigated the relationship between the CDI signature and clinical variables, published prognosis biomarkers, immune cell infiltration, functional enrichment pathways, and immunity biomarkers. In vitro assay showed that HNRNPF and FGF2 promoted lung cancer cell proliferation and migration and were also involved in cell death. Therefore, as a robust prognosis biomarker, CDI signatures can screen for patients who might benefit from immunotherapy and improve diagnostic accuracy and LUAD patient outcomes.

Fig. 1. The 15 types of cell death-related gene expression in high and low-risk groups and associations of cell death-related genes with LUAD prognosis using multivariate Cox regression.

Red represented the high-risk group while blue represented the low-risk group. *P < 0.05, **P < 0.01, ***P < 0.001.

Fig. 2. The prognosis value of CDI signature in LUAD.

a, b A comparison of 1-, 2-, 3-, 4-, and 5-year AUC values among 15 types of cell death signature showed the superiority of CDI signature. c A comparison of ROC curve in CDI signature with necroptosis and immunologic cell death signature. d The 1-, 3-, and 5-year ROC curve of CDI signature suggested that all AUC values were over 0.70. e–g The Kaplan–Meier survival curve with log-rank test demonstrated the relationship between OS and CDI signature, necroptosis, and immunologic cell death signature, respectively. h–j Clustering analysis showed gene classification in high and low-risk groups based on CDI signature, necroptosis, and immunologic cell death signature, respectively. k Diagram of gene interaction network showed that CDI signature was tightly associated with necroptosis and immunologic cell death-related genes. l The distribution of survival status based on necroptosis, immunologic cell death signature and CDI signature.

Fig. 3. The interaction of CDI signature with clinical variables.

a, b Univariate and multivariate Cox regression models for associations of risk score and clinicopathological factors with LUAD prognosis. c A nomogram consisting of CDI and N stage for predicting 1-, 3-, and 5- years survival for LUAD patients. d–f The 1-, 3-, and 5-year ROC curve of the nomogram, CDI, and N stage. g The CDI of LUAD histological phenotypes including adenocarcinoma with mixed subtypes, adenocarcinoma (NOS), bronchiolo-alveolar carcinoma, and papillary adenocarcinoma. h–k The K–M curve between high and low CDI group in adenocarcinoma with mixed subtypes, adenocarcinoma (NOS), bronchiolo-alveolar carcinoma, and papillary adenocarcinoma, respectively. l–o The K–M curve between high and low CDI group in LUAD with I, II, III, and IV stages. *P < 0.05, **P < 0.01, ***P < 0.001.

Fig. 4. Comparison with other risk signatures.

a ROC curve of Necroptosis signature, ICD signature, Pyroptosis signature, Ferroptosis signature. b K–M survival curve of the four signatures. c 1–5 years AUC values of CDI signature and four signatures. d C-index of our CDI signature compared with the four other signatures. e Restricted mean survival (RMS) curves for the five risk signatures.

Fig. 5. Correlation of CDI with immune cell infiltration.

a The Wilcoxon rank-sum test compared the absolute abundance scores of 8 immune cells and 2 stromal cell populations in high and low CDI groups using the MCP counter. b The relationship of CDI with 8 immune cells and 2 stromal cell populations using MCP counter. c The difference of 22 immune cell infiltration levels between high and low CDI groups using the CIBERSORT algorithm. d The relationship of CDI with 22 immune cells using the CIBERSORT algorithm. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

Fig. 6. The investigation of the immune landscape in high and low CDI groups using several algorithms.

a The correlation of CDI with infiltration level of immune cells using CIBERSORT, EPIC, QUANTISEQ, TIMER, and XCELL. b The difference in immune cell infiltration level between high and low CDI groups using CIBERSORT, EPIC, QUANTISEQ, TIMER, and XCELL. c–f The difference of ESTIMATE Score, IMMUNE Score, Stromal Score, and Tumor Purity between high and low CDI groups, respectively. g–j The relationship of CDI with ESTIMATE Score, IMMUNE Score, Stromal Score, and Tumor Purity, respectively. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

Fig. 7. Correlation of CDI with immunotherapy-related biomarkers.

a The difference in TIDE score between high and low CDI groups. b The relationship of CDI with TIDE score. c The percentage of immunotherapy responders and non-responders in high and low CDI groups using the TIDE score. d The difference of TMB between high and low CDI groups. e, h, k The Kaplan–Meier survival curve showed the relationship between OS and CDI in patients receiving immunotherapy with LUAD from GSE135222, metastatic urothelial carcinoma from IMvigor210 and melanoma from GSE78220, respectively. f, i, l The percentage of immunotherapy responders and non-responders in high and low CDI groups in patients receiving immunotherapy with LUAD from GSE135222, metastatic urothelial carcinoma from IMvigor210 and melanoma from GSE78220, respectively. g, j, m The difference of CDI between immunotherapy responders and non-responders in patients from GSE135222, IMvigor210, and GSE78220, respectively. n The difference of IPS between high and low CDI groups. o The relationship of CDI, MHC, EC, SC, and IPS in the TCGA cohort. *P < 0.05, **P < 0.01, ***P < 0.001.

Fig. 8. Validation of the potential function of HNRNPF and FGF2 in tumors by in vitro assays.

a HNRNPF and FGF2 expression in five paired tumor tissues (T) and their adjacent normal tissues (N). b Comparison of HNRNPF and FGF2 expressions in human bronchial epithelial and lung cancer cells. c The expression of HNRNPF and FGF2 in 1299 and A549 cells, respectively, after transferring with siRNA. d–f Knockdown of HNRNPF and FGF2 inhibited the proliferation and migration of lung cancer cells by CCK-8 assay d, EDU assay e, and wound-healing assay (f). g The effects of cisplatin treatments on necroptosis in the 1299 cells were determined, and cisplatin-induced necroptosis was further increased in the HNRNPF silencing 1299 cells. h The effects of oxaliplatin treatments on ICD in the A549 cells were determined, and oxaliplatin-induced ICD was further enhanced in the FGF2-silencing A549 cells. *P < 0.05, **P < 0.01, ***P < 0.001.