Genomic and Immunological Characterization of Pyroptosis in Lung Adenocarcinoma

Abstract

Pyroptosis is a programmed cell death that may either promote or hinder cancer growth under different circumstances. Pyroptosis-related genes (PRGs) could be a useful target for cancer therapy, and are uncommon in lung adenocarcinoma (LUAD). The expression profiles, mutation data and clinical information of LUAD patients were included in this study. A pyroptosis-related prognostic risk score (PPRS) model was constructed by performing Cox regression, weighted gene co-expression network analysis (WGCNA), and least absolute shrinkage and selection operator (LASSO) analysis to score LUAD patients. Somatic mutation and copy number variation (CNV), tumor immunity, and sensitivity to immunotherapy/chemotherapy were compared between different PPRS groups. Clinical parameters of LUAD were combined with PPRS to construct a decision tree and nomogram. Red module was highly positively correlated with pyroptosis. Seven genes (FCRLB, COTL1, GNG10, CASP4, DOK1, CCR2, and AQP8) were screened from the red module to construct a PPRS model. Significantly lower overall survival (OS), higher incidence of somatic mutation and CNV, elevated infiltration level of the immune cell together with increased probability of immune escape were observed in LUAD patients with higher PPRS, and were more sensitive to Cisplatin, Docetaxel, and Vinorelbine. We constructed a new PPRS model for patients with LUAD. The model might have clinical significance in the prediction of the prognosis of patients with LUAD and in the efficacy of chemotherapy and immunotherapy.

Figure 1

Genetic and expression variations of PRGs in TCGA cohort. (a) The waterfall diagram shows the somatic mutations of 27 PRGs from the LUAD sample of TCGA. (b) Kaplan-Meier survival plot of two groups with mutant and wild type (WT) PRGs. (c) GSEA of hallmark pathways by comparing mutant group to WT group in LUAD samples. (d) The CNV fraction of PRGs in LUAD samples. (e) Comparison of CNV difference in 27 PRGs in LUAD samples. (f) Comparison of expression of 27 PRGs between normal and LUAD samples. Log-rank test was performed in (b). Kruskal-Wallis test was performed in (e) and Wilcoxon test was performed in (f). ns, no significance. ^∗P < 0.05, ^∗∗P < 0.01, ^∗∗∗P < 0.001, ^∗∗∗∗P < 0.0001.

Figure 2

Identification of modules related to pyroptosis by WGCNA. (a) Hierarchical clustering of the genes of LUAD samples in the clustering analysis. (b) Analysis of network topology for various soft-thresholding powers. (c) The dendritic map of gene clusters generated by the dynamic tree cut package. (d) The number of genes contained in each module. (e) Correlation analysis of module-pyroptosis. The upper part represents the hierarchical clustering of the whole module, and the lower part represents the correlation between the module and pyroptosis. (f) The scatter plot of module membership (MM) and gene significance (GS) for pyroptosis in the red module.

Figure 3

Construction and evaluation of PPRS model. (a) A total of 73 promising candidates were identified among hub genes extracted from the red module. (b) 10 of the 73 genes were retained by application of LASSO-Cox regression model with a minimum of λ (λ = 0.0295). (c) FCRLB, COTL1, GNG10 and CASP4 were risk factors, while DOK1, CCR2 and AQP8 were protective factors. (d) PPRSs and corresponding living state of the samples obtained in ascending order and expression of 7 genes of the samples. (e) The survival rate of the high-PPRS group and the low-PPRS group in the TCGA-LUAD cohort. (f) The ROC curve of the PPRS model in the TCGA-LUAD cohort. (g) In cohorts GSE31210, the difference in OS and predictive efficacy was validated. (h): The difference in OS and predictive efficacy was validated for GSE31210 cohort. Log-rank test was conducted in (e g, and h).

Figure 4

Genomic mutation in PPRS risk group. (a) Aneuploidy score, homologous recombination defects, fraction altered, number of segments and tumor mutation burden of high-PPRS group and low-PPRS group were compared by Wilcox test. (b) The relation between PPRS and aneuploidy score, homologous recombination defects, fraction altered, number of segments, tumor mutation burden, respectively. (c) The waterfall map shows the incidence of somatic mutation and copy number variation in the high-PPRS group and the low-PPRS group. ^∗∗P < 0.01, ^∗∗∗P < 0.001, ^∗∗∗∗P < 0.0001.

Figure 5

Enrichment pathway and immune characteristics of PPRS risk group. (a) Correlation analysis matrix between KEGG pathway and PPRS. (b) The NESs of high-PPRS group was higher than that of low-PPRS group. (c) The relative proportion of immune cells in TCGA-LUAD between high-PPRS group and low-PPRS group. (d) The array of Pearson correlation analysis showed the correlation between PPRS in TCGA-LUAD and 22 kinds of immune cells. (e) Stromal score, immune score and ESTIMATE score of high-PPRS group and low-PPRS group in TCGA-LUAD cohort. Wilcoxon test was conducted. ns, no significance. ^∗P < 0.05, ^∗∗P < 0.01, ^∗∗∗P < 0.001, ^∗∗∗∗P < 0.0001. ? indicates undetected expression levels.

Figure 6

Prediction of response to immunotherapy and chemotherapy in patients with different PPRS. (a-c) The expression of 21 immune checkpoints between high-PPRS group and low-PPRS group in TCGA-LUAD (a), GSE31210 (b) and GSE72094 (c) cohorts. (d) The relative frequency of immune checkpoints expressed highest in the high-PPRS group. (e) Differences for samples with different PPRS in myeloid-derived suppressor cells (MDSC), cancer associated fibroblasts (CAF), M2 macrophages (TAM.M2), T cell exclusion, T cell dysfunction, TIDE scores. (f) Comparison for sensitivity of Paclitaxel, Cisplatin, Docetaxel and Vinorelbine in high- and low-PPRS groups of LUAD samples. Wilcoxon test was conducted. ns, no significance. ^∗P < 0.05, ^∗∗P < 0.01, ^∗∗∗P < 0.001, ^∗∗∗∗P < 0.0001.

Figure 7

Establishment and evaluation of the decision tree and nomogram. (a) Decision trees were constructed based on age, gender, T stage, N stage, M stage, clinical stage and PPRS. (b) The survival curve of the four subgroups defined by the decision tree. (c) The distribution of PPRS in the four subgroups defined by the decision tree. (d) The proportion of patient survival and death in the four subgroups defined by the decision tree. (e) Univariate and multivariate Cox regression analysis of clinical characteristics and PPRS. (f) Nomogram constructed by combining clinical features with PPRS. (g) The calibration curve showed the consistency between the nomogram predicted OS and the actual OS in the TCGA cohort. (h) DCA of the PPRS and clinicopathological features. (i) ROC curve for clinicopathological features and nomogram. ANOVA was conducted in (c and d). Log-rank test was conducted in (e). ^∗P < 0.05.