Pyroptosis-Related Gene Model Predicts Prognosis and Immune Microenvironment for Non-Small-Cell Lung Cancer

Abstract

Non-small-cell lung cancer (NSCLC) has a high incidence and mortality worldwide. Moreover, it needs more accurate means for predicting prognosis and treatments. Pyroptosis is a novel form of cell death about inflammation which was highly related to the occurrence and development of tumors. Despite having some studies about pyroptosis-related genes (PRGs) and cancer, the correlation has not been explored enough between PRGs and immune in NSCLC. In this study, we constructed a PRG model by WGCNA to access the prognosis value PRGs have. The testing cohort (n = 464) with four datasets from the GEO database conducted a survival analysis to confirm the stability of the prognostic model. The risk score and age are examined as independent prognostic factors. Based on the PRGs, we found multiple pathways enriched in immune in NSCLC. Separating samples into three subtypes by consensus cluster analysis, Cluster 3 was identified as immune-inflamed phenotype with an optimistic prognostic outcome. A three-gene PRG signature (BNIP3, CASP9, and CAPN1) was identified, and BNIP3 was identified as the core gene. Knockdown of BNIP3 significantly inhibited the growth of H358 cells and induced pyroptosis. In conclusion, the model construction based on PRGs provides novel insights into the prediction of NSCLC prognosis, and BNIP3 can serve as a diagnostic biomarker for NSCLC.

Figure 1

The flow chart of the overall study.

Figure 2

Coexpression network construction and identification of module related with clinical traits in non-small-cell lung cancer (NSCLC) patients. (a) Analysis of the scale-free fit index for different soft-thresholding powers. (b) Analysis of the correlation between mean connectivity and various soft-thresholding powers. (c) Dendrogram of the genes after merging modules and reclustering. Every color represents a module. (d) The relationships between modules and clinical characteristics of NSCLC; red means positive correlation, and blue means negative correlation. (e) Analysis of gene significance for module membership in the dark green module. The red line is a fitted curve. (f) The PPI network of pyroptosis-related genes in the “MEdarkgreen” module. The color and size of genes represent the degree value (database annotated). The big blue dots represent genes with a high degree, the small yellow dots are the genes with a low degree, and pink dots are the genes without a degree. The grey lines links dots represent the combined score and the thicker with higher relevance.

Figure 3

The prognostic genes identification and the correlation between risk score and survival condition with validation. (a) The forest plot of prognostic genes is based on the univariate Cox regression analysis. HR: hazard ratio; HR.95L, HR.95H: hazard ratio 95% confidence interval. (b) Distribution of LASSO coefficients of the three prognostic genes in the training cohort. (c) Selection of the best parameter (lambda) in the lambda sequence. (d) The distribution of risk groups based on risk score in samples in the training cohort. (e) The survival status of NSCLC patients in the training cohort in different risk groups. (f) The expression level of prognostic genes in the high- and low-risk groups in the training cohort. (g–i) Validation for aforementioned analyses about risk score based on GEO dataset.

Figure 4

Survival analysis for training cohort and testing cohort. (a) Kaplan-Meier survival analysis of samples in different risk groups in the training cohort. (b) ROC curve for 1, 3, and 5 years of survival time prediction. (c, d) The testing of survival analysis in validation dataset from GEO database. (e) ROC curve for comparing the prognostic model with other published models.

Figure 5

Construction of nomogram module and calibration. (a) Nomogram construction with independent prognostic factors (^∗p value < 0.05; ^∗∗p value < 0.01). (b) The Schoenfeld residuals of two clinical factors (age and risk scores) for proportional hazard assumption to test the nomogram. (c–e) The calibration curve for the nomogram model with 1, 5, and 8 years, respectively. OS: overall survival.

Figure 6

Gene set enrichment analysis. (a) Gene set enrichment analysis (GSEA) for “MEdarkgreen” module genes. (b) GSEA for all pyroptosis-related genes based on training cohort. (c) GSEA for pyroptosis-related genes in the “MEdarkgreen” module. (d) GSEA for genes after intersecting the “MEdarkgreen” module and validation cohort.

Figure 7

Identification of immune subtype and KEGG (Kyoto Encyclopedia of Genes and Genomes) function enrichment analysis. (a) The relationship between CDF (cumulative distribution function) area and consensus index. (b) The relative change in area under CDF curve for k-means. (c) The survey of consensus matrix when k = 3 and sample distribution. (d) Survival analysis of immune subtype in the training cohort. C1: cluster 1; C2: cluster 2; C3: cluster 3. (e, f) Kaplan-Meier survival analysis of cluster 1 and cluster 2. (g–i) KEGG functional enrichment analysis for three clusters, respectively.

Figure 8

Immune infiltration analysis. (a) ESTIMATE analysis of training cohort with risk and immune subgroups. C1: cluster 1; C2: cluster 2; C3: cluster 3. (b, c) The enrichment situation of immune cells in training cohort (b) and testing cohort (c) by ssGSEA (ns: not significant; ^∗p value < 0.05; ^∗∗p value < 0.01; and ^∗∗∗p value < 0.001). (d) CIBERSORT analysis of entire training cohort. (e, f) The enrichment analysis result of immune functions in training cohort (e) and validation dataset (f) after ssGSEA (ns: not significant; ^∗p value < 0.05; ^∗∗p value < 0.01; and ^∗∗∗p value < 0.001).

Figure 9

Knockdown of BNIP3 induces pyroptosis in lung cancer cells. (A) 656, 683, 703, 714, and 715 are patient numbers, immunohistochemical assay to compare BNIP3 expression in normal lung tissue and lung cancer tissue. (b) The expression of normal lung tissue protein and lung cancer tissue protein expression were detected by western blotting assay. Densitometry of the ration of BNIP3 was shown as bar chart. All data were representative of at least three independent experiments and presented as mean ± SD, ^∗∗p < 0.01 compared with the control. 723, 719, 683, and 733 are patient numbers. (c) Using flow cytometry to detect cell death of H358 after knockdown BNIP3. (d) Pyroptosis-related proteins were detected by western blotting assay. Densitometry of the ration of protein was shown as bar chart. All data were representative of at least three independent experiments and presented as mean ± SD, ^∗p < 0.05, ^∗∗p < 0.01 compared with the control.