Pyroptosis-related gene signature: A predictor for overall survival, immunotherapy response, and chemosensitivity in patients with pancreatic adenocarcinoma

Abstract

Background: Pancreatic adenocarcinoma (PAAD) is a lethal malignancy with high levels of heterogeneity. Pyroptosis is thought to influence the development of various tumors. Nevertheless, the role of pyroptosis-related genes (PRGs) in prognostic risk stratification and therapeutic guidance for PAAD remains ambiguously.

Methods: Transcriptome profile and clinical information of PAAD patients were retrieved from The Cancer Genome Atlas (TCGA) as well as Gene Expression Omnibus (GEO) databases, followed by differential analysis. Patients were divided into distinct pyroptosis phenotype subtypes based on the characteristic of differently expressed PRGs (DEPRGs). Then a PRG signature was established through univariate analysis and LASSO algorithm in the training set to assess the prognostic risk, and its reliability was verified in the validation set using receiver operating characteristic(ROC) curve. The correlation of risk score with tumor microenvironment(TME), TMB and chemotherapeutic drug sensitivity were also analyzed. In addition, a nomogram was constructed to promote better clinical application.

Results: A total of 28 DEPRGs were determined in the integrated TCGA-GEO datasets. Patients were divided into three pyroptosis phenotype subtypes, Kaplan-Meier curve suggested patients in cluster B had a worse prognosis than those in cluster A and C. Then a price signature comprised of 8 PRGs was generated. TME analysis suggested that the low-risk subgroup displayed potential stronger antitumor immune effect and might respond better to immune checkpoint inhibitors (ICIs) therapy. Furthermore, PRG signature exhibited favorable discriminatory ability for TMB status and the sensitivity of multiple conventional chemotherapeutic agents including paclitaxel. Ultimately, we constructed a promising nomogram according to the risk score and N stage with good predictive accuracy compared with the actual overall survival (OS) probabilities.

Conclusion: We established an 8-gene signature that could be regarded as an independent prognostic risk factor for PAAD patients. The 8-gene signature could provide rationale for immunotherapy and chemotherapy, which might help clinicians make precise individualized treatment regimens.

Keywords: Immunotherapy; Pancreatic adenocarcinoma; Pyroptosis.

Fig. 1

Whole process of our research and elimination of batch effect between datasets (A) Flow diagram describing data collection and processing (B,C) The box and PCA plot of three datasets without batch effect elimination (D,E) The box and PCA plot of three datasets with removal of batch effect.

Fig. 2

| Determination of DEPRGs and pyroptosis phenotype subtypes using unsupervised clustering analysis (A) Heatmap of 46 PRGs expression in tumor and normal adjacent tissues, genes with yellow font represents the DEPRGs (B) The mutation characteristics of PRGs in TCGA-PAAD cohort (C) Consensus matrix heatmap for k = 3 in PAAD (D) CDF curve when k changing from 2 to 9 (E) PCA plot of three pyroptosis phenotype clusters (F) Kaplan-Meier curve for OS between three pyroptosis phenotype clusters

Fig. 3

| Establishment and assessment of the PRG signature in training and validation set (A) Forest map for prognostic significance of 14 DEPRGs in PAAD (B) LASSO coefficient trajectory of 14 prognostic DEPRGs (C)Tenfold cross-validation for selecting the optimal value of λ (D) The histogram exhibits the coefficients of the 8 hub prognostic genes (E) Expression heatmap of 8 signature genes in high- and low-risk subgroups (F–K) Distribution plot of survival status and time with increasing risk score; Kaplan−Meier curve for OS difference between risk subgroups; ROC curve of the PRG signature for predicting OS event in training and validation set.

Fig. 4

| GO and KEGG enrichment analysis (A) Heatmap illustrating the expression pattern of 1198 DEGs among risk subgroups (B) Volcano map showing distribution of the 1198 DEGs (C–F) Bar diagram and circle plot showing the top significant GO terms and KEGG pathways for the DEGs between risk subgroups (G)Summary of the top 20 genes enriched in the most significant pathway in terms of functional similarity.

Fig. 5

| GSEA for the pathway mechanism between risk subgroups based on transcription profile (A) raincloud map for the top 12 significant pathways (B) The top 5 pathways with the highest and lowest NSE (C) The top 5 pathways with the lowest NSE in GSEA (D–I) Enrichment map showing pyroptosis related pivotal pathways.

Fig. 6

| PPI and miRNA-mRNA regulatory network for DEGs comparing around risk subgroups (A) A PPI network of the interaction pairs using interaction score >0.7 as the cutoff criterion (B) The top 20 hub genes with high degrees of connectivity. The darker the red, the higher the degree (C) A miRNA-mRNA regulatory network (D) alluvial plot showing the miRNA-mRNA regulatory axes focused on the hub genes

Fig. 7

| Relationship of the PRG signature with immune characterization (A–D) Distribution of immune related scores in high- and low-risk subgroup (E, F) Infiltration abundance of immunocytes and expression pattern of key immune checkpoint genes between risk subgroups (G-H) Heatmap matrix showing the correlation of signature genes with differentially infiltrated immunocytes and key immune checkpoint genes.

Fig. 8

| Relationship of the PRG signature with mutation(A, B) Waterfall plot showed the top 20 mutated genes in both risk subgroups (C) Integrated CNV landscape of the top 10 genes (D) Scatter plot of the correlation between TMB and risk score(E) TMB distribution among risk subgroups.

Fig. 9

| Drug sensitivity analysis of 16 common chemotherapeutic agents (A–H) Common chemotherapeutic agents with a lower IC50 value in low-risk subgroup (I–P) Common chemotherapeutic agents with a lower IC50 value in high-risk subgroup.

Fig. 10

| Construction of a PRG signature-based nomogram for clinical application(A-D) Distribution of risk score between diverse clinicopathologic features (E,F) Forest plot showing the prognostic significance of multiple clinical parameters (G-J)The PRG signature-based nomogram and validated calibration curves for 1-, 2- and 3-year OS prediction perfomance