The value of a metabolic reprogramming-related gene signature for pancreatic adenocarcinoma prognosis prediction

Abstract

Pancreatic ductal adenocarcinoma (PDAC) is one of the most fatal malignancies worldwide. Extensive enhancement of glycolysis and reprogramming of lipid metabolism are both associated with the development and progression of PDAC. Previous studies have suggested that various gene signatures could convey prognostic information about PDAC. However, the use of these signatures has some limitations, perhaps because of a lack of knowledge regarding the genetic and energy supply backgrounds of PDAC. Therefore, we conducted multi-mRNA analysis based on metabolic reprogramming to identify novel signatures for accurate prognosis prediction in PDAC patients. In this study, a three-gene signature comprising MET, ENO3 and CD36 was established to predict the overall survival of PDAC patients. The three-gene signature could divide patients into high- and low-risk groups by disparities in overall survival verified by log-rank test in two independent validation cohorts and could differentiate tumors from normal tissues with excellent accuracy in four Gene Expression Omnibus (GEO) cohorts. We also found a positive correlation between the risk score of the gene signature and inherited germline mutations in PDAC predisposition genes. A glycolysis and lipid metabolism-based gene nomogram and corresponding calibration curves showed significant performance for survival prediction in the TCGA-PDAC dataset. The high-risk designation was closely connected with oncological signatures and multiple aggressiveness-related pathways, as determined by gene set enrichment analysis (GSEA). In summary, our study developed a three-gene signature and established a prognostic nomogram that objectively predicted overall survival in PDAC. The findings could provide a reference for the prediction of overall survival and could aid in individualized management for PDAC patients.

Keywords: metabolic reprogramming; pancreatic adenocarcinoma.

Figure 1

Flowchart presenting the process of establishing the gene signature in this study.

Figure 2

Identification of DEGs in pancreatic cancer between tumor and paracancerous tissues. (A) Volcano plots of DEGs in the 3 indicated datasets. (X-axis: log2(FC); Y-axis: -log10(FDR) for each gene. Genes with FDR <0.01 and FC >1 or <-1 were considered as DEGs in each series. Blue: down-regulated genes; Gray: non-differential genes; Red: up-regulated genes). (B) Upset Venn diagrams of the DEGs identified in 3 GEO datasets. (C) Top 10 enriched KEGG pathways of the DEGs.

Figure 3

Lasso analysis and Kaplan-Meier curve for the patients in the TCGA, ICGC and FUSCC cohorts. (A) Representative heatmap of the DEGs significantly related to OS time identified by the log-rank test in the TCGA cohort. (B) LASSO coefficient profiles of the 12 glycolysis and lipid metabolism-based genes. LASSO, least absolute shrinkage and selection operator method. (C) Risk score analysis of the differentially expressed DEG signatures of PDAC. Risk scores of DEG signatures (top); survival status and duration of cases (middle); low-score and high-score groups for the three genes (bottom). (D) The Kaplan-Meier plot (low risk vs. high risk PDAC cases) of 5-year overall survival in the TCGA cohort. (E) Time-dependent ROC analyses at 1, 3, and 5 years in the TCGA cohort. (F) The Kaplan-Meier plot (low risk vs. high risk PDAC cases) of 5-year overall survival in the ICGC cohort. (G) Time dependent ROC analyses at 1, 3, and 5 years in the ICGC cohort. (H) The Kaplan-Meier plot (low risk vs. high risk PDAC cases) of 5-year overall survival in the FUSCC cohort (I). Time dependent ROC analyses at 1, 3, and 5 years in the FUSCC cohort.

Figure 4

Validation of expression and alteration of the three genes in pancreatic cancer. (A–C) The MET, CD36 and ENO3 mRNA expression levels in TCGA pancreatic cancer tumor tissue and matching normal tissue from data on TCGA and GTEx. Data were obtained from the GEPIA (http://gepia.cancer-pku.cn/). (D–F) The MET, CD36 and ENO3 mRNA expression levels in GEO32676 and GEO15471 pancreatic cancer tumor tissue compared with non-tumor tissues. (G–I) The representative protein expression of the 3 glycolysis and lipid metabolism-based genes in pancreatic cancer tumor tissue. Data were obtained from the human protein atlas (https://www.proteinatlas.org/). (J–L) Survival analysis of patients with PAAD in terms of MET, CD36 and ENO3 in TCGA patients. (M) Genetic alterations of the three genes in the ICGC, QCMG, TCGA and UTSW pancreatic cancer datasets. Data were obtained from the cBioportal (https://www.cbioportal.org/).

Figure 5

(A) The ROC curves of the risk scores differentiating pancreatic cancer from normal tissues in the four validation GEO datasets. The clinical and tumor mutation relevance of the three gene signatures. (B–D) The distribution of the risk scores in different AJCC stages in the TCGA cohort. (E–H) The Kaplan-Meier plot (low risk- score vs. high risk- score) of 5-year overall survival in patients in the TCGA cohort. (I–L) The expression level of the risk score in different mutation statuses of KRAS, TP53, CDKN2A, and SMAD4 in the TCGA dataset.

Figure 6

Validation of the nomogram in predicting overall survival of pancreatic cancer in the TCGA dataset. (A) Forest plot summary of multivariable Cox regression analyses of the risk score, age, sex, grade and tumor stage in the TCGA cohort. The squares represent the hazard ratio (HR), and the transverse lines represent 95% CIs. CI, confidence interval. (B) A nomogram to predict survival probability at 1, 3 and 5 years for PDAC patients based on the results derived from the TCGA cohort. (C–E) Calibration curve for the nomogram when predicting 1, 3- and 5-year overall survival.

Figure 7

Gene set enrichment analyses. (A–C) Top 3 (HOXC6 target cancer, intestine probiotics, target of SMAD2 or SMAD3.) oncological signatures significantly C2 (hallmark gene sets) enriched in the high-risk group identified by gene set enrichment analysis. (D–F) Top 3 (nuclear periphery, apical junction assembly, nuclear matrix.) oncological signatures significantly C5 (biological process) enriched in the high-risk group identified by gene set enrichment analysis. (G–I) Top 3 (MTORC1 signaling, glycolysis, unfolded protein response.) oncological signatures significantly H (hallmark gene sets) enriched in the high-risk group identified by gene set enrichment analysis.