A 14-gene gemcitabine resistance gene signature is significantly associated with the prognosis of pancreatic cancer patients

Xing Wei,Yun Zhao,Yang He,Chunfang Xu,Zhen Weng,Xiaochong Zhou

doi:10.1038/s41598-021-85680-x

Abstract

To identify a gemcitabine resistance-associated gene signature for risk stratification and prognosis prediction in pancreatic cancer. Pearson correlation analysis was performed with gemcitabine half maximal inhibitory concentration (IC50) data of 17 primary pancreatic cancer lines from Genomics of Drug Sensitivity in Cancer (GDSC) and the transcriptomic data from GDSC and Broad Institute Cancer Cell Line Encyclopedia, followed by risk stratification, expression evaluation, overall survival (OS) prediction, clinical data validation and nomogram establishment. Our biomarker discovery effort identified a 14-gene signature, most of which featured differential expression. The 14-gene signature was associated with poor OS in E-MTAB-6134 (HR 2.37; 95% CI 1.75–3.2; p < 0.0001), pancreatic cancer-Canada (PACA-CA) (HR 1.76; 95% CI 1.31–2.37; p = 0.00015), and 4 other independent validation cohorts: pancreatic cancer-Australia (PACA-AU) (HR 1.9; 95% CI 1.38–2.61; p < 0.0001), The Cancer Genome Atlas (TCGA) (HR 1.73; 95% CI 1.11–2.69; p = 0.014), GSE85916 (HR 1.97; 95% CI 1.14–3.42; p = 0.014) and GSE62452 (HR 1.82; 95% CI 1.02–3.24; p = 0.039). Multivariate analysis revealed that the 14-gene risk score was an independent pancreatic cancer outcome predictor in E-MTAB-6134 (p < 0.001) and TCGA (p = 0.006). A nomogram including the 14-gene was established for eventual clinical translation. We identified a novel gemcitabine resistance gene signature for risk stratification and robust categorization of pancreatic cancer patients with poor prognosis.

Highlights

To identify a gemcitabine resistance-associated gene signature for risk stratification and prognosis prediction in pancreatic cancer
Many studies have been carried out to elucidate the possible mechanisms involved in gemcitabine resistance[7,8,9], which are considered to be related to transport and metabolism behavior and are thought to involve multiple enzymes and signaling pathways, including human concentrative nucleoside transporters, human equilibrative nucleoside transporters, deoxycytidine kinase, and ribonucleotide reductase (RR), or increased activity of the detoxifying enzyme cytidine deaminase (CDA), and PI3K/Akt, MAPK and NF-κB pathways
Total of 1208 (550 positively correlated and 658 negatively correlated genes) and 1983 (1294 positively correlated and 689 negatively correlated genes) gemcitabine resistance-related genes were identified from the E-MTAB-3610 and Cell Line Encyclopedia (CCLE) datasets, respectively (Fig. 1A)

Summary

Introduction

To identify a gemcitabine resistance-associated gene signature for risk stratification and prognosis prediction in pancreatic cancer. Many studies have been carried out to elucidate the possible mechanisms involved in gemcitabine resistance[7,8,9], which are considered to be related to transport and metabolism behavior and are thought to involve multiple enzymes and signaling pathways, including human concentrative nucleoside transporters (hCNTs), human equilibrative nucleoside transporters (hENTs), deoxycytidine kinase (dCK), and ribonucleotide reductase (RR), or increased activity of the detoxifying enzyme cytidine deaminase (CDA), and PI3K/Akt, MAPK and NF-κB pathways Based on these findings, changes in multiple genes at the functional and expression levels could be a reasonable phenomenon during the development of gemcitabine resistance. By including genes related to gemcitabine resistance in pancreatic cancer, we performed a systematic and comprehensive discovery and validation of biomarkers to identify and generate a gene expression signature for the effective prognostic prediction of patients with pancreatic cancer by using multiple datasets. We identified a novel 14-gene signature comprising genes related to gemcitabine resistance, which could offer excellent accuracy for patient risk stratification and prognosis prediction

Methods

Results

Conclusion