Differences between Well-Differentiated Neuroendocrine Tumors and Ductal Adenocarcinomas of the Pancreas Assessed by Multi-Omics Profiling

Abstract

Most pancreatic neuroendocrine tumors (PNETs) are indolent, while pancreatic ductal adenocarcinomas (PDACs) are particularly aggressive. To elucidate the basis for this difference and to establish the biomarkers, by using the deep sequencing, we analyzed somatic variants across coding regions of 409 cancer genes and measured mRNA/miRNA expression in nine PNETs, eight PDACs, and four intestinal neuroendocrine tumors (INETs). There were 153 unique somatic variants considered pathogenic or likely pathogenic, found in 50, 57, and 24 genes in PDACs, PNETs, and INETs, respectively. Ten and 11 genes contained a pathogenic mutation in at least one sample of all tumor types and in PDACs and PNETs, respectively, while 28, 34, and 11 genes were found to be mutated exclusively in PDACs, PNETs, and INETs, respectively. The mRNA and miRNA transcriptomes of PDACs and NETs were distinct: from 54 to 1659 differentially expressed mRNAs and from 117 to 250 differentially expressed miRNAs exhibited high discrimination ability and resulted in models with an area under the receiver operating characteristics curve (AUC-ROC) >0.9 for both miRNA and mRNA. Given the miRNAs high stability, we proposed exploring that class of RNA as new pancreatic tumor biomarkers.

Keywords: NGS; adenocarcinoma; mRNA; miRNA; neuroendocrine tumors; somatic mutation.

Figure 1

Venn diagram, depicting unique and shared gene numbers carrying deleterious variants across tested tissues. Parallel sequencing of 13 neuroendocrine tumors and 8 ductal adenocarcinomas of the pancreas identified 1288 unique variants across 234 genes, of which 50, 57, and 24 genes carried variants likely to be deleterious in pancreatic ductal adenocarcinomas (PDACs), pancreatic neuroendocrine tumors (PNETs), and intestinal neuroendocrine tumors (INETs), respectively.

Figure 2

Venn diagram with variant numbers found in at least 40% of samples in all three tumor types selected by the parallel sequencing of 13 neuroendocrine tumors (nine PNETs and four INETs) and 8 ductal adenocarcinomas of the pancreas.

Figure 3

Deep sequencing of the mRNA transcriptomes identified 2082, 2621, and 107 mRNAs, whose expression significantly differentiated between PDACs and PNETs, PDACs and INETS, and PNETs and INETs, respectively, as demonstrated by principal component analysis (PCA) using the DESeq2 function “plotPCA” with the default parameters.

Figure 4

Deep sequencing of the miRNA transcriptomes identified 258, 297, and 129 mRNAs, whose expression significantly differentiated between PDACs and PNETs, PDACs and INETS, and PNETs and INETs, respectively, as demonstrated by principal component analysis (PCA) using the standard R programming language function “prcomp.”.

Figure 5

Scatter plot of individual samples from the DIABLO PLS-DA analysis results.

Figure 6

Main contributors to the first two components of the DIABLO PLS-DA model for miRNA and mRNA.