Deep sequencing reveals distinct microRNA-mRNA signatures that differentiate pancreatic neuroendocrine tumor from non-diseased pancreas tissue

Abstract

Background: Only a limited number of biomarkers guide personalized management of pancreatic neuroendocrine tumors (PanNETs). Transcriptome profiling of microRNA (miRs) and mRNA has shown value in segregating PanNETs and identifying patients more likely to respond to treatment. Because miRs are key regulators of mRNA expression, we sought to integrate expression data from both RNA species into miR-mRNA interaction networks to advance our understanding of PanNET biology.

Methods: We used deep miR/mRNA sequencing on six low-grade/high-risk, well-differentiated PanNETs compared with seven non-diseased tissues to identify differentially expressed miRs/mRNAs. Then we crossed a list of differentially expressed mRNAs with a list of in silico predicted mRNA targets of the most and least abundant miRs to generate high probability miR-mRNA interaction networks.

Results: Gene ontology and pathway analyses revealed several miR-mRNA pairs implicated in cellular processes and pathways suggesting perturbed neuroendocrine function (miR-7 and Reg family genes), cell adhesion (miR-216 family and NLGN1, NCAM1, and CNTN1; miR-670 and the claudins, CLDN1 and CLDN2), and metabolic processes (miR-670 and BCAT1/MPST; miR-129 and CTH).

Conclusion: These novel miR-mRNA interaction networks identified dysregulated pathways not observed when assessing mRNA alone and provide a foundation for further investigation of their utility as diagnostic and predictive biomarkers.

Keywords: Biomarkers; MicroRNA; Pancreatic neuroendocrine tumors; mRNA; miR-mRNA interaction networks.

Fig. 1

Analysis of mRNA expression levels. (a) PCA analysis of tumor (n = 6) and non-diseased (n = 7) tissue samples. (b) Volcano plot of intersection DE analysis. The significance level was set to|Log2(fold change)| ≥ 2 and - Log10(adjusted P-value) < 0.05. (c) Hierarchically clustered heatmap of top 10% more abundant and less abundant genes based on their corresponding VST values.

Fig. 2

Clinically relevant mRNA biomarker expression. The balloon plot shows the expression of key biomarkers in individual PanNETs compared to non-diseased tissue. Both the heatmap and size of dots denote normalized gene expression (VST). Red rectangular boxes highlight key markers assessed in routine clinical practice.

Fig. 3

Analysis of miR expression levels. (a) PCA analysis of tumor (n = 6) and non-diseased (n = 7) tissue samples. (b) Volcano plot of intersection DE analysis. The significance level was set to|Log₂(fold change)| ≥ 2 and - Log10(adjusted P-value < 0.05. (c) Hierarchically clustered heatmap of all (60) DEMs based on their corresponding VST normalized values.

Fig. 4

Putative miR-mRNA interaction networks. (a) Venn diagram of the less abundant and more abundant DEGs and predicted targets. (b) Histograms of degree counts for top three least abundant DEMs (left) and top three most abundant DEMs (right). (c) Network of top three least abundant DEMs with predicted more abundant mRNA. (d) Network of top three most abundant DEMs with predicted less abundant mRNA targets. (e) Hierarchically clustered correlation dot–plot between selected miR and all their corresponding predicted mRNA pairs.

Fig. 5

Function analysis of miR-mRNA interaction network. (a) GO enrichment analysis and (b) KEGG pathway analysis of the 89 predicted gene targets (dot plot on the left, network plot on the right). The size and color of the dots represent the amount of DEGs enriched in the pathway and enrichment significance.