RNA Sequencing of Cystic Lesions of the Pancreas

Abstract

Introduction: Differentiating mucinous, or premalignant, cysts from benign serous or inflammatory cysts in the pancreas can be challenging. The types of proteins present in tissue can be investigated using geneexpression analysis, which investigates the amount of RNA that has been transcribed. Gene-expression analysis can be performed by directly sequencing RNA (RNAseq). RNAseq has not been performed on pancreas cysts. We previously generated a gene classifier that differentiated between benign and solid mass lesions of the pancreas with 89% accuracy. Our aims were to: 1) demonstrate that it is feasible to perform RNAseq on cystic lesions of the pancreas using fluid obtained via EUS-FNA, 2) determine the performance of our solid mass classifier on cystic lesions, and 3) establish if cysts could be differentiated by their gene profile. Methods: Adult patients presenting for EUS were enrolled prospectively. Cyst fluid was obtained for the study if enough was obtained to satisfy clinical purposes. Fluid was placed into Trizol and frozen, then extracted, amplified, and subjected to 2 cycles of double-stranded cDNA synthesis. The cDNA was directly sequenced using Illumina high-throughput sequencing. The true nature of each lesion was diagnosed by positive cytology or CEA >192 for mucinous lesions, or at least 12 months of benign clinical follow-up. Results: Seventy patients have been enrolled and 43 (n=28 solid lesions and n=15 cysts) have had tissue sequenced to date. Of the 15 cysts, 9 were benign/inflammatory and 6 were mucinous. The average amount of RNA isolated from cysts was 42.8 ng/mL (range: 8.2-165.8 ng/mL). Two mucinous lesions yielded small amounts of RNA and failed to generate acceptable cDNA. The solid mass gene classifier was applied to the cyst genetic data using PAM-R software. The classifier called all cysts benign, regardless of cyst type. The cyst data was then subjected to hierarchical clustering using a PCA analysis. This approach successfully separated mucinous and benign cysts. There were 165 genes that were significantly (>3 fold, using FDA cutoff <0.05) overexpressed in mucinous lesions versus benign. Pathway analysis was performed and the gene sets that were enriched were distinct from those seen with the adenocarcinomas. Conclusion: These preliminary results show 1) it is feasible to perform RNAseq on cystic lesions of the pancreas using fluid obtained via EUS-FNA, 2) changes that have occurred in cancer are not present in premalignant cysts, and 3) it is possible to separate benign from mucinous cysts based on genetic alterations. We plan further enrollment of cyst cases and further data analysis to see what actual gene pathways are activated in cancer but not yet in mucinous cysts. Supported by ASGE Endoscopic Research Award.