Stratification of pancreatic tissue samples for molecular studies: RNA-based cellular annotation procedure.

Abstract

Meaningful profiling of pancreatic cancer samples is particularly challenging due to their complex cellular composition. Beyond tumor cells, surgical biopsies contain desmoplastic stroma with infiltrating inflammatory cells, adjacent normal parenchyma, and "non-pancreatic tissues". The risk of misinterpretation rises when the heterogeneous cancer tissues are sub-divided into smaller fragments for multiple analytic procedures. Pre-analytic histological evaluation is the best option to characterize pancreatic tissue samples. Our aim was to develop a complement or alternative procedure to determine the cellular composition of pancreatic cancerous biopsies, basing on intra-analytic molecular annotation. A standard process for sample stratification at a molecular level does not yet exist. Particularly in the case of retrospective or data depository-based studies, when hematoxylin-eosin stained sections are not available, it supports the correct interpretation of expression profiles. A five-gene transcriptional signature (RNACellStrat) was defined that allows cell type-specific stratification of pancreatic tissues. Testing biopsy material from biobanks with this procedure demonstrated high correspondence of molecular (qRT-PCR and microarray) and histologic (hematoxylin-eosin stain) evaluations. Notably, about a quarter of randomly selected samples (tissue fragments) were exposed as inappropriate for subsequent clinico-pathological interpretation. Via immediate intra-analytical procedure, our RNA-based stratification RNACellStrat increases the accuracy and reliability of the conclusions drawn from diagnostic and prognostic molecular information.