Transcriptomic Analysis of Laser Capture Microdissected Tumors Reveals Cancer- and Stromal-Specific Molecular Subtypes of Pancreatic Ductal Adenocarcinoma

Abstract

Purpose: Pancreatic ductal adenocarcinoma (PDAC) lethality is multifactorial; although studies have identified transcriptional and genetic subsets of tumors with different prognostic significance, there is limited understanding of features associated with the minority of patients who have durable remission after surgical resection. In this study, we performed laser capture microdissection (LCM) of PDAC samples to define their cancer- and stroma-specific molecular subtypes and identify a prognostic gene expression signature for short-term and long-term survival. Experimental Design: LCM and RNA-seq analysis of cancer and adjacent stroma of 19 treatment-naïve PDAC tumors were performed. Gene expression signatures were tested for their robustness in a large independent validation set. An RNA in situ hybridization (RNA-ISH) assay with pooled probes for genes associated with disease-free survival (DFS) was developed to probe 111 PDAC tumor samples. Results: Gene expression profiling identified four subtypes of cancer cells (C1-C4) and three subtypes of cancer-adjacent stroma (S1-S3). These stroma-specific subtypes were associated with DFS (p=5.55E-07), with S1 associated with better prognoses when paired with C1 and C2. Thirteen genes were found to be predominantly expressed in cancer cells and corresponded with DFS in a validation using existing RNA-seq datasets. A second validation on an independent cohort of patients using RNA-ISH probes to six of these prognostic genes demonstrated significant association with overall survival (median 17 versus 25 months; p<0.02). Conclusions: Our results identified specific signatures from the epithelial and the stroma components of PDAC, which add clarity to the nature of PDAC molecular subtypes and may help predict survival.