DNA Methylation Profiling Identifies Different Prognostic Clusters of Pancreatic NeuroEndocrine Tumors
Michele Stefanoli, Daniela Furlan, Nora Sahnane, Stefano La Rosa, Chiara Romualdi, Fausto Sessa, Carlo Capella. University of Insubria, Varese, Italy; Ospedale di Circolo, Varese, Italy; University of Padova, Padova, Italy
Background: Pancreatic NeuroEndocrine Tumors (PanNETs) are rare and heterogeneous neoplasms accounting for about 1-2% of pancreatic tumors. To date, very few studies have been published about DNA hypermethylation in such tumors with a limited number of genes analyzed and there is no information about distinct hypermethylation profiles of PanNETs.
Design: We investigated the epigenetic profile of 58 morphologically and immunohistochemically well characterized PanNETs analyzing the hypermethylation status of 33 tumor suppressor genes using MS-MLPA technique and bisulphite pyrosequencing. The aims of the study were to evaluate the frequency of aberrant DNA methylation in PanNETs and the occurrence of specific methylation profiles with respect to the morphofunctional and clinical profile of each tumor. Unsupervised hierarchical cluster analysis was performed using Jaccard index as dissimilarity measure.
Results: A subset of 17 PanNETs (29%) showed high levels of aberrant DNA methylation (more than 25% of the promoters examined) and this phenotype was positively correlated with nonfunctioning profile (p=0.05), advanced ENETS stage (p=0.02), and poor prognosis (p=0.02). Unsupervised hierarchical clustering provided the best performance in subtype classification of PanNETs distinguishing three specific methylation profiles that were associated with prognosis in both univariable (p=0.008) and multivariable analysis (p=0.05). Cluster 1 and cluster 2 (13 and 25 patients, respectively) showed very low levels of gene methylation and were associated with good prognosis. By contrast, cluster 3 (20 patients) was positively correlated with poor prognosis compared to cluster 1 and cluster 2 and showed significant higher levels of methylation in the following genes: DAPK1, TIMP3, PAX5, HIC1, IGSF4, PYCARD, ESR1, VHL, RARB, WT1 (p<1-4).
Conclusions: PanNETs showing a widespread DNA methylation are a distinct clinico-pathological entity characterized by an aggressive tumor behavior. PanNETs with specific epigenetic profiles are strongly associated with poor prognosis, suggesting that the study of DNA methylation may be a promising tool to identify molecular markers of potential diagnostic and prognostic values.
Wednesday, March 6, 2013 1:00 PM
Poster Session VI # 63, Wednesday Afternoon