DNA methylation profiling identifies different prognostic clusters of pancreatic neuroendocrine tumors.

The occurrence and clinical relevance of global DNA hypomethylation and the hypermethylation of gene promoter regions in Pancreatic neuroendocrine tumours (PanNETs) are still largely unknown. The aim of this study was to evaluate the incidence of both epigenetic alterations in PanNETs in order to assess the relation of specific methylation profiles to chromosomal instability and tumour phenotypes. We analyzed a well-characterized series of 58 PanNETs using DNA from formalin fixed paraffin embedded tissues. Methylation-Sensitive Multiple Ligation-dependent Probe Amplification was performed to assess simultaneously the methylation status of 33 tumour suppressor genes and copy number alterations of 53 chromosomal regions. Global DNA methylation was quantified for Long Interspersed Nucleotide Element-1 using bisulfite pyrosequencing. Unsupervised hierarchical clustering resulted in three prognostically different hypermethylation groups of PanNETs (p=0.004) with significantly higher methylation levels in cluster 3 that was associated with poor prognosis compared with clusters 1 and 2. PanNETs were hypomethylated compared to normal pancreas. In average a modest degree of DNA hypomethylation was observed in cluster 1 and 3 tumours, most of which showed low levels of copy number alterations. By contrast, cluster 2 was a group of highly hypomethylated PanNETs showing the highest degree of chromosomal instability. DNA demethylation was strongly correlated with poor prognosis (p<0.0001) and remained the only independent significant predictor of outcome in multivariate analysis that included stage, grading, Ki67 proliferative index and DNA hypermethylation clusters (p=0.006). The combination of global DNA demethylation and gene hypermethylation analyses allows to define biologically and prognostically distinct subsets of PanNETs. Both alterations can be found in these tumors and each one can promote tumorigenesis by independent processes.