Abstract 3094: Epigenetic classification of ependymal brain tumors across age groups

Since it has become evident that histopathological grading of ependymoma according to the WHO classification of CNS tumors is not capable of accurately classifying patients into meaningful strata, a broadly accepted molecular classification scheme with prognostic significance is desperately needed. In recent years, ependymomas were classified into molecular subgroups based on transcriptomic alterations. In tumors localized within the posterior fossa, two distinct biological entities of ependymoma were delineated by several studies (designated posterior fossa A and posterior fossa B), which show striking differences in genetic characteristics and clinical outcome. A similar consensus for supratentorial and spinal ependymoma is lacking. We studied genome-wide DNA methylation (Illumina HumanMethylation450 (450k) array) in 180 primary ependymal tumors (80 with corresponding gene expression profiling data generated by Affymetrix 133plus2.0 arrays), including ependymomas (posterior fossa, supratentorial, spinal), subependymomas (SE), myxopapillary ependymoma (MPE), pineal parenchymal tumors of intermediate differentiation (PPTID), and papillary tumors of the pineal region (PTPR). We performed hierarchical clustering to identify robust molecular subgroups. Independent gene expression profiling datasets from previously published ependymoma studies (Johnson et al.; Wani et al.; Witt et al.) were used as validation cohorts. DNA methylation data showed that ependymal brain tumors can be classified into eight molecular subgroups. Notably, MPE, SE, PPTID and PTPR tumors formed robust distinct clusters, as did posterior fossa Group A and Group B ependymomas. Supratentorial ependymomas can be classified into two principle molecular subgroups, one of which displays a dismal prognosis, and comprises predominantly children and infants, and is associated with highly recurrent gene fusion. Notably, a significant number of ependymomas previously classified by histology as WHO Grade II/III look like SE by methylation, and also have extremely good survival. In summary, using genome-wide DNA methylation and transcriptome analysis we could decipher robust molecular subgroups of ependymal brain tumors including supratentorial ependymoma. Diagnoses of tumors with challenging histopathological features can now be supported by this technology. Hence, this approach offers the possibility to replace the unambiguous histological grading system that is currently in use with a robust molec