Ependymoma, the third most common pediatric brain tumor is currently treated by maximal safe surgery, followed by irradiation. No chemotherapeutic regimen has shown an increased survival for patients with ependymoma. A lack of available ependymoma cell lines or animal models has stifled attempts to better understand the disease, and develop novel targeted therapies. At a genetic level, many pediatric posterior fossa ependymomas have a balanced karyotype, suggesting that they are initiated and maintained by very small copy number aberrations, small sequences changes, or epigenetic events such as DNA promotor methylation or histone modifications. Traditional methods to identify methylated genes in cancer include treating in vitro cell lines with a de-methylating agent such as 5-azacytidine, and determining which genes increase in expression. This method is unavailable for ependymoma due to the lack of available cell lines. We fragmented genomic DNA from 27 primary pediatric posterior fossa ependymomas, 10 normal brain samples, and 4 ependymoma xenograft cell lines and subjected them to MeDIP (methylation dependant immunoprecipitation). The IP fraction (enriched for methylated DNA) was then compared to the wash fraction (impoverished for methylated DNA) by differential labeling and hybridization to Nimblegen Promotor tiling arrays (MeDIP-ChIP). Regions of the genome in which the ratio of IP/Wash DNA was >2 in one or more tumor samples, but not in normal brain samples were identified as methylated. Expression analysis was then performed on an overlapping set of 100 ependymomas using Affymetrix Exon arrays. Increased promoter methylation was correlated with decreased gene expression. We identified methylation and silencing of numerous genes involved in transcriptional regulation (HOXD3, HIC1, FOXD4, RASSF2), apoptosis (BMF, TNF, TRADD, BCL2L12, TNFSF9) in addition to several other gene families. We subsequently analyzed 4/27 of the primary ependymoma samples by MeDIP-Seq using a next generation sequencing strategy. Briefly, tumor DNA was fragmented and fractionated into IP and wash fractions by MeDIP, and then subsequently analyzed by generating >30 million short (35 bp) reads for each fraction, followed by mapping and alignment to the genome on the UCSC genome browser. Regions of the genome which showed significantly more reads in the IP fraction versus the wash fraction were identified. There was a significant overlap between genes identified as methylated by MeDIP-ChIP and MeDIP-Seq. MeDIP-Seq also identified a number of regions of intragenic and intergenic tumor specific methylation that are currently of uncertain significance. Our data demonstrate that MeDIP-ChIP and MeDIP-Seq are useful tools to delineate the cancer epi-methylgenome. We suggest that by avoidance of high passage cell lines, the data provided may be more relevant to human tumors in vivo.

Citation Information: In: Proc Am Assoc Cancer Res; 2009 Apr 18-22; Denver, CO. Philadelphia (PA): AACR; 2009. Abstract nr 2197.

100th AACR Annual Meeting-- Apr 18-22, 2009; Denver, CO