Abstract
Glioblastomas exhibit widespread molecular alterations including a highly distorted epigenome. Here, we resolve genome-wide 5-methylcytosine and 5-hydroxymethylcytosine in glioblastoma through parallel processing of DNA with bisulfite and oxidative bisulfite treatments. We apply a statistical algorithm to estimate 5-methylcytosine, 5-hydroxymethylcytosine and unmethylated proportions from methylation array data. We show that 5-hydroxymethylcytosine is depleted in glioblastoma compared with prefrontal cortex tissue. In addition, the genomic localization of 5-hydroxymethylcytosine in glioblastoma is associated with features of dynamic cell-identity regulation such as tissue-specific transcription and super-enhancers. Annotation of 5-hydroxymethylcytosine genomic distribution reveal significant associations with RNA regulatory processes, immune function, stem cell maintenance and binding sites of transcription factors that drive cellular proliferation. In addition, model-based clustering results indicate that patients with low-5-hydroxymethylcytosine patterns have significantly poorer overall survival. Our results demonstrate that 5-hydroxymethylcytosine patterns are strongly related with transcription, localizes to disease-critical genes and are associated with patient prognosis.