Retrotransposon mobilization in cancer genomes

The Cancer Genome Atlas project was initiated by the National Cancer Institute in order to characterize the genomes of hundreds of tumors of various cancer types. While much effort has been put into detecting somatic genomic variation in these data, somatic structural variation induced by the activity of transposable element insertions has not been reported. Transposable elements (TEs) are particularly relevant in cancer in part because of several known cases in which a TE insertion is directly linked to cancer formation and studies linking the epigenetic status of retrotransposons to carcinogenesis and patient outcome. Additionally, evidence for somatic retrotransposition in eukaryotic genomes suggests that some tissues and therefore some cancer types may be disposed to increased retrotransposition. We built upon previous work to develop a highly efficient computational pipeline for the detection of non-reference mobile element insertions from high-throughput paired-end whole genome sequencing data that is capable of detecting breakpoints through a local assembly strategy. Using this, we analyzed 33 whole genome tumor datasets with paired normal samples from TCGA across 3 different cancer types: glioblastoma multiforme (GBM), ovarian serous cystoadenocarcinoma (OV) and colorectal adenocarcinoma (COAD). We detected 72 insertions in colon samples, almost all of them LINE-1 elements, and none in GBM or OV. The amount of somatic retrotransposition varies widely between samples with 61 insertions present in one case. The lack of somatic retrotransposon insertions in GBM and OV samples suggests that TE activity in cancer is restricted to certain cancer types.