Carcinogenesis results from aberrations in gene regulatory networks at the genetic and epigenetic levels. Epigenetic changes such as global hypomethylation along with hypermethylated tumor-suppressor gene promoters are hallmarks of many cancers. Transposable elements (TEs) constitute the majority of non-coding sequence in the human genome and 50% of the total human genome sequence; they are thought to be suppressed by DNA methylation in somatic tissues. TEs often carry transcription factor binding sites and can function as transcriptional enhancers. Mis-regulation of DNA methylation can lead to abnormal gene activity from TE-based enhancers. Although many studies have profiled cancer DNA methylomes, there has not been a focused study to investigate the changes in TE DNA methylation in cancer and its effect on cancer gene expression. We hypothesize that mis-expression of genes from TE-based enhancers contributes to carcinogenesis. To test this hypothesis, we generated complete DNA methylomes of endometrial cancer samples and the matched normal tissue. We then used computational algorithms to identify differentially methylated transposon subfamilies. Our approach consists of using two complementary sequencing-based methods to map the DNA methylomes (MeDIP-seq and MRE-seq). To map unmethylated sites at single CpG resolution, we sequence DNA, digested by three methylation-sensitive restriction enzymes (MRE). We use the HpaII (C^CGG), Hin6I (G^CGC) and AciI (C^CGC) enzymes to access a very large fraction of potentially non-methylated CpG sites. In addition, to map methylated regions we use an anti-methylcytosine antibody to immunoprecipitate the methylated fraction of DNA fragments and then sequence them. Analyses of sequencing data, generally discards reads that map to repetitive regions in the genome, thus significantly reducing our ability to investigate TEs. Our computational approach, harnesses reads that map to repetitive regions by re-aligning them to TE subfamilies. We then measure the enrichment of reads for each TE subfamily from both the sequencing methods, in the cancer tissue compared to the matched normal tissue. Our findings reveal that, as expected, TEs in normal endometrial tissues are mostly methylated. However, contrary to the common belief, we found that majority of TEs remain methylated in endometrial cancer samples. A few dozen TE subfamilies, including several ERV/LTR subfamilies (endogeneous retrovirus) are demethylated. More interestingly, when we compare our three type I cancer samples with the three type II cancer samples, we detect a clear cancer type-specific, TE-subfamily-specific DNA demethylation pattern. This observation, coupled with the idea that different TE subfamilies may encode different enhancer elements, suggests that demethylated TEs may rewire cancer type-specific gene regulatory networks and contribute to cancer-specific phenotypes.

Citation Format: {Authors}. {Abstract title} [abstract]. In: Proceedings of the 103rd Annual Meeting of the American Association for Cancer Research; 2012 Mar 31-Apr 4; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2012;72(8 Suppl):Abstract nr LB-379. doi:1538-7445.AM2012-LB-379