Extraordinary Cancer Epigenomics: Thinking Outside the Classical Coding and Promoter Box

Cancer Epigenetics and Biology Program (PEBC), Bellvitge Biomedical Research Institute (IDIBELL), Avinguda Gran Via de L’Hospitalet 199–203, L’Hospitalet del Llobregat, Barcelona, Catalonia, Spain Department of Physiological Sciences II, School of Medicine, University of Barcelona, Barcelona, Catalonia, Spain 3 [3_TD$DIFF]Institucio Catalana de Recerca i Estudis Avançats [4_TD$DIFF], Barcelona, Catalonia, Spain

. Integration of these studies with the functional annotations in ENCODE [ 7 5 _ T D $ D I F F ] shows that 88% of SNPs associated with [ 7 6 _ T D $ D I F F ] a risk of cancer are at positions that lie outside the coding regions of genes or their promoters, suggesting severe consequences of altered functions of the noncoding genome [2].

Trends
There is a new appreciation for the noncoding regions of the human genome as their functions in gene regulation are delineated through the use of functional genomics.
The cataloging of epigenetic marks and their impact on cellular function is proceeding rapidly.
The intersection of epigenetics and functional genomics reveals the myriad ways by which the expression of the human genome is regulated.
Understanding how epigenetic marks on the noncoding genome can promote tumor progression can lead to new therapeutic strategies.  relationships between SNPs, gene expression, and disease risk that have advanced our understanding of a wide variety of diseases[ 1 2 3 _ T D $ D I F F ] [12][13][14][15]. Extending this approach to epigenetic modifications such as DNA methylation can help clarify how epigenetics and the [ 1 2 5 _ T D $ D I F F ] noncoding genome interact in cancer. Sites of differential methylation that correlate with gene expression changes are called methylation [ 1 2 6 _ T D $ D I F F ] eQTLs ([ 1 2 7 _ T D $ D I F F ] meQTLs) and represent a promising approach to the study of cancer epigenetics and [ 1 2 8 _ T D $ D I F F ] its impact on gene expression [16][17][18]. Analysis of [ 1 3 6 _ T D $ D I F F ] 3649 primary human tumors created a meQTL catalog of DNA methylation associations for 21% of interrogated cancer risk polymorphisms [18]. This study linked risk alleles to genes previously characterized as having known roles in cancer in addition to as yet unidentified cancer genes. The association between breast cancer risk allele rs2380205 and the FBXO18 gene is of particular note. FBXO18 is an F-box protein helicase that actively participates in the formation of double-strand breaks and activation of tumor [ 1 3 8 _ T D $ D I F F ] protein-dependent apoptosis following DNA replication stress [20]. Genome-wide binding studies of mediator and cohesin components revealed that these factors concentrate at a subset of regulatory elements that group genomically to form what are termed 'super-enhancers' [23].

. Methods to Identify Enhancers
Identifying the genomic location of enhancers is an important goal for functional genomics studies. Chromatin immunoprecipitation sequencing (ChIP-seq) studies of transcription factor binding and histone modifications are typically used in lieu of [ 7 5 2 _ T D $ D I F F ] functional data to infer the activity of cis-regulatory modules. The drawback to using ChIP-seq data [ 7 5 3 _ T D $ D I F F ] is that binding does not perfectly correlate with function; a recent analysis of Encyclopedia of DNA Elements (ENCODE)predicted enhancers found only 30% with genuine enhancer/CRM activity [131].[ 7 5 6 _ T D $ D I F F ] A further drawback [ 7 5 7 _ T D $ D I F F ] to these approaches is that they identify large segments of the genome (typically greater than 1 kb) that are not amenable to motif analysis.
Several approaches have been developed to overcome these obstacles. The relationship between DNA methylation and super-enhancers can [ 1 6 9 _ T D $ D I F F ] also be measured using functional genomics data [29,30]. For example, Heyn and colleagues analyzed the methylation status of over

Internal Promoters
Epigenetic changes can also redirect the transcriptional network. By comparing methylation levels between primary tumors and metastases, Visoso et al. found a hypomethylation event that reactivates a cryptic transcript of TBC1D16, a Rab GTPase-activating protein [33]. The novel short isoform of TBC1D16 exacerbates melanoma growth and metastasis by targeting RAB5C and regulating EGFR[ 1 9 4 _ T D $ D I F F ] and confers poor clinical outcome while showing greater sensitivity to BRAF and MEK inhibitors than cells lacking the short transcript. This is an example of the contribution of epigenetics to metastasis and to  [35]. However, the prevalence of this effect in the cancer patient population at large remains unclear. The relationship between promoter methylation and usage of cryptic internal promoters to generate oncogenic isoforms requires further investigation.

The Genome 3D Structure
The human genome is packaged into the nucleus in a 3D structure with [ 2 2 2 _ T D $ D I F F ] distinct properties and characteristics [ 2 2 3 _ T D $ D I F F ] for each cell type. It is now apparent that this 3D organization impacts gene regulation [36,37]. Methods for determining the 3D structure of the genome are reviewed in Box 2. One of the most important consequences of this 3D organization is the direct interaction of enhancers with their target promoters through the formation of chromatin loops, which allows them to operate at  [38]. Their[ 2 3 1 _ T D $ D I F F ] studies show that mediator, a transcriptional coactivator, forms a protein complex with cohesin to loop the DNA between enhancers and promoters. Further, Nipbl, a cohesin-loading factor, is associated with these complexes, providing a mechanism by which the cohesin can be loaded [ 2 3 2 _ T D $ D I F F ] onto DNA to form and maintain chromatin loops. Both the mediator and cohesin genes have known roles in cancer [39][40][41][42][43]. DamID is an approach[ 7 8 3 _ T D $ D I F F ] used to catalog genomic sequences that interact with lamina-associated domains of the inner face of the nuclear membrane [137][138][139]. In DamID, bacterial adenine methyltransferase is fused to a protein of interest and allowed to interact with physically[ 7 8 5 _ T D $ D I F F ] proximal DNA. Sequences containing methylated products are enriched by Damspecific restriction enzymes, then sequenced and mapped back to the reference genome thus demarcating [ 7 8 6 _ T D $ D I F F ] genomic regions that interact with the protein of interest.
Another group of widely used techniques revolve around chromosome conformation capture (3C) methods coupled to HTP sequencing [140]. In principle, these methods quantify the relative spatial proximity between individual genomic loci through the digestion and re-ligation of  [46,49,[141][142][143][144] and those that analyze interactions of targeted loci (e.g., 3C, circularized chromatin conformation capture, chromatin interaction analysis by paired-end tag sequencing) [145][146][147][148]. subtype of glioblastoma [55]. Gliomas with a putative CpG island methylator phenotype (G-CIMP) are linked to isocitrate dehydrogenase (IDH) gain-of-function mutations [56][57][58][59]. Mutant IDHs produce 2-hydroxyglutarate, an oncometabolite that interferes with the function of iron-dependent hydroxylases [60]. IDH mutants have severely altered methylomes,  [66][67][68] and [ 3 2 6 _ T D $ D I F F ] in metastasis [69]. An integrated analysis of TCGA data revealed a master miRNA regulatory network that governs the ovarian cancer mesenchymal subtype, which [ 3 2 7 _ T D $ D I F F ] has poor survival [70]. One of the first such was the discovery that miR-124 contributes to pathogenesis through its role in diminishing levels of CDK6 [71]. Analysis of miRNAs in the colorectal cancer cell line HCT-116 found   , is sensitive to LED levels [116]. This work provides a novel regulatory interaction between enhancer function and lncRNAs and is a promising avenue for future studies.
Transcribed ultraconserved regions represent an interesting subclass of lncRNAs. They are a subset of genomic elements that are absolutely conserved between orthologous regions of the human, rat, and mouse genomes. These sequences can be located

Outstanding Questions
How much of the noncoding genome is subject to active epigenetic regulation?
What epigenetic marks are most descriptive of the underlying cancerregulatory network? Can these marks be utilized in personalized medicine?
What are the roles of retroelements in normal tissue and how do they contribute to oncogenesis?
What is the order of events among epigenetic changes, gene expression changes, and cancer phenotype?
Delineating causation [ 5 0 6 _ T D $ D I F F ] among epigenetic changes, gene expression changes, and the cancer phenotype is also critical for future targeted therapies. For example, consider the case of [ 5 0 7 _ T D $ D I F F ] IDHmutant G-CIMP gliomas, which feature a hypermethylated genome and activated PDGF due to methylation of  [129,130]. As epigenetic profiling becomes more accessible it is increasingly likely that we will be able to leverage [ 5 2 7 _ T D $ D I F F ] singlecell profiling to accurately determine both the cell type of origin and the precise oncogenic lesion, allowing highly specific and effective therapies to be administered. Thinking outside [ 5 2 8 _ T D $ D I F F ] the classical coding and promoter box by studying the epigenome of normal tissue and cancer cells will lead to better therapies for the public health crisis that is cancer.