Species-specific CG dinucleotide clustering and periodicity: Its genomic context, epigenomic influences, and relationship to human disease

Abstract

CG dinucleotides occupy a unique position among nucleotide pairs in vertebrate genomes. As the near-exclusive sites of DNA methylation, the cytosines of CG dinucleotide pairs represent the physical link between DNA sequence and epigenetic regulation. In general, DNA with a high density of methylated cytosines is associated with the formation of heterochromatin and silencing of gene expression. However, recent broad surveys of DNA methylation indicate that it is more than just a simple promoter on/off switch. It has tissue-specific patterns, and changes both as a cell differentiates and ages. While regions of unusual CG density annotated as CpG islands are often relatively hypomethylated, methylation of some CpG islands is essential to normal cellular function. Regions of unusual CG density can also exist away from promoters and in repetitive sequence.;Here we describe a set of novel, unbiased analyses of genomic CG patterns and the influence they have on epigenetic regulation starting with a species-specific definition of CG clustering. Our CG cluster annotation identifies genomic that are unusually CG-dense relative to the surrounding sequence, and allows meaningful comparisons can be made between CG clusters in different genomes independent of heterogeneous underlying sequence composition.;De novo DNA methyltransferases were not thought to be sequence-specific, instead relying on DNA-binding proteins to direct them to the appropriate sequence targets. Jia et al.'s recent identification of a DNMT3a-3L complex that has a preference for CG dinucleotides spaced 8 bp apart represents the first known sequence-specific enzymatic preference. Here we identify the specific sequence compartments targeted by this enzyme and the role of 8 bp periodicity in genomic imprinting -- the function traditionally associated with DNMT3L.;Mouse knockout experiments and rare human diseases have illustrated the essential nature of the information conveyed in DNA methylation to normal cellular function. However, nowhere is the dysregulation of DNA methylation more common than in neoplasia. Here we report the first comprehensive analysis of DNA methylation changes at promoter-associated CG clusters, which has implications for disease diagnosis, prognosis, and treatment.