CG dinucleotide periodicities recognized by the Dnmt3a-Dnmt3L complex are distinctive at retroelements and imprinted domains.

The Dnmt3a and Dnmt3L genes are critical mediators of cytosine methylation during gametogenesis, with major actions noted at transposable elements and imprinted loci. The Dnmt3a-Dnmt3L complex was recently described to have preferential activity at CG dinucleotides located 8-10 bp apart. Because cytosine methylation is heterogeneously distributed in the genome, we tested whether this relative sequence preference explains the effects of mutation of the Dnmt3a and Dnmt3L genes using bioinformatic analysis. We found that the human and mouse genomes are significantly enriched in a CG dinucleotide periodicity of 2 bp, leading to an increased frequency of CGs spaced 8 bp apart that represent widespread targets for this protein complex. When we broke down the human and mouse genomes by annotation, we found that this significant 2-bp periodicity and increased 8-bp periodicity are maintained in Alu SINEs in both species. The 8-bp periodicity was mapped genome-wide, identifying enrichment at the promoters of both paternally and maternally methylated imprinted genes and at CG dinucleotide-enriched sequences. We conclude that CG dinucleotide periodicity helps to explain some but not all of the relative sequence specificity of mutations of Dnmt3a or Dnmt3L in the establishment of germline cytosine methylation patterns.