Barlas, A. B.; Karaca, E.

Mammalian de novo DNA methylation is integral to gene regulation. This epigenetic modification is exerted by DNMT3A and DNMT3B enzymes predominantly on CpG islands. Over their enzymatic domains, DNMT3A/B shares over 90% of sequence similarity. Even so, DNMT3A prefers to methylate preferentially CGC, while DNMT3B prefers CGG sequences. To elucidate the mechanistic basis of this selective methylation profile, we performed extensive molecular dynamics simulations of DNMT3A/B enzymes bound to all possible CGX[C/G/T/A] motifs. As an outcome, we observed that the DNMT3 sequence selectivity at +2 position relies on the likelihood of a stable Arg/Lys-Guanine base-specific hydrogen bond formation. In the case of DNMT3A, Arg836 forms stable hydrogen bonds with a single guanine that is complementary to the second C in the CGC motif. In the case of DNMT3B, Arg836 is replaced by Lys777, where DNMT3B-Lys777 switches to hydrogen bonding with the two consecutive guanines in the CGG motif. The loss of branching in DNMT3B-Lys777 is further compensated by Lys777's cooperative action with Asn779. Our analysis also showed that the single nucleotide substitutions in CGX motifs lead to dramatic changes in the DNMT3A-DNA selective hydrogen bonding profile compared to the ones in DNMT3B. This finding is well correlated with the higher CpG specificity of DNMT3A compared to DNMT3B. All in all, we propose a simple explanation for a complex phenomenon, where the sequence selectivity of de novo DNMTs is directed by the guanine interaction capacity of these enzymes.