Mammalian DNA methylation patterns are established by two de novo DNA methyltransferases, DNMT3A and DNMT3B, which exhibit both redundant and distinctive methylation activities. However, the related molecular basis remains undetermined. Through comprehensive structural, enzymology and cellular characterization of DNMT3A and DNMT3B, we here report a multi-layered substrate-recognition mechanism underpinning their divergent genomic methylation activities. A hydrogen bond in the catalytic loop of DNMT3B causes a lower CpG specificity than DNMT3A, while the interplay of target recognition domain and homodimeric interface fine-tunes the distinct target selection between the two enzymes, with Lysine 777 of DNMT3B acting as a unique sensor of the +1 flanking base. The divergent substrate preference between DNMT3A and DNMT3B provides an explanation for site-specific epigenomic alterations seen in ICF syndrome with DNMT3B mutations. Together, this study reveals distinctive substrate-readout mechanisms of the two DNMT3 enzymes, implicative of their differential roles during development and pathogenesis.

Somatic DNMT3A mutations at R882 are frequently observed in AML patients including the very abundant R882H, but also R882C, R882P and R882S. Using deep enzymology, we show here that DNMT3A-R882H has more than 70-fold altered flanking sequence preferences when compared with wildtype DNMT3A. The R882H flanking sequence preferences mainly differ on the 3′ side of the CpG site, where they resemble DNMT3B, while 5′ flanking sequence preferences resemble wildtype DNMT3A, indicating that R882H behaves like a DNMT3A/DNMT3B chimera. Investigation of the activity and flanking sequence preferences of other mutations of R882 revealed that they cause similar effects. Bioinformatic analyses of genomic methylation patterns focusing on flanking sequence effects after expression of wildtype DNMT3A and R882H in human cells revealed that genomic methylation patterns reflect the details of the altered flanking sequence preferences of R882H. Concordantly, R882H specific hypermethylation in AML patients was strongly correlated with the R882H flanking sequence preferences. R882H specific DNA hypermethylation events in AML patients were accompanied by R882H specific mis-regulation of several genes with strong cancer connection, which are potential downstream targets of R882H. In conclusion, our data provide novel and detailed mechanistic understanding of the pathogenic mechanism of the DNMT3A R882H somatic cancer mutation.

DNA methylation maintenance by DNMT1 is an essential process in mammals but molecular mechanisms connecting DNA methylation patterns and enzyme activity remain elusive. Here, we systematically analyzed the specificity of DNMT1, revealing a pronounced influence of the DNA sequences flanking the target CpG site on DNMT1 activity. We determined DNMT1 structures in complex with preferred DNA substrates revealing that DNMT1 employs flanking sequence-dependent base flipping mechanisms, with large structural rearrangements of the DNA correlating with low catalytic activity. Moreover, flanking sequences influence the conformational dynamics of the active site and cofactor binding pocket. Importantly, we show that the flanking sequence preferences of DNMT1 highly correlate with genomic methylation in human and mouse cells, and 5-azacytidine triggered DNA demethylation is more pronounced at CpG sites with flanks disfavored by DNMT1. Overall, our findings uncover the intricate interplay between CpG-flanking sequence, DNMT1-mediated base flipping and the dynamic landscape of DNA methylation.

Epigenome editing is a promising technology, potentially allowing the stable reprogramming of gene expression profiles without alteration of the DNA sequence. Targeted DNA methylation has been successfully documented by many groups for silencing selected genes, but recent publications have raised concerns regarding its specificity. In the current work, we developed new EpiEditors for programmable DNA methylation in cells with a high efficiency and improved specificity. First, we demonstrated that the catalytically deactivated Cas9 protein (dCas9)-SunTag scaffold, which has been used earlier for signal amplification, can be combined with the DNMT3A-DNMT3L single-chain effector domain, allowing for a strong methylation at the target genomic locus. We demonstrated that off-target activity of this system is mainly due to untargeted freely diffusing DNMT3A-DNMT3L subunits. Therefore, we generated several DNMT3A-DNMT3L variants containing mutations in the DNMT3A part, which reduced their endogenous DNA binding. We analyzed the genome-wide DNA methylation of selected variants and confirmed a striking reduction of untargeted methylation, most pronounced for the R887E mutant. For all potential applications of targeted DNA methylation, the efficiency and specificity of the treatment are the key factors. By developing highly active targeted methylation systems with strongly improved specificity, our work contributes to future applications of this approach.

DNA methyltransferases interact with their CpG target sites in the context of variable flanking sequences. We investigated DNA methylation by the human DNMT3B catalytic domain using substrate pools containing CpX target sites in randomized flanking context and identified combined effects of CpG recognition and flanking sequence interaction together with complex contact networks involved in balancing the interaction with different flanking sites. DNA methylation rates were more affected by flanking sequences at non-CpG than at CpG sites. We show that T775 has an essential dynamic role in the catalytic mechanism of DNMT3B. Moreover, we identify six amino acid residues in the DNA-binding interface of DNMT3B (N652, N656, N658, K777, N779, and R823), which are involved in the equalization of methylation rates of CpG sites in favored and disfavored sequence contexts by forming compensatory interactions to the flanking residues including a CpG specific contact to an A at the +1 flanking site. Non-CpG flanking preferences of DNMT3B are highly correlated with non-CpG methylation patterns in human cells. Comparison of the flanking sequence preferences of human and mouse DNMT3B revealed subtle differences suggesting a co-evolution of flanking sequence preferences and cellular DNMT targets.