DNA methylation and hydroxymethylation introduce relatively small changes in DNA bases, yet there are many proteins that distinguish between unmethylated, methylated and hydroxymethylated DNA. In the case of cytosine methylated DNA, solvation/desolvation effects are thought to play a major role in methylation specific binding. Using restriction endonuclease R. DpnI as an example, we have studied specific binding to adenine methylated DNA. From biochemical data, two crystal structures and hydrogen-deuterium experiments, we conclude that the high specificity of the enzyme relies on a “double readout” of methylation by two separate domains (catalytic and winged helix). Each domain relies partially on desolvation for methylation detection, but in addition, adenine methylation in the GATC context generates DNA methyl groups in close proximity, which enforces DNA deformation. We suspect that R. DpnI specifically recognizes the deformed conformation, and hence ultimately relies on a methyl-methyl clash in the substrate to specifically bind adenine methylated DNA. As a model for hydroxymethylcytosine specific DNA binding, we have used the PvuRts1I, a bacterial enzyme, which specifically cleaves the DNA with 5-hydroxymethylcytosine bases. A crystal structure of PvuRts1I shows a previously unrecognized SRA domain in the enzyme, which can be expected to flip the modified base. Although our own and another group’s biochemical data regarding nucleotide flipping are still confusing, currently the most plausible model is that specific 5hmC binding relies on scrutiny of the flipped base, which is highly reminiscent of the way how many DNA repair proteins detect damaged bases.