Water-mediated interactions enable transcription factors to specifically recognize dinucleotides

Transcription factors bind to DNA by recognizing specific bases within their binding motifs. Binding to each DNA mononucleotide within the motif often contributes independently to total binding energy. However, some TFs can bind to DNA more specifically than predicted by this model, by directly recognizing DNA dinucleotides. To understand this process, we have solved the structures of the basic helix-loop-helix protein MYF5, and the homeodomain protein BARHL2 together with DNA containing a set of dinucleotides that have different affinities to the proteins at very high resolution (< 1 Å). Although dinucleotides are known to determine DNA shape and rigidity, our results show that DNA backbone shape does not explain the observed differences in TF affinity. In contrast, we observe that dinucleotides are recognized by extensive water networks that connect the adjacent bases. Our results uncover the enigma of how TFs can recognize more complex local features than mononucleotides, and suggest that considering water-mediated recognition will be important in predicting affinities of macromolecules from their sequence.