Specificity of MLL1 and TET3 CXXC domains towards naturally occurring cytosine modifications.

CXXC domains have traditionally been considered as CpG specific DNA binding domains that are repelled by cytosine modifications. This view has recently been challenged by the demonstration that CXXC domain of TET3 has relaxed sequence specificity and binds with the highest affinity to symmetric DNA duplex containing 5caCpG. Here, we present a comparative analysis of the MLL1-CXXC and TET3-CXXC sequence specificity and tolerance to cytosine modifications (5-methyl, 5-hydroxymethyl, 5-formyl, 5-carboxyl) in CpG and non-CpG context. For the first time, we take into consideration possible interference from cytosine bases elsewhere in the sequence. We show that despite similar overall structure, MLL1-CXXC has greater sequence and modification specificity than TET3-CXXC. MLL1-CXXC is specific only for CpG and does not tolerate any cytosine modifications. In contrast, TET3-CXXC does not require the CpG context of cytosine bases. Methyl-, formyl- and carboxyl-modifications are tolerated by TET3-CXXC, but only preceding G. Based on our and other data we propose a parsimonious model of MLL1-CXXC and TET3-CXXC DNA binding. This model explains why the binding of modified DNA duplexes by TET3-CXXC requires in some cases a register shift and is therefore context-dependent.