Most of the amino acid residues contacting the substrate are
Most of the amino IOX-2 residues contacting the substrate are within the nucleotide-recognition lid (mauve or magenta in Fig. 4A). Two particularly important residues, W69 and Y76, are conserved in eubacteria and in some but not all of eight human AlkB homologs ,  (Fig. 4B). The tryptophan residue packs against the methylated base and may facilitate the catalytic mechanism by π–π and cation-π interactions with the substrate and reaction intermediates. The Y76 residue is at the centre of a H-bonding network and forms a bidentate H-bond bridging two phosphates of the nucleotide backbone that are 5’ to the methylated base helping to pin the substrate into position. Some residues within the catalytic core of AlkB also appear to tether the substrate, for example R161 in a flexible surface loop also makes a salt bridge to the phosphate 5’ to the methylated deoxyadenosine. These observations explain why good AlkB substrates including 1me-dAMP need a phosphate 5′ to the methylated base . The X-ray structure also reveals why 1meA and 3meC are better substrates than 1meG and 3meT in DNA , , . This is because D135 within the catalytic core can form a stabilizing H-bond with the external N6 vs. N4 groups of the former but not the latter bases. The N-terminal extension of the protein (cyan in Fig. 4A) forms a bridge between the catalytic core and nucleotide-recognition lid. The importance of this extension is not clear but it could stabilize the catalytic core and nucleotide-recognition subdomains whilst facilitating subtle conformational re-adjustments between these regions. Further details of the reaction mechanism have been derived from an enzyme–substrate complex exposed to limited levels of oxygenation that show partial occupancy by 2OG and 2OG/succinate. These details offer insight into the catalytic mechanism at an atomic level, for example the likely route taken by the O2 molecule as it tunnels towards the Fe2+ centre, and how the coordination geometry around the metal centre changes as the highly reactive oxy-ferryl intermediate is generated. However, these crystal structures form only snapshots of the catalytic process. It would be interesting to undertake computer simulations to demonstrate how substrate turnover may synchronise with O2 diffusion and the subtle conformational fluctuations within and between the protein subdomains. The three regions identified in the AlkB structure, the cofactor binding residues and some but not all of the substrate-binding residues are conserved in eight human AlkB homologs (Fig. 4B). This suggests that all homologs could have the potential to bind alkylated nucleic acids, however this has not been observed for ABH1  and studies of ABH4-8 have not been reported (see further discussion on AlkB homologs below). The first X-ray structure of the human AlkB homolog, ABH3, became available recently . A structural model of the catalytic core of this protein was constructed several years ago based on crystal structures of other 2OG-Fe2+-dependent dioxygenases , . The overall topology and location of conserved residues within the catalytic-site of the model are in good agreement with the X-ray structure. Although the new ABH3 structure was not generated whilst bound to DNA or RNA, significant differences to AlkB can be seen within the substrate-binding pocket. The pocket of ABH3 is considerably more polar than that of AlkB, and lacks a potential nucleotide stacking sidechain at the position corresponding to W69; however there is a Trp residue within the binding site (a few residues upstream, Fig. 4B) that could play a similar stacking role. Due to variability of the sequence and secondary structure within the nucleotide-recognition site of the human AlkB homologues (Fig. 4B), the resolution of their structures with substrates bound is needed before a clear picture of alternative binding modes can emerge. Such studies should ultimately explain why ABH2 and ABH3 are inefficient in repairing large alkylated bases in DNA or small substrates, such as 1-medAMP and the trinucleotide T(1meA)T. This could possibly result from less flexibility in the lid region and alternative requirements for sugar-phosphate backbone recognition .