Most of Type II REases characterized to date were found to belong to the PD-(D/E)XK superfamily of nucleases , although theoretical predictions have been made that some REases belong to different, unrelated superfamilies. To date, predictions have been experimentally confirmed for REases from three non-PD-(D/E)XK superfamilies: R.BfiI belongs to the PLD superfamily , R.PabI exhibits a new 'half-pipe' fold  and R.KpnI and a few other enzymes belong to the HNH superfamily . Here, we present a refined alignment and the first structural model of the R.Eco29kI structure, and the first set of experimental data that support the prediction that REases may belong also to the GIY-YIG superfamily of nucleases.
According to the model, the active center of R.Eco29kI comprises amino acid residues Y49, Y76, R104, H108, E142, and N154. Alanine substitutions of the corresponding residues in I-TevI, the archetypal member of the GIY-YIG superfamily, have been shown to be critical for nuclease activity of the enzyme . The analysis of the crystal structure of the catalytic domain I-TevI confirmed the importance of these residues for the catalytic activity of the enzyme, having demonstrated that the putative catalytic residues are located on a shallow concave surface, and E75 (E142 of R.Eco29kI) was identified as a divalent cation-binding site . We constructed mutants of R.Eco29kI with amino acid substitutions Y49A, R104A, H108F, E142A and N154L, and R86A. Proteins with Ala and Phe substitutions of Y76 and H108A and N154A were not obtained. We did not find plasmids with required substitutions possibly because of toxicity of the resulting mutant proteins. We selected Phe as a replacement for His and Leu as a replacement for Asn because they are of nearly the same size and are chemically inert. The Ala substitution of R86 residue, predicted in this work to be functionally irrelevant, has no effect on the catalysis or binding as it should. The data on the specificity and mechanism of binding show that, although the introduced substitutions change the efficiency of binding of mutant protein forms, the spatial structure of constructed mutant forms of R.Eco29kI undergoes no dramatic changes.
Substitutions Y49A, H108F, and E142A completely eliminate the nuclease activity of R.Eco29kI just as I-TevI mutants Y6A and E75A. Further, substitutions R104A and N154L greatly reduce the R.Eco29kI catalytic activity, but do not abolish it completely. It is noteworthy that the residual nuclease activity was also found in the mutants of I-TevI, in which R27 and N90 (homolog's of R104 and N154 in R.EcoR29kI), were replaced by Ala [13, 14]. The efficiency of DNA-binding of Y49A, R104A and E142A mutants proved to be slightly reduced, whereas in N154L and in particular in H108F it is more significantly compromised. These data show that all the above-mentioned residues are to some extent involved in DNA-binding, but in the case of Y49, R104, and E142 the effect of mutations concerns mostly catalysis, while H108 and N154 are probably mostly important for DNA binding (as of today, it is difficult to determine the contribution of these residues to catalysis, the efficiency of which obviously depends on binding). It should be emphasized that residues R27 and E75 in I-TevI, homologous to residues R104 and E142 of R.Eco29kI, are of critical importance for catalysis . All these results are in very good agreement with the structural model presented in this work and with our prediction that R.Eco29kI, unlike all other REases studied to date, belongs to the GIY-YIG superfamily of nucleases . Importantly, modeling combined with mutagenesis allowed us to verify and refine the previous prediction of the active site in R.Eco29kI enzyme, which was based solely on the sequence considerations , demonstrating the utility of the methods for protein structure prediction. We would like to emphasize that our analysis relies primarily on the detection of similarities between R.Eco29kI and I-TevI, and does not represent an attempt to identify the functionally important residues of R.Eco29kI ab initio. There is no direct biochemical and biophysical data on the precise role of individual residues in the catalytic mechanism of DNA cleavage by I-TevI, therefore we refrain from making excessive speculations about the exact roles of the corresponding residues in R.EcoR29kI. On the other hand, any refinement of the I-TevI mechanism of action and the studies on the function of its residues may be reflected in appropriate adjustments of the interpretation of the R.Eco29kI model presented here.
It is noteworthy that at such a low level of similarity as between R.Eco29kI and I-TevI, one can expect some local structural divergence to occur between protein structures. In particular, secondary structure elements that are not directly hydrogen-bonded are prone to mutual translations or rotations induced by repacking of side chains in the protein core. Such shifts of homologous helices have been indeed observed between the remotely related GIY-YIG nucleases I-TevI and UvrC . Therefore, the root mean square deviation of the presented R.Eco29kI model from the true (currently unknown) structure of this protein can turn out to be substantial, in the range of similarity between I-TevI and UvrC, i.e. 2.9 Å for the 74 structurally superimposable core residues. Nonetheless, we are confident that the mutual orientation of major secondary structure elements that constitute the GIY-YIG fold, and the functionally important residues of R.Eco29kI studied in this work, is very accurate.
Extreme problems with the aggregation of R.Eco29kI at high concentrations (data not shown) hampered the attempts to obtain crystals or concentrated solutions suitable for NMR analyses, therefore the theoretical model of R.Eco29kI structure developed and validated in this work will serve as a starting point for computational docking and experimental analyses aimed at the identification of residues important for specific recognition of DNA by R.Eco29kI and understanding of the cleavage mechanism and architecture of the protein-DNA complex.