Structural and Dynamic Features of the MutT Protein in the Recognition of Nucleotides with the Mutagenic 8-Oxoguanine Base*

Escherichia coli MutT hydrolyzes 8-oxo-dGTP to 8-oxo-dGMP, an event that can prevent the misincorporation of 8-oxoguanine opposite adenine in DNA. Of the several enzymes that recognize 8-oxoguanine, MutT exhibits high substrate specificity for 8-oxoguanine nucleotides; however, the structural basis for this specificity is unknown. The crystal structures of MutT in the apo and holo forms and in the binary and ternary forms complexed with the product 8-oxo-dGMP and 8-oxo-dGMP plus Mn2+, respectively, were determined. MutT strictly recognizes the overall conformation of 8-oxo-dGMP through a number of hydrogen bonds. This recognition mode revealed that 8-oxoguanine nucleotides are discriminated from guanine nucleotides by not only the hydrogen bond between the N7-H and Oδ (N119) atoms but also by the syn glycosidic conformation that 8-oxoguanine nucleotides prefer. Nevertheless, these discrimination factors cannot by themselves explain the roughly 34,000-fold difference between the affinity of MutT for 8-oxo-dGMP and dGMP. When the binary complex of MutT with 8-oxo-dGMP is compared with the ligand-free form, ordering and considerable movement of the flexible loops surrounding 8-oxo-dGMP in the binary complex are observed. These results indicate that MutT specifically recognizes 8-oxoguanine nucleotides by the ligand-induced conformational change.

that, interestingly, OGG1 and MutM do not recognize the O8 atom, which is the most characteristic feature of the 8-oxoG moiety, and the interaction observed between the O8 atom and the main-chain atom of MutY is relatively weak. Alternatively, OGG1, MutM, and MutY commonly discriminate 8-oxoG from G by the protonation at N7 accompanied by the oxidation of C8. Structural studies on various enzymes that recognize 8-oxoG have succeeded in explaining the mechanism by which 8-oxoG is discriminated from normal G in DNA, but one of the most interesting questions to be elucidated is the mechanism by which MutT acquires extremely high substrate specificity for 8-oxoG compared with the other enzymes.
MutT belongs to the Nudix (nucleoside diphosphate linked to some other moiety, X) hydrolase family (17). Nudix family members have a highly conserved MutT signature (Nudix motif); i.e. GX 5 EX 7 REUXEEXGU, where U is a hydrophobic residue and X is any amino acid. Current genome analyses have found a large number of open reading frames containing the MutT signature, but their functions, i.e. their substrates, are not identified in the case of almost all these proteins because of a lack of homology outside the MutT signature. MutT is the most examined protein in this family. Its structure was first determined by NMR (18) and has greatly contributed to the study of the Nudix hydrolase family. NMR studies of MutT with its product, 8-oxo-dGMP, have predicted several recognition models of 8-oxo-dGMP (19). However, the precise recognition mechanism of 8-oxoG nucleotides remains unclear. Therefore, it is necessary to determine the crystal structures of MutT to explain the extremely high substrate specificity of MutT for 8-oxoG nucleotides.
Here, we present x-ray crystallographic analyses of the apo enzyme; the Mn 2ϩ -bound holo enzyme (MutT-Mn 2ϩ ); the binary complex with 8-oxo-dGMP, a reaction product (MutT-8-oxo-dGMP); and the tertiary complex with 8-oxo-dGMP and Mn 2ϩ (MutT-8-oxo-dGMP-Mn 2ϩ ). These structures have revealed the mechanism of the extremely high substrate specificity of MutT for 8-oxoG nucleotides and have allowed us to propose the exact roles of some conserved residues in the MutT signature.

EXPERIMENTAL PROCEDURES
Protein Expression and Purification-The E. coli strain BL21 (DE3) harboring a newly constructed pET8c/MutT plasmid was used for the expression of native and selenomethionine (SeMet)-substituted MutTs. Native MutT was overexpressed in Luria-Bertani (LB) broth, and SeMet MutT was overexpressed in LeMaster broth containing seleno-DL-methionine instead of methionine with sufficient amounts of isoleucine, lysine, and threonine to inhibit the methionine pathway (20,21). This condition was also present in the overexpression of SeMet hMTH1 (22). Purification of MutT was carried out by almost the same procedure (except that the hydroxyapatite column chromatography step was skipped), as described previously (23). DEAE-Sepharose and the HiPrep 16/60 Sephacryl S-200 HR column were substituted for DEAE-Sephacel and the Sephadex G-75 column, respectively. The purified protein solution was concentrated to ϳ6 mg/ml. Crystallization-The native and SeMet-substituted apo forms and all complexes were crystallized by hanging drop vapor diffusion at 288 K. Crystals of native and SeMet-substituted forms were obtained from a droplet containing 3 mg/ml protein, 10 mM Tris-HCl (pH 7.5), 0.5 mM EDTA, 2.5% glycerol, 0.5 mM 2-mercaptoethanol, 0.7 M potassium sodium tartrate, and 44 mM HEPES-NaOH (pH 7.5) equilibrated against a reservoir containing 1.4 M potassium sodium tartrate and 87 mM HEPES-NaOH (pH 7.5). Crystals of MutT-Mn 2ϩ were obtained in the same manner, as described above, except that 10 mM MnCl 2 was added to the droplet. The crystallizations of MutT-8-oxo-dGMP and MutT-8-oxo-dGMP-Mn 2ϩ were described previously (24). The crystals were transferred to a cryosolution of each reservoir containing 30% sucrose and were then flash frozen.
Data Collection, Processing, Phasing, and Structure Refinement-Diffraction data were collected at 100 K on beamline 18B of the Photon Factory (Tsukuba, Japan) and on beamlines 41XU, 44XU, 38B1, and 40B2 of SPring-8 (Harima, Japan). The data for native and SeMet derivative forms were processed and scaled by DSP/MOSFLM and SCALA (25). The data for MutT-Mn 2ϩ were processed and scaled by DENZO and SCALEPACK (26). There are two molecules in the asymmetric unit with V M of 2.2 (native MutT) and 2.5 (MutT-Mn 2ϩ ) Å 3 ⅐Da Ϫ1 (27). Data collection statistics of the best data used for structure determination and refinements are listed in Table 1. Data collection statistics of MutT-8-oxo-dGMP and MutT-8oxo-dGMP-Mn 2ϩ are quoted from the reference by Nakamura et al. (24).
The positions of eight selenium atoms were determined using SOLVE (28). The initial phases were calculated using MLPHARE (29) and improved using DM (30). The initial model was built using TOM (31) and O (32). The model was refined using X-PLOR (33) and CNS (34). Using the model of the SeMet derivative, the successive refinement of native MutT converged at an R value of 20.4% and an R free of 23.1% for reflections in the resolution range 20 -1.8 Å. The structure of MutT-8-oxo-dGMP was solved by molecular replacement with AMoRe (35) using the structure of the native apo form as a search model. The 2F o Ϫ F c maps after CNS refinements clearly showed the density for 8-oxo-dGMP and the conformationally changed loop regions (L-A and L-D). These regions were manually built and fitted into the density with O. The structure of MutT-8oxo-dGMP-Mn 2ϩ was refined starting with the coordinates of the MutT-8-oxo-dGMP. The structure of MutT-Mn 2ϩ was solved by molecular replacement with AMoRe by using the structure of the apo form as a search model. The stereochemical qualities of the structures were checked by PROCHECK (36); the refinement statistics are listed in Table 2. Superposition of MutT structures were carried out using Lsqkab (37). All molecular graphics were prepared using PyMOL (38).

RESULTS AND DISCUSSION
Overall Structures of MutT and MutT-8-oxo-dGMP-The crystal structures of the MutT apo and MutT-8-oxo-dGMP complex forms were determined at a resolution of 1.8 and 1.96 Å, respectively. MutT is composed of two ␣-helices (␣-1 and ␣-2) and six ␤-strands (␤-1 to ␤-6) ( Figs. 1 and 2A); it adopts an ␣-␤-␣ sandwich structure that is conserved among members of the Nudix family. The Nudix motif (23 residues from Gly-38 i.e. the MutT signature (GX 5 EX 7 REUXEEXGU), adopts the characteristic strand-loop-helix-loop (SLHL) structure formed by ␤-3Ј, L-B, ␣-1, and L-C (39,40). The crystal of the apo form contains two protein molecules per asymmetric unit, and they are very similar to each other with root mean square deviation (r.m.s.d.) of 0.5 Å for the corresponding 121 C␣ atoms. For simplicity, only one molecule will be referred to in all further discussions. MutT exists as a monomer, which is found in the MutT-8-oxo-dGMP crystal.
In the apo form, the electron densities of L-A connecting ␤-2 and ␤-3 are not available, indicating that the L-A loop region has a highly flexible conformation ( Fig. 2A). On the other hand, in the MutT-8-oxo-dGMP complex, the ordering of the flexible     (18,19).
Recognition Scheme of 8-oxo-dGMP by MutT-When 8-oxo-dGMP binds to MutT, large ligand-induced conformational changes occur in the L-A and L-D regions, namely, the ordering of the flexible L-A loop and considerable movement of L-A and L-D to the surrounding 8-oxo-dGMP (Figs. 2C and 3A). The side chains of Arg-23 and His-28 on L-A form hydrogen-bonding interactions with phosphate and sugar moieties of 8-oxo-dGMP, respectively, whereas Arg-78 interacts with the sugar moiety through a water-mediated hydrogen bond. Asp-77 and Arg-78 make two hydrogen bonds between their side chains. The conformational change of the L-A and L-D regions also produces the water molecule-mediated interaction between His-28 and Asp-77 and the CHinteraction between His-28 and Phe-75 ( Fig. 3A and supplemental Fig. S2). Thus, the loops L-A and L-D connect to each other, resulting in the formation of a cave composed of ␤-1, ␤-3, ␤-3Ј, ␤-5, and ␣-2 for substrate binding (Fig. 3B). 8-Oxo-dGMP is inserted deeply into the cave in which the wall on one side is filled with hydrophobic residues (Leu-4, Ile-6, Val-8, Ile-80, and Leu-82), and the other side and bottom include some polar residues (Arg-23 and Asn-119). The 8-oxoG base and the deoxyribose are perfectly buried, and the phosphate group faces the solvents (Fig. 3B). The glycosidic conformation of 8-oxo-dGMP bound to MutT is syn. This fact is consistent with the first suggestion by Bessman et al. that MutT may recognize the syn conformation, because the 8-substituted purine nucleotides were better substrates compared with the normal purine nucleotides (42). The sugar ring puckering and the sugar-phosphate backbone conformation of 8-oxo-dGMP are C2Ј-endo and gauche Ϫ -trans, respectively. These conformations are generally observed in 8-substituted purine nucleosides and 5Ј-nucleotides (43).
These ligand-induced conformational changes result in the strict recognition of the overall structure of 8-oxo-dGMP by MutT through a number of hydrogen bonds (Fig. 3C). The characteristic features of the 8-oxoG base are the oxygen atom (O8) at C8 and the hydrogen atom (N7-H) at N7 accompanied by oxidation. MutT recognizes N7-H of 8-oxoG by a hydrogen bond with O␦ of Asn-119 (Fig. 3C, a red dashed line).  82, Ile-80, and Ala-118 (Fig. 3D). In addition, the carbonyl oxygen of Gly-37 forms water molecule-mediated hydrogen bonds with N2-H of 8-oxoG and the phosphate oxygen. The side chain of His-28 is directly hydrogen-bonded to the O3Ј atom of the deoxyribose. The O3Ј atom also forms a hydrogen bond with a water molecule, binding to the side chain of Arg-78. The phosphate group forms a hydrogen bond with the side chain of Arg-23 and a water molecule-mediated hydrogen bond with the main chain of Lys-39. In summary, 8-oxo-dGMP is surrounded by 12 types of hydrogen bonds. The hydrogen-bonding interactions with the pyrimidine moiety and the ␣-phosphate group in MutT-8-oxo-dGMP are similar to those with the corresponding pyrimidine moieties and ␣-phosphate groups in the structures of BdRppH-GTP-Mg 2ϩ and BdRppH-dGTP (BdRppH-(d)GTPs) (40). The positions of the base moieties of (d)GTPs with the syn conformation in BdRppH-(d)GTPs accord with that of 8-oxo-dGMP in MutT-8-oxo-dGMP with an r.m.s.d. of 0.6 Å for the corresponding 11 atoms when proteins are superimposed. BdRppH with Arg-40, Phe-52, and Asn-136 residues corresponding to Arg-23, Phe-35, and Asn-119 of MutT, respectively, recognizes N1-H, N2-H, and O6 of the pyrimidine moiety by four hydrogen bonds with Phe-52 and Asn-136; P␣-O of the ␣-phosphate group is recognized by a hydrogen bond with Arg-40. This recognition mode is the same as that observed in MutT. Apart from the similarities, differences are found in recognition of the imidazole moiety of the base and the sugar moiety as well as in the ligand-induced conformational change. The imidazole and sugar moieties of (d)GTPs in BdRppH-(d)GTPs do not form hydrogen bonds with any residues in BdRppH. In addition, although ligand-induced conformational change with loop ordering is observed in BdRppH, the change is significantly small (ϳ2-4 Å) as compared with that in MutT. The large ligand-induced conformational change observed in MutT does contribute to its high affinity for 8-oxoG nucleotides, as discussed below. There is a large discrepancy between the K m values of 0.081 and 268 M for the hydrolysis of 8-oxo-dGTP by MutT and of dGTP by BdRppH, respectively (7,44). This may derive from these structural differences, in addition to the unfavorable syn conformation of (d)GTPs in BdRppH-(d)GTPs.
As a result of the strict recognition of 8-oxo-dGMP with the large conformational change, there are low B factors and unambiguous electron densities around 8-oxo-dGMP (supplemental Fig. S1B and Fig. 3E). The average B factor of 8-oxo-dGMP is 12.4 Å 2 and that of the residues involved in the recognition of 8-oxo-dGMP, Arg-23, His-28, Phe-35, Asp-77, Arg-78, and Asn-119 is 13.6 Å 2 . These low B factors and unambiguous electron densities represent the small thermal motion and/or the ordered positioning of 8-oxo-dGMP and the residues of the active site in the crystal lattice. This phenomenon explains isothermal titration calorimetry experiments (45), indicating that the tight binding of 8-oxo-dGMP to MutT (⌬G ϭ Ϫ9.8 kcal/ mol) is driven by a highly favorable enthalpy (⌬H ϭ Ϫ39.0 kcal/ mol) with an unfavorable entropy (ϪT⌬S ϭ 29.2 kcal/mol). The unfavorable entropy would be a result of the conformational rigidity generated from the connection of loops L-A and L-D with large ligand-induced conformational changes. On the other hand, the more favorable enthalpy would be produced by the large number of hydrogen bonds and van der Waals interactions formed between 8-oxo-dGMP and MutT; this is sufficient to compensate for the unfavorable entropy and to bind tightly.
Furthermore, the hydrogen bond-mediated recognition mode found in MutT-8oxo-dGMP is comparable with the results of mutational studies in which it was found that the R78A, N119D, and N119A mutants show 7-, 37-, and 1650-fold decreases in affinity for 8-oxo-dGMP in comparison with the wild type and that they lose binding free energies (⌬⌬G) of 1.1, 2.1, and 4.3 kcal/mol, respectively, as measured by the increases in K I (46). According to previous reports, the contribution of the hydrogen bond to protein stability can be estimated as ϳ2 and 1.2 kcal/mol for hydrogen bonds between protein residues and for those between a water molecule and a protein residue, respectively (47,48). Judging from the MutT-8oxo-dGMP structure, the N119D and N119A mutants lose one hydrogen bond between the O6 of 8-oxo-dGMP and the N␦-H of Asn-119 and two hydrogen bonds involving the O6 and N7-H of 8-oxo-dGMP and the amide group of Asn-119, respectively (Fig. 3C). The R78A mutant loses the hydrogen bond to a water molecule (Fig. 3C); that is, ⌬⌬G losses of 1.2, 2, and 4 kcal/mol are estimated for the R78A, N119D, and N119A mutants, respectively; this agrees perfectly with the experimental data (46).
The feature of the substrate-binding site in MutT-8-oxo-dGMP is also consistent with reports that MutT hydrolyzes both deoxyribose and ribose derivatives of 8-oxoG nucleotides with similar efficiency (2,7). This is because, despite a number of hydrogen-bonding interactions between 8-oxo-dGMP and MutT, there is space for a hydroxyl group instead of the hydrogen atom at the 2Ј position of the sugar ring (Fig. 3B). This recognition mechanism of 8-oxo-dGMP by MutT is different from any models predicted by NMR studies (PDB IDs: 1PPX, 1PUN, 1PUQ, 1PUS, and 1MUT) (supplemental Fig. S3).
Discrimination of 8-oxoG Nucleotides from G Nucleotides-The K m values of E. coli MutT for 8-oxo-dGTP and 8-oxo-GTP are ϳ3,800to 14,000-fold lower than the values for the corresponding G nucleotides (7). These data agree with the observation that the K d value (52 nM) between 8-oxo-dGMP and MutT is 34,000-fold lower than that (1.76 mM) between dGMP and MutT (45). Thus, the most important question that this study addresses is the mechanism by which MutT obtains high substrate specificity for 8-oxoG nucleotides as compared with G nucleotides.
According to the recognition scheme of 8-oxo-dGMP by MutT, the major difference in the recognition of 8-oxoG versus G is whether the single hydrogen bond between O␦ of Asn-119 and N7-H of 8-oxoG occurs or not. This situation is similar to those of OGG1, MutM, and MutY (13,15,16). If the side-chain conformation of Asn-119 in the MutT complex with G nucleotides was the same as that in the MutT-8oxo-dGMP structure, the two lone pairs at N7 of G and O␦ of Asn-119 would be repulsive (supplemental Fig. S4). To avoid this repulsion, a rotation about the side-chain torsion angles in Asn-119 should be required. Thus, the difference in the number of hydrogen bonds formed between MutT-G and MutT-8-oxoG complexes is only one. The contribution of one hydrogen bond to ⌬⌬G is estimated to be 2 kcal/mol (47,48).
The syn glycosidic conformation of 8-oxo-dGMP must also be one of the elements contributing to the substrate specificity of MutT, because 8-oxoG nucleotides favor a syn conformation by the steric hindrance between O8 and the sugar moiety; this is in contrast with G nucleotides that adopt both syn and anti conformations (49,50). Because of the lack of quantitative data on the preference of the syn conformation about the 8-oxoG nucleotides, it is difficult to estimate the free energy difference between the syn and anti conformations. However, it is unlikely that the preference of MutT for the syn conformation over the anti conformation is more than 10-fold (⌬⌬G Ϸ 1.4 kcal/mol), because the anti conformation for 8-oxoG nucleotides is sometimes observed in the crystal structures of 8-oxoG recognition complexes such as OGG1 and MutY (15,16).
These discrimination factors cannot by themselves explain the roughly 34,000-fold difference between the binding affinity of MutT for 8-oxo-dGMP and dGMP (⌬⌬G Ϸ 6 kcal/mol). When 8-oxo-dGMP binds to MutT, large ligand-induced conformational changes with an ordering of loop regions are observed. On the other hand, in the binding of dGMP to MutT, the thermal parameters were ⌬G ϭ Ϫ3.7, ⌬H ϭ Ϫ3.3, and ϪT⌬S ϭ Ϫ0.4 kcal/mol; the changes in backbone 15 N and NH chemical shifts in 22 residues; and the slowing down of the NH exchange with D 2 O of 20 residues are remarkably different when compared with the changes (the thermal parameters of ⌬G ϭ Ϫ9.8, ⌬H ϭ Ϫ39.0, and ϪT⌬S ϭ 29.2 kcal/mol; the changes in backbone 15 N and NH chemical shifts in 62 residues; and the slowing down of the NH exchange with D 2 O of 45 residues) involved in the binding of 8-oxo-dGMP to MutT (19,45). These facts suggest that no significant conformational change in MutT is observed when dGMP binds to MutT. The large ligand-induced conformational change in MutT also contributes to the discrimination of 8-oxoG nucleotides from G nucleotides.
A comparison of the amino acid sequences of MutT-related enzymes suggests that the enzymes with higher substrate specificity for 8-oxoG are only MutT homologs from closely related species with conserved amino acids in the positions that participate in the recognition of 8-oxoG and the conformational change (Fig. 1, green asterisk). In fact, these amino acids are not highly conserved among E. coli Orf135 (51), Bacillus subtilis YtkD (52), and hMTH1 (8)  MutT-8-oxo-dGMP-Mn 2ϩ and MutT-Mn 2ϩ , to determine metal-binding sites at the MutT signature of MutT. The crystal of MutT-Mn 2ϩ contains two proteins per asymmetric unit. Their overall structures are very similar, with an r.m.s.d. of 0.3 Å for the corresponding 118 C␣ atoms; for simplicity, only one molecule will be referred to in all further discussion. The structure of MutT-Mn 2ϩ is similar to that of the apo form, with an r.m.s.d. of 0.6 Å for the corresponding 118 C␣ atoms.
In the MutT signature having an SLHL structure (Fig. 4A), Gly-38, Glu-44, Arg-52, Glu-53, Glu-56, Glu-57, and Gly-59 are completely conserved among the members of the Nudix family (Fig. 1). The SLHL structure of MutT is similar to those of other enzymes in the Nudix family. For example, the conserved 23 residues can be superimposed on those found in the Pyrobaculum aerophilum Nudix protein with an r.m.s.d. of 0.5 Å (53). A characteristic feature in the SLHL structure is the hydrogenbonding network centering on the converged Arg-52 that anchors ␣-1 to its connecting loop. The side chains of Glu-44 and Arg-52, which form two hydrogen bonds with each other, participate in hydrogen bonding to the main chain atoms in the nonconserved Glu-41 and Lys-39 residues, respectively. Arg-52 also interacts with the side chains of the conserved Glu-53 and Glu-56 residues (Fig. 4A). This hydrogen-bonding network is roughly the same as that in the other Nudix proteins and contributes to the conformational stability of the SLHL structure (39,40,(53)(54)(55)(56)(57)(58)(59)(60)(61)(62)(63)(64)(65). In the electron density maps of the MutT-Mn 2ϩ crystal produced by co-crystallization, there were three peaks corresponding to the metal ions near the MutT signature. Judging from the peak heights, B factors, bond lengths, and bond angles, we determined that two peaks were Mn 2ϩ ions (nearly ideal octahedral coordination and an average bond length of ϳ2.2 Å) and one was Na ϩ (distorted octahedral coordination and an average bond length of ϳ2.5 Å) (Fig. 4B). Furthermore, the two Mn 2ϩ sites were also confirmed from significant densities (Ͼ10 level) on the anomalous difference Fourier map at ϭ 1 Å (data not shown). On the other hand, the densities of the Na ϩ site were less than noise level. The refined MutT-Mn 2ϩ structure reveals that two Mn 2ϩ ions form a binuclear metal center with a bridged water molecule. One Mn 2ϩ coordinates to the oxygen atoms of Glu-53, Glu-57, three water molecules, and Asp-77 in another molecule (Asp-77*). The coordination partners of another Mn 2ϩ are Glu-53, L-tartrate (a crystallization reagent), and four water molecules. Na ϩ coordinates to the oxygen atoms of Gly-37, Glu-57, Asp-77*, and two water molecules. Asp-77* and L-tartrate bind to the MutT signature through metal ions, but do not distort the SLHL structure.
In the MutT-8-oxo-dGMP-Mn 2ϩ crystal (prepared by soaking MutT-8-oxo-dGMP crystals in reservoir supplemented with 1 mM MnCl 2 ), Mn 2ϩ binds to the six oxygen atoms of the main chain of Gly-37, the side chain of Glu-57, the phosphate group, and three water molecules with nearly ideal octahedral coordination (Fig. 4A). The binding of Mn 2ϩ to the MutT-8oxo-dGMP binary complex makes the phosphate group move slightly toward Mn 2ϩ (the P atom moves by 0.7 Å). The position of Mn 2ϩ observed in MutT-8-oxo-dGMP-Mn 2ϩ is close to that of Na ϩ in MutT-Mn 2ϩ (at a distance of 1.4 Å). The three Mn 2ϩ binding sites consisting of the Mn 2ϩ in MutT-8-oxo-dGMP-Mn 2ϩ and two additional Mn 2ϩ ions in MutT-Mn 2ϩ (Fig. 4C) are located at common metal-binding sites observed in the Nudix family (60) and correspond to those observed in the ternary complexes of E. coli ADP-ribose pyrophosphatase, Mycobacterium tuberculosis ADP-ribose pyrophosphatase, Caenorhabditis elegans diadenosine 5Ј,5ЉЈ-P 1 ,P 4 -tetraphosphate pyrophosphohydrolase, Xenopus laevis X29, human NUDT5, Thermus thermophilus Ndx2, and BdRppH (40, 55, 57, 64 -67). Thus, the three sites are considered to be candidates for metal binding in 8-oxo-dGTP hydrolysis. Previous kinetic studies have shown that MutT binds to one Mn 2ϩ in the absence of nucleotides and two Mn 2ϩ ions in the presence of a nonspecific substrate analog, AMPCPP (68). The middle Mn 2ϩ (Mn2 in Fig.  4C) may be prebound to the active site, judging from the number of coordination partners in MutT and the Mn 2ϩ ion (Mn1 in Fig. 4C) found in MutT-8-oxo-dGMP-Mn 2ϩ ; otherwise, it and other Mn 2ϩ ion (Mn1 and Mn3, respectively in Fig. 4C) would be recruited with the substrate. The probability of the three metals binding to MutT in the presence of the real substrate 8-oxo-dGTP cannot be neglected because of the fact that the number of binding metals depends on the kinds of substrate analogs (40, 54, 60, 61, 64 -66).
The structures of MutT-8-oxo-dGMP-Mn 2ϩ and MutT-Mn 2ϩ suggest structural insights into some essential or important residues for the 8-oxo-dGTP hydrolysis. Glu-53 and Glu-57 are essential for the suppression of spontaneous A:T to C:G transversion mutations (69), and E53Q and E57Q mutants decrease k cat by 10 4 to 10 5 -fold (70). On the other hand, Glu-56 is nonessential for the suppression of the mutations (69), and E56Q and E98Q mutants have relatively small effects (Ͻ24fold) on k cat (70). These results agree with our structural studies showing that essential residues, Glu-53 and Glu-57, directly bind to metal ions, whereas important residues, Glu-56 and Glu98, make water-mediated interactions with metal ions (Fig.  4, A and B). Gly-37 and Gly-38, which are located at the surface of the ligand-binding site, are also revealed to be essential residues for the suppression of the mutations (69). The side chains of any residues except Gly in positions 37 and 38 would contact the base moiety of the nucleotide ligand and the essential residue for the catalysis, Glu-53, respectively (Fig. 4, A and C). For this reason, to express 8-oxo-dGTPase activity, residues 37 and 38 must be Gly, which has the smallest side chain.
A number of kinetic, mutational, and NMR studies of MutT using dGTP and/or a substrate analog, AMPCPP, have been reported, and a catalytic mechanism is proposed by Mildvan and coworkers (71). Compared with our structural data, there appear to be some differences in the metal-binding sites. Gly-38 is involved in metal coordination in their model, but Gly-37, instead of Gly-38, is a metal ligand in the structure of MutT-8-oxo-dGMP-Mn 2ϩ (Fig. 4, A and C). The all-crystal structures of Nudix proteins show that the carbonyl oxygen of the corresponding residue to Gly-38 participates in the formation of ␤-sheet, whereas that of Gly-37 binds to a metal ion (40, 55, 57-59, 61, 64 -67). Glu-56 and Glu98, which are metal ligands in the model proposed by Mildvan and coworkers, interact with water molecules bound to metal ions in our structures. The metal coordination scheme changes in the active site during the reaction, and a more proper enzymatic mechanism activated by metal ions might be examined by kinetic protein crystallography.