Structural Characterization of Viral Ortholog of Human DNA Glycosylase NEIL1 Bound to Thymine Glycol or 5-Hydroxyuracil-containing DNA*

Background: Nei is a DNA glycosylase of the base excision repair pathway. Results: We present two crystal structures of an Nei bound to thymine glycol or 5-hydroxyuracil. Conclusion: Mutational analysis of active site residues suggests that lesion recognition happens before the damaged base is everted into the active site. Significance: These are the first structures of any Nei in complex with a damaged base. Thymine glycol (Tg) and 5-hydroxyuracil (5-OHU) are common oxidized products of pyrimidines, which are recognized and cleaved by two DNA glycosylases of the base excision repair pathway, endonuclease III (Nth) and endonuclease VIII (Nei). Although there are several structures of Nei enzymes unliganded or bound to an abasic (apurinic or apyrimidinic) site, until now there was no structure of an Nei bound to a DNA lesion. Mimivirus Nei1 (MvNei1) is an ortholog of human NEIL1, which was previously crystallized bound to DNA containing an apurinic site (Imamura, K., Wallace, S. S., and Doublié, S. (2009) J. Biol. Chem. 284, 26174–26183). Here, we present two crystal structures of MvNei1 bound to two oxidized pyrimidines, Tg and 5-OHU. Both lesions are flipped out from the DNA helix. Tg is in the anti conformation, whereas 5-OHU adopts both anti and syn conformations in the glycosylase active site. Only two protein side chains (Glu-6 and Tyr-253) are within hydrogen-bonding contact with either damaged base, and mutating these residues did not markedly affect the glycosylase activity. This finding suggests that lesion recognition by Nei occurs before the damaged base flips into the glycosylase active site.

Various types of oxidative DNA damage, including abasic sites, base lesions, and strand breaks, are caused by reactive oxygen species generated by endogenous metabolism from a cell and from exposure to ionizing radiation. The resulting DNA damage can interfere with both the efficiency and fidelity of DNA replication and transcription and therefore can lead to mutagenesis, carcinogenesis, and aging. To prevent the toxic and mutagenic effects of oxidative DNA damage, a variety of DNA repair pathways has evolved in all species. One of these pathways is base excision repair, a process initiated by DNA glycosylases, which recognize and remove oxidized DNA damage (1)(2)(3)(4)(5). These DNA glycosylases are divided into two families, the HhH superfamily and Fpg 3 /Nei family, based on the presence of particular structural motifs. The enzymes in both families are also classified based on their substrate specificity. Oxidized purines like 8-oxoguanine (8-oxoG) are substrates for Fpg (Fpg/Nei family) and Ogg (HhH superfamily). Oxidized pyrimidines, however, are recognized and removed by endonuclease VIII (Nei; Fpg/Nei family) and endonuclease III (HhH superfamily).
Members of the Fpg/Nei family share a two-domain architecture. The N-terminal domain consists of a two-layered ␤-sandwich with two ␣-helices, and the C-terminal domain contains four ␣-helices, two of which are involved in a conserved helix-two-turn-helix motif and two ␤-strands that make up the zinc finger motif (6 -10). Escherichia coli formamidopyrimidine DNA glycosylase (Fpg) was the first identified member of this family and was characterized by its ability to recognize and remove methyl-formamidopyrimidines (11). It was later concluded that the principal biological substrate for Fpg is 8-oxoG (12,13). Recent studies have shown that FapyG is also an important biological substrate (14). Structural studies of Fpg complexed with DNA containing an oxidized base have provided insight into the general features of lesion presentation in the active site of Fpg (15,16). The everted 8-oxoG is recognized and stabilized by the ␣F-␤9/10 loop, also called the 8-oxoG capping loop (17). Recently, the hypothesis that the recognition of 8-oxoG might occur before it flips out of the DNA helix was put forward (17).
Mimivirus Nei1 (MvNei1) is an ortholog of human NEIL1, and both enzymes belong to the Nei family of DNA glycosylases (18 -21). Nei enzymes are bifunctional glycosylases, with glyco-* This work was supported, in whole or in part, by National Institutes of Health sylase and lyase activities. Nei glycosylases mainly recognize and cleave oxidized pyrimidines and have also been shown more recently to cleave the further oxidation products of 8-oxoG, spiroiminodihydantoin, and guanidinohydantoin (22)(23)(24)(25). Several groups have shown that 8-oxoG itself is a poor substrate (18, 24, 26 -28), which is in contradiction with work that showed that NEIL1 can remove the oxidized guanine (20,29). In at least one of these cases, the discrepancy may be attributed to a difference in the position of the lesion within the DNA oligonucleotide (29). Among the commercially available substrates tested, MvNei1 recognizes and removes thymine glycol (5,6-dihydro-5,6-dihydroxythymine (Tg)) and 5-hydroxyuracil (5-OHU) and displays negligible activity toward 8-oxoG (26). Tg is the most common oxidation product of thymine (30,31), and 5-OHU is generated by the oxidative deamination of cytosine (32, 33) (Fig. 1). We reported previously the crystal structures of MvNei1 unliganded and in complex with a DNA oligomer containing tetrahydrofuran (THF), a stable AP site analog (34). The structure of MvNei1 was found to be very similar to that of hNEIL1, EcoNei, and other Fpg/Nei enzymes. The MvNei1⅐THF complex illustrated how this enzyme binds DNA, although the question of how it recognizes a lesion was left unanswered. Here, we present the structures of MvNei1 in complex with DNA oligomers containing either Tg or 5-OHU, which are the first crystal structures of any Nei with a damaged base. The structures of MvNei1E3Q⅐Tg:C and MvNei1E3Q⅐5-OHU:G show that the damaged base makes few interactions with MvNei1. The paucity of contacts to the damaged base suggests that lesion recognition occurs before base flipping.
DNA Preparation and Complex Formation-For crystallization, a 13-mer oligodeoxynucleotide 5Ј-CGTCCAXGTC-TAC-3Ј (X ϭ Tg or 5-OHU) was used. The oligodeoxynucleotides (oligo) containing TG and 5-OHU were purchased from Midland Certified Reagent Co. (Midland, TX) and gel-purified. The commercially available thymine glycol is predominantly the 5R,6S stereoisomer (87%) (Glen Research, Sterling, VA) (35,36). Annealing of the damage-containing oligodeoxynucleotide with the complementary strand and complex formation with the enzyme were performed as reported previously (34). For the kinetics experiments, a 35-mer 5Ј-TGTCAATAGCAAGXG-GAGAAGTCAATCGTGAGTCT-3Ј (X ϭ Tg or 5-OHU) oligonucleotide and its complementary strand were used. An oligonucleotide with uracil at the X position was used to generate a site of base loss (AP site). The modified strand was labeled at the 5Ј-end using [␥-32 P]ATP with T4 polynucleotide kinase (New England Biolabs, Ipswich, MA) and annealed to the complementary strand in a 1:1 ratio. To minimize the error because of the uncertainty in the DNA concentration after ethanol precipitation, the labeled oligonucleotide for the glycosylase assay was diluted 10-fold with the corresponding unlabeled damaged oligonucleotide (26). For the AP lyase assay, the labeled oligodeoxynucleotide was used undiluted. The double-stranded oligonucleotide containing a uracil was treated with uracil DNA glycosylase (New England Biolabs, Ipswich, MA) for 1 h at 37°C to generate an AP site.
Kinetics Experiments-Kinetics experiments were performed under single-turnover conditions ([Enzyme] Ͼ [DNA]), using the 35-base pair duplex to evaluate the glycosylase and lyase activity of MvNei1 variants. In each case, the total reaction volume was 100 l with a final duplex DNA concentration of 2.5 nM in the reaction buffer (20 mM Hepes-KOH, pH 7.8, 75 mM KCl, 1 mM EDTA, 0.1 mg/ml BSA, 1 mM DTT, 5% (v/v) glycerol) and 300 nM MvNei1 WT, 600 nM Y253F, or 900 nM E6A at 37°C. For the Tg:A reaction, aliquots were removed from the reaction at various time points, 1/3, 2/3, 1, 2, 5, 10, 30, and 60 min, and quenched with 2 l of 1 N NaOH following a heat treatment at 90°C for 2 min. For the 5-OHU:G reaction, aliquots were removed from the reaction at various time points, 1/10, 2/10, 1/3, 2/3, 1, 2, 5, 10, 20, and 40 min, and quenched with 2 l of 1 N NaOH following heat treatment at 90°C for 2 min. Because of the fast lyase reaction, kinetics experiments were performed using an RQF-3 rapid quench flow instrument (KinTek Corp., Austin, TX). The time points for AP:A were 0.01, 0.02, 0.05, 0.1, 0.2, 0.5, 1, 2, 5, and 20 s. The reaction samples were quenched using a 1% (w/v) SDS solution, and the mixture was incubated on ice for 1 h. After the incubation, the supernatant was taken for quantification. The kinetic parameters for both activity experiments (glycosylase and lyase) were obtained from three independent experiments and fitted using Prism (GraphPad Software, La Jolla, CA).
Crystallization, X-ray Data Collection, and Structure Determination-Crystallization conditions for MvNei1E3Q in complex with Tg or 5-OHU are similar to those reported for the MvNei1⅐tetrahydrofuran (THF) complex (34). However, sodium formate (0.15 M) was used as a salt instead of 0.15 M magnesium nitrate to crystallize MvNei1E3Q in complex with 5-OHU. The pH of the crystallization solution was measured to be 7.0. The base opposite the lesion is a C for Tg and a G for 5-OHU. Although a Tg:A base pair is more biologically relevant, crystals were obtained readily with C opposite Tg, although attempts to crystallize a complex with an oligonucleotide containing Tg opposite A yielded only small needles. Both crystallographic data sets of the MvNei1E3Q⅐Tg:C and MvNei1E3Q⅐5-OHU:G complexes were collected at 100 K at ϭ 1.0332 Å at beamline 23-ID-B of the Advanced Photon Source, and additional data for MvNei1E3Q⅐Tg:C were collected at copper radiation on a Mar345 detector (MarResearch, Hamburg). Both data sets were processed, merged, and scaled with Denzo/Scalepack (37).
Both structures were determined by difference Fourier methods using phases from the MvNei1⅐THF model devoid of all non-protein atoms (Protein Data Bank code 3A46 (34)). The refinement of the MvNei1E3Q⅐Tg:C complex was performed using CNS (38). The MvNei1E3Q⅐5-OHU:G complex was refined using CNS (38) followed by phenix.refine (39). The final models for both structures were found to exhibit good geometry, as determined using Procheck (40). The original Fo MvNei1E3Q⅐Tg:C or MvNeilE3Q⅐5-OHU:G -Fo MvNei1⅐THF isomorphous difference Fourier maps indicated that the putative lesion recognition loop (residues 217-245) in both MvNei1E3Q⅐Tg:C and MvNei1E3Q⅐5-OHU:G complexes was not well ordered throughout this region. After building the damaged bases (Tg and 5-OHU) and further refinement, the electron density map corresponding to the loop was judged adequate for building main-chain atoms only. There are no residues in the disallowed region of the Ramachandran plot. Refinement statistics are shown in Table 1. All structure figures were prepared using PyMOL (41).
Protein Data Bank Accession Codes-Coordinates and structure factors for MvNei1E3Q bound to Tg and 5-OHU have been deposited in the Protein Data Bank with accession codes 3VK8 and 3VK7, respectively.

Structure Determination of MvNei1 in Complex with DNA
Containing Tg:C or 5-OHU:G-To date, there is no crystal structure of any Nei glycosylase in complex with a damaged base, which prompted our effort to crystallize MvNei1 with DNA containing a lesion. We tested several glycosylase-defi-cient variants, including P2G and E3Q. The residual glycosylase activity of the P2G and E3Q variants was negligible. The E3Q variant was selected for crystallization trials because it was previously documented that the lack of the catalytic proline could affect the position of the ␣F-␤9/10 loop in Fpg (16,42). The E3Q variant was crystallized with 5-OHU and Tg, which are commercially available and are good substrates for Nei enzymes, including MvNei1, hNEIL1, and EcoNei ( Fig. 1) (24,26,43,44). Although both are oxidized pyrimidines, 5-OHU is planar, and Tg is not (45). Crystals of MvNei1E3Q with Tgcontaining DNA were obtained using crystallization conditions similar to those used for the complex of MvNei1 with DNA containing an abasic site analog (THF) (34). Diffraction data were collected to 2.0 Å resolution at the Advanced Photon Source at Argonne National Laboratory (Argonne, IL) and at our home x-ray source. Because of the high isomorphism between the MvNei1⅐THF and MvNei1E3Q⅐Tg:C complexes (overall cross-R of 0.18 on amplitudes), the initial phases were obtained by using the model of MvNei1 in complex with THF devoid of all non-protein atoms. The structure of MvNei1E3Q with Tg-containing DNA was refined to R-factor and R free values of 20.0 and 23.0%, respectively (Table 1 and Fig. 2A).
A single crystal of MvNei1E3Q in complex with 5-OHU:Gcontaining DNA was used to collect a 2.1 Å resolution data set at the Advanced Photon Source synchrotron. The MvNei1E3Q⅐5-OHU:G data set was also isomorphous with the MvNei1⅐THF complex (cross-R on amplitudes was 0.28). The initial phases for MvNei1E3Q⅐5-OHU:G were also obtained using the MvNei1⅐THF model devoid of all non-protein atoms. The residual F o Ϫ F c map after several cycles of refinement revealed that two conformations of 5-OHU were bound in the active site with ϳ70% occupancy for the syn conformation and ϳ30% occupancy for the anti conformation in molecule A. The structure of MvNei1E3Q⅐5-OHU:G was refined with CNS (37) followed by phenix.refine (39) to R-factor and R free values of 21.3 and 26.6%, respectively (Table 1 and Fig. 2B). Although the crystal asymmetric unit comprises two MvNei1E3Q⅐DNA complexes, the description below will focus mainly on one of the complexes, molecule A, because both molecules have comparable average B-factors.
Overall Structure of MvNei1E3Q in Complex with Tg:C or 5-OHU:G-The overall models of the MvNei1E3Q⅐Tg:C and MvNei1E3Q⅐5-OHU⅐G complexes are nearly identical to that of MvNei1⅐THF (root mean square deviation 0.24 Å (calculated on 288 C␣) and 0.34 Å (calculated on 287 C␣), respectively). The differences between these structures are localized in the active site and are due to slight changes in the position of three residues in the lesion binding pocket (Pro-2, Glu/Gln-3, and Glu-6), and changes in the conformation of the putative lesion recognition loop (residues 217-245 in MvNei1).
In the structures of the MvNei1E3Q⅐Tg:C and MvNei1E3Q⅐5-OHU:G complexes, the lesion is flipped out of the DNA double helix and located in the lesion-binding site (Fig. 2). The density observed for Tg indicated that Tg is in the anti conformation in the MvNei1E3Q⅐Tg:C complex. However, the electron density observed for 5-OHU in the MvNeiE3Q⅐5-OHU:G complex suggested that both the syn and anti conformations are possible, with the respective occupancies refined to ϳ0.70 and ϳ0.30. The catalytic amino acid residues Pro-2 and Gln-3 (Glu-3 in the wild-type enzyme) in the MvNei1E3Q⅐Tg:C and /5-OHU:G complexes adopt positions that are nearly identical to those in the MvNei1⅐THF:C structure, although there are slight changes in the side chain orientations. The side chain of Gln-3 rotates away from O4Ј on the deoxyribose (Fig. 3) The Pro-2 ring also adopts slightly different rotamers. The distance between the amine of the proline and C1Ј of the deoxyribose is 4.0 Å in the complex with Tg, 4.4 Å in the complex with 5-OHU, and 3.7 Å in the THF complex. The carboxylate moiety of Glu-6 in MvNei1E3Q⅐Tg:C and MvNei1E3Q⅐5-OHU:G is rotated toward the lesion (Tg or 5-OHU) (Fig. 3).
Interactions of MvNei1 with Tg or 5-OHU-The DNA binds to MvNei1 in the MvNeiE3Q⅐Tg:C or MvNei1E3Q⅐5-OHU:G complexes in a manner that is very similar to that described for the MvNei1⅐THF complex (34), with the exception of the interactions involving the lesions (Tg or 5-OHU) and the estranged base guanine. The 13-mer duplex is bound, as expected, in the DNA binding cleft between the two domains and lies perpendicularly to the long axis of MvNei1 (Fig. 2). Arg-277, which is well conserved among the Fpg/Nei family and plays an important role in the glycosylase but not the lyase activity, contacts the oxygen of two phosphates (P 0 and P Ϫ1 ) adjacent to the lesion (Fig. 4) (46). In the MvNei1E3Q⅐Tg:C complex, there are very few hydrogen bonding interactions involving the Tg base. A direct hydrogen bond interaction between Tg and MvNei1 was observed, between the main-chain amide of Tyr-221 and O4 of Tg (Figs. 4A and 5A). (The electron density for the side chain of Tyr-221 was unclear and therefore was not built.) Two additional watermediated hydrogen bond interactions were also observed, con-necting the hydroxyl group of Tyr-253 and O5 of Tg and the main-chain carbonyl group of Leu-84 with O6 of Tg. Three aromatic residues, Tyr-174, Phe-250, and Tyr-253, surround Tg. Tyr-174 and Tyr-253 are within van der Waals distance and may therefore help stabilize the damaged base (Fig. 4A).
In the MvNei1E3Q⅐5-OHU:G complex, 5-OHU adopts two conformations, anti and syn. In both conformations, a hydrogen bond between the main-chain amide of Tyr-221 and the O4 of 5-OHU is maintained, reminiscent of the MvNei1E3Q⅐Tg:C

structure (Figs. 4 and 5).
In the anti conformation, a watermediated hydrogen bond between the hydroxyl group of Tyr-253 and O5 of 5-OHU was also observed in molecule A, as in the MvNei1E3Q⅐Tg:C structure (Fig. 4, A and B). In the syn conformation, an additional hydrogen bond was observed between the carboxyl group of Glu-6 and the hydroxyl group at position 5 in 5-OHU (Figs. 4C and 5C). A water-mediated hydrogen bond from the hydroxyl group of Tyr-253, which was seen in the MvNei1E3Q⅐Tg:C complex, was not observed in the syn conformation of the MvNei1E3Q⅐5-OHU complex. In contrast, a hydrogen bond between the main-chain carbonyl group of Leu-84 and the lesion (O6 of Tg and O2 of 5-OHU) was maintained (Fig. 4C). As seen with Tg, three aromatic residues, Tyr-174, Phe-250, and Tyr-253, are in the vicinity of 5-OHU, and the two tyrosines are within van der Waals distance of the damaged base (Fig. 4, B and C).
Interactions of MvNei1 with the Estranged Base-In both the MvNei1E3Q⅐Tg:C and MvNei1E3Q⅐5-OHU:G complexes, Leu-84, Arg-114, and Phe-116 were observed filling the void created by the extrusion of the damaged base, as described before (34). Leu-84 takes the place of the excised damaged base and participates in a hydrogen bond interaction with the damaged base (Tg and 5-OHU-syn) via its main-chain carbonyl (Figs. 4 and 5). Phe-116 is wedged between the estranged base (cytosine or guanine) and its 5Ј neighbor. In the MvNei1E3Q⅐Tg:C complex, Arg-114 interacts with the orphaned cytosine via two hydrogen bonds, involving O2 and N3 of the estranged C and N⑀H and NH of Arg-114 (supplemental Fig. S1A), as described previously (34). In the MvNei1E3Q⅐5-OHU:G complex, Arg-114 interacts with the orphaned guanine also via two hydrogen bonds (O6 and N7 of estranged G and N⑀H and NH of Arg-114) (supplemental Fig. S1B), as reported in the Bacillus stearothermophilus (Bst) Fpg complex with DNA containing a reduced abasic site (47).
Activity Assays of MvNei1WT, E6A, and Y253F Variants-Very few hydrogen bond interactions were observed between the lesion and MvNei1 in the complexes of MvNei1E3Q with either Tg:C or 5-OHU:G (Fig. 4). Tyr-253 interacts with both Tg and anti-5-OHU via a water molecule (Figs. 4 and 5). We hypothesized that Tyr-253 might thus contribute to either lesion recognition or lesion stabilization after the lesion has flipped out into the glycosylase active site. The tyrosine also contacts phosphate P 0 and could therefore play a part in the lyase reaction. In addition, Glu-6 is located within hydrogen bonding distance of Tg and 5-OHU in the anti conformation, although formation of such an H-bond would require the glutamate to be protonated (Fig. 4, A and B). However, Glu-6 interacts with the hydroxyl 5-OH of syn-5-OHU via a weak H-bond  (ϳ3.5 Å). To evaluate the potential role of the only protein side chains seen contacting the lesions in either the glycosylase or lyase reaction, activity assays of MvNei1 with DNA containing Tg:A, 5-OHU:G, or AP:A were performed with enzyme variants mutated at Glu-6 and Tyr-253. The MvNei1 Y253F variant was designed because it lacks the ability to hydrogen bond with a water molecule while maintaining the aromatic ring. The MvNei1 E6A variant forfeits any possibility of a hydrogen bond from its side-chain atoms with the lesion.
An initial glycosylase activity assay showed that the Y253F and E6A variants exhibited a level of activity similar to that of MvNei1 WT (data not shown). This result indicated that Tyr-253 and Glu-6 were not key residues in the recognition of the lesion. However, further kinetics experiments of MvNei1 WT and the two variants (Y253F and E6A) with Tg:A and 5-OHU:G were performed to analyze differences in reaction velocity. When performing pre-steady state kinetics on DNA glycosylases, either a single-or double-exponential model can be used to fit the data depending on the substrates tested. The Tg:A data were fit to a single-exponential model, as observed for Candida albicans Fpg with a double-stranded Tg substrate (48). We note that another group reported that kinetics data obtained with human NEIL1 and Tg-containing DNA were fit to a doubleexponential model, presumably because of the interconversion of the cis-and trans-Tg epimers (49). Mutating Tyr-253 and Glu-6 in MvNei1 decreased the velocity of the glycosylase reaction by 6.7-and 3.6-fold, respectively (Table 2). No significant difference in the velocity of the glycosylase reaction was observed with 5-OHU:G (Table 2). Furthermore, the glycosylase reaction using 5-OHU fits a double-exponential model, instead of a single-exponential model as observed with Tg:A. It presumably corresponds to the presence of both syn and anti conformations observed for 5-OHU in the MvNei1E3Q⅐5-OHU:G complex.
To check the effect of mutating Tyr-253 and Glu-6 on the lyase reaction, kinetics experiments with MvNei1 WT and the two variants (Y253F and E6A) with AP:A were performed. The result of the lyase reaction, which fit the double-exponential model presumably because a naturally occurring AP site can adopt open and closed ring conformations, indicated no drastic difference arising from mutating either of these residues. The E6A variant exhibited a k obs that was decreased by almost 2-fold, although the k obs for Y253F was slightly higher than that of wild-type MvNei1.

DISCUSSION
One of the pressing questions regarding Nei glycosylases is how these enzymes locate and recognize a damaged base (50,51). Unlike Fpg, an enzyme of the same family with a narrow substrate specificity, Nei enzymes efficiently recognize a wide variety of oxidized lesions (12,13,26,(52)(53)(54)(55). We therefore set out to crystallize MvNei1E3Q in complex with two of its best commercially available substrates, Tg and 5-OHU (26), in an attempt to answer this question. Our work describes the first crystal structure of any Nei in complex with a damaged base. Several structures of Fpg in complex with a lesion (8-oxoG, ring-opened formamidopyrimidine, and 5-hydroxy-5-methylhydantoin) have been reported (15,16,42,56), and these structures were compared with the MvNei1 complexes to gauge how lesions are recognized. Among the Fpg structures, the BstFpgE3Q⅐8oxoG:C complex was chosen as a reference because the E3Q variant was employed in both BstFpg and MvNei1 structures (15).
The structures of MvNei1E3Q⅐Tg:C and MvNei1E3Q⅐5-OHU:G complexes revealed that the lesion (Tg or 5-OHU) is extruded from the DNA helix and located in the active site of the enzyme. In these structures, the electron density for the tip (residues 219 -223) of the putative lesion recognition loop is disordered, and as a result only the main-chain atoms of these residues were built. However, the ␣F-␤10 loop in the BstFpgE3Q⅐8oxoG:C complex is well ordered. Few hydrogen bond interactions between MvNei1 and the damaged base are observed in the MvNei1E3Q⅐Tg:C and MvNei1E3Q⅐5-OHU:G complexes. In contrast, 8-oxoG is stabilized by several strong hydrogen bond interactions in the BstFpgE3Q⅐8oxoG:C complex (15). In particular, the ␣F-␤10 loop wraps around 8-oxoG providing several important main-chain interactions to O6 and the protonated N7. Verdine and co-workers (17) showed that a BstFpg variant in which the ␣F-␤10 loop was deleted showed no glycosylase activity toward 8-oxoG, underscoring the crucial role of this protein segment in the recognition of the oxidized guanine.
A superposition of MvNei1E3Q⅐Tg:C onto the BstFpgE3Q⅐8-oxoG:C complex reveals that O4 and N3 of Tg are located approximately at the same position as O6 and N7 in 8-oxoG. O4 of Tg participates in one H-bond interaction with the mainchain amide of Tyr-221. N3, however, does not partake in any H-bond interaction with the putative lesion recognition loop. The loop is shorter than that of Fpg and as a result it cannot wrap around the lesion ( Fig. 6 and supplemental Fig. 2). This finding suggests that the MvNei1 loop does not play a key role in either the recognition of the damaged base nor its stabilization. Moreover, the structure of the MvNei1E3Q⅐5-OHU:G complex revealed that 5-OHU can adopt both anti and syn conformations in the active site, suggesting that the lesion in the active site of MvNei1 is not well anchored, in contrast to 8-oxoG in the BstFpgE3Q⅐8-oxoG:C complex. The putative lesion recognition loop in MvNei1 is shorter than the ␣F-␤9/10 loop in Fpg as mentioned before, and this segment in hNEIL1 is even shorter than in MvNei1 (supplemental Fig. S2). Nei enzymes prefer oxidized pyrimidines and the further oxidation products of 8-oxoG, such as guanidinohydantoin and spiroiminodihydantoin, although several groups have shown that 8-oxoG itself is removed very inefficiently (18, 24, 26 -28). Taken together, these observations suggest that the long ␣F-␤9/10 loop in Fpg has evolved to stabilize and cap its target lesion, 8-oxoG.
Previous work on E. coli Nei had predicted that the binding pocket for the everted base would be formed by three aromatic residues, namely Phe-228, Phe-230 and Tyr-170 (9). The authors further predicted that the plane of the tyrosine would be perpendicular to that of Tg. This is exactly what we have observed with the corresponding residues (Phe-250, Phe-253, and Tyr-174) in the everted lesion binding pocket of MvNei1 (Fig. 4A). Hydrogen bonding interactions were also predicted between Tg and Pro-2, Glu-6, and Arg-213. Pro-2 and Glu-6 are in close proximity of the lesion in the MvNei1 structure. Arg-213, which was predicted to contact O4 and O5, is not conserved, and we do not observe any residue capable of making these H-bond interactions. Another group also used molecular dynamics calculations to predict which hNEIL1 residues might

-OHU:G, or AP:A) determined under single-turnover conditions
The kinetic parameters were obtained from three independent experiments. Data obtained with 5-OHU:G and AP:A were fit with a double-exponential model, which yields two rate constants (k 1 and k 2 ). Standard deviations are shown. NA means not applicable.  (57) published their work, mutational data for Glu-6 was not available, and this residue was predicted to be important for glycosylase activity. However, we found in this study that mutating this glutamate to alanine does not adversely affect the glycosylase activity.
Mutating the other residue whose side chain is in close contact to Tg (Tyr-253) does not abolish the glycosylase activity either. In addition, mutating either residue to alanine has little effect on the AP lyase activity. The decreased velocity of the Tyr-253 variant for the Tg:A substrate suggests that Tyr-253 might participate in stabilizing the lesion via its water-mediated hydrogen bond interaction. The decreased velocity of the E6A variant for Tg:A indicates that Glu-6 affects the stability of the lesion. However, H-bond interaction between Glu-6 and O2 of Tg would require the glutamate to be protonated. The pK a of glutamate is 4.5, and the pH of the crystallization solution is 7.0, meaning that Glu-6 should in theory carry a negative charge. There are, however, a number of variables that can alter the pK a of a carboxylate group, including whether it is buried and how many hydrogen bonds with which it participates (58). The exact nature of the interaction between Glu-6 and O 2 of Tg remains to be elucidated. No significant change of the kinetic rate for 5-OHU:G was observed after mutating either Tyr-253 or Glu-6. This result suggests that 5-OHU is not significantly stabilized by these protein side chains. This is consistent with the observation that two conformations of 5-OHU (anti and syn) coexist in the MvNei1E3Q⅐5-OHU:G complex.
Our mutational studies suggest that the two protein residues whose side chains are in close contact to the damaged base are not involved in lesion recognition. The question then of how Nei enzymes recognize their target substrates still remains. In an attempt to answer this question, an unmodified thymine was modeled into the active site of MvNei1 (supplemental Fig. S3). A modeled thymine was substituted for Tg in the MvNei1E3Q⅐Tg:C complex structure (supplemental Fig. S3). Although two water-mediated H bonds involving the 5-and 6-hydroxyl groups of Tg would be lost, the hydrogen bond interaction between O4 of thymine and the main chain amide of Tyr-221 of MvNei1 would likely remain, as in the MvNei1E3Q⅐Tg:C and /5-OHU:G complexes (Fig. 4). The hydrophobic interactions with Tyr-253, Phe-250, and Tyr-174 would remain essentially unchanged. Although the superposition suggests that once a normal base finds it way into the active site it would likely be cleaved because no obvious discrimination between the damaged and nondamaged base is observed, cleavage of normal thymine by MvNei1 is not observed (data not shown). This leads to the hypothesis that the lesion might be recognized before the lesion flips out of the DNA helix. The same speculation was originally put forward by Qi et al. (17) for the recognition of 8-oxoG by BstFpg. The authors obtained crystal structures of BstFpg cross-linked to DNA containing G or 8-oxoG within the double helix and reported that the C2Јendo sugar pucker of guanine undergoes a pseudorotation to an alternative pucker (C4Ј-exo) when G is replaced by 8-oxoG. This localized reorganization of the DNA backbone was proposed to be a key element in the way Fpg discriminates between a target lesion and a regular base. A local conformational perturbation may also be at play here. The presence of (5R,6S)-Tg within double-stranded DNA is known to hinder proper stacking of the 5Ј base, especially when that base is a purine (45,59,60). Furthermore, the nature of the base opposite the lesion was shown to modulate repair of Tg by human NEIL1 (49); Tg occupies the wobble position when mispaired with G (61) and is excised much more rapidly than Tg:A (49). Importantly the Tg:G mispair in DNA results from oxidative damage to 5-methylcytosine followed by deamination. Similarly the 5-OHU:G mispair arises after oxidative deamination of cytosine. Whether Nei enzymes sense local conformational perturbations in the DNA to recognize their target lesions will require further investigation.
We presented here the first crystal structures of an Nei enzyme in complex with a DNA duplex containing either Tg or 5-OHU. This study showed that Tg in the active site is in the anti conformation, whereas 5-OHU can adopt both anti and syn conformations. Unexpectedly, very few hydrogen bond interactions were observed between the enzyme and either lesion, and mutating two of the residues in contact with the lesion did not affect the glycosylase activity. This finding suggests that lesion recognition is likely performed before the lesion is flipped out from the DNA double helix.