DNA Damage Processing by Human 8-Oxoguanine-DNA Glycosylase Mutants with the Occluded Active Site*

Background: Oxoguanine-DNA glycosylase (OGG1) removes highly mutagenic 8-oxoguanine from DNA. Results: OGG1 mutations C253I and C253L occlude the active site and distort the OGG1-DNA precatalytic complex but retain some activity. Conclusion: Active site of OGG1 possesses flexibility that partially compensates for distortions. Significance: Active site plasticity may be important for dynamic recognition of multiple DNA lesions by DNA glycosylases. 8-Oxoguanine-DNA glycosylase (OGG1) removes premutagenic lesion 8-oxoguanine (8-oxo-G) from DNA and then nicks the nascent abasic (apurinic/apyrimidinic) site by β-elimination. Although the structure of OGG1 bound to damaged DNA is known, the dynamic aspects of 8-oxo-G recognition are not well understood. To comprehend the mechanisms of substrate recognition and processing, we have constructed OGG1 mutants with the active site occluded by replacement of Cys-253, which forms a wall of the base-binding pocket, with bulky leucine or isoleucine. The conformational dynamics of OGG1 mutants were characterized by single-turnover kinetics and stopped-flow kinetics with fluorescent detection. Additionally, the conformational mobility of wild type and the mutant OGG1 substrate complex was assessed using molecular dynamics simulations. Although pocket occlusion distorted the active site and greatly decreased the catalytic activity of OGG1, it did not fully prevent processing of 8-oxo-G and apurinic/apyrimidinic sites. Both mutants were notably stimulated in the presence of free 8-bromoguanine, indicating that this base can bind to the distorted OGG1 and facilitate β-elimination. The results agree with the concept of enzyme plasticity, suggesting that the active site of OGG1 is flexible enough to compensate partially for distortions caused by mutation.

The structure of human OGG1 has been determined for the free enzyme and a number of its complexes with DNA approximating various points along the reaction coordinate (8 -19). OGG1 belongs to the endonuclease III structural superfamily (20), possessing a characteristic helix-hairpin-helix motif and an extended loop rich in Gly and Pro and containing an absolutely conserved catalytic Asp-268 residue. Another absolutely conserved residue, Lys-249, acts as a nucleophile that attacks C1Ј of the damaged nucleotide during the reaction and forms a transient covalent intermediate, the Schiff base (21). The enzyme features a wide positively charged groove where DNA binds. Upon binding, DNA is kinked by ϳ70°, and the 8-oxo-G base is everted from the double helix. A narrow deep pocket within the DNA binding groove accommodates the damaged base, whereas the void left in DNA after the eversion is filled by Asn-149 and Asn-150.
In the base-binding pocket, the flipped out 8-oxo-G is tightly sandwiched between the aromatic system of Phe-319 and the thiol group of Cys-253, which, as quantum mechanical calculations show (14), likely exists as a deprotonated thiolate anion, stabilized by interactions with the positively charged Lys-249. The edges of the 8-oxo-G base form several hydrogen bonds with Gly-42, Asp-168, and Gln-315 (Fig. 1A).
In addition to its ability to excise 8-oxo-G, OGG1 possesses a slower activity that nicks DNA by the ␤-elimination mechanism (AP lyase activity). The DNA substrates for this activity may contain either 8-oxo-G or preformed AP sites. In the latter case, the structure of OGG1 bound to DNA containing an uncleavable AP site analog (3-hydroxytetrahydrofuran-2-yl)methyl phosphate shows that the base-binding pocket remains empty despite that the sugar moiety is everted in much the same way as the damaged base.
To understand the importance of the interactions within the base-binding pocket for different reactions catalyzed by OGG1 on different substrates, we have constructed site-directed mutants replacing the Cys-253 residue with either leucine or isoleucine (Fig. 1A). The hypothesis was that the added steric bulk and inability to form the Cys-253(thiolate)-Lys-249 dipole would exclude 8-oxo-G from the pocket and disable its excision but still allow the enzyme to act on AP sites. We have characterized the reactions by these active site occlusion mutants using high resolution stoppedflow kinetics with fluorescence detection and molecular dynamics simulation. Overall, our results suggest that the mutations distort the active site to the disadvantage of the catalytic step of the reaction, whereas the recognition of 8-oxo-G is still efficient. However, the active site of OGG1 appears to be flexible enough to retain some residual repair activity.

EXPERIMENTAL PROCEDURES
Oligonucleotide Synthesis and Purification-Oligodeoxyribonucleotides (ODNs; Fig. 1) were synthesized on an ASM-700 synthesizer (BIOSSET, Novosibirsk, Russia) using phosphoramidites purchased from Glen Research (Sterling, VA) and purified by anion exchange high performance liquid chromatography (HPLC) on a Nucleosil 100 -10 N(CH 3 ) 2 column followed by reverse-phase HPLC on a Nucleosil 100-10 C 18 column (both columns from Macherey-Nagel, Düren, Germany). The purity of the ODNs was assessed by 20% denaturing PAGE after staining with Stains-All (Sigma). The concentrations of the ODNs were determined from their absorbance at 260 nm.
The AP-containing ODN was prepared by incubating 0.1 mmol of the ODN containing dU at the intended place of the AP site for 14 h at 37°C with 15 units of uracil-DNA glycosylase (New England Biolabs, Ipswich, MA) in 150 l of the buffer containing 20 mM Tris-HCl (pH 8.0), 1 mM EDTA, 1 mM dithiothreitol (DTT), and 0.1 mg/ml bovine serum albumin. The reaction product was purified by reverse-phase HPLC on a Nucleosil 100-5 C 18 column using a linear gradient of 0 -20% acetonitrile in 0.1 M triethylammonium acetate (pH 7.0). The pooled fractions were concentrated and then converted to the lithium salt form using a Sep-Pak Plus C 18 cartridge (Waters, Milford, MA). The integrity of the AP-containing ODN was assessed by PAGE followed by Stains-All staining. To confirm the presence of the AP site in the ODN after the treatment with uracil-DNA glycosylase, samples were treated with 10% aqueous piperidine at 95°C and were completely cleaved at the modified site. When needed, the modified strands were 32 P-labeled using [␥-32 P]ATP and phage T4 polynucleotide kinase (SibEnzyme, Novosibirsk, Russia) according to the manufacturer's protocol, purified by 20% denaturing PAGE, and annealed to the complementary strand.
Enzymes-Site-directed mutants were constructed using a QuikChange multisite-directed mutagenesis kit (Agilent Technologies, Santa Clara, CA); pET-15b plasmid carrying the OGG1 isoform 1a insert (22) served as a template The mutant proteins were induced and purified as described for the wildtype OGG1 (22). As the mutants proved to retain their activity on the AP substrates, concentration of the active form was determined from burst phase kinetics using the AP substrate as described (23).
Stopped-flow Measurements-Stopped-flow measurements with fluorescence detection were carried out using a model SX.18MV stopped-flow spectrometer (Applied Photophysics, Leatherhead, UK) fitted with a 150-watt xenon arc lamp. All experiments were carried out at 25°C in a buffer containing 50 mM Tris-HCl (pH 7.5), 50 mM KCl, 1 mM EDTA, 1 mM DTT, and 9% (v/v) glycerol. The enzyme in one syringe was rapidly mixed with the substrate solution in the other syringe. The substrate and enzyme solutions were equilibrated for 45 min at FIGURE 1. A, impact of the occluding mutations on the structure of OGG1. The 8-oxo-G nucleotide is shown as a van der Waals sphere model, and amino acid residues Gln-315 and Phe-319 forming the 8-oxo-G-binding pocket are shown as ball-and-stick models colored according to the atom type (cyan, carbons; blue, nitrogens; red, oxygens; brown, phosphorus atoms). The amino acid residue at position 253 is color-coded: yellow, Cys-253 (wild type); cyan, Ile-253; magenta, Leu-253. The rest of the protein is shown as a semitransparent schematic model. The structure was built by direct replacement of Cys-253 using the mutagenesis tool of PyMOL (47). B and C, sequence of oligodeoxyribonucleotides and chemical nature of the lesions (X) used in the kinetic experiments (B, AP site; C, 8-oxo-2Ј-deoxyguanosine).
25°C before mixing. The concentrations after mixing were varied in the 0.5-2 M range for the ODN duplexes, and the concentration of OGG1 in all experiments was 1 M. Excitation wavelength ex ϭ 283 nm was used, and the fluorescent emission was monitored using a 320-nm long pass filter (WG-320, Schott, Mainz, Germany) to detect the fluorescence of the protein's Trp residues. The dead time of the instrument was 1.4 ms. Each trace obtained is the average of at least four individual experiments.
Data Processing-The approach is based on the fluorescence intensity changes in the course of the reaction due to the formation of the enzyme-DNA complex and its subsequent transformation to conformers corresponding to different intermediates of the substrate recognition and processing (22,24,25). The kinetic parameters were obtained by numerical integration of the system of kinetic differential equations (Equation 1) and the least squares global nonlinear fitting of the total fluorescence (F, total fluorescence; F b , background fluorescence; f i , fluorescence coefficients of individual OGG1 conformers; [E i ], concentrations of individual OGG1 conformers, where i ϭ 0 corresponds to the free enzyme) using the DynaFit package (BioKin, Pullman, WA) (26).
In the evaluated mechanisms, except for the first bimolecular step and the product release step, all other reactions were first order. The kinetic parameters were obtained by global fitting of sets of fluorescence curves obtained at different concentrations of the reactants. During the fitting procedure, all relevant rate constants for the forward and reverse reactions as well as the specific molar response factors for all intermediate complexes were optimized. The minimal kinetic mechanism (Scheme 1) was confirmed using a scree test as described for the wild-type OGG1 (24).
OGG1 Cleavage Assay and Product Accumulation Kinetics-To analyze the products formed by OGG1, the reaction was performed under the same conditions as the stopped-flow experiments except the substrate ODNs were 32 P-labeled. If needed, the reaction mixture also contained 0.5 mM 8-bromoguanine. The reaction was terminated at required time points by adding the loading dye solution containing 7 M urea. To determine the rate of DNA nicking (the AP lyase activity), the aliquots were directly analyzed by 20% denaturing PAGE. To analyze the rate of AP site formation from the substrate containing 8-oxo-G (the N-glycosylase activity), 10 -12 volumes of 2% LiClO 4 in acetone were added to the aliquots, and the precipitates were washed with 100 l of 85% ethanol, then twice with 100 l of acetone, dried, and treated with 10% aqueous piperidine at 95°C for 30 min to cleave all AP sites introduced by the enzyme. The products were again precipitated, washed, and dried as describe above. The samples were dissolved in 3 l of the loading dye solution and analyzed by 20% denaturing PAGE. The gels were exposed to Agfa CP-BU x-ray film (Agfa-Geavert, Mortsel, Belgium); the autoradiograms were scanned and quantified by scanning densitometry using Gel-Pro Analyzer software (Media Cybernetics, Rockville, MD). The rate constants were obtained by numerical integration of systems of mass balance and kinetic differential equations assuming a fast equilibrium single-turnover model (Scheme 2, Equation 2; see under "Results" for details of the kinetic model choice) and fitting the product accumulation curves to the solution using DynaFit.
For Scheme 2A ] are the concentrations of complexes of the enzyme with 8-oxo-G, AP site-containing, and nicked DNA, respectively; K a is the association constant; k glyc is the rate constant of base excision (the glycosylase reaction), and k ly is the rate constant of strand nicking (the AP lyase reaction). Molecular Dynamic Simulations-The atomic structure of human OGG1 (PDB code 1EBM) containing a 15-mer-DNA duplex was taken as a starting structure for modeling (see Fig.  5A for the sequence and numeration of the duplex). The K249Q mutation, present in the structure to inactivate the enzyme, was converted back to Lys-249 by manually editing the PDB file. The residues 80 -82 missing in the 1EBM structure were restored using the following protocol. The initial backbone conformation of the residues 80 -82 was taken from the free OGG1 structure (PDB entry 1KO9). The fragment was aligned for the best fit to the source structure and incorporated into the PDB file. Then the side chains of residues 80 -82 were restored, and their orientation was optimized in 500 steps of Fletcher energy optimization algorithm. The final refinement was done by 500 ps of simulated annealing using the GUI-BioPASED molecular modeling suite (27). It should be noted that residues 80 -82 reside on the protein surface far from the active site. The obtained structure (model WT) contains cytosine opposite the lesion. The C253I and C253L mutant models were manually constructed by side-chain replacement and adjustment of 8-oxo-G base position, followed by energy optimization and molecular dynamics-simulated annealing. The mutant structures were finally validated using the PDB validation module of GUI-BioPASED (27). Following the results of quantum mechanics/molecular mechanics data (14) and OGG1 mechanism considerations (12), some amino acid residues were simulated in nonstandard protonation state to approximate the situation immediately preceding the nucleophilic attack. Cys-253 was modeled in the deprotonated state as the thiolate anion, and Lys-249 and Asp-268 were modeled with neutral amino group and carboxyl group, respectively. The force field parameters of these charge states were derived from AMBER ff99. The force field parameters for the 8-oxo-2Ј-deoxyguanosine-5Ј-phosphate residue were taken from Ref. 28. A 10-ns molecular dynamics simulation was performed using the BioPASED molecular dynamics modeling software (29) using the AMBER ff99 force field and the analytical Gaussian solventexclusion implicit solvent model (30), with an integration time step of 1 fs. The system was gradually heated from 10 to 300 K over 50 ps and equilibrated at 300 K. A classic molecular dynamics trajectory was generated in the NTV ensemble with harmonic restraints of 0.001 kcal/A 2 for the protein atoms, 0.25 kcal/A 2 for atoms of the terminal nucleotides, and 0.0025 kcal/A 2 for the rest of the DNA atoms. Coordinates of each atom of the system (snapshots) were saved every 2 ps, thus producing a trajectory of 5076 files. All trajectories were analyzed using MDTRA trajectory analysis software (31).

Stopped-flow Kinetics of C253I and C253L OGG1-
The overall shape of Trp fluorescence traces observed for the interaction between C253I and C253L OGG1 and oxo-G substrate ( Fig. 2) was quite similar to each other and to the Trp traces for wildtype OGG1 processing the same substrate (22,24), yet the characteristic times were different. The fluorescence quickly dropped until ϳ1 s. This step was followed by a much more gradual fluorescent signal decrease with a less pronounced change in the amplitude, and then the fluorescence started to grow at ϳ200 -300 s, but this process was still not complete by 3000 -4000 s. The time scale of the slow fluorescent changes is shifted toward longer times compared with wild-type OGG1. The initial fast decrease in the fluorescent signal can be associated with DNA-enzyme complex formation, and no notable differences in fluorescent traces for mutant and wild-type Damage Processing by OGG1 with the Occluded Active Site OCTOBER 4, 2013 • VOLUME 288 • NUMBER 40 OGG1 were detected at this step. According to the PAGE analyses discussed below, at this time scale (up to 4000 s) no nicked product was formed; therefore, the observed growth of fluorescence probably reflects the abasic product release from the enzyme-substrate complex.
Processing of the 8-oxo-G substrate by both mutant forms was described using minimal Scheme 1, which was the same as for the wild-type OGG1 (22). Values for the individual rate constants were determined by global fitting (Table 1). Notably, k 4 , the rate constant of the irreversible step, was 15-fold lower for both mutant forms in comparison with the wild-type enzyme (0.0046 and 0.0042 s Ϫ1 for OGG1 C253L and OGG1 C253I, respectively, in contrast to 0.06 s Ϫ1 for wild-type OGG1) (22). According to the PAGE analysis of C253I and C253L enzymes, the k 4 rate constant is likely related to the AP site product formation, which is followed by DNA release; the same attribution of this reaction step was made earlier for wild-type OGG1 (22).
In contrast to the catalytic step, the preceding stages of the reaction, namely binding and structural adjustment of the enzyme-substrate complex, are affected to a lesser degree. The general association constant K a , calculated using Equation 3, was 1.9 ϫ 10 7 and 1.6 ϫ 10 7 M Ϫ1 for OGG1 C253I and OGG1 C253L, respectively, only ϳ1.5-fold lower than for the wildtype enzyme (K a ϭ 2.9 ϫ 10 7 M Ϫ1 ) (22). Thus, it seems likely that both mutations predominantly affect the catalytic ability of OGG1 rather than its affinity for DNA or its capacity to distort DNA toward the precatalytic complex.

C253I and C253L Product Formation Rates-Presteady-state
stopped-flow analysis with fluorescence detection permits the detection of conformational changes in the enzyme molecule. To associate the conformational transitions with particular chemical steps of the catalytic cycle, we used PAGE to detect accumulation of abasic and nicked products for C253I and C253L OGG1 (Fig. 3). No considerable difference in the product formation rates between the two mutant forms was registered. However, in the OGG1 mutant forms both N-glycosylase and AP lyase activities were significantly retarded compared with the wild-type OGG1. With the 8-oxo-G substrate, the abasic product accumulation rates were ϳ180and ϳ90-fold slower for C253I and C253L OGG1, respectively, than for the wild-type enzyme under the same conditions when OGG1 and 8-oxo-G substrate were both taken at 1 M (Fig. 3A). The rate of accumulation of the nicked product for the 8-oxo-G substrate was affected by the mutations to a lesser degree, decreasing 4 -8-fold only (Fig. 3B). For the AP substrate, as expected from the relaxed requirement for the base-binding pocket, the nicked product accumulation was faster than for the 8-oxo-G substrate for both C253I and C253L OGG1 (Fig. 3B).
Rate constants or product formations were estimated from the PAGE analysis results. Because the concentrations of both the enzyme and the substrate (1 M) greatly exceeded K d values (ϳ53 and 63 nM for C253I and OGG1 C253L; K d was calculated as 1/K a , where K a was from Equation 3) and the catalytic rate was apparently low, the single-turnover model was used, and the observed rate constant was taken as an approximation for the rate constant of the catalytic step. The N-glycosylase reaction with the 8-oxo-G substrate and the AP lyase reaction with AP substrate were fitted as one irreversible step (Scheme 2, A and B, respectively), whereas the combined N-glycosylase and AP lyase reactions with 8-oxo-G substrate were fitted as two sequential irreversible steps (Scheme 2C), and the rate constant of AP lyase reaction for 8-oxo-G substrate was calculated using the k glyc constant obtained for Scheme 2A. The calculated rate constants are listed in Table 2. The apparent values of k glyc and k ly for both C253I and C253L OGG1 were at least 1 order of magnitude lower than for wild-type OGG1 (22). The observed  k glyc constants were of the same order of magnitude as the k 4 constants determined from stopped-flow experiments but exceeded the k ly constants 20 -80-fold, indicating that in the stopped-flow experiments, the released product was indeed the AP site rather than nicked DNA. Notably, the k ly constant for the AP substrate was 2.8 -4.4-fold higher than for the 8-oxo-G substrate, again consistent with the idea that the AP site does not enter the base-binding pocket deeply, and therefore its cleavage should be less affected by the active site occlusion.
Effect of 8-Bromoguanine on the Interaction of OGG1 with 8-Oxoguanine and AP Substrates-We and others have previously shown that the AP lyase reaction catalyzed by OGG1 is stimulated by adding free 8-bromoguanine base (BrG) to the reaction mixture (12,22,24). The effect of BrG on the AP lyase activity of C253I and C253L OGG1 was studied using PAGE analysis (Fig. 4). For both mutant forms, a notable increase in the nicked product formation rates was observed during interaction with both AP-and 8-oxo-G substrates in the presence of 0.5 mM BrG. In the case of the AP substrate (Fig. 4, A and C), BrG enhanced the reaction rate 5-15-fold for both mutants and wild-type OGG1. The stimulation of strand nicking was observed both under the single-turnover (Fig. 4A) and for the multiple turnover conditions (Fig. 4C). With the 8-oxo-G substrate, two types of experiments were conducted (Fig. 4B). In one setup, BrG was added to the reaction mixture a few seconds after the protein, and the substrates were combined. In the other setup, OGG1 was premixed with BrG for 45 min before the substrate was introduced. As with the AP substrate, a significant acceleration of the AP lyase reaction was observed for both C253I and C253L mutant forms.
Of note, the ␤-elimination rate did not depend on the order of mixing. This may indicate that even if BrG binds to the occluded OGG1 active site in the absence of DNA, this binding is not tight enough to completely prevent subsequent entrance and coordination of 8-oxo-G substrate.
Molecular Dynamics of OGG1 C253I and C253L-To gain insight into the structural consequences of the Cys-253 replacement, we have performed molecular dynamic studies of wild-type OGG1 and the mutants bound to DNA. As the experiments suggested that the initial stages of DNA binding and lesion recognition are probably unaffected by the mutations, and they slow down the reaction by perturbing the Michaelis complex, our main goal was to probe the conformational space available to the enzyme with the base everted from DNA and poised for excision. Therefore, we used a crystal structure of catalytically inactivated OGG1 bound to DNA with 8-oxo-G fully in the lesion recognition pocket (1EBM in PDB) as a starting model and an implicit solvent system to accelerate the sampling of available conformations of the active site (32). The dynamics of the active site can be affected by the charges of ionizable groups inside. In the OGG1 reaction intermediate immediately preceding the nucleophilic attack at C1Ј, Lys-249 has to be deprotonated to expose a lone electron pair for the attack, whereas Asp-268 is juxtaposed to a partially negatively charged O4Ј, which receives a proton to open the deoxyribose ring (12), so Asp-268 is likely protonated. Accordingly, we chose to model both Lys-249 and Asp-268 in the neutral state in all three structures. Furthermore, in wild-type OGG1, we have simulated Cys-253 as a thiolate anion, following the quantum   (14). Replacement of charged Lys-249 to the neutral one abandons the Lys-249 ϩ -Cys-253 Ϫ dipole suggested to exist in wild-type OGG1 (14), but these two residues are still held together by a strong hydrogen bond (Ϫ2.35 kcal/mol, 2.9 Å), existing over 99% of the simulation.
All three models evolved slowly for the initial stage of the run, up to 5-7 ns, and were quite stable after 7.5 ns, as evidenced by one-and two-dimensional r.m.s.d. plots (Fig. 5, B-E); overall, the heavy atom r.m.s.d. did not exceed 1.8 Å. Introduction of a bulky substituent into the lesion recognition pocket, as expected, displaced the 8-oxo-G base from the proper position in the pocket. In the wild-type OGG1, Phe-319 forms one wall of the pocket, stacking with the everted 8-oxo-G. Notably, in OGG1 C253I, ϳ80% of the area of the Phe-319 side chain overlaps with 8-oxo-G planar rings through the complete trajectory of the dynamics, whereas in the C253L mutant, this overlap is ϳ20% through the whole dynamics (Fig. 6A). Wild-type OGG1 demonstrated an intermediate behavior, starting at ϳ60% overlap and then gradually decreasing to ϳ20% at 4 -6 ns. Moreover, the inclination of the Phe-319 ring relative to the plane of 8-oxo-G is ϳ20°more in C253L OGG1 than in the other two models. The extra bulk of the isoleucine and leucine side chain is accommodated in different ways. In the C253I mutant, the side chain of Ile is folded to be generally parallel to the 8-oxo-G plane, which may additionally restrain the base movement through van der Waals interactions, whereas the side chain of Leu in C253L is extended and pushes the base further into the wall formed by Phe-319 (Fig. 6F).
Close examination of the structural parameters of DNA revealed that the mutations hardly affect the conformation of the oxo-dG nucleotide and the immediately adjacent DNA residues. Covalent bond torsion angles and deoxyribose phase angle in these residues were almost identical in all three simulations, with minor changes in the ␤ angle in C253L OGG1 (approximately Ϫ20°turn in the second half of the simulation) and angle in C253I OGG1 (stays stable at ϳ270°through the simulation, whereas wild-type and C253L models turn ϳ20°in the second half of the simulation). Thus, paradoxically, it seems that the mutations in OGG1 that were supposed to occlude the basebinding pocket and prevent 8-oxo-G from entering it do not prevent full movement of the damaged nucleotide but rather distort the pocket itself to accommodate the everted 8-oxo-G.
The distance N(Lys-249)-C1Ј(oxo-dG 0 ) is stable and similar (ϳ4 Å) in all three models. However, the angle N(Lys-249)-C1Ј(oxo-dG 0 )-N9(oxo-dG 0 ) is stably different between the wild type (ϳ75°) and both mutants (ϳ110°) (Fig. 6B). This angle is important for proper overlap of the orbitals during the nucleophilic displacement of 8-oxo-G by the amino group of Lys-249, and generally, such reactions tolerate no more than 10 -15°displacement from the optimal alignment (33). However, the alignment in the direction N(Lys-249)-C1Ј(oxo-dG 0 )-O4Ј(oxo- FIGURE 6. A, overlap of Phe-319 and 8-oxo-G aromatic rings during the molecular dynamics in wild-type (purple), C253I (magenta), and C253L OGG1 (blue). The traces show moving average over 50 snapshots. B, angle N(Lys-249)-C1Ј(oxo-dG (0) )-N9(oxo-dG (0) ) of approach of catalytic Lys-249 of wild-type (purple), C253I (magenta), and C253L (blue) OGG1 to the reacting C1Ј of oxo-dG. C-E, population of the catalytically favorable conformation in wild-type OGG1 (C), OGG1 C253I (D), and OGG1 C253L (E). In the catalytically favorable conformation (top right cluster on each plot, very scarcely populated in E), Asp-268 makes a hydrogen bond with O4Ј of the damaged nucleotide and is far from O1P of the adjacent dGMP, whereas Lys-249 approaches C1Ј of the damaged nucleotide at a favorable angle for nucleophilic attack in the C1Ј-N9 direction. F, superposition of a representative simulation snapshot (5.9 ns) showing the overlap of the residues sandwiching 8-oxo-G in wild-type, C253I, and C253L structures. Note the position of Leu, which pushes 8-oxo-G and Phe-319 aside from their positions in wild-type and C253I OGG1. G, superposition of a representative simulation snapshot (7.0 ns) showing the distortion at the complementary DNA strand introduced by active site-occluding mutations. Three central base pairs are shown. Tyr-203 wedge lies behind dC (ϩ1) . Note that dC (0) and dT (Ϫ1) are more disordered than other overlapping bases. F and G, carbon atoms are colored according to the model: wild-type (purple), C253I (magenta), and C253L (blue). Oxygen atoms are colored red; nitrogen, dark blue; phosphorus, orange; and sulfur, yellow. Tyr-203 is completely colored according to the model. OCTOBER 4, 2013 • VOLUME 288 • NUMBER 40 dG 0 ) (data not shown) is similar in all three models, consistent with the idea that N-glycosidic bond, and not C1Ј-O4Ј bond, is broken first. The side chain of the second catalytic amino acid, Asp-268, in the wild-type enzyme remained near O4Ј 67% of the simulation and formed an alternative bond with a nonbridging oxygen O1P of dG Ϫ1 for 7% of the simulation. In the mutants, however, Asp-268 was predominantly contacting the O1P(dG Ϫ1 ) rather than O4Ј(oxo-dG 0 ) (45 and 39% of the time, respectively, for C253I OGG1, 60 and 27% of the time, respectively, for C253L OGG1; see Fig. 5A for the numeration of deoxynucleotides in the DNA part of the model). Plotting the whole trajectory in three dimensions with the coordinates N(Lys-249)-C1Ј(oxo-dG 0 )-N9(oxo-dG 0 ) angle, O␦2(Asp-268)-O4Ј(oxo-dG 0 ) distance, and O␦2(Asp-268)-O1P(dG Ϫ1 ) distance, one can see that the active site populates four clusters of conformational space (Fig. 6, C-E), of which only one (top right) is favorable for catalysis with respect to Lys-249 attack angle and Asp-268 stabilization. This cluster is the most populated in the wild-type enzyme (Fig. 6C), less visited by the C253I active site (Fig. 6D), and is almost empty in C253L (Fig. 6E). Thus, the combination of misalignment of Lys-249 and a decrease in the population of the catalytically competent Asp-268 conformation together may account for the compromised catalytic activity of the OGG1 mutants.

Damage Processing by OGG1 with the Occluded Active Site
Contacts with the strand opposite to the lesion are also important for forming the proper Michaelis complex by OGG1, reflected in the strong specificity of this enzyme for cytosine at this position (34,35). Two structural features of OGG1-DNA interactions are the most eminent in fixing the proper conformation of the opposite strand. First, Tyr-203 side-chain wedges between C opposite to 8-oxo-G (dC (0) ) and the base 5Ј to it (C in 1EBM, dC (ϩ1) ), which helps to kink DNA. Second, the residues Asn-149, Arg-154, and Arg-204 make contacts with dC (0) . Analysis of dynamics of Tyr-203 inclination and overlap with the interacting cytosines shows that all three structures do not undergo major changes in this region. The distortion of DNA in the mutants is highest at the dC (0) and especially dT (Ϫ1) nucleotides (Fig. 6G). However, this seems to affect only the backbone, because all 6-bp parameters involving dT (Ϫ1) (shear, stretch, stagger, buckle, propeller, and opening) are overall not significantly different between wild-type and mutant OGG1. Regarding contacts with dC (0) , the guanidine group of Arg-154 in the C253I mutant drifts away from the base early in the simulations, whereas in the C253L OGG1, Asn-149 is far from the cytosine in the first 80% of the dynamics but then assumes the wild-type-like conformation (data not shown).

DISCUSSION
A long-standing problem in the field of DNA repair is how DNA repairs enzymes, DNA glycosylases in particular, and discriminates their target-damaged nucleotides from the normal ones, which are present in a great excess yet, with very rare exceptions, are all but refractory to the enzymes' action. DNA glycosylases, as many other nucleic acid-dependent enzymes, enforce significant conformational changes upon their substrates. In the precatalytic complex, the helical axis of DNA is kinked at the point of the lesion (ϳ70°in OGG1-DNA complex 1EBM), and the damaged nucleotide is rotated from its position inside the double helix into the enzyme's active site. An emerging model of glycosylase specificity acknowledges that nucleotides in DNA are sampled in several sequential steps, with normal nucleotides rejected early in the process, whereas cognatedamaged nucleotides successfully clear all checkpoints and adopt a specific conformation and position in the active site allowing the N-glycosidic bond breakage to ensue (25,36). The checkpoints are transition energy barriers between the conformers of the enzyme-substrate complex and may correspond to DNA kinking, eversion of the damaged nucleotide, insertion of the enzyme's amino acids in the void left after the eversion, and accommodation of the damaged nucleotide in the enzyme's active site, etc.
The substrate recognition and catalysis trajectory of OGG1 is well sampled by x-ray crystallography and molecular dynamic simulation. In particular, four sequential conformations of the nucleotide eversion are known as follows: the early structure where the target base is disengaged from its Watson-Crick interactions, unstacked, and slightly pushed toward the major groove (15); the intermediate structure where the target base falls in a so-called "exo-site" at the entrance to the active site (14); the advanced structure where the target base is almost at its final position but has not yet formed all specific bonds with the base-binding pocket of OGG1 (16,17,19); and the precatalytic structure with the damaged base poised for removal (8,11,16). Steered molecular dynamics with umbrella sampling indicates that all intermediate conformations are energetically similar with 2-4 kcal/mol barriers between them, whereas the final fully everted conformation is 10 -12 kcal/mol lower than the intermediates. 5 For E. coli Fpg, another 8-oxoguanine-DNA glycosylase is structurally unrelated to OGG1, and the eversion transition barrier directly determined from stopped-flow data had enthalpy ⌬H 0 ϭ Ϫ15.5 kcal/mol, although that was partially offset by negative entropy attributed to the enzyme-structure complex becoming more rigid upon base eversion and void filling (25).
In the precatalytic structure (8), the 8-oxo-G base stacks against the aromatic system of Phe-319 and is pressed by the thiolate Cys-253 Ϫ from the other side of the plane. The Watson-Crick-forming groups are all engaged in the interactions with the enzyme's active site; N 2 donates one hydrogen bond to O⑀1 of Gln-315 and another to either O␦1 of Asp-268 or the main chain O of Pro-266; N1 donates a hydrogen bond to O⑀1 of Gln-315, and O 6 accepts two hydrogen bonds from structured water molecules that interact with other amino acid residues of the active site. Outside of the base-binding pocket but still in the active site, Lys-249 is poised for the nucleophilic attack at C1Ј, and Asp-268 forms a hydrogen bond with O4Ј of the target nucleotide, assisting the deoxyribose ring opening and stabilizing the positive charge that develops on the ring when the base leaves. The only bond specific for 8-oxo-G versus G is formed by N7, which, in the pyrrolic type as in 8-oxo-G, donates a hydrogen bond to the main chain O of Gly-42 but cannot do so in the pyridinic type as in G. Interestingly, if G is forcibly presented to the active site in nearly the same conformation as 8-oxo-G, it is still not excised (19). As the N-glycosidic bond in oxo-dG is chemically more stable than in dG (37), this can only be explained by small but multiple perturbations in the active site structure when the base in the pocket is normal.
Our interest in the mechanisms of substrate discrimination by DNA glycosylases led us to address the question how a distortion of the base-binding pocket can affect the action of OGG1. We have chosen to replace Cys-253 with Leu or Ile, which are bulkier than Cys and should presumably displace 8-oxo-G from its proper position within the active site. In addition, the mutation should eliminate the negative charge on the thiolate anion, predicted from quantum mechanical modeling of the OGG1-DNA complex (14). The partial occlusion of the active site of OGG1 by a Q315F mutation leads to a severely reduced activity (16). Of note, OGG1-Q315F cross-linked to DNA at two different places produced two structures as follows: one with 8-oxo-G in the exo-site and another with 8-oxo-G in an advanced intermediate state; however, the amount of the residual activity was almost the same in both cases (16). Our unpublished data 6 on the Q315W mutant of OGG1 confirm that collision of the active site with the Watson-Crick edge interferes with base excision. A recent report on another active site-occluding mutant shows that the replacement of Cys-253 with Trp inactivates the glycosylase function of OGG1 but spares the AP lyase activity (18).
The minimal general kinetic scheme of OGG1 Cys-253 mutants turned out to be the same as the scheme for the wild-type OGG1 (22). Moreover, the profiles of fluorescence increase and decrease were consistent in the three forms of the enzyme, suggesting that the kinetic steps correspond to the same conformational transitions. The rate constants for individual conformational transitions varied between the respective constants for the wild-type enzyme; the differences were rather small at the early stages of the reaction, but the irreversible step, ascribed to the catalytic process of N-glycosidic bond breakage, was severely affected in the mutants. We have observed a similar behavior of OGG1 and Fpg in our studies on substrate perturbation, in which good substrates (such as 8-oxo-G:C pairs) and poor substrates (such as 8-oxo-G:A pairs) were processed by these glycosylases following the same kinetic scheme but with the rates of one step being very different between good and poor substrates (24,38). However, in the case of the poor 8-oxo-G:A substrate, the affected step was attributed to the insertion of OGG1 residues into the void formed by damaged base eversion (24); these residues form highly specific discriminating bonds with C (8). The large effect of pocket-occluding mutations on the N-glycosidic bond breakage clearly reflects the requirement for precise positioning of catalytic residues relative to the reacting atoms of the 8-oxo-G deoxynucleoside.
A quite unexpected finding was stimulation of the activity of the occlusion mutants by BrG. When this phenomenon was first described for the wild-type OGG1 and its AP lyase activity, it was assumed that the 7H-tautomer, assisted by the electronegative bromine substituent, binds in the enzyme's active site in the same mode as the excised 8-oxo-G, and it acts as a general base to pull a hydrogen off C2Ј to induce ␤-elimination (12). Later, however, it was shown that BrG also efficiently stimulates the AP lyase activity of OGG1 when the substrate contains 8-oxo-G, suggesting that the excised 8-oxo-G does not stay tightly bound in the pocket but is in fast exchange with BrG (22). From the data we present here, it is apparent that the initiation of ␤-elimination by BrG has relaxed geometric requirements, because it is as efficient with the occluded basebinding pocket (5.5-and 15.2-fold stimulation for C253I and C253L, respectively) as with the wild-type one (9.2-fold stimulation). This is consistent with the observation that BrG-induced ␤-elimination has mixed stereochemistry of proton abstraction from C2Ј, with an ϳ4:1 ratio of pro-R to pro-S hydrogens removed, indicating some flexibility of the active site at this reaction step (12). Although we did not model OGG1-AP-DNA-BrG ternary complexes, the nearly identical conformations of the oxo-dG nucleoside in the active site of wild-type and mutant OGG1 suggest that BrG also can be accommodated in a similar way in all three structures.
Another inference that can be made from the efficient stimulation of the OGG1 C253L and C253I by BrG concerns the depth of the 8-oxo-G sampling by the occlusion mutants. Indeed, it is highly unlikely that 8-oxo-G would be excised in the exo-site, where the distances from C1Ј to the catalytic amino group of Lys-249 and carboxyl of Asp-268 are too long. If even with the occluded (i.e. energetically unfavorable) active site, ␤-elimination remains the rate-limiting step and can be enhanced by BrG, then 8-oxo-G populates its correct or nearly correct position in the active site of C253L and C253I with a non-negligible probability, much like in the case of the Q315F occlusion mutant with a distal disulfide cross-link (16) or the substrate containing photocaged 8-oxo-G (17). This inference is also supported by our modeling results, in which the catalytically optimal spatial arrangement of Lys-249 and Asp-268 is less populated but still not completely prohibited in the mutants (Fig. 6, C-E).
Overall, both our experimental results and computational simulations resonate well with the concepts of active site plasticity of enzymes (39,40) and rugged free energy landscape of enzymatic catalysis (41). Plasticity is characteristic of many enzymes that have a range of substrates or catalyze different reactions, thus having evolved to adopt a variety of conformations and to stabilize multiple, albeit similar, transition states. If this is a case, mutation of a seemingly critical residue often is not detrimental for at least some of the enzyme's functions to the extent that could be expected from the energy of the lost interactions, because the active site is flexible enough to compensate partially for the misfit. Among DNA glycosylases, E. coli MutYs have been shown to compensate for a loss of Lys-20, a residue partially accounting for its weak ␤-elimination activity, by using another active site amine, that of Lys-142 (42). In our own work with E. coli endonuclease VIII, 7 the mutations abolishing critical interactions with DNA phosphates and greatly diminishing the glycosylase activity have no effect on the AP lyase activity due to restructuring of the active site that Damage Processing by OGG1 with the Occluded Active Site OCTOBER 4, 2013 • VOLUME 288 • NUMBER 40 fixes the substrate DNA in a catalytically competent conformation. OGG1 has the ability to accommodate a moderately diverse set of modified bases, including 8-oxo-G, 8-oxoadenine, 2,6-diamino-4-oxo-5-formamidopyrimidine, and several other substituted purines, and catalyzes base removal and ␤-elimination, the latter capable of proceeding either with the excised base or with no base in the active site (35,(43)(44)(45)(46). Strikingly, the active site of OGG1 is flexible enough to compensate for the positional exchange of Lys-249 and Cys-253, resulting in an enzyme with a significant residual activity (18). We conclude that although base-binding pocket occlusion distorts the active site OGG1 with the bound substrate and greatly decreases the catalytic proficiency of the enzyme, it does not fully prevent damaged base sampling and excision, likely due to the intrinsic flexibility of the protein and the strong energetic preference for 8-oxo-G eversion once the enzyme is bound to its preferred substrate DNA.