Substrate Specificity and Reaction Mechanism of Murine 8-Oxoguanine-DNA Glycosylase*

Genomic DNA is prone to oxidation by reactive oxygen species. A major product of DNA oxidation is the miscoding base 8-oxoguanine (8-oxoG). The mutagenic effects of 8-oxoG in mammalian cells are prevented by a DNA repair system consisting of 8-oxoguanine-DNA glycosylase (Ogg1), adenine-DNA glycosylase, and 8-oxo-dGTPase. We have cloned, overexpressed, and charac-terized mOgg1, the product of the murine ogg1 gene. mOgg1 is a DNA glycosylase/AP lyase belonging to the endonuclease III family of DNA repair enzymes. The AP lyase activity of mOgg1 is significantly lower than its glycosylase activity. mOgg1 releases 8-oxoG from DNA when paired with C, T, or G, but efficient DNA strand nicking is observed only with 8-oxoG:C. Binding of mOgg1 to oligonucleotides containing 8-oxoG:C is strong ( K D 5 51.5 n M ), unlike other mispairs. The aver- age residence time for mOgg1 bound to substrate containing 8-oxoG:C is 18.3 min; the time course for accu-mulation b D -galactopyranoside-inducible ex- a mOgg1. sequence verified using Novagen. Determination of K D by Gel Shift Assay— Reaction mixtures con- tained 1 n M 32 P-labeled oligonucleotide duplex, 25 m M sodium phosphate, pH 7.5, 100 m M NaCl, 2 m M Na-EDTA, 10% glycerol, and varying amounts of mOgg1 in a total volume of 10 m l. The enzyme was diluted as described above. Reaction the following operations Enzyme and m to nondenaturing (17 long) prerun in 0.5 3 TBE at 300 V at least h. Loading was at V, and a tracer dye (bromphenol blue, 0.5 3 TBE, 10% glycerol) was loaded in a separate lane. After 10 min, the voltage was reduced to 190 V, and the gel was run until the dye migrated approximately 10 cm. Gels were quantified using a Molecular Dynamics PhosphorImager system. Binding con- stants were calculated from three independent experiments using a Jandel SigmaPlot version 5.00 nonlinear fit routine. Half-life reaction 10 n M oligonucleotide duplex, 25 m M phos- phate, pH 7.5, 100 m M NaCl, 2 m M 0.1 mg/ml bovine serum albumin, and m M mOgg1 in a total volume of 8 m l. reaction was incubated for varying times (80 s to h) at and NaBH 4 was added to a final concentration of m M reaction was proceed for 2 min and then terminated by adding 10 m l of SDS loading buffer and heating for 5 min at 95 °C. were analyzed by 12% discontinuous SDS-PAGE. After plotting the initial velocities of cross-link formation versus the time of preincubation, obtained from three independent experiments, the half-life of the covalent complex was calculated as t 1 2 5 (ln 2)/ k , where k was determined by

The reaction mechanism and substrate specificity of Escherichia coli Fpg has been investigated extensively (for a recent review, see Ref. 28). However, in view of marked sequence differences, models for Fpg cannot be extrapolated directly to Ogg1. For example, while Fpg nicks DNA by ␤and ␦-elimination, Ogg1 performs only ␤-elimination (12,18). Likewise, Fpg efficiently removes a naturally occurring lesion, Me-FaPy-G, from DNA (29), while Ogg1 is significantly less active on this substrate (12,18,30).
Detailed mechanistic studies have not been reported for Ogg1; most of the data published were obtained using nonuniform assays (30 -32). In the present study, we analyze the substrate specificity of mOgg1, a murine counterpart of Ogg1 cloned in our laboratory (18), using as substrates nucleotide analogs prepared for previous studies of GO system enzymes (11,(33)(34)(35). Based on the relative efficiency of enzymatic processing, we propose a scheme for lesion recognition and reaction catalysis by mOgg1. We also report marked differences in the efficiency of several steps in the mOgg1-catalyzed reaction with damaged DNA.

EXPERIMENTAL PROCEDURES
Enzymes-Fpg protein was purified from an overproducing strain of E. coli (35). E. coli EndoIII was a kind gift from Dr. Richard Cunningham (SUNY Albany). T4 polynucleotide kinase, Klenow fragment, T4 DNA ligase, and restriction endonucleases were purchased from New England Biolabs, Pfu DNA polymerase and PCR reagents from Stratagene, and uracil-DNA glycosylase from Life Technologies, Inc.
Oligodeoxynucleotides-Unmodified and modified oligodeoxynucleotides ( Fig. 1) were prepared from the corresponding phosphoramidites by solid state synthesis using an Applied Biosystems model 394 auto-mated DNA synthesizer. Phosphoramidites were either purchased from Glen Research or synthesized by published methods (36 -42). Oligodeoxynucleotides were purified by denaturing PAGE and reverse phase chromatography. Oligodeoxynucleotides used as substrates for mOgg1 contained a normal or modified nucleotide at the 11th position of the 23-mer, d(CTCTCCCTTCXCTCCTTTCCTCT), where X represents dG, 8-oxodG, 8-oxodI, 8-oxodN, 8-MeOdG, 6-OMe-8-oxodG, 8-oxodA, 8-ami-nodG, 8-oxo-carbadG, dU, or F. Oligodeoxynucleotides were labeled at the 5Ј-end using [␥-32 P]ATP (Amersham Pharmacia Biotech) and T4 polynucleotide kinase and then annealed to the corresponding complementary strand in a 1:1.3 ratio. To obtain a duplex oligonucleotide containing an AP site opposite C, the corresponding duplex containing a U:C mispair was treated with 0.007 units/l of uracil-DNA glycosylase for 30 min at 37°C in the buffer used for the mOgg1 reaction (see below). To obtain a duplex oligonucleotide nicked by ␤-elimination and containing trans-4-hydroxy-2-pentenal, a duplex containing a U:C mispair was treated with uracil-DNA glycosylase, as described above, followed by treatment with 50 g/ml EndoIII.
Overexpression and Purification of mOgg1-mOgg1 was amplified from a mOgg1 cDNA template (18) by PCR, using Pfu DNA polymerase. The upstream primer contained a NdeI restriction site and an ATG start codon. The downstream primer contained a translation stop codon and a BamHI restriction site. The amplified gene was subcloned into the NdeI-BamHI site of the T7 expression vector pET13a (43). This construct allows isopropyl-1-thio-␤-D-galactopyranoside-inducible expression of a full-length, nonfusion recombinant mOgg1. The sequence was verified using standard fluorescent DNA sequencing techniques. The strain for mOgg1 overexpression, BL21(DE3), was obtained from Novagen.
To purify recombinant mOgg1, 4 liters of YTϫ2 medium containing 25 g/ml kanamycin were inoculated with 100 ml of an overnight culture of BL21(DE3) pET13a-mOgg1 E. coli. The culture was grown at 37°C with shaking at 240 rpm until A 260 reached 0.6. The temperature was then reduced to 15°C, and the culture was allowed to equilibrate for 1 h with shaking. Isopropyl-1-thio-␤-D-galactopyranoside was added to a final concentration of 0.2 mM, and cells were incubated further for 16 h at 15°C with shaking, harvested by centrifugation for 20 min at 12,000 ϫ g at 4°C, and stored frozen at Ϫ80°C. After thawing, the pellet (20 g, wet weight) was resuspended in 20 ml of B/PER reagent (Pierce) supplemented with Complete Protease Inhibitor Mixture (Roche Molecular Biochemicals). The concentration of NaCl was adjusted to 1 M, and the sample was incubated at room temperature for 20 min with gentle stirring. Cells were treated by ultrasound using a Biosonik IV sonicator (VWR Scientific) equipped with a 3 ⁄8-inch tip. Three pulses of 20 s at maximal power were delivered at 4-min intervals in an ice bath. The sample was diluted with 4 volumes of buffer C (50 mM Tris-HCl, pH 7.5, 1 mM Na-EDTA, 1 mM dithiothreitol), and cell debris was removed by 20-min centrifugation at 12,000 ϫ g at 4°C. The supernatant (crude extract) was directly applied to a Q Sepharose Fast Flow (Amersham Pharmacia Biotech) column (50 ml) equilibrated in buffer C with 200 mM NaCl. The column was washed with 2 volumes of buffer C containing 200 mM NaCl, and flow-through and wash fractions were pooled and precipitated with 80% ammonium sulfate. The pellet was collected by centrifugation, dissolved in buffer A (50 mM HEPES-NaOH, pH 7.5, 1 mM Na-EDTA, 1 mM dithiothreitol) supplemented with 200 mM NaCl (fraction I), and loaded onto an S Sepharose Fast Flow (Amersham Pharmacia Biotech) column (75 ml) equilibrated with the same buffer. The protein was eluted in a 1000-ml gradient of 200 -800 mM NaCl in buffer A. Fractions containing DNA glycosylase activity eluted at about 490 mM NaCl. These were pooled (fraction II), diluted with an equal volume of buffer A, and loaded onto a 20-ml heparin Sepharose (Amersham Pharmacia Biotech) column equilibrated in buffer A containing 200 mM NaCl. The column was developed by a 300-ml gradient of 200 -800 mM NaCl in buffer A. Fractions containing the desired activity eluted at approximately 530 mM NaCl and were pooled, diluted with an equal volume of buffer A (fraction III), and loaded onto a MonoS HR 5/5 column (Amersham Pharmacia Biotech) equilibrated in buffer A containing 200 mM NaCl. A 40-ml gradient of 200 -600 mM NaCl in buffer A was applied to the column. Fractions across the main absorption peak at 280 nM were analyzed by 12% SDS-PAGE with Coomassie Blue staining. Fractions containing protein of the expected size eluted at about 500 mM NaCl. These were concentrated on a Centricon-10 device (Amicon) to approximately 0.7 ml (fraction IV) and then subjected to gel filtration on a Superdex 75 HR 10/30 column (Amersham Pharmacia Biotech). Fractions included in the main absorption peak at 280 nM were analyzed by SDS-PAGE; fractions of at least 95% purity were pooled (fraction V), dialyzed against buffer A containing 250 mM NaCl and 50% glycerol, and stored at Ϫ20°C. Concentrations of protein solutions were determined by Bradford assay using bovine serum albumin as a standard. The apparent molecular mass of mOgg1 was determined by 12% SDS-PAGE using Bio-Rad SDS-PAGE molecular weight standards (low range) and by gel filtration on a Superdex 75 HR 10/30 column using Combithek calibration kit (Roche Molecular Biochemicals). Protein sequencing was performed on an Applied Biosystems 475A Protein Sequencer, employing standard Edman degradation chemistry.
Standard Assay of mOgg1 Activity-Reaction mixtures included 20 or 50 nM 32 P-labeled oligonucleotide duplex, 25 mM sodium phosphate, pH 7.5, 100 mM NaCl, 2 mM Na-EDTA, and 20 nM mOgg1 (or 2 l of a chromatographic fraction) in a total volume of 10 l. The enzyme was diluted to working concentrations in 0.5ϫ reaction buffer containing 0.5 mg/ml bovine serum albumin. Reactions were initiated by adding enzyme and allowed to proceed for 5 min at 37°C. To measure base release, 2.5 l of 0.5 M putrescine-HCl, pH 8.0, was added; the reaction mixture was heated at 95°C for 5 min and then mixed with 6.25 l of formamide dye loading buffer, followed by heating for 1 min at 95°C. To measure DNA strand nicking, 5 l of formamide dye loading buffer was added, and the reaction mixture was heated at 95°C for 1 min. Aliquots (5 l) were analyzed by 20% PAGE in 8 M urea. Products were analyzed quantitatively using a Molecular Dynamics PhosphorImager system.
Determination of Kinetic Parameters-Reaction mixtures and conditions used for kinetic studies were identical to the standard activity assay using varying amounts of the appropriate 32 P-labeled oligonucleotide duplex. The enzyme concentration and reaction time was adjusted so as to cleave no more than 10% of the substrate. Kinetic parameters were calculated using a Jandel SigmaPlot version 5.00 nonlinear fit routine. Three independent experiments were performed for each analysis.
Determination of K D by Gel Shift Assay-Reaction mixtures contained 1 nM 32 P-labeled oligonucleotide duplex, 25 mM sodium phosphate, pH 7.5, 100 mM NaCl, 2 mM Na-EDTA, 10% glycerol, and varying amounts of mOgg1 in a total volume of 10 l. The enzyme was diluted as described above. Reaction mixtures were preequilibrated at 4°C, and the following operations were performed at this temperature. Enzyme was added and allowed to bind for 3 min, and 5-l aliquots were subjected to 8% nondenaturing PAGE (17 cm long) prerun in 0.5ϫ TBE at 300 V for at least 2 h. Loading was done at 300 V, and a tracer dye (bromphenol blue, 0.5ϫ TBE, 10% glycerol) was loaded in a separate lane. After 10 min, the voltage was reduced to 190 V, and the gel was run until the dye migrated approximately 10 cm. Gels were quantified using a Molecular Dynamics PhosphorImager system. Binding constants were calculated from three independent experiments using a Jandel SigmaPlot version 5.00 nonlinear fit routine.
Determination of Half-life for the Covalent Complex-The reaction mixture included 10 nM oligonucleotide duplex, 25 mM sodium phosphate, pH 7.5, 100 mM NaCl, 2 mM EDTA, 0.1 mg/ml bovine serum albumin, and 1 M mOgg1 in a total volume of 8 l. The reaction was incubated for varying times (80 s to 10 h) at 37°C, and NaBH 4 was added to a final concentration of 100 mM. The reaction was allowed to proceed for 2 min and then terminated by adding 10 l of SDS loading buffer and heating for 5 min at 95°C. Products were analyzed by 12% discontinuous SDS-PAGE. After plotting the initial velocities of crosslink formation versus the time of preincubation, obtained from three independent experiments, the half-life of the covalent complex was calculated as t1 ⁄2 ϭ (ln 2)/k, where k was determined by fitting data from the descending part of the curve to the equation E ϭ a ϩ (E max Ϫ a)e Ϫkt (where E represents cross-linking at a given time point; a is a parameter used to correct for a kinetically irreversible enzyme-substrate complex; E max is the highest cross-linking achieved in the experiment; k is the first order rate coefficient; and t is time), using a Jandel SigmaPlot version 5.00 nonlinear fit routine.

RESULTS
Purification of Recombinant mOgg1-mOgg1 protein was overexpressed in E. coli and purified as a full-length nonfusion protein (Table I). After passing through a Q Sepharose column, followed by ammonium sulfate precipitation, the observed recovery of mOgg1 activity was Ͼ100%, suggesting the presence of uncharacterized inhibitors in the crude extract. Separation from host Fpg protein was achieved at the S Sepharose step; the band representing ␦-elimination, characteristic of Fpg, was absent in activity assays with fraction II and in later chromatographic steps. Furthermore, Fpg and mOgg1 eluted from the heparin-Sepharose column at different salt concentrations. The ratio of specific activities measured by strand nicking and base release end points dropped from 26.2% in the crude extract to 3.33% in fraction II and remained largely unchanged until purification was complete, suggesting that after resident Fpg protein is removed (i) mOgg1 catalyzes strand nicking with much less efficiency than base release and (ii) this strand nicking activity copurifies with base release activity. The highest specific activity at both end points was attained after the S Sepharose step; however, at this stage mOgg1 comprised only about 60% of total protein content. At the end of the purification procedure, mOgg1 was estimated to be at least 95% pure by SDS-PAGE analysis with Coomassie Blue staining (Fig. 2).
The relatively low purification factor of 40 is not unusual for purification of recombinant proteins from strongly inducing systems, such as the T7 RNA polymerase-dependent pET family of expression vectors, since the initial amount of target protein is high.
SDS-PAGE analysis of purified recombinant mOgg1 showed an apparent molecular mass of 40.7 kDa, in good agreement with the predicted molecular mass of 38.8 kDa. The protein deviated from normal mobility during gel filtration in 1 M NaCl (Stokes radius, 32 Å; molecular mass determined from gel filtration, 54.3 kDa), suggesting moderately elongated shape. Identity of the purified material was confirmed by Edman degradation, which revealed an N-terminal peptide of the expected sequence, MLFRSWL; the initial formylmethionine was not removed after polypeptide synthesis.
Optimal Reaction Conditions-The salt and pH dependence of purified recombinant mOgg1 were investigated. The enzyme demonstrated an optimum of 100 mM NaCl for base release (Fig. 3A). Efficiency of DNA strand nicking decreased linearly with increasing salt concentration (Fig. 3A). No significant differences were observed between KCl and NaCl (data not shown). Base release exhibited a broad pH optimum of 7.5; DNA strand nicking was less dependent on pH (Fig. 3B). Tris-HCl and sodium phosphate buffers were found to stimulate moderately base release activity compared with HEPES-NaOH; DNA strand nicking was highest in Tris-HCl buffer, probably due to its nucleophilicity (data not shown). MgCl 2 inhibited strand nicking at concentrations of Ͼ1 mM and base  Dependence of Binding and Cleavage of Substrates on the Opposite Base-In a time course experiment using an 8-oxoG: C-containing substrate, the rate of base release in reactions catalyzed by mOgg1 was significantly higher than the rate of DNA strand nicking (Fig. 4). This allowed kinetic parameters to be determined for both the DNA glycosylase and AP lyase steps of the reaction. To quantify base release and strand nicking independently, the reaction products were treated with putrescine, leading to quantitative cleavage of AP sites without additional cleavage at 8-oxodG (procedure modified from Ref. 45). Cleavage of DNA backbone after combined treatment with mOgg1 and putrescine reflects the glycosylase action of the enzyme, while cleavage in the absence of putrescine reflects combined glycosylase/AP lyase activity. Purified recombinant mOgg1 was used to explore the kinetic mechanism of this enzyme and its dependence on the base positioned across the lesion.
With respect to base release, 8-oxodG:dC was a preferred substrate for mOgg1; however, processing of 8-oxodG:dT and 8-oxodG:dG was more efficient than DNA strand nicking of these substrates (Fig. 5). Kinetic parameters for cleavage of 8-oxodG-containing substrates are summarized in Table II. For all substrates, base release is kinetically more efficient than base release combined with nicking, with k cat contributing heavily to this effect (for 8-oxodG:dG, the contribution of K m to this decrease is more pronounced). Assuming a two-step reaction mechanism, E ϩ S^ES 3 ESЈ 3 E ϩ P, an overall k cat for DNA strand nicking corresponds to k 3 (rate constant for the step ESЈ 3 E ϩ P) within a 15% margin of error.
Substrate binding by mOgg1 was determined by a gel mobility shift assay using oligonucleotides containing 8-oxodG, AP site, and F, a noncleavable analog of an AP site; these data are presented in Table II. The best ligands contain C opposite the lesion.
Stability of mOgg1-oligonucleotide Covalent Complex-During catalysis of base removal, Ogg1 forms a Schiff base covalent complex with 8-oxodG; this intermediate can be irreversibly trapped with NaBH 4 (13,17). To further investigate mOgg1 substrate specificity and dependence of the reaction on the base opposite the lesion, the time course of covalent complex formation and decay was studied for substrates containing 8-oxodG positioned opposite each of the four DNA bases. A 100-fold excess of the enzyme was incubated with substrate for varying times, NaBH 4 was added to the reaction, and the initial rate of cross-linking was measured (Fig. 6). The amount of covalent intermediate formed with 8-oxodG:dC and 8-oxodG:dT decreased exponentially (Fig. 6, B and D) with a half-life of 18.3 and 14.6 min, respectively. The 8-oxodG:dG pair demonstrated a two-phase time course, first an exponential decay (half-life 2.3 min) and then an increase in the amount of the covalent intermediate (Fig. 6C). The 8-oxodG:dA pair had no significant exponential decay phase, and, after a lag period, concentration of the covalent complex increased linearly over 10 h (Fig. 6A). The absolute rate of NaBH 4 -induced reduction at the first time point measured, reflecting the initial rate of covalent complex formation, was highest for 8-oxodG:dC, followed by 8-oxodG:dT (approximately 2 times less), 8-oxodG:dG (10 times less), and 8-oxodG:dA (20 times less) (Fig. 6). Kinetically irreversible enzyme trapping was observed for all four substrates, albeit at different levels.
Binding and Cleavage of Base Damage Analogs-To elucidate roles for various functional groups of the 8-oxoG moiety, we conducted binding, base release, and strand nicking assays on a series of substrates containing modified purines. Two types of experiments were performed; relative efficiencies of base release and strand nicking were determined for a matrix of unmodified or modified base pairs, and kinetic and binding parameters were determined for selected modified purines paired with C.
In conditions. Base pairs that showed less than 2% cleavage were judged to be resistant to mOgg1. Those showing higher (2-5%) levels of cleavage not affected by increasing enzyme concentration were also regarded as nonsubstrates. These results are summarized in Fig. 7. Unmodified bases (A, C, G, T, and U) were not subject to removal by mOgg1 (Fig. 7A). Some purine analogs were resistant to mOgg1 (8-MeOG), and others (8-aminoG, 8-oxoN) were processed with low efficiency when placed opposite C. 8-OxoA and 8-oxoI were removed efficiently when paired with C. 8-OxoA was totally resistant to cleavage when paired with any other base; for 8-oxoI, efficiencies for the other pairs were much lower. 8-OxoG and 6-OMe-8-oxoG were removed to varying extents from all base pairs (except 6-OMe-8-oxoG:A) (Fig. 7A). Base removal was not observed in singlestranded DNA for any purine analog tested. Interestingly, among all 8-oxoG:purine combinations tested, 8-oxoG:8-oxoG was much better than 8-oxoG:A or 8-oxoG:G. However, this probably was due to conversion of such a lesion into 8-oxodG opposite AP site, since when 8-oxodG in the unlabeled strand was replaced by the uncleavable derivative, 8-oxo-carbadG, the efficiency of the enzyme sharply dropped (data not shown).
Base pairs associated with efficient base release were also tested for strand nicking (Fig. 7B). Here 8-oxodG, 8-oxodI, 8-oxodA, and 6-OMe-8-oxodG opposite dC were the best substrates. When dT, dU, or 8-oxodG was substituted for dC, marginal cleavage was observed at 8-oxodG and 6-OMe-8-ox-  odG. Only 8-oxodG-containing oligonucleotides were cleaved when the lesion was placed opposite F. Since 8-oxodG:dC is presumed to be the natural substrate for mOgg1, binding and kinetic experiments were performed on a subset of purine analogs paired with dC (data for 8-oxodA:dT also were obtained). Results of these experiments are summarized in Table II. The affinity of mOgg1 for 8-oxodG was 51.5 nM, close to that observed for Fpg (33). Loss of the 2-amino group (8-oxodI), loss of both the 2-amino and the 6-keto group (8-oxodN), replacement of the 6-keto function with an amino or methoxy group (8-oxodA, 6-OMe-8-oxodG), and replacement of the 8-keto group with methoxy (8-MeOdG) did not compromise affinity and even improved it in some cases by 1-2 orders of magnitude. However, removal of the 8-keto group (dG) or replacement of it with an amino function (8-aminodG) decreased binding significantly, as did replacement of the 6-keto group with an amino group (8-oxodA) in the context of the 8-oxoA:T pair. Replacement of the deoxyribose heterocyclic oxygen with a methylene group (8-oxo-carbadG) yielded a noncleavable ligand with a K D close to that of a natural substrate.
When the kinetic parameters of base release were measured, effects of functional group substitutions on K m and k cat were observed. The removal of the 2-amino group (8-oxodI) had little effect on substrate preference, but the additional removal of the 6-keto group (8-oxodN) resulted in a very poor substrate, affecting both K m and k cat . The removal of the 8-keto group (dG) or its replacement (8-MeOdG, 8-aminodG) fully abolished (dG, 8-MeOdG) or significantly decreased (8-aminodG) catalytic activity. In contrast, modifications of the 6-keto group (6-OMe-8-oxodG, 8-oxodA) resulted in much better substrates for mOgg1, with improved K m and k cat . The ability to cleave 8-oxodA was completely lost when the base was placed opposite dT.
At the level of strand nicking, most changes in substrate specificity were due to changes in K m . The only exception was the removal of the 2-amino group (8-oxodI), which caused a 9-fold increase in K m combined with a 6-fold increase in k cat , leaving the overall specificity constant close to that of a 8-ox-odG-containing substrate. 8-OxodN and 8-aminodG were poor substrates for base nicking as well as base release. Unlike in the base release assay, 6-OMe-8-oxodG and 8-oxodA were not significantly better substrates than 8-oxoG.

Enzymatic Activities of mOgg1
DNA glycosylases have been classified as monofunctional (uracil-DNA glycosylase, AlkA, Tag, and others) or bifunctional (EndoIII, Fpg, bacteriophage T4 endonuclease V, and others) based on their ability to catalyze ␤-elimination after base excision and to form borohydride-sensitive covalent intermediates (28). Most bifunctional DNA glycosylases possess coupled glycosylase and AP lyase activities, catalyzing strand nicking at approximately the same rate as base release, and are readily cross-linked to their substrates by treatment with NaBH 4 . However, at least one borohydride-sensitive DNA glycosylase, MutY, has no AP lyase activity (46,47).
In the present study, we compare the relative glycosylase and AP lyase activities of mOgg1 and find that the latter is about 1 order of magnitude less. We attribute this difference to a delay between base excision and ␤-elimination during the reaction. The borohydride-sensitive species, presumably a covalent Schiff base intermediate, is stable (half-life 18 min). Therefore, mOgg1 represents a different class of DNA glycosylases: a bifunctional DNA glycosylase with uncoupled AP lyase activity. The independence of DNA strand nicking and base release was suggested for human Ogg1 (21) based on the observation that this enzyme removed 8-oxoG from all mispairs but cleaved DNA only when 8-oxoG or an AP site was positioned opposite C. The rate of base removal for the human enzyme was approximately 2-fold higher than DNA strand nicking (21). Here we present kinetic evidence that this independence occurs when ␤-elimination at the nascent AP site fails to proceed in concerted fashion with base removal. It is not clear whether AP lyase activity contributes to the biological function of the bifunctional DNA glycosylases, since AP endonucleases can process abasic sites during base excision repair; however, AP lyase activity appears to influence the choice between short patch and long patch base excision repair pathways in mammalian cells (48). Following ␤-elimination, the 3Ј terminus at the site of the nick exists as an ␣,␤-unsaturated aldehyde and must be removed so that the DNA polymerase can catalyze a gap-filling reaction. Some bifunctional DNA glycosylases, such as Fpg or endonuclease VIII, catalyze two sequential ␤-elimination steps (␤,␦-elimination), releasing the sugar moiety as trans-4-oxo-2pentenal (49,50). This process generates a 3Ј-phosphate, which requires further processing by a phosphatase to provide a primer end. An alternative mechanism was described for yeast Ogg1 and Drosophila S3 ribosomal protein (a protein with 8-oxoguanine glycosylase activity whose in vivo function is not clear). This reaction is Mg 2ϩ -dependent, removes the sugar moiety as trans-4-hydroxy-2-pentenal-5-phosphate, and is likely to be hydrolytic (44,51). mOgg1 does not possess these activities. We observed ␤-elimination but not ␤,␦-elimination at the intact abasic site, and the 3Ј-terminal ␣,␤-unsaturated aldehyde was not processed in the presence or absence of Mg 2ϩ . mOgg1 activity was inhibited slightly at millimolar concentrations of Mg 2ϩ (Fig. 3). With respect to its lack of ␤,␦-elimination activity, mOgg1 is similar to other enzymes of the EndoIII superfamily that catalyze only ␤-elimination (28). Inconsistencies between mOgg1 and yeast Ogg1 with regard to their ability to release trans-4-hydroxy-2-pentenal-5-phosphate require further investigation. Reexamination of the activities of Drosophila S3 and yeast Ogg1, using structural methods applied to studies of Fpg (52), would seem to be in order.
Assessment of the importance of various functional groups of 8-oxoG for recognition and catalysis by mOgg1 must take into account the lactim-lactam equilibria at the N 1 -O 6 , N 7 -O 8 (C 8 ), N 2 (C 2 )-N 1 and N 2 (C 2 )-N 3 tautomeric centers. Of the purine analogs used in this study, only G, 8-oxoG, 8-oxoA, and 8-MeOG were studied in this respect, using ab initio computational methods (54,55) or NMR (3, 56 -58). No significant differences between unmodified purines and 8-substituted derivatives were found at tautomeric centers other than N 7 -O 8 (C 8 ). The latter has been shown to exist in the lactam configuration, with N7 protonated and a keto group at C8. We assume all 8-oxo derivatives employed in this study obey this rule and that 8-aminoG exists predominantly in the enamine configuration, similar to exocyclic amino groups of adenine, cytosine, and guanine.
The effect of the nucleotide positioned opposite the lesion on base release and strand nicking activity of Ogg1 has been described for the related enzyme of human origin (21). To explain the lack of DNA nicking activity on substrates other than 8-oxoG:C, it has been hypothesized that keeping the damaged strand topologically intact when the opposite base is not C would prevent erroneous base excision repair and allow for postreplicative repair by recombination (21). However, it is not clear why the same considerations would not also apply to Fpg-initiated repair in bacteria (33). We have conducted a detailed kinetic investigation for the opposite base dependence of mOgg1, which allows direct comparison with Fpg, an enzyme with the same general substrate specificity but, presumably, a different mechanism of action. Preferences of mOgg1 and Fpg for the opposite base are markedly different. In the case of Fpg, replacement of dC opposite the lesion with dG or dT leads to a 10 -30-fold increase in the specificity constant, attributed to destabilization of hydrogen bonding at the site of damage (33). Such an effect could facilitate flipping out of the base to be excised (59). In contrast, mOgg1 specificity, measured by DNA strand nicking, decreased by at least 2 orders of magnitude when dC opposite the lesion was replaced by any other nucleotide. The glycosylase activity of mOgg1 was more tolerant to substitutions opposite the damaged base, but a preference for C is still clear (Table II). Binding experiments employing the tetrahydrofuran analog of an AP site indicate that the affinity of mOgg1 for DNA decreases after base excision unless C is positioned across from the lesion. Therefore, mOgg1 is similar to MutY with respect to its ability to recognize not only the damaged strand but also the complementary strand. It was recently shown that removal of the C terminus of MutY (approximately one-third of the full length extending beyond the core EndoIII homology region) leads to loss of preference for 8-oxoG as an opposite base, compared with G (60). mOgg1 also contains domains outside the core EndoIII homology region that may contribute to recognition of the base opposite the lesion. Such recognition may involve establishing hydrogen and van der Waals bonds with the lesion through the major or minor groove, and contacting the complementary base with a segment of the enzyme inserted into the helix while flipping the damaged base or both the damaged and complementary base out of the helix.
Opposite Base Preference and Ligand Recognition-Our K D data show that a substituent at C8 is crucial for mOgg1 binding. The nature of this substituent is also important; 8-amin-odG is a rather poor ligand, and there appears to be a requirement for an electronegative element or hydrogen bond acceptor. For Fpg, a critical element of substrate recognition is O 8 in the major groove (33). This configuration is present in 8-oxodG:dC and absent from 8-oxodG:dA, where 8-oxodG is in the syn conformation (34). No data are presently available concerning the structures of duplex DNA in aqueous solution containing the other types of purine analogs used in this study. It is generally believed that Watson-Crick base pairing provides sufficient stabilization energy to retain 8-oxodG in the anti conformation (61,62). If this is the case, replacing C as the opposite base with T, G, or A is expected to cause a progressive decrease in the occupation of the anti conformation of 8-oxodG and lead to an increase in apparent K D . This is the case both for Fpg (33) and mOgg1 (this paper). However, when purine analogs other than 8-oxoG are considered, this simple model does not accommodate mOgg1. For example, in the 8-oxodA:dC pair, both the steric effect of O 8 and reduced base pairing capacity should drive 8-oxodA into a syn conformation (55); nevertheless, this substrate is an excellent ligand for mOgg1, while 8-oxodA:dT, with 8-oxodA in an anti conformation (63), is not. Likewise, 8-oxodN, which cannot be stabilized in anti by hydrogen bonding, also is a good ligand. Therefore, a particular arrangement of hydrogen bond donors and acceptors in the grooves of the DNA is probably not the only major factor in the recognition of DNA damage by mOgg1. A model has been proposed (64) in which the enzyme first detects a local distortion and then probes the damaged and complementary bases, either while both are flipped out of the helix or when one base is flipped out and the other contacts protein inserted into the helix. It seems likely that the complementary base is checked before the damaged base is positioned for the catalytic reaction; otherwise, catalysis would not be influenced by the complementary base. If the complementary base is C, catalysis begins; otherwise, the complex may dissociate. Such a model also can explain the high affinity of mOgg1 and Fpg for DNA containing an abasic site or its analogs, an effect that was not readily accommodated in terms of a specific atom(s) in the major groove.
Catalytic Steps-DNA containing purine analogs that are good ligands but poor substrates may be used to analyze positions important for catalysis (33,34). For mOgg1, such analogs include 8-oxodN, 8-MeOdG, and 8-oxo-carbadG. 8-AminodG also may be considered with this group, since the observed decrease in substrate specificity (k cat /K m ) is higher than could reasonably be explained by the observed increase in K D . Among these analogs, 8-oxo-carbadG is resistant to glycosidic bond cleavage due to the amine rather than the hemiamine acetal character of C1Ј, as reported for carbocyclic analogs with Fpg and MutY (34,35). 8-OxoN, a good ligand but a poor substrate, differs from 8-oxoI, a good ligand and good substrate, by lacking a 6-keto group and a proton at N1. 8-OxoA and 6-OMe-8-oxoG also lack a proton at N1 but are excellent substrates. Therefore, we assume that the substituent at C6 is important for catalysis, while the N1 proton is not. Moreover, since the N1 proton has increased acidity in C8-keto purines (56,57,65), the tautomeric equilibrium at the N 1 -O 6 center may favor the lactim configuration in the enzyme active site; perhaps, 8-oxoG even exists there as a C6-enolate anion if a suitable anionic amino acid residue is present in the active site. If this is the case, the enzyme may utilize N1 as a hydrogen bond acceptor rather than a donor, providing an additional mechanism for discrimination against a C8-unsubstituted guanine. A substitution at C6 also destabilizes the proton at N7 (57), which may help to delocalize the positive charge accumulated on the base during the excision step. A third possibility is that a group at C6 may interact sterically with the enzyme active site to position the base properly for catalysis.
8-MeOG and 8-aminoG differ from 8-oxoG, a good ligand and a good substrate, by the nature of the C8 substitution and also by having a pyridinic rather than pyrrolic N7. This observation leads us to suggest another feature involved in the catalytic release of 8-oxoG as being either the keto function at C8 or a proton at N7. Both atoms may form hydrogen bonds; in addition, O 8 may participate in general acid catalysis or form an oxyanion (53).
The possible catalytic mechanisms of mOgg1 are summarized in Fig. 8. The mechanism of base excision was not included in the originally proposed unified model of glycosylase/AP lyase action, although base protonation has been postulated (66,67). For Fpg acting on an 8-oxodG-containing substrate, protonation at O 6 , combined with nucleophilic attack at C1Ј of the closed deoxyribose ring, was proposed to initiate base release (33,68). If base protonation also operates in the case of mOgg1, O 8 is a more likely site (see below) (Fig. 8A). The other mechanism to initiate base excision, however, is to protonate the heterocyclic oxygen atom of deoxyribose (Fig. 8B). Evidence supporting this mechanism emerged recently. For example, Fpg, EndoIII, and bacteriophage T4 endonuclease V are active on O-alkoxyamine-modified, ringopened abasic sites that share a double bond between C1Ј and the hydroxylamine nitrogen (69). KCN was found to inhibit DNA strand nicking and base release by mOgg1 (18). In addition, modified oligonucleotides with a pyrrolidine analog of deoxyribose, carrying a positive charge on the heteroatom, are ligands of very high affinity for many DNA glycosylases, presumably imitating the transition state (70,71). These findings indicate first that the glycosylase action may involve a C1Ј-base iminium intermediate with an open deoxyribose ring and, second, that the ring opens following protonation at the heterocyclic oxygen. S N 2 base displacement with formation of an oxy- carbenium ion was proposed as a mechanism (Fig. 8C), based on the affinity of some glycosylases for tetravalent positively charged pyrrolidine and pyrrolidine homonucleoside transition state analogs; nucleoside hydrolases appear to act in this way (71)(72)(73). However, the geometry of the protonated pyrrolidine analog is different from that of a planar oxycarbenium ion and more closely resembles an oxonium ion; thus, the evidence does not support this mechanism well. More likely, following ring opening, an amino group of the enzyme carries out imine exchange, displacing the base and forming an enzyme-DNA covalent complex (53,69). Base protonation may still be important, facilitating imine exchange.
In 8-oxodG, protonation may occur at the C6 and C8 keto groups, at N3 or at heterocyclic oxygen. Since mOgg1 acts efficiently on 8-oxoA and 6-OMe-8-oxoG, O 6 is unlikely to be the main protonation site. Although N 6 of 8-oxoA can be protonated as an imino tautomer, this mechanism cannot account for the protonation of 6-OMe-8-oxoG. A good ligand, 8-MeOdG, which allows protonation at N3, is not a substrate, indicating that N3 is not the main protonation site. Thus, participation of O 8 or a heterocyclic oxygen seems to be required for the catalytic activity of the enzyme. Although S N 2 base displacement cannot definitely be excluded, it seems less likely because of the existence of good ligands  that are poor substrates; the same considerations apply to protonation at a heterocyclic oxygen (Fig. 8, pathways C and B, respectively). The protonating amino acid (if any) cannot be determined from these experiments; one hypothesis is that it may be His 270 , which is conserved in Ogg1 and EndoIII enzymes throughout all species and, based on structures of other proteins in the EndoIII family, lies near the site of the damaged nucleotide binding (74,75). The Schiff base most probably forms with Lys 249 , as in the case of yeast and human Ogg1 (31,76).

Kinetic Mechanism of Cleavage by mOgg1
The distinctly different time course and kinetic parameters of base release and DNA strand nicking (Fig. 4, Table II) suggest a two-stage kinetic mechanism for mOgg1, qualitatively similar to that described for esterases and serine proteases (77). First, the damaged base is excised, and the enzyme remains covalently linked to substrate through a Schiff base (Fig. 9, steps 1 and 2). In a slower step, when C or T is positioned opposite the lesion, ␤-elimination takes place, and the imino intermediate is hydrolyzed (Fig. 9, step 3a). The short half-life of the covalent intermediate involving the 8-oxodG:dG pair, combined with the observation that base release from this substrate is more efficient than nicking, indicates that mOgg1 dissociates from 8-oxodG:dG shortly after the glycosylase step ( Fig. 9, step 3b). It is noteworthy that the k value calculated from covalent complex decay experiments for 8-oxodG:dG is closer to the independently measured k cat of base release, while k off for 8-oxodG:dC and 8-oxodG:dT is similar to the k cat of DNA strand nicking (Fig. 6, Table II), suggesting that mOgg1 does not dissociate from these substrates until ␤-elimination has occurred, leaving product release as the rate-determining step. When 8-oxodG:dA is the substrate, the absence of an exponential decay phase is probably due to the very low rate of formation of the covalent complex; the same model for dissociation can be applied to this pair. This dissociation event may not be coupled with physical partition of the enzyme and DNA, but the Schiff base disappears.
The paradoxical increase in the amount of NaBH 4 -sensitive covalent intermediate over time for 8-oxodG:dA and 8-ox-odG:dG substrates (Fig. 6) also may be explained by a twostage reaction mechanism. After the Schiff base is hydrolyzed, the enzyme is in effect presented with another substrate, an AP site opposite A, G, or T. The second phase of the reaction curve, an increase in the amount of the Schiff base, may be explained by formation of a second covalent complex that is kinetically irreversible when A or G is opposite the lesion (Fig. 9, steps  4b-4d). In MutY, another slow-turnover catalyst, a significant portion of the imino intermediate is kinetically trapped (46). The existence of such dead end covalent intermediates may be biologically significant, since only 8-oxodG:dC can be repaired by excision of 8-oxoG without a mutagenic event. Other mispairs involving 8-oxoG can only be repaired error-free through the MutY pathway or by recombination repair.

Biological Implications
Ogg1 is the DNA glycosylase primarily responsible for repair of 8-oxoG in mammals (26). In this respect, Ogg1 is a functional analog of Fpg in the bacterial GO system. Other components of this system have been identified in mammalian cells (78,79). In addition to 8-oxoG, Ogg1 may remove other damaged bases from DNA; however, such action could be mutagenic. For example, Ogg1 efficiently excises 8-oxoA paired with C (Ref. 80,FIG. 9. Proposed kinetic mechanism of mOgg1. 1, mOgg1 binds 8-oxoguanine (oG) positioned opposite the four canonical DNA bases with different efficiency. 2, the enzyme excises the damaged base and forms a covalent complex with the substrate containing an abasic site. 3a, if ␤-elimination occurs, the Schiff base is hydrolyzed, and the enzyme dissociates, leaving a nick in DNA. 4a, this nick is further processed by base excision repair, and Ogg1 catalyzes another round of repair. 3b, if the Schiff base is hydrolyzed without ␤-elimination, the enzyme remains bound to DNA containing an abasic site. Such a complex may form a second Schiff base (4b), nick DNA, and return in the productive enzymatic turnover (4c), or the enzyme may become irreversibly kinetically trapped (Ogg1*) in the complex (4d). The relative thickness of arrows reflects efficiencies of various steps. Ogg1 separated from (oG:X) by a horizontal line symbolizes noncovalent complexes; Ogg1 separated from (oG:X) by a vertical line symbolizes covalent complexes. this study). Although oxidation of adenine in DNA is expected to generate 8-oxoA:T pairs, A 3 G transitions are observed during replication of 8-oxoA-containing plasmids in vivo (81), indicating transient formation of a 8-oxoA:C mispair. Thus, the effect of Ogg1 would be to fix this mutation. Although 8-ami-noG, an important mutagenic adduct (82), is a poor substrate for Ogg1, it could be repaired by this enzyme in a backup pathway. One also can envision 8-oxoI formed from adenine as a result of clustered oxidative damage; this lesion also is a substrate for Ogg1.
Ogg1 has strong glycosylase activity but is not an efficient AP lyase in kinetic terms. In mammalian cells, Ogg1 may act in concert with other components of base excision repair, such as an AP endonuclease and DNA polymerase ␤, which could enhance its intrinsic lyase activity or displace the enzyme from the complex and process the AP site hydrolytically. At least one such reaction has been reported; human T-G mismatch glycosylase binds tightly to the AP site after base removal and is displaced by AP endonuclease (83). Unlike lesions removed by monofunctional DNA glycosylases, repair of 8-oxoG mainly follows a DNA polymerase ␤-dependent single-base patch pathway (48). Human AP endonuclease is physically associated with DNA polymerase ␤ (84); perhaps, when Ogg1 is bound to the processed lesion, the enzyme directs the assembly of an AP site-AP endonuclease-DNA polymerase ␤ complex, as opposed to an AP site-AP endonuclease-DNA polymerase ␦-PCNA complex; alternatively, it could limit directly the size of a DNA repair patch.

Structure-Function Correlates
After this work was complete, the three-dimensional structure of catalytically impaired human Ogg1 in complex with DNA containing 8-oxoG was solved at 2.1-Å resolution (85). Given the considerable degree of homology between the human and murine enzymes, structural features of the two enzymes are expected to be preserved. In the complex, 8-oxoG is flipped out of the DNA helix and is inserted deep into the enzyme active site, while side chains from Asn 142 , Arg 154 , Arg 204 , and Tyr 203 make contacts with C opposite the lesion. 8-OxoG forms hydrogen bonds via N7, O 6 , N1, and N 2 with Gly 42 , Gln 315 , and two water molecules. Surprisingly, O 8 forms no contacts with atoms in its vicinity. Binding of 8-oxodA and 6-OMe-8-oxodG by mOgg1 can be reconciled with the model as follows. If the Gln 315 side chain is rotated 180°, positions of hydrogen bond donors and acceptors will be reversed, allowing 8-oxoA to be accommodated in the active site pocket. 6-OMe-8-oxodG can be accommodated, albeit with strain, by either configuration of the Gln 315 side chain. This flexibility is analogous to that proposed for the accommodation of inosine into the active site of MutY (75). The model predicts a catalytic mechanism involving Lys 249 as a nucleophile, Asn 268 as an activator of Lys 249 , and His 270 either in protonation at O1Ј or in restoring protonation of Asn 268 after Lys 249 activation.
The three-dimensional structure is generally consistent with our data but does not explain certain findings, e.g. stability of the Schiff base intermediate or the existence of some good ligands that are poor substrates for Ogg1. Also, while the structure lends little support to the possibility of O 8 protonation, a mechanism that seems more favorable based on kinetic data, it does not distinguish between S N 2 base displacement and heterocyclic oxygen protonation. Further mechanistic studies of Ogg1 will be required to characterize the initial steps of substrate recognition and catalysis.