An Oxidative Damage-specific Endonuclease from Rat Liver Mitochondria*

Reactive oxygen species have been shown to generate mutagenic lesions in DNA. One of the most abundant lesions in both nuclear and mitochondrial DNA is 7,8-dihydro-8-oxoguanine (8-oxoG). We report here the partial purification and characterization of a mitochondrial oxidative damage endonuclease (mtODE) from rat liver that recognizes and incises at 8-oxoG and abasic sites in duplex DNA. Rat liver mitochondria were purified by differential and Percoll gradient centrifugation, and mtODE was extracted from Triton X-100-solubilized mitochondria. Incision activity was measured using a radiolabeled double-stranded DNA oligonucleotide containing a unique 8-oxoG, and reaction products were separated by polyacrylamide gel electrophoresis. Gel filtration chromatography predicts mtODE’s molecular mass to be between 25 and 30 kDa. mtODE has a monovalent cation optimum between 50 and 100 mm KCl and a pH optimum between 7.5 and 8. mtODE does not require any co-factors and is active in the presence of 5 mm EDTA. It is specific for 8-oxoG and preferentially incises at 8-oxoG:C base pairs. mtODE is a putative 8-oxoG glycosylase/lyase enzyme, because it can be covalently linked to the 8-oxoG oligonucleotide by sodium borohydride reduction. Comparison of mtODE’s activity with other known 8-oxoG glycosylases/lyases and mitochondrial enzymes reveals that this may be a novel protein.

Reactive oxygen species have been shown to generate mutagenic lesions in DNA. One of the most abundant lesions in both nuclear and mitochondrial DNA is 7,8dihydro-8-oxoguanine (8-oxoG). We report here the partial purification and characterization of a mitochondrial oxidative damage endonuclease (mtODE) from rat liver that recognizes and incises at 8-oxoG and abasic sites in duplex DNA. Rat liver mitochondria were purified by differential and Percoll gradient centrifugation, and mtODE was extracted from Triton X-100-solubilized mitochondria. Incision activity was measured using a radiolabeled double-stranded DNA oligonucleotide containing a unique 8-oxoG, and reaction products were separated by polyacrylamide gel electrophoresis. Gel filtration chromatography predicts mtODE's molecular mass to be between 25 and 30 kDa. mtODE has a monovalent cation optimum between 50 and 100 mM KCl and a pH optimum between 7.5 and 8. mtODE does not require any co-factors and is active in the presence of 5 mM EDTA. It is specific for 8-oxoG and preferentially incises at 8-oxoG:C base pairs. mtODE is a putative 8-oxoG glycosylase/lyase enzyme, because it can be covalently linked to the 8-oxoG oligonucleotide by sodium borohydride reduction. Comparison of mtODE's activity with other known 8-oxoG glycosylases/lyases and mitochondrial enzymes reveals that this may be a novel protein.
Reactive oxygen species are generated in cells as a by-product of cellular respiration. Reactive oxygen species react with proteins, lipids and DNA causing cellular damage. When DNA is the target, a variety of DNA adducts are formed. Among these, 8-oxoG is one of the most abundant lesions generated (1, 2). 8-oxoG 1 is considered to be a premutagenic lesion because it can mispair with adenine during DNA replication (3)(4)(5). In the absence of DNA repair, this mispairing results in G to T transversion mutations. Since many reactive oxygen species are generated by oxidative processes that occur in mitochondria, it is of great interest to understand the oxidative DNA damage processing in these organelles.
Mitochondrial DNA (mtDNA) is composed of a 16.5-kilobase pair circular genome, encoding 13 structural genes, 22 tRNAs, and two rRNAs. The DNA lies in close proximity to the free radical-producing electron transport chain, and it has been reported that mtDNA contains a higher level of oxidative DNA damage than nuclear DNA (6). Since mtDNA is subjected to relatively high levels of oxidative damage, mitochondria need DNA repair mechanisms to maintain their DNA genomes.
Mitochondrial DNA repair mechanisms differ from those in the nucleus. Evidence to suggest that mitochondria lacked DNA repair mechanisms came from the observation that UV damage is not repaired in mitochondria (7,8), while it is efficiently processed in the nucleus by nucleotide excision repair. In addition, damage caused by cisplatin and nitrogen mustard, agents that are known to induce DNA adducts that are substrates for the nuclear nucleotide excision repair pathway, is inefficiently repaired in mitochondria (8). From these studies it has been widely assumed that mitochondria have no DNA repair capacity. Clearly, nucleotide excision repair, as it exists in the nucleus, does not appear to be present in mitochondria. However, it still remains to be determined whether nucleotide excision repair factors participate in mitochondrial DNA repair.
This laboratory and others have demonstrated that mitochondria repair some forms of DNA damage. DNA damage induced by monofunctional alkylating agents, alloxan, streptozotocin, and acridine orange, agents that generate lesions known to be repaired by base excision repair (BER), is removed from mtDNA (8 -13). Further support for BER in mitochondria comes from the fact that enzymes involved in BER such as uracil DNA glycosylase, apurinic/apyrimidinic (AP) endonucleases, and a methyl transferase have been purified from mitochondria (14 -18). Recently, recombination repair has been reported in mitochondria (8,19). Thus, it is evident that mitochondria contain some base excision and recombinational DNA repair mechanisms.
BER is an important mechanism that cells use for the removal of oxidative DNA damage. In the process of BER, a damage-specific glycosylase recognizes a damaged base and then cleaves the N-glycosylic bond between the sugar and the base, generating an AP site. Some glycosylases have an associated AP lyase function that cleaves the DNA phosphate backbone, while others rely on AP endonucleases for strand cleavage. Next, a phosphodiesterase excises the 3Ј-terminal unsaturated sugar derivative. The resulting one-nucleotide gap is resynthesized by a DNA polymerase, and the ends are sealed by a DNA ligase (20).
The isolation of oxidative damage repair glycosylases from higher eukaryotes has been challenging due to the fact that these enzymes are typically expressed at very low levels in cells. Only recently have genetic approaches proven successful in the cloning of the yeast, mouse, and human enzymes (21)(22)(23)(24)(25)(26). The purification of enzymes from mitochondria is even further complicated by the necessity to isolate pure mitochondria in sufficient quantities for such studies. This together with the low abundance of these enzymes has been the reason for the lack of progress in the isolation of mitochondrial repair enzymes. Thus, no oxidative damage-processing enzymes from mitochondria have been purified or characterized extensively.
Previous studies from this laboratory and others suggested that mitochondria may contain a glycosylase/lyase enzyme responsible for the removal of oxidative damage (13,17). In our recent study on the repair of oxidative damage in mitochondria, Chinese hamster ovary cells were exposed to acridine orange plus light for the purpose of introducing 8-oxoG and other oxidative DNA damage lesions (13). The DNA repair profile showed that these lesions were rapidly removed from the mitochondrial genome with more than 60% of the lesions repaired within 4 h, suggesting that mammalian cells possess mitochondrial enzymes that recognize and remove 8-oxoG. In this study, we have partially purified a mitochondrial oxidative damage endonuclease that is responsible for the recognition and incision of 8-oxoG and abasic sites. This is the first report of the purification of an 8-oxoG DNA damage-processing enzyme from mitochondria.

EXPERIMENTAL PROCEDURES
Materials-Livers were removed from 6-month-old male white Wistar rats (Gerontology Research Center animal colony). The columns and chromatography matrices were obtained from Pharmacia Biotech Inc. Protease inhibitors leupeptin, aprotinin, and E-64 were from Boehringer Mannheim. Benzamidine HCl, dithiothreitol, bovine serum albumin, pepstatin, and phenylmethylsulfonyl fluoride were from Sigma. Radioisotopes were from NEN Life Science Products. G-25 spin columns were purchased from 5 Prime 3 3 Prime, Inc. (Boulder, CO). T4 polynucleotide kinase was obtained from U.S. Biochemical Corp. Fpg protein was provided by Dr. Arthur Grollman (Stony Brook, NY). Endonuclease III and endonuclease IV were purchased from Trevigen, Inc. Uracil DNA glycosylase was obtained from Boehringer Mannheim.
Mitochondrial Enzyme Assays-To determine the localization of mtODE within the mitochondria, Percoll gradient-purified mitochondria were subjected to a nonspecific protease treatment. The Percoll gradient-purified mitochondria were separated into two samples; one was incubated with Pronase (7.5 mg/ml) and resuspended in 120 mM KCl and 20 mM HEPES, pH 7.4, while the other fraction was incubated with buffer. mtODE activity was assayed from both fractions with protease inhibitors present as described below.
To confirm the localization of mtODE within mitochondria, the Percoll gradient-purified mitochondria were subjected to fractionation by digitonin according to the method of Ragan (27). Digitonin fractionation separated the mitochondria into two compartments, the outer membrane and inner membrane/matrix fractions. Once separated, the composition of the individual fractions was examined by enzymatic analysis. The activity of monoamine oxidase, a marker for the outer membrane, was measured spectrophotometrically according to the method described by Ragan (27). Cytochrome c oxidase activity was measured spectrophotometrically as a marker for the inner membrane/ matrix compartment as described by Sottocasa et al. (28). Also, lactate dehydrogenase activity was measured to determine the level of cytosolic contamination in the Percoll gradient-purified mitochondria as described by Wroblewski et al. (29). mtODE activity was measured as described below.
Oligonucleotides-All oligonucleotides were from Midland Certified Reagent Co., except for the TG oligonucleotide and its complementary strand, which were synthesized by Paragon Biotech, Inc. In addition, the TOG oligonucleotide, containing the 8-oxoG, was provided by Dr. Francis Johnson (Stony Brook, NY). The sequences of the oligonucleotides utilized in this study are displayed in Table I. Oligonucleotides containing either the DNA lesion of interest or an unmodified base were 5Ј-end-labeled using T4 polynucleotide kinase and [␥-32 P]ATP. To separate the free [␥-32 P]ATP from the oligonucleotides, the reaction mixture was passed through a G-25 spin column. Complementary oligonucleotides were annealed in 100 mM KCl, 10 mM Tris, pH 7.8, and 1 mM EDTA by heating the oligonucleotides to 80°C and then allowed to cool slowly to room temperature.
Oligonucleotide Incision Assay-Incision reactions (20 l) contained 40 mM HEPES-KOH (pH 7.6), 75 mM KCl, 2 mM dithiothreitol, 1 mM EDTA, 0.1 mg/ml bovine serum albumin, 5.9 nM 32 P-labeled duplex, and the column fractions at the amounts indicated for each experiment. The reaction was incubated overnight at 32°C and then terminated by the addition of 0.8 l of 5 mg/ml proteinase K and 0.8 l of 10% SDS and incubated for 15 min at 55°C. The DNA was ethanol-precipitated by the addition of 4 l of 11 M ammonium acetate, 1 g of glycogen, and 60 l of ethanol. The DNA was pelleted, dried, then resuspended in formamide dye containing 90% formamide, 0.002% of bromphenol blue, and 0.002% xylene cyanol, heated to 80°C for 2 min, and subjected to electrophoresis on a denaturing 20% polyacrylamide gel containing 7 M urea. Samples containing the AP oligonucleotide were heated to 55°C, instead of 80°C, prior to loading due to the heat lability of the AP site. After electrophoresis, the gel was subjected to autoradiography at Ϫ80°C. Quantitation of results employed a Molecular Dynamics Phos-phorImager and ImageQuant NT software. We define 1 unit of 8-oxoG endonuclease activity as 1 fmol of oligonucleotide incised during an 18-h incubation at 32°C.
Preparation of Substrates-The AP oligonucleotide listed in Table I was originally synthesized with a uracil at the 11th position. To generate the abasic site oligonucleotide, uracil DNA glycosylase (1 unit) was incubated with the uracil-containing oligonucleotide duplex for 30 min at 37°C. To demonstrate that AP sites were generated, endonuclease III (1 unit), IV (1 unit), or Fpg protein (0.8 unit) were added and incubated with the duplex for 2 h, under the conditions described for oligonucleotide incision assay. More than 97% of the oligonucleotide was incised by each enzyme.
To generate the ring-opened guanine adduct, FapyG, the FC oligonucleotide (Table I) was treated with dimethyl sulfate and alkali according to Chetsanga and Lindahl (30). The resultant oligonucleotide product was designated FapyG. To demonstrate that this treatment induced the FapyG adduct, the oligonucleotide was duplexed with its complementary oligonucleotide and incubated with Fpg protein (0.8 unit) for 2 h at 37°C under oligonucleotide incision assay conditions. Greater than 57% of the oligonucleotide was sensitive to Fpg cleavage. To determine whether any abasic sites were generated in the course of preparing the substrate, the FapyG duplex was incubated with endonuclease IV (1 unit) or endonuclease III (1 unit), resulting in 3 and 6% of the oligonucleotide being incised, respectively. Preparation of Mitochondria-Mitochondria were purified from rat liver using a combination of differential and Percoll gradient centrifugation (modified from Ref. 31). All procedures were carried out at 4°C. Briefly, fresh liver was minced and homogenized in 1 ϫ M-SHE buffer (0.21 M mannitol, 0.07 M sucrose, 10 mM Hepes (pH 7.4), 1 mM EDTA, 1 mM EGTA, 0.15 mM spermine, 0.75 mM spermidine). The following protease inhibitors were added just before use: 1 mM dithiothreitol, 2 g/ml leupeptin, 2 M benzamidine-HCl, 1 mM phenylmethylsulfonyl fluoride. Unbroken cells and nuclei were pelleted at 500 ϫ g. The supernatant, containing the mitochondria, was centrifuged at 9,500 ϫ g to pellet the mitochondria. The mitochondria were resuspended and washed twice with 1 ϫ M-SHE buffer. Mitochondria (2 ml) were layered onto a Percoll solution (37.5 ml; 50% Percoll and 50% 2 ϫ M-SHE) and centrifuged at 50,000 ϫ g in a Ti-60 rotor for 1 h. The brown mitochondrial band was collected either by fractionating the gradient or by direct syringe aspiration. The collected mitochondria were pooled, diluted 10-fold with 1 ϫ M-SHE, and then pelleted by centrifugation. The mitochondria were resuspended in 50 ml of 1 ϫ M-SHE, washed, pelleted, and then frozen at Ϫ80°C until lysis.
Purification of Mitochondrial Oxidative Damage Endonuclease-The clarified mitochondrial supernatant (fraction I, 453 mg) was applied to a 15-ml DEAE-Sepharose Fast Flow column equilibrated with buffer A containing 300 mM KCl. After applying the sample, the column was washed with 50 ml of the same buffer. Five-ml fractions were collected. The majority of the mtODE activity did not bind the column. Fractions containing the bulk of the protein were pooled, and the salt concentration was adjusted to 100 mM KCl with buffer A and designated fraction II. Fraction II (384 mg) was loaded directly onto a FPLC 10/10 HR Mono S column equilibrated with buffer A containing 100 mM KCl. The column was washed with equilibration buffer until a stable base line was attained and then eluted with a 40-ml linear KCl gradient (100 mM to 1 M KCl). One-ml fractions were collected. Fractions were dialyzed against buffer A containing 100 mM KCl overnight and assayed for mtODE activity. The 8-oxoG oligonucleotide incising activity eluted after the bulk of the protein as a single peak at approximately 650 mM KCl (fraction III). At this stage in the purification, mtODE has an absolute requirement for detergent in the dialysis buffer. Fraction III (0.586 mg) was dialyzed against buffer B (20 mM Hepes, pH 7.6, 300 mM KCl, 5% glycerol, 0.015% Triton X-100, and 2 mM dithiothreitol) and loaded onto a Superdex 75 HR 10/30 gel filtration column that had been equilibrated with buffer B. The column was calibrated with blue dextran 2000, albumin, ovalbumin, chymotrypsinogen A, and ribonuclease A (low molecular weight standards; Pharmacia). The column fractions (0.4 ml) were dialyzed against buffer A overnight and then assayed for activity. The elution profile of standard proteins was used to calculate the molecular weight of mtODE. The active fractions were pooled, concentrated, and stored in 20 mM Hepes, pH 7.6, 100 mM KCl, and 50% glycerol at 4°C (fraction IV, 0.097 mg).
Cross-linking of mtODE to the 8-oxoG Oligonucleotide by Sodium Borohydride Reduction-Reactions were carried out as described above except that the experiments were performed in the presence of 5 mM EDTA, 50 mM NaBH 4 (freshly prepared), and 20 mM KCl instead of 75 mM KCl (modified from Ref. 32). To monitor the extent of the incision, 2 l of the reaction was added to 5 l of formamide dye, heated to 80°C for 2 min, and then electrophoresed through a 20% polyacrylamide gel containing 7 M urea. To detect the covalent complex, 15 l of the reaction was added to 15 l of 2 ϫ SDS-PAGE loading buffer (Novex), boiled for 10 min, and then separated by a 12% PAGE gel (Novex). Quantitation of the results was performed using a Molecular Dynamics PhosphorImager.

RESULTS
Purification of Rat Liver Mitochondria-The major source of DNA repair enzymes is the nucleus; therefore, purification of mitochondria is a crucial step in the isolation of mitochondrial DNA repair enzymes. Using mitochondria isolated by differential and Percoll gradient centrifugation, we identified an enzy-matic activity that specifically incised an 8-oxoG-containing oligonucleotide (Fig. 1A, lanes 3 and 4). To determine whether the mtODE activity was contained within the mitochondria, the Percoll gradient-purified mitochondria were treated with a nonspecific protease, Pronase, at 7.5 mg/ml for 9 min on ice. This exposure to protease causes extensive digestion of myofibrils and inactivates cytosolic enzymes in routine preparation of heart mitochondria. When compared with mitochondria that had not been exposed to protease, 84% of mtODE's activity remained (Fig. 1A, compare lanes 4 and 6).
To further demonstrate that the mtODE activity is contained within mitochondria, submitochondrial fractions were prepared by treating the Percoll gradient-purified mitochondria with digitonin as described by Ragan (27). Fractionation by digitonin produced two fractions, an outer membrane and a mitoplast fraction. The mitoplast fraction contains both the mitochondrial inner membrane and matrix compartments. The mitochondrial outer membrane fraction was associated with enhanced monoamine oxidase activity and low cytochrome c oxidase and mtODE activity (Table II). When compared with the outer membrane compartment, the mitoplasts had lower monoamine oxidase activity and higher cytochrome c oxidase and mtODE activity. The majority of mtODE's activity (88%) co-localized to the mitoplast compartment along with the cytochrome c oxidase activity. Lactate dehydrogenase was assayed in the crude liver homogenate and in the Percoll gradientpurified mitochondrial fractions to determine the level of cytosolic protein contamination. The activity declined 14-fold from 8.4 Ϯ 0.1 to 0.6 Ϯ 0.1 mol/mg/min, respectively, suggesting limited amounts of cytosolic contamination.
Purification of mtODE-An oligonucleotide duplex containing a unique 8-oxoG was employed as a substrate to follow the purification of an 8-oxoG incising activity from rat liver mitochondria. mtODE was purified approximately 500-fold from isolated mitochondria, and the purification is summarized in Table III. The most purified fraction has three bands migrating at approximately the 30,000 molecular weight range as detected by silver staining of a SDS-polyacrylamide gel.
Size and Catalytic Properties-As determined by gel filtration, the active enzyme eluted at a position corresponding to a molecular weight range of 25,000 -30,000, assuming a globular protein (Fig. 2). mtODE has an optimum KCl concentration between 50 and 100 mM KCl. While it is active across a broad pH range, its pH optimum is between pH 7.5 and 8.0. Divalent   FIG. 1. mtODE localization. A, lanes 1 and 2 contain buffer and duplexes only. One hundred g of protein from the non-proteasetreated (lanes 3 and 4) or protease-treated (lanes 5 and 6) Percoll gradient-purified mitochondria were assayed for 8-oxoG oligonucleotide (OG) incision activity as described under "Experimental Procedures." Lanes 1, 3, and 5 contain the control duplex, G. Lanes 2, 4, and 6 contain OG. B, each lane contains 100 g of protein incubated with either the G duplex (lanes 1, 3, and 5) or the OG duplex (lanes 2, 4, and 6). Lanes 1 and 2, contain protein from Percoll gradient-purified mitochondria (PG), lanes 3 and 4 contain the mitoplast protein fraction (MP), and lanes 5 and 6 were incubated with the outer membrane protein fraction (OM). Quantitation of the gel in panel B is shown in Table II. cations are not required, and mtODE is resistant to 5 mM EDTA.
Substrate Specificity-The substrate specificity of mtODE (fraction IV) was examined. The oligonucleotide incision assay was carried out using a panel of substrates including singlestranded oligonucleotides, duplexed oligonucleotides containing 8-oxoG, an abasic site, FapyG adducts, or their respective unmodified control oligonucleotide duplexes (Fig. 3). mtODE does not incise 8-oxoG in single-stranded DNA (lane 2); however, duplexed 8-oxoG and abasic sites are cleaved by mtODE (lanes 4 and 6, respectively). Duplexes containing the ringopened guanine adduct, FapyG, were also not recognized by the mtODE. As seen in Fig. 3, there is a weak incision product produced in the FapyG lane (4% of oligonucleotide incised); however, this is due to AP sites in the oligonucleotide and not to cleavage of FapyG residues, because both endonuclease IV and III were able to generate the same level of incision (3 and 6% of oligonucleotide incised, respectively).
The lack of FapyG cleavage was confirmed by another assay. We measured the release of [ 3 H]FapyG from calf thymus DNA by the assay described by de Oliveira (33). Fpg protein served as a positive control. mtODE was not able to release [ 3 H]FapyG above background, while Fpg consistently released greater than 90% of the label (data not shown). Therefore, mtODE recognizes and incises at 8-oxoG and abasic sites within doublestranded DNA but does not cleave 8-oxoG within single-stranded DNA or the ring-opened guanine adduct, FapyG.
Reaction Products-To determine whether mtODE made a single-stranded break or a double-stranded break, we employed duplexes labeled on both 5Ј-ends. mtODE made a singlestranded break on the DNA strand containing the 8-oxoG lesion (data not shown). To define the 3Ј-cleavage product, the 5Ј-end-labeled 8-oxoG-containing oligonucleotide duplex was incubated with fractions from the purification of mtODE. Fig. 5 compares the reaction products generated by mtODE fractions III and IV with those produced by treating an abasic site oligonucleotide with endonuclease III, endonuclease IV, or Fpg protein. Endonuclease III is known to generate a 3Ј-terminal unsaturated sugar derivative; endonuclease IV generates a 3Ј-hydroxyl group; and Fpg generates a 3Ј-phosphate group. mtODE produces a band that is consistent with a 10-mer containing a 3Ј-terminal unsaturated sugar derivative, the same product generated by endonuclease III (compare lane 1 with lanes 4 and 5). As can be seen in lane 4, there is a weak cleavage product corresponding to a 10-mer with a 3Ј-phosphate, like that generated by the Fpg protein. However, the more purified gel filtration fraction of mtODE, fraction IV, a Enzyme activities were measured as described under "Experimental Procedures." Data are presented as the mean Ϯ S.E. b Percoll gradient purified mitochondria were fractionated by treatment with digitonin as described by Ragan (27).  ). The oligonucleotides were as follows. ss G, single-stranded G oligonucleotide; ss OG, single-stranded 8-oxoG oligonucleotide; G, duplexed G; OG, duplexed OG oligonucleotide; APC, duplexed AP control oligonucleotide; AP, duplexed abasic site oligonucleotide; FC, untreated control oligonucleotide; FapyG, duplexed FapyG oligonucleotide. The reaction products were separated on a 20% urea-polyacrylamide gel, and then the gel was subjected to autoradiography. produces only the 3Ј-terminal unsaturated sugar derivative.
To determine the 5Ј-cleavage product, the 8-oxoG containing oligonucleotide was labeled on the 3Ј-end and then duplexed with its complementary oligonucleotide. Cleavage of this substrate revealed that mtODE generated a 5Ј-phosphate product, which is identical to that produced by the Fpg protein (data not shown).
Thus, mtODE makes a single-stranded break at the site of the 8-oxoG delimited by a 3Ј-terminal unsaturated sugar derivative and a 5Ј-phosphate ( Fig. 5 and data not shown). Both of the products are consistent with a ␤-elimination reaction mechanism, like that employed by endonuclease III (34 -36).
Catalytic Mechanism-In an attempt to support the proposed ␤-elimination mechanism for mtODE, we performed a sodium borohydride trapping experiment (32,37). If the enzyme acts like the bacterial enzymes endonuclease III or Fpg protein, which utilize covalent enzyme-DNA Schiff base intermediates, then sodium borohydride should be able to reduce this putative intermediate and covalently link the enzyme to the 32 P-end-labeled DNA, which can be detected by an oligonucleotide band shift. If the enzyme utilizes a noncovalent mechanism to cleave the DNA, then no labeled band should be revealed. Fraction IV was incubated with duplexes containing either a G:C or 8-oxoG:C base pairs in the presence of 50 mM sodium borohydride. Fig. 6A demonstrates that the enzyme is active under these conditions. Fig. 6B shows the trapping of mtODE to the 32 P-labeled oligonucleotide and shows that the band was dependent on the presence of 8-oxoG in the oligonucleotide. Incubation of mtODE with the DNA produced one labeled band, which migrated as a 54-kDa product. As expected, the Fpg protein was able to be covalently attached to the 32 P-labeled 8-oxoG-containing oligonucleotide. Under these reaction conditions, the Fpg protein, a 30-kDa protein, migrated with an apparent molecular mass of 48 kDa. This experiment provides support that mtODE cleaves the DNA via a covalent DNA-enzyme intermediate, a Schiff base intermediate. Taken together, Figs. 5 and 6 support our suggestion that mtODE proceeds through a ␤-elimination mechanism.
The sodium borohydride trapping experiment also suggests that mtODE may be an 8-oxoG glycosylase/lyase; however, due to the low abundance of the enzyme, we have not demonstrated 8-oxoG glycosylase activity by release of the free 8-oxoG base. DISCUSSION To examine the processing of oxidative DNA damage in mitochondria, we have partially purified an 8-oxoG-specific DNA endonuclease from rat liver mitochondria. We named this protein mtODE to reflect the fact that the enzyme recognizes 8-oxoG and abasic sites.
Although we expect this activity to be present in the nucleus, we are convinced that the 8-oxoG incision activity seen here is not due to a nuclear contamination. To demonstrate that mtODE is localized within the mitochondria, Percoll gradientpurified mitochondria were treated with a nonspecific protease or fractionated into submitochondrial fractions. mtODE activity was protease-resistant (Fig. 1A) and co-localized to the mitoplasts with the mitochondrial inner membrane marker enzyme, cytochrome c oxidase ( Fig. 1B and Table II). The localization of mtODE to the mitoplast compartment does not distinguish between an inner membrane and matrix localization. mtODE has been purified approximately 500-fold from isolated rat liver mitochondria. Based on the gel filtration and chemical cross-linking analysis, it has an estimated molecular weight in the range of 25,000 -30,000. It cleaves at 8-oxoG and abasic sites in double-stranded DNA but does not incise 8-oxoG in single-stranded DNA or the ring-opened formamidopyrimidine adduct, FapyG. Mismatched guanine duplexes are also not a substrate for the enzyme. mtODE makes a single-stranded break in the strand containing the lesion, and its reaction products are consistent with a ␤-elimination mechanism of cleavage.
DNA incision can be strongly influenced by the base opposite the 8-oxoG, and mtODE's base pairing preference is 8-oxoG opposite C Ͼ Ͼ T, G Ͼ A. The preference for 8-oxoG:C suggests that the enzyme functions more effectively before replication than during or after, when 8-oxoG:A mispairs would be introduced by polymerase ␥. In bacterial and mammalian cells, three proteins participate in the repair of guanine damage: an 8-oxoG DNA glycosylase/AP lyase (Fpg protein or mut M), an adenine DNA glycosylase (mut Y) that recognizes and removes adenine when base paired with 8-oxoG and an 8-oxo-dGTPase enzyme (mut T) that hydrolyzes the 8-oxodG triphosphate to the monophosphate so that it cannot be incorporated into newly synthesized DNA (for a recent review, see Ref. 38). Thus far, two repair enzymes for guanine damage have been isolated from mitochondria. mtODE is a functional homolog of the Fpg protein, and a mitochondrial mut T homolog has previously been identified (39). It remains to be determined whether mitochondria also contain a mut Y homolog to process 8-oxoG:A mispairs or whether this damage is allowed to persist.
Comparison of mtODE with other 8-oxoG-processing enzymes that have been isolated from a variety of sources, other than mitochondria, suggests that it may be a novel protein. In Escherichia coli, the Fpg protein is responsible for the removal of 8-oxoG from DNA (40,41). Fpg is a DNA glycosylase with an associated AP lyase activity. As shown in Fig. 5, mtODE and Fpg process the damage differently. Fpg uses a ␤,␦-elimination mechanism (42,43), while mtODE only employs a ␤-elimination step.
Three proteins that process 8-oxoG DNA damage have been identified from Saccharomyces cerevisiae, Fapy DNA glycosylase, OGG1, and OGG2, and all are proposed to be DNA glycosylase/lyase enzymes (21,22,33). It is most intriguing that yeast OGG1, oxoguanine glycosylase 1 (21,22), and mtODE share the same preference for 8-oxoG:C base pairs. Both enzymes are active at high salt concentrations (100 mM KCl). In addition, both enzymes can be covalently linked to the DNA when the reaction proceeds in the presence of sodium borohydride. It should be noted that yeast OGG1 prefers 8-oxoG 12-fold more than FapyG (21), while mtODE does not cleave FapyG. However, the differences in substrate utilization may be species-specific or relate to the purity and/or integrity of the mtODE preparation.
Recently, the human and mouse homologs of the yeast OGG1 have been cloned (23)(24)(25)(26). mtODE shares many characteristics with the mouse OGG1. Additionally, Rosenquist et al. (26) have identified a putative mitochondrial targeting sequence within the human and mouse genes. mtODE may be the mitochondrial form of OGG1.
Several nuclear repair activities for 8-oxoG have been identified in human cells. Bessho et al. (44) described two activities from HeLa nuclear extracts: an 8-oxoG glycosylase and an 8-oxoG endonuclease activity. Since the abundance of the mtODE is very low, we have not directly determined whether mtODE contains a glycosylase activity. However, since mtODE is capable of being trapped by sodium borohydride, we predict that it is a glycosylase. mtODE differs from both of these nuclear enzymes described by Bessho et al. because they resolve the 3Ј and 5Ј single-stranded break ends differently. mtODE produced a 3Ј-deoxyribose moiety and a 5Ј-phosphate, while the nuclear glycosylase/AP endonuclease generated a 3Ј-phosphate and a 5Ј-hydroxyl group, and the nuclear 8-oxoG endonuclease cleaved the oligonucleotide, leaving a 1-nucleotide gap delimited by 3Ј-and 5Ј-hydroxyl groups.
Yamamoto et al. (45) also described an 8-oxoG endonuclease that produced different 3Ј-ends than mtODE. This enzyme shares some properties with mtODE including being EDTAresistant and having a preference for 8-oxoG:C base pairs.
Another mammalian enzyme that cleaves at 8-oxoG is the Drosophila ribosomal protein S3 (46). Its proposed mechanism of action is via a ␤,␦-elimination similar to that of the Fpg protein; however, S3 only displays a weak ␦-elimination. Unlike mtODE, the S3 protein was reported to cleave at FapyG adducts (46). It should be noted that the human ribosomal S3 protein has not been shown to process 8-oxoG or FapyG adducts (47). 2 mtODE is separable from the major endonuclease previously identified in mitochondria, mitochondrial endonuclease G (48). A role for endonuclease G in oxidative damage processing has been proposed (49). The reaction conditions for mtODE and endonuclease G differ dramatically. Endonuclease G is activated in reaction buffers of low ionic strength and requires a divalent cation, while mtODE is activated in higher ionic strength buffers and is resistant to EDTA. mtODE is not endonuclease G.
Other DNA damage-processing enzymes have been purified from mitochondria, but it is unclear how they compare with mtODE. Tomkinson et al. (17) described three separable AP endonuclease activities (designated I, II, and III) from mouse plasmacytoma mitochondria (17). All three were reported to be bifunctional enzymes containing glycosylase and AP endonuclease activities. In addition, all three enzymes appeared to cleave the DNA substrates like the bacterial enzyme endonuclease III. Mitochondrial endonuclease II shares some properties with mtODE including its small size and KCl optimum (17). However, no specific substrates have been defined for these mitochondrial endonucleases, and therefore we cannot compare their substrate specificities with that of mtODE.
The presence of mtODE in mitochondria supports our previous findings that mitochondria repair oxidative damage. This is the first report to describe the isolation and characterization of an 8-oxoG incising activity from mitochondria. mtODE shares a significant amount of functional similarity to the OGG1 proteins and may be the rat mitochondrial homolog. Future experiments will need to be done to determine whether they are encoded by the same gene and then transported to 2 S. Linn, personal communication.
FIG. 6. Sodium borohydride crosslinking of mtODE to the 32 P-labeled 8-oxoG oligonucleotide. mtODE (fraction IV, 0.4 g) or Fpg protein (0.04 units) was incubated with control duplex (G) or 8-oxoG-containing duplex (OG) overnight at 32°C in the presence of 50 mM NaBH 4 in a modified incision reaction buffer. A, to monitor the extent of the incision reaction, 2 l of the reaction mixture was added to 5 l of formamide dye and then separated on a 20% urea-polyacrylamide gel. B, to detect the covalently linked protein-DNA complexes, 15 l of the reaction mixture was added to 15 l of 2 ϫ SDS-PAGE loading buffer (Novex), boiled for 10 min, and then separated on a 12% PAGE gel. The positions of the protein standards are indicated by hash marks. different subcellular compartments. Large scale preparations of mtODE are under way for protein microsequencing so that we can compare its sequence to OGG1 and other 8-oxoG-processing enzymes.