Compromised Incision of Oxidized Pyrimidines in Liver Mitochondria of Mice Deficient in NTH1 and OGG1 Glycosylases*

Mitochondrial DNA is constantly exposed to high levels of endogenously produced reactive oxygen species, resulting in elevated levels of oxidative damaged DNA bases. A large spectrum of DNA base alterations can be detected after oxidative stress, and many of these are highly mutagenic. Thus, an efficient repair of these is necessary for survival. Some of the DNA repair pathways involved have been characterized, but others are not yet determined. A DNA repair activity for thymine glycol and other oxidized pyrimidines has been described in mammalian mitochondria, but the nature of the glycosylases involved in this pathway remains unclear. The generation of mouse strains lacking murine thymine glycol-DNA glycosylase (mNTH1) and/or murine 8-oxoguanine-DNA glycosylase (mOGG1), the two major DNA N-glycosylase/apurinic/apyrimidinic (AP) lyases involved in the repair of oxidative base damage in the nucleus, has provided very useful biological model systems for the study of the function of these and other glycosylases in mitochondrial DNA repair. In this study, mouse liver mitochondrial extracts were generated from mNTH1-, mOGG1-, and [mNTH1, mOGG1]-deficient mice to ascertain the role of each of these glycosylases in the repair of oxidized pyrimidine base damage. We also characterized for the first time the incision of various modified bases in mitochondrial extracts from a double-knock-out [mNTH1, mOGG1]-deficient mouse. We show that mNTH1 is responsible for the repair of thymine glycols in mitochondrial DNA, whereas other glycosylase/AP lyases also participate in removing other oxidized pyrimidines, such as 5-hydroxycytosine and 5-hydroxyuracil. We did not detect a backup glycosylase or glycosylase/AP lyase activity for thymine glycol in the mitochondrial mouse extracts.

Under normal metabolic conditions, reactive oxygen species are generated primarily as a by-product of the electron transport chain in mitochondria (1). The mtDNA represents a particularly susceptible target for reactive oxygen species, as it is closely associated with the mitochondrial inner membrane where most reactive oxygen species are generated (2). Various laboratories (3,4), including ours (5), have demonstrated that oxidative DNA base damage accumulates more in the mitochondrial DNA than in the nuclear DNA.
The endonuclease III (NTH) 1 family of glycosylases is a highly conserved class of repair enzymes that are found from bacteria to man (6). NTH1 is the main endonuclease III homologue found in mice and humans (7)(8)(9). Its glycosylase/AP lyase component has very broad substrate specificity, and it removes several different types of oxidized bases from DNA (10,11). The primary target for NTH1 is oxidized pyrimidines in DNA, and NTH1 is involved in the repair of thymine glycol (5,6-dihydroxy-5,6-dihydrothymine (Tg)) in vivo (12).
Our laboratory has previously demonstrated the existence of a repair system specific for thymine glycols in mammalian mitochondria (13). A Tg glycosylase/AP lyase (mtTGEndo) was isolated from rat liver mitochondria to near-homogeneity. Based upon functional studies, we concluded that this protein, with an apparent molecular mass of Ϸ37 kDa, was the rat mitochondria homologue of endonuclease III.
Although there is good evidence for the presence of an endonuclease III homologue in Saccharomyces cerevisiae mitochondria (Ntg1p) (14), the situation in mammalian mitochondria is still not clear. By using a FLAG-tagged construct and anti-FLAG antibodies, Takao et al. (15) observed sorting of hNTH1 to both the nucleus and mitochondria. Cytosolic labeling of HeLa cells with anti-hNTH1 also suggested mitochondrial localization (16). On the other hand, Luna et al. (11) failed to detect any mitochondrial localization of an enhanced green fluorescent protein-hNTH1 in live HeLa cells. So far, no studies have addressed the possible mitochondrial localization of the murine NTH1.
Very recently, Takao et al. (16) demonstrated the presence of a Tg incision activity in mouse liver mitochondrial extracts that was lost upon disruption of the mNTH1 gene. The authors also reported the presence of another Tg incision activity in mitochondrial extracts from mNTH1-deficient mice, which they called TGG1 (thymine glycol glycosylase 1). TGG1 seems to be a monofunctional glycosylase that is induced upon the knockout of mNTH1.
Here, we have used extracts from mNTH1-deficient mice to investigate the role of this protein in mitochondrial DNA repair of various oxidized pyrimidine DNA base damage. We also examined the possible overlapping roles of the mNTH1 and the mOGG1 proteins in mitochondria using extracts from mOGG1deficient and double-knock-out animals. This is the first characterization of BER enzymes in mitochondria from doubleknock-out mice deficient in the two major DNA glycosylases for the repair of oxidized bases.

EXPERIMENTAL PROCEDURES
Materials-Percoll was obtained from Amersham Biosciences. Endonuclease III was obtained from Trevigen (Gaithersburg, MD); T4 polynucleotide kinase was obtained from Stratagene (La Jolla, CA). Recombinant UDG and poly(dI-dC) were obtained from Roche Diagnostics. [␥-32 P]ATP (3000 Ci/mmol) was obtained from PerkinElmer Life Sciences. All other reagents were ACS grade.
Animals-The generation and characterization of mOGG1 Ϫ/Ϫ mice have been described elsewhere (17). The generation of the mNTH1 Ϫ/Ϫ mice has been reported (18). The [mNTH1, mOGG1] Ϫ/Ϫ double-knockout mice were obtained by crossing the single knock-outs to produce mice heterozygous for both mNTH1 and mOGG1. These mice were then bred to produce mice of all possible genotypes: wild type, mNTH1 Ϫ/Ϫ , mOGG1 Ϫ/Ϫ , and [mNTH1, mOGG1] Ϫ/Ϫ used in the work described herein. For the isolation of mitochondria, the animals were killed by cervical dislocation, and the livers were removed immediately and processed. All animal procedures were performed in accordance with the Animals (Scientific Procedures) Act 1986 of the United Kingdom.
Preparation of Mouse Liver Mitochondrial Extracts-Mouse liver mitochondria (MLM) were isolated from fresh livers by using a combination of differential centrifugation and Percoll gradient separation, as described previously (19).
Oligonucleotides-Control and uracil-containing oligonucleotides were obtained from Midland Certified Reagent Co. (Midland, TX). 5-OHdC-and 5-OHdU-containing oligonucleotides were kindly provided by Michelle Ham (Massachusetts Institute of Technology, Cambridge, MA). All oligonucleotides contained the sequence 5Ј-ATAT-ACCGCGGXCGGCCGATCAAGCTTATT-3Ј, where X represents either the unmodified base or the specified base damage in the 30-mer sequence. AP-containing oligonucleotides were prepared by treating the uracil-containing substrate with 1 unit of recombinant uracil-DNA glycosylase for 1 h at 37°C. The oligonucleotide containing a single thymine glycol lesion was prepared by OsO 4 treatment of the oligonucleotide 5Ј-GAACGACAGATGACACGACAGACAAGCA-3Ј, as described previously (13). The substrates were 5Ј-end-labeled using T4 polynucleotide kinase and [␥-32 P]ATP. To separate the unincorporated free [␥-32 P]ATP, the reaction mixtures were spun through a Sephadex G-25 column. Complementary oligonucleotides were annealed in 10 mM Tris-HCl, pH 7.8, 1 mM EDTA, 100 mM KCl by heating the samples at 80°C for 5 min and allowing them to cool slowly to room temperature.
Oligonucleotide Incision Reactions-Incision reactions (20 l) contained 40 mM HEPES, pH 7.6, 5 mM EDTA, 75 mM KCl, 2 mM DTT, 0.5 mM MgCl 2 , 0.5 g poly(dI-dC), and 100 fmol of 32 P-oligonucleotide. The reactions were initiated by the addition of the extracts and incubated for 2 h at 37°C. The reactions were terminated by the addition of 1 l each of 5 mg/ml proteinase K and 10% SDS. After incubation at 55°C for 30 min, an equal amount of formamide loading dye was added. The samples were incubated at 90°C for 2 min and separated on a 20% polyacrylamide,7 M urea gel. The products were visualized with a Phos-phorImager (Amersham Biosciences), and incision percentage was quantified by using the ImageQuant® software.
Sodium Borohydride Trapping of the Incision Intermediates-Reactions were carried out as described above, with the addition of 50 mM NaBH 4 (freshly prepared). To detect the covalent complex, 10 l of the reaction was mixed with 10 l of 2 ϫ SDS-PAGE loading buffer (Invitrogen), boiled for 10 min, and separated on a 12% Tris/glycine-PAGE gel (Invitrogen). The gel was then dried, and the bands were detected in the PhosphorImager.

Normal Repair of Uracil in Extracts from NTH1-deficient
Mice-To investigate the role of the mNTH1 protein in mitochondrial BER, we prepared liver mitochondrial extracts from wild-type and mNTH1 Ϫ/Ϫ mice. Extracts from mOGG1 Ϫ/Ϫ mice and double-knock-out animals were also analyzed to determine whether the mOGG1 protein could replace mNTH1 in the repair of certain types of oxidative DNA base lesions. We tested for the presence of contaminating nuclear proteins by Western blot using monoclonal antibodies against the highly abundant nuclear matrix protein, lamin B2, and the mitochondrial protein, cytochrome oxidase subunit IV (COX IV), as a marker of mitochondrial content (Fig. 1). All of our mitochondrial extracts (lanes 1-4) contained negligible nuclear contamination (Ͻ5% of the COX IV signal). A lane with nuclear extracts (lane 5) was FIG. 1. Western blot analysis of mitochondrial and nuclear extracts. 10 g of mitochondrial (lanes 1-4) or nuclear (lane 5) extracts were subjected to electrophoresis on a 12% polyacrylamide-SDS gel. The proteins were transferred to a polyvinylidene difluoride membrane and blotted as described, using monoclonal anti-COX IV and monoclonal anti-lamin B2 antibodies. Lane 1, wild type; lane 2, included as a positive control of the lamin B2 antibody.
To demonstrate that the extracts obtained from the knockout animals were proficient in general DNA repair activities, we measured the incision of a uracil-containing substrate. Fig.  2, A and B, shows that all 4 mitochondrial extracts incise uracil-containing DNA with similar efficiency, as demonstrated by the appearance of the cleaved product band (panel A, lanes [3][4][5][6]. Uracil-containing oligonucleotides incubated in the absence of mitochondrial proteins (lane 1) as well as a control oligonucleotide (with a cytosine at position 11) incubated with wild-type MLM (lane 2) did not show any incision. Because mtUDG is a monofunctional glycosylase and the repair of uracil requires the coordinated action of an AP-endonuclease, we next measured AP-endonuclease activity in these extracts. To generate an AP-containing substrate, the uracil-containing oligonucleotide was incubated with recombinant UDG. The abasic site generated by this reaction was stable under our incubation conditions (Fig. 2C, lane 1) but was efficiently cleaved when incubated with APE1 (Fig. 2C, lane 2). All mitochondrial extracts efficiently incised the AP-containing substrate (lanes 3- 6). No decrease in the incision of an AP site containing oligonucleotide was observed in the mitochondrial extracts deficient in either mNTH1 or mOGG1. In contrast, both the single knock-outs and the double-knock-out extracts contained higher AP-endonuclease activity than the wild type (Fig. 2D), suggesting a possible compensatory mechanism, because mOGG1 and mNTH1 are bifunctional glycosylases with associated AP lyase activities. Together, these results suggest that the absence of mNTH1 or mOGG1 did not interfere with BER of lesions that are recognized by another glycosylase, such as UDG.
Thymine Glycol Incision in NTH1-deficient Mouse Extracts-To determine which glycosylases were involved in Tg removal from the mitochondrial DNA, we measured the incision of an oligonucleotide containing a single Tg:A lesion using the DNA glycosylase-deficient MLM extracts. Although wildtype and mOGG1 Ϫ/Ϫ extracts efficiently incised the Tg-containing substrate, all of the incision activity was abolished in the mNTH1 Ϫ/Ϫ as well as in the double-knock-out mouse (Fig. 3,  lines 2-5). This directly implicates mNTH1 as the main glycosylase/AP lyase for this base damage. Using this substrate, we did not detect any significant residual Tg incision activity in the mNTH1 Ϫ/Ϫ extracts. Under the same incubation conditions, no cleavage of the oligonucleotide substrate was observed in the absence of mitochondrial proteins (Fig. 3, lane 1).
Recently authors reported that the new Tg incision activity observed in whole cell extracts from their NTH1 Ϫ/Ϫ mouse strain was more active against Tg:G mispairs than Tg:A, whereas Takao et al. (16) described a novel monofunctional glycosylase in mitochondrial extracts, which they named TGG1. Therefore, because we did not detect any incision of a Tg:A-containing oligonucleotide by our mNTH1 Ϫ/Ϫ and [mNTH1, mOGG1] Ϫ/Ϫ mitochondrial extracts, we tested their ability to incise at a Tg:G mispair. Fig. 3 (lanes 6 -9) shows that although wild-type and mOGG1 Ϫ/Ϫ extracts had significant activity against a doublestranded oligonucleotide containing the Tg:G lesion, no significant incision of the Tg:G substrate was detected when either the mNTH1 Ϫ/Ϫ or [mNTH1, mOGG1] Ϫ/Ϫ mouse liver mitochondrial extracts were used. Thus, although a Tg:G mispair is clearly a substrate for NTH1, we find no evidence in mitochondrial extracts for a novel activity similar to that described by Ocampo et al. (20) in whole cell extracts.
To investigate whether the absence of Tg incision activity was caused by the reaction conditions, we incubated the Tg:Acontaining oligonucleotide with mNTH1 Ϫ/Ϫ mitochondrial extracts in the presence of varying KCl concentrations (25-150 mM). Here again, we did not detect cleavage at the Tg site by the mNTH1 Ϫ/Ϫ mitochondrial extracts under any of the conditions used (data not shown), in agreement with our earlier result of no backup incision activity.
To confirm that the mNTH1-deficient extracts lacked another glycosylase/AP lyase activity for Tg-containing DNA, we carried out a sodium borohydride-trapping assay, in which the intermediate Schiff base is reduced, thus covalently linking the enzyme-substrate complex. After incubation of the 32 P-Tg:Acontaining oligonucleotide with the mitochondrial extracts in the presence of sodium borohydride, the samples were analyzed in an SDS-PAGE gel to visualize the trapped complex. The SDS-PAGE gel (Fig. 4) showed that the formation of the oligonucleotide-enzyme complex took place only in the lanes incubated with wild-type and OGG1 Ϫ/Ϫ extracts (lanes 2 and 5). In agreement with the observed lack of incision activity toward Tg, no trapped complex was observed in the samples incubated with mNTH1 Ϫ/Ϫ or [mNTH1, mOGG1] Ϫ/Ϫ extracts (lanes 3 and 4, respectively). As a positive control, the substrate was incubated with bacterial endonuclease III (lane 6). No trapped complex was detected in wild-type samples incubated with a control oligonucleotide (lane 1).
Absence of a Monofunctional Glycosylase for Tg in NTHdeficient Mitochondrial Extracts-To investigate the possibility that Tg could be released without the concomitant cleavage of the phosphodiester backbone, we incubated the Tg:A-containing oligonucleotide with mitochondrial extracts and then added 25 ng of endonuclease IV to provide an exogenous source of AP lyase activity. The results are presented in Fig. 5. Panel A shows the results obtained in the absence of endo IV incubation, whereas panel B presents the results after incubation with 25 ng of endo IV. No differences between the Tg incision activities were observed in the presence or absence of the added AP lyase activity, indicating that no abasic sites had been left unresolved during the incubation with the mitochondrial extracts alone (Fig. 5, compare lanes 5 and 6 in panel A to panel  B). As a control, an AP-containing oligonucleotide (lane 8, both gels) was efficiently cleaved upon incubation with endo IV, demonstrating that the enzyme was active. A control oligonucleotide incubated with wild-type mitochondrial extracts was not cleaved either in the absence or in the presence of endo IV (Fig. 5, lanes 3, both panels). Control or Tg-containing substrates incubated in the absence of mitochondrial extracts were also not cleaved under the same incubation conditions (lanes 1  and 2).

NTH1-deficient Mouse Extracts Incise Other Oxidized
Pyrimidines-To evaluate the role of NTH1 in the removal of other oxidative damaged bases in mitochondrial DNA, we measured the incision activity toward 5-OHdC and 5-OHdU in mitochondrial extracts from wild-type and knock-out mice. This is shown in Fig. 6: panels A and B show the results obtained for 5-OHdC, whereas panels C and D present the results for 5-OHdU. Extracts from the wild-type and mOGG1 Ϫ/Ϫ animals cleaved the two oligonucleotides very efficiently. In both cases, extracts from mOGG1 Ϫ/Ϫ mice showed increased incision activity compared with the wild type, suggesting a possible up-regulation mechanism. In contrast to what we observed with the Tgcontaining oligonucleotide, a residual incision activity toward 5-OHdC and 5-OHdU could easily be detected in extracts from the mNTH1 Ϫ/Ϫ and the [mNTH1, mOGG1] Ϫ/Ϫ animals. Although lower 5-OHdU cleavage was observed in [mNTH1, mOGG1] Ϫ/Ϫ compared with the mNTH1-deficient extract, the differences were not statistically significant. This residual activity was not caused by non-enzymatic cleavage of the oligonucleotides, because incubation of the substrates under the same condition, but in the absence of mitochondrial extracts, did not result in any incision products (Fig. 6, lane 1, panels A  and C). In addition, control substrates containing an unmodified base were also not incised by wild-type extracts (lane 2, panels A and C), demonstrating that these incision activities were damage-specific. Thus, our results suggest that mouse mitochondria contain at least one other glycosylase/AP lyase that recognizes 5-OHdC and 5-OHdU. The incision observed in the double-knock-out extracts suggests that this activity cannot be attributed solely to mOGG1, and thus it may constitute a new enzymatic activity. DISCUSSION Mitochondria are the principal source of reactive oxygen species in the cell, and the mitochondrial genome is known to contain higher levels of oxidative DNA base damage than the nuclear genome (21). We have previously observed that the OGG1 glycosylase plays a more important role in mitochondria than in nuclear DNA, because the accumulation of 8-oxoG in FIG. 3. Incision of Tg-containing oligonucleotides by mouse liver mitochondrial extracts. A 30-mer oligonucleotide containing a single Tg lesion opposite either an adenine (lanes 1-5) or a guanine (lanes 6 -9) was incubated with 25 g of mitochondrial extracts for 2 h. The reactions were terminated, and the products were separated as described previously. A typical autoradiogram is presented. the mitochondrial DNA from OGG1 Ϫ/Ϫ was much more dramatic than in nuclear DNA when both were compared with wild type (22). Therefore, it is of great interest to characterize the function in mitochondria of another major glycosylase, NTH1, and we have done this by studying the incision of oligonucleotides containing DNA lesions that are known substrates for NTH1. The main oxidative DNA base lesions known to be substrates for NTH1 in vivo are Tg and other oxidized pyrimidines such as 5-hydroxycytosine and 5-hydroxyuracil. It is also important to characterize the BER activities in mitochondria from the double-knock-out animals that are deficient in both mOGG1 and mNTH1. These enzymes are the two main glycosylases for the removal of a wide spectrum of oxidative damages from DNA.
Although mitochondrial extracts from all four mouse strains tested incised a uracil or abasic site containing oligonucleotides with similar efficiency (Fig. 2), we failed to detect any incision activity against thymine glycol, when paired with either adenine or guanine, in extracts from mice deficient in mNTH1 (Fig. 3). These results are in apparent contradiction with the study by Takao et al. (16), who described a monofunctional TGG1 in partially purified liver mitochondrial extracts from their mNTH1 knock-out mouse. However, these authors only observed significant Tg incision activity after at least one step of protein purification. This could indicate that, even with the induced activity observed in their mNTH1 Ϫ/Ϫ fractions, TGG1 is still a very minor activity. Alternatively, there may be a competing or inhibiting protein, or other cofactors, present in the mitochondrial extracts that are lost during the purification procedure.
The lack of a backup Tg glycosylase in our mNTH1 Ϫ/Ϫ mitochondrial extracts is further supported by the observation that sodium borohydride-trapped complexes of Tg-containing oligonucleotides were only detected in extracts containing NTH1 (Fig. 4, lanes 2 and 5), suggesting that mNTH1 Ϫ/Ϫ mitochondrial extracts lack a DNA glycosylase/AP lyase that could recognize that base damage. The trapped complex in wild-type and mOGG1 Ϫ/Ϫ extracts was of slightly higher molecular mass in an SDS-PAGE than the endo III control, which was to be expected because mNTH1 is Ϸ37 kDa and endo III is Ϸ23 kDa.
Another important consideration that might help to explain the differences between our results and those reported by Takao et al. (16) concerns the method used to generate the thymine glycol-containing oligonucleotides used in the different studies. Osmium tetroxide used in this study results in the production of both the 5S,6R and 5R,6S diastereoisomers of cis-thymine glycol, and this resembles the situation after ␥ radiation (23,24). However, KMnO 4 oxidation, as used by Takao et al. (16), results predominantly in the formation of the 5R,6S isomer. Previous work by Asagoshi et al. (25) showed that mNTH1 incised oligomers containing either of the diastereoisomers with similar efficiencies, in good agreement with its known broad substrate specificity. However, it is possible that the lack of activity against our Tg oligonucleotide by extracts lacking mNTH1 may be a function of the low abundance of any novel Tg glycosylase, compounded by its preference for the 5R,6S cis isomer of Tg, which is present in lower amounts in our substrate. This could also help to explain our observation that oligonucleotides containing 5-OHdC and 5-OHdU are incised by these extracts, provided that the enzyme has broad substrate specificity.  1 and 3) was incubated with mitochondrial extracts, as described previously. At the end of the 2-h incubation, the samples were divided into two aliquots and either mock-incubated (panel A) or incubated with 25 ng of endonuclease IV (panel B) for 30 min at 37°C. The products were then resolved on a 20% polyacrylamide, 7 M urea gel, as described previously. As positive control for endo IV AP lyase activity, we used a 26-mer oligonucleotide containing a tetrahydrofuran residue at position 16 (lane 5).
The same extracts that lack Tg incision were not devoid of activity against all oxidized pyrimidines, as we did observe incision of oligonucleotides containing 5-OHdC and 5-OHdU in both the single mNTH1 and the double-knock-outs (Fig. 6). The slightly lower incision activity in the extracts from [mNTH1, mOGG1] Ϫ/Ϫ animals in comparison to the mNTH1 Ϫ/Ϫ suggests that, in the absence of mNTH1, mOGG1 may participate in the repair of oxidized pyrimidines. However, mOGG1 does not account for all the residual incision, suggesting the existence of another glycosylase that may recognize those lesions.
Recently, a number of groups have reported the identification and partial characterization of a novel human glycosylase with homology to the bacterial Fpg/Nei class of proteins (26 -28). This enzyme, now termed NEIL1, has broad substrate specificity and is characterized by its ␤/␦-elimination mechanism of cleaving the phosphodiester backbone. However, this latter trait discounts the possibility that NEIL1 is responsible for the residual activity against 5-OHdC and 5-OHdU observed in our mitochondrial extracts. Fig. 6 shows that the product in the mNTH1 Ϫ/Ϫ and [mNTH1, mOGG1] Ϫ/Ϫ lanes (panels A and C, lanes 5 and 4, respectively) migrates to the same position as that of the wild-type and mOGG1 Ϫ/Ϫ -derived extracts, indicating that the product is a result of ␤-elimination; mNTH1 does not carry out the ␦-elimination reaction, which would result in a faster migrating product. In support of this, Morland et al. (28) found that the fluorescence from enhanced green fluorescent protein-hFPG1 (hNEIL1) fusion was associated only with the nuclei of transiently transfected HeLa S3 cells.
The identity and characterization of the protein(s) responsible for the residual incision activity in our mNTH1-deficient mitochondrial extracts is being actively pursued in our laboratories. Indeed, experiments with whole cell sonicate have revealed the presence of a novel glycosylase with a broad substrate specificity that acts through a ␤-elimination reaction mechanism. 2 It will be interesting to determine whether this protein is identical to the backup enzyme activity from NTH1 Ϫ/Ϫ thymic sonicates described by Ocampo et al. (20) and whether this protein is localized to the mitochondrion.