Repair of Oxidative Damage to Nuclear and Mitochondrial DNA in Mammalian Cells*

Reactive oxygen species (ROS) are generated in cells as a byproduct of cellular metabolism. ROS react with proteins, lipids, and DNA. DNA base modifications, abasic sites, deoxyribose damage, and single and double strand breaks are all induced following various forms of oxidative stress. This review will focus on DNA repair of oxidative lesions by base excision repair (BER) and nucleotide excision repair (NER). We will focus on the mammalian BER enzymes that have recently been cloned and characterized. Mitochondrial DNA repair mechanisms for oxidative damage will also be discussed. Although sugar damage and double strand breaks are critical lesions induced by ionizing radiation and bleomycin, repair of these lesions will not be discussed here (see Refs. 1–3 for recent reviews).

Many experimental methods have been used to expose cells to oxidative damage, all attempting to mimic endogenous processes (4,6). Some studies have used hydrogen peroxide, which generates a large spectrum of lesions. Ionizing radiation also generates a wide spectrum of lesions including base damage and single and double strand breaks in DNA. Methylene blue plus visible light exposure primarily generates singlet oxygen damage, and osmium tetroxide generates primarily thymine glycols. For more discussion of this see Refs. 4 and 6. It is important to distinguish between the different types of oxidative stresses when evaluating experimental results.
Technical differences in the methods used for DNA isolation may well result in differences in the analysis of the DNA adducts. A recent review compared the various methods used to detect oxidative damage in DNA (13). One of the conclusions that emerged from the comparison was that there is a great need for methods to be more standardized and thus to provide more consistent results between different laboratories when comparing different but related techniques.
One aspect that is common to many methods used to detect oxidative damage is that the DNA modifications are measured as averages in the total cellular DNA. This is of limited value since advances in recent years have shown that DNA damage processing and the biological consequences of DNA lesions vary considerably depending upon where a lesion is situated in the genome. For example, UV-induced photoproducts are processed differently whether situated in an active gene or in a non-transcribed region, and this may also be the case for oxidative lesions.
The gene-specific repair assay (GSR) employs various DNA repair enzymes to detect specific lesions, and this assay has provided new insights about the heterogeneity of DNA repair in the nucleus (14) and more recently about the repair mechanisms for mitochondrial DNA (see below). For example, endonuclease III (endo III) can detect oxidized pyrimidines, and the Fapy DNA glycosylase (Fpg protein) can detect oxidized purines. Endo III-sensitive sites have been assayed in the general genome (15), and more recently, Fpg protein has been used to detect lesions in specific genes (16,17).

Base Excision Repair of Oxidative Damage
BER is initiated by DNA glycosylases, a class of enzymes that recognize a specific set of modified bases such as 8-oxoG or thymine glycol (TG). Glycosylases cleave the N-glycosylic bond between the modified base and the sugar. There are two classifications of glycosylases: simple glycosylases that only cleave the N-glycosylic bond and glycosylase/AP lyase enzymes, which cleave the N-glycosylic bond and the DNA-phosphate backbone. Following the glycosylase step, AP endonucleases are required to remove the 3Ј-deoxyribose moiety and generate a 3Ј-hydroxyl group, which can be extended by a DNA polymerase. The process is completed by a DNA ligase rejoining the free DNA ends (for reviews see Refs. 18 and 19).
Recently, the human and the mouse 8-oxoguanine glycosylase/AP lyase (human OGG1 or mouse Ogg1) genes have been cloned by their homology to yeast ogg1 (23)(24)(25). Human OGG1 gene was localized to the short arm of chromosome 3, 3p26.2 (23,25). Expression of the human gene in E. coli lacking mutM and mutY suppressed the spontaneous mutator phenotype of these cells (24,25). Human OGG1 (also called MutM homolog) was shown to cleave the DNA by a ␤-elimination mechanism preferentially at 8-oxoG:C base pairs (23,24). Several conserved domains have been identified in the yeast, mouse, and human genes including the a helix-hairpin-helix (HhH) and Gly/Pro-rich-Asp motif (GPD motif) (22,23,25). In addition, Arai et al. (25) reported that the yeast Ogg1 and human OGG1 contained a putative C 2 H 2 zinc finger-like motif, although in the yeast sequence one of the histidines was an arginine. Alignment of the ogg1 genes with other DNA repair glycosylases suggests that these enzymes may represent a DNA repair superfamily (18,22,23,25).
Nash et al. (22) identified another yeast protein, which preferentially interacted with the substrate 8-oxoG:G; they called the activity Ogg2. This same substrate preference was observed for the yeast Fapy DNA glycosylase previously isolated by de Oliveira et al. (26). Whether these two proteins are the same or not remains to be determined. In human extracts, an 8-oxoG endonuclease was identified from human polymorphonuclear neutrophils, which cleaved 8-oxoG but not the ring-opened guanine adduct, FapyG (27). One distinguishing feature of this enzyme was that it was magnesium-dependent. Another study identified two repair activities, an 8-oxoG glycosylase and an 8-oxoG endonuclease, from HeLa cell nuclear extracts (28). The 8-oxoG base pairing preferences for these enzymes were similar to that of yeast Ogg1. Further experiments are required to determine whether these proteins are human OGG1 or novel enzymes.
In E. coli, the MutY protein is an adenine DNA glycosylase that removes adenine when base paired with 8-oxoG. Using purified DNA polymerases, it has been demonstrated that the replicative polymerases incorporate adenine opposite 8-oxoG (29). A human MutY activity has been purified from calf thymus cells (30). The protein removes adenine mispairs including A:G, A/8-oxoG, and A:C. The glycosylase co-purified with a AP nicking activity, which was inhibited by neutralizing MutY antibodies. Recently, the gene for a human MutY homolog was cloned (31).
In cells, the deoxyribonucleotide pools are also subjected to oxidative damage. dGTP can be converted to 8-oxodGTP and incorporated into nascent DNA strands opposite adenine. To avoid such damage, cells possess an 8-oxodGTPase, which hydrolyzes the triphosphate to the monophosphate so that it can no longer be incorporated into DNA. In bacteria, the MutT gene product is the 8-oxodGTPase enzyme. A human MutT homolog has been cloned from a human cell line (32).
Repair of Thymine Glycols and Ring-saturated Pyrimidines-Another major adduct generated by oxidative stress is TGs (cf. Fig.  1). Unlike 8-oxoG, TGs block DNA and RNA polymerases and are thought to be lethal (33). Endo III is one of the bacterial enzymes responsible for recognition and removal of TGs; however, cells lacking endo III are not hypersensitive to H 2 O 2 or x-rays (34). Subsequently it was shown that bacteria contain another endonuclease that recognizes TG, endonuclease VIII (35). In addition, the Uvr ABC complex was shown to recognize TGs in vitro (36). Recently, a yeast homolog of endo III has been cloned, NTG1 (endonuclease three-like glycosylase 1) (37). NTG1 has a unique substrate specificity; not only does it remove oxidized purines, but it also recognizes and incises the ring-opened guanine adduct, FapyG. However, it does not incise the 8-oxodG adduct (37). Deletion of yeast NTG1 renders the cells sensitive to H 2 O 2 and menadione (37).
A mammalian TG glycosylase activity has been purified from extracts of calf thymus and bovine cells (38). More recently another gene for the human endonuclease III homolog was cloned (39). Like the bacterial enzyme endo III, the human enzyme acts on urea and TG residues. It also contains an iron/sulfur cluster and a helixhairpin-helix motif.
Repair of Abasic Sites and Sugar Damage-Apurinic or apyrimidinic sites (AP sites) are collectively called abasic sites. These are generated as a consequence of normal spontaneous hydrolysis of the N-glycosylic bond, by the action of DNA glycosylases or by oxidative damage to the sugar residues in DNA. AP endonucleases are enzymes that function to generate suitable DNA ends for DNA resynthesis or ligation (for reviews see Refs. 40 -42). One major AP endonuclease has been purified from human cells called HAP1 (also called APE, APEX, and Ref-1) (reviewed in Ref. 42). The enzyme cleaves 5Ј to the AP site leaving a 5Ј-deoxyribose moiety and a 3Ј-hydroxyl group on the DNA ends. In addition to the AP endonuclease activity, the enzyme possesses several other activities including a 3Ј-phosphatase, 3Ј-phosphodiesterase, and a very weak exonuclease activity (42)(43)(44).
AP endonucleases are the major pathway whereby AP sites in DNA are repaired; however, nucleotide excision may also participate (42). In E. coli, oxidized abasic sites are poorly recognized by various repair endonucleases (45). It would be interesting to know whether the mammalian HAP1 also displays this reduced recognition and incision of oxidized AP sites. If so, then what other pathways participate in the repair of oxidized AP sites? Do NER proteins recognize and repair oxidized AP sites?
It is already apparent that repair of oxidative damage by a BER mechanism in mammalian cells is more complex than in bacteria. Due to the high levels of endogenous oxidative damage, mammalian cells may have had to evolve multiple repair mechanisms to survive the daily insults. Therefore, it may be difficult to define what significance a particular protein has on the repair of specific types of DNA damage, and generation of single-gene knock-out mice may not be informative. Multiple single-gene knock-out mice may have to be crossbred before a phenotype is observed.

Nucleotide Excision Repair of Oxidative Damage
In bacteria and mammalian cells, the repair of oxidative damage is mediated by both BER and NER mechanisms (46,47). NER employs a complex set of proteins that remove damage from DNA (recently reviewed in Refs. 48 and 49). There are two components of NER, a global repair element and a TCR mechanism. Transcriptionally active genes are repaired at a faster rate than genes in non-transcriptionally active domains of the genome and with a strand bias favoring the transcribed DNA strand (50). The coupling between DNA repair and transcription is mediated via the basal transcription factor, TFIIH, which contains at least two DNA repair genes (50). Three genetic disorders have been identified that have defective NER, xeroderma pigmentosum (XP), Cockayne's syndrome (CS), and trichothiodystrophy.
Patients with XP are characterized by their acute sun sensitivity and the development of carcinomas at an early age. Seven complementation groups of XP have been identified, and several XP and XP/CS groups develop neurological abnormalities. Studies were performed to determine whether the oxidative damage repair capacity of the XP cells correlated with their neurological phenotypes (51,52). A host cell reactivation assay was used to measure the capability of two XP-A cell lines to repair viral DNA, which had been damaged by singlet oxygen (52). The XP-A cells showed no difference from normal cells. In another study, a chloramphenicol acetyltransferase reactivation assay was employed to evaluate the survival of a methylene blue plus light-treated DNA in XP-A, XP-C, XP-D, and XP-E cell lines. Using seven normal cell lines, the investigators first defined a normal response range and then determined whether the XP cell lines fell within this normal range. Of the cell lines tested, only the XP-C (3 out of 4) cell lines showed reduced chloramphenicol acetyltransferase reactivation. Whereas UV repair is compromised in all XP complementation groups to varying degrees, the repair of singlet oxygen damage is not. XP-A cells show the greatest sensitivity to UV damage, and most patients with XP-A have demonstrable neurological abnormalities. However, XP-A cells appear to be normal in their repair of singlet oxygen-mediated damage. This suggests that there is no direct relationship between the accumulation of singlet oxygen damage in DNA and the development of neurological defects.
It is possible that an oxidative DNA lesion other than those generated by singlet oxygen may be critical in the development of neurological defects in XP cells. This is supported by a study which investigated whether repair of lesions other than the major oxidative adducts was defective in NER-deficient XP cell lines. Satoh et al. (46) treated DNA plasmids with either ␥-irradiation or hydrogen peroxide plus copper and then removed the major adducts by treating the plasmids with the Fpg protein and endo III (46). They then assayed whether XP-A, XP-B, XP-C, XP-D, and XP-G cell extracts were able to perform DNA repair synthesis on such substrates. All XP cell lines showed reduced DNA repair synthesis as compared with normal cell extracts. The specific lesion in the plasmid DNA, which was dependent on NER for repair, was proposed to be purine dimers. The authors concluded that although repair of the major oxidative lesions is not impaired in XP cell extracts, there could be some endogenous oxidative lesions that are formed at low levels, require NER for repair, that accumulate in XP patients, and lead to neurological defects. CS patients are characterized by dwarfism, premature aging, sensitivity to sunlight, and mental retardation (53). Patients with features of both XP and CS patients have been identified for XP groups, XP-B, XP-D, and XP-G (53). CS cells are deficient in the preferential repair of active genes and TCR of UV-induced pyrimidine dimers (54,55). There are two complementation groups of CS, CS-A and CS-B. The genes CSA and CSB are cloned, but how these genes facilitate TCR is still not understood (for a recent review see Ref. 53). Recently, the CSB protein has been found to play a direct role in transcription, both in vitro (56) and in vivo (57).
CSA, CSB, and XP-A cells were tested for their sensitivity to ionizing radiation and for their ability to perform TCR after such damage (58). Both CS-A and CS-B cells showed radiosensitivity to ␥-rays whereas XP-A and normal cells did not. Using an antibody that recognizes repair patches, gene-specific repair was measured, and it was observed that CS-A and CS-B cells were defective in preferential repair of 10 gray of ␥-rays. Normal and XP-A cells showed normal TCR of ␥-rays. In a subsequent paper, TGs were shown to be repaired by a transcription-coupled repair mechanism, and this repair was defective in CS-B and XP-G/CS cells (59). The TCR of thymine glycols in XP-A, XP-F, and XP-G cells was shown to be normal, demonstrating that TCR of this oxidative lesion can continue in the absence of NER. Thus, there appears to be a NERindependent transcription-coupled repair pathway that utilizes CS-A, CS-B, and XP-G. The exact roles of the CS and XP-G proteins are under investigation in a number of laboratories.
Whereas we have learned much about the heterogeneity, fine structure, and transcription coupling of repair after UV repair, there is very little information about such events after oxidative damage. The repair of UV-induced pyrimidine dimers is known to begin at the nuclear matrix (60), and it would be interesting to understand the nuclear organization of the repair of oxidative DNA damage. It may be useful to focus on individual lesions rather than on the assessment of total damage since different lesions are repaired via different pathways (61).

DNA Damage Processing in Mitochondrial DNA
Because most ROS are generated by the oxidative phosphorylation processes that occur in mitochondria, it is of great interest to understand the oxidative DNA damage-processing mechanisms in these organelles. It has been observed that mtDNA contains a higher steady-state amount of oxidative DNA damage than nuclear DNA (62). Because mtDNA is subjected to relatively high amount of oxidative damage, it seems that mitochondria would need efficient DNA repair mechanisms to remove oxidative damage from its DNA.
One of the controversies in the study of oxidative DNA damage concerns the amount of 8-oxodG present in mtDNA. Although there appears to be a consensus about an increase of damage with age, the amounts of oxidative base modifications measured by various methods (liquid chromatography/mass spectroscopy, high pressure liquid chromatography, and enzymatic) do not agree with one another (13). There is a need to make concerted efforts to measure oxidative lesions and their repair under identical conditions using different methods to assess the same changes. There have been reports of 10-fold higher steady-state levels of 8-oxodG in mtDNA than in nuclear DNA (62). This finding has become one of the cornerstones of the mitochondrial theory of aging, but other observations suggest that it may not be true for cells in culture (63). There are many possibilities for artefactual formation of 8-oxodG in mtDNA during the purification (13), and this needs thorough investigation. Comparative studies would determine if variation were dependent on the type of method used. The early finding of the absence of repair of UV-induced pyrimidine dimers in mtDNA led to the general notion that there was no DNA repair mechanisms in mitochondria (64). Although it has been confirmed that UV-induced pyrimidine dimers are not repaired in mitochondria, recent studies clearly show that mitochondria repair their DNA.
Several observations support that mitochondria are capable of oxidative DNA damage repair. The GSR assay (65) has been modified to detect repair of oxidative lesions (66). With this approach, it is not necessary to isolate mtDNA. Bacterial repair enzymes that recognize and cleave the DNA at specific lesions are used. Oxidative lesions can be detected in the entire mitochondrial genome or in parts of it and be compared with the lesions present in the nuclear DNA from the same biological sample. Also strand bias, or TCR, can be assayed with this approach.
Using the GSR assay and a variety of damaging agents, repair of strand breaks and alkali-sensitive sites has been demonstrated in rodent and human mitochondrial DNA (66,67). In the latter report (66), the investigators used alkali in combination with either endo III or Fpg, and the initial damage frequencies for alkali plus enzyme were not significantly different than alkali alone, suggesting that they were primarily measuring repair of AP sites or strand breaks. From these studies it is apparent that AP sites are efficiently repaired from mtDNA. Consistent with this observation is the fact that an AP endonuclease has been partially purified from mitochondria (68).
Oxidative damage, as detected by the Fpg protein, is repaired in mtDNA from rat cells (16) and Chinese hamster ovary cells (17). In the study by Taffe et al. (17), acridine orange plus light was used as a method to generate oxidative damage, and Fpg protein was used in the GSR assay to assess repair of Fpg-sensitive sites. The acridine orange plus light-induced DNA damage was repaired from both mitochondrial and nuclear DNA sequences. Approximately 65% of the lesions were repaired within 4 h, and the repair in the mtDNA was as fast or faster than in the nuclear dihydrofolate reductase gene. The efficient repair of the Fpg-sensitive sites suggested that mitochondria contain a base excision repair protein, such as a Fpg homolog. Recently, in this laboratory, an 8-oxodG incising activity has been partially purified and characterized from rat liver mitochondria (69).
A polymerase extension assay has been utilized to evaluate H 2 O 2 damage induction and removal (70). Both nuclear and mitochondrial sequences were evaluated. This assay is not measuring a specific lesion but rather any damage that blocks the progression of the polymerase. Repair of H 2 O 2 -mediated damage in mitochondria was as efficient as in nuclear DNA when a brief H 2 O 2 exposure was used, whereas at higher levels of exposure mitochondrial DNA repair was not observed within 3.5 h of post-treatment.
Whereas in vitro DNA repair assays have proven very useful in the exploration of the NER mechanisms, no such assays have been available for the mitochondrial DNA until very recently. In one study, the DNA repair incorporation assay was used to demonstrate that plasmid DNA treated with hydrogen peroxide was repaired using Xenopus mitochondria protein extracts (71). More in vitro studies are needed.
Interestingly, complex lesions such as cisplatin interstrand cross-links, which are thought to be repaired via a recombination repair pathway, are removed from hamster mtDNA (72). This would suggest that mitochondria possess recombinational activities. Recently there has been support for this hypothesis; an in vitro assay has demonstrated recombination between plasmids using human mitochondrial protein extracts (73). Another complex lesion that is removed from mitochondrial DNA is 4-nitroquinoline DNA damage (74). 4-Nitroquinoline damage is generally thought to be removed via a NER pathway. However, NER as it exists in the nucleus does not exist in mitochondria. Whether any NER proteins participate in mitochondrial repair remains to be explored.
By several different methods, it is apparent that mitochondria possess the ability to remove oxidative DNA damage. Whether mitochondria utilize the same proteins as the nucleus or different ones will await their biochemical purification and cloning.

Conclusions and Perspective
DNA accumulates oxidative damage as a consequence of its inherent instability and constant insult by reactive oxygen species generated by endogenous and exogenous sources. The maintenance of the DNA genomes, both nuclear and mitochondrial, is dependent upon proficient repair mechanisms that remove oxidative damage. The repair of oxidative damage is well defined in bacteria, whereas in higher eukaryotes the genes and proteins responsible for repair have not been extensively characterized.
Contrary to the existing notion, mitochondria possess DNA repair mechanisms. Indeed, some oxidative damage, such as strand breaks and 8-oxoG, is repaired in mtDNA. Exploration of mitochondrial DNA damage processing is going to be an important area of research in the next years.
New insight into the repair mechanisms of oxidative DNA damage is emerging very quickly. It clearly involves a contribution of both BER and NER and other pathways. There may be well over 100 types of oxidative base modifications in mammalian DNA, and the accumulation of a single type or of multiple types of oxidative lesions may have deleterious consequences for the cells, such as disease or neurological deficits.