Disparity between DNA Base Excision Repair in Yeast and Mammals: Translational Implications¹

Effective treatment of cancer requires the continued development of novel chemotherapeutic agents to kill tumor cells. Currently, many of the Food and Drug Administration approved cancer-treating drugs act by forming DNA adducts, which lead to cell death. A component of cancer research has been devoted to understanding DNA repair pathways in hopes of defining the factors that confer resistance to anticancer drugs and developing strategies for modulating repair capacity as a means of overcoming resistance or enhancing sensitivity to cancer treatments.

Historically, yeast, particularly Saccharomyces cerevisiae, has been used as a model system to evaluate drug efficacy and selectivity, and to identify new targets for antitumor drugs. The advantage of this model is 2-fold; S. cerevisiae is a simple organism of which the entire genome is sequenced, and its DNA repair proteins have been considered generally well conserved among eukaryotes. One approach to chemotherapeutic drug identification has been to screen antitumor agents in mutant strains of yeast and associate drug sensitivity with a specific defect. For example, cisplatin was selectively toxic to yeast defective in the RAD6/RAD18-controlled cell-cycle pathways, indicating that cisplatin would be an appropriate therapy against tumors defective in postreplication repair (1, 2). Although S. cerevisiae has been an informative model for investigating the molecular biology and biochemistry of DNA repair in eukaryotes, the emergence of data over the past 5 years regarding mammalian DNA repair processes has revealed significant differences between the yeast and human pathways. This divergence challenges the appropriateness of using yeast to screen for chemotherapeutic agents. In this review, we focus specifically on the differences in the BER³ pathway of yeast and mammals, particularly as it relates to assessing DNA damaging-agent resistance.

BER involves the sequential activity of several repair proteins acting in concert to excise the target damage and restore DNA to its unmodified state (Fig. 1). After substrate base removal by a DNA glycosylase, an AP endonuclease initiates repair by incising the DNA backbone 5′ to the resulting abasic damage, leaving a 3′-hydroxyl group and a 5′-dRP moiety flanking the nucleotide gap (3, 4, 5). AP lesions are also common products of spontaneous or chemically induced (e.g., oxidation and alkylation) hydrolytic base release. APE1, the major human AP endonuclease (also known as HAP1, APE, hAPE, HAP, or Ref-1), and APN1, the predominant yeast AP endonuclease, while sharing some overlapping substrate specificity, share no structural similarity and exhibit significant differences in functional capacity (4, 6). APN1 belongs to the Escherichia coli endonuclease IV family, and constitutes ∼97% of the total AP endonuclease/3′-diesterase repair activities of S. cerevisiae (7). The S. cerevisiae counterpart to E. coli exo III (the second AP endonuclease family), named APN2/ETH1, comprises the remaining activity, but unlike Apn1 mutants, Apn2 mutants exhibit normal sensitivity to methyl methanesulfonate and show no increase in spontaneous mutation rates, suggesting a less prominent role in the repair of AP sites (8, 9, 10). In contrast, APE1, the human exo III equivalent, constitutes >95% of the total AP site incision activity (11), and a mammalian endonuclease IV/APN1 homologue has yet to be discovered. A weak AP endonuclease activity has been associated with a second human exo III homologue, APE2, which appears more like the APN2 protein of yeast (12).

Both APE1 and APN1 hydrolyze the phosphodiester backbone 5′ to the AP site, leaving a 3′-hydroxyl group and a 5′-dRP group flanking the nucleotide gap (Fig. 1; Refs. 3, 4, 5). In addition, both proteins exhibit a 3′ phosphodiesterase activity to remove 3′-blocking damages, e.g., 3′ phosphate or 3′ phosphoglycolate, which are associated with ionizing radiation and free-radical attack of DNA (13). However, the 3′ repair activity of APE1 is ∼100-fold lower than its AP endonuclease activity (14), whereas these two activities are roughly equal in APN1 (7). APE1 has been shown recently to possess a 3′ mismatch exonuclease activity and an excision function for certain anticancer/viral nucleoside analogs as well (15, 16).

3′ blocking groups are also formed by complex DNA glycosylases, such as OGG1 and NTH1, which not only excise oxidized base adducts, like 8-oxoG, thymine glycol, 5-hydroxyuracil and 5,6-dihydrothymine (Table 1; Refs. 17, 18), but exhibit an AP lyase activity that cleaves the phosphodiester bond 3′ to an AP site. This 3′ incision, which differs from the 5′ incision of APN1 and APE1, leaves behind a 5′-phosphate and a 3′-unsaturated aldehyde (4-hydroxypentenal) by a β-elimination reaction (19).

In contrast to what might be predicted based on the biochemical repair activities described above, APE1 knockout mouse blastocysts and antisense-expressing human cells are hypersensitive to γ rays and oxidizing agents, whereas Apn1 yeast mutants are at best mildly sensitive to γ rays and oxidizing (H₂O₂) agents (20). These results indicate that there exist unknown, species-specific factors or processes and alternative repair mechanisms that need to be acknowledged when selecting either yeast or mammalian cell lines to assess DNA-damaging agent responsiveness.

Another significant difference between yeast and mammals is the enzymatic processing that follows AP endonuclease incision of the DNA backbone (Fig. 1). In humans, DNA β-pol typically inserts a single nucleotide during repair synthesis and excises the 5′-dRP moiety remaining after APE1 incision (8, 21, 22). Depending on the nature of the 5′-moiety and the DNA substrate, more extensive strand displacement can occur that involves incorporation of up to six nucleotides and requires the processing activity of Fen1 (19, 20, 23). An alternative BER pathway has also been characterized in mammals that invokes the participation of polymerase ε or δ, as well as replication factor C, PCNA, and Fen1 (Fig. 2). To complete either short-patch (single nucleotide) or long-patch BER, DNA ligase I or a complex of DNA ligase III/XRCC1 act to seal the final nick.

In contrast, yeast does not contain a gene equivalent to ligase III, and although DNA β-pol and XRCC1-like proteins exist, a similar role for these factors in yeast BER seems remote (24). Specifically, in yeast, β-pol operates more prominently in DNA end-joining whereas yeast XRCC1 contributes mainly to S phase onset and M phase restraint, i.e., cell cycle control (25, 26). Polymerization in yeast is instead performed predominantly by DNA polymerase ε, and RAD27, the Fen1 homologue, removes the 5′-dRP-containing flaps after strand displacement. CDC9, the ligase I homologue in yeast, ligates the phosphodiester backbone (Fig. 1; Refs. 27, 28, 29). In total, the enzymes required for execution of BER in humans are significantly different from those in yeast, and, thus, pathway coordination will proceed differently as well. For instance, in humans, APE1 communicates with β-pol to facilitate DNA binding and accelerate excision of 5′-dRP residues (30), whereas neither protein plays a prominent role in yeast BER.

In addition to the significant repair activity differences described above, there exist complex regulatory mechanisms superimposed on the nuclease properties of APE1, not seen with APN1. For example, recent work indicates that APE1 autoregulates its own gene expression by binding to its own promoter and inhibiting transcription (31). In addition, phosphorylation and redox modification of APE1 also affect its repair and redox functions (32, 33, 34, 35, 36, 37). In both in vivo and in vitro studies, reactive oxygen species have been found to induce APE1 expression as part of an adaptive response (38, 39, 40, 41, 42, 43, 44). Furthermore, it has been shown in a B lymphocyte model system that after oxidative stress, APE1 translocates from the cytoplasm to the nucleus, followed by an increased level of APE1 protein synthesis (44). APN1 does not appear to be regulated in any of these manners, but this has never really been closely evaluated (7, 45).

Another difference between APE1 and APN1 is that APE1 is a multifunctional protein that impacts a wide variety of cellular functions. For example, APE1 (also known as REF-1 in this capacity) stimulates the sequence-specific DNA binding activity of numerous transcription factors that have physiological functions as diverse as cell cycle control, apoptosis, angiogenesis, cellular growth, cellular differentiation, neuronal excitation, hematopoiesis, and development (20, 36, 46). Consequently, APE1 is a pivotal signaling factor involved in coordinating the cellular adaptation to a wide array of environmental stimuli (20). Beyond the concerns stated above when considering yeast as a model system in screening for novel human therapeutics, there are also notable differences in the cellular response mechanisms of humans and yeast that will not be covered in this review, such as the absence of p53, DNA protein kinase, and so forth. This is again of importance when it is recognized that APE1, acting as a modifier of p53 activity, interacts with systems besides BER; clearly not something observed with APN1 in yeast.

APE1, XRCC1, and β-pol knockout mice are lethal, whereas APN1 null yeast strains are not (19). At least for APE1, the inviability could be because of either its repair or redox role, or both. Additionally, whereas mammalian β-pol knockout cells exhibit extreme hypersensitivity to alkylating agents and increased sensitivity to oxidizing agents (47, 48), yeast harboring a deletion in the Pol IV gene (the yeast β-pol counterpart) display no enhanced DNA damaging-agent sensitivity (27). There are similarly marked differences in the cellular phenotypes of the mammalian and yeast XRCC1 (cut5+/rad4+ in Schizosaccharomyces pombe) mutant lines (26, 49, 50). Thus, recognizing that animal and cell viability are different, the above data make a case that considerable disparity exists in the biological roles of the mammalian and yeast BER gene counterparts.

Another major difference between BER and the BER-related pathways of yeast and humans is in the handling of oxidative base damages. 8-oxoG is a major product of oxidative damage to DNA. The occurrence and cellular level of 8-oxoG is important in terms of mutagenesis and carcinogenesis, because it can pair with adenine and cause GC to TA transversions (51, 52). Resistance to the mutagenic effects of 8-oxoG is conferred by the 8-oxoguanine repair system (53), which, in humans, consists of three surveillance proteins (Table 1). These proteins are OGG1, a DNA glycosylase that removes oxidized guanine from DNA; MYH, which excises adenines misincorporated opposite 8-oxoG in DNA; and MTH, which converts 8-oxo-dGTP in the nucleotide pool to the monophosphate, preventing chromosomal misincorporation of 8-oxo-dGTP (54, 55, 56). However, the 8-oxoG (GO-like) surveillance system of yeast contrasts significantly with this scheme. Most notably, yeast lacks the biochemical activities corresponding to MYH or MTH, and consequently, mispaired adenines and 8-oxo-dGTP are handled by other corrective systems. For example, misincorporated adenine opposite an 8-oxoG is most likely corrected via the mismatch repair pathway, involving the yeast homologues of the mutS and mutL proteins (57, 58).

Although yeast OGG1 mutant cells show a significant increase in mutation frequency, OGG1-deficient mice exhibit only a moderately elevated spontaneous mutation rate in nonproliferative tissues. Ogg −/− mice do not develop malignancies and show no marked pathological changes. These animals are viable; however, their cells, in particular liver cells, exhibit a higher steady-state level of endogenous 8-oxoG in their genomes (59, 60). In light of the fact that it was shown recently that ogg−/− mouse embryonic fibroblasts are not deficient in the repair of 8-oxoG in the actively transcribed DNA strand, it is likely that the increase in the cellular level of 8-oxoG may be localized to the nontranscribed DNA strand or regions of the genome that are not actively transcribed (61). A transcription coupled repair of 8-oxoG (and also thymine glycol) has not been observed in yeast. This is yet another major difference between yeast and mammalian cells.

Notwithstanding the important role of MTH1 in the removal of 8-oxo-dGTP, the increase in hypoxanthine-guanine phosphoribosyltransferase mutations detected in mouse mth1−/− cells is only 2-fold higher than the control MTH1 wild-type cells (62). However, although only a modest defect was observed in the hypoxanthine-guanine phosphoribosyltransferase assay, 18-month-old MTH1 knockout mice exhibited an increase in spontaneous tumorigenesis in liver, lung, and stomach emphasizing the importance of this protein in cancer prevention (62). It is likely that in human cells, there are redundant enzymes present that can compensate for the lack of MTH1 protein in the cells. Redundancy of repair enzymes, in particular those that are involved in BER, appear to be common in mammalian cells. Because yeast does not have a MutT homologue, it is not clear how yeast prevents the misincorporation of 8-oxoG.

In addition to the OGG1 protein, human cells possess an additional activity that removes 8-oxoG, the E. coli endonuclease VIII homologue, NEIL1 (63, 64, 65). In contrast to OGG1, which prefers an 8-oxoG/C pair, NEIL1 exhibits a much higher activity on DNA containing an 8-oxoG/A or 8-oxoG/G pair (64). Furthermore, NEIL1 mRNA levels appear to be cell-cycle regulated and are highest during the S phase. Thus, it was suggested that NEIL1 activity might be coupled with DNA replication to preferentially remove 8-oxoG misincorporated opposite either A or G during replicative synthesis. In contrast, an endonuclease VIII homologue is absent in yeast. Instead, S. cerevisiae NTG1, a repair protein homologous to human NTH1, removes 8-oxoG residues from DNA originating specifically from replicative incorporation of 8-oxo-dGTP (66).

Oxidized pyrimidines such as 5-OHU are recognized by E. coli endo III homologs, i.e., NTH1 in human, and NTG1 and NTG2 in yeast (67, 68, 69, 70, 71). These proteins are highly conserved and are members of a superfamily of repair glycosylases that use a helix-hairpin-helix domain for DNA binding. NTG1 and NTG2, as well as human NTH1, after the excision of damaged pyrimidines, catalyze a β-elimination reaction that cleaves the phosphodiester bond 3′ to the AP lesion, generating a 3′-unsaturated aldehyde.

In addition to recognizing 8-oxoG, human NEIL1 also efficiently recognizes 5-OHU. However, although NEIL1 targets similar oxidized pyrimidines to that of the yeast proteins Ntg1 and Ntg2, their structure and reaction mechanisms are quite different from one another. NEIL1 belongs to a new family of repair enzymes that use a helix-two turns-helix and, in some cases, a zinc finger motif for DNA binding (72). In contrast to the endo III and OGG1 homologues, the AP lyase activity that is associated with NEIL1 catalyzes a β,δ-elimination reaction, generating a one base gap containing 3′ and 5′ phosphoryl ends. In humans, it is believed that polynucleotide kinase or APE1 plays an important role in the removal of the 3′ phosphate before the participation of repair polymerase and ligase in the subsequent BER process (65, 73, 74). The absence of a β,δ-AP lyase activity as a backup for the repair of AP sites in yeast could compromise the ability of yeast to handle excessive amounts of abasic damage.

It is interesting to note that yeast Ntg1 and Ntg2 single or double mutants did not exhibit significant increase in cellular sensitivity toward oxidants such as menadione and hydrogen peroxide, and did not show an appreciable increase in mutant frequency over the wild-type cells (75). Similarly, NTH1 −/− mouse cells also did not show significant increase in sensitivity toward menadione or hydrogen peroxide, and NTH1 knockout mouse showed no significant phenotypic difference from the wild-type mouse (76). Although the phenotype of both yeast and mammalian endo III knockout do not show any appreciable changes, the reason for the absence of a noticeable phenotype is different. In the yeast Ntg1 Ntg2 double-mutant, unrepaired oxidized bases are repaired efficiently by alternative repair pathways such as recombination or translesion synthesis. However, in mammalian cells, the unrepaired oxidized pyrimidines in NTH1 −/− cells are most likely repaired by back-up (or redundant) glycosylases that can also efficiently remove these lesions. As indicated in NTH1 −/− mouse liver cells, there were two additional activities that can help to remove thymine glycol (76), and these could possibly be the newly identified NEIL1 and NEIL2 proteins. As discussed earlier, these repair proteins are absent in yeast cells.

Finally, a major difference in the BER activity between yeast and human lies in their ability in handling uracil, the deamination product of cytosine. Uracil is repaired predominantly by UNG. The enzyme is highly conserved from bacteria to yeast to humans. UNG can remove uracil from both single- and double-stranded DNA, and will remove uracil efficiently from duplex DNA irrespective of the opposing base. However, unlike humans, yeast lacks a second uracil glycosylase, i.e., either the double-stranded DUG or the T/G mismatch glycosylase of humans. DUG recognizes uracil when it is opposite a G, a conformation that exists only when cytosine is deaminated from a normal G/C pair. More importantly, the best substrate for DUG is DNA containing ethenocytosine. Thus, the lack of a DUG activity in yeast raises the question as to how the organism handles ethenocytosine, a highly mutagenic cytosine lesion that is readily formed after lipid peroxidation (77, 78).

In conclusion, the BER pathway is essential for the repair of damaged DNA induced by oxidizing and alkylating agents, which are the majority of chemotherapeutic drugs used currently in the clinic today (79). The importance of this pathway in processing of DNA damage makes its members potential targets for novel chemotherapeutic agents. However, because the BER process and its main players are remarkably divergent from S. cerevisiae to humans, particularly in terms of repairing oxidative DNA damage, it is worth keeping these differences in mind if yeast continues to be used as a model or primary system in the screening for potential new human therapeutics (1, 2). The deviation in the BER process of S. cerevisiae may stem from the adaptation of yeast to a facultative anaerobe, which caused a lack of incentive to acquire a comprehensive system, which is essential in aerobes to battle against the multitude of oxidative base damages. Lastly, the emerging differences in the recombination repair and cell cycle checkpoint responses may extend the concerns raised within this review to nonoxidizing or alkylating agents as well.