Base excision repair capacity in informing healthspan

Summary

We review the range and scope of the endogenous DNA damage problem; the mechanistic aspects of the subpathways of BER; evidence that indicates that defects in BER can affect healthspan and the technical approaches to address whether altered BER capacity associates with human disease susceptibility.

Abstract

Base excision repair (BER) is a frontline defense mechanism for dealing with many common forms of endogenous DNA damage, several of which can drive mutagenic or cell death outcomes. The pathway engages proteins such as glycosylases, abasic endonucleases, polymerases and ligases to remove substrate modifications from DNA and restore the genome back to its original state. Inherited mutations in genes related to BER can give rise to disorders involving cancer, immunodeficiency and neurodegeneration. Studies employing genetically defined heterozygous (haploinsufficient) mouse models indicate that partial reduction in BER capacity can increase vulnerability to both spontaneous and exposure-dependent pathologies. In humans, measurement of BER variation has been imperfect to this point, yet tools to assess BER in epidemiological surveys are steadily evolving. We provide herein an overview of the BER pathway and discuss the current efforts toward defining the relationship of BER defects with disease susceptibility.

Introduction

DNA is under constant attack from exogenous (environmental) and endogenous ‘DNA-damaging agents’. Exogenous DNA damage occurs through exposure to chemicals or radiation originating from outside the organism. Endogenous DNA damage arises from modification by intracellular chemicals or through natural spontaneous decomposition (1), with the overall steady-state level of DNA damage being influenced by lifestyle practices and external factors (e.g. exposures, stress, etc.). The consequences of persistent DNA damage can vary from mutagenesis, potentially giving rise to carcinogenesis, to activation of cell death responses, which can contribute to degenerative disease (2,3). Consequently, repair of DNA damage is one of the most crucial biological processes in virtually all cell types and living organisms.

DNA repair systems are divided broadly into the following major pathways: direct reversal, mismatch repair, base excision repair (BER), nucleotide excision repair (NER) and recombinational repair. All of these pathways or their related subpathways—several in conjunction with checkpoint regulatory mechanisms that arrest the cell cycle to allow time for repair—have been designed to clear a particular subset of DNA damage from the genome, although some overlap in substrate specificity does exist. Defects in these protective processes have been genetically linked to development failings, cancer predisposition, neurological deficiencies and premature aging, emphasizing the importance of preserving one’s genetic material (4).

In this review, we begin by presenting the range and scope of the endogenous DNA damage problem. We next describe the mechanistic aspects of BER (and its subpathways), a major system for coping with many forms of endogenous DNA damage. Subsequently, we highlight evidence that indicates that defects in BER can affect healthspan. We close by discussing the technical approaches to address the extent to which altered BER capacity contributes to disease susceptibility.

Part 1. Endogenous DNA damage

DNA-damaging agents are widespread and include many environmental physical and chemical agents—such as ionizing radiation, ultraviolet light and numerous electrophilic compounds—as well as intracellularly produced reactive chemical species. Endogenous DNA damage is the most common and consists of three broad types: hydrolytic, oxidative or alkylative. Such damage is unavoidable, since it arises as natural products of the intrinsic instability of nucleic acid or as products of reactions with naturally produced intracellular metabolites, namely reactive oxygen species (ROS) or reactive nitrogen species and alkylating molecules. Despite decades of research, measuring the frequency of occurrence of DNA lesions in the genome remains difficult, as evidenced by the wide range in reported levels of endogenous DNA damage (5). This effort is further complicated by the variability in cellular and tissue environments, and the effect of time. Despite the limitations, it is widely accepted that over 50 000 endogenous lesions of the forms listed in Table I are generated within each human genome per day (1).

Table I.

Sources, types, frequency and consequences of several BER substrates

Lesion source	Lesion type	Lesion frequency (per 10e6 bases)	Toxicity	Mutagenicity	Repair enzyme/pathway
Hydrolysis	Deoxyuracil	1		+++	UNG, SMUG1
	AP site	3–20	+++	+	APE1
Alkylation	7-methyl guanine	1			Not repaired
	3-methyl adenine	0.1	++		MPG
	O6-methyl guanine	0.01	+	+++	Direct reversal
Oxidation	8oxoG	1–10	+	+++	OGG1, MUTYH
	Ethenoadenine	0.05	+++	+++	AlkB homologs
	Ethenocytosine	0.2	+++	+++	AlkB homologs
	Hydroxycytidine	1		++	NTH1, NEIL1
	Thymine glycol	2	+++	+	NTH1, NEIL1
	FaPyG	0.6a		++	OGG1, NTH1, NEIL1
	FaPyA	0.8a	++	+++	NEIL1

Table is adapted from references (2,5), with frequency data tailored specifically to frequency of lesions in leukocytes, except for the FaPy lesions. The toxicity and mutagenicity columns are presented as general guidance (with degree increasing with + symbol). The measured levels of any given DNA lesion tend to vary by orders of magnitude depending on sampled tissue and measurement technique. FaPy, formamidopyrimidine; SMUG1, single-strand selective monofunctional uracil glycosylase. ^aData from other cell types.

We point out that not all endogenous DNA damage is a product of direct chemical modification. Misincorporation events involving ribonucleotides, and ionic or tautomeric bases, can also contribute significantly to the inherent rate of mutation (6,7). Though misincorporated tautomers are not known to be repaired by the BER response, ribonucleotides can be processed by an initial cleavage event involving RNase H2, which shunts the incision product into a pathway that resembles long-patch BER (see Part 2).

Hydrolytic damage

Hydrolytic damage can occur when DNA spontaneously reacts with water. This chemical reaction can produce two common forms of DNA damage: the apurinic/apyrimidinic (AP or abasic) site and deoxyuracil. Spontaneous hydrolytic glycolysis occurs when water or naturally occurring ROS protonate the DNA base, inducing a spontaneous break in the linkage between the base moiety and the sugar backbone, creating an AP site (8). Pyrimidines are more rarely protonated, due to the specific pKa of the bases, and thus depurination is ~20 times more frequent than depyrimidination (9). In addition, DNA glycosylases intentionally catalyze the same reaction to remove damaged bases from DNA and initiate the BER response (see Part 2). Abasic sites can be measured in existing genomes by various techniques, but they are perhaps one of the most difficult lesions to quantify accurately since they can be generated spontaneously even without oxidative reactions. Nakamura et al. (10) have attempted to tackle this problem and report that in aggregate, the average somatic cell has a steady-state level of 50 000–200 000 AP sites, depending on the tissue and condition of the organism (5), although a mathematical model has predicted far fewer steady-state lesions (11). As AP sites lack the instructional information of the base, they present mutagenic templates to RNA and DNA polymerases, yet are often potent blocks to nucleic acid synthesis (12).

Of the unmodified nucleosides, it is primarily cytosine that is vulnerable to hydrolysis, spontaneously converting to deoxyuracil via deamination (13). Methylcytosine, an epigenetic base modification prevalent in the regulation of gene expression, is especially susceptible to deamination, undergoing hydrolysis at three to four times the rate of unmodified cytosine, resulting in the formation of thymine:guanine mispairs in DNA. Given that both uracil and thymine pair with adenine during chromosome replication, it is perhaps not surprising that C to T transitions are one of the most common mutations in human cells (14).

The rest of the DNA molecule is relatively resistant to hydrolytic attack, despite the presence of phosphoester bonds in the backbone, for example. In comparison with RNA, removal of the 2′ hydroxyl group from the ribose ring prevents self-cleavage reactions and changes the electrostatic landscape of DNA to be more resistant to spontaneous breakage. In fact, under physiological-like conditions, DNA will break into shorter and shorter fragments over the course of thousands of years, but the primary route of decay is breakage at sites that were first depurinated (1).

Alkylative damage

The most frequent route of alkylative DNA damage is through inadvertent reactions with S-adenosyl-methionine, a naturally occurring compound that is used as a cosubstrate in methyl group transfer reactions (15). In particular, S-adenosyl-methionine can readily methylate guanine to generate 7-methylguanine. This is a low toxicity lesion, yet it can spontaneously convert to an AP site through the breakage of its glycosyl bond; this reaction path is likely a significant contributor to the overall generation of the non-coding AP sites discussed above. A more minor, yet important, reaction product of S-adenosyl-methionine is the mutagenic, highly toxic 3-methyladenine lesion, which can strongly block DNA or RNA polymerases. Other harmful alkylative endogenous damages, which are not repaired by BER, but by direct reversal via a methyltransferase (i.e. O⁶-methylguanine-DNA methyltransferase), include the O-family products, most notably O⁶-methylguanine, O⁴-methylthymine and O⁴-ethylthymine (16).

Oxidative damage

Oxidative (and nitrative) DNA damage is mainly produced by reactions with ROS (or reactive nitrogen species) generated as direct or indirect products of oxygen metabolism that occurs in mitochondria during the production of adenosine triphosphate (ATP). However, ROS/reactive nitrogen species, such as hydrogen peroxide, superoxide, the hydroxyl radical or peroxynitrite, can also be produced by conditions (e.g. inflammation) or external exposures (e.g. ionizing radiation) that induce oxidative stress (17). Oxidative modification of DNA generally results in ring-oxidized or broken bases. One of the most common oxidative lesions is 8-oxoguanine (8oxoG), which is often used as a biomarker of oxidative stress and is produced by direct oxidation of the double bond of the minor guanine ring. 8oxoG is relatively non-toxic, yet is highly mutagenic, since it has the propensity to base-pair with adenine, causing G to T transversion mutations (18,19). Other non-bulky base lesions generated via direct oxidation include thymine glycol, hydroxycytosine, formamidopyrimidines and hydroxymethylcytosine, to name a few (20). In addition, more complex, helix-distorting oxidative products can be created via secondary reactions, i.e. when ROS peroxidize cellular lipids, which in turn react with DNA bases. The major base products of this cascade include pyrimido[1,2-]purin-10-(3H)-one (M1dG), 3,N4-ethenocytosine and 1,N6-ethenoadenine, which are substrates for BER, NER or other DNA repair mechanisms (21). Direct oxidation of DNA also contributes to the generation of AP sites and single-strand breaks, since ROS can attack the relatively labile glycosidic bond or the sugar ring and promote base release or strand cleavage, respectively. In the case of hydroxyl radical-mediated DNA strand breakage, the 3′ termini often harbor non-conventional chemistries, such as phosphoglycolates or phosphates, which require removal prior to repair synthesis or ligation.

Consequences of DNA damage

The result of any given DNA damage depends on two factors: its mispairing/miscoding potential and its polymerase-blocking potential. If a lesion has a propensity to promote erroneous base-pairing during DNA replication, it poses a mutagenic threat and thus has the potential to introduce a permanent genetic change, typically a missense or non-sense mutation. Such genetic alterations can initiate the carcinogenic process if they occur within the coding region or critical regulatory elements of a crucial tumor-suppressor gene. For example, the TP53 oncogene, which encodes the tumor-suppressor cell cycle checkpoint protein p53 and is mutated in approximately half of all cancers, is unusually susceptible to missense mutations, undergoing a high degree of spontaneous G:C to A:T transitions and G:C to T:A transversions, presumably arising from replication bypass of deaminated cytosine or unrepaired 8oxoG lesions, respectively (22). If the DNA modification is a powerful block to either a progressing DNA or RNA polymerase, it can result in replication or transcription arrest, which can give rise to chromosome instability or cell death responses (23–25). The former outcome is a common attribute of cancer cells (i.e. genome rearrangements or aneuploidy), whereas the latter likely contributes to degenerative diseases, such as disorders associated with neuronal cell loss. In general, lesions that strongly distort the double-helical structure of DNA tend to be more toxic, because they arrest replicating polymerases, inducing more ‘drastic’ cellular outcomes.

Part 2. BER of endogenous DNA damage

The proteins that deal with the non-bulky, endogenous DNA lesions described above are primarily part of the BER pathway. BER was a term originally coined to refer to a medley of enzymatic steps that carry out repair of chemically modified DNA bases. We describe next the various subpathways of BER, which are related through common chemistry or shared components, as well as the core players that function within these processes (summarized in Figure 1).

Fig. 1.

Short-patch BER pathway. A DNA glycosylase excises an aberrant base (e.g. uracil = U or thymine glycol = Tg). Monofunctional glycosylases create an AP site, which is incised by APE1 (left), whereas bifunctional glycosylases cleave the DNA backbone at the AP site via β-elimination or β,δ-elimination, creating a 3′-terminus that must be removed by APE1 (center) or polynucleotide kinase/phosphatase (PNKP) (right), respectively. Following end-cleanup (highlighted in red), single-nucleotide repair synthesis is conducted by POLβ and nick ligation is carried out by an XRCC1/LIG3α complex (highlighted in blue). See text for additional details.

DNA glycosylases

The first step in the repair of non-bulky base damage is detection by enzymes known as DNA glycosylases (26,27). DNA glycosylases are split into two branches, monofunctional and bifunctional, which share a common goal: to excise unusual or modified bases from the genome. Monofunctional glycosylases leave behind an AP site when they break the N-glycosyl bond, whereas bifunctional glycosylases can cleave the DNA backbone at the AP site product to generate a single-strand break. Glycosylases operate as scanning enzymes that slide along DNA, likely in the minor groove, looking for small helical distortions or reduced duplex stability by using a probing amino acid (28). At the site of a compatible distorted structure, they ‘flip-out’ the substrate base into their active site and cut the N-glycosyl bond by attacking carbon-1 of the associated sugar. The DNA glycosylase that is chosen to initiate repair is determined by the lesion and can influence the subpathway of BER that takes place, since different glycosylases produce distinct repair intermediates.

Monofunctional glycosylases

Monofunctional glycosylases include N-methylpurine-DNA glycosylase (MPG), uracil N-glycosylase (UNG), mutY homolog glycosylase (MUTYH), thymine-mismatch-specific thymine-DNA glycosylase (TDG), methyl-CpG-binding domain protein 4 (MBD4) and single-strand selective monofunctional uracil glycosylase (SMUG1). These glycosylases excise the substrate base through an aspartic acid-activated water molecule that performs a nucleophilic attack on the sugar carbon-1 (29). After base release, the DNA glycosylase can dissociate, leaving behind an AP site. However, there is evidence that certain monofunctional glycosylases can remain bound to the abasic intermediate to protect the lesion from undergoing spontaneous strand cleavage and/or to coordinate with the next enzyme in the BER response, an AP endonuclease (see below) (30).

Bifunctional glycosylases

Bifunctional glycosylases include endonuclease III-like protein 1 (NTH1), 8oxoG glycosylase (OGG1) and the endonuclease VIII-like protein family members (NEIL1, NEIL2 and NEIL3). To catalyze base release, the bifunctional glycosylases attack the same carbon atom as targeted by the monofunctional enzymes. However, instead of activating water, the conserved aspartic acid is used to activate the amino group of an active site lysine residue, which forms a catalytic Schiff base intermediate that reversibly links the protein to the DNA strand (31). At this point, the glycosylase can (i) be released from the DNA backbone leaving behind an abasic site or (ii) carry out one of two reactions to generate a DNA strand break prior to release. Specifically, bifunctional glycosylases execute either β-elimination, which results in a strand break with a 3′ α,β-unsaturated aldehyde (e.g. OGG1 and NTH1), or β,δ-elimination, which produces a 3′ phosphate terminus (e.g. NEIL1). In the former situation, APE1 (see next section) excises the 3′ blocking group via its phosphodiesterase functionality. In the latter situation, the 3′ phosphate is processed by polynucleotide kinase/phosphatase, which directs an APE1-independent repair reaction (32). Once the 3′ end has been converted to a hydroxyl group, a polymerase is recruited to carry out DNA synthesis and a ligase operates to seal the nick (31) (Figure 1).

Short-patch BER

AP endonucleases are central enzymes within BER and are responsible for initiating repair of both spontaneous hydrolytic AP sites and those generated by DNA glycosylases. The major human AP endonuclease, APE1, searches the genome for AP sites in a quasi-processive manner, which involves transitory scanning of DNA and probing for duplex instability (33). Although there is evidence suggesting that APE1 is recruited to AP sites by bound DNA glycosylases, it is more likely that APE1 locates naked AP-DNA. Once APE1 has identified an abasic lesion, it kinks the double helix, bringing the AP residue into its active site and organizing the phosphodiester backbone to facilitate strand cleavage via an activated water molecule and a metal cofactor, likely Mg²⁺. This incision event leaves behind a 3′ hydroxyl priming group and a 5′ deoxyribose phosphate (dRP).

Next, DNA polymerase β (POLβ) performs 5′ dRP excision and gap-filling synthesis, potentially in a coordinated manner with product-bound APE1 (34). POLβ is a specialized polymerase, lacking the processivity and exonuclease proofreading capacity of the replicative polymerases (35). However, POLβ possesses the dRP lyase functionality, which allows it to remove the 5′ blocking group generated by APE1 to facilitate nick ligation. In short-patch BER (perhaps more accurately termed, single-nucleotide BER), POLβ is thought to recruit a complex of X-ray repair cross-complementing protein 1 (XRCC1) and DNA Ligase 3α (LIG3α). Typically, after one-nucleotide extension, POLβ stalls due to its low processivity and poor strand-displacement activity, allowing XRCC1/LIG3α to close the backbone and complete the short-patch repair event (Figure 1). However, under circumstances where LIG3α fails to seal the nick, the alternate long-patch BER pathway comes into play (36).

Long-patch BER

Long-patch BER is called upon under specific circumstances, such as those involving certain DNA substrates or possibly during periods of active replication, when the relevant protein factors are more abundant. For example, specific oxidized or reduced chemical forms of AP sites, such as the C4 oxidized AP site, are not substrates for the lyase activity of POLβ after strand cleavage by APE1 (37). In these situations, strand-displacement, multinucleotide synthesis must occur to generate a DNA intermediate that can be processed by the flap-endonuclease, FEN1, to remove the 5′ blocking group to establish a ligatable substrate. Additionally, in conditions where cellular ATP levels are low—such as when excessive levels of DNA damage exist that promote poly(ADP-ribose) polymerase-driven ATP depletion or under specific conditions of metabolic stress—mammalian ATP-dependent DNA ligases will not efficiently catalyze nick sealing (36). In this situation, POLβ can carry out strand-displacement synthesis, or more likely, POLε or POLδ are recruited to promote long-patch BER. Completion of long-patch BER is presumably conducted by FEN1 and its interacting protein partners, proliferating cell nuclear antigen and LIG1, which cooperate to seal the DNA backbone after 5′ flap excision (38) (Figure 2).

Fig. 2.

Long-patch BER pathway. The black star represents a 5′ end modification that is not a substrate for the lyase activity of POLβ. In this scenario, or in situations of low intracellular ATP concentration where ligation is impeded, strand-displacement synthesis is carried out by POLβ, or one of the replication-associated polymerases, POLε or POLδ, to generate a 5′ flap structure. This structure is processed by FEN1, and repair is completed, likely in cooperation with its interacting protein partners proliferating cell nuclear antigen (PCNA) and LIG1.

Pathway kinetics

The kinetics of the BER pathway can be variable, influenced by the target substrate, the overall protein levels and the intracellular environment. For example, certain DNA glycosylases, such as OGG1, are considerably slower in catalyzing base release than others, namely UNG (31). As a result, certain base lesions have a greater probability of accumulating, particularly under situations of high DNA damage load, such as during conditions of oxidative stress. While repair of some base lesions must surely be dictated by the excision step, once an AP site or a single-strand break is generated, the limiting step in BER is less clear. In vitro reconstitution experiments suggest that the dRP lyase reaction of POLβ is rate-limiting (39). However, the use of only POLβ and LIG1, without XRCC1/LIG3 or poly(ADP-ribose) polymerase 1, makes this claim less certain. Mathematical modeling of in vitro kinetic data and intracellular protein concentrations for several of the BER enzymes supports the idea that the POLβ reactions are rate-limiting (40). Moreover, the in vivo experiments of Sobol et al. reinforce this conclusion, since overexpression of just the dRP lyase domain of POLβ rescues cellular sensitivity to the alkylating agent, methanesulfonate (MMS) (41,42). However, POLβ is not always the enzyme that carries out end-cleanup, and when APE1 is anointed this function, its 3′ phosphodiesterase activity can be rate-limiting (43). Furthermore, when poly(ADP-ribose) polymerase 1 is responsible for coordinating the repair of strand-break damage, ligation can become rate-limiting, particularly when the reaction is impaired by low ATP concentrations (44). In total, since situations arise in which virtually any step of BER could be rate-limiting, and since there exists variability in the rate of repair among individuals (see Part 5)—either due to genetic variation, such as single-nucleotide polymorphisms, or to disparate intracellular or external conditions—it is likely that a deficiency in any of the functional components of BER could produce a phenotype of increased susceptibility, if not outright symptoms of disease.

Part 3. BER defects in disease and susceptibility

Unlike some of the other DNA repair systems—such as NER where mutations in the core genes are linked to the sun-sensitive, cancer-prone disorder xeroderma pigmentosum (45,46)—mutations in the individual components of BER are not associated with a defined genetic disorder. However, there are instances were mutations in genes related to BER or one of its subpathways have been causally linked to inherited disorders involving cancer predisposition, immunodeficiency or neurological defects (47). In particular, mutations in (i) the DNA glycosylase MUTYH have been tied to colorectal cancer (48); (ii) the uracil DNA glycosylase UNG have been connected to hyper-IgM syndrome (49) and (iii) tyrosyl-DNA phosphodiesterase 1, aprataxin and polynucleotide kinase/phosphatase, three genes involved in single-strand break processing, have been linked to spinocerebellar ataxia with axonal neuropathy, ataxia with oculomotor apraxia type 1 and microcephaly with seizures, respectively (50–52). Tyrosyl-DNA phosphodiesterase 1 handles 3′ topoisomerase abortive intermediates, aprataxin copes with 5′ adenosine monophosphate abortive ligation intermediates, and polynucleotide kinase/phosphatase deals with 3′ phosphate residues. All of these proteins, as well as APE1, which can excise 3′ α,β-unsaturated aldehydes and phosphoglycolates, function as part of the DNA single-strand break repair subpathway of BER (53).

The scarcity of inherited disorders tied to BER likely relates to the essential nature of the repair system. In particular, deletion of both alleles of the core components of BER leads to early stage embryonic lethality or postnatal lethality in mice (54). This finding underscores the importance of BER in coping with the high level of endogenous DNA damage and suggests that severe loss-of-function alleles in at least a subset of the BER genes may not be present within the population. The diversity of the inherited disorders connected to BER also likely reflects the wide-spanning importance of the pathway in protecting the genome. A real challenge for the scientific community has been to address whether more subtle defects in BER associate with disease risk. The strongest evidence supporting the idea that partially reduced BER function will contribute to disease susceptibility comes from a series of ‘easily’ controlled studies using genetically defined, heterozygous (haploinsufficient) mouse models.

Part 4. BER haploinsufficient mouse models

DNA glycosylases

Unlike several of the core BER proteins that will be discussed below, deletion of both alleles of nearly all of the DNA glycosylases is compatible with life (55,56); the exception is TDG, a protein that has important roles in DNA demethylation epigenetic programming (57). In fact, many of the single glycosylase null animals show little or no phenotype. It is believed that the existence of multiple DNA glycosylases provides protective redundancy, and/or that base modifications are more easily tolerated or are handled by an alternative repair mechanism. Whatever the case, the lack of major clinical pathologies being associated with glycosylase null animals has discouraged extensive analysis of heterozygous knockout mice. We review here some of the work done using glycosylase haploinsufficient mouse models, yet point out that recent studies using complete knockout mice have revealed both expected and unexpected roles for DNA glycosylases in cancer predisposition, hypermetabolic syndrome, tissue inflammation and neurodegeneration (56,58).

MPG (a.k.a. AAG) is responsible for removing alkylative base lesions, such as 3-methyladenine, as well as several base modifications induced by ROS, including 1,N6-ethenoadenine, hypoxanthine, oxanine and 8oxoG. Meira et al. (59) reported, expectedly, that MPG^{− /−} mice exhibit reduced disease-free lifespan following treatment with a sublethal dose of the alkylating agent, MMS, in comparison with wild-type animals. However, they surprisingly found that the wild-type mice displayed a greater degree of retinal degeneration than their MPG^{− /−} counterparts when exposed to MMS, suggesting that MPG-initiated BER renders retinal photoreceptors sensitive to alkylation-induced cell death. Haploinsufficient (MPG^+/−) mice treated with MMS exhibited an intermediate level of retinal degeneration, indicating a glycosylase dosage-dependent outcome. Although studies assessing the comparative tumor susceptibility of heterozygous and homozygous MPG-deficient mice have not been conducted to our knowledge, work by Samson and colleagues indicates that MPG null animals display an increased risk of azoxymethane- or dextran sodium sulfate-induced, chronic inflammation-associated colorectal tumorigenesis (60–62).

OGG1 is the main enzyme responsible for the removal of the mutagenic base lesion 8oxoG. The OGG1^{− /−} mouse model was first described by Klungland et al. (63), and they reported an accumulation of 8oxoG in the nuclear genome and a corresponding elevated spontaneous mutation rate. However, the authors found that the null animals displayed no noticeable phenotype, including no increase in tumor formation. Cell-free tissue extracts from OGG1^+/− heterozygous mice were found to display a roughly 2-fold reduced incision activity against 8oxoG:C substrates or release of formamidopyrimidine base modifications, consistent with a repair haploinsufficiency. Similar to what was done by Liao et al. (64)—who evaluated chronic ulcerative colitis-induced carcinoma in dextran sodium sulfate-treated in both OGG1 knockout and haploinsufficient animals—it may be worthwhile to characterize more comprehensively heterozygous and null glycosylase mice for their dose–response profiles following a germane environmental challenge or pathological condition, as such information could allow us to better predict individual risk to a particular, quantifiable exposure. We point out that studies using double knockout OGG1^{− /−} MUTYH^{− /−} mice have shown an age-dependent increase in 8oxoG damage in several tissues, as well as an increased incidence of lung and small intestinal cancer, supporting the notion that accumulation of endogenous base damage can drive neoplastic transformation (65).

APE1

Since APE1 is essential for mouse embryogenesis, attention has been directed at characterizing the APE1^+/− animals. Meira et al. (66) were the first to interrogate the heterozygous mice, reporting that APE1^+/− tissues express roughly 50% of the normal level of messenger RNA and protein, and that isolated APE1^+/− cells exhibit increased sensitivity to the oxidizing agents menadione and paraquat. The investigators also observed a reduction in the number of heterozygous pups from what was expected from normal Mendelian inheritance, suggesting that a deficiency in APE1 adversely affects survival. Since the addition of antioxidants to the diet of the pregnant females increased the proportion of viable APE1^+/− progeny, it would appear that the inherent survival defects stem from deleterious consequences of oxidative stress that arises during gestation. Although the haploinsufficient mice did not appear to have an altered life expectancy, they did show an increased incidence of spontaneous microscopic tumors, as well as cardiac abnormalities.

Huamani et al. (67) reported that APE1^+/− mice display normal body weight and testis weight and have no major histological differences in testis, spleen or liver. However, upon crossing the heterozygous mice with a lacI transgenic mouse, which permits assessment of in vivo spontaneous mutation rates, they found that the APE1^+/− lacI⁺ mice exhibit an age-dependent increase (~2-fold) in spontaneous mutations in the liver, spleen and spermatogenic cells. Subsequent work by Vogel et al. (68) revealed that both the nuclear and mitochondrial genomes of heterozygous mice accumulate a higher level of DNA damage in comparison with the wild-type controls. In a separate study, Unnikrishnan et al. (69) reported that the liver of APE1^+/− mice exhibit both elevated DNA lesions and apoptosis when treated with the oxidizing agent and well-known heptocarcinogen, 2-nitropropane. This increased sensitivity to oxidative stress was shown to likely involve both impaired BER and an altered redox response, since APE1 operates not only as a DNA repair enzyme, but also as a transcriptional regulator via its ability to modulate transcription factor DNA-binding activity (70). Jeon et al. (71) found that heterozygous APE1^+/− mice exhibit impaired endothelium-dependent vasorelaxation and hypertension, likely due to the reduced capacity to accurately govern endothelial nitric oxide synthase activity and vascular nitric oxide levels through a redox-sensitive transcriptional mechanism involving H-ras-PI3-K/Akt. Finally, Sengupta et al. (72) reported that APE1^+/− mice have higher kidney renin messenger RNA and plasma-renin levels as compared with wild-type mice, suggesting another mechanism by which APE1 may regulate blood pressure homeostasis. In total, evidence indicates that a deficiency in APE1 can have adverse consequences on both spontaneous and exposure-dependent pathological endpoints.

POLβ

In the case of DNA POLβ, deletion of both alleles in mice leads to postnatal lethality, likely due to dysfunctional neurogenesis (73). Thus, like APE1, work has been focused on characterizing the haploinsufficient animals. Heydari and colleagues documented that POLβ^+/− mice express roughly 50% of the normal transcript and protein levels and also showed that the animals maintain a higher steady-state level of endogenous DNA damage as assessed by the Comet assay (74,75). In addition, the heterozygous mice hyperaccumulate DNA single-strand breaks following exposure to 2-nitropropane, consistent with a repair (gap processing) defect. Notably, a folate-deficient diet, which can induce oxidative stress and lead to an increase in uracil incorporation into the genome, was shown to promote single-strand break accumulation in POLβ^+/− mice, further evidence for a BER deficit (76). Cabelof et al. (77) reported that 24- to 26-month-old POLβ^+/− mice exhibit at least a 5-fold increase in the incidence of spontaneous lymphomas relative to their wild-type counterparts, as well as an elevated frequency of adenocarcinomas. Moreover, POLβ^+/− mice had an accelerated age-dependent rate of mortality, although their mean and maximum life expectancy was similar to their wild-type counterparts. Consistent with the increased tumorigenesis in the POLβ^+/− animals, bone marrow cells obtained from heterozygous mice displayed inherent chromosome instability (75,77), and POLβ^+/− germ cells exhibited a 2-fold higher spontaneous mutation frequency (78,79). POLβ^+/− lacI⁺ mice also showed increased mutagenicity in response to the DNA-damaging agent dimethyl sulfate (75). More recently, Ventrella-Lucente et al. (80) demonstrated that heterozygous animals injected intraperitoneally with 1,2-dimethylhydrazine, an established colon and liver carcinogen, display a higher level of aberrant colonic crypt formation, an early indicator of carcinogenesis. Thus, the studies to date indicate that POLβ is likely a low penetrance gene that when combined with a high penetrance insult, such as a relevant exposure or nutritional deficiency, can lead to increased DNA damage accumulation and pathology. We mention that (i) several missense variants in DNA POLβ, which are quite common in tumors, have been shown to drive cellular transformation and thereby likely contribute to the carcinogenic process (81) and (ii) impaired POLβ activity associates with pathologies that resemble systemic lupus erythematosus, an autoimmune disease (82).

XRCC1

Mice null for both alleles of the non-enzymatic scaffold protein XRCC1 die early during embryogenesis (83). Thus, work conducted in our laboratory characterized several endpoints related to disease and aging in XRCC1^+/− heterozygous mice, including fat deposition, blood chemistries, tumor formation and genomic stability, motor coordination and memory, and lifespan (84). By and large, haploinsufficient animals, which expressed roughly 50% XRCC1 protein, exhibited similar phenotypes to their wild-type littermates under normal (protective) husbandry conditions. A small number of XRCC1^+/− mice did incur spontaneous abdominal organ ruptures (of the bladder, liver or pancreas), possibly due to failed endoderm or mesoderm development, potentially linking BER-related defects to congenital gastrointestinal malformations. In addition, a subset of the heterozygous animals displayed clinical phenotypes consistent with a stroke, in-line with the emerging evidence that BER may modulate stroke risk or recovery (85). Finally, XRCC1^+/− mice treated with the alkylating agent azoxymethane showed an increase in liver toxicity and precancerous colonic lesions, supporting the hypothesis that reduced BER capacity increases susceptibility to a relevant genotoxic exposure (84). Saribasak et al. (86) found that XRCC1 happloinsufficiency affects somatic hypermutation and class switch recombination, events critical to the B-cell immune response.

Tebbs et al. (87) created a XRCC1 hypomorph model, where the XRCC1^{− /−} background was complemented with a randomly integrated XRCC1 minitransgene that expressed the native protein at roughly 10% the normal level. The fact that these animals were born and reached adulthood indicates that 10% XRCC1 expression is sufficient to support embryonic and postnatal development. Although this initial study described the hypomorph progeny as being healthy, fertile and ostensibly normal, a follow-up report indicated that the XRCC1 hypomorph mice were consistently smaller than their wild-type littermates, with an ~25% reduction in body weight at weaning that was carried into later life (88). Since the hypomorphs displayed no robust lethal tumor phenotype, a 10% XRCC1 protein level is seemingly adequate to support sufficient BER capacity to cope with the normal rates of endogenous DNA damage formation. Future work could include quantitatively mapping out the sensitivities of the various BER-deficient (e.g. haploinsufficient) mouse models to relevant environmental exposures in comparison with control animals, with an eye toward identifying where a deficit in BER might increase human disease susceptibility.

Part 5. Assessing BER capacity

Because BER defects may predispose to disease and premature aging phenotypes, it is of interest to determine the rate and/or effectiveness of the pathway among individuals. Toward this end, two general approaches have been pursued: specific (often single) step assays or whole pathway assays (summarized in Table II). We review here some of the evolving strategies to assess BER capacity within the population and highlight a few key advances thus far.

Table II.

Overview of the main assays to measure BER

Single-step assays
	Overview	Pros	Cons	Application
Oligo (radiological)	1. Damage-containing oligo substrate is incubated with a cell or tissue protein extract	• Custom synthesis of a site-specific defined DNA substrate	• Radioactive	In vitro
			• Not high-throughput
	2. Reactions are resolved on a denaturing gel, and substrate processing is quantified	• Highly sensitive	• Limited to extract-based studies
		• Well-established methodology
		• Can be adapted to look at repair intermediates
Oligo (molecular beacon)	1. Incubate damage-containing oligo substrate with protein extract (a) or transfect into cell (b)	• Custom synthesis of a defined DNA substrate	• Somewhat limited sensitivity	In vitro or in vivo
	2. Quantify amount of fluorescence by standard readout methods (a) or microscopy (b)
		• Real-time readout	• Main application currently for extract-based assays
		• Can be adapted to solid support strategy
		• High-throughput and multiplex capabilities	• Localization issues in cells
Pathway assays
	Basic principle	Pros	Cons	Application
HCR	1. Prepare damage-containing DNA substrate	• Plasmid-based substrate [can be damage-specific (oligo) or more general (chemical treatment)]	• Substrate preparation laborious	In vivo
	2. Transfect substrate into cells		• Certain cells not well suited to plasmid transfection
	3. Measure reporter gene output (e.g. luciferase or fluorescence)	• Localization issues minimal
			• Episomal (i.e. not the genome)
		• Well-established methodology
		• High-throughput capabilities
Comet	1. Isolate cells (before and/or after genotoxin treatment) and prepare single-cell suspension	• Genome measurement	• Background signal can complicate repair measurements
		• Low threshold for damage detection
		• Small sample size required (single-cell analysis)	• Conversion to repair capacity evaluation is cumbersome
	2. Prepare agarose gel mixture and lyse cells
		• Kits are common
			• Labor-intensive quantitation and interpretation
			• Reproducibility difficult
	3. Perform electrophoresis and quantify tail length by microscopy

Listed is a brief procedure for the different single-step and pathway assays, as well as the pros/cons and application to human studies. See Figures 3 and 4 for assay schematics. HCR, host cell reactivation.

Single-step assays

The early studies measuring BER capacity employed what we have deemed ‘single-step’ assays. These methods interrogate a single or a limited number of steps of the pathway, typically by examining the activity of a cell or tissue protein extract on a defined damage-containing DNA substrate. One of the earliest studies looked at variation in uracil DNA glycosylase activity using a well-trodden method that involves the conversion of radioactive modified bases, which had been introduced into calf thymus DNA, from ethanol-insoluble to the free ethanol-soluble form (89). This study revealed a nearly 60-fold variation in uracil DNA glycosylase activity in extracts from small intestine biopsies, but a less dramatic 3-fold variation in liver extracts. In additional early work, Frosina and colleagues, using a classic plasmid relaxation assay, which records nicking of damage-containing, supercoiled plasmid DNA, reported a 2.4-fold variation in AP endonuclease activity in lymphocytes of 10 normal individuals (90). This analysis was expanded to 23 healthy women and 20 women with breast cancer, revealing a similar range of AP site incision variability among the subjects, yet no detectable change with age or the presence of tumors (91,92).

Since the advent of oligonucleotide synthesis and damage-specific phosphoramidites, BER activities have been measured primarily using defined DNA substrates. For example, Paz-Elizur et al. (93) performed a scaled-up epidemiological analysis of 8oxoG excision/incision among 120 individuals using a radiolabeled oligonucleotide substrate assay that assesses the combination of glycosylase base release and AP site cleavage (Figure 3A). These investigations revealed an up to 2.8-fold variation in the rate of 8oxoG processing that was independent of gender and smoking, but was lower in the tissue extracts of cancer patients relative to matched healthy controls, even in non-affected tissues. Excision of ethenoadenine, a glycosylase activity primarily associated with MPG, has been examined using a similar strategy, and a >10-fold variation was observed in peripheral mononuclear blood cells of 80 individuals (94). Our laboratory has performed an assessment of the later steps of BER, i.e. AP site incision, single-nucleotide gap filling and nick ligation, using a collection of defined radiolabeled oligonucleotide substrates (47). These studies, which involved lymphoblastoid cell line extracts from 23 individuals, revealed interindividual variation of 1.9-, 1.3- and 3.4-fold, respectively.

Fig. 3.

Single-step BER assays. (A) Radiolabeled substrate assay. The synthetic site-specific, damage-containing (primarily a base modification or AP site) oligonucleotide is end-labeled (green star) and annealed to a complementary unlabeled strand. The duplex substrate is incubated with a protein extract, and the strand cleavage (repair) efficiency is quantified following separation of substrate (S) and product (P) on a denaturing polyacrylamide gel (right). (B) Molecular beacon assay. A damage-containing substrate (depicted as a stable hairpin structure) is synthesized to harbor both a fluorophore (yellow star) and a compatible quench (black dot). Following processing by repair enzymes (intermediate product in brackets), the short product spontaneously dissociates and the fluorophore signal is measured by a plate reader (in vitro) or microscopy (in vivo). (C) The solid support microarray assay. This method, which works off a similar excision/incision principle to the assays above, measures the ability of an extract to reduce the fluorescent signal by processing an affixed damage-containing DNA substrate. Following treatment (step 1) and subsequent washing steps (step 2), the percentage of fluorophore remaining bound to the solid support is quantified to determine efficiency of repair. See Table II for further details about these assays.

With advances in conditional fluorescence, methods, such as the molecular beacon approach, have been developed to measure single-step BER activities in either protein extracts or within live cells (95–97). In this assay, a fluorescent reporter is positioned opposite a compatible quench, yet upon processing of the damage and strand cleavage, the fluorophore is released from the substrate to reveal its signal (Figure 3B). This strategy has thus far been designed to primarily measure DNA glycosylase and AP endonuclease functions, yet to our knowledge, has not been applied to defined cohorts. A variation of this approach is to attach a fluorescent damage-containing substrate (without a quench) to a solid support, such as a microarray plate (Figure 3C). The fluorescent oligonucleotides are then exposed to a protein extract, and after a wash step, the degree of loss of fluorescence indicates the efficiency of strand cleavage, i.e. repair. A variety of substrate arrays can be printed to measure different DNA glycosylase or AP endonuclease activities, and this approach has been applied to characterize the effects of aging on BER in fibroblast extracts from 29 subjects, where it was surprisingly found that 8oxoG excision increases with age (98). Notably, this method is suited to multiplexing with different DNA substrates harboring distinct fluorescent readouts, permitting simultaneous quantification of multiple BER activities in a single extract.

Since the assays above target only one or two steps of the BER process, they are limited in the information that can be gathered. In particular, single-step assays cannot reveal impaired coordination within the pathway, which could arise due to incompatible relative protein concentrations/activities. Early work by Glassner et al. (99) has indeed demonstrated that an imbalance in the steps of BER can lead to the accumulation of harmful DNA intermediates, resulting in deleterious genomic consequences. Moreover, since the specific BER step measured may not represent the rate-limiting component of the pathway, any variation detected may not accurately reflect the ultimate repair capacity phenotype. For example, an individual who incises AP-DNA faster could easily complete the repair process slower due to a defective copy of POLβ. With this in mind, a number of assays have been developed to assess BER as a whole (Table II).

Pathway assays

The earliest BER pathway assays focused on extract-based radionucleotide incorporation (Figure 4A). The landmark paper by Lindahl’s group on the choice between long- and short-patch BER is a good example, where a radioactive DNA product arises only when the initial unlabeled, damage-containing substrate has been processed and repair has been initiated (100). Extract assays, however, tend to be sensitive to the specifics of the extraction procedure and the in vitro reaction conditions and are unable to capture the complex pathway interactions that take place in cells. Because of this, there has been an emphasis on the development of in vivo repair assays.

Fig. 4.

BER pathway assays: (A) Radiolabeled nucleotide incorporation assay. An unlabeled damage-containing DNA substrate (S; oligonucleotide substrate depicted) is incubated with a protein extract in the presence of a radiolabeled nucleotide (red bar-green star). BER processing results in the formation of radiolabeled repair intermediates (P1) and the radiolabeled fully repaired product (P2), which can be visualized following resolution on a denaturing polyacrylamide gel (right). (B) HCR assay. As depicted, the damage introduced blocks transcription of the reporter gene, resulting in low signal following transfection. Upon repair, the reporter gene signal is restored, allowing quantification of luciferase activity or fluorescence by plate reader technologies. (C) Comet assay. Cells are harvested with or without DNA-damaging agent treatment, and the level of chromosome damage is quantified by embedding single cells into an agarose gel, lysing them and subjecting their DNA to electrophoresis. A greater comet tail relative to the comet head is reflective of a higher level of DNA damage. See Table II for further details about these assays. HCR, host cell reactivation.

One of the earliest techniques for measuring cellular DNA repair capacity is the so-called host cell reactivation assay (101) (Figure 4B). This method involves the use of a plasmid substrate that has been modified to contain a particular form(s) of DNA damage, either by specific chemical treatment or strategic introduction of a damage-containing oligonucleotide. The idea is that the damage will lead to blocked transcription or transcriptional mutagenesis, rendering the reporter gene (typically luciferase or a fluorescent protein) within the plasmid ‘inactive’. If the DNA lesion is repaired completely and accurately, the reporter gene output is restored and can be quantified. While the assay has mainly been employed to assess NER, Nagel et al. (102) have recently adapted the technique to measure 8oxoG repair. In their assay, successful 8oxoG processing reverses the transcriptional mutagenesis induced by the mutagenic lesion, with repair inactivating the fluorescent protein (going from on to off). The basic method, which was modified for high-throughput, multiplex analysis of several DNA repair mechanisms, was applied to several human cell lines, although BER was not explicitly assessed.

One of the mainstays of DNA damage research has been the Comet assay (103) (Figure 4C). The Comet assay measures steady-state levels of DNA damage, yet can also record repair capacity over time after an initial genotoxic insult. The basic strategy involves embedding single cells, which can come from a variety of sources (e.g. cell culture, blood samples and tissues), into low-melting point agarose on a slide, and after lysis, subjecting their genomic DNA to electrophoresis. While intact undamaged (or repaired) double-stranded DNA translocates through the gel slowly, damaged (or unrepaired) DNA, especially DNA suffering from strand breaks, migrates out and creates a ‘comet tail’ that can be visualized with standard stains, such as ethidium bromide. The length of the comet tail and the proportion of DNA within the tail relative to the ‘head’ determine the level of genomic damage.

There are two major subflavors of the Comet assay, which differ in the cell lysis method applied and thus the types of DNA damage detected. The neutral comet detects double- and single-strand breaks, whereas the alkaline comet also detects AP sites and other alkaline-labile lesions. In addition, if a repair enzyme, such as the Escherichia coli DNA glycosylase FPG, is incorporated to treat cells after lysis, oxidative base lesions can be converted into single-strand breaks and measured. Variations of the Comet assay have been applied to several human-based studies—which have examined the variability of steady-state damage or BER responses among specific cohorts, as well as the effects of aging or environmental, lifestyle or occupational exposure on BER capacity—and we refer the reader to the following review for additional details of the relevant literature (104). A major problem that has plagued the Comet assay, however, is the high degree of interassay and interlab variability due to the sensitivity of the technique to slight variations in sample preparation or protocol execution, for example. In addition, Comet assays are fairly laborious, although with advances in commercial kits and materials science, they are becoming more mainstay tools of genotoxicity testing (105). Finally, we note that other methods have been developed to measure levels of chromosome damage, such as chromatography and Raman spectroscopy to quantify base lesions (e.g. 8oxoG) (106,107), direct chemical labeling of AP sites (108) and a fluorescent dye-based detection strategy for DNA strand breaks (109), yet these have their own limitations and technical challenges that will not be expounded upon here.

Part 6. Closing thoughts

It is well understood that endogenous DNA damage poses a significant threat to the integrity of the genome and thus the health and functionality of the organism. And while the basic foundation of the BER pathway is well characterized and understood, further investigations are likely to reveal additional molecular complexity and regulatory mechanisms for BER in the coming years. Genetically defined mouse models have provided important insight into the biological roles of the various BER components and have served as useful tools to determine the potential contribution of BER defects to spontaneous and exposure-dependent susceptibility. However, the differences in mouse and human pathophysiology underscore the importance of more comprehensive studies using defined cohorts to better delineate the relationship of BER defects with human disease (110).

In human cancer studies, due to the advent of high-throughput and massive parallel technologies, great headway has been made in determining what types of mutations likely participate in carcinogenesis. Although it has been known for decades that gross rearrangements like the Philadelphia chromosome are key in cancer development, recent whole-genome and whole-exome sequencing has revealed that there is a significantly higher rate of single-nucleotide mutations in cancer cells. As noted earlier, one of the most common mutations observed in human tumors is the C to T transition, which presumably arises from replication bypass of uracil or thymine residues generated by spontaneous deamination of cytosine or 5-methylcytosine, respectively, particularly within CpG islands (22). In addition, many, but not all cancers, have unusually high rates of G:C to T:A transversions, which likely arise from replication through an 8oxoG lesion (14,111). As such, evidence is compiling suggesting that endogenous DNA lesions can be important contributors to cancer initiation and/or progression, and that defects in BER could potentially promote tumorigenesis.

Although strides have been made to develop effective methods to assess BER capacity among individuals, each of the current methods has its own strengths and limitations (Table II). Nevertheless, as many of the strategies become more amendable to high-throughput analysis and new methods come online, the approach of combining multiple, complementary techniques on a single sample will likely lead to a better understanding of the subject’s repair potential. It indeed appears that the foundation is being set for more extensive BER epidemiological studies in the near future.

Funding

Intramural Research Program at the National Institutes of Health, National Institute on Aging (Z01 AG000750).

Glossary

Abbreviations:

8oxoG

8-oxoguanine

ATP

adenosine triphosphate

BER

base excision repair

dRP

deoxyribose phosphate

LIG3α

DNA ligase 3α

MMS

methanesulfonate

MPG

N-methylpurine-DNA glycosylase

MUTYH

mutY homolog glycosylase

NER

nucleotide excision repair

OGG1

8oxoG glycosylase

POLβ

DNA polymerase β

ROS

reactive oxygen species

UNG

uracil N-glycosylase

XRCC1

X-ray repair cross-complementing protein 1.