Endogenous base damage as a driver of genomic instability in homologous recombination-deficient cancers

Abstract

Homologous recombination (HR) is a high-fidelity DNA double-strand break (DSB) repair pathway. Both familial and somatic loss of function mutation(s) in various HR genes predispose to a variety of cancer types, underscoring the importance of error-free repair of DSBs in human physiology. While environmental sources of DSBs have been known, more recent studies have begun to uncover the role of endogenous base damage in leading to these breaks. Base damage repair intermediates often consist of single-strand breaks, which if left unrepaired, can lead to DSBs as the replication fork encounters these lesions. This review summarizes various sources of endogenous base damage and how these lesions are repaired. We highlight how conversion of base repair intermediates, particularly those with 5′or 3′ blocked ends, to DSBs can be a predominant source of genomic instability in HR-deficient cancers. We also discuss how endogenous base damage and ensuing DSBs can be exploited to enhance the efficacy of Poly (ADP-ribose) polymerase inhibitors (PARPi), that are widely used in the clinics for the regimen of HR-deficient cancers.

1. Introduction

Amongst the varied chemical and physical genomic lesions that a cell incurs, double-strand breaks (DSBs) are the most lethal. To resolve these toxic DSBs, at least three mechanistically distinct repair pathways have been identified which include: non-homologous end joining (NHEJ), homologous recombination (HR) and microhomology-mediated end joining (MMEJ) [1–3]. Amongst these pathways, HR has the highest fidelity because it entails the repair of the DSB using a homologous template. Briefly, repair of DSB by HR entails nuclease-mediated resection to generate a 3′-single-stranded DNA. BRCA1 has been implicated in promoting end-resection at DSBs. Additionally, BRCA1 together with BRCA2 promote loading of Rad51 on the 3′single-stranded DNA which then performs a homology search. Once a homologous sequence is found, DNA is copied, followed by dissolution and annealing of the invading strand back to its original template [4–6] (Fig. 1a). Due to its reliance on a homologous template, HR is only functional during the S and G2 phases of the cell cycle, during which it is the primary method of DSB repair. Loading of NHEJ proteins, Ku and DNA-dependent protein kinase catalytic subunit (DNA-PKcs) is negatively regulated at one-ended replication-coupled DSBs, further favoring HR during S-phase [7,8]. Lack of the high-fidelity HR pathway channels repair to more error-prone mechanisms such as MMEJ, invariably resulting in high genomic instability and increased predisposition to several cancer types. Of note, synthetic lethality between HR and MMEJ deficiency is observed even in the absence of any exogenous DNA damage [9,10]. This observation highlights that there should be continuous generation of DSBs in HR-deficient cells that become reliant on MMEJ for repair. This review discusses endogenous mechanisms that can fuel the generation of DSBs and hence dictate genomic stability in HR/BRCA-mutant cancers. In particular, the review focuses on the cell-intrinsic base damage as a source of single-strand breaks (SSBs) and argues that replication-coupled conversion of these lesions into DSBs accounts for a major fraction of genomic instability in HR-deficient settings. In parallel, we also discuss how base damage and resulting increased SSB frequency can be modulated to enhance therapeutic efficacy of Poly (ADP-ribose) polymerase inhibitors (PARPi), which are FDA-approved for the treatment of HR-deficient cancers [11] (Fig. 1b).

Fig. 1.

Endogenous base damage can result in the generation of replication-coupled DSBs. (a) Simplified schematic of homologous recombination at the replication fork. (b) Various cell-intrinsic base damage can introduce single-strand breaks (SSBs). These SSBs are converted to DSBs if a replication fork encounters them before the break is repaired. Additionally, cleavage of abasic site (AP) or fork collapse owing to bulky adducts can result in DSBs at fork head. BRCA-mediated homologous recombination (HR) can repair these lesions. In BRCA-deficient cells, DSBs persist which can lead to increased trapping of Poly (ADP-ribose) polymerase inhibitors (PARPi) at fork head, resulting in increased sensitivity to the drug.

PARPi are believed to function by trapping Poly (ADP-ribose) polymerases 1 and 2 (PARP1 and 2) at DNA breaks [12]. This can be particularly challenging for S-phase cells when the trapped PARP1/2 enzymes lie ahead of the fork, creating an impediment for DNA replication. Upon encountering the trapped PARP1/2 enzymes, forks can collapse to generate DSBs and necessitate BRCA1/2-mediated repair for survival. This synthetic lethal interaction between PARP1 and BRCA1/2-mediated repair underlies the clinical use of PARPi in the regimen of HR-deficient ovarian, breast, and prostate cancers [13]. Despite being a promising targeted therapeutic, clinical success with PARPi is impeded by emergence of resistance and severe side-effects which includes an increased risk of secondary malignancies, such as acute myeloid leukemia (AML) and myelodysplastic syndrome (MDS) [14,15]. These limitations highlight an urgent need to devise approaches to improve the therapeutic efficacy of this targeted therapy. Recent studies uncovered that enhancing base damage can be an effective approach to rapidly kill BRCA-mutant cells with PARPi while reducing the effective drug dose and overcoming resistance [16,17]. The review summarizes these studies and discusses how various forms of base damage can be potentially exploited for therapeutic utility.

2. Sources of endogenous base damage and their repair pathways

2.1. Abasic sites

Amongst the 70,000 lesions human cells encounter per day, approximately 25 % are abasic purinic/pyrimidic sites, hereafter referred to as AP sites [18,19]. AP sites are generated upon the loss of nitrogenous base from the deoxynucleotide triphosphate (dNTP) backbone. While AP sites can be formed by the enzymatic removal of damaged bases by specialized glycosylases, most AP lesions result from non-enzymatic hydrolytic cleavage of the glycosyl bonds. The glycosidic bond in purines is more labile than that of pyrimidines, accounting for a 20-fold excess of apurinic over apyrimidinic sites [20]. Moreover, the frequency of depurination is four-times faster in single-stranded DNA compared to its double-stranded counterparts [21]. Given the increased propensity of replication stress and hence single-stranded DNA in BRCA-mutant cells, it is reasonable to postulate that HR-deficiency is associated with a higher volume of AP sites. AP sites, unlike damaged bases, cannot be replicated via replicative DNA polymerases, posing a challenge to fork progression [22]. HR proteins can potentially contribute to replication-coupled repair of AP sites via at least three distinct modes (Fig. 2a):

1. Replication forks stalled by AP sites can be re-started via PrimPol mediated re-priming, leaving behind a gap that necessitates template switching-mediated repair by HR-proteins [23].

2. AP sites can be cleaved at the replication forks resulting in DSBs and hence portray dependency on HR-repair. Intriguingly, Apurinic/Apyrimidinic Endonuclease1 (APE1) can cleave AP sites even when single-stranded DNA is bound to RPA, making both gaps and fork junctions susceptible to nucleolytic cleavage and DSB generation [24].

3. When present on two opposite strands, AP sites can cross link. PARP1 dependent NEIL3 recruitment can unhook the AP sites mitigating the generation of DSBs [25,26]. In contrast, PARPi abrogates NEIL3 recruitment to AP-crosslinks, resulting in Fanconi Anemia (FA) pathway-dependent strand incision [25], and generation of replication coupled DSB, thereby increasing reliance on HR repair.

Fig. 2.

Unrepaired abasic sites increase reliance on BRCA-mediated repair at replication forks. (a) Abasic sites stall replication forks. These stalled forks either resume by re-priming leading to formation of gaps or they can be processed by APE1 resulting in generation of replication-coupled DSBs. Abasic sites when present across from each other on the two complementary strands can crosslink ahead of the fork. NEIL-3 can unhook the cross-linked AP sites in a PAR-dependent manner. In the presence of PARPi, cross-linked AP sites are processed by the FA pathway resulting in replication-coupled DSBs. Both gaps and replication-coupled DSBs rely on BRCA1/2-mediated repair. (b) Hypothesis for PARPi hypersensitivity in ALC1-deficient BRCA-mutant cells: APE1 has limited accessibility to chromatin buried abasic sites. Nucleosome sliding by ALC1 makes the abasic site accessible to APE1. In ALC1-deficient cells, APE1 cannot access the chromatin buried abasic sites. Abasic sites, however, are readily accessible to APE1 cleavage at replication forks where nucleosomes are evicted. APE1-mediated cleavage of the abasic site at the fork results in the generation of replication-coupled DSBs. PARPi trap PARP1/2 at these breaks resulting in fork collapse and PARPi hypersensitivity.

Recent studies provide mounting evidence for a direct correlation between gap frequency and PARPi response [27]. However, while AP sites result in gap formation [23,28], loss of APE1, the key enzyme involved in the repair of AP sites does not score as a top candidate in conferring PARPi sensitivity in HR-deficient cancers [29]. This could either be due to redundancy in function with the AP lyase role of bifunctional glycosylases or Translesion Synthesis (TLS)-mediated replication of AP sites. The latter seems to be supported by the reported synthetic lethality between HR and TLS-polymerase Rev1 and enrichment of TLS-mutational signatures in BRCA-mutant cells [23,30]. Dispensability of APE1 for PARPi response in HR-deficient cells could also suggest that rather than intact AP sites, cleavage of these lesions and formation of ligation non-compatible broken ends could be a stronger sensitizing factor. Cleavage of AP sites by APE1 generates ligation-incompatible 5′ deoxyribose phosphate which necessitates clearance by either lyase activity of Polymerase ß or by Fen1-mediated 5′ flap cleavage (Table 1). During S-phase, AP sites are believed to be protected from APE1 cleavage by the 5-Hydroxymethylcytosine Binding, ES Cell Specific (HMCES) [31]. Therefore, loss of HMCES could perhaps increase APE1-mediated AP site cleavage, correspondingly increasing PARPi sensitivity in BRCA-mutant cells. It is noteworthy that the cleavage of abasic sites has also been implicated as a key source of DNA damage in micronuclei, a common feature in HR-mutant cancers [32].

Table 1.

Chemical nature of various 5′ blocks, their sources and repair mechanisms. Ligation non-compatible blocks are indicated in red.

Chemical Nature of the 5’ block	Endogenous source	Enzymes involved in removing the block
	ROS	Kinase activity of PNKP converts 5’OH to 5’P
	Ribonucleotide excision repair	Fen1 removes the flap with 5’block during long-patch repair
	APE1 cleavage of abasic site	Polymerase ß lyase activity Fen1 removes the flap with 5’block during long-patch repair

The remarkable potency of targeting abasic site repair as a new therapeutic strategy for HR-mutant cancer was uncovered by several studies which found that Amplified in Liver Cancer 1 (ALC1) is a key contributor to genomic stability and PARPi response in BRCA-deficient settings [16,17,33,34]. ALC1 function is primarily restricted to base excision repair (BER) and is dispensable for both single- and double-strand break repair. Mechanistically, ALC1 is a nucleosome sliding enzyme that makes chromatin accessible to repair factors involved in BER [35–37]. Its role in BER has been reported to be epistatic with APE1, suggesting that it may have a role in generating accessibility of APE1 to chromatin buried abasic sites [16,17]. Indeed, a recent study uncovered that APE1 can more readily process abasic sites on naked DNA, followed by those on solvent-exposed nucleosome-wrapped substrate, and minimally on a solvent-occluded position [38]. However, it is puzzling that loss of ALC1 confers a more profound enhancement in PARPi sensitivity in BRCA-mutant cells in contrast to APE1 depletion. One possible explanation could be that ALC1 is required for recruitment of other abasic site processing enzymes, such as bifunctional glycosylases. Alternatively, unrepaired chromatin bound abasic sites in ALC1-deficient cells could be exposed at replication fork junctions, rendering them accessible to APE1 for cleavage, thereby generating replication-coupled DSBs and increasing reliance on HR for repair (Fig. 2b). Hence, PARPi hypersensitivity in ALC1-deficient cells should be reliant on the ability of APE1 to cleave abasic sites at the forks, resulting in the generation of replication-coupled DSBs. This mechanism would imply that loss of APE1 nuclease should confer PARPi resistance in BRCA-mutant cancer cells. If this mechanism is true, it makes ALC1 a more favorable target because APE1 is essential for survival and hence there would be a minimal probability for the emergence of resistance [39]. Notably, dispensability of ALC1 for multiple DNA repair pathways and transcriptional regulation makes it non-essential for organismal survival, suggesting that targeting ALC1 in HR-deficient cancers is likely to have a larger therapeutic window [16,17]. Very recently, a lead inhibitor of ALC1, EIS-12656, entered Phase I/II clinical trial for PARPi resistant advanced solid tumors. These findings therefore provide a strong foundation for exploiting cleavage of abasic site as an effective approach to improve PARPi efficacy and develop new targeted therapeutics for HR-deficient cancers.

2.2. Oxidized bases

Reactive oxygen species (ROS) represent one of the most frequent sources of SSBs in the genome. Major sources of intracellular ROS include mitochondria, NADPH oxidases, uncoupled nitric oxide synthase, cytochrome P450, peroxidases, endoplasmic reticulum and cyclooxygenases [40]. ROS can also be generated during metabolism of hormones such as estrogen [41]. In addition, oxidized dNTPs can also be incorporated into the DNA by X- and Y-family of polymerases [42]. Oxidized bases are highly mutagenic and can also impede fork progression. The importance of fixing oxidative damage during replication is underscored by a recent study which uncovered the existence of a specialized ROS sensor which is constitutively associated with the replisome. This sensor consists of oligomerized peroxiredoxin 2 protein. Elevated ROS levels result in disruption of the peroxiredoxin 2 oligomer, which in turn results in the dissociation of TIMELESS from the replisome, slowing the fork speed in response to oxidative DNA damage. Intriguingly, peroxiredoxin 2 loss sensitizes cancer cells to replication stress at a higher magnitude compared to their normal counterparts, highlighting that ROS induced damage represents a cancer cell specific vulnerability [43].

Amongst the four nitrogen bases in the DNA, Guanine is highly susceptible to ROS due to its low reduction potential, making 8-oxo-dG the most common product of oxidative damage. Additionally, ROS can result in the generation of pyrimidine lesions, such as FapyA and FapyG, thymine glycol and 5-hydroxycytosine [44]. DNA glycosylases involved in the removal of oxidized DNA bases as OGG1, NTHL1, NEIL1 and NEIL2 are bifunctional, meaning that they can both excise the base and cleave the abasic site generated to form a nick with a 3′ aldehyde, which is then processed into 3′OH by APE1 (Table 2). Unlike OGG1 and NTH1 which only act in the context of double-stranded DNA, bifunctional glycosylases NEIL1 and NEIL2 preferentially excise damaged bases from single-stranded DNA [45], [46]. NEIL1 has been shown to be active in the S-phase of the cell cycle [47]. This raises a possibility that increased oxidative damage can result in excision of the base and perhaps cleavage of the abasic sites at the fork, resulting in replication-coupled DSBs and necessitating HR-mediated repair.

Table 2.

Chemical nature of various 3’ blocks, their sources and repair mechanisms. Ligation non-compatible blocks are indicated in red.

Chemical Nature of the 3’ block	Endogenous source	Enzymes involved in removing the block
	ROS	Phophatase activity of PNKP converts 3’P to 3’OH
	Oxidative damage (sugar disintegration by ROS)	APE1 acts on blunt and recessed 3’ end TDP1 acts on protuding 3’ end to generate 3’ P which is then converted to 3’OH by PNKP
	Oxidative damage (bifunctional glycosylase)	APE1 exonuclease activity
	Top1 clevage at DNA mismatch and base damage	TDP1 hydrolyzes the phosphotyrosyl DNA bond to generate 3’P which is then converted to 3’OH by PNKP APE2 exonuclease activity
	Top1 cleavage of ribonucleotides	Top1 cleaves adjacent to the nick and introduces deletions APE2 exonuclease activity

ROS can also directly react with deoxyribose resulting in the direct disintegration of oxidized sugar and the generation of breaks with ligation non-compatible broken ends (Tables 1,2). For instance, reaction of ROS with ribose at the C4′ and C2′ position results in the generation of 3′ phosphoglycolate and 3′ erythrose respectively [48]. It is imperative that these ligation non-compatible DNA ends are processed in a timely manner. Absence of these end-processing enzymes would result in the formation of replication-coupled DSBs, generating substrates for PARPi trapping and hyper-reliance on HR. This notion is bolstered by the findings that loss of Polynucleotide kinase/phosphatase, PNKP, a key enzyme involved in processing various 3′ and 5′ blocking DNA ends generated upon oxidative damage, hyper-sensitizes cells to PARPi even in the absence of exogenous DNA damage [49]. In a DNA break with 5′OH and 3′ phosphate, the phosphatase activity of PNKP takes precedence [50]. However, it remains unclear whether the 5′ kinase or the 3′ phosphatase activity of PNKP is critical for PARPi response. It would be reasonable to assume that 3′ phosphatase activity would be more critical given that a 5′ block can be alternatively removed as a flap by Fen1 during long-patch base excision repair. In contrast, a similar process is not known for the removal of 3′ blocks. Owing to the essentiality of Fen1 in removing the 5′ block to generate ligation-compatible ends as well as its role in MMEJ, loss of this endonuclease is synthetically lethal with HR-deficiency [51]. Given that ROS levels tend to be higher in cancer cells compared to their normal counterparts, enhancing oxidative damage could be a potential approach to enhance the therapeutic index of PARPi.

2.3. Top1 cleavage complexes (Top1-cc)

Topoisomerase 1 (Top1) enzyme plays a critical role in relaxing DNA supercoils generated during transcription, replication and chromatin remodeling. Top1 achieves this activity via a 2-step process which entails the generation of single-strand breaks, permitting the rotation of double-stranded DNA around the intact strand, followed by ligation of the broken ends. During strand-cleavage, Top1 forms a covalent complex with DNA called the Top1 cleavage complex (Top1-cc) (Table 2). Because the ligation step is much faster than the cleavage step, Top1-cc are rarely detectable under normal conditions. Intriguingly, multiple endogenous lesions have been shown to either enhance Top1 binding or prevent the re-ligation step resulting in “trapping” of Top1-cc on the 3′-end of the DNA [52]. These endogenous lesions include DNA mismatches, various base damage and their repair intermediates, including 8-oxoG, abasic sites and single-strand breaks [53]. Collision of replicative machinery with these lesions can result in fork collapse and DSB generation, resulting in HR reliance[54]. Removal of Top1-cc from the 3′ end is achieved via tyrosyl-DNA phosphodiesterase 1 (TDP1)-mediated cleavage of the covalent bond between the tyrosine residue of Top1 and the 3′ phosphate of DNA. TDP1 activity leaves behind a 3’ phosphate which is processed by PNKP to a 3′hydroxyl (Table 2) [55]. Intriguingly, despite cell intrinsic sources that result in generation of Top1-cc complex, TDP1 does not emerge as a key candidate either as a synthetic lethal factor with HR-deficiency or as a strong sensitizer of PARPi response. This dispensability could perhaps be explained by a role of apurinic/apyrimidinic endonuclease 2 (APE2) as a downstream factor which can clear up the 3′ block in the absence of TDP1 or many other 3′ end processing enzymes [56]. It is noteworthy that APE2 has also been implicated in polymerase theta (POLθ)-mediated MMEJ [57]. Perhaps owing to this broad functionality, APE2 is essential for the proliferation of p53-proficient cells [56]. Of note, APEX2 knockout mice are viable, but they suffer from severe growth and lymphopoiesis defects [58]. Therefore, unlike hindering POLθ, targeting APE2 may not have a broad therapeutic index. Given that a fork collapsed at Top1-cc complex can be a platform of PARP1/2 trapping, enhancing endogenous base lesions that prevent re-ligation step of Top1 can be potentially considered in enhancing the efficacy of PARPi.

2.4. Ribonucleotides

The riboNTP pools in an actively dividing eukaryotic cell is 10–100 times higher than the dNTP pool [59], which increases the probability of replicative DNA polymerases to insert rNMP every few thousand base pairs [60]. This increased frequency of rNMP is further underscored by the embryonic lethality of mice deficient in RNAseH2, a key enzyme that processes rNMP incorporated in genomic DNA [61]. All three components of RNAseH2 complex: RNaseH2A/B/C are therefore constitutively associated with the replisome as a safeguarding mechanism [62]. Nicking of rNMP by RNAseH2 initiates the ribonucleotide excision repair (RER) pathway which results in a 5′ flap with ribonucleotide (Table 1). Fen1 subsequently cleaves the 5′ flap resulting in error-free removal of ribonucleotide in DNA [63]. In the event of defective RNAseH2 activity, Top1 mediated cleavage results in a 3′ DNA end with 2′−3′ cyclic phosphate necessitating further processing to generate ligation compatible DNA ends [64] (Table 2). Notably, the excision of mis-incorporated rNMP will occur behind the replication fork and hence it is unlikely to result in DSB generation due to the replisome encountering a broken template. However, Top1 can also incise the strand opposite to rNMP, resulting in a DSB and necessitating HR-mediated repair (Fig. 3) [65]. It is likely that this dual nicking activity of Top1 underlies synthetic lethality between RNAseH2 loss and BRCA1/2-deficiency [29]. Furthermore, the delayed kinetics of repair of DNA ends with 2′−3′ cyclic phosphate can act as a locus for trapping PARP1/2 enzymes resulting in enhanced PARPi sensitivity regardless of HR-status [29]. Notably, even in the context of a functional RER pathway, ID4, a mutational signature of TOP1 mediated repair can be detected in mammalian cells perhaps reflecting the abundance of rNMP incorporation during replication [66]. Top1-mediated generation of breaks and subsequent PARPi trapping was shown to extend the utility of PARPi in a fraction of chronic lymphocytic leukemia which display collateral loss of RNAseH2B gene. Therefore, the concept of increased base damage intermediates with ligation non-compatible ends can be considered as a potential means to broaden the application of PARPi beyond HR-deficient cancers.

Fig. 3.

Schematic showing how processing of ribonucleotide by Top1 can potentially lead to DSBs. Ribonuclease activity of Top1 at ribonucleotide results in 2′−3 cyclic phosphate. Top1 can subsequently cleavage the opposite strand resulting in a DSBs behind the replication fork.

2.5. Epigenetically modified cytosines

DNA methyltransferases (DNMT)-mediated methylation of deoxycytidine (dC) to form 5′-methyl-2 deoxycytidine (5mdC) is a well-established mechanism for the regulation of gene expression. Removal of 5mdC entails ten eleven translocases (TET) enzyme-mediated conversion of 5mdC to 5-hydroxymethyl deoxycytidine (5hmdC), which is further oxidized in a stepwise manner to 5-formyl deoxycytidine (5fdC) and 5-carboxyl deoxycytidine (5cadC) [67,68]. 5fdC and 5cadC are substrates for thymine DNA glycosylase (TDG) that can excise the modified cytosine resulting in formation of AP sites [69–71]. 5hmdC when present in the context of single-stranded DNA can undergo deamination via AID/APOBEC to form 5hmdU which is excised by single-strand specific monofunctional uracil DNA glycosylase 1, single-strand-selective monofunctional uracil-DNA glycosylase 1 (SMUG1) [72]. It is noteworthy that unlike TDG, SMUG1 can excise hmdU in the context of single-stranded DNA [73]. If excision happens in the front of the replication fork, it will result in an abasic lesion at the fork head. As discussed above, an AP site on template DNA can either stall the fork or can be processed by APE1, resulting in replication-coupled DSBs and increased reliance on BRCA-mediated repair. If the excision occurs on the template DNA in between the unligated Okazaki fragments, this could result in either gaps or DSBs necessitating HR-mediated repair (Fig. 4). A potential way by which cells may mitigate such toxic events is by limiting AID/APOBEC activity, and hence precluding the generation of hmdU from 5hmdC.

Fig. 4.

Schematic depicting how epigenetically modified cytosines can lead to DSBs. Left: Coordination between TET and BER enzymes convert 5mdC back to dC. 5hmdC when present in the context of single-strand DNA (eg. during replication) can be deaminated by APOBEC/AID to 5hmdU which can then be processed by SMUG1 and APE1 resulting in breaks. Right: Salvaged nucleoside pools of 5hmdC can be converted to 5hdUTP which can be incorporated in genomic DNA.

Given that cells recycle nucleotides released from dying cells or other metabolic processes, salvaged nucleosides also contain epigenetic modified cytosines. While 5mdC is salvaged as thymidine, 5hmdC and 5fdC can be deaminated by cytidine deaminase (CDA) to 5-hydroxymethyl deoxyuridine (5hmdU) and 5-formyl deoxyuridine (5fdU), respectively. 5hmdU and 5fdU can be phosphorylated by thymidine kinase (TK1) into monophosphate, followed by thymidylate kinase (DTYMK) to form di- and tri-phosphate analogs, 5hmUTP and 5fdUTP, respectively. Notably, DNA polymerase can incorporate both 5hmUTP and 5fdUTP in the genome [74]. Alternatively, 5hmdC can be phosphorylated by deoxycytidine kinase (DCK) to hmdCMP, which can then be deaminated by deoxycytidylate deaminase (DCTD) to hmdUMP followed by phosphorylation via DTYMK to 5hmdUTP [75]. As discussed above, incorporation of hmdUTP into genomic DNA can be highly toxic given the robust activity of SMUG1 at single-stranded DNA. To avert the toxic incorporation of hmdUTP, cells are equipped with “nucleotide sanitizing” enzyme DNPH1 which hydrolyzes hmdUMP to 5hmdU and decoy ribose phosphate (2-dRP) [76] (Fig. 4). A recent study reported that loss of DNPH1 results in hypersensitization of BRCA-mutant cells to PARPi. This phenotype could be rescued by SMUG1 loss, highlighting abasic sites as a key underlying factor in conferring PARPi hypersensitivity [77]. While targeting DNPH1 was shown to overcome various limitations of PARPi, it would be worthwhile to consider that a loss of function mutation in SMUG1 would be a straightforward means for cancer cells to develop resistance to a potential DNPH1 inhibitor. Furthermore, expression levels of CDA and DCK should be key considerations in targeting DNPH1 as a therapeutic strategy. SMUG1 has also been implicated in the generation of replication-coupled gaps observed in BRCA-mutant cells, suggesting that incorporation of uracil and hmdU is a frequent endogenous event [23]. Hence, exploiting the pool of non-canonical nucleosides in cancer cells can be a potential approach to increase replication-coupled gaps and breaks and thereby enhancing PARPi response in repair-deficient HR cells.

2.6. Aldehydes

Metabolic processes like lipid peroxidation can result in the generation of various forms of aldehydes, including formaldehyde, malondialdehyde, acrolein and crotonaldehyde [78]. Aldehydes typically react with the exocyclic amino group in guanine, cytosine and adenine bases (Fig. 5). In contrast, thymine is less sensitive to these reactions as it lacks an exocyclic amino group. Malondialdehyde reacts with guanine base to form exocyclic pyrimidopurinone called M1dG, which is one of the most abundant exocyclic adducts in humans. Owing to the bulkiness of the M1G, it is repaired by nucleotide- rather than base excision repair [79]. M1dA and M1dC are formed at a much lower level [80]. Acrolein and crotonaldehyde can be metabolized to epoxides which can react with guanine, cytosine and adenine bases to form their respective etheno-derivatives. TDG plays a key role in repairing these base lesions via BER [81]. These exocyclic adducts have been estimated to be 10–5400 per cell and to disrupt Watson-Crick pairing, impeding DNA replication and resulting in generation of DSBs specifically in S-phase cells [82]. Formaldehyde is generated via lipid peroxidation, demethylation of histone H3 lysine 4 by LSD1 [83] and repair of N-linked alkylation adducts m1A and m3C by AlkB family of enzymes [84]. Free formaldehyde can crosslink protein to DNA forming DNA-protein crosslinks (DPC), which are potent impediments for replication fork progression. It is imperative to note that formaldehyde likely forms DPC more readily on single-stranded than double-stranded DNA because the latter would entail first unwinding of the double helix to expose DNA exocyclic amines [85]. Hence, there is an increased likelihood of formaldehyde-induced DPC generation during DNA replication when there is an abundance of single-stranded DNA. The importance of HR protein in the repair of endogenously generated DPC is underscored by the observation that BRCA-deficient cells are hypersensitive to plasma level of formaldehyde [86]. DPC can either result in fork reversal or collapse, both of which necessitate the role of HR proteins for repair. Along similar lines, chemical inhibition of aldehyde dehydrogenase, which detoxifies toxic aldehydes, was shown to enhance replication-coupled damage and kill BRCA-mutant cells that have acquired resistance to PARPi [87]. Intriguingly, formaldehyde can also selectively deplete BRCA2 protein levels by proteasomal degradation, which can be a driver factor in tumorigenesis in cells with BRCA2 haploinsufficiency [88]. Hence, cell intrinsic aldehydes either via generation of nucleotide/base damage intermediates or formation of DPC can result in replication-coupled damage and perhaps also induce BRCAness. It would be worthwhile to examine whether clinically available inhibitors of aldehyde dehydrogenase can synergize PARPi response.

Fig. 5.

Chemical nature of various DNA base adducts generated upon reaction with aldehydes resulting from lipid peroxidation. Regions derived from the aldehydes are indicated in red.

3. Outstanding questions

The review summarizes the potential of targeting endogenous base damage in improving and developing new therapeutic avenues for HR-deficient cancers. Clinical application of this approach still requires addressing several fundamental questions which include: Is the frequency of endogenous base damaging events higher in BRCA-mutant cancer cells? Can we develop sensitive and reliable methods to detect various forms of base damage? How does genome and chromatin reorganization impact the frequency of base damage and single-strand breaks? The advent of multiple advanced technologies including CRISPR genetic screens, deep genome sequencing, quantitative mass-spectrometry, high-resolution imaging, sequencing and powerful computational algorithms to predict mutational signatures provide the foundation to address these long-standing questions at the interface of base damage and HR repair.

Data availability

No data was used for the research described in the article.