Unpaved roads: How the DNA damage response navigates endogenous genotoxins

Abstract

Accurate DNA repair is essential for cellular and organismal homeostasis, and DNA repair defects result in genetic diseases and cancer predisposition. Several environmental factors, such as ultraviolet light, damage DNA, but many other molecules with DNA damaging potential are byproducts of normal cellular processes. In this review, we highlight some of the prominent sources of endogenous DNA damage as well as their mechanisms of repair, with a special focus on repair by the homologous recombination and Fanconi anemia pathways. We also discuss how modulating DNA damage caused by endogenous factors may augment current approaches used to treat BRCA-deficient cancers. Finally, we describe how synthetic lethal interactions may be exploited to exacerbate DNA repair deficiencies and cause selective toxicity in additional types of cancers.

1. Introduction

Ubiquitous DNA damaging events caused by environmental and endogenous genotoxins constantly threaten genome integrity. In turn, cells possess a myriad of pathways that function to recognize and repair DNA lesions to ensure accurate genome duplication in replicating cells. Defective or inaccurate repair of damaged DNA can lead to cell death, or in other instances, malignant transformation and tumorigenesis. Hereditary mutations in several DNA repair genes are also responsible for wide ranging diseases, including premature aging, immunological deficiencies, neurodegeneration, and intellectual disability [21].

Intensive study of cellular responses to exogenous genotoxic insults such as ultraviolet light, ionizing radiation (IR), and chemotherapeutic agents have led to a detailed understanding of the genomic lesions that arise and the pathways that function to repair such damage. However, how these same pathways function to resolve endogenous clastogenic insults is less well understood. A greater understanding of the endogenous sources of DNA damage is emerging. Replication-transcription conflicts, repetitive DNA sequences that form secondary structures, nucleotide misincorporation, base-modifying metabolites and reactive crosslinking aldehydes all constitute bona fide sources of DNA damage that invoke requirements for specific DNA repair mechanisms (Fig. 1A). In this review, we summarize the current understanding of endogenous sources of damage with a special focus on how homologous recombination (HR) and Fanconi anemia (FA) DNA repair networks contribute to their resolution. We highlight how repair of endogenous sources of DNA damage has contributed to our understanding of disease etiology, both of genetic disorders and cancer development, while also pointing to new therapeutic strategies to enhance regenerative processes and effectively treat cancer.

Fig. 1.

Endogenous sources of DNA damage and mechanisms of repair. A. Endogenous sources of DNA damage include genomic incorporation of enzymatically or chemically modified nucleotides (denoted by square with red star), which when repaired, generate abasic sites and single-strand breaks (SSBs). Removal of genomic ribonucleotides also generates SSBs. Reactive aldehydes generate inter-strand crosslinks (ICLs) and base damage, and repetitive sequences and R-loops can interfere with replication fork progression. B. Base excision repair (BER) removes enzymatically or chemically modified nucleotides from the genome. The resulting SSBs are bound by PARP1, which promotes their repair. PARP inhibitors (PARPi) covalently trap PARP1 on chromatin, and this causes cell death in HR-deficient cancer cells. ALC1 deficiency perturbs chromatin remodeling and causes an accumulation of unrepaired BER intermediates, increases PARP trapping, and enhances PARPi toxicity in HR-deficient cells. DNPH1 or ITPA loss causes increased genomic incorporation of modified bases, and the resulting increase in SSBs generated by BER potentiate PARP trapping and enhance PARPi toxicity in HR-deficient cells. C. Reactive aldehydes damage DNA by forming ICLs. Proper ICL repair requires the Fanconi anemia (FA) pathway, but in FA-deficient cells, ICLs are converted into DSBs. Subsequent DSB repair by non-homologous end-joining (NHEJ) or microhomology-mediated end joining (MMEJ) frequently results in mutations (denoted by red stars), increased genome instability, and in FA patients, hematopoietic stem cell (HSC) dysfunction and bone marrow failure. D. In cancer cells with microsatellite instability (MSI), expansion of TA-dinucleotide repeats causes formation of non-B form DNA that stalls replication forks. Stalled forks collapse into double-strand breaks (DSBs) in WRN-deficient cells, causing cell death. E. R-loops form at sites of transcription, DSBs, and replication-transcription conflicts. Multiple DNA repair proteins, including several homologous recombination (HR) and FA factors ensure proper R-loop resolution, thereby maintaining genome stability.

2. Base damage as a vulnerability in BRCA mutant cells

Of the estimated 70,000 DNA lesions that occur daily in human cells, the majority are constituted by DNA single-strand breaks (SSBs) and abasic sites [36]. While these lesions can be caused by IR or reactive oxygen species (ROS) generated during cellular metabolism, SSBs and abasic sites are also intermediate structures generated during base excision repair (BER) [7]. The poly(ADP)ribose polymerase enzymes, PARP1 and PARP2, function as SSB sensors and are rapidly recruited to damage sites. Upon DNA binding, PARP1 and PARP2 enzymatic activity increases several thousand-fold to polymerize nicotinamide adenine dinucleotide (NAD+) into poly(ADP)ribose (PAR) chains [78]. These branched structures serve to recruit proteins with PAR binding domains, which both decondense chromatin and repair the underlying genomic lesion. PAR associated functions are mediated by recruitment of the chromatin remodeling enzyme ALC1 and by SSB repair (SSBR) factors XRCC1 and DNA Ligase III (Lig3). Accurate SSBR also depends on PARP1/2 auto-PARylation, which mediates timely PARP1/2 dissociation from damaged chromatin [19]. In addition, PARP1 dissociation from damaged chromatin depends on XRCC1, and in XRCC1-deficient cells, trapped PARP1 inhibits BER [28]. This defective PARP1 regulation underlies the cerebellar neurodegeneration and ataxia caused by hereditary XRCC1 mutations as these symptoms can be reversed upon PARP1 deletion in XRCC1-mutant mice [48]. Cerebellar ataxias also result from mutations in the XRCC1-interacting proteins tyrosyl-DNA phosphodiesterase (TDP1), Aprataxin (APTX), and Polynucleotide kinase-phosphatase (PNKP), all of which function in SSBR [104,22,80]. The neurodegenerative symptoms caused by defects in these genes suggest proper SSBR is essential for neuron function, and indeed, recent evidence indicates post-mitotic neurons accumulate high levels of SSBs that are repaired in a PARP1- and XRCC1-dependent manner [105].

Aberrant PARP1 chromatin retention is deleterious in the context of inherited neurodegenerative disorders, but PARP trapping can be exploited for therapeutic gain in other instances such as to achieve synthetic lethality in BRCA-mutant cancers. However, because patients with BRCA-mutant tumors frequently develop PARP inhibitor (PARPi) resistance, developing strategies to augment PARPi responses remains paramount. Data from several genetic screens indicates that defective BER potentiates PARP trapping and may improve PARPi efficacy. Loss of 2’-deoxynucleoside 5’-monophosphate N-glycosidase (DNPH1) and Inosine triphosphatase (ITPA) were identified in a whole-genome CRISPR screen to hypersensitize BRCA-deficient cells to PARPi. Both enzymes modulate BER by reducing genomic incorporation of abnormal bases (Fig. 1B). DNPH1 limits genomic incorporation of the cytotoxic nucleoside 5-hydroxymethyl deoxyuridine (hmdU) by hydrolyzing hmdU monophosphate (hmdUMP). This base metabolite is produced following Ten Eleven Translocation (TET)-mediated hydroxylation of 5-methylcytosine (5mC) and subsequent degradation of DNA that contains 5-hydroxymethylcytosine (5hmC). DNPH1 loss causes a three-fold increase in genomic hmdU incorporation, and subsequent removal of hmdU by the SMUG1 glycosylase results in increased PARP trapping on chromatin and replication fork collapse in HR-deficient cells [35] SMUG1 function also generates ssDNA gaps in BRCA-deficient cells in a PRIMPOL-dependent manner [93], suggesting a DNA replication associated re-priming event downstream of SMUG1 action contributes to PARPi sensitivity. Consistently, SMUG1 loss eliminates the PARPi hypersensitivity of DNPH1- and HR-double deficient cells commensurate with a reduction in ssDNA gaps, which are thought to trap PARP1 in BRCA-deficient cells [35,93].

A similar mechanism of synthetic lethality likely also explains the PARPi hypersensitivity of HR-and ITPA-double deficient cells. ITPA limits genomic incorporation of deoxyinosine(dI) by dephosphorylating deoxyinosine triphosphate (dITP) [60], and genomic dI is removed by the N-methylpurine DNA glycosylase (MPG) [3]. Thus, increased genomic dI incorporation upon ITPA loss followed by excision by MPG would increase levels of PARP trapping lesions in an analogous fashion to SMUG1 action on incorporated hmdU.

The aforementioned whole-genome CRISPR screen and a separate CRISPR screen targeting chromatin remodelers also found that ALC1 loss enhances PARPi sensitivity of HR-deficient cells [100,12,35,46] (Fig 1B). ALC1 binds tightly to PARylated nucleosomes using a carboxy-terminal macrodomain, allowing its amino-terminal imitation switch (ISWI) like ATPase/helicase domain to slide nucleosomes and decondense chromatin. ALC1 deficiency prevents chromatin decondensation at DNA damage sites and access of base damage repair factors. Consequently, the unrepaired BER intermediates in ALC1-deficient cells lead to increased ssDNA gaps and PARP trapping, and ultimately cause irreparable DNA damage in BRCA-deficient cells [100,12,46]. Similar to the resistance mechanism associated with DNPH1 loss, the PARPi hypersensitivity of BRCA- and ALC1-double deficient cells is partially suppressed by loss of MPG, which in addition to removing dI, also excises alkylated bases [46]. Future studies will be required to determine how alkylated bases, as opposed to other forms of base damage, cause synthetic lethality in BRCA- and ALC1-double deficient cells. A second question of interest is what are the metabolites that create endogenous base adducts that ALC1 responds to.

In addition to damaged DNA bases, genomic misincorporation of ribonucleotides also threatens chromosomal integrity and can modulate PARPi sensitivity of BRCA-deficient cells. Ribonucleotide incorporation is the most commonly occurring endogenous lesion in dividing cells; thus, targeting ribonucleotide excision repair (RER) may have therapeutic value when it comes to treating HR-deficient cancers. RER is catalyzed by RNASEH2, which cleaves the DNA backbone on the 5’ side of a misincorporated rNMP to create an SSB. Subsequent strand displacement synthesis by Pol δ, flap removal by FEN1 or EXO1, and ligation repairs the genomic template [52]. However, in RNASEH2-deficient cells, ribonucleotides are processed in a topoisomerase 1 (TOP1)-dependent manner, and the resulting SSBs increase PARP trapping and underlie the PARPi hypersensitivity of BRCA- and RNASEH2-double deficient cells [110]. Removal of genomic ribonucleotides also occurs in an APE2-dependent manner, and not surprisingly, APE2 loss is synthetic lethal with BRCA1/2 mutation, purportedly due to an accumulation of 3’ blocked DNA ends, which require HR for accurate repair [4] (Fig. 2A).

Fig. 2.

DNA repair pathways that are required for viability in HR-deficient cells. A. Genomic ribonucleotides (denoted by red square) are primarily removed by ribonucleotide excision repair (RER), and to a lesser extent, by topoisomerase 1 (TOP1)-mediated cleavage. TOP1 cleavage generates SSBs with 3’ blocks (denoted by triangle), which are removed by APE2. In RNASEH2-deficient cells, ribonucleotide removal by TOP1 causes a large increase in SSBs that are bound by PARP1, which when combined with PARPi treatment, enhances PARPi toxicity in HR-deficient cells. In APEX2-deficient cells, 3’ blocks generated by TOP1 cleavage are not removed by APE2, which eventually causes replication fork collapse and cell death in HR-deficient cells. B. DSBs that are resected in HR-deficient cells rely on Polθ-dependent microhomology mediated end-joining (MMEJ) for repair. DSB repair by MMEJ is mutagenic, but promotes survival of HR-deficient cells. Polθ inhibition blocks MMEJ, causing toxicity in HR-deficient cells. C. BRCA-deficient cells contain ssDNA gaps due to incomplete DNA replication that if not stabilized, cause chromosome segregation errors in mitosis. CIP2A, in complex with TOPBP1, binds under-replicated DNA in mitotic BRCA-deficient cells to tether chromatid fragments. In the absence of CIP2A, chromatid breakage causes cell death.

Functional RER is also essential for genome stability in non-cancerous cells as inherited mutations in all RNASEH2 subunits (RNASEH2A, RNASEH2B, and RNASEH2C) cause the autoimmune disorder Aicardi-Goutieres syndrome (AGS), which is characterized by diverse symptoms that include physical and intellectual disability, microcephaly, seizures, and skin rashes. However, the most reliable indicator of AGS is increased expression of interferon-stimulated genes (ISGs) in the blood, which occurs in almost all patients at all stages of life [26]. Perpetual ISG expression in AGS patients with RNASEH2 mutations is due to cGAS-STING activation, which is likely stimulated by accumulation of damaged cytoplasmic DNA that contains embedded ribonucleotides [63,45,64]. Currently, it remains unclear how, despite being ubiquitously expressed, RNASHE2 mutations cause selective central nervous and immune system dysfunction. Such tissue specificity may reflect a higher level of DNA damage caused by genomic ribonucleotides in these cell types.

The increased PARPi sensitivity of BRCA-deficient cells caused by genomic incorporation of modified deoxynucleotides, or ribonucleotides, has been attributed to DSBs that arise when replication forks collide with repair intermediates where PARP is trapped [30]. However, several recent studies have concluded that ssDNA gaps, not DSBs, are the primary lesions that mediate PARPi cell killing [24,31,76]. ssDNA gaps occur spontaneously in BRCA-deficient cells [93], and in response to DNA damaging agents, including PARPi, this phenotype is exacerbated due to unrestrained replication fork progression [24,76]. In addition, ssDNA in BRCA-deficient cells treated with PARPi can be further increased upon ALC1 loss [100], indicating that base damage contributes directly to gap formation. However, excessive base damage also contributes to DSB formation as DNPH1 loss increases DSBs in BRCA2-deficient cells treated with hmdU and/or PARPi. BRCA2-, DNPH1-double deficient cells undergo apoptosis upon prolonged PARPi treatment, and when apoptosis is inhibited, DSB formation increases [35]. This data suggests, at least in the context of DNPH1-deficiency, that apoptosis of PARPi-treated BRCA2-deficient cells is due to DSBs. Conversely, another study found that DSBs in cisplatin-treated BRCA2-deficient cells are dependent on apoptosis, and that when apoptosis is inhibited, DSBs do not form [76]. Clearly, whether DSBs or ssDNA gaps are the primary cause of the PARPi sensitivity of HR-deficient cells warrants further investigation, as discriminating between these possibilities may allow more precise predictions of patient treatment responses.

3. Reliance on secondary repair pathways for viability in BRCA-deficient cells

HR-deficient cells depend not only on accurate repair of base damage for survival, but proper functioning of additional DNA repair pathways as well. One such pathway is microhomology mediated end-joining (MMEJ), also known as alternative end-joining (Alt-EJ). Once thought to be solely a backup pathway that functioned in the absence of HR or NHEJ, it is now clear that proper MMEJ is required for genome stability in cells without any other DSB repair defects [14]. Like HR, DSB repair by MMEJ is dependent on end-resection. Subsequent annealing of DNA strands that exhibit microhomology followed by fill-in synthesis, ligation, and flap removal completes repair. DNA polymerase theta (Polθ) mediates MMEJ fill-in synthesis [66], and Polθ depletion enhances the cisplatin, mitomycin C (MMC), and PARPi sensitivities of FANCD2-deficient cells and the PARPi sensitivity of BRCA1-deficient cells [16]. In vivo, although the embryonic lethality associated with combined FANCD2 and Polθ loss is not complete, FANCD2−/− Polθ−/− pups are born with severe congenital abnormalities and die shortly after birth [16]. Consistent with the genetic data, a recently developed Polθ inhibitor, ART558, impairs the survival of BRCA-deficient cells (Fig. 2B). Interestingly, ART558 also impairs the growth of PARPi resistant BRCA1-deficient cells that lack 53BP1 or SHLD2 expression by causing excessive resection [107]. Reduced end resection causes resistance to ART558, implicating persistent ssDNA levels as a cause of Polθ inhibitor toxicity. Another Polθ inhibitor, novobiocin (NVB), was also found to be synthetic lethal with HR-deficiency, to synergize with PARPi, and to overcome acquired PARPi resistance in a patient-derived xenograft (PDX) model [109]. Thus, Polθ inhibition may be a useful treatment strategy for not only BRCA-deficient cancers, but also for BRCA-deficient cancers with acquired PARPi resistance.

The ssDNA gaps that arise in BRCA-deficient cells, if not repaired in interphase, must be stabilized in mitosis. Cells with incompletely replicated DNA that enter mitosis activate a repair pathway termed mitotic DNA synthesis (MiDAS), which as the name implies, results in mitotic DNA replication, often at difficult to replicate sequences such as common fragile sites (CFS) [68]. MiDAS is dependent on TRAIP, which promotes CMG unloading [29,88], and on RAD52, which mediates recruitment of the MUS81 endonuclease and POLD3 to CFS for repair. However, MiDAS does not require BRCA2 or RAD51 [11], in agreement with their inactivation during mitosis. Accordingly, MiDAS is active in HR-deficient cells, but current data indicates MiDAS disruption is not synthetic lethal with HR-deficiency [1]. Instead, under-replicated DNA that undergoes mitotic breakage in HR-deficient cells is stabilized by CIP2A. CIP2A, together with TOPBP1, localizes to damaged DNA in HR-deficient mitotic cells to tether broken chromatids and prevent mis-segregation of acentric chromatid fragments and cell death (Fig. 2C). CIP2A mitotic localization to DNA lesions relies on its coiled-coil domain, possibly indicating that oligomerization and/or high-order assembly of CIP2A occurs during mitotic chromatid tethering [1]. Given that TOPBP1 liquid-liquid condensates potentiate ATR signaling [34], it is plausible phase separation of CIP2A-TOPBP1 complexes may prevent chromatid mis-segregation. Future studies will also need to address how CIP2A-TOPBP1 mitotic localization to damaged chromatin occurs. While it was previously shown that TOPBP1 localizes to DSBs in mitosis through an interaction with MDC1 to maintain chromosomal stability [56], mitotic localization of CIP2A-TOPBP1 complexes to damaged DNA occurs independently of MDC1 [1].

Cells with defects in BRCA1 function also rely on alternative HR pathways for survival. In BRCA1 heterozygous cells, RNF168, which canonically functions in the chromatin ubiquitination cascade that recruits 53BP1 to DSBs, promotes PALB2 recruitment to resected DSBs and subsequent RAD51 loading. RNF168 is also essential for RAD51 recruitment and HR in BRCA1- and 53BP1-double deficient cells [111]. Similarly, a previous study found that BRCA1-deficient cancer cells exhibit residual RAD51 chromatin recruitment in response to DNA damage that is dependent on PALB2; however, the function of RNF168 in this setting is not clear. Residual RAD51 foci formation in BRCA1-deficient cells is also highly dependent on ATR, as ATR inhibition reduces RAD51 foci formation in BRCA1-deficient cells to a greater extent than in BRCA1-proficient cells [106]. ATR phosphorylates PALB2 in response to DNA damage [2,15], and as such, may be essential for PALB2 and RAD51 recruitment in BRCA1-deficient cells.

4. Transcriptional stress/R-loops

Nuclear RNA species such as R-loops can threaten genome integrity. R-loops are formed when nascent RNA hybridizes to the template DNA strand and displaces the non-template DNA strand to generate single-stranded DNA (ssDNA). R-loop formation regulates gene expression, transcription termination, and DSB repair, but aberrant formation or resolution of these structures causes DNA damage and disease [25]. Several DNA repair proteins facilitate R-loop resolution, including components of the Fanconi anemia (FA) pathway. FANCD2 foci formation in response to MMC treatment is largely R-loop dependent [40], and FANCM translocase activity is required for R-loop resolution [82]. Consistent with these results, FANCA−/− and FANCD2−/− patient cells have increased R-loop levels. Furthermore, overexpression of RNASEH1, the enzyme that resolves RNA:DNA hybrids, significantly attenuates the DNA damage observed in FANCD2-depleted cells [82]. More recent data indicates the FANCD2-FANCI complex directly binds the ssDNA of R-loops and that FANCD2 recruits the RNA processing enzymes hnRNPU and DDX47 to promote R-loop resolution [57,72]. Altogether, this data suggests R-loop resolution is a primary function of the FA pathway and that dysregulation of this process likely contributes to FA disease manifestation (Fig. 1E).

Like the Fanconi proteins, canonical HR proteins BRCA1, BRCA2, and RAD51 regulate R-loop resolution (Fig. 1E). Both BRCA1 and BRCA2 modulate R-loop formation through an association with RNA Polymerase II (RNAPII). By regulating RNAPII promoter proximal pausing, BRCA1 and BRCA2 prevent aberrant R-loop formation and DNA damage at actively transcribed genes [85,108]. TREX-2-dependent BRCA2 recruitment to transcription sites is also important for proper R-loop resolution [10]. At transcription termination sites, BRCA1 limits R-loop formation by recruiting the RNA-DNA helicase Senataxin (SETX), and this prevents SSBs and an accumulation of mutations [44]. SETX hereditary mutations cause the neurodegenerative disorders ataxia oculomotor apraxia type 2 (AOA2) and amyotrophic lateral sclerosis 4 (ALS4) [8,42], and SETX dysfunction likely also enhances the genome instability in BRCA1-deficient cancer cells. This may also partly explain the cell-type specificity of BRCA1-hereditary breast cancers, the majority of which arise from luminal progenitor cells [59,69]. Data from BRCA1-mutant patient samples indicate that R-loops are elevated in breast luminal epithelial cells, but not in basal epithelial or stromal cells [108]. Furthermore, elevated estrogen levels, which correlate with increased breast cancer predisposition, have been shown to cause R-loop-dependent DSBs [91]. Estrogen also regulates recruitment of DNA repair proteins to promoters of estrogen-responsive genes [50]; therefore, altered gene expression patterns due to deregulated estrogen levels may contribute to R-loop-dependent genome instability in a subset of BRCA-mutant cancers as well.

Cockayne syndrome (CS) is an inherited neurodegenerative disease that also results in extreme photosensitivity and premature aging. The gene most often mutated in CS is ERCC8, which encodes the CSB protein [89]. CSB canonically functions in transcription-coupled nucleotide excision repair (TC-NER) and also promotes R-loop resolution by recruiting RAD51 to actively transcribed genes. R-loops induced at transcription sites by reactive oxygen species (ROS) are bound directly by CSB, and through an interaction with RAD52, CSB mediates RAD51 loading onto the displaced ssDNA to promote R-loop resolution and transcriptional fidelity. Interestingly, this function of RAD51 does not depend on BRCA1 or BRCA2 [95]. Although R-loop resolution is an important function of BRCA1, R-loop accumulation has also been shown to reduce BRCA1-mediated repair [41]. As such, R-loop processing by the CSB-RAD52-RAD51 pathway may be especially important for limiting genome instability in BRCA1-deficient cancers.

Excessive R-loop formation at transcription sites can be deleterious, but properly regulated R-loop formation at DSBs facilitates accurate repair. In fission yeast, RNAPII recruitment to and subsequent transcription at DSBs prevents intrachromosomal recombination and loss of repetitive sequences. Promoter driven gene transcription by RNAPII is silenced in an ATM dependent manner on chromatin that is contiguous with DSBs [83]. However, promoter independent RNAPII-driven transcription emanating from DSBs may also promote resection via chromatin remodeling [71]. In human cells, RNAPII-dependent damage-induced long non-coding RNAs (dilncRNAs) synthesized at DSBs hybridize with the resected ssDNA and are recognized by BRCA1. Subsequent BRCA2-dependent RNASEH2 recruitment induces R-loop resolution [27]. SETX localization to R-loops at DSB sites also facilitates accurate repair, in part by increasing RAD51 recruitment [23]. Downstream of end resection, R-loop formation at DSBs promotes repair by enhancing RAD51-strand invasion and identification of homologous sequences on the undamaged sister chromatid [73]. Pre-existing RNA transcripts also promote accurate DSB repair, especially at sites of active transcription [6]. One previous study found that classical non-homologous end-joining (C-NHEJ) factors utilize pre-formed RNA to promote error-free DSB repair [17], while another found that in budding yeast, RNA-DNA recombination promotes DSB repair in cells deficient in both RNaseH1 and RnaseH2 [53].

5. Reactive aldehydes

Byproducts of cellular metabolism, such as reactive aldehydes, also threaten genome integrity. Reactive aldehydes are ubiquitous molecules that readily react with and damage DNA. Lipid peroxidation generates reactive aldehydes in cells, and environmental sources include cigarette smoke and alcohol consumption [70]. The human carcinogen formaldehyde is produced upon consumption of the artificial sweetener aspartame [96], and formaldehyde is also produced endogenously by oxidative demethylation reactions [84]. In the nucleus, histone and nucleic acid demethylation generates formaldehyde that can crosslink DNA [103,61,84].

The FA pathway has key functions in resolving R-loops, but genetic data from animal studies indicate that resolution of reactive aldehyde-induced DNA damage is another essential function of FA proteins. When acetaldehyde catabolism is disrupted in mice via aldehyde dehydrogenase 2 (ALDH2) knockout, FANCD2 becomes essential for proper embryonic development and hematopoiesis. Additionally, ALDH2−/− FANCD2−/− mice develop acute leukemia at high frequency, while ALDH2−/− mice do not [54]. ALDH2−/− FANCD2 −/− mice display elevated DNA damage in hematopoietic stem and progenitor cells and a 600-fold decrease in the hematopoietic stem cell (HSC) pool [37]. Further analysis of DNA damage in HSCs from ALDH2−/− FANCD2−/− mice revealed that error-prone end joining repairs DSBs and that mutagenic DNA repair in HSCs likely contributes to the bone marrow failure observed in FA [38] (Fig. 1C). Recent biochemical data also indicates acetaldehyde inter-strand crosslinks (ICLs) can be repaired by a REV1-dependent, FA-independent pathway in which the crosslink itself is cleaved [49]. Although this pathway is mutagenic, it does not cause DNA breakage, and thus may limit gross chromosomal rearrangements caused by error-prone end joining in FA-deficient cells.

In addition to generating ICLs, acetaldehyde also causes base damage when it reacts with DNA with the most common lesion being N²-ethyl-deoxyguanosine (N²-ethyl-dG) [92]. Although it is not known how N²-ethyl-dG is tolerated or repaired, this adduct accumulates in ALDH2-deficient mice upon prolonged alcohol consumption and is believed to contribute to tumorigenesis [67]. Acetaldehyde also converts dG to crotonaldehyde-derived propano-dG (CrPdG) and etheno-dG (NεdG), both of which are mutagenic. CrPdG inhibits DNA synthesis and increases the frequency of G to T transversion mutations [90]. NεdG is produced upon acetaldehyde induced lipid peroxidation and inhibits DNA synthesis by Pol δ, although NεdG adducts can be bypassed by translesion polymerases [20,39]. In vitro experiments indicate NεdG is most readily bypassed by Pol η, which incorporates G opposite the adduct [20]. Thus, in cells, NεdG likely results in increased G to C transition mutations. High resolution mass spectrometry analysis of DNA isolated from oral cells of volunteers who consumed alcohol also revealed that acetaldehyde reacts with deoxyadenosine (dA) and deoxycytidine (dC) to produce N⁶-ethyldeoxyadenosine and N⁴-ethyl-deoxycytidine respectively [43]. The same study identified 22 DNA adducts total induced by acetaldehyde, indicating that alcohol metabolism generates numerous distinct DNA lesions.

Endogenous formaldehyde is metabolized by alcohol dehydrogenase 5 (ALDH5), and in ALDH5 knockout cells, FANCD2 is similarly required for proper HSC function. Interestingly, in ALDH5−/− FANCD2−/− mice, not only is HSC function compromised, but liver and kidney functions are as well, indicating that repair of formaldehyde-induced DNA damage by the FA pathway maintains systemic metabolic homeostasis [81]. Proper formaldehyde catabolism is also essential for maintenance of BRCA2 expression [94]. Exogenous formaldehyde induces proteasome-dependent BRCA2 degradation. In BRCA2 heterozygous cells, BRCA2 loss caused by formaldehyde treatment induces replication fork degradation and genome instability. The formaldehyde-induced replication stress in BRCA2 heterozygous cells is likely due to aberrant R-loop formation as RNASEH1 overexpression suppresses the chromosomal abnormalities that would otherwise form [94]. Thus, formaldehyde not only induces DNA damage directly via generation of crosslinks, but also indirectly by decreasing BRCA2 protein levels. Such changes in BRCA2 expression due to formaldehyde exposure may contribute to cancer development in individuals with inherited BRCA2 mutations.

6. Repetitive sequences

A large portion of the human genome is composed of repetitive elements, including satellite repeats and telomeric sequences. Due to their propensity to adopt complex secondary DNA structures, repetitive sequences can stall replication forks [97] and require specialized DNA helicases and/or translocases for their resolution. This requirement is the basis for emerging data that repetitive DNA places unique requirements on the DNA damage response that can be exploited as synthetic lethal relationships. Prolonged fork stalling at repetitive sequences can lead to fork collapse and genomic instability, and inappropriate expansion of repeat sequences causes diseases such as Fragile X syndrome, Huntington disease, and Lynch syndrome [65].

Lynch syndrome, also known as hereditary non-polyposis colorectal cancer (HNPCC), results from inherited mutations in mismatch repair (MMR) genes and is characterized by microsatellite instability (MSI) [13]. Somatic mutations in MMR genes can also cause colon, gastric, ovarian, and endometrial MSI cancers. Like the synthetic lethality observed in BRCA-deficient cells treated with PARPi, MMR-deficient MSI cancer cells require function of the WRN helicase for survival [18,51,58,9]. Expansion of TA-dinucleotide repeats in MSI cancer cells causes accumulation of non-B form DNA that stalls replication forks, and WRN helicase activity is required for resolution of these structures. In the absence of WRN, MUS81-dependent cleavage of the expanded repeats causes chromosome breakage and apoptosis [99] (Fig. 1D). Although MSI cancers frequently exhibit sensitivity to immune checkpoint inhibitors due to high neoantigen levels [55], WRN inhibition may emerge as an alternative treatment strategy in instances where immune checkpoint blockade is ineffective.

Telomeres, the DNA at the end of linear chromosomes, also contain repetitive DNA sequences and specifically, the telomeric repeat sequence TTAGGG. This, combined with the propensity of telomeres to form t-loops and G-quadraplexes, causes replication stress. The telomeres of dividing cells normally shorten with each cell cycle, but in cancerous cells, telomere length is maintained due to telomerase expression, or a recombination-based mechanism known as Alternative Lengthening of Telomeres (ALT). Cancer cells that rely on ALT for telomere maintenance exhibit increased DNA damage at telomeres, and as such, are more vulnerable to telomere-specific replication stress than their telomerase-expressing counterparts.

FANCM, which canonically functions to recruit the FA core complex to DNA inter-strand crosslinks (ICL), also functions during ALT to suppress telomeric replication stress [62,74,87]. Association between FANCM and the FA core complex is not required for telomere maintenance in ALT cells [62]. Instead, telomere maintenance in ALT cells depends on FANCM translocase activity, the interaction between FANCM and its obligate binding partners FAAP24, MHF1, and MHF2, and an association between FANCM and the BLM-TOP3A-RMI (BTR) complex [62,74]. The FANCM-BTR interaction limits BLM accumulation at telomeres, and in the absence of FANCM, excess BLM telomere localization contributes to ALT cell death. It is unclear if excessive ALT or an aberrant form of ALT that creates toxic recombination intermediates is the cause of death in FANCM-deficient cells; however, current data indicates R-loop accumulation plays a role as RNASEH1 overexpression reduces telomeric replication stress in both FANCM-depleted and FANCM and BLM co-depleted cells [87].

Telomeric R-loops are produced in ALT cells upon RNAPII-dependent transcription of the long-noncoding RNA species TERRA. TERRA is important for chromosome end protection and maintenance of telomere length in ALT cells, but increased TERRA-dependent R-loops cause replication stress and genome instability [5,86]. BRCA1 regulates TERRA R-loop formation by directly binding to TERRA RNA, and this interaction mediates BRCA1 localization to TERRA promoters, which reduces TERRA R-loop levels [101]. Interestingly, RAD51 binding to TERRA RNA has the opposite effect and facilitates TERRA R-loop formation [33]. Such differences likely reflect the proper balance between TERRA R-loop formation and resolution needed for ALT cell viability.

7. Conclusions/future directions

The importance of functional DNA repair mechanisms is underscored by the myriad of genetic diseases that result from mutations in DNA repair genes. As discussed above, defects in SSBR genes frequently result in developmental neurodegenerative disorders, but recent data also indicates functional SSBR is essential for proper function of mature neurons [105]. Given that the human brain consumes approximately 20% of bodily oxygen, accurate repair of oxidative damage likely protects against age-related cognitive decline. A recent study found that Histone deacetylase 1 (HDAC1) stimulates genomic 8-oxoguanine (8-oxoG) removal by OGG1 in the brain and that disruption of this process causes neuronal DNA damage and cognitive impairment in aged mice. Spontaneous loss of HDAC1 activity and increased 8-oxoG is also observed in the brains of Alzheimer’s disease (AD) mice, but pharmacologic HDAC1 activation reverses these effects and improves cognition [77]. As such, future development of AD treatments capable of reactivating neuronal DNA repair pathways may prove efficacious in slowing disease progression.

Modulating SSBR is also likely to have therapeutic value for the treatment of BRCA-deficient cancers. Although PARPi often effectively treat BRCA-mutant tumors, acquired or inherent PARPi resistance has become increasingly common. Inhibiting the nucleotide sanitizers DNPH1 and ITPA or the chromatin remodeler ALC1 may augment PARPi treatment regimens and overcome known mechanisms of PARPi resistance by creating toxic BER intermediates that potentiate PARP trapping [100,12,35,46]. Why certain BER intermediates appear to be more potent PARP trappers than others warrants further investigation. Additionally, given the tissue-specificity of BRCA-mutant cancers, identifying the most prevalent tissue-specific base lesions may lead to the development of more targeted therapies in the future.

Augmenting endogenous base damage to enhance PARPi sensitivity of BRCA-deficient cancers may further increase the susceptibility of these cancers to immunotherapies as well. PARP inhibition induces T-cell infiltration in BRCA1-deficient models of triple negative breast cancer (TNBC) and high grade serous ovarian cancer (HGSOC) [32,75], and PD-1 blockade enhances this effect [32]. In addition, CTLA-4 blockade was shown to synergize with PARPi in BRCA1-deficient ovarian cancers [47]. As such, increased neoantigen presentation due to elevated base damage may potentiate such combinatorial treatments.

MSI is frequently observed in multiple types of cancers. Compared to MSS tumors, MSI cancers are less likely to become metastatic, but generally respond worse to traditional chemotherapies [98]. The identification of WRN as a synthetic lethal target of MSI cancers represents a novel treatment strategy, especially since WRN depletion causes toxicity in MSI cancer cells that are resistant to either chemotherapy or immunotherapy [79]. However, given that human cells contain multiple RecQ helicases that exhibit structural similarity, optimizing WRN specific inhibitors and assessing any off-target effects in non-cancerous cells will be an important step before the promise of WRN inhibition is realized in clinical settings [98].

Identifying genes that are synthetic lethal with the ALT pathway represents yet another strategy for maximizing tumor cell killing while limiting non-cancer cell death. Multiple studies have identified FANCM as an ALT synthetic lethal target, but development of a FANCM-specific inhibitor has yet to materialize. Targeting FANCM protein-protein interactions to disrupt its function may represent an alternative strategy as overexpression of the FANCM MM2 domain, which blocks the interaction of endogenous FANCM and BLM, is selectively toxic to ALT cells [62]. Similarly, the small molecule PIP-199, which disrupts the FANCM-BLM interaction [102], is also toxic to ALT cells [62].

In summary, proper repair of DNA damage caused by endogenous sources is essential for proper tissue and organismal homeostasis. As we learn more about tissue-specific sources of DNA damage, we may be able to more effectively treat cancer and develop novel treatments for patients with DNA repair-associated genetic diseases.