Replication-coupled DNA Repair

Summary

The replisome quickly and accurately copies billions of DNA bases each cell division cycle. However, it can make errors especially when the template DNA is damaged. In these cases, replication-coupled repair mechanisms remove the mistake or repair the template lesions to ensure high fidelity and complete copying of the genome. Failures in these genome maintenance activities generate mutations, rearrangements, and chromosome segregation problems that cause many human diseases. In this review I provide a broad overview of replication-coupled repair pathways explaining how they fix polymerase mistakes, respond to template damage that acts as obstacles to the replisome, deal with broken forks, and impact human health and disease.

Introduction

Spontaneous chemical changes and DNA lesions from endogenous and environmental sources are ubiquitous threats to the information stored in the billions of DNA bases in each human cell. These challenges are acutely problematic during DNA replication and compounded by other forms of replication stress such as conflicts with transcriptional machineries.

Replication-coupled repair, defined as mechanisms that process damaged DNA in coordination with the replisome, works to overcome these challenges and maintain genome stability. Multiple pathways that operate in overlapping layers of repair are engaged based on the type, location, and context of the problem. They include repair activities that remove misincorporation errors, mechanisms to overcome DNA polymerase blocking lesions like base damage, pathways to deal with replicative helicase blocking obstacles including interstrand (ICL) and DNA-protein (DPC) crosslinks, and double-strand break (DSB) repair that protects stalled forks from degradation and restarts broken forks (Figure 1). In addition, the replication machinery can bypass some forms of damage and postpone lesion removal until after DNA synthesis is complete. These are not rarely used mechanisms, but rather essential functions needed every cell division cycle.

Fig. 1.

DNA lesions that are resolved by replication-coupled DNA repair pathways. (A) A simplified diagram of a eukaryotic replication fork. (B) Lesions generated or encountered by the replisome.

Copying errors and other replication failures cause the genome changes that underlie many human diseases. Most notably, this genetic instability drives cancer development and generates resistance to therapeutic intervention. However, it also provides a difference between cancer and normal cells that can be targeted by chemotherapeutic agents, synthetic lethal approaches with DNA repair inhibitors, and immunotherapy approaches. Here I survey replication-coupled repair mechanisms and the consequences of their inactivation and identify important unanswered questions.

Base misincorporation and damage in the newly synthesized DNA strands

Replicative polymerases are extremely accurate copying machines with self-correction capabilities. Error rates in vitro are higher than what is observed in cells due to additional layers of mismatch correction (Kunkel and Erie, 2015). The priming polymerase Polα is the most error prone, but it also performs the least amount of DNA synthesis, and most of the bases it incorporates are removed during Okazaki fragment maturation. The major lagging strand polymerase, Polδ has intermediate fidelity while the leading strand polymerase Polε has the highest fidelity of less than one error per 10⁶ bases. Mutations in POLE and POLD1 that reduce proofreading and increase base misincorporation are frequent in a variety of cancer types. A genetically engineered mouse model of one of these POLE mutations confirmed that it greatly increases mutation burden and causes cancer (Li et al., 2018). Importantly, polymerase mutations may also make these cancers more amenable to treatment since they correlate with a favorable prognosis. One possible explanation is that the high mutation rate generates neoantigens that improve immune system recognition.

Replication-coupled mismatch repair

Misincorporation errors are corrected primarily by mismatch repair (MMR). The detailed mechanisms of MMR will not be covered here; however, a few points are worth noting. First, MMR is more active on the lagging strand than leading, perhaps because of physical interactions between the MutSa and MutLa MMR proteins with PCNA and the prevalence of nicks in the newly synthesized DNA that can act as strand-discrimination signals to tell the MMR machinery which base is incorrect (St Charles et al., 2015). This increased MMR action on the lagging strand compensates for the reduced fidelity of Polα and Polδ compared to Polε. Second, the MMR machinery is tethered to the replisome via its interaction with PCNA so it is positioned to fix misincorporation errors as they are made (Hombauer et al., 2011; Kleczkowska et al., 2001). PCNA-mediated tethering of repair mechanisms to the replication fork is a common theme in replication-associated repair as exemplified by the large number of repair proteins in the PCNA interactome (Srivastava et al., 2018). Third, interactions of PCNA with MMR proteins also can stimulate and direct MutLα- incisions to the nascent strand providing strand discrimination (Pluciennik et al., 2010). Fourth, MMR also repairs strand slippage problems that can yield frameshift mutations. Thus, MMR defects cause microsatellite instability (MSI) due to failures in repairing misalignment problems in homonucleotide repeat sequences. This characteristic mutation pattern is used to diagnose tumors caused by MMR deficiencies. Importantly, the increase in neoantigens caused by MSI makes MMR-mutant tumors hypersensitive to treatment with immune checkpoint inhibitors (Le et al., 2017).

Ribonucleotide excision repair

In addition to inserting incorrect nucleotides, polymerases can also insert damaged bases. Thus, repair mechanisms operate to sanitize the nucleotide pools. However, these mechanisms do not recognize ribonucleotides, which are incorporated in the growing nascent strands thousands of times more frequently than incorrect DNA bases (Williams et al., 2016). The primary mechanism to remove ribonucleotides is ribonucleotide excision repair (RER). RNAseH2 recognizes the ribose sugar in duplex DNA and cuts the phosphate backbone immediately 5’ to the ribonucleotide. Pold then performs strand displacement synthesis and the flap endonuclease FEN1 removes the short ssDNA fragment containing the ribonucleotide. Finally, DNA ligase seals the nick. RER is coupled to DNA replication because the RNaseH2B subunit contains a PIP box that allows it to interact with PCNA (Bubeck et al., 2011). Thus, like MMR proteins, RNAseH2 co-localizes with sites of DNA replication. Both MMR and RNAseH2 have also been seen to be part of the replication fork proteome by the iPOND (Isolation of Proteins on Nascent DNA) method that purifies and quantitates the proteins in the replisome and associated with the newly synthesized DNA (Dungrawala et al., 2015).

RER inactivation is extremely rare in cancer, but RNAseH2 mutations do cause Aicardi- Goutiéres syndrome (Crow et al., 2006). When RNAseH2 is inactivated, TOPO1 can incise the backbone and promote a backup repair pathway (Sekiguchi and Shuman, 1997), but with a significant risk of generating deletions that presumably contributes to the etiology of this disease. The presence of alternative repair mechanisms that contribute to viability but at the expense of genome stability is another common theme of replication-coupled repair and forms the foundation of synthetic lethal approaches to cancer therapy.

Other types of newly synthesized DNA repair

In addition to misincorporation errors, failures in Okazaki fragment maturation can threaten genome stability. Unligated Okazaki fragments activate PARP1 to promote recruitment of single-strand break repair machinery including XRCC1 and an alternative DNA ligase—LIG3 (Hanzlikova et al., 2018). This function of PARP1 may contribute to the therapeutic effects of PARP inhibitors in cancer.

Replisome obstacles—Polymerase blocking lesions in template DNA

Most base damage will not block the replicative CMG helicase; however, these lesions will often pause polymerases. The most common base lesions that interfere with DNA synthesis include abasic sites, base oxidation, and base methylation. Approximately 10,000–20,000 abasic sites are formed each day in every human cell through spontaneous base loss and glycosylase removal of uracil or damaged bases. Base oxidation, methylation, thymine glycols, and lipid peroxidation products are additional lesions that form at a rate of a total of ~20,000 per day. Exposure to environmental contaminants can increase this DNA damage burden, and in tissues exposed to sunlight, UV photoproducts are also common lesions.

While most abasic sites and base lesions are repaired by base excision repair (BER) or nucleotide excision repair (NER) in duplex DNA, these repair systems will not always remove the damage prior to the arrival of a DNA replication fork. The CMG helicase is largely insensitive to their presence, but they can be potent blocks to replicative polymerases.

Lesion skipping

The response to a base lesion depends on whether it is on the leading or lagging template strand (Figure 2). Lagging strand lesions may stall Pola-primase or Pold, but are generally not thought to stall the replication fork or continued DNA synthesis since new primers on the lagging strand naturally facilitate bypass of the problem (Figure 2A). Thus, lagging strand lesions will usually generate gaps that can be repaired or filled in by trans-lesion bypass polymerases post- replicatively—a prediction that has been experimentally verified with reconstituted yeast replication proteins (Taylor and Yeeles, 2018). As yet verification of this lesion-skipping model in vertebrate cells has not been possible because of the difficulty of directing damage specifically to the lagging strand template.

Fig. 2.

Differences in responses to leading and lagging strand base damage. (A) Most lagging strand damage will generate ssDNA gaps but may not stall DNA elongation due to rapid repriming by Pola-primase. (B) Translesion synthesis, PrimPol-dependent repriming, and fork reversal are three pathways to overcome leading strand lesions that cause uncoupling of replisome activities.

Lesions on the leading strand template would be expected to be more persistent blocks to synthesis since Pola-primase is much less capable of making a new primer on the leading strand (Taylor and Yeeles, 2018). Thus, leading strand lesions will often cause the CMG helicase and DNA synthesis to become functionally uncoupled (Figure 2B).

The amount of uncoupling in a human cell is not known and by itself, it may not be particularly unusual or problematic. In E. coli there is surprisingly little coordination between leading and lagging strand synthesis with or without DNA damage (Graham et al., 2017). Furthermore, helicase unwinding is reduced ~80% when a polymerase pauses. This so-called “dead man’s switch” prevents the helicase from running too far ahead of DNA synthesis (Marians, 2018). Whether the same regulation happens in eukaryotic cells is unknown, although there is evidence from replicating DNA in Xenopus extracts that uncoupling between helicase and polymerase slows the helicase (Sparks et al., 2019). Nonetheless, uncoupling in the Xenopus system can be quite extensive leading to unwinding of kilobases of DNA (Byun et al., 2005). Such large amounts of ssDNA has not been observed in vertebrate cells suggesting the Xenopus egg extracts may not reflect a typical cellular response, but even small amounts of uncoupling can generate a platform for recruitment of ssDNA binding proteins that initiate ATR- dependent replication stress signaling or recruit other repair proteins (Figure 2B).

Eventually, a new primer is made on the leading strand template allowing DNA synthesis to resume, leaving a ssDNA gap on the leading strand. Repriming in E. coli happens within minutes and does not require a specialized primase (Yeeles and Marians, 2011). Thus, both lagging and leading strand damage is skipped by the bacterial replisome although with different kinetics. The eukaryotic replisome has some ability to skip leading strand lesions, but appears to preferentially use a specialized primase called PrimPol (Bianchi et al., 2013; Garcia-Gomez et al., 2013; Mouron et al., 2013).

Once a DNA lesion has been skipped, two alternative pathways can finish resolving the problem. First, bypass polymerases may be recruited to perform translesion DNA synthesis (Sale et al., 2012). This will often be mutagenic because of the lack of proper coding information on the damaged template and the translesion bypass polymerases (TLS polymerases) lack proofreading activity. Alternatively, a template switching mechanism could utilize the information from the undamaged sister chromatid as a template for synthesis in an error-free manner.

If most DNA lesions generate gaps behind the replication fork, this would explain how the ATR replication checkpoint is readily activated since TOPBP1-dependent activation requires a 5’ DNA end adjacent to ssDNA to be activated by the TOPBP1 (Saldivar et al., 2018). In cases where there is no 5’ DNA junction, an alternative ATR activation pathway dependent on the ETAA1 ATR activator may still operate as long as there is a region of RPA-coated ssDNA, but this pathway appears to be less involved in replication-stress induced signaling than the TOPBP1 pathway and relatively more important for controlling cell division processes independent of DNA damage (Bass and Cortez, 2019; Bass et al., 2016; Haahr et al., 2016; Saldivar et al., 2018). Activated ATR signals to a large number of pathways to promote repair, fork restart, and delay progression through the cell cycle (Saldivar et al., 2018).

Fork reversal

An alternative mechanism for dealing with fork stalling lesions is fork reversal, which is catalyzed by fork reversal enzymes including SMARCAL1, ZRANB3, and HLTF (Betous et al., 2012; Blastyak et al., 2010; Ciccia et al., 2012; Kile et al., 2015; Kolinjivadi et al., 2017; Vujanovic et al., 2017). Fork reversal involves migrating the three-way fork junction backwards to displace and anneal the nascent DNA strands to form what is called a chicken foot structure (Figures 2B and 3). Once thought to be rare in eukaryotes since reversed forks were only observed by electron microscopy (EM) in replication checkpoint-deficient budding yeast cells (Sogo et al., 2002), recent studies indicate that fork reversal may be quite frequent in vertebrates. In fact, EM analyses of replication forks purified from mammalian cells revealed that ~25% of the detected forks are reversed in cells treated with agents that induce nucleotide depletion, oxidative base damage, UV photoproducts, topoisomerase cleavage complexes, or DNA crosslinks (Zellweger et al., 2015). This important result suggests that fork reversal is common even when the fork encounters lesions that should be relatively easy to skip. One caveat to this interpretation is that the EM analysis assumes equivalent frequencies of psoralen- dependent stabilization and subsequent purification of replication intermediates to derive a quantitation of fork reversal frequency. Furthermore, the method is resource and time intensive, limiting how much data can be acquired. Thus, while the relative levels of fork reversal reported are informative, the field would benefit from development of an independent method to ascertain the relative and absolute percentages of reversed forks. In any case, based on the phenotypes associated with inactivating the fork reversal enzymes SMARCAL1, ZRANB3, or HLTF, fork reversal must be a significant replication stress tolerance mechanism. Inactivating each of these genes individually causes increased sensitivity to replication stress, changes in replication fork progression, and evidence of genome instability (Bansbach et al., 2009; Blastyak et al., 2010; Ciccia et al., 2009; Ciccia et al., 2012; Yuan et al., 2009, 2012).

Fig. 3.

Illustration of the complexities of fork reversal. Fork reversal places a template DNA lesion back into duplex DNA where it can be removed by excision repair. The ATP-dependent DNA translocases, SMARCAL1, HLTF, and ZRANB3 are regulated by RPA, 3’ DNA ends, and poly-ubiquitylated PCNA respectively to catalyze fork reversal. In some cases these enzymes may work sequentially, although the order of action depicted in the diagram is speculative. ATR signaling controls SMARCAL1 and also may stimulate reversal of undamaged forks. RAD51 acts through an unknown mechanism to promote reversal, and then RAD51 filaments stabilized by BRCA2 and additional homology directed repair (HDR) proteins prevent nucleases from degrading the nascent DNA strands.

Fork reversal, has several potential benefits. First, it may be a way to place a template DNA lesion back into the context of duplex DNA where excision repair mechanisms can operate. While intuitively attractive, there is little direct data on whether excision repair is actually coupled to reversal so experimentally testing this model of replication-coupled repair is a high priority. One possibility is that ZRANB3 directly participates in removal of the DNA lesion since it contains an endonuclease activity in addition to its fork remodeling function (Weston et al., 2012). This endonuclease is unusual in that it depends on both the ATPase motor and DNA substrate recognition domain of ZRANB3, which may ensure that DNA scission is coupled to fork reversal (Badu-Nkansah et al., 2016; Weston et al., 2012). Second, fork reversal could be a mechanism of template switching using the newly synthesized DNA strand as an undamaged template. Third, fork reversal may be a way of sequestering the stalled fork until a converging fork initiated from another origin finishes DNA synthesis of the region. Fourth, reversal is an essential step in some repair mechanisms such as when two forks converge on an inter-strand crosslink (Amunugama et al., 2018). Finally, fork reversal could be an intermediate in a recombination pathway of fork restart.

Recombination proteins including RAD51, BRCA1, and BRCA2 do have critical functions in fork reversal pathways. RAD51 is needed to generate reversed forks (Zellweger et al., 2015). However, what it does to promote reversal is unknown. One model is that RAD51 binds to the ssDNA on the template strands to do some kind of coordinated annealing reaction in cooperation with the fork reversal enzymes (Figure 3) (Neelsen and Lopes, 2015). Alternatively, RAD51 could act to capture the nascent ssDNA as it is formed by the reversal enzymes, essentially shifting an equilibrium in favor of fork reversal (Bhat and Cortez, 2018). There are still other possibilities since RAD51 can also bind double-stranded DNA and has many protein- protein interactions.

The reversed fork is not a static structure. It can be processed by DSB repair nucleases like MRE11 or DNA2 or directly reset to a three-way junction by DNA translocases. Nuclease processing may remove end-binding proteins and promote fork restart (Teixeira-Silva et al., 2017). In fact, the MRE11-RAD50-NBS1 (MRN) nuclease localizes to replication forks even in the absence of any added exogenous DNA damage and interacts with Replication Protein A (RPA) (Dungrawala et al., 2015; Maser et al., 2001; Seeber et al., 2016; Xu et al., 2008). However, end processing is typically restricted by RAD51 to prevent excessive nascent strand degradation (Hashimoto et al., 2010; Kolinjivadi et al., 2017; Lemacon et al., 2017; Mijic et al., 2017; Schlacher et al., 2011). This process is called fork protection (Figure 3). Stabilization of the reversed fork may be sufficient in most cases to allow proper merging with a converging fork (Carr and Lambert, 2013). Alternatively, since the reversed nascent DNA contains homology to the rewound parental DNA, RAD51 bound to the reversed fork could promote strand invasion to generate a D-loop. This action has the potential to restart DNA replication by re-generating a replication fork (Hashimoto et al., 2012; Petermann et al., 2010). How frequently this occurs is unknown.

The actions of RAD51 at forks are highly regulated. Additional recombination proteins including BRCA1, BRCA2, and the FANC proteins positively promote RAD51-dependent fork protection although they are not essential for fork reversal (Bhat and Cortez, 2018). Other proteins including RADX, FBH1, and BLM regulate RAD51 at forks to prevent excessive RAD51 activity and regulate fork protection (Bhat et al., 2018; Chu et al., 2015; Dungrawala et al., 2017; Fugger et al., 2009; Lorenz et al., 2009). The need for RAD51 antagonists illustrates the paradigm that repair mechanisms must be directed only to act when needed and must be coordinated in ways to ensure resolution since the intermediates in repair pathways can often be more deleterious than the original lesion.

There are many unanswered questions about fork reversal. Perhaps most pressing, why there are so many reversal enzymes and how they work with each other is unknown. SMARCAL1, ZRANB3, and HLTF are expressed in proliferating cells and they all catalyze similar biochemical reactions, but they do not act redundantly and their inactivation can yield different phenotypic outcomes. For example, SMARCAL1 mutations cause Schimke immunoosseous dysplasia while ZRANB3 and HLTF deficiencies may cause cancer (Poole and Cortez, 2017). SMARCAL1 but not ZRANB3 or HLTF has been implicated in regulating telomere replication in ways that impact alternative lengthening of telomeres (ALT) mechanisms that contribute to cancer cell immortalization (Cox et al., 2016; Diplas et al., 2018; Poole et al., 2015). The key to their differences may be in their substrate specificities and regulation which are dictated by accessory domains that bind DNA, RPA, and ubiquitylated PCNA (Figure 3) (Betous et al., 2013; Ciccia et al., 2009; Ciccia et al., 2012; Kile et al., 2015; Yuan et al., 2012). In addition, many other helicases and DNA translocases can catalyze fork reversal at least in vitro adding even more complication.

What happens to the replisome machinery during the reversal step is also unclear. There is little evidence that the replisome dissociates from the fork in response to fork stalling (De Piccoli et al., 2012; Dungrawala et al., 2015), and most forks remain competent to resume DNA synthesis rapidly even after a prolonged blockage. Thus, it seems likely that the CMG helicase remains bound since it cannot be reloaded during S-phase, but how reversal happens with it circling a template DNA strand and where it is located is unclear.

Fork reversal and lesion skipping may be two alternative pathways that could operate independently of each other. In bacterial systems transcriptional induction by the SOS DNA damage response likely increases TLS at the expense of increased mutation frequency. There is relatively little evidence that transcriptional responses control pathway choice in human cells, but this may partly be because most studies are done in cancer cell lines. Post-translational modifications including ubiquitylation and sumoylation especially to PCNA are important regulators that control pathway choice (Moldovan et al., 2007). ATR signaling is also important since fork reversal enzymes like SMARCAL1 are direct ATR substrates. ATR-dependent SMARCAL1 phosphorylation reduces its fork remodeling activity (Couch et al., 2013). This result is consistent with the observation that fork reversal is increased in budding yeast with mutations in the ATR pathway (Sogo et al., 2002). SMARCAL1 phosphorylation only happens after it binds to fork junctions, so phosphorylation may promote switching to another step in a fork reversal pathway or allow resetting of the reversed fork instead of turning off fork reversal altogether (Figure 3). Surprisingly, ATR is also reported to promote fork reversal by signaling from a stalled fork to forks that have not encountered a lesion at all, inducing their reversal (Mutreja et al., 2018). How reversal of undamaged forks would be valuable to the cell is not clear since it would tend to delay completion of DNA synthesis and the chicken foot structure is prone to nuclease processing that can generate genetic instability. Perhaps this is only a mechanism that operates after acute exposure to genotoxic agents that induce large numbers of lesions. Unfortunately, most experiments are done with these types of perturbations since the less frequent need for replication-coupled repair that is more physiological is more difficult to study. Further analyses are needed to explain how fork reversal, lesion skipping, and other template damage tolerance mechanisms are regulated and which pathway predominates in different contexts.

Additional mechanisms to repair polymerase stalling damage

Additional specialized mechanisms operate at specific types of polymerase stalling lesions. For example, a new mechanism of dealing with abasic sites has recently been discovered that depends on a protein called HMCES (Mohni et al., 2019). HMCES binds PCNA and uses an N- terminal cysteine to generate a DNA-protein crosslink (DPC) with the abasic site. This unusual mechanism shields abasic sites in ssDNA from TLS polymerases and endonucleases. How the HMCES-abasic site crosslink is resolved is unknown, but this mechanism of abasic site recognition and repair reduces mutation frequencies and provides protection from a wide-range of DNA damaging agents that generate abasic sites. Furthermore, the mechanism appears to be evolutionarily ancient, with HMCES proteins found in eubacteria, archaebacteria, and eukaryotes. How this mechanism provides a significant advantage over alternatives and whether fork reversal or template switching is involved in the error-free resolution of the HMCES-DPC is unknown.

Replisome obstacles: Helicase-blocking DNA lesions

ICLs and large DPCs especially on the leading strand where the CMG helicase translocates would be expected to be potent fork blocks since they interfere with helicase unwinding. While much less frequent than polymerase blocking damage, these lesions pose especially difficult challenges to replication.

Interstrand crosslinks are formed in response to DNA damaging agents used in cancer chemotherapy including nitrogen mustards and mitomycin C. They are also formed as minor products of ionizing and UV radiation. Endogenously, aldehydes generated by lipid peroxidation and sugar ring opening at abasic sites can yield ICLs. Many of the agents that induce ICLs such as UV, IR, aldehydes, and nitrogen mustards can also cause DPCs. Finally, interrupting the catalytic cycle of enzymes that utilize a covalent protein-DNA intermediate such as topoisomerases also generate DPCs that often are removed by replication-coupled repair mechanisms.

Helicase Traverse and Bypass

Surprisingly, neither ICLs nor DPCs are absolute impediments to replication fork movement. CMG encircles the leading strand template and uses and an accessory protein, MCM10, that helps ensure lagging strand DPCs do not interfere with helicase movement (Langston et al., 2017). However, DPCs on the CMG-tracking, leading strand template would be expected to be more problematic since they should not be accommodated within the CMG channel that is wide enough for duplex DNA but not for a bulky protein. But CMG can bypass this lesion as well (Sparks et al., 2019). In this case, an accessory DNA helicase RTEL1 unwinds the DNA past the lesion to provide a short patch of ssDNA. Somehow this allows the MCM complex to bypass the DPC, which then promotes proteolysis-dependent repair of the DPC lesion.

The CMG helicase can also “traverse” an ICL (Figure 4) (Huang et al., 2013). The CMG may be remodeled in a way to allow it to move past the ICL and facilitate resumption of DNA synthesis past the intact crosslink (Bellani et al., 2018). Two additional DNA helicases, FANCM and BLM, and the ATR kinase are required although their exact functions in this process are unknown (Ling et al., 2016; Rohleder et al., 2016). Traverse of the crosslink is relatively rapid – taking between 5–7 minutes. As in DPC bypass, the ICL may block the CMG helicase from unwinding DNA, but if another DNA helicase generates ssDNA adjacent to the lesion, then the MCM ring may be able to open, slide past the lesion, and then reclose perhaps with the help of additional accessory proteins (Trakselis et al., 2017). Further testing of this model is needed to understand if and how CMG gains this flexibility. Another interesting question in the case of ICL traverse is why it is not observed in the Xenopus extract replication system. Perhaps the very high concentration of replication proteins in the extract that are poised to promote rapid cell division cycles in early embryogenesis disfavors the traverse mechanism and instead favors fork convergence (Figure 4).

Fig. 4.

Possible replication-coupled ICL repair mechanisms. (A) Two converging forks approach the ICL. Short-chain TRAIP-dependent ubiquitylation promotes NEIL3 recruitment and strand unhooking. Long-chain ubiquitylation causes replisome unloading followed by fork reversal and unhooking by endonuclease incisions via the Fanconi anaemia pathway. (B) ICLs can be traversed with the help of the accessory DNA helicases FANCM and BLM. Remodeling of the CMG helicase may allow it to move past the ICL. Repriming then allows continued DNA synthesis leaving the ICL to be repaired through the Fanconi anaemia pathway.

Both ICL traverse and DPC bypass would in principle lead to a situation very similar to base lesions in which the polymerase would be stalled but the CMG helicase would be free to continue to unwind DNA. Repriming could then allow the ICL or DPC to be repaired post- replicatively. While the DPC may not be lethal on its own, failure to remove the ICL would interfere with chromosome segregation.

ICL repair

ICL traverse or fork convergence at the ICL both end up forming very similar X-shaped DNA structures around the crosslink (Figure 4). Unhooking the crosslink is then essential to allow completion of DNA synthesis and chromosome segregation. Unhooking and repair take place through at least two pathways depending on the type of ICL. Psoralen and abasic site-induced crosslinks can be unhooked without incision of the phosphodiester backbone through the action of the NEIL3 glycosylase (Semlow et al., 2016). NEIL3 recruitment is regulated by short-chain CMG ubiquitylation catalyzed by TRAIP (Wu et al., 2019). This unhooking reaction would leave an abasic site on one template strand and a mono-adduct or adenosine on the other, which would no longer pose an obstacle to the CMG helicase. Additional steps such as TLS would then replicate the gaps.

Other types of ICLs are not good substrates for NEIL3 and are instead repaired via a FANC protein-dependent mechanism. The FANC proteins are named because their inactivation causes Fanconi anaemia, a disorder characterized by cellular hypersensitivity to interstrand crosslinking agents (Ceccaldi et al., 2016). Three notable characteristics of this pathway are the requirements that the CMG helicase be unloaded, the DNA backbone be incised to allow unhooking, and involvement of fork reversal (Amunugama et al., 2018; Long et al., 2014; Zhang et al., 2015).

CMG helicase unloading at the ICL was initially reported to depend on the tumor suppressor BRCA1; however, new data indicate this was a mistake and that TRAIP catalyzes more extensive CMG ubiquitylation to promote p97-dependent unloading (Wu et al., 2019). CMG removal happens prior to fork reversal suggesting that replisome disassembly is needed for fork reversal in this context (Amunugama et al., 2018; Long et al., 2014). Fork reversal happens on only one of the converged forks and is followed by strand incision and repair via the Fanconi anaemia pathway that is reviewed elsewhere (Ceccaldi et al., 2016). The need to unload the helicase for fork reversal to occur at a crosslink does not easily fit the observation in human cells that fork reversal is a highly frequent response to psoralen-induced crosslinks, even happening at forks that do not encounter the lesion directly (Mutreja et al., 2018). Sorting out these experimental system differences is needed to fully understand the importance of fork reversal to ICL repair and other replication stress tolerance mechanisms.

DPC repair

After helicase bypass of a DPC, the lesion must still be removed to complete gap filling replication. Since large DPCs block the action of NER, at least two proteolysis-dependent mechanisms operate that depend on SPRTN or the proteasome (Larsen et al., 2019; Stingele et al., 2016; Vaz et al., 2016). Proteasome degradation is triggered via DPC ubiquitylation. Like CMG removal at an ICL, the initial DPC ubiquitylation is directed by the E3 ligase TRAIP but additional ligases may be required to generate proteasome-targeting ubiquitin chains (Larsen et al., 2019). Sumoylation also regulates DPC repair (Borgermann et al., 2019; Stingele et al., 2014), but the complete mechanisms that control DPC repair such as how nearby DNA binding proteins are protected from removal remain poorly understood. Once degraded, the residual small peptide-crosslink lesion can be excised by NER. However, the gap in the daughter strand would first need to be closed using a combination of Pold and TLS polymerases. Thus, this process is expected to be mutagenic.

A specialized case of DPC repair is removal of topoisomerase proteins that are unable to complete their catalytic cycle. Drugs like etoposide, topotecan, and ciprofloxacin are topoisomerase poisons used to treat cancer and infectious diseases that greatly increase the frequency of these covalent topoisomerase-DNA complexes. These DPCs can be removed by tyrosyl-DNA phosphodiesterases (TDP1 or TDP2) (Pommier et al., 2014). Like other DPC repair, sumoylation is an important regulator of these specialized topoisomerase removal systems (Hudson et al., 2012; Schellenberg et al., 2017).

Other fork obstacles

In addition to ICLs and DPCs, other obstacles that block fork movement include tightly bound, non-covalent protein complexes. For example, some proteins act as fork barriers that help ensure unidirectional replication in loci that otherwise would be prone to head-on transcription- replication conflicts. While these are not DNA lesions needing repair, investigators have co- opted these systems to design site-specific replication blocks that have been especially useful to investigate how recombination-based repair mechanisms act to control genome stability during replication (Willis et al., 2018).

Fork breakage and restart

Another serious threat to genome stability is fork breakage. DSBs are generated during replication via multiple mechanisms (Figure 5). Any repair process that involves DNA strand cutting can generate a DSB if the incision happens in ssDNA. In addition, single-strand breaks in the template DNA can be converted to DSBs via CMG unwinding (Hashimoto et al., 2012). Collisions of forks with poisoned topoisomerase cleavage complexes generate breaks. DSBs are also caused by structure specific endonucleases including MUS81 that process persistently stalled forks (Hanada et al., 2007). The end of the nascent-nascent strand duplex of a reversed replication fork mimics a DSB, and the reversed and stalled forks may also be cleaved by nucleases like ARTEMIS, XPF, and nucleases scaffolded by SLX4 (Bhat and Cortez, 2018).

Fig. 5.

Four mechanisms for generating a single-ended double-strand break at replications forks. Single-ended breaks are repaired by break induced replication mechanisms that depend on an alternative replisome.

In contrast to DSBs that form in quiescent or G1 phase cells, replication-associated breaks are typically “single-ended” necessitating recombination as the preferred repair mechanism. In fact, end-joining must be actively avoided since it could cause genome rearrangements and toxicity. Thus, while homology-directed repair (HDR) of DSBs is frequently studied as a mechanism to repair two-ended breaks during S and G2 phases, the most frequent need for these repair proteins is likely at replication forks. This need at least partly explains the utility of drugs like PARP inhibitors that increase the frequency of fork breakage in treating HDR-deficient cancers.

Break Induced Replication

Resection of the DNA end at a collapsed fork followed by strand invasion can regenerate a fork structure and facilitate resumption of DNA synthesis using a modified replisome assembled on a migrating D-loop (Figure 5) (Anand et al., 2013; Kramara et al., 2018). This process is called break-induced replication (BIR) and is best understood in yeast systems. As might be expected for a mechanism involving strand invasion, yeast BIR is often RAD51-dependent. An additional Pold subunit, yeast Pol32 or human POLD3, is added to the BIR replisome (Lydeard et al., 2007), and there may be additional changes including the use of an alternative helicase.

The length of BIR synthesis can be restricted by convergence of a replication fork from a nearby origin and by the ability of nucleases to process the D-loop structure (Mayle et al., 2015). Restricting the length of BIR synthesis helps maintain genome stability because BIR synthesis is error prone (Deem et al., 2011; Sakofsky et al., 2014; Smith et al., 2007). Errors in BIR can arise because of improper recombination with highly similar sequences and the BIR replisome tends to make more misincorporation, frameshift, and template switching errors. In part, this mutagenic synthesis is due to extensive utilization of the modified Pold polymerase, which frequently dissociate from the template increasing the risk of strand slippage and improper template switching. Some replication-coupled repair mechanisms also may not operate efficiently in the context of the BIR replisome. Finally BIR synthesis happens at a migrating D- loop. Thus, more ssDNA is generated which is less chemically stable and more prone to damage by chemicals and enzymes like APOBECs that generate a cancer mutational signature that is associated with the ssDNA on lagging strand templates (Haradhvala et al., 2016).

The mechanisms controlling BIR are not well understood in vertebrate cells. Although RAD51 is critical to restart stalled replication forks in many circumstances, the relative contributions of RAD51-dependent and -independent BIR pathways are unknown. RAD51-independent BIR requires RAD52, and may be engaged at fragile sites, when large amounts of replication stress generate broken forks, or when other fork stability mechanisms are inoperable (Bhowmick et al., 2016; Costantino et al., 2014; Lemacon et al., 2017; Sotiriou et al., 2016). For example, BIR at common fragile sites happens after MUS81 cleaves unreplicated DNA during mitosis. Thus, this form of BIR is also called MiDAS for mitotic DNA synthesis (Bhowmick et al., 2016). This illustrates the idea that in addition to restarting broken forks, BIR is a mechanism to prepare for mitosis and facilitate chromosome segregation. Finally, BIR may be useful for telomere extension in cells utilizing ALT (Dilley et al., 2016). In this case, telomere damage induces a BIR replisome to replicate telomere sequences. This ALT mechanism promotes telomere maintenance thereby facilitating tumorigenesis.

Additional Perspectives

Coordinating DNA repair with DNA replication ensures that the genome is copied completely while minimizing errors. This coordination poses some special difficulties since most repair mechanisms depend on an intact DNA duplex to function. Thus, any strand incisions must either avoid ssDNA at the fork or be coupled with a DSB repair mechanism. Replication also provides some advantages. It provides a sensing mechanism for DNA lesions, and DNA synthesis generates a duplicate copy of the genome from which information can be utilized to complete repair. Furthermore, any single lesion often can be repaired or tolerated via more than one replication-coupled pathway. This partial redundancy provides robustness, thereby ensuring that one way or another cells complete replication. It also forms the foundation of synthetic lethal approaches in cancer treatment when one pathway is inactivated by mutation and a drug inhibits another.

The details of many of these repair mechanisms still must be worked out. Reconstitution of origin-dependent eukaryotic replication with purified proteins and the use of Xenopus egg extracts provide powerful systems that can be coupled with site-specific DNA lesions to generate high-resolution information. However, cellular systems to selectively target damage to leading or lagging strand templates at specific genomic locations are needed to better dissect context-dependent repair processes. Elucidating the details of these mechanisms yields insights that are generalizable to other areas of biology and identifies vulnerabilities that can explain dysfunction in disease or provide opportunities for drug development. Furthermore, unbiased approaches like iPOND proteomics and CRISPR-Cas9 genetic screens continue to add to the inventory of replication-coupled repair proteins emphasizing how much is left to learn.