Prevention and repair of DNA damage is essential for maintenance of genomic stability and cell survival. DNA replication during S-phase can be a source of DNA damage if endogenous or exogenous stresses impair the progression of replication forks. It has become increasingly clear that DNA-damage-response pathways do not only respond to the presence of damaged DNA, but also modulate DNA replication dynamics to prevent DNA damage formation during S-phase. Such observations may help explain the developmental defects or cancer predisposition caused by mutations in DNA-damage-response genes. The present review focuses on molecular mechanisms by which DNA-damage-response pathways control and promote replication dynamics in vertebrate cells. In particular, DNA damage pathways contribute to proper replication by regulating replication initiation, stabilizing transiently stalled forks, promoting replication restart and facilitating fork movement on difficult-to-replicate templates. If replication fork progression fails to be rescued, this may lead to DNA damage and genomic instability via nuclease processing of aberrant fork structures or incomplete sister chromatid separation during mitosis.
In dividing cells, faithful and complete replication of the genome during S-phase is essential for maintenance of genomic stability and cell survival. If the progression of DNA replication forks is impaired, this can lead to the formation of aberrant DNA structures and generation of DNA damage, including DSBs (double-strand breaks). Spontaneous replication-associated DNA damage may inhibit cell growth during development and promote genomic instability that leads to cancer. Consequently, cellular pathways that prevent and repair damage at replication forks are important during development and can act as tumour suppressors. Thanks to advances in the measurement of replication dynamics using DNA fibre approaches and Xenopus egg extracts [1,2], it has become increasingly clear in recent years that DNA-damage-response and -repair pathways do not only repair damage generated at the replication forks, but also modulate the initiation and progression of DNA replication. Pathways that are activated by perturbed or stalled replication forks, such as checkpoint signalling and homologous recombination, modulate replication initiation and fork progression in response to DNA-damaging treatments, but they also perform similar functions in the absence of exogenously induced damage. Replication dynamics and the incidence of replication-associated DNA damage are therefore often notably altered in cells with defects in these pathways, which may underlie the developmental and/or cancer-prone genetic disorders that can be associated with such defects. The present review will discuss the molecular mechanisms by which the DNA damage response may regulate replication dynamics in vertebrate cells, on the basis of findings from Xenopus, chicken DT40 cell and mammalian systems while also drawing on central findings from yeast and bacteria. We have aimed to concentrate on the most recent findings, as some of these topics have been covered in our earlier reviews [3,4].
EUKARYOTIC DNA REPLICATION
DNA replication initiates at genomic sites that are termed origins of replication. In higher eukaryotes, these are organised into initiation zones or clusters that are activated at different times during S-phase [5,6]. Initially, the DNA double helix is unwound to form a ‘replication bubble’ before two replication complexes are assembled (Figure 1A). Two individual replication forks (which form the replicon) then move bi-directionally away from the origin, unless a RFB (replication fork barrier) is present. Origins are selected through binding of the ORC (origin recognition complex), a six-subunit complex consisting of ORC1–ORC6 [7,8]. Origin selection seems to be dependent on the ORC1 subunit [9,10] and, at least in higher eukaryotes, is sequence-independent . Factors influencing the spacing of origin firing include chromatin structure and epigenetics, nuclear organization or matrix, developmental stage, and gene expression . During G1-phase and prior to S-phase, CDC6 (cell division cycle 6), CDT1 (chromatin licensing and DNA replication factor 1) and the MCM (mini-chromosome maintenance) 2–7 helicase complex are recruited to sites of ORC binding in a co-ordinated fashion to form the pre-RC (pre-replication complex) . Replication is then initiated in early S-phase as the pre-RC becomes specifically phosphorylated by the cyclin E CDK2 (cyclin-dependent kinase 2) and DDK (Dbf4/Drf1-dependent CDC7 kinase) [14,15]. In budding yeast, the pre-RC components phosphorylated by CDK2 are Sld3 and Sld2 [16,17], with the possible metazoan homologues of these proteins being RECQL4 (RecQ protein-like 4), GEMC1 (geminin coiled-coil domain-containing 1) and Treslin [18–22]. DDK, on the other hand, phosphorylates the MCM2–7 complex, showing a preference for the MCM2 subunit . Phosphorylation of these pre-RC components promotes loading of CDC45 and the GINS (go-ichi-ni-san) complex, which are essential for the actual initiation of replication (origin firing) and fork progression [23–25].
Overview of vertebrate DNA replication
Replication fork progression
The MCM2–7 hexamer and CDC45 form the active replicative helicase required for DNA unwinding during replication (Figure 1B). As the MCM complex unwinds the parental DNA in an ATP-dependent manner, the GINS complex maintains protein–protein interactions within the replication complex (replisome) [26–28]. The replisome also contains the clamp loader, RFC (replication factor C), which helps to load the sliding clamp, PCNA (proliferating cell nuclear antigen), on to primed DNA. PCNA tethers the DNA polymerases to the chromosome, enabling processive and high-speed DNA replication . The replisome contains three DNA polymerases, the polymerase α-primase and the replicative polymerases δ and ϵ, which replicate the lagging and leading strands respectively. RFC and PCNA regulate the switch from the polymerase α-primase to the replicative polymerases , and also act as a loading platform for FEN-1 (flap endonuclease-1) and DNA ligase I, which process and seal Okazaki fragments.
If replication forks encounter obstacles on the DNA template, they may stall or collapse. Such obstacles can include spontaneous DNA damage or secondary structures that may arise in repeated A-T- or G-C-rich sequences [31,32]. Replication fork stalling may also occur when the replisome collides with transcription machinery [33,34] or when it meets an RFB, a DNA-bound protein or protein complex that actively promotes fork pausing or stalling. Such RFBs promote mating-type switching in fission yeast  or maintenance of ribosomal DNA repeat copy number from yeast to humans [36,37]. A stalled replication fork is halted in its movement, but able to resume progression when the blockade is removed. A collapsed fork, however, has become inactivated, for example by running into a single-strand break causing generation of a DSB or by dissociation of the replication machinery (reviewed in ). To stabilize forks, cells must maintain viable fork structures and assembly of the replication machinery until the replication block is removed, as will be discussed below. Replication forks that have been stalled for many hours become inactivated and are unable to restart. Under these circumstances, global replication restart is achieved by new origin firing . Inactivation of replication forks also coincides with the generation of fork-associated DSBs by the structure-specific endonuclease complex of MUS81 (MUS81 endonuclease homologue) and EME1 (essential meiotic endonuclease 1 homologue 1) [38,39].
Termination of DNA replication in eukaryotes occurs when two opposing replication forks converge. In bacteria, the circular chromosome contains a specific terminator site (TER) opposite to the origin of replication . Efficient termination at these sites requires an RFB posed by the Tus protein, which blocks progression of the replicative helicase . Less is known about replication termination in vertebrates, but in Xenopus egg extracts, at least, it appears that termination occurs at random sites . A large amount of evidence from SV40 (simian virus 40) and yeast systems suggests that sister chromatids become intertwined (catenated) at replication termination sites and that the resolution of these structures, for successful completion of termination, requires DNA TOPII (topoisomerase II) [43–45]. Similar mechanisms are likely to operate in higher eukaryotes (Figure 1B) .
Chromatin remodelling during DNA replication
Efficient replication fork progression requires that the replisome gain access to the DNA through remodelling of the chromatin structure. Nucleosomes are disassembled ahead of the replication fork, leading to eviction of parental histones, which are recycled on to the daughter strands behind the replication fork, combined with de novo deposition of newly synthesized histones (Figure 1C). Histone eviction probably requires ATP-dependent chromatin-remodelling enzymes, whereas histone chaperones sequester the released histones to facilitate recycling . Histone chaperones such as ASF1 (anti-silencing function 1) and the FACT (facilitates chromatin transcription) complex interact with the replisome and act during replication. FACT may be involved in disrupting nucleosomes ahead of the fork, whereas ASF1 acts as an acceptor of parental and newly synthesized histones, and is a donor for the major histone loader, CAF-1 (chromatin assembly factor 1) [47–51]. CAF-1 deposits histones behind the replication fork, and its interaction with PCNA couples chromatin assembly with DNA replication [47,52]. ASF1 and CAF-1 are required for S-phase progression [53–55], and both the FACT subunit SSRP1 (structure-specific recognition protein 1) and ASF1 promote efficient replication fork progression [49,50].
The newly replicated sister chromatids need to be tethered together to promote the proper segregation of chromatids during mitosis. This process, termed sister chromatid cohesion, requires cohesin, a ring-shaped complex consisting of four SMC (structural maintenance of chromosome) subunits [56,57]. The complex is loaded on to DNA in a CDC7-dependent manner during G1–S-phase . Its role in maintaining a close distance between chromosome arms is key in facilitating sister chromatid-dependent modes of DNA damage repair . Recruitment of the complex to sites of DNA damage is checkpoint-dependent [60–62], whereas mutations in the complex lead to an increase in chromosomal aberrations .
DNA-DAMAGE-RESPONSE PATHWAYS AT REPLICATION FORKS
Obstacles on the DNA that specifically affect the progression of the DNA polymerases can lead to uncoupling of the replicative polymerase and helicase activities, with the helicase continuing unwinding to generate excessively long stretches of ssDNA (single-stranded DNA) [63,64]. In addition, exonuclease resection of damaged DNA structures such as DSBs can generate ssDNA, which occurs especially during the S- and G2-phase, as resection is controlled by CDK activity [65–67]. This ssDNA is recognised as DNA damage by the cell cycle checkpoint machinery. Specifically, RPA (replication protein A)-coated ssDNA recruits and activates the checkpoint kinase ATR [ATM (ataxia telangiectasia mutated)- and Rad3-related] and its interaction partner ATRIP (ATR-interacting protein) (Figure 2A) . Although loss of ATR and other components of the ATR pathway is embryonic lethal, hypomorphic mutations in ATR and other genes of the pathway can cause growth and developmental disorders, such as ATR-Seckel syndrome [69,70]. Activated ATR phosphorylates a large number of downstream targets involved in checkpoint signalling, DNA repair and apoptosis, such as histone H2AX  and tumour protein p53 . One important role of ATR is the phosphorylation and activation of the effector kinase CHK1 (checkpoint kinase 1) [73,74]. Efficient phosphorylation of CHK1 by ATR is promoted by several other replication and checkpoint proteins, including RAD9 (Rad9 homologue), RAD17 (Rad17 homologue), TOPBP1 [TOP (DNA) II-binding protein 1], claspin, and the complex of TIM (timeless homologue) and TIPIN (TIMELESS-interacting protein) [75–80]. CHK1 in turn phosphorylates the cell cycle phosphatases CDC25-A, CDC25-B and CDC25-C to prevent activation of cyclin E/A-CDK2 and cyclin B-CDK1. This slows progression through S-phase and prevents mitotic entry, providing the cell with time to restart or repair the stalled replication forks [81–83]. In budding yeast, the functional homologue of CHK1, Rad53, suppresses late origin firing in response to DNA damage by phosphorylating Sld3 and Dbf4 [84–86]. Other substrates of CHK1 include p53 and DNA repair proteins such as RAD51, BRCA2 (breast cancer early onset 2) and FANCE [FA (Fanconi anaemia), complementation group E] (see below) [87–90].
DNA-damage-response pathways at replication forks
The ATM checkpoint kinase, mutated in the neurodegenerative and cancer-prone genetic disorder ataxia telangiectasia, is activated in response to DSBs. This requires the MRE11 (meiotic recombination 11)–RAD50–NBS1 complex in concert with MDC1 (mediator of DNA-damage checkpoint 1), 53BP1 (p53-binding protein 1) and a host of other signalling and mediator proteins. Its effector kinase is CHK2, which similarly to CHK1 targets CDKs and p53 (reviewed in ). ATM is considered to be less important in the response to replication blocks than ATR. However, ATM is upstream of ATR activation as it promotes exonuclease processing of DSBs, and can itself be phosphorylated and activated by ATR in response to replication blocks [67,92]. Indeed, ATM signalling also slows DNA replication in response to DNA damage, probably by inhibiting origin firing, and plays roles in the stabilization and repair of damaged replication forks [93–96].
HR (homologous recombination) was initially identified as the pathway mediating genetic recombination in meiosis. It has become clear, however, that HR also repairs DNA DSBs, single-strand gaps and damaged replication forks in mitotic cells, as well as having roles during normal DNA replication [97,98]. HR repairs DSBs by using homologous DNA sequences (the homologous chromosome, sister chromatid or homologous sequences elsewhere in the genome) as a template for DNA polymerase-mediated re-synthesis of the sequence containing the break (reviewed in [99,100]). Similar mechanisms are thought to be involved in the functions of HR during replication. In eukaryotes, HR activity is controlled by CDKs, restricting it to S-phase and G2phase of the cell cycle [65,66,101]. HR during S-phase and G2-phase mainly uses the identical sister chromatids, produced by DNA replication, as homologous templates for repair, limiting the potential for large-scale genomic rearrangements.
The central HR factor is RecA in bacteria and RAD51 in eukaryotes (Figure 2B). RAD51 is recruited to DNA ends with 3′-single stranded overhangs and possibly ssDNA gaps, where it forms protein–DNA filaments . RAD51 filament formation and stability is regulated by a large number of RAD51-interacting proteins, including the RAD51 paralogues XRCC2 (X-ray repair complementing defective repair in Chinese hamster cells 2), XRCC3, RAD51B, RAD51C, RAD51D and BRCA2, the protein product of the breast-cancer susceptibility gene [103–106]. The requirement for ssDNA makes HR repair of DSBs dependent on DNA end resection by nucleases such as MRE11 and EXO1 (exonuclease 1), a process that is controlled by CDK activity and therefore restricted to S-phase and G2-phase of the cell cycle [65,66,107]. Once bound to the overhang, RAD51 catalyses homology search and recombines the 3′-overhang into the homologous double-stranded DNA to form a D-loop (displacement loop), which generates a HJ (Holliday junction) . Upon completion of DNA repair synthesis, the remaining HJ structures are removed by HJ resolution by HJ resolvases such as GEN1 (Gen endonuclease homologue 1), or HJ dissolution by a complex of BLM (Bloom syndrome, RecQ helicase-like), TOPIIIa and hRMI1 (RecQ-mediated genome instability 1, homologue) [109,110].
Efficient HR requires the FA pathway, so called because its components are mutated in patients with the autosomal recessive disorder FA. Mutations in the FA pathway cause extreme sensitivity to DNA cross-linking agents, cancer predisposition, developmental defects and anaemia . In the current understanding of the FA pathway, its central factors appear to be FANCD2, which is phosphorylated by ATR in response to replication stress, followed by ubiquitylation by the activated FA core complex , and the tumour suppressor BRCA2 (FANCD1). Although the FA pathway seems to interact with both TLS (translesion synthesis) (see below) and HR, its role during HR is characterized better; in particular, BRCA2 directly promotes RAD51 function by facilitating RAD51 filament formation and stability . BRCA2 also interacts with the FA protein FANCN (PALB2) and together they promote RAD51-mediated D-loop formation [114,115].
Alternative helicases and polymerases
In addition to the MCM2–7 replicative helicase, other DNA helicases play roles at replication forks encountering DNA damage. These helicases can remodel perturbed replication fork structures or difficult-to-replicate secondary structures to aid progression of the replisome. In humans, these include two of the known RecQ helicases, WRN (Werner syndrome, RecQ helicase-like) and BLM (Figure 2C). Individuals carrying non-functional copies of the BLM or WRN genes suffer from the rare autosomal recessive disorders Bloom and Werner syndrome respectively, which are characterized by a strong predisposition to all types of cancer [116,117]. Both BLM and WRN have functions in regulating HR; they can disrupt intermediate structures such as the HJ, thus preventing spontaneous HR and genomic instability at perturbed replication forks [118,119], and BLM also suppresses genetic rearrangements via its HJ dissolution activity . The SMARCAL1 (SWI/SNF-related, matrix-associated, actin-dependent regulator of chromatin, subfamily a-like 1) helicase, mutated in the genetic disorder Schimke Immunoosseous dysplasia, is an unusual ATP-dependent ssDNA-annealing helicase that can potentially reverse previous unwinding by the MCM complex . A further two DNA helicases with roles at perturbed forks are the FA pathway components FANCJ/BRIP1 and FANCM [121–123]. Both can function similarly to BLM and WRN by remodelling stalled DNA replication forks [121,123], and FANCJ can also unwind secondary structures [122,124]. FANCJ potentially works alongside specialized polymerases , which replace the conventional replicative polymerase in such circumstances; a process known as TLS (Figure 2D). These polymerases can replicate past a variety of non-coding DNA lesions as their active sites can tolerate a distorted template (reviewed in [126,127]). So far there are five known TLS polymerases: DNA polymerases η, κ, ι, ζ (the latter comprising two subunits REV3 and REV7) and REV1 (REV1 homologue). REV1 interacts with all other known TLS polymerases via its C-terminus  and is believed to be a key regulator of lesion bypass. As low-fidelity TLS polymerases may also increase the chance of mutations, TLS is only activated at sites of DNA damage through monoubiquitination of PCNA by RAD6 (ubiquitin-conjugating enzyme E2B) and the E3 ubiquitin ligase RAD18 (RAD18 homologue) [129,130]. This allows switching from replicative to TLS polymerases, as the latter have an increased affinity for monoubiquitinated PCNA through their ubiquitin-binding motifs . Polyubiquitination of PCNA at the same residue by Rad5 in yeast, in contrast, activates not TLS but an alternative ‘error-free’ damage avoidance mechanism thought to involve HR-like activities such as template switching (reviewed in ). Mammalian cells contain two Rad5 homologues, HLTF (helicase-like transcription factor) and SHPRH [SNF2 histone linker PHD (plant homeodomain) RING helicase] [133–135].
REGULATION OF REPLICATION BY THE DNA DAMAGE RESPONSE
Checkpoint signalling controls replication initiation
ATR and ATM signalling pathways down-regulate CDK activity, thereby reducing the overall levels of origin firing during normal S-phase, and inhibition of these pathways increases origin firing . This pathway is particularly well studied in the case of ATR/CHK1 signalling. Reduced CHK1 activity leads to accumulation of CDC25A in undamaged cells [82,83]. This is accompanied by increased cyclin E-CDK2 activity, increased loading of CDC45 on to chromatin and increased frequencies of origin firing [137–139]. CHK1 also regulates Cyclin A-CDK1 activity, which specifically promotes late origin firing . It has been proposed that the ssDNA that is transiently generated during replication initiation causes low levels of checkpoint activation, thus generating a feedback loop to down-regulate further initiation . In Xenopus egg extracts, checkpoint inhibition shortens the overall duration of S-phase , although this is not necessarily the case in non-embryonic vertebrate cells, possibly due to the simultaneous fork slowing observed [141,142].
Changes in CDK activity affect the activation of whole replication clusters more strongly than the activation of origins within clusters, and checkpoint inhibition therefore predominantly increases the number of simultaneously active clusters . At the same time, other signalling pathways downstream of the checkpoint counteract the suppression of origin firing, possibly only within clusters. PLK1 (Polo-like kinase 1) is recruited during replication stress in an ATR-dependent manner and can phosphorylate ORC2 to counteract the checkpoint and allow some amount of origin firing even in presence of replication blocks [144,145]. This is probably necessary because activation of dormant origins within replication clusters serves as an important mechanism to allow completion of global replication and prevent genomic instability [146–149]. The above observations suggest that the checkpoint selectively allows origin firing within clusters, but not activation of new clusters . If, however, replication cluster activation is increased due to checkpoint inhibition, this causes DNA damage .
Checkpoint signalling controls replication fork progression
In addition to increased replication initiation, vertebrate cells that are defective in ATR/CHK1 signalling display reduced average speeds of replication fork progression during an unperturbed S-phase. This applies to cells with defects in ATR, CHK1, claspin or TIM [78,150–153]. These observations suggest that, just as the S-phase checkpoint regulates origin firing in unperturbed cells, it also modulates the progression of replication forks during normal replication.
Evidence from our previous work suggests that replication initiation control by the checkpoint may underlie its promotion of replication fork progression. Preventing excessive origin firing in CHK1-defective cells by simultaneous inhibition of CDK2 using roscovitine, or siRNA depletion of CDC7, can restore normal speeds of replication fork progression . This suggests that inhibition of CDK (and possibly DDK), to regulate origin firing, may be an important mechanism by which CHK1 promotes normal replication fork progression. In agreement with this, the generation of spontaneous DNA damage in CHK1-depleted cells depends on CDK2 and CDC25A [139,154]. In another study, overexpression of active cyclin A2-CDK1 increased origin firing and slowed replication fork progression, whereas loss of CDK1 increased speeds of fork progression . Several mechanisms by which increased origin density might perturb replication fork progression can be envisaged (Figure 3). The increase in active replication forks may deplete replication factors such as nucleotides . Changes in origin density could also interfere with the spatial co-ordination of replication initiation with transcription, which may increase conflicts between replication and transcription activities [157–160]. In addition to controlling origin firing, ATR/CHK1 signalling may facilitate replication fork progression by promoting fork stability via any of the pathways discussed below; reduced stability of spontaneously stalled forks is likely to result in slower average replication fork speeds.
Control of replication fork progression and stability by checkpoint factors
Intriguingly, evidence suggests that in the presence of DNA damage, the checkpoint can also act to slow down replication fork progression, which would be a useful mechanism to reduce collisions of the replication machinery with DNA lesions. CHK1 and TIPIN have been shown to slow replication fork progression in presence of the DNA-damaging agents camptothecin and UV respectively [78,161]. The molecular mechanism of checkpoint-dependent replication fork slowing is not yet understood, but it could involve phosphorylation of replication proteins such as the MCM complex .
The checkpoint controls replication fork stability
Checkpoint signalling stabilizes replication forks, i.e. it prevents the accumulation of aberrant fork structures or DNA damage during replication, both in cells treated with replication inhibitors and during unperturbed S-phase. Components of both the ATR and the ATM pathways have been shown to stabilize replication forks stalled by replication inhibitors: ATR, ATM, MRE11, CHK1, claspin, TIM and TIPIN [96,163–165]. ATR, ATM, MRE11 and CHK1 also protect from DNA breakage during an unperturbed S-phase, possibly by stabilizing spontaneously stalled forks in addition to promoting DNA repair [96,139,166,167]. Checkpoint signalling might stabilize replication forks through a variety of downstream mechanisms (Figure 3). Checkpoint kinases phosphorylate the MCM complex  and it has been observed that the checkpoint prevents replication proteins from dissociating from stalled or collapsed forks . Other checkpoint targets that promote replication fork stability include the BLM and WRN helicases [168–170]. Furthermore, checkpoint kinases can potentially regulate HR by phosphorylating RAD51, BRCA2, FANCE and especially FANCD2 [87,89,90,112]. In our hands, RAD51 depletion and CHK1 inhibition had an additive effect on inhibiting replication fork restart, suggesting that CHK1 activity stabilizes stalled replication forks through mechanisms other than promoting HR . However, the replication fork restart assay cannot distinguish between effects on replication fork progression and stability .
The checkpoint also down-regulates the activity of nucleases that could otherwise inactivate stalled replication forks by aberrant resection or processing into DSBs: the structure-specific endonuclease Mus81–Eme1 in fission yeast and EXO1 in mammalian cells [171,172]. Spontaneous DSB formation during replication in mammalian cells deficient in CHK1 or WEE1 (a checkpoint kinase that acts downstream of CHK1) is MUS81-dependent, but a direct regulation of MUS81–EME1 by checkpoint kinases has not yet been found [173,174].
Another target of the checkpoint is the transcription machinery. In budding yeast, the functional homologue of CHK1, Rad53, prevents stalled forks from reversing into X-shaped structures representing HJ formation . This fork reversal is thought to indicate fork collapse and has been explained by topological strains on stalled replication forks that result from the tethering of transcription to nuclear pores. In response to replication stress, checkpoint signalling targets nuclear pore components to disrupt this tethering and stabilize stalled forks .
The checkpoint may also stabilize stalled forks by promoting cohesion [60–62], or regulating histone dynamics at forks. The histone chaperone ASF1 is potentially regulated by ATM and CHK1, which inactivate TLK1 (tousled-like kinase 1) and TLK2, leading to dephosphorylation of ASF1 [177,178]. During replication inhibition, ASF1 buffers excess histones so that a pool of histones is readily available for deposition once the block has been removed and replication forks restart [48,51]. Changes in the regulation of ASF1 could potentially affect its promotion of replication fork progression , and evidence from yeast suggests that proper histone supply promotes the stability of replication forks .
Claspin, TIM and TIPIN may promote replication independently of ATR signalling
Certain components of the ATR pathway may promote replication fork progression independently of ATR signalling itself. Although claspin promotes replication fork progression in a manner similar to CHK1, it does not appear to control origin firing via CDC25 or CDK and seems to work in a parallel pathway to CHK1 [152,164]. Accordingly, down-regulating origin firing in claspin-deficient cells has only a small effect on increasing replication fork speeds . Claspin, as well as TIM and TIPIN, may instead promote replication fork progression through their direct interactions with the replication machinery [181–183]. These proteins are orthologues of budding yeast Mrc1 (Claspin), Tof1 (TIM) and Csm3 (TIPIN), which form a complex that promotes both replication fork progression and stability [184–186]. One way in which the Mrc1–Tof1–Csm3 complex is thought to achieve this is by helping to couple polymerase and helicase activities to prevent excessive unwinding and ssDNA formation [186,187], and a similar function has been proposed for its vertebrate homologues. TIM depletion increases spontaneous formation of ssDNA, supporting the idea that TIM/TIPIN could also counteract polymerase–helicase uncoupling . This function appeared to be independent of ATR signalling, just as some functions of Tof1 and Csm3 are independent of the ATR homologue Mec1 . Current evidence suggests that TIM and TIPIN may promote replication fork progression and stability by promoting proper sister chromatid cohesion through their interaction with the cohesion factor AND-1 (WD repeat and HMG-box DNA-binding protein 1; WDHD1) [165,190–192]. TIPIN and AND-1 also promote loading of DNA polymerase α, which could facilitate replication fork restart by re-priming downstream of a lesion .
Homologous recombination promotes replication fork stability and restart
Our models of HR-dependent pathways of replication fork restart are largely informed by observations from Escherichia coli (Figure 4) [193,194]. HR-dependent fork restart involves HJ and D-loop intermediates, which can be formed either directly from a stalled fork structure or after generation of a DSB at the fork. This process allows loading of the replication machinery in E. coli and budding yeast . Similarly, HR plays an important role in fork stabilization and restart in vertebrate cells. Defects in the HR proteins BRCA2, RAD51, MRE11, FANCA or FANCD2 lead to accumulation of spontaneous and replication inhibitor-induced DSBs, suggesting that these proteins stabilize stalled replication forks [166,167,195–197]. The HR proteins MRE11, RAD51 and XRCC3 have been reported to facilitate efficient replication fork restart after treatment with replication inhibitors [38,96,198]. MRE11 is recruited to stalled replication forks [199–201] and may perform end processing so as to generate the lagging strand gap, or 3′-overhang, which would be required for RAD51 loading. In addition, MRE11 could perform a number of other functions. It interacts with BLM and WRN and might collaborate with these helicases in functions other than resection (see below) [199,202]. Budding yeast Mre11 has been suggested to promote fork restart by facilitating sister chromatid cohesion .
Replication fork restart by homologous recombination
Once the correct single-stranded overhangs are available, XRCC3 could then promote RAD51 loading for D-loop formation and fork restart. RAD51-mediated fork restart seems to differ from HR-mediated DSB repair in that it does not involve RAD51 foci formation, which would be indicative of very long RAD51 filaments, or detectable long patch recombination in a reporter construct . Recombination-free restart could be supported by the BLM–TOPIIIa–hRMI1 complex, which resolves double HJ in a process that avoids crossing over . If stalled forks collapse and are processed into DSBs after longer replication blocks, RAD51-dependent HR is required for the repair of the breaks [38,170].
The role of HR at stalled replication forks in higher eukaryotes is complicated by the many additional factors that play roles during HR, such as BRCA1, BRCA2, PARP-1 [poly(ADP-ribose) polymerase 1] and the FA complex. BRCA2 and PARP-1 promote replication fork stabilization or restart, which is attributed to roles in the recruitment and regulation of MRE11 and/or RAD51. PARP-1, absent in yeast but the major PARP enzyme in vertebrate cells, is activated by DNA damaged structures, including ssDNA gaps, and has a variety of signalling and recruitment functions in the DNA damage response. PARP-1 interacts with MRE11 [198,204] and, in response to replication inhibitor treatment, it promotes MRE11 foci formation, ssDNA generation and replication fork restart (Figure 4), suggesting that PARP-1 promotes replication fork restart through MRE11-dependent DNA resection .
Supporting the idea that MRE11 may promote resection at stalled replication forks, the tumour suppressor BRCA2 was found to prevent excessive MRE11-dependent resection at stalled forks . RAD51 itself has a similar preventive effect, and BRCA2 is thought to prevent resection by regulating RAD51 loading and the stability of RAD51 filaments [205,206]. Absence of BRCA2 has not been found to impair replication fork restart, but it leads to fork collapse and increased genomic rearrangements after replication inhibitor treatment [195,205].
The effects of HR proteins on the speed of replication fork progression are still poorly understood. There is evidence that HR-deficient cells display decreased fork progression during normal S-phase [205,207]. This could result from a requirement for HR to stabilize or restart forks that stall when encountering endogenous obstacles. On the other hand, the presence of RAD51 and its paralogues XRCC3 and XRCC2 actively slow fork progression on templates containing bulky lesions induced by cisplatin. RAD54, which is involved in later stages of HR, was not required for this fork slowing. It therefore seems that it is RAD51 loading on to the damaged template that slows fork progression directly or indirectly . This suggests that RAD51 activity does not serve to promote fork progression in this situation but rather the opposite, although HR is important for survival of cisplatin damage through its repair function .
Fork-remodelling helicases promote replication fork progression and restart
The WRN, BLM, FANCM and SMARCAL1 DNA helicases have all been implicated in replication fork progression and restart (Figures 5A and 5B) [209–213]. Deficiency in both BLM and WRN causes slow fork progression in untreated cells, as well as defects in the restart of replication forks stalled by replication inhibitors or DNA damage [211–215]. BLM and WRN have the ability to unwind difficult-to-replicate secondary structures in concert with FANCJ, which may aid replisome progression [125,216,217]. In addition, BLM has been suggested to promote fork progression by indirectly modulating nucleotide pools. BLM-deficient cells display decreased levels of CDA (cytidine deaminase), which leads to a relative increase in cellular dCTP levels (Figure 5B). Re-balancing of pyrimidine pools by expression of CDA or the addition of deoxyuridine improved replication fork speeds, but not spontaneous fork stalling in BLM-deficient cells. This suggests that slow fork progression and replication fork restart defect in the absence of BLM may be two separate phenotypes . BLM-dependent replication fork restart is assisted by RIF1 (RAP1-interacting factor homologue), a DNA-binding protein of unknown biochemical function .
DNA helicases and REV1 promote replication fork progression and restart
BLM and WRN might both promote fork restart after replication blocks through their modulation of HR (Figure 5A). Both helicases could promote D-loop-mediated restart by facilitating fork regression to form an HJ [220,221]. BLM is also involved in one of the two MRE11-dependent pathways of DNA resection, which may promote RAD51 loading during fork restart .
Although promoting RAD51 function appears important for replication fork restart, BLM and WRN also possess anti-recombinogenic activities that could promote fork restart if HR activities are detrimental for restart under some circumstances (see above). BLM and WRN could prevent HR by reversing fork regression through their HJ migration activity and repressing aberrant formation of RAD51 filaments at stalled or collapsed forks [118,119,222]. Interestingly, ATR and ATM differentially phosphorylate WRN in response to replication blocks: ATR phosphorylation of WRN stabilizes stalled forks early during replication blocks, whereas ATM phosphorylation promotes HR repair of collapsed forks after long replication blocks . This could suggest that WRN has an anti-recombinogenic function at stalled forks, and raises the possibility that preventing HR can be a requirement for fork stability under some circumstances. BLM- and WRN-deficient cells treated with replication inhibitors display increased RAD51 foci formation, which could be interpreted as elevated unscheduled RAD51 filament formation . Alternatively, however, increased RAD51 foci formation could indicate an increased need for the repair of forks collapsed into DSB, as MUS81-dependent DSB formation at forks was observed in WRN-deficient cells  or a failure to complete HR. When remodelling replication forks, BLM and WRN may work together with the annealing helicase SMARCAL1. SMARCAL1 is recruited to ssDNA-containing lesions such as stalled replication forks, and its absence leads to both ssDNA accumulation and defective replication fork restart, supporting the idea that SMARCAL1 may promote fork stability and restart by re-annealing long stretches of ssDNA generated at stalled forks [210,224].
The FANCM helicase seems to have particularly intriguing effects on replication. DT40 cells deficient in FANCM (but not other FA proteins) displayed not decreased, but increased replication fork restart after release from camptothecin that was dependent on ATR, CDK and PLK activity . FANCM-deficient cells also displayed increased ATR-dependent phosphorylation of MCM2 and increased origin firing. This suggests that the ATR- and PLK1-dependent pathway of promoting origin firing during replication stress becomes activated in the absence of FANCM to allow replication restart in the vicinity of stalled forks (Figure 5A) [144,225]. Human cells lacking FANCM, or FANCM ATPase activity, displayed increased replication fork speeds and reduced accumulation of ssDNA during unperturbed S-phase and in the presence of replication inhibitors, suggesting that FANCM helicase activity slows replication fork progression (Figure 5B) . In contrast to the first report, however, this group found that the absence of FANCM reduced replication fork restart when cells were treated with camptothecin . One potential reason for this discrepancy could be that new origin firing was able to rescue replication in DT40, but not human (HeLa), cells.
REV1 and HLTF facilitate replication fork progression and restart
The TLS pathway can maintain replication fork progression on templates containing bulky adducts by enabling lesion bypass directly at the fork (early response), but it also performs post-replicative filling of ssDNA gaps left behind the replication fork (late response) . The REV1 polymerase in particular has been implicated in performing TLS directly at the fork, as it promotes replication fork progression on templates containing bulky adducts in chicken DT40 cells and mouse embryonic fibroblasts (Figure 5C) [226,227]. It seems that, in DT40 cells at least, this REV1-mediated mode of fork-associated lesion bypass works independently of RAD18 and PCNA ubiquitination, which instead promotes post-replicative gap filling by TLS . This process is aided by the WRN helicase . REV1 can also promote fork progression through secondary structures in the DNA, together with FANCJ [125,229]. In REV1-deficient cells, which are unable to bypass such lesions directly and rely on post-replicative gap filling instead, histone deposition at the site of lesions becomes uncoupled from replication fork progression. In this instance, specific silencing histone modifications are lost, which, should the lesion be located around a promoter region, can result in a loss of gene silencing [125,229].
The RAD5 homologue HLTF, which may be involved in the alternative ‘error-free’ mechanism, has also been implicated in replication fork progression (Figure 5D). Simlarly to RAD5, HLTF is a RING-domain ubiquitin ligase and SWI/SNF helicase. Via their helicase activity, Rad5 and HLTF can branch-migrate HJ and reverse replication forks in vitro, which has been suggested to enable template switching during the alternative pathway [230,231]. Indeed, the HLTF helicase activity promotes timely replication fork restart after release from the methylating agent MMS (methyl methanesulfonate) . In apparent contradiction to this, MMS treatment also leads to degradation of HLTF, which in turn promotes the activity of the second Rad5 homologue SHPRH to suppress MMS-induced mutagenesis . Since promotion of fork restart by HLTF was observed at earlier time points than HLTF degradation [133,231], it seems possible that HLTF and SHPRH may be involved in early and late responses to MMS damage at forks, such as fork progression and post-replication gap filling respectively.
Resolution of termination intermediates prevents genomic instability
Proper replication termination, and resolution of catenated DNA structures that may arise at termination sites, is ultimately required for the correct separation of sister chromatids during mitosis. Incomplete sister chromatid separation can be observed as DNA bridges that connect the daughter nuclei during anaphase. Especially, the late-replicating centromeres frequently form such anaphase bridges even during normal cell growth, though the bridges disappear towards telophase, suggesting that replication intermediates are still resolved during mitosis . In addition to TOPII, the complex of BLM, TOPIIIα and hRMI1 is required for the resolution of anaphase bridges . It was suggested that the latter complex helps resolve termination structures by a similar mechanism to its dissolution of double HJ structures during HR [232,233].
Interestingly, there is increasing evidence that additional anaphase bridges can arise from impaired replication fork progression or a failure to activate dormant origins [147,233,234]. This presumably occurs because cells enter mitosis without having completed replication. Anaphase bridges resulting from replication inhibition do not localise to centromeric regions and are marked by foci of FANCD2 and FANCI, suggesting that the FA pathway is activated by the perturbed replication structures that give rise to these bridges [233,234]. The FA pathway seems to promote BLM recruitment to help resolve non-centromeric bridges . Increased levels of unresolved replication intermediates and anaphase bridges have been implicated in the generation of DNA damage and genomic instability. Inhibition of replication fork progression can cause symmetrical DNA lesions during the following mitosis and G1-phase . Breakage of anaphase bridges can lead to the exclusion of chromosome fragments from the daughter nuclei to form micronuclei, resulting in loss of genetic material [147,234]. That this could be a common mechanism underlying genomic instability is supported by the observation that anaphase bridges and micronuclei localize to common fragile sites, late-replicating genomic loci that frequently display instability in cells treated with replication inhibitors and are often re-arranged in FA, Bloom's syndrome and cancer cells [233,234,236].
Our insights into the effects of DNA-damage-response factors on the initiation and progression of DNA replication have greatly increased in recent years, mainly due to advances in measuring replication dynamics. We are also starting to gain a better understanding of how changes to replication dynamics cause DNA damage and genomic instability. The emerging picture suggests that DNA-damage-response pathways are activated during normal S-phase by both endogenous DNA lesions and ssDNA intermediates at replication forks. The DNA damage response then promotes the proper progression of replication forks by: (i) regulating aspects of replication such as initiation, stability of the replication complex and sister chromatid cohesion; (ii) stabilizing transiently stalled forks by regulating a variety of cellular pathways such as transcription to preventing aberrant re-arrangements and processing of these forks; (iii) promoting replication restart by remodelling of damaged replication forks and promoting firing of dormant origins; and (iv) taking over from the standard replication machinery to facilitate fork movement on difficult-to-replicate templates. If fork progression is impaired, this can lead to increased generation of DSBs at forks through the action of fork-processing nucleases. In addition, prevention of normal replication termination due to impaired fork progression and restart can lead to incomplete sister chromatid separation followed by generation of DNA damage and genomic instability during mitosis. However, although we can observe the effects of the DNA damage response on replication dynamics, the exact nature of the molecular transactions at forks that underlie these observations are still largely subject to speculation, and further investigations and improved methodologies will be needed to test these speculative models.
One major issue that remains to be addressed is the nature of the DNA structures formed at perturbed replication forks in vivo, for example whether transactions such as fork regression occur in human cells. Finding ways to answer these questions would constitute a major breakthrough. Further open questions include whether the DNA damage response promotes proper DNA replication by co-ordinating and integrating it with other cellular mechanisms such as transcription or cell signalling. What are the mechanisms and consequences of replication-associated DNA damage in stem cells and during development? How do genetic defects in factors that promote replication fork progression lead to phenotypes from growth failure (e.g. ATR-Seckel syndrome) to cancer predisposition (e.g. Bloom syndrome)? An interesting new light has been shed on the latter question by reports that activation of oncogenes can cause perturbed replication fork progression, which may be important for genomic instability during cancer development [237–239]. This suggests that studying replication-associated DNA damage will advance our understanding of cancer development and may open up new possibilities for cancer therapy.
WD repeat and HMG-box DNA-binding protein 1
anti-silencing function 1
ataxia telangiectasia mutated
ATM- and Rad3-related
Bloom syndrome, RecQ helicase-like
breast cancer early onset
chromatin assembly factor 1
cell division cycle
chromatin licensing and DNA replication factor 1
Dbf4/Drf1-dependent CDC7 kinase
essential meiotic endonuclease 1 homologue 1
facilitates chromatin transcription
FA, complementation group E
Gen endonuclease homologue 1
helicase-like transcription factor
RecQ-mediated genome instability 1, homologue
meiotic recombination 11
MUS81 endonuclease homologue
origin recognition complex
proliferating cell nuclear antigen
replication fork barrier
replication factor C
SNF2 histone linker PHD RING helicase
SWI/SNF-related, matrix-associated, actin-dependent regulator of chromatin, subfamily a-like 1
Werner syndrome, RecQ helicase-like
X-ray repair complementing defective repair in Chinese hamster cells
We apologise to all authors whose work we could not cite due to space restrictions.
We thank Cancer Research UK and the Royal Society for financial support.