Controlling DNA replication origins in response to DNA damage – inhibit globally, activate locally

In bacteria, DNA replication generally initiates at a single, well-defined site (origin) on a circular chromosome. As a consequence, the time required to replicate a bacterial chromosome is proportional to its size: a twofold increase in chromosome size would be accompanied by a twofold increase in the time required for its replication. By contrast, eukaryotes replicate their genomes from multiple origins that are distributed on multiple chromosomes. So, the time required to replicate even relatively large genomes can be quite brief. For example, during the earliest embryonic cell cycles, the entire Drosophila melanogaster genome is replicated in 3–4 minutes (Rabinowitz, 1941). This is because replication origins are spaced quite close together (∼3.4 kb), and most or all origins initiate within the first 30 seconds of S phase (Blumenthal et al., 1974). In this case, the time required to replicate the entire genome is directly related to the distance between origins rather than the genome size. In most eukaryotic cells, however, origins do not fire synchronously at the beginning of S phase but instead are activated throughout S phase according to a cell-type-specific programme. This lengthens S phase, but affords the cell with additional opportunities to coordinate replication with other nuclear functions such as gene expression. The ability to replicate the genome from multiple origins was probably a crucial event in the evolution of eukaryotes that allowed genomes to increase in size and complexity, which is necessary for multicellularity. Nevertheless, the presence of multiple origins presents challenges to ensure that all parts of the genome are fully replicated in each S phase, but that no origin initiates for a second time in one cell cycle. These challenges are met by a two-step mechanism that consists of two distinct biochemical processes – ‘origin licensing’, which involves the loading of the replicative helicase at future origins, and ‘origin firing’, which involves the subsequent activation of the replicative helicase. These processes are separated temporally and tightly coupled to distinct phases of the cell cycle. This has important implications for how replication might be regulated during S phase when on-going replication forks encounter a problem.

Here, we review our current understanding of replication initiation by defining the main players in origin licensing and origin firing. We describe the mechanisms regulating replication during the cell cycle and discuss the significance of these mechanisms in the context of DNA damage. Finally, we summarise our current understanding of how origin firing is regulated in response to DNA damage, and we suggest ways in which the different effects described above are achieved and why this is important for maintaining genome stability.

Origin licensing involves the assembly of pre-replicative complexes (pre-RCs) that contain the core replicative helicase component Mcm2-7, consisting of the minichromosome maintenance proteins 2 to 7 (Mcm2 to Mcm7) at all potential replication origins (Fig. 1). This assembly requires the origin recognition complex (ORC), comprising six subunits, Orc1 to Orc6, as well as cell division cycle 6 (Cdc6) protein and the DNA replication factor Cdt1 (cell division cycle 10-dependent transcript 1) (for reviews, see Araki, 2010; Bell and Dutta, 2002; Blow and Dutta, 2005; Boos et al., 2012; Diffley, 2004; Masai et al., 2010; Remus et al., 2009; Sclafani and Holzen, 2007; Tanaka and Araki, 2010; Tanaka and Araki, 2011). ORC is required for the binding of all pre-RC components and is thus a key player in establishing the sites of all potential origins (see Box 1). In both budding yeast and Xenopus, two copies of the Mcm2-7 complex are loaded onto DNA as a double-hexamer (Evrin et al., 2009; Gambus et al., 2011; Remus et al., 2009). Both in vivo and in vitro, once it is bound to DNA, the Mcm2-7 complex is extremely stable, with a half-life of binding exceeding several hours (Evrin et al., 2009; Kuipers et al., 2011; Remus et al., 2009). This is important because the pre-RC must be retained at origins for many hours between the time of its assembly at the end of mitosis and the time of origin firing during S phase.

In contrast to Mcm2-7, ORC has a shorter residence time on chromatin (McNairn et al., 2005; Prasanth et al., 2010). It is therefore possible that a single ORC might be able to load multiple Mcm2-7 double-hexamers onto DNA at different sites during G1 phase. Indeed, estimates suggest that a 5- to 20-fold excess of Mcm2-7 double-hexamers is loaded onto chromatin relative to either the number of chromatin-bound ORCs or the number of functional origins from yeast to human (Donovan et al., 1997; Edwards et al., 2002; Hyrien et al., 2003). The importance of excess origin licensing will be discussed below.

During S phase, the Mcm2-7 double-hexamer becomes activated in a reaction known as ‘origin firing’. It is now clear that robust helicase activity requires the association of two additional factors with Mcm2-7: Cdc45 and the heterotetrameric GINS complex (for ‘Go, Ichi, Nii and San’; ‘five, one, two and three’ in Japanese) (Ilves et al., 2010). All of the components of this CMG (Cdc45–Mcm2-7–GINS) complex are essential, not just to initiate replication, but also during the elongation phase of replication (Gambus et al., 2006; Ilves et al., 2010; Kanemaki et al., 2003; Labib et al., 2000; Moyer et al., 2006; Pacek et al., 2006; Pacek and Walter, 2004; Tercero et al., 2000). A set of ‘firing factors’ is required for the formation of the CMG complex in vivo, but these factors are not subsequently required for elongation. Firing factors, which have so far been best characterised in yeast, include Sld2, Sld3, Sld7, Dpb11 and two protein kinases – cyclin-dependent kinase (CDK) and Dbf4-dependent kinase (DDK). DNA polymerase ε, the leading strand DNA polymerase, is also required for this step. Metazoan orthologues of all these proteins except for Sld7 have been identified. Most have the same names in yeast and humans, but the metazoan Sld2 orthologue is RecQL4 (Matsuno et al., 2006; Sangrithi et al., 2005), the Sld3 orthologue is treslin-TICRR (Kumagai et al., 2010; Sanchez-Pulido et al., 2010; Sansam et al., 2010) and the Dpb11 orthologue is TopBP1. There are also additional factors acting at this step that do not have obvious yeast orthologues, including GEMC1, Idas, DUE-B and Mcm9 (Balestrini et al., 2010; Chowdhury et al., 2010; Lutzmann and Méchali, 2008; Pefani et al., 2011).

The processes of origin licensing and origin firing must be separated during the cell cycle to ensure that origins fire just once in each cell cycle. Licensing only occurs during a window of time from late mitosis through early G1 phase, whereas origin firing cannot take place during G1 phase. As a consequence, once an origin fires during S phase, it cannot be relicensed, and consequently cannot refire, until the cell passes through the subsequent mitosis. Mechanisms that limit licensing to G1 phase in eukaryotes have been reviewed thoroughly and discussed elsewhere (Araki, 2010; Bell and Dutta, 2002; Blow and Dutta, 2005; Boos et al., 2012; Diffley, 2004; Diffley, 2010; Diffley, 2011; Masai et al., 2010; Remus and Diffley, 2009; Sclafani and Holzen, 2007; Tanaka and Araki, 2010; Tanaka and Araki, 2011). Briefly, licensing is restricted to G1 phase because three different cell-cycle-regulated inhibitors collaborate to prevent licensing outside of G1 phase. These inhibitors include CDKs, the small geminin protein and the Cul4-containing ubiquitin ligase. CDK inhibits multiple pre-RC components by a variety of mechanisms [discussed elsewhere (Diffley, 2004)], geminin is a stoichiometric inhibitor of Cdt1, and Cul4 ubiquitin ligase targets Cdt1 for ubiquitin-mediated proteolysis. Geminin, like the mitotic cyclins, is targeted for degradation by ubiquitin-mediated degradation by the anaphase-promoting complex/cyclosome (APC/C). Because the APC/C is active during G1 phase, two of the inhibitors, geminin and mitotic cyclins, are not present then. The third inhibitory pathway is primarily restricted to S phase because Cul4-dependent degradation of Cdt1 requires chromatin-bound proliferating cell nuclear antigen [PCNA, the replisome ‘sliding clamp’ (Havens and Walter, 2011)]. Therefore, during G1 phase, all three inhibitory pathways are inactive, which allows licensing to occur only during this period.

Although CDK inhibits licensing, it is nevertheless essential for origin firing in S phase. In budding yeast, phosphorylation of Sld2 and Sld3 by CDK is necessary and sufficient to promote initiation of replication by bridging phosphorylated Sld3, Dbp11 and phosphorylated Sld2 and thus recruiting Cdc45 and GINS (Tanaka et al., 2007; Zegerman and Diffley, 2007) (Fig. 1). Therefore, origin licensing and origin firing are temporally separated during the cell cycle by the reciprocal oscillations of APC/C and CDK activities. This ensures that replication origins fire just once in each cell cycle, and, as we will explain, it has implications for the DNA damage responses discussed below.

There are further levels of regulation of origin firing during S phase. First, more origins are licensed during G1 than are used during S phase; most remain ‘dormant’ and are passively replicated in S phase. And second, the time at which individual origins fire during S phase is not random, and different origins fire at different times during S phase (Pope et al., 2013). For convenience, origins are often described as being ‘early’ and ‘late’ firing, although, at least in budding yeast, origins fire in a continuum through S phase (Raghuraman et al., 2001). Experiments and mathematical modelling in fission yeast have indicated that this apparent hierarchy in origin firing is actually caused by underlying differences in the efficiencies of origins (Bechhoefer and Rhind, 2012; Rhind et al., 2010; Xu et al., 2012). Origin timing is a two-step process: the temporal programme is first determined early in the cell cycle (Dimitrova and Gilbert, 1999; Raghuraman et al., 1997), probably by establishing specific chromatin states around origins (see Box 2), and second, the time of origin firing during S phase is determined by the affinity of origins for limiting amounts of firing factors (Box 3).

Once in S phase, replisomes can encounter obstacles that hinder their progression and stability. These obstacles include DNA lesions, secondary DNA structures and large protein complexes bound to the DNA. In such cases, replication fork progression is blocked and the replisome is said to have ‘stalled’. Replication fork stalling can also be caused by other types of replication stress, such as the inhibition of the replicative polymerases, for example by aphidicolin, or a reduction in the available deoxyribonucleotide triphosphate (dNTP) pools by inhibition of ribonucleotide reductase (RNR) with hydroxyurea (HU). There are a number of cellular mechanisms that deal with replication obstacles, the discussion of which is outside the realm of this article and can be found elsewhere (Labib and De Piccoli, 2011; López-Contreras and Fernandez-Capetillo, 2010; Zegerman and Diffley, 2009).

When a stalled replication fork is not able to resume replication, a replication fork from a downstream origin might replicate up to the inactive fork, completing replication of this replicon (Fig. 2). If, however, the fork approaching from the downstream origin also fails, the DNA between the two forks cannot be replicated without re-establishment of replication forks. Although Escherichia coli has evolved several ways of reloading the replicative helicase at stalled forks (Heller and Marians, 2007), similar mechanisms have not yet been described in eukaryotes. Thus, replication of this region might only be possible by establishing forks at new replication origins. However, as discussed above, new replication origins cannot be established between these stalled forks because origin licensing is blocked during S phase. This problem is partially mitigated by the licensing of an excess of origins before S phase (Fig. 2) (Blow and Ge, 2009). Most of these origins are normally ‘dormant’, but a licensed dormant origin between the two stalled forks can be activated and can rescue both stalled forks. Consequently, origins can be thought of as playing a role not only in the normal replication of the genome, but also in fork rescue.

For this to work effectively, dormant origins must be controlled in two important ways. First, they must only be used when needed; and second, they must be inactivated if they are replicated ‘passively’ by replisomes originating from distal origins. How is this achieved? The ‘licensing factor model’ initially proposed by Harland and Laskey predicted that unreplicated chromatin is ‘licensed’ with a mark required for replication that is erased by the act of replication, thus distinguishing replicated and unreplicated portions of the genome (Harland, 1981; Harland and Laskey, 1980). We now know that this ‘license’ is the Mcm2-7 double-hexamer. Work with yeast has shown that dormant origins assemble ORC- and Cdc6-dependent pre-RCs, similar to active origins. Moreover, these pre-RCs are lost after the origin is passively replicated (Santocanale and Diffley, 1996), consistent with a replication-dependent ‘erasure’ of the license. Yeast studies have also provided a possible explanation for why dormant origins are normally inactive: they appear to be inherently later firing than the surrounding origins and thus can only be induced to fire if their passive replication is prevented (Kalejta et al., 1998; Mesner et al., 2003; Santocanale et al., 1999; Vujcic et al., 1999). Therefore, dormant origins, at least in yeast, are dormant because they are passively replicated from nearby, earlier-firing origins, which displaces pre-RCs before they have an opportunity to fire.

There is considerable evidence that dormant origins are activated when replication forks are slowed in metazoans. Using DNA fiber autoradiography after pulse-labelling with tritiated thymidine, it was originally shown that a decrease in the rate of progression of replication forks was accompanied by an increase in the number of origins fired (Ockey and Saffhill, 1976; Taylor, 1977). This result has been confirmed and extended using the technique of molecular combing, where long DNA molecules can be spread on glass coverslips. Newly synthesised DNA can be visualised by pulse labelling with halogenated nucleotides, which can be detected by immunofluorescence, and specific DNA sequences can be localised with fluorescence in situ hybridisation (FISH). Studying replication dynamics in an amplified locus in a modified Chinese hamster lung fibroblast cell line, Anglana and colleagues found that slowing replication fork progression with HU had complex effects on origin firing – one normally dominant origin was suppressed, but several nearby dormant origins became activated (Anglana et al., 2003). Similarly, inter-origin distances decrease in Xenopus egg extracts and human cells after replicative stress (Ge et al., 2007; Woodward et al., 2006), consistent with the activation of additional origins. This suggests a model in which a decrease in fork speed is compensated by an increase in the number of origins fired, so that the overall length of S phase can remain unchanged.

There is now very strong correlative evidence indicating that the availability of dormant origins is important for cell viability and genome stability under conditions of replicative stress. For instance, experimental reduction in the levels of individual Mcm2-7 subunits by small interfering RNA (siRNA) causes a decrease in the activation of dormant origins and results in increased cell lethality after treatment with agents that interfere with replication in Caenorhabditis elegans and human cells (Ge et al., 2007; Ibarra et al., 2008; Woodward et al., 2006). Moreover, mice expressing reduced levels of Mcm2 (Pruitt et al., 2007) or an Mcm4 mutant protein (Shima et al., 2007) exhibit a number of phenotypes, including early onset of cancer. Interestingly, the viability of mice expressing mutant Mcm4 is further compromised in background strains that have an intrinsically lower number of active origins, suggesting that the phenotypes observed in these mice are due to a lower number of licensed origins (Kawabata et al., 2011). It should be noted that, in all of the above studies, origin numbers were reduced by compromising the Mcm2-7 complex. Thus, although the phenotypes described above correlate with reduced numbers of dormant origins, it is also formally possible that they are attributable to some other function of Mcm2-7. For example, it has been shown that Mcm2-7 also has a role in origin-independent break-induced replication (BIR) in yeast (Lydeard et al., 2010). It will be important to confirm these results by reducing the loading of Mcm2-7 with a different approach, for example by reducing the levels of licensing factors not involved in BIR, such as ORC or Cdc6.

Although the previous section described experiments showing an increase in origin use after DNA damage, there is extensive evidence that DNA damage and replication stress can also inhibit further origin firing in budding yeast and mammalian cells through the activation of DNA damage checkpoints (although this might not be the case in fission yeast) (Hayashi et al., 2007; Heichinger et al., 2006; Mickle et al., 2007). The inhibition of origin firing by the S phase checkpoint was first inferred from studying cells from patients with the rare genetic disorder ataxia-talangiectasia (AT), which is caused by mutations in the ataxia telangiectasia mutated (ATM) gene, one of the apical PI-3-kinase-related kinases in metazoan checkpoints. Ionising radiation (IR) significantly inhibits DNA synthesis in normal cells, but, in AT cells, replication is not as severely inhibited. This DNA synthesis seen specifically in checkpoint-deficient cells is known as radio-resistant DNA synthesis (RDS). Using sucrose gradient fractionation, it was shown that RDS correlates with a failure to suppress the appearance of low-molecular-weight nascent DNA, postulated to arise from new initiation events (Painter, 1977; Painter, 1985; Painter and Young, 1980). Through the use of molecular techniques to examine origin firing at specific loci, it has been shown that some RDS in AT cells is due to premature entry of G1 cells into S phase, and some of it is also due to an inability of S phase cells to inhibit origin firing (Larner et al., 1999; Lee et al., 1997). Moreover, an active checkpoint was shown to prevent the appearance of the ‘late S’ replication foci after replication stress, indicating a global checkpoint-dependent shutdown of late origin firing (Dimitrova and Gilbert, 2000; Zachos et al., 2003). Finally, evidence that DNA damage and replication stress inhibit origin firing has come from studies, in both human cells and Xenopus egg extracts, that used DNA fibre labelling to assess replication dynamics at the level of single replication forks (Marheineke and Hyrien, 2001; Marheineke and Hyrien, 2004; Merrick et al., 2004).

Studies in budding yeast have led to a molecular understanding of how the S phase checkpoint inhibits origin firing. After treatment with the alkylating agent MMS, S phase progression of yeast cells is slowed substantially, which depends on Mec1, the orthologue of the human ATM-related kinase (ATR), and its downstream kinase Rad53, the yeast functional analogue of CHK1 (Paulovich and Hartwell, 1995). It was later shown that the rate of fork progression upon MMS treatment is similar in wild-type cells and in checkpoint mutants (Tercero and Diffley, 2001); however, late origins become activated only in checkpoint mutants (Shirahige et al., 1998; Tercero and Diffley, 2001), indicating that the checkpoint-dependent slowing of S phase is due to a global inhibition of origin firing. The checkpoint also blocks late origin firing in HU-arrested cells, and it was shown that late origins in wild-type cells remain in the pre-replicative state (Santocanale and Diffley, 1998; Santocanale et al., 1999). This indicates that the checkpoint-dependent block of replication initiation occurs at an early stage in the transition from the pre-RC to the activation of the Mcm2-7 helicase. Two reports have shown that Rad53 prevents late origin firing by phosphorylating and inhibiting the two key firing factors Sld3 and Dbf4 (Fig. 1). Yeast strains harbouring mutants in Sld3 and Dbf4 that can no longer be phosphorylated by Rad53 are unable to prevent late origin firing in the presence of replicative stress such as HU or MMS (Lopez-Mosqueda et al., 2010; Zegerman and Diffley, 2010).

There is evidence in metazoans that origin firing might be prevented by similar mechanisms. Similar to the situation in yeast (Zegerman and Diffley, 2010), the CDK-dependent interaction between the Sld3 and Dpb11 orthologues treslin-TICRR and TopBP1 is blocked by HU in HeLa cells, and this block is relieved by treatment with a small-molecule inhibitor of CHK1 (Boos et al., 2011). This inhibition might be due to direct phosphorylation of treslin-TICRR, but it is also possible that the inhibition works by reducing CDK and/or DDK activity. There is evidence that CDK activity is inhibited in response to DNA damage in human cells through the CHK1-dependent degradation of Cdc25A, a phosphatase that is required for CDK activation in S phase (Falck et al., 2001). There is evidence from Xenopus egg extracts and human cells that DDK might also be regulated in response to the activation of DNA damage checkpoints during S phase (Costanzo et al., 2003; Heffernan et al., 2007; Matsuoka et al., 2007).

At first glance, these two effects of replicative stress on origin activation seem contradictory: some experiments suggest that dormant origins become activated by stalled forks, whereas others suggest that origin firing is inhibited by stalled forks. It is possible that these differences are influenced by the different experimental systems used. However, these differences more likely reflect differences in the methodology used to measure origin firing. In all of the experiments that demonstrate origin activation after fork stalling, the dormant origins that became activated were always in close proximity to labelled (and presumably stalled) replication forks, whereas in the experiments demonstrating checkpoint-dependent origin inhibition, total (i.e. global) origin firing was monitored. This suggests that dormant origins near the stalled forks are activated, whereas the majority of unfired origins distal to the stalled forks are inhibited.

This makes sense: the inhibition of distal, late-firing origins might prevent the generation of additional replicative stress in unreplicated regions of the genome, whereas local activation of dormant origins might direct replication towards regions actively struggling with stalled forks, ensuring the completion of replication in these domains. This implies that dormant origins around stalled forks manage to fire under conditions in which the general origin firing is inhibited by the S phase checkpoint. The molecular mechanism underlying this paradox is unknown, but below we suggest some possibilities. Underpinning all of the models below is the assumption that origins within a local ‘cluster’ fire stochastically (Blow and Ge, 2009). That is, all origins within a cluster have a fixed probability of firing per unit of time, which is probably based on their relative affinities for limiting firing factors. Firing of dormant origins within the cluster in the presence of replication stress can then be explained as follows: when a fork from the first fired origin is slowed down or becomes stalled, adjacent dormant origins have more time, and therefore a higher probability to fire before being passively replicated and inactivated. But the question now arises at to why these origins are not inhibited by the DNA damage checkpoint. It should be noted that the models described below are not mutually exclusive, and a combination of these mechanisms might be used by cells (Fig. 3).

Local dormant origins might escape checkpoint inhibition because of timing, which we describe as being ‘passive’. One possibility is that there is a short delay between the initial local fork stalling and the activation of the global S phase checkpoint, and this short delay gives dormant origins more time to become activated (model IA). This model predicts that the origins that are inhibited globally by the checkpoint are later-firing than the origins within the cluster where the fork has stalled.

Alternatively (model IB), dormant origins within the local cluster might have not yet fired but could have passed the step in origin activation that is inhibited by the checkpoint before initial fork stalling and checkpoint activation. There is currently no direct evidence for either of these models in metazoans, but there is some support for model IB in budding yeast. As described above, Sld3 and Cdc45 are recruited to early-firing origins early in G1 phase, long before any origin firing (Aparicio et al., 1997; Gambus et al., 2006; Kamimura et al., 2001). It has recently been shown that this recruitment requires DDK (Tanaka et al., 2011). As the checkpoint works, at least in part, through inhibition of DDK (Lopez-Mosqueda et al., 2010; Zegerman and Diffley, 2010), it is possible that these early origins are refractory to checkpoint inhibition because DDK has already executed its function at these origins.

The second model postulates the existence of a ‘protective factor’ that is localised to active replication clusters and prevents checkpoint inhibition of dormant origin firing. One might imagine that the recruitment of an anti-checkpoint factor, for example a checkpoint-kinase inhibitor or a phosphatase that counteracts checkpoint phosphorylation, to stalled forks might make local firing factors more refractory to checkpoint inhibition, thus allowing local dormant origin activation even in the presence of an active checkpoint pathway.

A third model posits that the checkpoint actively promotes local dormant origin firing, while still blocking global origin firing. Consistent with such a role for the S phase checkpoint in the activation of dormant origins, studies in Xenopus egg extracts have suggested that phosphorylation of Mcm2-7 by ATR leads to the recruitment of another cell-cycle-regulated kinase, Plk1, to the chromatin, which in turn suppresses the inhibition of origin firing by the S phase checkpoint (Trenz et al., 2008).

Further work is required to determine, which, if any, of these models is responsible for the observed differences in the regulation of local and global origin firing by the S phase checkpoint. Regardless of the precise mechanism, this strategy allows the firing of dormant origins to rescue locally stalled forks, while globally preventing further initiation events under conditions of general replication stress.

Eukaryotic cells have developed a sophisticated system to protect against the potential dangers of being unable to license origins after G1 phase by packing the genome with an excess of potential origins, before carefully orchestrating their activation during S phase to ensure complete genome replication, while at the same time preventing any re-replication. There is still a great deal to learn about the mechanisms that are involved in origin selection, licensing and activation in unperturbed cells and in response to replication stress. Further mechanistic insights should inform system-wide approaches to model origin activation, which could provide further insights into this complex process.