Dephosphorylation of the pre-initiation complex during S-phase is critical for origin firing

Genome stability requires complete DNA duplication exactly once before cell division. In eukaryotes, cyclin-dependent kinase (CDK) plays a dual role in this regulation by inhibiting helicase loading factors before also activating origin firing. CDK activates initiation by phosphorylation of two substrates, Sld2 and Sld3, forming a transient and limiting intermediate – the pre-initiation complex (pre-IC). The importance and mechanism of dissociation of the pre-IC from origins is not understood. Here we show in the budding yeast Saccharomyces cerevisiae that CDK phosphorylation of Sld3 and Sld2 is specifically and rapidly turned over during interphase by the PP2A and PP4 phosphatases. Inhibiting dephosphorylation of Sld3/Sld2 causes dramatic defects in replication initiation genome-wide, retention of the pre-IC at origins and cell death. These studies not only provide a mechanism to guarantee that Sld3 and Sld2 are dephosphorylated before helicase loading factors but also uncover a novel positive role for phosphatases in eukaryotic origin firing.


Introduction
Eukaryotes replicate their genomes from multiple origins that must initiate only once during the cell cycle. This is achieved by complete separation of the DNA loading of the replicative Mcm2-7 helicase (also known as licensing) and the activation of this helicase into different phases of the cell cycle 1 . Cyclin-Dependent Kinase (CDK) plays a vital dual role in this regulation both as an inhibitor of licensing and together with Dbf4-dependent kinase (DDK) as an activator of the helicase in S-phase 2 .
Although the exact details of how CDK and DDK mediate replication initiation is not fully understood, DDK directly phosphorylates Mcm2-7 double hexamers to generate a binding site for the initiation factors Sld3/Sld7, while CDK phosphorylates Sld3 and an additional initiation factor Sld2, which results in their phospho-dependent interaction with the BRCT repeats of Dpb11 2 . This complex formed by DDK and CDK on Mcm2-7 double hexamers is called the pre-initiation complex (pre-IC) and forms transiently at origins during the initiation reaction 3,4 . As Sld2, Sld3 and Dpb11 are not part of the replisome, they must be released by some mechanism during origin firing 2 . It is not currently known whether the release of Sld2, Sld3 and Dpb11 is in itself an important step in the assembly of the replisome.
To ensure the complete replication of the entire genome, the number and timing of replication initiation events across chromosomes is also regulated during S-phase in eukaryotes 5 . Although factors such as chromatin context determine the timing and likelihood of origin firing in S-phase 5 , we and others have shown that key components of the pre-IC, including Sld2, Sld3 and Dpb11 as well as the Dbf4 subunit of DDK are stoichiometrically low in abundance and limit the number of simultaneous initiation events [6][7][8] . This suggests that the active removal of the pre-IC from origins may be necessary to release and recycle these low abundance factors for further origin firing in a single S-phase ( Figure 1A).
Here we set out to analyse the mechanism of pre-IC regulation during S-phase in the budding yeast Saccharomyces cerevisiae. We show that the two critical CDK targets Sld2 and Sld3 are actively and specifically dephosphorylated during S-phase. The counteraction of Sld3/Sld2 phosphorylation is mediated by the PP2A and PP4 phosphatases and failure to dephosphorylate these pre-IC factors results in dramatic defects in replication initiation in vivo. Together we show that beyond the known role for kinases in origin firing, the dephosphorylation and dissolution of the pre-IC is a novel critical step in the eukaryotic replication reaction.

Sld2 and Sld3 are rapidly and specifically dephosphorylated in S-phase
As the CDK-dependent pre-IC complex is a transient intermediate during origin firing, we hypothesised that a phosphatase might be required for the dissolution of this complex during S-phase ( Figure 1A). To test this in budding yeast we inhibited CDK specifically in S-phase by synchronising cells containing the analogue sensitive (as) allele of the CDK catalytic subunit Cdc28, which is rapidly inhibited by the addition of 1-NM-PP1 9 ( Figure 1B). Analysis of CDK targets, Sld3, Sld2, Sli15, Yen1 and Orc6 revealed that these proteins are hypo-phosphorylated in G1 phase and phosphorylated as CDK levels rise on entry into S-phase (20 minutes, Figure 1C), as expected. Importantly, after addition of 1-NM-PP1 at 25 minutes, both Sld2 and Sld3 were rapidly dephosphorylated, but other CDK targets remained phosphorylated ( Figure 1C). Unlike the rapid dephosphorylation of Sld2/Sld3, the slow accumulation of unphosphorylated Orc6 ( Figure 1C) was reduced by cycloheximide treatment suggesting that this is due to new protein synthesis (Supplementary Figure 1A). Sld3 has 12 CDK sites, two of which are essential for replication initiation, T600 and S622 10,11 . To determine whether the rapid dephosphorylation of Sld3 after inhibition of CDK occurs at these essential sites we analysed the phosphorylation of Sld3 mutants that either have all CDK sites mutated to alanine (12A), or just the 10 nonessential sites (10A, Figure 1D). As expected, the wild-type protein was phosphorylated in S-phase and dephosphorylated upon addition of 1-NM-PP1, while the 12A mutant was not detectibly phosphorylated ( Figure 1D). Significantly, the 10A mutant, which retains the essential CDK sites, was still phosphorylated and rapidly dephosphorylated upon addition of 1-NM-PP1 ( Figure 1D), suggesting that the essential CDK sites in Sld3 that are required for replication initiation are rapidly dephosphorylated in S-phase in the absence of CDK.
Previous studies have shown that phosphorylation of Sld2 at multiple CDK sites allows phosphorylation of the CDK site (T84) required for binding to Dpb11 12 .
Consistent with this co-dependence, any combination of mutations in the CDK sites of Sld2 abrogated all detectible phosphorylation (Supplementary Figure 1B-C), which prevented us from narrowing down which CDK sites in Sld2 are specifically dephosphorylated in vivo. Despite this, these data show that the CDK phosphorylation of both Sld3 and Sld2 are rapidly reversed in S-phase if CDK is inhibited.

Sld2 and Sld3 are specifically dephosphorylated throughout S-and G2-phase independently of Rif1 and Cdc14
Yeast strains that lack the cyclin Clb5 have a temporal gap in CDK activity in Sphase 13 and in accordance with Figure 1 we wondered whether this gap would result in Sld2/Sld3 dephosphorylation. As expected, both Sld2 and Sld3 became phosphorylated in early S-phase in the clb5∆ mutant strain (20 minutes, Figure 2A), likely due to the activity of the alternative S-phase cyclin Clb6. Importantly, by 30 minutes both Sld2 and Sld3, but not Orc6, became dephosphorylated ( Figure 2A).
Interestingly, Sld2 phosphorylation increased again from 40 minutes, perhaps due to the accumulation of mitotic cyclin-CDKs, but Sld3 did not, suggesting different specificities of the CDK complexes ( Figure 2A). This rapid and specific dephosphorylation of the key initiation factors Sld2/Sld3 in mid-S-phase may explain why only early origins initiate in clb5∆ strains 14 .
To address the nature of this Sld2/Sld3 phosphatase, we explored when during the cell cycle this phosphatase activity was detectable. By synchronising the cdc28-as1 strain in G1-phase and releasing into S-phase, we observed that addition of 1-NM-PP1 during early S-phase, late S-phase or G2 phase ( Figure 2B) all resulted in rapid dephosphorylation of both Sld3 and Sld2, suggesting that the phosphatase activity is present during these periods of the cell cycle. Sld2 and Sld3 dephosphorylation was also not dependent on DNA replication as inhibition of helicase loading, using a temperature sensitive allele of Cdc6 (cdc6-1), prevented replication but not dephosphorylation (Supplementary Figure 2A).
It has been previously shown that the phosphatase required for mitotic exit in budding yeast, Cdc14, can dephosphorylate Sld2 and Orc6 in vivo 15,16 . Cdc14 is sequestered in the nucleolus until anaphase 17 , suggesting that this is not the activity that is dephosphorylating Sld3/Sld2 from early S-phase onwards ( Figure 2B). Despite this, some non-nucleolar activity has been detected for Cdc14 before anaphase 18 .
To test whether Cdc14 is responsible for Sld2/Sld3 dephosphorylation before mitotic exit, we utilised a temperature sensitive allele of Cdc14 (cdc14-1) to inhibit this phosphatase in nocodazole arrested cells. Importantly inhibition of cdc14-1 prevented the dephosphorylation of Orc6, but not Sld2 or Sld3 ( Figure 2C and Supplementary Figure 2B). This suggests that unlike Orc6, Sld2 and Sld3 are dephosphorylated independently of Cdc14 before mitosis.
The PP1 phosphatase, through binding to Rif1, has been shown to dephosphorylate DDK substrates 5 . To address whether PP1/Rif1 might also be required for Sld2/Sld3 dephosphorylation we performed an experiment as in Figure 1C, but in a strain lacking Rif1. Significantly, while DDK-dependent Mcm4 phosphorylation was insensitive to inhibition of CDK, Sld2 and Sld3 were rapidly dephosphorylated in a rif1∆ strain (Supplementary Figure 2C). This again demonstrates the specificity of the targeted dephosphorylation of Sld2/Sld3 over other phosphorylated replication factors and shows that neither PP1/Rif1 nor Cdc14 are involved in Sld2/Sld3 dephosphorylation in S-phase.

Sld2 and Sld3 are dephosphorylated in S-phase by PP2A and PP4
To narrow down the phosphatases responsible for Sld2/Sld3 dephosphorylation we used a chemical genetics approach (Supplementary Figure 3A). Two broad specificity phosphatase inhibitors, cantharidin and 9,10-phenanthrenequinone (PQ), appeared to abrogate Sld3 dephosphorylation in vivo (Supplementary Figure 3B).  23 and therefore strains with PP2A mutants are also swe1∆ to overcome these defects 23,24 . The loss of SWE1 does not affect the dephosphorylation phenotypes we observe (e.g Figure 3D).
PP2A is a trimeric complex including a catalytic (C, Pph21/Pph22 - Figure 3C), scaffold (A, Tpd3) and a specificity (B) subunit (Cdc55 or Rts1) 25 . As with the pph21∆ pph22∆ strains, tpd3∆ resulted in hyper-phosphorylation of Sld3 ( Figure 3D Although PP2A mutants have a dramatic effect on Sld3 dephosphorylation, some dephosphorylation was still observed in the presence of 1-NM-PP1 (e.g Figure 3A), which was not the case after cantharidin treatment (Supplementary Figure 3D). It has been shown that in Pph21/Pph22 deficient cells another cantharidin-sensitive PP2A family phosphatase Pph3, which is the catalytic subunit of the PP4 complex, can contribute residual phosphatase activity 26,27 . To test whether Pph3 contributes to Sld2 and Sld3 dephosphorylation we combined null mutations of the PP2A scaffold subunit tpd3∆ with pph3∆. Importantly while loss of Pph3 alone had a small effect on Sld3 or Sld2 dephosphorylation ( Figure 3E and 3F) the tpd3∆ pph3∆ double mutant completely lacked any Sld3 dephosphorylation in the presence of 1-NM-PP1 ( Figure   3E, right panel), similar to cantharidin (Supplementary Figure 3D). This phosphorylation of Sld3 was not due to spurious Rad53 activation (Supplementary Figure 4F). For Sld2, we still observed some dephosphorylation in the tpd3∆ pph3∆ double mutant ( Figure 3F), but quantification of the hypo-phosphorylated forms of Sld2 revealed that the dephosphorylation of Sld2 was dramatically reduced in the tpd3∆ pph3∆ strain ( Figure 3G).
PP2A and Pph3 dephosphorylate the essential CDK sites of Sld3 because the 10A mutant, which only retains the 2 essential CDK sites ( Figure 1D), was not dephosphorylated in the pph3∆ tpd3∆ double mutant strain ( Figure 3H). Together, these data show that both PP2A RTS1 and PP2A CDC55 , but also PP4 (Pph3) are responsible for the dephosphorylation of the CDK sites in Sld3 and also in large part for the dephosphorylation of Sld2 ( Figure 3I).

Sld3 interaction with Rts1 contributes to dephosphorylation
Having established that PP2A and Pph3 control Sld3/Sld2 dephosphorylation in Sphase we wanted to determine the physiological role of this regulation. Unfortunately, due to the large number of functions of these phosphatases, combined mutations of PP2A and Pph3 are extremely sick 26,27 . To examine specifically the functional importance of Sld3/Sld2 dephosphorylation we set out to identify phosphatase-interaction mutants of Sld2/Sld3 that would be defective only in the regulation of these CDK targets. A short linear motif (SLiM) has been identified that is required for the recruitment of the mammalian orthologue of Rts1 (B56) to substrates 28,29 . Using the B56 binding-site prediction software 29 , we identified a high probability Rts1 binding site (LxxIxE) in Sld3, starting at amino acid 616 ( Figure 4A and Supplementary Figure 5A). Interestingly, this LxxIxE motif is immediately followed by one of the essential CDK sites in Sld3, S622 ( Figure 4A).  Figure 4D). This triple mutant that lacks the interaction with PP2A RTS1 and lacks PP2A CDC55 and PP4 (Pph3) also resulted in abundant phosphorylation of Sld3 in G1 phase ( Figure 4D), perhaps due to CDK phosphorylation remaining from previous cycles (see Discussion). Importantly, loss of Sld3 dephosphorylation in the sld3-R cdc55∆ pph3∆ strain was not due to inappropriate Rad53 activation ( Figure 4D), occurred at the essential CDK sites ( Figure 4E) and did not affect Sld3 protein stability (Supplementary Figure 5F).
Together these data show that we have identified a separation of function mutant of Sld3 that can no longer be targeted by PP2A RTS1 and in combination with mutation of PP2A CDC55 and PP4 results in abrogation of the dephosphorylation of Sld3 in Sphase.
Since Sld2 is also dephosphorylated in a PP2A-dependent manner ( Figure 3F/G) we wondered whether the Rts1-binding mutant of Sld3 might affect Sld2 phosphorylation in trans. Although fully defective in Sld3 dephosphorylation ( Figure 4D), the sld3-R cdc55∆ pph3∆ strain was only partially defective in Sld2 dephosphorylation ( Figure   4F and Supplementary Figure 5E).

Failure to dephosphorylate Sld3 and Sld2 causes a defect in origin firing
To specifically test for S-phase defects resulting from the failure to dephosphorylate  Figure 6C with Figure   4D). This suggested that the conditional degron alleles of Cdc55 and Pph3 combined with the sld3-R mutant is a suitable inducible strain to analyse specifically the Sphase functions of the dephosphorylation of Sld3/Sld2.
Flow cytometry analysis revealed that failure to dephosphorylate Sld3 (cdc55-AID pph3-AID sld3-R) resulted in a dramatically slower S-phase (purple versus grey, Figure 5A and 5B), which was not due to defects in the G1-S transition (as determined by budding index, Figure 5C), nor due to aberrant Rad53 activation ( Figure 5D). To analyse this defect in S-phase progression in more detail we determined the dynamics of DNA replication genome-wide by high-throughput sequencing and copy-number analysis. Plotting the median replication time (Trep) produces a profile whereby peaks delineate origins and troughs represent termination zones ( Figure 5E, Chromosome XIV is shown as an example). While the control strain revealed peaks of early and later origin firing (e.g ARS1415 and ARS1411 respectively, black line Figure 5E), the cdc55-AID pph3-AID sld3-R mutant that cannot dephosphorylate Sld3 showed a delay in early origin firing and a dramatic defect in late origin firing (purple line, Figure 5E). Significantly we did not detect any difference in replication elongation rates (as determined by the slope of Trep between origins), suggesting that the slow S-phase in the cdc55-AID pph3-AID sld3-R strain is due to initiation not replisome progression defects (data not shown).
Analysis of all origins, binned into quintiles according to their normal Trep, revealed that every origin group was delayed in the cdc55-AID pph3-AID sld3-R strain, even the earliest firing origins (yellow boxes, Figure 5F), suggesting that dephosphorylation of Sld3 is required for all origin firing. Phosphatase mutant combinations with only a partial defect in Sld3 dephosphorylation also showed a partial defect in S-phase progression by flow cytometry (orange/blue lines, Figure 5B) and by copy number analysis ( Figure 5E/F). These combinations reveal that the replication phenotype is not simply due to any one of the cdc55-AID pph3-AID sld3-R alleles having a defect in origin firing, but instead is an additive effect, similar to Sld3 dephosphorylation (e.g Figure 4C/D). Therefore, using the conditional Cdc55/Pph3 mutants in synchronised cells, combined with the separation of function allele of Sld3 that cannot bind to PP2A RTS1 , Figure 5 demonstrates that dephosphorylation of Sld3/ Sld2, is important for origin firing genome-wide.

Failure to dephosphorylate Sld3/Sld2 delays the release of the pre-IC from origins
Since dephosphorylation of the pre-IC proteins Sld3/Sld2 is important for origin firing ( Figure 5), we wondered whether this dephosphorylation is important for pre-IC release from origins ( Figure 1A). Detection of the Mcm2-7 complex at origins through Mcm4 ChIP-seq ( Figure 6A) showed that Mcm2-7 is loaded at all origins in G1 phase in both the control strain (cdc55-AID) and the strain that is defective in the dephosphorylation of Sld3/Sld2 (cdc55-AID pph3-AID sld3-R). Importantly this demonstrates that the reduction in origin firing observed in the cdc55-AID pph3-AID sld3-R strain ( Figure 5) is not due to a licensing defect. By 40 minutes after release from G1 phase the Mcm2-7 complex becomes delocalised from the earliest origins in the control strain, as initiation occurs ( Figure 6A, note that the heatmaps are ordered from early to late firing origins). In the cdc55-AID pph3-AID sld3-R strain however, the movement of the Mcm2-7 complex away from origins was greatly delayed ( Figure   6A), consistent with a reduction in initiation at all origins when Sld3/Sld2 dephosphorylation is defective ( Figure 5).
To analyse pre-IC dynamics we first performed Sld3 ChIP-seq but we could not achieve sufficient enrichment for Sld3 by this method (data not shown). Instead, we analysed Sld3-ChIP by qPCR, as previously described 3 , at three early firing origins ( Figure 6B). In the control strain, Sld3 binds to early origins in G1 phase and is released from origins during S-phase ( Figure 6B). In the cdc55-AID pph3-AID sld3-R strain however, Sld3 accumulated at these origins during S-phase, relative to G1 ( Figure 6B), consistent with a delayed release of this protein from origins in the absence of dephosphorylation. For Sld2, we observed a transient interaction with origins during S-phase as expected 3 , which at 25 minutes was preferentially at later firing origins in the control strain ( Figure 6C). In the cdc55-AID pph3-AID sld3-R strain Sld2 was still detected at early firing origins at 25 minutes ( Figure 6C), reflecting the delay in initiation and delay in the release of Sld3 from early origins ( Figure 6B). Sld2 is released from origins to a greater extent than Sld3 in the cdc55-AID pph3-AID sld3-R strain ( Figure 6B versus 6C), likely because Sld2 is still dephosphorylated to a significant degree in this background ( Figure 4F). Together these ChIP data are consistent with a role for dephosphorylation of Sld2/Sld3 in the release of the pre-IC from origins during initiation.

Dephosphorylation of Sld3 and Sld2 is essential
Significantly, combination of cdc55∆, pph3∆ and sld3-R, which abrogates Sld3/Sld2 dephosphorylation ( Figure 4D/F) and causes genome-wide defects in replication initiation ( Figure 5) was largely inviable, except for a small number of microcolonies ( Figure 7A). This synthetic lethality was also observed with the cdc55-AID pph3-AID sld3-R strain in the presence of auxin (Supplementary Figure 6D).
If the phenotypes of sld3-R are due to loss of Rts1 binding ( Figure 4B), rather than another function of Sld3, then we reasoned that restoring an Rts1 binding site to the sld3-R mutant should rescue these phenotypes. Addition of the Rts1 SLiM of Sld3 (L616-N625) to the C-terminus of the sld3-R mutant (referred to as sld3-R+SLiM, Figure 7B) led to a complete rescue of spore viability and colony size when combined with cdc55∆ pph3∆ ( Figure 7B and Supplementary Figure 7A). This demonstrates that the lethality of the sld3-R mutant in combination with loss of PP2A CDC55 /PPH3 is very likely due to loss of a direct interaction between Sld3 with Rts1.
Since defects in the dephosphorylation of Sld2 and Sld3 leads to excess CDK phosphorylation of these targets, we wondered whether the partial inhibition of CDK might also counterbalance the defects in dephosphorylation (Supplementary Figure  7B). The cdc28-as1 mutant has reduced activity even in the absence of 1-NM-PP1 9 and importantly we observed that this allele partially suppressed the growth defect of the cdc55∆ pph3∆ sld3-R mutations (Supplementary Figure 7C). This suppression by cdc28-as1 explains how we obtained viable cdc55∆ pph3∆ sld3-R mutants for the analysis of dephosphorylation (e.g in Figure 3/4) and also strongly suggests that it is the hyper-phosphorylation of Sld3/Sld2 by CDK that causes the synthetic lethality of the combined phosphatase mutants in vivo.
As failure to dephosphorylate the pre-IC results in retention of this complex at origins ( Figure 6), we hypothesised that over-expression of subsets of pre-IC proteins might interfere with complex stability at origins, which might lead to suppression of the phosphatase mutant phenotype. To test this, we over-expressed the Sld3-R mutant that cannot bind to PP2A RTS1 with and without over-expression of Dpb11. Overexpression of Sld3-R significantly rescued the growth defect associated with the cdc55-AID pph3∆ sld3-R strain, which was further enhanced by over-expression of Dpb11 ( Figure 7C). This over-expression also rescued the S-phase defect associated with failure to dephosphorylate Sld3 and Sld2, without affecting the G1-S transition ( Figure 7D). Together this is consistent with pre-IC complex turnover at origins being a critical function for Sld3/Sld2 dephosphorylation in vivo.

Dephosphorylation of Sld2/Sld3 during S-phase
Across eukaryotes, CDK is essential for replication initiation and in budding yeast this is mediated by phosphorylation of Sld2 and Sld3. Here we show that these substrates are also actively and specifically dephosphorylated during S-phase, suggesting that the flux of phosphorylation-dephosphorylation is an important feature of these substrates 32 . The importance of this control is underlined by the fact that three separate phosphatases, PP2A RTS1 , PP2A CDC55 and PP4 (Pph3) are responsible for the dephosphorylation of Sld3 and at least partially of Sld2. One important function for rapid dephosphorylation of Sld2/Sld3 is to allow the release of the pre-IC for efficient origin firing genome-wide ( Figure 5/6). This positive role for phosphatases in replication initiation contrasts with the conventional view that phosphatases merely oppose kinase function. Further biochemical studies will be required to determine the exact mechanics of how Sld3/Sld2 dephosphorylation is regulated, how it causes dissociation of the pre-IC and why this dissociation is important for origin firing, but it is intriguing that the Rts1 binding site on Sld3 overlaps with the critical Dpb11 binding site ( Figure 4A), suggesting that phosphatase regulation and pre-IC formation are indeed mutually exclusive.
Although Sld3 and Sld2 are retained at early origins in the cdc55-AID pph3-AID sld3-R strain, they are still released over time ( Figure 6B/C). This might be due to the penetrance of our alleles, in particular due to the still considerable dephosphorylation of Sld2 in this background ( Figure 4F), and it is also possible that even in the absence of dephosphorylation there is still some off-rate of the pre-IC from origins. This latter possibility would explain why initiation does occur at a single origin in the reconstituted replication system in vitro, in the absence of phosphatase activity 33 , but in vivo these phosphatases are clearly very important for origin firing genome-wide ( Figure 5).

The dual role of dephosphorylation of Sld3/Sld2 throughout the cell cycle
Licensing factors, such as Orc6, must be phosphorylated and inhibited by CDK before Sld3 and Sld2 become phosphorylated in order to ensure that origins only fire once in S-phase 1 . PP2A has been shown to contribute to the ordering of dephosphorylation of CDK substrates during mitotic exit 34,35 and during interphase 24,36 . Loss of PP2A/PP4 resulted in Sld3 hyperphosphorylation in G1-phase ( Figure   4D), strongly suggesting that these phosphatases are critical to dephosphorylate Sld3 during mitotic exit. We also show that when CDK activity is interrupted in interphase, for example in a clb5∆ strain, the rapid and specific dephosphorylation of Sld3/Sld2 by PP2A/PP4 prevents re-replication by ensuring that these initiation factors are inactivated, while other CDK targets such as Orc6 remain phosphorylated ( Figure 2A). The presence of this Sld2/3 specific phosphatase activity explains the continual requirement for CDK activity for origin firing throughout S-phase 14 and may explain how multiple organisms, from fission yeast to humans, respond to DNA damage in S-phase by inhibiting CDK, without causing subsequent re-replication 37, 38 . This study therefore highlights a dual function for PP2A/PP4 phosphatases in replication control: to allow pre-IC recycling and origin firing in S-phase ( Figure 5/6), and to ensure that Sld2/Sld3 are rapidly dephosphorylated to prevent inappropriate initiation at any time when CDK activity drops, for example during mitotic exit ( Figure   4D).
Although PP2A/PP4 are required for Sld3 dephosphorylation in M/G1 phase, we believe that it is the replication initiation defects in S-phase ( Figure 5/6) that cause the loss of viability in the cdc55∆ pph3∆ sld3-R strain (Figure 7). Firstly, we see no evidence for re-replication in G1 phase in this strain, as determined by flow cytometry and DNA copy number analysis (data not shown), probably because although Sld3 remains phosphorylated in G1 phase in this strain, Sld2 is not ( Figure 4D and 4F). In addition, we can suppress both the growth defect and S-phase defect by overexpression of Sld3/Dpb11 ( Figure 7D), which may help to drive the dissolution of the pre-IC at origins.

Conservation of PP2A function in replication control
It has been shown in Xenopus egg extracts that PP2A plays a positive role in replication initiation 36,39 , critically during the CDK-dependent step of Cdc45 recruitment to chromatin 40 . It remains to be tested whether PP2A in vertebrates regulates pre-IC dynamics through dephosphorylation of the essential CDK targets.
Analysing this in human cells may be complicated by a potential negative role for PP2A in origin firing through the degradation of the Sld3 orthologue Treslin 41 , which we do not observe in yeast (Supplementary Figure 5F). Given the importance of inhibition of DNA replication as a chemotherapeutic strategy, the novel essential function for PP2A/PP4 in replication initiation described here may provide a new rationale for targeting these phosphatases in cancers 42 .

Cell synchrony experiments, growth assays and yeast strains
The yeast strains used in this work are listed in Supplementary Table 1

Protein extraction, western blotting and phosphorylation analysis
Protein was extracted from pelleted yeast cells by bead beating with glass beads in 20% TCA. The resulting precipitate was pelleted and resuspended in Laemmli buffer, neutralised with Tris base and boiled. For Rad53, Orc6 and Mcm4 phosphorylation, 7.5%, 10% and 6% SDS PAGE gels were used respectively. For higher resolution of protein phosphorylation, Phos-tag (Alpha Laboratories) SDS PAGE was employed.
Orc6 (for Figure 1C AB_880249). Blots were visualised using ECL or ECL prime system (Amersham) and quantified using FIJI. For Trep profile generation, yeast genomic DNA was extracted from 5 minute interval samples using the smash and grab method (https://fangmanbrewer.genetics.washington.edu/smash-n-grab.html). DNA was sonicated (Bioruptor Pico, Diagenode), and the libraries were prepared according to the TruSeq Nano sample preparation guide from Illumina. To generate replication timing profiles, the ratio of uniquely mapped reads in the replicating samples to the non-replicating samples was calculated following 43 . Trep was determined as the time-point (in minutes after G1 release) at which each genomic bin is halfway from one copy to two copies 44 . Replication profiles were generated using ggplot and smoothed using a moving average in R. The values of Trep for Fig. 5F were taken from OriDB 45 . were collected and washed once with lysis buffer, once with buffer 1 (50mM HEPES/KOH pH7.5, 1mM EDTA, 1% Triton X-100, 0.1% Sodium deoxycholate and 250mM NaCl), once with buffer 2 (50mM HEPES/KOH pH7.5, 1mM EDTA, 1% Triton X-100, 0.1% Sodium deoxycholate and 500mM NaCl), once with buffer 3 (0.25M LiCl, 0.5% NP-40, 0.5% Sodium deoxycholate, 1mM EDTA, 10mM Tris-HCl pH8) and once with TE pH8. Samples were eluted in elution buffer (0.85X TE pH8, 1% SDS, 0.25M NaCl) for 30 minutes at 65°C. Eluted materials were transferred to new tubes and treated with RNase A (100 μg/ml; Roche) for 2.5 hours at 37°C, then with Proteinase K (800 μg/ml; Roche) overnight at 65°C. DNA was purified with CHIP clean concentrator kit (Zymo). For CHIP-qPCR: qPCR was performed on diluted samples using LightCycler 480 SYBR green 1 master kit (Roche) using primers listed below.

CHIP-seq, CHIP-qPCR
Using the 2 ΔΔct method, CHIP recovery relative to Input was normalised to the corresponding recovery from an unreplicated region and/or centrometric region to calculate fold enrichment per sample. For each strain, enrichment was then reported relative to the G1 sample. EDTA was added and beads were incubated for 30 minutes at room temperature.
After three washes with 1ml HBS, beads were eluted in 30μl Laemmli buffer by boiling for 10 minutes at 95 o C.

Data and Software Availability
Sequencing data is available at GEO: GSE186490.          Only chromosome XIV is shown as an example, known origins are annotated above.
F. Box and whisker plot of all origins from E, separated into quintiles according to their normal Trep. **** represents P value <10 -11 from a t-test. B. qPCR analysis of Sld3-13myc ChIP as in A, for the early origins ARS305, ARS110, ARS922. The G1 enrichment signal was set to 1, error bars are SD, n=4, *** is a P value of <0.0005 from a t-test C. As in A, but for Sld2-6HA