ABSTRACT
DNA polymerase delta (Pol ∂) plays several essential roles in eukaryotic DNA replication and repair. At the replication fork, Pol ∂ is responsible for the synthesis and processing of the lagging strand; this role requires Pol ∂ to extend Okazaki fragment primers synthesized by Pol α-primase, and to carry out strand-displacement synthesis coupled to nuclease cleavage during Okazaki fragment termination. Destabilizing mutations in human Pol ∂ subunits cause replication stress and syndromic immunodeficiency. Analogously, reduced levels of Pol ∂ in Saccharomyces cerevisiae lead to pervasive genome instability. Here, we analyze the how the depletion of Pol ∂ impacts replication initiation and elongation in vivo in S. cerevisiae. We determine that Pol ∂ depletion leads to a dependence on checkpoint signaling and recombination-mediated repair for cellular viability. By analyzing nascent lagging-strand products, we observe both a genome-wide change in the establishment and progression of replication forks and a global defect in Pol ∂-mediated Okazaki fragment processing. Additionally, we detect significant lagging-strand synthesis by the leading-strand polymerase (Pol ɛ) in late-replicating regions of the genome when Pol ∂ is depleted. Together, our results are consistent with the presence of least two Pol ∂ complexes in the eukaryotic replisome under normal conditions. We propose that replisomes with sub-stoichiometric Pol ∂ are defective due to competition between nascent lagging-strand primers and terminating Okazaki fragments, and that this competition underlies both the replication and genome stability defects observed when cellular levels of Pol ∂ fall below a critical threshold.
INTRODUCTION
DNA polymerase delta (Pol ∂) is an essential replisome component in all known eukaryotes (1). During lagging-strand synthesis, Pol ∂ extends Okazaki fragment primers synthesized by DNA polymerase alpha/primase (Pol α) (2). Upon reaching the 5’ end of the preceding Okazaki fragment, Pol ∂ continues synthesis, iteratively displacing short 5’ flap structures that are nucleolytically cleaved by Fen1 and other nucleases (3–5) to replace most or all of the DNA incorporated by the error-prone Pol α (6, 7). Pol ∂ plays an additional role at replication origins, synthesizing a stretch of DNA on the nascent leading strand that is subsequently extended by DNA polymerase epsilon (Pol ɛ) (8–10). Thus, Pol ∂ is directly responsible for the synthesis of approximately half the nuclear DNA in eukaryotic genomes (11, 12) and is intimately involved in every step of the replication program. Multiple suppression mechanisms exist to maintain the specificity of Pol ɛ and Pol ∂ for leading- and lagging-strand synthesis, respectively (13, 14). However, Pol ∂ can effectively synthesize the entire leading strand in both budding and fission yeast when the catalytic activity of Pol ɛ is abrogated (9, 15–17), and recent work suggests that Pol ∂ may take over leading-strand synthesis from Pol ɛ during replication termination under normal conditions (18).
Mutations affecting the stability or catalytic activity of Pol ∂ are associated with various human diseases. Alleles with reduced replication fidelity, including exonuclease-deficient alleles, are driver mutations for heritable cancers (19–21) and can induce catastrophic genome instability when expressed in yeast (22). A heterozygous active site mutation that abolishes Pol ∂ catalytic activity gives rise to developmental defects including lipodystrophy (23). In addition, mutations that reduce the stability of the Pol ∂ complex have recently been reported as causal for a syndromic immunodeficiency associated with replication stress (24). Some of the phenotypes associated with Pol ∂ deficiency overlap with those reported for Pol ɛ hypomorphy in a mouse model (25).
The depletion of Pol ∂ in budding yeast via transcriptional repression of the catalytic Pol3 subunit leads to various manifestations of genome instability, including an increase in the frequency of both point mutations and gross chromosomal rearrangements (26–28). However, the underlying basis for these defects could in theory derive from any of the roles of Pol ∂ in DNA replication or repair (29). The DNA replication machinery is highly conserved from yeast to humans. Therefore, directly characterizing the behavior of the yeast replisome under conditions where Pol ∂ is scarce can provide insights into the mechanisms by which Pol ∂ hypomorphs lead to genome instability in multicellular eukaryotes including humans.
Here, we define the short-term effects of both mild and acute Pol ∂ depletion on the DNA replication program in budding yeast via the genome-wide analysis of lagging-strand synthesis, replication-origin firing and DNA polymerase usage. We find that Pol ∂ depletion impairs the firing of early, efficient replication origins, slows replisome movement, and leads to a dependence on checkpoint activation for survival through even a single cell cycle. We further demonstrate that cells with limiting Pol ∂ are defective for Okazaki fragment processing, and provide evidence that Pol ɛ can participate in lagging-strand synthesis in late-replicating regions of the genome.
RESULTS
In S. cerevisiae, Pol ∂ is a heterotrimeric complex (30), comprising the essential Pol3 catalytic subunit and two additional accessory subunits – Pol31 and Pol32. Of the two accessory subunits, only Pol31 is essential for viability as long as the PIP boxes on the other subunits are intact (31). Besides forming part of Pol ∂, Pol31 and Pol32 are integral components of the translesion DNA polymerase ζ (32): we therefore chose to titrate Pol3 to avoid as best as possible any indirect effects due to Pol ζ depletion. Titratable expression of POL3 from the GAL1 promoter has previously been used to study the long-term effects of Pol ∂ depletion (26, 27). However, because modulating transcription using galactose is relatively slow and GAL1-POL3 strains have been reported to revert to high expression with detectable frequency (27), we chose to titrate Pol3 via targeted proteolysis. The resulting rapid degradation of Pol3 was allows us to investigate the immediate consequences of Pol ∂ depletion within a single cell cycle.
In vivo titration of Pol ∂ in vivo via controlled proteolysis of Pol3
To rapidly deplete Pol3 from cycling S. cerevisiae cells, we tagged the protein with an C-terminal auxin-inducible degron (33) and added a C-terminal 9xMyc tag for detection of Pol3 by western blot. Pol3 depletion in a strain carrying TIR1 (33) was readily detectable after 30 minutes (Fig. S1A) and treatment with indoleacetic acid (IAA) at concentrations between 0.01 mM and 1 mM allowed Pol3 levels to be titrated between wild-type and essentially undetectable levels (Fig. 1A). Long-term growth in the presence of IAA was impaired as assayed by serial dilution of cells on solid media (Fig. 1B) and cell density in liquid media (Fig. S1B). Cells were viable even at 1 mM IAA (Fig. 1B, S1B). Analysis of DNA content by flow cytometry in synchronized cultures indicated that low Pol3 levels delayed both entry into S-phase and extended its duration (Fig. 1C). Consistent with these data, asynchronous cultures treated with high concentrations of IAA accumulated in G1 and S-phase (Fig. S1C).
Conditional depletion of Pol ∂ leads to checkpoint activation, DNA damage, and dependence on recombination-mediated repair
We assayed checkpoint activation during Pol3 depletion by western blot against Rad53 following 2h of treatment with increasing concentrations of IAA (Fig. 2A). Phosphorylation of Rad53 was apparent from 0.2 mM IAA – the lowest concentration that leads to a growth defect in POL3-AID strains (Fig. 1B). Rad53 phosphorylation can occur via the DNA replication checkpoint (DRC), dependent on Mrc1, or the DNA damage checkpoint (DDC), mediated by Rad9 (34). We tested Rad53 phosphorylation in POL3-AID strains deleted for either MRC1 or RAD9, or for the apical kinase MEC1, which is required for both pathways (Fig. 2A). As expected, Rad53 phosphorylation was virtually abolished in a mec1Δ;sml1Δ strain. Deletion of RAD9 substantially reduced the extent of Rad53 phosphorylation, especially at lower concentrations of IAA, while deletion of MRC1 had a much weaker effect, (Fig. 2A). We therefore conclude that Pol ∂ scarcity leads to checkpoint activation, predominantly through the DDC but with a contribution from the DRC.
We next tested whether checkpoint activation was required to maintain the viability of POL3-AID cells during Pol3 depletion. Both POL3-AID;mec1Δ;sml1Δ and POL3-AID;rad53Δ;sml1Δ strains showed severe growth phenotypes even at 0.05 mM IAA (Fig. 2B), when Pol3 concentrations are only slightly reduced (Fig. 1A). Furthermore, consistent with the relative contributions of the DDC and DRC to Rad53 phosphorylation (Fig. 2A), the POL3-AID;rad9Δ strain had a stronger growth defect than POL3-AID;mrc1Δ at moderate concentrations of IAA (Fig. 2B). Depletion of Pol3 for 2-4h was sufficient to cause widespread cell death in the mec1Δ;sml1Δ background, as assayed by serial dilution spot tests of cells allowed to recover on YPD plates after transient treatment with 1 mM IAA (Fig. 2C). Under these conditions, G1-arrested mec1Δ;sml1Δ cells depleted for Pol3 were able to enter S-phase, but few if any cells reached G2 100 minutes after release from G1 (Fig. 2D).
The requirement for checkpoint activation, particularly via the DDC, suggested that DNA damage may accumulate during Pol3 depletion. We tested for the accumulation of DNA double-strand breaks (DSBs) via western blots against histone H2A phosphorylated at Serine 129 (phosphor-H2A) following 2h of IAA treatment (Fig. 2E). Phospho-H2A was detectable at IAA concentrations over 0.05 mM – the same concentration at which mec1Δ;sml1Δ cells begin to show an obvious growth phenotype (Fig. 2C). To test which pathway is required for the repair of DSBs caused by Pol3 depletion, we deleted the homologous recombination (HR) factors RAD51, RAD52 or MRE11, as well as the non-homologous end-joining factor YKU70. Deletion of RAD51, RAD52 or MRE11 substantially decreased viability at moderate IAA concentrations; by contrast, abrogation of end-joining via deletion of YKU70 slightly increased fitness under these conditions (Fig. 2F). This indicates that DSBs occurring due to Pol3 depletion are repaired by HR-mediated mechanisms.
Homologous recombination is largely restricted to late S and G2, while DNA damage arising as a result of Pol ∂ scarcity is likely to occur throughout both early- and late-replicating regions of the genome. The accumulation of single-stranded DNA in those regions could further contribute to checkpoint activation. To explore the possibility that single-stranded DNA is more abundant when Pol ∂ is scarce, and that gaps persist through S-phase in Pol3 depletion, we assayed the level of RPA bound to chromatin via western blot against Rfa1 following subcellular fractionation (Fig. 2G). After 2h of IAA treatment, both the total and chromatin-bound pools of RPA increased substantially. To exclude the possibility that this increase simply reflected the accumulation of IAA-treated cells in S-phase (Fig. S1C), we assayed Rfa1 levels after synchronous release into S-phase (Fig. 2H). Chromatin-bound RPA was highly enriched in both mid and late S-phase in IAA-treated cells compared to untreated controls, implying the persistence of single-stranded DNA throughout S-phase.
Together, these data suggest that depletion of Pol3, and consequently Pol ∂, rapidly leads to the accumulation of single-stranded DNA and double-strand breaks, leading to checkpoint activation and delayed progression through S-phase. HR-mediated DNA repair pathways are essential for cells with limiting Pol ∂ to complete S-phase and continue through the cell cycle.
Effects of limiting Pol ∂ on productive replication origin firing in vivo
Pol ∂ is responsible for the majority of DNA replication on the lagging strand (11, 12), but also plays a key role in the initiation of leading-strand synthesis at replication origins (8–10). To investigate the effect of Pol3 depletion on origin firing, we sequenced Okazaki fragments from asynchronous strains depleted for Pol3. We compared the effects of progressively lower Pol ∂ levels for two biological replicate strains grown in YPD alone or supplemented with 0.2 mM, 0.6 mM, or 1 mM IAA. Data showing the average behavior of two replicates are shown in Fig. 3, and individual replicates are shown separately in Fig. S2. Cultures were treated with IAA for 2h before the addition of rapamycin for 1h to deplete DNA ligase I (Cdc9) from the nucleus via anchor-away (35). As previously described (36), we quantified origin firing efficiency as the Origin Efficiency Metric (OEM), based on the fraction of Okazaki fragments mapping to either the Watson or Crick strand in a window ±10 kb around previously validated replication origins: OEM represents the fraction of cells in the population in which a given origin fires.
Analysis of Okazaki fragment strand bias across the genome suggested that the average direction of replication was relatively similar at most loci across all Pol3 depletion conditions (Fig. 3A). Surprisingly, quantitative analysis of origin firing (Fig. 3B) revealed a statistically significant increase in the average firing efficiency – expressed as OEM – upon Pol3 depletion (p = 0.0205 for 0 mM vs 1 mM by Wilcoxon signed rank test), accompanied by a change in the distribution of these efficiencies (p < 0.0001 for 0 mM vs 1 mM by Kolmogorov-Smirnov test). Direct comparison of firing efficiencies in YPD compared to YPD + 1 mM IAA indicated that Pol3 depletion slightly decreased the firing efficiency of the most efficient origins, while increasing the efficiency of normally weak origins (Fig. 3C). Because origin firing efficiency in S. cerevisiae correlates strongly with replication timing, we reasoned that the increased firing of inefficient origins could be related to their late replication. We assigned origins as early or late based on their median replication time (Trep) in a previous study (37), and separately analyzed the change in efficiency for early- and late-firing origins (Fig. 3D-E). Consistent with a strong effect of firing time, we observed a significant increase in the firing efficiency of late origins during Pol3 depletion (Fig. 3E. p < 0.0001 for 0 mM vs 1 mM by Wilcoxon signed rank test), accompanied by a decrease in early origin firing (Fig. 3D. p < 0.0001 for 0 mM vs 1 mM).
Replication-fork speed is reduced when Pol ∂ is limiting
The progressive increase in firing efficiency of late-replicating origins at lower Pol3 concentrations (Fig. 3E) could be at least partly explained by decreased replication-fork mobility, which would reduce passive replication of these origins and instead allow them to fire. To test this, we sequenced Okazaki fragments from strains synchronously released into S-phase at 25 °C in YPD ±1 mM IAA (flow cytometry data showing the timepoints selected for analysis are in Fig. S3). Okazaki fragment sequencing data from this time course are shown in Fig. 4A. The density of sequencing reads in cultures sampled from 45-60 minutes after release showed decreased replication of origin-distal regions upon Pol3 depletion (Fig. 4A), consistent with delayed origin firing and/or reduced fork speed.
To confirm that Pol3 depletion globally decreases replisome mobility, we carried out meta-analysis of both total (Fig. 4B) and strand-specific (Fig. 4C) Okazaki fragment density 20 kb up- and downstream of early-firing origins. Total Okazaki fragment density monitors the extent to which the lagging strand has been replicated, regardless of the direction of fork movement. By contrast, the stranded analysis considers only Okazaki fragments from leftward-moving forks to the left of the origin, and from rightward-moving forks to the right – these data therefore report more directly on the movement of forks emanating from the origins being analyzed. In both cases, Okazaki fragments from Pol3-depleted cells generated a sharper peak close to early-firing origins after 45 minutes, which propagated away from origins more slowly than for cells grown in YPD (Fig. 4B, 4C). We therefore conclude that the progression of replication forks through the genome is impaired when Pol ∂ is limiting. However, we note that our data do not distinguish between the global slowing of all forks and normal fork speed punctuated by widespread fork stalling.
Taken together, the global changes in replication-origin firing efficiency and replication-fork progression are most consistent with a model whereby severely limiting Pol ∂ globally impairs origin firing in vivo, albeit surprisingly mildly. Because fork progression is impeded when Pol ∂ is scarce, normally late-firing origins are less likely to be passively replicated and the net effect is an increase in their firing efficiency. An analogous global decrease in replication-origin firing efficiency has been observed in Pol ∂-deficient human fibroblasts (24).
Pol ∂ depletion alters lagging-strand biogenesis without affecting Okazaki fragment size
Given Pol ∂’s role as the major lagging-strand polymerase, we anticipated that one or more steps required for Okazaki fragment biogenesis – i.e. initiation, synthesis, and processing – could be impacted by depletion of Pol3. We first assayed the length of end-labeled Okazaki fragments obtained after Cdc9 anchor away during Pol3 depletion in asynchronous cultures. Surprisingly, we observed a that the overall length distribution of Okazaki fragments was unchanged, but the normal nucleosome-based phasing of fragment size (7) was lost at IAA concentrations greater than 0.1 mM (a representative gel is shown in Fig. 5A, and a replicate in Fig. S4A). We additionally verified this result in a strain carrying an AID-tagged version of the Pol31 protein. When Pol31 was depleted, nucleosome-based phasing of Okazaki fragment size was again absent at IAA concentrations above 0.1 mM (Fig. S4B).
We have previously observed a loss of nucleosome-based Okazaki fragment phasing in strains deficient for Okazaki fragment processing, including a Rad27 conditional depletion strain. When lagging-strand processing by nick translation is impaired by the absence of Rad27, Okazaki fragment termini are enriched on the replication-fork proximal sides of nucleosome dyads (38). We confirmed that this phenotype occurs in our POL3-AID strain. Increasing concentrations of IAA shifted the peak of Okazaki fragment 5’ and 3’ termini progressively upstream of the nucleosome dyad (Fig. 5B & replicate data in Fig. S4C), as previously observed for Okazaki fragment processing mutants (38). To further confirm the decreased nick translation during Pol3 depletion outside nucleosomes, we aligned Okazaki fragment termini around Abf1, Reb1 and Rap1 binding sites; these general regulatory factors present ‘hard’ barriers to strand-displacement synthesis by Pol ∂, leading to a sharp peak of Okazaki fragment termini (7) that is reduced when nick translation is impaired (38). Consistent with the other phenotypes described above, Okazaki fragment termini were progressively less enriched at Abf1/Reb1/Rap1 sites as Pol3 was increasingly depleted (Fig. 5C and replicate data in Fig. S4D). Taken together, our data are consistent with Pol3 depletion leading to a global defect in lagging-strand processing via a reduction in nick translation by Pol ∂.
Pol ɛ contributes to lagging-strand synthesis in late-replicating regions when Pol ∂ is limiting
The defects observed in nick translation and Okazaki fragment processing during Pol3 depletion suggested at least two possibilities that are not mutually exclusive. In the context of limiting Pol ∂, the polymerase might prematurely disengage from the terminating Okazaki fragment as opposed to remaining associated for iterative rounds of strand-displacement synthesis. Alternatively, one or more alternative polymerases may be contributing to bulk lagging-strand synthesis under these conditions. Extensive use of an alternative polymerase is expected to impair nick translation because Pol ∂ has a much higher strand-displacement synthesis activity than other DNA polymerases in yeast (39). Synthesis by Pol α is inhibited by RFC-dependent loading of PCNA (15, 40), but Pol ɛ has been reported to compete weakly with Pol ∂ for lagging-strand primers in vitro (13). In addition, any of the three translesion polymerases present in S. cerevisiae could contribute: indeed, Pol η has recently been reported to participate in lagging-strand synthesis to a measurable extent (41).
To test the possibility that translesion (TLS) DNA polymerases might contribute substantially to DNA replication during Pol3 depletion, we combined the POL3-AID allele with individual knockouts of REV1, REV3 and RAD30 (Pol η), as well as the rev1;rev3;rad30 triple mutant (ΔTLS). Individual or simultaneous deletion of the three TLS polymerases failed to exacerbate the growth phenotype observed during Pol3 depletion at any concentration of IAA (Fig. 6A). We conclude that growth during transient Pol ∂ depletion is not rescued by the widespread action of TLS polymerases.
In order to investigate the overall contribution of Pol ɛ to DNA replication in the context of Pol ∂ depletion, we combined POL3-AID with the pol2-M644G allele of the catalytic subunit of Pol ɛ: this allele causes increased incorporation of ribonucleotides into the nascent leading strand (42). In strains unable to remove single ribonucleotides from genomic DNA due to deletion of the RNase H2 subunit RNH202, ribonucleotide positions can be tracked by HydEn-seq. Ribonucleotides are cleaved by alkaline hydrolysis, followed by strand-specific high-throughput sequencing to identify their locations (11). Ribonucleotides incorporated by Pol ɛ are normally highly specific for the leading strand, so an increased contribution of Pol ɛ to lagging-strand synthesis during Pol ∂ depletion should result in a loss of strand bias for Pol ɛ-derived ribonucleotides (Fig. 6B).
We prepared HydEn-seq libraries from two biological replicate POL3-AID;pol2-M644G;rnh202Δ strains grown ± 1 mM IAA for 2h before genomic DNA preparation. Average results for the two strains at 0 mM and 1 mM IAA are shown in Fig. 6, and results for individual replicates in Fig. S5. We observed a subtle reduction in the strand bias of ribonucleotide incorporation by Pol ɛ within the region 10 kb up- and downstream of replication origins when Pol3 was depleted at 1 mM IAA (Fig. 6C).
Pol3 depletion has opposing effects on the firing efficiencies of early- and late-firing replication origins (Fig. 3E-F), which means that the same strand cannot necessarily be assigned as lagging at different auxin concentrations. Moreover, the stable association of scarce Pol ∂ complexes with elongating replisomes emanating from early-firing origins might cause replisomes assembling later in S-phase to be differentially sensitive to Pol3 depletion. We therefore separately analyzed ribonucleotide incorporation by Pol ɛ around early and late origins, taking into account the change in lagging-strand bias inferred from the Okazaki fragment sequencing experiments shown in Fig. 3.
Around early-firing origins – defined as origins with a Trep below the median for the 281 origins analyzed (37) – the increase in ribonucleotide density on the predicted lagging strand corresponded precisely to the change in lagging-strand identity arising from the reduced firing of these origins (Fig. 6D). Therefore, we conclude that during Pol ∂ depletion the ribonucleotide distribution in early replicating regions changed purely as a result of altered origin firing (as opposed to changes in the enzymology of the resulting replisomes). In contrast, we observed strikingly different results around late-firing origins (Fig. 6E). If strict segregation of Pol ɛ to the leading strand were maintained, the increased firing efficiency of these origins would result in increased ribonucleotide incorporation on the Crick strand to the left of origins, with a corresponding decrease to the right. By contrast, the change in ribonucleotide distribution had the opposite polarity to the change in average replication-fork direction (Fig. 6E). Therefore, although the polarization of lagging-strand identity around late origins is increased by Pol3 depletion, there is a substantial reduction in the strand bias of ribonucleotide incorporation by Pol ɛ that more than offsets the change in average replication direction. We therefore conclude that the discrimination of Pol ɛ for the leading strand is reduced in late-replicating regions of the genome when Pol ∂ is scarce, resulting in a detectable contribution of Pol ɛ to lagging-strand synthesis.
By subtracting the change in replication-fork polarity from the change in ribonucleotide incorporation, we estimated the contribution of Pol ɛ to lagging-strand synthesis in early- and late-replicating regions (Fig. 6F). We observe a ∼10% reduction in strand discrimination around late origins. Because complete synthesis of the lagging strand by Pol ɛ would lead to a 50% reduction in bias over many cell cycles (Fig. 3B) but only at most half of the DNA in the culture is nascent after 2h of IAA treatment, the maximum theoretical increase in Pol ɛ contribution under these conditions is 25%. We therefore propose that Pol ɛ is either synthesizing all of the lagging strand at a subset of replication forks, or a fraction of the lagging strand at most forks established in late S-phase. Our data cannot formally rule out the alternative scenario, not mutually exclusive with Pol ɛ replication on the lagging strand, whereby RNA primers are retained in the genome during Pol ∂ depletion. However, because nick translation appears to be similarly impaired in early- and late-replicating regions (Fig. S6) while ribonucleotide strand bias changes exclusively in late-replicating regions (Fig. 6E-F), we favor the interpretation that invokes Pol ɛ synthesis on the lagging strand.
DISCUSSION
By directly analyzing the products of lagging-strand synthesis following proteolytic titration of Pol3 in S. cerevisiae, we identified several immediate consequences of Pol ∂ depletion for DNA replication. Extremely low levels of Pol ∂ support viability if the checkpoint and recombination-mediated repair pathways are intact (Fig. 2). However, replication origin firing (Fig. 3), the progression of the resulting replication forks through the genome (Fig. 4), and lagging-strand processing (Fig. 5) are all progressively impaired as the concentration of Pol ∂ is reduced. In addition, we show that the enzymology of DNA replication is altered in late-replicating regions of the genome when Pol ∂ is scarce, such that Pol ɛ now contributes detectably to lagging-strand synthesis (Fig. 6).
Pol ∂ is required for normal replication origin firing, and synthesizes both the leading and lagging strands in the immediate vicinity of origins in vivo (9), consistent with in vitro data demonstrating that leading-strand synthesis can proceed via extension of the first Okazaki fragment on the lagging strand (8). The reduction in origin firing efficiency that we observe for early origins (Fig. 3) is surprisingly modest given this essential role, although the delayed entry into S-phase observed under these conditions is likely caused by slow recruitment of scarce Pol ∂. Although Pol ∂ deficiency reduces the disparity in firing efficiencies between replication origins compared to unperturbed conditions (Fig. 3), the overall replication timing program appears to be largely conserved (Fig. 4). Even if Pol ∂ recruitment becomes limiting for origin firing, this binding is still more favorable for early-firing origins. Estimates of Pol3 and Pol31 protein levels in S. cerevisiae vary substantially (43). The observation that our strain completes S-phase with no discernible delay at 0.2 mM IAA (Fig. 1C) – a concentration that reduces Pol3 levels at least tenfold (Fig. 1A) – implies that the normal cellular levels of Pol ∂ are greatly in excess of the minimal amounts needed for timely replication of the genome during rapid growth. However, checkpoint activation is required for survival after even a mild reduction in Pol3 levels (Fig. 2), suggesting that at least a proportion of forks encounter difficulties under these conditions.
Elongating replication forks in cells severely depleted for Pol ∂ have at least two obvious defects: they progress slowly (Fig. 4) and are impaired for normal Okazaki fragment maturation (Fig. 5). Extension of Okazaki fragments by Pol ɛ would decrease nick translation because Pol ɛ is deficient for strand-displacement synthesis (15, 39). But the changes in nick translation that we observe are global (Fig. 5A, S4A-B, S6), and are therefore unlikely to arise exclusively from lagging-strand synthesis by Pol ɛ since this only occurs at detectable levels in late-replicating regions (Fig. 6). Therefore, impaired fork progression and lagging-strand processing during Pol3 depletion must be reconciled with recently reported data in vitro (44) and in vivo (45) showing that Pol ∂ remains stably associated with replication forks for long periods of time while undergoing concentration-dependent exchange (44). How could depletion of such a stably associated Pol ∂ complex give rise to the phenotypes we observe?
Pol ∂ has two somewhat distinct roles during lagging-strand synthesis. It extends PCNA-bound primer-template junctions resulting from Pol α/primase synthesis (2), and also engages at nicks generated by iterative rounds of Pol ∂ strand-displacement synthesis followed by Fen1 cleavage during Okazaki fragment processing (5). Assuming that Okazaki fragment priming often occurs before processing of the prior fragment is completed, two competing Pol ∂ substrates repeatedly coexist within the same replisome. Thus, a single stably associated Pol ∂ complex must effectively ‘choose’ between the nick translation and elongation reactions. Delayed extension of Pol α products would lead to an accumulation of single-stranded DNA and primer template junctions, likely leading to both checkpoint activation and to slower fork movement. Conversely, premature dissociation from the Okazaki fragment terminus to extend the subsequent fragment would lead to an overall reduction in nick translation. We observe all three of the above phenotypes (fork slowing, checkpoint activation, and impaired nick translation). We therefore propose that one stably associated Pol ∂ complex is minimally required per replisome, but that two Pol ∂ are necessary for optimal synthesis by the S. cerevisiae replisome in vivo (Fig. 7). According to this model, the increasingly severe defects we observe as Pol3 levels are reduced reflect an increasing number of replisomes with only one Pol ∂. Single-molecule studies have shown a single Pol ∂ present in a reconstituted replisome assembled on a forked DNA substrate (44) – an observation that initially appears to be in conflict with our model. It is possible that additional factors that stimulate Pol ∂ recruitment are present in vivo, or indeed that physically coupled sister replisomes (46) may share multiple Pol ∂ complexes as recently proposed for Pol α (47). Our model speculates that replisomes with one Pol ∂ can complete replication, albeit with defects in fork progression and Okazaki fragment maturation. Furthermore, lagging-strand processing has not been directly investigated in the reconstituted replisomes shown to contain a single Pol ∂ (44). We therefore do not believe that our data contradict previous reports.
Assuming stable binding of Pol ∂ to the replisome, the concentration of free Pol ∂ will be further reduced in late S-phase by sequestration of the few available complexes in slowly progressing replisomes. Under these conditions, the suppression mechanisms that restrain Pol ɛ from lagging-strand synthesis (13) are apparently overcome; Pol ɛ therefore extends lagging-strand primers in a fraction of replisomes assembled at late-firing origins (Fig. 6). In theory, these replisomes could represent either complexes from which Pol ∂ has dissociated, or complexes that never recruited Pol ∂. Our data cannot distinguish between these possibilities.
Low levels of Pol ∂ in S. cerevisiae increase both point mutation frequency (26) and large-scale genomic rearrangements (27). In human fibroblasts, hypomorphic Pol ∂ results in the accumulation of Rad51 and 53BP1 foci in addition to altering replication dynamics (24). Our data suggest distinct underlying mechanisms for these phenotypes (Fig. 7). Depleting Pol ∂ reduces nick translation (Fig. 5). Elevated mutagenesis has previously been observed in S. cerevisiae mutants that impair nick translation during lagging-strand synthesis (48, 49), commensurate with the extent to which nick translation is perturbed (38) and presumably due to increased retention of Pol α-derived DNA in the daughter genome. Our model also invokes an increased lifetime for single-stranded DNA on the lagging-strand template, which would increase both the frequency of one-ended double-strand breaks at replication forks and the sensitivity of Pol ∂-deficient cells to DNA damaging agents that preferentially target single-stranded DNA. Finally, the persistent chromatin-bound RPA observed through S-phase (Fig 2H) is consistent with the presence of potentially recombinogenic single-stranded gaps. In human cells, such gaps would be substrates for post-replicative repair mediated by 53BP1 and Rad52(50). Thus, the distinct effects of Pol ∂ scarcity on lagging-strand primer extension and Okazaki fragment processing can each contribute to genome instability via related but independent mechanisms.
MATERIALS AND METHODS
Yeast strains
All strains are W303 RAD5+ background and contained additional mutations required for anchor-away depletion of Cdc9. The genotype of the wild-type strain is tor1-1::HIS3, fpr1::NatMX4, RPL13A-2xFKBP12::TRP1, CDC9-FRB::HygMX. The Pol3 depletion strain contains OsTIR1 integrated at the URA3 locus and POL3 tagged with IAA17(71-116)-9xmyc (Addgene #99522) as described in (51). Individual knockout strains were created by gene replacement. Ribonucleotide-hyperincorporating strains were a generous gift from Hannah Klein. All mutations were introduced into the POL3-AID background by cross, except for rad52. Biological replicate strains represent independent colonies derived from one or more crosses.
Cell growth, spot tests, and cell-cycle synchronization
Yeast were grown in YPD at 30C unless indicated otherwise. 3-indoleacetic acid (IAA) (Sigma I2886-5G) was dissolved in 95% ethanol to 200mM. For short term experiments, IAA was added to log phase cells for two hours, followed by addition of 1ug/mL rapamycin (Spectrum 41810000-2) for one hour for Okazaki fragment analysis. Spot tests were performed with exponentially growing cultures (OD 0.65), washed in sterile water, and diluted five-fold in sterile water in 96-well plates and spotted on plates at 30C for two days. For synchronized S-phase analysis, log phase cells (OD 0.2) were briefly sonicated to disrupt clumps and 10ug/mL alpha factor was added, followed by 5ug/mL every hour until >95% cells were fully shmooed. After complete arrest, IAA was added and 5ug/mL alpha factor was added every hour to maintain arrest. To release cells from arrest, cells were washed twice with deionized water and resuspended in YPD with or without IAA as required.
Flow cytometry
Cells were collected by adding 150uL of yeast culture to 350uL absolute ethanol and stored at 4C. Samples were treated with RNase by pelleting cells and resuspending in 500uL of 50mM sodium citrate with 42ug/mL RNase A and incubating at 50C for two hours, followed by addition of 100ug proteinase K for two additional hours. Equal volume of 50mM sodium citrate with 0.2uL SYTOX green (Fisher S7020) was added, samples were sonicated, and analyzed on a Becton Dickinson Accuri.
Subcellular fractionation and western blotting
Cells for subcellular fractionation were treated with 0.1% sodium azide and collected by centrifugation and stored at −20C. Cells were thawed on ice, washed and resuspended in spheroplasting buffer, and treated with 10uL of 20mg/mL zymolyase T20 and 1uL 1M DTT for 40 minutes at 30C before harvesting spheroplasts by centrifugation. Spheroplasts were resuspended in 300uL extraction buffer and divided into whole cell (50uL), soluble (50uL), and chromatin (200uL) fractions. The whole cell fraction was treated with 1.25uL 10% Triton-100, vortexed, incubated on ice for 5 minutes, treated with 1uL universal nuclease, incubated for 15 minutes on ice, and mixed with 20uL urea loading buffer. The soluble fraction was treated with 1.25uL 10% Triton-100, vortexed, incubated on ice for 5 minutes, centrifuged, and 20uL urea loading buffer was added to the supernatant. The chromatin fraction was treated with 1.25uL 10% Triton-100, vortexed and incubated on ice for 5 minutes before addition of 30% sucrose. After centrifugation, the previous step was repeated, and the resulting pellet was resuspended in 50µL EB with 1.25µL 10% Triton-100, treated with 1uL universal nuclease, incubated for 15 minutes on ice, and mixed with 20uL urea loading buffer.
Cells for western blotting were collected (2.0 OD units), washed, and resuspended in 200uL 20% TCA with glass beads, vortexed for 10 minutes, and the lysate was kept. The beads were washed with 600uL 5% TCA, combining with the lysate from the previous step. The lysate was centrifuged and the pellet was resuspended in 100uL urea loading buffer with 1uL 10M NaOH to restore color.
All western blotting samples were boiled for five minutes at 95C before loading onto an SDS-PAGE gel. Gels were blotted onto nitrocellulose membranes by wet transfer, blocked in 5% milk in PBS-T for one hour, and incubated with primary antibody at 4C overnight. After PBS-T washes, secondary antibody incubation for one hour, and PBS-T washes, blots were exposed with SuperSignal West Pico (Fisher 34080). The following antibodies were used in this study: anti-myc (1:1000 Genscript A00173-100), anti-Rad53 (1:1000 Abcam ab104232), anti-S129phosp H2A (1:1000 Abcam ab15083), anti-Rfa1 (1:50000 Abcam ab221198), anti-H4 (1:1000 Abcam ab10158), anti-PGK1 (1:1000 Fisher 459250), anti-rabbit IgG HRP (1:5000 GE Healthcare NA934V), anti-mouse IgG HRP (1:5000 Abcam ab97046).
Okazaki fragment preparation, labeling and sequencing
Okazaki fragments were purified, end-labeled and deep-sequenced as previously described (7), with minor modifications. Briefly, Okazaki fragments were purified from genomic DNA collected after CDC9 repression, using sequential elutions from Source 15Q columns. DNA was treated with RNase before adaptor ligation, second strand synthesis, and barcoding. Paired-end sequencing (2 × 75 bp) was carried out on an Illumina Next-seq 500 platform.
HydEN-seq
HydEN-seq was carried out as described (53). Briefly, DNA was cut with SbfI-HF, treated with 300mM KOH, and 5’ phosphorylated before adapter ligation, second strand synthesis and barcoding. Paired-end sequencing (2 × 75 bp) was carried out on an Illumina Next-seq 500 platform
Computational analyses
FASTQ files were aligned to the S288C reference genome using the Bowtie (v2.2.9). Low quality reads and PCR duplicates were removed and resulting data was converted to BEDPE files using the Samtools suite (v1.3.1). For Okazaki fragment sequencing, genome coverage was calculated using the Bedtools suite (v2.26.0) in a strand-specific manner. Origin efficiency metric analysis of predefined origins was carried out as previously described (36) with the origin list from the same source (included as table S1). Briefly, the fraction of Okazaki fragments mapping to the Watson or Crick strand in 10 kb windows to the left and right of the origin (WL, WR, CL, and CR) is calculated for each origin. OEM is calculated as WL/(WL+CL) – WR/(WR+CR). Classification of origins as early or late was based on Trep from(37). Okazaki fragment ends were aligned to consensus nucleosome dyads (54) or Abf1/Reb1/Rap1 sites(55) and normalized to the median end density in a 2kb window around the sites.
DATA AVAILABILITY
Sequencing data have been submitted to the GEO under accession number 141884.
ACKNOWLEDGEMENTS
We thank Tom Petes and Dirk Remus for helpful discussions, and Hannah Klein for sharing strains. We thank members of the Smith lab for discussions and critical reading of the manuscript, and NYU Gencore for assistance with Illumina sequencing. This work was supported by NIH R01 GM114340 to D.J.S.
Footnotes
Data deposition: Raw and processed sequencing data available on the GEO, accession GSE141884