Increased expression of Polδ does not alter the canonical replication program in vivo

Background: In vitro experiments utilising the reconstituted Saccharomyces cerevisiae eukaryotic replisome indicated that the efficiency of the leading strand replication is impaired by a moderate increase in Polδ concentration. It was hypothesised that the slower rate of the leading strand synthesis characteristic for reactions containing two-fold and four-fold increased concentration of Polδ represented a consequence of a relatively rare event, during which Polδ stochastically outcompeted Polε and, in an inefficient manner, temporarily facilitated extension of the leading strand. Inspired by this observation, we aimed to determine whether similarly increased Polδ levels influence replication dynamics in vivo using the fission yeast Schizosaccharomyces pombe as a model system. Methods: To generate S. pombe strains over-expressing Polδ, we utilised Cre-Lox mediated cassette exchange and integrated one or three extra genomic copies of all four Polδ genes. To estimate expression of respective Polδ genes in Polδ-overexpressing mutants, we measured relative transcript levels of cdc1 +, cdc6 + (or cdc6 L591G), cdc27 + and cdm1 + by reverse transcription followed by quantitative PCR (RT-qPCR). To assess the impact of Polδ over-expression on cell physiology and replication dynamics, we used standard cell biology techniques and polymerase usage sequencing. Results: We provide an evidence that two-fold and four-fold over-production of Polδ does not significantly alter growth rate, cellular morphology and S-phase duration. Polymerase usage sequencing analysis further indicates that increased Polδ expression does not change activities of Polδ, Polε and Polα at replication initiation sites and across replication termination zones. Additionally, we show that mutants over-expressing Polδ preserve WT-like distribution of replication origin efficiencies. Conclusions: Our experiments do not disprove the existence of opportunistic polymerase switches; however, the data indicate that, if stochastic replacement of Polε for Polδ does occur i n vivo, it represents a rare phenomenon that does not significantly influence canonical replication program.


Introduction
Unchallenged duplication of the eukaryotic genome requires the coordinated action of three replicative polymerase complexes: Polα-primase (hereafter referred to as Polα), Polδ and Polε (Burgers & Kunkel, 2017). According to the canonical model of eukaryotic replication, Polα and Polδ cooperate to discontinuously synthesise the lagging strand via the iterative production of short Okazaki fragments (OF), ca. 150bp, whereas Polε caries out continuous leading strand replication (Clausen et al., 2015;Daigaku et al., 2015;Miyabe et al., 2011). Interestingly, such strict division of labour does not always apply, and deviations have been documented (Guilliam & Yeeles, 2020a).
While polymerase activities of Polα and Polδ are indispensable for cell survival, the polymerase domain of Polε is not required for completion of replication in either Saccharomyces cerevisiae or S. pombe (Feng & D'Urso, 2001;Kesti et al., 1999). In both yeast experimental models it has been demonstrated that Polδ facilitates the leading strand synthesis when catalytically-inactive Polε is expressed (Garbacz et al., 2018;Miyabe et al., 2015). Such findings have found support in in vitro experiments utilising reconstituted replisome system (Yeeles et al., 2017), confirming that, under certain circumstances, Polδ is competent in the leading strand replication.
Indeed, it has been reported that Polδ replicates both DNA strands during homologous recombination restarted replication in S. pombe (Miyabe et al., 2015) and break induced replication in S. cerevisiae (Donnianni et al., 2019). Additionally, genomic analysis by polymerase usage sequencing (Pu-Seq) or HydEn-seq revealed that Polδ is involved in the initiation of the leading strand replication in unperturbed S. cerevisiae and S. pombe cells, respectively (Daigaku et al., 2015;Garbacz et al., 2018;Zhou et al., 2019). In agreement with such findings, PCNA-associated Polδ has been shown to play an important role in early stages of the leading strand replication in vitro (Aria & Yeeles, 2018;Yeeles et al., 2017). Moreover, it has recently been proposed that Polδ takes over the leading strand synthesis prior to replication fork termination (Zhou et al., 2019). The exact role of Polδ during the final stages of replisome progression is, however, yet to be clarified.
Apart from homologous recombination dependent DNA synthesis and replication initiation, Polδ-mediated leading strand synthesis has been shown to occur in the context of polymerase uncoupling. It has been reported that cyclobutane pyrimidine dimer driven disengagement of CMG-associated Polε from the leading 3'OH generates a gap, the efficient filling of which requires the translesion synthesis machinery, as well as the action of Polδ (Guilliam & Yeeles, 2020b). Additionally, it has been demonstrated that Polδ takes over the leading strand synthesis and performs an error-free bypass of oxidative DNA adducts thymine glycol and 8-oxoguanine (Guilliam & Yeeles, 2021). In further support of a more generic function of Polδ in the leading strand synthesis, Polδ has been shown to proofread errors introduced by Polε in hyper mutator pol2-M644G mutants (Bulock et al., 2020). In line with all aforementioned observations, it has been shown that CMG-associated Polε exists in two mutually-exclusive conformations, of which only one facilitates DNA synthesis (Zhou et al., 2017).
Intriguingly, according to in vitro studies of eukaryotic replication, two-fold and four-fold increase in Polδ concentration reduces the rate of the leading strand synthesis (Yeeles et al., 2017). It has been suggested that the observed retardation of leading strand replication represents a consequence of stochastic polymerase switching, during which Polδ outcompetes Polε and temporarily facilitates inefficient extension of the leading 3' end. Since the effect of Polδ concentration on replisome progression and the hypothetical phenomenon of leading strand polymerase switching has not been investigated in vivo, we aimed to test whether similar a phenomenon manifests in living cells, potentially shedding light on a yet uncomprehended promiscuity of replicative polymerases.

Methods
Yeast culture and transformation S. pombe cells were grown in yeast extract (YE) (Formedium, PCM0155) with supplements (Formedium, PSU0110) medium according to standard procedures (Petersen & Russell, 2016). Briefly, cells (25% glycerol stocks stored at -80°C) were streaked onto an agar plate and incubated at 30°C for 2-3 days. Next, cells were inoculated into a liquid medium and cultivated at 30°C for ca. 36 h in the ISF-1-W shaker (Kuhner) with constant shaking (180 rpm). Cultures were diluted accordingly two times during the course of cultivation. Then appropriate amounts of cells were collected (depending on experiment) and processed further. Cells were transformed by the lithium-acetate based method (Bähler et al., 1998). Optical density (OD) of liquid cell cultures was assessed by WPA CO8000 Cell Density Meter (Biochrom). Doubling times were calculated using the formula: DT = 1/k, where DT stands for doubling time and k represents the slope of linear regression computed from a time-series of log 2 -transformed OD measurements. A list of strains used in this study is provided in Table 1. Microscopy 1 mL of exponentially growing cells was centrifuged (1000 × g, 5 min, 25°C) and the cell pellet resuspended in 1 mL of 70% ethanol. 500 µL of fixed cells were collected by centrifugation (1000 × g, 5 min, 25°C) and re-suspended in 50 µL of H 2 O containing 1µM 4′,6-diamidino-2-phenylindole (DAPI). Cells were incubated at room temperature in the dark for at least 15 min, and then analysed by microscopy using a Nikon E400 system. Cell lengths were determined from DIC images by measuring the distance between the opposite poles of the cell using ImageJ software (version 1.51m9) (Schneider et al., 2012). At least 200 cells per sample were scored.

RT-qPCR
Total RNA was isolated from 1-2 mL of exponentially growing cells (OD 600 = 0.5; 5×10 6 cells/mL) using MasterPure Yeast RNA purification kit (Cambio Ltd, MPY03100 , where RNA target represents the relative transcript level of a given target gene, and Cq (target) and Cq (reference) stand for PCR cycle quantification values of target and reference genes, respectively. act1 was used as the reference gene. A list of qPCR primers (obtained from Integrated DNA Technologies) used in this study is provided in Table 3.

Pu-Seq library preparation
For all strains presented, two sets of Pu-Seq libraries were prepared. One set was prepared as described previously (Daigaku et al., 2015;Keszthelyi et al., 2015). Briefly, 10-20 µg of genomic DNA containing increased quantities of misincorporated ribonucleotides (rNMPs) was treated with 0.3M NaOH for 2 h at 55°C. Digested DNA was run on a 2% agarose gel (in 0.5× TBE). 300-500bp ssDNA fragments were gel extracted and subjected to complementary second strand synthesis primed by random 8-mers (obtained from Integrated DNA technologies). Resulting double-stranded (dsDNA) fragments were converted to Illumina sequencing libraries using NEBNext Ultra DNA library prep kit for Illumina (NEB, E7645S) and NEBNext multiplex oligos for Illumina (NEB, E7335). The second set of Pu-Seq libraries was prepared according to a modified version of the established GLOE-Seq protocol (Sriramachandran et al., 2020), which utilises two subsequent ligations of adapter/splinter oligonucleotides, first to rNMPdependent phosphorylated 5' ends and, following sonication, to 3' ends of the ssDNA fragments. rNMP-dependent 5' ends were generated from genomic DNA containing increased quantities of misincorporated ribonucleotides treated by RNAse H2 (NEB, M0288S) and subsequently denatured at 95°C. Sequencing was performed on an Illumina NextSeq 500 sequencer. Sequencing reads were mapped onto the reference genome using Bowtie2 (Langmead & Salzberg, 2012).

Pu-Seq data analysis
Polymerase tracks at any given 300bp bin were calculated using the equation PT = (RT -RB) / (RT + RB), where PT represents polymerase track, and R T and R B stand for rNMPs mapped to the top and the bottom DNA strands, respectively. Polymerase tracks were determined for each biological repeat separately, then averages of the two repeats were used for subsequent analysis.
Positions and efficiencies of origins of replication were determined from differential values of polymerase tracks, similarly to (Daigaku et al. (2015). Briefly, for all three datasets (Polδ, Polε, Polα), the difference of each neighbouring datapoint of polymerase track values (smoothed by simple moving average of 3) was calculated as Diff i = PT i -PT i-1 , where Diff i represents differential value at position i, and PT i -PT i-1 stand for smoothed polymerase track values at positions i and i-1, respectively. Differential value of the first bin on a given chromosome was assigned 0. Polε differentials and the opposites of Polδ and Polα differentials were averaged and smoothed by simple moving average of 3. Then, positive peaks (indicating sharp inclinations in the data) were selected. Differential peaks containing two or more distinct maxima separated by at least four bins were treated as independent peaks. Peaks with maxima bellow 30 th percentile were disregarded. Each independent differential peak represented an origin of replication, the efficiency of which was estimated as 50% of the sum of its values. 259 replication initiation regions and 147 termination zones were selected using wild-type (WT) origin efficiency data. For comparison purposes, origin efficiencies were normalised assuming that the efficiency of the most efficient origin was 100%. Data were analysed in R (https:// www.R-project.org; R Core Team, 2020) using a custom script (see Software availability).

Results
Brief overview of polymerase usage sequencing Pu-Seq methodology determines the genome-wide polymerase activities by detecting the traces of rNMPs misincorporated by mutated Polδ (cdc6 L591G ), Polε (cdc20 M630F ) or Polα (pol1 L850F ) (Daigaku et al., 2015;Keszthelyi et al., 2015). In Pu-Seq, respective polymerase mutant strains also carry a deletion of rnh201, the catalytic subunit of RNase H2 complex, disruption of which abrogates ribonucleotide excision repair (RER) and thus stabilises misincorporated rNMPs (Daigaku et al., 2015). To assess activities of individual replicative polymerases, we employed a strategy previously used to analyse Okazaki fragment sequencing data (Petryk et al., 2016). Briefly, activities of Polδ, Polε and Polα at any given locus are expressed as polymerase tracks, which are proportional differences of rNMPs misincorporated in the top and the bottom DNA strands (Figure 1).
To validate that 2×Polδ and 4×Polδ mutants displayed increased expression of Polδ genes, we measured relative transcript levels of cdc1 + , cdc27 + , cdc6 + /cdc6 L591G , and cdm1 + by RT-qPCR. In all genetic backgrounds tested, 2×Polδ and 4×Polδ mutants displayed a significant increase in relative transcript levels of all four Polδ genes ( Figure 2C). Unfortunately, due to the unavailability of commercial antibodies recognising Polδ subunits in S. pombe, we were unable to confirm that protein levels of the Polδ subunits were also increased. It has been previously reported, however, that plasmid-based over-expression of each of the four Polδ subunits is achievable in S. pombe (Kang et al., 2000;MacNeill et al., 1996;Reynolds et al., 1998). Consequently, we reasoned that  To determine the fundamental cellular consequences of Polδoverexpression, we assessed growth rate and cellular morphology of WT, 2×Polδ and 4×Polδ cells. Polδ-overexpressing mutants displayed WT-like growth parameters and did not develop any cellular or nuclear defects ( Figure 2D and 2E). Accordingly, increased Polδ expression did not alter the distribution of cell sizes ( Figure 2F). To assess whether increased Polδ expression influenced progression through S-phase specifically, we synchronised WT, 2×Polδ and 4×Polδ cells with a Cdc2 asM17 background in G2 by the addition of 3-Br-PP1 and analysed changes in DNA content in 15-min intervals after release. Progression through S-phase in 2×Polδ and 4×Polδ mutants was undistinguishable from WT cells ( Figure 2F), suggesting that the over-production of Polδ did not change S-phase progression. Taken together, a moderate increase in Polδ expression did not have a notable impact on cell cycle or replication progression.

Replication dynamics
To investigate the potential influence of Polδ-overexpression on replication dynamics in greater detail, we performed two independent Pu-Seq experiments, each of which addressed activities of Polδ, Polε and Polα, in WT, 2×Polδ and 4×Polδ cells. Overall, in all genetic backgrounds tested, Polδ, Polε and Polα tracks displayed very little variation (Figure 3), suggesting that increased Polδ levels did not dramatically alter the properties of replication. To capture a genome-wide view of replication, we examined regions around efficient origins of replication [characterised by estimated firing efficiency (Ori Eff ) of at least 40%] and regions constituting replication termination zones, which were defined by two efficient origins (Ori Eff > 40%) and did not contain any intermediary efficiency origins (20% < Ori Eff < 40%). Analysis of Polδ and Polε tracks associated with 259 efficient origins and 147 termination zones did not reveal any notable differences ( Figure 4A and 4B). We observed that Polα tracks in 2×Polδ cells displayed marginal deviation from the WT profile ( Figure 4A and 4B); however, considering that the observed difference was not reflected in 4×Polδ cells, we concluded this observation represented a technical, rather than biological phenomenon. We reasoned that if increased Polδ levels negatively affected replisome progression, 2×Polδ and 4×Polδ mutants would be expected to display increased activity of low and intermediary efficiency origins. Polδ-overexpressing cells, however, retained a WT-like distribution of genome-wide origin efficiencies, which further indicated normal replication progression ( Figure 4C and 4D). Taken together, we concluded that, in our experimental system, a moderate increase in Polδ levels did not result in any observable changes in replication dynamics.

Conclusions
In this study, we tested whether a moderate (2-4-fold) increase in Polδ expression impairs, or in any way alters, replication dynamics under normal conditions in S. pombe. The presented experiments were inspired by report that a two-fold and four-fold genes (cdc1 + , cdc27 + , cdc6 + /cdc6 L591G and cdm1 + ) in the indicated mutants measured by RT-qPCR. Mutants designated as pol1 L850F , cdc6 L591G and cdc20 M630F also carried rnh201∆. Individual points represent data from independent experiments. For WT, 2×Polδ and 4×Polδ cells, 19 independent measurements were taken (all genetic backgrounds combined). Horizontal lines represent means. Statistical significance was determined by the unpaired two-sample t-test. ** p ≤ 0.01; *** p ≤ 0.001;**** p ≤ 0.0001 (D) Representative growth curves of WT, 2×Polδ and 4×Polδ cells. Optical density (OD) was measured in 1h intervals for total 10 h. Time-series of log 2 -transformed OD measurements are presented. Red lines represent linear regression models. Slopes of linear regression models (k) and calculated doubling times are indicated. (E) Representative images of WT, 2×Polδ and 4×Polδ cells stained with DAPI. Composite images of DIC and DAPI channels are shown. Scale bar represents 5 µm. increase in Polδ concentration reduces the rate of the leading strand synthesis in vitro, hypothesised to be due to stochastic polymerase switching, during which Polδ outcompetes Polε and temporarily facilitates the extension of the leading strand (Yeeles et al., 2017).
We constructed a set of strains carrying either one or three extra copies of all Polδ genes and validated that these Polδ integrations resulted in increased transcription of the respective Polδ components: cdc1 + , cdc27 + , cdc6 + and cdm1 + . We were unable to explore if the Polδ subunits were upregulated at  the protein level. However, considering that successful ectopic over-production of Polδ subunits has been reported in the seminal literature (Kang et al., 2000;MacNeill et al., 1996;Reynolds et al., 1998), we argue that our experimental design conveyed a genuine Polδ over-production.
We determined that cells characterised by up to four-fold increased Polδ expression do not exhibit defects in growth and cell cycle progression. Furthermore, utilising Pu-Seq methodology, we demonstrated that genome-wide replication dynamics in 2×Polδ and 4×Polδ mutants is virtually indistinguishable from WT, arguing against the notion of stochastic polymerase switching or any other impairment of DNA replication induced by over-production of Polδ. Naturally, it is still possible that we simply did not reach the threshold of Polδ expression that is required for the polymerase-switch to occur at frequencies detectable by Pu-Seq. Higher cellular levels of Polδ could be achieved by ectopic or strong promoter-driven expression of Polδ genes; however, we argue that such an extensive Polδ over-production would constitute a non-physiological system, which would no longer be biologically relevant in relation to the reported in vitro data (Yeeles et al., 2017). Moreover, it has been shown that gross over-expression of cdc6 + is detrimental to overall cell physiology (MacNeill et al., 1996), which would likely make Pu-Seq experiments difficult to interpret or impossible to carry out. We also argue that promoter manipulation or plasmid-based over-expression would disrupt the stoichiometry of Polδ subunits, which could be detrimental to Polδ folding and function.
While we established that moderate over-expression of Polδ does not noticeably affect canonical replication, we acknowledge that presented data do not sufficiently disprove the natural occurrence of the stochastic switch from Polε to Polδ. Nevertheless, our data do imply that, if such events occur in vivo, they manifest at low frequencies and likely represent only a marginal disturbance to an overwhelmingly robust replication program.

Data availability
Underlying data Gene Expression Omnibus: Raw and processed Pu-Seq data, Accession number GSE165503; https://identifiers.org/geo: GSE165503. Zach and Carr have generated fission yeast strains that express additional copies of the 4 polymerase delta genes so that there are either 2 total copies or 4 total copies present in the genome. This design parallels the Yeeles in vitro work that similarly compared a 2x and 4x increase in polymerase delta.
The authors do not have antibody recognizing their polymerase delta proteins, but instead show that transcript levels of all 4 genes scale by qPCR. They characterize the effect of 2x, and 4x polymerase delta expression for its effects on: cell morphology, growth/doubling, and DNA content/cell cycle progression. Their calculation of growth uses absorbance readings, and is accompanied by cell length measurements to show that absorbance increase is not attributable to altered cell size. All results suggest that 2x and 4x increases in polymerase delta genes have no apparent effect on fission yeast cell fitness. Further, this over expression model does not noticeably slow DNA synthesis or cause replication-dependent effects on cell health and growth.
The methods are clearly written, and provide thorough descriptions of materials and equipment used. Interestingly, the authors used 2 separate methods for Pu-Seq library preparation. The impact of using 2 different methods is not described in the text, and data from 1 experiment is shown in Figure 1. The "...Means of two independent experiments are shown." in figure 3, and the data suggests that differences between prepared libraries is minimal. Indication of replicates would be helpful if not explicitly stated (elaborated below).
Overall, the authors provide a clear test of a specific hypothesis: that increased polymerase delta expression will impede synthesis. The data is clean, the premise is interesting, and the subject is worthy of study. The issue of not seeing protein level to characterize polymerase delta levels is minor in this initial in vivo characterization, given the characterization of transcript levels.
Minor comments: The methods are well-described and clear. Some elaboration of replicates would improve the text to be very clear throughout. Could experimental replicate number be indicated for 2C? Figure 3 suggests that the means of the 2 separate Pu-Seq methods are very close-is this the case? Are both Pu-Seq library replicates used in Figure 4 calculations? ○ If polymerase delta switching to the leading strand impacts replisome speed, could this be seen in polymerase epsilon tracks? Figure 3 suggests that while the patterns are very similar for Pol e under over expression conditions, the overlap may not be perfect; in contrast, there is apparently tight overlay for normal/2x/4x in polymerase delta and alpha tracks. Is a non-perfect overlap in polymerase epsilon tracks the residual effect of a transient situation?
Since Figure 3 shows the mean of 2 independent experiments, is the variation lost by combining individual experiments (prepared differently)?
○ I have no complaint with the use of qPCR to quantify expression of the polymerases delta components, and transcript levels appear to scale by ~2x and ~4x relative to the baseline amount of each, independent transcript. Based on known promoter activities, how does the authors' 2x or 4x level of transcript increase (from amplified copies driven by the native promoters) compare to ectopic promoter-driven expression from MacNeill (1996), Kang ○ may not be perfect; in contrast, there is apparently tight overlay for normal/2x/4x in polymerase delta and alpha tracks. Is a non-perfect overlap in polymerase epsilon tracks the residual effect of a transient situation? Since Figure 3 shows the mean of 2 independent experiments, is the variation lost by combining individual experiments (prepared differently)? We argued that, if replisome speed was affected, we would detect changes in origin firing. Origins efficiency profiles in WT and Polδ-overexpressing cells; however, are comparable (Figure 4 -C, D). It is true that, according to Figure 3, Polε tracks in WT and Polδoverexpressing cells do not display perfect overlap. We believe this is due to slightly higher inter-experimental variability in WT Polε datasets, as is now indicated in revised Figure 1. Intriguingly, although the canonical model divides the workload of leading strand synthesis and lagging strand synthesis between Polε and Polδ, in vitro and in vivo studies showed evidence of Pol δ activity on both strands under certain circumstances. Yet, in vitro reconstitution of replication indicated that although increased dosage of Polδ can out compete Polε on the leading strand, this is much less efficient for its synthesis. The current work tests the hypothesis that whether the same effect exist in vivo, which will provide more understanding on replication dynamics in living cells.
Using a carefully designed Cre-Lox system, the authors constructed strains with two-or four-fold increased Polδ holoenzyme with the ability to incorporate rNTPs for following assessment of polymerase usage. With various analysis on cell growth and genome-wide quantification of individual polymerase usage especially at origins and termination regions, they concluded that there is no detectable impact in replication dynamics or associated growth defect caused by increased amounts of Polδ.
Major Comments: It will be nice to also show an example for termination site identification through Pu-seq as the demonstration for origin in Fig 1. 1.
In the method section, the authors mentioned that an independent set of modified GLOEseq was also performed along side Pu-seq. Are the GLOE-seq results comparable with those of Pu-seq? Is this work still undergoing? 2.
Increased amount of Polδ transcripts were confirmed with the over-expression system, but increased usage of Polδ is barely evident as shown by this work. Although not required for the current work, it will be nice to be followed up by testing whether or not more Polδ subunits can be incorporated into the replisome using iPOND or similar techniques.

3.
Minor points: In the analysis of Pu-seq data, PT is calculated as (Rt+Rb)/(Rt-Rb). Is there certain mathematical conversion omitted here? Otherwise, how is the range of PT kept within [-1,1] as in Fig 1? 1.
It is concluded that Polα tracks in 2xPolδ is marginally different from the other two conditions. I assume it is demonstrated by the mean polymerase tracks in Fig 4A and 4B as a slight deviation of the curve of squares from the other two curves. It is better to make that clear in the figure legend, or with some statistical analysis.

2.
Is the work clearly and accurately presented and does it cite the current literature? Yes

Is the study design appropriate and is the work technically sound? Yes
Are sufficient details of methods and analysis provided to allow replication by others? Yes fact that denominator is always bigger than nominator ensures that PT falls within [-1,1]. Fig 4A and 4B as a slight deviation of the curve of squares from the other two curves. It is better to make that clear in the figure legend, or with some statistical analysis.

○
Minor deviations in Polα tracks are now signified by asterisks.

Additional changes to the manuscript:
In the introduction section, we now reference a recent paper discussing the role of Polδ in bypassing oxidative DNA lesions: