Tunability of DNA Polymerase Stability during Eukaryotic DNA Replication

SUMMARY

Structural and biochemical studies have revealed the basic principles of how the replisome duplicates genomic DNA, but little is known about its dynamics during DNA replication. We reconstitute the 34 proteins needed to form the S. cerevisiae replisome and show how changing local concentrations of the key DNA polymerases tunes the ability of the complex to efficiently recycle these proteins or to dynamically exchange them. Particularly, we demonstrate redundancy of the Pol α-primase DNA polymerase activity in replication and show that Pol α-primase and the lagging-strand Pol δ can be re-used within the replisome to support the synthesis of large numbers of Okazaki fragments. This unexpected malleability of the replisome might allow it to deal with barriers and resource challenges during replication of large genomes.

In Brief

Using single-molecule imaging, Lewis et al. show that the yeast DNA-replication machinery does not stably bind DNA polymerases but that their stability is tunable through their concentration in solution. Furthermore, they demonstrate that Pol δ and Pol α are recycled to support the synthesis of multiple Okazaki fragments.

Graphical Abstract

INTRODUCTION

To robustly synthesize genomic DNA, the eukaryotic replisome requires a large number of interacting protein factors with different enzymatic activities. Key components include the 11-subunit CMG helicase and three different multi-subunit B-family DNA polymerases: the leading-strand Pol ε, lagging-strand Pol δ, and Pol α-primase (Bell and Labib, 2016; Burgers and Kunkel, 2017). CMG unwinds DNA by translocating along one of the strands in a 3′ to 5′ direction while forming a complex with Pol ε (Langston et al., 2014; Sun et al., 2015) to support highly processive synthesis of DNA on the leading strand. On the lagging strand, Pol α-primase generates ~25-nt RNA-DNA primers that Pol δ extends to generate ~150-bp Okazaki fragments (Bell and Labib, 2016). The commonly accepted model of replication depicts Pol ε stably anchored to CMG but shows Pol δ not physically tethered to the replisome and subsequently replaced for the synthesis of each Okazaki fragment (Bell and Labib, 2016). Biochemical and structural studies have provided valuable insights into basic enzymatic activities and overall architecture of the eukaryotic replisome. However, the dynamic behavior of the various replisomal components is largely unexplored due to the challenges associated with averaging over ensembles of molecules that is needed to gain structural and functional insight.

We demonstrate here the reconstitution of the S. cerevisiae replisome by purified protein factors and the visualization of processive DNA replication at the single-molecule level. We show that all three replicative polymerases can remain stably associated with the replisome for the synthesis of tens of kilobases (kb) and that the DNA synthesis activity of Pol α-primase is dispensable under these conditions. Pol α-primase and Pol δ are retained at the fork while synthesizing large numbers of successive Okazaki fragments. This unexpected observation of recycling of the lagging-strand polymerases without dissociation from the replisome directly challenges textbook models and implies physical connectivity between Pol δ and the replisome and the formation of loops in the lagging-strand DNA. We demonstrate that this retention of Pol δ is facilitated through the Pol32 subunit of Pol δ (Huang et al., 1999; Johansson et al., 2004). Interestingly, when challenged with competing polymerases, all three polymerases are able to exchange from solution into a moving replisome in a concentration-dependent manner. This exchange is in contrast to the CMG helicase, which remains stably associated even when challenged with excess helicases in solution. We propose the balance between replisome stability and plasticity enables the complex to provide access to other binding partners during S phase while not compromising the stability and robustness of the replisome.

RESULTS

Single-Molecule DNA Replication

We visualized individual budding yeast replisomes using a real-time, single-molecule fluorescence assay that allows us to monitor simultaneously DNA synthesis and protein dynamics. We assembled linear, double-stranded DNA (dsDNA) molecules (18.3 kb in length) in a microfluidic flow cell placed onto a fluorescence microscope. The DNA is stretched and attached to the surface at both ends (Figures S1A and S1B). A pre-made synthetic replication fork at one end of the DNA enables direct loading of the replisomal proteins onto the DNA (Figure 1A, left). DNA synthesis is initiated by loading CMG and the Mcm10 initiation factor (CMGM) onto the template followed by the introduction of Ctf4, Mcm10, Mrc1-Tof1-Csm3 (MTC), PCNA, RFC, RPA, DNA polymerases α, δ, ε, Mg²⁺, the four dNTPs, and four rNTPs (Figure 1A, right) (Georgescu et al., 2015). Real-time synthesis trajectories were obtained from near-TIRF fluorescence imaging of SYTOX orange (S.O.)-stained dsDNA in the absence of buffer flow. As DNA synthesis proceeds, the leading strand appears as a diffraction-limited spot that moves unidirectionally along the template DNA. Initially weak in intensity, the spot increases in intensity as more dsDNA is generated at the leading strand and accumulates into a collapsed globular coil with dimensions smaller than the diffraction-limited resolution of the imaging system (Figure 1B, left; Video S1). The intensity per base pair of this coiled leading-strand product is similar to the intensity per base pair as measured over the stretched template (Figures S1C and S1D). To establish that the observed events correspond to simultaneous leading- and lagging-strand synthesis, we divided the DNA template into three regions: the leading-strand spot (“lead”), the length of DNA behind it (“lag”), and the length ahead of it (“parental”) (Figure 1B, right). For every time point, the DNA content is calculated from the integrated fluorescence intensity in each region (Ganji et al., 2018). During replication, the DNA content of the parental region decreased, while DNA content of the leading- and lagging-strand regions increased simultaneously (Figures 1B, right, and S1E). Importantly, in the absence of either Mg²⁺, nucleotides, CMG, or DNA polymerases, replication events were not observed (Figure S2A).

Figure 1. Single-Molecule Visualization of DNA Synthesis.

(A) Schematic representation of DNA-replication assay (see Method Details).

(B) Left: kymograph showing DNA replication on a single DNA substrate. The leading-strand tail appears as a bright spot that moves in a unidirectional manner while simultaneously increasing in intensity. Right: length of the lagging strand (“lag”), leading strand (“lead”), and parental DNA (“parental”) as a function of time, measured by the integrated intensity of the dsDNA.

(C) Single-molecule rate distribution. The bimodal distribution was fit with the sum of two Gaussians (black line) with rates of 8 ± 2 bp/s and 33 ± 2 bp/s (n = 96 molecules). Errors represent SEM.

(D) Schematic representation of the pre-assembly DNA replication assay (see Method Details).

(E) Left: kymograph showing activity of a pre-assembled replisome on a single DNA substrate in the absence of polymerases in solution. Right: length of the lagging strand (yellow), leading strand (blue), and parental DNA (gray) as a function of time.

(F) Example kymograph showing the DNA (gray) and labeled RPA (pink). RPA colocalizes with the leading-strand DNA spot.

(G) Example kymograph showing DNA (gray) and labeled RPA (pink) during leading-strand DNA synthesis. RPA accumulates on the single-stranded lagging-strand product.

See also Figures S1 and S2.

To quantify the instantaneous rates of replication, we tracked the position of the leading-strand spot (Tinevez et al., 2017). The measured population-averaged rate of 19 ± 6 bp/s (mean ± SEM; n = 96 molecules) (Figures S2B and S2C) is consistent with previously reported ensemble in vitro and in vivo measurements (Aria and Yeeles, 2018; Hodgson et al., 2007; Sekedat et al., 2010; Szyjka et al., 2005; Tourriére et al., 2005; Yeeles et al., 2017). Rates of DNA synthesis varied within individual replisomes (Figures 1C, S2B, and S2C), with the single-molecule rate distribution having two distinct peaks at 8 ± 2 bp/s and 33 ± 2 bp/s (mean ± SEM; n = 96 molecules, 315 segments). This bimodal rate distribution was reported in our previous single-molecule studies of leading-strand synthesis and is attributed to dynamic interaction of the MTC complex with the replisome (Lewis et al., 2017b). Observation of the same rate distribution here implies MTC acts in a similar fashion during combined leading- and lagging-strand DNA synthesis.

Pre-assembled Replisomes Replicate DNA as a Highly Stable Entity

To measure the stability of synthesizing replisomes on DNA, we carried out single-molecule, pre-assembly replication assays. In these assays, the replisome is first pre-assembled on surface-tethered DNA in the flow cell (Figure 1D, left). Next, the flow cell is washed and replication is initiated by introduction of a replication solution that omits all three polymerases. This protocol ensures only the initially assembled polymerases remain, eliminating the possibility of other polymerases associating with the replisome (Figure 1D, right). We hypothesized that the requirement for a new Pol δ for each Okazaki fragment would prevent processive synthesis. Surprisingly, these conditions support processive DNA replication, with synthesis rates and product lengths similar to those measured with excess polymerases in solution (Figures 1E and S2C–S2E). To confirm that the lagging strand does not contain single-stranded DNA (ssDNA) gaps, we repeated the reaction in the presence of fluorescently labeled RPA. The labeled RPA signal colocalizes with the leading-strand spot, and we do not see RPA binding on the lagging strand in the wake of the replication fork (Figure 1F). In contrast, when we omit proteins needed for lagging-strand synthesis, we see accumulation of RPA on the single-stranded lagging strand (Figure 1G). Together, these observations suggest Pol δ can remain stably associated to the replisome, challenging the current view that a new Pol δ holoenzyme is recruited to extend each Okazaki fragment (as reviewed in Bell and Labib, 2016). These results also reveal that the Pol α-primase is stably associated to the replication machinery and primes multiple Okazaki fragments.

Direct Visualization Confirms Presence of Single Polymerases in a Processive Replisome

To further explore the observation that Pol δ remains tethered to the replisome, we repeated our single-molecule, pre-assembly assay in the presence of fluorescently labeled polymerases (Figure 2A). Labeling did not affect the average rates of DNA synthesis (Figure S2C). The kymographs in Figure 2A show bright fluorescent spots for each of the labeled polymerases during DNA synthesis. The DNA polymerases colocalize with the leading-strand spot, consistent with the polymerases being stably incorporated into reconstituted replisomes. To determine the stoichiometry of Pol ε, we divided the intensity at the fork by the intensity of a single polymerase. We observe the continuous presence of one Pol ε, one Pol δ, and two Pol α-primases at actively synthesizing replication forks (Figure 2B). To further confirm this analysis, we repeated the experiment under conditions that allow for photobleaching of fluorophores covalently bound to Pol ε. We calculated the number of photobleaching steps of Pol ε at the fork using an unbiased fitting algorithm based on change-point theory (Figure S3A) (Duderstadt et al., 2016; Hill et al., 2018). Consistent with our previous analysis, we find one Pol ε at the fork (Figure S3B). DNA synthesis on the lagging strand is supported by both Pol δ and the DNA polymerase activity of the multifunctional Pol α-primases. To investigate the influence of DNA synthesis activity by Pol α-primase on the lagging strand, we repeated the pre-assembly assay with a mutant of Pol α-primase that is able to produce RNA primers but is unable to extend these into DNA (Pol α^Cat, Figure S4). Remarkably, the DNA polymerase activity of Pol α-primase is not required during processive leading- and lagging-strand synthesis (Figures 2C and S2C). Taken together, these observations are consistent with a model in which Pol δ supports lagging-strand synthesis (Aria and Yeeles, 2018; Yeeles et al., 2017) and Pol α-primase is only required for its primase activity.

Figure 2. Direct Visualization of Polymerase Stability and Stoichiometry during Processive DNA Synthesis.

(A) Example kymographs showing activity of pre-assembled replisomes on single DNA substrates in the absence of excess polymerases in solution. The DNA(left), labeled Pol ε, Pol δ, and Pol α-primase (middle), the polymerase intensities as a function of time (right). All polymerases colocalize with the leading-strand spot.

(B) Distribution of the number of Pol ε (blue), Pol δ (yellow), and Pol α-primase (turquoise) at the fork. A Gaussian fit (black line) gives 1.0 ± 0.2 (mean ± SEM; n = 70), 1.0 ± 0.2 (mean ± SEM; n = 70), and 1.8 ± 0.4 (mean ± SEM; n = 42), respectively.

(C) Left: kymograph showing DNA replication of pre-assembled replisomes containing Pol α^Cat. Right: length of the lagging strand (yellow), leading strand (blue), and parental DNA (gray) as a function of time.

See also Figures S3 and S4.

Direct Visualization of Concentration-Dependent Exchange of Pol ε

Our observations and previous studies (Douglas et al., 2018; Langston et al., 2014; Lewis et al., 2017b) suggest that Pol ε is a stable component of the replisome. However, recent single-molecule experiments have demonstrated that stable components of multi-protein complexes can undergo dynamic exchange when challenged with competing binding partners (Delalez et al., 2010; Graham et al., 2011; Li et al., 2019; Scherr et al., 2018). In particular, complexes held together by multiple weak interactions support high stability in the absence of competing factors but rapid exchange in the presence thereof (van Oijen and Dixon, 2015). We hypothesized that the multiple contact points between Pol ε and the replisome may similarly allow it to exchange dynamically in the presence of free polymerases in solution (Åberg et al., 2016). To visualize such polymerase dynamics, we repeated the assay but now with labeled polymerases in solution. The presence of excess polymerases in solution did not result in changes to replication kinetics (Figure S2C). In the presence of polymerases in solution, we detect on average one Pol ε at the replication fork (Figure 3A). Next, to identify whether exchange occurs, we carried out single-molecule FRAP (fluorescence recovery after photobleaching) experiments (Beattie et al., 2017; Lewis et al., 2017a; Spenkelink et al., 2019). Using labeled Pol ε, we photobleached all labeled polymerases at the replication fork using a pulse of high laser intensity after initiation of replication (Figure 3B, left). After bleaching, we monitored fluorescence recovery to see if new unbleached Pol ε from solution could exchange into the replisome (Figure 3B, right). Figure 3C shows a kymograph of the fluorescence recovery of Pol ε in the presence of 20 nM Pol ε in solution. We obtained an average exchange rate of ~1 per 10 min for these conditions (i.e., (12.5 ± 0.1) × 10⁻⁴ s⁻¹, Figure S5A). This rate is equivalent to exchange of one Pol ε over a length of ~15 kb. Decreasing the labeled Pol ε concentration from 20 to 2 nM, the mean exchange rate decreased (Figure 3D). These results demonstrate that Pol ε undergoes concentration-dependent exchange at the replication fork.

Figure 3. Visualization of Pol ε Dynamics.

(A) Number of Pol ε at the fork as a function of excess polymerases in solution. Errors represent SEM.

(B) Schematic representation of the FRAP assay.

(C) Example kymograph showing exchange of Pol ε at the fork with DNA (left), labeled Pol ε (middle), and the labeled Pol ε intensity (right).

(D) Exchange rate as a function of Pol ε concentration. The line represents a hyperbolic fit, giving a maximum exchange rate of (1.4 ± 0.5) × 10⁻³ s⁻¹ (mean ± error of the fit).

See also Figures S5 and S7.

As an internal control, we set out to characterize the exchange kinetics of the CMG helicase (Figure 4A). A number of mechanisms are in place during the cell cycle to achieve the precise loading of exactly one CMG per replication fork (Abid Ali et al., 2017; Bell and Labib, 2016; Douglas et al., 2018). Combined with structural data that point to a role of CMG as the central organizer of the replisome, we hypothesized that CMG would remain stably bound during elongation. Using fluorescently labeled CMG, we observe a single CMG remaining stably bound during DNA synthesis on timescales longer than 30 min, even when challenged with excess CMG in solution (Figures 4B and 4C).

Figure 4. CMG Is a Stable Component of the Replisome during Replication.

(A) Schematic representation of the assay.

(B) Example kymograph showing the DNA (left), labeled CMG (middle), and the CMG intensity as a function of time (right). CMG colocalizes with the leading-strand DNA spot.

(C). Average number of CMG molecules at elongating replication forks over time, giving 1.0 ± 0.7 (mean ± SEM; n = 89).

Direct Visualization of Pol α-primase Dynamics and Stoichiometry

To test whether Pol α-primase displays exchange behavior similar to Pol ε, we repeated the FRAP experiments using labeled Pol α-primases. Figure 5A shows a kymograph of the fluorescence recovery of Pol α-primase in the presence of 20 nM Pol α-primase in solution. We obtained an average exchange rate of ~1 per 10 min (i.e., (1.7 ± 0.1) × 10⁻³ s⁻¹) for these conditions. As with Pol ε, decreasing the labeled Pol α-primase concentration from 20 to 2 nM decreased the mean exchange rate (Figure 5B). Interestingly, at low polymerase concentrations, only one Pol α-primase molecule was observed at the fork (Figure 5C). This shows that the binding affinities for the two Pol α-primase molecules are different, suggesting a unique hierarchy of functional interactions that keep each Pol α-primase attached to the replisome. To understand how the Pol α-primase exchange rate compares to the Okazaki-fragment cycling time, we quantified the average length of Okazaki fragments under our experimental conditions. We used fluorescently labeled RPA to assess the amount of ssDNA as a measure of the Okazaki fragment size (Figure 5E, left). As shown before (Figure 1F), RPA was always colocalized at the replication fork. In agreement with biochemical studies (Georgescu et al., 2015), we observe that the number of RPA molecules is dependent on the concentration of Pol α-primase (Figure 5E, right). Given that the footprint of RPA is 30 ± 2 nt (Figure 5D) and at 20 nM Pol α-primase we see 3.7 ± 0.5 (mean ± SEM; n = 64) RPA molecules at the fork, we determine that the average Okazaki fragment length in our single-molecule assays is 111 ± 16 bp (mean ± SEM), consistent with in vivo studies (Bell and Labib, 2016). To further confirm that lagging-strand synthesis is unaffected by the omission of Pol α-primase polymerase activity, we also quantified the number of RPAs at preassembled replisomes containing Pol α^Cat. Under these conditions, we see 9 ± 1 RPAs at the fork. This number is as observed for conditions containing wild-type Pol α-primase (Figure 5E, right).

Figure 5. Visualization of Pol α-primase Dynamics.

(A) Example kymograph showing the DNA (left) and labeled Pol α-primase (middle) and corresponding labeled Pol α-primase intensity at the fork (right).

(B) Exchange rate as a function of Pol α-primase concentration. The line represents a hyperbolic fit, giving a maximum exchange rate of (2.3 ± 0.2) × 10⁻³ s⁻¹ (mean ± error of the fit).

(C) Number of Pol α-primase at the fork at as a function of excess polymerases in solution. Errors represent SEM.

(D) Histograms of the number of RPA molecules binding to either a 57-mer (purple, n = 1299) and 99-mer oligo (pink, n = 939). The average ssDNA-RPA footprint is30 ± 2 nt. Error represents SEM.

(E) Left: schematic representation showing the relationship between the number of RPA molecules and Okazaki-fragment length. Right: scatterplots of the number of RPA for preassembled Pol α^Cat (9 ± 1; n = 68), preassembled Pol α-primase (9. ± 1; n = 64), 2 nM (7.0 ± 0.9; n = 51), 20 nM (3.7 ± 0.5; n = 64), and 70 nM (1.5 ± 0.3; n = 60) Pol α-primase in solution. The black line and gray box represent the mean and SEM.

See also Figures S5 and S7.

Direct Visualization of Pol δ Dynamics

We next visualized labeled Pol δ in the presence of excess polymerases in solution. Interestingly, in 86% ± 5% (mean ± SEM; n = 55) of events DNA synthesis begins only after binding of Pol δ, consistent with a model where assembly of Pol δ into the replisome stimulates processive simultaneous leading- and lagging-strand synthesis. We repeated the FRAP experiments with labeled Pol δ (Figures 6A and S5B). We observe that the exchange rate of Pol δ is also dependent on the concentration of Pol δ in solution (Figure 6C). Surprisingly, even at the highest concentration of Pol δ that still allows visualization of single molecules (20 nM), the Pol δ exchange rate is such that it would correspond to the synthesis of many Okazaki fragments. At an average replication rate of 19 ± 6 bp/s (Figure S2C), our observation suggests that Pol δ is retained within the replisome for synthesis of 142 ± 78 Okazaki fragments and therefore must have stabilizing contacts with other protein factors within the replisome.

Figure 6. Visualization of Pol δ Dynamics.

(A) Example kymograph showing the DNA (left) and labeled Pol δ (middle) and corresponding labeled Pol δ intensity at the fork (right).

(B) Example kymograph showing the DNA (left) and labeled Pol δ³²⁻(middle) and corresponding labeled Pol δ³²⁻ intensity at the fork (right).

(C) Exchange rate as a function of polymerase concentration with Pol δ (yellow) and Pol δ³²⁻ (magenta). The lines represent hyperbolic fits, giving a maximum exchange rate of (1 ± 3) × 10⁻³ (mean ± error of the fit) for Pol δ (yellow) and (4 ± 8) × 10⁻³ (mean ± error of the fit) Pol δ³²⁻ (magenta).

See also Figures S5–S7.

We then set out to identify the mechanism through which Pol δ is retained in replisomes. The Pol32 subunit of Pol δ is documented to interact with other protein factors in the replisome, such as the Pol1 subunit of Pol α-primase (Huang et al., 1999; Johansson et al., 2004). Thus, we predicted that elimination of contact between Pol32 of Pol δ and the replisome would result in a change in exchange rate of Pol δ into the replisome. If the Pol32 subunit provides an important interaction between Pol δ and the replisome, the exchange of Pol δ³²⁻ (Pol δ lacking the Pol32 subunit) should be measurably faster than the exchange of the complete Pol δ holoenzyme. To test this hypothesis, we fluorescently labeled Pol δ³²⁻ and repeated the FRAP measurements (Figures 6B and S5C). Indeed, the exchange rate of Pol δ³² is 2.5-fold faster than the exchange of the Pol δ holoenzyme ((27.0 ± 0.5) × 10⁻⁴ s⁻¹ compared to (12 ± 5) × 10⁻⁴ s⁻¹) (Figure 6C). These data show that Pol32 plays an important role in stabilizing Pol δ in the replisome.

DISCUSSION

We report here the reconstitution and visualization at the single-molecule level of DNA synthesis by budding yeast replisomes. Our real-time fluorescence assay allows us to directly visualize replication kinetics and quantify protein dynamics in individual replisomes—observables that are not accessible via classical biochemical approaches. The average observed replication rates are similar to those previously reported in ensemble biochemical reactions (Aria and Yeeles, 2018; Georgescu et al., 2015; Kurat et al., 2017; Langston et al., 2017; Yeeles et al., 2017) and are within the range of replication fork rates inside the cell (Conti et al., 2007; Hodgson et al., 2007; Sekedat et al., 2010). We show that the eukaryotic replisome acts as a stable processive machine under dilute conditions. In the presence of excess polymerases, however, the replisome is not fixed in composition but instead is a highly dynamic complex continually exchanging major components. This model is founded upon three observations: (1) in the absence of all three replicative polymerases in solution, the replisome forms a stable complex able to support processive, concerted leading- and lagging-strand synthesis (Figure 2); (2) Pol δ can be retained at the replication fork for multiple Okazaki fragments, mediated at least in part through an interaction with Pol32 (Figure 6); and (3) the leading-and lagging-strand polymerases exchange during DNA synthesis in a concentration-dependent manner without affecting rates of replication (Figures 3, 6, and S2C).

In contrast to long-standing views that replisome architecture is static, single-molecule fluorescence experiments have documented dynamic exchange of components at physiologically relevant timescales in large multi-protein complexes across all domains of life (Geertsema et al., 2014; Liao et al., 2016). Our work provides the first direct evidence that all three polymerases (Pol α-primases, Pol ε, and Pol δ) are similarly exchanged from solution in a concentration-dependent manner during DNA synthesis. In the absence of polymerases in solution, the original polymerases are retained, and the replisome forms a highly stable complex resistant to dilution (Figure 7A). In the presence of excess polymerases, they are exchanged into the replisome at a rate dependent on their concentration (Figure 7B). Concentration-dependent exchange can be rationalized through an interaction network consisting of multiple weak interactions. Under dilute conditions, transient disruption of any one of these interactions would be followed by its rapid reformation to prevent dissociation. If, however, there are exogenous competitors in close proximity to the complex, one of these can bind at a transiently vacated binding site and, consequently, be at a sufficiently high local concentration to compete out the original protein.

Figure 7. Compositional Plasticity at the Eukaryotic Replication Fork.

(A) All three replicative polymerases can form a stable complex in the absence of free polymerases in solution, indicating that the lagging-strand Pol α-primase and Pol δ can recycle for synthesis of many Okazaki fragments (Pol ε, blue; Pol α-primases, turquoise; Pol δ, yellow; CMG, green; PCNA, red; RPA, pink). The Pol δ linkage is unclear, but its Pol32 subunit is involved. The Ctf4 and MTC components of the replisome are not shown for clarity.

(B) In the presence of excess polymerase, all three Pols undergo exchange in a concentration-dependent manner. However, CMG remains stably bound and does not undergo exchange.

A concentration-dependent exchange mechanism likely plays an important role in genomic integrity (Mueller et al., 2019). Through such a mechanism, the replisome has access to a plurality of molecular pathways to achieve and ensure continuous replisome progression under a variety of cellular stresses. Repriming of leading-strand synthesis past DNA damage is highly inefficient (Taylor and Yeeles, 2018). As a result, the replisome need to recruit specialized repair polymerases directly to the fork for efficient bypass of damage. Concentration-dependent exchange of Pol ε on the leading strand provides a simple molecular mechanism to grant repair polymerases access to the replisome during the initial response to DNA damage. While our biochemical and experimental knowledge of the eukaryotic replisome is not as detailed as it is for the bacterial and viral systems, our report implies bacterial and eukaryotic replication machinery utilize the same basic operating principles; however, they differ in the way in which they assemble and coordinate these processes (Yao and O’Donnell, 2016)

Dynamic exchange of polymerases also provides the replisome with a straightforward way to pass through cohesion rings that hold together replicated sister chromatids. Cohesin can move over obstacles ~11 nm in size (Stigler et al., 2016). Since the size of the replisome is much larger (Sun et al., 2015), it is difficult to envisage how it can pass through cohesin rings. We speculate that CMG (~10 nm) may fit through the pore and that dynamic exchange of other components allows the replisome to pass as a minimal complex to be re-joined with its polymerases immediately after.

Our results show that both the exchange rate and the stoichiometry of Pol α-primase at the fork are dependent on the concentration of Pol α-primase in solution. At low concentrations, only one Pol α-primase is present at the fork, likely bound to Ctf4 (Simon et al., 2014). At high concentrations, however, we observe two Pol α-primasess colocalizing with the replisome. Due to the diffraction-limited nature of our experiments, we cannot directly comment on the exact binding site of both Pol α-primasess. Furthermore, we also cannot determine if the second Pol α-primase participates in primer synthesis or if it is bound elsewhere near the fork. The redundancy of the DNA polymerase activity of Pol α-primase suggests it may perform other functions in the cell other than priming in DNA replication. Taken together, these results highlight the importance of determining the functional significance of these observations and whether a second Pol α-primase allows efficient bypass across obstacles.

It is generally assumed that Pol δ is not associated with CMG and that a new Pol δ is recruited for the synthesis of each successive Okazaki fragment. In contrast, our results show that Pol δ can be retained at the fork for synthesis of multiple Okazaki fragments. This observation implies that there are one or more interactions between Pol δ and a stable part of the replisome. We discover that one important interaction is through the Pol32 domain of Pol δ—without this domain, Pol δ exchange is ~2.5-fold faster (Figure 6). It has been shown that Pol32 interacts with the Pol1 subunit of Pol α-primases. Pol α-primase has a specific interaction with CMG (Georgescu et al., 2015) and can be tethered to CMG via interactions with Ctf4 and Mcm10 (Gambus et al., 2009; Warren et al., 2009). We, therefore, propose that Pol δ could be tethered to CMG, at least in part mediated by an interaction with Pol α-primases. Retention of Pol δ over the timescales presented here, combined with the RPA stoichiometry, implies that lagging-strand replication loops can be formed at the eukaryotic replication fork (Figure 7) (Chastain et al., 2003).

STAR★METHODS

Detailed methods are provided in the online version of this paper and include the following:

LEAD CONTACT AND MATERIALS AVAILABILITY

Further information and request for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Antoine van Oijen (vanoijen@uow.edu.au). We are glad to share all unique/stable reagents with reasonable compensation by requestor for its shipping.

EXPERIMENTAL MODEL AND SUBJECT DETAILS

Yeast and bacterial strains

Proteins were purified from Saccharomyces cerevisiae strain W303 (genotype: ade2–1 ura3–1 his3–11,15 trp1–1 leu2–3,112 can1–100 bar1Δ MATa pep4::TKANMX6), a gift from the Brian Chait lab (Rockefeller University), was modified to express the proteins listed in the Key Resources Table using linearized plasmids with standard genetic procedures. Proteins purified from Escherichia coli were transformed with plasmids for overexpression in BL21-DE3 codon plus RIL cells as detailed in the Key resources table.

METHOD DETAILS

Protein purification and labelling

Purification of Pol δ^32-

Pol δ was purified as previously described (Georgescu et al., 2014), the final step of which involves an elution from a sulphopropyl cation exchange column (GE Healthcare) with a 100–500 mM NaCl gradient in a buffer containing 350 mM potassium glutamate and 25 mM HEPES pH 7.5. Whereas, Pol δ (as a complete holoenzyme) elutes around 350 mM NaCl, Pol δ^32- elutes around 250 mM NaCl, allowing complete separation (Figure S6). Thus, peak fractions at 250 mM NaCl were aliquoted, flash frozen, and stored at −80°C.

Purification of Pol α^Cat

Pol α^Cat contains the DNA polymerase active site mutations Pol 1–D996S, D998S and was purified as previously described in (Georgescu et al., 2015) with few modifications. Briefly, we inserted a 2× strep tag at the N terminus of Pol 12 for elimination of N-terminally proteolyzed Pol 12. After batch purification from anti-FLAG agarose (Sigma) the eluent was mixed with 300 μL of StrepTactin resin (IBA Lifesciences) equilibrated with PBS buffer supplemented with 10% (v/v) glycerol and washed. Pol α^Cat was eluted in PBS buffer supplemented with 10% (v/v) glycerol and 10 mM desthiobiotin. Fractions were pooled, frozen in liquid N₂ and stored in aliquots at −80°C.

Production of SFP synthase reagents for protein labeling

SFP synthase was purified by nickel-NTA chromatography as previously described (Yin et al., 2006). Alexa Fluor 488 (AF488) was functionalized by Co-enzyme A (CoA) and purified by HPLC as previously described (Yin et al., 2006) and an LD650–CoA (a CoA-derivatized photostable version of Cy5) was purchased from Lumidyne Technologies (USA).

Preparation of LD650–Pol ε and AF488–Pol ε

To obtain Pol ε labeled with a single LD650 (Lumidyne Technologies) or Alexa Fluor 488 (Invitrogen) dye, we inserted the “S6” peptide (GDSLSWLLELLN) (Zhou et al., 2007) between the N terminus of Pol 2 and its 3 × FLAG tag. The resultant Pol ε–S6 plasmid was overexpressed and purified in S. cerevisiae as previously described (Georgescu et al., 2014). For labeling Pol ε–S6, SFP enzyme, and either LD650–CoA or AF488–CoA were then incubated at a 1:2:5 molar ratio for 1 h at room temperature in the presence of 10 mM MgCl₂. Excess dye and SFP enzyme were removed by purification on a Superose 6 column (GE Healthcare) with a buffer containing 25 mM Tris-HCl pH 7.5, 150 mM NaCl, 1 mM dithiothreitol, 5% (v/v) glycerol. Fractions were frozen in liquid N₂ and stored in aliquots at −80°C. The degree of labeling was determined to be ~100% for both LD650–Pol ε and A488–Pol ε by UV/vis spectrophotometry.

Preparation of LD650–CMG

CMG labeled with LD650 (Lumindyne Technologies) was produced as previously described in (Wasserman et al., 2019).

Preparation of Cy5–Pol δ

Pol δ was prepared in S. cerevisiae as previously described (Georgescu et al., 2014) and subsequently dialyzed into 250 mM NaCl, 5% (v/v) glycerol, 50 mM potassium glutamate, 25 mM HEPES pH 7.5, and 2 mM dithiothreitol. Following dialysis Pol δ was labeled with a 5-fold molar excess of Cy5–NHS (GE Healthcare) for 5 min at 4°C. Excess dye was removed by five buffer exchange steps through a centrifugal filter (50K MWCO; Millipore) following the manufacturer’s instructions. Labeled Cy5–Pol δ was frozen in liquid N₂ and stored in aliquots at −80°C. The degree of labeling was measured to be 5 fluorophores per Pol δ holoenzyme by UV/vis spectrophotometry.

Preparation of AF647–Pol δ^32-

Alexa Fluor 647 (Invitrogen), was used to label Pol δ^32-. Labeling reactions were carried out using 3-fold molar excess of dye with 58.5 μM Pol δ^32- in 320 μL of Pol δ^32- labeling buffer (30 mM Tris-HCl pH 7.6, 2 mM dithiothreitol, 300 mM NaCl, 50 mM potassium glutamate, 10% (v/v) glycerol) for 10 min at 4°C with gentle rotation. Immediately following the coupling reaction, excess dye was removed by sequential elutions from two 0.5 mL Zeba spin desalting columns (7K MWCO; Thermofisher) following the manufacturer’s instructions equilibrated in buffer containing 30 mM Tris-HCl pH 7.6, 2 mM dithiothreitol, 300 mM NaCl, 50 mM potassium glutamate, 10% (v/v) glycerol. Labeled AF647–Pol δ^32- was frozen in liquid N₂ and stored in aliquots at −80°C. The degree of labeling was measured to be 1 fluorophore per Pol δ^32- holoenzyme by UV/vis spectrophotometry.

Preparation of AF647–Pol α-primases

Alexa Fluor 647 (Invitrogen), was used to label Pol α-primases. Labeling reactions were carried out using 5-fold molar excess of dye with 3.38 μM Pol α-primase in 120 μL of Pol α-primase labeling buffer (50 mM HEPES pH 7.5, 2 mM dithiothreitol, 200 mM NaCl, 50 mM potassium glutamate, 10% (v/v) glycerol) overnight at 4°C with gentle rotation. Immediately following the coupling reaction, excess dye was removed by sequential elutions from three 0.5 mL Zeba spin desalting columns (7K MWCO; Thermofisher) following the manufacturer’s instructions equilibrated in buffer containing 30 mM Tris-HCl pH 7.6, 2 mM dithiothreitol, 300 mM NaCl, 50 mM potassium glutamate, 10% (v/v) glycerol. Labeled AF647– Pol α-primase was frozen in liquid N₂ and stored in aliquots at −80°C. The degree of labeling was measured to be 1 fluorophore per Pol α-primase holoenzyme by UV/vis spectrophotometry.

Preparation of AF647–RPA

Alexa Fluor 647 (Invitrogen) was used to label RPA. Labeling reactions were carried out using 5-fold molar excess of dye with 45 μM RPA in 550 μL of RPA labeling buffer (50 mM Tris-HCl pH 7.6, 3 mM dithiothreitol, 1mM EDTA, 200 mM NaCl, 10% (v/v) glycerol) for 2 h at 23°C with gentle rotation. Immediately following the coupling, excess dye was removed by gel filtration at 1 mL/min through a column (1.5 × 10 cm) of Sephadex G-25 (GE Healthcare), equilibrated in gel filtration buffer (50 mM Tris-HCl pH 7.6, 3 mM dithiothreitol, 1mM EDTA, 200 mM NaCl, 20% (v/v) glycerol). Labeled AF647–RPA was frozen in liquid N₂ and stored as single use aliquots at −80°C. The degree of labeling was measured to be 1 fluorophore per RPA trimer by UV/vis spectrophotometry.

DNA Substrate Preparation

Oligonucleotides and DNA

Oligonucleotides were purchased from Integrated DNA technologies (USA). Plasmid pSuperCos1 DNA was purified by Aldevron (USA).

Linear forked doubly-tethered DNA substrates

To make the doubly tethered linear fork DNA substrate, plasmid pSupercos1 DNA (van Loenhout et al., 2012) was linearized overnight at 37°C with 100 U of BstXI in 1 × Buffer 3.1 (New England Biolabs). The 18,284 bp fragment was purified with a Wizard SV gel and PCR clean up kit (Promega) and the concentration was measured. The fork junction was constructed by annealing 15.3 pmol of 160Ld, 91.8 pmol 99Lg, 1530 pmol of fork primer (Table S1) by heating at 94°C for 5 min before slowly cooling. Similarly, the biotinylated blocking duplex was generated by annealing 5.3 pmol of blockingLd and blockingLg by heating at 94°C for 5 min before slowly cooling. 1.5 pmol of the 18,284 bp linear DNA template was ligated to the pre-annealed fork junction and biotinylated blocking duplex in 1 × T4 ligase buffer (New England Biolabs) and 2000 U of T4 ligase (New England Biolabs) overnight at 16°C. The ligated linear forked DNA substrates were purified from excess DNA oligonucleotides by adjusting NaCl to 300 mM and loaded by gravity onto a Sepharose 4B (Sigma; 1 × 25 cm) column, equilibrated in gel filtration buffer (10 mM Tris-HCl pH 8.0, 1 mM EDTA, and 300 mM NaCl). Ligated biotinylated linear DNA substrates eluted as a single peak in the column void volume, fractions under the peak were analyzed by agarose gel electrophoresis. Fractions were pooled and dialysed overnight in 2 L of sterilized TE buffer, concentrated 2-fold in a vacuum concentrator and the concentration measured. Aliquots were stored at −80°C.

Primed linear DNA substrate

The primed linear substrate used to test the polymerase activity of Pol α^Cat (Figure S4A) was constructed as follows (Table S1). To create the 243 nt template, Near 143-mer and Far 100-mer were ligated together by first mixing with Near Far bridge in a 1:1:3 molar ratio in the presence of 50 mM NaCl and 5 mM Trisodium Citrate pH 7.0. Next, the duplex was annealed by heating to 94°C for 5 min before slowly cooling. Following this 1 mM ATP and 4,000 Units of T4 Ligase (New England Biolabs) were added and the reaction was incubated for 16 h at 15°C. The ligated product was purified on an 8% polyacrylamide, 8M Urea denaturing gel using SYBR Safe stain (Invitrogen) to visualize the single-stranded product. The DNA was recovered by crushing the gel and soaking in buffer TE pH 8.0 for 16 h at room temperature. Followed by spinning at 15,000 rpm for 5 min, to recover the supernatant. 30-mer Primer was 5′-end labeled with γ−³²P-ATP by T4 PNK (New England Biolabs) according to manufacturer instructions and purified on an S-200 HR microspin column (GE Healthcare). The radiolabeled primer was annealed to the 243 nt template in a 2:3 ratio.

Ensemble DNA replication assay

To test the polymerase activity of Pol α^{Cat Cat} (Figure S4A), 2 nM of the linear DNA substrate primed with a 5′−³²P-primer (see Primed Linear DNA Substrate, Method Details) was incubated with the indicated polymerase complex in the presence of 25 mM Tris-OAc pH 7.5, 5% (v/v) glycerol, 100 μg/mL BSA, 5 mM TCEP, 10 mM Magnesium Acetate, 50 mM Potassium glutamate, and 0.1 mM EDTA. Replication reaction was initiated with 80 μM each of dTTP, dATP, dCTP, and dGTP and proceeded for 20 min at 30°C. Reaction volumes were 20 μL. Reactions were stopped by mixing with an equal volume of 23 stop buffer containing 80% (w/v) formamide, 8 mM EDTA, and 1% (w/v) SDS and were subsequently run on a 10% polyacrylamide, 8 M Urea denaturing gel for 1 h and 30 min at 125 V. The gel was exposed to a phosphorimaging screen for 16 h and visualized with a Typhoon FLA 9500 scanner (GE Healthcare).

To confirm the absence of DNA polymerase activity in Pol α^Cat (Figure S4B), 2 nM of the linear DNA substrate primed with a 5′−³²P-primer (see Primed linear substrate) was incubated with either Pol α or Pol α^Cat in the presence of 25 mM Tris-OAc pH 7.5, 5% (v/v) glycerol, 100 μg/mL BSA, 5 mM TCEP, 10 mM Mg-OAc, 50 mM potassium glutamate, and 0.1 mM EDTA. Replication reactions were initiated by the addition of 80 μM each of dTTP, dATP, dCTP, and dGTP and allowed to proceed for 20 min at 30°C. Reaction volumes were 20 μL. Reactions were quenched by mixing with an equal volume of 2× stop buffer containing 80% (w/v) formamide, 8 mM EDTA, and 1% (w/v) SDS and were subsequently run on a 10% polyacrylamide, 8 M Urea denaturing gel for 1.5 h at 125 V. The gel was exposed to a phosphorimaging screen for 16 h and visualized with a Typhoon FLA 9500 scanner (GE Healthcare).

Single-molecule DNA replication assays

Flow cell preparation

Flow cells were prepared as described previously (Geertsema et al., 2015; Lewis et al., 2017a). Briefly, a polydimethylsiloxane (Sylgard) lid was placed on top of a PEG-biotin-functionalized microscope slide (24 × 24 mm, Marienfeld) to create a 1-mm-wide and 100-μm-high flow channel (volume 1 μL). Polyethylene tubes (PE-60: 0.76-mm inlet diameter and 1.22-mm outer diameter, Walker Scientific) were inserted to allow for a buffer flow. To help prevent nonspecific interactions of proteins and DNA with the surface, the chamber was blocked with blocking buffer (50 mM Tris-HCl pH 7.6, 50 mM KCl, 2% (v/v) Tween-20). The forked DNA substrates (20 pM) were flowed through the chamber for 20 min at 17 μL/min in the presence of 200 μM Chloroquine (Sigma). The DNA was visualized by flowing in replication buffer (25 mM Tris-HCl, pH 7.6, 10 mM magnesium acetate, 50 mM potassium glutamate, 40 μg/mL BSA, 0.1 mM EDTA, 5 mM dithiothreitol, and 0.0025% (v/v) Tween-20) with 150 nM Sytox Orange (S.O.; Life Technologies).

Replication reaction conditions

Conditions for the pre-assembly replication reactions were carried out in three steps. First, 30 nM CMG was loaded at 10 μL/min in CMG loading buffer (25 mM Tris-HCl, pH 7.6, 10 mM magnesium acetate, 250 mM potassium glutamate, 40 μg/mL BSA, 0.1 mM EDTA, 5 mM dithiothreitol, and 0.0025% (v/v) Tween-20) with 60 nM Mcm10 and 400 μM ATP. Following this, the replisome was assembled by introducing 20 nM Pol ε, 20 nM Pol δ, 20 nM Pol α, 20 nM Ctf4, 20 nM PCNA, 20 nM RFC, and 30 nM MTC (Mrc1-Tof1-Csm3) in replication buffer supplemented with 400 μM ATP, and 60 μM dCTP/dGTP at 10 μL/min for 5 min. Replication was initiated by washing the flow cell with 100 μL (100 flow cell volumes) at 50 μL/min with a solution containing 60 nM Mcm10, 20 nM PCNA, 20 nM RFC, 200 nM RPA (20 nM when using labeled RPA), and 30 nM MTC in replication buffer supplemented with 5 mM ATP, 125 μM dCTP, dGTP, dATP, and dTTP, and 250 μM CTP, GTP, ATP, and UTP, and 150 nM S.O.

Conditions for replication under the continuous presence of all proteins were performed in multiple stages. First, 30 nM CMG (or LD650–CMG) was loaded at 10 μL/min in CMG loading buffer, with 60 nM Mcm10 and 400 μM ATP. When LD650–CMG was used, the flowcell was subsequently washed under a continuous flow of CMG loading buffer supplemented with 500 mM NaCl at 10 μL/min for 10 min. Then, replication reactions were initiated by introducing 60 nM Mcm10, 20 nM Pol ε (unless specified otherwise), 20 nM Pol δ (unless specified otherwise), 20 nM Pol α (unless specified otherwise), 20 nM Ctf4, 20 nM PCNA, 20 nM or 2nM RFC, 200 nM RPA, and 30 nM MTC in replication buffer supplemented with 5 mM ATP, 125 μM dCTP, dGTP, dATP, and dTTP, and 250 μM CTP, GTP, ATP, and UTP, and 150 nM S.O. In CMG challenge experiments 10 nM CMG was added to the replication reaction.

Imaging conditions

All single-molecule assays were carried out on an inverted microscope (Nikon Eclipse Ti-E) fitted with a CFI Apo TIRF 100 × oil-immersion objective (NA 1.49, Nikon). The temperature was maintained at 31.2°C by an electrically heated chamber (Okolab). dsDNA was visualized every 10 s for 30 min by exciting the S.O. with a 568-nm laser (Coherent, Sapphire 568–200 CW) at 80 mW/cm². The red fluorescently labeled proteins were excited at 80 mW/cm² (800 W/cm² during a FRAP pulse) with a 647-nm laser (Coherent, Obis 647–100 CW). The AF488–Pol ε was visualized with a 488-nm laser (Coherent, Sapphire 488–200 CW) at 140 mW/cm². The signals were spectrally separated using appropriate filter sets (Chroma) and fluorescence signals collected on an Evolve 512 Delta EMCCD (Photometics). Typically, nine fields of view (five for the FRAP experiments) were selected for imaging. Single-molecule experimental results were derived from at least three or four technical replicates for each experimental condition.

Determination of RPA binding footprint

Flow cells were prepared as described above. Oligonucleotides 57-mer or 99Lg (Table S1) were incubated with AF647–RPA in replication buffer for 10 min at 25°C. DNA–RPA complexes were introduced on the surface of the flow cell and washed with 100 μL of replication buffer. The AF647–RPA were excited at 80 mW/cm² with a 647-nm laser (Coherent, Obis 647–100 CW). Imaging was carried out as described in ‘Single-molecule DNA replication assays’.

Measurement of the stoichiometry of fluorescently labeled proteins at the replisome

The average intensity of labeled proteins (Pol α-primases, Pol ε, Pol δ, CMG, or RPA) were quantified by immobilization on the surface of a cleaned microscope coverslip in replication buffer at 6 pM. Imaging was carried out under the same conditions used during the single-molecule replication experiments. We calculated the integrated intensity for every fluorescent protein in a field of view after applying a local background subtraction (Lewis et al., 2017b). The histograms obtained were fit with a Gaussian distribution function using MATLAB 2016b, to give a mean intensity. We calculated the total number of molecules at every time point during DNA replication by dividing their intensities by the intensity of a single molecule. Subsequent histograms were fit to Gaussian distribution using MATLAB 2016b. For Pol ε the stoichiometry was also determined through photobleaching analysis. LD650–Pol ε was excited at 8 W/cm² (100 fold higher than before) to allow for photobleaching during our experiment time. Photobleaching steps were fit using change-point analysis (Duderstadt et al., 2016; Hill et al., 2018). Subsequent histogram was fit to Gaussian distribution using MATLAB 2016b.

Analysis of single-molecule replication kinetics and protein dynamics

All analyses were carried out using ImageJ/Fiji (1.51w) and MATLAB 2016b, and in-house built plugins. The rate of replication of a single molecule was obtained by first tracking the position of the leading-strand spot using the Linear-motion LAP tracker in TrackMate v3.6.0 (Tinevez et al., 2017). Individual rate segments were identified using kinetic change-point analysis (Duderstadt et al., 2016; Hill et al., 2018; Lewis et al., 2017b). The rates obtained from this algorithm were weighted by the DNA segment length, to reflect the number of nucleotides that were synthesized at this rate. This places more significance on the longer rate segments, as they have a higher signal-to-noise ratio compared with shorter segments (Lewis et al., 2017b).

To measure the intensity of the leading-strand spot at the replication fork, we tracked the position of the leading-strand spot and integrated the intensity for all colors simultaneously over time. To obtain the characteristic exchange time τ from the FRAP experiments, the data were fit with a FRAP recovery function correcting for photobleaching (Beattie et al., 2017; Lewis et al., 2017a; Spenkelink et al., 2019) (Equation 1, where a is the amplitude of photobleaching, t_b is the photobleaching time (measured in Figure S7), and I₀ is the number of polymerases at the fork at steady state).

$$I=a{e}^{\frac{t}{t_b}}+l_0\left(1-e^{\frac{t}{\tau }}\right).$$ (Equation 1)

The maximum exchange rate was obtained by fitting the data with a hyperbolic equation (Equation 2, where R is the exchange rate, R_max is the maximum exchange rate, [Pol] is the polymerase concentration, and K_b is the characteristic binding constant).

$$R=\frac{R_{\max }[PoI]}{K_b+[Pol]}.$$ (Equation 2)

QUANTIFICATION AND STATISTICAL ANALYSIS

The number of molecules or events analyzed is indicated in the text or figure legends. Errors reported in this study represent the standard error of the mean (SEM) or the error of the fit, as indicated in the text or figure legends. Every single-molecule replication experiment was carried out at least in triplicate.

DATA AND CODE AVAILABILITY

Raw data is available upon request. All home-built ImageJ plugins used in this study are freely available on the Github repository for Single-molecule/Image analysis tools (https://github.com/SingleMolecule and https://github.com/LMSpenkelink/SingleMoleculeReplication) or available upon request.

KEY RESOURCES TABLE

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Bacterial and Virus Strains
S. cerevisiae strain W303	Brian Chait Lab (Rockefeller University)	N/A
Escherichia coli BL21-DE3 codon plus RIL	Agilent	Cat #230245
Chemicals, Peptides, and Recombinant Proteins
Alexa Fluor 647 NHS ester mono-reactive dye	Life Technologies	Cat# A20006
Alexa Fluor 488 maleimide mono-reactive dye	Life Technologies	Cat# A10254
Cy5 NHS ester mono-reactive dye	GE Healthcare	Cat# PA15101
LD650 maleimide mono-reactive dye	Lumidyne Technologies	Cat# LD650-MAL
Coenzyme A trilithium salt	Sigma-Aldrich	Cat# C3019
d-desthiobiotin	Sigma-Aldrich	Cat# D1411
Chloroquine diphosphate salt	Sigma-Aldrich	Cat# C6628
SYBR-Safe DNA stain	Life Technologies	Cat# S33102
SYTOX Orange nucleic acid stain	Life Technologies	Cat#S11368
γ-32P-ATP	Perkin Elmer	Cat# NEG002A
Polydimethylsiloxane (184 Silicone Elastomer)	Sylgard	Cat# 500GM184KIT
Biotin-PEG and MPEG (MW 5,000) bundle	Laysan Bio	Cat# BIO-PEG-SVA-5K-100MG & MPEG-SVA-5K-1g
3-Aminopropyl)triethoxysilane (APTES)	Alfa Aesar	Cat# A10668
dNTP Bundle	Jena Bioscience	Cat# NU-1005L
ATP solution	Jena Bioscience	Cat# NU-1010
NTP Bundle	Jena Bioscience	Cat# NU-1014L
3 × Flag-tag	EZBiolab	Cat# cp7204
Restriction endonuclease BstXI	New England Biolabs	Cat# R0113L
T4 DNA ligase	New England Biolabs	Cat# M0202L
T4 Polynucleotide Kinase	New England Biolabs	Cat# M0201L
NeutrAvidin protein	Life Technologies	Cat# 31000
PMSF Protease Inhibitor	ThermoFisher	Cat# 36978
CMG helicase	(Langston et al., 2014)	N/A
Pol ε	(Langston et al., 2014)	N/A
RFC clamp loader	(Langston et al., 2014)	N/A
PCNA	(Langston et al., 2014)	N/A
RPA	(Langston et al., 2014)	N/A
Mcm10	(Langston et al., 2017; Lewis et al., 2017b)	N/A
MTC	(Langston et al., 2017; Lewis et al., 2017b)	N/A
Pol α	(Georgescu et al., 2014)	N/A
Pol δ	(Georgescu et al., 2014)	N/A
Ctf4	(Georgescu et al., 2014)	N/A
Pol αCat	This study	N/A
LD650/AF488–Pol ε	This study	N/A
LD650–CMG	(Wasserman et al., 2019)	N/A
AF647–Pol δ32-	This study	N/A
AF647–Pol α	This study	N/A
AF647–RPA	This study	N/A
Oligonucleotides
See Table S1	IDT	N/A
Recombinant DNA
Sfp pet29b C-terminal His Tag	Addgene	Cat# 75015
pSuperCos1	Cees Dekker Lab (TU Delft)	N/A
Software and Algorithms
MATLAB 2016b	MathWorks	https://www.mathworks.com/products/matlab.html
ImageJ/FIJI (1.51w)	(Schindelin et al., 2012)	https://imagej.net/Fiji
TrackMate v3.6.0	(Tinevez et al., 2017)	N/A

Highlights.

• We use fluorescence imaging to visualize individual yeast replisomes in real time

• Polymerase lifetime at the fork is tunable through its concentration in solution

• Pol δ and Pol α are recycled to support the synthesis of multiple Okazaki fragments

• The Pol32 subunit of Pol δ plays a key role in tethering Pol δ to the replisome