CMG helicase and DNA polymerase ε form a functional 15-subunit holoenzyme for eukaryotic leading-strand DNA replication

Significance

All cells must replicate their chromosomes prior to cell division. This process is carried out by a collection of proteins, known as the replisome, that act together to unwind the double helix and synthesize two new DNA strands complementary to the two parental strands. The details of replisome function have been worked out for bacteria but are much less well understood for eukaryotic cells. We have developed a system for studying eukaryotic replisome function in vitro using purified proteins. Using this system, we have identified a direct interaction between the component that unwinds the DNA, the CMG (Cdc45-MCM-GINS) helicase, and the component that replicates the leading strand, DNA polymerase ε, to form a large helicase–polymerase holoenzyme comprising 15 separate proteins.

Abstract

DNA replication in eukaryotes is asymmetric, with separate DNA polymerases (Pol) dedicated to bulk synthesis of the leading and lagging strands. Pol α/primase initiates primers on both strands that are extended by Pol ε on the leading strand and by Pol δ on the lagging strand. The CMG (Cdc45-MCM-GINS) helicase surrounds the leading strand and is proposed to recruit Pol ε for leading-strand synthesis, but to date a direct interaction between CMG and Pol ε has not been demonstrated. While purifying CMG helicase overexpressed in yeast, we detected a functional complex between CMG and native Pol ε. Using pure CMG and Pol ε, we reconstituted a stable 15-subunit CMG–Pol ε complex and showed that it is a functional polymerase–helicase on a model replication fork in vitro. On its own, the Pol2 catalytic subunit of Pol ε is inefficient in CMG-dependent replication, but addition of the Dpb2 protein subunit of Pol ε, known to bind the Psf1 protein subunit of CMG, allows stable synthesis with CMG. Dpb2 does not affect Pol δ function with CMG, and thus we propose that the connection between Dpb2 and CMG helps to stabilize Pol ε on the leading strand as part of a 15-subunit leading-strand holoenzyme we refer to as CMGE. Direct binding between Pol ε and CMG provides an explanation for specific targeting of Pol ε to the leading strand and provides clear mechanistic evidence for how strand asymmetry is maintained in eukaryotes.

Replisomes are multisubunit protein complexes that coordinately unwind duplex DNA and duplicate both parental strands during chromosomal replication. Detailed studies of cellular and viral systems show that the basic functional units of replication—helicase, primase, and DNA polymerase (Pol)—are common to all replisomes whereas the evolutionary histories of the individual components in different kingdoms are distinctive and diverse (1). Accordingly, the sequence and structure of replisome components are unrelated, and thus connections and coordination among the different functional units can be expected to vary widely.

The most well-studied cellular replisome to date, bacterial Escherichia coli, uses multiple copies of a single DNA polymerase to replicate both parental strands, and the action of these polymerases is coordinated by a multifunctional clamp loader that also connects to the replicative helicase (2). For reasons that are still unclear, the eukaryotic replisome uses three different polymerases for normal chromosome duplication, including one for the leading strand (Pol ε) and two for the lagging strand (Pol α/primase and Pol δ) (3–5). Similarly, whereas the replicative helicase in E. coli is a homohexamer of DnaB, the eukaryotic CMG (Cdc45-MCM-GINS) helicase consists of 11 distinct subunits assembled on chromatin by loading of the heterohexameric Mcm2-7 helicase core at an origin and its subsequent activation by association with Cdc45 and the heterotetrameric GINS (Sld5-Psf1-Psf2-Psf3) complex at the onset of S-phase to form the CMG complex (6–8). Among other things, the complexity of the eukaryotic system reflects the need to restrict chromosome duplication to a single round in a normal cell cycle so that proper ploidy can be maintained across multiple chromosomes after cell division.

Detailed biochemical studies of the E. coli replisome show that the leading and lagging strand replicases are coupled and intimately linked to the replicative helicase, a feature also common to the well-characterized T4 and T7 bacteriophage replication systems (9–11). For this reason, it has been assumed that the same would be true of eukaryotic systems, and this notion has been strongly reinforced by the identification in yeast of replication progression complexes (RPCs), large multiprotein complexes containing, among other proteins, CMG, Mcm10, Mrc1, and Ctf4 (12, 13). The RPC also contains Pol α/primase under low-salt conditions, suggesting that it is more weakly bound, and binding of Pol α to the replisome is abolished in cells lacking Ctf4 or its metazoan counterpart, AND-1 (13–16). Ctf4 binds both the catalytic Pol1 subunit of Pol α and GINS in yeast and thus is thought to tether Pol α to CMG in the replisome (13, 16, 17).

Neither Pol δ nor Pol ε is found in the most highly purified RPCs, which are defined by mass spectrometry of proteins bound after sequential affinity purification of two separate CMG components from a cell extract (12, 13). However, the noncatalytic Dpb2 protein subunit of Pol ε is known to bind to the GINS component of CMG, and recent evidence suggests that this interaction helps maintain Pol ε at the replication fork (18, 19). Pol δ was shown to bind Pol α via its nonessential Pol32 subunit (20), suggesting that Pol δ might be recruited from solution to extend primers initiated by Pol α/primase and may only associate transiently with the core replisome.

To study the eukaryotic replisome in detail, we initiated a long-term project to purify the numerous components of the RPC/replisome from the model eukaryote Saccharomyces cerevisiae. Pioneering work on Drosophila and human CMG showed that an active helicase complex could be obtained by coexpression of all 11 subunits in insect cells (7, 21) so we cooverexpressed all 11 CMG subunits in yeast and purified the complex to homogeneity (22). We showed that, like its human counterpart, yeast CMG is capable of catalyzing replication of a model replication-fork substrate (21, 22). Using this system, we also showed that CMG enforces a preference for Pol ε over Pol δ in leading-strand replication whereas proliferating cell nuclear antigen (PCNA) enforces the opposite preference on the lagging strand (22). Preferential binding of Pol δ to PCNA has been clearly demonstrated and provides an explanation for the dominance of Pol δ in lagging-strand synthesis (23), but the nature of any interaction between Pol ε and CMG on the leading strand is poorly understood.

While purifying CMG from yeast, we identified a direct interaction between overexpressed CMG and native Pol ε to form a multifunctional eukaryotic leading-strand holoenzyme that we refer to as CMGE. Using separately purified CMG and Pol ε, we reconstituted a stable, 15-subunit CMGE and showed that it is an active helicase–polymerase in vitro. We also show that the Dpb2 subunit of Pol ε, which binds to the Psf1 protein subunit of GINS, promotes efficient Pol ε function with CMG. Direct binding of Pol ε to the full CMG complex has not been previously demonstrated, and this interaction provides a mechanistic foundation for preferential replication of the leading strand by Pol ε as part of a stable helicase–polymerase holoenzyme (4).

Results

Purified Yeast CMG Is an Active Helicase.

As part of our efforts to reconstitute an active eukaryotic replisome from separately purified components, we cooverexpressed all 11 subunits of the S. cerevisiae CMG helicase in yeast cells and purified the full CMG complex using separate tags on Mcm5 and Sld5 or Cdc45 and Mcm5, similar to the approach used to purify the Drosophila and human CMG complexes from insect cells (7, 21, 22). Gel filtration of the resulting CMG produced a highly pure complex with a peak at the expected position for an ∼785-kDa complex (Fig. 1A). The ability of CMG to unwind a model Y-substrate was tested, and the peak of unwinding activity comigrated with the peak of CMG (Fig. 1B, fraction 23), confirming that CMG forms a functional helicase complex as previously observed (7, 21).

Fig. 1.

Gel-filtration analysis of purified CMG. The 11-subunit CMG overexpressed in yeast was purified through two affinity columns, and 300 μL of purified protein was injected onto a Superose 6 column. (A) Indicated fractions from the Superose 6 column elution were separated on a 10% SDS/PAGE gel and stained with Coomassie Blue. (B) Two microliters of each CMG fraction from A was assayed for helicase activity in a 12-μL reaction using a 5′-radiolabeled fork substrate (see schematic at Left) for 20 min at 30 °C in the presence of 2 mM ATP. Products were separated on a 10% native PAGE gel and subjected to phosphorimagery (Top). The chart below shows the percent of substrate unwound by each indicated CMG fraction as determined by quantitative analysis of the phosphorimage. (C) An oligo was radiolabeled at its 5′ tail and annealed to M13mp18 circular ssDNA (see schematic at Left). The M13 substrate (0.5 nM) was incubated with 12 nM CMG and 1 mM ATP at 30 °C, and 12-μL aliquots were removed at the indicated times and analyzed as in B. The percentage of the substrate unwound by CMG is indicated below the phosphorimage. (D) A circular DNA of 200 nt was constructed as described in SI Materials and Methods. An oligo with a 5′ dC₃₀ tail and 70 nt complementary to the circle was radiolabeled at its 5′ end and annealed to the circle, which contains a dT₂₅ region adjacent to the annealed duplex to facilitate CMG loading (21). The substrate was incubated with CMG and ATP as in C, and 12-μL aliquots were stopped at the indicated times.

CMG Helicase Is Active on Large but Not Small Circular DNA Substrates.

Previous work in bacteria and phages has shown that rolling-circle replication is an ideal system for studying the action of replisomes, and indeed a previous report demonstrated rolling-circle replication supported by human CMG and Pol ε (21). Earlier studies showed that the Drosophila and human CMG complexes can unwind a short 5′-tailed oligo annealed to 7.3 kb circular bacteriophage M13mp18 ssDNA and that yeast CMG also has this property, as shown in Fig. 1C. The substrates used for rolling-circle replication are much smaller than M13, typically in the 100- to 200-nt range, and indeed a 200-nt circle was used for leading-strand replication in the human system although CMG-unwinding activity was not directly demonstrated on this substrate (21).

To determine whether yeast CMG can load efficiently and productively onto small circular DNAs, we annealed a 70-nt oligo with a radiolabeled 5′ dC₃₀ tail to a 200-nt circle similar to that used for replication in the human system (21) and monitored unwinding of the labeled oligo by CMG during a 40-min time course. In contrast to the M13 circle, we observed no unwinding by CMG on the 200-nt circle (Fig. 1D) or on a larger 300-nt version of the same circle (Fig. S1A). As a control for unwinding by CMG in this sequence context, we assembled a linear fork substrate with the same sequence as the circular forks by annealing the radiolabeled 5′ dC₃₀–tailed 70-mer to a linear oligo used to construct the small circular substrates. As shown in Fig. S1B, this linear substrate is progressively unwound by CMG over the course of 40 min. Taken together, the simplest explanation of these results is that yeast CMG loads poorly onto small circular substrates, at least under the conditions used in our assays.

CMG Functions in Leading-Strand Replication with Pol ε.

The inability of CMG to unwind small circular substrates led us to construct a linear forked DNA substrate for replication, as reported previously (22). Leading-strand replication on a forked linear DNA can be monitored by following extension of a radiolabeled primer annealed to the forked substrate, and the function of CMG with Pol ε over a 60-min time course is shown in Fig. 2. As shown in the scheme at the top, CMG is preincubated with the 2.8-kb forked linear DNA substrate in the absence of ATP for 10 min to permit CMG loading, replication is initiated by adding Pol ε, replication factor C (RFC), PCNA, and dNTPs, and the reaction is started by adding ATP to fuel the CMG helicase. Replication protein A (RPA) is added after initiation because it was previously shown to inhibit CMG loading in vitro (22). Control experiments showed that neither Pol ε nor CMG on its own was capable of leading-strand synthesis on this substrate (Fig. S2A) whereas CMG and Pol ε cooperatively extended the leading strand to full length (Fig. 2 and Fig. S2A). CMG did not stimulate synthesis by Pol ε on a singly primed ssDNA template (Fig. S3) so we inferred that CMG was acting through its helicase activity to unwind the fork rather than directly stimulating the polymerase activity of Pol ε.

Fig. 2.

Time course of leading-strand replication by CMG plus Pol ε. The reaction scheme is summarized above the gel. In a 180-μL reaction, 25 nM CMG was incubated 10 min at 30 °C with 1 nM 2.8 kb forked DNA substrate with a 5′-radiolabeled leading-strand primer followed by addition of 6 nM RFC, 20 nM PCNA, and 15 nM Pol ε and 60 μM dNTPs. After 5 min, the reaction was started with 4 mM ATP followed 1 min later by addition of 350 nM RPA. The 20-μL aliquots of the reaction were stopped at the indicated times after addition of ATP and analyzed in a denaturing polyacrylamide gel. Positions of radiolabeled molecular weight reference bands are shown to the left of the gel and to the right of the gel are lane profiles of the indicated time points generated by ImageQuant analysis of the scanned phosphorimage. The percentage of primers extended past the forked junction at each time point are shown in Fig. S2B.

As shown in the lane profiles to the right of Fig. 2 and in the graph in Fig. S2B, full-length 2.8-kb products appeared as early as the 7-min time point, indicating that a subpopulation of leading-strand replisomes was capable of replicating at an average rate of ∼400 bp/min, within four- to sevenfold of the in vivo rate (24, 25). However, most replisomes required 20–60 min after addition of ATP to form full-length products. This rate was ∼10- to 20-fold slower than the typical estimates for replication-fork progression in yeast but was comparable to the rate of replication using S-phase yeast extracts on linear and circular plasmid substrates (26–28). We presume that additional factors or posttranslational modifications not present in these reactions may increase the speed of the minimal replisome in vitro. We also note that the present reactions do not include Pol α/primase or Pol δ, the polymerases responsible for lagging-strand synthesis. In the absence of lagging-strand synthesis, ssDNA accumulates on the lagging strand, and this excess ssDNA may interfere with helicase progression, perhaps by binding to Cdc45 as part of the replication stress response as suggested (29). Indeed, we observed that leading-strand replication by CMG-Pol ε on our model fork substrates was modestly stimulated by RPA (22) and by E. coli single-strand DNA binding protein (SSB) (Fig. S4), both of which bind to ssDNA and perhaps alleviate the inhibition of helicase progression.

Identification of a Functional Complex Between Overexpressed CMG and Native Pol ε.

Having demonstrated that yeast CMG is a functional helicase that can cooperatively unwind and synthesize the leading strand with Pol ε, we sought to understand the underlying mechanism whereby CMG exhibits a preference for Pol ε over Pol δ in leading-strand synthesis as demonstrated in both the yeast and human systems (21, 22). While purifying the CMG complex, we frequently observed two unidentified proteins coeluting with CMG through two affinity columns (Fig. 3A). These proteins were not observed in preparations of overexpressed Mcm2-7 under similar conditions and thus are specific to CMG (Fig. S5). The copurifying proteins could be removed by extensive washing or further purification (Fig. 1A), but their persistence through two purification steps under a variety of conditions was noteworthy. Furthermore, one of these additional proteins was much larger than any of the CMG subunits, suggesting that it might be a native yeast protein copurifying with CMG. To investigate the identity of these proteins, we analyzed the gel in Fig. 3A by mass spectrometry and identified the two copurifying proteins as Pol2 and Dpb2, the two essential subunits of the Pol ε complex (30, 31). Although we analyzed only the visible Pol 2 and Dpb2 bands, we presumed that the small Dpb3/4 subunits of Pol ε were also present. Identification of a stable CMG–Pol ε complex was unexpected given that Pol ε was not previously detected in replication progression complexes after double-affinity purification (12, 13).

Fig. 3.

Identification of a Functional CMG–Pol ε complex. (A) Native Pol ε copurifies with overexpressed CMG. CMG was purified through two affinity columns (FLAG-Mcm5 and GST-Sld5), and the GST tag was cleaved. A sample of purified protein was separated by SDS/PAGE and stained with Coomassie Blue. Indicated bands were analyzed by mass spectrometry, and the protein(s) identified for each band is shown to the left (CMG) and right (Pol ε subunits) of the gel. The bands labeled * and ** are the Prescission Protease and the cleaved GST tag, respectively. Copurification of functional native Pol ε was observed in nine separate purifications of CMG. (B) Leading-strand replication by 30 nM CMG with copurifying Pol ε. The scheme of the leading strand replication reaction is in the gray box (Left). Reactions contained 2 nM RFC, 10 nM PCNA, and 5 mM ATP, and 200 nM E. coli SSB was used instead of RPA (21). Positions of radiolabeled molecular weight reference bands are shown to the left of the gel (Right) and lane scans of the indicated time points are shown to the right of the gel.

To determine whether this fortuitously purified CMG-Pol ε complex was functional, we used the linear replication fork assay described in Fig. 2 to monitor leading-strand replication by CMG–Pol ε. Because CMG-bound Pol ε has a 3′–5′ proofreading exonuclease activity, we added two dNTPs during the initial CMG-loading step to prevent Pol ε from completely digesting the primer or extending it. As shown in Fig. 3B, the CMG–Pol ε complex purified directly from yeast extended the radiolabeled leading-strand primer to full length after 30 min with some complexes reaching full length after only 10 min. Examination of the lane profiles to the right of the gel in Fig. 3B shows that leading-strand synthesis by the directly purified CMG–Pol ε proceeded with very similar kinetics to the reaction in Fig. 2 using separately purified CMG and Pol ε. These data provide strong evidence for a leading-strand CMG–Pol ε holoenzyme we refer to as CMGE.

Reconstitution of CMGE, a Multifunctional CMG–Polε Holoenzyme.

To further investigate complex formation between CMG and Pol ε, we analyzed the migration of Pol ε during ultracentrifugation in a glycerol gradient in the presence and absence of CMG. The four-subunit Pol ε complex is ∼380 kDa and CMG is ∼785 kDa so, if Pol ε binds to CMG, then the reconstituted complex should migrate as a much larger species, ∼1.15 mDa, causing a clear shift in the migration of Pol ε compared with the no-CMG control. As shown in Fig. 4, Pol ε alone migrated as a single species with a peak around fraction 32 (middle gel). When mixed with CMG, however, a portion of Pol ε shifted its migration to a much higher molecular weight, peaking at fraction 14, whereas excess Pol ε (not bound to CMG) migrated at the position expected for Pol ε alone (Fig. 4, top gel). These results demonstrate the formation of a specific complex between Pol ε and CMG with no other protein required to mediate the interaction. They also confirm, as shown in the expanded view of CMGE fraction 14 at the right of Fig. 4, that the Dpb3 and Dpb4 subunits of Pol ε are incorporated into a CMGE holoenzyme comprising 15 separate subunits. Densitometric analysis of Pol2/Dpb2 and Mcm2-7/Cdc45 from the CMGE complex peak (fractions 12–16 in Fig. 4) indicated that Pol ε and CMG were present in equimolar amounts in the CMGE complex (Fig. S6A).

Fig. 4.

Reconstitution of a Stable CMG–Pol ε Complex. Glycerol gradient elution profiles of CMG alone (Bottom), Pol ε alone (Middle), and a mixture of Pol ε and CMG (Top). Fractions were collected from the bottom of the centrifuge tubes so that the largest molecular weight species were collected first, and a sample of each indicated fraction was separated by SDS/PAGE and stained with Coomassie Blue. The elution peak of each complex (as determined by densitometry shown in Fig. S6A) is indicated in the gels. Molecular mass size standards (thyroglobulin 670 kDa, γ-globulin 158 kDa, and ovalbumin 44 kDa) were run in a parallel gradient, and the migration peaks of each standard are indicated at the top of the gels. (Right) An expanded view of CMGE Fraction 14 showing the presence of all 15 CMGE subunits. The migration of each subunit is indicated to the right of the gels. Reconstitution of the CMG–Pol ε complex was demonstrated by glycerol gradient a total of six times.

To determine whether the reconstituted CMGE was functional for leading-strand replication, we used an assay similar to that described in Fig. 2. to test the peak fraction of CMGE from a glycerol gradient. In the absence of added Pol ε, reconstituted CMGE was able to unwind the duplex and synthesize the leading strand to full length (Fig. S6B). Together with the data from Fig. 3, these results indicate that CMGE is a functional holoenzyme with both unwinding and DNA polymerase functions.

The Dpb2 Subunit of Pol ε Is Required for Efficient Leading-Strand Replication with CMG.

The Dpb2 subunit of Pol ε binds to both the catalytic subunit Pol2 and to the Psf1 subunit of GINS and, whereas this connection has been shown to be required to recruit Pol ε to the replication fork, it was not clear whether it is essential for ongoing Pol ε function with CMG during elongation (19, 30, 32, 33). Indeed, a three-subunit form of Pol ε lacking Dpb2 was shown to be a fully functional and processive polymerase on singly primed circular substrates so Dpb2 is not required for the polymerase activity of Pol ε (34, 35). To investigate the role of Dpb2 in Pol ε function with CMG at a replication fork, we separately purified Dpb2 and the Pol2 catalytic subunit (Fig. S7) and examined Pol2 activity in leading-strand replication assays with CMG in the presence and absence of Dpb2 (30, 36). As shown in Fig. 5, in the absence of Dpb2, Pol2 gave a weak full-length signal after 30 min (lane 2). Upon addition of increasing amounts of Dpb2, however, the ability of Pol2 to synthesize the full 2.8-kb leading strand was restored (Fig. 5, compare lanes 4, 6, and 8 with lane 2; also see plots above the gel), suggesting that the connection between Pol2 and CMG via Dpb2 is important for efficient Pol ε function in leading-strand replication with CMG.

Fig. 5.

Dpb2 links Pol2 to CMG for CMGE function. Leading-strand replication reactions are similar to Fig. 2 but using either Pol2 or Pol ε. As shown in the reaction scheme (Top), CMG was mixed with 1.0 nM DNA template separately from 5 nM Pol2 and Dpb2 (lanes 1–8) or 20 nM Pol ε (lanes 9 and 10). Both mixtures were preincubated for 10 min and then combined for 5 min. The reaction was started with ATP, stopped at the indicated times, and processed as in Fig. 2. (Middle) The graph shows ImageQuant lane scans corresponding to 30 min reaction time at different Dpb2 concentrations with 5 nM Pol2.

The foregoing experiments do not exclude the possibility that Dpb2 is a general stimulatory factor for CMG so, to determine whether this effect is specific to Pol2/Pol ε, we added Dpb2 to leading-strand replication assays containing Pol δ/CMG. Our previous results showed that Pol δ is not able to replicate the substrate to full length with CMG under any conditions tested (22), and indeed addition of Dpb2 had no effect on Pol δ function with CMG (Fig. S8). Taken together, these results strongly suggest that Dpb2 is a specific factor linking Pol ε to CMG in the leading-strand CMGE holoenzyme and that this connection is important not just for recruitment of Pol ε to origins but also for elongation during S-phase.

Discussion

Pol ε is recruited to prereplicative complexes along with GINS as part of a complex that is essential for CMG formation, and subsequent activation of CMG is the defining step in replication initiation (32, 33, 37). The Dpb2 subunit of Pol ε also binds to GINS, and this connection is essential for Pol ε recruitment to origins (32), but the mechanism by which Pol ε is specified for ongoing leading-strand synthesis is poorly understood. We have identified a stable complex between CMG helicase and Pol ε to form a 15-subunit holoenzyme, CMGE, that is functional for leading-strand replication in vitro (Figs. 3 and 4). CMGE concurrently unwinds and synthesizes DNA on the leading strand, and the tight connection between helicase and polymerase explains how Pol ε is specified for leading-strand replication over Pol δ and other polymerases both in vivo and in vitro (4, 21, 22).

The persistence of a functional CMG–Pol ε complex during purification through two highly selective affinity columns under high ionic strength conditions (Fig. 3) was unexpected for several reasons. Most significantly, Pol ε does not copurify with isolated yeast replication progression complexes (RPCs) after two affinity columns (12, 13). Indeed, Pol α can be isolated with purified RPCs provided low ionic strength is used, yet Pol ε is not observed even in these mild isolation procedures (13). Furthermore, mixing of S-phase yeast extracts containing tagged and untagged versions of Dpb2 followed by immunoprecipitation of CMG led to the conclusion that Pol ε binding at the replication fork was highly dynamic, implying that Pol ε does not stably associate with CMG (32). Finally, the C-terminal domain of the Psf1 subunit of GINS binds to Pol ε, yet a study of Drosophila CMG indicated that the C-terminal domain of Psf1 is essential for CMG formation and was putatively assigned to interaction with Mcm5 and potentially unavailable to bind other proteins (8).

Detailed time courses of CMG-dependent leading-strand replication by Pol ε show that full-length 2.8-kb products are observed after 7 min and accumulate steadily out to 30 min and beyond (Figs. 2 and 3) (22). Along with our previously published results, we show that leading-strand replication by Pol ε absolutely requires CMG (Fig. S2A) and is stimulated by RPA, yet loading of CMG onto DNA is inhibited by RPA (22). This inhibition of helicase loading is reminiscent of the E. coli system, where loading of the replicative helicase DnaB is strongly inhibited by SSB (38). Thus, if CMG were to fall off the DNA, it would be unlikely to rebind in the presence of RPA so we propose that CMG is loaded onto the forked DNA substrate before addition of RPA and remains bound until replication is complete.

The connection between Pol ε and CMG via Dpb2 contributes to the stability of the binary complex during replication (Fig. 5), but, given that both Pol ε and CMG bind to DNA, the contribution of the individual complexes to the stability of the holoenzyme during active replication remains to be determined. Although Pol δ does not bind DNA in the absence of the PCNA clamp, Pol ε binds DNA on its own with a preference for primed template (23, 39, 40). It will be interesting to investigate whether DNA binding by Pol ε helps to stabilize CMG on DNA, in which case the stability of the active CMGE holoenzyme may be greater than that of its individual components.

Studies of leading-strand replication by human Pol ε showed fivefold stimulation by addition of CMG, but, unlike the yeast system, replication did not absolutely require CMG (21). Furthermore, although RPA stimulated the processivity of unwinding by human CMG, it did not support CMG-dependent leading-strand replication on a rolling circle by human Pol ε, which was dependent instead on E. coli SSB (21). We did not observe helicase activity by yeast CMG on rolling-circle substrates (Fig. 1D), and we also show that leading-strand replication by CMG–Pol ε on linear fork substrates occurs in the absence of single strand-binding proteins and is modestly stimulated by both RPA and SSB (Fig. S4) (22). It is unclear whether these differences reflect a true divergence between the yeast and human systems or simple discrepancies between assay systems and conditions.

We do not yet know the processivity of Pol ε in our reactions, but the connection between Pol ε and CMG may tether the polymerase to the replisome so that, even if it falls off the primer terminus, it can quickly rebind and continue synthesis. One potential drawback to a stable connection between Pol ε and CMG is the need for dynamic exchange between high-fidelity and low-fidelity polymerases to overcome template blocks during fork progression (41). Temperature-sensitive mutants of the Dpb2 subunit of Pol ε that no longer bind stably to the Pol2 catalytic subunit confer a strong mutator phenotype that is partially dependent on Pol ζ, suggesting that low-fidelity polymerases have greater access to the leading strand when the CMG-Dpb2-Pol2 connection is disrupted (34, 42, 43). Together, these data suggest a model whereby the Pol ε–CMG connection is flexible, enabling translesion and other polymerases to bind the DNA temporarily without displacing Pol ε from the replisome. As illustrated in Fig. 6, we propose that the proximity of Pol ε tethered to the fork by CMG serves to limit the action of the low-fidelity polymerase and enable rapid recovery of the primer terminus by the high-fidelity Pol ε as soon as the lesion is bypassed. This hypothesis and the dynamics of processes that connect the leading-strand apparatus to lagging-strand enzymes remain exciting avenues for future exploration.

Fig. 6.

Hypothesis of Pol ε retention at a fork by binding CMG. Binding of Pol ε (green) to GINS (purple) in CMG may help retain Pol ε at the fork upon dissociation from the primer terminus. One implication of this action is to facilitate polymerase switching at template lesions, illustrated in the example shown: (i) CMGE approaches a lesion (red octagon) in the leading strand template and (ii) Pol ε vacates PCNA (red) and the primer terminus upon encountering the lesion but remains bound to CMG, allowing access to a translesion synthesis (TLS) polymerase (pink). (iii) The TLS Pol(s) bypasses the lesion whereupon (iv) Pol ε rebinds the primer terminus and resumes high-fidelity replication.

Materials and Methods

Experimental procedures are described in full in SI Materials and Methods.

CMG Expression and Purification.

Yeast strains expressing all 11 CMG subunits were induced with galactose, harvested, and frozen as pellets in liquid nitrogen. Extracts made from grinding frozen pellets were purified on two successive affinity columns (22) and further purified on Superose 6 where indicated. All oligonucleotides are given in Table S1.

Helicase Assays.

Reactions containing 0.5 nM radiolabeled DNA substrate were incubated at 30 °C, and products were separated on 10% (wt/vol) Native PAGE minigels. Amounts of added CMG helicase and incubation conditions are indicated in the figure legends.

Replication Assays.

Reactions contained 1.0 nM pUC19 primed fork with a 5′-radiolabeled leading-strand primer annealed to the fork. Reaction volumes, ATP added, proteins added, and their amounts are indicated in the figure legends. Unless otherwise noted, CMG was first incubated at 30 °C with the substrate for 10 min. Pol ε was added along with RFC/PCNA and 60 μM dNTPs and incubated a further 5 min to extend the primer into a short ssDNA gap at the fork where CMG is expected to load. ATP was then added to initiate unwinding by CMG, and SSB or RPA was added last. At the indicated times after addition of ATP, aliquots of the reaction were stopped and separated on an alkaline agarose gel. Exceptions to this protocol are indicated in the figure legends.

Reconstitution of CMGE.

To examine CMG–Pol ε complex formation, 320 pmol of CMG was mixed with an excess of Pol ε (480 pmol) for 30 min and separated in a 15–35% glycerol gradient at 4 °C for 18 h at 260,000 × g. Fractions (7 drops, ∼170 μL) were collected from the bottom of the centrifuge tubes, and 20-μL samples were analyzed by SDS/PAGE and stained with Coomassie Blue. Identical gradient analyses were performed for CMG and Pol ε alone. Gels were scanned in an ImageQuant LAS4000 (GE Healthcare) and analyzed using ImageQuant TL v2005 software.