CMG helicase and DNA polymerase e form a functional15-subunit holoenzyme for eukaryotic leading-strandDNA replicationLance D. Langstona,b, Dan Zhanga,b, Olga Yurievaa,b, Roxana E. Georgescua,b, Jeff Finkelsteina,b, Nina Y. Yaoa,b,Chiara Indianic, and Mike E. O’Donnella,b,1

aThe Rockefeller University, bHoward Hughes Medical Institute, New York, NY 10065; and cManhattan College, Riverdale, NY 10471

Contributed by Mike E. O’Donnell, September 23, 2014 (sent for review August 13, 2014)

DNA replication in eukaryotes is asymmetric, with separate DNApolymerases (Pol) dedicated to bulk synthesis of the leading andlagging strands. Pol α/primase initiates primers on both strands thatare extended by Pol e on the leading strand and by Pol δ on thelagging strand. The CMG (Cdc45-MCM-GINS) helicase surrounds theleading strand and is proposed to recruit Pol e for leading-strandsynthesis, but to date a direct interaction between CMG and Pol ehas not been demonstrated. While purifying CMG helicase overex-pressed in yeast, we detected a functional complex between CMGand native Pol e. Using pure CMG and Pol e, we reconstituted a sta-ble 15-subunit CMG–Pol e complex and showed that it is a functionalpolymerase–helicase on a model replication fork in vitro. On itsown, the Pol2 catalytic subunit of Pol e is inefficient in CMG-depen-dent replication, but addition of the Dpb2 protein subunit of Pol e,known to bind the Psf1 protein subunit of CMG, allows stable syn-thesiswith CMG.Dpb2does not affect Pol δ functionwith CMG, andthuswe propose that the connection betweenDpb2 and CMGhelpsto stabilize Pol e on the leading strand as part of a 15-subunit lead-ing-strand holoenzyme we refer to as CMGE. Direct binding betweenPol e andCMGprovides an explanation for specific targeting of Pol e tothe leading strand and provides clear mechanistic evidence for howstrand asymmetry is maintained in eukaryotes.

DNA replication | replication fork | helicase | polymerase | CMG

Replisomes are multisubunit protein complexes that co-ordinately unwind duplex DNA and duplicate both parental

strands during chromosomal replication. Detailed studies ofcellular and viral systems show that the basic functional units ofreplication—helicase, primase, and DNA polymerase (Pol)—arecommon to all replisomes whereas the evolutionary histories of theindividual components in different kingdoms are distinctive anddiverse (1). Accordingly, the sequence and structure of replisomecomponents are unrelated, and thus connections and coordinationamong 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 poly-merase to replicate both parental strands, and the action of thesepolymerases is coordinated by a multifunctional clamp loaderthat also connects to the replicative helicase (2). For reasons thatare still unclear, the eukaryotic replisome uses three differentpolymerases for normal chromosome duplication, including onefor the leading strand (Pol e) and two for the lagging strand (Polα/primase and Pol δ) (3–5). Similarly, whereas the replicativehelicase in E. coli is a homohexamer of DnaB, the eukaryotic CMG(Cdc45-MCM-GINS) helicase consists of 11 distinct subunits as-sembled on chromatin by loading of the heterohexameric Mcm2-7helicase core at an origin and its subsequent activation by associ-ation 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 sys-tem reflects the need to restrict chromosome duplication to a sin-gle round in a normal cell cycle so that proper ploidy can bemaintained across multiple chromosomes after cell division.

Detailed biochemical studies of the E. coli replisome showthat the leading and lagging strand replicases are coupled andintimately linked to the replicative helicase, a feature alsocommon to the well-characterized T4 and T7 bacteriophagereplication systems (9–11). For this reason, it has been assumedthat the same would be true of eukaryotic systems, and thisnotion has been strongly reinforced by the identification inyeast of replication progression complexes (RPCs), largemultiprotein complexes containing, among other proteins,CMG, Mcm10, Mrc1, and Ctf4 (12, 13). The RPC also containsPol α/primase under low-salt conditions, suggesting that it is moreweakly bound, and binding of Pol α to the replisome is abolishedin cells lacking Ctf4 or its metazoan counterpart, AND-1 (13–16).Ctf4 binds both the catalytic Pol1 subunit of Pol α and GINS inyeast and thus is thought to tether Pol α to CMG in the replisome(13, 16, 17).Neither Pol δ nor Pol e is found in the most highly purified

RPCs, which are defined by mass spectrometry of proteins boundafter sequential affinity purification of two separate CMG com-ponents from a cell extract (12, 13). However, the noncatalyticDpb2 protein subunit of Pol e is known to bind to the GINScomponent of CMG, and recent evidence suggests that this in-teraction helps maintain Pol e at the replication fork (18, 19). Polδ was shown to bind Pol α via its nonessential Pol32 subunit (20),suggesting that Pol δ might be recruited from solution to extendprimers initiated by Pol α/primase and may only associatetransiently with the core replisome.To study the eukaryotic replisome in detail, we initiated a

long-term project to purify the numerous components of the

Significance

All cells must replicate their chromosomes prior to cell division.This process is carried out by a collection of proteins, known asthe replisome, that act together to unwind the double helixand synthesize two new DNA strands complementary to thetwo parental strands. The details of replisome function havebeen worked out for bacteria but are much less well un-derstood for eukaryotic cells. We have developed a system forstudying eukaryotic replisome function in vitro using purifiedproteins. Using this system, we have identified a direct in-teraction between the component that unwinds the DNA,the CMG (Cdc45-MCM-GINS) helicase, and the componentthat replicates the leading strand, DNA polymerase e, toform a large helicase–polymerase holoenzyme comprising15 separate proteins.

Author contributions: L.D.L. and M.E.O. designed research; L.D.L., D.Z., R.E.G., N.Y.Y., andC.I. performed research; L.D.L., O.Y., R.E.G., and J.F. contributed new reagents/analytictools; L.D.L. and M.E.O. analyzed data; and L.D.L. and M.E.O. wrote the paper.

The authors declare no conflict of interest.

Freely available online through the PNAS open access option.1To whom correspondence should be addressed. Email: odonnel@rockefeller.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1418334111/-/DCSupplemental.

15390–15395 | PNAS | October 28, 2014 | vol. 111 | no. 43 www.pnas.org/cgi/doi/10.1073/pnas.1418334111

RPC/replisome from the model eukaryote Saccharomyces cerevisiae.Pioneering work on Drosophila and human CMG showed that anactive helicase complex could be obtained by coexpression of all 11subunits in insect cells (7, 21) so we cooverexpressed all 11 CMGsubunits in yeast and purified the complex to homogeneity (22). Weshowed that, like its human counterpart, yeast CMG is capable ofcatalyzing replication of a model replication-fork substrate (21, 22).Using this system, we also showed that CMG enforces a preferencefor Pol e over Pol δ in leading-strand replication whereas pro-liferating cell nuclear antigen (PCNA) enforces the opposite pref-erence on the lagging strand (22). Preferential binding of Pol δ toPCNA has been clearly demonstrated and provides an explanationfor the dominance of Pol δ in lagging-strand synthesis (23), but thenature of any interaction between Pol e and CMG on the leadingstrand is poorly understood.While purifying CMG from yeast, we identified a direct in-

teraction between overexpressed CMG and native Pol e to forma multifunctional eukaryotic leading-strand holoenzyme that werefer to as CMGE. Using separately purified CMG and Pol e, wereconstituted a stable, 15-subunit CMGE and showed that it is anactive helicase–polymerase in vitro. We also show that the Dpb2subunit of Pol e, which binds to the Psf1 protein subunit ofGINS, promotes efficient Pol e function with CMG. Directbinding of Pol e to the full CMG complex has not been pre-viously demonstrated, and this interaction provides a mechanis-tic foundation for preferential replication of the leading strandby Pol e as part of a stable helicase–polymerase holoenzyme (4).

ResultsPurified Yeast CMG Is an Active Helicase. As part of our efforts toreconstitute an active eukaryotic replisome from separately pu-rified components, we cooverexpressed all 11 subunits of theS. cerevisiae CMG helicase in yeast cells and purified the full CMGcomplex using separate tags on Mcm5 and Sld5 or Cdc45 andMcm5, similar to the approach used to purify the Drosophila andhuman CMG complexes from insect cells (7, 21, 22). Gel filtra-tion of the resulting CMG produced a highly pure complex witha peak at the expected position for an ∼785-kDa complex (Fig.1A). The ability of CMG to unwind a model Y-substrate wastested, and the peak of unwinding activity comigrated with thepeak of CMG (Fig. 1B, fraction 23), confirming that CMG formsa functional helicase complex as previously observed (7, 21).

CMG Helicase Is Active on Large but Not Small Circular DNA Substrates.Previous work in bacteria and phages has shown that rolling-circlereplication is an ideal system for studying the action of replisomes,and indeed a previous report demonstrated rolling-circle replica-tion supported by human CMG and Pol e (21). Earlier studiesshowed that the Drosophila and human CMG complexes can un-wind a short 5′-tailed oligo annealed to 7.3 kb circular bacterio-phage M13mp18 ssDNA and that yeast CMG also has thisproperty, as shown in Fig. 1C. The substrates used for rolling-circlereplication are much smaller thanM13, typically in the 100- to 200-nt range, and indeed a 200-nt circle was used for leading-strandreplication in the human system although CMG-unwinding activitywas not directly demonstrated on this substrate (21).To determine whether yeast CMG can load efficiently and pro-

ductively onto small circular DNAs, we annealed a 70-nt oligo witha radiolabeled 5′ dC30 tail to a 200-nt circle similar to that used forreplication in the human system (21) and monitored unwinding ofthe labeled oligo by CMG during a 40-min time course. In contrastto theM13 circle, we observed no unwinding byCMGon the 200-ntcircle (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 thecircular forks by annealing the radiolabeled 5′ dC30–tailed 70-merto a linear oligo used to construct the small circular substrates. Asshown in Fig. S1B, this linear substrate is progressively unwound byCMG over the course of 40 min. Taken together, the simplest ex-planation of these results is that yeastCMGloads poorly onto smallcircular substrates, at least under the conditions used in our assays.

CMG Functions in Leading-Strand Replication with Pol e. The inabilityofCMGtounwind small circular substrates led us to construct a linearforked DNA substrate for replication, as reported previously (22).Leading-strand replication on a forked linearDNA can bemonitoredby following extensionof a radiolabeledprimer annealed to the forkedsubstrate, and the function of CMG with Pol e over a 60-min timecourse is shown in Fig. 2. As shown in the scheme at the top, CMG ispreincubated with the 2.8-kb forked linear DNA substrate in the ab-sence of ATP for 10 min to permit CMG loading, replication is initi-ated by adding Pol e, 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 waspreviously shown to inhibit CMG loading in vitro (22). Controlexperiments showed that neither Pol e nor CMG on its own was ca-pable of leading-strand synthesis on this substrate (Fig. S2A) whereasCMG and Pol e cooperatively extended the leading strand to fulllength (Fig. 2 and Fig. S2A). CMGdid not stimulate synthesis by Pol eona singlyprimed ssDNAtemplate (Fig. S3) sowe inferred thatCMG

16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32

fxn # 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32

No

CM

G

Boi

led

0%

10%

20%

30%

16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32% s

ubst

rate

unw

ound

* dT60

dT39

dN30

dN31

Superose 6 fraction number

Thyroglobulin670 kDa

Mcm 2-7Cdc45 -

Sld5 -

Psf1/2 -Psf3 -

A

B

M137.3 kb

*

CdN39

dT80*200-mer

circle

dC30dN70dT25

*

No

CM

G

Boi

l

*

+ CMG

% unwound 0% 3% 14% 36%

+ CMG

% unwound 0% 0% 0%

D

*

Fig. 1. Gel-filtration analysis of purified CMG. The 11-subunit CMG overex-pressed in yeastwas purified through twoaffinity columns, and 300 μL of purifiedprotein was injected onto a Superose 6 column. (A) Indicated fractions from theSuperose 6 column elution were separated on a 10% SDS/PAGE gel and stainedwithCoomassieBlue. (B) Twomicroliters of eachCMGfraction fromAwasassayedfor helicase activity in a 12-μL reaction using a 5′-radiolabeled fork substrate (seeschematic at Left) for 20min at 30 °C in the presence of 2mMATP. Products wereseparatedona10%nativePAGEgelandsubjectedtophosphorimagery (Top). Thechart below shows the percent of substrate unwound by each indicated CMGfractionasdeterminedbyquantitativeanalysisof thephosphorimage. (C)Anoligowas radiolabeled at its 5′ tail and annealed to M13mp18 circular ssDNA (seeschematic atLeft). TheM13substrate (0.5nM)was incubatedwith12nMCMGand1 mM ATP at 30 °C, and 12-μL aliquots were removed at the indicated times andanalyzed as in B. The percentage of the substrate unwound by CMG is indicatedbelow the phosphorimage. (D) A circular DNA of 200 nt was constructed as de-scribed in SI Materials and Methods. An oligo with a 5′ dC30 tail and 70 nt com-plementary 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 CMGloading (21). The substrate was incubated with CMG and ATP as in C, and 12-μLaliquots were stopped at the indicated times.

Langston et al. PNAS | October 28, 2014 | vol. 111 | no. 43 | 15391

BIOCH

EMISTR

Y

was acting through its helicase activity to unwind the fork rather thandirectly stimulating the polymerase activity of Pol e.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 asthe 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 ofATP to form full-length products. This rate was ∼10- to 20-foldslower than the typical estimates for replication-fork progressionin yeast but was comparable to the rate of replication usingS-phase yeast extracts on linear and circular plasmid substrates(26–28). We presume that additional factors or posttranslationalmodifications not present in these reactions may increase thespeed of the minimal replisome in vitro. We also note that thepresent reactions do not include Pol α/primase or Pol δ, the po-lymerases responsible for lagging-strand synthesis. In the absenceof lagging-strand synthesis, ssDNA accumulates on the laggingstrand, and this excess ssDNA may interfere with helicase pro-gression, perhaps by binding to Cdc45 as part of the replicationstress response as suggested (29). Indeed, we observed thatleading-strand replication by CMG-Pol e on our model forksubstrates was modestly stimulated by RPA (22) and by E. colisingle-strand DNA binding protein (SSB) (Fig. S4), both ofwhich bind to ssDNA and perhaps alleviate the inhibition ofhelicase progression.

Identification of a Functional Complex Between Overexpressed CMGand Native Pol e. Having demonstrated that yeast CMG is a func-tional helicase that can cooperatively unwind and synthesize the

leading strand with Pol e, we sought to understand the underlyingmechanism whereby CMG exhibits a preference for Pol e over Polδ in leading-strand synthesis as demonstrated in both the yeast andhuman systems (21, 22). While purifying the CMG complex, wefrequently observed two unidentified proteins coeluting with CMGthrough two affinity columns (Fig. 3A). These proteins were notobserved in preparations of overexpressed Mcm2-7 under similarconditions and thus are specific to CMG (Fig. S5). The copurifyingproteins could be removed by extensive washing or further puri-fication (Fig. 1A), but their persistence through two purificationsteps under a variety of conditions was noteworthy. Furthermore,one of these additional proteins was much larger than any of theCMG subunits, suggesting that it might be a native yeast proteincopurifying with CMG. To investigate the identity of these pro-teins, we analyzed the gel in Fig. 3A by mass spectrometry andidentified the two copurifying proteins as Pol2 and Dpb2, the twoessential subunits of the Pol e complex (30, 31). Although we an-alyzed only the visible Pol 2 andDpb2 bands, we presumed that thesmall Dpb3/4 subunits of Pol e were also present. Identification ofa stable CMG–Pol e complex was unexpected given that Pol e wasnot previously detected in replication progression complexes afterdouble-affinity purification (12, 13).To determine whether this fortuitously purified CMG-Pol e

complex was functional, we used the linear replication fork assaydescribed in Fig. 2 to monitor leading-strand replication byCMG–Pol e. Because CMG-bound Pol e has a 3′–5′ proofread-ing exonuclease activity, we added two dNTPs during the initialCMG-loading step to prevent Pol e from completely digestingthe primer or extending it. As shown in Fig. 3B, the CMG–Pol ecomplex purified directly from yeast extended the radiolabeled

1.5 kb -

1.0 kb -

0.5 kb -

Lane # 1 2 3 4 5 6 7 8

Starting substrate

2.8 kb full length extension

CMG + 2.8 kb forked linear DNA

Pol ++ RFC/PCNA

+ dNTPs + ATP STOP

at tRPA

1) LOAD CMG

2) INITIATE DNA SYNTHESIS

3) START DNA UNWINDING

* * *

Fig. 2. Time course of leading-strand replication by CMG plus Pol e. Thereaction scheme is summarized above the gel. In a 180-μL reaction, 25nM CMG was incubated 10 min at 30 °C with 1 nM 2.8 kb forked DNAsubstrate with a 5′-radiolabeled leading-strand primer followed byaddition of 6 nM RFC, 20 nM PCNA, and 15 nM Pol e and 60 μM dNTPs.After 5 min, the reaction was started with 4 mM ATP followed 1 minlater by addition of 350 nM RPA. The 20-μL aliquots of the reaction werestopped at the indicated times after addition of ATP and analyzed ina denaturing polyacrylamide gel. Positions of radiolabeled molecularweight reference bands are shown to the left of the gel and to the rightof the gel are lane profiles of the indicated time points generated byImageQuant analysis of the scanned phosphorimage. The percentage ofprimers extended past the forked junction at each time point are shownin Fig. S2B.

Sld5

Cdc45

Psf1Psf2

Psf3

Mcm5Mcm4/7Mcm2/3

Mcm6

Pol2

Dpb2

A

1.5 kb

1.0 kb

0.5 kb

Complete leading strand

extension

Starting substrate

B

*

**

CMG-Pol + 2.8 kb DNA + dCTP/dGTP

+ ATP

STOP at t

SSB

+ RFC/PCNAa dATP/dTTP

Fig. 3. Identification of a Functional CMG–Pol e complex. (A) Native Pol ecopurifies with overexpressed CMG. CMG was purified through two affinitycolumns (FLAG-Mcm5 and GST-Sld5), and the GST tag was cleaved. A sampleof purified protein was separated by SDS/PAGE and stained with CoomassieBlue. Indicated bands were analyzed by mass spectrometry, and the protein(s)identified for each band is shown to the left (CMG) and right (Pol e subunits)of the gel. The bands labeled * and ** are the Prescission Protease and thecleaved GST tag, respectively. Copurification of functional native Pol e wasobserved in nine separate purifications of CMG. (B) Leading-strand replica-tion by 30 nM CMG with copurifying Pol e. The scheme of the leading strandreplication 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 shownto the left of the gel (Right) and lane scans of the indicated time points areshown to the right of the gel.

15392 | www.pnas.org/cgi/doi/10.1073/pnas.1418334111 Langston et al.

leading-strand primer to full length after 30 min with some com-plexes reaching full length after only 10 min. Examination of thelane profiles to the right of the gel in Fig. 3B shows that leading-strand synthesis by the directly purified CMG–Pol e proceededwith very similar kinetics to the reaction in Fig. 2 using separatelypurified CMG and Pol e. These data provide strong evidence fora leading-strand CMG–Pol e holoenzyme we refer to as CMGE.

Reconstitution of CMGE, a Multifunctional CMG–Pole Holoenzyme. Tofurther investigate complex formation between CMG and Pol e,we analyzed the migration of Pol e during ultracentrifugation ina glycerol gradient in the presence and absence of CMG. The four-subunit Pol e complex is ∼380 kDa and CMG is ∼785 kDa so, ifPol e binds to CMG, then the reconstituted complex should mi-grate as a much larger species, ∼1.15 mDa, causing a clear shift inthe migration of Pol e compared with the no-CMG control. Asshown in Fig. 4, Pol e alonemigrated as a single species with a peakaround fraction 32 (middle gel).Whenmixed with CMG, however,a portion of Pol e shifted its migration to a much higher molecularweight, peaking at fraction 14, whereas excess Pol e (not bound toCMG) migrated at the position expected for Pol e alone (Fig. 4,top gel). These results demonstrate the formation of a specificcomplex between Pol e andCMGwith no other protein required tomediate the interaction. They also confirm, as shown in the ex-panded view of CMGE fraction 14 at the right of Fig. 4, that theDpb3 and Dpb4 subunits of Pol e are incorporated into a CMGEholoenzyme comprising 15 separate subunits. Densitometricanalysis of Pol2/Dpb2 and Mcm2-7/Cdc45 from the CMGE com-plex peak (fractions 12–16 in Fig. 4) indicated that Pol e and CMGwere present in equimolar amounts in the CMGE complex(Fig. S6A).To determine whether the reconstituted CMGE was func-

tional for leading-strand replication, we used an assay similar tothat described in Fig. 2. to test the peak fraction of CMGE froma glycerol gradient. In the absence of added Pol e, 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 holoenzymewith both unwinding and DNA polymerase functions.

The Dpb2 Subunit of Pol e Is Required for Efficient Leading-StrandReplication with CMG. The Dpb2 subunit of Pol e binds to both thecatalytic subunit Pol2 and to the Psf1 subunit of GINS and,whereas this connection has been shown to be required to recruitPol e to the replication fork, it was not clear whether it is essentialfor ongoing Pol e function with CMG during elongation (19, 30,32, 33). Indeed, a three-subunit form of Pol e lacking Dpb2 wasshown to be a fully functional and processive polymerase onsingly primed circular substrates so Dpb2 is not required for thepolymerase activity of Pol e (34, 35). To investigate the role ofDpb2 in Pol e function with CMG at a replication fork, we sep-arately purified Dpb2 and the Pol2 catalytic subunit (Fig. S7) andexamined Pol2 activity in leading-strand replication assays withCMG in the presence and absence of Dpb2 (30, 36). As shown inFig. 5, in the absence of Dpb2, Pol2 gave a weak full-length signalafter 30 min (lane 2). Upon addition of increasing amounts ofDpb2, however, the ability of Pol2 to synthesize the full 2.8-kbleading strand was restored (Fig. 5, compare lanes 4, 6, and 8 withlane 2; also see plots above the gel), suggesting that the con-nection between Pol2 and CMG via Dpb2 is important for effi-cient Pol e function in leading-strand replication with CMG.The foregoing experiments do not exclude the possibility that

Dpb2 is a general stimulatory factor for CMG so, to determinewhether this effect is specific to Pol2/Pol e, we added Dpb2 toleading-strand replication assays containing Pol δ/CMG. Ourprevious results showed that Pol δ is not able to replicate thesubstrate to full length with CMG under any conditions tested(22), and indeed addition of Dpb2 had no effect on Pol δ functionwith CMG (Fig. S8). Taken together, these results strongly suggestthat Dpb2 is a specific factor linking Pol e to CMG in the leading-strand CMGE holoenzyme and that this connection is important

sedimentation

- Psf1/2

CMGEFraction14

- Pol2

- Dpb2

- Dpb3

- Cdc45

- Sld5

- Mcm6- Mcm3- Mcm2/4- Mcm7- Mcm5

- Dpb4

- Psf3

670 kDa 158 kDa 50 kDa

2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34 36 38 40 42 44 46 48 50 52 fxn#

- Pol2

- Dpb3- Dpb4

\ Cdc45

- Sld5

Mcm2-7

- Psf1/2- Psf3

- Dpb2

- Pol2

- Dpb3- Dpb4

- Dpb2

- Cdc45Mcm2-7

Fig. 4. Reconstitution of a Stable CMG–Pol e Complex. Glycerol gradientelution profiles of CMG alone (Bottom), Pol e alone (Middle), and a mixtureof Pol e and CMG (Top). Fractions were collected from the bottom of thecentrifuge tubes so that the largest molecular weight species were collectedfirst, and a sample of each indicated fraction was separated by SDS/PAGEand stained with Coomassie Blue. The elution peak of each complex (asdetermined 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 migrationpeaks of each standard are indicated at the top of the gels. (Right) An ex-panded view of CMGE Fraction 14 showing the presence of all 15 CMGEsubunits. The migration of each subunit is indicated to the right of the gels.Reconstitution of the CMG–Pol e complex was demonstrated by glycerolgradient a total of six times.

0

100

200

300

400

[Dpb2] nM 0 5 12.5 25

0.5 kb 1.0 1.5 2.0 3.0 kb

3.0 kb

2.0 kb

1.5 kb

1.0 kb

0.5 kb

Pol

Lane 1 2 3 4 5 6 7 8 9 10

Rel

ativ

e in

tens

ity

CMG + DNA

Pol2 + Dpb2+ RFC/PCNA + dNTPs

+ ATP

RPA

5 nM Pol2

[Dpb2]

12.5 nM25 nM

5 nM

0 nM

STOP at t

Fig. 5. Dpb2 links Pol2 to CMG for CMGE function. Leading-strand repli-cation reactions are similar to Fig. 2 but using either Pol2 or Pol e. As shownin the reaction scheme (Top), CMG was mixed with 1.0 nM DNA templateseparately from 5 nM Pol2 and Dpb2 (lanes 1–8) or 20 nM Pol e (lanes 9 and10). Both mixtures were preincubated for 10 min and then combined for5 min. The reaction was started with ATP, stopped at the indicated times,and processed as in Fig. 2. (Middle) The graph shows ImageQuant lane scanscorresponding to 30 min reaction time at different Dpb2 concentrationswith 5 nM Pol2.

Langston et al. PNAS | October 28, 2014 | vol. 111 | no. 43 | 15393

BIOCH

EMISTR

Y

not just for recruitment of Pol e to origins but also for elongationduring S-phase.

DiscussionPol e is recruited to prereplicative complexes along with GINS aspart of a complex that is essential for CMG formation, and sub-sequent activation of CMG is the defining step in replicationinitiation (32, 33, 37). The Dpb2 subunit of Pol e also binds toGINS, and this connection is essential for Pol e recruitment toorigins (32), but the mechanism by which Pol e is specified forongoing leading-strand synthesis is poorly understood. We haveidentified a stable complex between CMG helicase and Pol e toform a 15-subunit holoenzyme, CMGE, that is functional forleading-strand replication in vitro (Figs. 3 and 4). CMGE con-currently unwinds and synthesizes DNA on the leading strand,and the tight connection between helicase and polymeraseexplains how Pol e is specified for leading-strand replication overPol δ and other polymerases both in vivo and in vitro (4, 21, 22).The persistence of a functional CMG–Pol e complex during

purification through two highly selective affinity columns underhigh ionic strength conditions (Fig. 3) was unexpected for severalreasons. Most significantly, Pol e does not copurify with isolatedyeast replication progression complexes (RPCs) after two affinitycolumns (12, 13). Indeed, Pol α can be isolated with purifiedRPCs provided low ionic strength is used, yet Pol e is not ob-served even in these mild isolation procedures (13). Furthermore,mixing of S-phase yeast extracts containing tagged and untaggedversions of Dpb2 followed by immunoprecipitation of CMG ledto the conclusion that Pol e binding at the replication fork washighly dynamic, implying that Pol e does not stably associate withCMG (32). Finally, the C-terminal domain of the Psf1 subunit ofGINS binds to Pol e, yet a study of Drosophila CMG indicatedthat the C-terminal domain of Psf1 is essential for CMG forma-tion and was putatively assigned to interaction with Mcm5 andpotentially unavailable to bind other proteins (8).Detailed time courses of CMG-dependent leading-strand rep-

lication by Pol e show that full-length 2.8-kb products are observedafter 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 e absolutelyrequires CMG (Fig. S2A) and is stimulated by RPA, yet loading ofCMG onto DNA is inhibited by RPA (22). This inhibition ofhelicase loading is reminiscent of the E. coli system, whereloading of the replicative helicase DnaB is strongly inhibited bySSB (38). Thus, if CMG were to fall off the DNA, it would beunlikely to rebind in the presence of RPA so we propose that CMGis loaded onto the forked DNA substrate before addition of RPAand remains bound until replication is complete.The connection between Pol e and CMG via Dpb2 contributes

to the stability of the binary complex during replication (Fig. 5),but, given that both Pol e and CMG bind to DNA, the contribu-tion of the individual complexes to the stability of the holoenzymeduring active replication remains to be determined. Although Polδ does not bind DNA in the absence of the PCNA clamp, Pol ebinds DNA on its own with a preference for primed template (23,39, 40). It will be interesting to investigate whether DNA bindingby Pol e helps to stabilize CMG on DNA, in which case the sta-bility of the active CMGE holoenzyme may be greater than that ofits individual components.Studies of leading-strand replication by human Pol e showed

fivefold stimulation by addition of CMG, but, unlike the yeastsystem, replication did not absolutely require CMG (21). Further-more, although RPA stimulated the processivity of unwinding byhuman CMG, it did not support CMG-dependent leading-strandreplication on a rolling circle by human Pol e, which was dependentinstead on E. coli SSB (21). We did not observe helicase activity byyeast CMG on rolling-circle substrates (Fig. 1D), and we also showthat leading-strand replication by CMG–Pol e on linear fork sub-strates occurs in the absence of single strand-binding proteins andis modestly stimulated by both RPA and SSB (Fig. S4) (22). It isunclear whether these differences reflect a true divergence betweenthe yeast and human systems or simple discrepancies between assaysystems and conditions.We do not yet know the processivity of Pol e in our reactions,

but the connection between Pol e and CMG may tether thepolymerase to the replisome so that, even if it falls off the primerterminus, it can quickly rebind and continue synthesis. One po-tential drawback to a stable connection between Pol e and CMGis the need for dynamic exchange between high-fidelity and low-fidelity polymerases to overcome template blocks during forkprogression (41). Temperature-sensitive mutants of the Dpb2subunit of Pol e that no longer bind stably to the Pol2 catalyticsubunit confer a strong mutator phenotype that is partially de-pendent on Pol ζ, suggesting that low-fidelity polymerases havegreater access to the leading strand when the CMG-Dpb2-Pol2connection is disrupted (34, 42, 43). Together, these data suggesta model whereby the Pol e–CMG connection is flexible, enablingtranslesion and other polymerases to bind the DNA temporarilywithout displacing Pol e from the replisome. As illustrated in Fig.6, we propose that the proximity of Pol e tethered to the fork byCMG serves to limit the action of the low-fidelity polymeraseand enable rapid recovery of the primer terminus by the high-fidelity Pol e as soon as the lesion is bypassed. This hypothesisand the dynamics of processes that connect the leading-strandapparatus to lagging-strand enzymes remain exciting avenues forfuture exploration.

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

CMG Expression and Purification. Yeast strains expressing all 11 CMG subunitswere induced with galactose, harvested, and frozen as pellets in liquid ni-trogen. Extracts made from grinding frozen pellets were purified on twosuccessive affinity columns (22) and further purified on Superose 6where indicated. All oligonucleotides are given in Table S1.

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

Leading strand lesion

TLS Pol

Pol

(iii) Lesion bypass

(i) Pol encounters lesion

(ii) Polymerase switch

(iv) Pol resumes

Mcm2-7

PCNA

GINS

Fig. 6. Hypothesis of Pol e retention at a fork by binding CMG. Binding of Pol e(green) to GINS (purple) in CMGmay help retain Pol e at the fork upon dissoci-ation from the primer terminus. One implication of this action is to facilitatepolymerase switching at template lesions, illustrated in the example shown: (i)CMGE approaches a lesion (red octagon) in the leading strand template and (ii)Pol e vacates PCNA (red) and the primer terminus upon encountering the lesionbut remains bound to CMG, allowing access to a translesion synthesis (TLS)polymerase (pink). (iii) The TLS Pol(s) bypasses the lesion whereupon (iv) Pol erebinds the primer terminus and resumes high-fidelity replication.

15394 | www.pnas.org/cgi/doi/10.1073/pnas.1418334111 Langston et al.

Replication Assays. Reactions contained 1.0 nM pUC19 primed fork with a5′-radiolabeled leading-strand primer annealed to the fork. Reaction vol-umes, ATP added, proteins added, and their amounts are indicated in thefigure legends. Unless otherwise noted, CMG was first incubated at 30 °Cwith the substrate for 10 min. Pol e was added along with RFC/PCNA and 60μM dNTPs and incubated a further 5 min to extend the primer into a shortssDNA gap at the fork where CMG is expected to load. ATP was then addedto initiate unwinding by CMG, and SSB or RPA was added last. At the in-dicated times after addition of ATP, aliquots of the reaction were stoppedand separated on an alkaline agarose gel. Exceptions to this protocol areindicated in the figure legends.

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

ACKNOWLEDGMENTS. We thank the Rockefeller Proteomics ResourceCenter for mass spectrometry. We are grateful for support from NIH GrantGM38839 and the Howard Hughes Medical Institute.

1. Forterre P (2013) Why are there so many diverse replication machineries? J Mol Biol425(23):4714–4726.

2. O’Donnell M (2006) Replisome architecture and dynamics in Escherichia coli. J BiolChem 281(16):10653–10656.

3. Garg P, Burgers PM (2005) DNA polymerases that propagate the eukaryotic DNAreplication fork. Crit Rev Biochem Mol Biol 40(2):115–128.

4. Pursell ZF, Isoz I, Lundström EB, Johansson E, Kunkel TA (2007) Yeast DNA polymeraseepsilon participates in leading-strand DNA replication. Science 317(5834):127–130.

5. Nick McElhinny SA, Gordenin DA, Stith CM, Burgers PM, Kunkel TA (2008) Division of

labor at the eukaryotic replication fork. Mol Cell 30(2):137–144.6. Moyer SE, Lewis PW, Botchan MR (2006) Isolation of the Cdc45/Mcm2-7/GINS (CMG)

complex, a candidate for the eukaryotic DNA replication fork helicase. Proc Natl AcadSci USA 103(27):10236–10241.

7. Ilves I, Petojevic T, Pesavento JJ, Botchan MR (2010) Activation of the MCM2-7 helicase

by association with Cdc45 and GINS proteins. Mol Cell 37(2):247–258.8. Costa A, et al. (2011) The structural basis for MCM2-7 helicase activation by GINS and

Cdc45. Nat Struct Mol Biol 18(4):471–477.9. Patel SS, Pandey M, Nandakumar D (2011) Dynamic coupling between the motors of

DNA replication: Hexameric helicase, DNA polymerase, and primase. Curr Opin Chem

Biol 15(5):595–605.10. Manosas M, Spiering MM, Ding F, Croquette V, Benkovic SJ (2012) Collaborative

coupling between polymerase and helicase for leading-strand synthesis. Nucleic AcidsRes 40(13):6187–6198.

11. Kim S, Dallmann HG, McHenry CS, Marians KJ (1996) Coupling of a replicative poly-

merase and helicase: A tau-DnaB interaction mediates rapid replication fork move-ment. Cell 84(4):643–650.

12. Gambus A, et al. (2006) GINS maintains association of Cdc45 with MCM in replisomeprogression complexes at eukaryotic DNA replication forks. Nat Cell Biol 8(4):358–366.

13. Gambus A, et al. (2009) A key role for Ctf4 in coupling the MCM2-7 helicase to DNA

polymerase alpha within the eukaryotic replisome. EMBO J 28(19):2992–3004.14. Bermudez VP, Farina A, Tappin I, Hurwitz J (2010) Influence of the human cohesion

establishment factor Ctf4/AND-1 on DNA replication. J Biol Chem 285(13):9493–9505.15. Zhu W, et al. (2007) Mcm10 and And-1/CTF4 recruit DNA polymerase alpha to chro-

matin for initiation of DNA replication. Genes Dev 21(18):2288–2299.16. Tanaka H, et al. (2009) Ctf4 coordinates the progression of helicase and DNA poly-

merase alpha. Genes Cells 14(7):807–820.17. Miles J, Formosa T (1992) Protein affinity chromatography with purified yeast DNA

polymerase alpha detects proteins that bind to DNA polymerase. Proc Natl Acad SciUSA 89(4):1276–1280.

18. Sengupta S, van Deursen F, de Piccoli G, Labib K (2013) Dpb2 integrates the leading-strand DNA polymerase into the eukaryotic replisome. Curr Biol 23(7):543–552.

19. Takayama Y, et al. (2003) GINS, a novel multiprotein complex required for chromo-somal DNA replication in budding yeast. Genes Dev 17(9):1153–1165.

20. Johansson E, Garg P, Burgers PM (2004) The Pol32 subunit of DNA polymerase deltacontains separable domains for processive replication and proliferating cell nuclearantigen (PCNA) binding. J Biol Chem 279(3):1907–1915.

21. Kang YH, Galal WC, Farina A, Tappin I, Hurwitz J (2012) Properties of the humanCdc45/Mcm2-7/GINS helicase complex and its action with DNA polymerase epsilon inrolling circle DNA synthesis. Proc Natl Acad Sci USA 109(16):6042–6047.

22. Georgescu RE, et al. (2014) Mechanism of asymmetric polymerase assembly at theeukaryotic replication fork. Nat Struct Mol Biol 21(8):664–670.

23. Chilkova O, et al. (2007) The eukaryotic leading and lagging strand DNA polymerasesare loaded onto primer-ends via separate mechanisms but have comparable proc-essivity in the presence of PCNA. Nucleic Acids Res 35(19):6588–6597.

24. Raghuraman MK, et al. (2001) Replication dynamics of the yeast genome. Science294(5540):115–121.

25. Sekedat MD, et al. (2010) GINS motion reveals replication fork progression is re-markably uniform throughout the yeast genome. Mol Syst Biol 6:353.

26. Heller RC, et al. (2011) Eukaryotic origin-dependent DNA replication in vitro revealssequential action of DDK and S-CDK kinases. Cell 146(1):80–91.

27. On KF, et al. (2014) Prereplicative complexes assembled in vitro support origin-dependent and independent DNA replication. EMBO J 33(6):605–620.

28. Gros J, Devbhandari S, Remus D (2014) Origin plasticity during budding yeast DNAreplication in vitro. EMBO J 33(6):621–636.

29. Bruck I, Kaplan DL (2013) Cdc45-ssDNA interaction is important for stalling the heli-case during replication stress. J Biol Chem 288(11):7550–7563.

30. Morrison A, Araki H, Clark AB, Hamatake RK, Sugino A (1990) A third essential DNApolymerase in S. cerevisiae. Cell 62(6):1143–1151.

31. Araki H, Hamatake RK, Johnston LH, Sugino A (1991) DPB2, the gene encoding DNApolymerase II subunit B, is required for chromosome replication in Saccharomycescerevisiae. Proc Natl Acad Sci USA 88(11):4601–4605.

32. Sengupta S, van Deursen F, de Piccoli G, Labib K (2013) Dpb2 integrates the leading-strand DNA polymerase into the eukaryotic replisome. Curr Biol 23(7):543–552.

33. Muramatsu S, Hirai K, Tak YS, Kamimura Y, Araki H (2010) CDK-dependent complexformation between replication proteins Dpb11, Sld2, Pol (epsilon, and GINS in bud-ding yeast. Genes Dev 24(6):602–612.

34. Isoz I, Persson U, Volkov K, Johansson E (2012) The C-terminus of Dpb2 is required forinteraction with Pol2 and for cell viability. Nucleic Acids Res 40(22):11545–11553.

35. Bermudez VP, Farina A, Raghavan V, Tappin I, Hurwitz J (2011) Studies on humanDNA polymerase epsilon and GINS complex and their role in DNA replication. J BiolChem 286(33):28963–28977.

36. Hamatake RK, et al. (1990) Purification and characterization of DNA polymerase IIfrom the yeast Saccharomyces cerevisiae. Identification of the catalytic core anda possible holoenzyme form of the enzyme. J Biol Chem 265(7):4072–4083.

37. Li Y, Araki H (2013) Loading and activation of DNA replicative helicases: The key stepof initiation of DNA replication. Genes Cells 18(4):266–277.

38. LeBowitz JH, McMacken R (1986) The Escherichia coli dnaB replication protein isa DNA helicase. J Biol Chem 261(10):4738–4748.

39. Tsubota T, Maki S, Kubota H, Sugino A, Maki H (2003) Double-stranded DNA bindingproperties of Saccharomyces cerevisiae DNA polymerase epsilon and of the Dpb3p-Dpb4p subassembly. Genes Cells 8(11):873–888.

40. Tsubota T, et al. (2006) Double-stranded DNA binding, an unusual property of DNApolymerase epsilon, promotes epigenetic silencing in Saccharomyces cerevisiae. J BiolChem 281(43):32898–32908.

41. Langston LD, Indiani C, O’Donnell M (2009) Whither the replisome: Emerging per-spectives on the dynamic nature of the DNA replication machinery. Cell Cycle 8(17):2686–2691.

42. Jaszczur M, et al. (2008) Dpb2p, a noncatalytic subunit of DNA polymerase epsilon,contributes to the fidelity of DNA replication in Saccharomyces cerevisiae. Genetics178(2):633–647.

43. Kraszewska J, Garbacz M, Jonczyk P, Fijalkowska IJ, Jaszczur M (2012) Defect ofDpb2p, a noncatalytic subunit of DNA polymerase e, promotes error prone replicationof undamaged chromosomal DNA in Saccharomyces cerevisiae. Mutat Res 737(1-2):34–42.

Langston et al. PNAS | October 28, 2014 | vol. 111 | no. 43 | 15395

BIOCH

EMISTR

Y