Mechanism of replication origin melting nucleated by CMG helicase assembly

Abstract

The activation of eukaryotic origins of replication occurs in temporally separated steps to ensure that chromosomes are copied only once per cell cycle. First, the MCM helicase is loaded onto duplex DNA as an inactive double hexamer. Activation occurs after the recruitment of a set of firing factors that assemble two Cdc45-MCM-GINS (CMG) holo-helicases. CMG formation leads to the underwinding of DNA on the path to the establishment of the replication fork, but whether DNA becomes melted at this stage is unknown¹. Here we use cryo-electron microscopy to image ATP-dependent CMG assembly on a chromatinized origin, reconstituted in vitro with purified yeast proteins. We find that CMG formation disrupts the double hexamer interface and thereby exposes duplex DNA in between the two CMGs. The two helicases remain tethered, which gives rise to a splayed dimer, with implications for origin activation and replisome integrity. Inside each MCM ring, the double helix becomes untwisted and base pairing is broken. This comes as the result of ATP-triggered conformational changes in MCM that involve DNA stretching and protein-mediated stabilization of three orphan bases. Mcm2 pore-loop residues that engage DNA in our structure are dispensable for double hexamer loading and CMG formation, but are essential to untwist the DNA and promote replication. Our results explain how ATP binding nucleates origin DNA melting by the CMG and maintains replisome stability at initiation.

Fig. 1. Visualization of origin-dependent CMG assembly by electron microscopy.

a, Workflow for the assembly of CMG on a chromatinized origin of replication for electron microscopy (EM) imaging. HSW, high-salt wash; LSW, low-salt wash; NCP, nucleosome core particle. b, Left, 2D averages derived from NS-EM imaging of the CMG assembly reaction. Centre, raw images and right, in silico reconstitution (ReconSil) of the double hexamer (DH) or dCMGE particles on the chromatinized origin of replication. Bottom, representation of the double-hexamer-to-CMG conversion efficiency. c, Measure of inter-nucleosome distance matches the expected length of the ARS1 origin of replication (n = 444 origins for double hexamer; n = 186 origins for dCMGE). Error bars, mean ± s.d. d, Comparison between MCM loading on short DNA containing MH roadblocks. After HSW treatment, equal amounts of loaded MCM helicases are eluted from Strep-TactinXT beads. The black arrowhead indicates MH-bound DNA. For gel source data, see Supplementary Fig. 1. This experiment was performed twice. e, Analysis of the replication products by alkaline agarose gel electrophoresis indicates that short nucleosome- and MH-capped origins can be replicated. For gel source data, see Supplementary Fig. 1. This experiment was performed twice. f, Replication reaction performed as shown in d except on large ARS1 circular DNA of wild-type and mutant MCMs. Mutants include Mcm2 6A, which targets residues that are involved in DNA untwisting; Mcm6 2E, which targets the Mcm6 wedge insertion; and Mcm6 5E, which targets the safety latch. For gel source data, see Supplementary Fig. 1. This experiment was performed twice. g, ReconSil of dCMGE formation on a 6× ARS1 array built from loaded double hexamers. This experiment was performed three times.

Fig. 2. dCMGE formation reconfigures the double hexamer interface, resulting in a splayed dimer.

a, Surface rendering of the dCMGE complex. b, Double-hexamer-to-dCMGE conversion promotes a one-subunit register shift at the MCM dimerization interface. Circles represent ZnFs. Black circles connected by lines indicate ZnFs engaged in tight inter-ring interactions. c, Double-hexamer-to-dCMGE conversion promotes the disengagement of an Mcm7 α-helical extension that protects the Mcm5 A domain on the opposite ring. This structural change exposes a GINS-binding site on Mcm5. PDB 7P30 refers to the Protein Data Bank (PDB) accession code. d, The dCMGE dimer is held together by a Mcm6 homo-dimer as well as by the DNA duplex. The dCMGE splayed dimer exposes a stretch of twisted duplex DNA that intervenes between the two MCM rings. e, DNA digestion disrupts the dCMGE dimer into single isolated CMGs (sCMGs), while also promoting the disengagement of Pol ε. This experiment was performed twice. Mean values are shown.

Fig. 3. ATP and DNA binding in the dCMGE complex.

a, MCM nucleotide occupancy in the double hexamer and in the dCMGE complex. b, Surface rendering of the nucleotide in the six ATPase sites of MCM. c, Duplex DNA binding in the dCMGE complex (left) explains how the double-hexamer-to-dCMGE transition leads to selection of the translocation strand. The ATPase pore loops in the dCMGE complex only contact the leading-strand template. The density for the selected translocation strand (red on the right) has been extracted from the duplex DNA density (grey on the left). d, The leading-strand template extracted from the dCMGE structure superposed on the yeast CMG translocating on a DNA fork reconstituted on an artificial DNA fork (PDB 6U0M), bound to the fork stabilization complex (PDB 6SKL) or bound to SCF^Dia2 and duplex DNA (PDB 7PMK).

Fig. 4. dCMGE formation leads to the untwisting of duplex DNA and breaks at least three base pairs.

a, Cryo-EM density of origin DNA. b, 0.7 turns of the double helix become untwisted after dCMGE formation. Three orphan bases become stabilized by residues T423 and R424 from the Mcm6-specific N-terminal hairpin insertion. c, Within the Mcm2 ATPase domain, only residue K587 contacts duplex DNA in the double hexamer. In dCMGE, five additional Mcm2 ATPase residues contact the DNA, which promotes widening of the minor groove, untwisting of duplex DNA and disruption of base pairing. d, Topology footprint assay for DNA unwinding. Complete reactions contained all firing factors after MCM loading plus TopoI; omission of DDK blocks all untwisting. Omission of Mcm10 captures the initial untwisted state. This initially untwisted state generates topoisomers of −2 and −3 (cyan arrowheads) as previously observed. Additional negatively supercoiled topoisomers can be detected when Mcm10 is present, indicating further untwisting after ejection of the lagging strand from CMG. No topoisomers were observed with the Mcm2 6A mutant. For gel source data, see Supplementary Fig. 1. This experiment was performed twice. e, NS-EM CMG averages derived from the CMG assembly reaction using wild-type or Mcm2 6A MCMs. CMG assembly reactions were performed on a 2 × MH DNA template (Extended Data Fig. 1c) to reduce background single CMG particles that are present owing to incomplete roadblocking of DNA. dCMGE is the product of CMG assembly using wild-type proteins, whereas Mcm2 6A primarily forms single CMGs. A minority of particles are compatible with dCMG or dCMGE formation (although in the latter, Pol ε occupancy is only partial (dCMG(E))). f, Mcm2 6A MCMs are converted to CMG at wild-type levels. P values were determined by two-tailed Welch’s t-test; NS, not significant. This experiment was performed three times. Error bars, mean ± s.d. g, Mcm2 6A disrupts the dCMGE dimer mostly into single isolated CMGs. P values were determined by two-tailed Welch’s t-test; **P = 0.0030. This experiment was performed three times. Error bars, mean ± s.d.

Fig. 5. A change in nucleotide engagement promotes the coupled disruption of the double hexamer dimerization interface and the stabilization of three orphan bases in the origin DNA duplex.

a, The Mcm6-specific wedge insertion in the N-terminal β-hairpin forms part of the dimerization interface in the double hexamer. In this configuration, wedge residues T423 and R424 map on the outer surface of the double hexamer. b, The double-hexamer-to-dCMGE transition promotes a reconfiguration of the Mcm6 wedge insertion, with T423 and R424 transitioning from the outer MCM perimeter to the inner lumen of the MCM ring. c, Swinging of the Mcm6 wedge from the outer MCM surface to the inner lumen leads to the stabilization of three orphan bases in the untwisted origin DNA duplex. d, In the double hexamer, the Mcm4 and Mcm7 h2i pore loops face downwards, with Mcm4 pushing against the Mcm6 N-terminal β-hairpin. This functions as a latch that maintains the Mcm6 wedge packed against the double hexamer dimerization interface. After CMG assembly, global changes in the ATPase tier of MCM cause the Mcm4 and Mcm7 h2i pore loops to move upwards, which releases the safety latch of the Mcm6 wedge insertion. This change promotes the Mcm6 wedge to swing upwards, with the R423 and T424 elements entering the MCM lumen to stabilize three orphan bases.

Extended Data Fig. 1. Origin-dependent CMG assembly with purified proteins visualized by electron microscopy.

a. Purified MCM loading and firing factors (left), and additional factors required for DNA substrate preparation (right) analysed by SDS–PAGE with Coomassie staining. For gel source data, see Supplementary Fig. 1. Similar results were observed for at least two independent sample preparations. b. 6% PAGE gel of capped, origin DNA substrates used in this study. (Below) Cartoon representation of ARS1 origins of replication, containing the two inverted ORC-binding sites, ACS (high affinity, red arrow) and B2 (low affinity, orange arrow). ARS1 is flanked by nucleosomes (NCP) or covalently linked methyltransferases (MH). For gel source data, see Supplementary Fig. 1. Similar results were observed for three independent sample preparations. c. Representative NS micrograph and 2D averages of entire CMG assembly reactions used to generate the ReconSiled origins shown in Fig. 1b. 70% of CMG particles exist in a dimeric (dCMGE) trans configuration (light orange), with GINS positioned on opposed sides of MCM. 11% of dCMGE particles exist in a cis configuration (dark orange) that might derive from trans-dCMGE disengagement and rotation. This experiment was performed more than three times. d. Cartoon representation of 6x ARS1 array, containing the two inverted ORC-binding sites, ACS (high affinity, red arrow) and B2 (low affinity, orange arrow). Each ARS1 origin is separated by 40 bp linker DNA. Array is flanked by covalently attached MH. Blue arrows indicate MseI cut sites. e. Reaction scheme for CMG assembly reactions on DNA substrates containing a 6x ARS1 array. f. Representative NS micrograph and representative double hexamer and dCMGE 2D averages obtained from CMG assembly reaction on 6× ARS1 array. This experiment was performed three times. g. 6% PAGE gel of partial DNA digestion of 6x ARS array by MseI carried out under the same conditions as NS-EM experiments. Lane 1 contains unmodified 6× ARS1 array. Lane 2 contains MH-conjugated 6x ARS1 array DNA. Lane 3 contains MseI digested 6× ARS1 array DNA. For gel source data, see Supplementary Fig. 1. This experiment was performed twice.

Extended Data Fig. 2. Sample preparation and validation of dCMGE cryo-EM reconstructions.

a. Schematic of biochemical reconstitution used for cryo-EM samples. b. Representative cryo-EM micrograph of entire dCMGE assembly reaction with particles highlighted with white circles. Cryo-EM sample preparation was performed once; similar results were observed in at least three independent NS sample preparations. b c. Representative 2D class averages from NS-EM (left panel) and cryo-EM imaging (right panel). Box widths represent 500 Å. d. Fourier shell correlation plot for the C1-refined CMGE dimer map. e. Angular distribution plot for the C1-refined CMGE dimer map. f. CryoSPARC local-resolution estimate for the C1-refined CMGE dimer map. g. Fourier shell correlation plot for symmetry-expanded CMGE map used in model building. h. Angular distribution plot for the symmetry-expanded CMGE map used in model building. i. Model-to-map correlation graph. j. PHENIX local-resolution estimate for the symmetry-expanded CMGE map.

Extended Data Fig. 3. Cryo-EM data processing pipeline for dCMGE assembled on nucleosome-capped origin DNA.

a. Schematic shows the classification and refinement steps taken to achieve the symmetry-expanded refined CMGE cryo-EM structure. The software used during each processing step is listed. b. Schematic of the classification and refinement pipeline for the C1-refined CMGE dimer. c. Symmetry-expanded refined CMGE structure (coloured) docked into C1-refined CMGE dimer map (grey). Reported resolutions in all schematics are calculated based on the FSC = 0.143 criterion.

Extended Data Fig. 4. Quality of cryo-EM densities.

Example cryo-EM density of Mcm2 (a), Mcm4 (b), Mcm7 (c), Pol2 (d), Sld5 (e), Dpb2 (f), and Cdc45 (g). For each subunit, side chain density features are shown in the inset.

Extended Data Fig. 5. Analysis of protein–DNA interactions within the dCMGE complex.

a. MCM contacts with DNA in dCMGE. Phosphate backbone interactions are indicated with black arrows and base interactions are highlighted with blue arrows. Residues are coloured based on individual MCM subunits. b. Comparison of N-terminal domains of Mcm2–7. The N-terminal pore loops are highlighted in red. Mcm6 contains a unique insertion (‘wedge’) with residues that stabilize the orphan bases exposed upon DNA untwisting. c. Cryo-EM density of flexible Mcm6 wedge stabilizing three lagging-strand bases. d. Representative cryo-EM densities of DNA contacts for Mcm3 and Mcm6. e. Sequence alignment of N-terminal region of Mcm6. Conserved T423 and R424 are outlined with black boxes. f. Sequence alignment between the h2i and PS1BH motifs of yeast Mcm2 compared to selected higher eukaryotes. Mcm2 6A mutations are outlined with black boxes. All alignments are coloured using the ClustalX colouring scheme.

Extended Data Fig. 6. Supporting information for the mechanism of DNA-bubble nucleation.

a. Overview of the Mcm2 6A mutant in the context of ATPase domain. Residues the push the lagging-strand template are coloured gold and residues that pull the leading-strand template are coloured in grey. b. Reaction scheme for topology footprint assay for DNA unwinding. c. Over-exposed gel as seen in Fig. 4d. For gel source data, see Supplementary Fig. 1. This experiment was performed twice.

Extended Data Fig. 7. dCMGE sterically impedes the docking of the E3 ligase onto MCM.

When CMGE–SCF^Dia2 (7PMK) is superposed onto the double hexamer (7P30) and the dCMGE structure (this study), major clashes can be identified between SCF^Dia2 engaged to one ring and the Mcm3 subunit from the opposed ring in the CMG dimer. This clash explains why the CMG assembled around the origin DNA duplex during initiation is protected from disassembly before lagging-strand ejection.

Extended Data Fig. 8. Schematic representation of the steps that lead to replication origin firing.

From top to bottom: double hexamer is loaded onto duplex origin DNA in an ATP-hydrolysis dependent manner. ADP formed during double hexamer assembly remains bound to MCM. The Mcm6 wedge insertion forms part of the double hexamer dimerization interface (inset). After loading the double hexamer makes limited contacts with duplex DNA. DDK phosphorylation triggers the recruitment of firing factors that deposit Cdc45 (C in CMGE) and GINS (G) onto the MCM (M), in the context of the so-called pre-initiation complex, the formation of which requires Pol ε (E) and CDK kinase activity. ADP release and binding of new ATP by MCM triggers CMGE assembly. CMGE assembly leads to the reconfiguration of the double hexamer interface, resulting in hexamer separation and a 1 subunit register shift pivoting around the Mcm6 N-terminal domain. This movement results in the exposure of 1.5 turns of duplex DNA between two MCM rings, and nucleation of DNA melting within each MCM ring. DNA melting is promoted upon ATP-triggered untwisting of 0.7 turns of the DNA, through the action of Mcm2 (pink in the inset), which pushes on the lagging strand while simultaneously pulling on the leading strand. As DNA is untwisted, Watson and Crick base pairs are broken, and three orphan bases become stabilized by the Mcm6 wedge (orange in the inset), which moves from the double hexamer interface and invades the MCM central channel. Action by Mcm10 triggers ATP hydrolysis by CMG and ejection of the lagging strand through an unknown mechanism, resulting in helicase bypass and establishment of replication forks.