Protein purification

ORC, Cdc6, Mcm2–7/Cdt1, DDK, CDK, Sld2, Sld3–Sld7, Cdc45, Dpb11, Polε, Mrc1, Csm3/Tof1, Mcm10 and MH were purified based on established protocols9,10,11,44,46,47,48.

Cloning, expression and purification of Mcm2–7/Cdt1mutants

The pMG73 plasmid was generated using a QuikChange Site-Directed Mutagenesis Kit (Agilent) with oMG40 and oMG41 primers (Supplementary Table 1) according to the manufacturer’s protocol. The pMG73 (Supplementary Table 2) was integrated into the yAM20 strain (Supplementary Table 3), yielding the yMG44 strain (Supplementary Table 3) that was used to overexpress the Mcm6 4A (Mcm6 R614A D615A E616A E617A) mutant. Mcm6 4A was purified as Mcm2–7/Cdt1 wild-type.

Cloning, expression and purification of Mcm10 mutants

Mcm10 ΔCTD (Δ358–571) expression vector was cloned into pET302-NT-His (vector sequence provided in the Supplementary Information) using Azenta. T7 express cells (New England Biolabs) were transformed with Mcm10 ΔCTD expression plasmid (pSSH006). Transformant colonies were inoculated into a 250-ml LB culture containing ampicillin (100 µg ml−1), which was grown overnight at 37 °C with shaking at 200 r.p.m. The next morning, 2 × 2 L of LB containing ampicillin (100 µg ml−1) was inoculated with 100-ml overnight culture. The cultures were grown at 37 °C to optical density at 600 nm (OD600) of 0.5. The cultures were then moved to 16 °C, at 200 r.p.m. After 40 min, OD600 of 0.6 was reached. To then induce expression, 0.5 mM isopropyl β-d-1-thiogalactopyranoside (IPTG) was added, and cells were left shaking for 3 h. Cells were collected by centrifugation at 4,000 r.p.m. for 20 min. Before lysis, cell pellets were resuspended in 50 ml of lysis buffer (25 mM HEPES pH 7.6, 10% glycerol, 0.02% NP-40, 1 mM EGTA, 500 mM NaCl, Roche protease inhibitor tablets, 1 mM dithiothreitol (DTT) + 0.5 mM Pefabloc). Cells were lysed by sonication for 120 s (5 s on, 5 s off) at 40%. After centrifugation at 20,000 r.p.m. for 30 min, the supernatant was incubated with 1 ml of M2 Flag resin (Sigma-Aldrich). The resin was washed extensively with wash buffer (25 mM HEPES pH 7.6, 10% glycerol, 0.02% NP-40, 300 mM NaCl). Mcm10 ΔCTD was eluted by the addition of 0.25 mg ml−1 3× Flag peptide. Fractions containing Mcm10 ΔCTD were then incubated with 2 ml of Ni-NTA resin (Qiagen). After washing (25 mM HEPES pH 7.6, 10% glycerol, 0.02% NP-40, 300 mM NaCl, 10 mM imidazole), the Mcm10 ΔCTD was eluted using 200 mM imidazole. Protein fractions were concentrated and loaded onto a Superose 6 Increase (24 ml) equilibrated in 25 mM HEPES pH 7.6, 10% glycerol, 0.02% NP-40, 0.5 mM EDTA, 200 mM NaCl, 2 mM DTT. Mcm10 ΔCTD fractions were then pooled, concentrated, aliquoted and snap frozen in liquid N2 (Mcm10 ΔCTD yield, 100 µg).

Microscale thermophoresis

In a microplate (384-well, F-bottom, Greiner Bio-One), 15-µl reactions were set up, covering a concentration range of 2,122, 1,273, 636, 212, 64, 21, 6.4, 2.1, 0.6, 0.2 and 0 nM Mcm10 wild-type or Mcm10 CTD mutant in reaction buffer (25 mM HEPES-KOH pH 7.6, 100 mM K-glutamate, 10 mM magnesium acetate, 0.02% NP-40, 0.5 mM TCEP). Each 15-µl reaction contained 5 nM fluorescently labeled ssDNA, CCCCCCCCCCCC[FAM]. Reactions were incubated for 30 min before measurements. Microscale thermophoresis measurements were carried out with Monolith NT.115 Premium Capillaries and Monolith NT.115. Using a blue LED excitation, 100% laser power and 20% MST, temporal fluorescence intensity traces were recorded for the different protein concentrations at room temperature using NTControl version 2.2.1. Traces were analyzed using NT Analysis 1.5.41, showing DFluorescence [temperature jump] ≥ 10. Independent biological triplicates were normalized between 0 and 1. Dose–response curves including standard error were plotted in GraphPad Prism version 9.4.1. The binding model used to fit the DNA binding data was as follows: nonlinear regression (curve fit) using a ‘specific binding with hill slope’ model Y = Bmax × Xh /(Kdh + Xh). The data were normalized (from 0 to 1) for comparison purposes, by setting the lowest value in each replicate to 0 and the highest value to 1 (and then plotting the mean value with standard errors).

DNA template: short 168-bp MH-flanked origins

The native ARS1 origin of replication flanked by M.HpaII was amplified by polymerase chain reaction (PCR) and purified as previously described9. MH-flanked origins were prepared based on previously established protocols9,11.

In vitro CMG assembly and activation on short MH-flanked origins

CMG assembly and activation were carried out by adapting previously published protocols11. In brief, 20 nM ARS1 MH-flanked origin DNA was incubated with 52 nM ORC, 52 nM Cdc6 and 110 nM Mcm2–7/Cdt1 for 25 min at 24 °C in loading buffer (25 mM HEPES-KOH pH 7.6, 100 mM K-glutamate, 10 mM magnesium acetate, 0.02% NP-40, 0.5 mM TCEP) + 5 mM ATP. Next, 80 nM DDK was added to the reaction and incubated for a further 10 min at 24 °C. DNA-bound protein complexes were isolated by incubation with 4 µl of MagStrep ‘type3’ XT beads (IBA), pre-washed in 1× loading buffer, for 30 min at 24 °C, to pull on twin-strep-tagged MH. Non-DNA bound proteins were removed by washing the beads three times with 100 µl of wash buffer (25 mM HEPES-KOH pH 7.6, 5 mM magnesium acetate, 0.02% NP-40, 500 mM NaCl), followed by one wash with 100 µl of loading buffer. DNA-loaded, phosphorylated double hexamers were eluted in 20 µl of elution buffer (25 mM HEPES-KOH pH 7.6, 105 mM K-glutamate, 10 mM magnesium acetate, 0.02% NP-40, 0.5 mM TCEP, 27 mM biotin, 5 mM ATP) for 10 min at 24 °C. The supernatant was then removed, and 125 nM CDK was added and incubated for 2 min at 30 °C. A mix of firing factors was then added to a final concentration of 45 nM Dpb11, 150 nM GINS (either His-GINS or TwinStrepII-GINS), 120 nM Cdc45, 30 nM Polε, 45 nM Sld3–Sld7 and 75 nM Sld2, including or excluding 22 nM Mcm10 or Mcm10 mutants. After a 14-min incubation, the reaction was applied directly to grids.

DNA replication assays

Replication assays were carried out as described previously11 using pJY22 plasmid (Supplementary Table 2). Staged replication reactions containing Rad53, Mrc1 and controls were performed as previously described with the exception of Rad53:Mcm10 1:10 ratio17. Mrc1 and Csm3/Tof1 were used at 20 nM concentration. For pulse-chase experiments, the conditions were the same as for standard DNA replication assay, except that the concentration of dCTP in pulse was reduced to 4 μM, whereas, during the chase, it was increased to 600 μM. The chase was at either 8 min (Fig. 6e,f) or 9 min (Extended Data Fig. 4f). Quantitations were performed using ImageJ2/Fiji version 2.3.0.

Plasmid-based DNA unwinding assay

DNA unwinding assays were carried out using previously published protocols11. In brief, the DNA unwinding assay was performed using 3.2-kb pBS_ARS1_WTA plasmid49 following a published protocol10. Then, 25 fmol of plasmid DNA was treated with 20 nM Topo I for 30 min at 30 °C in a buffer containing 25 mM HEPES-KOH pH 7.6, 100 mM K-glutamate, 10 mM magnesium acetate, 0.02% NP-40-S, 5% glycerol, 2 mM DTT, 5 mM ATP. Next, 10 nM ORC, 50 nM Cdc6 and 100 nM Mcm2–7/Cdt1 were added for 20 min at 30 °C. The reaction was then supplemented with 50nM DDK, and incubation was continued for 30 min at 30 °C. Additional buffer was supplemented to achieve a final concentration of 250 mM K-glutamate, 25 mM HEPES, 10 mM Mg-acetate, 0.02% NP-40-S, 8% glycerol, 400 μg ml−1 BSA, 5 mM ATP, 1 mM DTT. 25 nM Topo I. The mix of firing factors was prepared before use and added at time 0, reaching a final concentration of 30 nM Dpb11, 20 nM GINS, 50 nM Cdc45, 20 nM Polε, 20 nM CDK, 10 nM Mcm10, 25 nM Sld3–Sld7, 50 nM Sld2, 50 nM RPA. After a 40-min incubation at 25 °C, the reaction was quenched using 13 mM EDTA, 0.3% SDS and 0.1 mg ml−1 Proteinase K (Merck) and incubated at 42 °C for 20 min. The sample was extracted once with phenol:chloroform:isoamylalcohol (25:24:1) and ethanol precipitated, and the DNA pellet was resuspended in 1× Tris-EDTA for analysis. Samples were run in 1.5% agarose gel with TAE, followed by staining with ethidium bromide.

Negative-stain EM sample preparation and data collection

Negative-stain sample preparation was conducted using previously published protocols11. Preparation of negative-stain samples was performed on either 300-mesh or 400-mesh copper grids with carbon film (EM Resolution or Agar Scientific, respectively). Grids were glow discharged for 60 s at 25 mA (GloQube Plus, Quorum), and 4 µl of sample was applied to the grids, followed by 2-min incubation. Grids were stained by two successive applications of 4 µl of 2% (w/v) uranyl acetate with quick blotting between the applications. The second stain application was blotted after 20 s to remove excess stain, and grids were stored before imaging. Data collection was carried out on a Tecnai LaB6 G2 Spirit transmission electron microscope (FEI) operating at 120 keV. Two cameras were used for micrograph collection: a 2,000 × 2,000 Gatan Ultrascan 100 camera at a nominal magnification of 30,000 (with a physical pixel size of 3.45 Å per pixel) and a 4,000 × 4,000 Gatan RIO at a nominal magnification of 29,000 (with a physical pixel size of 3.1 Å per pixel). Collections were carried out within a −0.5 µm to −2.0 µm defocus range. Digital Micrograph software was used for data acquisition.

Negative-stain electron microscopy analysis image processing

Negative-stain electron microscopy analysis image processing was carried out using approaches described previously11. A particle subset was manually picked using RELION 3.1 (ref. 50) and used to train a Topaz model for particle picking51. Negative stain image processing was performed using RELION 3.1. The contrast transfer function (CTF) was estimated using Gctf52, and particles were extracted and subjected to reference-free 2D classification in RELION 3.1. The same particle population trends were observed when a different team member re-analyzed the particle stacks performing 2D classification with cryoSPARC53.

When only sCMGE and double hexamer 2D classes were found, conversion efficiency for sCMGEs was calculated by dividing the number of sCMGE10 pairs by the number of double hexamers added to the number of sCMGE10 pairs—that is, (sCMGE / 2) / (DH + sCMGE / 2). This is because two sCMGE complexes would result from any given double hexamer. When sCMGE, dCMGE and double hexamer 2D classes were found, conversion efficiency for sCMGEs was calculated by dividing the number of sCMGE10 pairs by the number of double hexamers added to the number of sCMGE10 pairs added to the number of dCMGEs—that is, (sCMGE / 2) / (DH + sCMGE / 2 + dCMGE).

Graphene oxide grid preparation

UltrAuFoil R1.2/1.3 grids (Quantifoil Micro Tools) were freshly glow discharged for 5 min at 25 mA using a glow discharge unit (GloQube Plus, Quorum). Graphene oxide dispersion (2 mg ml−1, Sigma-Aldrich) was diluted by adding 80 µl of water to 10 µl of graphene oxide dispersion. Diluted dispersion was spun down at 500g for 1 min. The top layer of dispersion was transferred to a new tube, avoiding aggregates. For the grid preparation, 4 µl of diluted graphene oxide dispersion was applied to the freshly glow discharged UltrAuFoil R1.2/1.3 grid and incubated for 2 min. Both sides of the grid were blotted. Droplets of water, 20 µl each, were picked up with the grid and blotted, twice for the front and once for the back of the grid. Grids were then dried upside down.

Cryo-EM sample preparation and data collection

Cryo-EM sample preparation was carried out by adapting previously published protocols11. CMG assembly and activation reactions (reconstituted as described in in vitro CMG assembly and activation on short MH-flanked origins) were frozen on UltrAuFoil R1.2/1.3 grids (Quantifoil Micro Tools) with a freshly prepared graphene oxide layer. All grids were prepared as detailed above before freezing. Samples were prepared by applying 4 µl of undiluted CMG assembly and activation reaction on grids, incubated for 2 min at 25 °C in 90% humidity. Excess sample was subsequently blotted away for 4.5 s or 5.0 s, and grids were plunge frozen in liquid ethane using a Vitrobot Mark IV (FEI Thermo Fisher Scientific).

Data collection was performed on an in-house Thermo Fisher Scientific Titan Krios transmission electron microscope operated at 300 kV equipped with a Gatan K2 direct electron detector camera and a GIF Quantum energy filter (Gatan). Images were collected automatically using EPU software (Thermo Fisher Scientific) in counting mode with a physical pixel size of 1.08 Å per pixel, with a total electron dose of 51.4 electrons per Å2 during a total exposure time of 10 s, dose fractionated into 32 movie frames (Table 1). We used a slit width of 20 eV on the energy filter and a defocus range of −1.8 μm to −3.3 μm. A total of 71,117 micrographs were collected from two separate sessions.

Cryo-EM image processing

Data processing was carried out in RELION 3.1 (ref. 50) or cryoSPARC version 4.0.3 (ref. 53) (Extended Data Fig. 3). Correction for movie drift and dose weighting was performed using MotionCorr2 (ref. 54). CTF parameters were estimated for the drift-corrected micrographs using CtfFind4 (ref. 55) in RELION 3.1. First, dataset 1, which was collected first, was processed separately and, at a later stage, combined with dataset 2. For the first dataset, particles were manually picked from 2,000 micrographs using cryoSPARC version 4.0.3 (ref. 53). These particles were extracted with a box size of 4202 pixels for 2D. Starting with this subset of particles across the entire defocus range, a Topaz model51 was trained iteratively to improve particle picking. In RELION 3.1, particles were picked from both datasets (first and second collection, total 71,116 micrographs) with a select threshold of 0.

The two datasets were combined, and a total of 2,016,248 particles were picked, binned by 2 and extracted with a box size of 1802 pixels. Picked particles were subjected to 2D classification to remove remaining smaller particles and contaminants. With the remaining particles, we carried out a 3D classification with two or three subclasses, angular sampling of 7.5°, regularization parameter T of 4 using a low-pass filtered initial model from previous ab initio and processing steps on dataset 1 of CMGE complexes (Extended Data Fig. 3). At the same stage, all particles were subjected to 2D classification to select for high-resolution 2D classes. Selection of 2D and 3D classes were combined, yielding 244,182 particles. These particles were un-binned to then perform another round of 3D classification with two subclasses, angular sampling of 7.5°, regularization parameter T of 8 using a low-pass filtered model from previous 3D classification. The resulting 172,552 CMGE particles were refined to yield maps with resolutions of 4.4 Å.

These particles were subjected to several rounds of CTF refinement and two rounds of Bayesian polishing. After this, CTF-refined and polished particles were refined with local searches with a mask encompassing the entire CMGE density to 3.7 Å resolution, the ‘consensus’ structure (Extended Data Fig. 2 and Table 1).

From 3D classifications, we found that some 3D classes contain double-stranded DNA (dsDNA) inside the central channel of CMG toward the C-terminus of CMG. 3D classification, however, gave poor particle separation of CMGE particles on ssDNA and CMGE particles with dsDNA in the C-terminal region. To better separate these two particle sets, we used signal subtraction in combination with reference-free 2D classification without alignment in RELION 3.1 (refs. 8,50). Using this approach, we separated side views; a subset of 13,174 particles was selected as being threaded on only ssDNA, and a subset of 23,337 particles was selected with dsDNA in the C-terminal region. All signal-subtracted particles were then reverted. To separate top and bottom views of CMGE that contain ssDNA from CMGE that contain dsDNA, we carried out a 3D classification of the remaining particles and particle side views with dsDNA. The 3D classification was performed with 20 subclasses, angular sampling of 7.5° and regularization parameter T of 16, and resolution was limited to E = 12, using a low-pass filtered model from previous refinement. 3D classes were categorised into CMGEs with either ssDNA or dsDNA in C-terminus of the central channel. Particles from 2D classified side views mentioned above were also added accordingly to the particle sets of CMGE on ssDNA or dsDNA. These two particle sets were refined using local searches to 3.9-Å resolution for ssDNA-bound CMGE complexes and 3.9-Å resolution for particles with dsDNA in the C-terminal region (Table 1).

All refinements were performed using fully independent data half-sets, and resolutions are reported based on the Fourier shell correlation (FSC) = 0.143 criterion (Extended Data Fig. 2). FSCs were calculated with a soft mask. Maps were corrected for the modulation transfer function of the detector and sharpened by applying a negative B-factor as determined by the post-processing function of RELION or using a higher B-factor to prevent overfitting. PyEM (https://github.com/asarnow/pyem) and bsfot (https://cbiit.github.io/Bsoft/) were used for format conversion.

Model building and refinement

To generate a suitable starting point for model building, deposited coordinates for CMGE (Protein Data Bank (PDB) 7QHS) were split into MCM subdomains, each of which was rigid body docked into a refined consensus volume in UCSF Chimera56, along with chains corresponding to the GINS subcomplex and Cdc45. DNA polymerase epsilon catalytic subunit A coordinates AA 1551–1585 were obtained from PDB 7PMK. Refined maps were converted to MTZ format using the mrc2mtz module. The coordinates were then rebuilt and extended in Coot57 according to the density of blurred and sharpened outputs from mrc2mtz. To address steric clashes and geometric outliers, the model was further adjusted using ISOLDE58, resulting in a base set of coordinates. As data processing revealed an ssDNA-bound and a nexus-bound state within the consensus particle set, maps for these subclassified volumes were also produced for model building. The base coordinates were then docked into these maps and adjusted according to the density. Each set of coordinates was then refined separately in PHENIX59. Atomic model geometries were evaluated using MolProbity webserver60. Mcm10 docking was performed using UCSF Chimera56 and locally adjusted using ISOLDE58.

Map and model visualization

Maps and all model illustrations were visualized and prepared using Chimera or ChimeraX56,61.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.