Structural mechanism for the selective phosphorylation of DNA-loaded MCM double hexamers by the Dbf4-dependent kinase

Abstract

Loading of the eukaryotic replicative helicase onto replication origins involves two MCM hexamers forming a double hexamer (DH) around duplex DNA. During S phase, helicase activation requires MCM phosphorylation by Dbf4-dependent kinase (DDK), comprising Cdc7 and Dbf4. DDK selectively phosphorylates loaded DHs, but how such fidelity is achieved is unknown. Here, we determine the cryogenic electron microscopy structure of Saccharomyces cerevisiae DDK in the act of phosphorylating a DH. DDK docks onto one MCM ring and phosphorylates the opposed ring. Truncation of the Dbf4 docking domain abrogates DH phosphorylation, yet Cdc7 kinase activity is unaffected. Late origin firing is blocked in response to DNA damage via Dbf4 phosphorylation by the Rad53 checkpoint kinase. DDK phosphorylation by Rad53 impairs DH phosphorylation by blockage of DDK binding to DHs, and also interferes with the Cdc7 active site. Our results explain the structural basis and regulation of the selective phosphorylation of DNA-loaded MCM DHs, which supports bidirectional replication.

Fig. 1. Protein–DNA contacts and nucleotide occupancy in the DNA-loaded MCM DH.

a, A cut-through view of the MCM DH revealing the atomic model of the double helix. Density and atomic model for selected DNA interacting residues. b, Isosurface representation of the 3.0-Å-resolution structure of the DNA-loaded MCM DH. c, Bottom view of ATPase sites with nucleotides are depicted in black. Nucleotide density at the Mcm6–Mcm2 interface is consistent with ATP rather than ADP binding. Segmented density for ATP and the magnesium ion in the active site provides confidence about the assignment. Apo means no nucleotide.

Fig. 2. Molecular basis for kinase recognition and phosphorylation of the MCM DH substrate.

a, SDS–PAGE gel of unmodified and DDK-phosphorylated MCM DHs, tethered to DNA beads. The phosphorylation-dependent shifts of Mcm4 and Mcm6 are highlighted in green and orange, respectively. Notably, after a low-salt-wash step, DDK remains bound to the MCM. Representative of at least n = 3 independent experiments. b, 2D class average of the isolated DDK shows subnanometer-resolution features, indicating that DDK is a suitable cryo-EM target. Scale bar, 10 nm. c, 3.3-Å-resolution structure of the MCM DH–DDK complex, showing the catalytic core of DDK (Cdc7 bound to C-terminal Dbf4), engaged to the Mcm4 subunit of one MCM ring in the DH. Two cut-through views of the kinase core are shown to highlight how the atomic model matches cryo-EM density. d, DDK docks onto the Mcm4 A domain via the Dbf4 zinc finger C domain, and onto the Mcm4 B domain via the Dbf4 M domain. Middle: active site of the human DDK crystal structure, which was cocrystallized with an MCM substrate peptide in the active site of Cdc7. The first resolved N-terminal residue (R155) of Mcm4 in the cryo-EM map neatly aligns with the C-terminal end of the MCM peptide. The Mcm4 N-terminal tail in our structure is therefore suitably poised for phosphorylation by the Cdc7 active site. An N-terminal Mcm4 segment (P155 to R176), which is invisible in the absence of DDK, is partially stabilized in the DH–DDK complex. e, One known DDK phosphosite in Mcm4 (S171) maps within the DDK-stabilized N-terminal segment visible in our structure. Five additional known phosphosites and others detected by mass spectrometry map upstream of the modeled N-terminal region of Mcm4. This agrees with the notion that active site access requires extended structural flexibility of the phosphorylation substrate. Sites reported to be important for recruitment of the firing factor Sld3 are highlighted. The uncropped gel image for a is available as Source data with the paper online. Source data

Fig. 3. The Dbf4 BRCT domain docks onto the A domain of Mcm2 and its truncation does not affect Cdc7 catalytic activity.

a, Signal subtraction of the MCM ATPase domains, followed by focused classification, 3D refinement and LAFTER filtering, allowed resolution of the docking site of DDK (shown in purple) onto the MCM DH. b, The crystal structure of the Dbf4 BRCT domain (PDB entry 3QBZ) was fitted to the newly resolved docking site, which contacts the A domain of the Mcm2 subunit. c, Wild-type (WT) DDK (lanes 1–7) and DDK containing a Dbf4 BRCT truncation (lacking residues 119–219, lanes 8–14) show comparable autophosphorylation efficiency, which can be reverted by lambda phosphatase treatment (lanes 7 and 14). Representative of at least three independent experiments. d, Autoradiograph of a kinase assay using a well-characterized substrate of DDK kinases (residues 35–47 of human Mcm2). e, Quantification of the kinase assay. The average of three biological replicas is plotted and error bars show s.d. Reads were normalized to the 45-min time point of wild-type DDK. The DDK mutant shows wild-type levels of MCM phosphorylation. As described for the human ortholog, phosphorylation by Cdc7 requires prephosphorylation of Ser41 (P + 1). Uncropped gel images for c and d are available as Source data with the paper online. Source data

Fig. 4. DDK docking onto the MCM DH in cis is required for MCM phosphorylation in trans.

a,b, Composite map of DH structure bound by the DDK core particle and LAFTER-filtered BRCT Dbf4 density shows an extended configuration adopted by DDK, with docking onto Mcm2 in cis and phosphorylation of Mcm4 in trans; composite map and atomic model (a). The network of protein contacts that support DH phosphorylation by DDK is illustrated by artificial separation of the different interactors (b). c, Titration of wild-type DDK shows efficient phosphorylation of DNA-loaded MCM DHs (lanes 6–9). When the ΔBRCT DDK mutant was tested in the same concentration range, no phosphorylation was detected. The effect was observed in three independent experiments. d, Negative-stain 2D class averages of MCM DHs incubated with wild-type or ΔBRCT DDK. In the absence of the Dbf4 BRCT docking domain, no DDK density can be visualized bound to the MCM DH. e, Model illustrating the importance of DDK docking in cis for MCM phosphorylation in trans. Source data

Fig. 5. DDK preautophosphorylation prevents peptide and MCM–Cdt1 phosphorylation, but not DH phosphorylation and engagement.

a, MCM peptide phosphorylation drops by 92% following DDK preautophosphorylation. b, MCM–Cdt1 phosphorylation is abrogated following DDK preautophosphorylation. c, DH phosphorylation is virtually unperturbed following DDK preautophosphorylation. a–c, Representative images of three independent experiments. d, 2D averages derived from negatively stained particles indicating that DDK binding to DHs is unperturbed following DDK preautophosphorylation. Uncropped images for a and c are available as Source data with the paper online. Source data

Fig. 6. DDK inhibition by Rad53.

a, Rad53 phosphotargets on Dbf4. S518–S528 phosphorylation blocks origin firing in cells. b, Autoradiograph and quantification of a kinase assay comparing peptide phosphorylation after treatment of DDK with Rad53 or an enzymatically inactive version (Rad53^KD). The average of three biological replicas is plotted and error bars shows s.d. c, Model showing how phosphorylation by Rad53 interferes with the Cdc7 active site, independent of DH engagement. d, Phosphorylation of DNA-loaded MCM DHs is inhibited after Rad53-dependent phosphorylation of DDK, but not by the presence of Rad53^KD. The effect was observed in three independent experiments. e, Table summarizing engagement of DDK with MCM DHs in the presence of Rad53 and Rad53^KD when imaged by negative-stain EM. Representative 2D class averages of DHs and DDK-engaged DHs are shown from 6,033 (no Rad53), 7,018 (WT Rad53) and 5,519 (kinase-dead Rad53) picked DHs. Blue arrowheads indicate DDK bound to the DH and the red cross indicates DH without DDK. f, Model showing how Rad53-dependent phosphorylation of Dbf4 prevents DDK from docking onto the MCM and inhibits MCM phosphorylation. Uncropped images for b and d are available with the paper online. Source data

Extended Data Fig. 1. MCM loading and phosphorylation intermediates visualised by cryo-EM.

a MCM double hexamer loading and phosphorylation was reconstituted in vitro and analysed by cryo-EM. As a result, several intermediates on route to double hexamer loading could be visualised, alongside instances where DDK interacts with double hexamers. These intermediates include DNA-bound ORC, MCM–Cdt1 and the MCM–ORC (MO) intermediate. b A cartoon representation of the origin licensing. c Representative aligned movie sum of yeast DDK analysed by cryo-EM (9,469 movies were collected). Scale bar is equivalent to 50 nm. d 2D averages of DDK showing high-resolution features. Scale bar is equivalent to 10 nm.

Extended Data Fig. 2. Cryo-EM structure of the MCM double hexamer isolated and bound to DDK.

a Representative aligned movie sum. b 2D averages containing recognisable double hexamers. c Angular distribution of the MCM double hexamer structure. d Resolution of the MCM double hexamer estimated using gold-standard Fourier Shell Correlation. e Three rotated and one cut-through view of the MCM double hexamer 3D structure, color-coded according to the local resolution. f Angular distribution of the MCM double hexamer–DDK complex. g Resolution of the MCM double hexamer–DDK complex estimated using gold-standard Fourier Shell Correlation. h Three rotated views and one cut-through view of the MCM double hexamer–DDK complex 3D structure, color-coded according to the local resolution. Source data

Extended Data Fig. 3. Protein-DNA contact map and illustration of the ATPase sites in the DNA-loaded MCM double hexamer.

a Model of the residues within the MCM hexamer that contact the DNA. Density and atomic model for selected interactions are shown. b Illustration of the nucleotides bound in the different ATPase sites. c Segmented density for each of the nucleotides and the coordinated magnesium ion.

Extended Data Fig. 4. Image processing for the MCM double hexamer–DDK complex.

a Processing pipeline. b Illustration of the symmetry expansion protocol. Partial DDK occupancy on the two-fold symmetric MCM double hexamer hampers structure determination of a DDK–MCM double hexamer complex. To align asymmetrically bound DDK molecules, symmetry expansion was performed. According to this procedure, a copy of the dataset was appended to the MCM–DDK particles, after rotation around the C2 symmetry axis. The signal of one DDK molecule was subtracted from the rotated images to prevent particles from reverting to the original orientation. DDK-bound double hexamer particles were identified by using focused 3D classification skipping angular searches.

Extended Data Fig. 5. DDK inhibition by Rad53.

a Phospho-mimicking mutations of T188/S192, which map at the Dbf4–Mcm2 interface, do not interfere with the auto-phosphorylation efficiency of DDK. b Quantification of the kinase assay comparing wild type and phospho-mimicking DDK. The phospho-mimicking mutant efficiently phosphorylates the MCM substrate peptide. c MCM double hexamer phosphorylation by Cdc7–Dbf4^T188D,S192D is comparable to the wild type kinase. d-f A possible mechanism of phosphorylation-independent inhibition of DDK by Rad53. d The crystal structure of Rad53(FHA1)/Dbf4(BRCT)/Cdc7 (PDB entry 5T2S) superposed to DH–DDK shows no clash between Mcm2 and Rad53. e Modelled configuration would not support Mcm4 phosphorylation in trans. f Model for Rad53 engagement sequestering DDK in a cis configuration. Rad53 has been reported to interfere with DH phosphorylation by DDK via two mechanisms, a recently discovered antagonistic binding of Rad53 to DDK that can prevent DH engagement, and the direct phosphorylation of DDK, which blocks late origin firing^,. Our DDK-DH-DNA structure informs both molecular mechanisms (Fig. 5a). To address the phosphorylation-independent, antagonistic binding, we integrated our new cryo-EM data with the published crystallographic information on the Rad53-DDK interaction. The Dbf4 BRCT domain is known to interact with the non-catalytic Forkhead-associated 1 (FHA1) domain of Rad53. When we superposed BRCT-FHA1 co-crystal structure (PDB entry 5T2F) to our DNA-DH-DDK structure we detected no steric clash between FHA1 and the Mcm2 A domain, indicating that FHA1 engagement could still support BRCT docking onto Mcm2. However, when using the structure of a ternary complex, with the FHA1 domain simultaneously bound to Dbf4 and a Cdc7 peptide, we concluded that Rad53 binding would be incompatible with Mcm4 phosphorylation by DDK acting in trans. In fact, the Rad53-interacting peptide of Cdc7 (residues 480-491, including phospho-Thr484) maps only 10 residues downstream of the C-terminal end of the Cdc7 atomic model in our DH-DDK structure. A stretch of 10 residues would cover a distance no longer than 3.1 nm, which is shorter than the 9.2 nm distance modelled between the FHA1-engaged Cdc7 peptide and the Cdc7 C-terminus in our DH-DDK atomic model. Therefore, based on our analysis, concomitant binding of Dbf4 and Cdc7 by FHA1 Rad53 could cause the DDK particle to fold back onto itself. When sequestered by Rad53, DDK would be impaired from visiting the extended state that straddles across the two helicase rings. Blocking Mcm4 engagement in trans would prevent DH recognition, in turn impairing any activating phosphorylation. However, antagonistic binding of Rad53 to DDK for blocking DH phosphorylation, independent of Rad53 catalytic activity, would require a stoichiometric interaction between the two kinases. This explains why the Rad53-phosphorylation dependent block of origin firing is more readily detected^,. Source data