Introduction

Eukaryotic chromosomal DNA replication in the cell cycle, a tightly regulated process that ensures precise gene copying, begins with the unwinding of double-stranded DNA (dsDNA) into signal-stranded DNA (ssDNA), including the formation of an active helicase Cdc45–MCM–GINS (CMG) complex [1,2]. In the G1 phase of the yeast cell cycle, two inactive helicase Mcm2–7 hexamer rings (Mcm2–6–4–7–3–5) are loaded onto autonomously replicating sequences (ARSs) of dsDNA to form a double hexamer (MCM DH) at the replication origin [3–5]. To facilitate this loading, Cdt1 binds to the Mcm2–7 hexamer ring and stabilizes a gap formation between the Mcm2 and Mcm5, which serves as a dsDNA entry gate [6–9]. Next, the MCM DH on dsDNA is phosphorylated by Dbf4-dependent protein kinase (Cdc7 kinase or DDK) (Supplementary Figure 1A), allowing the key factor Sld3 and its partner Sld7 to recruit Cdc45 (Sld7–Sld3–Cdc45) to each MCM (Supplementary Figure 1B) [10–13]. Subsequently, the phosphorylation of Sld3 by cyclin-dependent kinase (CDK) regulates the assembly of GINS onto MCM DH with Dpb11, CDK-phosphorylated Sld2, and DNA polymerase ε to form CMG (Supplementary Figure 1C) [14–19]. Finally, the factors Sld2, Sld3, Sld7, and Dpb11 are released by elusive mechanisms from an active CMG and bidirectional replication by translocating along the 3′ to 5′ direction of the DNA strand (Supplementary Figure 1D) is facilitated [20, 21].

As a central regulator of CMG formation in Saccharomyces cerevisiae, Sld3 functions as a bridge protein in the Sld7–Sld3–Cdc45 complex to recruit Cdc45 to DDK-phosphorylated MCM DH. This Sld3 contains three domains each of which binds mainly to one of Sld7, Cdc45, and MCM. The N-terminal domain of Sld3 (Sld3NTD: M1–L116) binds to the N-terminal domain (M1–D130) of Sld7 (Sld7NTD) (Sld7 C-terminal domain [Sld7CTD: K168–S257] is a self-dimerization domain) [22]. Following the NTD of Sld3, the middle part is a central portion known as the Cdc45-binding domain (Sld3CBD: S148–K430) [12, 23]. A previous study also demonstrated that a region (L510–R530) in the Sld3 C-terminal domain (Sld3CTD: T445–T679) binds to Mcm4 and 6 [24]. CDK-dependent phosphorylation of two residues (T600 and S622) downstream of Sld3CTD initiates the binding of GINS-carrying Dpb11 to Cdc45–MCM [19, 25]. In addition to its role as a bridge in the recruitment of Cdc45 and GINS, Sld3 binds to two single-stranded DNA (ssDNA) fragments of ARS1 which was identified as an origin of DNA replication (ssARS1-2 and ssARS1-5), but not to the corresponding double-stranded ARS1 (dsDNA) [26]. The specific association between Sld3 and ssDNA is not affected by CDK phosphorylation. Furthermore, the structures of the Sld7CTD dimer (PDBID:3X38) [22], Sld7NTD–Sld3NTD (PDBID:3X37) [22], Sld3CBD (PDBID:3WI3) [23], MCM DH (6F0L) [27], CMG (PDBIDs:3JC5, 3JC6, 3JC7) [28], CMG-DNA-polLJ (PDBID:7Z13) [29], CMG-DONSON-DNA (PDBID:8Q6O) [30], and so on, were determined through crystallography and cryogenic electron microscopy (cryo-EM). Cdc45 belongs to the DHH superfamily of proteins defined by the conserved triad motif DHH (Asp-His-His), and there is a DHH-associate domain (DHHA1: R523–L650) at the C-terminus [31]. Recent single-molecule biochemical assays have reported the stepwise recruitment of multiple Cdc45s to the MCM DH [32]. However, how Sld3–Sld7 brings Cdc45 into the MCM for the formation of CMG to regulate the initiation of DNA replication remains unclear.

In the present study, we aimed to understand how Sld3 recruits Cdc45 to the MCM DH with Sld7 for CMG formation through structure and particle analyses. We determined the structure of S. cerevisiae Sld3CBD–Cdc45 at 2.6 Å resolution and presented the detailed interactions between Sld3 and Cdc45, confirmed through in vitro and in vivo mutant analyses. Compared to the monomer structures, the conformation of Sld3CBD and Cdc45 in the Sld3CBD–Cdc45 complex changed significantly to bind each other. Based on the structural similarity of Cdc45 in Sld3CBD–Cdc45 and CMG, we modelled Sld3CBD–Cdc45–MCM–dsDNA and SCMG–dsDNA (Sld3CBD–CMG–dsDNA) as a snapshot of how helicase CMG forms. The models demonstrated that Sld3CBD, MCM, and GINS bind to different sites on Cdc45, indicating that Sld7–Sld3 could remain at Cdc45–MCM until CMG formation after GINS loading. Consistency between the particle size of Sld7–Sld3ΔC–Cdc45, (Sld3ΔC: M1–K430; truncated the C-terminal domain) as per spectroscopic analysis, and the distance of Sld3CBDs in the Cdc45–MCM dimer suggested that the ternary complex of Sld7–Sld3–Cdc45 forms a dimer off/on the MCM DH for recruiting Cdc45. Furthermore, ssDNA-binding analysis of ARS1 fragments implied that the release of Sld3–Sld7 could be associated with ssARS1 unwound by CMG. Our findings illustrate the recruit–release function of Sld3 in CMG formation and expand our knowledge of the initiation process of DNA replication.

Results

The overall structure of Sld3CBD and Cdc45 complex

We obtained a recombinant Sld3CBD–Cdc45 complex of S. cerevisiae and determined its crystal structure at 2.6 Å resolution using molecular replacement (Figure 1A, Supplementary Figure 2 A, 3). Only one complex exists within an asymmetric unit. Similar to monomer Sld3CBD (PDBID: 3WI3 with RMSD of 0.50 Å for 181 Cα atoms) [23], Sld3CBD (Y154–P420) in the Sld3CBD–Cdc45 complex is an α-helical structure with two disordered regions of R317–S336 and P364–A369, but with a significant difference in that a disordered part in monomer Sld3CBD was visualized as a C-terminal part of bent long helix α8 (F294–R316; referred to as α8CTP) (Supplementary Figures 3, 4A and 5). Cdc45 (M1–L650) is an α/β structure composed of three β-sheets (anti-parallel: β1-β6-β5-β4-β2-β3, anti-parallel: β7-β8, mixed: β9-β10-β11-β13-β12) surrounded by 21 α-helices (Supplementary Figure 3). Owing to the poor electron density map, the regions D106–K110, D166–K227, S306–V310, and T438–D460 on the molecular surface could not be built. In contrast to the monomer human Cdc45 (huCdc45) and the CMG form (Cdc45 in the yeast CMG complex), more than the N-terminal half of the protruding long helix α7 was disordered in the Sld3CBD–Cdc45 complex (Supplementary Figures 4B and 6).

Figure 1.

Conformation changes in Sld3CBD and Cdc45 for binding each other

Sld3CBD binds to Cdc45 in a way similar to a toothed gear (Figure 1B), with a contact surface of 1808 Ų accounting for approximately 13.3% and 7.0% of the total surface of Sld3CBD and Cdc45, respectively. Two helices (α8 and α9) of Sld3CBD formed a plier structure and gripped the C-terminal domain DHHA1 (R523–L650) of Cdc45 through hydrophobic interactions and hydrogen bonds. The C-terminal loop (L646–L650) and the C-terminal part of α7 (E232–S242) of Cdc45 sandwiched the helix α8CTP of Sld3CBD. Compared to the isolated form (PDBIDs: 5DGO for huCdc45 [31] and 6CC2 for EhCdc45 [33]) and the CMG form (PDBID: 3JC6 [28]), the Cdc45 in the Sld3CBD–Cdc45 complex changed the conformation of DHHA1 to form pockets on its two sides for binding Sld3CBD α8CTP and α9. The remaining part of Cdc45 (∼K517) retained a structure with RMSD values of 1.29 Å (isolated huCdc45), 1.81 Å (isolated EhCdc45), and 1.295 Å (in CMG) for 243, 251, and 361 Cα atoms, respectively.

The helix α8CTP (F294–R316) of Sld3CBD was surrounded by the C-terminal domain DHHA1 and a C-terminal part of the protruded long helix α7 (E232–S242) of Cdc45 but completely disordered in Sld3CBD alone (PDBID: 3WI3) [23] (Figure 1B and Supplementary Figure 4A). Therefore, the helix α8CTP seems to be an intrinsically disordered segment when Sld3 alone but folds into a helix coupled to the binding partner Cdc45 in the Sld3CBD–Cdc45 complex. This α8CTP is essential for binding to Cdc45, as its deletion inhibited cell growth in a previous study [23]. Furthermore, proline substitution for Cdc45 Ser242 (Cdc45-35), which interacts with L307 and T310 in Sld3CBD α8CTP (Figure 1C), conferred temperature-sensitive growth to yeast cells (Supplementary Figure 7). Although an increase in the number of copies of Sld3–Sld7 could weakly suppress cell growth defects, it did not recover the disrupted interaction.

The α9 (L337–E360) of Sld3CBD is the secondary Cdc45-binding region, which is located in a shallow dent formed by the C-terminal helix and a loop with the following β-strand (K520–Q531) of Cdc45 DHHA1 (Figure 1B, 1D and 1E). Three hydrophobic residues (I352, I355, and L356) in Sld3CBD α9 interact with the C-terminal sheet (L522, L527 and V529) and helix (L641, and L647) of Cdc45 DHHA1 (Figure 1D), while the side chains of two negatively charged residues (D344 and D348) form hydrogen bonds through a water molecule to the main and side chains of R523 on a loop of Cdc45 DHHA1 (Figure 1E). A disordered region, L527–V529, of Cdc45 DHHA1 in the isolated form forms a β-sheet in the Sld3CBD–Cdc45 complex and binds to the I352 and I355 of Sld3CBD α9. We substituted single, double, or ternary positively charged or hydrophilic residues at D344, D348, I352, I355, and L356 of Sld3CBD α9 (D344R/D348R: Sld3-2R, I352Y: Sld3-Y, I352S/I355S/L356S: Sld3-3S, I352E/I355E/L356E: Sld3-3E). Based on the CD spectra, we confirmed that these mutants retained the structural elements of Sld3CBD (Supplementary Figure 8). Double and ternary mutants Sld3-2R, Sld3-3S, and Sld3-3E eliminated Cdc45-binding affinity, whereas the single mutant Sld3-Y seemed to retain a faint interaction with Cdc45 (Figure 2A). We also substituted the single, double, or ternary residues R523, L522, L527, V529, L637, and L641 of Cdc45 on Sld3CBD α9 binding sites (L637S/L641S: Cdc45-IIS, L637E/L641E: Cdc45-IIE, L522S/L527S/V529S: Cdc45-IIIS, L522E/L527E/V529E: Cdc45-IIIE, and Cdc45-A: R523A). All Cdc45 mutants disrupted the binding between Sld3CBD and Cdc45, except for Cdc45-A, as surmised from the Sld3CBD–Cdc45 structure (Figure 2B). Although mutant Cdc45-A has reduced three hydrogen bonds with D344 of Sld3CBD, the remaining hydrogen-bond network maintains contact between Sld3CBD and Cdc45. Furthermore, in vivo genetic studies confirmed the importance of these Sld3 residues. By expressing Sld3-3S, Sld3-3E, and Sld3-2R in Sld3, the variant strains caused complete no growth, whereas the Sld3-Y strain maintains cell growth (Figure 2C), indicating that these residues are required to work together to bind Cdc45 and the loss of the Cdc45-binding affinity of Sld3 inhibits the recruitment of Cdc45 and the subsequent formation of the active replicative helicase CMG for DNA replication.

Figure 2.

In comparison with Cdc45 alone (huCdc45) or CMG-form (in CMG complex), domain DHHA1 of Cdc45 changed conformation significantly for binding to Sld3CBD (Supplementary Figure 4C). The loop I595–N604 in Cdc45 DHHA1 changed conformation to interact with Sld3CBD α8CTP. Subsequent the following helix α19 (F605–E615) rotated the C-terminus by 25 degrees, which altered the conformation of the next two β-strands (β12 and β13) in the mixed β-sheet (β9-β10-β11-β13-β12) (Supplementary Figure 3 and 4C), allowing Sld3CBD α8CTP to enter the binding pocket. Interestingly, the Sld3CBD-Cdc45 structure showed that the Sld3CBD binding site of Cdc45 completely differed from that of GINS and MCM bound to Cdc45, suggesting that the Sld3CBD, Cdc45 and GINS could bind to MCM together (Supplementary Figure 9A). Furthermore, we conducted a mutation analysis on two Cdc45 residues involved in binding to MCM (Cdc45 G367D) and GINS (Cdc45 W481R) [34], and found that these mutations did not disrupt the binding of Sld3CBD-Cdc45 complex (Supplementary Figure 9B).

Cdc45 recruitment to MCM DH by Sld3 with partner Sld7

Except for the Sld3 binding region DHHA1, the N-terminal domain of Cdc45 (Cdc45NTD) maintained a structure similar to that of the monomer, Sld3CBD–Cdc45 complex, and CMG complex. Therefore, we modelled Sld3CBD–Cdc45–MCM–dsDNA and SCMG–dsDNA (Figure 3) by superposing Cdc45NTD structures between Sld3CBD–Cdc45 and CMG dimers respectively, which were modelled by superposing Mcm2–7 structures between CMG (PDBID: 3JC6) [28] and MCM–dsDNA (PDBID: 5BK4) [35]. In the models, two Sld3CBDs are located in each monomer of the Cdc45–MCM dimer over 230 Å apart (Figure 3A).

Figure 3.

To investigate how Sld7–Sld3 brings Cdc45 to MCM DH, we attempted particle analysis for the Sld7–Sld3–Cdc45 complex using the purified recombinant Sld7–Sld3ΔC–Cdc45. The approximate molecular weight of Sld7–Sld3ΔC–Cdc45 was estimated to be >400 kDa according to a weight calibration of size-exclusion chromatography (Supplementary Figure 2B). Given that the molecular weight calculated from their amino acid sequences was 158 kDa, the purified complex should be a dimer. Considering the Sld7–Sld3ΔC–Cdc45 dimer should be a long rod-sharped molecule, the estimated value from SEC could be larger than theoretical values. Subsequently, using dynamic light scattering, the particle size (hydrodynamic diameter) of the tripartite complex was estimated to be around 232 Å (Figure 3B, Supplementary Figure 10), which is consistent with the distance of Sld3CBDs in the model of Sld3CBD–Cdc45–MCM dimer. Furthermore, considering that the domains of Sld7 (NTD: Sld3NTD-binding, CTD: self-dimerization) and Sld3 (NTD: Sld7NTD-binding, CBD: Cdc45-binding, CTD: MCM-binding) function independently [22,23], we estimated a dimer model of Sld7–Sld3–Cdc45, as shown in Figure 3B. Because the domains of Sld7–Sld3–Cdc45 are linked by long loops, the dimer of Sld7–Sld3–Cdc45 is flexible with long-shape.

Our SCMG–dsDNA model demonstrated that Sld3CBD neighbors Mcm2 and binds to Cdc45 on the opposite side of GINS binding (Figure 3C), indicating that Sld7–Sld3 bound to the Cdc45–MCM dimer did not contact GINS and could be remained for CMG formation. A recent study also reported a structure of the DONSON (Sld2 homolog) dimer with CMG (8Q6O), showing that the DONSON dimer delivers GINS to MCM and reconfigures the MCM motors in the double CMG [30]. The DONSON dimer is loading at the GINS site on MCM, which is the opposite of Sld3CBD in the SCMG-dsDNA model. Interestingly, in the Sld3CBD–Cdc45–MCM–dsDNA and SCMG–dsDNA model, we found that a part (S358-D383) of Sld3CBD containing the C-terminal of α9, a disordered fragment, and α10 was in contact distance with Mcm2CTD. In addition, the Sld3CBD-bound Cdc45 DHHA1 appeared to become close to Mcm2CTD. In contrast, the Cdc45 DHHA1 does not contact with Mcm2–7 or GINS in the CMG structure (Supplementary Figure 11). The conformational change in Cdc45 DHHA1 not only facilitates binding with Sld3CBD, but could also cause contact with Mcm2 without affecting the interaction between Cdc45 and GINS.

ssDNA binding affinity of Sld3 depended on complex formation with Cdc45 and Sld7

Previous studies showed that Sld3 binds directly to single-strand DNA fragments (ssARS1-2 and ssARS1-5) of ARS1 identified as an origin of DNA replication [26]. ARS1 was divided into three 80-bp segments: ARS1-12, ARS1-34, and ARS1-56, and these dsDNA segments could unwind into six signal-stranded DNA fragments with 80b length: ssARS1-1, 2, 3, 4, 5, and 6 (Supplementary Figure 12A). Given that Sld3 binds to Sld7 and Cdc45 on the MCM–DNA complex during CMG formation, we investigated whether Sld7 and Cdc45 affect the ssDNA-binding affinity of Sld3. Therefore, we performed an ssDNA binding assay using the Sld3CBD, Sld3CBD–Cdc45, Sld7–Sld3ΔC–Cdc45 and Sld7–Sld3ΔC (Sld3ΔC: binding part M1–K430 to Sld7 and Cdc45) complexes against different regions of ssARS1 for comparison. Cdc45 binds tighter to long ssDNA (>60 bases) with a litter sequence specificity [36]. Therefore, we fragmented each ssARS1 into fragments of 40 bases to prevent this nonspecific interaction (Figure 4A and Supplementary Figures 12B).

Figure 4.

To investigate the specificity of the ssDNA binding affinity of Sld3, we employed an electrophoretic mobility shift assay (EMSA) using non-denaturing PAGE (native-PAGE) with ssDNAs or proteins alone as controls. For the Sld3CBD and ssDNA mixtures at a molar ratio of 1:1, the band corresponding to ssDNA disappeared, indicating a strong binding affinity (Figure 4B). The ssDNA band remained (Figure 4B) and new bands corresponding to the ssDNA–protein complex appeared in CBB staining PAGE (Supplementary Figures 13) when the Sld3CBD–Cdc45 complex was mixed with ssDNA at the same ratio, indicating that the binding affinity of Sld3CBD–Cdc45 for ssDNA was lower than that of Sld3CBD alone. Additionally, the disappearance of ssDNA bands and the appearance of new bands corresponding to the ssDNA–protein complex correlated with an increase in the concentration of Sld3CBD–Cdc45 in the mixture.

In the presence of the Sld7–Sld3ΔC–Cdc45 complex, the band corresponding to ssDNA disappeared in the equimolar ratio of protein and ssDNA, similar to that observed for Sld3CBD (Figure 4B and Supplementary Figures 13). This result demonstrates that the presence of Sld7 and Sld3NTD could increase the ssDNA-binding affinity to a level comparable to that of Sld3CBD. Also, Sld7–Sld3ΔC showed a ssDNA-binding affinity similar to that of Sld7–Sld3ΔC–Cdc45, implying that ssDNA-binding of Sld7–Sld3ΔC is independent of Cdc45. Furthermore, the results revealed no significant difference in binding affinity between the 40-base fragments of ssARS1-2-1, 2, 3 and ssARS1-5-1, 2, 3 for Sld3CBD, Sld3CBD–Cdc45, Sld7–Sld3ΔC–Cdc45, and Sld7–Sld3ΔC, indicating that there is no stronger binding-region specific to ssARS1-2 or ssARS1-5 fragments. For sequence specificity, we also analyzed other fragments of ssARS1-1, 3, 4, and 6 and dsARS1, and all of them showed no binding (Supplementary Figure 14). Such binding experiments can serve as a negative control, confirming that the absence or presence of ssDNA bands is contingent on protein binding.

The surface charge of the Sld3CBD–Cdc45 structure, shows that Cdc45 covers the main positive charge region of Sld3CBD α8CTP (Supplementary Figure 15A), which may weaken the binding affinity of Sld3CBD–Cdc45 to ssDNA. Conversely, on the Sld3–Sld7 structure, there is a large positive charge area strip on Sld7NTD (Supplementary Figure 15B). Considering that ssARS1 is unwound from dsARS1 by the activated helicase CMG complex formed after loading Cdc45 and GINS, Sld3–Sld7 having a stronger ssARS1-binding affinity may provide an advantage for the dissociation of Sld7–Sld3 from the CMG complex.

Discussion

As a central regulator of helicase CMG formation, Sld3 has attracted interest in studies aiming to understand the initiation of DNA replication. Owing to the lack of structural information on Sld3 complexed with Cdc45 or MCM, how Sld3 regulates helicase activation with other factors remains unknown. Here, we present the structure of the Sld3CBD–Cdc45 complex and particle size of the Sld7–Sld3–Cdc45 complex to examine the detailed mechanisms underlying CMG formation.

Sld3 is highly conserved in eukaryotes, whereas its functional homolog/counterpart in metazoans Treslin (also known as Ticrr) has a distinct size and sequence, except for Sld3CBD. Sequence alignment of Sld3CBD among Sld3 and Treslin revealed that all Cdc45-binding residues in α8 and α9 identified in our study were almost conserved or exhibited conserved changes (Supplementary Figures 4 and 5), indicating that these regions provide a similar mode of interaction between Sld3CBD and Cdc45 in the regulation of metazoan DNA replication. Therefore, we hypothesize that they load Cdc45, as observed in yeast Sld3 and Sld7.

Sld3CBD and Cdc45 significantly change their conformation to bind to each other in the Sld3CBD–Cdc45 complex compared with their conformation as monomers. Interestingly, the conformational changes in Cdc45 DHHA1 upon binding to Sld3CBD also caused contact with Mcm2NTD in the Sld3CBD–Cdc45–MCM–dsDNA and SCMG–dsDNA models, whereas DHHA1 interacted with neither MCM nor GINS in the CMG structure. Taken together, Sld3CBD contacts Mcm2NTD in the Sld3CBD–Cdc45–MCM–dsDNA and SCMG–dsDNA models and then seems to play a guiding role in helping Cdc45 bind to MCM at the right position. Sld3CBDs in each monomer of the Sld3CBD–Cdc45–MCM dimer were located at a distance of more than 230 Å, which is consistent with the results of a particle size analysis of the Sld7–Sld3ΔC–Cdc45 complex off MCM DH in solution. Therefore, we propose that a binding manner of Sld7–Sld3–Cdc45 in a flexible long-sharped dimer Cdc45–Sld3–(Sld7)₂–Sld3–Cdc45 off/on MCM DH is advantageous for efficiently recruiting two Cdc45 molecules to an MCM DH, consequently leading to the formation of a pair of CMG helicases.

In the SCMG–dsDNA complex model, Cdc45 bound to Sld3CBD, MCM, and GINS on different sides (with contact surfaces of 6.7, 4.8, and 5.1% of the total Cdc45 surface, respectively). Our structure of Sld3CBD-Cdc45 and model shows that these bindings occur at different sites of Cdc45, suggesting that Sld3CBD, Cdc45 and GINS could bind to MCM together, and do not directly interfere with each other. The competition of Sld3 and GINS for binding to Cdc45 or Cdc45-MCM reported by using in vitro binding experiments [26], may be caused by the conformation change of Cdc45 DHHA1 or without other initiation factors. In particular, Sld3 and GINS bind to Cdc45 and MCM rings at opposite positions (Mcm2–4–6 and Mcm5–3–7 vs. Sld3 and GINS, respectively), suggesting that the GINS-recruitment protein should cross a long-distance in an MCM monomer or MCM DH to recognize the phosphorylation site (T600 and S622) of Sld3. Furthermore, our SCMG–dsDNA model revealed that Sld3CBD on CMG appears to contact an N-terminal helix of Mcm2CTD and Sld3CTD could extend to bind to Mcm4NTD through interaction with the NTDs of Mcm2, and 6, rather than the CTD tier (Supplementary Figure 16), indicating that the binding of Sld3–Sld7 to MCM does not appear to affect the ability of AAA+ motors to open and close the MCM ring with a gap between Mcm2CTD and Mcm5CTD [27, 28, 35, 38]. Taking our findings and those of previous studies together, we propose a detailed process for helicase CMG formation from inactive MCM, as depicted in Figure 5 A–C, in which Sld7–Sld3 brings Cdc45 into the MCM as a Sld7–Sld3–Cdc45 dimer (Cdc45–Sld3–[Sld7]₂–Sld3–Cdc45), which remains until GINS loading.

Figure 5.

The following inquiry concerns the dissociation of Sld3 and other factors. Interestingly, the mutant analysis showed that disrupting just one binding site between Sld3CBD and Cdc45 could cause the dissociation of Sld3CBD and Cdc45, indicating that the binding between Sld3CBD and Cdc45 could be broken easily. Our binding analysis of ssARS1 fragments to Sld3CBD, Sld3CBD–Cdc45, and Sld7–Sld3ΔC–Cdc45 demonstrated the sequence specificity to ssARS1-2 and ssARS1-5 fragments, not others including dsARS1. Considering that ssARS1-2 and ssARS1-5 are on both sides of MCM DH-bound dsDNA at replication origin [39], the origin unwinding by CMG generates ssDNA and further sequesters the Sld7-Sld3 complex onto ssDNA to remove Sld7-Sld3 from CMG. As a bridge protein, Sld3 recruits Cdc45 to MCM, and phosphorylated Sld3 regulates the recruitment of GINS [40]. Finally, the release of Sld3 and Sld7 from CMG could be associated with the binding of ssARS1 unwound by CMG and may also be related to the dissociation of Dpb11–Sld2 from CMG (Figure 5 D, E) [29, 41]. Further structural investigation of Sld3–Sld7–Cdc45–MCM in the GINS-recruiting stage is required to validate the complete CMG formation process.

In conclusion, our structural and biochemical studies of Sld3CBD–Cdc45 revealed a detailed process of CMG formation and the subsequent release of the central regulator Sld3 and other factors associated with single-stranded DNA, leading to a deeper understanding of the initialization mechanism of DNA replication.

Materials and Methods

Preparation of proteins

The C-terminal His-tagged Sld3CBD of S. cerevisiae was expressed in Escherichia coli and prepared as previously described [23]. For co-expression of S. cerevisiae Sld3CBD (S148–K430) and Cdc45 (M1-L650) (Sld3CBD–Cdc45) in E. coli, the Sld3CBD DNA fragment containing the His₆-tag (LEHHHHHH) at the C-terminus and the Cdc45 DNA fragment were cloned in the co-overexpression vector pETDuet-1 between the NcoI and SalI restriction sites and the NdeI and XhoI restriction sites, respectively. The primer sequences for Sld3CBD–Cdc45 are listed in Supplementary Table 1.

To confirm the interactions obtained from the Sld3CBD–Cdc45 structure, we constructed four types of single or multi-mutants for Sld3CBD (Sld3-3S: I352S/I355S/L356S, Sld3-3E: 352E/I355E/L356E, Sld3-2R: D344R/D348R, and Sld3-Y: I352Y), and three types of single or multi-mutants for Cdc45 (Cdc45-IIIS: L522S/L527S/V529S, Cdc45-IIIE: L522E/L527E/V529E, Cdc45-IIS: L637S/L641S, Cdc45-IIE: L637E/L641E, and Cdc45-A: R523A) using the Quick Change site-directed mutagenesis method and the inverse PCR method with pETDuet-1–Sld3CBD–Cdc45 as template DNA. The primer sequences for the mutant strains are listed in Supplementary Table 1.

As S. cerevisiae Sld7–Sld3–Cdc45 could not be co-overexpressed in E. coli, we attempted to clone it from several other fungal sources. The complex of Sld7 (M1-T268), Sld3ΔC (M1-K430, truncated C-terminal domain), and Cdc45 (M1-I666) from the budding yeast Kluyveromyces marxianus, which belongs to the same family as S. cerevisiae (Sld7–Sld3ΔC–Cdc45), was cloned into pETDuet-1. We also added a His₆-tag (LEHHHHHH) to the C-terminus of Sld3ΔC. The primer sequences for Sld7–Sld3ΔC and Sld7–Sld3ΔC–Cdc45 are listed in Supplementary Table 1. As Cdc45 mutants can lose the ability to bind to Sld3, we overexpress Sld7–Sld3ΔC by using multi-mutants of Cdc4 (Cdc45-IIS) in Sld7–Sld3ΔC–Cdc45. The Ser-substitution residues (L654S/L658S) of Cdc45-IIS from K. marxianus were selected based on sequence conservation. We constructed mutant Sld7–Sld3ΔC–Cdc45-IIS using the Quick Change site-directed mutagenesis method and the inverse PCR method with pETDuet-1–Sld7–Sld3ΔC–Cdc45 as template DNA. The primer sequences for the mutant strains are listed in Supplementary Table 1.

To overexpress Sld3CBD, Sld3CBD–Cdc45, Sld3CBD–Cdc45 mutants, Sld7–Sld3ΔC–Cdc45 and Sld7–Sld3ΔC, the vector of pET26Sld3CBD, pETDuet-1-Sld3CBD–Cdc45, pETDuet-1-Sld3CBD–Cdc45 mutants, pETDuet-1-Sld7–Sld3ΔC–Cdc45 or pETDuet-1-Sld7–Sld3ΔC was transformed into E. coli strain BL21 (DE3) through electroporation, followed by preculturing in 5 ml of Luria–Bertani medium containing 100 μg/mL ampicillin at 310 K overnight. The culture was then transferred to 3 L Luria–Bertani medium and grown until the OD₆₀₀ reached 0.6. After cooling the culture for 30 min on ice, overexpression of each sample was induced by the addition of isopropyl-β-D-1-thiogalactopyranoside to a final concentration of 0.5 mM, and then cells were grown for an additional 12 h at 293 K. Cells were harvested by centrifugation at 4,000 g for 20 min at 283 K and then resuspended in a buffer containing 20 mM Tris-HCl pH 7.5, 300 mM NaCl, 10% glycerol, 0.1 mg/ml DNase, and 1× protease-inhibitor cocktail (cOmplete EDTA-free; Roche, Basel, Switzerland).

The harvested cells were crushed through sonication and centrifuged at 40,000 g for 30 min at 283 K. After filtration through a 0.45-µm filter (Sigma-Aldrich/Merck Millipore, Burlington, MA, USA), the supernatant was then loaded onto the HisTrap HP column equilibrated with buffer A [20 mM Tris-HCl, pH 7.5, 300 mM NaCl, 10% glycerol], and the column was washed with buffer A. The His-tagged target protein was eluted with imidazole at 20% (100 mM), 30% (150 mM), and 50% (250 mM), followed by a linear gradient of 250–500 mM imidazole in buffer B [20 mM Tris-HCl, pH 7.5, 300 mM NaCl, 10% glycerol, 500 mM imidazole]. We analyzed the fractions using sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE). Except for mutants of Sld3CBD–Cdc45, the pooled fractions were then purified through size-exclusion chromatography using a buffer A-equilibrated Superdex 200 16/60 column (GE HealthCare, Chicago, IL, USA). Protein purity was confirmed through SDS-PAGE. The purified samples were concentrated to 10 mg/ml and stored at 193 K.

Crystallization and diffraction data collection

The crystallization screening of Sld3CBD–Cdc45 was performed at 293 K using the sitting-drop vapor diffusion method. The drop was a mixture of 1.0 µl of a 10 mg/ml protein solution with the equivalent volume of reservoir buffer from the commercial crystallization kits (JCSG core I-IV, classics, classics II, PEGs, PEGsII, and MPD suite from Qiagen, Venlo, the Netherlands). The initial needle- and plate-shaped crystals appeared under eight and one conditions, respectively. After optimizing the conditions by altering the type and concentration of precipitant, the salt reagent, and the pH of the buffer, the best crystals were obtained in a solution of 0.2M sodium acetate, 0.1M Bis-Tris propane (pH 6.5: pH 8.5 = 3:7), 20%(w/v) PEG3500, and a drop containing 2.0 µl of a 10 mg/ml protein solution mixed with the equivalent volume of reservoir buffer.

Diffraction data were collected using a beamline BL-17A at the Photon Factory, Tsukuba, Japan. Before data collection, the crystals were cryoprotected by soaking them in a reservoir buffer supplemented with 5% (v/v) glycerol and then flash-cooled under a nitrogen gas stream at 100 K. XDS/XSCALE was used to index, integrate, scale, and merge the dataset [42]. Supplementary Table 2 summarizes the data collection and data processing statistics.

Structure determination and refinement

The structure of Sld3CBD–Cdc45 was determined using the molecular replacement method with the AutoMR wizard in Phenix [43, 44] and human Cdc45 (PDBID: 5DGO) [31] and Sld3CBD (PDBID:3WI3) [23] structures were used as search models. Several rounds of refinement were performed using the Phenix_refine program in Phenix, interleaved with manual building and fitting according to the electron density maps of 2Fo-Fc and Fo-Fc using the Coot program [44, 45]. Supplementary Table 2 shows the final refinement statistics and geometry defined by MolProbity [46]. All structural diagrams were generated using PyMol [47].

Mutant analysis of Sld3 and Cdc45

To analyze the binding sites of Sld3CBD-Cdc45, in conjunction with Cdc45 binding sites to MCM and GINS, we performed co-express pull-down binding assay. We constructed four variants of Sld3CBD and five variants of Cdc45 according to the binding information from the obtained Sld3CBD–Cdc45 structure. We co-express all mutations of Sld3CBD-Cdc45 (Sld3CBD-C-histag) and load them onto the HisTrap HP column under the same conditions as in [Preparation of proteins]. After extracting the samples through Ni-affinity chromatography, we concentrated each eluted sample to 0.005 mg/mL (by nanodrop A280) by centrifugation and checked the binding status of the mutants using SDS-PAGE. Both Sld3CBD and Cdc45 should be observed in the elution group if they formed a complex. The overexpress level of the Cdc45 was checked by - IPTG and +IPTG.

To check whether mutations affected the structure, we performed circular dichroism (CD) spectrometry measurement of Sld3mutants–Cdc45. CD spectra were collected using a J-805 spectropolarimeter (JASCO, Tokyo, Japan) in a quartz cell with a path length of 1 mm in an atmosphere of N₂ at 298K. For CD measurements, the samples were dialyzed in a buffer [20 mM Tris-HCl, pH 7.5, 50 mM NaCl] and adjusted to a concentration of 0.5 mg/mL through absorption. CD spectra for the wavelength range of 190–300 nm were obtained by averaging the results of four scans. The results are given in molar ellipticity per residues [θ] mrw (× 10⁻³) vs the wavelength/nm [48]. The secondary structures of each sample were estimated using the K2D3 method [49]. For wild-type Sld3CBD, we calculated secondary structures from the obtained Sld3CBD–Cdc45 structure.

Growth of mutant cells

The isolation and plasmid shuffling of the temperature-sensitive yeast strain YYK13 (for Sld3 mutant analysis) and Cdc45-35 (for Cdc45 mutant analysis) have been described in a previous study [12]. To investigate Sld3 mutants, strain YYK13 was transformed with the YCplac22 plasmid containing multiple variants of SLD3 mutant genes. The transformants were streaked on SD plates lacking Trp and Leu (SD-Trp, Leu) and SD-Trp, Leu containing 5-fluoroorotic acid (5-FOA-Trp, Leu), and then incubated at 298 K for 3 days. To analyze the Cdc45 mutant, strain Cdc45-35 containing the CDC45 mutant gene was re-transformed with the YEplac195 plasmid containing SLD3 and the YEplac122 plasmid containing SLD7. The transformants were streaked on yeast extract–peptone–dextrose plates and incubated at 298 or 305 K for 3 days.

Modelling of complexes

We constructed the models of Sld3CBD–Cdc45–MCM–dsDNA (Sld3CBD–Cdc45–MCM dimer complexed with dsDNA) and SCMG–dsDNA (SCMG dimer complexed with dsDNA) using a two-step process. First, the structure of the MCM–NTDs in the CMG monomer (PDBID: 3JC6) [28] was superimposed on that of MCM DH complexed with dsDNA (5BK4) [35] to obtain a dimer of Cdc45–MCM–GINS complexed with dsDNA (Cdc45–MCM–GINS–dsDNA). Next, the models of Sld3CBD–Cdc45–MCM–dsDNA and SCMG–dsDNA were obtained by superimposing Cdc45NTD from Sld3CBD–Cdc45 onto Cdc45–MCM–GINS–dsDNA.

Dynamic light scattering

We estimated the particle size of the Sld7–Sld3ΔC–Cdc45 complex using dynamic light scattering in terms of peak size [50]. To avoid the effect of concentration, we measured four samples of the Sld7–Sld3ΔC–Cdc45 complex. Each sample was overexpressed independently and purified through size-exclusion chromatography. All samples were concentrated to 10 mg/ml, and stored at 193 K. Before measurement, protein samples were dialyzed in a buffer containing 20 mM Tris-HCl pH 7.5 and 300 mM NaCl and then filtered through a 0.22-μm filter. The pure buffer was measured as a background control and 10 μM lysozyme was measured as a standard control. Dynamic light scattering data were collected and analyzed using a Malvern Zetasizer Nano-ZS instrument (Malvern Panalytical, Malvern, UK) for 0.2–0.5 mg/ml protein samples 500 μl in a microquartz cuvette (Malvern-ZEN0112). An automatic duration model was used to collect data. The Zetasizer 6.20 was utilized for data analysis from three measurements to estimate the particle size of each sample.

Electrophoretic mobility shift assay for ssDNA binding

Sld3 binds to single strands of DNA (ssARS1-2 and ssARS1-5) [26]. To determine the specificity of the ssDNA binding affinity of Sld3, we conducted an electrophoretic mobility shift assay (EMSA) using non-denaturing PAGE (native-PAGE) [51]. Given that Cdc45 binds ssDNA with a nonspecific sequence at lengths greater than 60 bases [36], we designed three fragments of ssDNA 40 bases in length (first half 1–40 bp, second half 41–80 bp, and middle half 21–60 bp) for each ssDNA 80-bp segment. All ssDNA fragments of S. cerevisiae were synthesized by Sigma-Aldrich. The 40-bp dsDNA fragments (dsARS1-34_1:ssARS1-3_1/ssARS1-4_2) were converted by annealing them using the PCR protocol and then checked through polyacrylamide gel electrophoresis. The loaded samples were incubated overnight at 277 K in TMK buffer (20mM Tris-HCl pH8, 100mM MgCl₂, 200mM KCl) containing synthesized ssDNA and varying amounts of proteins at ssDNA: protein molecular ratios of 1:0, 1:1, 1:2, 1:4, and 0:1 with 10 pM as 1 unit. After incubation, the mixtures were loaded onto a polyacrylamide gel (5% (w/v) acrylamide (39:1), 10% 10 × running buffer (0.25 M Tris, 1.92 M Glycine), 0.1% Ammonium peroxodisulphate, 0.06% (v/v) TEMED) without denaturing (native-PAGE). EMSA was performed at 10 mA/200 V per gel for 40 min at 277 K in 1 × running buffer. After electrophoresis, the reaction products were visualized using SYBR safe (SYBR safe:1 × running buffer = 0.0001:1) to stain the ssDNA and Coomassie brilliant blue (G-250) to stain the proteins. We repeated the EMSA experiments three or more times to confirm the readability. Considering the functional similarity of ARS1-core, the EMSA of Sld7–Sld3ΔC and Sld7–Sld3ΔC–Cdc45 of K. marxianus used ssDNA fragments of S. cerevisiae.

Data Availability

Structures and associated experimental data for Sld3CBD-Cdc45 complex (8J09) were deposited in the Protein Data Bank Japan.

Acknowledgements

The authors thank Mr. Naofumi Sakurai for his help in the purification of proteins. We would like to thank the beamline staff of the Photon Factory and SPring-8 for collecting X-ray diffraction data (Proposal No. 2016A2562, 2017A2551, and 2018A2508)

Additional information

Author Contributions

Hao Li: Investigation, Formal analysis, Writing—original draft & editing. Izumi Ishizaki: Investigation, Formal analysis. Koji Kato: Formal structure analysis. XiaoMei Sun: Investigation. Sachiko Muramatsu: Investigation. Hiroshi Itou: Investigation, Writing—review & editing. Toyoyuki Ose: Formal analysis. Hiroyuki Araki: conceptualization, Formal analysis, Writing—review & editing. Min Yao: conceptualization, Formal analysis, Methodology, Writing— original draft & editing.

Funding

This work was supported by KAKENHI Grant-in-aid for Scientific Research(B), Japan Society for the Promotion of Science [No. 21H01754 to M. Y.]; and Platform Project for Supporting Drug Discovery and Life Science Research, Japan Agency for Medical Research and Development [JP18am0101071, JP19am0101083].

Additional files

Supplemental Figures and Tables