Cdk1-mediated phosphorylation of Cdc7 suppresses DNA re-replication

ABSTRACT

To maintain genetic stability, the entire mammalian genome must replicate only once per cell cycle. This is largely achieved by strictly regulating the stepwise formation of the pre-replication complex (pre-RC), followed by the activation of individual origins of DNA replication by Cdc7/Dbf4 kinase. However, the mechanism how Cdc7 itself is regulated in the context of cell cycle progression is poorly understood. Here we report that Cdc7 is phosphorylated by a Cdk1-dependent manner during prometaphase on multiple sites, resulting in its dissociation from origins. In contrast, Dbf4 is not removed from origins in prometaphase, nor is it degraded as cells exit mitosis. Our data thus demonstrates that constitutive phosphorylation of Cdc7 at Cdk1 recognition sites, but not the regulation of Dbf4, prevents the initiation of DNA replication in normally cycling cells and under conditions that promote re-replication in G2/M. As cells exit mitosis, PP1α associates with and dephosphorylates Cdc7. Together, our data support a model where Cdc7 (de)phosphorylation is the molecular switch for the activation and inactivation of DNA replication in mitosis, directly connecting Cdc7 and PP1α/Cdk1 to the regulation of once-per-cell cycle DNA replication in mammalian cells.

Introduction

In eukaryotes, multiple origins are required to replicate the entire genome in the time afforded by the cell cycle. The challenge of utilizing multiple origins is ensuring that each origin fires only once per cell cycle, defects of which may lead to aneuploidy¹ and the development of diseases such as cancer.² At least three cooperative and overlapping mechanisms exist to strictly control the once-per-cell cycle replication of the entire genome in metazoan cells.¹ In particular, Cdk-mediated protein phosphorylation is utilized by cells to prevent DNA re-replication, in which much of the focus has been on the role of Cdk2. However, chemical biology and genetic evidence suggest that the mitotic Cdk (i.e., Cdk1/cdc2/cdc28) plays an important role in preventing DNA re-replication.^1,3 For example, Cdk1 may prevent re-replication in fission yeast, budding yeast and metazoan cells by removing replication proteins from the chromatin and nucleus through phosphorylation of these proteins. In all cases where specific phosphorylation sites have been identified, multiple phosphorylation events are required to remove the protein from chromatin, suggesting that there is a common mechanism for Cdk1-dependent hyperphosphorylation in the regulation of DNA replication.^4-7

Dbf4 degradation at the end of mitosis has previously been suggested as the major regulation mechanism in limiting the activity of the DNA replication activator Dbf4/Drf1-dependent kinase (DDK; Cdc7) function to S phase, at least in budding yeast.^8-11 This proposal was mainly based on the following observation: Although the level of yeast Cdc7 protein is relatively constant throughout the cell cycle, that of Dbf4 fluctuates as it is low in G1 and the post-metaphase/anaphase transition but high during S and G2 phases.^8-11 Therefore, it was suggested that Dbf4 binds to origins and activates Cdc7 to initiate DNA replication when the level of Dbf4 is high.^12-14 As a cell reaches mitotic phase, Dbf4 is degraded by the anaphase promoting complex (APC) to prevent DNA re-replication.^8-11 However, more recent evidence shows that not all yeast Dbf4 molecules are degraded at the end of metaphase, suggesting that a pool of Dbf4 may present in the cell throughout the cell cycle.¹⁵

Published data indicated that human DDK may be regulated in a similar fashion to that of budding yeast since both the organisms showed a similar pattern of Dbf4 expression throughout the cell cycle.^16,17 However, the levels of Xenopus Dbf4 and Cdc7 proteins are found to be constant throughout the cell cycle.^18,19 Interestingly, the Xenopus Dbf4 remains chromatin bound throughout the cell cycle, and Dbf4 on the chromatin recruits Cdc7 to the position upon pre-RC formation,^18,19 suggesting that the regulation of DDK in Xenopus is based on the chromatin association and disassociation of Cdc7 rather than Dbf4 proteolysis.

Data from in vitro studies suggested previously that mammalian Cdc7 is phosphorylated by a Cdk.²⁰ Interestingly, however, the same authors also found that Cdc7 kinase activity was not necessarily stimulated by Cdks.²⁰ In contrast, neither budding yeast nor Xenopus Cdc7 requires Cdk-mediated phosphorylation for DNA replication.^21,22 Finally, other groups have successfully used recombinant human DDK in in vitro kinase assays in the absence of Cdk phosphorylation.^23,24 Thus, the significance and function of Cdk-dependent Cdc7 phosphorylation in vivo is still to be unambiguously determined.

In this study, we revisit Cdk-dependent phosphorylation on human Cdc7. We find that Cdc7 is phosphorylated in prometaphase in both drug-arrested and normally cycling cells. Furthermore, we have identified that Cdc7 is phosphorylated on multiple sites, and phosphorylation can be abrogated by genetically modifying Cdk1 consensus sequences. The Cdk1-dependent phosphorylation of Cdc7 removes it from origins, resulting in the prevention of replication initiation, although Dbf4 remains associated with origin/chromatin throughout the cell cycle. Furthermore, the phosphorylation of Cdc7 prevents SP600125-induced re-replication in G2. Finally, we have also found that Cdc7 is dephosphorylated by PP1α as a cell exits mitosis. Taken together, our data indicates that the phosphorylation of Cdc7 by mitotic Cdk plays an essential role in ensuring that the initiation of DNA replication is confined to S phase by preventing Cdc7 from interacting with origin-associated Dbf4 during G2/M, and that DNA replication in mammalian cells is regulated by Cdk1-mediated phosphorylation and PP1α-mediated dephosphorylation of Cdc7 in a cell cycle dependent manner, rather than by the cell cycle-dependent fluctuation of Dbf4 protein levels.

Results

Cdc7 is phosphorylated during prometaphase in a Cdk1-dependent manner

To gain insights into the mechanism of Cdc7 regulation during the cell cycle progression, we examined its phosphorylation in asynchronously growing HeLa S3 cells (async) and those arrested in S phase with hydroxyurea (HU, 2 mM for 18 h) or in prometaphase with nocodazole (NZ, 50 ng/mL for 18 h). Our data obtained from polyacrylamide gel electrophoresis (PAGE) and subsequent Western blotting showed that Cdc7 is present as a slower migrating isoform in NZ-arrested cells when compared to those in asynchronous and synchronized by HU (Fig. 1A). It should be noted that cells synchronization in prometaphase were confirmed by both flow cytometry (Fig. S1) and the enrichment of phosphorylated histone H3 at Ser10 (pHistone H3^S10) (Fig. 1A).²⁵ As shown in Fig. 1B, the slower migrating Cdc7 band was not shown when a sample was treated with lambda phosphatase, indicating that the band contains phospho-Cdc7. To make sure that the phosphorylation of Cdc7 in prometaphase is not an artifact caused by NZ, we examined the Cdc7 phosphorylation after HeLa S3 cells were released from G1/S arrest by double thymidine (DT) treatment (Fig. S1B). Cdc7 phosphorylation was observed by 8 h post-release, when most cells reach G2/M (Fig. 1C). Taken together, our data indicate that Cdc7 is hyperphosphorylated at or near prometaphase.

Figure 1.

Cdc7 is phosphorylated during mitosis in a mitotic Cdk-dependent manner. (A) Cdc7 is modified in a cell-cycle dependent manner. Extracts from normally cycling HeLa cells (async) or cells arrested at G1/S border with hydroxyurea (HU) or at prometaphase with nocodazole (NZ) were subjected to Western blot analysis using antibodies specific for Cdc7, pHistone H3^S10 or tubulin. (B) Cdc7 is phosphorylated at prometaphase. Extracts from cells arrested at prometaphase were treated with λ-phosphatase (PPase) and processed for Western blotting using an anti-Cdc7 antibody. Histone H1 was used as a loading control. (C) Cdc7 is phosphorylated in G2/M. Extracts from cells collected 8 h post-double thymidine (DT) arrest-and-release were treated with lambda phosphatase, and processed for Western blotting using an anti-Cdc 7 antibody. γ-tubulin was used as a loading control. (D) Cdc7 is phosphorylated in a Cdk1-dependent manner during mitosis. HEK293T cells arrested by NZ were treated with 25 µM roscovitine for 2 h, followed by Western blotting with an anti-Cdc7 antibody. “-” denotes a sham control treated with DMSO alone. β-tubulin was used as a loading control. (E) Cdc7 contains at least five Cdk1 minimal consensus sequence motifs (S/T-P) throughout the protein. Residues are indicated by identity (S, serine; T, threonine) and positions in the primary sequence. (F) Cdc7 phosphorylation sites by Cdk1 were confirmed by a recombinant DNA approach. A plasmid construct containing wild type Cdc7 (WT) or those containing mutations at all five potential Cdk1 phosphorylation sites were transiently transfected into HEK293T cells, followed by Western blotting with an anti-Flag tag antibody. Cdc7^×5A and Cdc7^×5D are Cdc7 mutated at all five potential Cdk1 phosphorylation sites to alanine and aspartic acid residues, respectively. (G) Cdc7^×5A resembles dephosphorylated Cdc7^WT. Lysates from HEK293T cells transfected with either mycFLAG-Cdc7^WT or Cdc7^×5A were subjected to Western blot analysis using anti-FLAG antibody. (H) Cdc7 is phosphorylated on multiple residues. HEK293T cells transfected with mycFLAG-Cdc7^WT or mycFLAG-Cdc7^×5A were arrested at prometaphase by NZ, followed by separation of proteins by two-dimensional PAGE and, subsequently, Western blotting with an anti-FLAG antibody. Black arrows indicate Cdc7 isoforms. Predicted pH values of phosphorylated mycFLAG-Cdc7 shown are according to the ProMoST tool.²⁹ Note that endogenous Cdc7 is still present. (I) GFP-Cdc7 phospho-alleles show activity in an in vitro kinase assay. ATP:ADP conversion was compared against GFP-Cdc7^WT. The population of different GFP-Cdc7 alleles showed a significant difference (one way ANOVA, F=14.15, df = 4, 18, p < 0.05, α level = 0.05) with * denoting significant different population via a Dunnett's test compared to GFP-Cdc7^WT. Data are mean ± SEM, n ≥ 3.

Since it was previously reported by an in vitro study that Cdc7 may be phosphorylated by Cdk1,²⁰ and because Cdk1 is a mitotic Cdk in mammalian cells, we asked if we could inhibit Cdc7 phosphorylation by inhibiting Cdk1 in prometaphase. We found that phospho-Cdc7 band was no longer observed when HeLa S3 cells arrested at prometaphase by NZ were treated with the Cdk inhibitor roscovitine at a concentration of 25 µM for 2 h. Therefore, we have concluded that Cdc7 is phosphorylated in a Cdk1-dependent manner during mitosis (Fig. 1D).

Computer-aided search identified five potential Cdk1 phosphorylation sites on human Cdc7: Ser16, Ser302, Thr376, Thr472 and Thr503 (Fig. 1E). Using PCR-based site-directed mutagenesis, we altered the amino acid sequences of potential Cdk1 phosphorylation sites to either alanine or aspartic acids as follows: Cdc7^T376A (single), Cdc7^T472A/T503A (double), Cdc7^{S16A/S302A/T376A} (triple), Cdc7^×5A (phospho-dead) and Cdc7^×5D (phosphomimetic) alleles. We verified the expression of different alleles of Cdc7 recombinant proteins by Western blot analysis using extracts from HEK293T cells that had been transiently transfected with mycFlag-tagged plasmid constructs, among which three major alleles containing wild-type Cdc7 (Cdc7^WT), phospho-dead Cdc7^×5A and phosphomimetic Cdc7^×5D are shown in Fig. 1F. As expected, the migration patterns of Cdc7^×5A and Cdc7^×5D were similar to those of non-phosphorylated and phosphorylated Cdc7, respectively. When hyperphosphorlated Cdc7^WT was treated with lambda phosphatase, the slower migrating isoform was no longer observed, and only band observed was at the same molecular weight as Cdc7^×5A (Fig. 1G).

Multiple large-scale proteomic analyses have indicated that Cdc7 is phosphorylated at Ser16, Ser302, and Thr503 in prometaphase,^26-28 which is consistent with our prediction (Fig 1E.). To further confirm it, we determined the extent of hyper-phosphorylation in vivo by two-dimensional PAGE with mycFLAG-tagged Cdc7^WT and Cdc7^×5A expressed in HEK293T cells. In Cdc7^WT, at least five isoforms are present (Fig. 1H), with the most acidic isoform found to have a pI consistent with seven predicted phosphorylation events, based on the prediction by ProMoST.²⁹ Most of these events are abrogated in Cdc7^×5A. However, three acidic isoforms are detected even in Cdc7^×5A (Fig. 1H), suggesting that there may exist up to three additional phosphorylation sites on Cdc7. This may be due to Cdc7 autophosphorylation¹⁴ or phosphorylation by other kinases such as polo-like kinase (Plk), as large scale proteomics studies indicated that Plk phosphorylates Cdc7 at Serine 27 and 277 in prometaphase.^26,28,30-33 In any event, our data demonstrate that Cdc7 is phosphorylated at multiple sites in a mitotic Cdk-dependent manner around prometaphase.

We next examined the kinase activity of Cdc7^T376A along with other relevant recombinant proteins, since it had been inferred but not directly shown that phosphorylation at Thr376 is required for the activation of Cdc7 kinase, based on homology to the primary structures of other serine/threonine kinases.²⁰ We noted that the homology-based rationale may not have a solid logical ground since recent crystallization studies indicated that Cdc7 does not have a threonine in its activation segment.²³ Instead, the study showed that a glutamic acid (Glu217) has replaced the threonine, making Cdc7 an atypical member of the serine/threonine kinase family.²³ Using a luciferase-based assay that has previously been successfully utilized to assay Cdc7 activity,²⁴ we compared the relative kinase activities of Cdc7^WT, Cdc7^×5A, Cdc7^×5D and Cdc7^T376A in relation to the Cdc7^1-431 negative control, an allele that is deleted for its Dbf4 interacting domain.³⁴ Under our experimental conditions, all three mutant alleles (Cdc7^×5A, Cdc7^×5D and Cdc7^T376A) showed substantial kinase activity close to the level of wild type, although the difference in kinase activities between the mutants and Cdc7^WT are still significant (Fig. 1I, p < 0.05). Importantly, the kinase activities of the three mutant alleles were not significantly different each other (Fig. 1I). This result indicates that the kinase activity of Cdc7 is somewhat attenuated but not abrogated even in the presence of multiple amino acid changes to the primary protein sequence. Furthermore, our data demonstrate that Thr376 is not required for Cdc7 kinase activity (Fig. 1I).

Cdk-dependent phosphorylation regulates Cdc7 interaction with origins

The finding that Cdc7 is phosphorylated by mitotic Cdk during mitosis led us to hypothesize that non-phosphorylated Cdc7 may be the isoform activating DNA replication. We therefore sought to determine the phosphorylation status and timing of dephosphorylation of Cdc7 as HeLa S3 cells exit mitosis. We found that Cdc7 started to become dephosphorylated approximately 2 h post-NZ (Fig. 2A) when cells begin to exit mitosis, and the majority of Cdc7 is dephosphorylated by 6 h post-NZ (Fig. 2A; for relative cell cycle positions, see Fig. S1).

Figure 2.

Cdk1-mediated phosphorylation regulates the interactions of Cdc7 with origins. (A) Cdc7 becomes de-phosphorylated as cells enter next cell cycle. HeLa S3 cells arrested by NZ were released into cell cycle, and samples were collected at indicated time points and biochemically fractionated. Western blotting was carried out with antibodies specific for Cdc7 or Orc2. “AS” denotes asynchronously growing HeLa cells. pCdc7 denotes phosphorylated Cdc7. (B) Schematic illustration of MCM4 and DBF4 origins used in this study. Approximate distances (kb) between primer pairs of origin (black) and non-origin control (white) are not drawn to scale. (C) Cdc7 returns to origins 6 h after release from prometaphase arrest. Asynchronous and NZ-based synchronized cells were collected for ChIP analysis using an antibody specific for Cdc7. Cdc7 enriched at the DBF4 origin was expressed as a ratio of origin:non-origin. For cell cycle position, see Fig. S1. (D) Phosphorylation at multiple residues may be required to dissociate Cdc7 from the chromatin. Asynchronous HEK293T cells expressing GFP-tagged alleles of Cdc7 wild type (WT), phospho-dead Cdc7 (X5A), or partially mutated phosphorylation sites (Cdc7^S16A, Cdc7^S302A, Cdc7^T376A and Cdc7^T472A, and Cdc7^T503A) were subjected to a ChIP assay to measure their binding abilities to the MCM4 and DBF4 origins, expressed as the ratio of origin:non-origin DNA. The relative binding ability between X5A and WT is significant (DBF4 ori: 1 way ANOVA; F=5.028, df 3,11 p < 0.05, α level = 0.05; MCM4 origin: 1 way ANOVA; F = 5.421, df = 3,11 p < 0.05, α level = 0.05.). * denotes significant increase when compared to WT, according to Dunnett's Multiple Comparison Test. Data are mean ± SEM, n > 3. (E) Phosphorylation of Cdc7 regulates its ability to bind to origins. Asynchronous HEK293T cells expressing GFP-tagged alleles of Cdc7 WT and phosphomimetic X5D were subjected to a ChIP assay to measure their binding abilities to origins. The relative amounts of Cdc7 bound to origins were determined by qPCR in triplicate, and expressed as the percentage of input DNA. A one-tailed Student's t-test was performed for origin and non-origin bound samples. * denotes significant enrichment. Data are mean ± SEM, n = 3.

We next examined the association of origins with Cdc7 as the function of cell cycle progression from prometaphase. Data from a chromatin immunoprecipitation (ChIP) assay using the well-characterized Dbf4 origin in HEK293T cells ^35,36 showed that Cdc7 was not enriched at the DBF4 origin until 6 h post-NZ (Fig. 2B and C) the time when cells are the new cell cycle. We next determined the binding abilities of different phospho-Cdc7 alleles using origins at the DBF4 and MCM4 loci.^35-37 As shown in Fig. 2D, phospho-dead Cdc7 allele (Cdc7^×5A) showed the highest affinity for both the MCM4 and DBF4 origins. The double (Cdc7^T472A/T503A) and triple (Cdc7^{S16A/S302A/T376A}) mutants showed more modest increases in affinity for the origin, compared to Cdc7^WT. (* p < 0.05; 1 way ANOVA: Dunnett's test). Therefore, we used the Cdc7^×5A, which was compared with phosphomimetic Cdc7^×5D allele (see below), for the subsequent experiments.

Data from ChIP assays using asynchronous HEK293T cells transiently transfected with GFP-Cdc7^WT or GFP-Cdc7^×5D construct showed that the association of the latter with origins was significantly lower than wild type (* p < 0.05; Student's T-test; Fig. 2E). Taken together, our data suggest that mitotic Cdk phosphorylation of Cdc7 is largely responsible for preventing the association of Cdc7 with origins.

Dbf4 remains chromatin bound throughout the cell cycle

Since Cdc7 is regulated by Dbf4 for the activation of DNA replication, we examined the relationship between the level of Dbf4 and timing of Cdc7 phosphorylation. HeLa cells arrested at G1/S by DT treatment reaches prometaphase by ∼6 h when they are released into cell cycle (Fig. S1). By 10 h post-G1/S, the majority of cells are in (early) G1 of the next cell division cycle. As shown in Fig. 3A, the level of Dbf4 protein does not notably fluctuate throughout the cell cycle, except a slightly higher level at 10 h post-DT. To gain a better understanding of Dbf4 regulation during mitosis, we carried out a detailed study using HeLa cells synchronized at prometaphase by NZ. Data from this study showed that the level of Dbf4 was about one third lower in prometaphase (0-2 h post-NZ) than asynchronous cells (Fig. S1; Fig. 3B). However, the level gradually increases as cells progress through late M to early G1, demonstrating that there is actually a degree of fluctuation. The decrease of Dbf4 levels precedes the destruction of cyclin B (Fig. 3B), suggesting that it occurs prior to the metaphase to anaphase transition, and may not be the result of APC-directed proteolysis. To determine if Dbf4 remains bound to origins in prometaphase, we examined its interaction with MCM4 and DBF4 loci using ChIP assays on asynchronous cells and those synchronized by NZ treatment. Interestingly, Dbf4 was found from both origins and non-origins, in an approximately equal amount (Fig. 3D). These results suggest that Dbf4 remains chromatin (origin)-bound during mitosis.

Figure 3.

Dbf4 remains origin bound during prometaphase. (A) The levels of Dbf4 are relatively constant throughout the cell cycle. Dbf4 protein levels were assessed after release from the DT-based HeLa cell synchronization at G1/S. γ-tubulin was used as a loading control. (B) The levels of Dbf4 and cyclin B are shown after cells arrested at prometaphase were released into cell cycle. GAPDH was used as a loading control. (C) The level of Dbf4 proteins only slightly decreases at the end of mitosis. The densitometric data are mean ± SEM, n=3, with panel B representing average blot image. (D) Dbf4 associates with origins and non-origin chromatin at equivalent levels. Asynchronous cells and those arrested in prometaphase were collected and subjected to ChIP analysis using an antibody specific for Dbf4. Levels of Dbf4 enriched at indicated regions of the chromatin were measured with qPCR. (E) Cdc7 and Dbf4 interaction is not substantially affected by Cdk1-dependent phosphorylation of Cdc7. Dbf4 protein was immunoprecipitated with an antibody specific for Dbf4 (ab124707; Abcam) from asynchronous HeLa cell extracts or those arrested at prometaphase by NZ (Dbf4 IgG), and the level of co-immunoprecipitated Cdc7 was measured by Western blotting.

Since Cdk1-mediated phosphorylation leads to the exclusion of Cdc7 from origins, we asked if Cdc7 phosphorylation affects its interaction with Dbf4. We immunoprecipitated Dbf4 from extracts prepared from asynchronous or prometaphase-arrested HeLa cells, and examined the level of co-immunoprecipitated Dbf4. Surprisingly, the amount of Cdc7 co-purified with Dbf4 was not substantially different between asynchronous and NZ-arrested cells (Fig. 3E). Because reducing and gel running conditions employed to observe Cdc7 phosphorylation on Western blots are not conducive to overnight immunoprecipitation conditions, we do not see robust isoforms present in samples derived from NZ-arrested cells. However, since most Cdc7 is hyper-phosphorylated at prometaphase, this evidence suggested that Cdc7 phosphorylation may only slightly decrease its affinity for Dbf4. These data suggest that although Cdk1-dependent phosphorylation of Cdc7 may influence the interaction between Cdc7 and Dbf4, it does not prevent these proteins from interaction. Taken together, it appears that the role of phosphorylation is to prevent Cdc7 from interacting with the chromatin (origin in particular) rather than directly regulating DDK activity.

The phosphorylation of Cdc7 prevents DNA (re-)replication

We next determined the functional significance of mitotic Cdk-mediated Cdc7 phosphorylation, as it was previously suggested that phosphorylation is required for Cdc7 activation.²⁰ To assess the effect of Cdc7 phosphorylation on DNA replication, we replaced endogenous Cdc7 with GFP-Cdc7^WT, GFP-Cdc7^×5A or GFP-Cdc7^×5D in HEK293T. The replacement was largely completed by 72 h post-co-transfection of Cdc7 siRNA and GFP-tagged recombinant Cdc7 constructs (Fig. 4A). To determine functional Cdc7 directly in relation to DNA replication initiation, we examined the phosphorylation of Mcm2 on Ser53, as it was previously reported that the phosphorylation on this residue is carried out by Dbf4-Cdc7 in a Cdk2-independent manner.³⁸ As well, Cdc7 dependent-Ser53 phosphorylation is stimulated by the conserved Treslin/Sld3 replication initiation protein at the beginning of S phase and serves as an efficient and direct biochemical readout of helicase formation and activation and the initiation of DNA replication.^39,40 Data from Western blotting with an anti-pMcm2^S53 antibody showed that the phosphorylation of Mcm2 on Ser53 was significantly reduced when endogenous Cdc7 was removed or replaced with GFP-Cdc7^×5D (* p < 0.05, 1 way ANOVA, Dunnett's test, Fig. 4B and C). This reduction is not likely due to cell cycle changes, as the overexpression of GFP-Cdc7 alleles alone or with siRNA does not lead to an obvious cell cycle arrest or a change in Mcm2 phosphorylation (Fig. S2). We did find a limited number of cell death and >4N DNA content in the cell population, but cell death is often seen when Cdc7 is depleted.⁴¹ As well, a certain degree of stress and death did occur in a portion of transfected samples; however, this did not translate into a change in Mcm2 phosphorylation. Our data thus suggest that phospho-Cdc7 may not effectively support DNA replication. Therefore, we determined replication activities in cells expressing different Cdc7 alleles under the same conditions as in Fig. 4A. We found that approximately 60% of cells expressing GFP-Cdc7^WT and GFP-Cdc7^×5A effectively incorporated the thymidine analog 5-ethynyl-2'-deoxyuridine (EdU). In contrast, only ∼26% cells transfected with GFP-Cdc7^×5D replicated DNA (Fig. 4D and E; boxes i vs xiii in panel D: p < 0.05), confirming that non-phosphorylated Cdc7, but not phospho-Cdc7, effectively supports DNA replication. This conclusion is also consistent with data shown in Fig. 2, which demonstrates that non-phosphorylated Cdc7, but not phosphorylated Cdc7, strongly associates with origins. Our data further indicate that, although Cdc7^×5A may have slightly reduced kinase activity in vitro, it is perfectly capable of activating DNA replication in vivo. Thus, the difference between Cdc7^×5A and Cdc7^×5D in promoting DNA replication is not because of defects in kinase activity caused by the changes in underlying amino acid sequence.

Figure 4.

Cdk1-dependent phosphorylation of Cdc7 impedes Mcm2 phosphorylation and DNA (re-)replication. (A) Endogenous Cdc7 was replaced with exogenous recombinant GFP-Cdc7 alleles by 72 h post-transfection. HEK293T cells were co-transfected with Cdc7 siRNA (to remove endogenous Cdc7) and GFP-Cdc7 constructs indicated. A scrambled siRNA was used as a control. Cdc7 denotes endogenous Cdc7. (B) The phosphorylation of Mcm2 at Ser53 is notably reduced when endogenous Cdc7 is ablated by Cdc7 siRNA or replaced with phosphomimetic Cdc7^×5D. Cells treated as in (A) were collected 72 h post- transfection and proteins were subjected to Western blotting with antibodies specific for phospho-Mcm2^S53, total Mcm2 or GAPDH. (C) Quantification of Mcm2^S53 phosphorylation. Levels of phosphorylated Mcm2 (shown in panel B) were quantified by densitometry and normalized against total Mcm2 protein which was first normalized against a GAPDH loading control. A significant difference was found among the population with different Cdc7 alleles, using a one-way ANOVA. F=6.094, df=4, 10, p<0.05. A * denotes a significant difference compared to scrambled siRNA control determined with a Dunnett's Multiple Comparison Test. Data are mean ± SEM, n=3. (D) DNA replication is strongly supported by Cdc7^×5A but not by Cdc7^×5D. HEK293T cells treated as in panel A were labeled with EdU-Chromeo 543 (an indirect readout of DNA replication) and quantified. A significant difference in EdU labeling was found in the population, as indicated by one-way ANOVA. F=101.9, df=4, 10, p<0.05, α level = 0.05. Data are mean ± SEM, n=3. 100 cells over at least 4 fields were counted for each replicate. Groups sharing the same letters, A, B or C, are not significantly different each other. (E) Representative images described in panel D, labeled with Chromeo 543 (EdU; red) and DAPI (DNA; blue). (F) SP600125-induced over-replication is reduced when Cdc7 is absent or endogenous Cdc7 is replaced with phosphomimetic Cdc7^×5D. HEK293T cells transiently transfected with GFP-Cdc7 alleles or vector alone were collected after Cdc7siRNA and SP600125 co-treatment (20 μM, 36 h). The subsequent flow cytometry profiles were gated based on results from asynchronous cells and those treated with SP600125. Greater than 4N DNA content is quantified on histograms and expressed as a percentage of total events (n>10,000). The population is significantly different based on one-way ANOVA; F = 60.02, df = 4, 15, p < 0.05, α level = 0.05. (Tukey levels not sharing the same letter are significantly different p<0.05). Data are mean ± SEM, n=3. (G) Representative histograms from the experiment described in panel F. Events were gated to prevent counting of doublets.

Together, our data suggest that Cdk1-mediated phosphorylation of Cdc7 plays a critical role in preventing re-initiating DNA replication by dissociating it from origins. To gain a better understanding about the role of Cdk1-dependent phosphorylation of Cdc7, we used SP600125-mediated DNA re-replication to determine whether constitutively phosphorylated Cdc7 can inhibit DNA re-replication in mitosis. The SP600125 Cdk1/JNK inhibitor was previously shown to cause DNA re-replication during G2, producing >4N cells.³ Our data showed that SP600125-mediated multiple DNA replication in HEK293T cells occurs in a Cdc7-dependent manner, as drug-treated cells with wild type Cdc7 (Cdc7^WT and scrambled samples in Fig. 4 F and G; 20 μM SP600125, 36 h) showed >4N phenotype but Cdc7 knockdown cells did not (Fig. 4 F and G). In the absence of Cdk1-dependent phosphorylation of Cdc7 (Cdc7^×5A), cells still re-replicate their DNA. However, in the presence of SP600125, DNA re-replication was largely suppressed in HEK293T cells that had been replaced their endogenous Cdc7 with phosphomimetic Cdc7^×5D (Fig. 4F and G; p < 0.05, 1 way ANOVA, Tukey's HSD). Our data thus demonstrate that the phosphorylation of Cdc7 is a key component of the Cdk1-dependent prevention of DNA re-replication during G2/M.

PP1α dephosphorylates Cdc7 as cells exit mitosis

As a cell exits mitosis to enter a new cell division cycle, it must de-repress DNA replication machinery. As a part of this process, our data suggest that Cdc7 is dephosphorylated and re-localized to the origins (Fig. 2). In this context, the enzyme responsible for Cdc7 dephosphorylation can function as a critical molecular switch to make cells become competent for DNA replication. Several phosphatases including Cdc25B/C, PP1 and PP2A are involved in cell cycle regulation by dephosphorylating proteins during G2-M.⁴² When we treated cells with okadaic acid (a PP1/PP2A inhibitor, 100 nM), Cdc7 dephosphorylation was reduced (Fig. 5A, 2 h, +OA), suggesting that Cdc7 may be dephosphorylated by either PP1 or PP2A phosphatase. Therefore, we carried out a co-immunoprecipitation assay to determine whether Cdc7 associates with PP1 or PP2A. We found that Cdc7 co-purifies with PP1α in asynchronous HeLa cells (Fig. 5B). However, we were unable to co-purify PP2A with Cdc7, leading us to test PP1α further. We then carried out an in vitro phosphatase assay with a recombinant PP1 catalytic subunit on Cdc7 using extracts from HeLa cells arrested in prometaphase by NZ. Data from this study indicate that phospho-Cdc7 is dephosphorylated by PP1 phosphatase (Fig. 5C). Finally, we examined the timing of the interaction between PP1α and Cdc7 by co-immunoprecipitation during mitotic exit. Cdc7 isolated from cells arrested in mitosis, which contain mostly phospho-Cdc7, was not notably co-purified with PP1α (Fig. 5D, time 0). In contrast, Cdc7 and PP1α strongly interact by 2 h post-NZ (Fig. 5D). This result is consistent with the timeline of Cdc7 dephosphorylation (Fig. 2A and 5D). Taken together, our data indicate that phospho-Cdc7 is dephosphorylated by PP1α as cells exit mitosis. This conclusion is consistent with a previous report that PP1α dephosphorylates Cdk1-target proteins after the metaphase-to-anaphase transition.⁴² Thus, our data connects PP1 phosphatase to the activation of DNA replication.

Figure 5.

Cdc7 is dephosphorylated by PP1α as cells exit mitosis. (A) Cdc7 de-phosphorylation is moderately reduced by okadaic acid (OA, 100 nM, 2 h). (B) PP1α interacts with Cdc7. Immunoprecipitation (IP) was carried out with an anti-PP1α antibody on HeLa cell extracts (PP1α IgG). Anti-goat IgG was used as a negative control (Con IgG). Co-immunoprecipitated Cdc7 was detected by Western blotting (WB). (C) Phospho-Cdc7 is dephosphorylated by PP1. Proteins isolated from HeLa S3 cells arrested in prometaphase by NZ were incubated at 30°C for 30 min with 5 units of PP1, followed by Western blotting with an anti-Cdc7 antibody. (D) Phospho-Cdc7 interacts with PP1α as cells exit M phase. HeLa S3 cells arrested at prometaphase by NZ were subjected to immunoprecipitation with an anti-Cdc7 antibody at 0, 1, 2 and 3 h post-NZ. Co-immunoprecipitated PP1α was detected by Western blotting. An anti-mouse IgG was used as control (Con). * indicates non-specific IgG fragment. IgG_HC and IgG_LC denote heavy and light chains of Mouse IgG, respectively.

Discussion

Utilizing single mutant (T376A), double mutant (T472A/T503A), triple mutant (S16A/S302A/T376A), phospho-dead, and phosphomimetic recombinant Cdc7 alleles in the canonical Cdk1 recognition sites, we have established that Cdc7 phosphorylation is not required for the initiation of DNA replication. Thus, our data clearly demonstrates that the dephosphorylation of Cdc7 is required for its function in the activation of DNA replication. In fact, the phosphorylation of Cdc7 is the major mechanism of preventing DNA (re)replication by DDK, as the affinity of phospho-Cdc7 for origins is almost completely abrogated. Only when Cdc7 is dephosphorylated by PP1α at the end of mitosis can Cdc7 move back to the origin of replication. If Cdc7 is not removed from origins, then inappropriate DNA rereplication can result. Overall, our data indicates that the phosphorylation of Cdc7 by mitotic Cdk is an important regulatory mechanism to ensure once-per-cell-cycle DNA replication.

In light of our data presented here, we propose a model where Cdk1-dependent phosphorylation of Cdc7 is an essential mechanism for preventing DNA re-replication during mitosis. According to this model, Cdk1 plays a key role for turning off the molecular switch (i.e., Cdc7) by phosphorylating and removing it from origins, resulting in the prevention of DNA re-replication during G2-M period. A question is: how may the phosphorylation of Cdc7 promote its dissociation from origins? In this regard, it should be noted that phosphorylation on multiple sites adds considerably more negative charges to Cdc7. In particular, Cdc7 is a very basic, positively charged protein, whose binding to the negatively charged chromatin (origins) can be stabilized by electrostatic forces. The addition of phosphates increases in the negative charge to Cdc7, resulting in its repulsion from the negatively charged DNA backbone. If the binding of Cdc7 to an origin is at least in part through its direct association with chromatin, the phosphorylation-mediated dissociation is likely through this mechanism.

In this control mechanism, other (replication) proteins may also be involved, by which the specificity of Cdc7 binding to origins can be achieved. For example, it has been shown in budding yeast that Cdc7 and Dbf4 interact through the MCM complex: Dbf4 with one or more of the Mcm2, Mcm4 and Mcm6 subunits^43-45 and Cdc7 with Mcm4 and Mcm5.⁴⁴ Coupled with the observation that subunits of the Mcm complex are removed from the chromatin via Cdk1 phosphorylation,^46,47 the Cdk1-dependent phosphorylation of Cdc7 on multiple sites prevents Cdc7 from interacting with the chromatin and associated proteins, leading to its removal from origins at G2/M. As cells exit mitosis, they will have to gradually switch back their system to make themselves competent for DNA replication during the next cell division cycle. One of the key steps for this process is the restoration of Cdc7 function to make it competent for DNA replication. This critical step is regulated by PP1α, which dephosphorylates phospho-Cdc7 during exit from mitosis and the M-G1 transition. Thus, PP1α plays an essential role in switching on Cdc7 to restore its competency for DNA replication. As such, our data directly connect PP1α to the activation of DNA replication. Together, our data presented in this paper establishes a novel model for the regulation of mammalian DNA replication, in which the critical switch is phosphorylation and dephosphorylation of Cdc7 by Cdk1 and PP1α, respectively.

Materials and methods

Plasmid constructs

The pEGFP-huCdc7 recombinant plasmid was constructed by cloning a full-length cDNA encoding the entire human Cdc7 into the Sma I site of pEGFP-C1 (Clontech). To study putative phosphorylation sites within Cdc7, the changes of amino acid sequence were generated by Quickchange PCR-based site-directed mutagenesis (Stratagene), using oligonucleotides listed in Table S1. mCherry constructs were made by exchanging the gene encoding GFP in the plasmid construct with the gene encoding mCherry, which was cloned from Addgene plasmid 18985 ⁴⁸ utilizing NheI and BsrGI sites.

In vitro kinase assay

GFP-Cdc7 alleles were affinity purified using the GFP-nAb single domain antibody system, as per manufacturer's protocols (Chromotek). Equivalent amounts of recombinant GFP-Cdc7 protein bound to GFP-nAb agarose beads were resuspended in 1× kinase buffer (New England Biolabs), supplemented with 200 ng of recombinant Dbf4 and 100 ng of recombinant Mcm4 proteins (Cedarlane Biotechnologies). 1 mM of ATP (Promega) was added and incubated at 37°C for 1 h. Finally, the reaction was collected and ATP consumption was assayed using an ADP-Glo assay kit as per suggested by the manufacturer (Promega), with luminescence assayed on a Synergy H4 Hybrid 96-well plate reader (Biotek).

Phosphatase treatment

Lambda protein phosphatase (λ-PPase) and PP1 phosphatase were purchased from New England Biolabs. HeLa cells treated with NZ were lysed in NP-40 lysis buffer. Cell extracts were then incubated with phosphatase for 30 min at 30°C according to the manufacturer's protocol. The reactions were stopped by adding 2× SDS sample buffer, followed by protein separation/identification by SDS-PAGE/Western blotting.

Co-immunoprecipitation assay

Co-IP from whole cell lysates was performed as described previously.⁴⁹

EdU labeling

5-ethynyl-2'-deoxyuridine (EdU; Invitrogen) labeling of replicating DNA was carried out as described previously.⁵⁰ Chromeo 546 azide was purchased from Active Motif. Cells were imaged on a Carl Zeiss LSM 510 or Axiovert 100 using a 20× objective. Cells were deemed to be positive for EdU labeling when there was signal observed consistently throughout the entire nucleus above the maximum exposure threshold set by using the negative Edu alone/no Chromeo 546 control in conjunction with no Edu/Chromeo 546 alone.

Chromatin immunoprecipitation (ChIP) assays

ChIP was performed as described previously.^36,49,51 qPCR primers used are listed in Table S2.

Western blot analysis

PAGE and Western blotting were carried out as described previously.⁵² For the resolution of Cdc7 phospho-isoforms, sample buffer was supplemented with 8M urea and SDS-PAGE was carried out utilizing a 10% resolving gel and at least 1 cm 3% stacking gel, run at 125 volts for 2.5 h to maximize resolution. Representative images of at least two independent experiments are shown. Densitometry was carried out on images obtained using the FluorChem FC3 system (Protein Simple, USA). AlphaView software was used, with non-saturated images and local background set on equal area measurements to determine percent difference based on a control band, as indicated for each experiment.

Flow cytometry

Flow cytometry using propidium iodide was carried out as described previously,⁵³ using a Beckman Coulter Cytomics FC500 cytometer. The total counts were gated to account for potential doublets.

Statistical methods

One-way ANOVA and Student's t-tests was carried out with GraphPad Prism 5. At least three replicates were performed for each experiment. The sample size for cells was n > 100; the sample size for flow cytometry was total events n > 10,000. The α level was 0.05. The values obtained for each ANOVA are listed in the associated figure legend. A post hoc test was performed as indicated in the associated figure legend, p < 0.05.

Abbreviations

CDK

Cyclin Dependent Kinase

DDK

Dbf4/Drf1-dependent kinase

ChIP

chromatin immunoprecipitation

Co-IP

co-immunoprecipitation

Mcm

minichromosome maintenance

NZ

nocodazole

ORC

origin recognition complex

PPase

phosphatase

Pre-RC

pre-replication complex

siRNA

small interfering RNA

WB

Western Blot

WCL

whole cell lysate; X5A, Cdc7^×5A; X5D, Cdc7^×5D

Disclosure of potential conflicts of interest

No potential conflicts of interest were disclosed.

Acknowledgments

We thank Miroslav Dundr for providing a plasmid containing mCherry via Addgene (Cambridge, MA).

Author contributions

HL conceived the research project, analyzed data and wrote the final version of this manuscript; JK designed and carried out experiments, analyzed data and wrote initial manuscript; BJK made important contribution to the formulation of the Cdc7 regulation mechanism, and designed and carried out experiments; AM carried out some experiments.

Funding

This work was supported by funds from the Canadian Institutes of Health Research (CIHR #79473), Natural Sciences and Engineering Research Council of Canada (NSRC #203528-2013), Northern Ontario Heritage Funds Corporation (NOHFC), and the City of Greater Sudbury to HL. JK thanks to Cancer Care Ontario for the Peter Crossgrove postdoctoral fellowship. BJK greatly acknowledges for Ontario Graduate scholarship.