|Home | About | Journals | Submit | Contact Us | Français|
Cyclin-dependent (CDK) and Dbf4-dependent (DDK) kinases trigger DNA replication in all eukaryotes, but how these kinases cooperate to regulate DNA synthesis is largely unknown. Here, we show that budding yeast Mcm4 is phosphorylated in vivo during S phase in a manner dependent on the presence of five CDK phosphoacceptor residues within the N-terminal domain of Mcm4. Mutation to alanine of these five sites (mcm4-5A) abolishes phosphorylation and decreases replication origin firing efficiency at 22°C. Surprisingly, the loss of function mcm4-5A mutation confers cold and hydroxyurea sensitivity to DDK gain of function conditions (mcm5/bob1 mutation or DDK overexpression), implying that phosphorylation of Mcm4 by CDK somehow counteracts negative effects produced by ectopic DDK activation. Deletion of the S phase cyclins Clb5,6 is synthetic lethal with mcm4-5A and mimics its effects on DDK up mutants. Furthermore, we find that Clb5 expressed late in the cell cycle can still suppress the lethality of clb5,6Δ bob1 cells, whereas mitotic cyclins Clb2, 3, or 4 expressed early cannot. We propose that the N-terminal extension of eukaryotic Mcm4 integrates regulatory inputs from S-CDK and DDK, which may play an important role for the proper assembly or stabilization of replisome–progression complexes.
Eukaryotic chromosome replication initiates throughout S phase from multiple origins, and it is controlled in large part by cyclin-dependent kinases (CDKs). CDKs have both positive and negative roles for DNA replication. They promote the initiation of DNA synthesis from competent origins, but they also prevent the reassembly of prereplicative complexes (preRCs) at origins that have already fired. This inhibition of preRC formation by CDKs allows chromosome replication to be coupled to the cell cycle (Diffley, 2004 ). PreRC assembly is well understood: it begins with origin recognition complex binding to origin DNA, followed by recruitment of Cdc6 and Cdt1, which in turn permit loading of the Mcm2-7 complex, the likely replicative helicase (Blow and Dutta, 2005 ; Takeda and Dutta, 2005 ). Binding of Cdc6 and Cdt1 is a highly regulated step ensuring that origin licensing is restricted to the G1 phase and that origins do not fire twice during the same cell cycle (Blow and Dutta, 2005 ). CDKs inhibit origin licensing by targeting several preRC components (Cdc6, Cdt1, and Mcm3) for degradation or nuclear exclusion (Liku et al., 2005 ). Phosphorylation by CDK of Orc2 and Orc6 in Saccharomyces cerevisiae (Nguyen et al., 2001 ) and Orp2 in Schizosaccharomyces pombe (Vas et al., 2001 ) also participates in inhibiting preRC formation during the S-to-M phase period.
Besides preventing preRC assembly CDKs also have a positive, yet less well understood role for origin firing. The maturation of preRCs requires activation of two evolutionary conserved S/T protein kinases: Dbf4-Cdc7 (Dbf4-dependent kinase or DDK) and S-phase CDK (Clb5,6-Cdk1 in S. cerevisiae or CycE,A-Cdk2 in higher eukaryotes). These two kinases promote the recruitment of the Sld3-Cdc45 and Sld2-Dpb11 heterodimers, the GINS complex and finally, RPA and DNApolα/primase to the site of initiation (Tanaka and Nasmyth, 1998 ; Zou and Stillman, 2000 ; Kamimura et al., 2001 ; Masumoto et al., 2002 ; Takayama et al., 2003 ). Some of these initiation factors (Mcm2-7, Cdc45, and GINS) move along with replication forks, indicating that they may be part of the active helicase complex (Aparicio et al., 1997 ; Labib et al., 2000 ; Tercero et al., 2000 ; Takayama et al., 2003 ). Accordingly, minichromosome maintenance (MCM) and Cdc45 are both required for DNA unwinding in Xenopus egg extracts (Pacek and Walter, 2004 ). The GINS complex maintains association of MCM helicase with Cdc45 and other replication factors, such as the checkpoint mediator Mrc1, the fork-pausing complex Tof1-Csm3 as well as DNA polymerase-associated proteins (Gambus et al., 2006 ; Kanemaki and Labib, 2006 ). Thus, the replication initiation and progression complexes are large molecular entities that contain numerous potential targets for CDK and DDK, but most studies have first focused on the MCM complex because it is well conserved among eukaryotes and carries helicase activity.
Several in vivo phosphorylation sites in the N-terminal region of Mcm2, 4, and 6 have been mapped, which can be phosphorylated in vitro either by CDK or DDK (Komamura-Kohno et al., 2006 ; Masai et al., 2006 ; Montagnoli et al., 2006 ; Sheu and Stillman, 2006 ). However, mutation of these sites to either Ala or Glu does not cause lethality, and the importance of these modifications for the proper execution of S phase remains to be evaluated. Recently, a role for DDK-dependent Mcm4 phosphorylation in promoting interaction with Cdc45 was demonstrated (Masai et al., 2006 ; Sheu and Stillman, 2006 ). There is also evidence suggesting that Mcm4 phosphorylation by CDK might be inhibitory: in Xenopus, Mcm4 hyperphosphorylation by CDK was correlated with decrease of its binding to chromatin (Hendrickson et al., 1996 ; Findeisen et al., 1999 ) and the in vitro helicase activity of Mcm4-6-7 was inhibited when Mcm4 was phosphorylated by Cdk2 (Ishimi et al., 2000 ). In vivo studies have determined that budding yeast Clb5,6-Cdk1 acts positively on DNA replication by phosphorylating Sld2 (Masumoto et al., 2002 ) and the DNAPolε subunit Dpb2 (Kesti et al., 2004 ). An Sld2 mutant in which all CDK phosphoacceptor sites are changed to Ala is lethal, shows strong defects in S phase progression, and it was demonstrated that phosphorylation of Thr84 is solely responsible for stabilizing the Sld2–Dpb11 interaction (Tak et al., 2006 ). A breakthrough came from the recent discovery that phosphomimetic forms of Sld2 combined to constitutive Sld3–Dbp11 complex formation can bypass all minimal requirement of CDK for DNA replication (Tanaka et al., 2007 ; Zegerman and Diffley, 2007 ). That Sld2 and Sld3 phosphorylation is sufficient implies that phosphorylation of MCM by CDK is not essential for DNA replication. Although not a prime player, the MCM complex is clearly targeted by CDK and DDK in several eukaryotes, where fine-tuning of replisome assembly and helicase activity might be biologically important. In contrast, Archaea use only a subset of initiation factors found in eukaryotes and neither orthologues of Sld3, Cdc45, Sld2, and Dpb11 nor of CDK/DDK can be found. The homohexameric MCM complex from Methanobacterium thermoautotrophicum has strong helicase activity in vitro, whereas MCM complexes or subcomplexes isolated from eukaryotes have, at best, a weak activity (Kelman et al., 1999 ; Chong et al., 2000 ; Lee and Hurwitz, 2001 ; Shin et al., 2003 ). Interestingly, archaeal Mcm proteins also lack the S/T-rich N-terminal extensions of eukaryotic Mcm2, 4, and 6, which are proposed targets for regulation by CDK and DDK. Here, we provide in vivo evidence that phosphorylation of CDK consensus sites within the N-terminus of S. cerevisiae Mcm4 contributes to efficient origin firing. We also find that preventing Mcm4 N-ter phosphorylation is severely deleterious when combined to gain of function DDK mutations, suggesting that a proper balance between CDK and DDK activities on the MCM complex is necessary for efficient replisome assembly or progression.
tetCDC7 plasmid (D577) was described previously (Nougarede et al., 2000 ), and it is a YCplac22(TRP1) derivative. The DBF4 open reading frame (ORF) was inserted into pCM189(URA3) (Gari et al., 1997 ) to obtain the tetDBF4 plasmid. Mutations of putative CDK phosphorylation sites in ScMcm4 (Ser 7, 17, 32, 69, and 145) were generated using the QuikChange mutagenesis kit (GE Healthcare, Chalfont St. Giles, United Kingdom), substituting serines for alanines. Mutation of Ser7 was marked with a HaeII restriction site, and Ser17 by a KasI site. Integration of all five mutations at the MCM4 locus to generate the mcm4-5A allele was done by direct gene replacement. A URA3 marker was first integrated 350 base pairs before the start codon of the MCM4 ORF. Transformation with a mutant MCM4 gene fragment and selection with 5-fluorootic acid allowed popping out of the URA3 gene and integration of the mutated allele (strain E1448). The presence of mutations was verified by DNA sequencing. The bob1-1 allele was marked by replacing nucleotides 25–128 downstream of MCM5 stop codon by the Kluyveromyces lactis TRP1 gene. Strains expressing tagged versions of wild-type Mcm4 and mutant Mcm4-5A proteins were constructed by replacing the stop codon of the MCM4 and mcm4-5A genes by the TEV-PrA-7His-Sp.his5+ cassette of plasmid pYM10 (Knop et al., 1999 ). A 3HA-tagged CLB5 gene under the control of the GALS promoter was generated by inserting the GALS-3HA-nat cassette of pYM-N32 (Janke et al., 2004 ) just downstream of the initiator ATG codon of the CLB5 gene. Chemical inhibition of CDK activity in the cdc28-as1 strain was done in YEPD medium containing 0.5 mM 4-amino-1-tert-butyl-3-(1′-naphthylmethyl)pyrazolo[3,4-d]pyrimidine (1-NMPP1) (Bishop et al., 2000 ). Table 1 lists yeast strains used in this study.
In vivo 5-bromo-2′-deoxyuridine (BrdU) incorporation, DNA combing and detection were performed as described previously (Lengronne et al., 2001 ), except that BrdU stretches and DNA fibers were labeled concomitantly with two different fluorophores. BrdU was detected with a rat anti-BrdU antibody (clone BU-75; SeraLab, Crawley, United Kingdom) and a secondary antibody coupled to Alexa 488 (Invitrogen, Paisley, United Kingdom), whereas DNA was revealed with a mouse anti-guanosine antibody (clone GK-21; Argene, Varilhes, France) and a secondary antibody coupled to Alexa 546 (Invitrogen). Interorigin distances (IODs), defined as the distance between the center of two successive colinear BrdU tracks, were measured and plotted as a distribution of IOD range categories. Box-and-whiskers plots were generated using GraphPad Prism software to visualize the main parameters of the distributions. Hypotheses that two distributions are equal or not were verified using the Mann–Whitney statistical test.
Whole-cell extract proteins (15 μg) prepared using the Tri-chloro-acetic acid (TCA) method were loaded on standard 6% SDS-polyacrylamide gel electrophoresis (PAGE) gels (10 × 10 × 0.08 cm) containing 25 μM PhosTag ligand (AAL-107; NARD Institute, Amagasaki, Japan) and 50 μM MnCl2, according to Kinoshita et al. (2006) , with special care to avoid any traces of phosphate in buffers or molecular weight markers. Gels were run at 40 mA for 1 h 30 until bromophenol blue runs out, rinsed twice for 10 min in transfer buffer (Tris-glycine, SDS, and ethanol) containing 1 mM EDTA to chelate MnCl2, and once in the same buffer without EDTA. Proteins were transferred on ProTran membrane (Whatman Schleicher and Schuell, Dassel, Germany) by semidry blotting for 75 min at 0.1 mA/cm2. The protein A tag was revealed using peroxydase anti-peroxidase (PAP) antibody (P1291; Sigma Chemical, Poole, Dorset, United Kingdom) at dilution 1:4000.
The N-terminal domain of eukaryotic Mcm4 contains a cluster of putative CDK phosphorylation sites that is conserved throughout evolution (Figure 1). This feature is striking in view of the weak overall sequence conservation of this domain. In S. cerevisiae, there are two SPxK/R (S7 and S145) and three SP motifs (S17, S32, and S69) within the first 150 residues. Interestingly, all five CDK motifs are preceded by one or two serine residues that, according to Masai et al. (2006) and Montagnoli et al. (2006) , could correspond to DDK phosphoacceptor sites. This juxtaposition of potential DDK/CDK phosphoacceptors is a conserved feature of Mcm4 N termini as all species aligned in Figure 1 contain at least four of them. The remainder of the protein consists mostly of highly conserved domains (AAA+ helicase motifs) with very few potential CDK sites. We performed phosphoproteomic analysis with an allele of MCM4 tagged at its C-terminus with protein A and introduced at the natural locus. However, assessing the in vivo phosphorylation status of yeast Mcm4 turned out to be difficult because the protein did not show obvious mobility shift during the cell cycle, in standard or modified western blot conditions (Figure 2B, top). Two-dimensional gel electrophoresis and mass spectrometry analyses were also unsuccessful (data not shown), perhaps due to the fact that only a small fraction of cellular MCM molecules is phosphorylated in vivo (Sheu and Stillman, 2006 ). Recently, a phosphate ligand (PhosTag) was introduced, which significantly slows down the migration of phosphoproteins in SDS-PAGE (Kinoshita et al., 2006 ). Using PhosTag, we found very reproducibly that ~10–20% of Mcm4 molecules migrate more slowly, depending on cell cycle position (Figure 2B, bottom). These Mcm4-specific (Figure 2A) slower migrating species were present in α-factor-arrested cells, decreased at 15 min and reached their maximum 30 min after release, concomitant with DNA replication. Better inspection of samples run without PhosTag (Figure 2B, top) reveals a broadening of the Mcm4 band at 30min, which likely corresponds to the phosphorylated forms seen using PhosTag. These forms decreased in G2/M to increase again upon S phase in the following cycle (75 min). We conclude that a fraction of yeast Mcm4 is phosphorylated in vivo during S phase. To test whether this phosphorylation depends on CDK sites within the N terminus, serines within all five SP or SPxK/R motifs (S7, 17, 32, 69, and 145) were substituted to alanine. This allele (mcm4-5A) was subjected to PhosTag analysis as mentioned above. Mutation of these sites caused disappearance of all slower migrating bands (Figure 2C), demonstrating that Mcm4 phosphorylation in vivo depends on one or more of these five CDK sites clustered within Mcm4's N-terminus. However, the complex pattern of Mcm4 phosphorylation during the cell cycle precluded any simple assessment of the kinase phosphorylating these sites. Specific inhibition of CDK during an α-factor release by using a small molecule (1-NMPP1) in a cdc28-as1 strain led to a complete disappearance of slower migrating bands during the first 30 min of the time course (Figure 2E). Although these cells never exited G1 (no budding, no DNA replication), slower migrating bands reappeared at later times, indicating that Mcm4 phosphorylation can also occur in a CDK-independent manner. This suggests that several kinases can phosphorylate Mcm4 but that CDK inhibition prevents or slows down the initial phosphorylation of Mcm4 in late G1.
The biological importance of these five CDK phosphoacceptor sites (S7, S17, S32, S69, and S145) within Mcm4 was assessed using flow cytometry, cell viability, and minichromosome maintenance assays, which did not point toward obvious DNA replication defects in the mcm4-5A mutant strain (data not shown). Thus, these sites cannot be essential CDK targets, but it does not rule out that Mcm4 sites may act synergistically with other targets. Indeed, we found that the mcm4-5A allele is lethal in a strain lacking the S phase cyclins Clb5 and 6, which on its own has normal viability despite delayed DNA replication (Figure 3A) (Schwob and Nasmyth, 1993 ). Using a conditional allele of CLB5 driven by the GALs promoter, we were able to test the consequences for DNA replication and cell division of depleting Clb5 in a clb6 mcm4-5A strain. Figure 3B shows that GALs-CLB5 clb6 mcm4-5A cells on glucose replicate DNA more slowly than GALs-CLB5 clb6 control cells (compare the 45- and 150-min times in both strains). The triple mutant cells underwent one or two additional divisions and then died without ever forming a colony (Figure 3A). The additive effects (colethality) of mcm4-5A and clb5,6Δ mutations implies both that the N-terminal Mcm4 sites can be phosphorylated by other kinases, possibly M phase CDKs and that Clb5,6-CDK (obviously) targets other key proteins. The lethality of mcm4-5A clb5,6Δ cells would thus stem from the impossibility to phosphorylate Mcm4 combined to the absence (or delay) of phosphorylation of replication factors normally targeted by Clb5,6-Cdk1, such as for example Sld3 (Zegerman and Diffley, 2007 ).
We also combined the mcm4-5A allele with various other mutations in replication initiation factors (including putative CDK sites within Mcm2 and Mcm10). One double mutant, mcm4-5A bob1, turned out to be cold-sensitive (cs) for growth (Figure 4A). bob1-1 is an allele of MCM5 (changing Pro83 to Leu) that bypasses the requirement of CDC7 and DBF4 for DNA replication (Hardy et al., 1997 ). The mcm4-5A bob1 double mutant ceases cell proliferation at 22°C (and below) after one to three cell divisions. To see whether these cells have problems to initiate DNA replication, bob1 and mcm4-5A bob1 cells were synchronized with α-factor at permissive temperature (32°C) to allow preRC formation, and then they were released at 22°C (restrictive temperature). Flow cytometry analysis revealed that although S phase started on schedule (30 min after αF-release) in the double mutant, it progressed very slowly, reaching apparent completion only at 105 min (instead of 60 min in the bob1 control; Figure 4B). Figure 4C shows that the mcm4-5A bob1 double mutant is also highly sensitive to hydroxyurea (HU) at permissive temperature. These cells did not die in the first cell cycle (unlike rad53 mutants on HU) but formed microcolonies composed of 10–20 cells, suggesting that they are capable of exiting mitosis under chronic HU exposure but suffer from gradual loss of viability.
The slow S phase in the mcm4-5A bob1 mutant at 22°C could be due either to a decrease of the number of active origins or to elongation proceeding at a slower pace. To monitor origin firing efficiency, we analyzed BrdU-labeled DNA fibers straightened on a microscope glass slide by using DNA combing (Lengronne et al., 2001 ). On release from α-factor at 22°C into medium containing HU, BrdU gets incorporated into short 10- to 20-kb regions surrounding origins (Figure 5A). The density of active origins was then calculated by measuring the distance separating the center of two successive BrdU tracks (IOD) in >200 fibers. Figure 5B shows that the mean IOD in the mcm4-5A bob1 double mutant is more than twice that of bob1 and wild-type strains (90 kb instead of 42 kb), implying that the number of fired origins (at least for the subset of early origins) is significantly lower in the double mutant. The notion that mcm4-5A bob1 cells have a slow S phase because of initiation, not elongation defects is supported by the absence of a significant delay in S phase completion when mcm4-5A bob1 cells are first arrested in HU at permissive temperature (allowing for early origins to fire) and then released at restrictive temperature (Supplemental Figure S1). We conclude that the lengthening of S phase in the mcm4-5A bob1 double mutant is almost entirely due to a failure to activate origins.
One trivial explanation for the synthetic effects between mcm4-5A and mcm5/bob1 could be that each mutation causes a slight loss of function that together would affect the functionality of the MCM complex. A more attractive hypothesis is that the N-termini of Mcm4 and Mcm5 (where the mcm4-5A and bob1 mutations reside) integrate regulatory signals conveyed by CDK and DDK. In this scenario, contradictory inputs brought about by the mcm4-5A (loss of CDK function) and bob1 (DDK bypass) mutations would make the initiation mechanism inefficient. In keeping with the latter hypothesis (contradictory inputs) for the colethality of mcm4-5A and bob1, we found that ectopic DDK expression recapitulates the phenotypes obtained with mcm5/bob1. Plasmids bearing the CDC7 and DBF4 genes under control of a Tet-repressible promoter were introduced in the mcm4-5A strain. As shown in Figure 4D, the mcm4-5A mutant did not grow at 22°C in the absence of doxycycline, when both CDC7 and DBF4 are coexpressed. This lethality was not observed when only one of the two DDK subunits was overexpressed, or at 32°C at which temperature cells grew equally well in the presence or absence of doxycycline (data not shown). These findings indicate that the lethality of the mcm4-5A bob1 strain likely stems from precocious or constitutive DDK-dependent activation of the MCM complex, not from a loss of function of the Mcm5Bob1 subunit. The cold-sensitive phenotype suggests that, in mcm4-5A bob1 cells at 22°C, a fraction of preRCs might be locked in a conformation that is refractory to inducers of DNA replication.
These observations prompted us to evaluate the possibility that the mcm4-5A bob1 lethality was due to an imbalance between CDK and DDK activities. One possibility is that forced activation of the preRC by DDK without previous Mcm4 phosphorylation by CDK leaves preinitiation complexes in a state that is unstable or refractory to origin firing. If this scenario were correct, delaying the activation of S phase CDK in a bob1 background should produce the same phenotype as the mcm4-5A bob1 double mutant. Deletion of CLB5 and CLB6 leads to a 30- to 40-min delay in the initiation of DNA replication, which is then triggered by Clb1–4 cyclins (Schwob and Nasmyth, 1993 ). As predicted by the above-mentioned hypothesis, the clb5,6Δ bob1 triple mutant was also found cold sensitive for growth (Figure 6A), with DNA replication progressing very slowly at the nonpermissive temperature (Figure 6B). Moreover, the clb5,6Δ bob1 triple mutant showed HU sensitivity similar to that of the original mcm4-5A bob1 mutant (Figure 6C), indicating that ablation of Clb5 and 6 recapitulates the effects caused by removing Mcm4 phosphoacceptor residues in a bob1 context. We further explored whether, in this situation of delayed CDK activation, hyperactivation of DDK could substitute for the bob1 mutation. Indeed, Figure 6D shows that clb5,6Δ cells cannot grow when both CDC7 and DBF4 are overexpressed. The strong similarity of phenotypes obtained with mcm4-5A and clb5,6Δ in conditions of DDK bypass or up-mutations is consistent with residues S7, S17, S32, S69, and S145 of Mcm4 being in vivo targets of CDK. It also suggests that S-CDK and DDK must act in a balanced and coordinated manner to fire origins in an efficient way.
In the experiments described above (Figure 6), the lethality of bob1 clb5,6Δ cells at low temperature could stem either from a delay in phosphorylating S-CDK substrates (if mitotic cyclins can substitute) or a complete lack thereof (if Clb1–4/Cdk1 cannot productively phosphorylate these targets). To distinguish between these possibilities, we introduced in the bob1 clb5,6Δ strain a version of CLB5 under control of the G2/M-specific SWI5 promoter, which causes Clb5-Cdk1 kinase activity to occur at the same time as that of Clb2-Cdk1 (Supplemental Figure S2). Figure 7A shows that the SWI5pr-CLB5 allele completely suppressed the cold lethality of the bob1 clb5,6Δ mutant. Thus, although Clb5 expressed in G2/M can rescue the replication defects of the bob1 clb5,6Δ triple mutant, mitotic Clb1–4 cyclins that are expressed and active at the same time cannot. We conclude that it is not the timing of CDK activation that is critical for the viability of bob1 cells, rather the specificity of the cyclin–Cdk complex. Clearly, Clb5,6/Cdk1 does something to bob1 cells that Clb1–4/Cdk1 cannot. To strengthen this conclusion, we performed the reciprocal experiment by expressing mitotic cyclins at a time matching that of the Clb5,6 S phase cyclins. To this aim, CLB2, 3, and 4 open reading frames were each introduced at the CLB5 locus (under control of the CLB5 promoter) in a bob1 clb6Δ swe1Δ strain. It was shown previously that SWE1 deletion significantly increases the ability of clb5::CLB2, 3, or 4 to drive S phase to almost wild-type kinetics in a clb5,6Δ strain (Hu and Aparicio, 2005 ). Furthermore, it is known that Clb2 protein made from the CLB5 promoter is expressed at similar levels and timing as endogenous Clb5, while carrying stronger kinase activity (Cross et al., 1999 ; Loog and Morgan, 2005 ). Despite this, we found that none of the mitotic cyclins expressed early could suppress the cold sensitivity of the bob1 clb5,6Δ swe1Δ strain (Figure 7B). Thus, although either S or M phase cyclins can trigger DNA replication in wild-type cells, only Clb5 and 6 can do so in a strain where DDK regulation has been bypassed, suggesting that Clb5,6/Cdk1 has the unique property to interface with DDK activity.
What could be the role of Mcm4 N-ter phosphorylation during the normal process of initiation? Although the mcm4-5A mutation did not confer obvious growth defects in otherwise wild-type cells, careful analysis of S-phase progression by flow cytometry and DNA combing revealed a significant and reproducible lengthening of S phase in the mcm4-5A mutant strain at 18°C. DNA replication was completed only 120 min after release from G1 compared with 90 min in wild type (Figure 8). This lengthening was also present at 22°C, but less important (data not shown). Using DNA combing, we found that the mean distance between active origins was 61 kb in the mcm4-5A single mutant at 22°C, compared with 42 kb in wild-type and bob1 cells (Figure 5 and Supplemental Figure S3), indicating that chromosomes are duplicated from fewer origins in mcm4-5A cells. These results demonstrate that, although not essential in laboratory conditions, the phosphorylation of the Mcm4 N-terminus contributes to the efficacy of origin firing.
Mcm4 is highly conserved in eukaryotes. In multiple alignments covering divergent species, the similarity is high throughout most of its length, reaching 61% identity over a 137-amino acids stretch. The similarity drops dramatically, however, within the first 170 residues of Mcm4, which are rich in potential CDK phosphorylation sites. Using phosphospecific antibodies Ishimi and coworkers showed that at least seven of the 11 N-terminal S/TP sites are phosphorylated in vivo in human cells (Komamura-Kohno et al., 2006 ). In vitro studies revealed that these phosphorylations are CDK-dependent (Ishimi and Komamura-Kohno, 2001 ). Finally, phosphopeptide mapping showed that CycB-Cdk1 was responsible for most if not all Mcm4 phosphorylation in M phase Xenopus extracts (Hendrickson et al., 1996 ; Pereverzeva et al., 2000 ). Here, we show that a fraction of Mcm4 molecules is phosphorylated during S phase in S. cerevisiae (Figure 2). Although we could not provide a formal demonstration that CDK is indeed responsible for these phosphorylations, this idea is supported by the strong similarities between the phenotypes caused by the elimination of five CDK phosphoacceptor residues within the N-terminus of ScMcm4 and those obtained by deleting the S-phase cyclins, when these mutants are combined to DDK gain-of-function mutations.
We provide the first in vivo evidence that phosphorylation of Mcm4 contributes to an efficient initiation of DNA replication. This phosphorylation of Mcm4 is not essential for initiation because the mcm4-5A mutant strain, which lacks all five CDK phosphoacceptor sites, displays only subtle replication defects in laboratory conditions. Yet, at temperatures of ≤22°C, which are common for yeast ecotypes, the mcm4-5A allele causes a clear lengthening of S phase, which correlates with a reduced number of fired origins determined by DNA combing. Interestingly for our understanding of the mechanism of replication initiation, we show that more profound effects are observed when the mcm4-5A allele is combined to mutations that bypass (mcm5/bob1) or up-regulate (ectopic Cdc7-Dbf4 expression) DDK activity. In this context, mcm4-5A cells proceed exceedingly slowly through S phase at low temperatures, leading to cell death within the first three cell divisions. The terminal morphology consists of large dumbbell cells with DNA trapped in the neck or stretched between the mother and daughter cells (data not shown). The decreased origin firing measured on single DNA molecules in the mcm4-5A bob1 strain points to an activating, not inhibitory, role of Mcm4 phosphorylation by CDK in initiating DNA replication. BrdU tracks were not shorter, nor was S phase significantly lengthened when mcm4-5A bob1 cells were shifted to restrictive temperature after initiation, arguing against a role of Mcm4 N-ter phosphorylation in the elongation of DNA synthesis. In contrast to Mcm4 that we show here to only contribute to efficient origin firing, Sld2 and Sld3 are essential to promote DNA replication in a CDK-dependent manner (Tak et al., 2006 ; Tanaka et al., 2007 ; Zegerman and Diffley, 2007 ). We suggest that, in addition to promoting the formation of Sld2–Dpb11 and Sld3–Cdc45 complexes required for recruitment of GINS and replisome progression complexes, S-CDKs also perform a nonessential regulatory function on the MCM complex that interfaces with DDK function.
Our finding that Clb5 expressed late in the cell cycle (from the SWI5 promoter), but not any of the mitotic cyclins Clb2, 3, or 4 expressed early (from the CLB5 promoter in a swe1Δ background), can suppress the cold lethality of the bob1 clb5,6Δ mutant indicates that the defects of this strain stem from a lack, not simply a delay of phosphorylation of one or more Clb5,6-specific substrates. It is known that several CDK substrates, among which are some key replication factors such as Orc6, Cdc6, Mcm3, and Sld2, are more efficiently phosphorylated in vitro by Clb5-Cdk1 than by Clb2-Cdk1 (Loog and Morgan, 2005 ). This specificity was shown to depend on the presence of a hydrophobic patch in Clb5, and, in some cases, of a Cy or RXL motif in the substrate (Wilmes et al., 2004 ). Given the genetic evidence presented here, we propose that Mcm4 could be another member of this class of preferential Clb5–CDK targets. However, Mcm4 is not phosphorylated exclusively by S-CDKs, as revealed by the synthetic lethal interaction between the mcm4-5A and clb5,6Δ mutations and by the residual phosphorylation after chemical inhibition of Cdc28. The kinases responsible for these phosphorylations are likely mitotic CDKs and DDK, which could modify Mcm4 but to a lower level than in the presence of S-CDK. These activities would be sufficient for viability of clb5,6Δ cells, but not for the same cells in the context of DDK gain-of-function mutations.
It is thought that bob1, an allele of MCM5, bypasses the requirement of CDC7 and DBF4 for DNA replication by inducing a conformational change in the MCM complex that mimics activation by Dbf4-Cdc7 kinase. In keeping with this interpretation, it was shown that bob1 cells exhibit abnormally high amounts of chromatin-bound Cdc45 in α-factor arrested G1 cells (Sclafani et al., 2002 ). Cdc45 binding to chromatin depends on DDK, and it is therefore very low in wild-type G1 cells. The notion that mcm4-5A bob1 defects are linked to bypass of DDK regulation is corroborated by our finding that ectopic expression of CDC7 and DBF4 in the mcm4-5A mutant has the same effect as the bob1 mutation. Thus, it seems that MCM complexes preactivated by DDK cannot stably promote origin firing, unless Mcm4 is also phosphorylated on its N terminus by S-CDK. This suggests that Mcm4 phosphorylation by CDK compensates for a conformational defect caused by the bob1 mutation or by DDK hyperactivation (Fletcher and Chen, 2006 ). Another way to think about these data is that there might be an order in the molecular events leading to preRC activation by CDK and DDK. In this view, forced preRC activation by DDK (either through bob1 or ectopic DDK expression) without previous Mcm4 phosphorylation by S-CDK might render origins refractory to firing by affecting replisome assembly or stabilization.
Why should ectopic DDK activity in conditions in which Mcm4 cannot be phosphorylated by CDK be deleterious for preRC activation? One hint could be that Cdc45 binds to chromatin much earlier in G1 in bob1 than in wild-type (WT) cells (Sclafani et al., 2002 ). Binding of Cdc45 before Mcm4 phosphorylation by Clb5,6-Cdk1 might lead to the assembly of a abnormal preinitiation complex that might be locked, at least at low temperatures, in a conformation that cannot easily trigger initiation. Another hypothesis is suggested from recent studies on Mcm2 phosphorylation (Montagnoli et al., 2006 ), in which six in vivo phosphorylation sites on human Mcm2 have been identified, three dependent on DDK and three on CDK. Two of these modifications affect adjacent serines, with DDK phosphorylating S40 and CDK phosphorylating S41. Strikingly, it was found that although both kinases could phosphorylate their cognate serines (S40 and S41) in vitro when the neighboring Ser was unphosphorylated, only DDK could do so when the other serine was already phosphorylated. That is, CDK is unable to phosphorylate S41 when S40 is already modified. Reciprocally, it was shown that Mcm2 phosphorylation by DDK is facilitated by prior phosphorylation by CDK (Masai et al., 2000 ). This suggests that CDK and DDK must act sequentially (CDK first) to phosphorylate hMcm2 on Ser41 and 40, respectively, which also fits the notion that DDK targets S/T residues that are either embedded in acidic stretches or next to residues already carrying (negatively charged) phosphate moieties. Priming of a DDK substrate by prior CDK phosphorylation has been recently demonstrated for the yeast meiotic recombination protein Mer2, and it could be a common mechanism for CDK-DDK coregulation (Wan et al., 2008 ). Mcm2 has an evolutionary divergent N-terminal domain that contains several such potential DDK-CDK biphosphorylation SSP (or STP) motifs. In fact, most of the potential CDK sites located in the Mcm4 N-terminal domains of many evolutionary distant species, as well as all five sites that we mutated in ScMcm4 belong to this category (Figure 1). It was shown recently that Mcm4 is phosphorylated by DDK in Xenopus and budding yeast (Takahashi and Walter, 2005 , Masai et al., 2006 ; Sheu and Stillman, 2006 ). Together with our data, these results suggest that one or more subunits of the MCM complex are phosphorylated both by CDK and DDK, whereby CDK might favor the action of DDK but, conversely, in which precocious phosphorylation by DDK might forestall the action of CDK on the preRC. Such a dependency in kinase function might provide for directionality along the sophisticated path of origin firing, which entails the orderly addition of various subcomplexes to the preRC before DNA synthesis actually begins. We propose that the amino-terminal extensions of eukaryotic Mcm2, 4, and 6 integrate regulatory signals conveyed by S-CDK and DDK. Failure to coordinate CDK and DDK activities on the MCM complex might destabilize the replisome and cause abortive firing of replication origins. We speculate that the regulation of the N-terminal extensions of MCM by S-CDK and DDK may also account for the modulation of fork progression rates seen in eukaryotes (Raghuraman et al., 2001 ) compared with Archaea (Lundgren et al., 2004 ).
We thank Drs. K. Shokat for providing cdc28-as1 and O. Aparicio for swe1Δ clb5::CLB2, 3, and 4 strains; G. Ammerer for mass spectrometry analysis; Dr. M. Matsuo for providing PhosTag; the Montpellier DNA Combing Facility for preparing surfaces; and T. Gostan for help with statistical analysis. This work was supported by grants from the Association pour la Recherche contre le Cancer (ARC 4704), the French Ministry of Research (ACI BCMS 0230), and Cancéropôle Grand Sud-Ouest.
This article was published online ahead of print in MBC in Press (http://www.molbiolcell.org/cgi/doi/10.1091/mbc.E07-06-0614) on March 5, 2008.