|Home | About | Journals | Submit | Contact Us | Français|
Origins of DNA replication are licensed in G1 by recruiting the mini-chromosome maintenance (MCM) proteins to form a pre-replicative complex. Prior to initiation of DNA synthesis from each origin, a pre-initiation complex (pre-IC) containing Cdc45 and other proteins is formed. We report that Cdc7-Dbf4 protein kinase (DDK) promotes assembly of a stable Cdc45-MCM complex exclusively on chromatin in S phase. In this complex, Mcm4 is hyper-phosphorylated. Studies in vitro using purified DDK and Mcm4 demonstrate that hyper-phosphorylation occurs at the Mcm4 N-terminus. However, the DDK substrate specificity is conferred by an adjacent DDK docking domain, sufficient for facilitating efficient phosphorylation of artificial phospho-acceptors in cis. Genetic evidence suggests that phosphorylation of Mcm4 by DDK is important for timely S-phase progression and for cell viability upon overproduction of Cdc45. We suggest that DDK docks on and phosphorylates MCM proteins at licensed origins to promote proper assembly of pre-IC.
To ensure genome inheritance, eukaryotes must duplicate all their chromosomes accurately and segregate them evenly to daughter cells in each cell division cycle (reviewed in Bell and Dutta, 2002; Diffley and Labib, 2002; Stillman, 2005). Duplication of chromosomal DNA in eukaryotes initiates from multiple replication origins in a temporally controlled manner during S phase. However, the mechanism of origin activation and hence the regulation of S-phase progression remain elusive.
The stepwise assembly of replication factors at origins provides multiple points of control to ensure efficiency and fidelity of DNA replication (reviewed in Bell and Dutta, 2002; Diffley and Labib, 2002; Stillman, 2005). In S. cerevisiae, the origin recognition complex (ORC) binds to replication origins throughout the cell cycle. At around the M/G1 transition when cyclin-dependent kinases (CDKs) become inactive, licensing factors recruit to origins the minichromosome maintenance (MCM) complex, forming a pre-replicative complex (pre-RC). The MCM complex is a heterohexamer composed of Mcm2, Mcm3, Mcm4, Mcm5, Mcm6 and Mcm7 and is likely a component of the helicase that unwinds DNA during replication (reviewed in Forsburg, 2004; Labib and Diffley, 2001; Tye and Sawyer, 2000). Accumulation of CDK activity during and after the G1/S transition prohibits pre-RC assembly. Conversely, S-phase CDKs, together with the Cdc7-Dbf4 kinase (also known as DDK, for Dbf4-Dependent Kinase (Nasmyth, 1996)) are required for origin activation. The activities of these kinases further recruit replication factors such as Cdc45, Sld2, Sld3 and the GINS complex to form a preinitiation complex (pre-IC) (Kamimura et al., 2001; Kanemaki et al., 2003; Masumoto et al., 2002; Takayama et al., 2003; Zou and Stillman, 1998).
Tight association of Cdc45 with an origin coincides with its temporal order of firing and thus Cdc45 is a marker for origin activation (Aparicio et al., 1999; Zou and Stillman, 2000). Cdc45 is also important for recruiting the polymerase α/primase complex to the origin to start DNA synthesis (Zou and Stillman, 2000). After origin firing, some of the initiation factors, such as Cdc45, GINS and the MCM complex, move with replication forks to carry on the elongation step of DNA synthesis (Aparicio et al., 1997; Gambus et al., 2006; Kanemaki et al., 2003; Labib et al., 2000; Takayama et al., 2003; Tercero et al., 2000). Moreover, Cdc45 and MCM are both required for DNA unwinding in Xenopus egg extracts (Pacek and Walter, 2004), suggesting that Cdc45, MCM and perhaps other pre-IC components function together as an active replicative helicase. Thus, in order to understand the mechanism of origin activation, it is important to study how Cdc45 recruitment to origins is regulated.
The S-phase kinase DDK is a conserved serine/threonine kinase composed of the catalytic subunit Cdc7 and the regulatory subunit Dbf4 (reviewed in Jares et al., 2000; Masai and Arai, 2002; Sclafani, 2000). Genetic evidence has suggested that DDK functions after CDK activation (Nougarede et al., 2000). Moreover, instead of functioning as a global switch, DDK is required for activation of individual origins throughout S phase to promote timely S-phase progression (Bousset and Diffley, 1998; Donaldson et al., 1998a; Pasero et al., 1999). DDK interacts with origins of DNA replication (Dowell et al., 1994), but the target of the interaction is not clear. Since DDK is also required for recruiting Cdc45 and other pre-IC components to origins (Kanemaki et al., 2003; Takayama et al., 2003; Zou and Stillman, 1998; Zou and Stillman, 2000), it is likely that DDK needs to be recruited to individual origins and phosphorylate crucial substrates there to convert each pre-RC into a pre-IC.
Several lines of evidence suggest that MCM proteins are prime targets for DDK. Genetic evidence showed that a mutation in the MCM complex, mcm5-bob1, partially bypasses the essential role of DDK (Hardy et al., 1997). Furthermore, an allele of DBF4 has been isolated as an allele-specific suppressor of mcm2-1 (Lei et al., 1997). In vitro kinase assays demonstrated several MCM subunits were substrates of DDK (Lei et al., 1997; Weinreich and Stillman, 1999), but the physiological significance is not known.
In this paper, we report the identification of a stable Cdc45-MCM complex exclusively on S-phase chromatin as a consequence of DDK action. DDK is recruited through a DDK docking domain on Mcm4 to facilitate hyper-phosphorylation of a separate domain at the Mcm4 N-terminus that is important for S-phase progression. We suggest that DDK is part of a molecular switch that activates the inert MCM apoenzyme by regulating assembly of the active replicative helicase complex at origins. Thus, DDK recruitment might be a key determinant of temporal and spatial control of origin firing.
To understand how Cdc45 is recruited to chromatin during S phase, we examined components associated with Cdc45 on the S-phase chromatin. Log-phase yeast cells were arrested in G1 phase using α-factor and in S phase using hydroxyurea (HU). A large-scale chromatin fractionation procedure (Fig. 1A), modified from a previous protocol was used (Liang and Stillman, 1997). Crude yeast extract was separated into a soluble protein and another fraction containing chromatin-bound proteins released after exhaustive DNase I digestion of the chromatin pellet (SN2). Cdc45 was immunoprecipitated from SN2 using polyclonal antibody against Cdc45. Analysis of the Cdc45 immunoprecipitate (IP) from the S-phase SN2 revealed several specific protein bands that were absent in the Cdc45 IP from the G1 SN2, in which Cdc45 was not present, or in the control IP using 12CA5 antibody (Fig. 1B. lanes 2 & 3). Mass spectrometry analysis demonstrated that these bands correspond to Cdc45 and all six MCM subunits. This result was confirmed by immuno-blot analyses (Fig. 1C, middle panel and data not shown). The stoichiometry of Cdc45 and each MCM subunits is not known, however, the intensity of each band in the silver-stained gel was comparable. In addition, the yeast Mcm3, Mcm5 and Mcm7 proteins individually co-purified with the yeast Cdc45 when co-expressed in the insect cells (unpublished results), suggesting direct interaction of Cdc45 with multiple MCM subunits.
MCM proteins were not detectable in the Cdc45 IP of SN1 fraction prepared from G1 and S-phase cells even though all the components are abundant (Fig.1C, left and middle panels). Moreover, the complex was not present on G1-phase chromatin since Cdc45 was not there. Thus the stable Cdc45-MCM interaction occurs only on chromatin and only during S phase. These observations suggest that assembly of the Cdc45-MCM complex is a regulated process. Interestingly, a portion of total Mcm4 on the S-phase chromatin was modified and exhibited a slower mobility in SDS polyacrylamide gel electrophoresis (SDS-PAGE) (Fig. 1C, left panel). The slower form of Mcm4 was greatly enriched after Cdc45 IP (Fig. 1C, right panel). Lamda phosphatase treatment of the Cdc45 IP demonstrates that this modification was phosphorylation (Fig. 1D). The mobility shift was prominent considering the size of the protein (~110 kDa) and the phosphatase treatment sometimes resulted in multiple intermediate bands, suggesting that the Mcm4 subunit in the Cdc45-MCM complex isolated from S-phase arrested cells is hyper-phosphorylated on multiple sites.
Because the HU-induced arrest in S phase also triggers a DNA damage checkpoint response, it raises the concern that hyper-phosphorylation of Mcm4 is a response to HU treatment rather than a normal S-phase event. To address this issue, yeast cells were arrested in G1 using α-factor and then released into fresh YPD medium. Cells were collected at different times after release and fractionated for chromatin-bound proteins (Fig. 1E). Immuno-blot analysis showed that a small portion of Mcm4 was hyper-phosphorylated as cells entered S phase (about 40–60 min after release), as measured by budding index and CDK-dependent phosphorylation of Orc6 (Liang and Stillman, 1997; Nguyen et al., 2001). Hyper-phosphorylation of Mcm4 occured concomitant with Cdc45 binding to chromatin. Cdc45 IP of SN2 prepared from the time-course samples co-precipitated the MCM complex containing mainly the hyper-phosphorylated Mcm4 in S phase (Fig. 1F). Thus, hyper-phosphorylated Mcm4 was specifically present in the Cdc45-MCM complex on chromatin during normal cell cycle progression in S-phase.
Because DDK is required for Cdc45 loading and origin activation and purified DDK can phosphorylate Mcm4 in vitro (Lei et al., 1997; Weinreich and Stillman, 1999; Zou and Stillman, 2000), we asked whether DDK was required for hyper-phosphorylation of Mcm4 on chromatin and Cdc45-MCM complex formation during S phase. Although the essential role of DDK can be bypassed in the mcm5-bob1 mutant (Hardy et al., 1997), mcm5-bob1 cdc7Δ cells proliferate much slower than mcm5-bob1 cells (Weinreich and Stillman, 1999); Fig. 2A), suggesting a partial bypass. Flow cytometry analysis of DNA content in cells released from G1 arrest showed that, compared with congenic mcm5-bob1 cells, mcm5-bob1 cdc7Δ cells were defective in S-phase entry and progression (Weinreich and Stillman, 1999); Fig. 2B). We isolated SN2 from mcm5-bob1 and mcm5-bob1 cdc7Δ cells arrested in HU. Because mcm5-bob1 cdc7Δ cells are larger than mcm5-bob1 cells, same volume of mcm5-bob1 cdc7Δ cell pellet contained roughly half the number of mcm5-bob1 cells. When same amount of total SN2 protein from each of mcm5-bob1 and mcm5-bob1 cdc7Δ cells was compared by immunoblot analysis, similar amounts of Mcm4, Mcm2, Mcm3 and Cdc45 were detected (Fig. 2C & D). While a small portion of Mcm4 was hyper-phosphorylated in SN2 from mcm5-bob1 cells, hyper- phosphorylation of Mcm4 was not detectable in SN2 from mcm5-bob1 cdc7Δ cells (Fig. 2C). The presence of Cdc45 in the mcm5-bob1 cdc7Δ chromatin is consistent with the idea that mcm5-bob1 mutation allows Cdc45 loading independent of DDK (Sclafani et al., 2002). However, when Cdc45 from each of these SN2 fractions was immunoprecipitated, much less MCM subunits were present in the purification from mcm5-bob1 cdc7Δ chromatin. This result suggests that, in the absence of DDK activity, the Cdc45-MCM complex is not formed efficiently or is less stable. Thus, DDK plays an important role in regulating Mcm4 phosphorylation and proper Cdc45-MCM interaction in vivo, in addition to promoting S-phase progression.
As illustrated in Fig. 3A, a major part of Mcm4 is very conserved among orthologs and contains distinct motifs of the AAA+ ATPase family and zinc finger motifs that are likely involved in double-hexamer formation (reviewed in Forsburg, 2004; Labib and Diffley, 2001; Tye and Sawyer, 2000). In contrast, the N-terminus of Mcm4 is not conserved in primary sequence, but is extremely S/T rich in all eukaryotic Mcm4 orthologs (29% in S. cerevisiae Mcm4). We thus term this region NSD (for N-terminal Serine/threonine-rich Domain).
To determine which region of Mcm4 is important for phosphorylation, various Mcm4 fragments were expressed and purified from bacteria and used as substrates for the yeast DDK expressed in and purified from insect cells (Fig. 3; in this article, substrates refer to protein substrates for the kinase; substrate ATP in this study is constant at 100 μM). Mcm4 fragments without NSD (mcm4175–933 and mcm4175–333) were not phosphorylated to a detectable level, indicating that NSD was essential for phosphorylation. NSD (mcm41–175) alone was phosphorylated by DDK albeit much less efficiently than the full-length protein (Fig. 3B, compare substrates 1 and 2). Nevertheless, this result indicateed that NSD contained DDK phosphorylation sites. Interestingly, mcm41–333 was phosphorylated as efficiently as the full-length protein (compare substrates 1 and 4), suggesting that, in addition to NSD, an additional element within amino acid residues 175–333 was important for efficient phosphorylation.
Notably, most of the 32P incorporation in mcm41–333 migrated slower than the bulk of the mcm41–333 protein in SDS-PAGE. The mobility shift was substantial and reminiscent of the Mcm4 protein shift in the Cdc45-MCM complex isolated from yeast chromatin. The shift is most likely due to high stoichiometric phosphorylation (i.e. multiple phosphorylation on a single molecule). Hence, we refer this shifted band as the hyper-phosphorylated form and the band co-migrating with the input protein as hypo-phosphorylated.
mcm4146–333, which contains a small portion of NSD, was phosphorylated efficiently. Mutation of multiple S/T to A in this portion of NSD (mcm4146–333, 5A+2A) abolisheed the phosphorylation, consistent with the presence of phospho-acceptor sites at the NSD. In contrast to mcm41–333, phosphorylation of mcm4146–333 did not result in a mobility shift. Thus, hyper-phosphorylation required an extended NSD.
Taken together, we identified two regions important for defining Mcm4 as the substrate of DDK. The NSD contains phospho-acceptor sites for DDK and an element in the conserved region next to NSD facilitates efficient phosphorylation, but is not itself a phosphorylation target.
Phosphorylation of mcm41–333 produced two major bands of distinctive gel mobility and few intermediate bands, suggesting that the kinase and the substrate did not dissociate after each phosphorylation event before the completion of stoichiometric phosphorylation (Duyster et al., 1995). Thus, hyper-phosphorylation of mcm41–333 was likely processive.
To further address the mechanism of hyper-phosphorylation, kinase assays were performed using various substrate concentrations (Fig. 4A). If phosphorylation was distributive, rather than processive, increasing the concentration of the substrate should result in a decrease in hyper-phosphorylation because the high concentration of unphosphorylated substrate would compete with the partially phosphorylated substrates each time substrate and enzyme dissociate (Burack and Sturgill, 1997). In contrast, increasing the concentration of mcm41–333 in the kinase assay did not cause a decrease in the amount of the hyper-phosphorylated product (Fig. 4A and B), consistent with non-distributive phosphorylation. Only when substrate concentration was near or lower than the enzyme concentration did the ratio of the two phosphorylated products change (Fig. 4C), suggesting that the hyper-phosphorylated form was produced from processive phosphorylation as a result of high affinity interaction between the substrate and the kinase through an additional interaction distinct from that of the catalytic site and the phospho-acceptor.
To determine whether Mcm4 binds to DDK, as is implied by the processive nature of phosphorylation, we expressed and purified various GST fusions of mcm4 fragments to test binding to purified DDK in vitro. Using a GST-pull down assay, GST-mcm41-333, GST-mcm4175–333, GST-mcm4146–333, and GST-mcm4146–333, 5A+2A, but not GST or GST-mcm41–175 bound to purified DDK (Fig. 4D). These interactions occurred both in the presence or absence of ATP (data not shown). In addition, GST-mcm4175–333, GST-mcm4146–333 and GST-mcm4146–333, 5A+2A interacted with DDK, suggesting that DDK phosphorylation sites were not required for the interaction. Therefore, Mcm4 bound to DDK through an element within amino acid residues 175–333.
The fact that mcm4175–333 bound DDK (Fig. 4D) and functions in cis, but not in trans, for efficient phosphorylation of NSD (Fig. 3B and data not shown) strongly suggested that this element promoted phosphorylation by binding to the kinase. This idea predicted that, if added in trans, mcm4175–333 could compete with mcm41–333 for binding to DDK and thereby interfere with its phosphorylation.
To test the prediction, the phosphorylation of mcm41–333 in the presence of various amounts of mcm4175–333 was analyzed. In the kinase assay using 30 nM kinase and 0.25 μM mcm41–333, mcm4175–333 inhibited phosphorylation of mcm41–333 in a concentration-dependent manner (Fig. 4E, lanes 7–11). The inhibition was evident even at 0.25 μM of mcm4175–333 fragment (compare lanes 6 and 7). In contrast, auto-phosphorylation on the Dbf4 subunit was not inhibited, but rather, increased in the presence of mcm4175–333, suggesting that the inhibitory effect of mcm4175–333 was specific to mcm41–333. The functional significance of auto-phosphorylation and its stimulation require further studies. Unlike mcm4175–333, NSD fragment (mcm41–175) only caused minor inhibition on the phosphorylation of mcm41–333 and only at high concentrations (Fig. 4E, lanes 1–5), probably due to competition between phospho-acceptors for the catalytic site on DDK.
These results suggest that mcm4175–333 contains a DDK-docking domain (DDD) and, together with phospho-acceptor sites within NSD, define Mcm4 as a specific substrate for DDK. As an added proof, we are able to create an artificial DDK substrate by attaching a heterologous phospho-acceptor sequence to the N-terminus of DDD (Fig. 3B, substrate 9).
The kinase docking model also provided mechanistic explanation for the processive, high stoichiometric phosphorylation of mcm41–333 by DDK. This model predicted that mcm4175–333 should have little effect on the high stoichiometric phosphorylation of mcm41–333 if the DDK-mcm41–333 complex was allowed to form first. The inhibitory effect was eliminated when mcm41–333 was pre-incubated with DDK before mcm4175–333 and ATP were added (Fig. 4F, compare lane 7 to lanes 8–12). For unknown reasons, pre-incubation reduced the overall phosphorylation. Nevertheless, addition of excess amounts of mcm4175–333 had little effect on hyper-phosphorylation of mcm41–333. This result argued that DDD-kinase interaction was the underlying mechanism for processive phosphorylation of NSD by DDK.
To further attribute elements in DDD that promote NSD phosphorylation, deletions of DDD from the C-terminus of mcm41–333 were prepared for in vitro kinase assay. Serial deletions down to mcm41–200 did not result in a significant decrease in phosphorylation levels (Fig. 3C). Rather, as the DDD region shortened, the relative levels of high stoichiometric phosphorylation decreased, along with the accumulation of bands corresponding to phosphorylation of intermediate stoichiometry, suggesting decreased processivity. Thus, shortening the DDD resulted in weaker binding of enzyme to substrates. Even though mcm41–200 has only 25 more additional residues than the NSD fragment (mcm41–175), it was phosphorylated much more efficiently than NSD alone. Therefore, residues 176–200 might constitute part of the DDK docking sites but additional sequences are needed for more efficient recruitment of DDK.
To determine whether the NSD is important for the biological function of Mcm4, we constructed various alleles of MCM4 with truncations of the NSD region and tested whether these mutants could rescue the lethality of mcm4Δ using a plasmid shuffle assay (Fig. 5A). A mutant, mcm4Δ2-174, which lacked the NSD, rescued mcm4Δ poorly. In contrast, mutant alleles with partial NSD (mcm4Δ2-73, mcm4Δ74-174, mcm4Δ2-112 and mcm4Δ2-145) rescued mcm4Δ as well as the full-length MCM4, suggesting that NSD, although dispensable, is important for the fitness of cells. The fact that either mcm4Δ74-174 or mcm4Δ2-73 can function normally in vivo suggests that residues 2-73 (20 S/T residues) and residues 74-174 (31 S/T residues) are redundant. Moreover, one of the alleles, mcm4Δ2-145, which supports normal cell growth, has only 29 residues (146–174) of NSD, with 10 S/T residues. The functional redundancy can explain why the primary sequence of NSD, apart from being S/T rich, is not conserved among Mcm4 orthologs.
The divergent nature of NSD prompted us to investigate whether phosphorylation sites from another DDK substrate can functionally replace NSD. Like Mcm4, Mcm2 has been shown to be a DDK substrate in vitro (Lei et al., 1997; Weinreich and Stillman, 1999). Genetic interaction between MCM2 and DBF4 has suggested Mcm2 as a likely physiological target for DDK (Lei et al., 1997). However, the functional significance of Mcm2 phosphorylation was not demonstrated. To test whether potential DDK target sites from Mcm2 could function in the context of Mcm4, we replaced the Mcm4 NSD with the N-terminal 200 residues of Mcm2 to generate the allele mcm21-200-mcm4Δ2-174. This fusion protein functionally replaced Mcm4 in vivo (Fig. 5A). In addition, the fusion protein mcm21-200-mcm4175-333 was a good substrate for DDK in vitro (data no shown). Further phenotypic analysis showed that the fusion protein functioned as well as the wild-type Mcm4 in supporting normal cell growth and timely S-phase progression (see below; Fig. 5B-E). These results further strengthen the idea that phosphorylation of multiple MCM components by DDK is of physiological importance.
The phenotypes of cells containing different alleles of MCM4 were further investigated. Two independent isolates of mcm4Δ2-174 cells grew very slowly in liquid YPD media at 30oC, compared with cells harboring the wild-type MCM4 (Fig 5B). In contrast, the growth curves of mcm4Δ2-145 and mcm21-200-mcm4Δ2-175 cells were very similar to that of the wild type. On YPD plates, the two mcm4Δ2-174 isolates grew slowly at 25oC, 30oC and 37oC and did not grow at 15oC and 16oC (Fig. 5C, ,6C6C and data not shown) while mcm4Δ2-73, mcm4Δ74-174, mcm4Δ2-112, mcm4Δ2-145 and mcm21-200-mcm4Δ2-175 cells grew just like wild-type cells at those temperatures. Thus, mcm4Δ2-174 cells were cold-sensitive.
Flow cytometry analysis of asynchronous cell populations revealed that mcm4Δ2-174 cells accumulated more S-phase cells compared to MCM4 and mcm21-200-mcm4Δ2-175 cells (Fig. 5D). After treatment with HU, MCM4 and mcm21-200-mcm4Δ2-175 cells arrested mostly in the early S phase, while mcm4Δ2-174 cells arrested at all stages across the S phase, displaying a broader peak between 1C and 2C DNA content. This result suggested that mcm4Δ2-174 cells take more time going through S phase than MCM4 and mcm21-200-mcm4Δ2-175 cells, explaining why HU induced a broad S phase pattern of DNA content.
To further assess the S-phase progression of these cells, we arrested cells in G1 using α-factor, released them into fresh medium and collected cells for flow cytometry analysis at indicated time points (Fig. 5E). While MCM4 and mcm21-200-mcm4Δ2-175 cells were well into S phase 30 min after release and entered the next G1 by 105 min, mcm4Δ2-174 cells just entered S phase at 45 min and progressed slowly across S phase and most cells still did not reach the next G1 by 150 min. The delayed entry and progression of S phase is reminiscent of what was observed for mcm5-bob1 cdc7Δ cells (Weinreich and Stillman, 1999); Fig. 2B). Taken together, NSD of Mcm4 is essential for timely entry and progression through S phase.
Since mcm4Δ2-145 was able to rescue mcm4Δ as efficiently as the full-length Mcm4, we asked whether the remaining 10 S/T residues in this region were important for the function of the protein. Different combinations of S/T to A mutations were made in this short NSD (Fig. 6A) and their ability to rescue mcm4Δ determined (Fig. 6B). Mutation alleles mcm4Δ2-145, 2A and mcm4Δ2-145, 3A+2A rescued mcm4Δ as efficiently as mcm4Δ2-145 while mcm4Δ2-145, 5A functioned less than mcm4Δ2-145 but better than mcm4Δ2-174. Mostly noticeably, mcm4Δ2-145, 5A+2A, which has 7 S/T to A mutations, rescued mcm4Δ as poorly as mcm4Δ2-174. Like mcm4Δ2-174, mcm4Δ2-145, 5A and mcm4Δ2-145, 3A+2A exhibited a cold-sensitive phenotype and did not grow at 15oC (Fig. 6C, sectors 5 and 6). Moreover, as shown earlier in the in vitro kinase assay (Fig 3B), an mcm4 fragment containing the same short NSD as mcm4Δ2-145 (mcm4146-333) was phosphorylated efficiently by DDK while the substrate with the same 7 S/T to A mutations (mcm4146-333, 5A+2A) was not. Thus, the in vivo function of the Mcm4 N-terminus correlates with its capacity as a DDK substrate.
Unexpectedly, addition of the 3xHA sequence, which contains no S/T residues, to the N-terminus of mcm4Δ2-174 abolished the function of the protein (Fig. 6A and 6B). We asked whether re-introducing serine residues into this heterologous sequence might allow phosphorylation of the resulting fusion protein SS/HA-mcm4Δ2–174 (Fig. 6A) and thereby render it functional. Indeed, SS/HA-mcm4Δ2-174 not only functionally complemented mcm4Δ, but it grews better than mcm4Δ2-174 and was rescued from the cold-sensitive phenotype (Fig.6B and 6C, sectors 7 and 8). In addition, in vitro kinase assay demonstrated that SS/HA-mcm4175-333 was phosphorylated by DDK while 3xHA-mcm4175–333 was not (Fig. 3B, substrates 8 and 9). Taken together, these data demonstrate that phosphorylation of Mcm4 at its N-terminus by DDK is required for cell growth.
Since phosphorylation of Mcm4 by DDK is important for cell growth and DDK is required for formation of the stable Cdc45-MCM complex on the S phase chromatin, it is likely that these processes are functionally linked. Such link may be revealed in by synthetic genetic interactions.
We examined the effect of Cdc45 overproduction in MCM4, mcm4Δ2-174, mcm4Δ2-145 and mcm4Δ2-145, 3A+2A cells. Control vector derived from a CEN-based plasmid, pRS416, carrying only GAL1,10 promoter sequence or the vector containing GAL1,10::CDC45 were introduced into these cells. On the SC-URA plate, where the GAL1,10 promoter was repressed, MCM4, mcm4Δ2-174, mcm4Δ2-145 and mcm4Δ2-145, 3A+2A cells carrying either vectors grew similarly to each other (Fig. 7, left panel). On the SC-URA-Gal plate, where the GAL1,10 promoter is induced, MCM4 and mcm4Δ2-145 cells carrying either vectors also grow similar to each other (Fig. 7, right panel, compare sectors 1 and 2 and sectors 5 and 6), indicating that overproduction of Cdc45 did not affect the growth of these cells. In contrast, on the same plate, mcm4Δ2-174 and mcm4Δ2-145, 3A+2A cells carrying GAL1,10::CDC45 grew poorly and formed micro-colonies that eventually stopped growing, while cells harboring control vectors grew just fine (Fig. 7, right panel, compare sectors 3 and 4 and sectors 7 and 8). Thus, overproduction of Cdc45 impairs the growth of NSD mutants in MCM4. This genetic interaction, is consistent with the hypothesis that phosphorylation of the MCM complex by DDK is important for the regulated assembly or activity of the Cdc45-MCM complex.
We show here that Cdc45 and MCM form a DDK-dependent complex that exists only in S phase and exclusively on chromatin, consistent with prior reports on co-immunoprecipitation (IP) of Cdc45 with several subunits of the MCM complex (Dalton and Hopwood, 1997; Hopwood and Dalton, 1996; Zou and Stillman, 1998). A recent study reported isolation of large replisome progression complexes (RPCs) that contains MCM, Cdc45, GINS and other regulators of the replication fork (Gambus et al., 2006). Our preliminary data suggests that the Cdc45-MCM complex isolated in this study also contains GINS components (unpublished). Since Cdc45 and MCM are required for DNA unwinding in the Xenopus egg extract system (Pacek and Walter, 2004) and a weak helicase activity was detected in the MCM immunoprecipitates of chromatin assembled in the presence of Cdc45 (Masuda et al., 2003), it is possible that this complex may contain the active replicative helicase holoenzyme.
Cdc45 interacts individually with Mcm3, Mcm5 and Mcm7 when co-expressed in insect cells (unpublished). These observations are also supported by the identification of cdc46-1 (an allele of MCM5) and cdc47-1 (an allele of MCM7) as allele-specific suppressors of cold-sensitive cdc45-1 (Dalton and Hopwood, 1997; Moir et al., 1982). Together, these results suggest that binding of Cdc45 to the MCM complex is direct.
DDK is required for activating individual licensed origins and for Cdc45 loading onto origins (Aparicio et al., 1999; Bousset and Diffley, 1998; Donaldson et al., 1998a; Pasero et al., 1999; Zou and Stillman, 1998; Zou and Stillman, 2000). Thus regulation of Cdc45-MCM complex formation is likely an underlying mechanism for temporal control of origin activation. A portion of the chromatin bound Mcm4 subunit is hyper-phosphorylated in a cell cycle-regulated, DDK-dependent manner. Furthermore, hyperphosphroylated Mcm4 is dramatically enriched in the Cdc45-associated MCM complex. Importantly, phosphorylation of Mcm4 by DDK is crucial for normal cell growth and S phase progression. The physiological role of Mcm4 phosphorylation is relevant to Cdc45 function as overproduction of Cdc45 is detrimental in NSD mutants lacking DDK phosphorylation sites.
A weak interaction of Cdc45 with origins of DNA replication can be detected in late G1 phase, before DDK activation, but only in the presence of protein cross-linking reagents, suggesting that Cdc45 may be recruited to origins by another protein, perhaps Sld3 (Aparicio et al., 1999; Kamimura et al., 2001; Kanemaki and Labib, 2006). Thus the relationship between Mcm4 phosphorylation and Cdc45-MCM complex assembly may be more complicated than a simple phospho-regulated interaction between Cdc45 and Mcm4. One possibility is that DDK phosphorylation of MCM may activate helicase activity. Another possibility is that DDK phosphorylation of DDK may promote stable Cdc45-MCM complex assembly by promoting association or dissociation of another protein, perhaps GINS or Sld3.
In addition to CDK and DDK, origin binding of Cdc45 also requires other proteins, such as Sld3 and the GINS complex (Kamimura et al., 2001; Kanemaki and Labib, 2006; Kanemaki et al., 2003; Takayama et al., 2003). Phosphorylation of MCM by DDK may affect origin association of these factors that in turn are required for Cdc45-MCM complex assembly, or it might promote dissociation of factors such as Sld3. Overproduction of Cdc45 restores origin association of a hypomorph sld3-5 mutant, which has reduced origin-association (Kamimura et al., 2001). In contrast, overproduction of Cdc45 exacerbates the defect in mcm4 NSD mutants. This is not due to the toxicity of Cdc45 overproduction because this synthetic phenotype does not occur in wild type. A recent report shows that Sld3 is displaced from origin while GINS is recruited to origins, and, as a result, Cdc45 and GINS become stably engaged and move away with replication fork during initiation (Kanemaki and Labib, 2006). Thus, it is possible that Sld3 may need to be removed before the Cdc45 and GINS are stably engaged at origins and that this does not happen efficiently when Cdc45 is over-expressed in an mcm4 NSD mutant.
Kinase assays have suggested that, in addition to Mcm4, other MCM subunits, such as Mcm2, may be physiological substrates for DDK (Hardy et al., 1997; Lei et al., 1997; Weinreich and Stillman, 1999). The N-terminus of Mcm2 can functionally substitute in vivo for the Mcm4 NSD. Mcm6 might also be a good candidate as a physiological substrate of DDK since it also has an extended S/T-rich N-terminal region. The presence of multiple DDK targets in the MCM complex allows certain biological robustness and explains why NSD mutants of MCM4, although defective, are still viable and replicating DNA. It may require simultaneous elimination of multiple MCM N-termini to achieve a complete loss of function.
In this study, we show that two regions in Mcm4, NSD and DDD, together define Mcm4 as a substrate of processive phosphorylation by DDK. DDD rather than NSD is the main determinant of DDK substrate specificity and the main role of NSD is to provide phospho-acceptors for DDK. DDD can act in cis to facilitate efficient phosphorylation of an artificially created phospho-acceptor sequence attached to its N-terminus. Importantly, this artificial phospho-acceptor sequence can functionally substitute for the NSD of Mcm4 in vivo, consistent with the degenerate nature of the Mcm4 NSD sequence during evolution.
The degeneracy of the NSD suggests that its function is regulatory, rather than structural. It is noteworthy that the archaeal MCM proteins lack an N-terminus analogous to the NSD in Mcm4. Interestingly, the purified hexameric MCM complex from archaeal species has DNA helicase activity, but the hexameric eukaryotic MCM lacks such an activity in vitro (Tye and Sawyer, 2000). Thus, it is possible that the extended amino termini of eukaryotic MCM complex are inhibitory to the intrinsic MCM helicase activity and DDK-mediated loading of Cdc45 and other factors and/or hyper-phosphorylation of MCM subunits relieves this inhibition.
The ability of the DDD fragment mcm4175-333 to increase DDK auto-phosphorylation raises additional possibility that binding of DDD to DDK might cause a conformational change in DDK or removal of an auto-inhibitory domain that increases its activity. Either model of kinase activation is not necessarily exclusive of the kinase-recruitment mechanism. However, kinase activation model cannot solely account for the processive phosphorylation because DDD does not increase phosphorylation of NSD by DDK when added in cis.
The kinase recruitment model also predicts an additional substrate recognition domain in DDK distinct from the catalytic site in the kinase. A similar mode of substrate recognition has been reported for other kinases. For example, in the case of CDKs, the “hydrophobic patch” of the cyclin plays a main role in substrate recognition (Adams et al., 1996; Kelly et al., 1998; Schulman et al., 1998). Certain members of Polo-like kinases use the C-terminal Polo-box domain to interact with specific substrates (Lowery et al., 2005).
The DDD region of Mcm4 may contain multiple kinase docking sites to ensure highly processive phosphorylation in vivo. Serial deletions from the C-terminus of to mcm41-333 down to mcm41-200 show that, as the DDD region shortens, phosphorylation becomes less processive and products corresponding to phosphorylation of intermediate stoichiometry accumulate. The additional 25 residues in mcm41-200 render it a much more efficient substrate for DDK than the NSD fragment (mcm41-175), although phosphorylation is less processive. Residues 176-200 retain enough of the DDK docking function to promote semi-processive phosphorylation of NSD. Further analyses are needed to pinpoint the minimal unit and the consensus sequence for DDK interaction and thereby to predict other potential substrates for DDK. We notice that sequences related to a short motif “FRNFLMSF” within residues 176 to 200 of Mcm4 are present in other MCM subunits. Preliminary studies show that mutating multiple aromatic and hydrophobic residues within this motif abolishes the function of Mcm4. However, the functional relevance of this motif in kinase recruitment remains to be addressed.
Processive phosphorylation of Mcm4 by DDK is likely to be physiologically relevant because it allows a switch-like activation of origins via concerted phosphorylation of an MCM complex at a particular origin, rather than by gradual phosphorylation is the kinase reaction was distributive. This switch-like mechanism may explain why origins fire individually in a defined temporal pattern throughout S phase rather than fire randomly in a distributive manner upon activation of DDK. We speculate that early origins may preferentially recruit DDK and thus are activated earlier. The processive nature of phosphorylation also raises the possibility that prior phosphorylation events may increase the affinity of kinase-substrate interaction. Although the in vitro kinase assay clearly indicates that prior CDK phosphorylation is not essential for Mcm4 phosphorylation by DDK, it does not rule out the possibility that CDK still has a facilitating role. Since Clb5-CDK is required for late origin firing (Donaldson et al., 1998b), we speculate that CDK phosphorylation may increase the probability that late origins are activated, therby ensuring complete replication of the genome before M phase.
To conclude, we incorporate our findings and other relevant reports and propose a model for activation of licensed origins (Fig. 7B). At the heart of this model is DDK recruitment, which is likely the key determinant of temporal and spatial control of origin firing. In this model, Cdc45 and Sld3 cooperate to establish weak association with origins during G1 (Kamimura et al., 2001; Kanemaki and Labib, 2006). This weak association is CDK-independent and is thus presumably DDK-independent because DDK can exert its function in DNA replication only after CDK is activated (Nougarede et al., 2000). At the G1/S transition, and throughout S phase, DDK is recruited to individual origins through interaction, at least in part, with DDD on Mcm4, to facilitate concerted phosphorylation on the N-terminal side of multiple MCM subunits. As a consequence of concerted phosphorylation, MCM may assume a change in conformation or surface charge. We speculate that this change in MCM may displace Sld3 and allow stable and active engagement of Cdc45 and GINS to the MCM complex on the origin. The stable complex containing MCM, Cdc45 and GINS is likely part of the active replicative helicase complex that unwinds DNA during S phase.
Yeast media and cell synchronization were as described previously (Weinreich and Stillman, 1999). Samples for flow cytometry analysis were prepared as described (Haase and Reed, 2002). The processed samples were analyzed using LSRII cell analyzer (BD Biosciences).
Yeast strains and plasmids are described in supplementary materials. Mcm4 N-terminal deletion mutants and site directed mutants were created using fusion PCR. The mcm4 mutant PCR products were digested with SphI/StuI, cloned into SphI/StuI sites of pEM54.3 (pMCM4/LEU2). Constructs were confirmed by DNA sequencing.
The mcm4 mutants were tested for function in vivo using plasmid shuffle assay, in which the LEU2-containing plasmids were transformed into YS960 (mcm4Δ::TRP1 + pMCM4/URA3) and selected on SC-Leu-Ura plates. The transformants were isolated and grown in YPD overnight and 10-fold serial dilutions starting from 105 cells were spotted onto 5-FOA plates to select for loss of URA3 plasmid carrying the wild-type MCM4. The same dilutions were spotted on YPD plates as control sets. FOA-resistant clones were isolated for further phenotypic analyses.
Synthetic peptide Cdc45-I2 (CDEDEEDEDETISNKRGN) derived from Cdc45 with extra cysteine at the N-terminus were conjugated to KLH (Pierce) and used for inoculation into rabbits for polyclonal antibodies production (Covance).
Chromatin pellet of YB468 was prepared as described previously (Liang and Stillman, 1997). The soluble fraction is SN1. The pellet contains chromatin. Chromatin-bound proteins were released into solution by treating the pellet with 100U/ml of DNaseI for at least 30 min on ice. For buffer condition and detail see supplementary materials.
For Cdc45 IP, about 10 ml of SN2 was passed through 1 ml of Protein A-Sepharose (Amersham), incubated with 50 μl Protein A-Sepharose coupled with purified polyclonal antibodies against Cdc45, washed 7 times with 10 ml EBX buffer. The bound proteins were eluted with 2 mg/ml Cdc45-I2 in EB buffer. See supplementary materials for buffer composition. 10 % of the elution was analyzed by SDS-10% PAGE and silver stain. Immuno-blot analysis was performed as described previously (Weinreich and Stillman, 1999).
For mass spectrometry analysis of proteins from Cdc45 IP, SYPRO Ruby (Bio-Rad) stained bands were excised from the gel, digested with modified trypsin, and analyzed by Matrix-assisted laser desorption ionization-time of flight (MALDI-TOF). The mass spectrometry was done by Dr. M. Myers in the Cold Spring Harbor Lab.
Full-length or truncated Mcm4 fragments were amplified by PCR and cloned into the NdeI/NotI sites of pET-21a. The resulting constructs express proteins with a C-terminal His6 tag. The proteins were expressed from BL21 CodonPlus (DE3) by induction with 1 mM IPTG at 15~18oC for 20 hours. Typically, 350 ml of induced culture was collected and resuspended in 10 ml of lysis buffer (50 mM Tris pH 7.4, 100 mM NaCl, protease inhibitor tablets (Boehringer), and 1 mM PMSF). Lysozyme (Sigma) was added at 1 mg/ml and the cell suspension was incubated on ice for 30 min, sonicated 3 times (20% amplitude for 15 sec and rest on ice for 30 sec), and centrifuged at 17, 000 g (12,000 rpm in SS34) for 20 min at 4oC. The supernatant was collected and NaCl was added to a final concentration of 500 mM. The extract was centrifuged again and the supernatant was collected and filtered through 0.45-μm grids. The HIS6-tagged proteins were affinity-purified using Talon IMAC resin (Clontech) as described by manufacturer.
GST-mcm41-333 was constructed by sub-cloning the NdeI/NotI fragment containing mcm41-333 into pGEX-KG. GST-fusion proteins were expressed from BL21 CodonPlus (DE3) cells and crude extract was prepared as described above. Proteins were affinity purified on glutathione-agarose beads (Pharmacia) as described by the manufacturer.
The purification of DDK and in vitro kinase assay was performed as described (Weinreich and Stillman, 1999) with indicated amount of DDK and protein substrates. 10 μl of the kinase reaction was fractionated by SDS-11 % PAGE, stained with Coomassie Blue, dried and autoradiographed or analyzed using FLA-5100 Imaging System (FUJIFILM Life Science).
In vitro binding reactions contained 5 μg of GST fusion protein immobilized on 10 μl of glutathione-agarose beads and 2 μg of purified DDK in 100 μl of binding buffer (40 mM Tris-HCl pH 7.6, 100 mM NaCl, 0.1 mM EDTA, 10% (v/v) glycerol, 0.1 % Triton X- 100, 10 mM MgCl2, 200 μM ATP, 2 mM DTT, 1 mM PMSF and protease inhibitors) containing 1 mg/ml bovine serum albumin. Binding reactions were incubated for 1 h at 4°C with mixing and washed 5 times with 500 μl binding buffer. The beads were resuspended in 20 μl sample buffer, boiled and 30 % of the binding reaction was fractionated by SDS-10 % PAGE. Pull-down of DDK was analyzed by immuno-blot as described previously (Weinreich and Stillman, 1999).
This work was supported by a National Institutes of Health grant (GM4436) and Y.-J.S. was a Helen Hay Whitney postdoctoral fellow. We thank Dr. Michael Myers and the Cold Spring Harbor Proteomics shared resource for mass spectrometry sequencing and members of the Stillman lab for helpful discussions. We also thank anonymous reviewers for excellent comments.