Recapitulating the G1 to S phase events of replication in vitro
A major obstacle to the recapitulation of eukaryotic DNA replication initiation in vitro
is the incompatibility of the cell cycle conditions required for helicase loading (G1) and activation (S). To overcome this hurdle, we simulated the G1 to S phase transition using a combination of S. cerevisiae
extracts, similar to the approach used for nucleus- and origin-independent replication using Xenopus
egg extracts (Walter et al., 1998
). First, we used G1-arrested extract supplemented with purified Cdc6 to load the replicative helicase onto immobilized ARS1
origin DNA (Bowers et al., 2004
; Seki and Diffley, 2000
). The loaded Mcm2-7 complexes were isolated from the G1 extract and activated by incubation with an S-phase extract ().
An Assay for Replisome Assembly in vitro
S phase extracts were prepared from cells modified in two ways to enhance their replication capacity. First, these cells contained a temperature-sensitive allele in the DDK catalytic subunit Cdc7 and were arrested at the non-permissive temperature before extract preparation. Thus, the arrested cells are poised for replication initiation with unreplicated DNA but elevated S-CDK levels. To compensate for a lack of DDK activity, we treated loaded Mcm2-7 with purified DDK prior to addition of S phase extract. Second, these cells overproduced Sld2, Sld3, Dpb11, and Cdc45, which are normally expressed at low levels (Ghaemmaghami et al., 2003
). Thus, after origin loading in the G1 extract, the Mcm2-7 helicase is exposed to both essential replication-activating kinases and an extract containing a robust source of the proteins required for origin activation.
After sequential treatment of the loaded Mcm2-7 with DDK and S-phase extract, we observed origin association of the helicase activators Cdc45 and Psf2 (a GINS subunit) as well as Mcm10 (). These associations were dependent on the addition of S-phase extract (, lanes 1-2), an intact origin sequence (A-B2-, lane 3), and prior Mcm2-7 loading (-Cdc6, lane 4). If the temperature-arrested S phase extract was replaced with extracts prepared from hydroxyurea (HU) or G1-arrested cells overexpressing Cdc45, Dpb11, Sld2 and Sld3 then Cdc45, GINS and Mcm10 failed to associate (, lanes 5-6). Interestingly, providing additional DDK to these extracts restored recruitment of all three proteins to the HU extract and Cdc45 recruitment to the G1 extract. In contrast, a nocodazole-arrested extract overexpressing the same four proteins (, lane 7) showed a similar pattern of protein recruitment with or without added DDK, suggesting that when all factors are present, there is no M-phase barrier to replisome assembly. Together, these properties mirrored the hallmarks of origin activation in vivo.
Distinct requirements for Cdc45 and GINS origin recruitment
We investigated the interdependencies of replication protein recruitment to origin DNA () by immunodepleting individual factors from the S phase extract and assessing the ability of other replication proteins to associate with the origin. In each case, addition of purified forms of the depleted protein (Fig. S1
) restored replication protein recruitment, indicating that the depleted extracts remained active and that other essential proteins had not been co-depleted.
Interdependent Recruitment of Replisome Proteins
Analysis of the depleted extracts uncovered distinct requirements for the recruitment of the helicase activating proteins Cdc45 and GINS (). Only Sld3 depletion resulted in a loss of Cdc45 association, although depletion of GINS showed reduced Cdc45 recruitment. In contrast, Cdc45, Sld3, Sld2 and Dpb11 were each required for stable GINS recruitment. Finally, unlike studies of Xenopus
Mcm10 (Wohlschlegel et al., 2002
), Cdc45, Dpb11 and GINS associated with the origin DNA in the absence of Mcm10 (). In addition, Mcm10 recruitment was eliminated by depletion of any of the other proteins tested.
A Biochemical Assay for Origin-dependent Replication Initiation
Given that helicase-activating proteins were recruited to the origin-containing DNA template, we probed the reaction for the completion of later steps in the replication initiation process. A ~1 kb linear template poorly supported Pol α recruitment and nucleotide incorporation ( and data not shown). In contrast, a larger, 5.9-kb ARS1 plasmid robustly supported both activities (). Reactions containing the plasmid template included six-fold fewer copies of ARS1 (due to less efficient bead attachment) than reactions with the 1 kb linear template (, compare ORC levels). Nevertheless, the templates showed similar levels of Mcm2-7 loading yet Pol α and replication levels were much higher for the plasmid template. Thus, plasmid DNA was more efficient for helicase loading, polymerase loading, and replication initiation.
Long DNA Templates Supports Polymerase Loading and Replication Initiation
To exclude the possibility that the observed nucleotide incorporation is the result of non-specific repair events, we tested for properties expected for genuine replication products. Nucleotide incorporation was dependent on prior pre-RC formation and ATP hydrolysis (, lanes 1-3). Examination of the replication products by native agarose electrophoresis revealed that nucleotide incorporation was associated with a shift in the mobility of the plasmid (), consistent with formation of theta replication intermediates. Replication products synthesized in the presence of BrdUTP migrated at the position expected for heavy:light DNA in a CsCl gradient, indicating that replication was semi-conservative (). In addition, no reaction products were detected in the presence of aphidicolin, a potent inhibitor of eukaryotic replicative DNA polymerases (, lane 4), and nucleotide incorporation accumulated for at least 60 minutes of incubation (). Most importantly, the replication reaction depended on defined origin sequences. DNA templates with an ARS1 origin lacking an ORC binding site (A-B2-) showed dramatically reduced replication (, lanes 5-11).
DNA length rather than supercoiling or the circular nature of the plasmid DNA was required for replication. Direct comparison of replication using a 7.6 kb ARS1 plasmid (randomly biotinylated) or a linearized version of the same plasmid (biotinylated at one end) showed that the linear template replicated 2-fold more efficiently than its circular counterpart (, lanes 8-11), due to longer replication products. This finding suggests that the random attachment of the circular DNA to the magnetic bead inhibited replication by impeding replication forks. We determined total nucleotide incorporation and found that ~ 3% of the total plasmid DNA is replicated in the assay. Incomplete Mcm2-7 loading and replication elongation appear to be the primary reasons for the low levels of incorporation (see discussion).
To determine if the over-expression of Cdc45, Dpb11, Sld2 and Sld3 was important for DNA replication, we tested S-phase extracts from cells with endogenous protein levels. These extracts failed to either initiate replication or recruit GINS or Mcm10 (, lane 2). By adding purified and active forms of the limiting proteins (see ) to the S phase extract, we observed that Cdc45, Sld2 and Dpb11 were each limiting for both events (). Together, these data indicate that this assay accurately recapitulates replication initiation, displaying a dependence on a defined origin, the replicative DNA polymerases and multiple essential replication initiation proteins.
DDK and CDK Are Required For Distinct Steps In Origin Activation
We next asked how the DDK and CDK kinases affected replication factor recruitment and replication initiation. We eliminated DDK activity by omitting purified DDK and S-CDK activity by addition of the S-CDK inhibitor Sic1. Both kinases were essential for replication initiation but their loss had distinct effects on replication protein recruitment (). In the absence of DDK, none of the factors examined associated with origin DNA (). In contrast, S-CDK activity was required for Mcm10, Dpb11 and GINS association but not for Sld3 and Cdc45. The association of Cdc45 and Sld3 in the absence of S-CDK suggested that DDK drives the formation of an initial complex (DDK-dependent complex) that is then acted upon in an S-CDK-catalyzed event to recruit Dpb11, GINS and Mcm10. Consistent with a more robust association of Cdc45 with origins upon entry into S phase (Aparicio et al., 1999
), salt extraction experiments showed that Cdc45 association is stabilized by the recruitment of the S-CDK dependent factors (Fig. S2
DDK and S-CDK Are Required For Distinct Stages of Origin Activation
Our findings predict that Sld3 and Cdc45 origin recruitment depends on DDK, however, Sld3 and Cdc45 associate with early origins in G1 (Aparicio et al., 1999
; Kamimura et al., 2001
; Kanemaki and Labib, 2006
), a time when Dbf4 is a target of APC-dependent degradation (Sclafani and Holzen, 2007
). To address the role of DDK during the G1 recruitment of Cdc45 and Sld3, we compared their origin association in CDC7
wild type and temperature sensitive (cdc7-4
) cells. Due to reduced Cdc45 and Sld3 origin binding at 37°C in wild-type cells (data not shown), we performed this analysis at 25°C (Cdc45) or 32°C (Sld3). Using either ChIP-Chip (Cdc45) or ChIP-PCR (Sld3), we found that association of Cdc45 and Sld3 with early-firing origins during G1 (ARS305, ARS306
) was reduced in cdc7-4
cells (). Weak Cdc45 association with some late-firing origins was not reduced by the cdc7
mutation (, ARS316
), potentially due to residual Cdc7 activity at 25°C. Analysis of sites of Cdc45 binding reduced in the cdc7-4
background identified 49 origins, most of which initiate in the first 20% of S phase (Table S1
). Thus, DDK is active in late G1 cells and drives the association of Cdc45 and Sld3 with early origins prior to the S-CDK-dependent recruitment of Sld2, Dpb11 and GINS.
Additional experiments support a model in which DDK acts prior to CDK at the origin. We observed the highest levels of replication protein origin association and replication initiation when the loaded helicases were exposed to DDK first, then exposed to CDK in the S-phase extract (, DDK →CDK). Addition of DDK to the S phase extract exposed the loaded helicases to both kinases simultaneously (DDK with CDK) and resulted in lower protein association and initiation. Finally, if loaded helicases were exposed to CDK and S-phase extract followed by CDK inactivation and addition of DDK (CDK →DDK), we observed no replication initiation (). Intriguingly, under these conditions association of Cdc45, GINS and Mcm10 were dramatically reduced () even though equivalent DDK phosphorylation of Mcm4 was observed (, lanes 6 and 7). This suggests that prior exposure to CDK prevents subsequent DDK phosphorylation of Mcm2-7 from driving origin recruitment of Cdc45 or GINS.
Distinct Requirements for Leading and Lagging DNA Polymerase Recruitment
Three DNA Polymerases act at the eukaryotic replication fork but the assembly of these enzymes at the replisome is poorly understood. Because of their affinity for ssDNA templates, DNA polymerase recruitment could require origin unwinding but this has only been addressed for DNA Pol α (Walter and Newport, 2000
). Alternatively, DNA polymerases could directly or indirectly interact with loaded helicase even in the absence of DNA unwinding. Mcm10 has been shown to interact with and stabilize Pol α/primase, however, its role in the initial recruitment of DNA polymerases is unknown.
We first asked if Mcm10 was involved in DNA polymerase recruitment. Depletion of Mcm10 dramatically reduced Pol α loading and DNA replication (). These effects were not due to the destabilization of Pol α in the absence of Mcm10 (Ricke and Bielinsky, 2004
), because addition of purified Mcm10 restored both events. Mcm10 depletion had little effect on Pol ε recruitment but reduced Pol δ association by half. Notably, addition of purified MBP-Mcm10 stimulated Pol δ recruitment and DNA synthesis compared with the unperturbed extract (, lanes 7-9, 5B), suggesting that Mcm10 facilitates Pol δ origin recruitment and that Mcm10 was limiting for this event.
Mcm10 is Required for the Recruitment of Pol α and Pol δ to Origin DNA
To investigate the connection between origin unwinding and replisome assembly we monitored association of the ssDNA binding protein RPA with the template DNA. RPA association with the circular template was dependent on ATP hydrolysis (, +ATPγS), pre-RC formation (, -Cdc6), and Cdc45 (). ATPγS was added after DDK phosphorylation of the Mcm2-7 complex. Consistent with DDK functioning in the ATPγS reaction, Sld3 and Cdc45 are recruited to the origin under these conditions (Fig. S4
). Thus, Cdc45 and Sld3 do not require DNA unwinding for their recruitment, consistent with studies showing that inactivation of Mcm2-7 ATP binding motifs does not interfere with Cdc45 recruitment (Ying and Gautier, 2005
ATP Hydrolysis is Required for RPA Loading and for the Loading of a Subset of Polymerases
Because it was added to the reaction after DDK phosphorylation of loaded Mcm2-7, ATPγS could prevent origin unwinding in two ways: inhibition of CDK activity and/or inhibition of Mcm2-7 ATPase activity. Since we knew the effects of CDK inhibition (), we sought conditions in which ATPγS specifically inhibited Mcm2-7. To this end, we exploited a mutant in Cdk1 (Cdk1-AS) that preferentially binds and hydrolyzes modified ATP (Ubersax et al., 2003
). We added purified Clb5-Cdk1-AS and a hydrolyzable form of the modified ATP (6-benzyl-ATP) along with ATPγS to the S phase extract (). Importantly, in these conditions we observed CDK-dependent phosphorylation of Orc6 () but only background levels of RPA association with origin DNA. Thus, in these conditions, ATPγS inhibits an event downstream of CDK function required for DNA unwinding, most likely ATP hydrolysis by Mcm2-7.
To determine which replication proteins and DNA polymerases required DNA unwinding for origin recruitment, we assessed replication protein recruitment in the presence of ATPγS, Clb5-Cdk1-AS and 6-benzyl ATP (). Consistent with the restoration of CDK activity, the CDK-dependent recruitment of Mcm10 and GINS (Psf2) was not blocked under these conditions (, lanes 4 and 13). Thus, DNA unwinding is not required for Mcm10 or GINS origin association. Even though Mcm10 is present at the origin and is required for the loading of Pol α and Pol δ (), these polymerases were not recruited in the absence of DNA unwinding. In contrast, in the same conditions Pol ε was present at the origin DNA at similar levels as GINS. Both proteins show reduced recruitment in the presence of ATPγS relative to ATP, most likely due to incomplete restoration of S-CDK activity. Thus, our findings support a model in which DNA unwinding is required for Pol α and δ recruitment but that Mcm10 and Pol ε are recruited independently of this event.