|Home | About | Journals | Submit | Contact Us | Français|
The licensing of eukaryotic DNA replication origins, which ensures once per cell cycle replication, involves the loading of six related minichromosome maintenance proteins (Mcm2-7) into prereplicative complexes (pre-RCs). Mcm2-7 forms the core of the replicative DNA helicase, which is inactive in the pre-RC. The ATP-dependent Mcm2-7 loading reaction requires the Origin Recognition Complex (ORC), Cdc6 and Cdt1. We have reconstituted Mcm2-7 loading with purified budding yeast proteins. Using biochemical approaches and electron microscopy, we show that single heptamers of Cdt1·Mcm2-7 are loaded cooperatively into stable, head-to-head Mcm2-7 double hexamers connected via N-terminal rings. DNA runs through a central channel in the double hexamer, and, once loaded, Mcm2-7 can slide passively along double-stranded DNA. Our work has significant implications for understanding how eukaryotic DNA replication origins are chosen and licensed, how replisomes assemble during initiation and how unwinding occurs during DNA replication.
In eukaryotic cells, DNA replication initiates from multiple replication origins distributed along multiple chromosomes. This allows cells to replicate large genomes in relatively short periods of time. However, origin usage must be carefully coordinated to ensure the genome is completely replicated in each cell cycle, but no region of the genome is replicated more than once. A two-step mechanism underpins once per cell cycle replication in eukaryotes (Bell and Dutta, 2002; Blow and Dutta, 2005; Diffley, 2004). In the first step, known as licensing, the six subunit Origin Recognition Complex (ORC) together with Cdc6 and Cdt1 load Mcm2-7 onto DNA in a reaction requiring ATP hydrolysis. Mcm2-7, comprising six related polypeptides, is believed to be the engine of the replicative helicase but is inactive in this prereplicative complex (pre-RC). The loading of Mcm2-7 can only occur during G1 phase when cyclin dependent kinase (CDK) activity is low and the anaphase promoting complex/cyclosome (APC/C) is active. In budding yeast, CDKs prevent Mcm2-7 loading by directly phosphorylating and inhibiting each pre-RC component. Pre-RCs are activated in S phase by the combined action of two protein kinases, CDK and the Dbf4-dependent protein kinase (DDK) comprising a heterodimer of Dbf4 and Cdc7. CDKs work by phosphorylating Sld2 and Sld3, which generates binding sites for tandem pairs of BRCT repeats in Dpb11 (Tanaka et al., 2007; Zegerman and Diffley, 2007), whilst DDK phosphorylates Mcm2-7 directly (Sheu and Stillman, 2006). These events together lead to the loading of Cdc45 and the GINS complex into a pre-initiation complex (pre-IC), which is required for the activation of the Mcm2-7 helicase.
Although we know the order in which initiation factors are recruited to origins, relatively little is known about the biochemical mechanism of pre-RC and pre-IC assembly and how this leads to origin unwinding and replisome assembly. To understand these mechanisms in detail, it will be necessary to reconstitute these reactions with purified proteins.
We set out to purify pre-RC components for reconstitution experiments. Where possible, proteins were expressed and purified from budding yeast cells arrested in G1 phase, a period of competence for pre-RC assembly in vivo (Piatti et al., 1996) and in vitro (Seki and Diffley, 2000). ORC was purified from G1 phase cell extracts from a yeast strain that over-expressed all six ORC subunits (Bowers et al., 2004) (Figure 1A and Supplementary Figure 1A). Because Cdc6 protein is highly unstable during G1 phase in budding yeast (Drury et al., 2000), it was expressed in insect cells from a baculovirus vector and purified as an apparent monomer (Figure 1B and Supplementary Figure 1C). The purified Cdc6 migrated in SDS-PAGE as a phosphatase-sensitive doublet (Figure 1C lanes 1,2) indicating that it is phosphorylated. We, therefore, also expressed and purified a mutant Cdc6 lacking all eight CDK phosphorylation sites (Figure 1C lanes 3,4). This protein migrated as a fast-migrating single band in the presence and absence of lambda phosphatase.
To purify Mcm2-7, the Mcm4 subunit was tagged at its C-terminus with a 3XFLAG epitope and purified from the soluble (non-chromatin-bound) fraction of G1 phase yeast extracts by α-FLAG immuno-affinity chromatography followed by Superdex 200 gel-filtration chromatography. Approximately half of the Mcm4 eluted from the Superdex 200 column in a high molecular weight complex that co-fractionated with a series of other polypeptides (Figure 1D). Immunoblot (Figure 1E) and mass-spectrometric analysis (data not shown) identified the co-eluting polypeptides in the complex as Mcm2, Mcm3, Mcm5, Mcm6, Mcm7, and Cdt1. The Cdt1·Mcm2-7 complex eluted in the same fraction as thyroglobulin (670kDa) (Figure 1D), consistent with the predicted molecular weight of a Cdt1·Mcm2-7 heptamer (676kDa). The existence of a stoichiometric complex between Cdt1 and Mcm2-7 is consistent with earlier biochemical and genetic analysis (Kawasaki et al., 2006; Tanaka and Diffley, 2002).
Transmission electron microscopy of negatively-stained, purified Cdt1·Mcm2-7 revealed a relatively homogeneous distribution of globular particles, approximately 15 nm in diameter (Figure 1F). Reference-free classification of ~10,000 particles yielded views that we interpret broadly as top/bottom and side views (Figure 1F). The round top/bottom views feature a central hole or cavity. Presumptive side views feature two distinct, parallel layers of protein density (Figure 1F, white arrows in bottom average panel) separated by a gap and connected by thin protein bridges. These are similar to views of the homohexameric complex of the archaeal Mcm from Methanothermobacter thermautotrophicus which assembles from a single Mcm subunit (MthMcm) (Pape et al., 2003). Extra protein density located on the side of the particles and not found in the archaeal complex (arrowheads, Figure 1F) may be due to the non-Mcm Cdt1 subunit. We are currently analysing these images to determine the 3D structure of Cdt1·Mcm2-7. We conclude from EM images and gel filtration results that Cdt1 and Mcm2-7 form a hetero-heptameric complex. Significantly for the work described herein, double heptamers were never seen.
To begin to examine pre-RC assembly in vitro, we analysed the recruitment of these purified proteins to linear, origin-containing DNA fragments (1kb) coupled to paramagnetic beads (DNA beads). Figure 2A (lanes 1-3) shows that, after incubation with purified proteins in reactions containing ATP, these DNA beads (but not beads lacking DNA – data not shown) bind ORC, Cdc6, Cdt1 and Mcm subunits. Mcm2-7 loading in vivo and in extracts has previously been defined by the generation of Mcm2-7 complexes that remain bound to DNA even after high salt treatment (Bowers et al., 2004; Donovan et al., 1997). Figure 2A (lanes 4-6) shows that high salt (0.5M NaCl) extraction of DNA beads from reactions containing ATP quantitatively removed ORC but not the Mcm subunits. Mcm2-7 loading is distinguished from simple recruitment because loading requires ATP hydrolysis by ORC and Cdc6 (Bowers et al., 2004; Klemm and Bell, 2001; Perkins and Diffley, 1998; Randell et al., 2006; Seki and Diffley, 2000; Weinreich et al., 1999). When ATP-hydrolysis was prevented by incubation with the ATP analogue ATPγS (Figure 2A, lanes 7-12), ORC, Cdc6, Cdt1 and Mcm subunits were all bound to DNA beads after a low salt wash; however, the Mcm subunits along with ORC and Cdt1 were all quantitatively removed from DNA beads by high salt extraction. Figure 2B shows that all six Mcm2-7 subunits are loaded onto DNA in a salt-resistant manner in the presence of ATP, but in the presence of ATPγS, all six Mcm subunits are removed by high salt extraction, suggesting that Mcm2-7 subunits remain associated in one complex during recruitment and after loading onto DNA. From these observations we conclude that a salt-resistant, DNA-bound Mcm2-7 complex is generated upon co-incubation of DNA with purified ORC, Cdc6, and Cdt1·Mcm2-7 in a reaction that requires ATP hydrolysis. Together, these results show that ORC, Cdc6 and Cdt1 are sufficient to perform Mcm2-7 loading in vitro, consistent with previous work (Gillespie et al., 2001; Kawasaki et al., 2006).
After high salt wash of the ATP-containing reactions, Cdc6 and Cdt1, along with Mcm2-7, remained associated with the beads (Figure 2A and B), To determine whether these proteins are actually bound to DNA, we treated the salt-washed beads with EcoR1, which cleaves the DNA at a site near the biotin-streptavidin linkage. Figure 2C shows that Cdc6 and Cdt1 remained bound to beads after removal of >95% of the DNA with EcoR1. Mcm2-7, however, were quantitatively removed from beads with the DNA. Thus, retention of Cdc6 and Cdt1 on DNA beads is due to non-specific interactions with the beads after Mcm2-7 loading. By contrast, Mcm2-7 proteins appear to be loaded directly on DNA. Maximum levels of Mcm2-7 loaded onto DNA were reached between 25-40 minutes of incubation (Figure 2D), similar to reported results in extracts (Bowers et al., 2004).
To further examine Mcm2-7 loading, DNA bead-bound proteins were visualised by silver staining after SDS-PAGE. Figure 2E shows that very similar amounts of Mcm2-7 and ORC were loaded onto DNA during the course of the reaction (lanes 2-4). In separate experiments, up to 20% of the input Mcm2-7 was converted into salt-resistant DNA-bound complexes (data not shown). We conclude that the Mcm2-7 loading reaction with purified proteins is relatively efficient, unlike the loading of Mcm2-7 in cell extracts (Seki and Diffley, 2000).
Figure 3A shows that Mcm2-7 subunits were neither recruited nor loaded onto DNA in the absence of Cdc6 protein (lanes 1, 5, 9, 13) indicating that ORC alone cannot efficiently recruit Cdt1·Mcm2-7. Similarly, Mcm2-7 subunits were neither recruited nor loaded in the absence of ORC (Figure 3B, lanes 1, 5, 9, 13). In both Figure 3A and 3B, ATP-dependent Mcm2-7 loading was rescued by titrating the missing component back into these reactions. These experiments show that ORC and Cdc6 are both required to recruit Cdt1·Mcm2-7 to DNA.
As shown in Figure 1C, the Cdc6 protein used here is at least partly phosphorylated, presumably by endogenous insect cell CDKs. Figure 3C shows, however, loading efficiencies of Cdc6-wt were indistinguishable from those of Cdc6-8A across a range of concentrations. Therefore, CDK phosphorylation of Cdc6 does not directly inhibit Mcm2-7 loading in this system, consistent with previous work showing that CDK inhibition of Cdc6 function requires additional protein co-factors, including SCFCDC4 and Clb2-Cdc28 (Drury et al., 2000; Mimura et al., 2004).
The initiation of DNA replication in budding yeast, unlike metazoans, depends on specific DNA sequences. These origins serve as specific binding sites for ORC and are also the sites at which pre-RCs assemble. The A element of the well-characterised ARS1 contains the essential ARS consensus sequence (ACS), which, together with the non-essential B1 element, defines the primary ORC binding site (Bell and Stillman, 1992; Marahrens and Stillman, 1992; Rao and Stillman, 1995; Rowley et al., 1995). The B2 element contains a 9/11 match to the ACS and can bind ORC in the absence of the ACS (Bell and Stillman, 1992). We therefore compared ORC binding and Mcm2-7 loading on wild type ARS1 and a A-B2- double mutant, which removes all known ORC binding sites. Surprisingly, Figure 3D shows that, in the absence of competitor DNA, ORC is recruited equally well, even at sub-stoichiometric concentrations, to the wild type ARS1 and the A-B2- double mutant (Figure 3D, lanes 1,2,5,6). Figure 3D also shows that Mcm2-7 loading efficiency was indistinguishable between wild type and the A-B2- mutant ARS1 in the absence of competitor DNA (lanes 1-8). Addition of non-specific competitor DNA strongly suppressed the binding of ORC as well as Mcm2-7 to both wild type and A-B2- mutant DNA (data not shown), and, under these conditions, ORC specifically associated with the wild-type ARS1 DNA (Figure 3D, compare lanes 9-10 and 13-14). Mcm2-7 loading was also preferentially observed on wild-type origin DNA in these reactions (Figure 3D, lanes 9-16). These data suggest that purified ORC, Cdc6, and Cdt1·Mcm2-7 can mediate sequence-specific Mcm2-7 loading in vitro in the presence of competitor DNA, yet specific sequences are not mechanistically required for either ORC binding or Mcm2-7 loading.
As shown in Figure 2C (lane 4), the loaded Mcm2-7 complexes can be purified after high salt wash by removal of the DNA from beads with EcoR1. Upon examination by EM after negative staining, the majority of the Mcm2-7 particles present in this fraction were roughly rectangular and homogeneous in size (150 Å × 230 Å)(Figure 4A and Supplementary Figure 4A,B). Image analysis and classification of individual particles showed them to be composed of four parallel layers of protein density (Figure 4C, upper panel), with the two outer layers being thicker than the two inner layers. Protein density was reduced along the long axis through the center of the particles, suggesting that a channel runs through the particle. The height of this four layer Mcm2-7 complex is roughly twice that of the two-layer Cdt1·Mcm2-7 complex. Thus, the loaded complex appears to be a double hexamer of Mcm2-7.
To test whether the double hexamer was the result of enforced hydrophobic interactions induced by exposure to high-salt, we omitted the high-salt wash and analyzed the entire loading reaction in solution rather than on beads. Reactions were incubated with free 1 kb DNA fragments and spotted directly onto carbon-coated copper grids for electron-microscopic analysis. As shown in the overview image of Figure 4B, the same rectangular double-hexameric particles were readily seen against a background of smaller, more heterogeneous complexes that likely correspond to free Cdt1·Mcm2-7 complexes and ORC. Image analysis confirmed that these abundant, larger particles were structurally indistinguishable from the Mcm2-7 particles purified after high salt wash (Figure 4C, lower panel). Double hexamers were not detected in reactions that lacked either DNA, ORC, or Cdc6. They were also not detected in complete reactions that were performed in the presence of ATPγS (Supplementary Figure 2).
These double hexamers are similar to 2D class averages of electron-microscopic images of double-hexameric assemblies of the archaeal MthMcm (Costa et al., 2006). In the case of MthMcm, molecular modelling indicated that the thicker outer tiers of the double-hexamer contain the AAA+ domains, whereas the two inner tiers correspond to the double-hexameric ring of the N-terminal domain that mediates hexamer-hexamer association. The Mcm2-7 double hexamer appears to have a similar organisation. To obtain independent evidence of the orientation of the individual hexamers, we took advantage of the FLAG epitope tag on the Mcm4 C-terminus. We performed loading as in Figure 4A except reactions contained purified α-FLAG antibody. Figure 4D shows that the extra density resulting from the antibody-labelling of the Mcm4 C-terminus was specifically associated with the outer tiers of the structure and the sites of antibody attachment on either end of the double-hexamers appeared to be spatially related by rotational symmetry. Terminally bound IgG was also observed in cases where only one IgG was bound to an Mcm2-7 double hexamer or where one IgG was simultaneously bound to two Mcm2-7 double hexamers (Supplementary Figure 3). Thus, Mcm2-7 are loaded as head-to-head double hexamers with the C-terminal AAA+ domain located in the thicker tiers on the outside of the double hexamer.
We calculated a 3D reconstruction of the Mcm2-7 double hexamer to ~30 Å resolution from 2292 particles out of a 3700 particles dataset (Supplementary Figure 4). Surface representations of a series of angular rotations around the long axis of the particles and an end-on view are shown in Figure 5A-E. The overall barrel-like structure (150 Å × 230 Å) exhibits four ring-like tiers of protein density consistent with the density described for the 2D class averages above. 2D reprojections of the 3D model agree well with the 2D class averages, supporting the validity of the 3D map (Supplementary Figure 4B). The two hexamers are connected by multiple thin protein bridges at the inner and outer circumference of the N-terminal rings (Figure 5 and Supplementary Figure 4F). This differs slightly from the X-ray structure of the double hexameric MthMcm N-terminal ring, which is connected primarily by interactions near the inner circumference of the N-terminal rings (Fletcher et al., 2003). Several eukaryotic Mcms (2,4,6) have extended N-terminal domains not found in Archaea, which may contribute to these inter-hexamer interactions. Similar to the MthMcm hexamer, side openings are found between the N-terminal and AAA+ rings in each hexamer (Figure 5A-D). The end-on view (Figure 5E), the cut-away view (Figure 5F), and the density sections (Supplementary Figure 4C) show that a central channel runs through the entire length of the double hexamer. Both the cut-away view of the surface-rendered 3D model and serial density sections through the double-hexameric structure reveal that the channel bends at the interface between the two hexameric halves due to off-register stacking of the hexamers (Figure 5F and Supplementary Figure 4C).
The presence of the central channel suggests a path for DNA through the Mcm2-7 double hexamer, although DNA density was not directly recovered in the 3D reconstruction. We used direct mounting and tungsten rotary shadowing of purified, loaded Mcm2-7 complexes to visualize the DNA associated with Mcm2-7 double hexamers. Mcm2-7 particles bound to linear 1kb DNA fragments (Figure 6A) exhibited similar shape and dimension (166 ± 15 Å × 249 ± 13 Å [n = 75]) to the Mcm2-7 double hexamer described above. The small increase in overall dimensions of the double hexamer is likely due to deposition of tungsten. Very few DNA molecules contained more than one double hexamer. In specimens where we could unambiguously trace the contour of the DNA molecules, the DNA appeared to enter and exit from opposite ends of the Mcm2-7 double hexamer (Figure 6A). In addition, free DNA molecules exhibited mean DNA contour lengths that were not significantly different from those bound by Mcm2-7 double hexamers (318 nm ± 16 nm [n=30] versus 321 nm ± 31 nm [n=43], respectively). These observations are consistent with DNA passing through the central channel of Mcm2-7 double hexamers. We do not exclude the possibility that Mcm2-7 may also engage DNA by other mechanisms (Costa et al., 2008).
If DNA passed through this central channel, Mcm2-7 would be topologically linked to the DNA. To test this, we asked whether circularisation of the bound DNA prevented dissociation of Mcm2-7. Figure 6B shows that Mcm2-7 exhibited a half-life on DNA of only approximately 10 minutes on 1 kb linear DNA in high-salt (0.5M NaCl) buffer (Figure 6B). This could reflect the overall instability of the complex, or it could reflect the loss of Mcm2-7 from the ends of these short, linear molecules. To distinguish between these possibilities, we compared residence times of Mcm2-7 double hexamers on a 1 kb linear DNA template to a circularised version of the same 1 kb DNA fragment. Strikingly, Figure 6C shows that Mcm2-7 double hexamers exhibited a half-life on the circular DNA that greatly exceeded 60 minutes, compared to 10 minutes on linear DNA (Figure 6C, lanes 6-10, and graph below). Figure 6C also shows that the loss of Mcm2-7 from the linear DNA occurred at the same rate in the presence or absence of ATP. These data strongly suggest that Mcm2-7 double hexamers can slide off the ends of a linear DNA molecule in an ATP-independent fashion and are therefore topologically linked to the DNA. At a more physiological salt concentration (0.1M NaCl), the half-life of Mcm2-7 double hexamers on linear DNA increased to approximately 35 minutes (Figure 6D, lanes 1-5 and graph below). This longer half-life on linear DNA is presumably due to electrostatic interactions between DNA and Mcm2-7. However, this is still significantly shorter than the half-life of Mcm2-7 double hexamers on the circular DNA template, which was greater than 60 minutes (Figure 6D, lanes 6-10 and graph below). Again, the rate of Mcm2-7 loss was not affected by the presence of ATP. Taken together, these experiments show that Mcm2-7 double hexamers encircle and bind DNA in the central channel, and that once bound to DNA, the double hexamers are mobile.
Because this sliding activity might have led to an underestimate of double hexamer formation by EM, we reinvestigated the binding to circular DNA by EM using rotary shadowing. Most DNA molecules contained either zero or one double hexamer (examples of one double hexamer in Figure 6E, i, ii, iii). Although there were DNA molecules with two or three double hexamers (Figure 6E, iv, v), these accounted for fewer than 5% of the protein:DNA complexes arguing that loading of multiple double hexamers is unlikely to be processive. The appearance of relaxed circular DNA bound to Mcm2-7 is consistent with the idea that the double hexamers do not contain extensive single-stranded DNA.
Our results provide the first evidence that ORC and Cdc6 load the Mcm2-7 proteins from single Cdt1·Mcm2-7 heptamers into pre-RCs as head-to-head double hexamers. DNA, probably double stranded, passes through the long, central channel of this double hexamer. And, once loaded, the double hexamer is mobile, capable of passive one-dimensional diffusion or ‘sliding’ along DNA. These features of the pre-RC have implications for how origins are chosen and how replisomes assemble during initiation.
The loading of Mcm2-7 requires ORC, Cdc6 and hydrolysable ATP, consistent with requirements in vivo. The requirement for Cdt1 was not tested because it is an integral component of the Mcm2-7 complex. The interaction of Cdt1 with both Mcm2-7 and Orc6 (Chen et al., 2007), suggests that it may act as a bridge between ORC and Mcm2-7. However, our results demonstrate that Cdc6 is also essential to recruit Mcm2-7 to origins, indicating that additional interactions are involved in this recruitment.
Surprisingly, loading of Mcm2-7 in vitro does not require specific ORC binding sites. Our results may contribute to resolving the long-standing issue of how orthologues of ORC can act on specific DNA sequences in yeast, but show little or no sequence preference in metazoans (Gilbert, 2004). Our results indicate that even yeast ORC has no inherent mechanistic requirement for specific DNA sequences in the loading of Mcm2-7. The sequence specific DNA binding of the budding yeast ORC may be an evolutionary adaptation designed to ensure sufficient origin activity in a genome containing very little intergenic DNA (Brewer, 1994). Sequence specificity appears to be an integral part of the S.cerevisiae core ORC whilst sequence specificity of Schizosaccharomyces pombe ORC is conferred by an extended AT hook domain on the Orc4 subunit (Chuang et al., 2002; Kong and DePamphilis, 2001; Lee et al., 2001). Recruitment of ORC in metazoans may also involve interactions with additional sequence specific DNA binding proteins like TRF2 (Atanasiu et al., 2006; Deng et al., 2007; Tatsumi et al., 2008). Consistent with this idea, recruitment of ORC to a GAL4 DNA binding site array via fusion of ORC subunits or Cdc6 to the GAL4 DNA binding domain is sufficient to create a functional replication origin in human cells (Takeda et al., 2005).
In E.coli, the ATP-bound DnaA initator binds to 9-mer sequences in oriC causing localised DNA unwinding in adjacent 13-mer sequences (Bramhill and Kornberg, 1988). The two DnaB helicase hexamers are then recruited to load around the unwound 13-mer strands by slightly different mechanisms (Mott et al., 2008). Unlike DnaA binding, ORC binding does not induce any detectable DNA unwinding. Thus, the eukaryotic initiator does not generate single-stranded DNA for the Mcm2-7 helicase to encircle. Moreover, previous analysis of yeast origins in vivo failed to detect any KMnO4-reactive DNA in G1-arrested cells (Geraghty et al., 2000) and data not shown) indicating that initial origin melting does not occur when Mcm2-7 is loaded onto DNA but, instead, occurs when the helicase is activated during S phase. Consistent with this, the ability of loaded Mcm2-7 to slide freely on double-stranded DNA in an ATP-independent manner (Figure 6B-D) indicates that Mcm2-7 loaded in vitro encircles double-stranded, not single-stranded DNA. Delaying the generation of single stranded DNA within the Mcm2-7 double hexamer until S phase may protect origin DNA from damage during G1 phase.
The loading of the double hexamer appears to be highly cooperative. Double Cdt1·Mcm2-7 heptamers were never seen prior to loading and single Mcm2-7 hexamers were never detected on DNA after loading. These results indicate that the two hexamers are loaded together in a concerted reaction. The stoichiometry of ORC and Cdc6 in this loading reaction is presently unknown. Although yeast origins generally contain a single high affinity ORC binding site, a second potential ORC binding site resides within the B2 element of ARS1 (Wilmes and Bell, 2002). Additionally, archaeal replication origins, which may initiate replication by mechanisms similar to eukaryotes, generally have ORC binding sites arranged symmetrically around an A/T rich region (Wigley, 2009). The lack of stringent sequence requirement suggests that a second ORC binding site need not be a specific sequence element. We note that this concerted loading reaction requires the coordinated breaking and reforming of four separate rings, two in each hexamer. Further work will be required to elucidate the mechanism of this complex loading reaction and to determine the functionality of the loaded Mcm2-7 double hexamer.
The binding of Mcm2-7 around double stranded DNA has implications for how DNA unwinding is ultimately catalysed by the Cdc45/Mcm2-7/GINS (CMG) complex (Moyer et al., 2006). Mcm2-7 may act in unwinding analogously to the eukaryotic viral SF3 initiator/helicases including the SV40 large T antigen (TAg) and the papillomavirus E1 protein. The TAg double hexamer can bind to double stranded DNA, and this binding can induce the generation of a short (8bp) stretch of melted DNA specifically within one of the two hexamers (Borowiec and Hurwitz, 1988). Although TAg and E1 can assemble as double hexamers around double stranded DNA (Dean et al., 1992; Fouts et al., 1999; Liu et al., 1998), current models indicate that they act during unwinding as classical helicases by encircling single stranded DNA (Enemark and Joshua-Tor, 2008; Schuck and Stenlund, 2005). If Mcm2-7 act analogously to these proteins, CDK- and DDK-dependent events must promote remodelling of the Mcm2-7 complex to encircle single stranded DNA during origin melting.
Alternatively, Mcm2-7 may act during replication as a double-strand DNA translocase (Laskey and Madine, 2003). In this model, Cdc45 and/or GINS would play a direct, structural role in strand separation, perhaps acting as a ‘plough’ or ‘pin’ into which DNA is pumped by Mcm2-7 (Takahashi et al., 2005). This is analogous to the bacterial RuvAB Holliday junction branch migrating enzyme in which two RuvB hexamers pump double stranded DNA through a tetramer of RuvA, which coordinates the separation and re-annealing of strands (West, 2003). We currently favour this second model because it does not require topological reorganisation of Mcm2-7 subunits during initiation and because it provides a potential biochemical function for Cdc45 and/or GINS during replication. The helicase activity of archaeal Mcm (Chong et al., 2000; Kelman et al., 1999; Shechter et al., 2000) as well as eukaryotic Mcm complexes (Bochman and Schwacha, 2008; Ishimi, 1997; Moyer et al., 2006) on ssDNA substrates need not reflect their mode of action in vivo: even double stranded DNA translocases like RuvB can function in standard helicase assays (Tsaneva et al., 1993), presumably because they can translocate along one strand of DNA and displace annealed oligonucleotides.
By analogy to the mode of action of both SV40 large T antigen and RuvB, the orientation of the Mcm2-7 hexamers suggests that DNA is translocated from the AAA+ domains towards the N-terminal domains in each hexamer (Li et al., 2003; VanLoock et al., 2002; Wessel et al., 1992; West, 2003). Consistent with this, the archaeal MCM hexamer from Sulfolobus solfataricus (SsoMCM) can be loaded onto the 3’ single-stranded overhang of a forked substrate, with its AAA+ domain facing the fork (McGeoch et al., 2005).
The purified Replisome Progression Complex (RPC), a protein assemblage containing the CMG complex as well as other replisome components, contains just a single copy of Mcm4 (and presumably the other Mcm2-7 subunits) (Gambus et al., 2006). Thus, the functional unit of Mcm2-7 in elongation may be the single hexamer. We note that DDK, required for helicase activation, acts on N-terminal tails of Mcm2, 4 and 6 (Sheu and Stillman, 2006). Moreover, the mcm5-bob1 allele, which bypasses the requirement for DDK, is a mutation in the N-terminal domain of Mcm5 near the interface between hexamers (Fletcher et al., 2003). We speculate that DDK may activate the helicase by inducing separation of the two hexamers and the mcm5-bob1 allele may mimic this by destabilising interactions between hexamers.
The length of an Mcm2-7 double hexamer (230 Å) corresponds to approximately 68bp of B form DNA. This is similar to the size of the pre-RC footprint at several origins in vivo (Diffley et al., 1994; Santocanale and Diffley, 1996). However, based on data from Donovan et al. (1997), the equivalent of approximately 5 double hexamers are loaded per origin in yeast. The pre-RC footprint cannot account for this number of double hexamers. Indeed, the loading of multiple Mcm2-7 complexes per origin appears to be a common feature in eukaryotes (Hyrien et al., 2003). Once loaded, Mcm2-7 double hexamers may slide away from origins, allowing repeated loading. Although multiple Mcm2-7 complexes are loaded per origin, there is currently no evidence that this loading is processive and our results indicate that loading in vitro is distributive. Additional activities would presumably be required to allow the double hexamers to slide past nucleosomes bordering origins. Since ORC and Cdc6 can promote sequence-independent loading of Mcm2-7 (Figure 3D), it is also possible that some Mcm2-7 is loaded outside of origins, perhaps at random locations in vivo, which might not have been detectable in ChIP-chip experiments (Wyrick et al., 2001). These extra Mcm2-7 complexes may act as ‘spare’ origins to rescue stalled replication forks (Ge et al., 2007; Woodward et al., 2006).
The ability of the double hexamer to slide on double stranded DNA provides a mechanism by which supernumerary Mcm2-7 double hexamers may be pushed ahead of the replisome during normal elongation. These excess double hexamers could play a role in restoring an active helicase at stalled forks. This might provide an important mechanism for replication restart that does not require the reloading of soluble Mcm2-7 and, therefore, does not compromise mechanisms designed to ensure once per cell cycle replication.
Expression and purification protocols are included as supplementary data.
Loading reactions were performed in 40 μl of 25 mM Hepes-KOH pH 7.6/0.1M K-glutamate/0.02% NP-40/10 mM Mg(OAc)2/5% glycerol/1 mM DTT and 5 mM ATP or ATPγS. DNA beads were included at 1 pmol (25 nM) of DNA molecules. Reactions were mixed on ice and incubated at 30°C. Beads were washed once with 0.4 ml low-salt wash buffer (25 mM Hepes-KOH pH 7.6/0.3 M K-glutamate/0.02% NP-40/5 mM Mg(OAc)2/1 mM EDTA/1 mM EGTA/10% glycerol/1 mM DTT) and once with 0.4 ml high-salt wash buffer (as low-salt buffer, but 0.5 M NaCl instead of 0.3 M K-glutamate).
Mcm2-7-DNA complexes were purified from a standard loading reaction, followed by cleavage of the DNA in 40 μl of 25 mM Hepes-KOH pH 7.6/0.1 M K-glutamate/0.02% NP-40/5 mM Mg(OAc)2/5% glycerol with EcoR1 for 25 minutes at 30°C. IgG-labelled Mcm2-7 double hexamers were purified similar to unlabelled complexes, except that 1 μmg of a-FLAG M2 antibody (Sigma) was added to a loading reaction 10 minutes before the wash. Sample preparation and image processing were carried out as described in supplementary methods.
For rotary shadowing, Mcm2-7- complexes bound to linear DNA were eluted from beads with EcoR1 in 25 mM Hepes-NaOH pH 7.6/ 50 mM NaCl/5 mM Mg(OAc)2/5% glycerol. Mcm2-7 complexes bound to circularized DNA were purified from loading reactions in solution by gel filtration on Superose 6 in EcoR1 digestion buffer. The sample was prepared as described (Griffith and Christiansen, 1978) with modifications (supplementary methods).
This work was funded by Cancer Research UK (DR, JFXD, FB, EPM), ICR (FB and EPM) and NIH grants to JDG (GM31819 and ES13773). DR was supported by an EMBO long term fellowship. We are grateful to Dale Wigley for help and advice in the early stages of this project and to Dr Paula da Fonseca for advice on image processing. We thank Mark Skehel and colleagues for mass spectrometry and Karim Labib for antibodies. We are also grateful to members of the Diffley laboratory for helpful discussions.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.