|Home | About | Journals | Submit | Contact Us | Français|
The eukaryotic origin recognition complex (ORC) interacts with and remodels origins of DNA replication prior to initiation in S phase. Here we report single particle cryo-EM-derived structure of the supra-molecular assembly comprising of S. cerevisiae ORC, the replication initiation factor Cdc6 and double strand ARS1 origin DNA in the presence of ATPγS. The six subunits of ORC are arranged as Orc1:Orc4:Orc5:Orc2:Orc3 with Orc6 binding to Orc2. Cdc6 binding changes the conformation of ORC, particularly re-orientating the Orc1 N-terminal BAH-domain. Segmentation of the 3D map of ORC•Cdc6 on DNA and docking with the crystal structure of the homologous archaeal Orc1/Cdc6 protein suggest an origin DNA binding model in which the DNA tracks along the interior surface of the crescent-like ORC. Thus ORC bends and wraps the DNA. This model is consistent with the observation that binding of a single Cdc6 extends the ORC footprint on origin DNA from both ends.
Before a cell can divide, its genome must be duplicated (Kornberg and Baker, 1992). DNA polymerases, the workhorses that synthesize the new strands of the double helix work only at an established replication fork in which the double helix DNA has already been unwound by a replicative helicase complex. However, the close-ring, hexameric structured helicase cannot load onto DNA by itself and needs to be actively assembled onto DNA by the concerted actions of initiator proteins, which comprise ORC, Cdc6 and Cdt1 in eukaryotes (Bell and Dutta, 2002; Bell and Stillman, 1992; Bochman and Schwacha, 2008; Evrin et al., 2009; Kawakami and Katayama, 2010; Remus and Diffley, 2009; Stillman, 2005).
MCM2-7 was found recently to be loaded onto double strand DNA as a head-to-head double hexamer in vitro, with the double stranded DNA likely running through the center of the barrel shaped MCM2-7 hexamers (Evrin et al., 2009; Remus et al., 2009). The double hexamer of MCM2-7 is loaded as an inactive complex, with the two MCM2-7 hexamers primed for activation to form bi-directional DNA replication forks (Botchan and Berger, 2010). It is known that ATP binding by ORC and Cdc6 and then hydrolysis, first by Cdc6 and then by ORC, are required for ordered assembly of pre-Replication Complexes (pre-RC) with multiple MCM2-7 hexamers loading per origin and then for subsequent initiation of DNA replication (Bowers et al., 2004; Klemm et al., 1997; Lee et al., 2000; Randell et al., 2006; Speck et al., 2005; Speck and Stillman, 2007). Unwinding of the origin DNA only occurs after activation at S phase, but the mechanism is not known. In contrast, the bacterial DnaA, which has limited similarity to some subunits of ORC and to Cdc6, is monomeric is solution, but 10 - 20 of them oligomerize at the origin DNA during replication initiation and form a right-handed super-helix when crystallized at high concentration and in the absence of DNA (Erzberger et al., 2006; Mott and Berger, 2007). It was proposed that the bacterial origin DNA might wrap around the super-helical proteinaceous DnaA core, forming a positively super-coiled structure on DnaA and consequently causing the nearby AT-rich region to unwind (Duderstadt et al., 2010; Erzberger et al., 2006). DnaC then helps to load the hexameric DnaB helicase onto the melted region of DNA (Kawakami and Katayama, 2010). The eukaryotic Orc1 subunit and the Cdc6 protein share considerable amino acid similarity with each other and are related in sequence to the Orc1/Cdc6 initiator protein in archaeal species, but archaea lack proteins related to Orc2, Orc3, Orc4, Orc5 and Orc6 (Duncker et al., 2009). The crystal structures of the archaeal Orc1/Cdc6 in complex with the origin DNA reveals a bi-partite origin recognition mechanism (Dueber et al., 2007; Gaudier et al., 2007). Archaeal Orc1 binds origin DNA with both a C-terminal winged helix domain (WHD) and a helix-loop-helix insertion in the N-terminal ATPase domain. The C-shaped archaeal Orc1 acts like two claws of a lobster to grip, bending, and deforming the origin DNA (Dueber et al., 2007; Gaudier et al., 2007; Remus and Diffley, 2009). The helix-loop-helix insertion in the ATPase domain distinguishes the initiator clade from other AAA+ proteins, and is termed initiator-specific motif (ISM) (Iyer et al., 2004). Therefore, it is possible that some of the eukaryotic AAA+ type ORC subunits also bind the origin DNA with two claws.
Structural analyses of eukaryotic replication initiators lag behind the prokaryotic systems, likely due to the fact that the eukaryotic initiators form pre-existing large multi-protein complexes and that these complexes are flexible, rendering crystallization of the intact complexes difficult. Single particle EM has been applied to reveal the overall architecture of ORC from Saccharomyces cerevisiae (Speck et al., 2005) and Drosophila melanogaster (Clarey et al., 2006). The S. cerevisiae ORC (ScORC) is roughly a two-lobed and crescent-like structure. The approximate positions of the individual yeast ORC subunits were localized by systematically fusing the subunits at either the amino or carboxyl terminus, one subunit and one terminus at a time, with a bacterial maltose binding protein (Chen et al., 2008). The DmORC has a similarly elongated shape (Clarey et al., 2006). Interestingly, it was found that hyper-phosphorylation, a major regulatory event for DmORC function, did not markedly change its overall structure (Clarey et al., 2008).
The yeast replication initiator Cdc6, upon binding to ORC that is bound to origin DNA, was shown to turn on an ATP-hydrolysis-dependent molecular switch in ORC (Speck et al., 2005), although the physical basis for the switch is unknown. Furthermore, it is not known how the ORC interacts with and remodels origin DNA in preparation for loading of the replicative helicase component MCM2-7 to form a pre-RC. Existing structural analyses on eukaryotic ORC were done in the absence of DNA and by negative stain EM that has limited resolution (Chen et al., 2008; Clarey et al., 2008; Clarey et al., 2006; Speck et al., 2005). Cryo-EM methods avoid stain-associated artifacts and have the potential to achieve better resolution. We therefore launched a significant cryo-EM effort to overcome difficulties associated with imaging the relatively small ORC particles. We have now determined the first cryo-EM structures of yeast S. cerevisiae ORC and ORC-Cdc6 in the presence of origin DNA and ATPγS. Comparison of these structures reveals a series of conformational changes in ORC upon DNA and Cdc6 binding and suggests a model of how ORC binds origin DNA. This work represents an important step towards understanding the biochemistry of how the pre-RC is assembled.
ScORC is a bi-lobed structure (Speck et al., 2005). In the previous study, we were able to assign Orc1:Orc4:Orc5 to the top lobe of the ScORC and determine that the lower lobe contained Orc2, Orc3, and Orc6 (Chen et al., 2008). However, we were unable to distinguish between Orc2 and Orc3, thus it was unclear if Orc6 bound to Orc2 or Orc3. Therefore we analyzed binary interactions between GST-Orc6 and each of the other ORC subunits that were translated in vitro. In human ORC, Orc3 associates with Orc6 in vitro (Siddiqui and Stillman, 2007), but contrary to our expectations, only Orc2 of the S. cerevisiae Orc1/2/3/4/5 bound to GST-Orc6 (Figure 1A).
A domain mapping experiment showed that the N-terminal region of Orc2 (amino acids 1 - 265) retained affinity for GST-Orc6, but C-terminus did not (Figure 1B). Further deletion of only 35 - 40 residues from either terminus of the 1-265 Orc2 fragment reduced the interaction with GSTOrc6 and further deletions from either terminus eliminated the interaction (Figure 1C, 81-265 or 1-170), suggesting that a relatively large region(s) within the Orc2 N-terminal third (amino acids 1 - 265) participates in Orc6-binding. This new information enables us to more accurately interpret the difference observed in the negative stain EM images of ORC (Orc1-6) and Orc1-5 sub-complex (missing Orc6) (Chen et al., 2008). In projection, the lower lobe of ORC is composed of three density peaks, labeled with ε, ζ, and η (Figure 1D). In the Orc1-5 sub-complex, all of the three densities are present, indicating that none of these densities can be exclusively attributed to Orc6. However, the densities ε and η are much weaker in Orc1-5. Given the new information that Orc6 binds to Orc2, and the knowledge that Orc1-5 binds origin DNA in a sequence-specific and ATP-dependent manner that is indistinguishable from that of the intact ORC (Lee and Bell, 1997), we conclude that the density ε and η are largely from Orc2, and Orc6 binds on top of Orc2. This assignment of Orc6 is consistent with the 3D difference mapping between ORC and Orc1-5 showing two different peaks, the larger one near the bottom of the structure, and the smaller one in the middle region (Chen et al., 2008). Consequently, density ζ at the bottom left can be assigned to Orc3, unaffected by the presence or absence of Orc6. In conclusion, we suggest the assignment of the overall architecture of ScORC as Orc1:Orc4:Orc5:Orc2:Orc3 with Orc6 binding to Orc2, on the top of Orc2 in this view (Figure 1E).
In order to visualize how ORC interacts with the origin DNA and the replication initiator Cdc6, cryo-EM single particle three-dimensional reconstruction was determined on three complexes: purified S. cerevisiae ORC alone; ORC bound to ARS1 origin-containing 66-bp long double stranded DNA (dsDNA); and ORC bound to ARS1 origin DNA and S. cerevisiae Cdc6 (see Experimental Procedures) (Speck et al., 2005). All three samples were maintained in buffer containing 1 mM ATPγS, a slowly hydrolyzable ATP analogue. Although yeast ORC at 414 kDa is small for cryo-EM, by using a secondary layer of continuous thin carbon film over the primary lacey carbon film, and by strictly selecting EM-grid regions with the thinnest vitreous ice, we were able to obtain electron micrographs with adequate particle contrast (Figure 2A). We found that a small percentage of ORC particles (< 10%) formed dimer-like aggregates in solution and on EM grids. It was therefore necessary to manually select particles from micrographs to avoid the occasional putative ORC dimers. Fig 2B shows five class averages and their corresponding 2D projections of ORC-DNA (See also Figure S1). The 3D EM map of ORC alone is relatively featureless (Figure 2C) while the larger ORC assemblies, ORC-Cdc6-DNA and ORC-DNA, containing more structural detail (Figures 2D and and3).3). This probably indicates that ORC is flexible on its own and interaction with its functional partners, DNA and Cdc6, stabilizes the ORC structure. Fourier shell correlations (FSC) of the 3D reconstructions indicate that the data self-consistency ranges from ~ 1 nm to 2.5 nm (Figure S2). But FSC is not an accurate measurement of resolution (Grigorieff, 2000). Based on structural features, the EM map of ORC alone may have a resolution of 2.5 nm, and that of ORC-DNA and ORC-Cdc6-DNA of 1.5-nm resolution. The cryo-EM map of the largest assembly, ORC-DNA-Cdc6, represents a significant improvement over the previously published negative stain EM map (Chen et al., 2008): the previous map reveals only the shape of the ORC complex, but contains no structural features for the components, whereas in the new cryo-EM map, several of the ORC subunits are resolved. See below for more description.
The cryo-EM 3D map of ORC is similar in overall architecture to the elongated, crescent-like, two-lobed shape of the negative stain EM structure (Speck et al., 2005) (Figure 2C). The approximate locations of the ORC subunits are marked, based on our previous negative stain EM mapping results in which the maltose-binding protein was fused to individual subunits (Chen et al., 2008). We group the upper and middle regions into the top lobe consisting of Orc1, Orc4, and Orc5. The bottom region forms the second lobe and is composed of Orc2, Orc3, and Orc6. The cryo-EM structure of ORC is slightly twisted out of plane as compared with the flat shape in the negative stain EM map. The flattening was likely caused by air-drying and by the preferred orientation of ORC in stain salt on carbon film (Speck et al., 2005). In the EM maps (Figures 2C and 2D, front views), a 3nm-sized density region at the left side in the middle region is tentatively assigned to the N-terminal BAH domain of Orc1 (1-NTD), based on the proximity to the Orc1 main density and the fit to the crystal structure of this domain (see below).
The structure of ORC bound to a 66-bp double-stranded ARS1 origin DNA was determined to be better resolution than ORC alone as evidenced by increased structural details (Figure 2D). When the lower lobes of the ORC and ORC-DNA structures are aligned, the upper lobe of ORC-DNA appears to be rotated as a rigid body by ~ 20° relative to ORC alone (Figures 2C and 2D). Importantly, the upper lobe is composed of Orc1, Orc4, and Orc5, thus containing all the potential ATP binding and hydrolysis sites (Bowers et al., 2004; Klemm et al., 1997; Speck et al., 2005). The large conformational change in the ATPase-containing top lobe of ORC may explain the long-standing observation that ORC binding to origin DNA is an ATP-dependent event (Bell and Stillman, 1992). Due to the limited resolution (~ 1.5 nm), the 66-bp DNA could not be demarcated in ORC-DNA structure (Figure 2D).
The replication initiator Cdc6 can bind ORC in the presence of both origin DNA and ATP. In the absence of origin DNA, ATP hydrolysis disrupted the ORC-Cdc6 complex (Speck et al., 2005). Thus a stable ORC-DNA-Cdc6 complex was assembled in the presence of the slowly hydrolyzing ATPγS. Our previous negative stain EM analysis showed that Cdc6 bound to the side of the crescent-like ORC structure in the absence of origin DNA (Speck et al., 2005). In the current studies, we first asked if Cdc6 binds to the same location in the presence of origin DNA. Comparison of the reference-free 2D class averages of stained ORC-DNA particles with that of ORC-DNA-Cdc6 clearly shows that Cdc6 still binds to the same position, transforming the crescent-shaped structure into a ring-like structure (Figures 3A and 3B, thick white arrows). However, we found an important difference in the presence of origin DNA: the orientation of the density assigned to Orc1 appears to be changed when Cdc6 binds to ORC-DNA (Figures 3A and 3B, thin white arrows).
To visualize the nature of the Orc1 rearrangement, we determined the cryo-EM structure of ORC-DNA-Cdc6 to ~ 1.5 nm resolution (Figure 3). A side-by-side comparison of the ORCDNA structure (Figure 3C, the same as Figure 2D but with additional side and back views) with that of ORC-DNA-Cdc6 (Figure 3D) clearly reveals the density of the bound Cdc6 (highlighted in purple), with the remaining density belonging to ORC and DNA. The positions of the individual ORC subunits are again labeled. Orc1 is composed of a highly curved C-shaped main body (highlighted in light blue) and a separate NTD (highlighted in cyan). As already hinted in the 2D averages, upon Cdc6 binding the Orc1 main body rotated, as indicated by a pair of blue arrows (Figure 3D, best viewed in the front and back images). Furthermore, the Orc1-NTD flipped towards the back of the ORC crescent, as indicated by a cyan arrow in the top and back views in Figure 3D. The flip of Orc1-NTD appears to make room for the incoming Cdc6, which binds ORC from the same side. The rotation of the Orc1 main body appears to make better contact with the incoming Cdc6 (see the front and the back panels in Figure 3C and 3D; also see below for docking of the archaeal Orc1/Cdc6 crystal structure). Orc1 reorientation also rearranges Orc4 and pushes Orc4 slightly away from Orc1 (Figure 3D, black arrow).
Strikingly, upon Cdc6 binding a small piece of density crops out towards the front of the ORC-DNA complex, protruding from the lower lobe where Orc2, Orc3 and Orc6 reside. The corresponding densities in ORC-DNA and in ORC-DNA-Cdc6 are highlighted in red (Figure 3C, 3D; Movie S1). The identity of this density is currently unknown. However, because Orc6 has been determined to bind on top of Orc2 and is located approximately in this region, we propose that the protrusion may be a part of Orc6. If true, the central location of Orc6 may enable Orc6 to reach to Orc1, which is located on the opposite end from Orc2 in our model. The experimental paradox with this idea is that the full-length Orc1 has no significant affinity for GST-Orc6 (Figure 1A). One interpretation is that the flexible Orc1 N-terminal domain, which can be rotated drastically upon Cdc6 binding to the ORC•DNA complex (Figures 3C and 3D), may interfere with the binary interaction between Orc1 and Orc6. To test this possibility, we subjected a series of deletion constructs of Orc1 to GST-Orc6 pulldown assays. Deletion of the first 100 or 200 residues (101 - 914 and 201 - 914) had no effect on GST-Orc6 binding (Figure 4A). In contrast, Orc1 lacking the first 300 residues (301 - 914) showed interaction with GST-Orc6 (Figure 4A) that was Orc6-dependent (Figure 4B). Deletion of a further 100 residues (401 - 914) abolished this binding (Figure 4A), implying that the residues 301 - 400 are important for Orc6-binding. Indeed, a short peptide of Orc1 (301 - 400) retained affinity for Orc6. Fusion of the N-terminal 300 residues to the short peptide (1 - 400), however, eliminated this binding activity (Figure 4B). These results suggest that there is binary interaction between Orc6 and Orc1301 - 400, which could be repressed by the neighboring NTD of Orc1. Taken together, Orc6 has affinity for Orc1, in addition to Orc2 (Figure 1), both of which are located at opposite ends of ORC in our model. This conclusion lends support to our tentative assignment of Orc6 to the central protrusion in ORC-DNA-Cdc6 structure.
Therefore, our cryo-EM study shows that the most prominent changes in ORC-DNA structure upon Cdc6 binding are confined to the rearrangement of Orc1 and probably Orc6. This observation is in good agreement with a previous work reporting that Cdc6 binding to the ORCDNA complex increased the sensitivity of Orc1 and Orc6 to trypsin digestion, whereas the Orc3, Orc4 and Orc5 subunits were relatively resistant (Mizushima et al., 2000).
Proteins involved in the assembly of the pre-RC are mostly AAA+ proteins, many of whose archaeal homolog structures have been determined by crystallography (Mott and Berger, 2007). Of particular relevance to the eukaryotic initiators are the archaeal Orc1/Cdc6 initiator protein that is analogous to the eukaryotic Orc1 subunit of ORC and the Cdc6 protein, which are related in amino acid sequence to each other (Dueber et al., 2007; Gaudier et al., 2007; Liang et al., 1995; Liu et al., 2000; Singleton et al., 2004). The structure of the Pyrobaculum aerophilum Orc1/Cdc6 in the absence of DNA is in a linear and extended conformation and was proposed to bind DNA primarily via the C-terminal winged helix domain (WHD) (Liu et al., 2000) (Figure 5A). However, the structures of the Aeropyrum pernix and the Sulfolobus solfataricus Orc1/Cdc6-like initiator proteins, either not bound or bound to DNA, formed a curved, C-shaped conformation with both the N-terminal ATPase domain and the C-terminal WHD interacting with origin DNA when it is present (Figure 5B) (Dueber et al., 2007; Gaudier et al., 2007; Singleton et al., 2004). In order to dock the homologous crystal structures, we first segmented the 3D density map of ORC-DNA-Cdc6 by using a watershed based semi-automatic program Segger (Pintilie et al., 2010) (Figure 5C). The electron densities of Orc1, Orc4, Orc5, and Cdc6 are well resolved and separated from their respective neighbors, and the program Segger was able to demarcate the boundaries of these four proteins automatically without ambiguity. The archaeal structure can be fitted as single rigid body into the segmented subunit density (Figure S3). However, the remaining three proteins (Orc2, Orc3, and Orc6) at the lower lobe are densely packed with little separation, consequently, Segger was unable to automatically demarcate their boundaries. Therefore, the boundaries depicting Orc2, Orc3, and Orc6 are speculative (Figure 5C). We were unable to trace the 66-bp DNA. This may be due to the possibilities that DNA density intermingles with protein densities in the EM map at the modest resolution; and bound DNA may be partially flexible thus is partially averaged out in 3D reconstruction.
From the above-described EM density segmentation, it appears that the large crescent-like ORC structure is actually formed by a chain of smaller C-shaped protein subunits, Orc1 through Orc5, except for the Orc6 at the front of the lower lobe (Figure 5C). In the absence of a crystal structure for any of the yeast replication initiators, we used the highly curved Orc1/Cdc6 structure (PDB 2qby, Figure 5B) to model Orc1-5 (Figures 5D and S3). We confirmed with the online protein homology recognition program Phyre (Kelley and Sternberg, 2009) that S. cerevisiae ORC subunits 1 - 5 are indeed orthologs of the archaeal Orc1 structures (Clarey et al., 2006; Dueber et al., 2007; Gaudier et al., 2007; Speck et al., 2005). The fitting is good at the top lobe where Orc1, Orc4, and Orc5 are located, but less well in the lower lobe because of our inability to clearly separate Orc2 and Orc3 densities, and probably because of the fact that Orc2 and Orc3 contain extra domains unrelated to the AAA+ and WHD domains (Figure 5D). The yeast Cdc6, located at the side of ORC, takes on a nearly linear configuration and bridges the gap between the two ends of ORC (Figure 5C). Hence the extended Orc1/Cdc6 structure (Figure 5A, PDB 1fnn) was used to model the yeast Cdc6 protein (Liu et al., 2000) (Figures 5D and S3). The archaeal protein fits well with the segmented Cdc6 density, with its large ATPase domain contacting Orc1 at the top, and its WHD pointing towards the bottom near Orc3 (Figure 5D).
In summary, five copies of the C-shaped archaeal Orc1 structure (PDB ID: 2qby) (Dueber et al., 2007) were docked as individual rigid bodies to the densities assigned to Orc1 -Orc5, with the ATPase domain facing toward the front, the WHD towards the back, and the middle helical domain at the ridge pointing outwards (Figure 5D). We refrained from adjusting the domains of the archaeal Orc1 structure to improve fitting to the EM map because the sequence identity between the archaeal Orc1/Cdc6 and the yeast Orc1-5 subunits are not high, ranging from 8% for Orc2 to 21% for Orc1. The crystal structure of yeast Orc1-NTD is known and fits with the assigned density (PDB id 2m4z, Figures 5D and S3, Movie S1) (Zhang et al., 2002). We note that the backside location of Orc1-NTD BAH domain away from the ORC main body is consistent with the knowledge that deletion of this domain doesn't affect assembly of ORC complex and that Orc1-NTD is not involved in DNA binding, but is involved in nucleosome binding and origin preference within chromatin (Bell et al., 1995; Chen et al., 2008; Muller et al., 2010; Onishi et al., 2007; Zhang et al., 2002).
With the afore-described homolog docking, the putative site of the arginine finger of Orc4 ATPase domain faces the putative nucleotide-binding site of Orc1. However, these two sites are too far apart to enable their cooperative ATP hydrolysis (Figure 5D, black double arrow in the front view). This situation is reminiscent of the crystal structure of DNA bound archaeal Orc1/Cdc6 heterodimer in which one subunit is too distant to activate the ATPase of the other (Dueber et al., 2007). Thus, our structure suggests that the Orc1 ATPase is inhibited in the ORCDNA-Cdc6 complex, and additional conformational change in Orc1 would be required to bring the Orc1 closer to Orc4, in order for Orc1 ATP hydrolysis to occur. We propose that the required conformational change of Orc1 may be triggered by ATP hydrolysis of Cdc6, which is known to occur before ORC ATPase activity (Randell et al., 2006). Therefore the ORC-DNACdc6 complex structure may define the order of ATP hydrolysis, Cdc6 being the first followed by Orc1.
From segmentation and docking of the archaeal homolog structures into the cryo-EM map, the two ends of all of the five C-shaped large ORC subunits, i.e., their WHD and ATPase domains, all point towards the inside of the crescent-like ORC structure, with the middle helical domains at the ridge of the large crescent facing outside (Figure 5D). Based on this model, the segmented subunit densities were pulled apart to better visualize their shapes and inter-connectivity (Figure 6A). Both WHD and ATPase domain in each subunit are predicted to bind DNA (Dueber et al., 2007; Gaudier et al., 2007). This structural insight, together with the biochemical data indicating that Orc1 through Orc5 all directly bind to origin DNA (Lee and Bell, 1997), suggests that the origin DNA tracks along the interior surface of ORC (Figure 6A). Because the inside surface is curved, the origin DNA would then be bent when bound to ORC. We have found that the ARS1 origin DNA is indeed negatively super-coiled when bound to ORC (Mitelheiser and Stillman, unpublished), similar to DNA wrapping induced by S. pombe and Drosophila ORC (Houchens et al., 2008; Remus et al., 2004). Interestingly, a 72-base pair DNA can be modeled along the proposed interior surface of ORC (Figure 6B and Movie S1). The length of the model DNA is comparable to the 80-bp DNaseI footprint of the origin DNA in the presence of ORC and Cdc6 (Bell and Stillman, 1992; Speck et al., 2005). Perhaps, the most satisfying feature of our origin DNA binding model is that the DNA enters and exits ORC at the same side of ORC where Cdc6 binds. This observation explains the puzzle that although only one Cdc6 binds to ORC, Cdc6 extends the DNase I footprint of ORC at both ends of the origin DNA (Figure 6A) (Speck et al., 2005). An earlier metal shadowing EM observation of the ORC bound to a long ARS1-containing dsDNA showed that some of the ORC-bound DNA molecules were indeed highly bent at the ORC binding site (Chastain et al., 2004).
In this report, we present cryo-EM analyses of the eukaryotic ORC and its complexes with origin DNA and the other replication initiator Cdc6. With improved resolution, several of the protein subunits take on distinctive shapes as compared with the featureless blobs seen in the previously reported negative stained EM structures (Clarey et al., 2006; Speck et al., 2005). The improved resolution allows docking of the single subunit archaeal ortholog Orc1/Cdc6 proteins and suggests an emerging view of origin DNA binding to the highly curved interior surface, but not to the exterior surface, of the eukaryotic six-subunit ORC.
We have identified the locations and approximate boundaries of most protein components in the cryo-EM 3D map of the ORC-DNA-Cdc6 assembly. Identification of these subunits is based mainly on the improved structural features, a previous subunit mapping study (Chen et al., 2008), and our new in vitro pulldown data. We also used biochemical knowledge to resolve ambiguities in distinguishing between Orc4 and Orc5 in the upper lobe of the ORC structure since density features and mapping pattern alone were not sufficient to determine whether Orc4 or Orc5 follows the largest Orc1 subunit in the top lobe. Based on the fact that Orc1 ATPase is activated by an arginine finger in Orc4 (Bowers et al., 2004; Speck et al., 2005), Orc4 was assigned to the density next to Orc1.
The assigned spatial arrangement of these subunits is in good agreement with DNA cross-linking data showing that Orc1 and Orc4 bind near the essential A element of ARS1 origin DNA and Orc2 and Orc3 bind closer to the B elements (Lee and Bell, 1997). Importantly, binding of Cdc6 on the side of the crescent adjacent to Orc1 at the top and Orc3 at the bottom of ORC could explain the peculiar DNase I footprint protection pattern that extends from both ends of the ORC binding site on ARS1 origin upon Cdc6 binding (Speck et al., 2005; Speck and Stillman, 2007). We note that in the current ORC-DNA-Cdc6 architecture, Orc1 and Orc3 are located at the two ends of the tightly packed structure, and Orc6 potentially at the front projecting away from the main structure.
The spatial arrangement of the S. cerevisiae ORC is largely consistent with what is known about human ORC (Dhar et al., 2001; Siddiqui and Stillman, 2007; Vashee et al., 2001), raising the possibility that all eukaryotic ORC shares a similar architecture. Indeed, the ScORC and DmORC structures have the same dimension, closely resemble each other, and share the essential features of the half-ring structure and mid-body location of Orc5 (Figure S4). However, the DNA binding mode we propose for ScORC differs from that proposed for Drosophila ORC (Clarey et al., 2008). This is not surprising, because the DNA binding mode of DmORC is known to be different from ScORC: Whereas ScORC recognizes specific origin DNA sequences, albeit with low binding specificity, DmORC binds DNA with little sequence specificity, and DmORC was proposed to recognize the negatively super-coiled DNA topology, rather than the DNA sequence (Remus et al., 2004; Remus and Diffley, 2009). We also note that the function of human Orc2 and Orc3 seems to be swapped compared to ScOrc2 and ScOrc3 in terms of Orc6 binding (Figure 1; Siddiqui and Stillman, 2007). As the molecular masses of HsOrc2 and HsOrc3 are also swapped, i.e. HsOrc2 is smaller than HsOrc3, certain functional switching between Orc2 and Orc3 might have occurred during evolution.
The super-helical arrangement of the bacterial DnaA oligomer as revealed by crystallography functions to unwind DNA and can with the help of DnaC load the bacterial DnaB helicase (Erzberger et al., 2006). Upon binding to DNA, the bacterial initiator DnaA locally unwinds the DNA helix to allow assembly of the helicase. In contrast, ORC loads multiple MCM2-7 double hexamers onto the origin and there is no evidence for duplex DNA unwinding in the pre-RC. The yeast ORC forms a slightly twisted and highly curved crescent-like structure that is partially flexible in solution. ORC undergoes a conformational change in the presence of ATP, particularly at the Orc1 subunit when it binds origin DNA (Figure 2), which in S. cerevisiae has very weak sequence specificity and even less specificity in other eukaryotes. Such a conformational change might increase the affinity of ORC to the origin DNA or allow efficient Cdc6 binding where ORC undergoes even further conformational changes. The Cdc6-induced ring structure sitting on top of the Orc2-Orc3 stem is the main feature of the ORC-DNA-Cdc6 super-assembly. The Orc1 N-terminal BAH domain is at the back of the ring. Orc1 BAH is known to interact with either the silencing factor Sir1 at the silent mating type loci (Bell et al., 1995; Hsu et al., 2005) or histones (Onishi et al., 2007). Our observation that the BAH domain localizes to the back of the ORC-Cdc6-DNA structure suggests that it might interact with histones in nucleosomes without interference with the origin DNA binding activity. ORC is known to interact with and organize nucleosome positioning at origins of DNA replication (Eaton et al., 2010).
The large conformational changes in ORC induced by Cdc6 binding is probably the physical basis for the molecular switch that transforms ORC from a recognizer of origin DNA sequences into the MCM2-7-loading machine, projecting what appears to be Orc6 to the front of the ORC-Cdc6 ring to engage Cdt1 that is bound to the MCM2-7 hexameric helicase (Chen et al., 2007). The hetero-hexameric yeast MCM2-7 is a near symmetric ring with a similar dimension as the ring formed in ORC-Cdc6 structure (Speck et al., 2005). At the front surface of ORC-Cdc6, the initial interaction between Orc6 and Cdt1 would bring the helicase in close contact with ORC-DNA-Cdc6 assembly (Chen et al., 2007), thus inducing subsequent interaction between Cdc6 and Mcm2 (Jang et al., 2001), a critical interaction that results in the eventual loading of the helicase onto the origin DNA (Evrin et al., 2009; Remus et al., 2009; Tsakraklides and Bell, 2010). The path of the DNA we predict has both ends exiting the ORC-Cdc6 structure adjacent to the Orc1 and Cdc6 ATPase subunits and on one side of the ring. This is consistent with the large bend observed in ORC-DNA complexes (Chastain et al., 2004).
Orc6 bound near the B elements in DNA crosslinking experiments (Lee and Bell, 1997) and it has two predicted repeat domains that are similar in structure to the C-terminal repeats of the transcription factor TFIIB (Chesnokov et al., 2003). The C-terminal domain of TFIIB binds promotor DNA in cooperation with the TATA binding protein TPB (Tsai and Sigler, 2000) and TFIIB was recently shown to regulate the closed to open promotor transition when bound to the RNA polymerase (Liu et al., 2010). Since the protrusion in the ORC-Cdc6-DNA structure is likely to consist of Orc6 (Figure 3C), we suggest that following the interaction between Cdt1 and Orc6, the ORC-Cdc6 complex undergoes additional conformational changes to load the MCM2-7 hexamer and bring the B-elements of the origin nearer to Orc6. In a recent study, it is suggested that the two domains of Orc6 can associate with two Cdt1 during the recruitment of Cdt1-MCM2-7 to ORC (Labib, 2011; Takara and Bell, 2011).
The improved cryo-EM structure of ORC-Cdc6 on origin DNA is a step forward in determining how this protein machine loads another AAA+ protein machine onto DNA to mark the location for initiation of DNA replication in S phase of the cell cycle. As discussed, the structure suggests an origin DNA binding model, and provides a framework for understanding pre-RCs assembly.
Chunyan Tang participated in the initial stage of cryo-EM work. We thank Sylvain Mitelheiser for suggestions on the in vitro translation of ORC subunits and Patty Wendel for technical assistance in ORC prep. This work was supported by National Institutes of Health Grants GM45436 (to B.S.) and GM74985 (to H.L.), and the United Kingdom Medical Research Council (to C.S.). H.K. was supported by Postdoctoral Fellowships for Research Abroad from the Japan Society for the Promotion of Science and the Uehara Memorial Foundation.
Pulldown assays were performed as previously described (Chen et al., 2008; Siddiqui and Stillman) except as follows. GST-Orc6 was cloned and overexpressed using the pET system (Novagen). The template DNAs for in vitro transcription/translation of ORC subunits were derived from pCITE-2a(+) (Novagen) carrying individual full-length ORC gene (Chen et al., 2008): the templates for deletion constructs were prepared using the QuikChange site-directed mutagenesis (Stratagene) or PCR. The TNT T7 Quick system for PCR DNA (Promega) was used for transcription/translation when the PCR-amplified DNAs were templates.
The ScORC and Cdc6 were expressed and purified as described (Speck et al., 2005). The ARS1-containing 66-bp double stranded DNA was prepared by PCR. We incubated ORC at 1.9 mg/ml concentration with DNA at 1:1.2 molar ratio in a buffer containing 0.5 mM ATPγS for 30 min. The mixture was diluted to 0.19 mg/ml of the ORC concentration with the buffer containing 50 mM Hepes/KOH pH7.5, 1 mM EDTA, 1 mM EGTA, 1 mM DTT, 100 mM KGlu. Purified Cdc6 was added to the diluted ORC-DNA mixture with the molar ratio of 1:1.3 for ORC:Cdc6, incubated on ice for additional 10 min. The solution was further diluted by 5× just before flash-freezing.
The lacey carbon grids (EM Science, Hatfield, PA) were first coated with a thin layer of continuous carbon film. The grids were glow-discharged for 30 s in a vacuum evaporator (Edwards). A Vitrobot plunge freezer (FEI, Hillsboro, OR) was used with temperature set to 11 °C and relative humidity set to 70%. Three μl diluted sample was pipetted on to the EM grids, waited for 30 s, blotted, and then plugged the grids in liquid ethane to obtain vitreous sample. Cryo-EM was carried in a JEM 2010F TEM (JEOL USA, Peabody, MA) equipped with a Gatan 626 cryo-specimen holder and a Gatan 4K×4K UltraScan CCD camera (Gatan, Pittsburg, PA). Electron micrographs were recorded in low-dose mode on Kodak SO-163 negative films at magnification of 60,000×. To produce adequate contrast for these small particles, we selected regions of EM grids with very thin ice, and used relatively large under-focus values of 3 - 6 μm. The negative films were developed in Kodak D-19 solution, and digitized with a Nikon Supercool Scanner 8000ED at a step size of 6.35 μm (4000 dpi), corresponding to a pixel size of 1.06 Å at the sample level.
All images from ORC, ORC-DNA, and ORC-DNA-Cdc6 samples were computationally processed in a similar manner. We wrote several python scripts invoking a series of EMAN commands to facilitate processing a large number of images (Ludtke et al., 1999). We first produced high contrast images for manual selection of particles with a script that processed all scanned raw images automatically. For each image, the image format is changed from TIF to MRC, and the image size is shrunk by a factor of 4. Then low-pass filtered (25 Å) images are calculated. To pick particles based on the low-pass filtered images, we used the semi-automatic mode in “boxer” and manually removed the “bad” particles. The boxer size is 128 × 128 pixel. For contrast transfer function (CTF) correction, we first made a structure factor file by “ctfit” using several particle sets at different defocuses, and then find the CTF parameters with “fitctf”, and flipped the phase of the images with “applyctf”. Particles from individual images were eventually combined into one file, contrast inverted and high-pass filtered.
Although the particles were carefully selected initially, many bad particles could still enter the dataset. We subsequently used reference-free classification on the low-pass filtered and shrunk particles to carry out a second round of particle selection: particles that did not produce good class averages were rejected at this stage. For each dataset containing more than 100,000 particles, we performed classification with class number set to 500. Raw particles producing well-defined class averages were pooled, and those did not were removed. The final number of particles used for projection-based 3D refinement was 40,000 for ORC, 36,000 for ORC-DNA, and 54,000 for ORC-DNA-Cdc6. Thus slightly over half of the initially selected particles were rejected. We used the published 20-Å resolution 3D map of ORC (EMDB ID 1156) as the starting model. All three cryo-EM datasets used this same starting model. This model was determined previously by random conical tilting method from negatively stained EM images (Speck et al., 2005). We first refined the model with dataset at reduced sample level (4.24 Å/pixel) on a Linux workstation with four dual-core processors. The resulting 3D map was further refined with finer sampled dataset (2.12 Å/pixel). The “amask” option of the EMAN's “refine” command was used to make mask from the envelope of the model and the mask was subsequently applied to the new volume after each refinement cycle. The Fourier shell correlations of the EM maps were calculated by “eotest” in EMAN. 3D EM map visualization and docking of the crystal structures were carried out in UCSF Chimera (Pettersen et al., 2004). Density segmentation used the semi-automated program Segger (Pintilie et al., 2010), a program now incorporated in the UCSF Chimera. To maintain objectivity and reproducibility, only the program default values were used during segmentation. Sub-domains were grouped into putative subunits based on the molecular shape and the subunit localization information.
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.
The cryo-EM 3D density map of ScORC-Cdc6-DNA has been deposited in the Electron Microscopy Data Bank (EMDB) with access code EMD-5381.