Bidirectional eukaryotic DNA replication is established by quasi-symmetrical helicase loading

See allHide authors and affiliations

Science  21 Jul 2017:
Vol. 357, Issue 6348, pp. 314-318
DOI: 10.1126/science.aan0063

Getting loaded—make mine a double!

Chromosomal DNA replication initiates bidirectionally by loading two ring-shaped helicases onto DNA in opposite orientations. How this symmetry is achieved has been puzzling because replication initiation sites contain only one essential binding site for the initiator, the origin recognition complex (ORC). Coster and Diffley now show that both helicases are loaded by a similar mechanism. Efficient loading requires binding of two ORC complexes to two ORC binding sites in opposite orientations. Natural origins were found to be partially symmetrical, containing functionally relevant secondary ORC sites. Sites can be flexibly spaced, but introducing an intervening “roadblock” prevented loading, suggesting that individual helicases translocate toward each other on DNA to form a stable double ring.

Science, this issue p. 314


Bidirectional replication from eukaryotic DNA replication origins requires the loading of two ring-shaped minichromosome maintenance (MCM) helicases around DNA in opposite orientations. MCM loading is orchestrated by binding of the origin recognition complex (ORC) to DNA, but how ORC coordinates symmetrical MCM loading is unclear. We used natural budding yeast DNA replication origins and synthetic DNA sequences to show that efficient MCM loading requires binding of two ORC molecules to two ORC binding sites. The relative orientation of these sites, but not the distance between them, was found to be critical for MCM loading in vitro and origin function in vivo. We propose that quasi-symmetrical loading of individual MCM hexamers by ORC and directed MCM translocation into double hexamers acts as a unifying mechanism for the establishment of bidirectional replication in archaea and eukaryotes.

The motor of the eukaryotic replicative helicase—the heterohexameric minichromosome maintenance (MCM) complex—is loaded onto replication origins as an inactive, head-to-head double hexamer during the G1 phase of the cell cycle (13). During S phase, the double hexamer is converted into two active CMG (Cdc45-MCM-GINS) helicases (46) that nucleate assembly of the two bidirectional replisomes. MCM loading begins with the binding of the origin recognition complex (ORC) to DNA. ORC, together with Cdc6 and Cdt1, then recruits the first MCM hexamer (7, 8). Adenosine triphosphate (ATP) binding by ORC, Cdc6, and MCM is required for this recruitment, and ATP hydrolysis by MCM then drives double hexamer assembly (9, 10). But it is unclear how the second hexamer is recruited and loaded in the correct position and orientation (fig. S1A).

The recruitment of the first MCM hexamer to ORC/Cdc6 is mediated by the C terminus of Mcm3 (11). Studies using single-molecule approaches have suggested that the second MCM hexamer is directly recruited by interaction with the first hexamer, aided by a second Cdc6 but not a second ORC molecule (8, 12). If this were the case, point mutants in the C terminus of Mcm3 (e.g., Mcm3-13) that cannot be recruited as the first hexamer should still be able to load as a second hexamer in the presence of wild-type MCM. However, such mutants fail to load even when mixed with wild-type MCM, which suggests that both hexamers must be able to bind ORC/Cdc6 for loading (11). It remains possible that Mcm3-13 loading still occurs, but at a level below detection limits. We therefore developed a more sensitive and quantitative assay for MCM loading based on a fusion of Mcm3 to luciferase (Fig. 1A and fig. S2). The Mcm3-13 mutation had no effect on MCM complex stability (fig. S1B) or subunit stoichiometry (fig. S1C). However, despite the improved sensitivity, no loading of Mcm3-13 was observed in the absence or presence of wild-type MCM over a range of ratios (Fig. 1B and fig. S1, D and E).

Fig. 1 A single ORC binding site is insufficient for optimal MCM loading and origin function.

(A) MCM recruitment and loading are unaffected by a luciferase tag and are defective in Mcm3-13. (B) Untagged wild-type (WT) MCM does not rescue the loading defect of Mcm3-13. (C to E) A synthetic ORC binding site does not support plasmid replication in vivo (C), even in the presence of Abf1 binding sites (D) or poly(dA) stretches of increasing length (E), as assayed by the formation of yeast colonies after transformation. (F) ORC binding to the synthetic site is ATP-dependent and comparable to natural origins. (G) Quantification of (F). (H) The presence of chloride during loading (80 mM KCl) drives sequence-specific MCM loading and reveals that natural origins are more efficient than a single synthetic site.

Although this result suggested that both MCM hexamers must be able to bind ORC/Cdc6, budding yeast origins generally contain a single high-affinity ORC binding site located at one end of a nucleosome-free region (13). To test whether these features are sufficient for origin activity in vivo, we constructed synthetic sequences bearing a single T-rich ORC binding site in a GC-rich background with or without nucleosome-excluding sequences associated with origins—either binding sites for the transcription factor Abf1 or different lengths of polydeoxyadenosine [poly(dA)] sequences (13, 14). Yeast cells transformed with these constructs failed to sustain growth on selective media, indicating that they are not functional origins (Fig. 1, C to E). The synthetic ORC binding site supported ATP-dependent (Fig. 1F) and salt-stable (fig. S1F) ORC binding in vitro comparable to or better than several natural origins (Fig. 1, F and G). However, MCM loading on this synthetic sequence was poor relative to natural origins, even those such as ARS606 with weaker ORC binding (Fig. 1H).

Budding yeast origins also often contain multiple, degenerate ORC binding sites in the opposite orientation to the high-affinity site (15). We constructed additional templates that contained two ORC sites separated by 100 base pairs (bp) in all possible orientations (fig. S3A). Whereas ORC binding and MCM recruitment in adenosine 5′-O-(3-thiotriphosphate) (ATPγS) were unaffected by the relative orientation of the two sites (Fig. 2, A and B), MCM loading was greatly stimulated (by a factor of >7) when the two sites were in a head-to-head orientation (Fig. 2C). This synergy required an intact second site (fig. S3B) and resulted in loading efficiencies equivalent to natural origins (see below). Two sites in the head-to-head orientation also generated functional origins in vivo, with or without Abf1 sites (Fig. 2, D and E). Furthermore, changing only three bases within the poly(dA) stretch created a close match to the ORC binding sequence that supported ORC binding in vitro (fig. S3C) and converted all of the inactive poly(dA) templates (Fig. 1E) into active origins (Fig. 2F). Thus, when two ORC binding sites are present, they display orientation-dependent synergy that is necessary and sufficient for maximal MCM loading in vitro and origin function in vivo.

Fig. 2 Two ORC sites are necessary and sufficient for maximal MCM loading and origin activity.

(A to C) ORC binding (A), MCM recruitment (B), and MCM loading (C) onto synthetic substrates containing one or two ORC binding sites in all possible orientations. (D and E) In vivo origin activity of synthetic two-site substrates in the absence (D) or presence (E) of Abf1 binding sites. (F) Three base-pair changes that create a second ORC site are sufficient to convert the poly(dA) substrates from Fig. 1E into active origins.

Templates with single high-affinity ORC binding sites supported low levels of MCM loading in vitro (Fig. 2C), indicating either that a single ORC binding event is sufficient for MCM loading or that loading on single-site templates involves two ORC binding events: one at the high-affinity site and one nonspecifically in adjacent sequences. Several lines of evidence support the latter possibility: (i) ORC binding to a specific site (fig. S1F) and MCM loading onto two-site templates (Fig. 3A) were resistant to salt concentrations of up to 100 mM NaCl, whereas nonspecific binding of ORC (fig. S1F) and loading onto templates lacking ORC sites (Fig. 3A) were both sensitive to salt in this range. MCM loading on one-site templates exhibited intermediate salt sensitivity, which suggested that both specific and nonspecific ORC binding contribute to loading onto these templates. (ii) Consistent with this idea, MCM loading on the one-site template was almost completely eliminated by an oligonucleotide competitor containing an ORC binding site (Fig. 3B), whereas the two-site template was largely resistant to competition (factor of ~23 difference). (iii) MCM loading on the one-site template was also inhibited by excess template (fig. S4); this finding suggests that excess high-affinity binding sites sequester ORC, thereby preventing nonspecific binding needed for MCM loading.

Fig. 3 Natural origins use one high-affinity site and additional secondary sites.

(A) Salt sensitivity of one-site loading suggests the usage of a second nonspecific site. The indicated concentration of salt was present during loading. All reactions were subsequently washed with high salt (1 M NaCl). (B) One-site loading is more sensitive to competitor DNA than two-site loading. (C) MCM recruitment and loading as a function of ORC concentration. Data are means ± SEM. (D) Natural origins harbor secondary ORC binding sites. (E) MCM loading with synthetic versus natural origins, as well as deletion mutants (see fig. S6). (F) Mutations in secondary sites impair MCM loading in ARS600.1 and ARS1216.

If occupancy of two ORC sites in opposite orientations is required for loading, it is likely to be especially inefficient at low ORC concentrations. Indeed, mathematical modeling predicts a simple linear relationship between ORC concentration and MCM loading in a one-ORC mechanism, but predicts a more complex, sigmoidal relationship if two ORC molecules are required for loading (fig. S5). As expected, in the presence of ATPγS under nonspecific (no NaCl) conditions, MCM recruitment increased linearly with increasing ORC concentration and fitted well to a one-ORC model (Fig. 3C and fig. S5), consistent with the fact that there is a single MCM in the recruited complex. However, in the presence of ATP, MCM loading did not increase linearly with increasing ORC concentration but instead generated a sigmoidal curve that fit well to the two-ORC model but not the one-ORC model (Fig. 3C and fig. S5). Thus, efficient MCM loading requires two ORC molecules.

We next examined the role of secondary ORC binding sites in natural origins. We selected several well-studied origins, all of which contained multiple near-matches to the ORC binding site 3′ and on the opposite strand relative to the essential site (fig. S6). In each case, the essential site was the closest match to the consensus sequence (fig. S6) and was the site with the highest affinity, as robust ORC binding was lost upon deletion (fig. S7A). However, under milder wash conditions, ORC bound above background levels to the remaining sequences (Fig. 3D). This binding was stabilized by ATP (fig. S7B), a hallmark of sequence-specific ORC binding (16). Moreover, MCM was loaded on these sequences at levels greater than the no-site control and comparable to the synthetic one-site construct (Fig. 3E). ARS606, which had the weakest high-affinity site (Fig. 1, F and G), had the strongest secondary sites (fig. S7B) and loaded MCM better than other origins with stronger high-affinity sites (Fig. 3E). To test the importance of secondary ORC sites, we identified two origins, ARS600.1 and ARS1216, with small numbers of secondary ORC binding sites (fig. S7C). Mutation of these secondary sites greatly reduced MCM loading in vitro and impaired origin activity in vivo (Fig. 3F and fig. S7, D and E). Thus, natural origins use one high-affinity ORC binding site with one or more lower-affinity secondary sites in the opposite orientation.

To investigate why two ORC binding events promote MCM loading but the distance between high-affinity and secondary sites within natural origins varies, we tested the effects of distance between two specific sites. Neither ORC binding nor MCM recruitment was affected by altering the spacing between sites (fig. S8). Synergistic MCM loading occurred over a wide range of distances from 25 to 400 bp, with a peak at 70 bp (Fig. 4A). Moreover, origin activity in vivo was seen with all two-ORC site templates at distances up to 400 bp apart (Fig. 4B). This suggested that MCM hexamers may translocate along DNA before forming a stable double hexamer. Such long-range directed translocation is used by type III restriction enzymes and mismatch repair proteins (17, 18) and can be blocked by an intervening obstacle. A covalent protein roadblock (19) impaired MCM loading when positioned between two ORC sites, but not when positioned 3′ of either one or two sites (Fig. 4, C and D, and fig. S9).

Fig. 4 Two-site loading exhibits flexible spacing and is sensitive to an intervening roadblock.

(A and B) Synergistic loading (A) and origin activity (B) exhibit flexible intersite spacing. (C) A covalent DNA-protein roadblock between two sites inhibits synergistic loading. (D) Only the complete roadblock reaction leads to a block in loading. (E) Proposed model for MCM loading. See text for details.

We propose (Fig. 4E) that each MCM hexamer in the double hexamer is loaded by the same mechanism in a reaction requiring ORC, Cdc6, and Cdt1. Our results do not address whether these are independent or concerted events. Budding yeast origins comprise a single high-affinity ORC binding site and multiple lower-affinity sites in the opposite orientation; we describe this arrangement as “quasi-symmetrical.” This role for multiple ORC binding sites, which is supported by the bimodal distribution of ORC bound at origins in vivo (20), may explain why there is a switch in sequence bias from T-rich to A-rich between the high-affinity ORC site and flanking sequence and how the positioning of nucleosomes at both ends of the nucleosome-free region at yeast origins can be affected by ORC (13). We note that archaeal origins have a related architecture with head-to-head high-affinity ORC binding sites, separated by 65 to 70 bp of AT-rich DNA (21)—close to the distance that yielded optimal synergistic loading in budding yeast and the length of DNA covered by the double hexamer (~68 bp) (1). Deletion analysis of a fission yeast replication origin in vivo also revealed a requirement for two ORC binding sites in MCM loading (22). Origins in higher eukaryotes show little sequence specificity (23), but they may act in a two-ORC mechanism similar to that of the yeast system under nonspecific conditions.

Why the second ORC was not detected in single-molecule experiments (8) is unclear; perhaps transient ORC binding at weak sites was missed for technical reasons, or unlabeled ORC may have been responsible for loading on the secondary site. Alternatively, there may be a second, inefficient, pathway in which a single ORC molecule hops between two sites to load each hexamer, which might be more prevalent at the low ORC concentrations used in these experiments and with the specific natural origin used. Such a “hopping” model preserves the inherent symmetry of our proposed model, but our results indicate that it cannot be a major pathway.

The formation of stable double hexamers from imprecisely spaced ORC binding sites requires the translocation of single MCM hexamers. Although this may occur by passive sliding, we note that MCM loading requires ATP hydrolysis by both MCM hexamers (9, 10) and that the predicted direction of movement of MCM loaded at distant ORC sites is the same as that of CMG during replication (24), leading us to suggest that translocation is an active process. Consistent with this idea, the eukaryotic MCM complex has DNA translocase activity without Cdc45 or GINS (25), and archaeal MCM can translocate over duplex DNA (26).

Supplementary Materials

Materials and Methods

Figs. S1 to S9

Tables S1 and S2

References (27, 28)

References and Notes

  1. Acknowledgments: We thank H. Yardimci for advice and reagents, H. Patel and G. Kelly for bioinformatic analyses, and A. McClure for critical reading of the manuscript. Supported by the Francis Crick Institute, which receives its core funding from Cancer Research UK (FC001066), the UK Medical Research Council (FC001066), and the Wellcome Trust (FC001066). This work was also funded by a Wellcome Trust Senior Investigator Award (106252/Z/14/Z) and a European Research Council Advanced Grant (669424-CHROMOREP) to J.F.X.D. The authors declare no conflict of interest. All data are contained within the paper and the accompanying supplementary materials. J.F.X.D. and G.C. designed the experiments, analyzed data, and wrote the manuscript; G.C. performed all experiments.
View Abstract

Navigate This Article