A mechanism of origin licensing control through autoinhibition of S. cerevisiae ORC·DNA·Cdc6

The coordinated action of multiple replicative helicase loading factors is needed for the licensing of replication origins prior to DNA replication. Binding of the Origin Recognition Complex (ORC) to DNA initiates the ATP-dependent recruitment of Cdc6, Cdt1 and Mcm2-7 loading, but the structural details for timely ATPase site regulation and for how loading can be impeded by inhibitory signals, such as cyclin-dependent kinase phosphorylation, are unknown. Using cryo-electron microscopy, we have determined several structures of S. cerevisiae ORC·DNA·Cdc6 intermediates at 2.5–2.7 Å resolution. These structures reveal distinct ring conformations of the initiator·co-loader assembly and inactive ATPase site configurations for ORC and Cdc6. The Orc6 N-terminal domain laterally engages the ORC·Cdc6 ring in a manner that is incompatible with productive Mcm2-7 docking, while deletion of this Orc6 region alleviates the CDK-mediated inhibition of Mcm7 recruitment. Our findings support a model in which Orc6 promotes the assembly of an autoinhibited ORC·DNA·Cdc6 intermediate to block origin licensing in response to CDK phosphorylation and to avert DNA re-replication. Cryo-EM structures of S. cerevisiae ORC bound to DNA and Cdc6 reveal an autoinhibited conformation and suggest a mechanism of origin licensing control in response to CDK phosphorylation.

T he replication of DNA in eukaryotic cells requires the bidirectional loading of replicative helicases onto origins of replication and their timely activation for replisome assembly (reviewed in refs. [1][2][3]. The hexameric initiator ORC first binds and bends DNA, and then associates with the coloader Cdc6 to form a ring-shaped, heptameric initiator·co-loader complex that is competent for recruiting the Mcm2-7 helicase motor and Cdt1 to origins (reviewed in refs. 4,5 ). Docking of hexameric Mcm2-7 onto ORC·Cdc6 guides DNA insertion into the central helicase pore through an opening in the Mcm2-7 ring (the Mcm2/5 gate) [6][7][8][9][10] . Stable association of Mcm2-7 with DNA requires the coordinated loading of two hexamers, oriented in a head-to-head fashion, to produce double hexamers that topologically entrap duplex DNA [11][12][13] . Origins licensed in this manner during the G1 phase of the cell cycle are primed for Dbf4dependent kinase (DDK) and cyclin-dependent kinase (CDK) activation in S phase 14 .
Apart from Cdt1 and Orc6 (a subunit of ORC), all loading factors belong to the superfamily of ATPases Associated with various cellular Activities (AAA+) 15,16 . ATP binding and hydrolysis are essential for Mcm2-7 recruitment and loading, driving the assembly of various initiation intermediates and the conformational changes needed for helicase deposition 1,5,17 . The AAA+ folds are appended by winged-helix (WH) domains in ORC, Cdc6, and some Mcm2-7 subunits that help stabilize interactions within and between different loading factors 8,18 . While all loading factors can be recruited in the absence of ATP hydrolysis to form an ORC·Cdc6·Cdt1·Mcm2-7 (OCCM) intermediate in which DNA is successfully inserted into the pore of the first Mcm2-7 ring 12,13,19,20 , conversion of ATP to ADP by Mcm2-7 is critical for helicase ring closure and double hexamer formation [21][22][23] . By contrast, ATP hydrolysis by ORC promotes repeated Mcm2-7 loading, while that of Cdc6 mediates disassembly of unproductive loading intermediates, but neither are necessary for double hexamer assembly in vitro 21,22,[24][25][26] . Nonetheless, ATP constitutes a critical co-factor for ORC and Cdc6, stabilizing key interactions within the ternary ORC·D-NA·Cdc6 complex [27][28][29][30][31] . Of particular relevance are the active ATPase centers between Orc1 and Orc4, and between Cdc6 and Orc1, which are the only ATP sites that contain all the signature sequence motifs required for catalysis (Walker A and B, sensor 1 and 2, arginine finger) 16,18 . Prior structural studies have found that these ATPase sites exist in closed conformations "primed" for hydrolysis 6,8,9,[32][33][34][35] ; yet it is important to suppress ATPase activity sufficiently to avoid prematurely destabilizing initiation intermediates, which would lead to abortive loading events and decreased loading efficiency. The physical basis for how ATPase activity by ORC·Cdc6 is kept low to prevent premature catalysis is poorly understood.
Helicase loading (origin licensing) and activation (origin firing) are temporally separated into different stages of the cell cycle to prevent re-replication of DNA 36 . Eukaryotes have evolved various strategies to restrict Mcm2-7 loading to late M and G1 phases 37,38 . In S. cerevisiae, CDK-mediated phosphorylation of ORC aids in preventing helicase loading outside of G1 by activating a quality control pathway that also helps release unproductive loading intermediates during origin licensing in G1 by stimulating the ATPase activity of ORC·Cdc6 24,[39][40][41] . However, whether and how posttranslational modifications alter the structure of the initiator·co-loader to prevent loading is not well characterized. Likewise, it is unknown whether defined intermediates exist that could specifically stall Mcm2-7 recruitment or deposition.
Here, we report cryo-EM structures of ARS1-bound S. cerevisiae (Sc) ORC and of the ternary complex with Cdc6 at sub-3 Å resolutions. Overall, the organization of ScORC·DNA and ScORC·DNA·Cdc6 resembles previous lower resolution structures 6,8,9,33,34 , but ScORC·DNA·Cdc6 is seen in two distinct ring conformations that modulate DNA contacts of the co-loader Cdc6. The Orc1·Orc4 and Cdc6·Orc1 ATPase sites reside in a pre-hydrolysis state, with water molecules in non-lytic positions and a disengaged sensor 1 at the Orc1·Orc4 site, and need to be further activated for catalysis to occur. Strikingly, the N-terminal domain of Orc6 docks onto the ORC·Cdc6 ring by binding to Orc1 and Cdc6 and is situated in a manner that sterically blocks Mcm7 attachment upon CDK-mediated ORC phosphorylation. Together, our findings reveal insights into ATPase site regulation of the initiator·co-loader and provide models for how an autoinhibited ScORC·DNA·Cdc6 state helps control timely Mcm2-7 loading.

Results and discussion
To determine a sub-3 Å structure of the eukaryotic initiator· co-loader assembly, we chose to study S. cerevisiae (Sc) ORC·DNA·Cdc6 because, in contrast to metazoan ORC, the yeast complex binds DNA in a defined sequence register 6,33,[42][43][44] , which we predicted would reduce heterogeneity and improve the attainable resolution. We reconstituted the ternary complex in vitro on an 84 bp ARS1 DNA fragment containing the ACS and B1 elements that are required for origin function 45 , followed by size exclusion chromatography to enrich for DNA-bound ORC·Cdc6 assemblies ( Supplementary Fig. 1). Cryo-EM analysis of the sample revealed the presence of Cdc6 in a subset of the particles; however, the weaker signal in this region of a consensus cryo-EM map (at an overall resolution of 2.3 Å) indicated compositional and/or conformational heterogeneity ( Supplementary  Fig. 2). Using a masked 3D classification approach of the Cdc6 region, we obtained structures of ScORC·DNA (OD) and ScORC·DNA·Cdc6 (ODC) assemblies at 2.5-2.7 Å resolution (Supplementary Figs. 2 and 3). The higher resolution of our structures compared to previous ones and the increased occupancy of the co-loader allowed de novo model building of Cdc6, correction of register shifts, and the placement of ordered solvent molecules (Supplementary Fig. 4 and Supplementary Table 1).
Our ScORC·DNA complex structure is similar to the previous 3.0 Å model with few exceptions 6 : first, we observe only weak density for the winged-helix (WH) domain of Orc2, which is sandwiched between Orc1 and Orc2 in the earlier structure of the DNA-bound yeast initiator obtained from crosslinked samples 6 ( Fig. 1a). This region is more flexible in the absence of crosslinker and may thus not stably occupy the region between Orc1 and Orc2. Second, Arg 254 in the basic patch of Orc2 makes a basespecific contact with a guanine in the major groove of the B1 element of ARS1; this side chain is flipped away in ARS305bound ORC, which contains an adenine instead of a guanine at this position 6 . Local differences in base recognition may thus contribute to observed variations of ScORC's affinity for different ARS elements 46 .
The ORC·Cdc6 ring adopts distinct conformational states. In our cryo-EM structure of the ternary complex, Cdc6 is seen to complete the ORC ring as expected by binding in between Orc1 and Orc2 (Fig. 1b). Surprisingly, masked classification of the Cdc6 region revealed that the ORC·Cdc6 ring adopts two distinct conformations that differ in pitch and the positioning of Cdc6 (Fig. 1c, d). The two states arise from a rocking motion of mainly Orc1 and Cdc6, which leads to ring flattening as the complex transitions from state 1 to state 2 (referred to as ODC1 and ODC2 hereafter, Fig. 1d). Both states also differ from the ring conformation seen in a loading intermediate containing DNA, ORC, Cdc6, Cdt1, and Mcm2-7 (OCCM) 8 , indicating this change in ring conformation may be linked to the functional ORC·Cdc6 cycle during Mcm2-7 loading (Fig. 1d).
Interestingly, the two ScORC·DNA·Cdc6 structures we determined differ in the DNA binding mode of the Cdc6 subunit. In ODC2, Cdc6, and in particular the B-loop element in its AAA + domain, a loop region that is located after the conserved catalytic Walker B motif, shifts by~5 Å and increases its contact area with DNA ( Fig. 1e-g). The rearrangement of Cdc6 results in an additional hydrogen bond between Asn 232 and the sugarphosphate backbone, and Arg 240 moves closer to form a salt bridge with a phosphate in the ACS (Fig. 1g) stabilizing the initiator on DNA 33 . Collectively, these results establish that the B-loop element is a bona fide DNA binding element in the AAA+ domain of eukaryotic Cdc6 co-loaders.
The Orc1·Orc4 and Cdc6·Orc1 ATPase sites reside in a prehydrolysis state. The initiator ORC contains one functional ATPase center that is formed at the interface between Orc1 and Orc4 protomers 26,47 . A second ATPase site is formed by Orc1 and Cdc6 upon recruitment of the co-loader 29,48 . The ATPase activity in ScORC·DNA and ScORC·DNA·Cdc6 is kept low through inhibition by DNA binding and the absence of stimulatory factors such as the Mcm3-WH domain 24,27 . In prior structures of yeast and metazoan initiator-containing assemblies, these interfaces formed closed ATPase sites with a trans-acting arginine finger, a residue essential for catalysis in AAA+ ATPases, engaging the ATP γ-phosphate, indicating these sites were primed for ATP hydrolysis 6,8,9,[32][33][34][35] . However, the structural features of the Orc1·Orc4 ATPase center in previous models fail to explain the suppressed ATPase activity of ORC upon engaging DNA. Our 2.55 Å resolution cryo-EM map of ScORC·DNA afforded the location of active-site water molecules (which were not resolved in any of the previous eukaryotic initiator structures due to lower resolutions) and a better definition of the side chain positions of active site residues (Fig. 2a). As expected, the conserved Walker A lysine of Orc1 (K485) bonds with the βand γ-phosphate, while the Orc4 arginine finger (R267) and Orc1 sensor 2 arginine (R704) engage the γ-phosphate (and also the α-phosphate in case of the sensor 2). The Walker B aspartate (Orc1-D566) binds magnesium through two water molecules (Fig. 2b). The catalytic glutamate (Orc1-E567), a residue that activates a lytic water for nucleophilic attack of the γ-phosphate during catalysis, points towards the ATP γ-phosphate, its position stabilized by tether residues in Orc4, Y232, and R263 32 , and a bonding interaction with a water molecule in the magnesium coordination sphere. Surprisingly, the side chain of sensor 1 (Orc1-N600), a conserved polar residue in AAA+ ATPases that supports ATP hydrolysis by helping to activate the lytic water and/or by bonding with the ATP γ-phosphate 15,49,50 , is flipped away from ATP ( Fig. 2a, b). Apart from the magnesium-coordinating waters, two additional ordered water molecules are observed in the active site in a cavity near the γ-phosphate (W1 and W2, Fig. 2a, b). W1 hydrogen bonds with the catalytic glutamate, while both W1 and W2 also directly bond with the γ-phosphate. However, neither of the ordered waters is correctly positioned for an in-line attack on the γ-phosphate. Together, these findings argue that the Orc1·Orc4 ATPase in DNA-bound ScORC prefers to reside in a pre-hydrolysis conformation. The presence of two active-site waters could suggest that ORC may employ a "multiple"-water mechanism for catalysis, which has been suggested for other nucleotide-hydrolyzing systems and appears energetically favorable over a single-water mechanism 51,52 .
How is the Orc1·Orc4 ATPase site then activated for catalysis? Mutagenesis of the Orc1-sensor 1 in budding yeast has shown that a polar residue in this position is important for cell viability, consistent with a role of Orc1-N600 in ATP hydrolysis 53 . Thus, we predict that the following rearrangements are needed for ATP hydrolysis: (1) a rotamer switch of the sensor 1 residue, (2) repositioning of at least one of the ordered water molecules in the active site for an in-line attack, and (3) a slight shift of the Orc4-R263 tether may also be needed. The Orc4-R263 tether may fulfill a role analogous to the arginine coupler in E. coli DnaC, which has been implicated in active site regulation of the bacterial helicase loader 54 . In the DNA-free DnaBC complex, the arginine coupler DnaC-R216 occupies the site of the lytic water prior to ATP hydrolysis; likewise, Orc4-R263 may impede correct positioning of the water molecule for catalysis ( Supplementary  Fig. 5). One possibility is that Cdc6 binding drives ScORC into a hydrolysis-competent conformation; however, inspection of the Orc1·Orc4 ATPase site configuration in our 2.5 and 2.7 Å cryo-EM structures of ODC1 and ODC2 reveals that the active site waters remain in a non-lytic position ( Supplementary Fig. 6a-c). The Orc1-sensor 1 is also flipped away from ATP in ODC1, although it appears to move closer to ATP in ODC2 ( Supplementary Fig. 6a-c). Regardless, the Orc1·Orc4 ATPase centers in all three states (OD, ODC1, and ODC2) are similar and preferentially persist in a pre-hydrolysis state. Thus, ORC's ATPase activation is likely coupled to a later event during Mcm2-7 loading to allow the release of ORC from DNA and recycling of the initiator 26,55 . Indeed, ORC's ATPase activity is not required for Mcm2-7 loading in vitro but is essential for ORC function in vivo 22,26,56 . Higher resolution structures of Mcm2-7containing loading intermediates than currently available 8,9,57 will be required to define the alterations that trigger ATP hydrolysis by ORC.
The second ATPase center in the ternary ScORC·DNA·Cdc6 complex at the interface between Orc1 and Cdc6 likewise appears to reside in a pre-hydrolysis state. Overall, the Cdc6·Orc1 sites in ODC1 and ODC2 are configured very similarly to each other and to that of Orc1 and Orc4, but the Cdc6-sensor 1 (N263), which is important for budding yeast viability and ATP hydrolysis by Cdc6 48,58,59 , points into the active site and bonds with the γ-phosphate (Fig. 2c, d and Supplementary Fig. 6d-f). A water molecule positioned for an in-line nucleophilic attack, however, is not seen. We postulate that the lytic water will be stabilized for nucleophilic attack upon docking of the winged-helix domain of Mcm3 onto ORC·Cdc6, which binds to the Cdc6 ATPase in the OCCM complex and has been shown to stimulate ATP hydrolysis of the ternary complex by several fold 8,24 . Considering the quality control function of the Cdc6 ATPase, it seems advantageous to keep its basal activity in ORC·Cdc6 low; hyperactivity would likely interfere with productive Mcm2-7 loading by prematurely releasing loading factors 21,22,24 . Intriguingly, Cdc6 binding to ScORC causes conformational rearrangements in the B-loop of Orc1 that help reorient the Orc1 arginine finger helix (with arginine finger R616) for completion of the Cdc6 ATPase site ( Supplementary Fig. 7). In Drosophila ORC, the Orc1-B-loop region has been identified as an element that contributes to the coordination of DNA-binding and ATP hydrolysis in the initiator and may help with inter-ATPase site communication 33 . Combined, these observations suggest that coupling conformational changes in the Orc1 B-loop to ATPase site configuration may be a universal feature of eukaryotic initiator-co-loader assemblies.
The N-terminal cyclin box domain of Orc6 docks onto the ORC·Cdc6 ring. During data processing of the ternary complex, we noticed that the cryo-EM maps contained weak, fragmented density at the cleft formed by the winged-helix and AAA+ domain tiers near the Orc1 subunit ( Supplementary Fig. 2c). Masked classification revealed that, in about half of the ScORC·DNA·Cdc6 particles, an additional domain composed of several α-helices was docked onto the ORC·Cdc6 ring, which has not been previously observed in the ternary complex. With a resolution range of 2.9-3.5 Å, this region of the cryo-EM map was sufficiently resolved for de novo model building (Supplementary Figs. 3f, 4b). To our surprise, we identified this region as the N-terminal cyclin box fold (CB N ) of Orc6 ( Fig. 3a, b). Orc6-CB N binds in a groove formed by the lid subdomain of the Orc1-AAA+ region and the winged-helix fold of Cdc6 (Fig. 3a, c). The binding of Orc6-CB N appears to stabilize the Cdc6-WH fold since the cryo-EM map shows higher resolution features in this region compared to the one without a docked Orc6-CB N . The tripartite binding site buries~990 Å 2 of solventaccessible surface area and is stabilized by hydrogen bonds and salt bridges formed between Orc6-CB N and the Orc1-AAA+ lid, and by van der Waals interactions between Orc6-CB N and Cdc6-WH (Fig. 3d). This bonding pattern explains why Orc6-CB N is not seen docked onto the ORC ring in ScORC·DNA without Cdc6. In contrast to Orc6-CB N , the C-terminal Orc6 cyclin box fold (CB C ) is bound at the opposite side of the ORC·Cdc6 ring and is sandwiched between the Orc3 insertion domain, the winged-helix domains of Orc3 and Orc5, and the bent DNA duplex as seen in previous structures of the yeast initiator and other loading intermediates 6,9,34,57 (Fig. 3a). This configuration separates the C-terminal residue of Orc6-CB N and the N-terminal residue of Orc6-CB C by 123 Å, which can easily be bridged by the 156 amino acid-long, flexible linker that connects both domains but is not resolved in our cryo-EM maps (Fig. 3a, b). Further subclassification of Orc6-CB N particles revealed that they can also adopt both ODC1 and ODC2 ring conformations, indicating that Orc6-CB N docking does not alter the equilibrium between both states, nor does it change the configuration of the ATPase centers ( Supplementary Fig. 8a).
To confirm the interaction between the N-terminal domain of Orc6, Orc1, and Cdc6 with an independent technique, we turned to crosslinking mass spectrometry. Using the cleavable crosslinker DSSO, we identified several crosslinks between lysines in the Orc6-CB N or the proximal Orc6 linker region (immediately following the CB N up to residue 150) and the Cdc6-WH domain or the Orc1-AAA+ lid region ( Fig. 3e and Supplementary Fig. 8b, c). The Orc1-WH domain also crosslinked to the proximal Orc6 linker and is likely located along the path the chain takes towards the Orc6-CB C . By contrast, the Orc6-CB C predominantly crosslinked to Orc2 on the opposite side of the ORC·Cdc6 ring ( Supplementary Fig. 8b, c). These findings agree with our cryo-EM data and establish that a docked Orc6-CB N onto the ORC·Cdc6 ring (referred to as ORC·DNA·Cdc6·Orc6-CB N hereafter) is a bona fide state of the initiator·co-loader complex adopted by a subset of ternary ScORC·DNA·Cdc6 assemblies.  60 (Fig. 4a). Structural alignment of Orc6-CB N as seen in the MO structure and in our structure revealed that the overall fold of the N-terminal Orc6 domain is not significantly altered between both assemblies (RMSD 3.25 Å), and is also similar to that of the human counterpart 57,61 ( Fig. 4b and Supplementary Fig. 4c). Intriguingly, mapping the binding sites for Mcm2/Mcm6 (in MO) and Orc1/Cdc6 (in ScORC·DNA·Cdc6·Orc6-CB N ) onto Orc6-CB N shows that there is a substantial overlap between the interacting regions, indicating that docking of Orc6-CB N onto Mcm2-7 and the ORC·Cdc6 ring is mutually exclusive (Fig. 4c and Supplementary  Fig. 9a) 9 . Surprisingly, comparison of the cryo-EM maps of this "semi-attached" OCCM and our assembly revealed that the map region previously assigned as Mcm7-WH corresponds to Orc6-CB N (Supplementary Fig. 10 39,40 , which reside in the proximal linker region (immediately following the Orc6-CB N ) that we identified to crosslink to Orc1 and Cdc6 ( Fig. 3e and Supplementary Figs. 1b, 8c, and 11). We thus hypothesized that posttranslational modification of phosphosites in Orc6 may promote docking of the N-terminal Orc6 domain onto the ORC·Cdc6 ring to prevent productive OCCM assembly.
To test this premise, we devised an assay to monitor recruitment of Mcm7 to the initiator·co-loader complex. Mcm7 and the Orc6-CB N domain interact with the same site in ScORC·DNA·Cdc6, and both binding events are thus expected to be mutually exclusive (Fig. 5a). In the presence of ATPγS, we observed weak but reproducible co-association of Mcm7 with ScORC·DNA·Cdc6 above background levels (no ATPγS reactions) in pulldowns (Fig. 5b)   c-e Deletion of Orc6-CB N abrogates the phosphorylation-dependent regulation of ScMcm7 binding to ScORC·DNA·Cdc6. Coomassie-stained SDS-PAGE gels of elutions from amylose pulldowns are shown in (b-e). The mean and standard error of the mean of ScMcm7 band intensities (n = 4 independent recruitment assays) are plotted in (c-e). Note that Mcm7 migrates as a doublet and that Orc2 and the unstructured linker region of Orc6ΔCB N are prone to proteolytic cleavage during expression and purification. Asterisks mark degradation products of ORC subunits that are stably integrated into ORC. In e, a higher percentage gel was used to visualize Orc6ΔCB N /Δlinker due to its low molecular weight. SDS-PAGE gels of inputs from b-e are included in Supplementary Fig. 12. Source data are provided as a Source Data file. phosphorylated by CDK (Fig. 5b). If these differences in Mcm7 recruitment are indeed coupled to phosphorylation-induced docking of Orc6-CB N onto the ORC·Cdc6 ring, then deletion of this domain should abrogate the phosphoregulation of Mcm7 recruitment. Congruent to this prediction, ORC assemblies that lacked the N-terminal Orc6 cyclin box fold, alone (Orc6 ΔCB N ) or in combination with the flexible linker region (Orc6 ΔCB N / Δlinker), recruited similar amounts of Mcm7 irrespective of phosphatase or kinase pretreatment (Fig. 5c-e). These results are consistent with an autoinhibitory role of Orc6-CB N during Mcm2-7 recruitment in response to ORC phosphorylation. Collectively, our biochemical and structural findings support a revised model for the regulation of Mcm2-7 loading (Fig. 6). Cdc6 binding to DNA-bound ORC completes the AAA+ ATPase ring, which can adopt two distinct conformations that are characterized by repositioning of Cdc6. In both states, the predominant pre-hydrolysis configuration of the ATPase sites would maintain the low basal ATPase activity of ORC·Cdc6, which likely prevents premature disassembly of the heptameric intermediate. When Orc6 is not phosphorylated (under conditions of low CDK activity), Orc6-CB N is flexibly tethered to ORC and the ORC·Cdc6 ring is available for full Mcm2-7 attachment in the pre-insertion OCCM 9 and subsequent loading of the first Mcm2-7 hexamer onto DNA. By contrast, high CDK activity leads to ORC phosphorylation and autoinhibition of ORC·DNA·Cdc6 through docking of the Orc6-CB N onto the initiator·co-loader ring, either in cis (involving the Orc6-CB N of the same ORC) or in trans (involving the Orc6-CB N of a second ORC). In both cases, Mcm2-7 recruitment to autoinhibited ORC·DNA·Cdc6 would stall at a partially attached OCCM stage 9 , stimulating Cdc6 ATPase activity (through Mcm3-WH binding), OCCM disassembly, and Mcm2-7/Cdt1 release as part of the quality control pathway 21,22,24 (Fig. 6). Such a mechanism may apply to the loading of the first and second Mcm2-7 hexamer, as recruitment of both occurs through a similar ORC/Cdc6-mediated mechanism 24,57,60,63 .
Additionally, an autoinhibited ORC·DNA·Cdc6 state could also impede loading by sequestering Orc6-CB N from other partner proteins, for example, Cdt1 39,64 or Mcm2/6 in MO 57 (after association of the second Cdc6 55 ), which could potentially uncouple the loading of the first and second hexamer. Within ORC, both Orc2 and Orc6 are substrates for CDK phosphorylation 39,40 . Future studies are needed to address whether phosphorylation of Orc6 alone or of both Orc2 and Orc6 contributes to ORC·DNA·Cdc6 autoinhibition; however, because of the proximity of the Orc6 phosphorylation sites to the Orc6-CB N , we favor a model wherein autoinhibition is regulated by Orc6 phosphorylation alone and Orc2 phosphorylation prevents origin licensing through a distinct but complementary mechanism.
At this point, it is unknown whether autoinhibition of ORC·DNA·Cdc6 extends to other eukaryotic systems, but several observations argue that this state may be limited to S. cerevisiae and closely related fungal species. First, structural studies of Drosophila ORC·DNA·Cdc6 have so far not resolved the  Orc6-CB N or -CB C 33 . Second, human Orc6 only weakly associates with the Orc1-5 subunits [65][66][67] . Third, the linker region joining Orc6-CB N and Orc6-CB C is substantially shorter in metazoan and in S. pombe Orc6 orthologs and not long enough to reach across the ORC·Cdc6 ring, although binding of Orc6-CB N in trans to a second ORC may be possible 67,68 ( Supplementary  Fig. 11). Not surprisingly, these systems also do not rely on Orc6 phosphorylation but employ other strategies to prevent Mcm2-7 loading in S phase and G2 37,38 . Nonetheless, it will be interesting to see whether functionally related inhibitory states that help to temporally coordinate the various steps involved in Mcm2-7 double hexamer formation can be identified in metazoans.

Methods
Expression and purification of S. cerevisiae ORC and Cdc6. Wild-type Saccharomyces cerevisiae (Sc) Orc1-6 subunits and Cdc6 were cloned into pFastBacderived baculovirus expression vectors (series 4 vectors, QB3 MacroLab, UC Berkeley) by ligation independent cloning (LIC). ScOrc1 and ScOrc4 were N-terminally tagged with hexa-histidine (6xHis) and maltose binding protein (MBP), respectively, while Cdc6 was fused to an N-terminal 6xHis-MBP tag. All affinity tags were followed by a Tobacco Etch Virus (TEV) protease cleavage site. For purification of MBP-ORC, full-length Orc6 and truncations (Orc6Δ1-120 and Orc6Δ1-269) were generated as C-terminal 6xHis fusions without a TEV site. Full-length ScORC was expressed in High5 insect cells using the Bac-to-Bac Baculovirus Expression System (Thermo Fisher Scientific). Individual baculoviruses for each ORC subunit and Cdc6 were generated in Sf9 cells by transfection of bacmids isolated from DH10Bac E. coli with Cellfectin II (Thermo Fisher Scientific). After amplification in Sf9 cells, viruses were used to co-infect 4 L High5 cell cultures for ScORC expression. Two days later, High5 cells were harvested by centrifugation and the cell pellet was resuspended in~140 mL lysis buffer (50 mM Tris-HCl (pH 7.8), 300 mM KCl, 10% glycerol, 50 mM imidazole, 1 mM ß-ME, 200 μM PMSF, 1 μg/mL leupeptin). The cell suspension was sonicated and the lysate clarified by ultracentrifugation at 142,032 × g for 45 min (Beckman Coulter Optima L-80 XP ultracentrifuge). The soluble fraction was subjected to ammonium sulfate precipitation (final 20% (v/v)) and subsequently cleared once more by ultracentrifugation. The supernatant was loaded onto a 5 mL HisTrap HP Nickel-affinity chromatography column (Cytiva) that was washed with 60 mL lysis buffer prior to ScORC elution with a 50-250 mM imidazole gradient. Peak fractions were further purified on 7.5-10 mL amylose columns (New England Biolabs) in 50 mM Tris-HCl (pH 7.8), 300 mM KCl, 10% glycerol, 1 mM ß-ME and eluted with 20 mM maltose. ScORC was incubated with 6xHis-tagged TEV protease overnight and then purified over another 5 mL HisTrap HP column (Cytiva, equilibrated in 50 mM Tris-HCl (pH 7.8), 300 mM KCl, 10% glycerol, 50 mM imidazole, 1 mM ß-ME) to remove the protease. The flow-through was concentrated and loaded onto a HiPrep 16/60 Sephacryl S-400 HR column (Cytiva) equilibrated in 25 mM HEPES-KOH (pH 7.6), 300 mM KCl, 10% glycerol, 1 mM DTT. ScORC peak fractions were pooled and concentrated in a 30 K Amicon Ultra-15 concentrator (Millipore), aliquoted, and flash-frozen in liquid nitrogen for storage at −80°C. For purification of MBP-ORC, TEV cleavage and the subsequent Nickel affinity step were omitted. Full-length ScCdc6 was expressed and purified as described for ScORC with the exception that all buffers were supplemented with 10 mM magnesium acetate and a HiLoad 16/600 Superdex 200 pg column (Cytiva) was used for final gel filtration chromatography. All purification steps were done at 4°C.
Cryo-EM data collection and image processing. Cryo-EM grids were prepared by applying 3.5 μL of concentrated ScORC·DNA·Cdc6 complex (supplemented with 0.001% TWEEN20) to 300-mesh R1.2/1.3 UltrAuFoil grids (Quantifoil Micro Tools GmbH), which had been freshly plasma-cleaned for 2 min at 5 W (18.9% H 2 /81.1% O 2 gas mixture). The sample was incubated on the grid for 10 s, blotted for 2 s and vitrified in liquid ethane using a Vitrobot Mark IV plunge freezer (Thermo Fisher Scientific). Cryo-EM data were recorded as dose-fractionated movies on a Titan Krios G2 cryo-electron microscope (Thermo Fisher Scientific) at an acceleration voltage of 300 kV using a post-GIF (Quantum LS imaging filter with a slit width of 20 eV) and a Gatan K2 summit direct electron detector. The microscope was also equipped with a spherical aberration corrector (CEOS GmbH Heidelberg, Germany). Automated data collection was performed with EPU software (Thermo Fisher Scientific) using fringe-free illumination with a beam-size of 750 nm and 3 exposures per 1.2 μm hole. The targeting defocus was set to −0.8 to −1.6 μm. Movies were recorded at a pixel size of 0.86 Å as 50 frames for a total of 7.5 s with a dose rate of 6.6 e − /Å 2 per second, yielding a total electron dose of 49.3 e − /Å 2 .
For image processing, dose-fractionated movies were first motion-corrected with MotionCor2 70 , and contrast transfer function parameters were determined with GCTF using non-dose-weighted, motion-corrected sums of movie frames 71 . Particles were auto-picked with GAUTOMATCH 72 from the dose-weighted movie-frame sums using low-pass filtered (to 30 Å) class averages obtained from a smaller test dataset as templates. Picked particles were extracted and normalized in RELION 2.1 73,74 with a box size of 300 × 300 pixels at a pixel size of 0.86 Å.
Extracted particles were further processed in RELION 2.1 74 and RELION 3.0 75 . 3D classification into four classes using a map of a previously recorded ScORC·DNA·Cdc6 test dataset (low-pass filtered initially to 60 Å) as a reference yielded one class with clearly resolved secondary structure elements that were refined using the autorefinement procedure in RELION. Visual inspection of the map showed that the density for Cdc6 was weaker than for ORC subunits, indicating that Cdc6 was substoichiometric or adopted multiple conformations. To sort particles further, the refined particle set was subjected to 3D classification within a soft-edged Cdc6 mask (generated from a Drosophila Cdc6 structure 33 docked into the cryo-EM map) without particle alignment. Particles in each class were subjected to 3D refinement, which yielded one map for ScORC·DNA (without Cdc6) and two maps for ScORC·DNA·Cdc6 that differed in the conformation of Cdc6 in the ORC·Cdc6 ring (states I (ODC1) and II (ODC2)). Both ScORC·DNA·Cdc6 maps contained additional weak and fragmented density near the Orc1 and Cdc6 subunits. Masked classification using a soft-edged mask encompassing this density allowed for sorting of ScORC·DNA·Cdc6 particles (after aligning all ScORC·DNA·Cdc6 particles to a consensus ScORC·DNA·Cdc6 map) into two groups that were further refined, one that contained a well-ordered N-terminal cyclin box domain of Orc6 docked onto the ORC·Cdc6 ring and another where this domain was flexible. Further subclassification and refinement revealed that the ternary complex with a docked Orc6 N-terminal domain could be subdivided into ODC1 and ODC2 states based on different conformations of Cdc6. Bayesian particle polishing and CTF refinement were used to generate shiny particles and optimize per-particle defocus parameters 75 . The final cryo-EM maps were B-factor sharpened with RELION's post-processing function applying a global, soft-edged volume mask low-passed filtered to 15 Å. Resolution values were calculated using gold-standard Fourier Shell Correlation (FSC) in RELION 76 .
Model building and refinement. Model building was initiated by docking the structure of ScORC•DNA (PDB 5zr1 6 ) into the cryo-EM maps of ScORC•DNA and ScORC•DNA•Cdc6, followed by manual rebuilding in COOT 77 . In contrast to previous low-resolution models of ScCdc6 and ScOrc6-CB N 8,9,34,57 , the high resolution of our cryo-EM maps enabled manual de novo building of Cdc6 and of the N-terminal cyclin box fold of Orc6 and clear identification of many side chain densities and unambiguous amino acid register assignment. The model coordinates were improved by iterative real-space refinement in PHENIX 78,79 (with secondary structure restraints for protein and DNA, and rotamer and Ramachandran restraints for protein) and manual rebuilding in COOT. Unsharpened, sharpened, and PHENIX density-modified cryo-EM maps 80 were all used for map interpretation, model building, and refinement and were smoothened by resampling in COOT by a factor of 1.5 to aid interpretation. The density-modified maps allowed clear resolution of well-ordered water molecules, which were placed in the cryo-EM maps using a combination of PHENIX.DOUSE and the "find water" option in COOT, followed by manual inspection. The final models were validated using MolProbity 81  Crosslinking mass spectrometry. ScORC and ScCdc6 proteins were dialyzed into 25 mM HEPES-KOH (pH 7.6), 250 mM potassium acetate, 10 mM magnesium acetate, 1 mM DTT at 4°C overnight. Proteins were centrifuged for 10 min at 21000 × g at 4°C to remove any precipitates that formed during dialysis. The ternary ScORC•DNA•Cdc6 complex was reconstituted as for structural studies but not further purified by gel filtration chromatography. For crosslinking, approximately 35 μg of reconstituted ternary complex was diluted with 73 μL of reconstitution buffer (25 mM HEPES-KOH (pH 7.6), 250 mM potassium acetate, 10 mM magnesium acetate, 1 mM ATP, 1 mM DTT), and CID-cleavable disuccinimidyl sulfoxide (DSSO, spacer arm 10.3 Å, Thermo Fisher Scientific) was added to the protein to a final concentration of 0.5, 2, or 4 mM. After 1 h of incubation at 10°C, crosslinking reactions were quenched by the addition of Tris-HCl (pH 6.8) to a final concentration of 50 mM and incubated for 1 h at room temperature. The crosslinking reagent was removed by three buffer exchanges into 8 M urea dissolved in 50 mM HEPES-KOH (pH 8.5) using 30 K Amicon Ultra-0.5 mL concentrators (with a wash volume of 400 μL each). Subsequently, the sample was reduced and alkylated for 30 min in 5 mM TCEP and 10 mM 2-chloroacetamide, followed by three buffer exchanges into 8 M urea (in 50 mM HEPES-KOH, pH 8.5) as described above. Subsequently, the sample was digested with Lys-C protease (Wako Chemicals, 1:100 enzyme to protein mass ratio) for 4 h at room temperature and with trypsin (Promega, 1:100 enzyme to protein mass ratio) at 37°C overnight. An additional 4 h incubation followed the next day with freshly added trypsin and 5% acetonitrile. The digest was acidified with trifluoric acid at 1% final concentration, sonicated, and stored at −80°C.
The digested sample was either injected, loaded, desalted, and then separated on a 50 cm uPAC C18 HPLC column (Pharmafluidics) connected to a modified Digital PicoView nano-source (New Objective), or injected and trapped onto a PepMap 100 C18 2 cm trap using an EASY nLC-1000 system (Thermo Fisher), and then separated on a 15 cm EASY-Spray C18 column (ES801) connected to an EASY-Spray source (all Thermo Fisher Scientific). The following chromatography method was used: 0.1% formic acid (buffer A), 0.1% formic acid in acetonitrile (buffer B), flow rate 500 nl/min (uPAC) or 250 nl/min (EASY), gradient 240 min in total, (mobile phase compositions in % B): 0-5 min 3-7%, 5-195 min 7-22%, 195-225 min 22-80%, 225-240 min 80%. MS spectra were collected on an Orbitrap Fusion Lumos mass spectrometer (Thermo Fisher Scientific) in "MS2_MS3" mode according to Liu et al. 89 . Peptide MS1 precursor ions were measured in the Orbitrap at 120 k resolution with advanced peak determination (APD) feature enabled, and those with assigned charge states between 3 and 8 were subjected to CID-MS2 fragmentation (25% CID collision energy) and fragments detected in the Orbitrap at 30 k resolution. Data-dependent HCD-MS3 scans were performed if a unique mass difference (Δm) of 31.9721 Da was found in the CID-MS2 scans with detection in the ion trap (35% HCD collision energy).
MS raw data were analyzed in Proteome Discoverer versions 2.4 and 2.5 (Thermo Fisher Scientific) using Sequest 90 search for linear peptides, including crosslinker-modifications, and XlinkX search to identify crosslinked peptides. MS2 fragment ion spectra not indicative of the DSSO crosslink delta mass were searched with the Sequest search engine against a custom protein database containing the expected protein components, a database of proteins previously identified in the Sf9 insect cell expression system 91 , as well as a database of contaminants taken from MaxQuant, cRAP, and those commonly identified during inhouse analyses, using the target-decoy strategy 92 . The following variable crosslinker modifications were considered: DSSO Hydrolyzed/+176.014 Da (K); DSSO Tris/+279.078 Da (K), DSSO alkene fragment/+54.011 Da (K); DSSO sulfenic acid fragment/ +103.993 Da (K), as well as Oxidation/+15.995 Da (M) and phosphorylation/ +79.966 Da (S, T, Y). Carbamidomethyl/+57.021 Da (C) was set as a static modification. Trypsin was selected as the cleavage reagent, allowing a maximum of two missed cleavage sites, peptide lengths between 6 and 150, 10 ppm precursor mass tolerance, and 0.02 Da fragment mass tolerance. PSM validation was performed using the Percolator node in PD and a target FDR of 1%. Phosphosites were considered localized above a PTM score of 50 as calculated with the IMP ptmRS node in PD.
XlinkX v2.0 was used to perform a database search against custom protein database containing the expected complex components to identify DSSOcrosslinked peptides and the following variable modification: DSSO Hydrolyzed/ +176.014 Da (K); Oxidation/+15.995 Da (M) and phosphorylation/+79.966 Da (S, T, Y). Crosslink-to-spectrum matches (CSMs) were accepted above an XlinkX score cutoff of 50, requiring a minimum of 3 MS3 spectra used for crosslink identification (both peptides must be identified by MS3, at least one peptide as thiol and alkene fragment for DSSO cleavage). MS2-only crosslink identifications were not accepted. The crosslinks from two repeats at three different DSSO concentrations (0.5, 2, and 4 mM) were merged and crosslink network maps generated using the xiNET webserver 93 . For visualization, only crosslinks with a CSM of 4 or greater were included.
Reporting summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability
The data that support this study are available from the corresponding author upon reasonable request. The PDB coordinates and cryo-EM maps have been deposited into the Protein Data Bank and Electron Microscopy Data Bank under the following accession numbers: PDB 7TJF and EMD-25924 for ScORC·DNA, PDB 7TJH and EMD-25925 for ScORC·DNA·Cdc6 state I (ODC1), PDB 7TJI and EMD-25926 for ScORC·DNA·Cdc6 state II (ODC2), PDB 7TJJ and EMD-25927 for ScORC·DNA·Cdc6·Orc6-CB N state I (ODC1·Orc6-CB N ), and PDB 7TJK and EMD-25928 for ScORC·DNA·Cdc6·Orc6-CB N state II (ODC2·Orc6-CB N ). The proteomics data have been submitted to the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifier PXD031033 94 . Source data are provided with this paper.