|Home | About | Journals | Submit | Contact Us | Français|
Opening and closing of two ring-shaped Mcm2–7 DNA helicases is necessary to license eukaryotic origins of replication although the mechanisms controlling these events are unclear. The origin-recognition complex (ORC), Cdc6 and Cdt1 facilitate this process, establishing a topological link between each Mcm2–7 and origin DNA. Using colocalization single-molecule spectroscopy and single-molecule FRET (Förster resonance energy transfer), we monitored S. cerevisiae Mcm2–7 ring opening and closing during origin licensing. The two Mcm2–7 rings are open during initial DNA association and close sequentially, concomitant with release of their associated Cdt1. ATP hydrolysis by Mcm2–7 is coupled to ring closure and Cdt1 release, and failure to load the first Mcm2–7 prevents recruitment of the second Mcm2–7. Our findings identify key mechanisms controlling the Mcm2–7 DNA-entry gate during origin licensing and reveal that the two Mcm2–7 complexes are loaded by a coordinated series of events with implications for bidirectional replication initiation and quality control.
During eukaryotic DNA replication, origins of replication are licensed when two copies of the ring-shaped, heterohexameric Mcm2–7 helicase topologically encircle origin DNA1. This linkage is established when the interface between Mcm2 and Mcm5 (Mcm2-Mcm5 gate) is opened to allow DNA to enter the central channel of the helicase and then closed to prevent DNA release2,3. The two Mcm2–7 complexes are loaded sequentially. One Mcm2–7, in complex with Cdt1, is initially recruited to origin DNA bound by the origin-recognition complex (ORC) and Cdc64–6. This intermediate rapidly releases Cdc6 and then Cdt16–8. A second Cdc6 and Cdt1-Mcm2–7 subsequently associate with ORC and the first Mcm2–78,9, followed by release of Cdc6, Cdt1 and ORC8. The net result is a head-to-head Mcm2–7 double hexamer that encircles the origin DNA and is poised for bidirectional initiation4,10.
ATP binding and hydrolysis is critical for helicase loading. ATP binding is required for the initial DNA association of the three helicase-loading proteins and Mcm2–711,12. ATP hydrolysis is required to move beyond this initial association and complete Mcm2–7 loading6,13,14. ORC, Cdc6 and Mcm2–7 all bind and hydrolyze ATP. Although not required for helicase loading, ORC ATP hydrolysis is required for the repetition of this event15. Cdc6 ATP hydrolysis is also not required for helicase loading13,14,16, however, it is required for a quality control mechanism that releases incompletely loaded Mcm2–7 from DNA13,14,17. Mcm2–7 ATP hydrolysis by at least a subset of the six Mcm2–7 ATPase motifs is required for helicase loading13,14, however, it remains unclear which event(s) depends on the action of these ATPases.
Although previous studies have revealed both the order of protein associations during helicase loading and their regulation18, the timing and mechanism of the key event of Mcm2–7 ring opening and closing remains unclear. ATP binding at the Mcm2-Mcm5 interface is proposed to close the Mcm2–7 ring3 and this is supported by EM studies of ATPγS-bound Mcm2–79. In contrast, in the presence of ATP structural studies show Mcm2–7 in an open state19,20. The status of the Mcm2-Mcm5 gate in the initially recruited Cdt1-Mcm2–7 complex is unknown. The sequence and structural similarity of ORC–Cdc6 to sliding clamp loaders has led to a hypothesis that binding to ORC and Cdc6 opens the Mcm2–7 ring7 but this remains to be tested.
Using a single-molecule FRET-based approach, we have examined the timing and mechanism of Mcm2–7 ring opening and closing and its relationship to other events of origin licensing. We find that Mcm2–7 is in an open state upon initial binding and that this state is independent of Cdt1 binding. Mcm2–7 ring closure occurs independently for each Mcm2–7 at a time that is concomitant with Cdt1 release. Interestingly, we find that ATP hydrolysis by Mcm5-Mcm3 is required for ring closure and Cdt1 release. Preventing these events inhibits recruitment of the second Mcm2–7 ring. Our findings provide important insights into the mechanism of helicase loading and reveal attributes of this event that favor double-hexamer formation and quality control.
Based on the closed Mcm2–7 ring structure21, we attached donor (D) and acceptor (A) fluorophores to Mcm2 and Mcm5 at positions where FRET should increase in the closed state (Figs. 1A and S1A). This fluorescent variant (Mcm2–725FRET) functioned at near wild-type levels in bulk helicase-loading assays performed with purified proteins (Supplementary Fig. 1B–D). To measure Mcm2–7 DNA association and changes in apparent FRET efficiency (EFRET) during helicase loading, we incubated surface-attached fluorescent origin DNA with purified Mcm2–725FRET, ORC, Cdc6, and Cdt18. Co-localization of the protein- and DNA-associated fluorophores was indicative of DNA binding. Alternating excitation of acceptor (Fig. 1B, Supplementary Figs. 2 and 3A, panel i) and donor (Figs. 1B and Supplementary Figs. 2 and 3A, panels ii-iv) allowed us to monitor Mcm2–7 binding to individual DNAs and calculate EFRET for bound Mcm2–725FRET. Long-lived sequential increases in Mcm2–7-associated fluorescence revealed the first and second Mcm2–7 binding events (Fig. 1B and Supplementary Fig. 3A8). We focused on events in which simultaneous increases in both acceptor-excited and total donor-excited fluorescence (e.g., Fig. 1B i and iii, black arrows) indicated that an Mcm2–725FRET with both D and A fluorophores was binding. After initial binding, Mcm2–725FRET showed relatively high D emission and weak A emission (e.g. Fig. 1Bii, ~850 s), producing a low EFRET value. Long-lived Mcm2–725FRET molecules subsequently displayed decreased D emission and increased A emission, indicative of increased EFRET (e.g., Fig. 1B ii, ~880 s).
Analysis of a large number of Mcm2–725FRET helicase-loading trajectories revealed evidence for two major types of DNA-Mcm2–7 complexes with distinct EFRET values (Fig. 1C). Early after DNA binding (<15 s), Mcm2–725FRET was predominantly in an EFRET state of ~0.18, (Fig. 1C, Supplementary Fig. 3B and Supplementary Table 1). At intermediate times (15–75 s), we observed a mixture of EFRET ~0.18 and EFRET ~0.36 states. At longer times (>75 s), we saw almost entirely EFRET ~0.36. A similar set of distributions was observed for binding of a second Mcm2–725FRET, except that an intermediate EFRET ~0.28 value was seen at early time points (Supplementary Fig. 3C and Supplementary Table 2). This intermediate value suggests that the first Mcm2–7 remains in the EFRET ~0.36 state while the second Mcm2–7 is initially in the EFRET ~0.18 state. These distributions are consistent with a transition by both the first and second Mcm2–7 complexes from an open Mcm2-Mcm5 gate (EFRET ~0.18) to a closed Mcm2-Mcm5 gate (EFRET ~0.36). A similar transition between discrete low and high EFRET states during helicase loading was observed for an alternative Mcm2–7 construct in which Mcm2 was fluorescently labeled at a different location (Supplementary Fig. 3D), consistent with the idea that EFRET increase is caused by a conformational change that reduces the distance between Mcm2 and Mcm5.
In support of the higher EFRET state representing a closed Mcm2-Mcm5 gate, Mcm2–725FRET DNA complexes with time-averaged EFRET values greater than 0.25 correlated with long-lived (>100 s) DNA associations (Fig. 1D). Consistent with higher EFRET being caused by changes within an individual Mcm2–7, elevated EFRET was seen only when D and A were on the same Mcm2–7 hexamer (Supplementary Fig. 2A). Furthermore, once molecules reached the EFRET ~0.36 state, no persistent excursions at or below EFRET = 0.18 were observed (>5 s; N=0/57), consistent with a stably closed Mcm2–7 ring4. These findings combined with subsequent observations lead us to conclude that the EFRET ~0.18 and ~0.36 states of individual Mcm2–725FRET complexes represent the open and closed conformation of the Mcm2-Mcm5 gate, respectively.
Because Cdc6 and Cdt1 sequentially dissociate from the DNA after facilitating initial Mcm2–7 binding8, we asked if either of these events correlated with Mcm2-Mcm5 gate closure. For the first Mcm2–7 association, we compared the times of gate closure, as assessed by attainment of the EFRET ~0.36 state, with previously determined distributions of first Cdc6 and Cdt1 release times (Fig. 2A and reference8). In each case, these times were measured relative to initial Mcm2–7 DNA association. The average time for Cdc6 release was much shorter than the average gate-closure time. In contrast, the distributions of times for Cdt1 release and gate-closure were similar, supporting a connection between these events. Because Cdt1 release is slower for the second Mcm2–7 than the first8, we asked if Mcm2-Mcm5 gate closure for the second Mcm2–7 was similarly delayed. Indeed, the times of gate closure after arrival of the second Mcm2–7 showed a similar distribution to the times of the second Cdt1 release (Fig. 2B). Thus, for both the first and second Mcm2–7, ring closure was concomitant with Cdt1 release.
For the comparisons between the previously determined Cdt1 release times and Mcm2–725FRET ring closure times to be valid, the kinetics of the Mcm2–725FRET loading reaction should be similar to that of the singly-modified Mcm2–7 (Mcm2–74SNAP, Mcm2–7 with SNAP at the N-terminus of Mcm4) used in the previous determination of Cdt1 release times8. To test for this, we made Mcm2–725FRET*, a preparation in which only Mcm5-SNAP is attached to a fluorophore but Mcm2-CLIP is still present in the complex. Whit only a single fluorophore on Mcm2–725FRET*, we could simultaneously measure the DNA association and dissociation of Mcm2–725FRET* and a second protein labeled with a second fluorophore. Although helicase loading was inhibited when Mcm2–725FRET* was combined with labeled Cdt1, labeled ORC and Cdc6 were compatible with Mcm2–725FRET*. Importantly, we observed similar release times of Cdc6 and ORC whether we used Mcm2–725FRET* or Mcm2–74SNAP (Supplementary Fig. 4A–C). Although we could not measure the kinetics of Cdt1 in the presence of Mcm2–725FRET, we note that the times of Cdc6 and ORC release encompass the times of Cdt1 release. Thus, the kinetics of the helicase loading reaction is not dramatically changed by the SNAP and CLIP proteins inserted into Mcm5 and Mcm2 present in Mcm2–725FRET.
The similar kinetics of Cdt1 release and Mcm2–7 ring closure suggested that Cdt1 binding to Mcm2–7 holds the Mcm2-Mcm5 gate open. To address this possibility, we monitored EFRET for Mcm2–725FRET in the absence of DNA, ORC and Cdc6. Whether Mcm2–725FRET was directly tethered to the slide in the absence of Cdt1 or tethered to the slide indirectly via Cdt1, the Mcm2-Mcm5 gate was predominantly in an open (EFRET ~0.16) state with smaller populations in higher EFRET states (Fig. 2C and Supplementary Table 3). Interestingly, the higher EFRET populations were reduced for Cdt1-bound Mcm2–7 (Supplementary Table 3), suggesting Cdt1 increases the already high bias of Mcm2–7 toward an open (EFRET ~0.16) state. Consistent with Cdt1-Mcm2–7 being in an largely open state in contrast to the closed state of loaded Mcm2–7, we measured solution EFRET values for free Cdt1–Mcm2–725FRET and loaded Mcm2–725FRET. Although the absolute values are different from those seen in the single-molecule experiments due to incomplete protein labeling (only doubly-labeled proteins were assessed in the single-molecule setting), we observed higher solution EFRET values for loaded Mcm2–725FRET (0.129 ± 0.004) than for free Cdt1–Mcm2–725FRET (0.076 ± 0.002). These findings are consistent with previous low-resolution structural studies showing that free Mcm2–7 has an open Mcm2-Mcm5 gate19,20. We conclude that Cdt1 is not required to prevent Mcm2–7 ring closure, but Cdt1-bound Mcm2–7 may more strongly favor the open state.
To simultaneously monitor in three-color experiments the status of the Mcm2-Mcm5 gate and release of fluorescently-labeled proteins in three-color experiments from individual DNAs, we labeled Mcm2 and Mcm5 with a fluorophore and a quencher, respectively (Mcm2–725quench, Fig. 3A). Bulk assays showed that Mcm2–725quench retains ~50% of wild-type helicase-loading activity (Supplementary Fig. 1C and 1D). Consistent with our Mcm2–725FRET studies, this complex showed high fluorescence upon initial DNA binding (gate open) and reduced fluorescence thereafter (gate closed, Fig. 3B and Supplementary Fig. 4D). Experiments combining Mcm2–725quench with differentially labeled Mcm2–7 (Mcm2–7JF646), showed that the Mcm2-Mcm5 gate of the first Mcm2–7 did not reopen once closed, including during loading of the second Mcm2–7 (Fig. 3C).
As with Mcm2–725FRET, labeled Cdc6 and ORC were compatible with Mcm2–725quench but fluorescently-labeled Cdt1 inhibited helicase loading when combined with this form of Mcm2–7. Consistent with Cdc6 being released before Cdt18, labeled Cdc6 was always released prior to Mcm2–725quench gate closure (Fig. 3D, 62/62 events). Combining labeled ORC with Mcm2–725quench revealed a connection between ORC release and closing of the second Mcm2–7 ring. In the majority of events (47/54), the single ORC involved in helicase loading8 was released at a time within experimental error of gate closure of the second Mcm2–7 (Fig. 3E). In the remaining events, ORC was retained on the DNA after second ring closure. Interestingly, relative to association of the second Mcm2–7, the average time until second Mcm2–7 ring closure was much longer than the previously determined average time until establishment of Mcm2–7-Mcm2–7 double-hexamer interactions (as measured by FRET between the N-termini of the first and second Mcm2–78, Supplementary Fig. 4E). Thus, an open second Mcm2–7 ring forms initial double-hexamer interactions with a closed first Mcm2–7 (Fig 3C), raising the possibility that the closed first Mcm2–7 ring could act as a template to facilitate closing of the second Mcm2–7. In combination, our data strongly suggest that ORC release only occurs after Mcm2–7 double-hexamer formation and closure of both Mcm2–7 rings, a mechanism that would ensure that ORC is retained until the completion of origin licensing.
To further investigate the mechanism of Mcm2-Mcm5 gate closing, we asked if the ATPase activities of Mcm2–7, Cdc6, or ORC control this event. Based on the temporal connection between gate closure and Cdt1 release, we focused on a mutation in the Mcm5-Mcm3 ATPase active site (mcm5-R549A, Supplementary Fig. 5A) that is defective for Cdt1 release and Mcm2–7 loading13,14. We incorporated this mutant into Mcm2–725FRET and monitored Mcm2-Mcm5 gate status. Strikingly, Mcm2–725FRET-5RA remains in an open-gate (EFRET ~0.18) state indefinitely after DNA association (Fig. 4A and Supplementary Fig. 5B). In contrast, using Cdc622 or ORC15 ATPase mutants did not prevent Mcm2-Mcm5 gate closure (Fig. 4B). The kinetics of Cdc6 release were unchanged by Mcm2–75RA (Fig. 4C). In contrast, the dwell time of Cdt1 associated with Mcm2–75RA was dramatically extended relative to wild-type Mcm2–7 (Fig. 4D). In most cases (87/109), Cdt1-DNA association was as long as Mcm2–75RA association, including many long-lived associations that ended with the simultaneous release of Mcm2–7 and Cdt1 (e.g., Supplementary Fig. 5C), as expected if the lack of Cdt1 release prevented ring closure. Interestingly, we did not observe any second Mcm2–7 associations (0/109) for the Mcm2–75RA mutant, suggesting release of Cdt1 and/or ring closure must be completed prior to second Mcm2–7 recruitment.
Our results support the initial conclusion that the EFRET ~0.18 and ~0.36 states of Mcm2–725FRET represent the open and closed Mcm2-Mcm5 gate. The Mcm2–7 ring was in the EFRET ~0.18 state before and immediately after DNA binding, consistent with an open Mcm2–7 ring allowing DNA access to the central channel. Similarly, all Mcm2–725FRET-5RA DNA associations remained in the open (EFRET ~0.18) state and were released by a high-salt wash that removes incompletely loaded Mcm2–7 (32/32 events12,23). Consistent with this conclusion, recent high-resolution cryo-EM structural studies of Mcm2–7 and Cdt1-Mcm2–7 found that both complexes are in an open-ring conformation24. The Mcm2–7 but not the ORC or Cdc6 ATPases are required for helicase loading13,14. Consistent with the higher EFRET state reflecting a loaded, closed-ring Mcm2–7, the Mcm5-Mcm3 ATPase mutant, but not mutations in ORC or Cdc6 ATPases, prevented formation of this state. Future studies will be required to determine if other Mcm2–7 ATPase mutants have the same effect. Finally, attainment of the EFRET ~0.36 state occurred independently for each Mcm2–7 complex, consistent with evidence that they are loaded one at a time 6–9. Although structural studies of the loaded double-hexamer suggest show a completely closed Mcm2–7 ring21, our findings do not exclude the possibility that the closed state we observe by FRET is sufficiently open to allow ssDNA to escape the loaded double-hexamer.
In addition to revealing the times of Mcm2–7 ring closure during helicase loading, the concomitant release of Cdt1 and closure of the Mcm2-Mcm5 gate and the inhibition of both events by the Mcm5-Mcm3 ATPase mutant support a model in which these events are causally linked (Fig. 5). We propose that the positively charged Mcm2–7 central channel and Cdt1 binding (Fig. 2C, Supplementary Table 3) favor an open conformation of the Mcm2–7 ring off the DNA. ORC–Cdc6 recruit an open Cdt1-Mcm2–7 ring such that it encircles DNA7, similar to recent studies of archaeal Mcm loading25. Although Cdt1 binding was not required to maintain an open Mcm2–7 off DNA, we propose that Cdt1 holds the Mcm2–7 ring open after negatively-charged DNA binds to the positively-charged Mcm2–7 central channel. Finally, after Cdc6 is released, we propose that Mcm5-Mcm3 ATP hydrolysis (and perhaps other Mcm2–7 ATPases) stimulates Cdt1 release triggering Mcm2-Mcm5 gate closure. It is also possible that Mcm5-Mcm3 ATP hydrolysis directly stimulates ring closure which causes Cdt1 release. Although ORC–Cdc6 has been proposed to function like a sliding-clamp loader during helicase loading7, of the known sliding clamp functions26, it appears this complex only retains the function of recruiting a protein-ring to the DNA. ORC and Cdc6 are not required to open the Mcm2–7 ring (Fig. 1 and and3)3) and ATP hydrolysis by ORC or Cdc6 is not required for ring closure (Fig. 4B). This does not eliminate other possible roles for ORC–Cdc6 including stimulating Mcm2–7 ATP hydrolysis6 or altering Mcm2–7 conformation to facilitate ring closure7.
The ordered release of Cdc6 and Cdt1 and the connection of the latter event to ring closure creates a window of time for Mcm2–7 loading quality control13,14,17. Cdc6 ATP hydrolysis is connected to the release non-productive Mcm2–7 complexes13,14. Because the Mcm2–7 ring is open throughout Cdc6 DNA association (Fig. 3D), this quality control mechanism would not require reopening of the Mcm2–7 ring. In addition, the ordered closure of rings would allow the first and second Mcm2–7 complexes to be assessed separately. Although the mechanism of this release is unclear, one simple hypothesis is that an ATP-dependent release of Cdc6 prior to Mcm2–7 ring closure leads to the simultaneous release of open non-productive Mcm2–7 complexes.
Our findings indicate that loading of the two Mcm2–7 complexes associated with origin licensing is the result of a single coordinated event rather than two independent Mcm2–7 loading events. Both the lack of second Mcm2–7 association for the Mcm2–75RA mutant and the finding that gate closure by the first Mcm2–7 always preceded DNA association of a second Mcm2–7 (Fig. 1B and S2A, 47/47 events), strongly suggest that recruitment of the second Mcm2–7 requires completion of the first loading event. This is inconsistent with models suggesting that two ORC molecules independently recruit and load one Mcm2–7. The connection between ORC release and closure of the second but not the first Mcm2–7 ring (Fig. 3E) also supports a coordinated mechanism. Importantly, these properties would ensure single Mcm2–7 loading events only occur as the first step in forming a Mcm2–7 double-hexamer.
The combination of fully reconstituted biochemical assays27 and detailed structural models of key replication intermediates7,9,21 has provided important insights into the events of eukaryotic replication initiation. Single-molecule studies complement these approaches by revealing reaction kinetics that are difficult to assay in asynchronous bulk reactions, identifying intermediates that are too short-lived or dynamic to analyze structurally and by monitoring changes in protein conformation in real time. Our findings show how the combination of single-molecule colocalization and single-molecule FRET can elucidate the complex and coordinated protein dynamics of helicase-loading events. More importantly, our findings reveal features of origin licensing that can reduce incomplete or incorrect events and, therefore, improve genome stability.
Cdc6SORT549 (pET-GSS-Cdc6), Cdt1SORT549–Mcm2–74SNAP (yST166) and ORC1SORT549 (yST163) were purified as described previously8. To monitor the Mcm2-Mcm5 gate, Cdt1–Mcm2–725FRET expressing strains were constructed by introducing an Asc I site after amino acid 721 of Mcm2 and amino acid 591 of Mcm52. A SNAP- (Mcm5, NEB) or CLIP-tag (Mcm2, NEB) was inserted with 10 amino acid linkers (GGSGGSGGSG) at each junction. To purify Mcm2–725FRET, Mcm2–721CLIP and Mcm5–591SNAP were expressed in conjunction with the remaining wild-type Mcm2–7 subunits and Cdt1 (yST229) or in the absence of Cdt1 (yST266, to make Mcm2–725FRET-biotin) and labeled with CLIP-Surface™ 647(NEB) and SNAP-Surface™ 549 (NEB) as described below. To monitor gate closure by quenching (Mcm2–725quench) or to create an alternate Mcm2-Mcm5 gate FRET pair (Mcm2–72C5FRET), Mcm2SORT (Mcm2 with LPETGG at its C-terminus) and Mcm5–591SNAP were co-expressed with the remaining wild type Mcm2–7 subunits and Cdt1 (yST220). Sortase was used to attach Mcm2SORT to the peptide NH2-GGGHH HHHHH HHHC-COOH coupled to maleimide-Dy549 and Mcm5–591SNAP was coupled to SNAP-BHQ-2 to form Mcm2–725quench (see below) or SNAP- Surface™ 649 to form Mcm2–72C5FRET. Mcm2-Mcm5 gate FRET was monitored in the context of the Mcm5-R549A mutant protein by incorporating the mutation into Mcm5 subunit of the Cdt1–Mcm2–725FRET expressing strain (yST299). The resulting mutant Cdt1-Mcm2–7 was labeled as described for Mcm2–725FRET to form Mcm2–725FRET-mcm5RA. The effect of Mcm5-R549A on Cdt1 release was monitored by purifying Cdt1-Mcm2–7 from a strain expressing Mcm5-R549A, Mcm4SNAP, Cdt1SORT and the remaining wild-type Mcm subunits (yST291). Mcm4SNAP was labeled with SNAP-JF646 (gift of Luke Lavis, Janelia Research Campus) and Cdt1SORT was coupled to the peptide NH2-GGGHH HHHHH HHHC-COOH coupled to maleimide-Dy549 to make Cdt1SORT549–Mcm2–74SNAP-mcm5RA. The effect of Mcm5-R549A on Cdc6 release was monitored by purifying Cdt1–Mcm2–7 from a strain expressing Mcm5-R549A, Mcm4SNAP, Cdt1 and the remaining wild-type Mcm subunits (yST289) to make Cdt1–Mcm2–74SNAP-mcm5RA.
S. cerevisiae (W303 background) strains yST229, yST220, yST299 or yST291 were grown to OD600 = 1.2 in 8 liters of YEP supplemented with 2% glycerol (v/v) at 30°C. Addition of 2% galactose (w/v) and α-factor (100 ng/mL) induced Cdt1-Mcm2–7 expression and arrested cells at G1. After 6 hours cells were harvested and sequentially washed with 50 ml of ice-cold MilliQ water with 0.2 mM PMSF followed by 150 ml buffer A (50 mM HEPES-KOH pH [7.6], 5 mM MgOAc, 1 mM ZnOAc, 2 mM ATP, 1 mM DTT, 10% glycerol, 0.02% NP-40) supplemented with 0.1 mM EDTA, 0.1 mM EGTA, 0.75 M potassium glutamate (KGlu) and 0.8 M Sorbitol. The washed pellet was resuspended in approximately 1/3 of packed cell volume of buffer A containing 0.1 mM EDTA, 0.1 mM EGTA, 0.75 M KGlu, 0.8 M Sorbitol, Complete Protease Inhibitor Cocktail Tablet (1 tablet per 15 mL total volume; Roche) and frozen dropwise in liquid nitrogen. Frozen cells were lysed in a SamplePrep freezermill (SPEX) and the lysate was clarified by ultracentrifugation in Type 70 Ti rotor at 45 krpm for 90 min at 4°C. The supernatant was applied to 2 ml anti-M2 FLAG resin (Sigma) pre-equilibrated in buffer A containing 0.1 mM EDTA, 0.1 mM EGTA and 0.75 M KGlu and incubated with rotation for 3 hours at 4°C. The resin was collected on a column and the flow-through was discarded. The resin was washed with 20 ml of buffer A with 0.3 M KGlu. Cdt1-Mcm2–7 was eluted with buffer A containing 0.3 M KGlu and 0.15 mg/mL 3xFLAG peptide. Peak fractions containing Cdt1-Mcm2–7 were pooled, and the protein was concentrated to ~ 1 mg/mL using a Vivaspin 6 centrifugal concentrator (molecular weight cutoff = 100 kDa, Sartorius) and aliquoted into 0.8 mL fractions. Starting with 8 L of cells, the yield is typically 2 mg of 95% pure Sort-Cdt1–Mcm2–7, according to SDS-PAGE.
SNAP or CLIP-tagged Cdt1-Mcm2–7 (Cdt1SORT–Mcm2–74SNAP, Cdt1–Mcm2–72C5FRET or Cdt1–Mcm2–725FRET) was labeled with SNAP-Surface™ 549 (NEB; Dy549), SNAP-BHQ-2, SNAP-JF646 (gift of Luke Lavis, Janelia Research Campus) or CLIP-Surface™ 647 by incubating with 1 nmol of dye at room temperature for 1hr. To make Mcm2–725FRET-biotin, SNAP-549-biotin was substituted for SNAP-Surface™ 549. For SORT-tagged Cdt1-Mcm2–7 (Cdt1Sort –Mcm2–74SNAP, Cdt1–Mcm2–72C5FRET or Cdt1–Mcm2–725quench), 1 mg of Cdt1-Mcm2–7 was incubated with equimolar amount of Srt5° evolved sortase28, purification described below) and CaCl2 was added to a final concentration of 5 mM in buffer A with 0.3 M KGlu. This was mixed with 100 nmol of peptide carrying a Sort-tag and labeled with Dy549 (Dyomics), dissolved in 200 µL of buffer A with 0.3 M KGlu (sequence and fluorescent labeling of the peptide are described below). The reaction was incubated at room temperature for 15 min, and then quenched with 20 mM EDTA. The net result of the sortase reaction is coupling of the fluorescently-labeled (or biotinylated) peptide to the N-terminus of Cdt1 with the sequence NH2-CHHHHHHHHHHLPETGGG followed by the remainder of the protein or to the C-terminus of Mcm2 with the sequence LPETGGGHHHHHHHHHHC-COOH.
For SNAP or CLIP-tagged Cdt1-Mcm2–7, after coupling the proteins to fluorophore(s), the reaction was applied to a Superdex 200 10/300 gel filtration column equilibrated in buffer A with 0.1 mM EDTA, 0.1 mM EGTA, and 0.3 M KGlu. Peak fractions containing Cdt1-Mcm2–7 were pooled, aliquoted and stored at −80°C.
For SORT-tagged Cdt1-Mcm2–7, after dye-coupling, the reaction was applied to a Superdex 200 10/300 gel filtration column equilibrated in buffer A with 0.1 mM EDTA, 0.1 mM EGTA, 0.3 M KGlu, and 10 mM imidazole. Peak fractions containing peptide-coupled Cdt1-Mcm2–7 were pooled and incubated with 0.5 mL of cOmplete His-Tag Purification Resin (Roche) pre-equilibrated in buffer A with 0.1 mM EDTA, 0.1 mM EGTA, 0.3 M KGlu, 10 mM imidazole, for 1 hour with rotation at 4°C. The flow-through was discarded and the resin was washed with 5 ml buffer A with 0.1 mM EDTA, 0.1 mM EGTA, 0.3 M KGlu and 10 mM imidazole. Peptide-coupled Cdt1-Mcm2–7 was eluted using buffer A with 0.1 mM EDTA, 0.1 mM EGTA, 0.3 M KGlu and 0.3 M imidazole. Peak fractions were pooled, aliquoted, and stored at −80°C.
Special note on handling of fluorescent dyes: light sources on all chromatography apparatuses (AKTA FPLC, HPLC) were turned off during preparative runs. Fractions containing fluorescently-labeled peptides and proteins were determined during previous analytical runs.
Fluorescently-labeling of peptides for Sortase coupling, as well as purification of the Sortase A pentamutant enzyme and Ulp1 were performed as reported previously8.
To determine the labeling efficiency of the SNAP and CLIP tag labeling approaches in the MCM2–725FRET context, we purified and labeled Mcm2–725FRET with SNAP-Surface™ 549 and CLIP-Surface™ 647 on the Mcm5 and Mcm2 subunits, respectively. We imaged a standard reaction containing 0.25nM ORC, 1nM Cdc6 and 2.5nM Cdt1–Mcm2–725FRET using the described protocol and monitored colocalization of Mcm2–725FRET fluorescence with DNA fluorescence (to ensure that we were monitoring fully assembled complexes). Each colocalization was scored as exhibiting both D and A fluorescence, only D, or only A. By assuming that the labeling reactions of the SNAP and CLIP tags in Mcm2–725FRET were independent, we calculated from the observed D and A colocalization frequencies that SNAP labeling efficiency was ~74%, and CLIP labeling efficiency was ~81%, yielding ~60% of Mcm2–725FRET complexes with both D and A fluorophores.
Commercially available compounds were used without further purification. Reaction yields were not optimized. Reversed-phase high-performance liquid chromatography (HPLC) was performed on Agilent LC-MS Single Quad System 1200 Series (analytical) and Agilent 1100 Preparative-scale Purification System (semi-preparative). Analytical HPLC was performed on Waters Atlantis T3 C18 column (2.1 × 150 mm, 5 µm particle size) at a flow rate of 0.5 mL/min with a binary gradient from Phase A (0.1 M triethyl ammonium bicarbonate (TEAB) or 0.1% trifluoroacetic acid (TFA) in water) to Phase B (acetonitrile) and monitored by absorbance at 280 nm. Semi-preparative HPLC was performed on VYDAC 218TP series C18 polymeric reversed-phase column (22 × 250 mm, 10 µm particle size) at a flow rate of 20 mL/min. Mass spectra were recorded by electrospray ionization (ESI) on an Agilent 6120 Quadrupole LC-MS or on an Agilent 6210 Time-of-Flight (TOF) or on a Thermo Scientific QExactive system.
SNAP-BHQ2 (BG-BHQ2, Supplementary Fig. 6A) was prepared by reacting the building block BG-NH2 (NEB) with commercially available BHQ-2 Succinimidyl ester (LGC Biosearch) as described (23). BHQ-2 Succinimidyl ester (2.5 mg, 4.1 µmol) was dissolved in anhydrous DMF (1.0 mL). BG-NH2 (1.1 mg, 4.1 µmol) and triethylamine (0.56 µL, 4.1 µmol) were added and the reaction mixture stirred overnight at room temperature. The solvent was removed under vacuum and the product purified by reversed-phase HPLC using 0.1 M TEAB/acetonitrile gradient (yield = 21%). BG-BHQ2: ESI-MS m/z 759.3104 [M+H]+ (calc. for C38H38N12O6, m/z 759.3110).
The bifunctional SNAP-549-Biotin (BG-549-biotin, Supplementary Fig. 6B) substrate was prepared by successive couplings of commercially available α-N-Fmoc-ε-N-Dde-lysine (Merck KGaA) with BG-NH2 (NEB), N-(+)-biotin-6-aminocaproic acid N-succinimidyl ester (Sigma-Aldrich) and DY-549 acid (Dyomics) according to synthetic route described by Kindermann et al29 and Smith et al30. SNAP-549-Biotin was synthesized as follows: BG-NH2 (250.0 mg, 0.92 mmol) was dissolved in anhydrous DMF (8 mL). HBTU (N,N,N′,N′-Tetramethyl-O-(1H–benzotriazol-1-yl)uronium hexafluorophosphate) (368.0 mg, 0.97 mmol), triethylamine (135 µL, 0.97 mmol), and Fmoc-Lys(Dde)-OH (515.5 mg, 0.97 mmol) were added and the reaction mixture stirred overnight at room temperature. The reaction mixture was poured onto water (80 mL). The white solid was collected by filtration, washed twice with water, and dried in desiccator under vacuum overnight (yield = 91%). BG-Lys(Dde)-Fmoc (8.0 mg, 10.2 µmol) was dissolved in anhydrous in DMF (1 mL). Et2NH (3.2 µL, 30.9 µmol) was added and the reaction mixture stirred overnight at room temperature. The solvent was removed under vacuum. Crude BG-Lys(Dde)-NH2 was dissolved in anhydrous DMF (1 mL). N-(+)-biotin-6-aminocaproic acid NHS (2.9 mg, 6.4 µmol) and triethylamine (1.0 µL, 7.0 mmol) were added and the reaction mixture stirred 1 h at room temperature. Reaction completion was monitored by LC-MS. A 2% solution of hydrazine in DMF (0.5 mL) was added and the reaction mixture stirred for 1 h at room temperature. The solvent was removed under vacuum and the product purified by reversed-phase HPLC using 0.1% TFA in water/acetonitrile gradient (yield = 76%). BG-Lys(NH2)-Biotin: ESI-TOFMS m/z 738.3 [M+H]+ (calcd. for C35H51N11O5S, m/z 738.4). BG-Lys(NH2)-Biotin·TFA salt (2.3 mg, 2.7 µmol) was dissolved in anhydrous DMF (1 mL). DY-549 acid (2.7 mg, 3.0 µmol), HBTU (1.2 mg, 3.0 µmol) and triethylamine (0.6 µL, 4.5 µmol) were added and the reaction mixture stirred 1 h at room temperature. The solvent was removed under vacuum and the product purified by reversed-phase HPLC using 0.1 M TEAB/acetonitrile gradient (yield = 78%). BG-549-PEG-Biotin: ESI-TOFMS m/z 767.7552 [M-2H]2− (calc. for C68H91N13O18S5, m/z 767.7532).
The micro-mirror total internal reflection (TIR) microscope used for multiwavelength single-molecule using excitation wavelengths 488, 532, and 633 nm has been previously described31,32. Biotinylated AlexFluor488-labeled 1.3kb-long DNA molecules containing an origin were coupled to the surface of a reaction chamber through streptavidin. Briefly, the chamber surface was cleaned and derivatized using a 200:1 ratio of silane-NHS-PEG and silane-NHS-PEG-biotin8. We identified DNA molecule locations by acquiring 4–7 images with 488 nm excitation at the beginning of the experiment. Unless otherwise noted, helicase loading reactions contained 0.5nM ORC, 2nM Cdc6 and 5nM Cdt1–Mcm2–7. Reaction buffer was as previously described13 except without any glycerol and with the addition of 2 mM dithiothreitol, 2 mg/ml bovine serum albumin (EMD Chemicals; La Jolla, CA), and an oxygen scavenging system (glucose oxidase and catalase) to minimize photobleaching32. After addition of protein to the DNA-coupled chamber, frames of one second duration were acquired. DNA was imaged before and immediately after adding the reaction to the slide but not throughout the experiment. The imaging protocol alternated between 1 s frames with the 532 laser on and 1 s frames with the 633 laser on over 20–30 minutes. Apparent EFRET was calculated as described33.
Because the events observed on each DNA molecule represent a independent measurement of the events being studied, all of the analyses evaluate many biological replicates.
Tethering experiments were done using Mcm2–725FRET-biotin in the absence of Cdt1 or Cdt1-biotin with Mcm2–725FRET purified in the absence of Cdt1. Complexes were added to the slide at a concentration of 0.04nM and coupling was briefly visualized. Free complexes were washed out using H-300mM KGlut, and imaged for 2 minutes by alternating between 1s frames with the 532 laser on, and 1s frames with the 633 laser on. Only complexes containing both 549 and 647 dyes were used for background subtraction and EFRET calculations.
Analysis of the CoSMoS data sets was similar to34. Specifically, we typically followed these four steps35: (1) defining the spatial relationship between the two images created at different excitation and emission wavelengths from the single field of view by the dual-view optical system (“mapping”), (2) correcting the data set for stage drift that occurred during the experiment (“drift correction”), (3) imaging the label on origin-DNA to identify the locations of single DNA molecules on the surface, and (4) integration of fluorescence emission from small regions centered at the pre-defined locations of coupled DNA locations in each acquired image to obtain plots of fluorescence intensity vs. time. These steps were carried out using custom image-processing software (https://github.com/gelles-brandeis/CoSMoS_Analysis) implemented in MATLAB (The Mathworks, Natick, MA). Confidence intervals for kinetic data were determined by bootstrapping.
Both the dual imaging optics and chromatic aberrations result in spatial displacement between fluorescent spot images of co-localized species that are labeled with different color dyes. Accurate co-localization of the differentially-labeled species therefore requires use of a mapping procedure. For each pair of colors a list of several hundred reference spot pairs were collected using a sample containing a surface-tethered oligonucleotide that was labeled with Alexa488, Cy3 and Cy5. Mapping the coordinates of a fluorescent spot to the equivalent location at a different color was performed using a transformation with fit parameters based on just the 15 nearest reference spots35.Drift correction and spot-detection were carried out as described in35. Fluorescence emission from labeled complexes was integrated over a 0.37 µm2 area centered at each drift corrected origin-DNA location, yielding for each DNA molecule a separate intensity time course for each color of fluorescent label being observed.
Images containing spots that were analyzed to produce a FRET time course were first mapped and drift-corrected (see above). By alternating between their laser excitation wavelengths we monitored the co-localization of donor and acceptor-labeled Mcm2–7 hexamers with the origin-DNA molecule. To determine the time until formation of the EFRET ~0.36 state, we noted the earliest time point at which the EFRET values increased by either 0.15 for the 1st Mcm2–7 or 0.1 for the 2nd Mcm2–7. Only Mcm2–7 molecules that were labeled with both fluorophores were used for analysis of the first Mcm2–7. For analysis of the second Mcm2–7, both the first and the second Mcm2–7 molecules had to be labeled with both fluorophores.
To calculate apparent FRET efficiencies, the baseline for each fluorescence intensity trace was first subjected to a low pass filter. That smoothed baseline was then subtracted from the starting trace, resulting in a fluorescence time record with a zero mean baseline31 (e.g. Figure 1B, panels i and ii). Apparent FRET efficiency was calculated using where IAcceptor and IDonor are the acceptor and donor emission intensities observed during donor excitation, respectively. No gamma correction was applied because no systematic change in (IAcceptor + IDonor) was observed upon changes in EFRET (e.g., Fig. 1B, S2A) or upon acceptor photobleaching.
Equench was calculated on baseline-corrected data (of single Mcm2–7 molecules) as described in Supplementary Fig. 4D.
To generate sufficient loaded Mcm2–725FRET, a large-scale helicase-loading reaction was performed with 20 pmoles of bead-attached, origin DNA 80 pmoles of Mcm2–725FRET, 20 pmoles of ORC and 40 pmoles of Cdc613. After a 20 minute incubation, the DNA beads were washed with a high-salt buffer (to remove incompletely loaded protein) and loaded Mcm2–725FRET was released from beads by DNAse I treatment as previously described13. The released, loaded Mcm2–725FRET or a similar concentration of unloaded solution Cdt1–Mcm2–725FRET were placed in a cuvette and excited at 549 nm. Flourescence emission was detected from 560 nm to 690 nm and the peak values of donor (574 nm) and acceptor (670 nm) emission were used to determine EFRET values. Reported uncertainties are the standard deviations of four separate experiments.
We are grateful to the members of the Bell laboratory for useful discussions, Bob Sauer for comments on the manuscript, and Luke D. Lavis (Janelia Research Campus) for providing Janelia Fluorophores. This work was supported by NIH grants R01 GM52339 (S.P.B.) and R01 GM81648 (J.G.) and a grant from the G. Harold and Leila Y. Matthews Foundation (J.G.). S.T. was supported in part by an NIH Pre-Doctoral Training Grant (GM007287). S.P.B. is an investigator with the Howard Hughes Medical Institute. This work was supported in part by the Koch Institute Support Grant P30-CA14051 from the NCI. We thank the Koch Institute Swanson Biotechnology Center for technical support, specifically the Biopolymers and Genomics cores.
The data that support the findings of this study are available from the corresponding authors upon reasonable request.
Author contributionsS.T. performed all experiments with feedback from J.G., L.J.F. and S.P.B. except ensemble FRET studies which were performed by K.C. I.R.C. prepared essential reagents. S.T., L.J.F. and J.G. analyzed the data. S.P.B. composed the paper with input from all authors. S.P.B. and J.G. directed the project.
Competing Financial Interests Statement
The authors declare no competing financial interests.