MCM Forked Substrate Specificity Involves Dynamic Interaction with the 5′-Tail*

The archaeal minichromosome maintenance protein MCM forms a homohexameric complex that functions as the DNA replicative helicase and serves as a model system for its eukaryotic counterpart. Here, we applied single molecule fluorescence resonance energy transfer methods to probe the substrate specificity and binding mechanism of MCM from the hyperthermophilic Archaea Sulfolobus solfataricus on various DNA substrates. S. solfataricus MCM displays a binding preference for forked substrates relative to partial or full duplex substrates. Moreover, the nature of MCM binding to Y-shaped substrates is distinct in that MCM loads on the 3′-tail while interacting with the 5′-tail likely via the MCM surface. These results provide the first elucidation of a dynamic nature of interaction between a ring-shaped helicase interacting with an opposing single-stranded DNA tail. This interaction contributes to substrate selectivity and increases the stability of the forked DNA-MCM complex, with possible implications for the MCM unwinding mechanism.

The eukaryotic minichromosome maintenance MCM2-7 complex plays a vital role in replication and is thought to be the replicative helicase (1,2). MCM2-7 is involved in various stages of the replication process, from licensing through initiation, replication, and termination (3,4). The MCM2-7 complex forms a heterohexamer, with a subcomplex (MCM4-6-7) functioning as a 3Ј-5Ј-helicase. The six subunits possess sequence similarity, suggesting a common evolutionary origin (1,5). The archaeal replication process represents a simpler form of that found in the Eukarya (6) and provides an excellent model system for studying replication (7,8). Most archaeal species encode a single MCM that is homologous to its eukaryotic counterparts (9) and that forms a homohexamer (10,11) or a double homohexamer (12).
The binding, translocation, and unwinding mechanisms of archaeal and eukaryotic MCM protein, and other replicative helicases have been the focus of many research efforts (13)(14)(15). MCM4-6-7 DNA unwinding shows a requirement for Y-shaped DNA substrate (16) or a 3Ј-tailed DNA with a 5Ј-bulky attachment (17), leading to the double hexamer model and steric exclusion model, respectively. Archaeal MCM proteins also display enhanced unwinding of a forked DNA substrate (18 -20). The importance of the opposing strand in efficient DNA unwinding is also evident in 5Ј-3Ј-replicative helicases such as Escherichia coli DnaB helicase (21) and bacteriophage T7 helicase (22). Although it is clear that a forked DNA is the unwinding substrate preferred by replicative helicases, debate exists over the nature and functional role of interactions between the enzyme and the single-stranded (ssDNA) 4 and double-stranded DNA components of the substrate, and many of the subtleties regarding replicative helicase binding, unwinding, and translocation remain elusive (13,14,23).
Single molecule techniques can reveal inner details of biochemical reactions and have been used to study mechanisms of various helicases (24 -30). Herein, we report on the mechanism of DNA binding and substrate selectivity of Sulfolobus solfataricus MCM (SsoMCM) obtained through single molecule fluorescence resonance energy transfer (smFRET) studies. smFRET (31,32) is a powerful method able to unravel biological events, including helicase activities, without ensemble averaging. smFRET was applied to probe the binding and orientation of MCM on DNA as well as the conformational changes induced in DNA upon MCM loading. MCM demonstrates a specificity of binding to forked DNA over single-tailed and duplex substrates. This results from a highly dynamic interaction between the 5Ј-tail and MCM, with a mechanistic implication on replicative helicases.

EXPERIMENTAL PROCEDURES
DNA Substrates and Annealing-Oligonucleotide substrates were purchased from Integrated DNA Technologies (Coralville, IA). Forked, partial, and duplex substrates were prepared by mixing the biotinylated and non-biotinylated strands * This work was supported in part by National Institutes of Health Grant GM065367 (to T. H.) and by the Medical Research Council (to S. D. B.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. in a 1:1.5 ratio at 1 M in buffer containing 10 mM Tris (pH 8) and 200 mM NaCl. The non-biotinylated strand was added in excess to minimize the chance of having residual biotinylated strand in the measurement. The annealing mixture was heated at 95°C for 2 min and allowed to cool slowly for 2 h in the heating block. (For substrates and sequences, see supplemental "Experimental Procedures.") Protein Purification and Protein Labeling-Wild-type SsoMCM and its C642A mutant were purified as described previously (10). Cy3 and Cy5 N-hydroxysuccinimide esters and maleimides were purchased from GE Healthcare. Fluorescence labeling of SsoMCM was also performed as described previously (10). Briefly, wild-type SsoMCM was incubated with the N-hydroxysuccinimide ester in 25 mM HEPES, 150 mM NaCl, and 10 mM MgCl 2 at pH 6.8 and then quenched with 10 mM Tris after 1 h (referred to as N-MCM). C642A mutant SsoMCM was incubated with the maleimide in the same buffer at pH 7.5 and then quenched with 10 mM ␤-mercaptoethanol after 4 h (referred to as C-MCM). Labeled MCM proteins were spun through a spin column and dialyzed extensively to remove any remaining free label. The degree of labeling was determined by UV spectrometry using the appropriate extinction coefficients and found to be essentially 1 for all cases.
Single Molecule Spectroscopy-smFRET measurements were performed using a wide-field total internal reflection fluorescence microscope (33,34). Total internal reflection excitation was done either using a prism or through an objective (Olympus UPlanSApo ϫ100 numerical aperture 1.4 oil). Images were acquired with a 30-ms time resolution using an electron-multiplying charge-coupled device camera (iXon DV887-BI, Andor Technology) and a homemade Cϩϩ program. FRET values were calculated as the ratio between the acceptor intensity and the sum of the intensities of the donor and acceptor, after correcting for cross-talk between the two detection channels and subtracting the background (32,35).
Single Molecule Measurements of Labeled MCM Complex-To reduce the possibility of FRET between one-donor DNA and an MCM hexamer labeled with multiple acceptors, wild-type MCM was mixed with labeled MCM to yield a final labeled/ unlabeled monomer ratio of 1:8. Labeled and unlabeled MCM proteins were incubated in the dark for at least 1 h at room temperature.
Surface-attached DNA molecules were incubated with the reported concentrations of MCM (all given in monomer concentration) for 5 min, ensuring sufficient time for binding.
Excess and unbound MCM was removed by a wash with buffer 4, followed by a wash with imaging buffer.

RESULTS
smFRET Analysis of SsoMCM Fork Binding Orientation-SsoMCM contains several functional domains with separable domain-dependent activities. The C-terminal half contains the AAA ϩ (ATPases associated with a variety of cellular activities) domain and a helix-turn-helix motif, and the N terminus includes the Walker A and B motifs (10). Recent studies of truncated SsoMCM complexes assigned the helicase activity to the C-terminal half and showed that the N-terminal half governs substrate selectivity and enzyme processivity (36).
For smFRET measurements, Cy3 was used as the donor and Cy5 as the acceptor. For DNA-to-MCM FRET analysis, MCM was labeled with Cy5 at the N terminus (N-MCM) or C terminus (C-MCM), and the various DNA substrates were labeled with Cy3. The activity of purified and labeled MCM was verified prior to measurements as described (10). The DNA molecules were attached to the surface of a flow chamber and incubated with MCM at a specified monomer concentration, followed by an extensive wash to remove unbound MCM (see "Experimental procedures"). The orientation of MCM was measured via the binding of acceptor-labeled C-MCM or N-MCM (8 M prewash monomer concentration) to forked DNA substrates that were donor-labeled at either the 3Ј-end or the junction. The normalized relative distance distributions, extracted from smFRET efficiency histograms, are shown for four different combinations: N terminus to 3Ј-end, C terminus to 3Ј-end, C terminus to junction, and N terminus to junction (Fig. 1, A-D, respectively). The data indicate that the N terminus is closer to the 3Ј-end (ϳ5.4 nm) than to the junction (Ͼ8.2 nm). The distances were calculated using the previously determined value of ϳ6.3 nm for the Cy3-Cy5 pair (35) and should be considered only as an approximation because of the relatively large fluorescence anisotropy of the protein-and DNA-conjugated fluorophores. Additionally, the C terminus is closer to the junction (ϳ7.0 nm) than to the N terminus (Ͼ8.2 nm). In combination, our data indicate the MCM orientation illustrated in Fig. 1; MCM binds to the 3Ј-tail with the C-terminal AAA ϩ helicase domain facing the junction and the N-terminal domain, which includes the substrate selectivity domain, closer to the 3Ј-end. These findings are in agreement with the previously reported organization of MCM on DNA (10) as measured in ensemble experiments, validating our smFRET assays.
SsoMCM Substrate Binding Specificity-We next aimed to probe the substrate binding preference of MCM. The fluorescence of surface-tethered donor-labeled DNA, either forked or with a 3Ј-tail (Fig. 2B), was measured after incubation with 3 M acceptor-labeled C-MCM. Fig. 2A shows donor (panel I) and acceptor (panel II) single molecule fluorescence images of MCM bound to forked DNA (panel I) and the single 3Ј-tailed partial duplex (panel II). The fraction of DNA bound by MCM was determined from the resulting smFRET histograms (see supplemental "Experimental Procedures"). MCM displayed a nearly 2-fold enhanced binding to forked DNA substrate over the 3Ј-tailed partial duplex substrates (Fig. 2B). We considered the following three explanations at to why MCM prefers a forked DNA: 1) MCM loads on the free 3Ј-tail while externally grabbing the 5Ј-tail, possibly through surface binding modes; 2) MCM preferentially encircles both the 3Ј-and 5Ј-tails in its central channel; and 3) as proposed previously for MCM (37), two hexamers are arranged in an eccentrically stacked double hexamer and bind separately to the 3Ј-and 5Ј-tails, each with its own polarity such that binding is improved when both tails are present.
To distinguish among these binding modes, we determined the efficiency of binding of MCM to other substrates. Five different donor-labeled substrates were measured: an 18-bp full duplex, a 5Ј-tailed partial duplex, and three forked substrates labeled in different locations (Fig.  2C). Binding efficiency was low for the full duplex and 5Ј-tailed duplex substrates and was significantly higher for the forked substrates (Fig.  2C). The low binding of MCM to duplex substrates disfavors the second loading mechanism. Hence, the observed MCM preferred loading on forked substrates may be either due to external binding of the 5Ј-tail or through a double hexamer binding both strands.
To further probe the loading mechanism, we designed three forked substrates labeled with a donor and an acceptor on opposing tails (Fig. 3), enabling us to investigate the behavior of the forked tails upon MCM loading. The DNA substrates were tethered to the surface, and unlabeled MCM at the indicated concentration was added (see "Experimental Procedures"). Fig.  3A shows the smFRET efficiency histograms of a forked DNA substrate with a donor at the 3Ј-tail end and an acceptor at the 5Ј-tail end as a function of MCM concentration. DNA showed only low FRET efficiency values (Fig. 3A, upper panel) because of a large separation of flexible ssDNA. With increasing MCM concentrations, higher FRET efficiency populations emerged, indicating that the 3Ј-and 5Ј-tails are being brought closer together. This effect was also observed with two other forked substrates with similarly labeled 5Ј-tails, but with longer and internally labeled 3Ј-tails (Fig. 3, B and C). The existence of a high FRET efficiency population even at an MCM concentration as low as 150 nM is indicative of a single MCM hexamer bound on the forked substrates, not double hexamers, because previous stoichiometric analyses showed that SsoMCM assembles on a forked DNA only as a hexamer at concentrations up to 5 M (10, 18). In addition, double hexamer loading would yield low FRET because the DNA tails would be spatially separated by the proteins. These results led us to propose that the 3Ј-tail of the forked DNA substrate is accommodated in the central channel of the MCM hexamer, whereas the DNA 5Ј-tail binds externally (Fig. 3D).
MCM Fork Binding and Interactions with the Opposite Strand-We next examined the dynamic nature of MCM binding to the forked DNA. To test the stability of the DNA-MCM complex, forked DNA (Fig. 3, substrate 1)    1). No significant dissociation of MCM from the DNA was observed, ensuring that our measurements were not influenced by protein dissociation.
Individual DNA molecules bound by MCM displayed temporal fluctuations in FRET efficiency in the absence of free MCM in solution (Fig. 4A, panel I), likely representing distance changes between the two ends of the forked substrate. Such continual fluctuation in distance with no apparent regularity was observed for the majority of molecules with all three forked substrates (81, 82, and 74% for substrates 1-3, respectively, in Fig. 3) and was still detected after ATP was added (supplemental Fig. 2). Similar behavior was observed for FRET efficiency between the donor at the 5Ј-tail and the acceptor at the N terminus of MCM (Fig. 4A, panel II), indicating a dynamic interaction of the 5Ј-tail with MCM. Furthermore, such complex interaction with MCM is unique to the 5Ј-tail, as we did not observe similar fluctuations in FRET between a donor at the 3Ј-tail and an acceptor on MCM.
To further analyze this phenomenon, having no apparent regular fluctuation period or amplitude, we applied an autocorrelation analysis, in which, by self-correlating the values of a FRET efficiency time trace at different time periods, the characteristic relaxation times and behaviors can be extracted (see supplemental "Experimental Procedures"). The autocorrelation functions of individual smFRET traces were calculated and averaged over many molecules, ranging from 20 to 100 molecules for each experiment, to extract the characteristic time scales (Fig. 4B). Both the MCM-induced tail/tail interactions (Fig. 4B, panels I-III) and the 5Ј-tail/MCM interaction (panel IV) yielded average autocorrelation functions of similar shapes and times scales. The autocorrelation fitted poorly with an exponential decay function (Fig.  4B, panel I, blue curve), whereas a stretched exponential type function (Fig. 4B, red curves), modeled previously for subdiffusion phenomena (38 -40), provided a good fit. The relaxation time was in the second range but with stretch coefficients of ϳ0.5, indicating a wide range of time scales for the dynamics, ranging from 60 ms to 5 s (Fig. 4B and supplemental "Experimental Procedures"). Subdiffusion is an anomalous diffusion phenomenon in which the mean square displacement has sublinear dependence on time, in contrast to regular diffusion with linear dependence on time (38,40,41). Our analysis therefore indicates that the interaction between the 5Ј-tail and MCM surface can occur in many conformational states, each with a unique trapping and relaxation time, resulting in an interaction governed by subdiffusion kinetics.

DISCUSSION
Using smFRET, we found that MCM possesses specificity for forked DNA substrates over single-tailed and duplex substrates. Furthermore, MCM loads on the forked DNA 3Ј-tail, with its C terminus facing the junction and N terminus toward the 3Ј-end, and externally binds to the 5Ј-tail. Previous ensemble FRET analysis of SsoMCM deduced the same binding orientation; however, a recent crystal structure of another hexameric AAA ϩ helicase, papillomavirus E1, shows binding to an ssDNA with opposite polarity (42). Our smFRET result further supports the apparent difference in DNA binding polarity between SsoMCM and papillomavirus E1.
The interaction between the 3Ј-tail-bound MCM complex and the 5Ј-tail was found to be dynamic, with fluctuations occurring on wide distance and time scales. This suggests that the 5Ј-tail does not simply bind the MCM surface and fall off, but instead experiences an intricate interaction landscape. Assuming there exists at least one external binding site in each monomer, the MCM hexamer would involve at least six such binding sites, with many possible conformations of the 5Ј-tail onto these sites. We propose that the external 5Ј-tail/MCM interaction is weak per binding site, enabling the 5Ј-tail to break off and diffuse between different sites. The diffusion steps between the different binding sites will therefore depend on the 5Ј-tail conformation and its arrangement on the various sites, with additional contributing factors such as ssDNA strain and the location of MCM on the 3Ј-tail. The net result would be a shallow but rugged energy landscape with many possible local minima, which would increase the stability of MCM on the replication fork and determine the distinctive selectivity of MCM for forked substrates.
The architecture of replicative helicases at the replication fork during DNA unwinding has been the subject of much debate. Some of the various suggested models are passive unwinding, in which the helicase translocates on one ssDNA and awaits the opening of the junction to translocate forward; and several dynamic models, in which the helicase directly destabilizes the junction (3,13,14,26,43). For MCM, diverse models have been proposed, including steric exclusion, rotary pump, T-antigen, and Ploughshare (23,37,44), suggesting an interaction with either single-or double-stranded DNA. In the steric exclusion model, the opposing strand (5Ј-tail) prevents the helicase from encircling and translocating along the duplex DNA and does not play an additional role once the helicase starts unwinding the DNA. In contrast, our data show that SsoMCM dynamically interacts with the 5Ј-end, which should also be the case during duplex unwinding. Such dynamic inter-action between a ring-shaped helicase and an opposing ssDNA tail has not been elucidated previously.
In light of our findings, we propose the following "opposite strand interaction" model for MCM/replication fork association. MCM loads with the 3Ј-tail accommodated in its central channel, and the 5Ј-tail binds and diffuses on its surface. This interaction would regulate the binding specificity for forked substrates and aid in the long-term stability of the MCM-replication fork complex and replication fidelity. It should be noted that the weak interaction of MCM with the opposite strand may be beneficial in preventing tangling of ssDNA or stalled replication by ensuring that the interaction is local, whereas it is possible that a single strand binding protein (single-stranded DNA-binding protein or replication protein A) also aids in preventing the MCM surface from binding to ssDNA at a distance. Additionally, as a 3Ј-5Ј-helicase MCM will translocate along the leading strand template (3Ј-tail), its dynamic interaction with the unwound lagging strand template (5Ј-tail) may facilitate coordination of leading and lagging strand synthesis events. In this regard, it may be relevant that the DNA primasebinding archaeal GINS complex interacts with the trailing N-terminal domains of SsoMCM (45). Additionally, this type of interaction may be incorporated into the unwinding mechanism of MCM, in which upon ATP hydrolysis, MCM will translocate forward on the ssDNA while drawing the 5Ј-tail. This combined forward translocation and opposite strand binding may result in sufficient destabilization of the junction, facilitating double-stranded DNA unwinding. Finally, we should also consider the possibility that the observed interaction may have inhibitory effects on the helicase activity (46), which may be lifted if another protein such as the DNA polymerase occupies the opposing strand (47,48).
Our smFRET study has revealed unique information about dynamic interaction between MCM and the opposite strand in the context of replication fork association with MCM. These findings likely bear relevance to the mechanism of replication fork binding and unwinding by MCM and other replicative helicases.