Differences in the Single-stranded DNA Binding Activities of MCM2-7 and MCM467

The MCM2-7 complex, a hexamer containing six distinct and essential subunits, is postulated to be the eukaryotic replicative DNA helicase. Although all six subunits function at the replication fork, only a specific subcomplex consisting of the MCM4, 6, and 7 subunits (MCM467) and not the MCM2-7 complex exhibits DNA helicase activity in vitro. To understand why MCM2-7 lacks helicase activity and to address the possible function of the MCM2, 3, and 5 subunits, we have compared the biochemical properties of the Saccharomyces cerevisiae MCM2-7 and MCM467 complexes. We demonstrate that both complexes are toroidal and possess a similar ATP-dependent single-stranded DNA (ssDNA) binding activity, indicating that the lack of helicase activity by MCM2-7 is not due to ineffective ssDNA binding. We identify two important differences between them. MCM467 binds dsDNA better than MCM2-7. In addition, we find that the rate of MCM2-7/ssDNA association is slow compared with MCM467; the association rate can be dramatically increased either by preincubation with ATP or by inclusion of mutations that ablate the MCM2/5 active site. We propose that the DNA binding differences between MCM2-7 and MCM467 correspond to a conformational change at the MCM2/5 active site with putative regulatory significance.

tion it must be separated into component single strands.
Although DNA polymerases require a single-stranded DNA (ssDNA) 2 template for activity, they have little or no intrinsic ability to unwind double-stranded DNA (dsDNA; reviewed in Ref. 1). DNA unwinding requires an ATP-dependent molecular motor termed the replicative helicase (reviewed in Ref. 2). In both prokaryotes and eukaryotes, the loading and activation of this helicase is a central and limiting event during DNA replication. Initiation culminates in replicative helicase loading, whereas the start of elongation requires extensive separation of duplex DNA by the helicase (reviewed in Refs. 3 and 4). Despite the critical importance of the replicative helicase, both its exact identity and mechanism remain controversial in eukaryotes.
Numerous studies implicate the minichromosome maintenance proteins (MCMs) as the replicative helicase. The MCMs are evolutionarily conserved from archaea to eukaryotes, with the archaea usually having a single MCM gene (5) and eukaryotes having six distinct and essential MCM genes (reviewed in Ref. 6). Each MCM protein (numbered 2-7) is an AAA ϩ ATPase, whose members include DNA helicases such as SV40 large T antigen and the papilloma virus E1 protein (7). Similar to prokaryotic replicative helicases (reviewed in Ref. 8), the six MCM subunits are both physically present in initiation and elongation complexes and functionally essential for both phases of DNA replication, evidence strongly suggesting that all six MCM subunits unwind DNA at the replication fork (reviewed in Ref. 3).
Despite the apparent dispensability of the MCM2, 3, and 5 subunits for in vitro helicase activity, their ATP active sites are essential in vivo. Mutational analysis of the Walker A ATPbinding motif indicates that all six MCM subunits require this motif for both viability and S phase progression (17,19). In vitro analysis of the corresponding MCM2-7 mutant complexes, however, indicates that the six MCM subunits fall into two functionally distinct subgroups (17). The Walker A motif in the MCM4, 6, and 7 subunits is essential for ATPase activity, whereas the ATP-binding motifs of the MCM2, 3 and 5 subunits contribute little to steady-state ATP hydrolysis hydrolysis (17). The discrepancy between the in vivo involvement of all six subunits in DNA replication and the in vitro participation of only a specific subgroup of MCM subunits in DNA unwinding remains unexplained. Recently, MCM2-7 has been isolated in vivo as part of a larger macromolecular complex having ATPdependent helicase activity; this complex additionally contains the essential GINS complex and CDC45, suggesting that these factors are activators of MCM helicase activity (20).
Although MCM467 has been extensively characterized biochemically, little work has been done with the MCM2-7 heterohexamer. To determine why the MCM2-7 complex lacks helicase activity, as well as to elucidate the function of the MCM2, 3, and 5 subunits, we have undertaken a comparative analysis of the ssDNA binding activity of the Saccharomyces cerevisiae MCM2-7 and MCM467 complexes. Our studies with MCM2-7 indicate that its lack of DNA helicase activity is not due to an inability to bind ssDNA, because both complexes have similar ssDNA binding affinities. However, we find an important difference between these complexes in ssDNA association rates. Because these two complexes differ in subunit composition (i.e. MCM2, 3, and 5), the functional difference between them suggests that the additional subunits regulate the manner in which the MCM2-7 complex interacts with DNA.

EXPERIMENTAL PROCEDURES
Buffers and Reagents-Buffers used include B1 (5 mM Tris-HCl, pH 7.5, 0.5 mM EDTA, 1 M sodium chloride, 100 g/ml bovine serum albumin (BSA)), B2 ( Table S1). Polynucleotide substrates were from GE Healthcare or Sigma. Nucleotide and DNA concentrations were calculated from absorbance at 260 nm. All other reagents were of the highest available purity.
Proteins and Purification-Hexameric MCM2-7 and MCM467 complexes were expressed and purified as described (17). The presence of individual MCM subunits were either directly visualized following separation by SDS-PAGE (for MCM6, 3, and 5) or using Western blot analysis with subunitspecific antibodies (Santa Cruz anti-MCM2 (sc-6680) and anti-MCM7 (sc-6688) and anti-MCM4 monoclonal antibody (AS6.1). 3 MCM subcomplexes were purified similar to the hexameric complexes and were made using specific mixtures of recombinant baculoviruses that each encode the desired MCM subunits, with one MCM subunit containing a C-terminal His tag to facilitate metal chelate chromatography. SV40 large T-antigen was expressed in insect cells and purified as a C-terminal Histagged protein; details are available upon request. All of the proteins were dialyzed against B2 buffer containing 100 mM potassium chloride and protease inhibitors. Protein concentrations were quantified using a Fuji FLA-5100 laser imager on SDS-PAGE-separated protein bands stained with Sypro orange (Molecular Probes, Eugene OR) using known amounts of BSA as a standard; protein concentrations unless otherwise noted are in pmol or nM of hexamer. MCM genes containing the Walker A (KA) or arginine finger (RA) alleles were completely sequenced to verify the constructs and used to make MCM complexes in a manner identical to the wild type complexes (17).
Helicase Assay-The helicase assay was performed essentially as described (21). Briefly, synthetic replication forks were prepared by annealing an equimolar mixture of oligonucleotides 233 and 235 (supplemental Table S1) and then filling in the resulting 5Ј overhang with [ 32 P]dATP in the presence of the other dNTPs using reverse transcriptase. The resulting forks were then gel-purified from an 8% native acrylamide gel following electroelution into a dialysis membrane, ethanol-precipitated, and resuspended to a concentration of ϳ1 M in 1ϫ TE. The reactions were incubated for 1 h at 37°C, stopped by the addition of unlabeled oligonucleotide 235 to 20 nM, proteinase K to 4 mg/ml, SDS to 0.4%, and a one-tenth volume of 10 ϫ stopload (25% w/v Ficoll (type 400), 100 mM EDTA, 0.1% SDS, 0.25% bromphenol blue, and 0.25% xylene cyanol) and heated at 50°C for 20 min. The samples were then separated on an 8% polyacrylamide gel at room temperature. The gel was subsequently dried, and the results were imaged and quantified using Fuji FLA-5100 phosphorimaging and Image Gauge software.
Double Filter Binding Assay-The DNA double filter binding assay was based on the work of Wong and Lohman (22). Nitrocellulose (BA85; Schleicher & Schuell) and DEAE-cellulose (DE81, Whatman) filters were prepared as described. The ssDNA substrates were 5Ј radiolabeled using T4 polynucleotide kinase and [␥-32 P]ATP. The dsDNA substrate was made by annealing an equimolar amount of nucleotide 510 to oligonucleotide 806 (supplemental Table S1), which was extended using exo Ϫ Klenow fragment (New England Biolabs) in the presence of unlabeled dNTPs spiked with [␣-32 P]dATP. Standard ssDNA binding reactions contained 4 nM of unlabeled nucleotide substrate spiked with a small quantity of 32 P labeled substrate, 120 nM MCM hexamer, 5 mM of either ATP or ATP␥S, 5 mM ␤-glycerophosphate all in 1ϫ buffer B2 with a final volume of 12.5 l. The reactions were incubated for 30 min (unless otherwise indicated) at 30°C and then spotted onto a filter stack and quickly washed with an additional 500 l of B2. After filtration using a FH 225V Filter Manifold (GE Healthcare), the nitrocellulose and DEAE membranes were separated and quantified by scintillation counting. The amount of DNA bound was calculated using Equation 1, where C NC and C DEAE are the radioactive counts retained on the nitrocellulose and DEAE membranes, respectively. The data points represent the averages of Ն3 repeats of the same experiment, and the error bars correspond to the standard deviations. The association kinetics were plotted as described (23,24) using Equation 2,  Table S1) was immobilized on the beads in buffer B1 for Ն30 min at 22°C, and the oligobound beads were separated by a magnet and washed to remove unbound oligonucleotide. Experiments using a radiolabeled oligonucleotide indicate that about 90 pmol of oligonucleotide were bound per 1 mg of beads. For each binding experiment, 20 l of beads were used. Following equilibration in buffer B3, the beads were resuspended in 25-l reactions containing B3, protein, and nucleotide as indicated and incubated for Ն30 min at 22°C. The beads were separated from unbound reaction components with a magnet, washed once with 50 l of either B3 or B3 supplemented with additional sodium chloride as indicated, resuspended in 10 l B3, and analyzed by SDS-PAGE and Sypro orange staining as described above.
Electrophoretic Mobility Shift Assay-Binding reactions contained protein as indicated, 4 nM of radiolabeled oligonucleotide 3 (supplemental Table S1), and 5 mM ATP␥S. The reactions were incubated for 30 min at 30°C and then separated by electrophoresis at 4°C through a prechilled 3.5% native polyacrylamide gel (30:1 acrylamide:bis containing 0.5ϫ TBE, 5% glycerol, 67 g/ml acetylated BSA, and 10 mM magnesium acetate) using 0.5ϫ TBE running buffer supplemented with 80 g/ml BSA and 10 mM magnesium acetate at 15 volts/cm. Protein⅐ssDNA complexes were imaged and quantified using Fuji FLA-5100 phosphorimaging.

MCM2-7 and MCM467 Form Toroidal Complexes of Differ-
ing Helicase Activity-Hexameric preparations of S. cerevisiae MCM2-7 and MCM467 were expressed in baculovirus-infected insect cells and purified to homogeneity (supplemental Fig. S1). Gel filtration and co-immunoprecipitation of the final preparations demonstrated that the complexes retained their hexameric size as was observed previously (17), and these preparations contained approximately equal stoichiometry of the specified MCM subunits as determined by two-dimensional gel electrophoresis and quantitative Western blotting (supplemental Fig. S1). As described below, these preparations demonstrated an ATP-dependent ssDNA binding activity; the average specific activity of our preparations for this activity is Ն50% of total protein (supplemental Fig. S1).
Using transmission electron microscopy, the MCM2-7 and MCM467 preparations appear largely homogeneous (Fig. 1A), with many individual complexes having a diameter (top) and height (side) of ϳ145 Å and an apparent central cavity 25-30 Å wide. Examination of hundreds of complexes from both preparations demonstrates that 20 -30% appear as ring-shaped structures, with about half containing six distinct lobes (Fig. 1A, small insets). Image reconstruction of the ring-shaped structures indicates that both complexes are of similar size and have pseudo 6-fold symmetry (Fig. 1A, large insets). These results are consistent with the published size and toroidal subunit organization of both eukaryotic MCM complexes (25)(26)(27), as well as the archaeal MCM complexes (9, 28 -30).
ATP-dependent ssDNA Binding Activity of the MCM Complexes-The ability of each MCM preparation to bind a short mixed sequence oligonucleotide (85 nt, oligonucleotide 510; supplemental Table S1) using a filter binding assay was examined. Both complexes demonstrate ATP-dependent ssDNA binding as a function of either MCM or ssDNA concentration (Fig. 2, A and B). Little ssDNA binding occurs in the absence of ATP (data not shown) or in the presence of GTP ( Fig. 2A). This interaction requires ATP binding but not hydrolysis; the poorly hydrolyzable ATP analog ATP␥S (17) stimulates the extent of ssDNA binding for both complexes relative to ATP ( Fig. 2A). Further, this binding activity is MCMdependent, because it is abolished upon preincubation of either the MCM2-7 or MCM467 complex with a monoclonal antibody that specifically binds the Walker B site in each MCM subunit (antibody AS1.1; data not shown (32)). 4 These data fit a 4 A. Schwacha and J. Bowers, unpublished observations.  NOVEMBER 16, 2007 • VOLUME 282 • NUMBER 46 hyperbolic function, consistent with a homogeneous population of molecules demonstrating noncooperative binding. Repeats of this assay with shorter probes (oligonucleotides 3 (44 nt) and 775 (40 nt); supplemental Table S1) generate similar results (data not shown). MCM phosphorylation appears to have little effect on ssDNA binding. Neither treating the complexes with lambda phosphatase to remove phosphorylation that may occur during insect cell expression nor adding phosphates with a recombinant preparation of CDC28/CLB5 (a gift from S. P. Bell, Massachusetts Institute of Technology) appreciably alters ssDNA binding levels (data not shown).

DNA Binding by the MCM2-7 and MCM467 Complexes
From these graphs, an apparent K d for ssDNA binding can be determined; using either ATP or ATP␥S, both MCM com-  Table S2). These values are in the range of those previously observed for MCM467 (apparent K d of ϳ2 nM using a 37-mer oligonucleotide (26)) and the archaeal MCM complex (apparent K d of ϳ150 -200 nM using a fork substrate (33)) and are typical of hexameric helicases (60 pM to 200 nM (2)).
ATP-dependent ssDNA binding was further examined as a function of nucleotide concentration (Fig. 2C). Both MCM complexes have similar ATP dependences (the k1 ⁄ 2(ATP) values for MCM2-7 and MCM467 being 248 Ϯ 152 and 177 Ϯ 82 M, respectively) with maximal ssDNA binding coinciding with physiological ATP concentrations (ϳ3 mM (34)). The nucleotide specificity of ssDNA binding by both MCM complexes was also tested (Fig. 2D). The triphosphate form of adenosine is needed to promote ssDNA binding, because neither ADP, AMP, dADP, nor any non-adenosine nucleotide support this activity (Fig. 2D).
The Identity of MCM Subunits within the MCM⅐ssDNA Complexes-The protein requirements for ssDNA binding were examined next. Filter binding experiments using various MCM subcomplexes demonstrated little or no ssDNA binding; however, when these subcomplexes were combined to generate either a MCM467 (MCM4/6 dimer ϩ MCM7 monomer) or MCM2-7 (MCM2/4/6 trimer ϩ MCM3/5/7 trimer) complex, high levels of ssDNA binding were recovered (data not shown and Fig. 3A). These results indicate that ssDNA binding is not an intrinsic property of any individual subunit but requires considerable oligomerization of MCM subunits. The only subcomplex that demonstrated substantial ssDNA binding was a MCM pentamer lacking MCM6 (Fig. 3A). This subcomplex is largely present as a variety of split ring structures rather than a closed toroid as visualized by electron microscopy (data not shown) and illustrates that although higher order MCM oligomerization is required for ssDNA binding, closure of the ring structure is not. To identify the MCM subunits that stably associate with ssDNA, binding experiments were performed using magnetic streptavidin beads coupled to biotinylated ssDNA (Fig. 3B). Following the addition of either the MCM2-7 or MCM467 complexes, the beads were washed in buffers of increasing salt concentration, and the proteins retained on the beads were analyzed by SDS-PAGE followed by either silver staining (Fig.  3B) or Western blotting to identify co-migrating MCM subunits (data not shown). In both cases, all input MCM subunits were retained on the beads in a ssDNA-dependent manner, strongly consistent with the notion that intact MCM hexamers bind ssDNA. Furthermore, these interactions are stable in moderate concentrations of salt as observed for MCM complexes on chromatin isolated during S phase (stable to Ͼ250 mM (35)), in contrast to the salt-sensitive chromatin binding observed during the G 1 phase (35,36).
To examine the oligomeric state of the MCM⅐ssDNA complexes, an electrophoretic mobility shift assay was used (Fig.  3C). The ability of MCM complexes to bind ssDNA in this assay varies widely; several studies only observe a shift of MCM467 in the presence of a cross-linking agent (15,26) or that the shift is inhibited in the archaeal MCMs by ATP␥S (9). However, other studies with MCM467 have observed ATP␥S-dependent mobility shifts (14,18). Using the MCM467 complex and radiolabeled oligonucleotide #3 (44 nt), we observe nucleotide-dependent binding (using ATP␥S); titrations with increasing amounts of MCM467 suggest a K d of a magnitude similar to the K d observed by filter binding (supplemental Table S2; ϳ40 nM). However, unlike the filter binding experiments, this activity is only supported by ATP␥S but not ATP (Fig. 3C, lane 5). Only one shifted band was present throughout the range of the protein titration, suggesting that the MCM467⅐ssDNA complex represents a single, defined species, although we cannot rule out the possibility of different co-migrating MCM⅐ssDNA complexes. This conjecture is further supported by experiments using the longer oligonucleotide 510 (85 nt) as a probe; again only a single shifted species was observed over the range of MCM concentrations (data not shown).
In contrast, under identical assay conditions, the MCM2-7 complex demonstrates little or no electromobility shift (Fig. 3C,  lanes 6 -8, maximum shift ϭ ϳ5%). These results suggest that even though the K d values for these two complexes are similar as determined by filter binding, the MCM2-7 complex is more susceptible to dissociation from ssDNA under these conditions than the MCM467 complex.
Polynucleotide Substrate Requirements for MCM⅐DNA Binding-The sequence specificity of MCM⅐ssDNA binding was assayed using the ability of unlabeled polynucleotides to compete with radiolabeled ssDNA for binding. Prior to addition of MCM complexes, labeled oligonucleotide was mixed together with a 1-fold (not shown), 10-fold (not shown), or 100-fold (Fig. 4A) weight excess of the indicated DNA or RNA homopolymers; the observed competition was dosage-dependent and increased with higher competitor levels. In general, the MCM2-7 complex demonstrates a higher degree of sequence specificity than MCM467. As previously reported (13,18), poly(dT) is the best competitor for MCM467; however, MCM2-7 demonstrates an even higher preference (about 2.5  Table S1). B, a magnetic bead binding assay containing biotinylated ssDNA. Lane 1, input protein (2.5 pmol) alone; lane 2, naked beads incubated with protein; lane 3, ssDNA-coated beads incubated with protein; lanes 4 -6, ssDNA-coated beads incubated with protein and 5 mM ATP. The reactions in lanes 1-4 were washed with buffer containing 50 mM NaCl; the sample in lane 5 was washed with buffer containing 250 mM NaCl, and the sample in lane 6 was washed with buffer containing 500 mM NaCl. Although the assay is ssDNA-dependent, it is not ATP-dependent. We speculate that the effective ssDNA concentration on the surface of the beads is sufficiently high to drive MCM binding even in the absence of ATP. C, electrophoretic mobility shift assay of MCM/ssDNA binding. The reactions contained radiolabeled oligonucleotide 3 (supplemental Table S1) and 5 mM ATP or ATP␥S with increasing amounts of MCM protein (see "Materials and Methods " for details). times better competition) for poly(dT) than MCM467. Similarly, poly(dC) competes for ssDNA binding with both MCM2-7 and MCM467, although not to as high a degree as poly(dT). In addition, polyribonucleotides are also capable of competing for binding, with poly(G) and poly(U) being the most effective. RNA binding by MCM467 has previously been shown (37).
The DNA length dependence of the MCM⅐ssDNA interaction was next tested (Fig. 3B). Competition experiments using unlabeled poly(dT) oligonucleotides of various defined lengths indicate that ssDNA of 15 nt is capable of competing, although maximum competition requires a length of Ն40 nt. The results were the same for both complexes and are similar to previous studies of ssDNA binding by MCM467 (between 37 and 50 nt (14,15,18,26)). These experiments do not indicate the location of the ssDNA-binding site within either MCM complex. However, assuming an axial rise of ϳ3.5 Å/base for single-stranded poly(dT) (38), a 40-nt oligonucleotide corresponds to a length of 122.5-140 Å. This is similar to the observed length of the central channel (ϳ145 Å; Fig. 1C), consistent with the likelihood that ssDNA binding occurs in this region.
The ability of the MCM2-7 and MCM467 complexes to bind blunt-ended dsDNA was also examined. Using a doublestranded form of our standard probe (85 bp, "Materials and Methods"), filter binding was performed as a function of MCM concentration (data not shown). The binding affinity of either complex for dsDNA was considerably less than for ssDNA, precluding a determining of a K d by this approach. Like the ssDNA binding activity, dsDNA binding is ATP-dependent and is blocked by preincubation with antibody AS1.1 (data not shown). To obtain an estimate of K d , a competition experiment was conducted to quantify the ability of dsDNA to compete for MCM ssDNA binding, yielding a K d of ϳ5.60 M for MCM2-7 and 2.12 M for MCM467 ( Fig. 4C and supplemental Table S2). MCM467⅐dsDNA binding has been previously observed for probes with 3Ј ssDNA tails (21, 39), but we find that probes containing either a 10-nt 3Ј or 5Ј extension are bound similarly to the blunt-ended probe (data not shown).

MCM2-7 and MCM467 Associate Differently with ssDNA; MCM2-7 Has an Additional ATP-dependent
Step-To further examine ssDNA binding by both complexes, the dissociation and association rates (k d and k a (apparent) , respectively) were measured. The observed dissociation for the two complexes was quite similar and slow (supplemental Fig. S2), with the MCM467 complex having a slightly slower k d than the MCM2-7 complex. Interestingly, their apparent association rates were quite different (Fig. 5A). Using ATP␥S, the MCM467 complex binds ssDNA quickly, with ϳ50% of total binding occurring in ϳ2.5 min; the MCM2-7 complex binds ssDNA very slowly, with 50% ssDNA binding occurring in ϳ12 min. Substituting ATP for ATP␥S gave similar results (not shown).
The difference in the association rates between these two complexes was further investigated. We reasoned that the relatively slow association of the MCM2-7 complex with ssDNA could reflect one or more additional ATP-dependent steps by MCM2-7 prior to productive ssDNA binding. To test this hypothesis, both complexes were separately preincubated with ATP␥S for 30 min, and then ssDNA was added. Although preincubation with nucleotide has little effect on the kinetics of MCM467 binding, nucleotide preincubation with MCM2-7 greatly accelerates its association rate to resemble that of MCM467 (Fig. 5B). Similar results were obtained using ATP (data not shown). These data demonstrate that the slow asso-  Table S1). Competitor and radiolabeled ssDNA substrates were mixed prior to protein addition. A, competition of 100-fold weight excess of unlabeled ssRNA, ssDNA, and dsDNA homopolymers for either MCM2-7 or MCM467 binding to labeled oligonucleotide. The results were plotted as fold competition, where a 10-fold competition corresponds to a 90% reduction in MCM binding to the labeled oligonucleotide. B, the addition of 100-fold molar excess of unlabeled poly(dT) competitor ssDNA oligonucleotides of length 15, 20, 25, 30, 35, 40, 50, 60, or 80 nt to standard binding reactions containing radiolabeled poly(dT) 80mer. C, the indicated amount of unlabeled dsDNA substrate was added to standard ssDNA binding reactions (4 nM 32 P-labeled oligonucleotide 510, 120 nM MCM hexamer, 5 mM ATP␥S, final volume of 12.5 l). The reactions were incubated and filtered as described under "Double Filter Binding Assay." The data were normalized to the "no competitor" reaction and plotted as hyperbolic decay using nonlinear regression. The IC 50 ϭ 683 Ϯ 78.6 nM and 360 Ϯ 26.5 nM for MCM2-7 and MCM467, respectively. ciation of MCM2-7 in the absence of ATP preincubation reflects an interaction between the MCM2-7 complex and ATP rather than slow ssDNA binding. To obtain association constants (supplemental Table S2), the results were replotted in the manner of von Hippel (Fig. 5C and Ref. 24).
It should be noted that even at the faster association rate, the ssDNA binding activity of the MCMs is considerably slower than would be anticipated for simple diffusion-limited bimolecular rates (10 7 -10 8 M Ϫ1 s Ϫ1 ) and is somewhat slower than what is commonly observed other hexameric helicases (2,40). Because ssDNA binding likely occurs within the central channel of the complex, the relatively slow binding that is observed may correspond to specific ssDNA loading into the channel rather than simple binding per se.
The Involvement of MCM ATP Active Sites in ssDNA Binding-To determine which MCM ATP active sites contribute to ssDNA binding, we tested mutant MCM2-7 complexes containing alanine substitution mutations of the universally conserved lysine within the Walker A ATP-binding motif (the MCM KA mutants; Fig. 6A). These mutant complexes were expressed and purified as stable heterohexamers in a manner identical to the wild type complexes. We previously characterized such mutant complexes for their effects on steady-state ATP hydrolysis; inclusion of any one such mutant subunit in the context of the remaining five wild type subunits largely abolishes ATP hydrolysis of the entire heterohexamer (17).
Seven mutant MCM2-7 complexes were tested for ATP-dependent ssDNA binding. Six complexes contain a single indicated KA mutant subunit in the presence of five other wild type subunits; the seventh complex contains the KA mutation in all six MCM subunits (6ϫKA). Fig. 6B shows that the 6ϫKA complex is completely unable to bind ssDNA, indicating that either the MCM Walker A mutations block nucleotide binding rather than hydrolysis or, alternatively, prevent the effects of ATP binding from being transmitted to the DNA-binding domain(s). In contrast, the other MCM complexes demonstrate a range of ssDNA binding activities. The complex containing the MCM4KA mutation is completely devoid of ATPstimulated ssDNA binding; complexes containing the mutation in MCM2 or MCM6 bind ssDNA at essentially wild type levels; and complexes with mutations in MCM3, MCM5, or MCM7 demonstrate intermediate levels of binding.
The ssDNA binding activity of these complexes was further explored by measuring their nucleotide dependence (Fig. 6C). In most cases, ssDNA binding is stimulated by both ATP and ATP␥S in a manner similar to the wild type MCM2-7 complex. However, we have repeatedly noticed that the MCM2-7 preparations containing the MCM7KA mutation demonstrate elevated levels of ATP-independent ssDNA binding; this binding is only stimulated slightly by ATP␥S but not by ATP. The significance of this observation is currently unknown. Nevertheless, these results indicate that unlike steady-state ATP hydrolysis (17), participation of the six MCM ATPase active sites in ssDNA binding is very different, with only the MCM4 subunit, and to a smaller extent the MCM7 subunit, being key to this activity.
The Difference in ssDNA Association Rates between MCM2-7 and MCM467 Depends upon the MCM2/5 Active Site-MCM2-7 contains three MCM subunits that MCM467 lacks: MCM2, 3, and 5. Because the MCM2-7 and MCM467 complexes differ in their association rate with ssDNA, the MCM2, 3, or 5 subunits  Table S1), and 120 nM MCM protein. B, effect of nucleotide preincubation on MCM⅐ssDNA association rates. Preparations of MCM2-7 or MCM467 were preincubated with ATP␥S for 30 min at 30°C prior to the addition of labeled oligonucleotide 510 at t ϭ 0. C, the data in A (no pre) and B (pre) were replotted in the manner of von Hippel (24); slope ϭ k a . are likely to be involved in this effect. Knowing that preincubation with ATP relieves the difference in ssDNA association rates between MCM2-7 and MCM467, we reasoned that a mutation in the ATPase active site of MCM2, 3, or 5 may alter the ssDNA association rate of MCM2-7. Of the KA mutant MCM complexes that still demonstrate ATP-dependent ssDNA binding, association kinetics were measured with and without ATP preincubation. Complexes that contain KA mutations in MCM2, 3, or 6 give results similar to the wild type MCM2-7 complex; in these cases, preincubation of the complex with ATP increases the association rate to MCM467 levels (supplemental Fig. S3). In sharp contrast, a MCM2-7 complex containing the MCM5 Walker A mutation binds ssDNA at a similar fast rate in both the presence and absence of ATP preincubation (Fig. 6D, left panel). This indicates that the MCM5 ATP active site is uniquely involved in the slow ATP-dependent step in MCM2-7 ssDNA association.
The MCMs, like most AAA ϩ ATPases, form ATP active sites at subunit interfaces; one subunit contributes a Walker A motif, whereas the adjacent subunit contributes an essential arginine (41). The MCM5 Walker A motif is predicted to form an active site with the essential arginine of MCM2 (41). To test the involvement of MCM2 in the MCM2-7 association rate, we generated MCM2-7 complexes with appropriate arginine to alanine substitutions (R3 A) 5 and assayed their association rates. Strikingly, only the MCM2-7 complex containing the MCM2RA mutation demonstrated an accelerated ssDNA association rate (data not shown and Fig. 6D, right panel). This result further substantiates the involvement of the MCM2/5 active site in MCM2-7 ssDNA association.

DISCUSSION
Our comparative analysis of the ssDNA binding properties of the MCM2-7 and MCM467 complexes reveals similarities and key functional differences both between these complexes and in relation to other helicases. In common with typical hexameric helicases, both MCM complexes demonstrate (pseudo)6-fold symmetry and ATP-dependent ssDNA binding. However, unlike typical homohexameric helicases that contain functionally equal subunits and bind ssDNA in a sequence-independent manner, our data indicate that the MCM2-7 subunits have differential involvement in ssDNA binding and a marked binding preference for poly(dT). Although both complexes have similar ssDNA binding affinities, they differ in their ssDNA association kinetics, evidence suggesting that the MCM2, 3, and 5 subunits regulate the loading or activation of the MCM2-7 complex in vivo. Further, our data demonstrate that the lack of helicase activity by the MCM2-7 complex is not due to an inability to bind ssDNA, as previously suggested (41).
The MCM Complex Has Unusual Properties for a Hexameric Helicase-Although helicases usually bind ssDNA in a sequence-independent manner (2), the MCMs prefer to bind to polypyrimidine tracts. Although this property was previously observed for MCM467 (14,18), we show that this preference is even stronger with MCM2-7 (Fig. 3A). Because eukaryotic replication origins are usually A/T-rich (42)(43)(44) and often contain poly(dT) tracts (45,46), an increased affinity to poly(dT) sequences could facilitate loading of the MCMs onto replication origins as previously proposed (18). However, this 5 M. Bochman, S. Bell, and A. Schwacha, manuscript in preparation. enhanced affinity for poly(dT) is also potentially disruptive during elongation in vivo; helicases need to freely translocate along DNA, and a high affinity toward poly(dT) might impede translocation and cause the replication fork to pause. Such events are deleterious; pausing leads to fork collapse and the production of potentially lethal DNA double strand breaks (47). Possible fork pausing by the MCM complex during normal replication may have broader implications for human health, because DNA replication in eukaryotes sometimes pathologically results in chromosome breaks at A/T-rich sequences referred to as fragile sites (reviewed in Ref. 48).
In contrast to homohexameric helicases, the individual MCM subunits contribute differentially to ssDNA binding. Analysis of MCM2-7 complexes containing Walker A mutant subunits reveals that, unexpectedly, only the active site on MCM4 is absolutely required for this activity, whereas the MCM7 active site is required to make the interaction ATP-dependent. Previous analysis supports these observations. In Schizosaccharomyces pombe, MCM complexes containing the MCM4KA mutation lose association with chromatin (19). In contrast, MCM6KA mutant complexes can still bind chromatin in a semi-purified Xenopus in vitro DNA replication system (15,49). One puzzling feature of the MCM2-7 complex is why it contains six distinct subunits. Our data suggest this arrangement has allowed individual subunits to evolve specialized functions, with some subunits specializing in ssDNA binding (MCM4 and 7), whereas other subunits may specialize in regulating this association (MCM2 and 5, below).
Key Differences between the Two MCM Complexes-Both the archaeal MCM complex (12, 50 -53) and the eukaryotic MCM467 complex (39,53) bind dsDNA. This ability has fueled speculation that MCM2-7 might function as a dsDNA pump (54). Although we demonstrate that both complexes have an ATP-dependent dsDNA binding activity, their affinity for dsDNA is ϳ100-fold lower than for ssDNA. Because MCM467 binds dsDNA better than MCM2-7, it suggests that the MCM2, 3, or 5 subunits may negatively affect dsDNA binding, raising the possibility that regulation of these subunits may facilitate dsDNA interaction under certain conditions. The poor affinity of MCM2-7 for dsDNA does not definitively rule out its involvement as a pump, however. Further mutational analysis of the MCM complex using mutants that specifically affect dsDNA binding (as have been developed in the archaeal MCM complex (12)) will be required to determine the in vivo significance of dsDNA binding.
Unexpectedly, MCM2-7 and MCM467 bind ssDNA with different kinetics. Although the MCM467 complex binds ssDNA relatively quickly, MCM2-7 binds approximately five times more slowly. The slow ssDNA association by MCM2-7 is due to an interaction involving ATP, because preincubation of the MCM2-7 complex with ATP removes this kinetic barrier. The effects of preincubation occur very slowly, requiring about 20 -25 min of MCM/ATP preincubation to stimulate maximal ssDNA binding (data not shown). This slow rate does not reflect slow ATP binding to the complex, because no noticeable time lag was observed in steady-state ATPase hydrolysis studies of the MCM2-7 complex. 6 We hypothesize that the slow step corresponds to an ATP-dependent conformational change by the MCM2-7 complex. Because the obvious differences between these two complexes are the MCM2, 3, and 5 subunits, it is reasonable to expect that these subunits are responsible for the difference in ssDNA association. This expectation is confirmed by our finding that the MCM5KA and MCM2RA mutant complexes bind ssDNA quickly without ATP preincubation, implying the involvement of these subunits in this slow ssDNA association step.
The Possible Nature of the MCM2-7 ATP-dependent Conformational Change-As previously demonstrated (16), the ATP active sites within the MCM complex lay at dimer interfaces, with one subunit contributing a Walker A motif, whereas the other subunit contributes a catalytically essential arginine "finger." This arrangement is typical for AAA ϩ ATPases (reviewed in Ref. 55). To fit these subunit associations onto the observed toroidal structure, MCM2 and MCM5 need to be juxtaposed to form an active site, with MCM5 contributing the Walker A motif and MCM2 contributing the essential catalytic arginine (16).
The effect of either the MCM5KA or the MCM2RA mutation on the association of MCM2-7 with ssDNA is puzzling. Both motifs likely ablate important contacts with ATP, suggesting that normally the slow association reflects an inhibitory nucleotide, or other possible inhibitor, bound at this site. In contrast, preincubation of MCM2-7 with ATP increases ssDNA association, suggesting that ATP binding is responsible for this effect. We suggest that both situations may serve to displace some inhibitory interaction at the MCM2/5 active site, either of a protein-protein or protein-ligand nature, resulting in a conformational change that increases the ssDNA association rate. One possible inhibitor could be ADP that has remained tightly bound to this site during protein purification, a common property of ATPases (56,57).
Is this proposed conformational change at the MCM2/5 interface physiologically relevant? In common with the other MCM genes, MCM2 and MCM5 are essential, indicating that the MCM467 helicase activity is insufficient to carry out in vivo DNA replication. Yet our previous analysis indicates that neither MCM2 nor MCM5 has critical involvement in ssDNA binding or steady-state ATP hydrolysis (17), suggesting that their essential in vivo function depends upon some yet undiscovered activity.
Available evidence supports a regulatory role for these two subunits in DNA association. Unlike the other five MCM subunits, MCM2 specifically binds chromatin (58). Further, both MCM2 and the MCM5 are linked to the CDC7/DBF4 regulatory kinase, which aids in cell cycle-dependent assembly of the elongation complex and functions immediately downstream of MCM2-7 loading at replication origins (3). Although the mechanistic role of CDC7/DBF4 phosphorylation is poorly understood, the MCM complex is likely to be the focus of its activity. A specific mutant in MCM5 (59) exists that bypasses the normally essential function of CDC7 in DNA replication, 6 A. Schwacha, unpublished observations. and the MCM2 subunit is a major substrate for this kinase (60,61). These results suggest that both MCM2 and MCM5 serve a regulatory function, possibly to activate MCM2-7 helicase activity. Perhaps MCM phosphorylation by the CDC7/DBF4 kinase causes a favorable conformational change at the MCM2/5 site that activates the DNA unwinding activity of the MCM2-7 complex in vivo.