Mcm10 Self-Association Is Mediated by an N-Terminal Coiled-Coil Domain

Minichromosome maintenance protein 10 (Mcm10) is an essential eukaryotic DNA-binding replication factor thought to serve as a scaffold to coordinate enzymatic activities within the replisome. Mcm10 appears to function as an oligomer rather than in its monomeric form (or rather than as a monomer). However, various orthologs have been found to contain 1, 2, 3, 4, or 6 subunits and thus, this issue has remained controversial. Here, we show that self-association of Xenopus laevis Mcm10 is mediated by a conserved coiled-coil (CC) motif within the N-terminal domain (NTD). Crystallographic analysis of the CC at 2.4 Å resolution revealed a three-helix bundle, consistent with the formation of both dimeric and trimeric Mcm10 CCs in solution. Mutation of the side chains at the subunit interface disrupted in vitro dimerization of both the CC and the NTD as monitored by analytical ultracentrifugation. In addition, the same mutations also impeded self-interaction of the full-length protein in vivo, as measured by yeast-two hybrid assays. We conclude that Mcm10 likely forms dimers or trimers to promote its diverse functions during DNA replication.


Introduction
DNA replication is carried out by multi-protein factories that in eukaryotes are assembled in stages to regulate the timing of DNA synthesis within the cell cycle [1,2,3]. Pre-replicative complexes (pre-RCs) are assembled at origins during G1 and are composed of origin recognition complex (ORC), Cdc6, Cdt1, and an inactive form of the minichromosome maintenance (Mcm) 2-7 helicase. The pre-RC is activated at the onset of S-phase by Dbf4dependent kinase (DDK) and cyclin-dependent kinase (CDK) activities [4]. In yeast, CDK phosphorylates Sld2 and Sld3 and facilitates their binding to Dpb11 [5,6,7] and DDK phosphorylates Mcm2 and Mcm4 [8,9] to promote the assembly of additional factors. Ultimately, pre-RC activation leads to the loading of Cdc45 and GINS (Go-Ichi-Nii-San), which form a functional helicase (CMG) complex with Mcm2-7 [10,11,12,13,14,15]. Unwinding of the origin is signified by loading of replication protein A (RPA), followed by recruitment of DNA polymerase a (pol a)-primase, which initiates DNA synthesis at the heads of the leading strands and each Okazaki fragment.
Mcm10 contains at least three functional domains [20]. An Nterminal coiled-coil (CC) domain (NTD) has been implicated in Mcm10 self-association [20] and the interaction with Mec3, a subunit of the 9-1-1 clamp (Alver and Bielinsky, unpublished results). In addition, the protein has a highly conserved internal (ID) and vertebrate-specific C-terminal domain (CTD) that bind DNA and the catalytic (p180) subunit of pol a [20,34,37]. The yeast orthologs have also been shown to interact with DNA and pol a despite the apparent lack of the CTD [16,19,21,38]. Thus, the ID is likely to mediate these interactions in S. cerevisiae. Moreover, recent evidence suggests that acetylation of the ID and CTD in human Mcm10 differentially controls their respective DNA binding and protein-protein interactions [39]. However, the details of this potential mechanism are still unclear.
The oligomeric state of Mcm10 has remained controversial, with reports ranging in size from 1-12 subunits [37]. scMcm10 was shown by size-exclusion chromatography to form large, 800-kDa homocomplexes consisting of ,12 molecules [40]. Selfassociation in that case was presumably dependent on the integrity of the zinc-finger (ZnF) motif within the ID, although purified ID from Xenopus laevis Mcm10 (xMcm10) was found to be monomeric [20]. Electron microscopy (EM) and single-particle analysis of the human protein showed a hexameric ring-shaped structure [41]. In contrast, asymmetric monomeric and dimeric forms of S. pombe Mcm10 (spMcm10) were reported [42,43]. Similarly, xMcm10 exhibited mass-dependent association into low molecular weight complexes that were presumed to represent Mcm10 dimers solely on the basis of dimerization of the isolated NTD [20]. Consistent with NTD-mediated self-assembly, scMcm10 showed a strong yeast two-hybrid interaction that was ablated when one binding partner carried a truncation of the first 100 amino acids. Moreover, these truncation mutants exhibited a striking sensitivity to the replication inhibitor hydroxyurea that was revealed in the absence of the 9-1-1 checkpoint clamp (Alver and Bielinsky, unpublished results). These observations agree with a report that demonstrates that scMcm10 is monomeric when bound to dsDNA, but capable of forming multi-subunit complexes on ssDNA [21].
Here, we studied the role of the NTD on xMcm10 self-assembly using structural, biophysical, and in vivo binding assays. We show that the CC region is necessary and sufficient to explain Mcm10-Mcm10 interaction, and is capable of forming both dimers and trimers in solution. The trimeric form of the CC was stabilized in a crystal structure, which revealed the residues at the subunit interface. Specific mutations at this interface disrupted dimerization of the isolated CC, the NTD, and eliminated self-association of the full-length protein by yeast-two hybrid interaction.

Protein Purification
Full-length xMcm10 was purified from baculovirus infected insect cells using the Bac-to-Bac expression system (Invitrogen). The gene was subcloned into pFastBac1 vector with a His 6 tag added to the C-terminus by PCR. Protein was expressed in Hi-5 insect cells for 48 hr. Cells were resuspended in lysis buffer (50 mM Tris buffer (pH 7.5), 500 mM NaCl, 10% glycerol) and hand homogenized. Protein was purified by nickel-nitrilotriacetic (NTA) acid affinity chromatography. Pooled Ni-NTA fractions were buffer exchanged into 50 mM Tris buffer (pH 7.5), 150 mM NaCl, and 10% glycerol and purified using Source Q (GE Healthcare) cation exchange, followed by gel filtration on a Superose6 (GE Healthcare) column equilibrated in 25 mM Tris buffer (pH 7.5), 150 mM NaCl, 5% glycerol, and 2mM b-mercaptoethanol (BME). Mcm10DN (aa 230-860) and Mcm10-NTD (aa 1-145) were expressed and purified as previously described [20,34]. Gene sequences encoding xMcm10 amino acids 95-124 and 95-132 were cloned into a pMALX(E) vector using NotI and BamHI restriction sites to generate CC fragments fused to the Cterminal end of a mutant form of maltose binding protein (MBP) with a short, uncleavable peptide linker as previously described [44]. The recombinant proteins were overexpressed in E.coli C41 cells for 3 hrs at 37uC in LB medium supplemented with 100 mg/ mL ampicillin with addition of 0.5 mM IPTG at mid-log phase. Cells were resuspended in lysis buffer and lysed under pressure (25,000 p.s.i.) using an EmulsiFlex-C3 homogenizer (Avestin, Inc.). Lysate was centrifuged at 35,0006g for 20 min. The supernatant was incubated with amylose resin (New England Biolabs) overnight at 4uC and washed with 15 column volumes of lysis buffer. Fusion proteins were eluted with 40 mM maltose in lysis buffer, concentrated, and further purified by size exclusion chromatography on a Superdex S200 column (GE Healthcare) equilibrated in 25 mM Tris buffer (pH 7.5), 150 mM NaCl, 5% glycerol, 4 mM BME, and 40 mM maltose. Purified MBP-CC proteins were flash frozen and stored at -80uC in 25 mM Tris buffer (pH 7.5), 150 mM NaCl, 0.2 mM tris(2-carboxyethyl)phosphine hydrochloride (TCEP), and 5 mM maltose.

X-ray Crystallography
Purified MBP-CC proteins were concentrated to 50 mg/mL using a 10-kDa MWCO Amicon spin concentrator and buffer exchanged into 25 mM Tris buffer (pH 7.5), 150 mM NaCl, 0.2 mM TCEP, and 5 mM maltose for crystallization. Crystals were grown by sitting drop vapor diffusion at 16uC by adding 2 ml protein to 2 ml reservoir solutions containing 0. , 23% PEG 3350, and 15% glycerol (MBP-CC 95-132 ) prior to data collection. X-ray diffraction data were collected at the Advanced Photon Source LS-CAT/sector 21 and processed using HKL2000 [45].
The structures of MBP-CC 95-124 (2.4 Å ) and MBP-CC 95-132 (3.1 Å ) were determined by molecular replacement using MBPX(E) from PDB ID 3H4Z as a search model [44,46]. Phases generated from three copies of MBP in the asymmetric unit revealed clear electron density for the Mcm10 coiled-coil in both cases. The models were built in COOT [47] and refined against a maximum likelihood target in PHENIX [48]. Although one additional turn of the a-helix was visible in the MBP-CC 95-132 structure, the side chains could not be unambiguously identified, and thus the lower resolution structure was not pursued further. Anisotropic motion was modeled using translation/libration/ screw-rotation (TLS) refinement, with each protomer defined as a TLS group. Individual anisotropic B-factors derived from the refined TLS parameters were held fixed during subsequent rounds of refinement. Adjustments to the model and addition of solvent was carried out iteratively through inspection of 2F o -F c , F o -F c and composite omit electron density maps. The final MBP-CC 95-124 model, consisting of MBPX(E) residues 1-367, the five-residue linker (AAAMG), and xMcm10 residues 95-122, was validated using PROCHECK [49]. 97.5% and 2.2% of residues reside in the favored and allowed regions of the Ramachandran plot, respectively. The remaining 0.3% in disallowed regions reside in the MBP-CC linker, MBP loops, or the extreme MBP amino terminus. The final model was deposited in the Protein Data Bank under accession number 4JBZ.  7), 150 mM NaCl, and 0.3 mM TCEP. Buffer viscosity, buffer density and partial specific volume were calculated using SEDNTERP [50]. Data was processed using c(s) analysis in SEDFIT [51,52].

Ultracentrifugation and Light Scattering
Molecular mass analysis of full-length Mcm10 and Mcm10DN by size exclusion chromatography and multi-angle light scattering (SEC-MALS) was carried out using a Superose6 10/300 GL column (GE Healthcare) operating at 0.4 ml/min in 25 mM Tris buffer (pH 7.5), 150 mM NaCl, 2% glycerol, and 2 mM BME. Absorbance, refractive index, and light scattering of the eluants were measured using a DAWN HELEOS II detector (Wyatt Technology) and data analyzed by ASTRA software.

Yeast Two-hybrid Assay and Immunoblotting
The Matchmaker 3 (Clontech) system was utilized to measure protein-protein interaction by yeast two-hybrid assay. Genes encoding the bait proteins were cloned into pGBKT7 to express Myc-tagged fusions with the Gal4-binding domain. Genes encoding the prey proteins were cloned into pGADT7 to express hemagglutinin (HA)-tagged fusions with the Gal4-activation domain. Plasmids were sequence verified and transformed into the reporter strain, AH109 (Clontech), in which the expression of ADE2 and HIS3 was under the control of a GAL1,10 promoter. Transformants were selected on medium lacking leucine and tryptophan, and protein interaction was scored on quadruple drop-out plates lacking adenine, leucine, histidine and tryptophan. To verify protein expression, total protein extracts were obtained from yeast cultures by trichloroacetic acid (TCA) preparation as described and separated by SDS-PAGE and subsequently transferred onto nitrocellulose membrane [53]. HA-tagged xMcm10 was visualized using a horseradish peroxidase (HRP)conjugated anti-HA antibody (Roche, 3F10). Myc-tagged

Mcm10 Self-associates through its N-terminal Domain
Full-length xMcm10 self-associates into low molecular mass complexes, which we previously hypothesized to form as a result of NTD dimerization [20]. In order to investigate the contribution of the NTD on self-association behavior, we purified a deletion mutant lacking the first 230 residues of xMcm10 (Mcm10DN) and analyzed molecular masses of full-length and Mcm10DN proteins by sedimentation velocity analytical ultracentrifugation and size exclusion chromatography coupled to multi-angle light scattering (SEC-MALS). The full-length protein showed a broad distribution of low and high sedimenting species indicative of multiple oligomeric states ( Figure 1B). The complex nature of the sedimentation profile precluded assignment of precise molecular mass to each peak. Similarly, SEC-MALS analysis of the fulllength protein showed a broad elution profile with at least three overlapping but distinct peaks and the majority of the protein existing as the lower molecular weight form ( Figure 1C). As with the sedimentation data, the overlapping nature of the peaks only allowed for an estimation -not a definitive assessment -of the respective molecular masses. From the light scattering data, the three major species were approximately 90.4 kDa (I), 189.3 kDa (II), and 322.7 kDa (III) in size, corresponding to 1, 2, and 3.4 Mcm10 subunits, respectively (the calculated mass from amino acid composition is 95.4 kDa). Although not strictly conclusive, these data are consistent with reports of dimeric and trimeric forms of yeast Mcm10 [21,42].
In contrast, Mcm10DN formed a single species corresponding unequivocally to a monomeric protein in both experiments. The molecular mass of the major (1.5S) peak from sedimentation velocity ( Figure 1B) was calculated to be 68.8 kDa, compared to 70.4 kDa calculated from the amino acid composition. The minor species observed at 2.3S did not increase with the protein concentration (Table S1) and was thus judged to be a contaminant. In addition, Mcm10DN eluted as a single, monodispersed species from a size exclusion column with a molecular mass of 75.160.8 kDa determined by MALS ( Figure 1D). Therefore, deletion of the NTD eliminated self-association of the full-length protein.

The Structure of the Mcm10 Coiled-coil Region
The NTD of the vertebrate and Saccharomyces cerevisiae Mcm10 orthologs contain a putative CC that we previously hypothesized accounts for dimerization of the NTD [20]. We tested the ability of this region to dimerize by fusing the peptide corresponding to xMcm10 residues 95-132 to maltose binding protein (MBP) and analyzing by sedimentation velocity ultracentrifugation. MBP alone sedimented as a monomer, in agreement with a previous  Table S1 in the Supporting Information. doi:10.1371/journal.pone.0070518.g005 determination [54]. In contrast, two species consistent with monomeric and dimeric forms of MBP-CC were present ( Figure  S1A,B). The dimeric form of MBP-CC persisted on SDS-PAGE gels even in the presence of high concentrations of reducing agents ( Figure S1C), a characteristic of coiled-coils observed in other proteins [55].
To verify this region of the protein as a bona fide CC, we determined the crystal structure of the MBP-CC fusion protein to a resolution of 2.4 Å (Figure 2). The highest quality diffraction data were obtained from a construct spanning xMcm10 residues 95-124 crystallized under low pH conditions (Table S2). The final model was refined to crystallographic residuals of 16.4% (R work ) and 20.5% (R free ). Surprisingly, the asymmetric unit consisted of a trimeric assembly with the Mcm10 residues at the center forming a parallel three-helix CC wrapped in a left-handed superhelix (Figure 2A,B). Trimer formation is not a crystallographic artifact, since we observed trimeric and dimeric MBP-CCs in solution under the same (low pH) conditions used for crystallization ( Figure 2C). Similarly, we verified that MBP did not influence trimerization since MBP alone is monomeric in solution at the low pH condition ( Figure 2D). Thus, the Mcm10 CC has the propensity to form both dimeric and trimeric helical bundles, consistent with our SEC-MALS analysis of the full-length protein ( Figure 1C).
CCs in other proteins have been shown to exist in multiple oligomeric states, a property largely dependent on the characteristics of the a and d hydrophobic side chains of the heptad repeat that form the helical interface [56]. For example, two-, three-, and four-stranded CCs in the GCN4 leucine zipper were engineered by mutating the a and d positions [57]. The Mcm10 CC helical region spans Glu98 to Leu122, two invariant residues in the human, frog, mouse, and budding yeast orthologs ( Figure 3A), although we did observe the helices to extend to at least Thr125 in lower resolution structures obtained from a longer 95-132 construct (data not shown). Most importantly, the high resolution of the structure enabled us to identify the residues of the CC interface as Leu104, Leu108, Met111, Met115, and Leu118 ( Figure 3B). This interface is entirely hydrophobic, with the side chains of each residue interacting with its equivalents on the other two helices through van der Waals packing around a three-fold rotation axis ( Figure 3C,D).
We expected the same residues to form the interface in a dimeric form of the CC based on other structures with both dimeric and trimeric propensities [57,58,59]. The conformation of the Mcm10 CC trimer is virtually identical to the isoleucine zipper variant of the GCN4 CC [58], with only a modest divergence at the N-terminal end ( Figure 4B), which likely results from nonhydrophobic heptad repeat a and d residues (Gln97, Lys101) and/ or the MBP tag ( Figure 4A). We therefore constructed a model of the dimeric Mcm10 CC using the GCN4 leucine zipper dimer as a template ( Figure 4C) [60]. As shown in Figure 4D, the dimer and trimer are related by a simple 60u rotation and 8 Å translation of one helix. The a and d positions are conserved between the two models, and the conformations of only two side chains (Leu104 and Leu108) needed to be adjusted to avoid steric collision across the dimer interface. Thus, only modest adjustments are required to interconvert between the CC dimer and trimer.

Mutations in the Coiled-coil Motif Disrupt Mcm10 Oligomerization
To validate the crystal structure as representative of a functional CC, we designed mutations aimed at disrupting self-interaction. We introduced electrostatic repulsion at the interface by substituting Leu104 and Leu108 with aspartate to create a L104D/ L108D double (2D) mutant. In addition, we eliminated side chains at positions 104, 108, 115, and 118 by alanine substitution to create a L104A/L108A/M115A/M118A quadruple (4A) mutant. Mutations were introduced into the MBP-CC 95-124 and NTD protein constructs and tested for dimerization using sedimentation velocity ( Figure 5). Both 2D and 4A mutants disrupted dimerization of the wild-type CC and NTD ( Figure 5A,B). Interestingly, replacing only Leu104 and Leu108 with alanine (2A mutant) was not enough to disrupt dimerization ( Figure S2A), suggesting that the remainder of the interface is sufficient to hold the CC together.
To confirm that the 2D and 4A mutations impeded Mcm10 dimerization in vivo, we conducted a yeast two-hybrid analysis ( Figure 6). Full-length xMcm10 as well as the 2D, 4A and Mcm10DN mutants were each fused to either a Gal4-binding oractivation domain. The interaction between T-antigen (T-ag) and p53 served as a positive control, whereas combinations of the respective activation domain fusions combined with an empty vector served as negative controls. Plasmid retention was evaluated by spotting cells onto double selection plates lacking leucine and tryptophan. We tested three independent strains for each of the two-hybrid pairs, as indicated in Figures 6A and B. The ability to interact was scored on quadruple selection plates. Full-length xMcm10 displayed strong self-interaction, almost at the level as the binding between T-ag and p53 (compare Figures 6A and B, right panels). In contrast, the interaction was eliminated by the 2D and 4A point mutations and the N-terminal deletion construct ( Figure 6A). As expected, empty vector controls did not show any viable colonies ( Figure 6B). Importantly, the lack of self-association between the respective mutants or the N-terminally truncated protein and full-length xMcm10 was not due to differences in protein expression, as analyzed by Western blot (Figure 6C,D). Since we also observed a significant difference in the sedimentation velocity profile of purified full-length wild-type protein and the 2D mutant ( Figure S2B), we conclude that the CC is the primary oligomerization motif in xMcm10.

Discussion
This work identifies an evolutionarily conserved CC motif in the N-terminus of xMcm10 and provides evidence that it is required for self-association. Our data also strongly suggest that Mcm10 exists in a dynamic equilibrium between multiple oligomeric states, which helps to explain the disagreement in the literature regarding the number of subunits. We observed a broad distribution of states of the full-length protein using two quantitative approaches, and consistently found the presence of both dimeric and trimeric species. It is striking that in addition to estimates of dimer and trimer formation of full-length xMcm10 by SEC-MALS, we detected a mixture of dimeric and trimeric forms of the isolated CC region, with dimers predominating in solution and a trimer in crystals.
The propensity of the Mcm10 CC to form multiple states can be explained by the particular CC sequence, since the rules governing the number of CC subunits as a function of the amino acids at the a and d positions within the heptad repeat is well understood [56,57,59,61]. Inclusion of methionine at the a position in the human, mouse, and frog Mcm10 CC (Met115 in our structure) likely destabilizes the dimer and would even favor parallel tetramers and pentamers [61]. This raises the possibility that the Mcm10 CC could accommodate higher order oligomers, formed either as a simple association between the helices or as more complex patterns such as a trimer of dimers [62,63,64]. Regardless of the oligomeric state, the residues lining the supercoil interface would remain the same. Indeed, our data indicate that the a and d residues identified in the trimer crystal structure are important for dimerization of the CC and the NTD in vitro and the selfassociation of the full-length protein in vivo.
The existence of dimers and trimers implies that the Mcm10 CC is metastable and therefore its oligomeric state is sensitive to environmental factors. In support of this, the Mcm10 CC trimer is stabilized at lower pH. Interestingly, pH dependent CC switches are important biological mechanisms by which proteins change conformation to drive various processes [56]. For example, viral glycoproteins adopt trimeric CCs in response to pH as a mechanism to fuse viral and cellular membranes [63,65,66]. In fact, the crystal structure of the human T cell leukemia virus type 1 transmembrane ectodomain, determined as an MBP fusion, formed a parallel trimeric CC required for proper function [67,68], further validating the importance of our trimeric MBP-Mcm10-CC structure. In addition, CC folding and remodeling in response to other environmental factors, including temperature and effector molecules is a general phenomenon [69,70,71,72,73,74,75]. On the basis of these examples and consistent with our data, we speculate that the Mcm10 CC exists mainly as an intrinsically disordered monomer or as a CC dimer, and has the propensity to attain other multimeric configurations in response to its environment.
It is intriguing to speculate that Mcm10 may adopt different oligomeric states to perform multiple roles during DNA replication. For example, higher-order oligomers may be used for sequestering Mcm10 at the replication fork. Upon pre-RC activation and origin melting, Mcm10 may reform as a dimer or trimer as DNA is denatured and replication factors recruited to the emerging fork. scMcm10 was reported to exhibit differential packing on ssDNA versus dsDNA [21], suggesting that a change in the Mcm10 conformation or its oligomeric state could facilitate strand separation. In this context, it is noteworthy that Mcm10 binds ssDNA with a 3-5-fold preference over dsDNA [19,20,21]. Oligomerization on ssDNA might thus assist in the initial unwinding step and aid what has been termed ''helicase activation'' [3] but may very well just be the coordinated stabilization of short stretches of unwound DNA after the separation of Mcm2-7 dimers [11,76].
Mcm10's modular architecture and lack of enzymatic activity suggest that it serves as a scaffold to orchestrate protein and DNA interactions within the replisome. Self-association would provide multiple points of contact between replication factors and DNA [37]. Indeed, Mcm10 is involved in multiple interactions, including but not limited to the replication and checkpoint clamps, PCNA and 9-1-1 [unpublished results and ref. 35], and pol a [16,17,20,33,34]. Protein-protein interactions could be mediated by the CC directly, similar to the interaction between Cdt1 and geminin [77,78,79]. Alternatively, dimerization could facilitate molecular interactions and recruiting proteins to the origin simply by increasing the number of possible binding sites on Mcm10 [37]. For example, the ID and CTD each bind DNA and pol a and could therefore be involved in a molecular hand-off, whereby Mcm10 is anchored to DNA via the ID while binding pol a at the CTD, and vice versa [34]. Additional Mcm10 subunits would enhance these interactions by increasing the number of ID and CTD present. Similarly, a parallel Mcm10 dimer could couple events on the leading and lagging strands or physically tether the helicase and pol a [16,18,27] while retaining the polarity necessary for fork progression. This would also explain why loss of the first 100 residues of scMcm10 confers such a strong sensitivity to hydroxyurea in the absence of the 9-1-1 clamp (Alver and Bielinsky, unpublished results).
Taken together, the dimerization or trimerization of xMcm10 in the absence of DNA reported here is consistent with previous work on spMcm10 [42], and the observation that three subunits of scMcm10 are bound to short ssDNA oligonucleotides, although these latter complexes were not shown to have the three-fold symmetry revealed in our crystal structure [21]. We did not find any evidence for Mcm10 hexamers, which were previously observed by EM of the human protein [41]. As discussed above we do not rule out a trimer of dimers, although this would not be consistent with the six-fold symmetrical EM structure reported.  (Table S1) Table S1. Although the precise masses cannot be accurately determined due to the complex nature of the sedimentation profile, the reduction of the 4S peak (marked with an asterisk) in the monomeric 2D mutant represents a significant difference from the WT. (TIF)