Structure and Mutagenesis Studies of the C-terminal Region of Licensing Factor Cdt1 Enable the Identification of Key Residues for Binding to Replicative Helicase Mcm Proteins

In eukaryotes, DNA replication is fired once in a single cell cycle before cell division starts to maintain stability of the genome. This event is tightly controlled by a series of proteins. Cdt1 is one of the licensing factors and is involved in recruiting replicative DNA helicase Mcm2–7 proteins into the pre-replicative complex together with Cdc6. In Cdt1, the C-terminal region serves as a binding site for Mcm2–7 proteins, although the details of these interactions remain largely unknown. Here, we report the structure of the region and the key residues for binding to Mcm proteins. We determined the solution structure of the C-terminal fragment, residues 450–557, of mouse Cdt1 by NMR. The structure consists of a winged-helix domain and shows unexpected similarity to those of the C-terminal domain of Cdc6 and the central fragment of Cdt1, thereby implying functional and evolutionary relationships. Structure-based mutagenesis and an in vitro binding assay enabled us to pinpoint the region that interacts with Mcm proteins. Moreover, by performing in vitro binding and budding yeast viability experiments, we showed that ∼45 residues located in the N-terminal direction of the structural region are equally crucial for recognizing Mcm proteins. Our data suggest the possibility that winged-helix domain plays a role as a common module to interact with replicative helicase in the DNA replication-licensing process.

In eukaryotes, DNA replication is fired once in a single cell cycle before cell division starts to maintain stability of the genome. This event is tightly controlled by a series of proteins. Cdt1 is one of the licensing factors and is involved in recruiting replicative DNA helicase Mcm2-7 proteins into the pre-replicative complex together with Cdc6. In Cdt1, the C-terminal region serves as a binding site for Mcm2-7 proteins, although the details of these interactions remain largely unknown. Here, we report the structure of the region and the key residues for binding to Mcm proteins. We determined the solution structure of the C-terminal fragment, residues 450 -557, of mouse Cdt1 by NMR. The structure consists of a winged-helix domain and shows unexpected similarity to those of the C-terminal domain of Cdc6 and the central fragment of Cdt1, thereby implying functionalandevolutionaryrelationships.Structure-basedmutagenesis and an in vitro binding assay enabled us to pinpoint the region that interacts with Mcm proteins. Moreover, by performing in vitro binding and budding yeast viability experiments, we showed that ϳ45 residues located in the N-terminal direction of the structural region are equally crucial for recognizing Mcm proteins. Our data suggest the possibility that winged-helix domain plays a role as a common module to interact with replicative helicase in the DNA replication-licensing process.
In eukaryotes, DNA replication is highly coordinated to retain the integrity of the genome. Whereas DNA replication in prokaryotes begins at a single site and stops at the end of the genome, eukaryotic genomes consist of multiple replication origins where DNA replication starts. These origins are synchronized so that they are activated only once in a single division cycle. A series of proteins, the origin recognition complex (ORC), 6 cell division cycle 6 homolog (Cdc6), chromatin licensing and DNA replication factor 1 (Cdt1), and minichromosome maintenance 2-7 (Mcm2-7) are known to play correlated roles in licensing (1)(2)(3)(4)(5). Oncogenic proliferation often causes abnormal expression of the proteins involved in the DNA-licensing process, thus emphasizing the importance of harmonious adjustments between these proteins (6). A complicated interaction network between these proteins has been reported (7), although the details at residue and atom levels remain largely unknown.
Formation of a pre-replication complex at each origin is the first event in the replication process. ORC proteins bind initially to each replication origin of DNA. The DNA sequences of the origins where ORC binds have not been identified, except for those in Saccharomyces cerevisiae (8), suggesting that there may be other factors involved in addition to the sequences (9). The DNA strand at the origin needs to be unpaired to begin replication; therefore, the existence of replicative helicase is essential. Mcm2-7 proteins are believed to function as the replicative helicase in eukaryotes. Each Mcm2-7 protein consists of a conserved AAAϩ ATPase type C-terminal helicase domain and a rather diverse N-terminal domain. The stoichiometry of Mcm2-7 while functioning remains unclear. Both hetero-hexameric and several additional complexes with different combinations of Mcm proteins have been found in vivo and in vitro (10). Meanwhile, reconstituted Mcm4/6/7 complex has helicase activity in vitro (11,12). The proteins that bridge between ORC and Mcm2-7 are Cdc6 and Cdt1. Cdc6, like Mcm proteins, belongs to the AAAϩ ATPase family and shows similarity to clamp loader proteins that load ring-shaped sliding clamps onto DNA (13). The result, Mcm2-7 forming a donut-like ring shape, led to the postulation that Cdc6 behaves as a clamp loader for Mcm2-7 (14,15).
Cdt1 was originally found in fission yeast (16), and its function as a factor involved in replication licensing was first characterized in Xenopus and fission yeast (17,18 Mcm2-7 both in vitro and in vivo (19 -22). Cdt1 can be divided  into three functional regions, namely, N-terminal, central, and  C-terminal regions, which have DNA-, geminin-, and Mcm2-7-binding activity, respectively (20,22). In addition to its role in binding to DNA, the N-terminal region is employed to interact with proliferating cell nuclear antigen (23) and Cdc7 and contains ubiquitination and acetylation sites. The function of Cdt1 is negatively controlled by a small protein termed geminin (24) and by the ubiquitin-mediated degradation pathway (25). In metazoans, the degree of sequence conservation in the C-terminal fragment of Cdt1 is higher than in the N-terminal and central regions (22). However, even in the C-terminal region, homology decreases abruptly in budding yeast, showing sequence identity of Ͻ20% compared with the mouse sequence.
In fact, S. cerevisiae Cdt1 could not be uncovered using only a simple database search (21). Of the three functional regions in Cdt1, structural information is available for the central region, which was determined by x-ray crystallography as a complex with geminin by Lee et al. (26). They also showed the long helical region of geminin is involved in hindering the interaction between Cdt1 and Mcm2-7.
Because of a resemblance in DNA replication initiation between archaea and eukaryotes, the data from archaea, particularly structural information, have assisted our understanding of the events in eukaryotes (27). The structures of ORC (28,29), Cdc6 (13), and Mcm (14) from archaea are very informative; however, it is impossible to deduce the function of Cdt1 using this strategy because of a lack of orthologues in archaea. So far, the regions of Cdt1 for Mcm2-7 binding have been characterized mainly by using deletion mutants (20,22). At a residuespecific level, to date, only one study has reported the importance of two conserved lysines (30).
To understand the function of Cdt1 in detail, we carried out a structural and mutagenesis study. In this study, we used multidimensional NMR spectroscopy to determine the structure of the C-terminal fragment (residues 450 -557) of mouse Cdt1. We then performed an in vitro binding assay with Mcm proteins with mutants prepared based on the NMR structure. Finally, we carried out in vitro and in vivo assays to demonstrate that the additional residues in the N-terminal direction from the structural region are equally important for the recognition of Mcm proteins.

EXPERIMENTAL PROCEDURES
DNA Cloning and Protein Purification for Structure Determination-DNA corresponding to the C-terminal fragment of mouse Cdt1 (residues 450 -557, mCdt1 C-WHD ) was cloned into a pGEX-6p-1 vector and then transformed into Escherichia coli BL21 strain. The fusion proteins were purified by GSH affinity chromatography. The N-terminal GST was removed by cleavage with recombinant rhinovirus 3C protease. Further purification steps using Q-Sepharose and Sephadex 75 column chromatography were applied. Finally, the buffer was replaced with 50 mM phosphate buffer, pH 6.5, for NMR measurement. Sitedirected mutagenesis was performed using the QuikChange mutagenesis kit (Stratagene) according to the manufacturer's instructions. All the mutants and the Mcm4/6/7 complex for the in vitro binding assay were cloned and purified using meth-ods identical to those described previously (22). We performed densitometric analysis for GST pulldown data by using Multi Gauge (Fujifilm) and quantified binding affinities between Cdt1 mutants and Mcm4/6/7 proteins.
Structure Calculation-Structures were calculated by CYANA 2.0 (34) and further refined by the AMBER 7 package (35). All the NOE cross-peaks were checked manually using NMRView (33) and assigned using the CANDID algorithm (36) of CYANA. Several tens of CANDID runs were carried out according to the criteria reported (37). A total of 1,826 meaningful NOE upper distance restraints was obtained by CANDID (415 intra-residual, 512 sequential, 523 medium range, and 375 long range). Backbone torsion angle restraints of 159 were derived by the TALOS program (38) and used for all the calculations. Finally, 100 structures that did not show significant violations against experiment restraints were generated using 15,000 steps torsion angle dynamics by CYANA and further refined by AMBER with an all-atom force field. To approximate solvent effects, the generalized Born implicit solvent model was used (39). The best 20 structures were selected and analyzed using AQUA and PROCHECK-NMR software (40). The structure closest to the mean coordinate was selected as a representative. The coordinates and NMR restraints for structure calculation were deposited in the PDB data base (ID: 2RQQ). All the figures were created by using PyMOL software (41).

Preparation of Stable Protein Fragment for NMR Study-
Initially the region containing residues 407-557 of mCdt1 (mCdt1 CTF ) was chosen for the NMR study. However, the residues 407-449 (mCdt1 C-N ) of the region were sensitive to protease and easily degraded at the protein concentration needed to measure three-dimensional NMR experiments in the required time. Therefore, we focused our attention on the region containing residues 450 -557 (mCdt1 C-WHD ) for structure determination, even though the region containing residues 407-557 is needed for binding to Mcm proteins. The construct of residues 407-557 was used instead for the in vitro binding assay. The sample of mCdt1 C-WHD was stable for several weeks at high concentration. The results of size-exclusion chromatography indicated that mCdt1 C-WHD exists as a monomer (data not shown).
Winged-helix Domain in C-terminal Fragment of Cdt1-Final structures are well refined, with a backbone and heavy atom root mean square deviation (r.m.s.d.) in residues Pro-454 to Glu-555 of 0.35 Å and 0.87 Å, respectively. The structures also have good geometry with 92.6% of residues in the most favored region of the Ramachandran plot (Table 1 and Fig. 1A). N-terminal residues to Cys-453 and C-terminal residues of Gly-556 and Leu-557 are not well converged in the ensemble due to absence of long range NOE. The secondary structures comprise four ␣-helices (H1, residues 455-480; H2, 487-496; H3, 504 -517; and H4, 541-552) and three ␤-strands (S1, 485-486; S2, 522-526; and S3, 531-535). Pro-467 is positioned in the middle of H1, causing a slight kink of H1 around the residue. The topology of the secondary structures, including the wing region between S2 and S3 (W1, 527-530) is H1-S1-H2-H3-S2-W1-S3-H4 and constitutes a typical winged-helix domain (hereafter WHD) (Fig. 1B). Residues located at the interior layer of each helix contribute to stabilization of the helices by forming hydrophobic interactions (Fig. 1C). Between helices H1 and H4, pairs of residues, Leu-463 and Val-552, and Leu-466 and Leu-548, are in contact with each other, respectively. For helices H1, H3, and H4, three residues, Leu-513, Leu-514, and Leu-517, of H3 interact with Leu-466 of H1 and Leu-548 of H4, respectively. Between H1 and H2, residues Leu-469 and Met-494 also form hydrophobic interactions. The Trp-521 on the loop connecting H3 and S2 stabilizes the secondary structures by contact with Ala-470 and Leu-473 of H1, and Asp-536, Val-539, and Leu-544 of H4. In addition to the hydrophobic interactions described, a hydrogen bond between N⑀1 of Trp-521 and backbone carbonyl of Asp-536 participates in tightening of the sec-ondary structures. The presence of this bond is supported by its observation in all the final 20 NMR structures. Intriguingly, the mCdt1 C-WHD sequence contains 19 leucines, with many of these residues being involved in hydrophobic interactions between secondary structures. Surface electrostatic charges reveal that, on the outward face from the end of helix H1 and the next loop connecting helices H1 and H2 and the loop between S3 and H3, there is a positively charged patch consisting of Arg-471, Arg-474, Arg-481, and Lys-537 of mCdt1 C-WHD . With the exception of this region, there is no noticeable charged patch (Fig. 1D).
Sequence Alignment and Conservation of Cdt1 C-WHD -To know sequence conservation from the viewpoint of function and evolution, the regions corresponding to mCdt1 C-WHD (residues 450 -557) were extracted from six representative species and aligned using MUSCLE software (42 (Fig. 1E). The sequence identities against mCdt1 C-WHD are 83% for human Cdt1 C-WHD , 59% for Xenopus, 44% for Drosophila, 24% for C. elegans, 20% for S. pombe, and 19% for S. cerevisiae. Two regions between S3 and H4 in C. elegans are not well aligned, and two yeast sequences show lower similarity than those from metazoan. In particular, the H1 region of S. cerevisiae is not well aligned with the others. Nonetheless, those residues that have important roles in forming the tertiary structure are well converged in ensemble conformers of mCdt1 C-WHD and are also conserved in sequence alignment. This suggests all the Cdt1 C-WHD s have similar structures.
Structures Sharing Similarities with Cdt1 C-WHD -The DALI server (43) was used to search for the protein structures similar to that of mCdt1 C-WHD . This search retrieved 562 WHD structures with Z-scores higher than 2.0 from the PDB data base in March of 2009 (44). Strikingly, the closest similarity was found in the structure of the C-terminal domain of the archael Cdc6 orthologue (PDB: 1FNN, Z score: 6.8, r.m.s.d.: 3.0 Å (over 81 residues)). WHDs in proteins, zinc uptake regulation protein Furb (2O03, Z score: 6.6, r.m.s.d.: 2.7 (76 residues)), Staphylococcus Sars protein (1P4X, Z score: 6.6, r.m.s.d.: 4.2 (76 residues)), and viral protein F93 (2CO5, Z-score: 6.5, r.m.s.d.: 2.8 (76 residues)), were followed. Despite not being included in the structures of closest similarity, it should be noted that the central region of mouse Cdt1 (residues 179 -365, mCdt1 M-WHD ) from the complex structure with geminin (PDB: 1WLQ) were also retrieved from the DALI server and had a Z-score of 4.9 and r.m.s.d. of 3.4 Å in 93 residues. To obtain detailed information from similar structures, we extracted 21 sequences in the structures with Z-scores higher than 6.0. Of these, considering sequence redundancies, we selected 16 and aligned the sequences based on the structural overlays. For comparison, we included the sequence from mCdt1 M-WHD . In each sequence any amino acid that had a corresponding mCdt1 C-WHD amino acid was extracted (Fig. 2A). The alignment revealed that residues in the loop regions of mCdt1 C-WHD were less aligned, suggesting that the loop conformations in mCdt1 C-WHD and other WHDs are diverse. Of the secondary structures, the helix H4 is hardly aligned even in the C-terminal domain structure of Cdc6 (1FNN) (Fig. 2). It is clear that the properties of hydrophobic residues involved in maintaining the tertiary fold in mCdt1 C-WHD are still conserved in the other WHDs despite overall lower sequence similarities, implying that their roles are to maintain tertiary folds. In the next section, we discuss WHDs in detail. mCdt1 M-WHD (1WLQ) has additional residues in N and C termini and longer inserted loop between S2 and S3 (Fig. 2). However, secondary structures in winged-helix regions are similar to each other and other WHDs in length and positions (Fig. 2).
Structure-based sequence alignment reveals 16% sequence identity between the two domains mCdt1 M-WHD and mCdt1 C-WHD , whereas sequence-only alignment could not align correctly even the secondary structural regions. Under "Discussion," we provide a detailed review of these findings from an evolutionary viewpoint.
In Vitro Binding Assays with C-terminal Regions of Cdt1 and Mcm4/6/7-In a previous experiment, we showed that the C-terminal fragment of Cdt1 binds to Mcm proteins (22). In the present study, we characterized the region in greater detail (Fig.  3, A-C). Because reconstituting Mcm2-7 in vitro is not possi- In three ␤-strands, S1 (485-486) is drawn in pink, S2 (522-526) in purple, and S3 (531-535) in cyan. The wing region between S2 and S3 is labeled as W1. B, a ribbon diagram of the representative NMR structure. The secondary structures are colored using the same code as in A. C, residues involved in stabilizing inter secondary structures are drawn as stick models. Asp-536 whose backbone carbonyl forms hydrogen bonds with N⑀1 of Trp 521 is labeled in red, and the others are in black. D, electrostatic surface calculated from APBS (56) is drawn in the range of Ϫ4k B /T (red) to 4k B /T (blue). All the figures of the structure were created by using PyMOL (41). All the coordinates were aligned to have the same direction. E, sequence alignment of Cdt1 C-WHD . Residues that correspond to 450 -557 of mouse Cdt1 were extracted in human, Xenopus, Drosophila, C. elegans, S. pombe, and S. cerevisiae and aligned using MUSCLE (42). The number of starting residues in each sequence line is written in parentheses. Secondary structures are drawn on the sequences. Residues showing identities of Ͼ70% are colored black, and those Ͼ50% are gray. ble, we selected the Mcm4/6/7 complex based on reports that reconstituted Mcm4/6/7 has DNA helicase activity (11,12,30). Initially, the region of mCdt1 that acted as a Mcm4/6/7 binding site was narrowed to a fragment of residues 407-557. This region showed comparable binding affinity toward Mcm4/6/7 to that of wild type. However, the fragment, whose structure was determined in this study (mCdt1 C-WHD , residues 450 -557), hardly bound to Mcm4/6/7 (Fig. 3, B and C, lane and bar  3). We selected only the protease-sensitive extra region of 407-458, but it also did not show binding affinity (Fig. 3, B and C,  lane and bar 5). We further prepared two additional constructs, the former or the latter half of mCdt1 C-N and the region of mCdt1 C-WHD (407-423/451-557 and 426 -450/451-557), and investigated their activities. Both constructs bound little with Mcm4/6/7 (Fig. 3, B and C, lanes and bars 2 and 5). Altogether, the existence of both residues 407-449 and 450 -557 are indispensable for binding to Mcm4/6/7. The PSIPRED software (45) predicted the secondary structure of residues 424 -448 as a helix, whereas the first part (407-423) that includes many prolines is likely to form neither helix nor strand. Judging from the sensitivity to protease, the region of mCdt1 C-N probably does not make stable interaction with the rest of the region in the construct 450 -557. Ferenbach et al. reported that the region required for Xenopus Cdt1 to bind to Mcm2/4/6/7 consists of residues 447-620 (20). However, residues 407-557 that we identified in mouse Cdt1 correspond to residues 469 -619 in Xenopus, suggesting that the residues 447-468 of Xenopus Cdt1 are not necessary for binding to Mcm2/4/6/7.

Regions of Winged-helix Domain for Binding to Mcm
Proteins-Both BLAST and PSI-BLAST searches when queried with residues 450 -557 of mouse Cdt1 could not retrieve meaningful homologous sequences other than those derived from Cdt1 of various species. Currently available bioinformatics tools do not list noticeable consensus motifs from the sequence (data not shown). As a result, knowing only the sequence is not informative for speculating the functional residues of Cdt1 C-WHD . Therefore, we combined structural information with sequence data. This procedure is based on the idea that the residues important for scaffold or function in protein are restrained during evolution. Therefore, searching for residues that are identi- The constructs having point mutations were generated with GST-mCdt1 (407-557). C, the amounts of bound Mcm4/6/7 and errors were quantified by repeating the reaction under identical conditions. D, surface-exposed and conserved residues that form a patch are shown. The colors in the order of blue to red indicate the lower to higher degrees of conservation. The first putative protein-protein interaction site is marked by a dashed circle, and the second site is marked by a continuous circle.
cal or have conserved properties on the exposed surface is a reasonable approach for deducing the functional portions from a clueless sequence. For this purpose, we used the WHISCY server (46). It selects only patches formed on conserved and solvent-accessible residues. In addition to the sequences used for alignment (Fig. 1E) and the current structure, we added nine more sequences from metazoans to increase precision. Two patches were scored as potential protein-protein interfaces (Fig. 3D). The first (dashed circle) was the N-terminal three residues of Met 450 , Thr 451 , and Arg 452 , and the other (continuous circle) was the region containing the end of helix H1 and its consecutive loop, which include Arg 471 , Arg 474 , Val 478 , and Arg 481 . Considering protease sensitivity around the region of residues 450 -452, the N-terminal three residues will be probably flexible even in the intact condition. The latter region has both hydrophilic and hydrophobic characteristics, overlapped partially with the positively charged surface of Cdt1 C-WHD , and is more conserved (Fig. 1). To identify the functional residues for Mcm2-7 proteins, we carried out an in vitro binding assay with Mcm4/6/7 proteins with a series of mutants, each of which has two to four amino acid substitutions in the region of Cdt1 C-WHD . Indeed, triple and quadruple mutants, R471Q/ R474Q/V478A, R474Q/V478A/R481Q, and R471Q/R474Q/ V478A/R481Q, showed markedly reduced binding activities to Mcm4/6/7 (Fig. 3, B and C, lanes and bars 9 -11). Considering that R471Q/R474Q influenced a little (Fig. 3B, lane 8), Val 478 will probably play the central role in interacting with Mcm proteins, further supplemented by electrostatic contacts with basic residues. In contrast, other mutants containing mutations in residues Met 488 , Leu 524 , Ile 527 , and Tyr 532 , chosen simply from sequence conservation, resulted in marginal changes in binding affinity compared with that observed in the wild type (Fig. 3, B  and C, lanes and bars 6 and 7).
Yeast Viability Assay with Mutants at Additional Fragment-To better characterize the flanking region, Cdt1 C-N (residues 407-449 in mCdt1), especially the highly conserved second half part, Leu 434 to Ala 448 , we performed a yeast viability assay. The alignment of eukaryotic sequences, including yeast sequences, reveals high sequence conservation of the region (Fig. 4A), which is distinct from Cdt1 C-WHD where sequence similarity is low between mammals and yeast (Fig. 1E). Hydrophobic residues (Leu, Met, and Ile) and basic residues (Arg) are particularly conserved. We undertook a systematic mutagenesis of the region for Cdt1, Tah11 (hereafter scCDT1), in S. cerevisiae and then observed cell viability. Note that scCDT1 is an essential gene. The promoter of endogenous scCDT1 gene in W303a was replaced with the GAL1 promoter. Integration of the mutant scCDT1 gene with the wild-type scCDT1 promoter and cyc1 terminator at the LEU2 locus allowed the strain to produce mutant proteins in the glucose medium. At first, a C-terminal truncated mutant (Fig. 4B, row 2) or mutants that lack the conserved region (Fig. 4B, rows 3 and 4) were checked for growth defects. Both deletion constructs were unable to support growth in glucose medium, while full-length scCDT1 (Fig.  4B-1) could be substituted for endogenous protein. These results indicate again that both regions of scCDT1, which correspond to Cdt1 C-WHD and highly conserved Cdt1 C-N , are essential for scCDT1 function. To examine the functional effects of substitutions in the conserved region, 19 alleles of scCDT1, each of which contains amino acid substitution(s), were then generated by site-directed mutagenesis (Fig. 4B, rows [5][6][7][8][9][10][11][12][13][14][15][16][17][18][19][20][21][22][23]. We substituted Ala for conserved Leu and Met residues, and charged residues, Arg and Asp, were replaced with neutral amino acid Asn. All the single amino acid substitutions proved to be viable. In contrast and intriguingly, five of eight double amino acid substitutions were nonviable. Also one of two triple amino acid changes for Arg was shown to be nonviable. Taken together, the existences of the hydrophobic residues at the positions of Leu 483 /Met 484 /Leu 487 in scCdt1 and Leu 434 /Leu 435 / Ile 438 in mCdt1, and the basic residues at Arg 486 /Arg 488 /Arg 490 in scCdt1 CTF and Arg 437 /Arg 439 /Lys 441 in mCdt1 CTF would be important for Cdt1 function. This was confirmed by in vitro GST pulldown experiments too. Whereas a single point mutant showed a nominal change (Fig. 3, B and C, lanes and bars 13), double and triple mutations in these residues caused the reduced affinities toward Mcm4/6/7 proteins (Fig. 3, B and C, lanes and bars [15][16][17], which is qualitatively in agreement with the results of yeast viability assay. You and Masai (30) reported that a double point mutant of mCdt1, K441A/K445A, failed to generate a complex with Mcm proteins. Lys 441 is one of those residues that we identified as an important residue in this study. Interestingly, the combined feature of hydrophobic and basic residues (Leu 434 , Leu 435 , Arg 437 , Ile 438 , Arg 439 , and Lys 441 ) is reminiscent of the residues pinpointed in mCdt1 C-WHD (Arg 471 , Arg 474 , Phe 477 , Val 478 , and Arg 481 ) for binding to Mcm proteins, likely suggesting similar contacting mode. On the other hand, the underlying role of the proline-rich region, residues 407-423, remains a subject for future studies.

DISCUSSION
WHD is a typical DNA recognition motif (47). When WHD binds to double strand DNA, the interactions occur at H3, S2, S3, and W1 in the majority of cases. H3 is the primary contact site and W1 acts as an additional and supportive site. To the best of our knowledge, a reversal in these roles of H3 and W1 has been reported once in an exceptional case (47). In addition to H3 and W1, the existence of positively charged residues at H3, S2, S3, and W1 are indispensable for DNA recognition. Cdt1 CTF binds to neither double nor single strand DNA (22). To determine the differences between mCdt1 C-WHD and other WHDs, we superimposed the coordinate of mCdt1 C-WHD into each WHD position of known DNA-protein complex structures and investigated the hypothetical contact sites between mCdt1 C-WHD and DNA. In contrast to WHDs that bind to DNA, the lengths of H1 or H4 of mCdt1 C-WHD are considerably longer, with these helices being more inclined to the direction of the plane that consisted of H3 and S2, S3, and W1. Because of these features, one or both of H1 and H4 of mCdt1 C-WHD would encounter steric clashes with DNA, if it binds to DNA in the same manner as the other WHDs do. In fact, as shown in the structure-based sequence alignment ( Fig. 2A), the residues in H1 and H4 are not fully aligned, meaning that they are unique to mCdt1 C-WHD . Furthermore, there is no particularly charged patch on the surface regions of H3, S2, S3, and W1 (Fig. 1C). It should be noted that the WHDs that share the closest similarities with mCdt1 C-WHD , those from PDB codes of 1FNN, 2O03, 1P4X, and 2CO5, are known to participate in protein-protein interactions, indicating that the structural features of WHD in Cdt1 C-WHD are distinct from other DNA binding domains. Taken together, the inability of Cdt1 C-WHD in DNA binding is consistent with its structural features, including long H1 and H4 helices and surface charge.
Accumulated data have revealed that some WHDs serve as a protein-protein interaction module. We need more examples to characterize in detail the region involved in protein-protein interaction, but at least in two such cases, the apoptotic protease-activating factor 1 (48) and ESCRT-II-III system (49), protein-protein interactions occur at similar areas of WHD to that identified as the Mcm2-7 binding site. The question then is, do other WHDs involved in DNA replication licensing use the same region to interact with Mcm2-7? The WHD of archael Cdc6 engages in interaction with Mcm proteins as well as DNA (50). As shown with typical DNA binding WHDs, H3 is a major DNA recognition site in Cdc6 WHD (51,52), but the region that interacts with Mcm proteins has not been characterized. The structural resemblance between WHDs of archael Cdc6 and mCdt1 C-WHD and the high degree of sequence homology between archael and eukaryotic Cdc6s (13) tempt us to presume that similar regions in WHDs are employed for Mcm proteins. In yeast, the Cdc6 WHD can interact with both DNA and Mcm2-7 simultaneously (53). When considering the steric effects and the large molecular sizes in DNA replication licensing protein, the prime candidate site of Cdc6 WHD for binding to Mcm2-7 will be placed at the opposite side from the H3 DNA binding site and likely include the region identified as the Mcm binding site in Cdt1 C-WHD . Yet there has been no report that Cdt1 Central has a helicase-binding activity. Nevertheless, both Cdt1 Central and Cdt1 CTF are required to load Mcm2-7 for licensing activity (20). Moreover, Lee et al. reported that the interaction between Cdt1 and Mcm2-7 is inhibited by steric hindrance caused by a contact between geminin that binds to the central region of Cdt1 and Mcm proteins (26). This raises the possibility that two fragments of Cdt1 Central and Cdt1 CTF are located in close proximity, at least when Cdt1 interacts with Mcm2-7, implying potential cooperation between the two regions.
Similarities in folding, sequence, and interacting site with partner protein led us to postulate that the central and C-terminal regions of Cdt1 were duplicated in tandem from a domain and have evolved together. This type of duplication occurs frequently in eukaryotic proteins and confers functional diversities (54). In addition to the region of WHD in the crystal structure (residues 179 -365 of mouse Cdt1), Cdt1 Central needs its flanking sequence to bind to geminin in vivo (20). Interest-ingly, this sequence is located in the N-terminal direction from the structural region of Cdt1 M-WHD and corresponds to residues 193-243 in Xenopus and 121-181 in mouse Cdt1, respectively. These positions are reminiscent of the relative positions between the protease-sensitive fragment, residues 407-448, and the structural part, residues 450 -557, of Cdt1 C-WHD in the current study. However, the central regions of Cdt1 are under weaker evolutionary pressure, because Cdt1 Central has lower sequence homology across species than those found in Cdt1 CTF (22). Evolutionary trends in Cdt1 fragments are also reflected by the presence or diverseness of their binding partners. Geminin, the partner of Cdt1 Central , is missing or elusive in budding yeast, whereas the presence of Mcm2-7 with which Cdt1 CTF mainly interacts is highly conserved in eukaryotes, consistent with evolutionary conservation in the fragments. In addition to the clues and questions required to understand the interaction between Cdt1 and Mcm2-7 raised by our data, it would be worthwhile studying the roles of WHDs in Cdt1 CTF , Cdt1 Central , and Cdc6 as a common evolutionary module.
In summary, we have shown that the C-terminal fragment of Cdt1 forms a WHD, but its structural features are different from other typical DNA binding WHDs. Although the structural region determined in the current study is insufficient to bind to Mcm proteins alone, we were able to characterize the residues of the WHD that are involved in interactions with replicative helicase. We also showed an additional region in the N-terminal direction of the WHD is equally important. How these two parts work cooperatively remains unanswered, and further experiments are required to elucidate this mechanism. Eukaryotic DNA replication initiation machinery is large and complicated, and presently, it is not possible to reconstitute the system in vitro. Therefore, dissecting each component and examining the parts is a reasonable strategy. We believe that our data will be a valuable addition to understanding the process and evolution of DNA replication licensing in eukaryotes.