Replication factors MCM2 and ORC1 interact with the histone acetyltransferase HBO1.

The minichromosome maintenance (MCM) proteins, together with the origin recognition complex (ORC) proteins and Cdc6, play an essential role in eukaryotic DNA replication through the formation of a pre-replication complex at origins of replication. We used a yeast two-hybrid screen to identify MCM2-interacting proteins. One of the proteins we identified is identical to the ORC1-interacting protein termed HBO1. HBO1 belongs to the MYST family, characterized by a highly conserved C2HC zinc finger and a putative histone acetyltransferase domain. Biochemical studies confirmed the interaction between MCM2 and HBO1 in vitro and in vivo. An N-terminal domain of MCM2 is necessary for binding to HBO1, and a C2HC zinc finger of HBO1 is essential for binding to MCM2. A reverse yeast two-hybrid selection was performed to isolate an allele of MCM2 that is defective for interaction with HBO1; this allele was then used to isolate a suppressor mutant of HBO1 that restores the interaction with the mutant MCM2. This suppressor mutation was located in the HBO1 zinc finger. Taken together, these findings strongly suggest that the interaction between MCM2 and HBO1 is direct and mediated by the C2HC zinc finger of HBO1. The biochemical and genetic interactions of MYST family protein HBO1 with two components of the replication apparatus, MCM2 and ORC1, suggest that HBO1-associated HAT activity may play a direct role in the process of DNA replication.

Eukaryotic DNA replication is a tightly regulated process that is strictly coupled to cell cycle progression, ensuring that DNA is replicated only during S phase and that each origin is used only once per cell cycle. This precise cell cycle coordination is the result of both positive and negative regulation of replication origin function. Genetic and biochemical studies in yeast and metazoans suggest that the initiation of DNA synthesis is a complex, multistep process that requires the participation of many proteins (1)(2)(3). This process involves the binding of the origin recognition complex (ORC) 1 to replication origins (4,5), the recruitment of Cdc6 and the six MCM proteins (MCM2-7; MCM for minichromosome maintenance) to form the pre-replicative complex (pre-RC) (6,7), and the activation of the pre-RC by protein kinases to initiate DNA synthesis (8).
MCM proteins were revealed to be involved in DNA replication as the result of genetic screens for mutants defective in progression through the cell division cycle (9 -12) or the replication of minichromosomes (13)(14)(15). Initial characterization of three genes, mcm2 (14), mcm3 (16), and mcm5/cdc46 (9,17), implicated each in DNA replication and showed they were related in sequence. This family rapidly grew to encompass the Schizosaccharomyces pombe mcm4/cdc21 ϩ (18,19) and mcm6/mis5 ϩ genes (15), and the Saccharomyces cerevisiae mcm7/cdc47 gene (20). Analysis of the complete S. cerevisiae genome sequence indicates that there are six MCM-encoding genes, and homologs of the MCM2-7 proteins have since been identified in all eukaryotes from yeast to humans (for reviews see Refs. [21][22][23]. MCM proteins play a critical role in the function and regulation of DNA replication. A loss-of-function mutation in any mcm gene impairs replication initiation in S. pombe, indicating that each MCM protein has a unique and essential function in DNA replication (24,25). Additionally, in mammalian cells microinjection of anti-MCM2 antibodies (26), anti-MCM3 antibodies (27), or antisense oligomers against MCM7 mRNA (28) blocks DNA replication. In a Xenopus laevis egg extract in vitro DNA replication system, immunodepletion of MCM3 protein prevents DNA synthesis (29,30). In Drosophila, mutations in the mcm2 or dpa/mcm4 genes inhibit cell proliferation and are homozygous lethal, with mutants showing prolonged S phases (31,32). Although MCM proteins are essential for normal DNA replication, it is not yet clear to what extent MCMs have a unique role in determining replication competence of chromosomal DNA or whether additional factors are involved in this process.
In a search for novel MCM-interacting factors, we identified a protein that interacts with the MCM2 subunit of the MCM complex. This protein was initially described as interacting with another component of the pre-RC, the ORC1 protein, and thus termed Histone acetyltransferase Binding to ORC1 (HBO1) (33). HBO1 shares sequence similarity with the MYST subfamily of histone acetyltransferases. MYST subfamily members include the human proteins MOZ (monocytic leukemia zinc finger protein) (34), MORF (MOZ-related factor) (35), and Tip60 (HIV Tat-interacting protein of 60 kDa) (36), Drosophila MOF (males-absent on the first) (37), and yeast proteins SAS2 and SAS3 (something about silencing) (38,39), and Esa1 (essential SAS family acetyltransferase) (40). MYST subfamily proteins are characterized by a highly conserved ϳ270 amino acid domain that consists of an atypical C 2 HC zinc finger motif and a putative histone acetyltransferase (HAT) domain.
In this study, we demonstrate that MCM2 interacts specifically with HBO1 both in vitro and in vivo and that this interaction is conserved in metazoans. Furthermore, we map the interacting domains of both MCM2 and HBO1 and describe a role for the C 2 HC zinc finger of HBO1 in this interaction. We also identify an interaction-defective allele of MCM2, as well as a compensating second site suppressor mutation in HBO1 that restores the interaction with the defective MCM2. These results indicate that two components of the pre-RC, MCM2 and ORC1, interact with a putative histone acetyltransferase and suggest a direct role for HAT activity in the process of DNA replication.
Construction of Plasmid DNAs-Full-length HBO1 cDNA was constructed in multiple steps. An SphI-HindIII HBO1 cDNA fragment was excised from clone TB305 and inserted into the corresponding sites of pUC119 to generate pUC119-HBO1(SphI-HindIII). The 5Ј region of HBO1 was amplified from a human placental cDNA library (Match-Maker; CLONTECH) by PCR using gene-specific primers TB27 (5Ј-GATCTCTAGACATATGCCGCGAAGGAAGAGGAATGCAG-3Ј; introduces XbaI and NdeI sites at 5Ј end of cDNA) and TB28 (5Ј-CTCTCCGTTGGTGCTGGTGCCTG-3Ј). This PCR fragment was digested with XbaI and SphI, and subcloned into the same sites of pUC119-HBO1(SphI-HindIII) to generate pUC119-HBO1. The nucleotide sequence of the PCR-amplified region of HBO1 was verified by sequencing. The HBO1 cDNA was subsequently excised using XbaI and AvrII and subcloned into the XbaI site of pBlueScript II SK ϩ (Stratagene, La Jolla, CA) in both orientations, creating pBS-HBO1(fwd and rev).
pAdTrack-HA 3 MCM2 was constructed in multiple steps. The mouse MCM2 cDNA was excised from pBS-mMCM2 (generous gift of H. Kimura, University of Oxford, Oxford, UK) as a KpnI-NotI fragment and inserted into the corresponding sites of pGEM-CMV (Nevins et al. (46)), generating pREC-mMCM2. A NotI site was inserted immediately following the mMCM2 initiating methionine in a PCR using primers MCM2-NotI (5Ј-CGGGGTACCATGAGCGGCCGCGCGGAGTCTTCT-GAGTCT-3Ј, initiator codon in bold and KpnI and NotI sites underlined) and MCM2-427 (5Ј-CTGCCAGCCTCACGGTCCCG-3Ј) which amplify nts 51-427 of the mMCM2 cDNA. The PCR product was digested with KpnI and BspEI and used to replace the corresponding fragment in pREC-mMCM2 to generate pREC-mMCM2-NotI. A fragment encoding three tandem copies of the HA epitope was released from pMT252 (generous gift of M. Tyers, Mount Sinai Hospital, Toronto, Canada) by digestion with NotI and inserted into the 5Ј NotI site of partially digested pREC-mMCM2-NotI to generate pREC-HA 3 mMCM2. The entire PCR-amplified region of the tagged mMCM2 was sequenced to confirm the construction and orientation. pAdTrack-HA 3 MCM2 was constructed by excising the tagged MCM2 construct from pREC-HA 3 MCM2 as an Asp718-XbaI fragment and inserting it into the corresponding sites of pAdTrack-CMV. pAdTrack-hORC1 was constructed by excising human ORC1 from pcDNA3-hORC1 (44) as a HindII-SmaI fragment and inserting into the HindIII-EcoRV sites of pAdTrack-CMV. AdTrack-AU5-HBO1 was constructed as follows. An N-terminal AU5 (TDFYLK) epitope tag was added to the HBO1 cDNA by insertion of annealed oligonucleotides TB43 (5Ј-CTAGACCATGGCCACCGACT-TCTACCTGAGGTT-3Ј, initiating codon in bold and XbaI site underlined) and TB44 (5Ј-TAAACTTCAGGTAGAAGTCGGTGGCCATGGT-3Ј) into the XbaI and NdeI sites of pBS-HBO1(rev), generating pBS-AU5-HBO1. An Asp718-NotI fragment was excised from pBS-AU5-HBO1 and ligated into the corresponding sites of pAdTrack-CMV, creating AdTrack-AU5-HBO1.
Cells and Viruses-C33A cervical carcinoma cells were grown in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum and infected in 1.5 ml (per 100-mm plate) of Dulbecco's modified Eagle's medium containing 25 mM HEPES (pH 7.5) for 75 min at 37°C. Following infection, Dulbecco's modified Eagle's medium plus 10% fetal bovine serum was added, and cells were incubated an additional 18 h prior to harvest. AdMCM2 (HA-tagged mouse MCM2), AdORC1 (human ORC1), and AdHBO1 (AU5-tagged human HBO1) recombinant adenoviruses were generated by the method of He et al. (45) after subcloning into pAdTrack-CMV shuttle vector. Adenoviral stocks were maintained as described (46) and purified by cesium chloride density gradient centrifugation.
Yeast Two-hybrid Liquid ␤-Galactosidase Assays-Quantitative twohybrid liquid culture ␤-galactosidase assays (47) were performed in yeast strains Y190 or MAV103. Expressions levels of the GAD and GBD fusion proteins were controlled by Western blot analysis for each experiment. Experiments were performed a minimum of three times. Values shown are representative results from individual experiments.
Chromatin-binding Assays-C33A cells were harvested by trypsinization and washed with phosphate-buffered saline. Cell pellets were resuspended to a final volume of 1 ml of CSK ϩ buffer (10 mM Pipes (pH 7), 100 mM NaCl, 300 mM sucrose, 3 mM MgCl 2 , 0.5% Triton X-100, 1 mM ATP, 0.1 mM PMSF, 1 g/ml aprotinin, 1 g/ml leupeptin, 1 g/ml pepstatin, 1 mM sodium orthovanadate, 25 mM ␤-glycerol phosphate). Fifty microliters of whole cell lysate (5%) was reserved for total protein control. Lysates were incubated on ice for 20 min, and then centrifuged at low speed for (1,000 ϫ g) for 5 min at 4°C. Chromatin pellets were washed once with 1 ml of CSK ϩ buffer for 5 min on ice and centrifuged again (1,000 ϫ g). Chromatin-containing pellets were resuspended in SDS sample buffer, resolved by SDS-PAGE, transferred to nitrocellulose membrane, and analyzed by immunoblotting.

RESULTS
Isolation of HBO1 as an MCM2-interacting Protein-To gain further insight into MCM function, we used a yeast two-hybrid system to screen a human placental cDNA library for proteins that interact with a Gal4 DNA-binding domain (GBD) fusion encoding the mouse MCM2 cDNA. Nine clones were identified by histidine nutritional selection and ␤-galactosidase activity. Sequence analysis revealed that two of these clones (TB305 and TB310) were identical to the C terminus (aa 218 -611) of a protein recently identified as interacting with ORC1, termed HBO1 (33). HBO1 was subsequently shown to interact with human androgen receptor and modulate androgen receptor transcriptional activity in a ligand-dependent manner (48). HBO1 is a 611-amino acid protein with a predicted molecular mass of 70 kDa (estimated pI 9.2) and contains a unique Nterminal serine-rich region (22% in aa 1-167) and a conserved 270-amino acid C-terminal domain characteristic of the MYST family of proteins (Fig. 1A) (34). The MYST domain contains a putative acetyl-coenzyme A (acetyl-CoA)-binding domain (Fig.  1B) and an atypical C 2 HC zinc finger. Throughout the MYST domain, HBO1 is highly conserved (ϳ40-65% identical and 60-80% similar) to several other MYST family proteins (Fig. 1A).
HBO1 Interacts with MCM2 and ORC1 in Vitro and in Vivo-To confirm the HBO1 interaction with MCM2, we con-structed full-length HBO1-(1-611) as a Gal4 activation domain (GAD) fusion and compared its binding activities to the Cterminal clone isolated in the two-hybrid screen (HBO1-(218 -611)). To generate a full-length HBO1 cDNA, a 0.7-kilobase N-terminal fragment of HBO1 was amplified from a human placental cDNA library using gene-specific primers and appended to clone TB305. As seen in Fig. 2A, the C-terminal clone of HBO1-(218 -611) specifically interacts with GBD fusions of mouse (GBD-mMCM2) and Drosophila MCM2 (GBD-dMCM2), as well as human (GBD-hORC1) and Drosophila ORC1 (GBD-dORC1), resulting in activation of the lacZ gene and production of ␤-galactosidase activity. Interestingly, whereas the fulllength GAD-HBO1 interacts relatively well with GBD-hORC1, it does not appear to produce ␤-galactosidase activity when coexpressed with the GAD-mMCM2, -dMCM2, and -dORC1 fusion proteins. Although this may be due to an inability of the full-length GAD-HBO1-(1-611) fusion protein to bind to mMCM2, dMCM2, and dORC1 in this system, it may also reflect the presence of an intrinsic transcriptional repression domain located in the N-terminal region of HBO1 (48). This N-terminal transcriptional repression activity may prevent detectable interactions when monitored in a transcription-based assay such as the yeast two-hybrid system and may become masked or inactivated when coexpressed with GBD-hORC1. We sought to examine further the differential binding activities observed in the yeast two-hybrid system and to determine whether full-length HBO1 is capable of interacting with MCM2 and ORC1. To accomplish this we performed GST pull-down experiments with in vitro synthesized, [ 35 S]methionine-labeled full-length HBO1 protein and bacterially produced GST fusion proteins GST-mMCM2, -dMCM2, -hORC1, and -dORC1. As seen in Fig. 2B, a strong and specific retention of HBO1-(1-611) protein was observed for the samples containing GST-mMCM2, -hORC1, and -dORC1, with a weaker retention observed for GST-dMCM2. These findings demonstrate that fulllength HBO1 is capable of interacting with MCM2 and ORC1 in vitro and indicate that the interactions of HBO1 with MCM2 and ORC1 are evolutionarily conserved.
To confirm the interactions of HBO1 with MCM2 and ORC1 under physiological conditions, we constructed recombinant adenoviruses expressing N-terminal AU5 epitope-tagged version of full-length human HBO1 (AdHBO1), N-terminal HAtagged full-length mouse MCM2 (AdMCM2), and full-length human ORC1 (AdORC1). Human C33A cervical carcinoma cells were infected with AdHBO1, AdMCM2, and AdORC1 (Fig.  3, as indicated), and whole cell extracts were prepared and immunoprecipitated with the indicated antibodies. The resulting immunoprecipitates were assayed for the presence of HBO1 or MCM2 by immunoblot analysis with anti-AU5 (HBO1) or anti-HA (MCM2) antibodies, respectively. In C33A cells, HBO1 polypeptide could be coimmunoprecipitated with endogenous human MCM2 (Fig. 3A, left panel), demonstrating that the HBO1-MCM2 interaction is conserved with human MCM2 in living cells. This interaction was dependent upon AdHBO1 infection of the cells and the presence of anti-MCM2 antibody in the immunoprecipitation reaction. In a reciprocal experiment, adenovirally produced mouse MCM2 protein coimmunoprecipitates with HBO1 (Fig. 3A, right panel). This interaction is dependent upon the production of exogenous HA-tagged MCM2 and AU5-tagged HB01, as well as anti-AU5 (HBO1) antibody. Finally, we confirmed the previously reported interaction between human ORC1 and HBO1 (33) using ectopically expressed ORC1 and HBO1 (Fig. 3B). Taken together, these results confirm the existence of a conserved interaction between MCM2 and HBO1 in vivo and demonstrate that fulllength HBO1 protein is capable of interacting with both MCM2 and ORC1.
HBO1 Localization-Based on our findings that HBO1 interacts with two components of the pre-RC in vitro and in vivo, and the possibility that HBO1-associated HAT activity may be important for the modulation of chromosomal histone function, we sought to determine the cellular localization of HBO1 protein. To accomplish this, we infected asynchronously growing C33A cells with AdHBO1 and prepared nuclear and cytoplasmic extracts. As demonstrated previously by immunofluorescent localization (48), immunoblot analysis revealed that ectopically expressed HBO1 protein is localized exclusively to the nuclear fraction (Fig. 4A, upper panel). By comparison, endogenous MCM2 protein is found both in the nuclear and cytoplasmic fractions (Fig. 4A, middle panel). Furthermore, we found that a fraction of the exogenous HBO1 protein is associated with a chromatin-enriched, low speed pellet (Fig. 4B), and that chromatin-associated HBO1 is resistant to release by nuclease digestion (data not shown). The nuclease resistance of chromatin-associated HBO1 protein is reminiscent of a study demonstrating that chromatin-bound human ORC is resistant to release by nuclease digestion (49). The observation that HBO1 protein is nuclear and associated with a chromosomal fraction is consistent with a role in regulation of chromatin dynamics, although the significance of the nuclease resistance of HBO1 and ORC remains to be determined.

HBO1 Zinc Finger Is Required for Binding to MCM2 and ORC1-We have demonstrated that HBO1 interacts with MCM2 both in vitro and in vivo.
However, the possibility remains that the interaction between MCM2 and HBO1 is mediated by ORC1 or that the ORC1-HBO1 interaction is mediated by MCM2. Therefore, we mapped the interaction domains of HBO1 and MCM2 to characterize further these interactions. The highly conserved MYST domain of HBO1 is defined by the presence of two distinct structural motifs, an atypical C 2 HC zinc finger and an acetyl-CoA-binding motif. Each member of the MYST family (except Esa1p) (40) conforms to the Cys-X 2 -Cys-X 12 -His-X 3 -Cys consensus. Single zinc fingers of functional significance are less common than repeated zinc finger motifs (50). Likewise, C 2 HC motifs are less common than other zinc finger motifs and typically occur singly (51), or as the first or last of a series of zinc fingers (52,53). The significance of this unusual motif remains to be determined, although zinc fingers may function in nucleic acid binding, and single zinc fingers are FIG. 2. HBO1 interacts with MCM2 and ORC1 in vitro. A, analysis of partial versus full-length HBO1 protein-protein interactions with MCM2 and ORC1. Partial (aa 218 -611) and full-length (aa 1-611) HBO1 fused to the GAL4 activation domain (GAD) were individually cotransformed into yeast strain Y190 with constructs encoding the GAL4 DNA-binding domain alone (GBD) or GBD fused to full-length mouse MCM2 (mMCM2), Drosophila MCM2 (dMCM2), human ORC1 (hORC1), or Drosophila ORC1 (dORC1), as indicated. Transformants were grown in selective medium, and ␤-galactosidase (␤-gal) activity was measured in a quantitative liquid culture assay. B, HBO1 was synthesized in vitro ( 35 S-labeled) and incubated with GST or GST-mMCM2, -dMCM2, -hORC1, or -dORC1 fusion proteins coupled to Sepharose beads and washed extensively. Proteins were resolved on 10% SDS-PAGE gel and analyzed by autoradiography. 10% of input HBO1 protein is indicated in left panel.
most likely to facilitate oligomerization or protein-protein interactions (50).
To map the region of HBO1 responsible for binding to MCM2 and ORC1, and to determine whether the C 2 HC zinc finger plays a role in these interactions, a series of HBO1 deletion mutants fused to the GAD were constructed, cotransformed into yeast strain Y190 with plasmids encoding GBD-dMCM2, -mMCM2, or -hORC1, and assayed for ␤-galactosidase activity (Fig. 5). As observed previously in Fig. 2A, GBD-dMCM2, -mMCM2, and -hORC1 interacted strongly with GAD-HBO1-(218 -611), whereas only hORC1 produced activity in conjunction with full-length GAD-HBO1-(1-611). Interestingly, the relative ability of each of the remaining GAD-HBO1 deletion mutants to interact with GBD-dMCM2, -mMCM2, and -hORC1 varied in a parallel manner. Removal of the C-terminal 69 amino acids of HBO1 (GAD-HBO1-(218 -542)) significantly reduced ␤-galactosidase activity for each GBD fusion protein.
To determine whether the C 2 HC zinc finger motif plays a role in protein-protein interactions with MCM2 and ORC1, we mutated the second cysteine of the zinc finger to an alanine Following immunoprecipitation (IP) from C33A whole cell extract with either ␣-MCM2 (left) or ␣-AU5 antibodies (right), as indicated, coprecipitated proteins were resolved by SDS-PAGE, transferred to nitrocellulose membrane, and probed with ␣-AU5 (left) or ␣-HA (right) antibodies. C33A extract (5% of input) is shown for AdHBO1 (left) and AdMCM2 (right) infection. B, following immunoprecipitation from C33A extract with ␣-ORC1 antibodies, coprecipitated proteins were resolved by SDS-PAGE, transferred to nitrocellulose membrane, and probed with ␣-AU5 antibodies. (Fig. 5, GAD-HBO1 C372A ). This single amino acid substitution abolished the ability of HBO1 to interact with dMCM2, mMCM2, and hORC1. These findings indicate that the C 2 HC zinc finger is necessary, but not sufficient, for interactions with MCM2 and ORC1 and provide insights into the functional role played by this highly conserved structural motif. Furthermore, these studies highlight the conserved nature of the proteinprotein interactions of HBO1 with Drosophila and mouse MCM2, as well as with human ORC1.
An N-terminal Domain of MCM2 Is Required for Binding to HBO1-To map the region of MCM2 responsible for binding to HBO1, a series of MCM2 deletion mutants fused to the GBD were constructed and assayed for HBO1 binding activity (Fig.  6). Given the highly conserved mechanism of interaction between HBO1 and MCM2, and the potential for translating observations made in these studies into a system that would allow future genetic analyses, Drosophila MCM2 was used for these analyses. In this two-hybrid assay, the first 72 amino acids of dMCM2 (GBD-dMCM2-(73-888)) were found to be dispensable for binding to HBO1. Further N-terminal truncations (GBD-dMCM2-(106 -888) and GBD-dMCM2-(133-888)) resulted in an incremental decrease in HBO1 binding to ϳ10% of full-length dMCM2. Finally, fusion proteins lacking the first 223 amino acids of dMCM2 were completely deficient in ␤-galactosidase activity, indicating that this N-terminal region is critical for MCM2-HBO1 interaction.
Isolation of an Allele of MCM2 Deficient in HBO1 Interaction-Having mapped the region of MCM2 important for interacting with HBO1 to the N terminus of MCM2 (amino acids 72-223), we wished to generate a mutant MCM2 allele defective for the ability to interact with HBO1. To accomplish this, we employed a reverse yeast two-hybrid system (42,56). This system employs an engineered yeast strain in which the interaction of two proteins expressed in the context of a two-hybrid system is deleterious to growth. Under these conditions, the mutation of a residue that is critical for this interaction provides a selective growth advantage, thereby facilitating detection. For this we used the counter-selectable yeast gene URA3, which encodes an enzyme involved in uracil biosynthesis, and also catalyzes the conversion of the nontoxic analog 5-FOA to a toxic product 5-fluorouracil. The expression level of this URA3 marker is tightly coupled to the formation of a functional transcription factor (through a pair of interacting GAD/GBD fusion proteins) and thus confers a Ura ϩ 5-FOA-sensitive (5-FOA S ) phenotype in the presence of transcriptional activity and Ura Ϫ 5-FOA-resistant (5-FOA R ) phenotype in its absence. Thus, in a large population of cells that express wild-type interacting hybrid proteins (5-FOA S phenotype), a few cells in which this interaction is disrupted can be selected on the basis of their 5-FOA R phenotype.
To generate a large library of random mutations in the FIG. 5. Mapping the region of HBO1 that interacts with MCM2 and ORC1. Human HBO1 deletion mutants fused to the GAD were individually cotransformed into yeast strain Y190 with constructs encoding full-length GBD-dMCM2, -mMCM2, or -hORC1. Transformants were grown in selective medium, and ␤-galactosidase activity was measured in a quantitative liquid culture assay. ␤-Galactosidase (␤-gal) activity of each GAD-HBO1 deletion construct is shown as a percentage relative to the activity of the respective full-length GBD fusion protein with GAD-HBO1-(218-618). Schematic representations of each GAD-HBO1 deletion construct are shown. Numbers correspond to amino acid residues. Highly conserved MYST domain, C 2 HC zinc finger, and acetyl coenzyme A-binding motif (HAT) are indicated.
region of dMCM2 previously identified as critical for HBO1 interaction, we performed PCR amplification of the cDNA encoding this region under conditions that favor nucleotide misincorporation (mutagenic PCR) (42). The mutagenized PCR fragment was then cotransformed into yeast cells with linearized GBD-dMCM2 expression vector containing homologous flanking sequences but lacking the HBO1-interacting region to allow incorporation by homologous recombination. Mutant allele MCM2 R2H101 was isolated from this mutagenized GBD-dMCM2 library in the presence of GAD-HBO1-(218 -611) and was found to contain three point mutations resulting in amino acid substitutions (R114S, V171A, and L222P). As shown in Fig. 7A, clone GBD-dMCM2 R2H101 is unable to interact with HBO1, resulting in a 5-FOA R phenotype and no ␤-galactosidase activity (compare rows 2 and 3). By introducing each of the three mutations individually into GBD-dMCM2, we determined that the leucine to proline mutation at position 222 of dMCM2 is responsible for the observed interaction-defective phenotype (Fig. 7A, compare rows 4 -6).
Alignment of MCM2 amino acid sequences from nine eukaryotic species as well as the archaeal MCM protein reveals that leucine residues occupy the equivalent position in each species (Fig. 7B), suggesting a functional importance for this leucine. To test whether this same leucine to proline mutation has a similar effect on HBO1 binding in the context of the mouse MCM2 protein, we generated a GBD-mMCM2 L237P mutant, and we assayed its ability to interact with HBO1 in the yeast two-hybrid system (Fig.  7C). As seen with the Drosophila MCM2 protein, the leucine to proline mutation at this position of mouse MCM2 results in a 13-fold decrease in ␤-galactosidase activity.
The leucine at position 222 of Drosophila MCM2 is predicted to be located in a turn region immediately flanked by an upstream ␤-sheet and a downstream ␣-helix (Chou-Fasman secondary structure analysis; MacVector 6.5.3). Insertion of a proline residue (which has a low degree of conformational freedom since the proline side chain nitrogen atom is covalently bonded to the main chain forming a rigid ring) at this position might be expected to produce a rigid conformational change in the secondary structure of this critical region of MCM2. As such, we constructed a more subtle leucine to alanine point mutation at position 222 of Drosophila MCM2 (GBD-dMCM2 L222A ), and we assayed its ability to interact with HBO1 in the two-hybrid system. As shown in Fig. 7D, GBD-dMCM2 L222A , like GBD-dMCM2 L222P , is defective for interac- FIG. 6. Mapping the region of MCM2 that interacts with HBO1. Drosophila MCM2 deletion mutants fused to the GBD were individually cotransformed into yeast strain Y190 with construct GAD-HBO1-(218 -611). Transformants were grown in selective medium, and ␤-galactosidase (␤-gal) activity was measured in a quantitative liquid culture assay. ␤-Galactosidase activity of each deletion construct is graphed as a percentage relative to the activity of full-length GBD-dMCM2 with GAD-HBO1-(218 -618). Schematic representations of each GBD-dMCM2 deletion construct are shown. Numbers correspond to amino acid residues. Known structural and functional motifs, including histone-binding domain (54), zinc fingers, and ATP-binding (ATPase) motifs (55), are indicated. tion with HBO1-(218 -611), suggesting that the absence of leucine at position 222, and not the insertion of a proline, is the critical determinant for disrupting HBO1 binding.
Isolation of an HBO1 Mutation That Restores Interaction with the dMCM2 L222P Mutant-We wished to obtain definitive evidence confirming that the interaction between MCM2 and HBO1 is direct and not mediated by a "bridging" protein such as ORC1. A classical genetic method of demonstrating a direct interaction between two proteins is to isolate a mutant allele of one protein that disrupts the interaction, and then to isolate a second site mutation in its partner protein that restores the ability to interact with the interaction-defective allele. We have demonstrated that mutant dMCM2 L222P is defective for the ability to interact with HBO1. By using a library of random mutations spanning the entire HBO1 coding sequence of GAD-HBO1-(218 -611), we screened for a mutant that restores the interaction with dMCM2 L222P in a yeast two-hybrid system. From this screen, a single clone (GAD-HBO1 SSS1 ) was isolated  (42)(43)(44)(45)(46)(47)(48)(49)(50)(51)(52). C, yeast strain Y190 containing GAD-HBO1-(218 -611) was transformed with either GBD-mMCM2 or GBD-mMCM2 L237P , grown in selective medium, and ␤-galactosidase activity was measured in a quantitative liquid culture assay. D, yeast strain Y190 containing GAD-HBO1-(218 -611) was transformed with either GBD-dMCM2, GBD-dMCM2 L222P , or GBD-dMCM2 L222A , grown in selective medium, and ␤-galactosidase activity measured in a quantitative liquid culture assay. and found to contain three mutations resulting in amino acid changes (K282R, I380T, and Y567H). As shown in Fig. 8, GAD-HBO1 SSS1 was capable of partially restoring dMCM2 interaction, as measured by ␤-galactosidase activity. Construction and analysis of each of the GAD-HB01 SSS1 point mutations individually indicates that a single amino acid change is responsible for the observed phenotype. Mutant GAD-HBO1 I380T restored the ability to interact with GBD-dMCM2 L222P to ϳ50% of the level of the wild-type MCM2 and HBO1 proteins. Interestingly, this isoleucine to threonine suppressor mutation maps to the X 12 spacer region of the C 2 HC zinc finger of HBO1. This is consistent with the finding that disruption of the second cysteine residue of the zinc finger (HBO1 C372A , Fig. 5) abolishes the ability of HBO1 to interact with MCM2 and ORC1 and further strengthens the notion that the C 2 HC zinc finger of HBO1 plays a critical role in direct protein-protein interactions with MCM2. DISCUSSION The MCM family of proteins are conserved in all eukaryotes and are essential for normal DNA replication. Nevertheless, the precise function of MCMs in DNA replication remains to be fully understood. In this study, we identified a novel interaction between the MCM2 protein, and HBO1, a putative histone acetyltransferase and MYST domain protein. These data, together with the previously characterized interaction between ORC1 and HBO1 (33), suggest that human HBO1 protein may play a direct role in DNA replication.
Several lines of evidence support the notion that the interaction between MCM2 and HBO1 is specific, physiologically significant, and evolutionarily conserved. First, the MCM2-HBO1 interaction was detected in several assay systems, including yeast two-hybrid, GST pull-down, and coimmunoprecipitation in human cells. Second, we have generated interaction-defective alleles of dMCM2 (dMCM2 L222P ) and HBO1 (HBO1 C372A ) that are defective in their ability to bind to their wild-type partner protein, respectively. Additionally, we have demonstrated a direct interaction between MCM2 and HBO1 by generating a mutant version of HBO1 (HBO1 I380T ) that restores the ability to interact with dMCM2 L222P . This second site suppressor mutation serves as genetic evidence of a direct interaction between MCM2 and HBO1.
The previous finding that HBO1 interacts with ORC1, together with the data presented here demonstrating a direct interaction with MCM2, suggests the possibility that HBO1 might be recruited to a pre-replication complex by multiple protein interactions. For the reasons stated above, we believe that these are likely independent interactions involving MCM2 and ORC1. Nevertheless, it is also quite possible that they are synergistic, resulting in the formation of a more stable complex involving HBO1.
Histone acetylation serves as a general mechanism to destabilize nucleosome core particles during various processes occurring in chromatin. Allfrey et al. (57) first reported that addition of acetyl groups to histone proteins is associated with transcriptionally active regions of the genome. More recently, many proteins that regulate transcription have been shown to possess or recruit HAT or histone deacetylase activities (reviewed in Ref. 58). Several MYST subfamily proteins have been implicated in a variety of transcriptional processes including transcriptional activation, dosage compensation, and silencing. MOZ was discovered due to its involvement in a chromosomal translocation that fuses MOZ to the transcriptional coactivator CREB-binding protein, resulting in a form of acute myeloid leukemia (34). Tip60 was cloned by its ability to interact with the viral activator HIV Tat in a two-hybrid system (36). MOF was discovered in a genetic screen for gene products necessary for hypertranscription and the acetylation of histone H4 at Lys-16 that is associated with Drosophila dosage compensation (37). Interestingly, the mof mutant identified in this screen contains a single point mutation located in the acetyl-CoAbinding (HAT) motif. SAS2 and SAS3 were identified as enhancers of sir1 epigenetic silencing defects in S. cerevisiae (38), and Esa1 was subsequently isolated by homology to SAS2 and SAS3 and shown to modulate genes important for cell cycle control (40,59). Among these proteins, Tip60 and Esa1 have been shown to possess intrinsic HAT activity (40,60).
The result of histone acetylation is a change in chromatin structure and a corresponding increase in the accessibility of the DNA by trans-acting factors. Given this notion, it might be expected that other DNA-dependent enzymatic activities requiring access to chromosomal DNA would also benefit from chromatin modification by histone acetylation, and this appears to be the case. McMurry and Krangel (61) recently used an antibody to acetylated histone H3 in a chromatin immunoprecipitation assay to demonstrate a link between histone acetylation and V(D)J recombination in the TCR␣/␦ locus. Furthermore, targeted deletion of the TCR␣ enhancer eliminates TCR␣ locus hyperacetylation and completely blocks V-to-J␣ rearrangement. Whereas the correlation between histone acetylation and accessibility to the V(D)J recombinase is striking, it remains to be determined how histone acetylation alters accessibility to specific gene loci, and whether altered enzymatic activity is a cause or effect of differential accessibility.
Histone acetylation also appears to play an important role in DNA repair. Ectopic expression of mutated MYST domain protein Tip60 lacking histone acetylase activity results in cells with defective double-strand DNA break repair and impaired apoptotic competence, indicating that Tip60-associated HAT activity plays a role in DNA repair and apoptosis (62).
By analogy, destabilization of chromatin structure through histone acetylation might also play an important role in the regulation of DNA replication. In proliferating rat liver cells, an increase in HAT activity is associated with the onset of DNA replication (63), and depression of histone acetylation by alkylating agents correlates with an inhibition of thymidine incorporation (64). Additionally, studies using Chinese hamster chromosomes demonstrate a close correlation between low levels of histone H4 acetylation and late DNA replication (65). More directly, core histone acetylation has been demonstrated previously to play a role in DNA replication efficiency (66). In this study, an SV40 in vitro replication system was used to analyze the replication efficiencies of SV40 minichromosomes associated with normal or hyperacetylated histones. Elongation of replication occurred with a higher efficiency in hyperacetylated minichromosomes, indicating that movement of the replication machinery through nucleosomal DNA is facilitated by acetylation of histone tails. This potential role of histone acetylation in the elongation stage of DNA replication is consistent with the finding that MCM proteins migrate with the replication fork during elongation (67). Additionally, Aoki and Schultz (68) recently examined the temporal and spatial distribution of sites of DNA replication in one-cell mouse embryos, and they demonstrated that inhibition of transcription with ␣-amanitin had no effect on the patterns of DNA replication. However, treatment of embryos with a specific inhibitor of histone deacetylase accelerated the completion of replication. This study suggests that acetylation of histones, but not transcriptional activity, is involved in the regulation of DNA replication in one-cell mouse embryos. These findings, along with the observations that HBO1 interacts with two components of the replication machinery, MCM2 and ORC1, make it interesting to speculate that HBO1-associated HAT activity may play a direct role in modulation of DNA replication.
Although the results presented here and elsewhere provide evidence for protein-protein interactions that could recruit a histone acetylase to a pre-replication complex, this possibility remains only a hypothesis that must be tested in the context of relevant biological systems. We believe the best approach to this question is a genetic strategy that might correlate potential physical interaction with biological function. Toward this end, our isolation of alleles of Drosophila and mouse MCM2 (dMCM2 L222P and mMCM2 L237P , respectively) that are defective for interaction with HBO1 provides an approach to this question. This interaction-defective allele can then be compared with its wild-type counterpart for the ability to functionally complement a knock-out in the corresponding gene or for the ability to function in an expression assay in the relevant cells. By taking this functional analysis one step further, it might then be possible to isolate a second site suppressor mutation, such as HBO1 I380T , that restores the ability of the interaction-defective allele to interact. This second site suppressor might then be used to functionally rescue the interaction-defective allele, restoring the wild-type phenotype. Clearly, a physiological link between HAT activity and origin function is an intriguing possibility, and establishing that link by functional assays is a critical next step in this investigation.