Architecture of the yeast origin recognition complex bound to origins of DNA replication.

In many organisms, the replication of DNA requires the binding of a protein called the initiator to DNA sites referred to as origins of replication. Analyses of multiple initiator proteins bound to their cognate origins have provided important insights into the mechanism by which DNA replication is initiated. To extend this level of analysis to the study of eukaryotic chromosomal replication, we have investigated the architecture of the Saccharomyces cerevisiae origin recognition complex (ORC) bound to yeast origins of replication. Determination of DNA residues important for ORC-origin association indicated that ORC interacts preferentially with one strand of the ARS1 origin of replication. DNA binding assays using ORC complexes lacking one of the six subunits demonstrated that the DNA binding domain of ORC requires the coordinate action of five of the six ORC subunits. Protein-DNA cross-linking studies suggested that recognition of origin sequences is mediated primarily by two different groups of ORC subunits that make sequence-specific contacts with two distinct regions of the DNA. Implications of these findings for ORC function and the mechanism of initiation of eukaryotic DNA replication are discussed.

In many organisms, the replication of DNA requires the binding of a protein called the initiator to DNA sites referred to as origins of replication. Analyses of multiple initiator proteins bound to their cognate origins have provided important insights into the mechanism by which DNA replication is initiated. To extend this level of analysis to the study of eukaryotic chromosomal replication, we have investigated the architecture of the Saccharomyces cerevisiae origin recognition complex (ORC) bound to yeast origins of replication. Determination of DNA residues important for ORC-origin association indicated that ORC interacts preferentially with one strand of the ARS1 origin of replication. DNA binding assays using ORC complexes lacking one of the six subunits demonstrated that the DNA binding domain of ORC requires the coordinate action of five of the six ORC subunits. Protein-DNA cross-linking studies suggested that recognition of origin sequences is mediated primarily by two different groups of ORC subunits that make sequence-specific contacts with two distinct regions of the DNA. Implications of these findings for ORC function and the mechanism of initiation of eukaryotic DNA replication are discussed.
The initiation of DNA replication is a complex process involving multiple regulated steps, including the selection of the initiation site on the DNA, unwinding of the DNA helix, and assembly of a multiprotein replication machine. Studies of the replication of bacteria, phages, and the genomes of viruses infecting eukaryotes have established that a protein called the initiator binds to the cognate origin of DNA replication in a sequence-specific manner. Once bound, initiator proteins often participate in other aspects of replication initiation; for instance, they facilitate origin unwinding and recruit other replication proteins to the origin (for a review, see reference 3). Detailed analyses of initiator proteins bound to their cognate origins have been important in the determination of how these proteins function during replication initiation. Our aim was to extend this type of analysis to a putative eukaryotic chromosomal initiator protein bound to an origin of DNA replication.
The strongest candidate for a eukaryotic initiator protein is the origin recognition complex (ORC) (7). ORC was first identified in the yeast Saccharomyces cerevisiae as a six-subunit complex that specifically binds to origin sequences in vitro in the presence of ATP. The six ORC subunits are referred to as Orc1p through Orc6p (in order of decreasing mass), and all six proteins are essential for the viability of yeast cells (5,6,30,32). ORC-DNA binding is required for the essential function of ORC, since mutations in origin sequences that reduce or eliminate origin function also reduce or eliminate ORC binding in vitro (7) and in vivo (1a, 42a). In addition, most conditional mutations in ORC genes lead to decreased origin usage (18,31) and decreased origin binding in vivo (1, 1a). A role for ORC as an initiator protein is further supported by its conservation in multiple eukaryotes. Proteins with amino acid sequence similarity to yeast ORC subunits have been identified in numerous other eukaryotes (for a review, see reference 17), and complexes of these ORC-like proteins have been purified from Xenopus laevis and Drosophila melanogaster (19,38). More importantly, ORC-related proteins are required for DNA replication in Schizosacharomyces pombe, Xenopus, and Drosophila cells (28; for a review, see reference 17).
Studies of proteins associated with yeast origins of DNA replication in vivo have demonstrated the existence of multiple ORC-dependent protein-DNA complexes (1a, 15, 40, 42a). During S phase, G 2 , and M phase, a postreplicative complex (post-RC) that produces a DNase I protection pattern similar to that seen in vitro with purified ORC is observed (7,15). As cells cross the M phase-to-G 1 boundary, however, additional proteins involved in replication, including Cdc6p and the MCM proteins, are recruited to origins to assemble a pre-RC (1a, 13, 40, 42a). At the G 1 -to-S transition, a third complex is formed that includes at least one DNA polymerase; this is referred to as the RC (1a). Studies of chromatin-associated proteins required for Xenopus DNA replication have demonstrated a clear order of assembly. ORC-related proteins bound to chromatin recruit X. laevis Cdc6p, and the resulting complex is required for the recruitment of X. laevis MCM proteins (14,37,38). Studies in S. cerevisiae suggest a similar order of assembly (1a, 42a). Thus, ORC forms the foundation for the assembly of critical higher-order origin structures that change during the cell cycle.
The best-defined eukaryotic origins of replication are those of S. cerevisiae. These elements were first identified as genomic DNA sequences capable of supporting the autonomous replication of episomal DNA (autonomous replicating sequences [ARSs]) (23,42). Many of these elements were subsequently shown to act as origins of replication in their normal chromosomal context (for a review, see reference 34). Yeast origins are modular in nature and contain an 11-bp ARS consensus sequence (ACS) that is essential for ORC-DNA binding and origin function in vivo, as well as additional elements that enhance origin function (generally referred to as B elements) (4). ARS1, the first well-characterized origin, has three such elements (B1, B2, and B3 [see Fig. 7]). In vitro and in vivo, DNase I protection assays of ARS1 demonstrated that ORC protects approximately 50 bp of DNA that include the ACS and B1 sequences (7,16,40). These two elements direct ORC-DNA binding at ARS1 and at ARS307 (35,39) and together represent the smallest functional region of either origin. We will refer to this minimal region required for ORC-DNA binding and origin function as the core origin.
We have a general understanding of the DNA sequence requirements for the association of ORC with origins (7,36,39), but previous studies have not examined the requirements of different ORC subunits for DNA binding. A thorough understanding of ORC bound to yeast origins of replication can address three important questions. (i) How does ORC interact with DNA? In particular, we would like to determine which of the approximately 50 protected base pairs are important for ORC-DNA binding, how the structure of origin DNA is affected by interaction with ORC, and which of the six essential ORC proteins are required for DNA binding. (ii) How do ORC subunits interact with each other? ORC is a preassembled complex in the absence of DNA, and we want to understand the organization of ORC subunits both in solution and in DNA complexes. An understanding of the spatial arrangement of ORC subunits is also relevant to the third question. (iii) How does the ORC interact with other proteins? In addition to their genetic and physical interactions with the CDC6 gene product (31), ORC subunits show multiple genetic interactions with other essential genes required for DNA replication (for a review, see reference 17). Since ORC plays a central role in assembling higher-order structures at origins, the arrangement of ORC subunits will undoubtedly influence the formation of these larger protein-DNA complexes.
In this study, we have used DNA modification to identify specific residues in ARS1 involved in ORC-DNA binding. DNA bending studies were used to investigate ORC-induced structural changes in origin DNA. In addition, we have used analysis of ORC complexes lacking one of the six subunits and protein-DNA cross-linking to determine which ORC subunits are required for DNA binding and how these subunits are arranged along the origin DNA. Together these studies provide a detailed view of ORC bound to origin DNA.

MATERIALS AND METHODS
Plasmids and competitor DNA. pDL01, used in modification-interference and missing-contact assays, was prepared by inserting the following sequence into the EcoRV site of pBluescript SK(ϩ): 5Ј AAGGATCCAA AGTGCACTTA ACT GCAGAAC TTTTGAAAAG CAAGCATTAA AGATCTAAAC ATAAAAT TTG TTAACTATCT AGATG 3Ј. This sequence encodes a minimal ORC binding site that contains 74 bp of ARS1 sequence, including the ACS and B1 elements. This DNA is bound by ORC in a manner similar to the binding of the wild-type ARS1 sequence as judged by DNase I protection and mobility shift assays (data not shown). Plasmids pARS1/GAG and pARS1/CTC were constructed by PCR-mediated mutagenesis of pARS1/WTA. pARS1/a Ϫ b2 Ϫ was generated by replacing the BglII-to-HindIII fragment of pARS1/858-865 (which has a linker substitution of the ACS) with the same fragment from pARS1/798-805 (which has a linker substitution of the B2 element) (these plasmids were previously described in reference 33). Plasmid stability assays were performed as described previously (6). Plasmids pARS1/WTA and pARS1/a Ϫ b2 Ϫ were used as templates for production of wild-type and ACS-negative, B2-negative competitor DNAs, respectively. Competitor DNAs were synthesized by 25 cycles of PCR with universal forward and reverse sequencing primers, and PCR products were purified on a 2% agarose gel (1ϫ Tris-borate-EDTA [TBE]). DNA was recovered by electroelution followed by ethanol precipitation. All other DNA fragments and DNA probes were purified by electrophoresis on a native 4.8% polyacrylamide gel (24:1 acrylamide/bisacrylamide, 1ϫ TBE) and recovered by electroelution.
Expression of ORC in insect cells. Expression and purification of wild-type ORC from insect cells was performed as described previously (27) except that 10 mM magnesium acetate was included in the chromatography buffers at all steps. The expression of partial ORC complexes lacking one subunit required the use of baculoviruses coexpressing two ORC subunits (6) and viruses expressing only one subunit. Baculoviruses expressing single ORC subunits were named bvORC1 through bvORC6. Each of the six partial ORC complexes was ex-pressed in Sf9 insect cells by coinfection with two viruses that each express two ORC subunits and one virus that expresses a single subunit. These partial ORC complexes were purified through the S-Sepharose step in the ORC purification scheme (27). The Sp1-Orc4p fusion protein was constructed by PCR amplifying DNA encoding the C-terminal 168 amino acids of the Sp1 transcription factor and fusing this PCR product to the coding sequence of the N terminus of Orc4p. Coding sequences for Orc3p and the Sp1-Orc4p fusion were cloned into pFast-Bac Dual (Gibco BRL) to generate plasmid pFBD/ORC3/N-SP1-ORC4, and baculovirus was produced from this plasmid by using the Bac-to-Bac Expression System (Gibco BRL). ORC complexes containing the Sp1-Orc4p fusion protein as the sole copy of Orc4p were expressed in insect cells and purified up to the Mono-Q chromatography step. Immunoblot analysis was performed as described previously (5) except that monoclonal antibodies directed against individual ORC subunits were used to detect ORC proteins (monoclonal antibodies SB16, SB46, SB3, SB6, SB5, and SB49 were used for detection of Orc1p through Orc6p, respectively).
Electrophoretic mobility shift assays. Unless otherwise noted, ORC-DNA binding conditions for all experiments were as follows. Reaction mixtures (15 l) contained 12.5 mM HEPES-KOH (pH 7.5), 2.5 mM magnesium acetate, 2.5 mM dithiothreitol, 5 mM EGTA, 0.66-mg/ml poly(dG-dC), 2-mg/ml bovine serum albumin (BSA), 1 mM ATP (where indicated), 20 ng of ARS1/a Ϫ b2 Ϫ competitor DNA, 0.22 ng of radiolabeled probe (300 cps), and 12 ng of ORC. Binding reaction mixtures were incubated for 10 min at room temperature. The ARS1 probe used in the mobility shift assay was generated by digesting pARS1/WTA with EcoRI, 3Ј-end labeling with the Klenow fragment of DNA polymerase I, and digesting with HindIII. The labeled 244-bp EcoRI-HindIII fragment was gel purified as described above. All ORC mobility shift assays were performed as described previously (36), except that gels and running buffers included 80-g/ml BSA and gels were run at 4°C for 4 h at 200 V.
Modification-interference and missing-contact assays. Chemical modifications of labeled pDL01 DNA was carried out as follows. Diethyl pyrocarbonate (DEPC) carbethoxylation was performed essentially as described previously (22), except that end-labeled DNA was heated for 5 min at 90°C prior to incubation with DEPC, which was carried out for only 5 min at 90°C. Following two ethanol precipitations, DEPC-modified DNA was resuspended in hybridization buffer (10 mM Tris-HCl, 1 mM EDTA, 30 mM NaCl; pH 8.0), heated to 95°C for 3 min, incubated at 65°C for 10 min, and allowed to reanneal by slowly cooling the suspension to room temperature. Formic acid depurination was performed as described previously (12). KMnO 4 modification was carried out as described previously (44), except that treatment with KMnO 4 was performed for 15 min at room temperature. Reannealing of melted DNA was performed as described above. Hydrazine depyrimidation was performed as described previously (29), except that incubation with hydrazine was carried out for 30 min at room temperature. Phosphate backbone ethylation with ethylnitrosourea (ENU) was performed as described previously (21).
Prior to separation of bound and unbound DNA molecules, an aliquot of each modified DNA sample was reserved for chemical cleavage (see input samples [I] in Fig. 1A). The remainder was incubated with purified ORC in the standard binding buffer, using between 6 and 10 ng of modified DNA and 600 ng of protein in a 30-l reaction mixture. Bound and unbound DNA molecules were separated by the electrophoretic mobility shift procedure described above, gels were exposed to film for 30 min, and bound and free DNAs were excised from the gels. DNA was recovered by electroelution and ethanol precipitation. Recovered DNA molecules (bound and free molecules, as well as input DNA aliquots) were cleaved at sites of modification by one of two methods. For ENU modification, ethylated phosphates were cleaved by heating in the presence of NaOH as described previously (21). For all other modifications, precipitated DNA was cleaved with piperidine (29). Cleaved DNA was resuspended in a formamide-dye mixture and separated on a 10% polyacrylamide DNA sequencing gel. The gels were dried and then exposed to film.
ORC-induced DNA bending assay. DNA fragments used for ORC-induced bending studies were generated by PCR from yeast genomic DNA (ARS305) or from plasmids (p19AB121 for ARS121; pARS1/WTA, pARS1/a Ϫ b2 Ϫ , and pARS1/858-865 for wild-type and mutant ARS1 fragments). Oligonucleotides were designed to add XbaI sites to both ends of the following regions amplified from each ARS: nucleotides Ϫ69 to ϩ145 of ARS305 (24), 285 to 498 of ARS121 (45), and 726 to 939 of ARS1 (33). PCRs were carried out for 25 cycles in the presence of 0.2 M [␣-32 P]dATP. The PCR products were digested with XbaI overnight, and the resulting 220-bp fragment was gel purified as described above. Bending probes were incubated with purified ORC according to the mobility shift assay ORC-DNA binding conditions with the following modifications. Reactions were performed in 40-l volumes including 13 ng of radiolabeled origin probe and a 50-fold molar excess of unlabeled competitor DNA (containing either the wild-type ARS1 sequence or the ARS1/a Ϫ b2 Ϫ sequence). One hundred fifty nanograms of ORC was added where indicated, and all reaction mixtures contained 96 g of BSA. Circularization of origin probes was performed as described previously (26), using 56 NEB U of T4 DNA ligase per reaction. Samples were removed at various times, reactions were stopped, and reaction products were electrophoresed on a 20-cm-long 5% native acrylamide gel (1:40 acrylamide/bisacrylamide; 1ϫ TBE) for 830 V-h. Circular DNA species were distinguished from linear molecules by their resistance to exonuclease III digestion. Gels were either dried and exposed to film or exposed to Molecular Dynamics PhosphorImager screens for quantitation by ImageQuaNT software.
Protein-DNA cross-linking. The production of 4Ј-azidophenacyl bromide (4Ј-AZPB) UV cross-linking probes combined methods described by Bell and Stillman (7) and Yang and Nash (47). The EcoRI-HindIII fragment of pARS1/WTA was subcloned into M13mp18 and M13mp19 replicative forms which had been cut with the same enzymes, and single-stranded DNAs were produced for use as templates for ARS1 cross-linking probes. An M13mp18 derivative of ARS305 was generated by cloning the ARS305 PCR product used for ORC-induced bending studies into this vector. The Ϫ40 universal sequencing primer and a second oligonucleotide were annealed to single-stranded templates. Extension using T4 DNA polymerase in the presence of ␣-S-dCTP or ␣-S-dTTP and labeled [␣-32 P]dATP or [␣-32 P]TTP resulted in the incorporation of one or two thiophosphate nucleotides and several radiolabeled nucleotides immediately following the second primer. The extension was chased with an excess of unlabeled, unmodified nucleotides to complete DNA synthesis. The resulting doublestranded circles were precipitated, and 4Ј-AZPB was coupled to the incorporated thiophosphates as described previously (47). Free 4Ј-AZPB was removed by using a 1-ml G-50 spin column. The purified DNA was digested with EcoRI and SalI to remove any modified or labeled dinucleoside triphosphates incorporated following the universal primer, and the liberated restriction fragment was gel purified. DNA fragments for bromodeoxyuridine (BrdU) cross-linking were synthesized as described previously (2), using the same single-stranded DNA templates as the 4Ј-AZPB cross-linking probes; they were then digested with EcoRI and SalI and gel purified.
ORC-DNA binding conditions for cross-linking experiments were as described above, with the following changes. The amount of purified protein added per binding reaction was increased twofold for 4Ј-AZPB cross-linking, and the amount of ARS1/a Ϫ b2 Ϫ competitor was increased three and fivefold for BrdU and 4Ј-AZPB cross-linking, respectively. Reaction mixtures were transferred to a microtiter plate and irradiated with a 254-nm light source (model UVG-54; UVP) at a distance of 1 cm for either 2 min (4Ј-AZPB cross-linking) or 30 min (BrdU cross-linking). Cross-linked proteins were treated with DNase I (Worthington) and micrococcal nuclease (Worthington) as described previously (7), precipitated with trichloroacetic acid, and resolved on sodium dodecyl sulfate-10% polyacrylamide gels. Gels were silver stained, dried, and exposed to film. We had previously reported that Orc1p did not cross-link to ARS1 (27); however, we subsequently discovered that the micrococcal nuclease used to digest DNA after cross-linking to ORC was contaminated with a protease. Orc1p and Orc2p were the most sensitive of the ORC subunits to this protease (data not shown). Boiling of the micrococcal nuclease eliminated the contaminating protease activity, and Orc1p and Orc2p cross-linking became observable.

Residues of ARS1 required for ORC-DNA binding.
To identify residues of ARS1 important for ORC-DNA binding in vitro, DNA modification-interference and missing-contact assays were performed. In both assays, the geometry of the DNA is altered at particular sites to identify important protein-DNA interactions. Modification-interference analysis involves changing the shape of the DNA by adding an adduct. In contrast, missing-contact assays involve the removal of a base (leaving the phosphate backbone intact). The two assays are complementary; if a particular site inhibits protein-DNA binding in both assays (either by being modified or by being removed), then it is likely to represent a region of the DNA that is important for association with protein. Since in the case of ORC-DNA binding we found that modification-interference and missing-contact experiments identified similar residues (Fig. 1), henceforth either the addition of an adduct or the removal of a base will be termed a modification of a residue. In these assays, ARS1 DNA fragments end labeled on either the top or bottom strand were modified with one of five reagents prior to ORC-DNA binding (see the legend to Fig. 1A for a description of the reagents and their resulting DNA modifications). The modified DNA was incubated with purified ORC and electrophoresed on a gel to separate bound and unbound DNA molecules. DNA modifications that inhibit ORC-DNA binding are reduced or absent in the bound DNA populations.
The individual residues whose modification most strongly inhibited ORC-DNA binding were located in a region spanning the ACS and B1 elements of ARS1 ( Fig. 1A and highlighted residues in Fig. 1B). ORC-DNA binding was extremely sensitive to modification of residues within the ACS, consistent with the essential role of this element in both ORC-DNA binding and origin function. In contrast, only some of the bases in the genetically defined B1 element inhibited ORC-DNA binding when modified, suggesting that the remainder of this element contributes to an origin function that does not involve ORC-DNA binding (36). The modification data also exhibited strand-specific differences. Modification of the top strand consistently affected ORC-DNA binding to a greater extent than did modification of the bottom strand. Within the ACS, topstrand modification interfered with ORC-DNA binding more than bottom-strand modification. Within the B1 element, the residues whose modification interfered with ORC-DNA association were exclusively on the top strand. The most striking strand-specific differences were in a region between the ACS and B1, where modification of the two strands had opposite effects. Top-strand modification interfered with ORC-DNA binding, whereas bottom-strand modifications were overrepresented in the bound DNA population.
To determine if the region of ARS1 between the ACS and B1 is important for origin function, we mutated residues 852 to 854, changing AGA to either GAG or CTC. Both mutations were tested to determine how they affected plasmid stability and ORC-DNA binding in vitro ( Table 1). The AGA-to-GAG mutant had no detectable defect in vivo or in vitro, whereas the AGAto-CTC substitution mutation resulted in a 50% decrease in both plasmid stability and binding in vitro. This region had not been identified in a previous genetic analysis of ARS1, presumably because a linker substitution of the sequences between the ACS and B1 did not lead to transversion mutations in the important AGA (33). Thus, this region contributed to ARS1 function and appeared to show a preference for purines on the top strand and for pyrimidines on the bottom strand.
ORC bends DNA at some, but not all, origins. Because some initiator proteins induce DNA bending at origins as a precursor to DNA unwinding (9, 11), we investigated whether ORC bends yeast origin DNA. A radiolabeled ARS1 fragment with compatible cohesive ends was incubated with DNA ligase, and the rate of formation of circular monomers, either with or without ORC, was monitored ( Fig. 2A). In the presence of ORC, the rate of formation of circular monomers was stimulated threefold, consistent with the interpretation that ORC bends ARS1 DNA, thereby facilitating the ligation of the two DNA ends. The observed stimulation of ligation required specific ORC-DNA binding, as demonstrated by competition experiments with wild-type and mutant competitor DNAs ( Fig.  2A; compare lanes 9 to 12 with lanes 5 to 8) and by experiments using a DNA-bending template containing a mutated ORC binding site ( Fig. 2A, lanes 13 to 24). To determine if bending of origin DNA was a general property of ORC, two other ARS elements were examined. ARS121 and ARS305 are both active chromosomal origins of replication (24,46). Although ORC bound strongly to these origins (data not shown), neither was bent by ORC (Fig. 2B). Thus, bending of DNA by ORC occurs at ARS1 but is not a general property of ORC interaction with origins of replication.
Orc6p is not required for DNA binding. To determine the roles of individual ORC subunits in ORC-DNA binding, we produced six different mutant ORC complexes, each missing a different ORC subunit (referred to as partial ORC complexes). Insect cells were coinfected with baculoviruses expressing five of the six ORC polypeptides, and these mutant ORC complexes were partially purified. The integrity of each of the six partial complexes was examined by determining the ORC subunits present in the partially purified fractions. In two cases, the omission of one subunit clearly compromised complex in-tegrity; the omission of Orc3p reduced the amount of Orc2p present, and the lack of Orc5p led to a loss of Orc4p from the complex (Fig. 3B). Protein fractions containing the six partial complexes were next assayed for ORC-DNA binding activity.
Of the six partial complexes tested for ARS1 DNA binding, five were inactive. A complex lacking Orc6p, however, was capable of binding origin DNA in a sequence-specific (data not shown) and ATP-dependent (Fig. 3A) manner. This protein-DNA complex is likely to contain all five of the ORC proteins present in the partial-complex fraction, as all five subunits coimmunoprecipitate and antibodies to three of the subunits (Orc1p, Orc2p, and Orc4p) can supershift the partial ORC-DNA complex (data not shown). Protein-DNA complexes formed with the protein fractions containing the other partial ORC complexes are likely due to other DNA binding proteins. None of these protein-DNA complexes was ATP dependent (Fig. 3A), DNA sequence specific, or supershifted with monoclonal antibodies directed against ORC subunits (data not shown). Thus, only Orc6p is dispensable for the formation of a stable DNA binding complex, and ORC-DNA binding activity is not readily attributable to any single polypeptide or small subassembly of polypeptides.
Organization of ORC subunits at ARS1. The partial complex experiments described above indicate that Orc1p, Orc2p,  Orc3p, Orc4p, and Orc5p are important for DNA binding. To define the subunits that are in close proximity to origin DNA and to determine how these subunits are arranged along the length of ARS1, we performed UV protein-DNA cross-linking studies. In these experiments, we generated a series of 10 ARS1 cross-linking probes that were modified by addition of a photoreactive azido group on one or two specific phosphate residues. The azido group is coupled to thiophosphate groups in the DNA backbone through the use of the reagent 4Ј-AZPB, which constrains the photoreactive group to a distance of approximately 10 Å from the phosphate backbone. Proteins cross-linked to a particular modified residue are therefore within this distance, and we will describe such proteins as being in close proximity to the DNA at this site. The portion of ARS1 examined using this set of 10 probes is 92 bp long and includes the entire region protected from DNase I digestion by ORC (Fig. 4B).
The cross-linking studies demonstrate that different ORC subunits are distributed over different regions of ARS1, with a clustering of subunits over the ACS (for a summary, see Fig. 7). Orc1p is cross-linked to the right half of the ACS and to regions flanking the ACS on the right (probes I and J). Orc2p FIG. 2. ORC-induced DNA bending at ARS1. (A) Circular monomer formation of radiolabeled ARS1 fragments in the presence or absence of ORC. After addition of DNA ligase, monomeric linear ARS1 fragments were converted to circular monomers, circular dimers, and linear dimers. The three electrophoretic species of linear dimers were the result of a weak intrinsic bend in the B3 element (removal of this DNA bend by the use of a B3 linker substitution did not affect the ORC-induced binding of ARS1 [data not shown]). The rates of circular monomer formation in the presence (lanes 5 to 8 and 9 to 12) and absence (lanes 1 to 4) of ORC were tested. ORC-ARS1 binding and ligation reactions were carried out in the presence of a 50-fold excess of unlabeled competitor DNA containing either a wild-type ARS1 binding site (lanes 9 to 12) or a mutant ARS1 binding site (lanes 5 to 8). A radiolabeled DNA probe containing a mutated ARS1 binding site (the ARS1/a Ϫ b2 Ϫ mutation) was also tested (lanes 13 to 24). (B) Circular monomer formation for ARS305 (left graph) and ARS121 (right graph) in the presence (å) and in the absence (ϫ) of ORC was measured. Both graphs also show quantitation of cyclization rates for wild-type ARS1 DNA in the presence (}) and in the absence (s) of ORC, and the left graph includes data for an ARS1 DNA fragment with a linker substitution in the ACS for comparison [pARS1/858-865 with ‫)ء(‬ and without (F) ORC]. The amount of circular monomer produced at each time point was expressed as the percentage of total counts in each lane. and Orc3p are colocalized to the entire ACS and to DNA that extends leftward toward the B1 element (probes E, F, G, and I); they are the only ORC subunits cross-linked to residues located between the ACS and B1 that contribute to ARS1 function (probes F and G) (Fig. 4). Because Orc4p and Orc5p have similar electrophoretic mobilities and are difficult to resolve, we modified Orc4p by fusing an additional 168 amino acids, derived from the Sp1 transcription factor, to its N terminus (Fig. 5A). This modified ORC4 protein migrated to a position above that of the Orc2p subunit. Cross-linking was performed with a wild-type ORC complex and a complex containing the modified Orc4p, and only one cross-linked protein exhibited altered mobility, allowing unambiguous identification of Orc4p and Orc5p (Fig. 5B). In the wild-type ORC complex, Orc4p is cross-linked strongly to the ACS (Fig. 4A, probe H) and more weakly to sequences to the right of the ACS (probe J). Orc5p is the only subunit cross-linked to the region of B1 important for ORC-DNA binding (Fig. 4A, probe D), and Orc6p is cross-linked to a residue at the end of the B1 element as well as to residues between B1 and B2 (probes B and C).
Not all ORC subunits were localized to a discrete region of ARS1. Although Orc1p, Orc3p, and Orc6p are cross-linked to only one region of the DNA, Orc2p, Orc4p, and Orc5p are each cross-linked to two distinct regions of ARS1, with no cross-linking to the intervening sequence ( Fig. 4A; for a summary, see Fig. 7). Orc2p, Orc4p, and Orc5p are cross-linked to sites on the DNA separated by 78, 54, and 35 bp, respectively. Cross-linking of subunits to two distinct regions of DNA could be explained by ORC subunits having an elongated shape, by the existence of multiple complexes or individual subunits bound at each origin, or by the occurrence of DNA bending at the origin. Based on comparison with protein-DNA cross-linking at ARS305 (see below), the latter possibility is the most likely (see Discussion).
ORC subunit organization at ARS305. To determine if the subunit organization of ORC is similar at other origins of DNA replication, ARS305 was examined with selected probes that placed cross-linkers at positions analogous to those tested for ARS1 (determined relative to the ACS). Four cross-linking Gels were silver stained to determine the positions of ORC subunits, dried, and exposed to film. In all cases, the observed protein-DNA cross-linking was ATP dependent. Cross-linking to ORC subunits was also UV dependent and sensitive to competition by a wild-type ARS1 fragment (data not shown). (B) Positions of cross-linking nucleotides in ARS1 probes A through J. Arrows identify the modified positions; the phosphate coupled to the photoreactive cross-linker is 5Ј to the residue indicated. Each probe has either one or two modified bases as indicated. Outlined residues between positions 839 and 868 are as described in the legend to Fig. 1B. probes used in the examination of the core binding region of ARS305 generated results similar to those seen at ARS1 (compare Fig. 4 and Fig. 6). Just as in ARS1, the ACS of ARS305 is cross-linked to Orc1p, Orc2p, Orc3p, and Orc4p (probe I 305 ). Orc1p and Orc4p also cross-link to the right of the ACS (probe J 305 ), and Orc3p cross-links near the B1 element of ARS305 (probe E 305 ) (Fig. 6). Importantly, in the central portion of the ARS305 B1 element, Orc5p is the sole cross-linked ORC subunit (probe D 305 ). At ARS305, the interaction of Orc2p with the DNA does not extend as far as that observed at ARS1, since this subunit is not cross-linked next to B1 by probe E 305 . Aside from this difference, the organization of ORC subunits over the core DNA binding site is similar at the two origins of replications (Fig. 7).
An interesting difference between ARS1 and ARS305 is that the Orc5p subunit is only cross-linked to one discrete site at ARS305 (it is present only at B1) and not to multiple sites as in the case of ARS1. To determine if the other subunits that cross-linked to more than one region of ARS1 (Orc2p and Orc4p) were also localized to only one region of ARS305, we examined sequences beyond the core ARS305 binding site. ARS305 cross-linking probes analogous to ARS1 probes A and B that cross-linked Orc2p and/or Orc4p, as well as a third probe that modified a nearby residue on the opposite strand, were synthesized (Fig. 6B). These three ARS305 probes showed no evidence of specific cross-linking (data not shown). Thus, these cross-linking data argue that the association of ORC subunits with the core origin is likely to be similar at all origins; however, ORC-DNA interactions outside of the core binding region do differ.
ORC subunits in the major groove of ARS1. Because the cross-linking experiments described above examine a region within a 5-to 10-Å radius from the phosphate backbone, they cannot distinguish between proteins that are directly interacting with DNA and those that are merely in close proximity to the DNA. To identify ORC subunits within van der Waals distance of thymines in the major groove of the DNA, BrdU cross-linking to ARS1 was performed. DNA fragments that incorporated BrdU and radiolabeled nucleotides into either the top or the bottom strand of ARS1 were generated and cross-linked to ORC. DNA fragments modified on either strand efficiently cross-linked to Orc2p and Orc4p (Fig. 8).
Orc1p cross-linked to the bottom strand of ARS1, although weak Orc1p cross-linking to the top-strand was occasionally observed (data not shown). BrdU cross-linking with both strands of ARS305 also detected these three ORC subunits (data not shown).

DISCUSSION
The interaction of ORC with origin DNA involves six proteins, with a combined molecular mass of 414 kDa, that specifically recognize a region of DNA approximately 30 bp in length. We found that the coordinate action of five of the six ORC subunits is required for this interaction (Fig. 9). Recognition of critical DNA residues at two genetically defined re-FIG. 5. 4Ј-AZPB cross-linking of a tagged Orc4p-containing ORC. (A) Protein composition of the Sp1-Orc4p fusion complex. We altered the electrophoretic mobility of Orc4p by fusing it to the C-terminal 168 amino acids of the Sp1 transcription factor. This fusion complemented a deletion of ORC4 in yeast cells, and an ORC complex with this Sp1-Orc4p hybrid as the sole copy of Orc4p showed normal DNA binding properties (data not shown). The Sp1-Orc4pcontaining complex was partially purified and electrophoresed beside purified wild-type (WT) ORC. Silver staining and immunoblot analysis indicate that the Sp1-Orc4p fusion migrates to a position above that of the Orc2p band, as expected for a protein of 82 kDa. Orc6p normally runs as doublet due to phosphorylation, and the hybrid complex contains mostly the phosphorylated form. We detect no differences in the in vitro properties of wild-type complexes before and after treatment with phosphatase (1b). (B) UV cross-linking of wild-type ORC and an ORC complex containing the Sp1-Orc4p fusion to ARS1 cross-linking probes B and D. Cross-linking was performed as described in the legend to Fig. 4A. gions of origins (the ACS and B1-like elements) is likely to be mediated by nonoverlapping ORC subunits or groups of subunits. Although we have no evidence of direct DNA contact by Orc5p, the failure to cross-link any other ORC subunit to the B1 regions of both ARS1 and ARS305 strongly implicates this subunit in the interaction with B1 residues. At the ACS, Orc1p, Orc2p, and Orc4p interact with the major groove, as all three subunits are cross-linked by BrdU to both ARS1 and ARS305 and are within 10 Å of the ACS of both origins. The ORC subunits bound at B1 and the ACS interact physically, as partial complex analysis indicates that Orc4p requires Orc5p to associate stably with the remainder of the complex. Together, our studies provide a picture of the ORC-origin association that forms a foundation for other structures assembled at origins.
Conservation of ORC-DNA interactions at the origin core. Our studies argue that the manners in which ORC interacts with its binding site at different yeast origins of replication are similar. A comparison of nine yeast origins showed that the four most highly conserved residues outside of the ACS fall within regions of ARS1 that strongly inhibit ORC-DNA binding when chemically modified (residues A839, A840, A852, and A854 [ Fig. 9]). These residues also fall within regions required for efficient origin function in origins whose structures have been well characterized. The residues at positions 839 and 840 of ARS1 are conserved in the B1-like elements of ARS305 (24) and ARS307 (35,43), and residues at positions 852 and 854 of ARS1 fall within the box 3Ј element of ARS305 (24), the extended A element of ARS307 (35,43), and the extended core region of ARS121 (46). The similar organizations of ORC subunits over the core binding site at ARS1 and ARS305 also argue that features of ORC-origin binding are conserved (Fig.   7) and that this view of the ORC-core origin interaction is likely to be generally applicable to all yeast origins.
The conservation of ORC subunit arrangement, however, does not extend beyond the core binding region. At ARS1, in addition to being cross-linked to a region within the core origin, Orc2p, Orc4p, and Orc5p are cross-linked to a second site in the flanking DNA. This interaction with two distinct regions of DNA is consistent with one of the following interpretations: (iv) the DNA at ARS1 is bent such that distant regions of the DNA are in close proximity to the same polypeptide. Although the differences in protein-DNA cross-linking observed at ARS1 and ARS305 could be explained by differences in stoichiometries of ORC or ORC subunits bound at these two origins, this interpretation is unlikely, as ORC-DNA complexes migrate similarly when ARS1 and ARS305 DNA fragments are used in electrophoretic mobility shift assays (data not shown). We favor DNA bending as the most likely explanation for the discrepancy in cross-linking results, an interpretation that is consistent with the ability of ORC to induce DNA bends at ARS1 but not at ARS305 (Fig. 2B). Thus, the association of ORC subunits with the core origin is likely to be similar at all origins, but higher-level interactions like DNA bending may strongly influence cross-linking outside of the core ORC-DNA binding site.
Implications of ORC-origin architecture for ORC function. ORC interacts with yeast origins by making multiple protein-DNA contacts, a strategy commonly used by initiator proteins to induce DNA distortion. For example, the Epstein-Barr virus initiator protein, EBNA1, forms a dimer that must bind cooperatively to two adjacent binding sites for origin function in vivo (20). The crystal structure of the EBNA1 dimer bound to DNA has been resolved, and modeling of two dimers bound to two adjacent binding sites has been performed (8). Because the DNA binding sites are separated by only 3 bp, two dimers cannot co-occupy these adjacent sites unless the DNA in between them is distorted to prevent collision of the proteins. Typically, initiator proteins are homomultimers that make multiple protein-DNA contacts by interacting with repeated DNA elements. ORC, however, is a heteromultimer which likely binds as a single complex to yeast origins. Thus, ORC makes multiple contacts by utilizing different sets of subunits to contact distinct regions of a large DNA binding site. It remains to be determined if these multimeric interactions function only in the specificity of the ORC-DNA interaction or are also required for downstream steps of DNA replication initiation (e.g., DNA unwinding).
The observation that ORC is much more sensitive to modifications of the top strand of ARS1 than it is to bottom-strand modifications is intriguing since preferred interactions with only one strand of DNA is a mechanism used by initiator proteins to stabilize an unwound region of the origin. After the Escherichia coli DnaA protein binds to its origin, it induces melting of adjacent DNA (the repeated 13-mer site) in a process called open complex formation (11). In the open complex, DnaA preferentially interacts with one of the single strands of the unwound 13-mer region (11,25). The simian virus 40 initiator, T antigen, forms a double-hexamer structure encircling two DNA elements that are subsequently unwound or untwisted (for a review, see reference 10). Although the Tantigen hexamers encircle both strands of the DNA in these distorted regions, each hexamer contacts only one strand (41). Since ORC is bound to origins throughout most of the cell cycle, ORC-DNA binding is unlikely to be sufficient for origin unwinding. We therefore imagine three possible models whereby preferred interaction with one strand of DNA may be utilized by ORC. (i) ORC is responsible for initial DNA unwinding at origins but must interact with or be modified by another protein at the appropriate time in the cell cycle for this activity to occur. (ii) Another protein(s) recruited to origins performs the unwinding function, but ORC stabilizes the melted region of the duplex by binding to one of the single strands. (iii) ORC is passive in the unwinding process but uses single-stranded DNA binding to remain associated with origins of replication after they are unwound. It also remains possible, however, that ORC simply prefers to interact with one strand of a duplex and does not bind single-stranded DNA.
ORC subunit interactions. Our studies provided information regarding ORC subunit arrangement within the complex (Fig. 9). The 4Ј-AZPB cross-linking studies identified ORC subunits in close proximity to the same modified DNA residue, and these results are consistent with the subunit organization that we derive from partial-complex experiments. The composition of the partial complex lacking Orc5p suggests that Orc4p and Orc5p physically contact each other (Fig. 3B), and 4Ј-AZPB cross-linking demonstrated that these two subunits can be cross-linked to the same DNA site (Fig. 4A, probe J). Similarly, a complex lacking Orc3p is also deficient in Orc2p (Fig. 3B), and both subunits are cross-linked by a number of ARS1 and ARS305 probes (Fig. 4A, probes E, F, I; Fig. 6A, probe I 305 ).
UV cross-linking studies can only describe the arrangement of ORC subunits in the presence of DNA. However, in the absence of DNA, ORC is a preassembled complex, and we believe that the conformation and relative positions of its subunits may be different when the ORC is free in solution or associated with origins. ORC binds to DNA in a sequencespecific manner only in the presence of ATP or ␥-S-ATP (7,27). Furthermore, binding of ORC to ATP and of ORC to DNA are coordinated processes, since specific DNA binding affects both association of Orc1p with ATP and the subsequent rate of ATP hydrolysis by this subunit (27). Such coordinate action is likely to be mediated by allosteric changes within the complex. Thus, we are interested in determining how the relative positions of ORC subunits are affected by nucleotide and/or origin DNA binding.
Our view of ORC-origin association is necessarily limited by the static nature of the DNA-binding assay in vitro. Our understanding must ultimately be expanded to incorporate various cell cycle contexts and the effects of proteins whose association with ORC is cell cycle regulated. These initial studies FIG. 9. Model of ORC-origin core interactions. Residues of ARS1 whose chemical modification strongly interferes with ORC-DNA binding are indicated, and matches to the conserved ARS consensus sequence are shown in boldface uppercase letters. Outside of the ACS, boldface uppercase letters represent residues that are conserved in seven of nine yeast chromosomal origins of replication examined (A839, A852, and A854 of ARS1). A residue conserved in six of nine origins is shown as an uppercase letter (residue A840). Hypothesized protein-DNA and protein-protein interactions are indicated by double arrowheads. Protein-DNA contacts are made primarily with the top strand of origins at three clusters of conserved DNA residues. Specific binding of the ACS is mediated by the combined major-groove interactions of Orc1p, Orc2p, and Orc4p. Orc5p is likely to contact important residues in B1, and Orc6p is dispensable for specific ORC-DNA binding. Physical interactions between Orc2p and Orc3p and between Orc4p and Orc5p are deduced from two lines of evidence: (i) omission of one subunit results in the loss of the second subunit from the complex (Fig. 3B, partial complexes lacking Orc3p and Orc5p, respectively), and (ii) Individual 4Ј-AZPB probes cross-link both subunits. The nine origins used for sequence comparison are ARS1, ARS121, ARS305, ARS306, ARS307, ARS1413, the 2m ARS, the histone H4 ARS, and the HMRE ARS.
will provide a foundation for understanding changes in ORC properties that are induced during the cell cycle. Functional elements within yeast origins are arranged asymmetrically, and accordingly, ORC binds to origins by distributing its subunits asymmetrically along the DNA. Higher-order complexes assembled at origins during G 1 also reflect this asymmetry, as comparisons with the post-RC demonstrate that the pre-RC has an added region of DNase I protection on only one side of the region protected throughout the cell cycle (15). How the asymmetry inherent in the ORC-origin complex is ultimately translated into the assembly of two symmetric replication forks at origins of bidirectional DNA replication remains to be understood.