INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

When Epstein-Barr virus (EBV) infects cells latently, its circularized chromosome is replicated only once per cell cycle (1) by cellular enzymes, presumably under the same regulatory control that limits initiation of replication on cellular chromosomes. Only one EBV-encoded protein, EBNA1, is known to participate in this replication, which it does by directing replication to initiate at oriP. oriP is a 1.7-kb region of the EBV chromosome that permits plasmids that carry it to be replicated and stably maintained in human cells that contain EBNA1 (22, 37, 40). Replication initiates at or near a 120-bp region called DS (14), for dyad symmetry, which contains four EBNA1 binding sites. DS is also the replicator of oriP; that is, it contains the cis-acting elements that lead DNA replication to initiate (16, 36, 38). The other functional component of oriP, FR, for family of repeats, is an array of 20 EBNA1 binding sites that prevents the loss of plasmids from dividing cells (19), apparently by tethering the plasmids via EBNA1 to condensed human chromosomes during mitosis (18, 24, 30).

Replication of oriP-dependent plasmids occurs only once within a cell cycle (39), so it has long been suspected that DS may recruit factors that are involved in the mechanisms of origin licensing and delicensing (4), which are believed to control initiation of replication for the chromosomes of all eukaryotic cells. Indeed, emerging data indicate that DS functions by recruiting the human homologues of the eukaryotic initiation factors ORC and MCM and that MCM, the ultimate licensing factor, associates at or near DS during G₁ and dissociates during S phase (9, 10, 28). It thus appears that replication initiates at or near DS and uses essentially the same set of factors and mechanisms that operate at the replication origins of cellular chromosomes. However, DS depends entirely on EBNA1 for its replicator activity, and its only essential sequences are its EBNA1 binding sites (38). EBNA1 and ORC can each be detected in immunoprecipitates of the other from cell lysates (10, 28), so it is likely that EBNA1 is critical either for recruiting the cellular factors or for assisting them in some manner.

In principle EBNA1 might assist the replication initiation factors in different ways. It might recruit the factors to the vicinity in the manner of a transcriptional enhancer without a strict positional requirement. Where replication initiates, at or near DS, is known only to a resolution of a few hundred base pairs (14). Alternatively, EBNA1 might participate more directly by helping to assemble factors locally in a more precise manner. Previously it has been speculated that EBNA1 might have a direct role in initiation by deforming or slightly untwisting DNA (5, 11), although it is clear that EBNA1 is not a DNA helicase (13, 25). Finally, EBNA1 might exclude nucleosomes from a region to permit access by replication factors or it might position them strategically, as the presence of a nucleosome adjacent to ORC has been found to be important for the efficiency of the Ars1 origin in Saccharomyces cerevisiae (21).

The arrangement of the EBNA1 binding sites at DS suggests that a precise architecture could be important for its function. These four sites are arranged such that the spacing between sites 1 and 2 and between sites 3 and 4 is 21 bp center to center, exactly two turns of the double helix, which places the bound EBNA1 dimers on the same side of the DNA (Fig. 1A). Either pair of sites, sites 1 and 2 or sites 3 and 4, supports replication, while other combinations of sites do not (16, 38). Increasing the spacing between the EBNA1 sites of either active pair by 5 or 10 bp was found to eliminate the activity of the pair (16). These alterations can be viewed as drastic, particularly the insertion of 5 bp, which places the neighboring EBNA1 dimers on opposite sides of the DNA. Nevertheless, the results demonstrated the importance of the spacing between EBNA1 sites and accounted for the fact that the FR component of oriP does not have detectable EBNA1-dependent replication activity (16, 36, 38), since EBNA1 sites at FR are spaced at 30-bp intervals.

View larger version (37K): [in this window] [in a new window]   FIG. 1.   EBNA1 binding sites at DS of oriP and mutations that alter their spacing. (A) EBNA1 dimers (double spheres) bound to their four sites at the ~120-bp replicator DS. Distances between the centers of the binding sites and some restriction sites are indicated. (B) Nucleotide sequences between the active pairs of EBNA1 sites 3 and 4 and 1 and 2 for wild-type (WT) and mutants, with the outermost two nucleotides of each EBNA1 site underlined. The boundaries of the 16-bp palindromic EBNA1 binding site were determined in a study of synthetic sites and from crystallographic data (3, 5). Substituted or inserted nucleotides are in bold; dots indicate deleted nucleotides.

For this study, we tested just how critical the spacing between EBNA1 binding sites might be for replicator function by altering the spacing by just 1 or 2 bp. Such mutations abolished activity, indicating that a very specific structure composed of adjacent EBNA1 dimers is essential. While measuring the binding of EBNA1 to sites with altered spacing, we found that EBNA1 bends DNA when it binds and estimated that binding of EBNA1 to an active pair of sites at DS bends the DNA by perhaps 90° or more.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Plasmids and oriP mutations. All mutations in oriP were constructed in pHEBo, a 7.2-kb plasmid that contains oriP and that confers on human cells resistance to hygromycin B (32). Plasmids harboring the dpm (double point mutation) mutations that inactivate individual EBNA1 binding sites and the 5- and 10-bp insertion mutations (16) were kindly provided by Janet Hearing. Plasmids with DS (p571), FR (p413), 2&4 (p658), and 1&2 (p740) alterations in oriP have been described (38). In p740, EBNA1 sites 1 and 2 at DS have been replaced by an XbaI site and the EcoRV site at nucleotide (nt) 8992 has been converted to an SstI site. An AvrII site was created between EBNA1 sites 3 and 4 of p740, as indicated in Fig. 1B, using divergent primers 5'-CGCCTAGGGTAACATATGCTATTGAATTAGGGT and 5'-CGCCTAGGGGAAGCATATGCTATCGAATTAGGGT to amplify DNA segments in each direction by PCR; the flanking primers used for amplification were ODJ4 (5'-GGAATCCTGACCCCATGT) and ODJ10 (5'-AACGTCAATCAGAGGGGC), which have 5' ends at nt 8943 and 9220, respectively. The two products were cut with AvrII, ligated together, cut with SstI and XbaI, and used to replace this section of p740. The 1-bp deletion at nt 9053 was found unintentionally in one clone that was tested for this construction. The other deletions and insertions between sites 3 and 4 were then made using primers containing the AvrII site and the desired mutations to amplify site 4 and sequences extending leftward beyond the SstI site. The 1&2 plasmids of Fig. 2C lack EBNA1 sites 3 and 4 at DS and are based on p715 (38); 1&2c (p718) is similar but has consensus mutations in sites 1 and 2 that slightly increase replication activity (38). The 1-bp insertion between sites 1 and 2 was introduced into p718 using a primer including the XbaI site to the left of site 2 and extending into site 1 to amplify DNA extending rightward beyond the HpaI site, which was used to replace the DNA between the XbaI and HpaI sites of p718. The resulting "+1" mutant plasmid also acquired a partial copy of this sequence, which added half of site 2 followed by an intact site 1 at the HpaI site. Mutant plasmid 1&2m contains the intended 1-bp insertion between sites 1 and 2 but also has 1 bp deleted from the center of site 2.

Plasmid replication assays. EBNA1-expressing 143B cells were described previously as 143/SVoB-H2.9, clone 4 (40). Assays for plasmid replication assays following transient transfection and for maintenance under selection were done as described in detail previously (38). Briefly, for transient transfections, duplicate 6-cm-diameter dishes of cells were transfected with 2.5 µg of plasmid as calcium phosphate coprecipitates followed by a glycerol shock, and the dishes in each pair were expanded into two 10-cm-diameter dishes the next day. To measure EBNA1-independent replication, transfected 293 cells were washed twice in phosphate-buffered saline containing 1 mM EDTA to remove extracellular plasmids before replating. Plasmids were extracted using a standard alkaline lysis protocol, digested with BamHI and DpnI, and detected by Southern analysis. For maintenance under selection, just 1/10 of the cells were replated into 6-cm-diameter dishes the day after transfection and cultured in the presence of 275 µg of hygromycin B per ml beginning the following day. Plasmids that failed to replicate gave primarily colonies that aborted by 2 weeks (due to the maintenance function of FR of oriP and EBNA1 [19, 27]), so approximately 4 weeks were required to expand these cultures enough to cover 10-cm-diameter dishes before harvesting plasmids. For replication-competent plasmids that gave stable colonies under selection, the cultures were split back as needed and harvested after 3 weeks under selection.

Electrophoretic mobility shift assays (EMSA). Assays were performed as described previously (38). EBNA1 N407 was a generous gift from David Mackey and Bill Sugden. End-labeled DNA fragments spanning DS were made using primers ODJ4 and ODJ10 to extend from nt 8943 to 9220 except when testing DS versions carrying the 47-bp deletion removing EBNA1 sites 1 and 2, in which case the boundary on the right was moved to nt 9266 to keep the size and center of the fragment similar to those of the other fragments.

Circular permutation assay. Using p740 (38), which lacks EBNA1 sites 1 and 2, as a template for PCR, the sequence from nt 8967 to 9295 was amplified using primers that added a BglII site to the left end and a BamHI site to the right end. The DNA fragment was digested with BstYI, which cuts both restriction sites, and inserted in tandem duplication between the BglII and BamHI sites of a derivative of pUC12 which contains a BglII site inserted at the EcoRI site. The duplicated region was amplified using standard sequencing primers and labeled internally with [-³²P]dATP at a specific activity of 3 µCi/nmol. The product was cut with different restriction enzymes, and the unit length fragments were purified by agarose gel electrophoresis and diluted with unlabeled fragments to a specific activity of 4,000 cpm per 15 fmol for use in EMSA with a 4% polyacrylamide gel. The electrophoretic mobilities were plotted using Origin, version 5, software, and curves were drawn using the spline curve function. The data were plotted twice in tandem to extend for two cycles so that the curves would be continuous at the ends. Only a single cycle is shown in Fig. 4C.

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

EBNA1 sites must be exactly 21 bp apart at their centers to support replication. To test the importance of the spacing between EBNA1 binding sites 3 and 4 at DS, we used a derivative of oriP-containing plasmid pHEBo, from which EBNA1 binding sites 1 and 2 at DS were deleted and which thus depends on sites 3 and 4 to replicate (38). Just 5 nucleotides, CGTTG, separate the boundaries of the 16-bp EBNA1 binding sites 3 and 4. The identity of the bases at these five positions is unlikely to be very important since sites 1 and 2 are separated by ACCCG. To facilitate the construction of mutants with altered spacing between these sites, 2 bp between sites 3 and 4 were altered to create an AvrII restriction site (Fig. 1). This altered DS half, called 3&4, supported plasmid replication about as well as wild-type DS during 2 to 3 days following transfection of EBNA1-containing 143B cells (Fig. 2A) and during 3 weeks while the cells were grown under selection with hygromycin B (Fig. 2B). Deletions of 1 or 2 bp and an insertion of 2 bp were introduced between sites 3 and 4, as shown in Fig. 1B. Each of the mutations that altered the spacing between the sites reduced replication activity to the background level in the transient transfection assay and eliminated plasmid maintenance under selection (Fig. 2A and B).

View larger version (49K): [in this window] [in a new window]   FIG. 2.   EBNA1 binding sites at DS must be exactly 21 bp apart, center to center, to support plasmid replication. (A) At 60 h after transfection of EBNA1-positive 143B cells, plasmids were isolated, digested with BamHI to linearize them and with DpnI to test for replication (loss of bacterial adenine methylation), and detected by Southern analysis. Arrow, full-length, DpnI-resistant plasmid. Plasmids carried oriP with DS intact (WT), with only EBNA1 sites 2 and 4 active (2&4), with only EBNA1 sites 3 and 4 present (3&4), or with 3&4 with altered spacing (1, 2, and +2). 3&4c and its derivatives, 2c and +2c, also contain the T-to-A consensus mutation at nt 9046 of site 4 (38). The average signals for the replicated plasmid relative to that for the wild type are shown above the blot image. Duplicate transfections were analyzed except for 1 (nt 9050) in lane 10. For lane 1, 4 ng of plasmid was mixed with DNA from nontransfected cells and treated similarly, as a control for DpnI digestion. (B) Test for maintenance of the same set of plasmids in cells grown under selection for 3 to 4 weeks. The Southern analysis of uncut plasmids extracted from 5 × 106 cells is shown. In the last two lanes, 50 and 250 pg of pHEBo was loaded, corresponding to 1.5 and 15 copies per cell. The supercoiled (S) and relaxed circular (RC) forms of the plasmid are indicated at the left. (C) Assay for replication of plasmids containing only EBNA1 sites 1 and 2 at DS during 50 h following transfection, as for panel A, but with only the most relevant part of the blot image shown. 1&2c contains consensus mutations in sites 1 and 2, as does the +1 mutant, for which duplicate transfections of two different plasmid preparations are shown (lanes 7 to 10). 1&2m has 1 bp deleted from the center of site 2. (D) Mutations at DS do not affect the inefficient replication of plasmids that can be detected in the absence of EBNA1. At 47 h after transfection of 293 cells, plasmids were extracted and analyzed as for panel A. Forty-five percent of the DNA was digested with BamHI and DpnI for the upper blot; for the lower blot, 5% of each was digested with BamHI to measure DNA uptake by cells. The average replication of each relative to that of pHEBo (WT), normalized for DNA uptake, is indicated above the upper image. The vector lacking oriP was tested for lanes 1 and 2.

The spacing mutants abolish DS activity by interfering with EBNA1 function. Some replication activity that is independent of EBNA1 has been attributed to oriP (2, 18), although such activity appears to be very inefficient compared to EBNA1-dependent replication (18, 38). To test whether the altered spacing between sites 3 and 4 might affect replication in the absence of EBNA1, plasmids were introduced into 293 cells and harvested 48 h later for analysis. In several experiments, one of which is shown in Fig. 2D, pHEBo, carrying wild-type oriP, replicated 2-fold to 3.5-fold more efficiently than vector pHyg did alone. The small magnitude of this difference and the weak signals made it necessary to normalize the amount of replicated, DpnI-resistant plasmids to the total amount of plasmids taken up by the cells (Fig. 2D, bottom). Deletion of FR from oriP lowered the efficiency of replication by nearly 50%, on average, but deletion of DS had no measurable effect, as noted previously (38). Likewise, in the absence of EBNA1, the 2-bp insertion between sites 3 and 4 had no effect on the inefficient replication of the plasmid carrying only these sites at DS.

Binding of EBNA1 to sites with altered spacing. For each functional pair of EBNA1 binding sites at DS there is one site with relatively high affinity, site 1 or 4, and one site with relatively low affinity, site 2 or 3. At limiting EBNA1 concentrations, binding to each weaker site can depend on binding to its partner site of higher affinity, revealing cooperativity (16, 33). We tested how the altered spacing between sites 3 and 4 affected their ability to bind EBNA1 using EMSA as described previously (38). We also examined the 5- and 10-bp insertions previously studied by Harrison et al. (16). It was necessary to use the DNA-binding domain of EBNA1 without the amino-terminal half of the protein, which causes aggregation of EBNA1-DNA complexes (25).

View larger version (24K): [in this window] [in a new window]   FIG. 3.   Binding of EBNA1 to sites 3 and 4 with normal or altered spacing. wt, normal spacing between sites 3 and 4; the changes in spacing are indicated by 2 to +10 (base pairs). (A) Electrophoresis of 15 fmol of end-labeled DNAs, with the indicated spacing between sites 3 and 4, through a 4% polyacrylamide gel after binding by EBNA1 N407 in a limiting amount (30 fmol of dimer; lanes 2 to 8) or in excess (60 fmol of dimer; lanes 9 to 15) or without EBNA1 (lane 1). The DNAs spanned DS but had EBNA1 sites 1 and 2 deleted (lanes 2 to 5 and 9 to 12) or inactivated by point mutations (lanes 6 to 8 and 13 to 15). In both cases, the DNAs with wild-type spacing were 277 and 278 bp long, respectively, with the center 11 bp to the right of site 3. At the left, the positions of free DNA (F) and the complexes with one site or two sites bound are indicated. (B) Helical phase analysis of DNA bending at sites 3 and 4. Shown are the mobilities of the complexes relative to that of free DNA, measured using the gel shown in panel A and a similar one. The mean values of three determinations (six for wild-type spacing) are indicated; brackets, range of measurement.

DNA bending by EBNA1 bound to one or multiple sites. The results of Fig. 3A indicated that EMSA could be used to gain information about the architecture of DS complexes with multiple sites bound by EBNA1. An initial step was to investigate the bending of DS DNA with EBNA1 bound to one or both sites of an active pair using a circular permutation assay (35). The basis for the assay is that a bend in DNA will reduce electrophoretic mobility the most when it is at the center of a linear fragment, whereas the closer the bend is to an end, the less effect it will have.

View larger version (20K): [in this window] [in a new window]   FIG. 4.   Circular permutation assay for EBNA1-induced DNA bending. Circularly permuted 288-bp DNAs containing the left half of DS were generated from the construct shown in panel B. EBNA1 sites 3 and 4 are indicated. (A) Internally labeled DNA (15 fmol) was excised at the restriction sites indicated, mixed with 60 fmol of EBNA1 N407, and electrophoresed through a 4% polyacrylamide gel. The positions of free DNA (F) and the complexes with one site or two sites bound are indicated at the left. For lanes 1 to 6, the DNA was excised with XbaI, MwoI, ApaI, BstYI, Bsu36I, and TaqI, respectively. (C) Mobilities of the complexes relative to that of unbound DNA, measured from two independent gels and plotted against the position of the end of each DNA within the nonpermuted sequence.

Changing the helical phase between two EBNA1-induced bends. With 10.5 bp per average turn of B form DNA (26), the 21-bp spacing between the centers of the EBNA1 binding sites of a functional pair is expected to place the two protein dimers in helical phase. This places both DNA bends in the same geometric plane, which makes the bend angles directly additive, giving twice the bend. (Of course, interactions between adjacent bound EBNA1 dimers could affect the bending.) Inserting 5 bp between the sites changes the helical phase by about 180 degrees, causing the two bends to cancel each other in a zigzag. Inserting 10 bp changes the phase very little, allowing the two bends to add directly, giving the greatest net bend. Insertions of intermediate sizes reduce net bending intermediately. This is the basis for helical phase analysis of DNA bending (35).

Bending of DS DNA by EBNA1 binding to up to four sites. When limiting amounts of the EBNA1 DNA-binding domain are allowed to bind to the entire DS, complexes with five different mobilities are observed, as shown in Fig. 5A, lane 2, rather than four, as would be expected without considering DNA bending and phasing. The explanation is that, when two sites are bound, two different amounts of net bending are possible, depending on whether the two sites are within an active pair, and thus are in helical phase, or not. Only one amount of net bending is possible when only one site is bound, when only one site is not bound, or when all four sites are bound. An experiment using point mutations to prevent binding to specific sites, shown in Fig. 5B, revealed that the complex called 2 IP (for in phase) contains the complexes formed when the sites of an active pair, 1 and 2 or 3 and 4, are both occupied. Other combinations of two sites, where one site is from each functional half of DS (primarily 1 and 4, the two stronger sites), are responsible for the complex called 2 OP (out of phase). 2 OP migrates faster than 2 IP because 33 bp separates the centers of sites 2 and 3, which means that sites 1 and 4 (or 2 and 3 or 1 and 3 or 2 and 4) are out of phase by about 1.5 bp or about 50° of rotation, causing the two DNA bends to be less than fully additive.

View larger version (36K): [in this window] [in a new window]   FIG. 5.   EMSA analysis of five types of complexes that result from binding of limiting amounts of EBNA1 to its four sites at DS. The complexes are labeled 1, 2 IP, 2 OP, 3, and 4 based on the number of EBNA1 binding sites that are bound by EBNA1 in each complex. IP and OP are explained in the text. (A) Effects of eliminating each EBNA1 site individually. The substitution mutations, 1-Bst, 2-Bst, etc., eliminate binding to each site (38). In lanes 3 to 6, the EBNA1 sites that are bound in the predominant 2 OP complex are indicated in each lane. At the right, the two EBNA1 sites that are occupied in each 2 IP complex are indicated. F, free DNA. (B) Effects of eliminating binding to different pairs of sites. The DNAs that were tested each contained double point mutations (dpm) in two different binding sites, as constructed by Harrsion et al. (16). For example, sites 2 and 3 are both inactive with dpm2 + 3. Lanes 6 to 10 are a repetition of lanes 1 to 5 except that the ratio of EBNA1 dimers to binding sites is doubled. (C) Effects of increasing the distance between sites 2 and 3 by 5 or 10 bp using the oriP mutants in2/3[5] and in2/3[10], respectively, described by Harrison et al. (16). (A to C) Fifteen femtomoles of each DNA was used with the following amounts of EBNA1 N407 dimer: A and C, none for lane 1, 60 fmol otherwise; B, 30 fmol for lanes 1 to 4, 60 fmol for lanes 5 to 9, 120 fmol for lane 10, none for lane 11.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

The oriP replicator, DS, appears to function through cellular initiation factors ORC and MCM (9, 10, 28), but it is completely dependent on EBV-encoded protein EBNA1 for activity. It is clear from this study that the replicator requires two EBNA1 binding sites that are spaced exactly 21 bp apart, center to center, implying that an exact structure must be formed by two adjacent EBNA1 dimers bound to DNA. The two EBNA1 dimers, adjacent on the DNA and in helical phase, bend the DNA substantially, so it is possible that EBNA1 performs an architectural role, shaping the DNA to facilitate an association with cellular factors. It is also conceivable that the large DNA bend is itself important. Alternatively, the DNA bend might simply allow two EBNA1 dimers to form a particular structure, which is then able to recruit ORC and associated replication factors.

The one complicating factor with this conclusion is that each functional pair of EBNA1 binding sites at DS contains one site of relatively low affinity, to which binding by EBNA1 is aided by the presence of EBNA1 at the adjacent, higher-affinity site (16, 33). This cooperative binding to the weaker sites is mediated, at least in part, by interactions between the DNA-binding domains of adjacent EBNA1 dimers, as shown in this study and previously (33). The crystal structure that was derived for the EBNA1 DNA-binding domain bound to DNA revealed that EBNA1 DNA-binding domains should be in close contact when bound to adjacent DS sites (5). As would be expected, this cooperative interaction was found to require correct spacing between the sites (Fig. 3). In contrast, in a study using full-length EBNA1, the cooperativity of binding to the weaker sites was not noticeably diminished by increasing the distance between adjacent sites by 10 bp or, even more surprisingly, by 5 bp (16). Presumably, this longer-range cooperativity arises from the interactions between the presumably flexible amino-terminal domains of EBNA1, which have been shown to mediate DNA looping between EBNA1 molecules bound to the two components of oriP (12, 31). Based on this, our 1- or 2-bp insertions between the EBNA1 sites of a functional pair might not be expected to result in a loss of EBNA1 binding to the weaker sites in vivo, but this cannot be assured.

To address this issue, we converted site 3 to a high-affinity site by mutating 2 bp. While this enabled EBNA1 to bind to the strengthened site 3 independently of its distance from site 4, replication still required the correct spacing (data not shown). This is consistent with there being a need for two EBNA1 dimers to form a specific structure. However, this issue cannot be resolved entirely without knowing how these mutations affect the binding of EBNA1 to sites 3 and 4 in vivo.

In 1996 and 1998 the crystal structure for the EBNA1 DNA-binding domain bound to its consensus recognition sequence was reported by Bochkarev et al. (5, 6). The DNA structure was described as a distorted B form wrapping smoothly around the protein, but a value for the net angle of bending around the protein dimer was not given. In discussing how EBNA1 dimers might interact at adjacent DS sites, the authors presented a model depicting DNA bent around the EBNA1 dimers and proposed that the DNA might need to "unbend" in the middle. However, to our knowledge it has not been stated in a publication that EBNA1 induces DNA to bend when it binds until now, so DNA bending by EBNA1 has remained largely unappreciated. For example, in a 1998 publication it was noted that EBNA1 binding to DS gives complexes with five different electrophoretic mobilities instead of four, corresponding to the four binding sites, and it was suggested that a fifth EBNA1 dimer might join the EBNA1-DS complex (41). Of course, the five distinct mobilities are simply a consequence of DNA bending at the sites of binding and the helical phasing between the bends (see Results).

The data presented in this report show that EBNA1 bends DNA in solution, confirming this aspect of the crystal structure. In addition, while we did not examine bending at each EBNA1 binding site at DS individually, the results indicated that EBNA1 induces bending at all sites to a similar extent. It is of interest that EBNA1 appeared to bend DNA at sites 3 and 4 to similar extents whether the two sites had their proper spacing or were separated by an additional 10 bp. This indicates that the inferred interaction between the EBNA1 dimers at adjacent DS sites does not change the DNA conformation to any great extent. It is also interesting that EBNA1 was still able to bind to sites 3 and 4 simultaneously when the sites were brought closer together by 1 or 2 bp. The binding was unstable and resulted in less bending, suggesting a collision between the adjacent dimers, but the fact that binding could be detected indicates that some flexibility of the EBNA1-DNA structure is possible despite a cost in binding energy.

The angle by which proteins induce DNA to bend can be estimated from the electrophoretic mobility of the complex through polyacrylamide gels using the equation of Thompson and Landy, µ_M /µ_E = cos /2, where µ_M and µ_E are the relative mobilities with the bend placed in the middle of the DNA fragment or at the end, respectively, and  is the angle of bending (34). This equation gave bending angles of 55° for EBNA1 bound to site 4 and 88° for EBNA1 bound to sites 3 and 4 using the results of Fig. 3C, which were obtained with a 4% polyacrylamide gel. The calculation is expected to underestimate the angle, particularly for the complex with both sites bound because the bend can only be placed near the end, not at it, so the effect of the bend cannot be eliminated from µ_E. Another problem is that we found that the ratio µ_M/µ_E decreased as the concentration of polyacrylamide in the gel was increased, giving calculated angles of 64° and 104° with 6% polyacrylamide and 76° and 112° with 8% polyacrylamide for the complexes with one site or both sites bound, respectively. Further work might reveal the reason for this effect, in which case a comparison to DNA standards with known angles of bending could give a reasonable estimate of the angle of bending induced by EBNA1 in solution. The DNA-binding domain of the E2 protein of bovine papillomavirus, which has a structure very similar to that of the EBNA1 core DNA-binding domain, was found to bend DNA by 50° in a cocrystal structure (17).

While one functional pair of EBNA1 binding sites at DS supports plasmid replication, the entire DS is needed for its full efficiency (38). The two functional pairs of EBNA1 binding sites at DS are out of helical phase with each other by 1.5 bp, assuming a helical pitch of 10.5 bp per turn, or 51° out of phase rotationally. If DNA bends 100° around each functional pair of EBNA1 dimers, this would mean that the flanking DNA helices would enter and leave the EBNA1-DS complex at an angle of about 70°. Consistent with this, Frappier and O'Donnell observed that complexes formed between EBNA1 and DS in vitro appeared as a small ball by electron microscopy, with the two arms of DNA protruding from the same side of the complex at an angle of 71° ± 45° to each other (12).

It is a common feature of DNA replication origins that their initiator proteins assemble into larger complexes with the origin DNA bent or wrapped around them. This was first noted for the O protein at the bacteriophage lambda origin and for the dnaA protein at oriC of E. coli, which participate in the initial unwinding of DNA for replication (8). The simian virus 40 initiator protein T antigen first binds to four close sites at the viral replication origin before additional molecules join to assemble into double hexamers that function as DNA helicases, with DNA bending in this case induced adjacent to the sites of T-antigen recognition (7, 29). At the replication origin of the distantly related bovine papillomavirus, cooperative DNA binding involving the T-antigen homolog, E1, and a second protein, E2, produces a sharp DNA bend (15). At replication origins of yeast chromosomes, DNA may wrap around the six-subunit origin recognition complex, ORC (20), which recruits other proteins to initiate replication.

It has been speculated that in some cases the bending of DNA at replication origins might facilitate DNA unwinding. But the most common principle might be a need for origin DNA to interact with a large complex of proteins, which DNA bending will often facilitate. This is the simplest explanation for the requirement for an exact spacing between two EBNA1 sites at the replicator of oriP; a specific structure involving two DNA-bound EBNA1 dimers might be needed to recruit ORC and associated replication initiation factors.