Studies on the Mechanism of DNA Linking by Epstein-Barr Virus Nuclear Antigen 1*

Epstein-Barr virus nuclear antigen 1 (EBNA1) can both bind to and link DNA. Dimers of EBNA1 bind specific sites, two clusters of which, the FR and DS, comprise the necessarycis-acting elements of the Epstein-Barr viral origin of plasmid replication. EBNA1-dimers can link FR and DS, looping out the intervening DNA. EBNA1 can also intermolecularly link DNAs to which it binds. Residues of EBNA1 that can mediate linking have been mapped to at least three, non-overlapping domains. These domains, when fused to the dimerization and DNA-binding domain of GAL4, can self-associate and thereby link DNAs bound site specifically by GAL4. Two disparate mechanisms could underlie self-association of linking domains: 1) linking domains could associate with other linking domains directly, or 2) linking domains could associate indirectly by binding to a common nucleic acid intermediate. We have found that EBNA1 can link DNA by each of these mechanisms, however, the linking domains associate directly with a greater apparent affinity than through a nonspecific nucleic acid intermediate.

The Epstein-Barr virus (EBV) 1 infects human B-lymphocytes and usually establishes a latent infection in them. In vivo and in vitro, the latently infected cells are induced to proliferate. Interestingly, the viral genome is maintained in these cells as a plasmid which is both replicated conservatively during S-phase and maintained efficiently at a stable copy number (1)(2)(3)(4). Only one of the latent viral gene products, EBNA1, and a small (1.8 kbp) cis-acting element (oriP) are required to recapitulate faithful plasmid replication in human and some other cells (5)(6)(7). Replication of oriP plasmids provides a useful model for studying control of initiation of replication, segregation of replicated DNAs, and maintenance of those DNAs in the mammalian nucleus.
Dimers of EBNA1 bind specifically to degenerate 20-bp sequences of DNA (8). The carboxyl-terminal one-third of EBNA1 contains the residues sufficient for both dimerization and DNA binding ( Fig. 1) (9 -12). (In this report dimerization of EBNA1 refers to protein-protein interactions between the DNA-binding domain of EBNA1, and not to protein-protein interactions be-tween linking domains.) The EBV genome contains 26 identified sites to which EBNA1 binds (13). Twenty-four of these sites are within two clusters which comprise oriP (5). Twenty sites with a high affinity for EBNA1 are embedded within a series of 30-bp repeats, termed the family of repeats (FR). The dyad symmetry element (DS), which is located 1 kbp away from FR contains 4 binding sites for EBNA1 with lower affinity than those in FR, two of which are part of a 65-base pair dyad (5,8,14,15). EBNA1, when bound to FR, can activate transcription of two viral promoters, one of which is 10 kbp away (16 -18). The ability of EBNA1 to bind to DNA is essential for its activation of replication and transcription through oriP (19,20).
In addition to binding to FR and DS, EBNA1 can also link them, forming a loop of the intervening DNA (21,22). Activities of EBNA1 other than DNA binding and DNA linking have not been identified. EBNA1 purified from insect and mammalian cells lacks detectable helicase or ATPase activity (23,24). EBNA1's apparent lack of enzymatic activities led several labs to search for proteins with which EBNA1 can interact. The most intriguing candidate thus far reported is EBNA1 itself. Does the ability of EBNA1 to link DNA contribute to its activation of transcription and replication? One study demonstrated that no small deletion within EBNA1, other than those which affect DNA-binding, abrogates the ability of EBNA1 to activate transcription or replication (19). The authors interpreted this finding to indicate that EBNA1 contains redundant activating domains. The linking domains of EBNA1 are redundant and therefore are reasonable candidates for its activating domains. Findings in another study support this contention (25). A derivative of EBNA1 lacking one of the three linking domains can activate transcription 9 -28% and replication 4 -28% as well as wild-type EBNA1. Derivatives of EBNA1 lacking all three linking domains fail to activate transcription or replication. DNA linking by EBNA1 is therefore likely to contribute to its activation of transcription and replication. EBNA1 can form intramolecular loops between FR and DS and can also intermolecularly link DNAs to which it binds (21, 22, 24, 26 -28). We have studied linking primarily in electrophoretic mobility shift assays (gel shifts) (28). Multiple derivatives of EBNA1 behave similarly in gel shift assays and three other described linking assays (26,28). In a gel shift assay, DNAs linked by EBNA1 do not migrate appreciably into a 4% polyacrylamide gel. The efficiency with which a DNA is incorporated into a linked complex increases with the number of EBNA1-binding sites it contains, but linking of DNAs containing only one site can be detected. We have identified three regions of EBNA1, amino acids 54 -89, 331-361, and 372-391, which mediate DNA linking. When fused to the dimerization and DNA-binding domain of GAL4 these fragments of EBNA1 mediate linking of DNAs containing five GAL4-binding sites. We also observed that increasing concentrations of linking protein decrease the efficiency of linking. This phenomenon, reminiscent of dissolution of antibody-antigen complexes by excess antigen, indicates that protein-protein interactions between the linking domains may underlie linking. We have tested this hypothesis.
Several mechanisms could explain how the DNA-linking domains of EBNA1 associate and thereby link DNA. Two possible mechanisms are: the linking domains associate through direct protein-protein interactions; and the linking domains associate indirectly by protein-nucleic acid interactions with common intermediates. The first putative mechanism invokes only protein-protein interactions. Because linking domains of EBNA1 function when fused to the exogenous DNA-binding protein, GAL4, models in which a linking domain interacts specifically with a DNA-binding domain cannot be supported. Therefore, if linking is mediated by protein-protein interactions, these interactions are likely to be between the linking domains themselves. The different linking domains of EBNA1 could associate heterotypically with one another or each could associate only homotypically. Linking domains could also associate by binding to nonspecific nucleic acid intermediates. In addition to the DNA bound site specifically by EBNA1, our linking assays contain several sources of nucleic acid: poly(dI)⅐poly(dC), nonspecific DNA flanking the EBNA1-binding site(s), and any nucleic acid contaminating our protein preparations. Binding of the linking domains to one or more of these nonspecific nucleic acids could incorporate specifically bound DNAs into a lattice. The linking domains are highly basic, and might therefore be anticipated to bind nucleic acids. RGG motifs found within some of the linking domains have been reported to bind RNA (29). Both mechanisms described can contribute to DNA linking by EBNA1.
The efficiency of DNA linking mediated by EBNA1 increases and then decreases as the concentration of EBNA1 increases relative to the DNA bound by it. Inhibition of linking is observed at concentrations of EBNA1 which are in excess relative to the EBNA1 bound to DNA site specifically, but are far less than the total DNA in the reaction. This observation indicates that the inhibition of linking is mediated by protein-protein interactions. Also consistent with this interpretation are our findings that proteins which contain linking domains can efficiently inhibit DNA linking and that peptides from EBNA1 which contain only linking domains can similarly inhibit linking. We also determined whether nucleic acids other than those bound site specifically by EBNA1 are essential for or can contribute to linking. Exclusion of poly(dI)⅐poly(dC) and of DNA flanking the EBNA1-binding sites from the linking reactions did not preclude linking. Treatment of the preparation of EBNA1 with RNase A and DNase I did not alter DNA linking. Nucleic acids, other than those bound site specifically by the DNA-binding domain of EBNA1, are not necessary for DNA linking by EBNA1. The presence of a high concentration of nonspecific DNA long enough to associate with multiple linking domains can, however, promote linking of DNAs bound by EBNA1. These results demonstrate that EBNA1 can link DNAs by protein-protein interactions between its linking domains.

EXPERIMENTAL PROCEDURES
DNAs-Two DNAs used as probes in this study were generated by endonuclease digestion of cloned DNAs. That with two EBNA1-binding sites is the 131 bp, SalI to PstI fragment derived from p880 (30); and that with five GAL4-binding sites is the 109-bp HindIII to XbaI fragment derived from G5BCAT (31). DNA fragments were purified from agarose gels. They were quantified by making serial dilutions in the presence of ethidium bromide (0.5 g/ml) and comparing their intensities, when exposed to ultraviolet light, to that of similar dilutions of a standard DNA. This method distinguishes less than 2-fold differences in DNA concentrations. These DNAs were labeled with the Klenow fragment of DNA polymerase I in the presence of dATP, dGTP, dTTP, and [␣-32 P]dCTP (Amersham), and precipitated three times to remove unincorporated label (32).
A.A. 40 -89 and A.A. 331-391 were generated by amplifying the fragment of EBNA1 encoding those amino acids with polymerase chain reaction primers containing SalI and XbaI sites at the 5Ј ends. Cloning of these fragments into the SalI and XbaI sites of pET-23bϩ (Novagen) resulted in fusions encoding an epitope (T7⅐Tag, Novagen), the EBNA1 amino acids, and 6 histidines. The sequence of both clones was verified using an ABI Prism automated sequencer (Applied BioSystems). Expression of the fusion proteins in BL21 bacteria containing a DE3 lysogen and carrying pLysS was done as described previously (28). Purification using Ni 2ϩ -NTA-agarose (Qiagen) followed standard protocols (The QIA expressionist, Qiagen). Following elution of A.A. 40 -89 and A.A. 331-391 in 500 and 150 mM imidazole, respectively, the peptides were dialyzed into a buffer containing 50 mM Tris (pH 7.5), 2.5 mM EDTA, and 0.3 M NaCl.
Nuclease Treatments-Treatments of 600 ng of N⌬360 with DNase I and RNase A were done in a buffer containing 50 mM Tris (pH 7.5), 0.25 mM EDTA, and 5 mM MgCl 2 . 1 g of RNase A (5 Prime 3 3 Prime Inc., Boulder, CO) and/or 10 units of DNase I (Boehringer Mannheim) were added to a 30-l reaction and incubated for 30 min at room temperature. This treatment degraded 95% of 300 ng of DNA to acid solubility. The amount of DNA contaminating 600 ng of N⌬360 is less than 3 ng as determined by comparison of the signal of intercalated ethidium bromide in a sample of N⌬360 with those in samples containing known amounts of DNA. EDTA was added to a final concentration of 16 mM either prior to or after incubation with DNase I.
Electrophoretic Mobility Shift Assays-Gel shift assays were done as described previously (28). All reactions contained 20 mM HEPES (pH 7.6), 2 mM EDTA, 10% glycerol, 0.1% Nonidet P-40, and 0.3 M NaCl. Proteins were mixed in volume of 15 l. The labeled DNA and 3 g of poly(dI)⅐poly(dC) (Pharmacia) in 8 l were added with mixing. The reactions were then incubated 30 -45 min at room temperature. Reactions were separated by electrophoresis at 15-25 mA through a 4% polyacrylamide gel in 0.5 ϫ TBE (45 mM Tris borate and 1 mM EDTA). Gels were dried on 3MM chromatography paper (Whatman) at 80°C for 1 h and then exposed to PhosphorImager screens (Molecular Dynamics). The percentage of DNA linked is determined by dividing the amount of linked DNA by the total amount of DNA in the lane.

RESULTS
Effect of Linking Domains on the Efficiency of DNA Linking-DNA linking by EBNA1 is inhibited at increasing concentrations of EBNA1 (28). Fusions of the DNA-binding domain of GAL4 and individual linking domains of EBNA1 also link DNAs containing GAL4-binding sites less efficiently as their concentration is increased beyond that required to yield maximal linking. We tested whether each type of linking protein could inhibit linking by the other type. The ability of GAL4(1-94)&54 -89 ( Fig. 1) to link DNAs containing five GAL4-binding sites was tested in the absence of any competitor and in the presence of a 2.5-fold molar excess of three derivatives of EBNA1 (Fig. 1). Derivatives of EBNA1 which contain linking domains, bEBNA1 and N⌬360, inhibited linking by the fusion protein ( Fig. 2A). N⌬389, which cannot link DNA, failed to inhibit linking by the fusion protein ( Fig. 2A). The converse experiment was also conducted (Fig. 2B). The ability of N⌬360 to link DNAs containing two EBNA1-binding sites was tested in the absence of any competitor and in the presence of a 4.5-fold molar excess of four GAL4 derivatives (Fig. 1). Derivatives of GAL4 fused to linking domains of EBNA1, GAL4(1-94)&54 -89, GAL4(1-94)&331-361, and GAL4(1-94)&372-391, inhibited linking by N⌬360. Unfused GAL4(1-94) did not inhibit linking by N⌬360. In these experiments there was a direct correlation between proteins that can inhibit DNA linking and proteins that contain DNA-linking domains.
Effect of Peptides Containing Linking Domains on DNA Linking by EBNA1-The observation that proteins with linking domains could inhibit DNA linking led us to test whether linking domains alone would be sufficient to inhibit DNA linking by EBNA1. For most experiments, N⌬360, which contains only one linking domain, was studied to facilitate detection of inhibition of linking. Two DNAs encoding fragments of EBNA1 (residues 40 -89 and 331-391, Fig. 1) were cloned into plasmids allowing efficient protein expression in bacteria. These fragments include all the amino acids from EBNA1 which are likely to contribute to DNA linking (28,34). The fragments of EBNA1 are fused at their amino terminus to an epitope and at their carboxyl terminus to 6 consecutive histidines. The epitope allowed identification of the desired peptides, and the 6-His tag facilitated their purification. Both peptides (A.A. 40 -89 and A.A. 331-391) were purified to near homogeneity and tested for their ability to inhibit DNA linking.
A.A. 40 -89 and A.A. 331-391 can each inhibit DNA linking. The efficiency with which N⌬360, a derivative of EBNA1 with only one linking domain, links DNAs containing two EBNA1binding sites was measured in the presence of a range of concentrations of A.A. 331-391 (Fig. 3A). The same range of concentrations of A.A. 331-391 was also tested in the absence of N⌬360. A.A. 331-391 can both inhibit N⌬360-mediated DNA linking and, because of its high positive charge, bind to DNA. However, its inhibition of linking occurred at a significantly lower concentration than did its DNA-binding. Similar experiments were conducted for A.A. 40 -89 and the results for A.A. 331-391 and A.A. 40 -89 are represented graphically in Fig. 3, B and C, respectively. The concentration of each peptide which reduced linking to 50% of that in the absence of competitor (K i ) was significantly less than the concentration which bound 50% of the DNA. These concentrations differ by 50-and 100-fold for A.A. 331-391 and A.A. 40 -89, respectively. These experiments demonstrate that A.A. 331-391 and A.A. 40 -89 can inhibit linking independently of their ability to bind to DNA. Even high concentrations of the peptides, which bind DNA detectably in gel shift assays, do not displace N⌬360 bound specifically to the probe DNA ( Fig. 3A and data not shown).
The inhibition of linking mediated by A.A. 331-391 and A.A. 40 -89 is specific. The peptides are quite basic (Fig. 1). A.A. 331-391 contains 21 basic residues and 6 acidic residues giving it a predicted net charge of ϩ15 at neutral pH. A.A. 40 -89 has 14 basic and 4 acidic residues for a predicted net charge of ϩ10 at neutral pH. To test whether nonspecific effects of these charges were solely responsible for the ability of A.A. 331-391 and A.A. 40 -89 to inhibit DNA linking, we sought a control peptide with a similar charge per molecule. Polylysine (Sigma) with a mass distribution of 1000 to 4000 daltons (Da) can be estimated to have an average mass of 2500 Da and an average charge per molecule of ϩ16 at neutral pH. The ability of polylysine, RNase A, and BSA to inhibit linking of DNAs containing two EBNA1-binding sites by N⌬360, a derivative of EBNA1 with one linking domain, was determined (Table I). The concentration of polylysine required to inhibit linking was less than that of BSA or RNase A. Polylysine, however, was less effective than A.A. 331-391 and A.A. 40 -89 at inhibiting linking of DNAs containing two binding sites by N⌬360. The K i of polylysine was 35-and 60-fold greater than that for A.A. 40 -89 and A.A. 331-391, respectively (Table I). These experiments demonstrate that the inhibition of DNA linking by A.A. 331-391 and A.A. 40 -89 is not mediated by their charge alone.
Linking of DNAs containing two EBNA1-binding sites by bEBNA1 is more resistant to competition than is linking by N⌬360. bEBNA1, which has all three linking domains, links DNAs containing two EBNA1-binding sites almost twice as efficiently as does N⌬360, which has only one linking domain (28). Linking of DNAs containing two EBNA1-binding sites by bEBNA1 was inhibited by A.A. 40 -89 and A. A. 331-391 (Table  I) bEBNA1 and the two binding site DNAs were also 6-fold more resistant to polylysine than similar complexes formed by N⌬360 (Table I). bEBNA1-mediated linking of DNAs with two binding sites was insensitive to RNase A and BSA in the range of concentrations tested ( Table I).
Effect of Nonspecific DNA on the Efficiency of DNA Linking by EBNA1-To determine whether the poly(dI)⅐poly(dC) used as a nonspecific competitor DNA contributes to linking in these assays, reactions were conducted in its absence. N⌬360 linked DNAs containing two EBNA1-binding sites in the absence of poly(dI)⅐poly(dC) (Fig. 4A). On average, the maximum percentage of this DNA linked by N⌬360 was 10 and 11% in the absence or presence of poly(dI)⅐poly(dC), respectively (Fig. 4, B  and C). The inhibition of linking by increasing concentrations of N⌬360 was also similar in the absence or presence of poly(dI)⅐poly(dC). At the highest concentrations of N⌬360 tested, we observed a complex migrating more slowly than that of the DNA with both specific EBNA1-binding sites occupied (designated C in Fig. 4A). When poly(dI)⅐poly(dC) was included in an otherwise identical reaction, this complex was not observed (data not shown). We interpret this complex to be DNA bound site specifically at both EBNA1-binding sites and nonsite specifically elsewhere. bEBNA1 also linked DNAs containing two EBNA1-binding sites similarly in the presence and absence of poly(dI)⅐poly(dC) (data not shown). These experiments demonstrate that the presence of nonspecific competitor DNA is not required for DNA linking by EBNA1.
Effect of Nucleic Acids Other Than EBNA1-binding Sites on the Efficiency of DNA Linking by EBNA1-Nucleic acids other than those bound site specifically by EBNA1 were introduced into the linking reactions from several sources. Poly(dI)⅐poly(dC) added to standard reactions was not required for linking (Fig. 4). DNA flanking the EBNA1-binding sites and any DNA or RNA contaminating the protein preparations were also added to linking reactions. The ability of N⌬360 which had been treated with RNase A and/or DNase I to link DNA in the absence of the other nonspecific nucleic acids was measured. First, poly(dI)⅐poly(dC) was excluded from the reactions. Second, a 41-bp DNA containing two EBNA1-binding sites, but lacking sequences flanking those sites (41BP) was used as the probe. This DNA is expected from x-ray crystallographic studies to be coated by EBNA1 with little or no uncovered flanking DNA (12). Third, the N⌬360 was treated with DNase I or RNase A or both prior to conducting the linking assay (treatments described under "Experimental Procedures"). N⌬360 tested under these conditions linked DNA (data not shown). The percentage of DNA linked increased then decreased with increasing protein concentration. Nuclease-treated and mocktreated N⌬360 did not differ in their activity. DNA linking mediated by N⌬360 is not substantially affected by treatment to exclude nucleic acids, other than those bound by EBNA1, from our assays.
The ability of N⌬360 to link 41BP in the absence and presence of poly(dI)⅐poly(dC) was measured (Fig. 5). 41BP behaved differently than the 131-bp DNA containing two EBNA1-binding sites and flanking sequences used formerly. In the absence of poly(dI)⅐poly(dC) the average maximum percentage of 41BP linked was 32%, approximately three times more than that of the 131-bp DNA with flanking sequences. In the absence of poly(dI)⅐poly(dC), linking of 41BP and of the 131-bp DNA was inhibited as the concentration of N⌬360 increased (Figs. 5 and  4). In the presence of poly(dI)⅐poly(dC) (130 ng/l; 3 g/reaction), 41BP was linked extremely efficiently, increasing to 80% at the highest concentration of N⌬360 tested (Fig. 5). Linking of 41BP in the presence of poly(dI)⅐poly(dC) differed from that of a similar DNA with nonspecific flanking sequences because it was not competed by excess N⌬360. The enhanced linking of 41BP by high concentrations of N⌬360 in the presence of FIG. 4. DNA linking by N⌬360 does not require poly(dI)⅐poly(dC). The efficiency with which N⌬360 links DNAs containing two binding sites in the presence or absence of poly(dI)⅐poly(dC) is similar. A, 20 fmol of a DNA with two EBNA1-binding sites were combined with a range of concentrations of N⌬360 prior to separation by electrophoresis through a 4% polyacrylamide gel. The positions of linked DNA, unbound DNA, DNA with one and two sites occupied (1 and 2, respectively), and a slower mobility complex (C) are indicated. PhosphorImage analysis was used to calculate the percentage of linked DNA of the total DNA in each lane, and this percentage is displayed at the bottom. B and C, DNA linking by N⌬360 in the absence (B) and presence (C) of 3 g of poly(dI)⅐poly(dC) per reaction was measured. Each graph plots the percentage of DNA with two EBNA1-binding sites linked against the concentration of dimers of N⌬360. Standard deviations are from four or five experiments in B and C, respectively. The data in C is a composite of that published previously (28) and additional new measurements. and bEBNA1 The ability of 6.5 nM dimers of either N⌬360 or bEBNA1 to link 20 fmol of a DNA with two EBNA1-binding sites was measured in the absence of competitor as well as in the presence of a range of concentrations of the indicated competitors. The concentration of competitor which inhibits linking to 50% of that in its absence (K i ) was determined. The highest concentrations of RNase A and BSA tested were 42,000 and 8,800 nM, respectively. poly(dI)⅐poly(dC), which averages ϳ500 bp in length, was only observed if at least 0.3 g of poly(dI)⅐poly(dC) per reaction was included (data not shown). Similar quantities of nonspecific DNAs other than poly(dI)⅐poly(dC) could also mediate this effect (data not shown). The ability of nonspecific DNAs to enhance linking of 41BP was mediated by DNAs of 30, 63, 108, and 888 bp in length, but not by DNAs of 8, 10, and 12 bp in length (data not shown). Linking of DNAs bound by N⌬360 has been found to be increased markedly by nonspecific DNAs only when the DNAs bound by EBNA1 lack flanking sequences and if the nonspecific DNAs are sufficiently long. DISCUSSION We have used gel shift assays to study the mechanism of DNA linking by EBNA1 and its derivative N⌬360. We had shown previously that EBNA1 in excess of that required to bind all the EBNA1-binding sites in a reaction inhibited linking in that reaction (28). Competition by excess EBNA1 could be detected in the presence of a 40-fold excess of competitor DNA fragments. The competing protein was in excess relative to the DNA-bound protein, but not relative to the competing DNA. This finding indicated that protein-protein interactions could mediate DNA linking. The competition was hypothesized to be mediated by the linking domains of proteins not bound to DNA interacting with the linking domains of proteins bound to DNA, thereby disrupting links between DNA-bound proteins. One prediction of this hypothesis is that only proteins with linking domains would compete with DNA linking. An extension of this proposal is that linking domains alone would compete with DNA linking.
These predictions have been tested and confirmed. Only proteins which contained linking domains could inhibit DNA linking when provided in excess (Fig. 2). DNA linking by GAL4(1-94)&54 -89 could be inhibited by bEBNA1 and N⌬360, but not by N⌬389. DNA linking by N⌬360 could be inhibited by GAL4(1-94)&54 -89, GAL4(1-94)&331-361, and GAL4(1-94)&372-391, but not by GAL4 . Because proteins with DNA-linking domains inhibit linking and derivatives without linking domains do not inhibit linking, it is likely that the linking domains themselves are mediating this inhibition. Proteins competed similarly whether they contained the same or different linking domain(s) than that of the protein with which they were competing. bEBNA1, which contains amino acids 54 -89, and N⌬360, which lacks amino acids 54 -89, could compete similarly for linking by GAL4(1-94)&54 -89 ( Fig. 2A). GAL4 fused to linking domains of EBNA1 competed equally well with DNA linking by N⌬360 whether the fusion contained the same or a different linking domain than found in N⌬360 (Fig. 2B). These observations indicate that EBNA1's linking domains can interact heterotypically with one another.
Two peptides (A.A. 40 -89 and A.A. 331-391, Fig. 1) which together contain all three of the identified linking domains inhibit DNA linking by EBNA1. Each of these peptides can also bind nonspecifically to DNA. The percentage of a DNA with two EBNA1-binding sites linked by N⌬360 could be halved by a 5-fold excess of A.A. 331-391 relative to N⌬360 (Fig. 3, A and  B). At this concentration of A.A. 331-391 (30 nM), the poly(dI)⅐ poly(dC) used as a competitor DNA was present in greater than 10-fold molar excess. Similarly, A.A. 40 -89 could mediate a 50% inhibition of linking at 55 nM. The peptides can efficiently inhibit linking at concentrations within 10-fold of the concentration of N⌬360 -peptide concentrations far less than that of the competitor DNA. These results indicate that inhibition of linking by the peptides is mediated by interactions with N⌬360 rather than nucleic acids.
We compared the ability of control proteins (BSA, RNase A, and polylysine) to inhibit DNA linking by N⌬360 (Table I). The ability of N⌬360 to link DNAs containing two EBNA1-binding sites was reduced only to 80% by 8.8 M BSA. Linking could be inhibited to 50% by 30 M RNase A. Approximately 1000-fold more RNase A than A.A. 331-391 was required to reduce linking by 50%. A.A. 40 -89 and A.A. 331-391 each are basic peptides with predicted net charges of ϩ10 and ϩ15 at neutral pH, respectively. The ability of polylysine with a similar charge per peptide to compete with linking mediated by N⌬360 was significantly greater than BSA or RNase A, but significantly less than A.A. 40 -89 or A.A. 331-391 (Table I). Linking could be inhibited to 50% by 1.9 M polylysine, a 35-or 60-fold higher concentration than that required for similar inhibition by A.A. 40 -89 or A.A. 331-391, respectively. This result demonstrates that the inhibition of N⌬360-mediated DNA linking by A.A. 40 -89 and A.A. 331-391 may be mediated in part by nonspecific affects of their charge, but specific affects of their sequence or structure also contribute to their inhibitory activity.
The efficiency with which DNAs are linked by EBNA1 likely reflects the valency of the complexes formed by the DNA and EBNA1. N⌬360, which has one of three linking domains of EBNA1, links greater than 95% of DNAs with 10 EBNA1binding sites (28), but only 12% of DNAs with two EBNA1binding sites (Fig. 3). A.A. 331-391 competes for linking by N⌬360 of the 10 binding site DNAs approximately 0.3% as effectively as for linking of the two binding site DNAs (data not shown). bEBNA1, which has all three linking domains, links DNAs containing two EBNA1-binding sites approximately twice as efficiently as N⌬360. A.A. 40 -89 and A.A. 331-391 compete approximately 1% as effectively with linking of a DNA with two binding sites by bEBNA1 as they do for linking of it by N⌬360 (Table I). The greater valency of contacts between linking domains within efficiently linked complexes presumably underlies the greater resistance of these complexes to being dissolved by peptides containing linking domains. The estimated concentration of EBNA1 in a nucleus with a 5-m diameter is 1 mM (35). Assuming looping of oriP is important for EBNA1-dependent activities of oriP, the large number of EBNA1-binding sites in FR may be necessary to stabilize looping to DS in the presence of such high concentrations of EBNA1. The ability of N⌬360 to link 41BP in the absence and presence of poly(dI)⅐poly(dC) is graphically represented. In the absence of poly(dI)⅐poly(dC), 41BP is linked by N⌬360 and linking becomes inhibited as the concentration of N⌬360 is increased. In the presence of poly(dI)⅐poly(dC) the efficiency of linking of 41BP increases over the range of N⌬360 concentrations tested. 20 fmol of 41BP were combined with a range of concentrations of N⌬360 in the presence or absence of 3 g of poly(dI)⅐poly(dC). The graph plots the percentage of 41BP linked against the concentration of dimers of N⌬360. Standard deviations are from two separate experiments.
The high positive charge of the linking domains permits them to associate with nucleic acids at neutral pH. That such binding is unessential for linking was demonstrated by treatment of protein with nucleases prior to testing that protein in linking assays from which all nucleic acids, other than those bound site specifically by EBNA1, had been excluded. Experiments with a DNA that contains two binding sites for EBNA1 but lacks flanking sequences, 41BP, demonstrated that sufficiently long, nonspecific DNAs can contribute to linking (Fig.  5). We hypothesize that this contribution can be masked by nonspecific DNA flanking the binding sites for EBNA1, because the high local concentration of this flanking DNA favors its association with the linking domains of bound EBNA1.
The apparent affinity of protein-protein interactions between linking domains is greater than the apparent affinity of the linking domains for nonspecific DNA. A large difference exists between the concentration of nonspecific DNA required to enhance linking of 41BP (40 -800 nM) and the concentration of N⌬360 required to link 41BP in the absence of nonspecific DNA (5 nM). Interaction of linking domains with DNA may occur in vivo. However, because the apparent affinity of the linking domains for other linking domains is significantly higher than for DNA (or for RNA (29)), we predict that proteinprotein interactions predominate in vivo.
Observations described in the Introduction support the assertion that DNA linking contributes to the activation of transcription and replication by EBNA1. That EBNA1 is likely to link DNA in vivo at oriP also supports this contention. Evidence for linking in vivo is indirect but strong. The length of oriP (1.8 kbp or 0.6 m for B-form DNA) dictates that both FR and DS, and the EBNA1 bound to them, are confined to a maximum volume of approximately 0.1 femtoliters. Therefore, the concentration of EBNA1-binding sites at oriP is minimally 300 nM. EBNA1 occupies all of its binding sites at oriP for at least the majority of the cell cycle (36). In vitro, DNAs bound by EBNA1 are linked at concentrations far less than 300 nM. In Fig. 4, lane 4, 7 fmol of DNA is bound by EBNA1; therefore, the maximum concentration of occupied EBNA1-binding sites is 0.6 nM, and linking is readily detected. Because the concentration of binding sites at oriP is approximately 500-fold higher, linking likely occurs in vivo.
Interactions between linking domains are mediated by specific sequences or structures of the linking domains. It is a distinct possibility that specific interactions between the linking domains of EBNA1 and other proteins also occur. Twenty amino acids of EBNA1 are sufficient to inhibit specifically linking of DNA by EBNA1 (GAL4(1-94)&372-391, Fig. 2).
Smaller molecules may also be effective inhibitors of linking domain interactions. EBV is associated with many diseases including several malignancies. In all of these diseases EBNA1 is expressed in infected cells and sometimes only EBNA1 is expressed. Therefore, small molecules which inhibit EBNA1's functions, such as DNA linking, could be clinically useful.