Competition for DNA Binding Sites between the Short and Long Forms of E2 Dimers Underlies Repression in Bovine Papillomavirus Type 1 DNA Replication Control

Plasmids.

pKSO has been previously described (43). The FLAG-E2 expression vector (pAcC13-FE2) was cloned from a previously described SP6-E2 expression vector (18) as follows. SP6-E2 was cut with BglII and SphI, and a double-stranded oligonucleotide (top strand, GATCTACCATGGACTACAAGGACGACGATGACAAGGAGACAGCATG; bottom strand, CTGTCTCCTTGTCATCGTCGTCCTTGTAGTCCATGGTA) coding for the FLAG epitope (Kodak, IBI) and the N-terminal four amino acids of E2 (minus the start methionine) was ligated into those sites (creating SP6-FE2). This FE2 fusion was removed from SP6 with BamHI and BglII and ligated to the pAcC13 vector at the BamHI and BglII sites. The GE2C expression vector was cloned into the pAcC13 vector by standard PCR methods; the Glu epitope has been previously described (12). The 339M mutation (35) was cloned from the pCG-E2-339M vector into the GE2C expression vector by using the KpnI and filled-in SacI sites. The WK33 mutation (10, 18) was cloned from the YEpE2B-WK33 expression vector (gift of E. Androphy) into the SP6-FE2 plasmid by using the SphI and KpnI sites; the FE2-WK33 reading frame was transferred to the pAcC13 vector as described above. After generation of the respective transfer vectors the infectious recombinant baculoviruses were selected by cotransfection and plaque selection in Sf9 cells (Baculogold; Pharmingen). pCLO and pCLOT were constructed by inserting the HindIII to BamHI fragment from pKSO or pKSOT, respectively, into these sites in the pACYC177 vector, which lacks E2 binding sites.

Cell culture.

To facilitate the transient-replication assays described below, we generated a stable cell line expressing BPV-1 E1 and E2 genes placed under control of tetracycline-regulated promoters. The principle of this gene expression system as well as the plasmids utilized was described previously (11). CHO cells were transfected with linearized plasmid pUHD15-1neo containing the tTA (tetracycline-controlled transactivator) gene under control of a cytomegalovirus (CMV) promoter and a neomycin selection marker. Stable transformants were selected with G418 (500 μg/ml). Individual clones were isolated and transiently supertransfected in the presence and absence of tetracycline with pUHC13-3, containing a tTA-responsive luciferase reporter gene. The CHO tTA⁺ clone showing the highest dynamic range of luciferase regulation was used for integration of the E1 and E2 genes. Both genes were cloned into the multiple cloning site of the tTA-responsive expression vector pUHD10-3. The resulting expression vectors, pUHD10-3/E1 and pUHD10-3/E2, were cotransfected with a hygromycin selection marker into CHO tTA⁺ cells with selection at a hygromycin concentration of 300 μg/ml. The functionality of the stable clones raised was analyzed by their capacity to promote transient replication of pKSO (see above) in a tetracycline-dependent fashion. The cell line chosen for all subsequent experiments was named CHO tTA-E1/E2 #6. The general tissue culture techniques employed, the transient transfections, and the analysis of replicated DNA have been previously described (23). Sf9 cells were maintained in suspension in Grace’s medium with 10% fetal calf serum, 0.1% pluronic acid, 0.33% yeastolate, and 0.33% lactalbumin. Sf9 cells were infected at a density of 5 × 10⁵ cells/ml; the multiplicity of infection was between 3 and 5 for both single infections and coinfections. Infections were allowed to progress for 48 h before harvesting. Cells were pelleted and washed once with phosphate-buffered saline before lysis.

Protein expression and purification.

The expression and purification of GE1 have been previously described (43). GE2C purification was similar to that of GE1, except that the Sf9 cells were lysed whole (omitting the nuclear isolation-extraction step). FE2 purification was essentially the same as that of GE2C, except that α-FLAG M2 monoclonal antibody-linked resin (Kodak, IBI) was used instead of α-Glu antibody. Batch elution from the α-FLAG column was accomplished with 200 μg of FLAG peptide/ml for 1 h at 4°C. Before elution, all columns were washed with at least 75 column volumes of buffer C (25 mM Tris [pH 8], 1 M LiCl, 1 mM EDTA, 10% glycerol) and at least 30 column volumes of buffer D (25 mM Tris [pH 8], 200 mM NaCl, 1 mM EDTA, 10% glycerol). For the purification of FE2-GE2C heterodimers, extracts from coinfected cells were first passed over the α-FLAG resin column; the column was washed and batch eluted as described above. The eluate was then incubated with α-Glu resin at 4°C with rotation. After the described washes, 20 mM triethylamine elution fractions were collected into tubes containing neutralizing volumes of 1 M HEPES, pH 7.5. Proteins were dialyzed against replication buffer (20 mM KPO₄ [pH 7.5], 1 mM EDTA, 1 mM dithiothreitol, 150 mM potassium glutamate, 10% glycerol), frozen in liquid N₂, and stored at −70°C.

DNase I protection assay.

Assays were performed as previously described with no carrier DNA (43). Protein-DNA mixtures were incubated at 37°C for 15 min; reactions were performed in a total volume of 100 μl at room temperature. The BamHI to EcoRI DNA fragment used for protection studies was derived from pKSO. Top and bottom strands were end labeled separately with ³²P in standard polynucleotide kinase reactions. The order of addition of the three proteins did not affect the appearance of the footprints.

In vitro DNA replication.

Assays were performed essentially as previously described (43). A typical reaction contained 50 ng of pKSO, pCLO, or pCLOT template and 10 μl of FM3A extract (5 to 10 mg/ml). The FM3A extract was prepared as previously described (43). Reactions included an ATP regenerating system, deoxynucleoside triphosphates, and [α-³²P]dCTP in a total volume of 25 μl. The amounts of GE1, FE2, and GE2C are indicated in the figure legends; the order of addition of the DNA, GE1, FE2, and GE2C did not affect the outcomes of the experiments. The GE1 protein was diluted to working concentrations with FM3A extract. Data shown are representative of reactions that were repeated several times, with different GE1, FE2, and GE2C preparations.

Heterodimers of E2-E2C are activators of in vitro DNA replication.

One reasonable mechanism by which the E2C protein might repress BPV-1 DNA replication is through formation of inactive heterodimers with the full-length E2. While dimers of E2 are stable against dissociation even at picomolar concentrations, it has been previously shown that heterodimers of short and long fragments of E2 can be created by cotranslation of the respective mRNA in vitro. Moreover, while mixing of the short and long forms does not create heterodimers, denaturation in 5 M urea followed by subsequent dialysis does result in heterodimeric species (30).

To explore this possibility we sought to purify E2-E2C heterodimers and to assay their potential for replication activation side by side with full-length E2 or E2C homodimers. To achieve this goal two different sorts of recombinant baculoviruses were created. The first virus (FE2) expresses a full-length E2 protein fused to the FLAG epitope (14), and the other (GE2C) encodes an E2C protein fused to the Glu epitope (12). Figure 2, lanes 1 to 3, shows the proteins in the crude extracts of infected Sf9 cells as detected in a Western blot with E2 monoclonal antibody. The cells were infected with either GE2C or FE2 separately or were coinfected with both viruses. The extracts from the coinfected cells were initially passed over an α-FLAG monoclonal antibody column, washed extensively, and batch eluted with FLAG peptide. The eluate containing the peptide, FE2 homodimers, and FE2-GE2C heterodimers was then incubated with the α-Glu resin; the resin was transferred to a column and washed extensively, and the remaining affinity-bound FE2-GE2C heterodimers were eluted and collected. Western blot and silver-stained analyses of this heterodimer preparation are shown in Figure 2, lanes 5 and 8, respectively. As a control, extracts from Sf9 cells infected singly with either GE2C or FE2 were mixed in equal proportion and then passed through the α-G agarose resin. After washes identical to those for the coinfected material, the eluted proteins were visualized by Western blotting (Fig. 2, lane 4) and by silver staining the material on sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) gels (Fig. 2, lane 7). In these circumstances the FLAG-tagged E2 could not be detected.

FE2-GE2C heterodimers and FE2 homodimers were tested in parallel in the BPV-1 in vitro DNA replication assay. As shown in Fig. 3, the heterodimers were as potent as the homodimers of E2 in activating E1-dependent DNA replication. Two independent heterodimer preparations were compared to the homodimers, and synthesis levels were quantitated (Fig. 3B). The heterodimer preparations showed equivalent activation compared to untagged E2, and homodimers of GE2C showed no activation of in vitro DNA replication (see below and data not shown). Furthermore, as the titrations of the heterodimers and the homodimers were essentially identical (Fig. 3 and data not shown), these experiments are not consistent with the counterargument that trace quantities of homodimers serve as activators. As a control, heterodimers of GE2C with a FLAG-tagged full-length E2 mutant, WK33, were prepared. The WK33 mutation of E2 (in the amino-terminal activation domain which is missing in E2C) has been shown to block replication function in cells (10), and homodimers of this protein are inactive in vitro (28). The heterodimer FE2/WK33-GE2C was inactive in the assays, establishing that the single intact activation domain in the “wild-type” heterodimer is critical for activation.

To probe these conclusions further we asked if heterodimers of FE2-GE2C could cooperatively bind with the E1 protein at the origin site as the homodimers of full-length E2 do. Serial dilutions of E1 protein were included in binding buffer with origin DNA fragments either alone or in combination with homodimer or heterodimer E2 protein. The DNA-protein complexes were then analyzed by DNase I footprint analysis. Figure 4 shows that the heterodimers bound slightly more weakly than did the E2 homodimers by themselves but that E1 stimulates binding to a similar extent. There is also a slight difference in the size of the footprint at BS11 (compare lanes 5 and 9). We can only speculate about the reasons why the heterodimers and homodimers display these small differences. One possibility is that the activation domains of the homodimers stabilize each other, resulting in less interference with binding, and that E1 can also stabilize the single E2 activation domain of the heterodimer. In any case, as shown in Fig. 4, the E2-E2C heterodimer and the E2 homodimer show equivalent cooperativity with E1 in DNA binding. With different preparations of E2 proteins we obtained qualitatively similar results via gel shift assays.

Homodimers of E2C inhibit E1-E2-dependent replication in vitro.

The implication of the above-described result is that the active form of the repressor is the homodimer of E2C. Liu et al. (29) have previously shown that in vitro DNA replication of HPV-11 origin templates can be inhibited slightly (less than twofold) by HPV-11 E2C protein preparations. To extend and confirm these conclusions we asked if E2-dependent in vitro DNA replication could be repressed by homodimers of the E2C protein and compared this data to a purified homodimer E2C repressor that harbors a mutation (339M) in the alpha helix of the DNA binding domain of E2 (1, 13). This latter protein was created by transferring the 339M mutation previously characterized in the full-length E2 protein (28) to the GE2C gene in a recombinant baculovirus (GE2C-339M). This mutation inactivates DNA binding but leaves dimerization unaltered. As expected neither the GE2C or GE2C-339 protein affected E1-mediated replication at any concentration of E1 or repressor protein tested when only these proteins were present in the reaction (data not shown).

The data presented in Fig. 5 clearly show that the E2C homodimer can repress DNA replication when synthesis is dependent upon the activator form of E2. Moreover, this repression requires the DNA binding domain of the repressor. Interestingly, to achieve a 50% inhibition, a 10-fold molar excess of E2C to E2 is required. This perhaps reflects the differences in cooperativity of binding with the E1 protein between the repressor form and the full-length form of E2 (see below for further discussion). To essentially shut off DNA synthesis in this reaction a 100-fold excess of the short forms is required (Fig. 5B). E2-activated BPV-1 DNA replication in vitro does not require PolII transcription, but E2C has been shown to repress gene expression in vivo. Therefore it was necessary to show that RNA polymerase transcription did not play a role in the repression detected. DNA replication reactions were thus assembled in the absence (Fig. 5C, lanes 1 to 4) and presence (Fig. 5C, lanes 5 to 8) of the transcription inhibitor, α-amanitin. Inhibition of DNA replication by GE2C was found to be independent of the effect of the drug.

An important observation made was that inhibition mediated by E2C was critically dependent not only on E2 levels but on E1 concentration. At threefold greater GE1 concentrations than that used in the reactions for Fig. 5, at which point replication becomes E2 independent, the net DNA synthesis was resistant to E2C repression in the range of titration shown for the repressor (data not shown). At threefold lower E1 concentrations no net synthesis was detected even in the absence of E2C. Thus, only at concentrations of E1 at which the replication was dependent upon the presence of E2 protein was inhibition by GE2C observed. These data are thus consistent with the model that E1 and E2 cooperative binding is crucial for efficient DNA replication and that E2C competes with this preinitiation complex by occupying cognate E2 sites. To test this point more directly we asked if E2C could interfere with E1 and E2 interaction on the DNA and if this competition is sensitive to E1 levels.

The experiments shown in Fig. 6 were designed to explore the proposal that E2C could block E1 binding if the latter protein’s DNA occupancy is dependent upon E2. DNase I protection experiments with origin templates were performed by assaying binding as shown by the protection patterns on the bottom strand in Fig. 6A and on the top strand in Fig. 6B. Serial dilution of E1 alone (lanes 4 to 6) compared to a parallel dilution in the presence of E2 (lanes 8 to 10) shows the end point of binding is enhanced by the addition of E2 protein. However, in the presence of E2C the same concentrations of E2 and E1 showed a reduced E1 binding as measured by less E1 protection at the origin palindrome (lanes 13 to 15). Specifically, in lanes 9 and 10, E1 protection of its recognition site was dependent upon E2 proteins; at these threshold concentrations of E1, addition of GE2C at an approximately 10-fold excess to E2 prevented E1 binding (lanes 14 and 15 show no stimulation over lanes 5 and 6, which contain E1 alone). To show this we needed to have GE2C included at a 10-fold excess to E2 to offset the DNA binding cooperativity of E1 and E2. However, at the highest E1 concentration, protection of the E1 binding site was not appreciably affected by GE2C, even at this 10-fold excess of repressor (compare lanes 4, 8, and 13). As a control we showed that the mutant form of the repressor GE2C-339M does not affect the formation of the E1-E2-DNA ternary complex, and at all E1 concentrations tested this protein does not interfere with E1 occupancy at the origin site. In this experiment mixing the mutant E2C with the wild-type E2 has a very small inhibitory effect upon E2 site 12 occupancy (compare lanes 7 and 17) perhaps due to nonspecific protein interactions, but in the presence of E1 this effect is not detectable. These data in sum thus establish that E2C can prevent the binding of E1 to the origin site when the initiator is absolutely dependent upon the E2 enhancer protein for loading.

E2 activation of BPV-1 DNA replication in vitro requires E2 DNA binding.

Previous data from our laboratory had shown that the in vitro DNA replication of BPV-1 could be activated by E2 in a manner that did not require flanking viral E2 binding sites in the templates (24, 34). It was found that activation showed little or no dependence upon mutations in these sites. One possible explanation for these results (28) was that the protein-protein interactions between E1 and E2 are dominant contributors to the ΔG driving the ternary complex. Thus E2-DNA interactions would not contribute to, or perturb when mutated, the equilibrium of the ternary complex on naked DNA substrates. While this thought may have some application, it would seem that to achieve repression as described above with the E2C protein, competition for some defined E2 binding sites would be required. This seems particularly true as an intact DNA binding domain of the repressor is required for achieving effective negative regulation. We therefore asked if a consensus E2 binding site present in the vector based on the ColE1 origin of replication might be responsible for this paradoxical set of data. We initially tried to mutate the E2 site in the vector (overlapped by the HgiEII site at nucleotide 1739 of pBluescript), but these attempts were unsuccessful, apparently because the sequences in this region are required for propagation in Escherichia coli (34a). To circumvent this problem we then searched for an E. coli vector that did not contain a consensus E2 binding site, and we found by computer scanning that prokaryote plasmid vectors based upon the p15A origin were suitable for these purposes. We therefore transferred the BPV-1 origin fragments from the templates pKSO and pKSOT to the vector pACYC177 (4), which replicates via this origin. pKSO contains BPV-1 sequence that spans both E2 binding sites 11 and 12, while the 50-bp viral sequences in pKSOT (for origin tiny) do not contain these sites. When these viral sequences were placed in the pACYC vector (resulting in the plasmids pCLO and pCLOT, respectively) the E2 activation effects for E2-positive or E2-negative binding site templates were markedly different. In fact these experiments do not prove that it was indeed the single E2 site at nucleotide 1739 in the pBR-based vectors that explicitly underlies the differences. However, they do show that the choice of recombinant vector backbone does influence the results and that with the appropriate vector E2 activation is indeed dependent upon cis-acting viral E2 binding sites.

Figure 7 shows that E2 activates the E1-limited in vitro DNA replication of pCLO 10- to 20-fold; in contrast pCLOT shows only marginal activation (less than 2- to 3-fold over the same range of E2 titration). Together with the other results in this report we must conclude that E2 DNA binding to its recognition motifs is important for activation in the in vitro system. In vitro these sites may either be provided by viral-encoded cognate sites (pKSO or pCLO) or vector backbone sites (pKSOT). Consensus sites as in pBR322 or nonconsensus and putatively weak E2 binding sites (those that have, for example, half-sites) contribute as might be anticipated by their affinities. Results shown in Fig. 7 indicate that an excess of E2C can down-regulate replication on the pCLO template (compare lanes 14 and 15 to 17 and 18). Quantitation of these replication products by PhosphorImager analysis shows that deletion of the E2 DNA site lowers E2 activation by 5- to 10-fold and subsequent E2C repression (lanes 5, 6, 8, and 9) is also lower for pCLOT than for pCLO. We should emphasize that where efficient replication is clearly dependent on an E2 site, E2C represses, but that the low level of replication activation detected in the vector without such an E2 site precludes a reliable quantitative measure of repression. We can only infer that nonconsensus E2 sites lead to the marginal activation and repression observed for the pCLOT templates.