Interaction of Epstein-Barr Virus BZLF1 C-Terminal Tail Structure and Core Zipper Is Required for DNA Replication but Not for Promoter Transactivation

Abstract

The Epstein-Barr virus (EBV) protein BZLF1 contains a bZIP DNA binding domain in which C-terminal tail residues fold back against a zipper region that forms a coiled coil and mediates dimerization. Point mutagenesis in the zipper region reveals the importance of individual residues within the ²⁰⁸SSENDRLR²¹⁵ sequence that is conserved in C/EBP for transactivation and EBV DNA replication. The restoration of BZLF1 DNA replication activity by the complementation of two deleterious mutations (S208E and D236K) indicates that the interaction of the C-terminal tail and the core zipper is required for DNA replication, identifying a functional role for this structural feature unique to BZLF1.

The induction of BZLF1 (also known as ZEBRA, Zta, and EB1) is the first step in the reactivation of the lytic cycle of Epstein-Barr virus (EBV) (1, 2). BZLF1 acts first as a transcription factor, inducing genes for other early lytic cycle proteins. As the lytic cycle progresses, BZLF1 also acts as a DNA replication factor and provides a lytic origin binding protein function for EBV replication (6).

BZLF1 is a bZIP transcription factor (3) with sequence similarity to c-fos and C/EBP in the dimerization and DNA binding domains. The results of domain-swapping experiments with the known bZIP protein GCN4 showed that the dimerization domain of BZLF1 homodimerizes in a way that is functionally equivalent to a leucine zipper but that BZLF1 is not able to heterodimerize with other bZIP proteins (7). The coiled-coil dimerization region was predicted to be shorter than those in other bZIP proteins because of a proline residue at position 223, expected to interrupt the helical coil structure (3, 7). The X-ray crystal structure of the BZLF1 dimerization and DNA binding domains bound to DNA (8) confirmed this prediction and additionally revealed that BZLF1 residues in the C-terminal direction from Pro223 fold back against the coiled coil, with stabilizing contacts between tail residues 228 to 236 and residues halfway up the coiled coil (Fig. 1A and B). Chemical cross-linking in the C-terminal region of a purified fragment of BZLF1 expressed in Escherichia coli and biophysical analyses of synthetic peptides were consistent with the X-ray structure (9). Results from transcription reporter assays indicated that the coiled-coil region is absolutely required for transactivation (presumably because of the requirement for dimerization) but that the C-terminal tail is not essential for this process (9).

FIG. 1.

Dimerization domains of BZLF1 and C/EBP. (A) Structural alignment of BZLF1 (yellow and green monomers; PDB accession no. 2C9L) and C/EBPβ (cyan and blue monomers; PDB accession no. 1GU5) (11, 12) on DNA (gray). Straight-line distances between the end of the C/EBPβ coiled coil and the conserved SSENDRLR residues are indicated. BZLF1 residues in the conserved motif and C/EBPβ C-terminal residues are labeled. The crystal structure of the C/EBPα bZIP domain (PDB accession no. 1NWQ) is nearly identical to that of C/EBPβ (root mean square deviation for 112Cα, 1.16 Å) but less complete. (B) Sequence alignment of the dimerization domains of BZLF1, C/EBPα, and C/EBPβ. Residues conserved between BZLF1 and C/EBPα are highlighted in gray; C/EBPβ residues identical to those in C/EBPα are shown by dots. C-terminal residues underlined and shown in italics are missing from the crystal structures. The conserved sequence motif is boxed; BZLF1 tail residues with which it interacts are overlined in green. (C) Straight-line (Cα-Cα) distances between the last ordered C/EBPβ residue and residues within the highly conserved motif. Corresponding (Corresp.) residues in C/EBPα are listed. Distances for the two monomers differ because the C/EBPβ homodimer structure is slightly asymmetric and because the last ordered residue in monomer 1 is Leu334 but that in monomer 2 is Lys332. The number of C-terminal residues required to span the distance as an extended polypeptide (assuming 1 residue/3.32 Å plus 2 residues to change direction) and the number extending beyond the coiled coil are listed as residues needed and residues available, respectively. No. res., number of residues. (D) Interaction between the C-terminal tail and coiled coil in the structure of the BZLF1-DNA complex (8). BZLF1 monomers are in green and yellow, and double-stranded DNA is in gray. The boxed area shows the region spanned by the conserved SSENDRLR motif and the hydrogen bond (dashed line) between S208 and D236. The bracketed area marked by an asterisk is shown in panel E. (E) Cross-sectional view of the S208-D236 interaction. The crystal structure of BZLF1 (top panel) and structural models of the S208E (middle panel) and S208E/D236K (bottom panel) mutants are shown. The predicted electrostatic repulsion is indicated by a double-headed arrow. WT, wild type.

We previously pointed out (7) an 8-residue motif in BZLF1 (²⁰⁸SSENDRLR²¹⁵) with striking sequence similarity to C/EBP (Fig. 1B). This motif is located midway along the coiled-coil dimerization region (8), precisely where the coiled coil contacts the C-terminal tail (Fig. 1A). In this study, we investigated the role of this conserved motif in BZLF1 function. The results suggest that S208 can play a key role in the BZLF1 structure required for EBV DNA replication. The reversal of the deleterious effects of the mutation of S208 by a compensatory mutation in the C-terminal tail provides experimental support for the X-ray structure and indicates that the binding of the tail structure to the coiled-coil region is required for DNA replication but not for transcription activation.

Conservation of the SSENDRLR motif is not due to conserved zipper-tail contacts.

In BZLF1, the SSENDRLR motif in the zipper region interacts with the C-terminal tail, raising the possibility that the sequence conservation in C/EBP reflects the conservation of zipper-tail contacts. To evaluate this possibility, we structurally aligned the dimerization domain of BZLF1 with the bZIP domain of C/EBPβ (Protein Data Bank [PDB] accession no. 1GU5; 82% sequence identity to the C/EBPα bZIP domain) (11, 12), which among known C/EBP structures lacks the fewest C-terminal residues of the protein (Fig. 1A and B). In the alignment, the coiled coil of C/EBPβ extends three helical turns beyond that of BZLF1. The distance between the last coiled-coil residue in C/EBPβ and residues within the conserved motif varies from 28 to 41 Å (Fig. 1C). This arrangement essentially rules out an interaction between C-terminal C/EBP residues and the conserved motif because a minimum of 11 to 15 residues is required for the polypeptide chain to fold back and contact the motif (assuming 2 residues to reverse direction and a fully extended conformation to span the remaining distance, an improbable scenario) whereas only 13 to 15 residues are present in the sequence on the C-terminal side of the coiled coil (Fig. 1C and legend). At best, only the last 2 to 4 residues in C/EBP may contact the conserved motif; however, these residues show no resemblance to BZLF1 tail residues (residues 228 to 236) that contact the motif, and so the stereochemical nature of the interactions would necessarily differ. We therefore conclude that the sequence conservation between BZLF1 and C/EBP is not due to the conservation of zipper-tail contacts; it may be related to intramolecular interactions required to form a stable zipper structure or reflect the conservation of a distinct, as yet unidentified function.

Conserved motif residues required for BZLF1 functions.

To further understand the role of the SSENDRLR motif, BZLF1 residues 208 to 214 were individually mutated to alanine. The serine residues at 208 and 209 were also mutated to glutamate. The resulting proteins, together with three previously described mutant proteins (K207A, K207A/R213A, and K207A/L214A) (7), were tested for their abilities to specifically bind to a known ZRE site in an electrophoretic mobility shift assay (EMSA) (Fig. 2A to C), to transactivate the promoter for BMRF1 (Fig. 2D), and to induce EBV lytic DNA replication (Fig. 2E).

FIG. 2.

Mutation of BZLF1 residues K207 to L214. (A) BZLF1, mutated BZLF1, or an empty vector control was in vitro translated in rabbit reticulocyte lysate with [³⁵S]methionine as a radioactive label. Products were fractionated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis and visualized by autoradiography. (B) Equal amounts of in vitro-translated products were tested for their abilities to bind to DNA by EMSA analysis with a ZRE probe (double stranded; cttcaTGAGTCAgtgcttc, where uppercase letters indicate the binding site). BZLF1 wild-type binding occurred in the presence of a 100-fold excess of a nonspecific competitor (ns) unrelated to the probe but not with a 100-fold excess of the same nonradioactive double-stranded sequence (comp) used as the probe. (C) Quantitation of EMSA results, averaged from three EMSA experiments, setting the level for BZLF1 at 100%. (D) Results from the BZLF1 transactivation assay. 293 cells were cotransfected with expression vectors for BZLF1 mutants, the pGL3EA reporter plasmid (expressing BMRF1), and a β-galactosidase reference vector. After 48 h, cell extracts were analyzed for luciferase activity, expressed in kilo-relative light units (kRLU), and results were adjusted for transfection efficiency based on the β-galactosidase reference values. pGL3 was the empty luciferase vector, and pBK2 was the empty BZLF1 expression vector. (E) Results from the BZLF1 replication assay. 293-ZKO cells were transfected with expression vectors for BZLF1 mutants and an empty vector control. After 48 h, total DNA was extracted and EBV DNA amplification was analyzed by quantitative real-time PCR using specific primers for the Rp promoter and actin (13). Results are expressed as the change (n-fold) in the level of EBV DNA, relative to that of the actin control. (F) Western blot of transfected 293 and 293-ZKO (ZKO) cells. Protein extracts were prepared 48 h after transfection, and equal amounts of extracts were analyzed by Western blotting with Z185 antibody for BZLF1.

The BZLF1 mutant proteins all gave similar levels of protein expression (measured by the incorporation of [³⁵S]Met) when translated in vitro in reticulocyte lysate (Fig. 2A), and all apart from K207A/L214A gave a measureable shift in the EMSA (Fig. 2B). The EMSA binding was specific since the shift occurred in the presence of an excess of a nonspecific oligonucleotide unrelated to the probe but not an excess of the same double-stranded oligonucleotide used as the probe (Fig. 2B). The EMSA was repeated three times, and the results were quantitated (Fig. 2C). The mutations had mostly small effects in this assay, causing changes of less than twofold in DNA binding ability. The L214A mutant had only about 40% of the wild-type activity, and the K207A/L214A double mutant had a very low level of binding, as we showed previously (7), consistent with the importance of the corresponding heptad positions in stabilizing the coiled-coil dimerization interface.

The transactivation assay (Fig. 2D) involved the cotransfection of 293 cells with a BZLF1 expression plasmid and a BMRF1 promoter-reporter plasmid (5). The EBV lytic replication assay (10, 13) used 293-ZKO cells which contain an EBV mutant lacking the BZLF1 gene (4); transfection with a BZLF1 expression plasmid causes the replication of the EBV DNA, which is measured by quantitative PCR (Fig. 2E). Some of the BZLF1 mutant proteins in transfected cells were unstable (Fig. 2F, right panel), so transactivation and replication data are shown only for mutants that gave protein expression similar to that of the wild-type BZLF1 (Fig. 2F, left panel). All the transactivation and replication assays for which results are presented in this paper were repeated four to six times, in each case with Western blotting analysis of the BZLF1 protein expression. Averaged data with standard deviations from transactivation and replication assays are shown, but just one representative Western blot is shown for the protein expression. Most mutants with protein expression levels similar to that of wild-type BZLF1 exhibited only modest effects on transactivation (changes of less than twofold compared to wild-type transactivation), the minor variations generally correlating with the protein expression level and/or DNA binding activity. Interestingly, the N211A mutation produced a fivefold reduction in DNA replication, perhaps reflecting the central position of N211 within a hydrogen bond network that stabilizes the interaction between the coiled coil and the C-terminal tail (8). Even more striking was S208E; this mutation caused a slight increase in transactivation (Fig. 2D) but severe reduction in the replication assay (Fig. 2E). Similar mutation of the adjacent S209 to glutamate had little effect on transactivation or DNA replication (Fig. 2D and E). We tested other amino acids in position 208 and found that S208D had effects similar to but slightly less pronounced than those of S208E (Fig. 3) and that the protein with the conservative mutation S208T differed little from the wild type. The truncation of the C terminus at residue 228 also prevented DNA replication but allowed transactivation (data not shown), consistent with a previous report (9) that a mutant truncated at 231 retains transactivation function. However, S208E is the first point mutation and the first mutation within the coiled-coil region that has this property of allowing transactivation but preventing DNA replication.

FIG. 3.

Compensation for the S208E replication defect. (A) Results from the DNA binding assay. Equal amounts of in vitro-translated proteins were tested for their abilities to bind DNA by EMSA analysis with the ZRE probe as described in the legend to Fig. 2B, and the results were quantitated as described in the legend to Fig. 2C. comp, competitor oligonucleotide; ns, nonspecific oligonucleotide. (B) Results from the BZLF1 transactivation assay. 293 cells were cotransfected with expression vectors for BZLF1 mutants, the pGL3EA reporter plasmid (expressing BMRF1), and a β-galactosidase vector, and cell extracts were analyzed for luciferase activity as described in the legend to Fig. 2D. pGL3 was the empty luciferase vector, and pBK2 was the empty BZLF1 expression vector. kRLU, kilo-relative light units. (C) Results from the BZLF1 replication assay. 293-ZKO cells were transfected with expression vectors for BZLF1 mutants and an empty vector control, and EBV DNA amplification was analyzed by quantitative real-time PCR as described in the legend to Fig. 2E. The S208E-D236K mutation compensated for the S208E replication defect, as indicated by arrows. (D) Western blot of transfected 293 and 293-ZKO (ZKO) cells. Protein extracts were prepared 48 h after transfection, and equal amounts of extracts were analyzed by Western blotting with Z185 antibody for BZLF1.

Residue S208 is of interest because it directly contacts the C-terminal tail. More specifically, S208 forms hydrogen bonds with (monomer 1) or is in close proximity to (monomer 2) residue D236 in the tail region (Fig. 1D and E, top panel), suggesting that the replacement of S208 by an acidic glutamate or aspartate perturbs the zipper-tail interface through an electrostatic repulsion effect (Fig. 1E, middle panel). To test this possibility, we investigated whether the charge reversal mutation D236K could rescue some of the replication activity of the S208E mutant. None of the single or double S208 or D236 point mutations tested significantly reduced either DNA binding (Fig. 3A) or transactivation (Fig. 3B), and the proteins with these mutations exhibited similar expression levels (Fig. 3D). The mutation of D236 alone to K also gave low replication activity (Fig. 3C), but the double mutant S208E/D236K had about 4.5 times the replication activity of the S208E mutant (Fig. 3C), suggesting that activity was rescued by the restoration of a stable zipper-tail interaction (Fig. 1E, bottom). The fact that the S208E-D236K double mutation did not restore replication activity completely to the wild-type level may indicate that the correct protein structure was not fully restored by the two complementary charge mutations or may reflect an additional role for residues within the SSENDRL motif in EBV DNA replication. Indeed, D236 as well as S208, S209, D212, and R213 within SSENDRL is partly solvent exposed and may conceivably interact with a replication factor that recognizes the hairpin structure of BZLF1. Although the C terminus of BZLF1 was already implicated in EBV DNA replication, our results strongly indicate that interaction between the tail and the coiled-coil region of BZLF1 is required for DNA replication. The results also provide direct functional support for this aspect of the X-ray structure in the context of the native protein in EBV-infected cells.