Effect of phosphorylation on the transactivation activity of Epstein-Barr virus BMRF1, a major target of the viral BGLF4 kinase

Epstein–Barr virus (EBV) is a gamma-1 human herpesvirus that infects most of the human population worldwide. EBV infection is associated closely with several human malignant diseases, including Burkitt's lymphoma, nasopharyngeal carcinoma (NPC) and Hodgkin's disease (Cohen, 2000; Young & Rickinson, 2004). Even though the virus is present in a latent form in most cancer tissues, elevated serum antibody titres against some EBV lytic antigens have been used as diagnostic markers for NPC (Chien et al., 2001) and the serum viral DNA load is an important prognostic parameter for NPC patients (Lin et al., 2004). Thus, the progression of lytic replication is important for the pathogenesis of EBV-associated malignancy.

Transcription-coupled viral DNA replication is a common phenomenon among human herpesviruses for the initiation of lytic replication. The promoters within lytic origin of replication (oriLyt) regions, which have been identified in human cytomegalovirus, Kaposi's sarcoma-associated virus (KSHV or human herpesvirus 8) and EBV, are crucial for viral lytic replication (AuCoin et al., 2004; Hammerschmidt & Sugden, 1988; Xu et al., 2004). Because the promoter regions within oriLyt are important for viral replication and can be substituted by other functional promoters, the transcription machinery has been suggested to remodel the viral chromatin structure and thereby facilitate the binding of the DNA polymerase complex to the viral DNA template (Xu et al., 2004).

In addition to DNA polymerase, functional replication machinery requires a DNA processivity factor (PF), which also is known as an accessory factor or sliding clamp, to stabilize DNA binding by the polymerase during replication (Ellison & Stillman, 2001). The EBV PF BMRF1 (EA-D) has single-stranded (ss) and double-stranded (ds) DNA-binding activities and increases the activity of the viral DNA polymerase BALF5 through direct interaction (Li et al., 1987; Tsurumi, 1993; Tsurumi et al., 1993). In addition to the processivity function, BMRF1 also has a transactivation activity on the viral oriLyt BHLF1 promoter and cellular gastrin promoter (Holley-Guthrie et al., 2005; Zhang et al., 1996, 1997). EBV oriLyt is located within the divergent promoter regions of two early genes, BHLF1 and BHRF1. The BHLF1 promoter is probably the most active EBV promoter and is activated by either Zta or BMRF1 alone or, maximally, by the synergistic effect of both proteins (Zhang et al., 1996). BMRF1 transactivates the oriLyt BHLF1 promoter through the Sp1/ZBP-89-responsive element within the downstream component (Zhang et al., 1997). Even without affecting Sp1 responsiveness, a mutation altering the sequence GATGG within the Sp1/ZBP-89 sites (–588 to –592 relative to the BHLF1 transcription-initiation site) reduced BMRF1 responsiveness by 75 % and abolished oriLyt replication (Zhang et al., 1997). Therefore, BMRF1-induced activation of the oriLyt region is considered to be crucial for viral DNA replication (Zhang et al., 1997).

Herpesvirus PFs have frequently been found to be phosphoproteins (Chan & Chandran, 2000; Chang & Balachandran, 1991; Chee et al., 1990; Gibson et al., 1981; Marsden et al., 1987). However, the phosphorylation sites and functional regulation of the herpesvirus PFs through phosphorylation are mostly unknown. BMRF1 was found to be phosphorylated by EBV BGLF4 kinase in vitro and in co-transfected cells (Chen et al., 2000; Gershburg & Pagano, 2002; Wang et al., 2005). BGLF4 induces the hyperphosphorylated isoforms of BMRF1 from pp50 (phosphoprotein 50 kDa) to pp52 and pp58 (Gershburg & Pagano, 2002; Wang et al., 2005) on SDS-PAGE and co-localizes with BMRF1 in the viral replication compartment (Wang et al., 2005). Knockdown of BGLF4 expression by RNA interference (RNAi) resulted in the absence of the hyperphosphorylated isoforms, confirming that the lytic phosphorylation of BMRF1 is contributed mostly by BGLF4 (Gershburg et al., 2007). However, the BGLF4-targeted phosphorylation site(s) on BMRF1 and the phosphorylation-mediated functional regulation of BMRF1 remain unclear.

BGLF4 is the only known viral serine/threonine protein kinase in EBV (Chen et al., 2000) and has been suggested to mimic cellular Cdc2 (CDK1) kinase activity (Kawaguchi & Kato, 2003; Kawaguchi et al., 2003; Kudoh et al., 2006; Lee et al., 2007). It can be detected in viral particles and was suggested to interact with BXLF1 (viral thymidine kinase) by a yeast two-hybrid study (Calderwood et al., 2007; Wang et al., 2005). In addition to BMRF1, BGLF4 is able to phosphorylate the viral nuclear proteins EBNA-LP (Kato et al., 2003), EBNA-2 (Yue et al., 2005) and BZLF1 (Asai et al., 2006), as well as the cellular translation elongation factor EF-1δ (Kato et al., 2001). Recently, BGLF4 was found, by using an RNAi approach, to be crucial for nuclear egress of virus (Gershburg et al., 2007).

In this study, we set out to explore the regulatory function of BGLF4-mediated phosphorylation on the proline-rich hinge region of BMRF1, which separates the DNA-processivity and nuclear-localization domains and contains seven possible Cdc2 sites. Functional regulation through phosphorylation of these mapped residues was analysed for the subcellular localization, DNA-binding abilities and transactivation activities of BMRF1.

Plasmids.
A BGLF4 expression plasmid containing the BGLF4 open reading frame (positions 110037–111326; GenBank accession no. AJ507799[GenBank] ) and kinase-dead mutant K102I, which contains a mutation at the catalytic lysine, were generated as described previously (Wang et al., 2005). A BMRF1-expressing plasmid (pYPW26 or pSG5-BMRF1) was generated by PCR from pBR322-BamM, which contains the B95-8 EBV BamHI M fragment, with primers LMRC 404 (5'-CCGGAATTCATGGAAACCACTCAGACTCT-3') and LMRC 405 (5'-CGCGGATCCTTAAATGAGGGGGTTAAAGGC-3'), and cloned into the BamHI and EcoRI sites of pSG5 (Stratagene) under the control of the simian virus 40 promoter. A plasmid expressing Flag–BMRF1 (pYPW21) was generated by PCR from pSG5-BMRF1 and cloned into the BamHI and EcoRI sites of pCMV-Tag2b (Stratagene). A plasmid expressing Flag–BMRF1(d316–377) was generated by PCR from the template containing a deletion from nt 946 to 1132 of BMRF1 (Chen et al., 1995). pCR3.1-BMRF1 was derived from plasmid Flag–BMRF1 by cloning the BMRF1 coding region into the BamHI and EcoRI sites of pCR3.1 (Invitrogen). All site-directed BMRF1 mutants were generated by using a single primer-based in vitro mutagenesis strategy (Makarova et al., 2000) with the primers and templates specified in Supplementary Table S1 (available in JGV Online). For the transactivation assay, the reporter plasmid pBHLF1-luc containing a DNA fragment (–1006 to +90 bp relative to the transcription start site of BHLF1; positions 41618–40521, GenBank accession no. DQ279927[GenBank] ) was PCR-amplified from the B95-8 BamH H fragment and cloned into pGL3-basic (Promega). The Zta expression plasmid pRC/CMV-Zta was described previously (Lu et al., 2000). The Rta expression plasmid RTS15 was a gift from Diane Hayward, Viral Oncology Program, Sidney Kimmel Cancer Center, Johns Hopkins School of Medicine, Baltimore, MA, USA (Ragoczy et al., 1998).

Cell lines.
HEK293T (293T) is a derivative of a human kidney epithelial cell line (ATCC no. CRL-1573). The EBV-negative cell line NPC-TW01 was established from a Taiwanese NPC (Lin et al., 1990) and NA is a recombinant Akata EBV-converted NPC-TW01 cell line (Chang et al., 1999).

Transfection, protein extraction and phosphatase treatment.
293T cells were transfected with appropriate amounts of DNA by using the calcium phosphate/BES method (Chen & Okayama, 1987); NA cells were transfected by using Lipofectamine 2000 (Invitrogen). Transfected cells were harvested at the time points indicated and resuspended in RIPA buffer as described previously (Wang et al., 2005). For phosphatase treatment, 3 µl lysate was incubated with phosphatase buffer in the presence or absence of 10 U calf intestinal alkaline phosphatase (CIP; New England BioLabs) at 37 °C for 1.5 h.

Immunoblotting.
Immunoblotting was performed in a manner similar to that described previously (Chen et al., 2000). The primary antibodies (Abs) used were anti-BGLF4 (2616; Wang et al., 2005), anti-Flag Ab (M2; Sigma), anti-BMRF1 Ab (Capricorn), anti-glyceraldehyde-3-phosphate dehydrogenase (GAPDH) Ab (Biodesign), anti-Zta 4F10 (Tsai et al., 1997), anti-caspase 3 Ab (Imgnex) and anti-poly(ADP–ribose) polymerase (PARP) Ab (BD Pharmingen).

In vitro transcription/translation.
Wild-type (WT) and mutant BMRF1 plasmids were expressed in vitro by using the TNT coupled reticulocyte lysate system (Promega) and T7 RNA polymerase, according to the manufacturer's protocol.

Subcellular fractionation.
The cellular fractionation protocol was modified from that described by Krajewski et al. (1993). The details were described previously (Wang et al., 2005).

DNA–cellulose chromatography.
WT and mutant BMRF1 were expressed in 293T cells. At 48 h post-transfection, cells were harvested and lysed in buffer A [25 % (v/v) glycerol, 20 mM Tris/HCl (pH 7.4), 1 mM EDTA, 1 mM EGTA]. Protein lysate (200 µg) was incubated with 500 µl 40 % dsDNA–cellulose or ssDNA–cellulose for 30 min at 4 °C. The protein–cellulose mixtures were then applied to chromatography columns (Bio-Rad) with a 200 µl bed volume. The column was washed with 500 µl binding buffer [20 mM Tris/HCl (pH 7.6), 5 mM MgCl₂] and then eluted with step gradients of 500 µl of each buffer containing 100, 150, 200, 250, 300, 400, 500, 600 or 1000 mM NaCl. The eluants were precipitated further by adding 50 µl 100 % trichloroacetic acid and 50 µl 0.15 % sodium deoxycholate, and subjected to immunoblotting analysis.

Luciferase assay.
293T cells were transfected with pBHLF1-luc, the amounts of transactivator plasmids indicated and Renilla luciferase reporter (pTK-Rluc) as a control for transfection efficiency. At 48 h post-transfection, cells were harvested and assayed for firefly and Renilla luciferase activities by using the Dual-Glo Luciferase Assay system (Promega). Promoter activities were determined from luciferase activity after normalization for Renilla luciferase activity in each reaction. Fold activation indicates the ratio of reporter activity to that of vector-transfected cells.

BGLF4 target residues are located within aa 316–378 of BMRF1
BGLF4 is known to co-localize with BMRF1 within the viral replication compartment and it is able to induce the hyperphosphorylation of BMRF1 in co-transfected cells (Gershburg & Pagano, 2002; Wang et al., 2005). To explore a possible regulatory role of BGLF4 on BMRF1 function, we searched first for possible BGLF4 target sites on BMRF1. According to the report that BGLF4 phosphorylates several substrates at cellular Cdc2 kinase target sites (Kawaguchi et al., 2003), seven putative Cdc2 sites on BMRF1 were identified (Ser-314, Ser-333, Ser-335, Ser-337, Thr-344, Ser-349 and Thr-355; Fig. 1a). To avoid variations in the detection efficiency of BMRF1 mutants using the BMRF1-specific Ab, Flag-tagged BMRF1 was used to analyse the phosphorylation sites. BGLF4-induced hyperphosphorylation of BMRF1 was detected by anti-Flag Ab. In the absence of BGLF4, Flag–BMRF1 displayed two major bands of approximately 55 kDa (Fig. 1b, lane 1, isoforms i and ii), which may be equivalent to the 50 and 52 kDa forms of authentic BMRF1 (Wang et al., 2005). As expected, expression of BGLF4 kinase induced hyperphosphorylated BMRF1 (Fig. 1b, lane 3, isoform iv). After CIP treatment, the molecular mass of BMRF1 was reduced (Fig. 1b, lanes 2 and 4, isoform iii), indicating that BMRF1 is phosphorylated by cellular kinase(s) and can be further phosphorylated by BGLF4. We refer to the isoforms equivalent to ii as hypophosphorylation and to those equivalent to i and iv as hyperphosphorylation. To determine whether the phosphorylated residues are within the predicted region, a deletion mutant, Flag–BMRF1(d316–378), which lacked six of the seven candidate Cdc2 sites, was co-expressed with BGLF4 in 293T cells. Co-expression of BGLF4 shifted the major species of Flag–BMRF1 from the hypophosphorylation to the hyperphosphorylation position (Fig. 1c, lanes 2–6), whereas no significant molecular mass change was found in Flag–BMRF1(d316–378) (Fig. 1c, lanes 8–12). This suggests that the region from aa 316 to 378 of BMRF1 contains the major BGLF4 phosphorylation sites.

(41K): Fig. 1. Identification of the region on BMRF1 responsible for BGLF4-induced hyperphosphorylation. (a) Functional domains of BMRF1 responsible for DNA binding, the nuclear-localization signal (NLS) and transactivation domains are indicated (Zhang et al., 1999). The BMRF1 protein sequence was analysed at . Seven residues, Ser-314, Ser-333, Ser-335, Ser-337, Thr-344, Ser-349 and Thr-355, are predicted to be possible Cdc2 target sites on BMRF1. Summary of Flag–BMRF1 point-mutant constructs are listed. For phosphorylation-defective mutants, serine (S) was substituted by alanine (A), and tyrosine (T) was substituted by valine (V). For phosphorylation-mimicking mutants, serine (S) was substituted by aspartic acid (D), and tyrosine (T) was substituted by glutamic acid (E). (b) Lysates from 293T cells transfected with Flag–BMRF1 and BGLF4 expression plasmids, treated or untreated with CIP and analysed by immunoblotting. i and iv indicate the hyperphosphorylated forms of BMRF1, ii the hypophosphorylated form and iii the non-phosphorylated form. (c) Indicated amounts of BGLF4 WT or kinase-dead (KD) plasmid were co-expressed with 0.2 µg Flag–BMRF1 WT or Flag–BMRF1(d316–378) plasmid in 293T cells. Protein expression was analysed by immunoblotting. Flag–BMRF1 was detected with anti-Flag M2 Ab. GAPDH was used as a loading control. hyper, Hyperphosphorylated isoforms; hypo, hypophosphorylated isoforms.

As revealed in Fig. 1(a), the region from aa 316 to 378 is a functionally uncharacterized domain between the processivity and nuclear-localization domains of BMRF1. This region encompasses a proline- and serine/threonine-rich domain that is very similar to the hinge regions identified in human IgA and other proteins (Kerr, 1990; Kokubo et al., 2000; Zhou et al., 1995). Hinge regions of these proteins are phosphorylated at their CDK target sites and are crucial for inducing conformational changes of these proteins (Knotts et al., 2001; Zhou et al., 1995). We were curious to know whether BGLF4 modulates BMRF1 function through this region.

Ser-337, Thr-344, Ser-349 and Thr-355 are the BGLF4 target sites of BMRF1
We then identified the BGLF4 phosphorylation sites on BMRF1 by introducing specific point mutations (Fig. 1a). The first set comprised Ser-314, Ser-333, Ser-337 and Thr-344. Flag–BMRF1 WT or mutants were transiently co-expressed with BGLF4 or vector control (VC) in 293T cells. Because S314A and S333A displayed similar BGLF4-induced hyperphosphorylation patterns to the WT (Fig. 2a, lanes 4, 8 and 10), these two residues were excluded as BGLF4 target sites. BGLF4-induced hyperphosphorylation isoforms decreased for S337A and T344V (Fig. 2a, lanes 12 and 14). A more obvious decrease in hyperphosphorylation was detected for the S337A/T344V double mutant (Fig. 2b, lane 4). Because residual BGLF4-induced hyperphosphorylation isoforms were observed in S337A/T344V, other residues might be targeted by BGLF4. We then determined whether Ser-349, Thr-355 or Ser-335 is targeted by BGLF4. Compared with S337A/T344V, both S337,49A/T344V and S337A/T344,55V (Fig. 2b, lanes 2, 4, 6 and 8), but not S335,37A/T344V (Fig. 2c, lane 6), displayed a further decrease in BGLF4-induced hyperphosphorylation. Therefore, Ser-349 and Thr-355, but not Ser-335, are also BGLF4 target residues. The quadruple mutant S337,49A/T344,55V (2A2V) was then generated and co-expressed with BGLF4 in 293T cells. We found that BGLF4-induced hyperphosphorylation was abolished in 2A2V (Fig. 2d, lane 6). To rule out a possible structural change affecting the ability of 2A2V to interact with BGLF4, both Flag–BMRF1 WT and 2A2V were co-expressed with BGLF4 and immunoprecipitated by using anti-Flag Ab. BGLF4 was apparently co-immunoprecipitated by both Flag–BMRF1 WT and 2A2V (Fig. 2e). Here, we conclude that Ser-337, Thr-344, Ser-349 and Thr-355 are the major BGLF4 target sites of BMRF1.

(57K): Fig. 2. Mapping of the BGLF4 target residues on BMRF1. (a–d) BGLF4 or VC (vector control) was co-expressed in 293T cells with Flag–BMRF1 WT or mutagenesis mutants as indicated. Protein expression was analysed by immunoblotting. (e) Lysates from 293T cells transiently expressing WT or KD BGLF4 and Flag–BMRF1 WT or 2A2V were immunoprecipitated with anti-Flag Ab and immunoblotted with anti-BGLF4 or anti-Flag Ab. Flag–BMRF1 was detected with anti-Flag M2 Ab. GAPDH was detected as a loading and size control. hyper, Hyperphosphorylated isoforms; hypo, hypophosphorylated isoforms.

Ser-337 and Thr-344 of BMRF1 are the major residues phosphorylated during EBV lytic replication
To determine whether the residues identified in BMRF1 were targeted during EBV lytic replication, Flag-tagged WT BMRF1 or mutants including S337A/T344V, 2A2V, 3A2V and S314,33A were expressed in EBV-positive NA cells. The lytic cycle was induced by transfection of an Rta expression plasmid and the expression of endogenous BGLF4 was confirmed by immunoblotting (Fig. 3a). Similar phosphorylation patterns were observed in WT and mutant S314,33A, confirming that Ser-314 and Ser-333 are not viral kinase target residues (Fig. 3a, lanes 2 and 10). No significant hyperphosphorylation bands were observed for 2A2V or 3A2V during the lytic stage (Fig. 3a, lanes 6 and 8). Similar to that observed in BGLF4 co-transfected 293T cells, a much weaker hyperphosphorylation band of S337A/T344V was seen in Rta-transduced NA cells (Fig. 3a, lane 4). Data here suggest that Ser-337 and Thr-344 are the major residues on BMRF1 that are phosphorylated during viral lytic replication. Furthermore, subcellular fractionation was performed to demonstrate that both Flag-tagged WT and 2A2V BMRF1 were expressed mainly in the nuclei of virus-replicating cells (Fig. 3b), suggesting that phosphorylation of Ser-337 and Thr-344 did not affect the nuclear localization of BMRF1 during viral lytic replication. Moreover, the abolishment of phosphorylation in 2A2V is not due to the loss of nuclear localization.

(39K): Fig. 3. Hyperphosphorylation of Flag–BMRF1 in EBV-replicating NA cells. (a) WT or mutant Flag–BMRF1, including S337A/T344V, 2A2V, 3A2V and S314,33A expression plasmids or vector, were transfected separately with Rta expression plasmid or vector into NA cells. Flag–BMRF1, Rta, GAPDH and BGLF4 were detected with specific antibodies. (b) Lysates from NA cells transfected with Rta and Flag–BMRF1 WT or 2A2V were fractionated into nuclear (N) and cytoplasmic (C) fractions. Poly(ADP–ribose) polymerase (PARP) and caspase 3 were detected with specific antibodies as markers of nuclear and cytoplasmic fractions, respectively.

Characterization of phosphorylation-mimicking BMRF1 mutants
To explore how phosphorylation may affect BMRF1 function, two phosphorylation-mimicking BMRF1 mutants, S337D/T344E (1D1E) and S337,49D/T344,55E (2D2E), were generated by substituting serine (S) by aspartic acid (D) and tyrosine (T) by glutamic acid (E) (Fig. 2a). The subcellular localization, expression patterns and abilities to bind dsDNA and ssDNA were compared among WT BMRF1, 2A2V, 1D1E and 2D2E.

The hyperphosphorylated form of BMRF1 was detected predominantly in the nucleus in our previous study (Wang et al., 2005). The nuclear-localization signal (NLS) was mapped to aa 383–397 of BMRF1, which is very close to the mapped BGLF4 target sites (Zhang et al., 1999). To determine whether phosphorylation affects NLS function, subcellular fractionation of Flag-tagged WT BMRF1, 2A2V, 1D1E and 2D2E was carried out in 293T cells. In essence, all BMRF1 proteins displayed a predominantly nuclear localization (Fig. 4a).

(58K): Fig. 4. Expression and characterization of phosphorylation-mimicking BMRF1. (a) Subcellular distributions of Flag–BMRF1 WT or mutants were analysed by fractionation. Lysates from 293T cells that transiently expressed Flag–BMRF1 WT, 2A2V and 2D2E were fractionated into nuclear or cytoplasmic fractions as described in the legend to Fig. 3(b). Flag–BMRF1, PARP (nuclear-fraction marker) and caspase 3 (cytoplasmic-fraction marker) were detected with specific antibodies. (b) Cell lysates from transiently transfected 293T cells expressing pSG5-based BMRF1 WT, 2A2V, 1D1E, 2D2E or BMRF1 WT with BGLF4 (lane 5) were resolved by SDS-PAGE and detected with anti-BMRF1 Ab. (c) BMRF1 WT, 2A2V and 2D2E (pSG5-BMRF1) expressed in transfected 293T cells or WT BMRF1 and 2D2E expression in reticulocyte lysate (IVT) were treated or not with CIP and analysed by immunoblotting. (d, e) Cell lysates from 293T cells expressing WT BMRF1 or mutants (pSG5-BMRF1), including 2A2V, 1D1E or 2D2E, were applied to dsDNA–cellulose (d) or ssDNA–cellulose (e) chromatography columns and eluted with buffers containing NaCl step gradients as indicated. Eluants were analysed by immunoblotting with anti-BMRF1 Ab. F, Flow-through; S, starting material; W, wash fraction.

Furthermore, we found both in vivo (Fig. 4b, lane 4; Fig. 4c, lanes 5 and 6) and in vitro (Fig. 4c, lanes 9 and 10) that the phosphorylation-mimicking mutant 2D2E, when displayed on SDS-PAGE, showed a much slower mobility (around 56 kDa) than WT BMRF1 (50 and 52 kDa) and 2A2V (50 kDa). To exclude the possibility that the obviously slower mobility was due to additional phosphorylation induced by the phosphorylation-mimicking residues, both in vivo- and in vitro (IVT)-expressed BMRF1 WT and 2D2E were treated with CIP. A slight increase in mobility was observed for in vivo-expressed 2D2E (Fig. 4c, lanes 5 and 6), whereas this was not observed for IVT 2D2E (Fig. 4c, lanes 9 and 10). This suggests that phosphorylation of the identified resides may induce additional phosphorylation of BMRF1 in vivo. Theoretically, the molecular mass of 2D2E will increase by 0.108 kDa compared with that of BMRF1. Therefore, the result here suggests that BGLF4-induced phosphorylation causes an anomalous retardation of migration of BMRF1. Furthermore, 2A2V displayed no obvious differences in phosphorylation compared with WT BMRF1 (Fig. 4c, lanes 1–4), suggesting that cellular kinase(s) also target these BGLF4 sites in the absence of BGLF4.

As a DNA polymerase PF, BMRF1 is able to bind both dsDNA and ssDNA non-specifically through the N-terminal 300 aa (Chen et al., 1995; Kiehl & Dorsky, 1995; Tsurumi, 1993). To determine whether phosphorylation affects the DNA-binding ability of BMRF1, 293T cell lysates expressing WT or mutant BMRF1 were applied to dsDNA–cellulose or ssDNA–cellulose chromatography columns and eluted with NaCl step gradients. WT BMRF1, 2A2V, 1D1E and 2D2E were all detected predominantly in the 250 mM NaCl fraction (Fig. 4d), indicating that the dsDNA-binding activity is not regulated through BGLF4-induced phosphorylation. Notably, 2A2V was detected in a broader range of eluant fractions. In ssDNA binding, the phosphorylation-mimicking mutants 1D1E and 2D2E displayed a slight decrease in binding ability compared with WT BMRF1. 2A2V also displayed a broad spectrum of binding ability to ssDNA–cellulose. However, WT BMRF1 and all of the mutants were eluted predominantly at 300 mM NaCl (Fig. 4e). Therefore, phosphorylation at Ser-337, Thr-344, Ser-349 and Thr-355 of BMRF1 seemed not to be crucial for the DNA-binding properties of BMRF1.

Phosphorylation modulation of BMRF1-induced BHLF1 activation
BMRF1 is known to transactivate the BHLF1 promoter within the oriLyt region (Zhang et al., 1996). This transactivation is suggested to be important for lytic replication. BMRF1 and BGLF4 are able to localize within the viral replication compartment; we therefore suggested that BGLF4 might regulate BMRF1-induced BHLF1 activation.

To test this possibility, increasing amounts of WT BMRF1, phosphorylation-defective mutant 2A2V or phosphorylation-mimicking mutant 2D2E were expressed in 293T cells together with the luciferase reporter plasmid (pBHLF1-luc) containing nt –1006 to +90 relative to the BHLF1 transcription start site (Fig. 5a). We found that 2A2V and 2D2E induced BHLF1 promoter activity in a dose-dependent manner, similar to WT BMRF1 (Fig. 5b). Similar transactivation abilities among the WT and mutants were observed, suggesting that phosphorylation of Ser-337, Thr-344, Ser-349 and Thr-355 does not affect the transactivation activity of BMRF1.

(34K): Fig. 5. Regulation by BGLF4 kinase of BMRF1-induced BHLF1 promoter activity. (a) Schematic structure of oriLyt; divergent promoters, enhancer elements and relevant protein-binding sites are indicated. This illustration is drawn according to published descriptions (Mitsouras et al., 2002; Schepers et al., 1996). The BHLF1 transcription start site is at position 40611 (GenBank accession no. DQ279927) of the viral genome. The BMRF1-responsive element is located at positions –588 to –592 relative to the BHLF1 transcription start site. The pBHLF1-luc reporter covers the region from +90 to –1006. (b) Increasing amounts of pCR3.1-based WT BMRF1, 2A2V or 2D2E expression plasmids were co-transfected with pBHLF1-luc reporter plasmids for promoter assay as described in Methods. Fold activation of indicated amounts of WT BMRF1 (black bars), 2A2V (white bars) or 2D2E (grey bars) is displayed. The remaining lysates from the duplicates were combined into RIPA buffer and analysed by immunoblotting (IB) with specific antibodies. (c) Increasing amounts of WT BGLF4 or KD expression plasmids were co-transfected with WT BMRF1 or 2A2V expression plasmids together with pBHLF1-luc into 293T cells. At 48 h post-transfection, cells were harvested and assayed for luciferase activities as described above. Each experiment was done in duplicate and similar results were observed three to five times.

Alternatively, WT BMRF1 or 2A2V was co-expressed with increasing amounts of the BGLF4 expression plasmid with pBHLF1-luc for the reporter assay. Co-expression of BGLF4, but not kinase-dead BGLF4, downregulated BMRF1 transactivation activity on the BHLF1 promoter in a dose-dependent manner (Fig. 5c). Interestingly, 2A2V responded to BGLF4 downregulation similarly to WT BMRF1, suggesting that BGLF4 inhibits BMRF1 transactivation activity on the BHLF1 promoter independently of phosphorylation of the mapped residues. As Sp1 and ZBP-89 are known to mediate BMRF1 transactivation activity (Holley-Guthrie et al., 2005; Zhang et al., 1997), we examined whether BGLF4 modulates the expression pattern of endogenous Sp1 or co-transfected ZBP-89 in transient transfection. No significant change of expression level or mobility shift of Sp1 or ZBP-89 was observed (see Supplementary Fig. S1, available in JGV Online).

BGLF4 upregulates Zta transactivation and the synergy of BMRF1 and Zta on the BHLF1 promoter
Maximal activation of the BHRF1 promoter depends on the synergy of BMRF1 and Zta (Zhang et al., 1996). As BGLF4 is able to phosphorylate Zta directly (Asai et al., 2006), we examined the effect of BGLF4 on the transactivation activity of Zta in 293T cells with pBHLF-luc reporter (Fig. 6a). In our experimental setting, expression of Zta alone gave a 56-fold transactivation of the BHLF1 promoter, whilst co-expression of BGLF4 stimulated it 124-fold. This enhancement of Zta transactivation was not observed in the presence of kinase-dead BGLF4 (Fig. 6a). In the synergistic experiment, Zta alone gave a 31-fold and BMRF1 gave a 121-fold transactivation, whereas Zta and BMRF1 together gave a 1356-fold transactivation. In the presence of BGLF4, the synergistic effect was increased further to 3236-fold, despite the fact that BGLF4 suppressed BMRF1 transactivation from 121-fold to 37-fold (Fig. 6b). The stimulatory effect of BGLF4 on the synergy of BMRF1 and Zta is dose-dependent and predominantly kinase activity-dependent (Fig. 6c).

(30K): Fig. 6. Effects of BGLF4 and phosphorylation of BMRF1 on synergistic activation of the BHLF1 promoter. Indicated plasmids were transfected with pBHLF1-luc reporter into 293T cells and analysed by luciferase assay. (a) Zta or vector was co-transfected with WT or KD BGLF4 expression plasmids. (b) Zta and BMRF1 were expressed alone or together for synergistic activation of the BHLF1 promoter. BGLF4 or vector was co-expressed with Zta, BMRF1 or both as indicated. (c) An increasing amount of WT BGLF4 or 0.2 µg KD BGLF4 was co-transfected with BMRF1 and Zta expression plasmids for luciferase assay. (d) Increasing amounts of WT BMRF1, 2A2V or 2D2E expression plasmids were co-transfected with Zta expression plasmid for reporter assay. Fold activations of indicated amounts of WT BMRF1 (black bars), 2A2V (white bars) or 2D2E (grey bars) with Zta were displayed. Protein expression was detected by immunoblotting (IB) with specific antibodies.

Phosphorylation-mimicking BMRF1 displays better synergistic activity of the BHLF1 promoter with Zta
Because BGLF4 targets both BMRF1 and Zta, it will be complicated to clarify whether BGLF4 enhances the synergy through the phosphorylation of BMRF1 by co-expressing BGLF4. To simplify the system, we co-expressed Zta with increasing amounts of WT BMRF1, phosphorylation-defective mutant 2A2V or phosphorylation-mimicking mutant 2D2E for reporter assay. 2A2V and 2D2E, as well as WT BMRF1, activated the BHLF1 promoter with Zta synergistically in a dose-dependent manner (Fig. 6d). Compared with WT BMRF1 or 2A2V, 2D2E reproducibly showed approximately 2.3- to 2.6-fold greater synergistic activation with Zta under similar protein expression levels, suggesting that the phosphorylation of BMRF1 enhances its synergistic transactivation activity with Zta on the BHLF1 promoter. Modification of viral PFs by phosphorylation is a phenomenon observed frequently in human herpesviruses. We aimed to identify the BGLF4 target sites on the EBV PF BMRF1 and to examine possible functional regulatory effects of phosphorylation. In this study, four residues, Ser-337, Thr-344, Ser-349 and Thr-355, were identified as the major BGLF4 kinase target sites on BMRF1 through a systematic point-mutation strategy (Fig. 2). We found that in vivo- or in vitro-expressed phosphorylation-mimicking (2D2E) mutants displayed anomalous mobility on SDS-PAGE; however, there were no obvious differences in localization or DNA-binding affinities compared with the phosphorylation-defective mutant (2A2V) and WT BMRF1 (Fig. 4). Similar transactivation activities on the BHLF1 promoter also were observed for 2D2E, 2A2V and WT BMRF1 (Fig. 5). Notably, the phosphorylation-mimicking mutant 2D2E had a greater ability to synergize with Zta to transactivate the oriLyt BHLF1 promoter (Fig. 6d). Moreover, BGLF4 enhanced the transactivation of Zta and the synergistic activation of BMRF1 and Zta of the BHLF1 promoter (Fig. 6a–c), even though it inhibits the BMRF1-induced activation of the BHLF1 promoter through a mechanism independent of targeting the proline-rich cluster of BMRF1 (Fig. 5).

According to a prediction using NetPhos 2.0 (), there are 19 serines, nine threonines and three tyrosines that constitute potential phosphorylation sites on BMRF1. BMRF1 was suggested to be targeted by viral or cellular kinases at multiple sites, due to the dramatic retardation of its mobility on SDS-PAGE, as virus replication proceeds. In this study, we demonstrated that four of the seven putative Cdc2 sites, localized within the region aa 337–355, were critical for BGLF4-induced hyperphosphorylation in vivo (Fig. 2). The phosphorylation-mimicking mutants of BMRF1 induced an anomalous molecular mass shift on SDS-PAGE, similar to that induced by BGLF4. This suggests that the multiple-phosphorylation isoforms of BMRF1 may be the result of different levels of phosphorylation at Ser-337, Thr-344, Ser-349 and Thr-355. Additional phosphorylation of the phosphorylation-mimicking mutant 2D2E was observed, as indicated by CIP treatment (Fig. 4c). Thus, phosphorylation at these residues may result in a conformational change of BMRF1 and facilitate further phosphorylation at other sites. In the in vitro immunoprecipitation kinase assay using recombinant baculovirus-expressed BGLF4, a much weaker phosphorylation signal was observed in glutathione S-transferase (GST)–BMRF1(d316–378) than in GST–BMRF1 (see Supplementary Fig. S2, available in JGV Online). Two additional Sp sites were found at aa 64 and 76 of BMRF1. Whether these two sites are phosphorylated by BGLF4 in vivo or contribute to BMRF1 function will need further investigation.

A possible scenario is proposed to illustrate the regulatory function of BGLF4 on BMRF1 transactivation during viral lytic replication. As revealed by the lytic gene-expression kinetics observed by microarray and Northern analysis, BMRF1 and BGLF4 are expressed after the expression of Zta (Lu et al., 2006). Once Zta binds to oriLyt, BMRF1 can then be recruited to oriLyt through interaction with Sp1/ZBP-89. At this time, BGLF4-induced BMRF1 phosphorylation may facilitate the initiation of viral DNA replication through the enhancement of Zta–BMRF1 synergy. As lytic DNA replication proceeds, BGLF4-induced inhibition of BMRF1 transactivation may prevent the dispensable transactivation activity of BMRF1 for correct processivity function at a later stage. Overall, we suggest that BGLF4 may enhance the efficiency of viral lytic replication by modulating the activation of the oriLyt BHLF1 promoter. This hypothesis is supported in part by the observation that knockdown of BGLF4 expression causes a 30 % decrease in the intracellular viral DNA copy number (Gershburg et al., 2007).

The precise mechanism by which BMRF1 synergizes with Zta to activate the BHLF1 promoter is not yet clear. It was found that direct interaction of Zta and BMRF1 is not required for Zta–BMRF1 synergy on the BHLF1 promoter (Zhang et al., 1996). Thus, the synergy may be mediated by the interplay and the association of the cellular transcription factors recruited by Zta and BMRF1. Zta can interact with cellular factors such as CREB-binding protein (CBP), p53, NF-κB and the basic transcriptional machinery (Sinclair, 2003). The interaction of Zta and TBP can stabilize the association of TFIID and the TATA motif (Lieberman & Berk, 1991). Zta is also known to stimulate the histone acetyltransferase activity of CBP (Chen et al., 2001). This stimulation could lead to chromatin remodelling and enhance the accessibility of oriLyt DNA around Zta-responsive elements. We propose that phosphorylated BMRF1 may form certain structures to stabilize the formation of the transcription-initiation complex containing Zta and BMRF1 around oriLyt, and therefore augment Zta–BMRF1 synergy on the BHLF1 promoter. Additionally, we demonstrate here that BGLF4 kinase activity can enhance the transactivation activity of Zta on the BHLF1 promoter. It is valuable to investigate whether BGLF4 also regulates other downstream viral promoters or Zta-responsive cellular promoters, such as TKT and MMP1 (Lu et al., 2000, 2003).

It has been reported that BGLF4 can downregulate EBNA-2 transactivation activity through direct phosphorylation (Yue et al., 2005). In this study, we demonstrate that BGLF4 downregulates BMRF1 transactivation ability independently of the identified residues on BMRF1. Thus, BGLF4 may regulate BMRF1 activity through other unidentified sites. Alternatively, BGLF4 may modulate other factors involved within the oriLyt region to suppress BMRF1 induction of the BHLF1 promoter. As expression of BGLF4 induces multiple mitotic events similar to those induced by Cdc2, including chromosome condensation and nuclear-envelope breakdown (Lee et al., 2007), it is also possible that BGLF4 may induce DNA structural change of the oriLyt region, leading to the suppression of BMRF1 transactivation activity.

The major BGLF4 target sites identified here are clustered within a hinge region-like domain of BMRF1. Hinge regions are usually flexible linkers with a high proline content and are crucial for transmitting conformational changes. Immunoglobulin hinge regions have been suggested to induce the conformational changes of the antigen-binding arms, allowing appropriate interactions with antigens (Tucker et al., 1981). In the human androgen receptor, substituting Ser-650 in the hinge region with Ala caused a 30 % reduction of the transcriptional activity (Zhou et al., 1995). Interestingly, Ser-337 and Thr-344 are encompassed in the conserved phosphorylation motifs of BMRF1 and another human herpesvirus PF, KSHV PF8, which are located within the functional uncharacterized domains between the processivity domain and NLS (aa 337–344 in BMRF1, aa 330–337 in PF8; see Supplementary Fig. S3, available in JGV Online). It is possible that the proline-rich regions of herpesvirus PFs act as hinge regions and are modulated by phosphorylation.

Taken together, our findings suggest that BGLF4 modulates the BMRF1 transactivation function through multiple mechanisms, which may ensure optimal initiation efficiency of lytic replication. This provides a possible model for herpesvirus kinases in regulating lytic replication through modulating oriLyt activation, and also gives an insight into the regulation of phosphorylation of the possible hinge region between the processivity domain and NLS among gammaherpesvirus PFs.

We thank Dr Shih-Tung Liu (Molecular Genetics Laboratory, Department of Microbiology and Immunology, Chang-Gung University, Taiwan) for plasmids expressing the BMRF1 deletion mutants. We also thank Dr Jen-Yang Chen, Dr Chih-Chung Lu, Mr Jiin-Trang Wang and Dr Chung-Pei Lee (National Taiwan University) for helpful discussions, and Ms Ling-Shih Chang for the generation of GST–BMRF1 proteins. This study was supported by grants NSC94-2320-B002-067 and NSC95-2320-B002-087-MY3 from the National Science Council, NHRI-EX95-9313BI from the National Health Research Institutes and NTU96R0312 from National Taiwan University, Taiwan. We are grateful to Dr Tim J. Harrison of University College London, UK, for critical reading and modification of the manuscript.