Repression of Epstein-Barr Virus EBNA-1 Gene Transcription by pRb during Restricted Latency

Abstract

During the restricted programs of Epstein-Barr virus (EBV) latency in EBV-associated tumors and a subpopulation of latently infected B cells in healthy EBV carriers, transcription of the EBV nuclear antigen 1 (EBNA-1) gene is mediated by the promoter Qp. Previously, two noncanonical E2F binding sites were identified within Qp. The role of E2F in the regulation of Qp, however, has been controversial and is undefined. Here we demonstrate that an E2F factor(s) within Burkitt lymphoma (BL) cells binds to a G/C-rich element [GGCG(C/G)] within the previously identified binding sites in Qp and prototypical E2F response elements. Furthermore, Qp-driven reporter gene expression could be efficiently repressed through either E2F binding site by the tumor suppressor pRb, a potent transcriptional repressor targeted to promoters during G₀ and the early G₁ phase of the cell cycle via its interaction with E2F; a mutant pRb (pRb⁷⁰⁶) lacking E2F binding capability was unable to repress Qp. However, we did not observe cell cycle variation in the expression of either EBNA-1 mRNA or protein in exponentially growing BL cells, consistent with previous predictions that Qp is constitutively active in these cells and with the extremely long t_1/2 of EBNA-1. By contrast, within G₀/G₁ in growth-arrested BL cells, EBNA-1 mRNA levels were twofold lower than in S phase, similar to the two- to eightfold differences in cell cycle expression of some cyclin mRNAs. Thus, although regulation of Qp is coupled to the cell cycle, this clearly has no impact on the level of EBNA-1 expressed in proliferating cells. We conclude, therefore, that the most important contribution of E2F to the regulation of Qp is to direct the pRb-mediated suppression of EBNA-1 expression within resting B cells, the principal reservoir of latent EBV. This would provide a means to restrict unneeded and potentially deleterious expression of EBNA-1 in a nonproliferating cell and to coordinate the activation of EBNA-1 expression necessary for EBV genome replication and maintenance upon reentry of the cell cycle in response to proliferative signals.

Epstein-Barr virus (EBV) establishes a latent (nonproductive) infection of B lymphocytes that persists for the life of its human host. Although infected B cells are capable of expressing at least 12 EBV latency-associated genes, only a subset of these are routinely expressed within the restricted latency programs characteristic of EBV-infected tumor cells and latently infected B cells of healthy EBV carriers (44). Soon after infection, the linear EBV DNA genome that was present within the virion circularizes via its terminal repeat elements, and it is thereafter maintained as an episome within the nuclei of latently infected cells (4, 11, 21). Unlike the replication of EBV DNA mediated by virus-encoded DNA polymerase and accessory proteins during the virus lytic cycle (productive infection), replication of the EBV episome in latently infected cells is mediated by the host cell DNA synthesis machinery and is therefore tightly coupled to the cell cycle (1, 17, 61). Nonetheless, long-term maintenance of the EBV episome in dividing cells does require the latency-associated EBV nuclear antigen 1 (EBNA-1), which binds in a sequence-specific manner to multiple sites within the origin of EBV episomal DNA replication, oriP (5, 6, 43, 62). The contribution of EBNA-1 to EBV genome maintenance, however, is unclear.

Because replication of the EBV genome occurs in synchrony with that of host chromosomes in latently infected cells, the issue of whether the expression of EBNA-1 is regulated in a cell cycle-dependent manner has been raised. Indeed, the demonstration by Sung et al. (56) that the EBNA-1 promoter Qp active during restricted latency can be bound in vitro by E2F transcription factors suggests that expression of EBNA-1 may be activated at the G₁/S boundary of the cell cycle, similar to cellular genes involved in DNA replication whose expression is activated by E2F (12, 14). The E2F binding sites in Qp, which differ significantly from consensus E2F binding sites but have not been well defined, overlap two EBNA-1 binding sites that mediate autorepression (Fig. 1) (45, 47, 49, 56). Based on their observation that an E2F-1 fusion protein can exclude EBNA-1 from binding to the autorepression domain of Qp in vitro, Sung et al. (56) proposed a model whereby E2F activates Qp by displacing EBNA-1 from the promoter during the G₁ phase of the cell cycle, presumably after phosphorylation-induced release of E2F-associated pocket proteins, such as the retinoblastoma susceptibility gene product pRb, that repress E2F-activated transcription (9, 14, 19, 54, 58, 59). In transient-transfection assays, however, overexpression of E2F-1 appeared to activate Qp equally well in EBV-positive and EBV-negative cells (56), and mutations within either putative E2F response element in Qp diminish promoter activity in the absence of EBNA-1 (39), suggesting that E2F can activate Qp independent of EBNA-1. Furthermore, association of pocket proteins such as pRb with E2F does not preclude a priori the binding of E2F to its response elements (2, 14, 35, 36, 51, 55), as is presumed in the proposed model.

FIG. 1.

Organization of the EBV promoter Qp. The nucleotide sequence of the EBNA-1 promoter Qp is shown from −60 to +55 relative to the major site of transcription initiation (+1; thick bent arrow); an alternative site of initiation at −31 is also indicated (thin bent arrow) (40). The major positive regulatory elements QRE-1 and QRE-2 (bound by IRF-1 and IRF-2) are overlined (37, 38, 48); the two EBNA-1 binding sites within the autorepression domain are indicated, as are the potential E2F-binding sites evaluated in this study (Qp5′, Qp_I, and Qp_II). The splice site indicated is the donor splice site of the 5′ exon of the EBNA-1 mRNA.

Although the demonstration that E2F can activate Qp in transient-transfection assays supported a role for E2F in the activation of Qp (56), subsequent studies from our laboratory and others (38, 39, 48) indicated that Qp is activated primarily by interferon regulatory factor 2, a constitutively expressed protein (18), through an element (QRE-2) immediately upstream of the transcription start site (Fig. 1). Because activation of Qp could occur independently of the putative E2F binding sites (37), we proposed that the primary function of E2F may be to target transcriptional repressors such as pRb to Qp to silence EBNA-1 expression in resting B cells (39). A study by Schaefer et al. (48), however, challenged whether E2F factors bind to Qp as initially reported and failed to find corroborating evidence that E2F regulates EBNA-1 expression. Specifically, when expression of a luciferase reporter gene under the control of Qp was analyzed in transiently transfected murine fibroblast cells after release from growth arrest induced by serum starvation, cell cycle periodicity in promoter activity was not observed. Notwithstanding potential differences between human and mouse cells with respect to the expression of E2F family members and their associated factors, this observation suggested that E2F does not play a significant role in the regulation of EBNA-1 expression.

To resolve these issues, we have evaluated binding of E2F to Qp and the functional significance of such an interaction. Our data indicate that an E2F factor(s) present in Burkitt lymphoma (BL) cells, which support EBNA-1 expression through Qp, does indeed bind to two noncanonical E2F binding sites in Qp that contain the core element 5′-GGCG(C/G)-3′, also present within the consensus E2F binding site [TTTT(G/C)(G/C)CG(G/C)]. In cotransfection experiments, Qp could be repressed by pRb through either E2F binding site, and repression required a functional E2F binding (pocket) domain in pRb. Consistent with this observation, we found that in growth-arrested BL cells, EBNA-1 mRNA levels were twofold lower in G₀/G₁ than in the S phase of the cell cycle. However, we observed no difference in EBNA-1 expression between G₀/G₁, S, and G₂/M, in cycling BL cells, indicating that EBNA-1 expression is constitutive in these cells, as previously predicted (38, 48). This suggests that the most significant contribution of E2F and its associated factors to the regulation of EBNA-1 expression during restricted latency is repression of Qp within resting B cells, a major latency-associated reservoir of EBV in vivo (34). This may provide a mechanism to limit unneeded and potentially deleterious expression of EBNA-1 and to coordinate the activation of EBNA-1 transcription upon reentry into the cell cycle in response to proliferative signals.

MATERIALS AND METHODS

Cell culture.

Saos-2 osteosarcoma cells (52), which are functionally null for pRb, were maintained in Dulbecco’s modified Eagle’s medium supplemented with 4.5 g of glucose per liter, 2 mM l-glutamine, and 10% fetal bovine serum (HyClone). The group I BL cell lines KemI and Akata, both of which utilize the EBV promoter Qp to express EBNA-1, were maintained in RPMI 1640 medium supplemented with 2 mM l-glutamine and 10% fetal bovine serum.

Preparation of cell extracts.

BL cells were washed in phosphate-buffered saline (PBS) and resuspended (∼10⁶ cells per ml) in lysis buffer (50 mM Tris-HCl [pH 8.0], 300 mM NaCl, 10% glycerol, 0.1 mM EDTA, 1 mM dithiothreitol [DTT], 0.5% Nonidet P-40, 0.5 mM phenylmethylsulfonyl fluoride, 1 μg each of aprotinin, pepstatin, and leupeptin per ml). Following a 60-min incubation on ice with agitation, the protein extract was clarified by centrifugation at 120,000 × g for 30 min at 4°C and the supernatant was dialyzed against 1 liter of buffer D (20 mM HEPES-KOH [pH 7.9], 0.1 M KCl, 20% glycerol, 0.2 mM EDTA, 0.5 mM DTT, 0.5 mM phenylmethylsulfonyl fluoride). The extract (128 mg of total protein) was then partially purified by heparin-Sepharose column chromatography. Bound protein was eluted from a 5-ml column (Bio-Rad) with a linear gradient of 0.1 to 1.0 M KCl in buffer D (total volume of 50 ml). Fractions (1 ml) were collected and the fractions with peak absorbance at 280 nm (fractions 16 to 30) were pooled and used in DNA binding assays.

EMSA.

Probes for electrophoretic mobility shift assays (EMSAs) were generated from double-stranded oligodeoxynucleotides containing 5′ overhangs by labeling with Klenow DNA polymerase in the presence of 1 mM each dGTP and dTTP and 50 μCi each of [α-³²P]dCTP and [α-³²P]dATP (3,000 Ci/mmol); unincorporated nucleotides were removed by passage through a NucTrap probe purification column (Stratagene). Binding reactions were performed in a 25-μl reaction mixture containing 10 mM HEPES-KOH (pH 7.5), 50 mM KCl, 1 mM EDTA, 0.1 mM DTT, 0.1% Triton X-100, 2.5% glycerol, 1 μg of bovine serum albumin, 1.0 μg of poly(dA-dT)-poly(dA-dT), and 5 μl of heparin-purified BL cell extract. After incubation at 25°C for 10 min, the ³²P-labeled DNA probe (0.5 ng) was added and incubation was continued for 20 min. Competition assays with unlabeled competitor oligodeoxynucleotides were performed by incubating the competitor (100 ng) with the extract for 10 min prior to addition of probe. The sequences of the oligodeoxynucleotides (sense strand) used in this study were as follows: Qp_I, 5′-GATCAAAAGGCGCGGGATAGGATC-3′; mtQp_I, 5′-GATAAAAttatCGGGATAGGATC-3′; Qp_II, 5′-TACCGGATGGCGGGTAATACATG-3′; mtQp_II, 5′-TACCGGATttatGGTAATACATG-3′; Qp5′, 5′-GATCAGATGGCGCGGGTGAGGATC-3′; mtQp5′, 5′-GATCAGATttatCGGGTGAGGATC-3′; E2F, 5′-ATTTAAGTTTCGCGCCCTTTCTCAA-3′; mtE2F, 5′-ATTTAAGTTTCGatCCCTTTCTCAA-3′; E2, 5′-CGTAGTTTTCGCGCTTAAATTTGAGAAAGGGCGCGAAACTAGTC-3′; mtE2, 5′-CGTAGTTcTaGtaCTTAAATTTGAGttAtctgagtAAACTAGTC-3′; c-myc, 5′-CAGAGGCTTGGCGGGAAAAAGAACGGAGGGAGGGATCGCGCTGAGTA-3′; cycA, 5′-TTCAATAGTCGCGGGATACTT-3′; mtcycA, 5′-TTCAATAGagcttGGATACTT-3′; Sp1, 5′-GATCATTCGATCGGGGCGGGGCGAGCGATC-3′; and mtSp1, 5′-ATTCGATCGGttCGGGGCGAGC-3′. In antibody supershift assays, 2 μl of anti-E2F or anti-Sp1 or 5 μl of anti-Rb or anti-p107 antibody was added to the binding-reaction mixtures 10 min after addition of the labeled probe and incubation was continued for an additional 20 min. Antibodies to E2F (H-111) and Sp1 (1C6) were obtained from Santa Cruz Biotechnology, antibody to pRb (Ab-1) was obtained from Oncogene Sciences, and antibodies to p107 (pool of SD2, SD4, SD6, SD9, and SD15) were kindly provided by N. Dyson. Protein-DNA complexes were resolved by electrophoresis in nondenaturing 5% acrylamide gels run at 4°C in 0.5× TBE buffer (1× TBE is 90 mM Tris-HCl, 88 mM boric acid, and 2 mM EDTA). Following electrophoresis, the gels were dried and processed by autoradiography.

Plasmids and site-directed mutagenesis.

Construction of the human growth hormone (hGH) reporter plasmid pOGH.006 containing the Qp promoter from bases −681 to +75 relative to the major transcription start site (+1) has been described, as has its derivative pOGH.006Δ34, which contains a 34-bp deletion from +10 to +43 that destroys the EBNA-1 binding domain (+10 to +53) of Qp (47). Four-base substitutions within Qp were generated in pOGH.006 by using the QuikChange system (Stratagene) as previously described (39). Expression plasmids encoding EBNA-1 (47), pRb (a gift of J. DeCaprio), or pRb⁷⁰⁶ (a gift of W. Kaelin) contained the respective coding sequence inserted into the eukaryotic expression vector pSG5 (Stratagene).

Transfections and reporter gene assays.

Saos-2 cells were transfected by a modified calcium phosphate procedure with N,N-bis(2-hydroxyethyl)-2-aminoethanesulfonic acid (BES; Calbiochem) as a buffer (7). One day prior to transfection, 100-mm tissue culture dishes were seeded with 8 × 10⁵ cells. Transfections were done in triplicate with 10 μg of Qp-hGH reporter plasmid, 5 μg of the appropriate expression vector, and 1 μg of a β-galactosidase expression vector (pCMV-βgal). In experiments involving variable amounts of expression vector, the total amount of DNA introduced was kept constant by addition of pSG5. The calcium phosphate-DNA precipitate was allowed to remain on the cells for 16 h at 35°C in a 3% CO₂ atmosphere. The cells were then rinsed twice with 10 ml of PBS, fed with 10 ml of fresh growth medium, and maintained for an additional 48 h at 37°C under 5% CO₂. The level of hGH in the culture medium was then determined in duplicate by using a radioimmunoassay kit (Nichols Institute). Differences in transfection efficiency were corrected by normalizing hGH values to β-galactosidase activities (adjusted for total protein assayed) present in transfected-cell extracts.

Cell cycle analysis of EBNA-1 expression.

Samples containing 10⁶ cells were washed in PBS, centrifuged, and resuspended in 1 ml of propidium iodide staining solution (0.05 mg of propidium iodide per ml, 0.1% sodium citrate, 0.1% Triton X-100) (27). Each sample was treated at room temperature with DNase-free RNase (0.005 mg/ml) for 30 min, filtered through 40-μm-pore-size nylon mesh, and subjected to fluorescence-activated cell sorter (FACS) analysis with a Becton Dickinson FACScan flow cytometer to determine the DNA content of nuclei. The percentages of cells within the G₀/G₁, S, and G₂/M phases of the cell cycle were determined by analysis with the computer program ModFit (Verity Software House). To isolate BL cells within specific phases of the cell cycle, Akata cells in either mid-log- or stationary (growth-arrested)-phase growth were labeled with 0.01 mM Hoechst 33342 (Sigma) at 37°C for 45 min and subjected to cell sorting with a Becton Dickinson FACS Vantage cell sorter. Cells within the G₀/G₁, mid-S, or G₂/M phase of the cell cycle were collected in sterile PBS at 4°C. For analysis of EBNA-1 protein expression, 1.25 × 10⁶ cells per sample were lysed in 30 μl sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) sample buffer (100 mM Tris-HCl [pH 6.8], 200 mM DTT, 4% SDS, protease inhibitor cocktail [Sigma], 20% glycerol, 0.2% bromophenol blue), and immediately heated at 100°C for 5 min. The proteins were fractionated by SDS-PAGE in 10% acrylamide gels, transferred to an Immobilon P membrane (Millipore), and processed by immunoblotting with an enhanced chemiluminescence detection system (Amersham). EBNA-1 was detected with a polyclonal rabbit antiserum (a gift of J. Hearing) followed by a secondary anti-rabbit antibody conjugated to horseradish peroxidase. The blots were subsequently stripped of antibody in 62.5 mM Tris-HCl (pH 6.8)–100 mM β-mercaptoethanol–2% SDS (50°C for 30 min) and probed with a mouse monoclonal antibody to actin (N350; Amersham) as a control for protein loading.

For measurement of EBNA-1 mRNA expression by semiquantitative reverse transcription-PCR (RT-PCR), total cellular RNA from sorted cells was isolated with RNAzol B as recommended by the manufacturer (Tel-Test) and extracted with an equal volume of phenol-chloroform and then chloroform prior to ethanol precipitation. Threefold serial dilutions of RNA (1,000 to 12.3 ng) were reverse transcribed at 42°C with Superscript-II reverse transcriptase (Gibco-BRL), as specified by the manufacturer, in 20-μl reaction mixtures containing 10 pmol each of primer specific for the EBNA-1 (Qp-derived) and ribosomal protein S14 mRNAs. The RT primers used were 5′-GTGGGTCCCTTTGCAGCCAA-3′ (EBNA-1) and 5′-ATCCGCCCGATCTTCATACC-3′ (S14). Control reactions with 1 μg of RNA but without reverse transcriptase were run in parallel. Following cDNA synthesis, samples were heated at 70°C for 15 min and diluted to 500 μl with 10 mM Tris-HCl (pH 8.0)–0.1 mM EDTA. Amplifications were performed in 50-μl reaction mixtures containing 10 mM Tris-HCl (pH 9.0); 2.5 mM MgCl₂; 50 mM KCl; 10% dimethyl sulfoxide; 25 pmol of each primer; 1 mM each dATP, dCTP, dGTP, and dTTP; 10 μl of diluted RT product; and 2.5 U of Taq DNA polymerase. One cycle of amplification consisted of 95°C for 40 s, 55°C for 2 min, and 72°C for 3 min; following the final cycle (30 cycles for EBNA-1, 25 cycles for S14), samples were maintained at 72°C for 15 min. The PCR primers used were as follows: EBNA-1, 5′-AAGGCGCGGGATAGCGTGCG-3′ (5′ primer) and 5′-GTCTTGGCCCTGATCCTGAG-3′ (nested 3′ primer); S14, 5′-GGCAGACCGAGATGAATCCTCA-3′ (5′ primer) and 5′-CAGGTCCAGGGGTCTTGGTCC-3′ (nested 3′ primer). One-tenth of each product was electrophoresed in a 1.5% agarose gel, transferred to a GeneScreen Plus membrane (Dupont), processed by standard Southern blot hybridization techniques, and quantitated by PhosphorImager analysis (Molecular Dynamics). To determine the relative levels of EBNA-1 mRNA expressed in each phase of the cell cycle analyzed, PhosphorImager values obtained for EBNA-1-specific signals within the linear range of detection were normalized to the corresponding S14 signal.

RESULTS

Characterization of the variant E2F binding sites within the EBNA-1 promoter Qp.

A previous report indicated, based on DNase I footprinting assays, that glutathione S-transferase–E2F-1 fusion proteins are capable of binding to two regions immediately downstream of the transcription initiation site of the EBNA-1 promoter Qp: 5′-AAAAGGCGCGGGA-3′ (+1 to +13) and 5′-TACCGGATGGCGGGTAATACA-3′ (+24 to +44) (56). Although it was noted that neither sequence contained a prototypical E2F binding site [TTTT(G/C)(G/C)CG(G/C)] (26), a reporter gene under the control of Qp did appear to be responsive to E2F in cotransfection experiments (56). However, in DNA binding experiments with extracts of Jurkat T cells, binding to the more downstream site (considered to be the higher-affinity binding site) could be competed only moderately or not at all with oligodeoxynucleotides containing known E2F binding sites from various promoters (56). This observation, in conjunction with a subsequent report from another laboratory that did not find convincing evidence of appreciable E2F-specific binding to this region of Qp (48), prompted us to reassess the binding of E2F factors to Qp.

Although the two regions of Qp protected by E2F-1 fusion proteins in DNase I footprinting assays (noted above) did not contain a canonical E2F binding site, comparison of the sequences within these footprints revealed a G/C-rich element in each (Qp_I and Qp_II [Fig. 1]) that was identical or nearly identical to the G/C component of either the consensus E2F binding site or known E2F binding sites within several E2F-responsive promoters (Table 1). To determine if these sites in Qp had binding properties consistent with an E2F response element, we performed a series of EMSAs with extracts from a group I BL cell line (KemI) in which EBNA-1 expression is driven by Qp (previous experiments were limited to non-EBV-infected T cells [56]). Although we detected some nonspecific binding to Qp_I and Qp_II, each element generated two specific complexes that were identical in mobility and relative intensity to two complexes generated under the same binding conditions with a probe containing a consensus E2F binding site (Fig. 2). Most importantly, the wild-type but not the mutated Qp_I, Qp_II, and E2F oligodeoxynucleotides could compete with each other for binding. Interestingly, the E2F binding site within the cyclin A promoter (50), a variant E2F binding site very similar to Qp_I and Qp_II (Table 1), could not compete for binding with either Qp site (Fig. 2). A third potential E2F binding site, Qp5′, located upstream of the transcription start site (Fig. 1), was also unable to compete with Qp_I and Qp_II (Fig. 3). Since the cyclin A and Qp5′ sites contain a core sequence similar (cyclin A) or identical (Qp5′) to Qp_I or Qp_II, the nucleotides surrounding the GGCG(C/G) core element obviously contribute significantly to the binding potential of these noncanonical E2F sites.

TABLE 1.

Comparison of E2F-like sites in Qp to known E2F binding sites

^aPutative (Qp) and known E2F binding sites in oligonucleotides used in this study. Boxed elements contain a partial or complete E2F binding site relative to the consensus [TTTT(G/C)(G/C)CG(G/C)]; the invariant CG pair present in known E2F binding sites is in boldface type. 

^bThe ability to compete (+) or not (−) for specific binding in EMSAs is indicated. 

^cOligonucleotides representative of the c-myc and adenovirus E2 promoters contained both E2F binding sites (I and II) from the respective promoter. 

FIG. 2.

Qp_I and Qp_II have protein binding properties identical to a consensus E2F binding site in BL cell extracts. EMSAs were performed with protein from the group I BL cell line KemI and double-stranded oligodeoxynucleotide probes containing the putative E2F binding sites Qp_I and Qp_II or a consensus E2F binding site (Table 1). Unlabeled competitor oligodeoxynucleotides added to binding-reaction mixtures in 200-fold excess of probe were the unmutated or mutated (mt) probes themselves or oligodeoxynucleotides containing the known E2F binding site(s) within the promoters for the genes encoding adenovirus E2 (22, 56), c-Myc (20, 57), or cyclin A (50). The Sp1 and mtSp1 competitors contained a consensus or mutated Sp1 binding site (23), respectively. Arrows indicate the complexes resulting from specific binding.

FIG. 3.

A potential upstream variant E2F binding site, Qp5′, has protein binding properties distinct from those of Qp_I and Qp_II. Protein-DNA complexes generated with KemI BL cell extracts and probes containing either Qp_I or Qp_II could be competed with unlabeled oligodeoxynucleotides containing either Qp_I or Qp_II but not the Qp5′ site (−41 to −34 [Fig. 1]) or mutated Qp_I and Qp_II sites. The specific complexes detected that were consistent with E2F binding as shown in Fig. 2 are indicated by arrows.

Finally, to confirm that E2F and E2F-associated factors such as the pocket proteins pRb and p107 (negative regulators of E2F function) were present within the complexes generated with Qp_I and Qp_II, we performed antibody supershift assays. As illustrated in Fig. 4, an antibody cross-reactive with E2F family members 1 to 5 could generate a supershifted complex containing either Qp_I or Qp_II, as did an antibody to pRb (the complex shifted with the pRb antibody requires longer exposure to be readily visible). Neither an antibody to p107 nor an antibody to the transcription factor Sp1 had an observable effect on complex formation or mobility. Furthermore, the lack of an effect of the Sp1 antibody (Fig. 4) and the inability of an oligodeoxynucleotide containing a consensus Sp1 site to compete with either Qp_I or Qp_II for binding (Fig. 2) indicated that these Qp elements, which are very similar to the G/C-rich binding sites of Sp1 (23), are not Sp1 binding sites. This is significant in that promoters that lack a TATA box, such as Qp, often contain multiple Sp1 binding sites (15, 42). Based on our DNA binding assays, we concluded that an E2F factor(s) expressed within B cells that maintain an active Qp is indeed capable of binding to Qp at two sites, Qp_I and Qp_II, consistent with the earlier observations of Sung et al. (56).

FIG. 4.

Detection of E2F-specific binding to Qp_I and Qp_II in BL cell extracts. Binding reactions were performed with probes containing either Qp_I or Qp_II in the absence or presence of antibodies broadly reactive with E2F family members 1 to 5 (αE2F), the E2F-associated pocket proteins pRb (αRb) and p107 (αp107), or the transcription factor Sp1 (αSP1). The antibody-shifted complexes detected are indicated by asterisks. The two complexes identified in Fig. 2 that result from specific binding to an E2F site in the probe are indicated by arrows.

Qp-mediated transcription can be repressed by pRb.

We next addressed whether Qp_I and Qp_II function as bona fide E2F response elements by testing whether pRb is capable of repressing a Qp-driven reporter gene. The Rb protein, which has no specific DNA binding capability itself, represses E2F-activated transcription upon being targeted to the promoter by a direct interaction with E2F, which binds to DNA in association with one of its three dimerization partner proteins DP-1, DP-2, or DP-3 (29, 30). Because endogenous pRb is normally in excess of E2F, these experiments were done in Saos-2 osteosarcoma cells, which are functionally null for pRb (52). As shown in Fig. 5A, in the presence of increasing amounts of cotransfected pRb expression vector (0 to 5.0 μg), reporter gene expression was repressed in a dose-dependent manner, with a maximal repression of ∼75%. Under the same cotransfection conditions, EBNA-1 repressed Qp activity by 90% (Fig. 5B, left). No repression was observed, however, upon cotransfection with an expression vector that encodes a mutant pRb (pRb⁷⁰⁶) unable to bind E2F, suggesting that this effect of pRb is indeed mediated through E2F. Identical results were obtained when these experiments were repeated with pRb-negative C33A carcinoma cells (data not shown).

FIG. 5.

Qp is repressible by pRb. (A) pRb⁻ Saos-2 cells were cotransfected with 10 μg of the Qp-hGH reporter plasmid pOGH.006 (Qp coordinates −681 to +75) and either 5 μg of empty expression vector pSG5 or an increasing amount (0.005, 0.05, 0.5, or 5 μg) of a pSG5-based pRb expression plasmid (pSG.Rb). Total DNA per transfection was kept constant by the addition of pSG5. (B) Saos-2 cells were transfected with 10 μg of pOGH.006 (Qp) or pOGH.006Δ34 (QpΔ34), which lacks the Qp autorepression domain and the Qp_I and Qp_II E2F binding sites, and 5 μg of either pSG5 or a pSG5-based expression plasmid encoding EBNA-1, pRb, or pRb⁷⁰⁶, a mutant unable to bind E2F factors. All transfection mixtures included 1 μg of a β-galactosidase expression plasmid. Expression of hGH was determined in duplicate at ∼40 h posttransfection, and hGH values were normalized to β-galactosidase activity to correct for differences in transfection efficiency. For each experiment, transfections were performed in triplicate; data presented are from a representative experiment in which hGH expression is given as percentage of control (Qp in the presence of empty pSG5).

To confirm that repression of Qp by pRb was mediated through Qp_I and/or Qp_II, we evaluated the ability of pRb to repress a Qp-driven reporter gene lacking these sites. As demonstrated in Fig. 5B (right), a construct (QpΔ34) containing a 34-bp deletion from +10 to +43 of Qp, which destroys EBNA-1 responsiveness (47) and completely removes Qp_II and the last three nucleotides of the G/C domain of Qp_I, was unresponsive to EBNA-1 and pRb. Introduction of the deletion alone resulted in at least a twofold greater level of Qp activity relative to the control reporter (Qp). Although the basis for this effect is not known, it is unlikely to be the result of decreased repression (by a non-pRb pocket protein) due to deletion of the E2F binding elements, since mutation of Qp_I and Qp_II alone or together did not have the same effect (see below). When the same mutations that destroy E2F-specific binding (Fig. 2) were introduced into either Qp_I or Qp_II, Qp activity could still be repressed by pRb, although to a lesser extent than in the unmutated construct (Fig. 6). However, when both Qp_I and Qp_II were mutated within the same construct, Qp was no longer repressed by pRb (Fig. 6). These data demonstrate, therefore, that repression of Qp by pRb is mediated through the variant E2F binding sites Qp_I and Qp_II.

FIG. 6.

Repression of Qp by pRb is mediated through Qp_I and Qp_II. Qp_I and Qp_II were mutated singly or together within the Qp-hGH reporter plasmid pOGH.006 by introducing 4-bp substitutions (see Materials and Methods) that destroyed E2F-specific binding to these elements (as in the experiment in Fig. 2). The pRb responsiveness of these mutated promoters (mtQp) relative to the wild-type promoter (wtQp) was assessed in cotransfection experiments, as described in the legend to Fig. 5, with 10 μg of the appropriate reporter plasmid and 5 μg of either pSG5 or pSG.Rb.

Cell cycle regulation of EBNA-1 expression.

If E2F and its associated factors play a significant role in the regulation of EBNA-1 expression during restricted latency, Qp activity should vary with respect to the cell cycle. Specifically, the negative effect that pRb and other pocket proteins have on E2F within G₀ and the early G₁ phase of the cell cycle dictates that an increase in EBNA-1 expression should occur as cells progress into S phase. To test this, we isolated Akata BL cells from the G₀/G₁, mid-S, and G₂/M phases of the cell cycle from a population of asynchronously growing (cycling) BL cells by a cell-sorting procedure based on the DNA content of viable cells that had been stained with Hoechst 33342 dye. The sorted cell subpopulations were then analyzed for the expression of EBNA-1 protein by immunoblotting or for EBNA-1 mRNA by semiquantitative RT-PCR. The advantage of this procedure over drug-induced cell synchronization and analysis of gene expression at time intervals following release of the cell cycle block is that group I BL cells readily undergo apoptosis in response to drugs commonly used for synchronization, e.g., aphidicolin and nocodazole, and that once the drug is removed these cells often do not synchronously reenter the cell cycle (46a). Thus, the chosen procedure enables one to obtain a high percentage of viable cells in specific phases of the cell cycle.

When cells isolated from a population of proliferating BL cells were analyzed for EBNA-1 expression, no variation in EBNA-1 protein or mRNA was observed among cells within the G₀/G₁, S, or G₂/M phase of the cell cycle (Fig. 7A). This was consistent with previous predictions by Schaefer et al. (48) and our laboratory (38) that Qp-mediated EBNA-1 expression in BL cells is constitutive. However, we reasoned that if Qp were repressed upon entering G₁, there may be insufficient time in cycling cells (analyzed in Fig. 7A) for a detectable turnover of EBNA-1 mRNA to occur in G₁ before the cells again pass into S phase. If this were true, a subpopulation of cells that had a higher proportion of cells in G₀ than in G₁ should express smaller amounts of EBNA-1 mRNA than will cells in S phase. To test this, we allowed Akata BL cells to reach stationary-phase growth (5 days after feeding) before isolating G₀/G₁- and mid-S-phase cells for analysis of EBNA-1 mRNA expression by RT-PCR. We have recently shown that in BL cells EBV induces a posttranscriptional downregulation of c-myc expression (the major growth-promoting factor in BL cells) upon reaching the stationary phase of the cell growth cycle, thus circumventing the proapoptotic properties of c-Myc under growth-limiting conditions (46). As shown in Fig. 7B, the majority of cells underwent growth arrest in G₀/G₁ upon reaching stationary phase. When the G₀/G₁ subpopulation of these cells (which should have a higher proportion of cells in G₀ or early G₁ prior to hyperphosphorylation of pRb) was compared to the mid-S-phase subpopulation, a twofold-higher level of EBNA-1 mRNA was detected in the S-phase cells (Fig. 7B; the G₀/G₁-to-S ratios in two independent experiments were 1:2.43 and 1:1.85). These data, therefore, were consistent with our observation that Qp is repressible by pRb and suggest that regulation of Qp is indeed coupled to the cell cycle. However, this aspect of Qp regulation clearly has little or no impact on the level of EBNA-1 expression in proliferating cells (Fig. 7A). We surmise, therefore, that the most important contribution of E2F and its associated factors to the regulation of Qp is to suppress EBNA-1 expression in a resting B cell.

FIG. 7.

EBNA-1 expression is regulated in a cell cycle-dependent manner in resting but not cycling cells. (A) Analysis of EBNA-1 protein and mRNA levels in cycling cells. Akata BL cells in log-phase growth were stained with Hoechst dye 33342, and cells within the G₀/G₁, S, and G₂/M phases of the cell cycle were isolated by FACS (brackets in the FACS profiles on right). EBNA-1 protein levels were analyzed by immunoblotting (1.25 × 10⁶ cells per sample); the blot was then stripped and reprobed for actin to detect differences due to unequal sample loading. EBNA-1 mRNA levels were assessed by semiquantitative RT-PCR coupled with Southern blot hybridization to enable quantitation of PCR products by PhosphorImager analysis; amplification of the S14 ribosomal protein mRNA served as an internal control to normalize EBNA-1 values. Lane 1 in each data set shows the negative control reaction that contained 1 μg of RNA template but lacked reverse transcriptase; lanes 2 to 6 represent threefold serial dilutions of the RNA template (1000 to 12.3 ng). A ratio of the relative levels of EBNA-1 mRNA detected in each phase of the cell cycle analyzed is presented to the right of the Southern blot. (B) Analysis of EBNA-1 mRNA levels in growth-arrested cells. Akata BL cells were harvested 5 days after feeding, and cells within the G₀/G₁ and S phases of the cell cycle were isolated and EBNA-1 mRNA levels were analyzed as described for panel A. Relative to a cycling-cell population, the majority of cells in the growth-arrested population (61%) were in G₀/G₁ (compare the FACS profiles of cells stained with propidium iodide [P.I.] in panels A and B). Data in panels A and B are representative of two independent experiments; the G₀/G₁-to-S ratio of EBNA-1 mRNA levels in growth-arrested cells in the two experiments were 1:2.43 (shown) and 1:1.85.

DISCUSSION

The E2F family of transcription factors activates the expression of numerous genes involved in DNA synthesis and regulation of the cell cycle. Here we have shown that pRb, a negative regulator of E2F without specific DNA binding capability itself (36), can repress the EBV EBNA-1 promoter Qp through two variant E2F binding sites. In its hypophosphorylated state in quiescent cells (G₀) and the early G₁ phase of the cell cycle, pRb and other pocket proteins (p107 and p130) are targeted to responsive promoters through their interaction with specific E2F family members, which bind to DNA as a heterodimer in association with one of three DP proteins (29, 30). As cells progress through G₁ into S phase, derepression of E2F-responsive promoters is believed to occur through the release of these repressors from the DNA-bound E2F-DP complex as the result of increased phosphorylation of the pocket proteins by cyclin-dependent protein kinases (9, 14, 19, 54, 58, 59).

Consistent with our observation that pRb can repress Qp in transient-transfection assays, we found that in latently infected BL cells, EBNA-1 mRNA levels were lower in growth-arrested cells, most of which are likely to be in either G₀ or early G₁ prior to hyperphosphorylation of pRb, than in cells within the S phase of the cell cycle. This suggests that the EBNA-1 promoter in these cells is subject to cell cycle-specific regulation and that the variant E2F-binding sites in Qp are indeed functional within the virus genome. The observed twofold difference in EBNA-1 transcript levels, although seemingly small, is consistent with the cell cycle-dependent differences in transcript levels (two- to eightfold) of other genes reported to be cell cycle regulated, including cyclins E, A, and C (13, 16, 33, 41). We did not, however, observe cell cycle periodicity in EBNA-1 mRNA levels within cycling cells, possibly due to insufficient turnover of transcripts in G₁ prior to reentering S phase. Regardless, because of the exceptional stability of the EBNA-1 protein (10, 32), any repression of EBNA-1 transcription in G₁ would be unlikely to significantly affect EBNA-1 levels in cycling cells. Therefore, negative regulation of Qp by pRb would appear to be manifested primarily in resting cells. This may explain the inability to consistently detect the expression of EBNA-1 in resting B cells, the major reservoir of EBV in the peripheral blood (34).

Recently, Davenport and Pagano (10) reported an increase in EBNA-1 mRNA expression in BL cells and a 2.4-fold increase in Qp-mediated reporter gene expression in NIH 3T3 fibroblasts as the cells entered and progressed through S phase following the release of cell cycle arrest with nocodazole (G₂/M), consistent with our observations in growth-arrested BL cells (Fig. 7B). It should be noted that these data (10) and our data indicating that Qp is downregulated in growth-arrested BL cells are in contrast to a previous report (48), which did not find an increase in Qp-driven reporter expression in murine fibroblasts upon reversal of growth arrest that had been induced by serum starvation. However, aside from potential differences in the regulation of Qp within the EBV genome and the various reporter plasmids used in the previous studies (10, 48), murine cells express considerably lower levels of pRb than do human cells (35, 55). Thus, if the E2F factor(s) that binds to Qp is one that preferentially interacts with pRb, one might not expect to observe efficient cell cycle-specific regulation of Qp in murine cells.

Previously, Sung et al. (56) proposed a model in which E2F factors activate Qp by displacing bound EBNA-1 from the autorepression domain of Qp as cells progress from G₁ to S phase. This model was based on the observations that E2F could exclude the binding of EBNA-1 to Qp in vitro and that coexpression of E2F-1 or a pRb binding form of E1A (but not a pRb binding mutant) activated Qp-driven reporter expression in transient-transfection assays. However, overexpression of E2F activated Qp equally well in the absence of EBNA-1 in EBV-negative cells (56). Thus, an alternative explanation of these transfection data, consistent with the results presented here, is that overexpression of E2F resulted in activation of reporter expression by titrating out negative regulators of E2F-mediated activation of Qp, such as pRb. That E2F factors contribute directly to activation of Qp is supported by our observation in this study with Saos-2 cells (Fig. 6) and previously with EBV-negative BL cells (39) that mutations which target the E2F binding sites in Qp reduce promoter activity by 20 to 60%, depending on the cell line used. Also, the previously proposed model assumes that E2F-DP complexes containing pRb or other pocket proteins are unable to bind DNA, which is not necessarily true (2, 14, 35, 36, 51, 55). Further, given the autoregulatory role of EBNA-1 (45, 47, 49, 56), it seems unlikely that additional expression of EBNA-1 (mediated by E2F displacement of EBNA-1) would be required if levels of EBNA-1 were already sufficient to occupy the autorepression domain of Qp. Therefore, we propose an alternative model whereby E2F contributes directly to activation of Qp in proliferating cells when EBNA-1 levels are insufficient to fully occupy the autorepression domain but ultimately mediates the repression of Qp by targeting pRb to the promoter in cells that have entered the resting state. Such a model is consistent with the data presented here and previously (39, 56) and is also compatible with the autoregulatory role of EBNA-1.

Since the essential, as yet undefined role of EBNA-1 in the maintenance (if not replication) of the EBV genome dictates that EBNA-1 be expressed in a proliferating cell (3), why would there be a need to repress EBNA-1 in a resting B cell? Although HLA class I- and II-restricted cytotoxic T lymphocytes specific for EBNA-1 have been isolated from the peripheral blood of healthy EBV carriers, cells that endogenously express EBNA-1 do not efficiently process and display EBNA-1 peptides in context with HLA antigens on their cell surface and therefore do not elicit a cytotoxic T-cell response (8, 24, 25, 31). Thus, it is unlikely that repression of EBNA-1 is critical to enable a latently infected cell to evade the anti-EBV immune surveillance of the host. Alternatively, repression may be important to limit deleterious physiological effects associated with sustained expression of EBNA-1 in a resting cell. Although EBNA-1 is not known to overtly affect cell growth or survival, it is not without oncogenic potential (28, 60). EBNA-1 has been reported to activate the expression of the recombinase activating genes 1 and 2 (RAG-1 and RAG-2) (53), and it is therefore conceivable that long-term expression of EBNA-1 could result in genomic instability as a result of inappropriate expression of the RAG genes. Whether the oncogenic potential of EBNA-1 is related to induction of genomic instability, however, has not been addressed. In summary, by usurping E2F and pRb as regulators of Qp during restricted latency, EBV appears to have adopted a mechanism whereby EBNA-1 expression can be coordinated with both the entry of a latently infected cell into a resting state (repression of EBNA-1) and its reentry into the cell cycle in response to a proliferative signal (activation of EBNA-1 expression). Given the pivotal role of EBNA-1 in EBV latency, such a mechanism is likely to contribute significantly to the long-term association of EBV with its host.