Epstein-Barr Virus Episome Stability Is Coupled to a Delay in Replication Timing

Abstract

The temporal regulation of DNA replication is thought to be important for chromosome organization and genome stability. We show here that Epstein-Barr virus (EBV) genomes replicate in mid- to late S phase and that agents that accelerate replication timing of EBV reduce viral genome stability. Hydroxyurea (HU) treatment, which is known to eliminate EBV episomes, shifted EBV replication to earlier times in the cell cycle. HU treatment correlated with hyperacetylation of histone H3 and loss of telomere repeat factor 2 (TRF2) binding at the EBV origin of plasmid replication (OriP). Deletion of TRF2 binding sites within OriP or short hairpin RNA depletion of TRF2 advanced the replication timing of OriP-containing plasmids. Inhibitors of class I histone deacetylases (HDACs) increased histone acetylation at OriP, advanced the replication timing of EBV, and reduced EBV genome copy number. We also show that HDAC1 and -2 form a stable complex with TRF2 at OriP and that HU treatment inhibits HDAC activity. We propose that the TRF2-HDAC complex enhances EBV episome stability by providing a checkpoint that delays replication initiation at OriP.

Epstein-Barr virus (EBV) is a human gammaherpesvirus that has been implicated as a causal cofactor of several human malignancies, including Burkitt's lymphoma, nasopharyngeal carcinoma, Hodgkin's disease, and lymphoproliferative diseases during immunosuppression (19, 21, 32, 46). The virus typically establishes a latent infection in B lymphocytes, where it persists as a multicopy circular minichromosome that replicates in synchrony with the cellular genome (1, 45). The EBV origin of plasmid replication (OriP) binds the virus-encoded EBNA1 protein to form an efficient replicon that is essential for the plasmid stability of latent episomes (reviewed in references 25 and 41). EBNA1 binds to a minimal replicator sequence within OriP, referred to as the dyad symmetry (DS) region, which consists of two pairs of EBNA1 binding sites flanked by telomere repeat factor (TRF) binding sites (2, 8, 22, 44). This minimal replicator has been shown to function indistinguishably from cellular origins that are licensed to replicate once per cell cycle (5, 9, 33, 34). Epigenetic events are necessary for the establishment of a stable episomal origin, but the nature of these modifications has not been identified (24, 29).

EBV episomes can be eliminated from some cell types by treatment with hydroxyurea (HU) (17, 40). HU has been used in clinical trials for treating EBV-positive central nervous system lymphomas (6, 38, 40). HU has also been used to eliminate the double minute satellite DNA that accumulates in many cancer cells and often carries amplified oncogenes or tumor suppressors, including c-myc and MDM2 (23, 39). The underlying mechanism of action of HU has been attributed to its ability to inhibit ribonucleotide reductase and slow cell cycle progression (43). The mechanism of HU-induced loss of double minutes and EBV episomes is not known but is thought to function through its effects on ribonucleotide reductase and DNA replication (18).

Chromatin organization and histone modifications have also been implicated in the regulation of OriP replication activity and plasmid stability. Previous studies have found that nucleosomes flank DS elements and undergo cell cycle changes in histone modifications (48). In particular, histone H3 acetylation was reduced in G₁/S and then enriched in mid-S phase. The histone H3 deacetylation in early S phase raised the question whether OriP was subject to temporal control of DNA replication. The temporal control of DNA replication is thought to be important for gene expression and chromosome organization (11, 35, 49). There is strong evidence that chromatin assembled early in S phase is hyperacetylated, and therefore more permissive for transcription, while chromatin assembled in late S phase is hypoacetylated (47). In this study, we explore the possibility that replication timing plays a role in the genome stability of latent EBV episomes. We investigate the effect of HU treatment on the replication timing of EBV genomes and explore the possibility that this is linked to the cell cycle deacetylation-acetylation of histone H3 at DS. Our findings suggest that histone deacetylation is linked to a delay in OriP replication timing and that this delay enhances the genome stability of latent EBV.

MATERIALS AND METHODS

Cell culture.

EBV-positive Burkitt lymphoma cell lines (Raji and MutuI) were maintained in RPMI medium supplemented with 10% fetal bovine serum, glutamine, penicillin, and streptomycin sulfate (Cellgro). D98/HR1 cells (EBV-positive adherent cells) and HeLa cells (EBV-negative adherent cells) were maintained in Dulbecco modified Eagle medium supplemented with 10% fetal bovine serum, glutamine, penicillin, and streptomycin sulfate (Cellgro). The A39 mini-EBV lymphoblastoid cell line was cultured as described previously (33, 34). HU and valproic acid (VPA) (Sigma) were used at the concentrations indicated in each figure legend. Trichostatin A (TSA) was used at 100 ng/ml for 4 h, and sodium butyrate (NaB) was used at 1 mM for 4 h. HeLa cells were transfected with Lipofectamine 2000 (Invitrogen, Inc.) and selected with hygromycin (100 μg/ml) for 10 days for generating stable pools containing OriP plasmids.

Antibodies.

The following rabbit polyclonal antibodies were used: rabbit polyclonal anti-EBNA1 and TRF2 were raised against a recombinant full-length EBNA1 and TRF2, and rabbit immunoglobulin G (IgG) (Santa Cruz), polyclonal ORC2 (BD Pharmingen), polyclonal acetyl H3 (AcH3) (Upstate), polyclonal MCM3 (Abcam), mSin3 (Santa Cruz Biotech), MTA (Santa Cruz Biotech), and monoclonal RRM2 (Abnova) were used according to the manufacturers' suggestions.

Plasmids.

OriP wild-type (wt) and mutant (nm⁻) plasmids have been described previously and consist of OriP sequences, EBNA1, enhanced green fluorescent protein, and hygromycin genes as a pREP10 (Invitrogen) derivative (8). Short hairpin TRF2 (shTRF2) has been described previously (7).

Replication Timing Assays.

Cells were incubated with 50 μM bromodeoxyuridine (BrdU) for 30 min. Labeled cells were then fixed in 70% ethanol and resuspended in propidium iodide (PI) staining buffer for 30 min. Stained cells were sorted at 50,000 cells per fraction for six fractions (G₁, S1 to S4, and G₂/M) by fluorescence-activated cell sorting (FACS). Three hundred microliters of lysis buffer I (50 mM Tris-HCl [pH 8.0], 1 M NaCl, 10 mM EDTA, 0.5% sodium dodecyl sulfate [SDS]; 0.2 mg/ml proteinase K, 10 μl heat-denatured sonicated single-stranded DNA [10 mg/ml]) was added to the collected cells and incubated at 50°C for 2 h. DNA was extracted with phenol-chloroform, precipitated with alcohol, and then dissolved in 500 μl 1× Tris-EDTA buffer. Sonication was used to generate fragments of ∼0.25 to 2 kb (average, 700 bp). DNA was then heat denatured at 95°C for 5 min and cooled down on ice. Fifty microliters was kept as input. Fifty microliters of 10× immunoprecipitation (IP) buffer (100 mM NaPO₄ [pH 7.0], 1.4 M NaCl, 0.5% Triton X-100) was added to the DNA solution to make a final 1× solution; 4 μl anti-BrdU (stock, 25 μg/ml; BD Pharmingen) was then incubated with the solution for 60 min at room temperature with rotation, and then 3.5 μg (10 μl) of rabbit anti-mouse IgG antibody (Sigma) was added to each tube and rotated for 30 min at room temperature. The precipitated DNA was centrifuged for 5 min at 14,000 rpm and washed twice in 750 μl 1× IP buffer, and the pellet was resuspended in 200 μl lysis buffer II (10 mM EDTA, 50 mM Tris-HCl [pH 8.8], 0.5% SDS, 0.25 mg/ml proteinase K). DNA was incubated at 37°C overnight, and then another 100 μl lysis buffer II was added and left for 1 h at 50°C. DNA again was phenol-chloroform extracted twice and precipitated with ethanol in the presence of glycogen. Real-time PCR was used to compare the BrdU incorporation on different regions with specific primers. Real-time PCR analysis of DNA was quantified using the standard curve method on an ABI 7000 thermocycler and normalized to the input of each fraction. The data were then normalized to the total signal (from G₁, to S1 to S4, to G₂/M, and set as 100%). At least three independent chromatin IPs (ChIPs) were performed for each data point. The error bars in the figures represent standard deviations from three real-time PCRs from the three ChIP experiments.

Primers used for real-timer PCR were as follows: for DS, ATGTAAATAAAACCGTGACAGCTCAT (forward) and TTACCCAACGGGAAGCATATG (reverse); for actin, GCCATGGTTGTGCCATTACA (forward) and GGCCAGGTTCTCTTTTTATTTCTG (reverse); for lamin B2, GTGCACAGCGCCAGGTTA (forward) and GTGCACAGCGCCAGGTTA (reverse); and for beta-globin, AGGAGAAGACTGCTGTCAATGC (forward) and TCCTTGAGCCTCTCTTATAACCTTGA (reverse).

Genome maintenance assay.

Cells (approximately 1 × 10⁶ cells per sample) were collected and resuspended in 100 μl SDS lysis buffer (1% SDS, 10 mM EDTA, 50 mM Tris, pH 8.0.). After brief sonication, IP dilution buffer (0.01% SDS, 1.1% Triton X-100, 1.2 mM EDTA, 16.7 mM Tris [pH 8.0], 167 mM NaCl) was added to 1 ml and then incubated with proteinase K for 2 to 3 h at 50°C. Three hundred microliters was removed and subjected to phenol-chloroform extraction and ethanol precipitation. Precipitated DNA was then assayed by real-time PCR using primers for the DS region of EBV and normalized by the cellular DNA signal at the actin gene locus.

HDAC activity assay.

A histone deacetylase (HDAC) activity assay kit (BioVision, CA) was used to measure the HDAC activity. Briefly, the HDAC fluorometric substrate and assay buffer were added to cell extract or IP samples in a 96-well format and incubated at 37°C for 30 min. The reaction was stopped by adding lysine developer, and the mixture was incubated for another 30 min at 37°C. A fluorescence plate reader with excitation at 355 nm and emission at 460 nm was used to quantify HDAC activity. The Bradford protein assay (Bio-Rad, CA) was used to measure the concentration of protein from input. The HDAC activity is presented as the relative fluorescence units per microgram input protein.

Additional methods.

ChIP assays were performed as described previously (48). Centrifugal elutriation was performed as described previously (10, 48). Cell cycle progression and length of S phase were measured by BrdU pulse-labeling combined with PI staining as described previously (3). RRM2 small interfering RNA was obtained from Dharmacon as a Smartpool and was transfected with Dharmacon transfection reagent according the manufacturer's protocol. TRF2 shRNA has been described previously (7).

RESULTS

Delay in replication timing of EBV.

Previous studies of the replication timing of EBV have been controversial. An earlier study reported that EBV replicates in early S phase (28), while a subsequent study using a different method found that EBV episomes replicate in late S phase (4). To help resolve this question, we assayed EBV genome replication timing by two different methods using Raji cells, an EBV-positive Burkitt lymphoma cell line that lacks functional viral DNA polymerase and replicates exclusively by episomal replication (Fig. 1). For the first method, Raji cells were pulse-labeled with BrdU for 30 min and then subjected to flow cytometry using FACS to fractionate cells based on their stage in the cell cycle. DNA synthesis (BrdU incorporation) was measured by BrdU-specific IP and real-time PCR analysis. Using this method, we compared EBV DNA replication to that of two cellular replicons with known temporal regulation of replication. The lamin B2 origin has been characterized as one of the earliest-firing origins, while the beta-globin locus origin is known to replicate late in the cell cycle. As expected, we found that lamin B2 replicates in the early fractions (S1), while beta-globin replicates in the late fractions (S4 and G₂/M). In contrast, the EBV DS region was found to replicate after lamin B2 (primarily in fraction S3) but not as late as beta-globin (Fig. 1B). To determine if other regions of the EBV genome replicated at the same time as DS, we compared six different EBV genome locations spanning essentially the entire 170 kb of EBV. We found that all regions tested replicated in the same S phase fraction as DS (Fig. 1C). This suggests that the EBV genome replicates within a single stage of the S phase, which is consistent with other reports that estimate that EBV latent genome replication is completed within a 20- to 40-min window in S phase (29).

FIG. 1.

EBV genomes replicate in mid to late S phase in Raji cells. (A) FACS analysis of PI-stained Raji cells was used to sort cells based on their position in the cell cycle. (B) The replication timing of DS, lamin B2, and beta-globin loci was determined by real-time PCR quantification of BrdU-containing DNA after pulse-labeling. Percent BrdU represents the relative distribution of DNA in each fraction of the cell cycle. (C) Different regions of EBV were amplified by real-time PCR, and results are presented as percent BrdU for each locus. (D) Raji cells were fractionated by centrifugal elutriation and then analyzed by FACS after PI staining. (E) Raji cells fractionated by centrifugal elutriation as shown in panel D were then assayed for replication timing of DS, lamin B2, and beta-globin loci. Error bars indicate standard deviations.

The S phase timing of replication was further evaluated by a second method using centrifugal elutriation, which separates cells based on their morphology. Centrifugal elutriation successfully separated Raji cells according to their stage of the cell cycle as determined by FACS analysis after PI staining (Fig. 1D). Cells were pulse-labeled with BrdU, fractionated by centrifugal elutriation, and then analyzed by BrdU IP and real-time PCR analysis for EBV DS region or cellular lamin B2 and beta-globin loci. As with the FACS method, centrifugal elutriation revealed that DS replicates in mid to late S phase, while lamin B2 replicates in the G₁ and early S fractions and beta-globin replicates in late S and G₂ fractions. The S phase delay in replication timing of EBV genomes was also observed in several different latently infected B-cell lines, including the MutuI Burkitt lymphoma cell line and the lymphoblastoid cell line containing the mini-EBV genome (A39) (see Fig. S1 in the supplemental material). The S phase delay in replication timing profile was further confirmed using three-color TaqMan probes to simultaneously measure DS and reference lamin B2 or beta-globin origins in the same PCRs (see Fig. S2 in the supplemental material). While the precise stage of S phase may vary among cell types and experimental designs, the delay in replication timing of EBV relative to lamin B remains invariant in all cell types and experimental protocols.

HU treatment advances the replication timing of EBV.

HU is the only pharmacological agent that has been reported to reduce and eliminate EBV genomes from latently infected cells. The mechanism of HU-induced episome loss is not known. We therefore explored whether HU treatment had any effect on the replication timing of EBV episomes (Fig. 2). First, we demonstrated that treatment of Raji cells with HU reduced EBV episome copy number using real-time PCR analysis of EBV DNA relative to a cellular DNA region in the actin open reading frame. We found that treatment with 50 μM HU for 6 days led to a ∼3-fold reduction in EBV genome copy number (Fig. 2A). Higher concentrations of HU (>200 μM) caused cell cycle arrest, and continuous growth in 50 μM HU reduced Raji cell viability (data not shown). We therefore assayed the replication timing of EBV relative to lamin B2 and beta-globin in untreated Raji cells (control) (Fig. 2B) or in cells treated with 50 μM HU for 6 days (Fig. 2C) or 100 μM HU for 3 days (Fig. 2D). We found that treatment with 50 μM and 100 μM HU caused a dramatic shift in EBV replication from the S3-S4 fraction in controls to the S1-S2 fraction in treated samples. Lamin B2 replication remained early in HU treatment, while beta-globin replication was slightly advanced in S phase with 50 μM and substantially advanced by 100 μM HU treatment. Similar effects of HU on OriP replication timing were observed in latently infected Mutu I cells (see Fig. S3 in the supplemental material). These findings indicate that HU treatment accelerates the replication timing of EBV, as well as a late-firing cellular replicon at the beta-globin locus.

FIG. 2.

HU advances replication timing of EBV. (A) Raji cells were treated with 50 μM HU or control for 6 days and then assayed for EBV genome copy number. EBV copy number was assayed by real-time PCR analysis of DS relative to cellular actin. (B to D) Replication timing of Raji cell DS, lamin B2, or beta-globin was analyzed by FACS under control conditions (B) or with 50 μM HU for 6 days (C) or 100 μM HU for 3 days. Error bars indicate standard deviations.

HU alters histone modification and TRF2 binding at OriP.

Earlier studies of OriP revealed that TRF2 and histones were closely positioned to the EBNA1 binding sites in DS (8, 48). Furthermore, histone H3 positioned at DS was deacetylated in the early S phase. We therefore examined whether HU treatment altered the TRF2 binding or histone H3 acetylation at DS (Fig. 3). Raji cells were treated with 50 μM HU for 5 days and then assayed by ChIP with antibodies specific for EBNA1, AcH3, TRF2, or the cellular replication factor MCM3. We found that HU treatment had a weak effect on EBNA1 binding, but increased AcH3, and decreased TRF2 binding by ∼2-fold (Fig. 3A). MCM3 binding did not change, but its cell cycle profile was not examined in these studies. The specificity for DS was demonstrated by comparing the binding at control region OriLyt, which is not an active origin during EBV latent infection in Raji cells. These experiments indicate that HU treatment leads to a remodeling of the DS region of OriP. Specifically, HU induced H3 acetylation and reduced TRF2 binding.

FIG. 3.

HU induces H3 acetylation and TRF2 dissociation from DS. Raji cells were treated with 0 μM (control) or 50 μM HU for 5 days and then assayed by ChIP with antibodies to EBNA1, AcH3, TRF2, MCM3, or control IgG. ChIP DNA was assayed by real-time PCR with primers specific for DS (A) or for OriLyt (B). Error bars indicate standard deviations.

TRF2 site deletion or shRNA depletion of TRF2 advances replication timing of OriP.

Previous studies have shown that shRNA depletion of TRF2 or substitution mutations in TRF2 binding sites in OriP inhibit replication and plasmid maintenance (7, 8). We next tested whether shRNA depletion of TRF2 altered the replication timing of OriP (Fig. 4A and B). For these experiments, we utilized the EBV-positive adherent cell line D98/HR1 because of its high transfection efficiency relative to Raji cells. Transfection of shTRF2 effectively reduced TRF2 protein without affecting the control protein SNF2h, as indicated by Western blotting (Fig. 4C). As in Raji cells, DS replicated in the later S phase fractions (S4) relative to lamin B2 (S1 to S2) in control transfected D98/HR1 cells. When TRF2 was depleted by TRF2-specific shRNA, DS replication advanced to the S3 fraction, while no change was observed in lamin B2 (Fig. 4A and B). To better evaluate the role of TRF binding sites in regulating OriP replication timing, we assayed the replication timing of plasmids containing wt OriP or a mutant OriP with substitution mutations in all three nanomer binding sites for TRF2 (OriP nm⁻) (Fig. 4D and E). Plasmids were transfected into HeLa cells and selected with hygromycin for 10 days. Replication timing was assayed with BrdU pulse-labeling and cell sorting, as in Fig. 1A. We found that plasmids in HeLa cells containing OriP wt replicated in S3 to S4, while plasmids with OriP nm⁻ replicated primarily in the S2 fraction. These results suggest that TRF2 binding sites in DS delay the replication timing of OriP-containing plasmids. Furthermore, these results suggest that the delay in replication timing correlates with the enhanced plasmid stability of OriP wt relative to OriP nm⁻.

FIG. 4.

TRF2 binding delays replication timing of OriP. (A and B) Replication timing was analyzed by the FACS method in EBV-positive D98/HR1 cells after transfection with control or shTRF2 expression plasmids. Replication timing was determined for DS (A) or lamin B2 (B) loci. (C) Western blot of D98/HR1 cells transfected with the control or shTRF2 plasmids used for the experiments shown in panels A and B. Immunoblots with anti-TRF2 (top panel) or anti-SNF2h (lower panel) are shown. (D) Replication timing of DS region on OriP wt or OriP nm⁻ plasmids in HeLa cells after 10 day of hygromycin (100 μg/ml) selection. (E) Same as in panel D, but replication timing was determined for the lamin B2 locus. Error bars indicate standard deviations.

VPA alters OriP protein binding and histone modification.

The other major effect of HU was an increase in histone H3 acetylation. We therefore tested the effect of several HDACs on the replication timing of EBV genomes in Raji cells. We found that TSA and NaB had a potent effect on the replication timing but were highly toxic to Raji cells (see Fig. S4 in the supplemental material). As an alternative, we used VPA, which inhibits class I HDAC and is less toxic to Raji cells. VPA was assayed for its effect on replication timing in Raji cells (Fig. 5A). VPA treatment (0.5 mM for 6 days) advanced the EBV replication from S4 to mostly S1 and S2. VPA had little effect on lamin B2 replication but caused an acceleration of the late-firing beta-globin from the G₂/M fraction to S2. This indicates that VPA treatment advances replication timing of EBV and a cellular late-firing replicon to early stages of the cell cycle.

FIG. 5.

VPA advances EBV replication timing and reduces viral genome copy number. (A) Raji cells treated with 0.5 mM VPA or control for 6 days and then assayed for replication timing at DS, lamin B2, and beta-globin loci. (B) ChIP assays with antibodies to EBNA1, AcH3, TRF2, ORC2, or control IgG by real-time PCR for binding at DS (top panel) or control OriLyt (lower panel) regions of EBV. (C) EBV genome copy number was assayed in Raji cells treated with 0.5 mM VPA for 3 or 6 days or with 1.5 mM VPA for 3 or 6 days or in control untreated cells, as indicated. (D) Southern blot of Raji cell DNA after mock treatment (lane 1) or treatment with 0.5 mM VPA (lane 2), 1.5 mM VPA (lane 3), or 0.05 mM HU (lane 4) for 6 days. DNA was cut with BamHI and hybridized with a probe to EBV OriLyt (top panel) or Alu repeats (lower panel). M, 1-kb molecular weight ladder. Error bars indicate standard deviations.

The effect of VPA treatment of OriP binding proteins was examined using ChIP assay. We found that treatment with 0.5 mM VPA for 6 days led to a ∼2-fold increase in AcH3 and a corresponding ∼2-fold decrease in TRF2 binding to the DS region of OriP. No significant changes were observed in EBNA1 or ORC2 binding, and no significant binding of any of these factors were observed at control region OriLyt (Fig. 5B, lower panel). These findings indicate that VPA affects protein binding and histone modification at OriP similarly to HU treatment (Fig. 3) and TRF2 depletion (Fig. 4).

VPA causes EBV episome loss.

Since VPA produced effects similar to those of HU on EBV replication timing and protein interactions, we tested whether VPA caused the loss of EBV genome copy number. Raji cells were incubated with 0.5 or 1.5 mM VPA for 3 or 6 days and then assayed by real-time PCR for EBV DNA relative to cellular actin DNA (Fig. 5C). We found that 1.5 mM VPA caused a ∼2- to 3-fold reduction in EBV copy number at 3 and 6 days. Similarly, we found that 0.5 mM VPA caused ∼3-fold reduction at 6 days, although there was little effect detected after 3 days. The EBV genome copy number reduction caused by VPA and HU treatment of Raji cells was also observed by Southern blot analysis (Fig. 5D). An OriLyt probe was used to detect the EBV genome (upper panel) and an Alu probe was used as a loading control for total cellular DNA (lower panel). These findings indicate that VPA eliminates EBV genome copy number in a dose- and time-dependent manner.

HDAC1 and -2 associate with TRF2 and are inhibited by HU.

HU treatment caused a significant increase in histone H3 acetylation at OriP (Fig. 3A). To determine if H3 acetylation at OriP is partially regulated by TRF2 interaction with HDAC complexes, we assayed TRF2 IPs for the presence of class I HDACs (Fig. 6A). We found that HDAC2 associates with TRF2 as measured by IP from Raji extracts (Fig. 6A, top panel). However, treatment with HU or VPA did not significantly reduce HDAC2 interaction with TRF2. TRF2 did not associate with mSin2 or MTA, two abundant components of well-characterized class I HDAC-containing complexes (Fig. 6A, lower panels). The effect of HU and VPA on HDAC activity was measured by a fluorescence-based HDAC assay (Fig. 6B). We found that both HU and VPA inhibited total cellular HDAC activity by ∼20 and ∼50% and inhibited to a lesser extent the HDAC activity associated TRF2 IPs (Fig. 6B). The association of HDAC1 and HDAC2 with OriP was measured by ChIP assay in Raji cells after HU, VPA, or mock treatment (Fig. 6C). Both HU and VPA caused HDAC1 and HDAC2 dissociation from OriP. This dissociation correlates with the partial loss of TRF2 binding at OriP (Fig. 2A). Taken together, these results suggest that HU, as well as VPA, disrupts TRF2 and HDAC association with OriP and that this dissociation correlates with advanced replication timing (Fig. 6D).

FIG. 6.

HU and VPA alter HDAC activity and association with OriP. (A) Raji cells treated with HU (0.05 mM) or VPA (0.5 mM) or mock treated for 6 days were subject to IP with control IgG or TRF2 antibody. Immunoprecipitates were then assayed by Western blotting with antibodies specific for HDAC2, HDAC1, TRF2, mSIN3, and MTA, as indicated. (B) HDAC activity in Raji cell extracts and IPs was measured by fluorescence assay (BioVision, Inc). Raji cells were treated with HU or VPA or mock treated as described for panel A. Input extracts were normalized by Bradford assay using bovine serum albumin as a standard. (C) ChIP assays with antibodies to HDAC1, HDAC2, or control IgG were used to measure binding to OriP (top panel) or control OriLyt (lower panel) in Raji cells after 6 days of HU, VPA, or mock treatment, as described for panel A. (D) Schematic model of TRF2-HDAC regulation of OriP replication in early G₁/S and its subsequent derepression of replication in mid/late S phase.

DISCUSSION

EBV latent infection has been estimated to contribute to ∼1% of all human cancers (31). The mechanisms that control latent cycle replication and episome stability are therefore of great interest for potential targets of therapeutic intervention. HU treatment has been used in a clinical setting to eradicate latent viral genomes, but its mechanism of action has remained poorly understood. In this work, we found that HU can alter a programmed delay in the replication timing of EBV genomes and OriP-containing episomes (Fig. 1 and 2). We found that HU treatment caused the dissociation of TRF2 from the DS region of OriP and altered the normal cell cycle pattern of histone deacetylation that occurs in early S phase at DS (Fig. 3). The importance of histone deacetylation was further substantiated by experiments with HDAC inhibitors (e.g., VPA), which prevented the early S phase histone deacetylation at OriP, advanced replication timing, and destabilized the latent EBV episome (Fig. 5). HU inhibited HDAC activity and eliminated TRF2-HDAC binding at OriP (Fig. 6). These findings suggest that a TRF2-HDAC complex mediates a replication timing delay important for EBV genome stability.

The latent episome of EBV can replicate for many generations without significant loss of copy number. Metaphase chromosome tethering correlates well with plasmid stability for EBV, Kaposi's sarcoma-associated herpesvirus, and human papillomavirus, but other factors must also contribute to this process (25). Efficient plasmid maintenance depends on unknown epigenetic changes in the viral chromosome and the coupling of DNA synthesis to plasmid segregation (24, 27). Therefore, it is likely that DNA and chromatin structures formed at OriP contribute to the plasmid maintenance process. Our previous studies of OriP found that DS is flanked by positioned nucleosomes that are subject to cell cycle changes in histone modifications (48). Genetic studies have demonstrated that the TRF2 binding sites within DS also contribute significantly to episome stability (7, 8, 26). More recent studies found that DNA recombination proteins and replication pausing at DS is also important for episome stability (10). The new findings from this study, which indicated that replication timing is coupled to EBV episome stability, further support a model that complex epigenetic regulation and checkpoint mechanisms contribute to EBV genome maintenance during latent infection.

The effect of HU on EBV episome maintenance has been well established, but its mechanism of action has not been elucidated (6, 18, 40). In this work, we found that HU treatment altered the replication timing and chromatin organization of OriP. Several mechanisms have been reported to regulate replication timing. In Saccharomyces cerevisiae, deletion of the HDAC Rpd3 caused late-firing origins to fire synchronously with early replication origins (42). Control of replication timing has been linked to the S phase checkpoint protein RAD53, indicating that replication timing was important for genome stability during S phase (12, 37). In higher eukaryotes, early replication timing correlates with histone acetylation, and tethering of HDAC2 delays replication timing (16). Replication timing is also controlled by the intra-S phase checkpoint kinases ATM and ATR, which regulate origin firing in response to replication stress (36). Moreover, dormant replication origins can be activated during replication stress induced by low levels of HU (15). This suggests that replication timing is regulated by mechanisms that link S phase checkpoints with histone acetylation. In concordance with these models, we found that HU treatment prevented the normal histone H3 deacetylation that occurs at the DS region of OriP in early S phase (Fig. 2 and 3). We propose that OriP is subject to a TRF2-HDAC checkpoint that regulates replication initiation at DS by histone modification. Histone deacetylation at DS may also explain why replication does not always initiate from OriP (29, 30).

HU is a known inhibitor of ribonucleotide reductase, which is required to generate deoxynucleoside triphosphates for DNA replication during S phase. At high concentrations (∼1 mM), HU can provoke significant replication defects and cause cell cycle arrest. However, at the low concentrations used in our experiments (∼50 μM), HU did not cause a cell cycle arrest (see Fig. S3 in the supplemental material) and did not elicit an S phase checkpoint response (data not shown). Nevertheless, low-level HU treatment alters the replication timing of OriP and destabilizes EBV genome maintenance. Our studies indicate that HU treatment operates through ribonucleotide reductase inhibition, since small interfering RNA depletion of the RRM2 subunit phenocopies HU effects on EBV replication timing, histone deacetylation at DS, and genome stability (see Fig. S7 in the supplemental material). HU treatment may affect EBV replication timing by altering the rate at which S phase progresses. However, low levels of HU had only modest effects on S phase duration and the percentage of cells in S phase (see Fig. S5 in the supplemental material). In contrast, HU treatment caused a significant loss of TRF2 binding and a corresponding increase in histone H3 acetylation at OriP (Fig. 3). We suggest that low-level HU mimics the nucleotide depletion observed in mid-S phase and that nucleotide depletion evokes a signal that reduces TRF2 binding, inhibits HDAC activity, and promotes replication initiation at OriP (Fig. 6D). Under normal conditions (no HU treatment), the TRF2-HDAC complex may prevent OriP from initiating among the first wave of cellular replicons. This would protect EBV genomes from replicating in cellular conditions that are mutagenic or suboptimal for the completion of DNA replication. The delay in replication timing may also favor the association of EBV genomes with chromatin and segregation machinery necessary for episome stability.

Epigenetic events have been implicated in the establishment of a stable EBV episome. Our previous studies indicated that nucleosome positioning and histone deacetylation at the DS region of OriP were important for plasmid stability. We now show that disruption of the normal cell cycle histone deacetylation pattern can cause genome instability. HDAC inhibitors (e.g., VPA, TSA, and NaB) promoted the loss of EBV genomes, similar to HU. Remarkably, HDAC inhibitors are well-established initiators of the EBV lytic cycle and induce apoptosis in cells where lytic replication is blocked (14, 20). In cells where lytic cycle replication is blocked, like for the replication defective genomes of Raji cells, HDAC inhibitors lead to a loss of EBV episomes. Our data indicates that HDAC inhibitors have a primary effect on chromatin organization at OriP and a fundamental role in maintaining the stable viral episome. This raises the possibility that lytic replication is a programmed response to cellular conditions that destabilize episome maintenance. If so, lytic induction therapies, which combine reactivation signals with inhibitors of viral lytic replication, may have the added benefit of eliminating nonreactivated latent virus genomes (13).

In conclusion, we have implicated the TRF2-HDAC complex in the regulated delay of replication timing of OriP. We provide evidence that TRF2-HDAC is an important target of HU, which is the only clinically tested pharmacological agent that eliminates EBV episomes from latently infected cells. We found that HU treatment causes a dissociation of TRF2-HDAC and an inhibition of HDAC activity. This corresponds to the loss of the programmed delay in replication timing and a decrease in episome stability. We propose that TRF2-HDAC provides OriP with a replication checkpoint that enhances viral genome stability. These finding provide new insights into the mechanism through which HU eliminates EBV genomes from infected cells, and they may provide new opportunities to disrupt latent EBV infection in human tumors.