Epigenetic Deregulation of the LMP1/LMP2 Locus of Epstein-Barr Virus by Mutation of a Single CTCF-Cohesin Binding Site

Abstract

The chromatin regulatory factors CTCF and cohesin have been implicated in the coordinated control of multiple gene loci in Epstein-Barr virus (EBV) latency. We have found that CTCF and cohesin are highly enriched at the convergent and partially overlapping transcripts for the LMP1 and LMP2A genes, but it is not yet known how CTCF and cohesin may coordinately regulate these transcripts. We now show that genetic disruption of this CTCF binding site (EBVΔCTCF166) leads to a deregulation of LMP1, LMP2A, and LMP2B transcription in EBV-immortalized B lymphocytes. EBVΔCTCF166 virus-immortalized primary B lymphocytes showed a decrease in LMP1 and LMP2A mRNA and a corresponding increase in LMP2B mRNA. The reduction of LMP1 and LMP2A correlated with a loss of euchromatic histone modification H3K9ac and a corresponding increase in heterochromatic histone modification H3K9me3 at the LMP2A promoter region in EBVΔCTCF166. Chromosome conformation capture (3C) revealed that DNA loop formation with the origin of plasmid replication (OriP) enhancer was eliminated in EBVΔCTCF166. We also observed that the EBV episome copy number was elevated in EBVΔCTCF166 and that this was not due to increased lytic cycle activity. These findings suggest that a single CTCF binding site controls LMP2A and LMP1 promoter selection, chromatin boundary function, DNA loop formation, and episome copy number control during EBV latency.

INTRODUCTION

Epstein-Barr virus (EBV) is a human herpesvirus that infects more than 90% of the adult population worldwide and establishes a long-term latent infection in B lymphocytes (1, 2). The latent infection is associated with several lymphoid and epithelial cell malignancies, especially the endemic forms of Burkitt's lymphoma and nasopharyngeal carcinoma (3, 4). During latent infection, EBV persists as multicopy minichromosomes (referred to as episomes) that express a highly restricted set of viral genes (5, 6). Latent cycle gene expression can vary depending on cell type, developmental stage, and environmental conditions (7). EBV latency genes are also coordinately regulated with each other and subject to complex feed-forward and feed-back mechanisms involving multiple virus- and cell-encoded factors (8–10). Viral latency products can drive host cell proliferation and survival, as well as drive B-cell immortalization and carcinogenesis. The mechanisms that control the coordinate regulation of latency gene products are therefore important for understanding EBV pathogenesis.

The latency membrane proteins LMP1, LMP2A, and LMP2B play essential roles in EBV latency and tumorigenesis (11–13). LMP2A and LMP1 can function to coordinately mimic B-cell receptor and CD40 coreceptor signaling in latently infected B cells (14). The LMP1 and LMP2A mRNA are generated from a common viral locus with convergent and overlapping primary transcripts (15). LMP2B is a shorter isoform of LMP2A, and its transcript initiates 5′ of the LMP1 transcription start site. Both LMP2A and LMP2B transcripts extend across the viral terminal repeats (TRs) to terminate in a common 3′ exon. The promoters for LMP2B and LMP1 are partially overlapping, and it is not know how divergent transcription from this locus is regulated during different forms of EBV latency (16). LMP1, LMP2A, and LMP2B transcription can be activated by viral encoded transcriptional regulators EBNA1 and EBNA2, in conjunction with host-cell specific factors, such as RBP-jK and Pu.1 (17–20). How LMP1, LMP2A, and LMP2B transcription is regulated to avoid RNA polymerase clashing and how these transcripts are further coordinated with latency gene programming is not yet known.

The cellular mechanisms that control coordinated gene expression are only partly understood. Chromatin organizing factors are thought to play a central role in regulating complex gene expression programs (21–23). The CCCTC-binding factor CTCF has been implicated in mediating long-distance DNA interactions and forming gene hubs important for gene regulation (24–28). CTCF is a nuclear DNA-binding protein that contains 11 zinc fingers and is well conserved among higher eukaryotes (29–31). CTCF is involved in different functions, including chromatin boundary formation, DNA loop formation, transcriptional activation and repression, and promoter-enhancer blocking activity (26, 27). CTCF colocalizes with cohesins at ∼15% of its binding sites on the human genome (22, 32, 33). Cohesin is a multiprotein complex that can form a ring-like structure capable of stabilizing interactions between DNA molecules, which is important for both sister chromatid cohesion in mitosis and promoter-enhancer interactions in transcriptional regulation (34–37). In EBV, CTCF binds at several key regulatory regions, many of which are co-occupied by cohesin (38–40). In particular, CTCF and cohesin can bind to the LMP1 and LMP2A control region at a position within the first intron of LMP2A and the 3′ untranslated region (3′UTR) of LMP1 (38) (Fig. 1A to C). Whether CTCF-cohesin binding at this position regulates LMP1 or LMP2A gene expression is not yet known.

FIG 1.

Generation of EBV bacmids that disrupt CTCF and cohesin binding to the LMP1/LMP2 control region. (A) ChIP-Seq peaks for CTCF, Rad21, SMC3, and IgG on the EBV genomes from B95.8 LCLs. (B) Close-up of ChIP-Seq peaks at position 166500 relative to the LMP2A first intron (hatched) and LMP1 3′UTR (black). (C) Sequence of a CTCF consensus binding site and its deletion used to generate EBV ΔCTCF bacmids. (D) EMSA results showing purified recombinant CTCF binding to a 45-bp fragment containing the CTCF consensus site at position 166500 (wt), a 45-bp fragment containing the 18-bp deletion of the CTCF consensus at position 166500 (ΔCTCF), a positive control with consensus CTCF (Ctrl+), or a negative control with no CTCF binding site (Ctrl−). (E) Ethidium bromide staining of an EcoRI restriction digest of EBV bacmids for EBV wt, EBV Kan-CTCF, EBVΔCTCF.1, or EBVΔCTCF.2 resolved on a 0.7% agarose gel by running for 16 to 20 h at 40 V. The marker is a 1-kb ladder. (F) ChIP assay in LCLs generated with EBV wt, EBVΔCTCF.1, or EBVΔCTCF.2 bacmid virus. ChIP with antibodies for CTCF, SMC1, SMC3, Rad21, or IgG control were assayed with primers that amplify EBV CTCF166 region. Error bars indicate standard deviations from the mean (SD) for three independent PCRs.

We investigated the function of the CTCF-cohesin binding site located within the LMP1/LMP2A control region. We have previously found that this CTCF-cohesin site formed a DNA-loop interaction with OriP (38). We used chromosome conformation capture (3C) and shRNA depletion studies to show that cohesin is important for DNA long-range interactions and that this protein complex also contributes to LMP1 and LMP2A transcription efficiency. Here we used bacmid recombineering and genetic disruption to further analyze the function of the CTCF-cohesin binding site within the LMP1/LMP2A control region. We show that this CTCF binding site is required for cohesin binding and for DNA loop formation between LMP1/LMP2A and OriP. Furthermore, we show that CTCF is required for the proper transcriptional regulation of LMP1, LMP2A and LMP2B and that the loss of CTCF results in the switch from euchromatic to heterochromatic epigenetic marks at the LMP2A and LMP1 promoter regions.

MATERIALS AND METHODS

Cells.

293T cells were cultured in Dulbecco modified Eagle medium with 10% fetal bovine serum and antibiotic in a 5% CO₂ incubator at 37°C. Lymphoblastoid cell lines (LCLs) and EBV-positive Burkitt's lymphoma cells were cultured in suspension in RPMI 1640 medium supplemented with 15% fetal bovine serum and antibiotics in a 5% CO₂ incubator at 37°C. 293T cells and LCLs carrying EBV bacmids were grown in medium containing 200 μg of hygromycin B/ml and were further selected for green fluorescent protein (GFP) expression. EBV bacmid virus was prepared from 293T cells and quantified as described previously (41, 42). LCLs containing EBV wt or EBVΔCTCF166 bacmids were generated by infecting B cells isolated from primary peripheral blood mononuclear cells with bacmid virus and selecting for hygromycin resistance. All phenotypic assays were performed with stable LCLs expanded for at least 8 weeks postinfection. EBV lytic gene expression was induced with 20 ng of TPA (12-O-tetradecanoylphorbol-13-acetate) and 3 mM sodium butyrate (NaB) for 72 h.

Construction of recombinant EBV virus with a CTCF166 binding site deletion.

EBV bacmid mutations were generated using the I-SceI two-step red-recombinase method (43). Briefly, primers for amplification of the kanamycin (Kan) gene from the pEPKanS plasmid used the forward primer 5′-CAACAAAGAACTTTGACCTGTTGTCCCTGAGATGTGAATTGGGATGGTGTGGATAACACCTTATTATTGATGTGACTTGTGATGCAATAAATAAAAGTACAGATAGATCAACCAATTAACCAATTCTGATTA-3′ and the reverse primer 5′-ATCTATCTGTACTTTTATTTATTGCATCACAAGTCACATCAATAATAAGGAGGATGACGACGATAAGTAGGG-3′. The PCR product consisting of Kan gene with one I-SceI site, flanked by 50 bp downstream and upstream of EBV sequence surrounding the designed CTCF deletion sequence at EBV position 166520 was gel purified by using a gel extraction kit (Qiagen). The ∼1.2K-bp PCR product containing the I-SceI-ΔCTCF-Kan product was electroporated into electrocompetent DY380 cells containing EBV BAC wild-type (wt) genome. EBV BAC wt with the correct Kan insertion (EBV BAC-Kan) was recovered, characterized by restriction digestion and DNA sequencing, and then transfected into the electrocompetent bacterial cells (GS1783), I-SceI-inducible cells. The Kan gene was then removed from the EBV BAC-Kan genome by homologous recombination wherein positive clones were screened for kanamycin sensitivity. The final CTCF deletion mutant was confirmed by restriction digest and DNA sequencing of the recombination site to demonstrate the integrity of EBV BAC wt, EBV BAC-Kan, and EBV BAC ΔCTCF.

3C assay.

Chromosome conformation capture (3C) assay followed the protocol of Hagege et al. (44), with minor modifications as described previously (45). Briefly, ∼10⁷ EBV-positive cells were fixed in 1% formaldehyde for 30 min and quenched with 0.125 M glycine. Cells were pelleted and resuspended in 0.5 ml of cold lysis buffer (10 mM Tris-HCl [pH 7.5], 10 mM NaCl, 5 mM MgCl₂, 0.1 mM EGTA) with freshly added 1× Complete protease inhibitors (Roche) and were lysed on ice for 10 min. The nuclei were collected by centrifugation at 500 × g for 10 min at 4°C, resuspended in 0.5 ml of 1.4× MseI buffer (New England BioLabs) containing 0.3% sodium dodecyl sulfate (SDS), and incubated for 1 h at 37°C with shaking at 1,200 rpm. SDS was then trapped by the addition of 2% (final concentration) Triton X-100, followed by incubation for 1 h at 37°C with shaking. Then, 500 U of MseI (50,000 U/ml; New England BioLabs) was added to the nuclei, and the samples were incubated at 37°C overnight with shaking. Next, 10-μl portions of the samples were collected before and after the MseI reaction to evaluate digestion efficiency. The reaction was stopped by the addition of 1.6% SDS (final concentration), followed by incubation at 65°C for 30 min with shaking at 1,200 rpm. The sample was then diluted 10-fold with 1.3× ligation buffer (Roche) plus 1% Triton X-100 and incubated for 1 h at 37°C with shaking at 900 rpm. Then, 100 U of T4 DNA ligase (5 U/μl; Roche) was added to the sample, and the reaction mixture was incubated at 16°C for 4 h, followed by 45 min at room temperature. A 300-μg portion of proteinase K was added to the sample, and the reaction was carried out at 65°C overnight. RNA was removed by adding 300 μg of RNase, and the sample was incubated for 1 h at 37°C. DNA was purified by phenol-chloroform extraction and ethanol precipitation. Purified DNA was then analyzed by quantitative PCR (qPCR). As a control for ligation products EBV wt bacmid was digested with 10 U of MseI overnight and then incubated with 10 U of T4 DNA ligase at 16°C overnight, followed by purification. 3C products were quantified by real-time PCR using the ΔC_T method with EBV bacmid C_T values as a control. The C_T values were normalized for each primer pair by setting the C_T value for 100 ng of the EBV bacmid control random ligation matrix DNA at 1. Primer sequences for quantitative PCR have been described previously and are available upon request (38, 45).

ChIP assay.

Chromatin immunoprecipitation (ChIP) assays were performed according to the protocol provided by Upstate Biotechnology, Inc., with minor modifications as previously described (45, 46). Additional modifications are as follows. Cells were fixed in 1% formaldehyde for 15 min. DNA was sonicated according to the manufacturer's specifications to between 200- and 350-bp DNA fragments on a Fisher Scientific Sonic Dismembrator. Real-time PCR was performed with SYBER Green in an ABI StepOne Plus using 1/100 of the ChIP DNA according to the manufacturer's specified parameters. Chromatin was immunoprecipitated with polyclonal antibodies for CTCF (Millipore), Smc3 (Abcam), Rad21 (Abcam) H3K4me3 (Active Motif), H3K9Ac (Active Motif), H3K9me3 (Active Motif), and IgG (Santa Cruz). The primer sequences for amplification are listed in Table 1.

TABLE 1.

Primers for qPCR-ChIP analysis of LMP1/LMP2 locus

Primer	Sequence (5′–3′)
	Forward	Reverse
LMP2Ap-1	GAGAGGAGCAGGTGCTTATTG	GTGTTGGCACTTCTGTGGATA
LMP2Ap-2	CACCTGAGCGTGGTGAAG	ACGGTGCATGTCACAGTAAG
LMP2Ap-3	GGCAAATGGCGGTGTTATG	CCAGGGCTTGGGAAGTG
LMP1p-4	AGGCAGTTGAGGAAAGAAGG	GGCCTACATCCCAAGAAACA
LMP1p-5	TGTGTGCATGTAAGCGTAGAAA	ACTTGGCCACCGCATTC

Western blot analysis.

A total of 2.5 × 10⁶ cells were collected and sonicated at 75 mA for 4 min (20 s, on/off) using a Fisher Scientific Sonic Dismembrator. Samples were run on a 4 to 20% Mini-Protean TGX precast gel (Bio-Rad) and transferred to an Immobilon-P transfer membrane. Membranes were incubated overnight with primary antibodies to LMP1 (a gift from E. Robertson), Zta (rabbit polyclonal to full-length bacterial protein), EBNA1 (rabbit polyclonal to bacterial protein lacking GA repeats), PCNA (Abcam), H3K9Ac (Active Motif), H3K9me3 (Active Motif), H3K4me3 (Active Motif), H3 (Abcam), CTCF (Abcam) GFP (Santa Cruz), and actin (Santa Cruz). Primary antibodies were used according to the manufacturer's specifications. Membranes were then incubated in the appropriate secondary antibody: mouse anti-rabbit IgG-horseradish peroxidase (IgG-HRP; Santa Cruz) or rabbit anti-mouse IgG-HRP (Thermo Scientific) for 1 h.

Methylated DNA immunoprecipitation (MeDIP).

The DNA was extracted from 2 × 10⁶ cells with a GeneJET genome DNA purification kit (Thermo Scientific) according to the manufacturer's instructions. A total of 440 μg of DNA was sonicated to between 200 and 300 bp. Samples were boiled for 10 min and immediately cooled on ice. To each sample, 50 μl of 10× immunoprecipitation buffer (IP buffer; 1.4 M NaCl, 0.5% Triton X-100) and 5 μg of 5-methylcytosine antibody (Active Motif) were added, followed by incubation overnight at 4°C. The immunocomplexes were precipitated by adding 50 μl of Dynabeads (Life Technologies) and rotated for 2 h at room temperature. Beads were collected with a magnetic rack and washed three times with 500 μl of 1× IP buffer. Beads were incubated at 50°C for 2 h with shaking in 500 μl of proteinase K digestion buffer (50 mM Tris [pH 8.0], 10 mM EDTA, 0.5% SDS, 1 mg of proteinase K/ml). DNA was extracted twice by phenol-chloroform, followed by ethanol precipitation. DNA was analyzed by real-time PCR. The primer sequences are listed in Table 1. A 8.8-μg portion of genomic DNA was used as input material.

RNA extraction and reverse transcription-PCR (RT-PCR).

RNA was extracted from 5 × 10⁶ cells using a Qiagen RNA extraction kit according to the manufacturer's protocol (Qiagen). After extraction, the RNA was incubated with 2 U of DNase I at 37°C for 30 min, followed by the enzyme inactivation at 65°C for 10 min. The RNA was quantified, and 2 μg of RNA was reverse transcribed using Super Script II reverse transcriptase from Invitrogen. Next, 50 ng of cDNA was analyzed by real-time PCR, using the ΔC_T method, with the GFP gene as the internal control. Primer sequences have been previously published and are available upon request (45, 46).

Isolation of EBV bacmid DNA and quantification of its copy number.

The intracellular EBV bacmid DNA copy number was determined by qPCR analysis of purified total genomic DNA and normalized to Namalwa genomic DNA. Briefly, 10⁶ EBV-positive cells were resuspended in SDS lysis buffer (1% SDS, 20 mM NaCl, 4 mM EDTA, 20 mM Tris [pH 8.0]) with proteinase K for at least 6 h at 50°C. The cell lysate was then subjected to phenol-chloroform extraction and ethanol precipitation. Precipitated DNA was then assayed by real-time PCR, using primers for the EBV DS region (5′-ATG TAA ATA AAA CCG TGA CAG CTC AT-3′ and 5′-TTA CCC AAC GGG AAG CAT ATG-3′) and for the cellular actin DNA signal (5′-GCC ATG GTT GTG CCA TTA CA-3′ and 5′-GGC CAG GTT CTC TTT TTA TTT CTG-3′) for normalization.

EMSA.

DNA fragments were labeled using T4 polynucleotide kinase (Roche) in the presence of [γ-³²p]ATP and purified using G-25 spin columns (GE Healthcare). Various amounts of purified CTCF protein (0, 100, 300, and 900 ng) were incubated with an equal amount of purified probe at room temperature for 30 min in a total volume of 25 μl of buffer containing 100 μM ZnSO₄, 10 mM Tris-Cl (pH 8.0), 60 mM KCl, 1 mM EDTA, 10 mM MgCl₂, 0.05 μg of poly(dI-dC)/μl, 0.5 μg of bovine serum albumin/μl, 0.05% NP-40, 35 mM β-mercaptoethanol, and 6% glycerol. The samples were then separated by electrophoresis on a native 5% polyacrylmide gel. Gels were dried and analyzed using a Typhoon PhosphorImager system. The oligonucleotides used in the electrophoretic mobility shift assay (EMSA) included a negative control (GATCCTGCTGTGCCAGAATACAAAATGCTAATAACTAGGGCAACTA) from subtelomere XqYq containing a mutated CTCF binding site, a positive control (GATCCTGCTGTGCCAGGGCGCCCCCTGCTGGCGACTAGGGCAACTA) containing a putative CTCF binding site from subtelomere XqYq, CTCF BS wt (AAGTCACATCAATAATAAGGGCGCCATCTAGCGGGAGATGTTATCCACACCATCCCAA), and CTCF BS ΔCTCF (TTGCATCACAAGTCACATCAATAATAAGGTGTTATCCACACCATCCCAATTCACATCT).

PFGE.

EBV-positive cells were resuspended in 1.0% agarose plugs and incubated in lysis buffer (0.2 M EDTA [pH 8.0], 1% SDS, 1 mg of proteinase K/ml) at 50°C for 48 h. The agarose plugs were washed twice in Tris-EDTA buffer (pH 7.5). Pulsed-field gel electrophoresis (PFGE) was performed for 22 h at 14°C, with a linear ramping pulse of 60 to 120 s through 120°C as described previously (Bio-Rad CHEF Mapper) (47). DNA was transferred to nylon membranes according to established methods for Southern blotting. The DNA was detected by hybridization with a ³²P-labeled probe specific for the EBV DS region or human chromosome 17 and visualized with a Molecular Dynamics PhosphorImager.

Fluorescence in situ hybridization (FISH).

For metaphase spreads, 0.5 × 10⁶ cells/ml were incubated in 0.1 μg of Colcemid (Roche)/ml for 4 h. Colcemid-treated cells were then spread onto slides and treated with 25 μg of RNase A/ml in 2× SSC (1× SSC is 0.15 M NaCl plus 0.015 M sodium citrate; Cellgro) for 40 min. The slides were washed quickly in 2× SSC to remove the RNase and placed in 70% ethanol (EtOH) for 3 min, then 90% EtOH for 3 min, and then 100% EtOH for 30 min, air dried for 10 min, and finally oven dried at 80°C for 5 min. Slides were then placed in denaturing solution (70% formamide in 2× SSC) in a 77°C water bath for 3 min, followed by 2.5 min each in an ethanol series (70, 90, and 100% EtOH), and air dried for 10 min. The slides were then incubated with 250 ng of EBV bacmid DNA probe labeled with biotin at 37°C overnight. Biotin-labeled EBV probe was prepared by using EBV bacmid as the DNA template with a nick translation kit (Invitrogen) according to the manufacturer's protocol. The next day, the slides were blocked in 5% milk 4× SSC solution and incubated with biotin α-avidin (Invitrogen) for 30 min at 37°C, followed by Oregon Green 488 nm-avidin (Invitrogen) for 30 min at 37°C. Slides were washed in 4× SSC-Tween and counterstained with DAPI (4′,6′-diamidino-2-phenylindole), and coverslips were placed with antifade and visualized with a ×100 lens on a Nikon E600 Upright microscope (Nikon Instruments) with ImageProPlus software (Media Cybernetics) and Adobe PhotoShop CS5 for image processing.

RESULTS

Generation of an EBV bacmid lacking CTCF166.

ChIP-Seq analyses of CTCF and cohesin binding to the EBV genome revealed a major co-occupied binding site within the LMP1/LMP2A transcription locus at the right end of the linear genome map (Fig. 1A). A consensus CTCF binding site was located at the center of the CTCF ChIP-Seq peak at position 166520 (EBV genome NC_007605.1) (Fig. 1B). To investigate the function of this CTCF binding site, we used recombineering methods to delete the core 18 bp of this site in the EBV bacmid genome, which spared any coding regions of the LMP1 or LMP2A open reading frames, annotated poly(A) sites, or splice acceptor-donor sites (Fig. 1C) (42). To determine whether the site and its deletion was sufficient to bind and disrupt CTCF, we performed EMSA with purified recombinant CTCF and oligonucleotides spanning the wt CTCF site or the same length oligonucleotide reconstructed with the 18-bp deletion (ΔCTCF) (Fig. 1D). We found that CTCF bound to the wt probe similarly to that of a canonical CTCF binding site (Ctrl+). CTCF did not bind efficiently to the ΔCTCF probe or to a control probe with no CTCF consensus binding site (Ctrl−). This demonstrates that the core sequence at position 166520 binds directly to CTCF and that an 18-bp deletion disrupts CTCF binding at this position.

The EBV bacmid genome containing the 18-bp deletion in CTCF166 was validated by restriction digestion (Fig. 1E), as well as by direct sequencing of the region surrounding the deletion. Bacmids for wt and ΔCTCF166 were transfected into 293HEK cells and used to generate infectious virus for immortalization of primary human B-lymphocytes and generation of LCLs. Two independently generated LCLs for EBV ΔCTCF166 (designated ΔCTCF.1 and ΔCTCF.2) were established (>8 weeks postinfection) and expanded for further characterization. To ascertain whether the deletion disrupted CTCF and cohesin binding in vivo, LCLs containing either EBV wt bacmid or ΔCTCF166 (ΔCTCF.1 or ΔCTCF.2) were assayed by ChIP for CTCF and cohesin binding (Fig. 1F). We found that CTCF and cohesin subunits Rad21, SMC1, and SMC3 were significantly reduced for binding to the region surrounding the CTCF166 site in EBV ΔCTCF.1 and ΔCTCF.2 relative to EBV wt LCLs. These results indicate that the ΔCTCF166 bacmid viral genomes can immortalize B lymphocytes and yet lack CTCF and cohesin binding at position 166520.

Slow proliferation rates and increase copy number in EBV ΔCTCF166 LCLs.

Although EBV ΔCTCF166 bacmids could immortalize primary B lymphocytes, we also observed that expansion of these cells was generally slower than for EBV wt bacmid genomes (Fig. 2A). We measured the growth curves of two independent LCLs derived from EBV wt (wt.1 and wt.2) and ΔCTCF166 (ΔCTCF.1 and ΔCTCF.2) bacmid virus and confirmed that EBV ΔCTCF166 LCLs proliferate at a lower rate than EBV wt bacmids (Fig. 2A). To further characterize EBV ΔCTCF166 LCLs, we examined the status of EBV episomes using PFGE, which can resolve the circular episomes from linear and integrated forms of viral DNA (Fig. 2B). We found that LCLs with EBV ΔCTCF166 were enriched for circular episomes relative to wt EBV LCLs (Fig. 2B). EBV ΔCTCF LCLs showed a slight increase in linear genomes as well, suggesting that the increase in copy number may be a consequence of partial lytic cycle replication or alternatively, aberrant latent cycle DNA replication. To further evaluate EBV copy number, we assayed episome number using metaphase FISH for EBV DNA (Fig. 2C and E). We found that ΔCTCF166 LCLs contained 2- to 3-fold more genomes per metaphase chromosome than wt EBV. We also observed that ΔCTCF166 LCLs had stronger intensity and enlarged foci, suggesting that these are multicopy or less compact than EBV wt genomes (Fig. 2C). To obtain a more quantitative measure of EBV copy number difference in ΔCTCF166 relative to wt EBV LCLs, we used qPCR with primers specific for viral and cellular DNA and normalized EBV DNA to Namalwa DNA from total genomic DNA preparations (Fig. 2D). Consistent with the PFGE and metaphase FISH results, we found that ΔCTCF166 LCLs had 4- to 8-fold-greater copy numbers than LCLs generated with EBV wt bacmid. These findings suggest that deletion of CTCF166 results in an increase in EBV genome copy number relative to wt genomes.

FIG 2.

EBV ΔCTCF166 has higher copy numbers and reduced proliferation rates in LCLs. (A) LCL population growth rates were measured by trypan blue staining for EBV wt.1, wt.2, ΔCTCF.1, or ΔCTCF.2 over a 16-day period. All LCLs were assayed >8 weeks post-B-cell infection. The average value of three independent cell counts were made for each time point, and the error bars show the SD. (B) PFGE of 10⁶ LCLs immortalized with EBV wt.1, wt.2, ΔCTCF.1, or ΔCTCF.2. EBV DNA (top panel) was detected in the gel well (chromatin trapped), episome position, and linear forms, as indicated. PFGE Southern blots were reprobed with cellular DNA for chromosome 17 as an internal loading control (lower panel). (C) FISH analysis of EBV DNA on metaphase chromosomes from EBV wt (left panels) or EBVΔCTCF (right panels) LCLs. (D) Quantification of EBV DNA FISH shown as representative examples in panel C. FISH signals were quantified as the average number of EBV DNA foci per total number of metaphase chromosomes. (E) qPCR of EBV DNA relative to cellular actin was quantified from total DNA derived from EBV wt.1, wt.2, ΔCTCF.1, or ΔCTCF.2 transformed LCLs.

Deregulation of LMP1, LMP2A, and LMP2B expression in EBV ΔCTCF166 LCLs.

To investigate whether EBV ΔCTCF166 had changes in viral gene expression, we compared EBV wt and ΔCTCF166 bacmid LCLs for viral gene expression using qRT-PCR (Fig. 3). We compared the relative mRNA levels of EBV wt.1, wt.2, ΔCTCF.1 and ΔCTCF.2 for latency transcripts LMP1, LMP2A, EBNA1, and EBNA2, as well as lytic transcripts for BNRF1 and gp350 (BLLF1). We found that LMP1 and LMP2A mRNA levels were substantially reduced (2- to 10-fold) in EBV ΔCTCF166 relative to the wt bacmid (Fig. 3A). In contrast, we found that LMP2B mRNA was increased 8- to 10-fold in EBV ΔCTCF166 relative to wt bacmid. The changes in LMP2A and LMP2B mRNA were confirmed with two independent primer sets that span intron-exon junctions (Fig. 3D). None of the other viral transcripts tested were altered as significantly as were LMP1, LMP2A, and LMP2B. We did not observe any significant increase in lytic transcripts, suggesting that lytic cycle gene expression is not occurring at detectable levels in these LCLs. Western blot analysis also revealed a reduction in LMP1 protein levels in ΔCTCF166 LCLs relative to the wt (Fig. 3B and C), with no apparent increase in the lytic viral protein Zta expression levels. LMP1 protein levels could be restored to wt levels when LCLs were treated with lytic inducing agents (e.g., TPA and NaB), indicating that these mutant genomes can produce LMP1 of similar size as the wt genomes (Fig. 3B). Interestingly, we did not observe any increase in basal Zta expression between wt and ΔCTCF166 LCLs, and we found a moderate reduction in Zta levels after TPA induction (Fig. 3B). These findings indicate that LMP1 and LMP2A mRNA and LMP1 protein levels are reduced, whereas LMP2B mRNA is elevated in EBV ΔCTCF166 relative to the wt bacmid.

FIG 3.

Deregulated LMP1, LMP2A, and LMP2B transcription in EBV ΔCTCF166 LCLs. (A) RT-qPCR for LMP2A, LMP2B, LMP1, EBNA1, EBNA2, BNRF1, and gp350 (BLLF1) mRNA were normalized to bacmid GFP for EBV wt.1, wt.2, ΔCTCF.1, or ΔCTCF.2. Error bars represent the SD for three independent RT-PCRs. (B) Western blot of EBV wt, ΔCTCF.1, or ΔCTCF.2 LCLs untreated (−) or treated (+) with lytic inducers TPA (20 ng/ml) and sodium butyrate (3 mM), and probed with antibody to Zta, LMP1, or actin. (C) Western blot of EBV wt or EBVΔCTCF.1 probed for EBNA1, LMP1, Zta, or PCNA. (D) Schematic of LMP1, LMP2A, and LMP2B transcripts and the primer pairs used for RT-qPCR.

Loss of euchromatic histone modifications at LMP promoters in EBV ΔCTCF166 LCLs.

To determine the effect of the CTCF166 deletion on histone modifications surrounding the LMP1 and LMP2 promoter regions, we performed ChIP assays to measure histone modifications H3K9ac, H3K4me3, and H3K9me3 (Fig. 4). We assayed ChIP DNA using primer pairs spanning the LMP1 and LMP2 5′ proximal promoter regions (Fig. 4A). We found that EBV ΔCTCF166 had substantial loss of H3K9ac at the LMP2A and LMP1 promoter regions (Fig. 4B). Interestingly, H3K4me3 levels were not significantly different than wt levels. In contrast, H3K9me3 levels were elevated at the LMP2A promoter region, especially in LCL ΔCTCF.2 genomes. Western blots indicated that the global levels of these modified histones were similar, whereas the LMP1 levels were reduced in ΔCTCF166 LCLs (Fig. 4C). We also examined these histone modifications in established LCLs (type III latency) and Mutu I (type I latency) cells and found that H3K9Ac and H3K4me3 were highly elevated at these positions in LCLs, where LMP1 and LMP2 are expressed, but not in Mutu I, where these genes are not expressed (Fig. 4D). These findings indicate that histone H3K9ac and H3K4me3 correlate with transcription activity and that CTCF deletion at this site results in a loss of histone acetylation and a corresponding increase in H3K9me3 at the LMP2A and LMP1 control regions.

FIG 4.

Loss of euchromatic histone modifications at the LMP1/LMP2 control region in EBV ΔCTCF166. (A) Schematic of EBV genome with position of CTCF166 binding site (star) and PCR primers used for ChIP assay indicated relative to LMP1, LMP2A, and LMP2B transcription start sites. (B) ChIP assay for H3K9Ac, H3K4me3, H3K9me3 with EBV wt, ΔCTCF.1, or ΔCTCF.2 derived LCLs. Error bars represent the SD for three independent PCRs. (C) Western blot of protein extracts from EBV wt, ΔCTCF.1, or ΔCTCF.2 LCLs probed with antibodies to H3K9Ac, H3K4me3, H3K9me3, H3, LMP1, actin, and GFP. (D) ChIP assays for H3K9Ac, H3K4me3, H3K9me3, and IgG for Mutu I and LCL.

Increased CpG DNA methylation at LMP promoters in EBV ΔCTCF166 LCLs.

We next tested whether the DNA methylation pattern was altered at the LMP promoter regions for EBV ΔCTCF166 LCLs (Fig. 5). DNA methylation was measured by MeDIP using an antibody specific for methyl cytosine (48). We found that DNA methylation was elevated in the promoter control regions for both LMP1 and LMP2 in EBV ΔCTCF166 LCLs relative to wt LCLs. Interestingly, and in contrast to H3K9me3, DNA methylation was enriched to a greater extent at the LMP1 and LMP2B promoter regions (primers 4 and 5) relative to the LMP2A promoter region (primers 1 to 3). The partial overlap of DNA methylation with H3K9me3 suggests that these two modifications are not identical forms of heterochromatin and that both epigenetic mechanisms may contribute to the transcriptional repression of LMP1 and LMP2A in ΔCTCF166 LCLs. We also found that cytosine methylation was highly elevated at the LMP1 and LMP2A promoter regions in Mutu I cells relative to LCLs (Fig. 5B), a finding consistent with its established role in transcriptional repression of LMP1 and LMP2 in type I latency (49).

FIG 5.

Increased DNA methylation at the LMP1/LMP2 control region in EBV ΔCTCF166. (A) MeDIP assay for EBV wt, ΔCTCF.1, or ΔCTCF.2 using primers as indicated in Fig. 4A. Error bars represent the SD for three independent PCRs. (B) MeDIP assay for Mutu I and LCL using primers described in Fig. 4A.

Loss of DNA loop interaction with OriP in ΔCTCF166 LCLs.

3C methods were used to measure the interaction between the LMP and OriP regions. Previous studies revealed that an LMP-OriP interaction exists and is sensitive to shRNA depletion of cohesin subunits Rad21 and SMC1 (38). We now show that 3C interactions between the LMP1/LMP2 control region and OriP are similarly disrupted in ΔCTCF166 LCLs (Fig. 6A). We used an anchor primer set at the MseI site near the CTCF166 position and measured 3C products by qPCR for interactions with three independent regions surrounding OriP or the latent viral promoter Cp. We found that a specific long-range interaction could be detected in LCL wt between the LMP region and OriP but not with the control Cp region. In contrast, no specific 3C products could be detected in ΔCTCF166 LCLs. 3C products were quantified as a percentage of qPCR products that could be measured from undiluted matrix reactions, indicating that the PCR products could be generated and were not affected by the deletion of the CTCF binding site. We conclude that CTCF166 is required for the formation of a stable DNA loop between LMP and OriP control regions. We also observed similar patterns of 3C interactions between LMP region and OriP in established LCLs and Mutu I cells (Fig. 6B). The relatively strong 3C signal in Mutu I cells, where LMP1 and LMP2 genes are epigenetically silenced, suggests that DNA loop formation is not sufficient for transcription activation of the LMP gene locus and may have different functions in different latency types.

FIG 6.

Loss of DNA loop formation between LMP and OriP control regions in EBV ΔCTCF166. (A) 3C assay for EBV wt, ΔCTCF.1, or ΔCTCF.2 LCLs using an anchor primer at the LMP1 promoter region (red arrow). Acceptor primer positions (black arrows) are indicated above, with three in the OriP region and three in the Cp/Wp region. MseI restriction sites are indicated. Error bars represent the SD for three independent PCRs. (B) 3C assay for LCL and Mutu I cells using acceptor primers at MseI restriction sites indicated on the x axis. Error bars represent the SD for three independent 3C reactions.

DISCUSSION

Coordinated gene regulation requires integration of multiple signals and communication between spatially separate elements. EBV gene expression is regulated at multiple levels, including alternative transcript initiation and mRNA splicing, multiple virus-encoded feed-forward and feed-back systems, and epigenetic repression of latent and lytic transcription. We show here that CTCF-cohesin confers an additional level of coordinate regulation to the expression of the EBV LMP genes in latently infected LCLs. In particular, we show that CTCF mediates the binding of cohesins to the CTCF166 site in the LMP2 first intron and LMP1 3′UTR (Fig. 1). We show that deletion of this site results in a decrease in LMP1 and LMP2A mRNA and a corresponding increase in LMP2B mRNA expression (Fig. 3). We found that CTCF166 is important for maintaining a euchromatic histone modification pattern at the promoter regions of LMP2A and LMP1 and that heterochromatic H3K9me3 and CpG DNA methylation increased at this control region when CTCF166 was deleted (Fig. 4 and 5). Finally, we show that CTCF166 is important for the DNA loop formed between the LMP region and the OriP enhancer (Fig. 6). These findings indicate that CTCF166 is an important site of transcriptional regulation and chromatin control element of the LMP locus. These findings also suggest that CTCF166 provides a mechanism for coordinate regulation of LMP genes with each other and with the transcriptional enhancer function of the OriP element (Fig. 7).

FIG 7.

Model depicting the function of CTCF166 in regulating LMP1/LMP2 chromatin and gene regulation. The loss of CTCF166 binding results in the accumulation of heterochromatic H3K9me3 and DNA methylation at the LMP2A and LMP1 promoter regions, the upregulation of LMP2B transcription, and loss of interaction with OriP enhancer.

CTCF has been implicated as a chromatin insulator and boundary factor (50–52). We have previously shown that CTCF can prevent epigenetic drift by blocking heterochromatin formation at the EBV Qp region (46). In the present study, we examined the role of CTCF at the LMP1/LMP2 region using EBV bacmid-immortalized LCLs. We found that deletion of the CTCF166 binding site resulted in an increase in histone H3K9me3, especially at the LMP2A promoter region (primers 1 to 3). We also found that histone H3K4me3 methylation was not significantly altered at the LMP1 and LMP2B promoter regions (Fig. 4B). In contrast, DNA methylation was more elevated at the LMP1 promoter region (primer 4 and 5) relative to the LMP2A promoter in ΔCTCF166 LCLs (Fig. 5A), although DNA methylation levels never achieved those observed in highly repressed Mutu I cells (Fig. 5B). Taken together, these results suggest that CTCF plays a complex role in regulating epigenetic modifications at the LMP locus. One possible function of CTCF at the LMP locus is to block the spread of heterochromatin that might be generated by the GC-rich repetitive DNA of the EBV terminal repeat (TR) region. Larger numbers of TRs have been shown to correlate with the epigenetic repression of LMP1 (53). Preventing the spread of heterochromatin from TR would account for the role of CTCF in activating LMP2A. However, it remains unclear how the loss of CTCF binding in ΔCTCF166 LCLs leads to the activation of LMP2B, which is in closer proximity to the TRs (Fig. 3A). This suggests that CTCF may have a more complex three-dimensional insulator function that is important for promoter selection and enhancer communication (Fig. 7).

CTCF and cohesins have been implicated in DNA loop interactions (22, 54, 55). Long-distance DNA interactions are essential to mediate communication between promoter and enhancer regulatory elements. EBV OriP has been implicated as a transcriptional enhancer of Cp and LMP1 promoter, but the mechanism of enhancer-promoter communication has not been well established (56, 57). Our recent studies using 3C methods indicate that OriP can physically interact with Cp in cell types where Cp is active (45). 3C has also shown that CTCF166 can interact with OriP, suggesting that it provides a mechanism for transcriptional activation of LMP1 promoter (38). Here, we show by genetic disruption that the loss of CTCF binding causes a loss of loop formation between LMP locus and OriP (Fig. 6). These findings support earlier biochemical and shRNA experiments and further demonstrate a correlation between OriP loop formation and transcriptional activation of LMP1 and LMP2A. How OriP functions as a transcriptional enhancer is not yet known. Several transcription factors can associate with OriP and Cp, such as Oct2 and OcaB (58), but it is not known how these factors may mediate differential interactions with other viral promoters, like LMP1, that can be enhanced by OriP. Previous studies have revealed that the region surrounding OriP is consistently elevated in euchromatic H3K4me3 (38, 59). We suggest that OriP drives euchromatin formation, either by direct recruitment of histone H3K4 methyltransferases, or altering the local chromatin environment to favor H3K4me3 formation. Interaction with other DNA elements, such as CTCF166, would enhance the formation of euchromatin at these interacting loci, perhaps by proximity to euchromatic histone methyltransferases (Fig. 7). However, we also found that OriP interacts with LMP locus in Mutu I cells where LMP1 and LMP2 are epigenetically silenced. This suggests that the OriP-LMP locus loop interaction is not sufficient for transcription activation and that factors other than loop formation must determine the latency type transcription profile.

CTCF166 is also located in the junction between two converging transcripts (LMP1 and LMP2A) that share some common template DNA. CTCF166 may function to help direct the RNA polymerase II traffic through these opposing transcripts. In the Kaposi's sarcoma-associated herpesvirus (KSHV) genome, a cluster of three CTCF sites lies within the first intron of the multicistronic latency transcript that encodes LANA, vCyclin, and vFLIP (60). This CTCF site cluster is highly reminiscent of the EBV CTCF166 site since it is also highly enriched for cohesin, regulates a latency gene cluster, and is situated in the first intron of a complex viral RNA transcript. In KSHV, the CTCF sites have been implicated in DNA looping with the KSHV lytic control region (61). In addition, the CTCF sites appear to alter RNA polymerase II function in mRNA processing by preventing alternative splicing and downstream promoter utilization (62). EBV CTCF166 may have similar functions in regulating the complex trafficking of RNA polymerase across the LMP1 and LMP2A transcripts. One simple mechanism could be the induced pausing of RNA polymerase II, and subsequent assembly of either splicing (LMP2A) or termination (LMP1) factors required to generate functional mRNA products.

CTCF binds to ∼19 sites in the EBV genome (38, 40). Several of these sites are co-occupied by cohesins, but the CTCF166 site appears to be unique for its high co-occupancy of both CTCF and cohesin. One remarkable phenotype of ΔCTCF166 LCLs is the elevated viral genome copy number (Fig. 2B to E). The increase in viral copy number was verified by several independent methods, including PFGE, metaphase FISH, and qPCR (Fig. 2). The increase viral DNA copy number is unlikely to be due to an increase in lytic cycle DNA replication since elevated lytic gene expression was not detected (Fig. 3). It is possible that elevated genome copy compensates for deficiencies in LMP1 mRNA expression. Alternatively, CTCF-cohesin may facilitate sister-chromatid cohesion and postreplicative repair through enhanced homologous recombination. The absence of this function may result in DNA replication defects, including compensatory amplification of viral genomes. Further studies will be required to resolve some of these more complicated functions of CTCF and to better understand how EBV has exploited CTCF binding sites to confer coordinate gene regulation and genome propagation in latent infection.