Differential Methylation of Epstein-Barr Virus Latency Promoters Facilitates Viral Persistence in Healthy Seropositive Individuals

Abstract

Epstein-Barr virus (EBV) establishes a life-long infection in humans, with distinct viral latency programs predominating during acute and chronic phases of infection. Only a subset of the EBV latency-associated antigens present during the acute phase of EBV infection are expressed in the latently infected memory B cells that serve as the long-term EBV reservoir. Since the EBV immortalization program elicits a potent cellular immune response, downregulation of viral gene expression in the long-term latency reservoir is likely to facilitate evasion of the immune response and persistence of EBV in the immunocompetent host. Tissue culture and tumor models of restricted EBV latency have consistently demonstrated a critical role for methylation of the viral genome in maintaining the restricted pattern of latency-associated gene expression. Here we extend these observations to demonstrate that the EBV genomes in the memory B-cell reservoir are also heavily and discretely methylated. This analysis reveals that methylation of the viral genome is a normal aspect of EBV infection in healthy immunocompetent individuals and is not restricted to the development of EBV-associated tumors. In addition, the pattern of methylation very likely accounts for the observed inhibition of the EBV immortalization program and the establishment and maintenance of a restricted latency program. Thus, EBV appears to be the first example of a parasite that usurps the host cell-directed methylation system to regulate pathogen gene expression and thereby establish a chronic infection.

Epstein-Barr virus (EBV), a gammaherpesvirus, is the etiologic agent of infectious mononucleosis, a self-limiting lymphoproliferative disease. In addition, EBV is associated with the development of several malignancies, including the endemic form of Burkitt’s lymphoma (BL), nasopharyngeal carcinoma, 30 to 50% of Hodgkin’s disease, and approximately half of AIDS-associated lymphomas (53). EBV infection of peripheral B cells in culture leads to the efficient establishment of a latent infection (absence of lytic virus replication), with concomitant growth transformation of the infected lymphocytes (immortalization). During latent infection, both in culture and in vivo, the viral genome is maintained as an autonomously replicating episome (14). EBV-immortalized B lymphoblasts are known to express nine viral antigens, including six Epstein-Barr nuclear antigens (EBNAs 1, 2, 3a, 3b, 3c, and 4) and three latency-associated membrane proteins (LMP1, LMP2A, and LMP2B). Transcription of the latency-associated EBNA genes occurs from a single complex transcriptional unit that spans nearly 100 kb of the viral genome (Fig. 1). One of two promoters, Cp or Wp, drives transcription of the EBNA genes, and a combination of alternative splicing and read through of 3′ processing signals leads to the synthesis of mature EBNA gene transcripts (Fig. 1) (67).

FIG. 1.

Schematic illustration of EBV EBNA gene transcription during immortalizing (type III) and restricted (types I and II) viral latency. Shown is a linear representation of the EBV genome, with the locations of the EBNA gene promoters Cp, Wp, and Qp indicated. The general organization of exons present in EBNA gene transcripts during immortalizing (type III) latency and restricted (type I and type II) latency is shown above the viral genome. Type III latency involves expression of all six EBNAs, generated by alternative splicing of long primary transcripts that initiate at either Cp or Wp. During type I or type II latency, only one of the EBNAs, EBNA 1, is expressed, and its transcription is driven from Qp (Cp and Wp are silent during restricted EBV latency). Fp is a lytic promoter located ca. 200 bp upstream of Qp (34, 64); oriP is the latency-associated origin of episomal replication; IR1, the major internal repeat, consists of 8 to 10 copies of the 3.0 kb BamHI W fragment. Note that the sizes of the small 5′ EBNA gene exons are exaggerated for the sake of clarity. EBNA 1 binding to oriP functions to upregulate Cp and Wp activity, as well as to assist with maintenance of the viral episome. EBNA 1 also functions to downregulate Qp through binding to two low-affinity sites downstream of Qp. TR, direct terminal repeats present at each end of the linear viral genome which mediate circularization of the viral genome upon infection.

EBV-immortalized lymphoblasts (type III latency) are detected during the acute phase of virus infection in infectious mononucleosis patients (72) but are eventually eliminated by a strong cellular immune response (directed against all the EBNAs except EBNA 1, which fails to undergo normal processing and presentation by major histocompatibility complex class I [36]). However, EBV manages to persist for the lifetime of the infected host by establishing a restricted program of latency characterized by extremely limited viral gene expression. During chronic asymptomatic infection in healthy individuals, EBV resides in memory B cells (2). Viral gene expression in these cells is restricted to expression of LMP2A or LMP2A plus EBNA 1 (2, 10, 50, 72). Notably, restricted forms of EBV latency (type I and type II latency) are also observed in the EBV-positive tumors that arise in immunocompetent hosts (53). Expression of EBNA 1 is observed, either alone (BL) or in combination with the LMPs (nasopharyngeal carcinoma and Hodgkin’s disease). During restricted latency, transcription of the EBNA 1 gene arises from a distinct promoter, Qp (Fig. 1A), whereas Cp and Wp are silent (46, 63). While transcription from Cp and Wp is positively regulated by EBNA 1, transcription from Qp is negatively regulated by EBNA 1 (Fig. 1) (49, 61, 65). Thus, the major feature that distinguishes immortalizing (type III) latency from restricted (type I/II) viral latency is the EBNA gene promoter(s) that is active (Cp/Wp or Qp).

We have begun to address the question of what factors enable EBV to establish restricted latency. Previous studies demonstrated that EBNA 1 activates transcription from Cp/Wp and represses Qp as the default pathway, unless Cp/Wp are inactivated (65). Furthermore, transfection of EBV-infected cell lines exhibiting a restricted latency program with Cp/Wp-driven reporter constructs leads to Cp/Wp activity from the reporter constructs, while the endogenous Cp/Wp remain silent (25, 57, 63, 65). The latter result indicates that these cells contain the required factors to drive Cp/Wp-initiated transcription. However, numerous studies have documented that Cp and Wp are extensively methylated in tumors and cell lines that display restricted EBV latency (1, 16, 25, 39, 41, 42, 57, 58, 65, 70), while Qp is protected from methylation (65, 71). Additionally, many EBV-positive BL cell lines drift during passage in tissue culture from a type I latency program to the type III immortalizing latency program, with a concomitant loss of viral genome methylation (42). Finally, treatment of BL cell lines exhibiting restricted EBV latency with 5-azacytidine, a methyltransferase inhibitor, in some cases leads to demethylation of the viral genome and subsequent activation of EBNA gene transcription from Cp and Wp (1, 25, 39, 57). Thus, since methylation of CpGs in the DNA of higher eukaryotes tightly correlates with transcriptional inactivity (8, 22, 32), methylation may be the primary mechanism for the establishment of the restricted EBV latency programs and is clearly required for maintenance of the type I latency program in BL cell lines.

As indicated above, methylation of CpGs in the DNA of higher eukaryotes correlates strongly with transcriptional inactivity. The exact mechanism by which methylation of a promoter leads to repression of gene expression remains unclear, but current evidence suggests that chromatin remodeling plays a role (5). Two groups have recently demonstrated that a methyl-CpG binding protein, MeCP2, binds specifically to methylated CpGs (27, 45). MeCP2 can displace histone H1 from nucleosomes and then, at least in part by recruiting histone deacetylases, allow chromatin to rearrange into a tight structure that excludes transcription factors (27, 45). A deficiency in the methyltransferase enzyme is embryonic lethal in mice, suggesting that methylation is required for development (37). Furthermore, methylation may function as a host defense strategy to inactivate parasitic elements (3, 4, 6). For instance, CpG methylation prevents activation of endogenous integrated retroviruses (18, 24, 73). A correlation between methylation of specific viral genes and transcriptional repression has also been observed for other viruses, including Marek’s disease virus (19, 28) and simian foamy virus type 3 (66).

Circumstantial evidence indicating that methylation of the EBV genome may be a normal aspect of EBV passage in vivo comes from inspection of the EBV genomic sequence, which reveals that the frequency of CpG dinucleotides is depressed (21, 30, 31). CpGs are also underrepresented in the genomes of three other gammaherpesviruses, bovine herpesvirus 4, herpesvirus saimiri, and murine gammaherpesvirus 68 (15, 31). The decreased frequency of CpG dinucleotides suggests that these viral genomes have been subjected to extensive methylation during growth in their host, resulting in the loss of CpG dinucleotides by mutation of methylated CpGs to TpGs. This implies that the observed methylation of the viral genome in EBV-associated tumors may reflect a normal component of the biology of viral infection.

Here we extend the previous analyses of viral genome methylation in EBV-positive tumors and tumor cell lines to demonstrate that the regulatory regions of the EBNA gene promoters in the EBV genome are differentially methylated in the long-term latency reservoir in normal seropositive individuals. The observed methylation pattern most likely accounts for the absence of detectable transcription of the EBNA genes associated with the type III immortalizing latency program. Thus, as observed in EBV-positive tumors, methylation of the viral genome appears to facilitate evasion of the host immune response and thus persistence of virus-infected cells.

MATERIALS AND METHODS

Genomic DNA isolation and bisulfite treatment.

Genomic DNA was isolated from the partially purified B cells (35) of three healthy EBV-seropositive human donors (donors 8, 9, and 10) and the EBV-negative BL cell line DG75 (negative control DNA). A Qiagen genomic DNA kit (Qiagen Inc., Santa Clarita, Calif.) was used. Prior to bisulfite treatment, genomic DNA from healthy seropositive donor B cells and from EBV-negative cells was digested overnight with PstI (Cp and Qp) or XhoI (Wp), ethanol precipitated, and bisulfite treated as previously described (12, 40, 51), with a few minor changes. Briefly, 5 μg of DNA was denatured in 0.3 M NaOH at 68°C for 15 min and incubated with 2.5 M sodium metabisulfite (pH 5.0) and 0.5 mM hydroquinone for 3 h in the dark at 55°C. DNA was purified and desalted with a GeneClean II kit (Bio 101, Inc., Vista, Calif.), desulfonated in 0.3 M NaOH, and precipitated with ammonium acetate, ethanol, and 1 μg of carrier glycogen. Bisulfite-mediated conversion of non-CpG cytosines into thymines was an average of 99.6% complete for Cp and 99.9% complete for Wp and Qp.

Bisulfite PCR analysis of EBV genome methylation.

EBV sequences were amplified by nested PCR from bisulfite-treated cellular DNA (1.5 to 2.5 μg) by using AmpliTaq Gold hot-start polymerase (Perkin-Elmer) and primers specific for the bottom strand of the viral genome. The sequences of the PCR primers used, and the assay conditions for each primer set, are outlined below. Bisulfite-treated DNA isolated from EBV-negative cells and water controls between donors were included as negative controls to ensure the absence of contamination. For the data compiled in Fig. 2, no false positives were detected in 50 DG75 EBV-negative DNA control PCRs and 93 water control PCRs.

FIG. 2.

In vivo methylation status of the EBV EBNA gene promoters Wp, Cp, and Qp in peripheral B cells. The genome coordinate for each CpG within the regions examined is given at the left. ●, methylated CpGs; ○, unmethylated CpGs. Each numbered column represents the data obtained from an independent bisulfite PCR reaction. Genomic DNA was isolated from the B cells of three healthy EBV-seropositive human donors (donors 8, 9, and 10) and from the EBV-negative BL cell line DG75 (negative control DNA). DNA was bisulfite treated and amplified by nested PCR. PCR products were cloned and sequenced. Conversion of non-CpG cytosines into thymines was at least 99.6% complete for all PCRs analyzed. For the compiled data shown, no false positives were detected in 50 control PCRs carried out with DG75 EBV-negative DNA and 93 water control PCRs. (A) CpG methylation in the region upstream of Cp. Eighteen independently generated bisulfite PCR products were analyzed. Also shown are the locations of the EBNA 2 enhancer (EBNA2-enh), the CCAAT boxes, and the Cp TATAA box and transcription start site. The P at genome coordinate 11210 in PCR 5 from donor 8 represents a polymorphism (CpG-to-TpG mutation). (B) CpG methylation in the region upstream of Wp. The analysis of 18 independently generated bisulfite PCR products is shown. The Cp/Wp enhancer, the CCAAT box, the Wp TATAA box and transcription start site, the W0 exon, and the W1 exon are indicated as well. (C) Hypomethylation of the region immediately upstream of Qp and hypermethylation of the region upstream of Fp. Sixteen independently generated bisulfite PCR products were analyzed. Also shown are the locations of the Fp TATAA box and transcription start site, the IRF1/IRF2 binding site, the Qp transcription start site, and the low-affinity EBNA 1 binding sites.

(i) Cp PCR.

For PCR amplication using 5′ outer Cp PCR primer (from genome coordinate bp 10843); 5′-ATATCCCAATTAAAAACCC and 3′ outer Cp PCR primer (from genome coordinate bp 11425); 5′-GTTAAGTGGGTTTATATGGT, reaction mixtures were incubated at 95°C for 10 min followed by 5 cycles of linear amplification with the 5′ primer and 45 cycles of amplification with both primers (94°C for 1 min, 60°C for 1 min, and 72°C for 2 min). For PCR amplification using 5′ inner Cp PCR primer (from genome coordinate bp 10,898; 5′-CATACACCCTAAACCAACC) and 3′ inner Cp PCR primer (from genome coordinate bp 11,385; 5′-ATGAGGGTTTTGGGGGTTT), reaction mixtures were incubated at 95°C for 10 min followed by 30 cycles of amplification (94°C for 1 min, 63°C for 1 min, and 72°C for 2 min).

(ii) Wp PCR.

For PCR amplification using outer Wp PCR primer (from genome coordinate bp 13796; 5′-CCCCCAAACTTTATCCAAATA) and 3′ outer Wp PCR primer (from genome coordinate bp 14660; 5′-TGGAGTGTTGGGTTTAGTAG), reaction mixtures were incubated at 95°C for 10 min followed by 5 cycles of linear amplification with the 5′ primer and 45 cycles of amplification with both primers (94°C for 1 min, 65°C for 1 min, and 72°C for 2 min). For PCR amplification using 5′ inner Wp PCR primer (from genome coordinate bp 13918; 5′-CCTATCACCAAACCTACCA) and 3′ inner Wp PCR primer (from genome coordinate bp 14486; 5′-GGGGAAAAGTTAGAAATTGGGT), reaction mixtures were incubated at 95°C for 10 min followed by 30 cycles of amplification (94°C for 1 min, 65°C for 1 min, and 72°C for 2 min).

(iii) Qp PCR.

For PCR using 5′ outer Qp PCR primer (from genome coordinate bp 61889; 5′-CCTCTATTTCTTCATCTATTAA) and 3′ outer Qp PCR primer (from genome coordinate bp 62486; 5′-GTTAAAATGTAAGGATAGTATG), reaction mixtures were incubated at 95°C for 10 min followed by 5 cycles of linear amplification with the 5′ primer and 45 cycles of amplification with both primers (94°C for 1 min, 59°C for 1 min, and 72°C for 2 min). For PCR using 5′ inner Qp PCR primer (from genome coordinate bp 62081; 5′-CAAATACAAAAACTTAAATCTC) and 3′ inner Qp PCR primer (from genome coordinate bp 62486; 5′-GTTAAAATGTAAGGATAGTATG), reaction mixtures were incubated at 95°C for 10 min followed by 30 cycles of amplification (94°C for 1 min, 59°C for 1 min, and 72°C for 2 min).

PCR cloning and sequencing.

PCR products were cloned into the pGEM-T vector (Promega, Madison, Wis.) and sequenced. Sequencing was performed with a dye terminator cycle sequencing kit (PE Applied Biosystems) and the following vector-specific or sequence-specific primers: pGEM-T vector primers SP6 and/or T7 (Promega), 5′CpBisSeq 1 (EBV genome coordinate bp 10898; 5′-CATACACCCTAAACCAACC), 5′CpBisSeq 2 (EBV genome coordinate bp 11075; 5′-CCAATAAAAAAACTCAAA), 5′WpBisSeq 1 (EBV genome coordinate bp 13918; 5′-CCTATCACCAAACCTACCA), 5′WpBisSeq 2 (EBV genome coordinate bp 14115; 5′-AACCTAAACCTAAAACCCC), and 5′QpBisSeq (EBV genome coordinate bp 62181; 5′-CTAACCTAACTAAAAATAAAAC).

Cell lines and transfections.

The EBV-positive type I BL cell line Rael (33), the lymphoblastoid cell line JY (75), the type III BL cell line Clone-16 (20), and the EBV-negative BL cell line DG75 were all maintained in RPMI 1640 with 10% fetal calf serum. Rael cells were electroporated with 10 μg of either the CW1CAT reporter plasmid (Fig. 4A) (76) or an irrelevant luciferase plasmid. Electroporations were carried out at 0.23 kV and 960 μF as described by Schaefer et al. (65). RNA was isolated at 72 h posttransfection as described below.

FIG. 4.

Cp-initiated transcription from a transiently transfected reporter construct in a BL cell line exhibiting restricted EBV latency. Rael, an EBV-positive BL cell line that displays restricted (type I) latency, was transfected with 10 μg of either the CW1CAT reporter plasmid (Fig. 4A) or an irrelevant luciferase (control [cntl]) plasmid. Total RNA was isolated at 72 h posttransfection, and the presence of Cp- and Wp-initiated transcripts was assessed by S1 nuclease protection. RNA was hybridized with ³²P-labeled Cp, Wp, or β-actin S1 probes. RNA samples from the EBV-infected lymphoblastoid cell line (LCL) JY, which drives EBNA gene transcription from Cp, and from the BL cell line Clone-16 (Cl-16), which initiates EBNA gene transcription at Wp, served as positive controls for Cp and Wp activity. Identical results were obtained in two independent experiments.

RNA isolation and S1 nuclease protection assays.

Total RNA was isolated according to the standard guanidium thiocyanate-phenol method (11), and the presence of Cp- and Wp-initiated transcripts was analyzed by S1 nuclease protection as previously described (48, 76). RNA (15 μg) was hybridized overnight at 42°C with γ-³²P-labeled Cp, Wp, or β-actin S1 probe (48, 76). The following day, samples were digested with S1 nuclease (100 U per ml of buffer; Promega) and analyzed on a 10% polyacrylamide–7 M urea gel. RNA samples from the EBV-infected lymphoblastoid cell line JY, which drives EBNA gene transcription from Cp, and from the BL cell line Clone-16, in which EBNA gene transcription arises from Wp, served as positive controls for Cp and Wp activity.

Methylation cassette assay, Southern blots, and chloramphenicol acetyltransferase (CAT) assays.

The methylation cassette assay was performed as described previously (55, 71) and outlined in Fig. 5A. The CW1CAT construct (see Fig. 5A) was either methylated in vitro with SssI (CpG) methylase (New England Biolabs, Boston, Mass.) according to the company’s specifications or mock methylated. The mock- and SssI-treated plasmids were digested with the methylation-insensitive enzymes BamHI and HindIII for 5 h with at least 10 U of enzyme per μg of DNA. The unmethylated digested vector, as well as the mock- and SssI-methylated W inserts, were isolated on agarose gels and purified by using a Qiaex II gel extraction kit (Qiagen). Diagnostic digestion of the mock-methylated and SssI-methylated W inserts with the methylation-sensitive restriction endonuclease SmaI demonstrated that the methylated W insert was completely resistant to digestion, while the unmethylated insert could be completely digested (data not shown; the W insert contains six SmaI cleavage sites).

FIG. 5.

Hypermethylation of the region upstream of Wp in the CW1CAT reporter plasmid inhibits Cp-driven CAT activity. (A) Outline of the methylation cassette assay. A diagram of the CW1CAT plasmid, showing the locations of Cp, Wp, and relevant restriction endonuclease digestion sites (B, BamHI; H, HindIII) is shown. FR, family of repeats; DS, dyad symmetry element; C1 and C2, first two exons of Cp-initiated transcripts. (B) Southern blot demonstrating that methylated and unmethylated W fragments ligate into the CW1CAT vector with equal efficiency. The methylation cassette assay was performed as outlined in panel A. Diagnostic digestion of the W inserts with the methylation-sensitive restriction endonuclease SmaI demonstrated that the methylated W insert was completely resistant to digestion, while the unmethylated insert could be completely digested (data not shown; the W insert contains six SmaI cleavage sites). The ligations were carried out as described in Materials and Methods. Prior to transfection, a small aliquot of each ligation was removed, digested with either BglII/XbaI or EcoRI, and probed with the ³²P-labeled BglII/HindIII fragment isolated from the CW1CAT construct. CW1CAT plasmid digested with BglII/XbaI or EcoRI served as a positive control (cntl). The 2,243-bp band in the BglII/XbaI digest and the 7,438-bp band in the EcoRI digest both represent correctly ligated DNA. The 1,540-bp band in the EcoRI digest represents inserts ligated only to the HindIII site of the vector. PhosphorImager (Molecular Dy- namics) quantitation demonstrated that there was no apparent difference between the ligation efficiencies of methylated and unmethylated DNA (the additional bands represent other minor ligation species that make up less than 10% of the total DNA in the ligation, as determined by PhosphorImager quantitation). (C) CAT activity recovered from cells transfected with either the unmethylated W, methylated W, or vector control ligations. The ligations were transfected into Rael cells, and CAT assays were carried out at 72 h posttransfection. The unacetylated and monoacetylated species of [¹⁴C]chloramphenicol were quantitated by PhosphorImager analysis, and the percentage of acetylated chloramphenicol is given below each lane. The experiment shown is representative of four independent experiments, and the observed decrease in CAT activity upon methylation of the W insert ranged from 8- to 16-fold.

The following vectors and inserts were ligated overnight at 16°C (1 to 2 μg DNA per ligation, containing a 1:1 molar ratio of vector to insert): (i) unmethylated W insert plus BamHI/HindIII-digested CW1CAT vector; (ii) methylated W insert plus BamHI/HindIII-digested CW1CAT vector; and (iii) BamHI/HindIII-digested CW1CAT vector alone. Prior to transfection of the ligation reaction products into Rael cells, a small aliquot of each ligation mixture was removed for Southern blot analysis. The ligation reaction mixture was digested with either BglII/XbaI or EcoRI, and the digested fragments were fractionated on a 1% agarose gel. The DNA was transferred from the agarose gel to a Nytran (Schleicher & Schuell) membrane, and the blot was probed with a ³²P-labeled BglII/HindIII fragment isolated from the CW1CAT construct. Prehybridization and hybridization were performed as previously described (77). CW1CAT plasmid digested with BglII/XbaI or EcoRI served as a positive control.

The ligation reaction products were transfected into Rael cells as described above, and CAT assays were carried out at 72 h posttransfection according to standard methods (17). The unacetylated and monoacetylated species of [¹⁴C]chloramphenicol were quantitated by PhosphorImager (Molecular Dynamics) analysis, and the percentage of acetylated chloramphenicol was determined.

RESULTS

To assess whether methylation of the EBV genome plays a role in regulating viral gene expression in the long-term latency reservoir in vivo, we examined methylation of the regulatory regions upstream of the EBNA gene promoters Cp, Wp, and Qp (Fig. 2). B lymphocytes from healthy leukopheresis donors were screened by PCR for the presence of EBV (the frequency of EBV genome-positive B cells in healthy seropositive individuals ranges from 1 in 10⁴ to 10⁶ B cells [44]). Three EBV genome-positive donors (donors 8, 9 and 10) were chosen for further analysis. Importantly, previous characterization of EBV-infected cells in the blood of healthy seropositive individuals determined that (i) no lytic EBV replication could be detected in peripheral blood lymphocytes (14) and (ii) the EBV genome-positive cells exclusively fractionated to the resting memory B-cell population (CD19⁺, CD23⁻, B7⁻, immunoglobulin D negative, surface immunoglobulin positive) (2, 43, 44). Thus, these analyses suggest that the infected B cells in the blood are a phenotypically homogeneous population. Because of the low frequency of EBV genome-positive B cells, bisulfite PCR analysis was used to assess the presence of methylated CpGs. Bisulfite PCR involves treating genomic DNA with sodium metabisulfite, which converts unmethylated cytosines to uracils but does not react with methylated cytosines (12, 40, 51). The bisulfite-modified DNA can then be PCR amplified (uracils are read as thymines by Taq polymerase), and the amplified product is cloned and sequenced.

Wp is hypermethylated in the long-term latency reservoir in vivo.

Analysis of 18 bisulfite PCR clones (each derived from an independent PCR), corresponding to the region encompassing Wp, revealed hypermethylation throughout the entire sequence (Fig. 2B; an average of 87% of the CpGs present in this region were methylated in each clone analyzed). It should be noted that each clone analyzed was generated from an independent PCR and therefore reflects amplification of independent viral genomes. The amplified Wp PCR product contained the Wp TATA and CCAAT boxes as well as an enhancer that upregulates transcription initiation from both Wp and Cp (48, 54). We also had a limited amount of B-cell DNA from a fourth donor, donor 7; two Wp bisulfite PCR clones obtained from this donor revealed extensive methylation around Wp (data not shown). Thus, the region around Wp was invariably hypermethylated in all donors examined (Fig. 2B and data not shown). Wp is located within the major internal repeat (IR1) of EBV (Fig. 1), and so each viral genome contains an estimated 8 to 10 copies of Wp. Since we have analyzed a relatively large number of independently generated bisulfite PCR clones and have not observed any variation in the level of methylation, this analysis suggests that all copies of Wp are hypermethylated.

Cp is sparsely to moderately methylated in the long-term latency reservoir.

Analysis of the region upstream of Cp revealed a variable pattern of methylation (Fig. 2A). Most of the clones isolated from donors 8 and 9 were moderately methylated, while clones isolated from donor 10 were relatively sparsely methylated. However, there was evidence of CpG methylation around Cp in all clones analyzed. In addition, we obtained four Cp bisulfite PCR clones from the small amount of DNA that we had from a fourth donor, donor 7 (mentioned above); this donor resembled donor 10 in that the four Cp bisulfite PCR clones analyzed showed a low level of methylation in the region upstream of Cp (data not shown). The region just upstream of the Cp transcription initiation site was the area most consistently and heavily methylated in the clones analyzed (Fig. 2A). In contrast, the region containing the EBNA 2-dependent enhancer (26, 49, 68, 74), which we have previously shown is critical for Cp-initiated transcription in EBV-immortalized lymphoblasts (77), was unmethylated in a majority of clones analyzed (Fig. 2A).

The observation that the EBNA 2-dependent enhancer is not heavily methylated is notable, since the only published analysis of EBV genome methylation in normal seropositive individuals focused on methylation of this enhancer (56). Consistent with our analysis, these investigators reported the presence of both methylated and unmethylated forms of the Cp EBNA 2-dependent enhancer (56). However, because the previous analysis did not assess methylation downstream of the EBNA 2-dependent enhancer or around Wp, these researchers could not determine the significance of unmethylated clones. These investigators interpreted the unmethylated clones as arising from EBV-infected B cells that had entered the lytic cycle (reactivation) and were replicating the viral genome, since rapidly replicating viral DNA would be expected to be hypomethylated. However, analyses of EBV gene expression in the peripheral B cells of healthy seropositive individuals have failed to detect any evidence of viral reactivation (2, 14), arguing against this possibility. In addition, as shown here, all clones analyzed exhibited methylation downstream of the EBNA 2-dependent enhancer and all EBV genomes analyzed were hypermethylated around Wp. The issue of whether the low to moderate level of methylation around Cp is sufficient to silence Cp transcription is addressed below.

Qp is hypomethylated in the long-term latency reservoir.

In contrast to the presence of methylated CpGs upstream of both Wp and Cp, analysis of Qp revealed a complete absence of methylated CpGs in the region extending from ca. −200 bp through the transcription start site (Fig. 2C; a single methylated CpG at bp 62282 was observed in one clone from donor 10). Importantly, every clone characterized was heavily methylated in the region upstream of Fp, a promoter active during the viral lytic cycle and quiescent during latent infection (34, 64). Thus, the absence of methylation around Qp does not reflect cloning rapidly dividing viral genomes (i.e., those arising from virus replication), since no methylation would be observed upstream of Fp in such clones. The tight boundary between methylated and unmethylated sequences suggests the presence of a cis element(s) that actively prevents encroachment of methylation into the Qp regulatory region.

Hypermethylation of sequences around Wp suppresses Cp-driven reporter gene activity.

Figure 3 summarizes the methylation status of the EBNA gene promoters in the long-term latency compartment. The analysis of viral genome methylation raises the question of whether the observed methylation around Cp and Wp is sufficient to prevent transcription from these promoters. In vitro methylation studies have shown that as little as 7% methylation of a reporter construct can lead to a 60 to 90% reduction in transcription, and higher levels of methylation almost completely abolish transcription, even in the presence of a strong transcriptional enhancer (8, 22, 32). These observations, combined with the results of in vivo reverse transcription-PCR (RT-PCR) studies demonstrating no Wp activity and little or no Cp activity in healthy donors (10, 50, 72), suggest that the extensive methylation around Wp in these donors is sufficient to repress Wp-initiated transcription, and even the moderate to low level of methylation around Cp most likely contributes to transcriptional repression of Cp. An intriguing possibility is that methylation around Wp reduces transcription from Cp, perhaps by preventing elongation of Cp-initiated transcripts through the BamHIW region (IR1 repeat) and/or by inhibiting transcription initiation from Cp (possibly by obscuring the shared Cp/Wp enhancer upstream of Wp). The ability of methylation to inhibit transcription elongation has been observed in Neurospora (60).

FIG. 3.

Summary of the methylation status of the EBNA gene promoters in the long-term latency reservoir in healthy seropositive individuals. The diagram schematically illustrates EBNA 1 gene transcription from Qp and also depicts EBNA 1 inhibition of Qp-initiated transcription. Notation is as for Fig. 1.

Using an EBV-positive BL cell line (Rael) that exhibits restricted type I EBV latency (33), we addressed whether Cp activity could be influenced by methylation of CpGs around Wp. Initially, we determined whether the necessary factors were present in the Rael BL cell line to drive transcription from Cp. Rael cells were transiently transfected with a reporter construct (CW1CAT [Fig. 5A]) containing both Cp and Wp (in their normal context within extensive surrounding viral sequences) or with an irrelevant control plasmid. S1 nuclease protection analysis using RNA isolated from the CW1CAT-transfected Rael cells demonstrated strong Cp-initiated transcription and the absence of detectable Wp-initiated transcripts (Fig. 4). However, no Cp- or Wp-initiated transcripts were detected in Rael cells transfected with the irrelevant control plasmid, demonstrating that the observed transcription arose from the transfected reporter construct (Fig. 4). These results indicate that the Rael cell line contains the necessary cellular and viral factors (i.e., EBNA1) to drive Cp-initiated transcription; in addition, it appears that Cp is exclusively used to drive reporter gene activity in this construct. This analysis confirms previous studies demonstrating the presence of the necessary factors to drive transcription from Cp in cell lines exhibiting restricted EBV latency (25, 57, 63, 65) and is consistent with the induction of Cp/Wp activity upon inhibiting methylation of the viral genome (1, 25, 39, 57).

To assess the impact of methylating sequences upstream of Wp on Cp-driven reporter gene activity, we used a methylation cassette assay (Fig. 5A) (55, 71). This assay utilizes the bacterial methylase SssI (CpG methylase; New England BioLabs) to methylate the CpGs present in a DNA insert from the reporter construct (Fig. 5A). The CW1CAT reporter construct was either mock methylated or methylated with SssI methylase (Fig. 5A). The W insert (extending from the junction of the EBV BamHI C and W fragments through to the 5′ end of the first W1 exon) was removed by digestion with BamHI and HindIII. Notably, this region extends several hundred base pairs upstream of the region that we analyzed for methylation in seropositive individuals (to bp 13215 of the EBV genome, compared to bp 14015 [Fig. 2]). However, based on the analysis of methylation of the viral genome in the Rael BL cell line (39), we would predict that methylation of the viral genome extends through the entire IR1 motif. The recovered inserts (methylated or unmethylated W inserts) were then ligated back into the BamHI and HindIII digested, unmethylated CW1CAT vector. Ligation efficiency was assessed by Southern analysis (Fig. 5B), and CAT reporter gene activity was determined by transfecting the ligation reaction products into the Rael cell line (Fig. 5C). Importantly, methylation has previously been shown to have no effect on transfection efficiency (32). In that study (32), it was demonstrated that levels of transcription from methylated and unmethylated reporter constructs were equivalent at 1 h posttransfection, whereas transcription from the methylated construct was greatly reduced by 16 h, and loss of transcription from the methylated construct correlated with loss of DNase hypersensitivity. These results indicate that assembly of a repressive chromatin structure on methylated DNA requires time and that the transfection efficiencies of methylated and unmethylated reporter plasmids are equivalent.

Southern blot analysis (Fig. 5B) demonstrated that both the mock-methylated and unmethylated inserts ligated into the vector with nearly identical efficiency (Fig. 5B). The presence of the 2,243-bp BglII/XbaI fragment is diagnostic of ligation of the insert at the HindIII site, while the presence of the 7,438-bp EcoRI fragment is diagnostic of complete ligation of the insert into the BamHI and HindIII sites in the vector (Fig. 5A and B). Quantitation of the intensity of the diagnostic fragments on the Southern blot indicated that >60% of both the mock-methylated and SssI-methylated inserts were correctly ligated into the vector. Analysis of the activities of the mock-methylated and Wp-methylated reporter constructs demonstrated that methylation of the sequences around Wp led to a >8-fold decrease in reporter gene activity (Fig. 5C). Thus, methylating a discrete region around Wp was sufficient to dramatically reduce Cp-driven reporter gene activity. It is likely that methylation of additional sequence (e.g., around Cp and within the BamHI W repeats) in the viral genome, as observed in the EBV genomes present in the long-term latency reservoir, results in an even greater inhibition of Cp-driven EBNA gene transcription. Notably, the observed impact of methylation on Cp-initiated reporter gene activity are likely to be position dependent based on previous results of Robertson and Ambinder (55) that demonstrated methylation of some regions upstream of Cp had little or no impact on promoter activity.

DISCUSSION

Here we have shown that the absence of expression of EBV antigens associated with B-cell immortalization in the long-term latency reservoir (10, 50, 72) correlates with methylation of two viral promoters (Cp and Wp) involved in driving transcription of these viral genes. Furthermore, the viral promoter Qp, involved in driving exclusive transcription of the EBNA 1 gene in EBV-associated tumors, is protected from methylation in the long-term latency reservoir. Figure 3 summarizes the proposed connection between EBV genome methylation and latency program utilization. Since EBNA 1 is the only viral antigen required for maintenance of the viral episome, as well as the only EBNA that fails to elicit a CD8 T-cell response, these findings imply that discrete methylation of the EBV genome (i) may drive the switch from immortalizing latency to restricted latency, (ii) may facilitate evasion of the potent host cellular immune response against specific EBV antigens expressed in virus-immortalized B cells, and (iii) ensures sufficient expression of EBNA 1 to allow propagation of the viral genome.

The extensive methylation detected around Wp in all donors (Fig. 2B) is consistent with the fact that Wp-initiated transcripts have not been observed in any of the nested RT-PCR studies of healthy donors (10, 50, 72). Such a high level of Wp methylation would most likely suffice to shut down transcription from Wp, as described in Results. Methylation around Wp could also function to repress transcription from Cp (as demonstrated in the methylation cassette assay in Fig. 5), possibly by inhibiting transcription elongation and/or by blocking the shared Cp/Wp enhancer located upstream of Wp. The fact that we failed to observe any unmethylated Wp clones suggests that all copies of Wp are methylated. Tissue culture studies of recombinant EBV containing a tag at the first Wp have shown that Wp activity comes from the first copy (77), and so it is formally possible that the first copy is unmethylated and the remaining copies are methylated. However, even if this is the case, the results of the methylation cassette assay imply that methylation in the downstream copies of Wp would inhibit transcription from the unmethylated first copy. The low residual CAT activity observed with the transfected methylated reporter construct (Fig. 5C) could represent either (i) a low level of Cp-initiated transcripts that traverse the methylated W region (note that Cp was not methylated in this assay, whereas it is moderately methylated in vivo, and only a single copy of Wp was present) and/or (ii) transcription of the methylated reporter construct prior to formation of a repressive chromatin structure (chromatin assembly on methylated DNA requires approximately 16 h [32]).

Cp-initiated transcripts have been detected by nested RT-PCR in the peripheral blood of some healthy seropositive individuals (10, 50), although transcription of the EBNA 2 gene was not observed (10, 50, 72). These results could reflect transcripts arising from Cp that are truncated, terminating within the BamHI W repeats. This possibility is consistent with the variable level of methylation around Cp observed here (Fig. 2A), coupled with an inability to transcribe through the hypermethylated W repeats (Fig. 2B and 5C). It is conceivable that the Cp-initiated transcripts detected in the in vivo RT-PCR studies (10, 50) arose from EBV genomes with a lower level of methylation, such as that observed around Cp in donor 10 and in donor 8 clone 8 (Fig. 2).

These findings lead to speculation about what triggers methylation of the EBV genome. Alterations in the methylation of cellular genes may occur during the differentiation of primary B cells into memory B cells, and EBV may be targeted during this process, through either direct infection of a germinal center B-cell population or EBV-driven differentiation of naive B cells into memory B cells (2). Another possibility is that EBV infection elicits viral genome methylation, perhaps as a host defense response. These two possibilities are not mutually exclusive. Interestingly, repeated sequences and secondary structures appear to be favored substrates for methyltransferases (4, 6). Thus, methylation of the EBV genome may initially be targeted to the BamHI W repeat region (IR1), which also contains extensive potential secondary structure (29), and then spread to other regions of the viral genome. Notably, expansion of methylation from integrated retroviruses into flanking cellular sequences has been observed (23). The identity of the methyltransferase that methylates the EBV genome is unknown, although it is likely a de novo methyltransferase; two mammalian de novo methyltransferases have recently been cloned (47).

We have also demonstrated that Qp remains unmethylated in vivo, as has been observed in multiple EBV-positive cell lines and tumors (65, 71). The sharp boundary between methylated and unmethylated DNA implies that Qp may be protected from methylation. This hypothesis is consistent with previous observations suggesting that the promoter of the aprt housekeeping gene, which is rich in CpGs, is protected from methylation by the presence of Sp1 binding sites at the 5′ boundary between methylated and unmethylated cellular DNA sequences (9, 38). Notably, Qp shares several common features with the promoters of housekeeping genes: (i) the Qp 5′ regulatory region is CpG rich; (ii) Qp is a TATA-less promoter, requiring minimal upstream sequences for activity (62); and (iii) Qp is active in a wide variety of cell types (62, 65). An alternative possibility is that Qp can become methylated and then targeted for demethylation; a demethylating enzyme was recently cloned (7). Thus, based on the analysis of EBV genome methylation and functional data (62, 65), we hypothesize that Qp serves as a default promoter that is protected from inactivation by host cell-directed methylation, thereby ensuring that sufficient levels of EBNA 1 (the only viral antigen required for maintenance of the viral episome) can be produced, even when Cp and Wp have been inactivated by methylation of the viral genome.

Consistent with this hypothesis, Qp-initiated EBNA 1 transcripts have been detected in peripheral B cells in some healthy donors (72). The variable detection of Qp-driven EBNA 1 gene transcripts in the long-term latency compartment likely reflects the fact that EBNA 1 negatively regulates Qp-initiated transcription. Thus, in the periphery where the memory B cells are quiescent (not cycling), sufficient levels of EBNA 1 protein may be expressed to repress transcription from Qp. In addition, there is evidence that Qp may be cell cycle regulated and thus inactive in noncycling cells (13, 69). As such, the sporadic detection of Qp-initiated transcripts may reflect the presence of memory B cells that have recently been stimulated to transiently proliferate.

Notably, sensitivity of Cp and Wp to host cell-directed methylation, along with protection of Qp from methylation, has been observed in most EBV-positive tumors. This argues that viral genome methylation in EBV-positive tumors reflects a normal component of EBV infection of immunocompetent hosts, as opposed to an oncogenic event limited to the development of EBV-associated tumors. The observation that the EBV genome is differentially methylated in the long-term latency reservoir raises the possibility that changes in host cell-directed methylation occur during the development of B-cell memory. Additionally, for those EBV-positive B-cell malignancies that develop in immunocompetent individuals, this finding suggests that such tumors arise through the same developmental pathway that leads to the establishment of the EBV long-term latency reservoir (2).

Does methylation of the viral genome drive the transition from immortalizing latency to restricted viral latency, or is restricted latency established by another mechanism? Clearly the current analysis cannot distinguish between these possibilities. However, it is notable that (i) in tumor cells exhibiting restricted EBV latency, reversal of viral genome methylation leads to transcription from Cp and Wp (1, 25, 39, 57); and (ii) transiently transfected Cp/Wp-driven reporter constructs display activity in all type I/II latently infected cell lines examined to date (25, 57, 63, 65), indicating that the necessary cellular and viral factors are present for Cp/Wp activity. Thus, it was this consistent relationship between methylation and inactivation of Cp/Wp that provided the basis for examining viral genome methylation in the long-term latency reservoir in vivo. However, in the absence of further data on the generation of the long-term latency reservoir, it will not be possible to critically assess the role of methylation in establishing the restricted latency program.

Finally, does methylation of the viral genome allow EBV to persist in the immunocompetent host, or does it reflect an effective host defense strategy? Analysis of EBV infection in immunocompromised patients and the effectiveness of anti-EBV cytotoxic T-lymphocyte therapy (59), coupled with the effective clearance of the immortalizing form of EBV latency in healthy individuals, argues that in the absence of a restricted viral latency program(s), the host would clear EBV infection. In addition, spontaneous outgrowth assays, in which B cells from a healthy seropositive donor are cultured with cord blood cells from an opposite-sex donor (52), have demonstrated that the EBV genomes in healthy seropositive donors are functional (as judged by their ability to reactivate and infect the surrounding cord blood cells) (2). Furthermore, analysis of the CpG dinucleotide frequency in the EBV genome (21, 30, 31) provides further evidence that methylation of the viral genome has not prevented efficient passage of this ubiquitous pathogen. Thus, unlike endogenous retroviruses which are inactivated by host-directed methylation (18, 24, 73), methylation of the EBV genome in the long-term latency reservoir does not prevent the persistence and passage of EBV. Taken together, these findings strongly suggest that EBV utilizes the host cell methylation system to its advantage, to regulate viral gene expression and establish lifelong latency in its host.