INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

Epstein-Barr virus (EBV), a human gammaherpesvirus, is largely B lymphotropic and possesses a unique set of latent cycle genes whose coordinate expression can drive the proliferation of latently infected B cells. This process can be studied in vitro, where experimental infection of resting B cells leads to the outgrowth of virus-transformed lymphoblastoid cell lines (LCLs) expressing the full range of EBV latent proteins; these include the nuclear antigens EBNA1, -2, -3A, -3B, -3C, and -LP and the latent membrane proteins LMP1, -2A, and -2B (25, 42). Immediately postinfection, viral transcription initiates from the latent cycle promoter Wp. The long primary transcripts thus produced are capable of generating all six EBNA mRNAs but appear to be preferentially, though not exclusively, processed to mRNAs encoding the EBNA2 and EBNA-LP proteins (2, 13, 62, 63). Their appearance leads to activation of an alternative upstream promoter, Cp, from which the full complement of EBNA proteins is expressed, as well as to activation of the more distant LMP promoters (1, 54, 60, 61, 67), thereby completing the full range of latent protein expression. The activation of Cp is followed by a gradual waning of Wp transcription such that Cp is dominant over Wp in most established LCLs (4, 63).

Virus-driven B-cell growth transformation is also observed in vivo during primary EBV infection. Thus, Wp, Cp, and LMP transcripts are detectable in the circulating B-cell pool of infectious mononucleosis (IM) patients (58). However, after resolution of the acute infection and establishment of a lifelong virus carrier state, virus-infected cells in the blood are only detectable within the resting memory B-cell pool (5, 32) and show a different program of viral transcription; in particular, Wp, Cp, and most if not all of the LMP promoters (with the possible exception of LMP2A) appear to be silent (41, 58). The mechanisms of promoter regulation which effect this switch to a more restricted form of virus latency are still not understood. However, some interesting potential clues have come from the study of EBV genome-positive malignancies, such as Burkitt's lymphoma (BL) and nasopharyngeal carcinoma, which also show more limited patterns of latent protein expression than do LCLs (42). In particular, Wp and Cp are silent in such tumors, and the only detectable EBNA, EBNA1, is expressed from a downstream EBNA1-specific promoter, Qp (48-50). Digestion of viral DNA from these tumors, using restriction enzymes that are either sensitive or insensitive to the presence of a methylated cytosine in the target sequence, indicated that several regions of the genome were more heavily methylated in tumor cells than in LCLs (3, 4, 11, 20, 22, 31, 51). These regions included the BamHI W and C fragments, within which Wp and Cp are situated, but not the region around Qp. Given the increasing evidence for methylation at CpG dinucleotides as a general mechanism of promoter silencing in eukaryotic cells (10), this raised the possibility that methylation of EBV latent promoters may be important in maintaining the restricted form of latency found in such tumors. This idea was indeed supported by earlier evidence that exposure of BL cell lines to 5' azacytidine, an inhibitor of DNA methylation, was capable of rescuing EBNA2 and LMP1 expression in a proportion of cells (16, 30).

With the development of the more sensitive technique of bisulfite genomic sequencing, it is possible to examine the methylation statuses of all CpG dinucleotides within any selected region of DNA rather than only those present within restriction enzyme sites (14). In this context, it is known that Cp can be activated by the binding of EBNA1 to the upstream oriP element (40, 53) and by the binding of two cellular factors, CBF2 and the EBNA2-interacting protein RBP-J/CBF1, to a more promoter-proximal EBNA2 response element (17, 19, 59, 66). While oriP remains unmethylated in BL and nasopharyngeal carcinoma tumor cells (12), Robertson et al. observed hypermethylation of a region of Cp encompassing the EBNA2 response element (46) and, in gel shift assays, showed that methylation of a particular CpG within that element abrogated CBF2 binding (45). This further implied that methylation is important in maintaining Cp in an inactive state in tumor cells. Whether methylation is coincident with, and therefore potentially involved in, the initial silencing of Cp in such cells remains an open question.

The present work describes similar studies in the context of Wp and takes advantage of two features of this promoter. One is the fact that promoter silencing can be followed in real time during the process of virus-induced B-cell transformation (63). The other is the recent identification of upstream regulatory sequences (7) which in reporter assays appear to be critical for Wp activity. These include a YY1 site within a region of the promoter responsible for its low baseline activity in a variety of cell lineages and, more importantly, binding sites for CREB family members (27), for RFX family members, and for the B-cell-specific activator protein BSAP/Pax5 (57a) within a region responsible for the promoter's high B-cell-specific activity. We noted that several of these sites contained CpG dinucleotides and were therefore potentially susceptible to the effect of methylation.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Tumor biopsy specimens, established cell lines, and recently-infected cells. The tumor biopsy specimens Sav, Ava, Isa, and Ali were from patients with endemic EBV-positive BL from East Africa (18). The cell lines tested included an EBV-negative sporadic BL line, DG75; the EBV-positive endemic BL cell lines Rael, Jad, and Eze (in which EBV antigen expression is restricted to EBNA1); the EBV-producing marmoset LCL B95.8; three long-established human LCLs, XMW, WBW, and JY; and the human T-leukemic cell line Jurkat. All cell lines were maintained in RPMI 1640 supplemented with 10% fetal calf serum and 2 mM glutamine. Before the cell lines were sampled for bisulfite sequencing analysis, the cells were grown for at least 2 weeks in the presence of 200 µM acyclovir in order to block any EBV genome replication in cells spontaneously entering the lytic cycle (28).

RNA extraction and RT-PCR analysis. RNA was prepared from cell pellets with RNAzol B extraction reagent (Cinna-Biotex) and subjected to reverse transcription (RT)-PCR analysis. In brief, cDNA synthesis was carried out on 1 µg of RNA using EBV-specific 3' primers and avian myeloblastosis virus reverse transcriptase (Roche) according to the manufacturer's instructions. The cDNA was then amplified with primer sets specific for Wp-initiated transcripts, Cp-initiated transcripts, BamHI Y3/U/K-spliced EBNA transcripts initiating from Wp or Cp, BamHI Q/U/K-spliced EBNA1 transcripts initiating from Qp or the immediately upstream lytic cycle promoter Fp (36, 50), and BamHI FQ/U/K-spliced EBNA1 transcripts initiating from Fp, and then the PCR products were detected on Southern blots with radiolabeled probes as described previously (36, 58). Quantitation was carried out using a Molecular Dynamics PhosphorImager.

Bisulfite genomic sequencing. Genomic DNA prepared by standard methods was subjected to bisulfite genomic sequencing (14). Briefly, 10-µg aliquots of DNA were digested with EcoRI, denatured in 0.2 M sodium hydroxide for 10 min at room temperature, neutralized in 0.3 M sodium acetate, and ethanol precipitated. The DNA pellet was resuspended in 1.2 ml of freshly made 3.1 M sodium bisulfite-0.5 mM hydroquinone (pH 5), overlaid with mineral oil, and incubated for 20 h at 50°C. The bisulfite-modified DNA was then purified using Qiaquick columns (Qiagen), denatured, neutralized, precipitated, and resuspended in 50 µl of sterile deionized water. For each sample, 2-µl aliquots of bisulfite-modified and unmodified DNA were amplified in strand-specific PCRs with primers specific for the regulatory regions of the Cp and Wp promoters as shown in Table 1. Bisulfite-modified DNA was amplified for 30 cycles of 94°C for 30 s, 50°C for 60 s, and 72°C for 90 s; unmodified DNA was amplified for 30 cycles of 94°C for 30 s, 60°C for 60 s, and 72°C for 90 s. The cell line samples were amplified with one round of PCR using the inner set of primers, whereas for some of the tumor biopsy and IM patient samples, a nested-PCR approach was required to obtain sufficient PCR product for cloning. In these cases, DNA was amplified with the outer primers, and then one-fifth of this first-round PCR product was amplified in a second round with the appropriate inner primers. The PCR products were gel purified and cloned into Escherichia coli XL-1 Blue using the pGEM-T Easy vector system I (Promega) according to the manufacturer's instructions. DNA was made from individual bacterial colonies using the Wizard plus SV miniprep kit (Promega) and sequenced using the universal primer 5'-GTAAAACGACGGCCAGT (Amersham-Pharmacia).

Bandshift assays. The preparation of nuclear extracts, the sequences of wild-type and mutant oligonucleotides used as bandshift probes, and the in vitro binding assay conditions have been described previously (7, 27; Tierney et al., submitted). Oligonucleotides incorporating 5-methylcytosine residues at specific CpG sites were synthesized by Alta Biosciences (University of Birmingham, Birmingham, United Kingdom).

Methylation cassette assays. Methylation cassette assays (43) were used to determine the effect of methylation of specific regulatory regions on Wp activity. Four types of cassette assay were carried out in which either the whole Wp regulatory region from 440 to +173 relative to the transcription start site, or the smaller regions 440 to 264, 264 to 170, and 170 to 16, were methylated in the context of an otherwise-unmethylated luciferase reporter vector. For each cassette assay, duplicate 20-µg aliquots of Wp440[NcoI]-GL2, a derivative of Wp 440-GL2 (7) in which an NcoI site had been introduced at 173 to 168 by site-directed mutagenesis, were digested overnight with 30 U of the appropriate enzymes, phenol-chloroform extracted, precipitated, and resuspended in 20 µl of Tris-EDTA buffer. One aliquot was methylated overnight at 37°C in a 30-µl reaction mixture with 10 U of SssI CpG methylase (New England Biolabs), 1× buffer 2 (New England Biolabs), and 160 µM S-adenosyl-methionine. The other aliquot (mock-methylated control) was treated in the same way except that the CpG methylase enzyme was omitted. The methylated and unmethylated cassette fragments and unmethylated vector fragments were gel purified using the Qiaquick gel extraction kit and eluted in 40 µl of sterile deionized water. The efficiency of the methylation reactions was confirmed by digestion of 1 µl of the methylated and mock-methylated vector fragments with 10 U of the methylation-sensitive enzyme SalI (Roche). The methylated or mock-methylated cassette fragments (10 µl) were then ligated back into 1 µl of the unmethylated vector in a 20-µl reaction mixture with 10 U of T4 DNA ligase (Roche). The efficiency of the ligation reaction was assessed by analysis of 1 µl of the ligation reaction mixture on a 1.5% agarose gel. The remainder of the ligation reaction mixtures were then cotransfected into 25 × 10⁶ DG75 or Jurkat cells with 1 µg of -galactosidase reporter plasmid using the DEAE-dextran method (47). The luciferase and -galactosidase activities were measured as described previously (7), with the -galactosidase values being used to normalize the luciferase data for variations in transfection efficiency.

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

Methylation status of Wp and Cp in BL and LCL cells in relation to promoter activity. Prior to methylation analysis, we first sought to establish a panel of reference EBV-positive cells in which the activities of Wp and Cp were known. As representatives of cells in which both Wp and Cp were likely to be silent, we selected four endemic BL biopsy cell preparations (from the tumors Sav, Ava, Isa, and Ali) and three endemic BL cell lines; one was the long-established cell line Rael, which has been used in several earlier studies of this kind (11, 31, 51), and the other two (Jad and Eze) were recently established lines in early passage with the classical group I phenotype and restriction of EBV latent protein expression to EBNA1. Preparations of RNA from these cell populations and, as a reference, from normal B cells 2 days following EBV infection in vitro were amplified by RT-PCR using primers specific (i) for Wp-initiated transcripts, (ii) for Cp-initiated transcripts, (iii) for BamHI Y3/U/K-spliced EBNA1 transcripts initiating from Wp or Cp, (iv) for BamHI Q/U/K-spliced EBNA1 transcripts initiating from Qp or from the immediately upstream early lytic cycle promoter Fp, and (v) for the BamHI FQ/U/K-spliced EBNA1 transcript initiating from Fp (36, 50, 58); the products were then run out on a gel and blotted using the appropriate transcript-specific oligonucleotide probe. As shown by the representative samples in Fig. 1, this confirmed that both Wp and Cp were silent and that Qp was active in BL biopsy cells and in BL cell lines. Note that such BL cell populations showed no spontaneous entry of cells into the virus lytic cycle, and accordingly, all Q/U/K-spliced transcripts reflected Qp activity with no detectable contribution from the early lytic cycle promoter Fp. Figure 1 also shows the corresponding results from recently infected B cells and from the four long-established LCLs (including B95.8) selected for this work. While recently infected cells showed only Wp transcription, all four long-established lines displayed the classical pattern of preferential Cp usage with little or no detectable Wp activity. Cp usage was further reflected by the presence of Y3/U/K-spliced transcripts, whereas the very low levels of Q/U/K-spliced transcription in these LCLs (most obvious in B95.8) could be explained by lytic cycle-associated Fp activity.

View larger version (41K): [in this window] [in a new window]   FIG. 1.   Analysis of EBV transcription in a representative BL biopsy specimen, SAV; a recently established BL cell line, EZE; and four long-established LCLs, WBW, XMW, JY, and B95.8. Corresponding data for experimentally infected B cells (2 days postinfection) are included as a reference. Shown are the results of RT-PCR analysis using primer-probe combinations specific for Wp-initiated transcripts, Cp-initiated transcripts, Wp- or Cp-initiated Y3/U/K-spliced EBNA1 transcripts, Fp- or Qp-initiated Q/U/K-spliced EBNA1 transcripts, and Fp-initiated FQ/U/K-spliced EBNA1 transcripts. Note that Wp, Cp, and Qp are latent cycle promoters and Fp is a lytic cycle promoter.

View larger version (65K): [in this window] [in a new window]   FIG. 2.   Methylation status of Cp and Wp regulatory regions in EBV-positive cells. (A) Diagram showing the main regulatory elements of Cp and Wp and the relative positions of CpG dinucleotides analyzed. (B, C, and D) Results of bisulfite sequencing of these regions in BL biopsy specimens (B), in BL cell lines (C), and in long-established LCLs (D). Several clones of the relevant PCR products were sequenced for each cell sample, and the individual CpG dinucleotides were identified as either methylated (+, shaded) or nonmethylated (); the overall percentage of methylated CpGs per sample is shown.

Effect of methylation on binding of transcription factors to Wp sequences. Methylation may inhibit transcription via a number of mechanisms, either by directly blocking the binding of transcription factors to DNA or through the mediation of methyl-CpG-binding proteins, which recruit histone deacetylases to the DNA, leading to remodeling of the chromatin into an inactive configuration (10). We therefore used recently developed bandshift assays (7, 27, 57a) to determine whether methylation of the relevant Wp oligonucleotide sequences affected transcription factor binding for each of the known sites in Wp. This was determined using the methylated derivatives as competitors of factor binding to a labeled wild-type probe; in each case the methylated sequence was compared with the wild-type competitor sequence and with a mutant sequence known to have lost binding activity.

View larger version (34K): [in this window] [in a new window]   FIG. 3.   Methylation sensitivity of transcription factor binding to Wp sequences in band shift assays. Shown are the results obtained for YY1 binding to the 308 to 279 probe (A), BSAP binding to the 242 to 215 probe (B), RFX binding to the 140 to 99 probe (C), BSAP binding to the 115 to 86 probe (D), and CREB binding to the 102 to 77 probe (E). In each case, the assays included a wild-type competitor, a mutant competitor sequence known to have lost transcription factor binding (7, 27, 57a), and derivatives of the wild-type sequence methylated at the CpG sites shown (). The lanes showing assays conducted with labeled probes in the absence of the DG75 cell nuclear extract served as controls.

Effect of methylation on the activities of Wp reporter constructs. The next set of experiments used a methylation cassette approach to examine the effect on promoter activity of methylating specific regions of Wp in a luciferase reporter construct carrying Wp sequences 440 to +173 relative to the transcription start site. All assays were conducted by transient transfection into a representative B-cell line, DG75, in which Wp is fully active, and into a representative non-B-cell line, Jurkat, in which Wp activity is regularly 10- to 20-fold lower (7). The results in Fig. 4 show the activities of the methylated constructs, in each case relative to that of a mock-methylated control construct tested in parallel in the relevant cell background.

View larger version (53K): [in this window] [in a new window]   FIG. 4.   Cassette replacement experiments to determine the effect of methylating individual regions of Wp sequence in the context of a Wp reporter. The methylated regions were nucleotides 440 to +173 (A), 440 to 264 (B), 264 to 170 (C), and 170 to 16 (D), in each case relative to the Wp transcription start site. The results are shown as histograms of mock-methylated (MOCK; set at 100% activity) versus cassette-methylated (METH.) reporter activities observed in the DG75 B-cell line and in the Jurkat T-cell line.

Methylation status of Wp and Cp during virus-induced B-cell transformation in vitro. Given the earlier evidence of Wp methylation in established Cp-using LCLs (Fig. 2D), we next carried out a time course study of Wp and Cp methylation status in relation to the usage of these promoters over a 4-week period following experimental infection of resting human B cells in vitro. Briefly, T-cell-depleted PBMCs from adult donors were exposed to a B95.8 virus preparation and then cultured in the presence of 200 µM acyclovir, and cells were harvested for DNA and RNA analysis at days 2, 7, 11, 15, 18, 21, and 28 of culture. In such experiments, foci of lymphoblastoid cells were apparent in infected (but not in uninfected control) cultures within the first week postinfection, and the first subculture could be made by days 7 to 10; thereafter, the emerging LCL could be subcultured 1:1 every 4 days.

View larger version (41K): [in this window] [in a new window]   FIG. 5.   Analysis of Wp and Cp transcription in recently-infected B cells. Gels showing Wp- and Cp-specific RT-PCR signals, in each case obtained by amplifying neat (N) and 10 (1)-, 100 (2)-, and 1,000 (3)-fold dilutions of W2-primed cDNA. Results are shown for B cells 2, 7, 11, 15, 18, 21, and 28 days postinfection, as well as for the B95.8 LCL and Rael-BL lines as controls. Shown below the gel are the quantitative results of signal intensity phosphorimage analysis.

View larger version (63K): [in this window] [in a new window]   FIG. 6.   Methylation status of Cp and Wp regulatory regions in B cells 4 to 21 days following B95.8 EBV infection. (A) Diagram showing the main regulatory elements of Cp and Wp and the relative positions of CpG dinucleotides analyzed. (B) Representative sequencing gels of bisulfite-modified DNA for Cp in Rael BL cells versus B cells at 21 days postinfection (p.i.) and for Wp in B cells at 2, 7, 11, 15, 18, and 21 days p.i. The diagrams on the left and right of the gels indicate the expected positions of methylated cytosine in the C sequencing track. (C) Summarized results of bisulfite sequencing of the Cp and Wp regions presented as in Fig. 2.

Methylation status of Wp and Cp in B cells during primary infection in vivo. Previous studies have shown that during primary infection in vivo, as seen in acute IM patients, both Wp-initiated and Cp-initiated transcripts can be amplified from latently infected B cells in the blood (58), consistent with these cells being in the process of virus-driven growth transformation. We used bisulfite genomic sequencing to screen cryopreserved PBMC preparations from seven such IM patients, each bled within 4 to 13 days of the onset of clinical symptoms, and the results are summarized in Fig. 7. In six of seven patients we found that Cp sequences were completely unmethylated, whereas there was almost complete methylation at Cp in the remaining case. Importantly, however, there was evidence of Wp methylation in every case, and in six of seven patients this involved 60% of the CpG sites analyzed from each patient.

View larger version (65K): [in this window] [in a new window]   FIG. 7.   Methylation status of Cp and Wp regulatory regions in EBV-positive cells from the blood of seven individual IM patients. The data are presented as in Fig. 2.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

Several earlier studies using methylation-sensitive restriction enzymes on EBV-positive tumors and established cell lines have shown a correlation between viral genome methylation in certain latent promoter regions and restricted patterns of latent gene expression (3, 11, 20, 22, 31, 51). These studies could not address whether methylation had occurred at the time of promoter silencing, in which case it might be a primary cause of transcriptional down-regulation, or had accrued later (26) as a means of maintaining the already-silenced state. In the present study, we have followed the methylation status of Wp during its natural down-regulation in vitro, as recently infected cells switch from Wp to Cp usage, and have focused on regions of Wp now known to contain key transcription factor binding elements. We chose the method of bisulfite genomic sequencing, since this allows all CpGs within the amplified regions to be analyzed rather than only those within restriction enzyme sites. Recent studies using this approach on tumor biopsies and cell lines have given a much more precise picture of CpG methylation status around the Qp, Cp, and LMP1 promoters (12, 46, 51, 56, 57) and have shown, for instance, that Cp silencing is not always accompanied by the type of blanket methylation of CpG sites (including that within the CBF2 binding sequence) initially inferred from restriction enzyme analysis (46). This result was in fact confirmed in the present work when, as an internal control, we analyzed the methylation of Cp sequences amplified from BL biopsy specimens and BL-derived cell lines (Fig. 2B and C). More importantly, Fig. 2 provides the first bisulfite sequence data on Wp methylation in tumor material and cell lines. Here, we were interested to note that the 13 CpGs in the 400 bp upstream of the transcription factor start site were not only consistently methylated in BL biopsy specimens and BL cell lines but also frequently methylated in established LCLs (Fig. 2D).

Such a result would not have been anticipated from earlier analyses of the BamHI W fragment in LCLs using the methylation-sensitive HpaII enzyme and its methylation-resistant isoschizomer, MspI. In these reports, although there was some heterogeneity in methylation patterns, the BamHI W region appeared to be significantly less methylated in LCLs than in BL lines (3, 11, 31). However, the HpaII enzyme is not informative about CpGs in the important 400-bp regulatory region of the promoter, and the only study to use the potentially more informative HhaI enzyme focused entirely on Wp in BL cell lines (22). It is worth noting here that, in order to simplify the interpretation of our initial screening assays on established LCLs, we confined our attention to cell lines showing classical Cp usage in the absence of detectable Wp activity; the situation in other established LCLs, where Cp is dominant but accompanied by persistent low-level Wp transcription (64), remains to be determined. We also were careful to study the cell lines after 2 weeks of maintenance in acyclovir, thereby ensuring that the analysis was not complicated by the presence of unmethylated progeny genomes produced by lytically infected cells in the culture (55). Such precautions over the choice and pretreatment of LCLs were not taken in earlier studies of BamHI W methylation by restriction enzyme analysis.

Recent work with reporter constructs, building on the initial studies of Jansson et al. (22), has identified several sites within a 400-bp region of Wp where the binding of cell transcription factors contributes to promoter activation. These include a YY1 site in a lineage-independent region of the promoter that is responsible for much of the low baseline activity of Wp in non-B cells and sites for BSAP, RFX family members, and CREB/ATF proteins which lie within the B-cell-specific region and which are all required for high-level promoter function in B cells (7, 27, 57a). We sought to determine whether methylation of CpGs in these areas might affect factor binding in in vitro bandshift assays and/or the activity of Wp reporter constructs in methylation cassette experiments. Such approaches showed that YY1 binding was not grossly affected by methylation, though we cannot rule out slight changes in affinity that would not have been detected in these particular assays, whereas cassette methylation of the lineage-independent region in toto caused a significant reduction of Wp activity in both B- and non-B-cell lines. Possibly methylation is affecting transcriptional activity by another route in these circumstances, perhaps by the recruitment of methyl-CpG-binding proteins to the DNA and induction of a more condensed configuration (10, 23, 34). Of the transcription factors conferring B-cell-specific activity, the binding of RFX proteins to their site (which contains two CpGs) was also unaffected by methylation; this is consistent with the literature on RFX proteins, which in some cases, as here, bind independently of methylation status and in others are selective for methylated sequence motifs (15, 52, 65). In contrast, the binding of BSAP to both its sites and of CREB/ATF proteins to their site was in each case abrogated by methylation of one particular CpG in the recognition sequence; this appears to be the first report of the methylation sensitivity of BSAP binding, whereas that of CREB/ATF binding is well documented in other contexts (21, 29). In line with earlier results from Wp constructs where individual binding sites in the B-cell-specific region had been mutated (7, 57a), cassette methylation of an AvrII/NcoI fragment of Wp which selectively targeted the more distal BSAP site caused a B-cell-specific impairment of promoter activity while methylation of the region containing the adjacent RFX, BSAP, and CREB binding sites caused a dramatic inhibition that was preferentially exerted in B cells. Collectively, these findings strongly suggest that methylation of Wp across the 400-bp regulatory region, were it to occur in the context of a viral infection, could contribute to promoter silencing.

Further experiments provide evidence that such methylation does occur in recently infected B cells at a time which is coincident with Wp-to-Cp switching. We first analyzed the growth transformation process in vitro; note that the precise kinetics of transformation will depend upon the virus dose as well as the particular conditions of B-cell culture, and even using the concentrated virus preparations employed here, there will be a degree of asynchrony at the single-cell level. Again, the cultures were maintained in the continual presence of acyclovir to avoid complications from recently infected cells entering the lytic cycle. Using semiquantitative RT-PCR assays of Wp and Cp transcription, we were able to show (in line with earlier reports [62, 63]) that Wp was activated first and reached its peak within the first week postinfection and then declined to much lower levels by days 21 to 28. By contrast, Cp was delayed in its activation and peaked slightly later but was maintained at a high level thereafter (Fig. 5). The important point is that, while Cp remained unmethylated throughout the process of LCL establishment, Wp became progressively methylated, with the first evidence of methylated residues appearing as early as day 7 and with >90% of CpGs affected by days 18 to 21 (Fig. 6).

Finally, we examined the methylation status of Wp and Cp in the circulating B cells of IM patients undergoing primary EBV infection. At this stage, the transient phase of virus-driven B-cell proliferation in vivo appears to be at its height, with latently infected LCL-like cells expressing the full spectrum of EBV latent proteins being detectable in blood and lymphoid tissues (35, 37, 58). In an earlier study from this laboratory, RT-PCR analysis of IM B cells using Wp- and Cp-specific assays of equal sensitivity showed strong expression of Cp transcripts in all 14 patients studied, whereas only 9 of 14 were positive for Wp transcripts, in many cases at barely detectable levels (58). This strongly suggests that the Wp-to-Cp transition is well advanced in the blood in many cases of acute IM. It is significant, therefore, that in the present work (Fig. 7) we observed extensive methylation of Wp but not Cp sequences in five of the seven patients studied and some methylation at Wp but not Cp in a sixth patient; the seventh patient showed methylation at both promoters, a pattern which other groups have found to be typical of latently infected cells in the memory B-cell pool of long-term virus carriers (38, 44). We could not quantitate Wp and Cp transcript levels in the PBMCs of these patients due to limitations of available cell numbers. However, the parallels between our present methylation analysis and our earlier transcriptional analysis, conducted on a different set of patients, strongly suggest that Wp down-regulation is coincident with methylation of the promoter in vivo just as it is in vitro.

In interpreting these findings, however, it is important to note that there are several (up to 10) copies of the Wp-containing BamHI W fragment in the EBV genome, constituting a large internal repeat which lies immediately downstream of the BamHI C fragment containing Cp (6). As in all earlier studies of Wp, our data on promoter methylation are therefore summative of all Wp copies in the resident EBV genome. The relevance of such studies might be questioned, therefore, if, as one result has implied (64), Wp transcription initiates mainly from the most 5' copy of the promoter. This result, however, came from a highly unusual EBV recombinant in which Cp is defective and in which there are only two copies of Wp. The observation was that EBNA transcription in LCLs established using this recombinant virus came from the 5' Wp copy, but the relative activities of the two Wp copies in freshly infected cells were never examined (64). It seems likely that in the early stages of B-cell infection with a wild-type virus, all copies of Wp are active; this is the most logical inference from the finding that EBNA-LP is hyperexpressed during transformation as a ladder of isoforms with different numbers of BamHI W-encoded repeat domains (13). In such circumstances, therefore, the mechanism of subsequent Wp down-regulation should apply to all copies of Wp. In any case, we would expect some 10% of our independently generated PCR clones to derive from the most 5' Wp region. In that context, in successive in vitro transformation experiments, we have sequenced more than 35 Wp clones at later time points and have found that every one was methylated at the majority of CpG sites. This strongly suggests that all copies of Wp, including the most 5' copy, are subject to methylation during transformation.

We would nevertheless stress that the temporal coincidence of Wp methylation and Wp-to-Cp switching during virus-induced B-cell transformation is merely consistent with, and cannot be interpreted as proof of, a causal relationship. It is possible that Wp is down-regulated by another primary mechanism, perhaps interference by transcription initiated at the upstream Cp (39, 40), and that this then rapidly induces methylation of the silenced promoters by de novo cellular DNA methylases (10). It may also be relevant that the BamHI W repeat region of the genome contains extensive potential secondary structure (24), a situation that appears to generate preferred substrates for methyltransferases (8, 9). Clearly, this methylation of Wp sequences does not spread to Cp in in vitro-transformed LCL cells, where the full pattern of latent gene expression is maintained. By contrast, methylation does eventually spread to Cp in at least some infected cells in vivo (38, 44). This almost certainly reflects the fact that additional physiological signals exist in vivo, probably linked to B-cell differentiation, which lead to Cp suppression and to the establishment of a reservoir of virus-carrying resting memory B cells. What the present work makes clear is that Wp methylation is a relatively early event in the process of B-cell transformation, occurring coincident with Wp-to-Cp switching, and that the ability to inhibit both transcription factor binding to Wp regulatory elements and Wp activity in reporter assays is consistent with an effector role for methylation in the silencing of this promoter.