EBNA 2 and 3 proteins target common sites and genes

To examine the mechanism of interplay in cellular gene reprogramming by EBNA 2, 3A, 3B and 3C proteins, we performed comparative analysis of the human genome regulatory elements targeted by these factors in a BL cell line expressing all EBV latent proteins (Mutu III) ([60] and Figure S1). Since EBNA 2 and 3 proteins do not possess direct DNA-binding activity and interact with their target sites through interactions with cellular DNA-binding proteins, the binding sites analysed in this study represent sites at which the EBNAs indirectly interact with DNA. ChIP-seq was carried out using an EBNA 2-specific monoclonal antibody and compared to ChIP-seq data generated in our laboratory using an anti-EBNA 3C polyclonal antibody that we found also independently precipitates EBNA 3A and 3B, albeit at a lower efficiency (Figure S2) [37] (ChIP-seq data are available through GEO Series accession number GSE47629). Our analysis identified 21,605 significant EBNA 2 binding sites and 7044 significant EBNA 3 binding sites in the human genome indicating that overall EBNA 3A, 3B and 3C proteins target fewer regulatory regions than EBNA 2. Our previous analysis of the human genome binding profile of the EBNA 3 proteins in the Mutu III cell line and recent analysis of EBNA 2 binding in an LCL carrying an EBNA 3B-deleted virus by Zhao et al revealed that only a small proportion of binding sites for these factors are proximal to gene transcription start sites (TSS) [17], [37]. Consistent with these observations, our analysis revealed that 75% of EBNA 2 sites and 84% of EBNA 3 sites were located distal (>4 kb) to TSSs (Figure 1A and B). Examination of the distances between genes and the closest binding sites for EBNA 2 and EBNA 3 proteins revealed that the closest EBNA 3 binding site was most often 10–50 kb from TSSs. In contrast, the closest EBNA 2 binding sites were found both proximal and distal to gene TSSs with similar frequency (Figure 1C). In conclusion, EBNA 2 and 3 proteins generally target distal regulatory elements rather than promoter sequences, with this being most apparent for the EBNA 3s.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 1. Analysis of ChIP-seq data for EBNA 2 and EBNA 3 proteins.

(A) Pie chart showing the distribution of all significant binding sites for EBNA 2 relative to gene TSSs. (B) Distribution of EBNA 3 family binding sites. (C) The frequency of EBNA 2 or EBNA 3 protein binding sites plotted as distance from the TSS of the closest gene. (D) Pie chart showing the proportion of sites identified for EBNA 2 and EBNA 3 family proteins that are shared or unique. (E) Comparison of genes closest to EBNA 2 binding sites with genes closest to EBNA 3 binding sites for sites within 2 kb of a TSS. (F) Comparison of genes closest to an EBNA 2 or EBNA 3 binding site located any distance from a gene TSS.

https://doi.org/10.1371/journal.ppat.1003636.g001

We next considered how EBNA 2 and 3 binding patterns might be related. Comparing binding we detected considerable overlap in the regulatory elements targeted by these proteins, with 25% of all highly significant sites identified bound by both EBNA 2 and the EBNA 3s (Figure 1D). Surprisingly, EBNA 3-only sites constituted only 8% of the sites identified in this analysis (Figure 1D). These data point to a key role of EBNA 3 proteins in the coregulation of cellular gene expression with EBNA 2.

We next sought to identify the genes targeted by EBNA 2 and 3 proteins via the binding sites we had mapped. Looking at binding sites located within 2 kb of a gene TSS, we found that EBNA 2 was associated with 3554 genes and EBNA 3 with 664 genes, consistent with the smaller number of EBNA 3 binding sites in the genome. Comparing genes with EBNA 3 binding sites within 2 kb of the TSS with genes within 2 kb of EBNA 2 binding sites revealed that 62% (412/664) of EBNA 3 proximal target genes were also bound by EBNA 2 (Figure 1E). In fact for 411 of these 412 genes, the proximal EBNA 2 and 3 binding sites were overlapping. Using more relaxed criteria to associate a binding site with a gene, we also identified the genes that were closest to a binding site irrespective of the distance from the site. Using this approach, we found that 80% (3157/3937) of genes closest to an EBNA 3 binding site were also the closest genes to an EBNA 2 binding site. Taken together our analysis indicates that EBNA 2 and 3 proteins generally target the same cellular genes and that a major role of the EBNA 3 proteins is in the co-regulation of genes with EBNA 2.

Comparison with gene expression array data links gene targeting with regulation

To obtain information on whether the potential gene targets we had identified through binding site analysis were regulated by EBNA 2 or EBNA 3 proteins, we examined data available from our own and other published gene expression array studies [11], [32], [37]–[39], [56]–[59], [61]–[62]. We found that 46% (299/654) of EBNA 2-regulated genes identified in these studies had EBNA 2 binding sites within 2 kb of a TSS. In contrast only 8% (199/2601) of documented EBNA 3-regulated genes had promoter-proximal EBNA 3 protein binding sites, likely reflecting the fact that gene regulation by the EBNA 3s is predominantly mediated via distal elements. Consistent with distal regulation of gene expression by the EBNA 3 proteins, the proportion of previously identified EBNA 3-regulated genes associated with an EBNA 3 binding site increased to 31% (802/2601) when we considered genes that were closest to a binding site, irrespective of the distance [32], [37]–[39], [61]–[62]. Concordantly, the proportion of previously identified EBNA 2-regulated target genes associated with an EBNA 2 binding site increased to 60% (393/654) when genes any distance from a binding site were included in the analysis.

EBNA 2 and EBNA 3 binding sites colocalize with histone modifications found at regulatory regions

To obtain information on the chromatin landscape at sites bound by EBNA 2 and the EBNA 3 proteins, we examined histone modifications mapped by ENCODE in the EBV-immortalised LCL GM12878 at the locations of the top 1000 most significant EBNA binding sites mapped in Mutu III BL cells. The GM12878 LCL expresses the same set of EBV latent genes expressed in Mutu III cells (Figure S1) and although there may be some variation in chromatin landscape between BL cells and LCLs, this LCL represents a useful proxy. We found that the majority of EBNA 2 binding sites could be broadly grouped into three clusters defined by high levels of acetylated lysine 27 on Histone H3 (H3K27ac) and the presence of mono-methylated lysine 4 on Histone H3 (H3K4me1), a profile indicative of active enhancers [63] (Figure 2A). The remaining EBNA 2 binding sites displayed the characteristic H3K4me1 enhancer mark, but lacked significant levels of H3K27ac, a profile indicative of poised/inactive enhancers [64] (Figure 2A). These poised EBNA 2 enhancer sites lacked detectable levels of H3K27me3 that can be found in association with some classes of poised enhancers [65]. Plotting average histone modification ChIP-seq signals across the top 1000 EBNA 2 binding sites confirmed enrichment of H3K4me1 and H3K27ac and the absence of H3K27me3 (Figure S3).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 2. Colocalization of histone modifications and transcription factor binding at EBNA 2 and 3 binding sites.

(A) Heatmap of EBNA 2, EBNA 3 and histone modification ChIP-seq signals at the top 1000 EBNA 2 binding sites. EBNA 2 and 3 ChIP-seq data from Mutu III BL cells was aggregated with ENCODE histone modification ChIP-seq data from the GM12878 LCL using hierarchical clustering. Each window displays the ChIP-seq signal −/+ 1 kb around the EBNA 2 binding site midpoint. Clusters of active enhancers (H3K4me1+, H3K27ac+) and poised enhancers (H3K4me1+, H3K27ac−) are indicated. (B) Heatmap of EBNA 3, EBNA 2 and histone modification ChIP-seq signals at the top 1000 EBNA 3 binding sites. (C) Position weight matrix and TF consensus prediction generated from unbiased motif searching using the top 300 EBNA 2 binding sites. Numbers show the p-value for site enrichment. (D) Position weight matrix derived from motif searching using the top 300 EBNA 3 family sites. (E) Position weight matrix for motif searching using all shared sites.

https://doi.org/10.1371/journal.ppat.1003636.g002

A proportion of the top 1000 EBNA 3 binding sites in the human genome also displayed the characteristics of active enhancers (H3K4me1+, H3K27ac+) (Figure 2B). It is noteworthy that these active EBNA 3 enhancer sites were also associated with high-levels of EBNA 2 binding (Figure 2B). The remaining EBNA 3 binding sites displayed the characteristics of poised enhancers (H3K4me1+, H3K27ac−) (Figure 2B). Plotting average histone modification ChIP-seq signals across the top 1000 EBNA 3 binding sites confirmed enrichment of H3K4me1 and H3K27ac and the absence of H3K27me3, similar to that observed for EBNA 2 binding sites (Figure S4). The chromatin landscape at EBNA 3 binding sites is consistent with the function of these proteins as both repressors and activators of transcription, but it is interesting that our analysis did not detect any colocalization of the H3K27me3 mark, or the K27 methyltransferase EZH2 that has been associated with EBNA 3-mediated gene repression [37], [48], [53], [55] (Figure 2B and S4). However, H3K27me3 is often broadly spread over promoter and gene bodies and the lack of H3K27me3 at EBNA 3 binding sites may be consistent with our recent observation that the repression of ADAM 28 and ADAMDEC1 by EBNA 3C via an intergenic enhancer was associated with larger increases in H3K27me3 within the genes than at the actual binding site [37].

EBNA 2 and 3 binding sites are also bound by B-cell specific transcription factors

To determine whether specific cellular transcription factor binding motifs were enriched at sites bound by EBNA 2, 3A, 3B and 3C we performed unbiased motif searching using the top 300 most significant EBNA binding sites. We found that EBNA 2 sites were enriched for RBP-Jκ and EBF1 motifs and that binding sites for the EBNA 3s were enriched for PU.1 motifs (Figure 2C and D). All shared sites were enriched for PU.1 and EBF1 motifs (Figure 2E). These data provide support for a role for RBP-Jκ and PU.1 in EBNA 2 and EBNA 3C gene targeting in vivo and corroborate recent reports of enrichment of EBF1 motifs at EBNA 2 binding sites in LCLs, implicating EBF1 in EBNA 2 targeting or binding stabilization [12]–[15], [17], [42]–[45].

We next examined transcription factor binding at EBNA 2 and EBNA 3 binding sites using recently published RBP-Jκ ChIP-seq data from the EBV-immortalised LCL, IB4 [17] and ENCODE GM12878 ChIP-seq data available for numerous transcription factors and regulators. As expected, the top 1000 EBNA 2 binding sites mapped in Mutu III cells were also extensively bound by both EBNA 2 and the EBNA 2 targeting partner RBP-Jκ in IB4 cells. EBNA 2 binding sites were however also bound by a plethora of transcription factors expressed in B cells that have not been previously implicated in EBNA 2 targeting, including BATF, Bcl11a, Bcl3, IRF4, PAX5, SP1 and TCF12 (Figure S5). Binding of the histone acetyl transferase p300, a co-activator characteristically associated with enhancers, is also evident at EBNA 2 sites. Plotting average ChIP-seq signals across EBNA 2 binding sites confirmed the binding of BATF, Bcl11a, Bcl3, IRF4, PAX5, SP1, TCF12 and p300 (Figure S6). These plots also detected association of the EBNA 2 binding protein PU.1 [15] with EBNA 2 binding sites. Interestingly, EBF1 ChIP-seq signals were high across EBNA 2 binding sites (Figure S6) consistent with the presence of EBF1 binding motifs at EBNA 2 sites (Figure 2C and E) and recent studies implicating EBF1 in EBNA 2 transcriptional regulation [17]. Our analysis revealed that the top 1000 EBNA 3 binding sites were also extensively bound by the same set of factors bound at EBNA 2 sites including RBP-Jκ, BATF, Bcl11a, Bcl3, IRF4, PAX5, SP1, TCF12, PU.1, EBF1 and p300 (Figures S5 and S7).

In summary, our co-association analyses detect the binding of multiple transcription factors at EBNA binding sites, with the binding of known targeting partners indistinguishable from that of a plethora of other transcription factors. These results are consistent with genome-wide ChIP-seq analyses by ENCODE that detect extensive context and cell-type specific co-association of multiple transcription factors with the same regulatory sites [66]. This is likely to reflect the accessibility of open chromatin at the binding site and protein-protein interactions between transcription factors. As a result, although our motif searching confirmed a potential role for EBF1 in EBNA 2 targeting, examination of the binding of particular transcription factors at EBNA binding sites through co-association analysis does not appear to be useful for the identification of the specific factors that may involved in targeting or stabilising the binding of EBNA 2 and 3 proteins to cellular DNA regulatory elements.

Investigating the role of co-incident EBNA 2 and 3 protein binding at key gene loci

Since our analysis of EBNA 2 and 3 binding had revealed a large number of shared binding sites, we set out to investigate the nature of this shared binding and its role in the control of epigenetic reprogramming by EBV. We carried out follow-up studies on three cellular genes (CTBP2, WEE1 and ITGAL) with highly significant common binding sites for EBNA 2 and EBNA 3 proteins with ChIP-seq peak heights of >10 sequence reads per million background subtracted reads. These genes are involved in regulating key processes relevant to cellular transformation including transcription repression, cell-cycle and cell adhesion and activation. C-terminal binding proteins (CtBPs) encoded by the closely-related genes CTBP2 and CTBP1 mediate transcriptional repression of a number of key tumour suppressor genes [67]. CtBPs play a key role in EBV transformation; CtBP binding by EBNA 3A and EBNA 3C is required for the epigenetic repression of p16^INK4a in infected cells [53]. WEE1 encodes a cell-cycle kinase and negative regulator of CDK1 that controls the transition from G2 into mitosis, a process that is deregulated at multiple potential levels by EBV [68]–[70]. The B-cell adhesion protein integrin alpha L encoded by the ITGAL gene forms part of the heterodimeric activation antigen, LFA-1, that is upregulated in EBV-infected cells [71]. These three genes also provide representative examples of co-incident binding sites at intragenic (CTBP2) and downstream (WEE1) long-range locations and at a gene promoter (ITGAL). CTBP2 and WEE1 represent novel EBV cellular gene targets identified in our study as the closest genes to EBNA 2 and 3 binding sites; ITGAL was previously identified as a gene repressed by EBNA 3B and 3C by White et al in microarray studies of BL cells infected with recombinant knock-out EBVs [38]. Expression of the LFA-1 heterodimer composed of integrin alpha L (CD11a) and beta 2 integrin has also been shown to be upregulated by the EBV membrane protein, LMP1 [72].

EBNA 3A, 3B and 3C bind the C-terminal binding protein 2 (CTBP2) enhancer

A single co-incident binding site for EBNA 2 and EBNA 3 proteins was mapped in the second intron of CTBP2, 95 kb downstream from the TSS (Figure 3). To confirm our ChIP-seq data we carried out ChIP-QPCR using EBNA 2-specific antibodies in the Mutu III cells used for ChIP-seq analysis using sets of primers designed to amplify the binding site, adjacent control regions and the peptidylprolyl isomerase 1 (PPIA) control gene, which is not bound or regulated by the EBNAs. This analysis confirmed that EBNA 2 bound specifically to the CTBP2 intragenic site in vivo (Figure 3B). Since our ChIP-seq analysis was carried out using an antibody that precipitated all three EBNA 3 proteins, we also carried out ChIP-QPCR in Mutu III cells to examine the binding of EBNA 3A, 3B and 3C proteins individually using anti-EBNA 3A, anti-EBNA 3B and anti-EBNA 3C antibodies that we verified do not cross-react with other family members (Figure S8). Our results demonstrated that EBNA 3A, 3B and 3C were all able to associate with the CTBP2 site in Mutu III cells (Figure 3B–E). We also extended our ChIP-QPCR analysis to examine binding of EBNA 2, 3A, 3B and 3C to the CTBP2 site in an EBV-immortalized LCL to investigate any potential cell-type specific differences in binding. As in Mutu III cells, we found that EBNA 2, 3A, 3B and 3C were all able to associate with the CTBP2 intragenic site in an LCL (Figure 3F–I).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 3. EBNA 2, 3A, 3B and 3C binding at the CTBP2 locus in EBV-infected cells.

(A) The number of EBNA 2 (green) and EBNA 3 (red) sequencing reads from immunoprecipitated Mutu III DNA are plotted per million background-subtracted total reads and aligned with the human genome. The direction of gene transcription is indicated by the red arrow. GM12878 LCL H3K27ac ChIP-seq data from ENCODE are shown at the bottom of the panel. Panels B-E show ChIP-QPCR carried out in Mutu III cells and panels F-I show data from the PER253 B95.8 LCL. Precipitated DNA was analysed using primer sets located at the binding site (set B) or regions on either side of the binding site (sets A and C). Primers spanning the transcription start site of the cellular gene encoding peptidylprolyl isomerase A (PPIA) that is not regulated or bound by the EBNAs provide a background binding control (indicated by dotted lines). (B) and (F) ChIP using anti-EBNA 2 antibodies. (C) and (G) ChIP using anti-EBNA 3A antibodies. (D) and (H) ChIP using anti-EBNA 3B antibodies. (E) and (I) ChIP using anti-EBNA 3C antibodies. Percentage input signals, after subtraction of no antibody controls, are shown as the mean +/− range of two independent experiments. (J) Q-PCR analysis of CTBP2 transcript levels using cDNA from wild-type LCLs (wt1, 2 and 3) and LCLs established from EBNA 3A knock-out viruses (mtB1, B2 and B3) in two different donor backgrounds (D2 and D3). Transcript levels were normalised to GAPDH levels and expressed relative to the level in D2 wt1 cells. (K) Q-PCR analysis of CTBP2 transcript levels using cDNA from wild-type LCLs infected with B95.8 virus (wt) and EBNA 3B knock-out LCLs (KO) in two different donor backgrounds (PER142 and PER253). Transcript levels were normalised to GAPDH levels and expressed relative to the level in wt cells for each donor. (L) Q-PCR analysis of CTBP2 transcript levels using cDNA from the ER/EB 2.5 LCL expressing EBNA 2 that is active in the presence of β-estradiol (+ est) and inactive in the absence of β-estradiol (−est). Cells were incubated in β-estradiol-free media for 4 days prior to re-addition of β-estradiol or DMSO control for 6 or 17 hrs. Transcript levels were normalised to GAPDH levels and expressed relative to the level in the absence of β-estradiol for each time course. All cDNA results (J–L) show the mean −/+ range of two independent QPCR reactions each performed in duplicate.

https://doi.org/10.1371/journal.ppat.1003636.g003

Previous expression microarray analyses have not identified CTBP2 as an EBNA 2, 3A, 3B or 3C-regulated gene. However, examination of microarray data from a study carried out by Hertle et al [32] (re-analysed by Dr R. White and colleagues and available at www.epstein-barrvirus.org.uk) using sets of LCLs infected with wild-type or EBNA 3A mutant EBV (Figure S9) revealed that expression of CTBP2 was upregulated in LCLs infected with EBNA 3A knock-out (KO) EBV. These data implicate EBNA 3A as a repressor of CTBP2 transcription. QPCR analysis of CTBP2 mRNA levels in these EBNA 3A KO LCLs confirmed that loss of EBNA 3A leads to a dramatic upregulation of CTBP2 transcript levels (Figure 3J). The predominant isoform of CTBP2 expressed in these cells initiates from the third most distal promoter (Figure S10). We also found that CTBP2 mRNA levels were higher in a latency I BL cell line expressing only EBNA 1 (Mutu I) compared to Mutu III cells, the latency III derivative of this cell line that expresses all EBV latent proteins (Figure 3J). To investigate the specific effects of EBNA 3B on CTBP2 mRNA expression we analysed pairs of LCLs infected with wild-type or EBNA 3B KO EBV (Figure S9). Consistent with our observations in EBNA 3A KO LCLs, we found that a lack of EBNA 3B expression also lead to an increase in CTBP2 transcript levels (Figure 3K). Taken together these data strongly implicate EBNA 3A and 3B binding at the CTBP2 locus in transcriptional repression of this gene. Interestingly, during our analysis we found that CTBP2 expression was cell-line specific; significant levels of CTBP2 transcripts were not detectable in a number of cell lines previously used in expression microarray studies by White et al [38]. These include a set of wild-type and EBNA 3B knock-out LCLs and the BL31 and BL2 cells used to generate a series of EBNA 3A, 3B or 3C knock-out cell lines [38], [54]. As a result we have been unable to analyse the effects of EBNA 3C gene knock-out on CTBP2 mRNA levels as the currently available EBNA 3C knock-out cell lines were generated in the BL31 or BL2 background [54].

To investigate any potential effects of EBNA 2 on CTBP2 expression that may not have been detected in previous expression studies, we examined CTBP2 mRNA levels in LCLs expressing a conditionally active estrogen receptor-EBNA 2 fusion protein (ER/EB 2.5) [3]. Our analysis did not detect any significant effect of the loss of EBNA 2 function on CTBP2 transcript levels (Figure 3L). Our data therefore correlate EBNA 3A and 3B binding at the CTBP2 site with the repression of CTBP2 transcription by these factors, but do not detect any effect of loss of EBNA 2 on CTBP2 mRNA levels in the presence of the EBNA 3s.

EBNA 3 proteins repress CTBP2 transcription by preventing enhancer-promoter looping

To determine the role of the long-range CTBP2 intragenic EBNA 2, 3A, 3B and 3C binding site in controlling CTBP2 transcription, we performed chromosome conformation capture analysis [73] to examine potential interactions between this site and the CTBP2 promoter through the formation of chromatin loops. Following digestion of cross-linked chromatin using an enzyme that cleaves intervening DNA, enhancer and promoter regions can be ligated together at low dilution using this technique only if these elements are in close spatial proximity (Figure 4A). We found that PCR amplification products using primers spanning the CTBP2 enhancer-promoter ligation junction could not be detected in LCLs generated from wild-type EBV where all EBNAs are expressed, but were readily detectable in EBNA 3A knock-out LCLs (Figure 4B). These data indicate that enhancer-promoter chromatin loops form in the absence of EBNA 3A but not in wild-type infected cells. Loop formation correlated with increased CTBP2 transcription in EBNA 3A knock-out LCLs indicating that looping at the CTBP2 locus is associated with active transcription (Figure 3J). These data are consistent with a model where EBNA 3A represses CTBP2 transcription by preventing the formation of chromatin loops that mediate gene activation. Concordantly, chromosome conformation capture experiments in LCLs expressing a conditionally active ER-EBNA 2 demonstrated that CTBP2 promoter-enhancer interactions were absent in these cells irrespective of whether EBNA 2 was functional (+/− β-estradiol) (Figure 4C and D) presumably as a result of the dominant binding and prevention of looping by the EBNA 3 proteins.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 4. The influence of EBNA 2 and 3A on chromosome looping at the CTBP2 locus.

(A) Diagram (not to scale) showing the EcoR1 restriction fragments at the CTBP2 locus that encompass the promoter (P), enhancer (E) and an intervening control region (con). The arrow indicates the direction of transcription. (B) Chromosome conformation analysis in LCLs infected with wild-type or EBNA 3A knock-out EBV using primer pairs that amplify across promoter-enhancer or promoter-control ligation junctions. Positive controls show PCR amplification from control digestion and ligation reactions carried out using PCR-amplified DNA fragments encompassing the promoter, enhancer and control regions. (C). Chromosome conformation capture analysis in the ER-EB 2.5 LCL expressing EBNA 2 that is active in the presence of β-estradiol (+ est) and inactive in the absence of β-estradiol (−est). (D) Model for the control of chromatin looping by EBNA 2 and 3 proteins at CTBP2. (E) Re-ChIP analysis using anti-EBNA 2 antibodies in the first round of ChIP followed by a second round of ChIP in absence of antibody or using anti-EBNA 2, EBNA 3A, EBNA 3B or EBNA 3C antibodies. Results show mean percentage primary input −/+ range of two independent Q-PCR reactions from a representative experiment. (F) Control re-ChIP analysis using anti-EBNA 3A antibodies in the first round followed by re-precipitation in the absence of antibody or using anti-EBNA 3A antibodies. (G) Control re-ChIP analysis using anti-EBNA 3B antibodies in the first round followed by re-precipitation in the absence of antibody or using anti-EBNA 3B antibodies. (H) Control re-ChIP analysis using anti-EBNA 3C antibodies in the first round followed by re-precipitation in the absence of antibody or using anti-EBNA 3C antibodies.

https://doi.org/10.1371/journal.ppat.1003636.g004

To directly examine whether EBNA 2 and 3 proteins can associate with the CTBP2 enhancer at the same time or compete for binding, we carried out re-ChIP analysis. Primary precipitation of chromatin was carried out using an EBNA 2-specific antibody and immunoprecipitated protein-DNA complexes were then eluted from the bead matrix and a second round of precipitation carried out using EBNA 2, 3A, 3B or 3C-specific antibodies (Figure 4E). Our analysis detected re-precipitation of EBNA 2 bound to the CTBP2 enhancer using EBNA 2-specific antibodies as expected, but EBNA 3A, 3B and 3C were not precipitated from EBNA 2-DNA complexes, despite the fact that they can be re-precipitated with the antibodies used (Figure 4F–H). Reciprocal re-ChIP experiments carried out using EBNA 3A, 3B or 3C antibodies in the first round and EBNA 2 antibodies in the second round confirmed that EBNA 2 could not be precipitated from EBNA 3-DNA complexes (Figure S11). Our results therefore indicate that EBNA 3A, 3B and 3C do not bind the CTBP2 enhancer at the same time as EBNA 2. Binding profiles obtained in ChIP therefore represent independent binding of EBNA 2 or EBNA 3 proteins to this site in different cells within the total cell population assayed. In summary, our results are consistent with CTBP2 transcriptional repression by EBNA 3 proteins via competitive binding at the CTBP2 intragenic enhancer (Figure 4D).

Cell-type specific differential binding of EBNA 3 proteins to two downstream distal enhancers at the WEE1 locus

ChIP-seq detected binding of EBNA 2 and 3 proteins to co-incident binding sites located in two clusters +27 kb and +39 to 44 kb downstream from the WEE1 TSS (Figure 5). The first cluster, enhancer 1 (+27 kb) has two distinct binding sites for EBNA 3 proteins (sites 1 and 2), with EBNA 2 binding detected predominantly at the most distal of these two sites (site 2). The second cluster, enhancer 2 (+39 to 44 kb) contains 3 binding sites (sites 3–5), with the highest binding signal for EBNA 3 proteins detected at the most distal site (site 5). Both clusters of EBNA 2 and 3 binding sites coincide with peaks of H3K27ac detected in ENCODE ChIP-seq experiments in the GM12878 LCL indicating that the binding sites are located within active gene regulatory regions (Figure 5).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 5. EBNA 2 and EBNA 3 protein binding at the WEE1 locus in EBV infected cells.

(A) EBNA 2 (green) and EBNA 3 (red) sequencing reads at the WEE1 locus (displayed as described in Figure 3). Panels B–E show ChIP-QPCR carried out in Mutu III cells and panels F–I show data from the PER253 B95.8 LCL. Precipitated DNA was analysed using primer sets located at the binding sites (sets B, D, F, H and J) or regions adjacent to the binding sites (sets A, C, E, G and I). Binding signals at the CTBP2 binding site in the same ChIP experiments are shown as a positive control and primers spanning the transcription start site of the cellular gene PPIA provide a background binding control (indicated by dotted lines). (B) and (F) ChIP using anti-EBNA 2 antibodies. (C) and (G) ChIP using anti-EBNA 3A antibodies. (D) and (H) ChIP using anti-EBNA 3B antibodies. (E) and (I) ChIP using anti-EBNA 3C antibodies. Percentage input signals, after subtraction of no antibody controls, are shown as the mean −/+ range of two independent ChIP experiments. (J) Q-PCR analysis of WEE1 transcript levels using cDNA from BL31 parental cells and BL31 cells infected with wild-type recombinant EBV (wtBac-2 and 3), EBNA 3C knock-out EBV (3C KO-3 and 6) or EBNA 3C revertant EBV (3Crev-2 and 4). Transcript levels were normalised to GAPDH levels and expressed relative to the level in parental BL31 cells. * indicates a p-value of <0.01 (students t-test) compared to the wtBac-2 cell line. (K) Q-PCR analysis of WEE1 transcript levels using cDNA from the ER/EB 2.5 LCL expressing EBNA 2 that is active in the presence of β-estradiol (+ est) and inactive in the absence of β-estradiol (−est). Cells were incubated in β-estradiol-free media for 4 days prior to re-addition of β-estradiol for 6 or 17 hrs. Transcript levels were normalised to GAPDH levels and expressed relative to the level in the absence of β-estradiol. All cDNA results (J–L) show the mean −/+ standard deviation of three independent QPCR analyses from two independent cDNA preparations.

https://doi.org/10.1371/journal.ppat.1003636.g005

ChIP-QPCR analysis confirmed EBNA 2 binding to all five sites in both Mutu III cells and an LCL (Figure 5). Consistent with ChIP-seq data, we detected the highest levels of EBNA 2 binding at site 2 in both cell lines (Figure 5B and F). EBNA 3A and 3B binding at WEE1 was low-level, with only small amounts of EBNA 3A binding at sites 4 and 5 in LCLs and weak EBNA 3B binding at sites 1 and 2 in Mutu III cells (Figure 5C, D, G and H). In contrast, EBNA 3C bound specific WEE1 enhancer sites at levels similar to that detected at CTBP2, although there were cell-type specific differences in the binding profile (Figure 5E and I). EBNA 3C bound predominantly at site 5 in Mutu III cells, but at both sites 4 and 5 in an LCL (Figure 5E and I).

Like CTBP2, WEE1 had not previously been documented as a gene regulated by EBNA 2 or EBNA 3 proteins. However, examination of microarray data obtained by White et al [38] from BL31 cells infected with a series of EBNA 3 knock-out viruses provided evidence of repressive effects of EBNA 3C (but not EBNA 3A and 3B) on WEE1 mRNA expression (www.epstein-barrvirus.org.uk). QPCR analysis of these BL31 cell lines confirmed that WEE1 transcript levels were modestly but significantly increased in cells infected with EBNA 3C knock-out EBV compared to cells infected with wild-type EBV or revertant viruses (Figure 5J). Examination of microarray data obtained by White et al [38] and Hertle et al [32] using EBNA 3B and EBNA 3A knock-out LCLs respectively, confirmed that EBNA 3A and EBNA 3B had no detectable effect on WEE1 mRNA levels. Our data are therefore consistent with EBNA 3C binding to the most distal enhancer region (enhancer 2) downstream of WEE1 playing a dominant role in the repression of WEE1 transcription. The low-level association of EBNA 3A and 3B at WEE1 enhancer sites however, does not detectably affect WEE1 mRNA levels in BL cells or LCLs.

To address the potential contribution of EBNA 2 in the regulation of WEE1 transcription we examined WEE1 mRNA expression in BL31 cells infected with an EBNA 2 knock-out virus and in LCLs expressing conditionally active EBNA 2. Our analysis revealed that WEE1 mRNA levels are reduced in the absence of EBNA 2 activity, implicating EBNA 2 in the positive regulation of WEE1 transcription (Figure 5K and L). Taken together our data indicate that the level of WEE1 transcription in EBV-infected cells is likely to be determined by a balance between gene repression by EBNA 3C that is counteracted by EBNA 2.

EBNA 3C represses WEE1 transcription by directing the formation of enhancer-promoter loops

To examine potential looping interactions between the downstream WEE1 enhancers and the WEE1 promoter and their regulation by EBNA 2 and 3C, we performed chromosome conformation capture analysis. In contrast to the situation at the CTBP2 locus, our results indicated that looping interactions at WEE1 were associated with transcriptional repression. Looping between both enhancer 1 (sites 1 and 2) and enhancer 2 (sites 3, 4 and 5) and the promoter was not detectable in parental EBV negative BL31 cells, but was detectable in wild-type EBV infected BL31 cells where WEE1 expression is repressed (Figure 5J). Consistent with the key role played by EBNA 3C in transcriptional repression of WEE1, looping interactions were absent in BL31 cells infected with EBNA 3C knock-out viruses indicating that EBNA 3C is required to maintain enhancer-promoter looping (Figure 6B).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 6. The influence of EBNA 2 and 3C on chromosome looping at the WEE1 locus.

(A) Diagram (not to scale) showing the EcoR1 restriction fragments at the WEE1 locus that encompass the promoter (P), two downstream enhancers (E1 and E2) and an intervening control region (con). The arrow indicates the direction of transcription. (B) Chromosome conformation analysis in BL31 parental cells and BL31 cells infected with wild-type recombinant EBV (wtBac-2), EBNA 3C knock-out EBV (3CKO-3) or EBNA 3C revertant EBV (3Crev-4) using primer pairs that amplify across promoter-enhancer or promoter-control ligation junctions. Positive controls show PCR amplification from control digestion and ligation reactions carried out using PCR-amplified DNA fragments encompassing the promoter and enhancers. (C) Chromosome conformation analysis in BL31 parental cells and BL31 cells infected with wild-type recombinant EBV (wtBac-2) or EBNA 2 KO EBV. (D) Chromosome conformation capture analysis in the ER-EB 2.5 LCL expressing EBNA 2 that is active in the presence of β-estradiol (+ est) and inactive in the absence of β-estradiol (−est). (E) Model for the control of chromatin looping by EBNA 2 and 3 proteins at WEE1. (F) Re-ChIP analysis in Mutu III cells using anti-EBNA 2 antibodies in the first round of ChIP followed by a second round of ChIP in absence of antibody or using anti-EBNA 2 or EBNA 3A, 3B or 3C antibodies. Primers at peak 5 in enhancer 2 were used for analysis. Results show mean percentage primary input −/+ range of two independent Q-PCR reactions from a representative experiment. (G) Control re-ChIP analysis using anti-EBNA 3C antibodies in the first round followed by re-precipitation in the absence of antibody or using anti-EBNA 3C antibodies.

https://doi.org/10.1371/journal.ppat.1003636.g006

The dominant role of EBNA 3C binding in mediating transcriptional repression of WEE1 via chromatin looping was supported by our observations that looping interactions were maintained in BL31 cells infected with EBNA 2 knock-out viruses and in an ER-EBNA 2 LCL when EBNA 2 function was inhibited through β-estradiol withdrawal (Figure 6C and D). Taken together these data support a model where EBNA 3C binding to a downstream enhancer element directs chromatin looping with the WEE1 promoter to facilitate transcriptional repression (Figure 6E). Given that loss of EBNA 2 decreases WEE1 transcription (Figure 5J), EBNA 2 may compete with EBNA 3C for binding to WEE1 enhancer sites and limit EBNA 3C-mediated repression in infected cells. It is likely that the semi-quantitative chromosome conformation capture assay is not able to detect the increased looping that may occur in the absence of EBNA 2. Consistent with competitive binding of EBNA 2 and 3 proteins and our observations at the CTBP2 locus, re-ChIP experiments did not detect co-association of EBNA 2 and 3C at WEE1 enhancer 2 (Figure 6F and G and S11).

Cell-type specific differential binding of EBNA 3 proteins to the ITGAL promoter

At ITGAL a series of three co-incident binding sites for EBNA 2 and EBNA 3 proteins are located in a 2 kb region encompassing the TSS that coincides with a peak of LCL H3K27ac (Figure 7). Consistent with ChIP-seq analysis, ChIP-QPCR experiments confirmed that EBNA 2 bound predominantly at site 3 in both BL cells and an LCL (Figure 7B and F). Interestingly, when we examined individual binding of EBNA 3 proteins to these promoter sites, we found that EBNA 3B also bound at site 3 in both BL cells and LCLs (Figure 4D and H), but there was no significant binding of EBNA 3A in either cell line, despite our ability to detect significant levels of EBNA 3A binding at the CTBP2 locus in the same ChIP samples (Figure 7C and G). We also found that although EBNA 3C bound site 3 in Mutu III cells at levels significantly above the background level detected at a control gene, there was no significant EBNA 3C binding in LCLs (Figure 7E and I). Remarkably, the differential and cell-type specific targeting of the ITGAL promoter by EBNA 3 family members we detected is entirely consistent with the documented effects of individual EBNA 3 proteins on ITGAL expression in these different cell backgrounds. Consistent with our observations in the Mutu III BL cell line, the White et al BL31 study [38] detected significant and reproducible increases in ITGAL expression in cell lines infected with EBNA 3B or EBNA 3C knock-out EBV, but not EBNA 3A knock-out EBV, compared to cells infected with wild-type EBV. EBNA 3B knock-out BL31 cells displayed a 3.09-fold (p-value: 2.77E-05) increase in ITGAL expression vs wild-type infected BL31 cells and EBNA 3C knock-out BL31 cells displayed a 2.72-fold increase (p-value: 1.09E-04) [38]. Interestingly, examination of microarray data from the White et al [38] and Hertle et al [32] LCL studies (www.epstein-barrvirus.org.uk) revealed a role for EBNA 3B, but not EBNA 3A in the repression of ITGAL expression in LCLs. These data are entirely consistent with our ChIP QPCR analysis where binding of EBNA 3B was detected in LCLs in the absence of any significant EBNA 3A or EBNA 3C binding (Figure 7G, H and I). We confirmed the repressive effects of EBNA 3B on ITGAL expression in LCLs using an additional independent set of wild-type and EBNA 3B knock-out cell lines (Figure 8A). Thus ITGAL is an EBNA 3B and 3C-repressed gene in BL cells and an EBNA 3B only-repressed gene in LCLs. Since EBNA 3A appears to play no role in ITGAL regulation, importantly our data demonstrate that it is the gene and cell-type specific binding of different members of the EBNA 3 protein family that determines the requirement for each EBNA 3 protein in the regulation of specific cellular genes.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 7. EBNA 2 and EBNA 3 protein binding at the ITGAL promoter in EBV-infected cells.

(A) EBNA 2 (green) and EBNA 3 (red) sequencing reads from immunoprecipitated Mutu III DNA plotted as in Figure 3. Panels B–E show ChIP-QPCR carried out in Mutu III cells and panels F–I show data from the PER253 B95.8 LCL. Precipitated DNA was analysed using primer sets located at the binding sites (sets B, D, and F) or regions adjacent to the binding sites (sets A, C, and E). Binding signals at the CTBP2 binding site in the same ChIP experiments are shown as a positive control and primers spanning the transcription start site of the cellular gene PPIA provide a background binding control (indicated by dotted lines). (B) and (F) ChIP using anti-EBNA 2 antibodies. (C) and (G) ChIP using anti-EBNA 3A antibodies. (D) and (H) ChIP using anti-EBNA 3B antibodies. (E) and (I) ChIP using anti-EBNA 3C antibodies. Percentage input signals, after subtraction of no antibody controls, are shown as the mean −/+ range of two independent ChIP experiments.

https://doi.org/10.1371/journal.ppat.1003636.g007

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 8. The Effect of EBNA 2 and 3 proteins on ITGAL expression.

(A) Q-PCR analysis of ITGAL transcript levels using cDNA from wild-type LCLs infected with B95.8 virus (wt) and EBNA 3B knock-out LCLs (KO) in two different donor backgrounds (PER142 and PER253). Transcript levels were normalised to GAPDH levels and expressed relative to the level in wt cells for each donor. Results show the mean −/+ range of two independent QPCR reactions each performed in duplicate. (B) Luciferase reporter assays carried out in DG75 cells transiently transfected with 2 µg of the control vector pGL3 basic, an ITGAL promoter-luciferase reporter (pGL3 ITGALp) or the EBV C promoter reporter (pCp1425GL2) (right panel) in the absence or presence of 10 or 20 µg of EBNA 2, 3A, 3B or 3C expressing constructs. Firefly luciferase signals were normalised to Renilla luciferase signals from the co-transfected control plasmid pRL-TK (1 µg). Results show the mean −/+ standard deviation of 3 independent experiments and are expressed relative to the pGL3 basic signal (left panel) or the Cp1425GL2 signal (right panel) in the absence of EBNA 2. Western blot analysis of EBNA 2, 3A, 3B and 3C expression levels in transfected cells. Each set of blots was also probed for actin as a loading control. (C) Re-ChIP analysis in Mutu III cells using anti-EBNA 2 antibodies in the first round of ChIP followed by a second round of ChIP in absence of antibody or using anti-EBNA 2, EBNA 3B or EBNA 3C antibodies. Primers at peak 3 were used for analysis. Results show mean percentage primary input −/+ range of two independent Q-PCR reactions from a representative experiment. (D) Control re-ChIP analysis using anti-EBNA 3B antibodies in the first round followed by re-precipitation in the absence of antibody or using anti-EBNA 3B antibodies. (E) Control re-ChIP analysis using anti-EBNA 3C antibodies in the first round followed by re-precipitation in the absence of antibody or using anti-EBNA 3C antibodies.

https://doi.org/10.1371/journal.ppat.1003636.g008

Since ITGAL is upregulated by EBV LMP1 [72], a direct target gene of EBNA 2 that is not expressed in the absence of EBNA 2, the effects of LMP1 and EBNA 2 on ITGAL expression cannot be separated using cells which lack functional EBNA 2. We therefore examined the regulation of the ITGAL promoter by EBNA 2 using ITGAL promoter-reporter constructs containing the 2 kb region encompassing all 3 binding sites. Transient transfection assays demonstrated that EBNA 2 activated the ITGAL promoter up to 5-fold (Figure 8B). The co-expression of EBNA 3A, 3B or 3C resulted in the inhibition of EBNA 2 activation (Figure 8B). These data are consistent with a model where EBNA 3 proteins can compete with EBNA 2 for binding sites in the ITGAL promoter to repress ITGAL transcription. It therefore appears that although EBNA 3A and EBNA 3C (in LCLs) do not bind the ITGAL promoter site significantly in vivo, these proteins are able to bind and compete with EBNA 2 for binding to this site when expressed at high levels in reporter assays. Competitive binding of EBNA 2 and 3 proteins at the ITGAL promoter was also supported by re-ChIP analysis in the Mutu III BL cell line, where we detected no simultaneous binding of EBNA 2 with EBNA 3B or EBNA 3C (Figure 8C–E and Figure S11).

EBNA 3 proteins also differentially bind EBNA 3-only promoter and enhancer sites

Our observations that the regulation of ITGAL and WEE1 expression by members of the EBNA 3 family was directed by the differential binding of subsets of these proteins to binding sites shared by EBNA 2 led us to investigate whether this was also the case at genes regulated by EBNA 3-only sites. The BCL2L11 (Bim) gene is known to be transcriptionally repressed by EBNA 3A and EBNA 3C, but not EBNA 3B [54] and recent studies using an epitope-tagged form of EBNA 3C have detected binding of EBNA 3C over a broad region from −3648 to +631 bps around the BCL2L11 TSS [48]. Although the peak binding signal of this tagged EBNA 3C in BL31 cells was detected with primer sets located −247 to −147 bps upstream of the BCL2L11 TSS, our ChIP-seq analysis fine-mapped significant EBNA 3 binding sites between +87 to +921 bps downstream of the BCL2L11 TSS. ChIP-QPCR analysis of the binding of individual EBNA 3A, 3B and 3C proteins to this site revealed that only EBNA 3A and 3C bind at significant levels in both the Mutu III BL and an LCL, consistent with the regulation of BCL2L11 expression by these two proteins (Figure 9B–G). We found that the peak of binding of EBNA 3A and 3C was detected by ChIP-QPCR using primer sets located +540 to +630 downstream from the TSS (set B) (Figure 9B–G).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 9. EBNA 3 protein binding at the BCL2L11 promoter in EBV-infected cells.

(A) EBNA 3 sequencing reads from immunoprecipitated Mutu III DNA plotted as in Figure 3. Panels B–D show ChIP-QPCR carried out in Mutu III cells and panels F–G show data from the PER253 B95.8 LCL. Precipitated DNA was analysed using primer sets located at the binding site (set B) or regions on either side of the binding site (sets A and C). Binding signals at the CTBP2 binding site in the same ChIP experiments are shown as a positive control and primers spanning the transcription start site of the cellular gene PPIA provide a background binding control (indicated by dotted lines). (B) and (E) ChIP using anti-EBNA 3A antibodies. (C) and (F) ChIP using anti-EBNA 3B antibodies. (D) and (G) ChIP using anti-EBNA 3C antibodies. Percentage input signals, after subtraction of no antibody controls, are shown as the mean −/+ range of two independent experiments.

https://doi.org/10.1371/journal.ppat.1003636.g009

We also examined differential binding of EBNA 3A, 3B and 3C to an EBNA 3-only distal regulatory element located between the ADAMDEC1 and ADAM28 genes that we have previously characterised [37]. We previously demonstrated that repression of ADAM28 and ADAMDEC1 transcription through H3K27me3 could be mediated through the action of EBNA 3C alone and detected EBNA 3C binding to this intergenic element in EBNA 3C-expressing cells [37]. Microarray studies have also implicated EBNA 3A in ADAMDEC1 repression in LCLs and BL cells and in ADAM28 repression in LCLs [32], [38]. Consistent with a role for both EBNA 3A and EBNA 3C in the repression of ADAMDEC1 transcription, we detected binding of EBNA 3A and EBNA 3C, but not EBNA 3B to the ADAM intergenic element in both BL cells and an LCL (Figure 10). Our data therefore indicate that the specificity of cellular gene regulation by members of the EBNA 3 family is directed by the differential binding of different family members to both unique EBNA 3 binding sites and binding sites shared with EBNA 2.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 10. EBNA 3 protein binding at the ADAM28/ADAMDEC1 intergenic enhancer in EBV-infected cells.

ChIP-QPCR carried out in Mutu III cells (A–C) and the PER253 B95.8 LCL (D–F). Precipitated DNA was analysed using primer sets located at the centre of the binding site (set B) or the edges of the binding site (sets A and C). Binding signals at the CTBP2 binding site in the same ChIP experiments are shown as a positive control and primers spanning the transcription start site of the cellular gene PPIA provide a background binding control (indicated by dotted lines). (A) and (D) ChIP using anti-EBNA 3A antibodies. (B) and (E) ChIP using anti-EBNA 3B antibodies. (C) and (F) ChIP using anti-EBNA 3C antibodies.

https://doi.org/10.1371/journal.ppat.1003636.g010

EBNA 3C binding at the ADAM intergenic site directs transcriptional repression via chromatin looping

To determine whether binding of EBNA 3 proteins to unique rather than shared sites was also able to direct chromatin looping, we used chromosome conformation capture to detect potential interactions between the ADAM intergenic EBNA 3 binding site and both the ADAMDEC1 and ADAM28 promoters in an EBNA 3C-expressing B-cell line. Our analysis revealed that the intergenic enhancer was in close proximity to both the ADAM28 promoter and the ADAMDEC1 promoter only in cells expressing EBNA 3C (BJAB E3C-3) and not in control cells (BJAB pz1) (Figure 11). These data indicate that EBNA 3C binding represses ADAMDEC1 and ADAM28 transcription by promoting juxtaposition of the intergenic binding element with the two distal ADAM gene promoters to facilitate H3K27me3 via the recruitment of polycomb repressor complexes. Although our data demonstrate that EBNA 3C is able to promote chromatin looping and gene repression alone, the binding of EBNA 3A to the intergenic site and the fact that the absence of EBNA 3A expression leads to increased ADAMDEC1 and ADAM28 expression also implicates EBNA 3A in the repression of ADAM gene transcription.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 11. The influence of EBNA 3C on chromosome looping at the ADAM28/ADAMDEC1 locus.

(A) Diagram (not to scale) showing the HindIII restriction fragments around the ADAM28 locus that encompass the promoter (P), the ADAM enhancer (E, located downstream of ADAM28) and two intervening control regions (con1 and con2). The arrow indicates the direction of transcription. (B) Chromosome conformation analysis of the ADAM28 locus in the pz1 control BJAB cell line (−) and the E3C-3 stable EBNA 3C expressing cell line (+) using primer pairs that amplify across promoter-enhancer or promoter-control ligation junctions. Positive controls show PCR amplification from control digestion and ligation reactions carried out using PCR-amplified DNA fragments encompassing the promoter, enhancer and control regions. (C) Diagram (not to scale) showing the AciI restriction fragments around the ADAMDEC1 locus that encompass the promoter (P), the ADAM enhancer (E, located upstream of ADAMDEC1) and an intervening control region (con). The arrow indicates the direction of transcription. (D) Chromosome conformation analysis of the ADAMDEC1 locus in the pz1 control BJAB cell line (−) and the E3C-3 stable EBNA 3C expressing cell line (+) using primer pairs that amplify across promoter-enhancer or promoter-control ligation junctions. Positive controls show PCR amplification from control digestion and ligation reactions carried out using PCR-amplified DNA fragments encompassing the promoter, enhancer and control region.

https://doi.org/10.1371/journal.ppat.1003636.g011

CTBP2

Vertebrate genomes contain two genes, CTBP1 and CTBP2, that encode the closely-related proteins CtBP1 and CtBP2, collectively referred to as CtBP. CtBP was first identified as a binding partner of the adenovirus oncoprotein E1A [79]–[80] and promotes cell survival, invasion and metastatic potential through transcriptional repression of a diverse range of tumour suppressors including E-cadherin, PTEN and p16^INK4a (reviewed in [67]). Interestingly, CtBP appears to play a key role in EBV-mediated transformation through interactions with EBNA 3A and 3C [46], [50]. Mutation of the CtBP binding motifs in EBNA 3A and 3C removes their ability to function as co-operative oncogenes in transformation assays and delays outgrowth of immortalized cells on primary infection [46], [50], [53].The epigenetic repression of p16^INK4a by EBNA 3A and 3C via CtBP has recently been shown to be a crucial step in the EBV transformation process [53], [62]. Although our data reveal that CTBP2 expression is repressed by the EBNA 3 proteins, there are no binding sites for EBNA 2 or 3 proteins at the CTBP1 locus indicating that CTBP2 is uniquely targeted. The control of total levels of CtBP levels by modulating CtBP2 expression may therefore be important in controlling EBV transformation, but it is also possible that CtBP2 could be performing functions distinct from CtBP1 in the cell that may limit immortalization. Recent reports have in fact identified a unique role for CtBP2 as a transcriptional activator [81].

Although we have shown that EBNA 3A, 3B and 3C bind the CTBP2 intragenic enhancer, our analysis of mRNA expression in EBNA 3 knock-out cell lines to date has only implicated EBNA 3A and EBNA 3B in the repression of CTBP2 transcription, due to the lack of CTBP2 expression in the EBNA 3C knock-out cell-lines available. The role of EBNA 3C in the regulation of CTBP2 expression therefore remains to be tested. Perhaps surprisingly, analysis of cell-lines conditionally expressing EBNA 2 also failed to detect differential regulation of CTBP2 mRNA levels despite the binding of EBNA 2 to the CTBP2 enhancer. The EBNA 2 ChIP-seq signal at CTBP2 is at least half of the signal at the WEE1 and ITGAL sites so it is possible that the affinity of EBNA 2 binding at the CTBP2 site is lower, making competition for binding with the EBNA 3 proteins less efficient. In EBV-infected cells the binding of the EBNA 3 proteins may therefore be dominant leading to effective inhibition of promoter-enhancer looping.

WEE1

The WEE1 gene encodes the Wee1 kinase, a key negative regulator of the G2/M cell-cycle transition [82]. Wee1 inactivates the mitotic cyclin-dependent kinase CDK1 during interphase through phosphorylation on tyrosine 15. This residue is subsequently dephosphorylated by the CDC25 phosphatase at the end of G2, triggering entry into mitosis. EBV infection results in the deregulation of the cell-cycle, with EBNA 3C playing a key role in overriding G1/S, G2/M and mitotic checkpoints through numerous potential mechanisms that include transcriptional repression and the regulation of protein stability [83]–[84]. Disruption of the G2/M checkpoint may be mediated through effects on the Chk2 kinase and we have recently described upregulation of an activator of CDK1, RGC-32, in EBV-infected cells [69], [85]. Although Wee1 has not previously been implicated in EBV-mediated cell-cycle disruption, a recent microarray study identified WEE1 as the fourth most regulated gene in the ‘cell proliferation’ gene ontology category in newly EBV-infected hyperproliferating B-cells [86]. WEE1 was upregulated in these infected proliferating cells and subsequently downregulated in the resulting immortalized LCLs, although levels remained higher than in uninfected cells. This pattern of expression was also observed for a key set of DNA damage response genes [86]. Interestingly, EBNA 3C was shown to be required for the repression of the DNA damage response induced in response to EBV infection, thus facilitating the outgrowth of permanently proliferating cells [86]. Our data demonstrating that EBNA 3C can repress WEE1 expression, an effect counteracted by EBNA 2, may now also implicate EBNA 3C in suppressing the negative effects of WEE1 on the G2/M checkpoint during the outgrowth of EBV- immortalized cells.

ITGAL

ITGAL encodes alpha L integrin (CD11a) that together with β2 integrin (CD18) forms the heterodimeric integrin, LFA-1 that mediates homotypic and heterotypic adhesion via binding to Intracellular cell adhesion molecules (ICAM) 1–3 [87]. LFA-1 plays an important role in the recruitment of immune cells to sites of infection and promotes T-cell migration and the activation of intracellular signalling cascades [88]–[89]. LFA-1 is upregulated on infection of primary B-cells by EBV along with ICAM1, LFA-3 and a number of B-cell activation molecules, mimicking the B-cell activation phenotype observed on exposure to antigen or mitogens [90]. Expression of the EBV-encoded constitutively active cell-surface receptor LMP1 alone in EBV negative B-cells is sufficient to induce LFA-1 expression on the surface of these cells, implicating LMP1-directed signalling in the activation of ITGAL expression [72], [91].

Our analysis has now revealed that EBNA 2 and a cell-type specific subset of EBNA 3 proteins bind to the ITGAL promoter in vivo implicating the co-ordinated activities of multiple EBV-encoded factors in the regulation of ITGAL expression in EBV-infected cells. EBNA 2 and EBNA 3 proteins appear to bind to this site competitively and EBNA 2 activation is inhibited by the presence of any one of the EBNA3s in reporter assays. Since in vivo only EBNA 3B binds the ITGAL promoter in LCLs and EBNA 3B and EBNA 3C bind in BLs, our reporter assays indicate that when overexpressed all EBNA 3s have the potential to interfere with EBNA 2 activation of the ITGAL promoter, presumably through competitive binding. It is therefore likely that the differential in vivo binding of the EBNA3s to the ITGAL promoter is regulated by the chromatin context and the binding of cellular transcription factors. Interestingly, previous analysis of B-cell lines stably expressing EBNA 2 or EBNA 3C alone reported no change in ITGAL (LFA-1) expression [72], [92]. The effects of EBNA 2 and 3C on the ITGAL promoter may therefore be context-specific, perhaps depending on the expression of cellular DNA targeting factors and the co-ordinated actions of multiple EBV-encoded latent proteins in infected cells. Consistent with this possibility and similar to WEE1, ITGAL expression is upregulated in the early stages of proliferation following EBV infection with expression subsequently reduced in the resulting LCLs [86]. Thus ITGAL expression may be regulated by LMP1 and the EBNAs in a temporal fashion during infection.

Mechanism of EBNA 3-mediated repression

Since the anti-EBNA 3C antibody we used for ChIP-seq analysis independently precipitates EBNA 3A and EBNA 3B, our analysis has mapped binding sites for all 3 EBNA 3 proteins in the human genome. Although these sites would be expected to be enriched for EBNA 3C binding sites since the efficiency of EBNA 3A and 3B precipitation by the antibody is lower (Figure S2), we have successfully confirmed EBNA 3A and EBNA 3B binding sites at multiple loci by Q-PCR. Our data thus provide a useful starting point for the study of gene regulation by all three EBNA 3 proteins.

Interestingly, our studies on shared EBNA 2 and EBNA 3 binding sites support a role for EBNA 3 proteins in the repression of all three EBNA target genes we investigated (CTBP2, WEE1 and ITGAL). EBNA 2 appears to compete for binding to these sites limiting EBNA 3-mediated repression to different extents depending on the gene locus. Thus at CTBP2, the EBNA 3 proteins appear to inhibit gene activation via the prevention of looping even in the presence of EBNA 2, with loss of EBNA 2 activity having little effect on CTBP2 gene expression. At WEE1 however, despite the establishment of chromatin loops that appear to mediate transcriptional repression by EBNA 3C in the presence of EBNA 2, loss of EBNA 2 results in decreased WEE1 expression indicating that EBNA 2 does limit EBNA 3C-mediated repression to some extent.

Our analysis also demonstrates that repression of gene transcription via binding of EBNA 3 proteins to enhancer elements results from the modulation of chromatin architecture in two different ways. At the CTBP2 locus, EBNA 3 proteins prevent the formation of chromatin loops that presumably activate the CTBP2 promoter and at WEE1, EBNA 3C is required for the formation of loops associated with transcriptional repression. At an EBNA 3-only binding site at the ADAM28/ADAMDEC1 locus EBNA 3C again promoted transcriptional repression via enhancer-promoter looping. Interestingly, consistent with our observations at CTBP2, the Drosophila Snail repressor protein has recently been shown to inhibit transcription via the prevention of enhancer looping (anti-looping) [93]. However, we are unaware of any previous reports of the establishment of chromatin loops by transcriptional repressor proteins that facilitate the repression of target genes. Our results therefore provide important new insights into the mechanism of action of transcriptional repressors.

Our previous analysis correlated binding of EBNA 3C to the intergenic distal binding site between the ADAM28 and ADAMDEC1 genes with increases in the polycomb repressor complex-associated H3K27me3 mark across the genes and EBNA 3C-mediated repression of mRNA expression [37]. Consistent with the repressive effects of EBNA 3A on ADAM28 and ADAMDEC1 expression [32], [38], our results now indicate that EBNA 3A co-operates in ADAM gene repression via association with this intergenic binding site. Thus repression directed via enhancer looping from EBNA 3A and EBNA 3C binding sites to gene promoters appears to be mediated in an analogous fashion to the H3K27me3-associated repression of genes that are targeted via promoter binding sites (BCL2L11 and p16^INK4a) [53], [55]. It will be interesting to determine whether EBNA 3C-directed looping at the shared EBNA 2 and 3C WEE1 enhancer binding sites also directs transcriptional repression via increases in H3K27me3, or through other potential effects on the recruitment or phosphorylation of RNA polymerase II.

Specificity of EBNA 3A, 3B and 3C targeting and gene regulation

Our studies have uncovered the molecular basis for the regulation of overlapping and distinct sets of cellular genes by different members of the EBNA 3 protein family. This is mediated by gene-specific, and in some cases cell-type specific, differential binding of EBNA 3A, 3B and 3C to gene regulatory elements. This differential binding is evident at both promoter and enhancer binding sites, although we have yet to identify the mechanism through which this differential binding is achieved.

Since EBNA 3 proteins do not bind DNA directly, binding specificity is likely to be directed through interactions with specific B-cell DNA-binding proteins. Unfortunately, analysis of ENCODE ChIP-seq binding data for a vast array of transcription factors was not useful in identifying the specific players involved in targeting EBNA 2 or 3 proteins to DNA. This analysis revealed that a plethora of B-cell transcription factors associate with EBNA 2 and 3 binding sites in LCLs consistent with an extensive network of protein-protein interactions between transcription factors at open chromatin sites. However, using motif searching we found that EBNA 3 binding sites were enriched for PU.1 (Spi-1) binding motifs highlighting an important role for PU.1 in cellular gene targeting by EBNA 3 proteins in vivo. PU.1 has been previously implicated in EBNA 3C-mediated co-activation of the LMP1 promoter with EBNA 2 via Spi-1/Spi-B sites [45] and indeed we also found that EBNA 2 and 3 shared sites were enriched for PU.1/Spi-1 binding motifs. Since our ChIP-seq analysis identified EBNA 3A, 3B and 3C binding sites it will be important to determine whether EBNA 3A and 3B can direct cellular gene regulation via PU.1 or whether enrichment of PU.1 binding motifs was detected in our study as a result of its enrichment at EBNA 3C binding sites only. It is of interest that EBNA 3C has recently been shown to bind and stabilise interferon regulatory factor 4 (IRF4) [94], a PU.1 binding partner. Dimerization between the DNA binding domains of PU.1 and IRF4 plays a key role in regulation via adjacent PU.1 and IRF4 motifs, such as those found in the mouse immunoglobulin λ light chain gene enhancer [95]. Although the PU.1-dependent co-activation of the LMP1 promoter by EBNA 3C was not dependent on an adjacent IRF4 binding site [45], it is possible that IRF4 may play a role in directing EBNA 3C-mediated cellular gene regulation, potentially in co-operation with PU.1. It is clear however, that the identification of the factors that direct the specificity of binding of EBNA 3A, 3B and 3C will necessitate a multi-pronged approach using a combination of proteomics to identify chromatin-associated binding partners and analysis of the genome-wide binding profiles of the individual EBNA 3 proteins and any co-associated transcription factor binding.

Although our data indicate that EBNA 2 is not bound to shared sites at the same time as EBNA 3 proteins, we have not been able to definitively determine whether the sets of EBNA 3 proteins bound to the same binding site are associated with the site simultaneously using the currently available antibodies. Recent studies have reported an interaction between EBNA 3A and 3C that could potentially explain the high degree of co-operativity between these two proteins in the regulation of cellular genes if these two proteins are bound to gene regulatory elements as a complex [48]. However, we have identified regulatory sites that are bound by EBNA 3C and not significantly by EBNA 3A at WEE1 and ITGAL, so these two proteins do not always appear to be present at the same site. Examination of the individual binding profiles of EBNA 3A, 3B and 3C and further re-ChIP analysis using new reagents will be required to determine whether these proteins do indeed bind simultaneously.

Given that the remaining EBV-encoded nuclear proteins, EBNA-1 and EBNA-LP also function as transcriptional regulators and multiple EBNA-1 binding sites in the human genome have been mapped in BL cells [96], it will be important in future studies to build up a composite picture of how the EBNAs target the human genome to direct cellular reprogramming by mapping shared and unique binding sites in the same cell background and addressing the combined impact of these factors on cellular gene expression. This will enable us to fully explore the way in which EBV deregulates cellular gene expression to further our understanding of cellular transformation.

Cell lines

The EBV-negative BL cell line DG75 and the EBV-positive latency III BL cell line Mutu III expressing all EBV latent proteins have been described previously [60], [97]. The EBV negative BL31 Burkitt's lymphoma cell lines infected with wild-type recombinant EBV Bacmids or EBNA 2 or EBNA 3C knock-out and revertant Bacmids have been described previously and were kindly provided by Dr R. White and Prof M. Allday [54], [98]. The control (pz1) and EBNA 3C-expressing (E3C-3) stable transfectants of the EBV-negative B-cell lymphoma cell line BJAB were previously described [72]. The sets of wild-type (wt1, wt2, wt3) and EBNA 3A knock-out LCLs (mtB1, mtB2, mtB3) have been previously described and were generated from B-cells derived from two different donors (donor 2 (D2) and donor 3 (D3)) using recombinant viruses [32]. The sets of wild-type and EBNA 3B knock-out LCLs were kindly provided by Dr H. Long and were generated using B-cells from two different donors (PER253 and PER142) infected with either wild-type B95.8 EBV derived from the marmoset B95.8 cell line or EBNA 3B-knock-out recombinant virus [99]. The LCL expressing conditionally active ER-EBNA 2 (ER/EB 2.5) has also been described previously [3]. For β-estradiol withdrawal and add back experiments ER/EB 2.5 cells were incubated in the absence of β-estradiol for 4 days and 1 µM β-estradiol was re-added for 6 or 17 hrs prior to cell harvest. All cell lines were routinely passaged twice-weekly and cultured using the conditions previously described for each line.

Chromatin immunoprecipitation

For ChIP-sequencing, EBNA 2 and EBNA 3 proteins were immunoprecipitated from 30×10⁶ cross-linked Mutu III cells as described previously [37]. EBNA 3 proteins were precipitated using sheep polyclonal anti-EBNA 3C antibodies (Abcam, ab16128) that we found also independently precipitate EBNA 3A and 3B as a result of cross-reactivity (Figure S2). EBNA 2 was precipitated using a 6-fold scale-up of previously described methods using 48 µg EBNA 2-specific mouse monoclonal antibody (PE2, gift from Prof M. Rowe), followed by an additional incubation with 81 µg rabbit anti-mouse antibodies (Dako) [100]–[101].

For ChIP-Quantitative PCR (ChIP-QPCR) EBNA 2 was immunoprecipiated from 5×10⁶ cross-linked cells using the PE2 monoclonal antibody as described previously [100]–[101]. EBNA 3A ChIP was carried out using 8 µg sheep polyclonal antibodies (Ex-alpha Biologicals, Inc., F115P) and EBNA 3B ChIP was performed using 8 µg sheep polyclonal antibodies (Ex-alpha Biologicals, Inc., F120P) following the protocol previously described for polyclonal antibodies [101]. EBNA 3C ChIP was carried out using 8 µg of E3cD8 monoclonal antibody [102] (gift from Prof M. Rowe) followed by an additional incubation with 13.5 µg rabbit anti-mouse antibodies (Dako).

Re-ChIP was performed essentially as previously described [101] using the antibody quantities described above and BSA-blocked protein A agarose beads (Merck).

Library preparation, sequencing and data analysis

EBNA 2 ChIP and input DNA was used to generate sequencing libraries that were then subjected to 35 bp single-end read sequencing with an Illumina Genome Analyzer IIx as described previously [37]. ChIP-seq data analysis was carried out as previously described [37] with initial significant peaks of binding identified with MACS (p<10⁻⁷) [103]. The distance between binding sites and RefSeq gene TSS were calculated and the nearest gene TSS identified. 200 bp regions centred on peak positions were considered as binding sites for further analysis. The distance between binding sites and RefSeq gene TSS were calculated and the nearest gene TSS identified. In unique versus shared site comparisons, for a site to be considered bound by one factor only, the p-value for binding had to be below 10-7 for that factor and above 10-2 for the other factor. For a site to be considered bound by both factors, the p-value for binding had to be below 10-6 for each and the two 200 bp binding regions also had to overlap by at least 100 bp.

Quantitative real-time PCR

For transcript analysis, total RNA was prepared using Trireagent (Sigma) and purified using RNeasy columns (Qiagen). cDNA was then synthesised using the ImProm-II reverse transcription system and random oligonucleotides (Promega). Real time PCR was carried out using an Applied Biosystems StepOnePlus PCR machine and the primers indicated in Table S1. Real time PCR reactions contained 3 µl of DNA, 7.5 µl of GoTaq qPCR mastermix (Promega) and 2.25 pmoles of each primer in a total volume of 15 µl. DNA was amplified by heating samples to 95°C for 10 minutes followed by 40 cycles of 95°C for 15 seconds and 60°C for 1 minute prior to dissociation curve analysis. Serial dilutions of cDNA or input DNA (for ChIP) were used to generate standard curves for each primer set. For ChIP analysis, the no antibody control signal was subtracted from the percentage input signal derived from the standard curve.

Chromosome conformation capture

Chromosome conformation capture assays were carried out essentially as described previously [104]. Cells were passed through a 70 µm filter to obtain a single cell preparation. 1×10⁷ cells were then fixed in 1% formaldehyde in the presence of 10% FCS for 10 mins at room temperature. The reaction was quenched with 0.125M glycine, and cells were collected by centrifugation at 230 g at 4°C. The pellet was resuspended in 0.5 ml cold lysis buffer (10 mM Tris-HCl, pH 7.7; 10 mM NaCl; 5 mM MgCl₂; 0.1 mM EGTA) with freshly added complete protease inhibitors (Roche) and lysed on ice for 10 mins. The nuclei were collected by centrifugation at 400 g for 10 min at 4°C, resuspended in 0.5 ml of 1.2X Buffer 4 (New England Biolabs) containing 0.3% SDS and incubated for 1 hr at 37°C while shaking at 900 rpm. Triton X-100 was then added to the nuclei to give a final concentration of 2% and the samples incubated for 1 hr at 37°C with shaking. 400 U EcoRI-HF, HindIII or AciI (New England Biolabs) were added to the nuclei and the samples incubated at 37°C overnight with shaking. The digestion reaction was stopped by addition of sodium dodecyl sulphate to a final concentration of 1.6% and incubation at 65°C for 25 mins with shaking. The sample was then diluted 10-fold with 1.15X ligation Buffer (New England Biolabs) containing 1% Triton X-100 and then incubated for 1 hr at 37°C with shaking. 100 U T4 DNA ligase (New England Biolabs) were added to the sample and the reaction was incubated at 16°C for 4 hrs and then 30 mins at room temperature. 300 µg of Proteinase K (Roche) were added to the sample and the reaction incubated at 65°C overnight. RNA was removed by incubation with 300 µg of RNAse for 45 mins at 37°C. Following two rounds of phenol-chloroform extraction, DNA was ethanol precipitated and analysed by PCR using primers designed to amplify across ligation junctions (Table S1).

As a control for ligation products, genomic DNA regions covering restriction sites of interest were amplified by PCR, purified, mixed in equimolar quantities and then digested with EcoRI-HF, HindIII or AciI (New England Biolabs) for 1.5 hours. Following heat inactivation at 65°C, the digested PCR products were incubated with 10 U T4 DNA-ligase (New England Biolabs) over a temperature range of 4 to 20°C overnight. The DNA was purified using the QIAquick Gel Extraction Kit (Qiagen). Purified positive control DNA was then analysed by PCR using the same primers used for chromosome conformation capture.

Plasmid construction, transient transfections and reporter assays

The ITGAL promoter region from −1756 to +174 bp encompassing all 3 binding sites was amplified from genomic DNA extracted from an LCL using primers designed to introduce SacI and HindIII sites at the 5′ and 3′ end respectively (Table S1) and cloned into pGL3 basic (Promega). The EBV C promoter luciferase reporter construct pCp1425GL2 was described previously [105].

DG75 cells were transfected by electroporation as described previously [100]. Cells were transfected with 2 µg pGL3 basic or pGL3-ITGALp, 1 µg of pRL-TK as a transfection control and 10 or 20 µg of pSG5-2A (expressing EBNA 2), pcDNA3 EBNA 3A, EBNA 3B or EBNA 3C expressing constructs (gift from Dr A.Bell). DNA amounts in each transfection were equalized using empty vector. Cells were harvested after 48 hrs and luciferase assays performed using the Dual luciferase assay kit (Promega) as described previously [100] using a Promega Glowmax multidection system.

Immunoblotting

Immunoblotting was carried out as described previously [100], [105] using the following antibodies: anti-EBNA 3C (mouse monoclonal, E3CA10 [102]), anti-EBNA 3A (sheep polyclonal, Ex-alpha Biologicals, Inc., F115P), anti-EBNA 3B (sheep polyclonal, Ex-alpha Biologicals, Inc., F120P or rabbit polyclonal, Bioss, bs-4698R), anti-EBNA-LP (Santa Cruz biotechnology, Inc sc-23537), anti-actin (Sigma), anti-TFIID (TBP) (Santa Cruz biotechnology, Inc sc-421). EBNA-1 was detected using M.S. human serum as previously described [101].