Genome-wide analysis of host-chromosome binding sites for Epstein-Barr Virus Nuclear Antigen 1 (EBNA1)

ChIP-Seq Analysis of EBV and human genomes

Raji Burkitt lymphoma cells were selected for EBNA1-ChIP-Seq experiments because they maintain a stable copy number of EBV episomes, and because the genomes are incapable of lytic replication (due to a mutation in BALF2), which might complicate ChIP analysis. Anti-EBNA1 monoclonal antibody and IgG control ChIP DNA was analyzed by Solexa-Illumina based deep sequencing methods. Sequence reads were mapped to the EBV or human genomes using the UCSC genome browser http://genome.ucsc.edu/cgi-bin/hgTracks, and a fold enrichment for EBNA1 relative to IgG control antibodies was calculated. A summary of the sequencing reads mapped to the human and viral genome is presented in Table 1. The EBNA1 enriched peaks that mapped to the EBV genome are shown in Figure 1A. We found three major peaks for EBNA1 mapping to the FR, DS and Qp region, as were predicted from earlier genetic and biochemical studies of EBNA1 binding to EBV DNA. No other regions were identified, indicating that these sites are likely to represent the major binding sites of EBNA1 in Raji genomes in vivo. Interestingly, the number of reads was greatest at the DS despite the fact the DNA replication does not consistently initiate from DS in Raji genomes [21, 22]. The DS peak extended into the adjacent Rep* region, suggesting that these auxillary EBNA1 binding sites contribute to the overall signal observed at the DS region [23]. Importantly, these results provide validation that EBNA1 ChIP Seq analysis was consistent with previous biochemical and genetic studies.

Sample	Solexa Illumina Pass Filtered sequence	Mapped to Human Genome	Mapped to EBV Genome	Unmapped
EBNA1	14268722	10783205(75.57%)	123764(0.87%)	3361753(23.56%)
IgG	11961444	8317994(69.54%)	35991(0.30%)	3607459(30.16%)

Initial inspection of EBNA1 binding sites across the human genome revealed a large number of candidate sites (4785 total sites with 903 showing >10 fold enrichment over IgG and peak score >8) with various positions relative to transcription start sites. Among the most remarkable was a cluster of highly enriched EBNA1 binding sites extending over ~40 kb region in chromosome 11, within the intergenic region upstream of the divergent promoters for the FAM55 D and FAM55B genes (Figure 1B and 1C). Numerous smaller peaks of EBNA1 binding were found in close proximity to the start sites of many cellular genes (e.g. MAP3K7IP2 and CDC7), as well at alternative promoter start sites (e.g. HDAC3), and repetitive elements (e.g. LINES) as shown in Figure 2. The density of EBNA1 peaks relative to transcription start sites was calculated (Figure 3A). We found that EBNA1 binding sites with 10 fold enrichment relative to IgG were elevated ~3 fold at the positions -500 to +500 relative to transcription start sites. This is consistent with the reported role of EBNA1 in the regulation of cellular gene expression. EBNA1 binding sites were also analyzed for overlap with repetitive DNA elements (Figure 3B). Over 50% of EBNA1 binding sites overlap with a repetitive element. LINE elements were the most prevalent sites of overlap (Figure 2D and 3B). We also found that EBNA1 was enriched ~2-3 fold at telomere repeat DNA (data not shown). This was intriguing since other studies have found evidence for biochemical interactions between EBNA1 and telomere repeat binding factors, as well as the incorporation of telomere repeat DNA into the DS region of OriP [24]. We also examined EBNA1 binding sites for overlap with reported histone modification patterns in lymphoblastoid and fibroblast cell lines from published ChIP-Seq (Figure 3C) and ChIP-ChIP (Figure 3D) data sets. We found that EBNA1 binding sites are predicted to overlap with major peaks of H3K4me3 (Figure 3C), but also with broader regions enriched in histone H3 K27me3, H4K20me1, and H3K9me1 (Figure 3D).

Identification of cellular EBNA1 binding sites in chromosome 11 and MAP3K7IP2 promoter region

To determine if some of the EBNA1 ChIP-Seq sites were bound directly by EBNA1, we assayed the ability of purified EBNA1 protein DNA binding domain (DBD) to bind candidate sequences in vitro using EMSA (Figure 4). The high occupancy EBNA1 binding sites throughout the genome (> 10 fold enrichment and peak score >8) were analyzed using the MEME web application http://meme.nbcr.net/meme4_3_0/cgi-bin/meme.cgi. Several candidate motifs are shown in Web LOGO format (Figure 4A), and the most common sequences were synthesized as 40 bp probes for use in EMSA. As a positive control, we assayed EBNA1 DBD for binding to the EBV FR DNA (Figure 4B, lanes 1-2). The most significant pattern found was a motif (Chr11.1) that was repeated 41 times in the chromosome 11 cluster. Other significant motifs (Motifs 2-5) were found scattered throughout the genome. We found that EBNA1 DBD bound with relatively high affinity to the Chr11.1 and Motif 2 (Figure 4B, lanes 2-6), but not to motifs Motifs 3, 4, or 5. We also analyzed the peak sequences enriched in ChIP Seq analysis at the CDC7, MAP3K7IP2, and HDAC3 binding sites (Figure 4B, lanes 13-18). Surprisingly, we found that only the MAP3K7IP2 binding site bound EBNA1 DBD directly. Other sites bound with affinities similar to that of a non-specific control for the EBV ZRE1/2 binding element (Figure 4B, lanes 19-20). The finding suggest that many of the EBNA1 peaks in ChIP Seq are either bound indirectly by EBNA1, or are not centered on the specific DNA recognition site bound by DNA.

To determine if EBNA1 bound to several distinct motifs, we rederived the consensus sites for Motif 2 (Figure 4C) and Chr 11 (Figure 4D) using a higher stringency for peak scores >10 and narrower window. We find that these consensus motifs are significantly different from each other and from previously established binding site consensus from EBV genome sites. The chr11 motif is found 771 times in the complete human genome, but is occupied by EBNA1 at only 23 of these sites (> 8 fold enrichment and peak score > 10). Motif 2 is found 429331 times in the human genome, but is occupied by EBNA1 at only 74 sites. These finding indicate that EBNA1 can bind directly to multiple cellular sites in the cellular genome, but actual binding may be restricted by chromatin context. These findings also indicate that EBNA1 can recognize a more degenerate DNA consensus site than previously appreciated. A similar conclusion was reached by Dresang et al. [17]. We also found that many EBNA1 ChIP-Seq binding sites were enriched for motifs that could not bind EBNA1. Among the most significant consensus motifs that did not bind directly to EBNA1 is shown in Figure 4E. Using search algorithms JASPER and TomTom to identify potential overlapping transcription factor recognition motifs, we found that Motif 4 contains a consensus Sp1 (p value .0011) and a Staf/Znf123 (p value .0023) recognition site. The identification of such consensus sites may help to identify cellular factors that mediate EBNA1 interaction with chromosomes through indirect mechanisms.

Validation of EBNA1 ChIP binding sites

To determine what percent of the EBNA1 binding sites determined by ChIP-Seq could be validated by independent ChIP analysis using real-time PCR methods, we assayed 26 independent loci that had varying enrichment signals in ChIP-Seq analysis. As expected, EBNA1 was highly enriched at DS (~4% of input DNA was recovered). Interestingly, a similar enrichment was found for the chromosome 11 cluster (Figure 5A). Almost all of the sites enriched by ChIP-seq were similarly enriched by real-time PCR relative to IgG. Several regions were not enriched, including those for EBV OriLyt, and cellular sites for GAPDH, HFM1, PMF1, and IL6R, which had low enrichment (< 10 fold) in ChIP Seq analysis (Figure 5B and 5C). To determine if EBNA1 enrichment was independent of the monoclonal antibody and the cell type, we assayed the binding of FLAG-EBNA1 after ectopic expression in EBV-293 cells (Figure 5D). We found that FLAG-EBNA1 bound with similar pattern and percent enrichment in Flag-EBNA1 transfected cells as did endogenous EBNA1 in Raji cells (Figure 5C). Similar results were also obtained in EBV positive lymphoblastoid cell lines (LCLs) (data not shown). This indicates that our results were not an artifact of the antibody to EBNA1 and not unique to Raji cells.

Regulation of cellular gene expression by EBNA1

To determine if cellular genes containing EBNA1 binding sites near the transcriptional start site were regulated by EBNA1, we assayed the effect of EBNA1 shRNA depletion. Raji cells were transfected with a plasmid expressing shEBNA1 or control shRNA (shControl), along with a GFP marker gene, and then selected by FACS for transfected cells (Figure 6). Western blot analysis indicated that EBNA1 levels were reduced to ~40% of control levels (Figure 6A) at 96 hrs post-infection. Since EBNA1 is required for Raji cell viability, we also observed a ~2 fold reduction in cell metabolic activity as measured by MTT assay (Figure 6B). To determine if EBNA1 depletion altered gene expression of any of the EBNA1 bound genes, we compared the RNA levels of several candidate genes by RT-PCR (Figure 6C and 6D). For genes with documented alternative promoter start sites, we generated primer pairs that would detect initiation at both transcription start sites. We found that EBNA1 depletion caused a significant reduction of several mRNAs, including HDAC3, MAP3K1, SIVA1, MYO1C, PBX2, NIN (uc001wyk), WASF2, and MDK. We did not find any genes that were upregulated by EBNA1 depletion, suggesting that EBNA1 does not function as a general transcriptional repressor of these tested genes in Raji cells. To further test the role of EBNA1 in cellular gene regulation, we assayed the ability of transiently transfected FLAG-EBNA1 to alter cellular gene transcription in an EBV negative Burkitt lymphoma cell line DG75 (Figure 7). Using this approach, we found that FLAG-EBNA1 transfection stimulated expression of CDC7, HDAC3, MAP3K1, MYO1C, TFEB, and PBX2. RT-PCR of EBNA1 mRNA was used as a positive control for EBNA1 expression. These results suggest that EBNA1 can activate a subset of genes when ectopically expressed in EBV negative Burkitt lymphoma cell lines.

Histone modifications at EBNA1 binding sites

To explore the possibility that EBNA1 may associate with chromatin enriched in a particular histone tail modification, we first assayed the overall correlations of EBNA1 binding sites with reported histone tail modifications in human lymphoid cells (Figure 3C and 3D). Based on this first analysis, we selected a set of histone tail modification-specific antibodies for ChIP assays at several EBNA1 binding sites identified in Raji cells (Figure8). We first assayed histone H3K4me2, which has been previously reported to be elevated at EBNA1 binding sites in the EBV genome [25]. As expected, we found that H3K4me2 was highly elevated at DS and Qp in the EBV genome (Figure 8A). H3K4me2 was also elevated at the cellular EBNA1 binding sites associated with CDC7 and PTPNB. Histone H4K20me1 was found to have a relatively high genome-wide correlation with EBNA1 binding (Figure 3D), and was indeed elevated at DS and Qp, as well as at the cellular EBNA1 binding sites associated with CDC7, Chr11, HDAC3, MAP3K7IP2, and MAP3K1 (Figure 8B). Histone H3K9me3, a mark associated with heterochromatin and repetitive DNA, was found to be highly elevated at the Chr11 repeat cluster (Figure 8C). Histone H3K4me3 and acetylated histone H3 (AcH3) and H4 (AcH4), which are associated with euchromatic and transcriptionally active regions, were elevated at cellular genes for CDC7 and PTPNB, while MAP3K1, MAP3K7IP2, and HDAC3 were found enriched just in AcH3 and AcH4 (Figure 8D-F). These findings suggest that EBNA1 binding may correlate with some histone modifications, but in a manner that is complex and context-dependent.

EBNA1 binding site close to the cMyc-IgG translocation break point in Raji Burkitt Lymphoma

Raji has a rearranged copy of the c-myc gene adjacent to the gamma 1 constant region gene of the human immunoglobulin heavy-chain locus, t(8;14) (q24;q32) [26]. Examination of EBNA1 binding sites in these translocated regions revealed peaks of >3 fold enrichment at the cMyc 3' end of chromosome 8 and >10 fold enrichment within the IgH locus of chromosome 14. In Raji Burkitt lymphoma, these two sites are fused together by a breakpoint in the cMyc and IgH 5' region, thus bringing the two EBNA1 binding sites in close proximity in the translocated allele. Although the mechanism of translocation is unknown, EBV has been considered a potential driving force for the Burkitt's translocations, and it is possible that these EBNA1 binding sites may link these sites to facilitate translocation.

EBNA1 can interact with a large number of cellular binding sites

In this study, we used ChIP-Seq to identify ~903 high occupancy (> 10 fold enrichment and peak score > 8), and ~4300 moderate occupancy (> 3 fold enrichment and peak score >5) binding sites for EBNA1 in the cellular chromosome of a human Burkitt lymphoma cell line. Several (~25) of the high and low occupancy binding sites identified by ChIP-Seq were validated for binding by conventional ChIP and real-time PCR (Figure 5). There was a good correlation between ChIP-Seq and ChIP-PCR for these binding sites, providing a high level of confidence in the ChIP-Seq data. Furthermore, only bonafide EBNA1 binding sites in the EBV genome (FR, DS, and Qp) scored positive in ChIP-Seq analysis (Figure 1A), suggesting that few false positives were generated by this method. Among the high-occupancy binding sites, we noted that ~7% are located within 500 bp of an annotated or predicted transcription start site (Figure 3A). We also noted that ~45% of EBNA1 binding sites overlapped with a repetitive DNA element (Figure 3B). Several DNA motifs could be identified in the high-occupancy EBNA1 binding site data set, but only two of these were found to bind directly to recombinant EBNA1 protein in vitro (Figure 4). Since EBNA1 is known to bind DNA directly through its carboxy-terminal DNA binding domain, and indirectly through its amino terminal tethering domain, it seems likely that many of the binding sites identified by ChIP-Seq represent a composite of these direct and indirect DNA-binding modes of EBNA1.

Identification of high affinity cellular binding sites

A remarkable finding from this study was the identification of a cluster of high-affinity EBNA1 binding sites in chromosome 11. The cluster represents ~10 kb of repetitive sequence situated between the divergent promoters for the Fam55B and Fam55 D genes. The function of the Fam55B and D proteins is not known, and shRNA depletion of EBNA1 had no detectable effect on Fam55B or D gene transcription. This region was elevated in histone H3 K9me3 (Figure 8) suggesting that it is largely heterochromatic and unlikely to be involved in transcription activation. We considered whether this site may represent a cellular origin of DNA replication, but we were unable to identify ORC2 or MCM protein binding at this site (data not shown). Purified EBNA1 DBD protein bound with high affinity to the major repeat elements in the chromosome 11 cluster, indicating that the binding is direct and mediated by the EBNA1 DNA binding domain. At present, it is not clear whether EBNA1 binding to this region of chromosome 11 has any functional significance.

Novel EBNA1 binding sites

Position weighted matrix (PWM) analysis and Web LOGO presentation revealed that many cellular EBNA1 binding sites are distinct from the consensus sites observed at EBV genome binding sites found at the FR, DS, or Qp regions. The chromosome 11 binding site consensus TGG[g/a]TAA[T/C][A/C]A[g/c]TGTT[G/A]CCT and the Motif 2 GG[C/T]AGCAtaT[A/G]CT[A/T][T/C]C do not resemble the consensus derived for previously known EBNA1 binding sites in the viral genome. However, our Motif 2 is similar the new consensus G[A/G][T/C]AGcATaTGCTaCC derived by Dresang et al using 70 viral and cellular binding sites [17]. In a separate study, Canaan et al. identified GaA[G/A]TAT[T/C] as a consensus site for EBNA1 binding at cellular genes subject to EBNA1 dependent regulation and association with EBNA1 protein by ChIP [18]. However, it was not clear from the Canaan et al. study whether these binding sites are bound directly or indirectly by the EBNA1 DNA binding domain. We also identified several motifs enriched in the EBNA1 ChIP-Seq peaks that did not bind directly to EBNA1 DNA binding domain in vitro. These sites may reflect indirect DNA binding by EBNA1, potentially through interactions with other sequence specific factors or chromatin-associated proteins. Cellular factors that have been implicated in mediating EBNA1 tethering to metaphase chromosomes, including EBP2 [27–29], histone H1 [16], and HMGA1 [14, 30], may be good candidates for interactions with some of these indirect binding motifs. In the future, it will be important to determine if there are functional differences between these different classes of EBNA1 binding sites, and what cellular factors mediate the indirect binding of EBNA1 to cellular chromosomes.

EBNA1 regulated cellular genes

EBNA1 may regulate to chromosome structure

The large number of direct and indirect EBNA1 binding sites distal to transcription start sites suggests that EBNA1 may be involved in modulating chromosome structure. EBNA1 is known to mediate long-distance interactions important for transcription of Cp during EBV infection of primary B-cells [8]. EBNA1 is also known to form highly stable homotypic interactions through its linking domains [32], which are thought to play important roles in tethering EBV episomes to metaphase chromosomes [33]. EBNA1 has also been a suspect in EBV-associated lymphomagenesis, especially Burkitt lymphoma where EBNA1 is expressed early and consistently throughout cancer cell evolution. Our ChIP-Seq data revealed that EBNA1 can bind to regions close to the chromosomal translocation break-points in both cMyc and IgG heavy chain enhancer regions (Figure 9), which represents the defining translocation associated with Burkitt's lymphoma. This provides a potential mechanism for EBNA1 in facilitating chromosome translocation by potentially mediating an interaction between these two loci. Further studies will be required to determine whether EBNA1 binding can mediate interactions between these two chromosomal sites in Burkitt's and non-Burkitt's lymphoma cells, and whether EBNA1 mediates long-distance interactions between other cellular and viral chromosomal sites identified in this study.

Cells and Plasmids

Raji cells (human EBV positive Burkitt lymphoma line) and DG75 (human EBV negative Burkitt lymphoma line) were maintained in RPMI containing 10% FBS and supplemented with glutamax (Invtitrogen) and antibiotics (penicillin and streptomycin). EBV-293 contains a hygromycin resistant EBV bacmid in human embryonic kidney (HEK) 293 cells (a kind gift of H. Delecluse) were maintained in RPMI containing 10% FBS, hygromycin (100 μg/ml), glutamax, and antibiotics. pCMV-Flag-EBNA1 was previously described [15]. shRNA directed against EBNA1 was generated by cloning the targeting hairpin sequence (gatatgtctcccctccctcctaggccactcaagcttcaatggcctaggagagaagggagacacatc) into the pENTR/D-Topo vector (Invitrogen).

Chromatin immunoprecipitation (ChIP) Assays

ChIP assays were performed as described previously [34]. Quantification of precipitated DNA was determined using real-time PCR and the standard curve method for absolute quantitation (ABI 7000 Real-Time PCR System). IPs were performed in triplicate for each antibody and the PCR reactions were repeated at least three times and standard deviations were indicated by error bars. Primers for ChIP assays are listed in Additional File 1 in the online Data Supplement. The following rabbit polyclonal antibodies were used for ChIP assays: anti-EBNA1 (305/10 wk), anti-IgG (Santa Cruz Biotechnology), anti-Flag (Sigma), anti-Acetylated histone H3 and H4 (Millipore), anti-dimethylated histone H3K4 (Abcam), anti monomethylated H4k20 (Abcam), and anti-trimethyl histone H3K9 and H3K4 (Millipore). Mouse monoclonal anti-actin (Sigma) and anti-EBNA1 (Advance Biotechnology) were used for Western Blotting.

Quantitative RT-PCR

Briefly, RNA was isolated from 2 × 10⁵ cells using RNeasy Kit (Qiagen) and then further treated with DNase I. Reverse transcriptase PCR (RT-PCR) was done as previously described. Real-time PCR was performed with SYBR green probe in an ABI Prism 7000 according to the manufacturer's specified parameters. Primer sequences for RT-PCR are listed in Additional File 2.

MTT assay

1 × 10⁴ Raji cells were plated in 96-well plates at 96 hrs post-transfection of shEBNA1 or Control shRNA. Cell viability was then measured by incorporation of 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyl tetrazolium bromide (MTT) (Millipore, Cell Growth Assay Kit), according to the manufacturer's protocol.

EMSA

Purified EBNA1 DNA binding domain (DBD) (aa 459-607) was expressed and purified from E. coli as a hexa-histidine fusion protein in Escherichia coli.

Protein-DNA binding reactions contained 10% Glycerol, 200 mM NaCl, 20 mM Tris-Cl pH 7.4, 1 mM DTT, 10 μg/mL BSA, 10 nM ³²P-labeled oligonucleotide DNA and 246 nM purified EBNA1 DBD. Samples were incubated for 20 min at 25°C then loaded onto a 6% polyacrylamide gel and electrophoresed for 90 minutes at 170 V in 1× TBE. Gels were dried and visualized by PhosphorImager. Oligonucleotides used for EMSA are listed in Additional File 3.

Chromatin-Immunoprecipitation for High Throughput Sequencing (ChIP-Seq)

Solexa ChIP-Seq experiments were performed with 2 × 10⁶ Raji cells per IP with either EBNA1 monoclonal antibody or control mouse IgG. ChIP methods were identical to conventional ChIP assays [34] with the exception that competitor salmon sperm DNA was excluded from all IP and wash buffers, and purified ChIP DNA was resuspended in 25 dH₂O. DNA fragments of ~150-300 bp range were isolated by agarose gel purification, ligated to primers, and then subject to Solexa sequencing using manufacturers recommendations (Illumina, Inc.).

Bioinformatics Analysis of ChIP-Seq