Identifying Sites Bound by Epstein-Barr Virus Nuclear Antigen 1 (EBNA1) in the Human Genome: Defining a Position-Weighted Matrix To Predict Sites Bound by EBNA1 in Viral Genomes

Abstract

We identified binding sites for Epstein-Barr virus (EBV) nuclear antigen 1 (EBNA1) in the human genome using chromatin immunoprecipitation and microarrays. The sequences for these newly identified sites were used to generate a position-weighted matrix (PWM) for EBNA1's DNA-binding sites. This PWM helped identify additional DNA-binding sites for EBNA1 in the genomes of EBV, Kaposi's sarcoma-associated herpesvirus, and cercopithecine herpesvirus 15 (CeHV-15) (also called herpesvirus papio 15). In particular, a homologue of the Rep* locus in EBV was predicted in the genome of CeHV-15, which is notable because Rep* of EBV was not predicted by the previously developed consensus sequence for EBNA1's binding DNA. The Rep* of CeHV-15 functions as an origin of DNA synthesis in the EBV-positive cell line Raji; this finding thus builds on a set of DNA-binding sites for EBNA1 predicted in silico.

Epstein-Barr virus (EBV) is a gammaherpesvirus infecting over 90% of the human population. EBV is associated with multiple malignancies including virtually all cases of endemic Burkitt's lymphoma (reviewed in reference 49) and undifferentiated nasopharyngeal carcinoma (4, 58), about 40% of Hodgkin's lymphoma cases (18), and about 14% of gastric carcinomas (53). EBV-positive cells express distinct subsets of viral genes; these distinct patterns of expression are called latency programs I, II, or III. EBV nuclear antigen 1 (EBNA1) is the only viral protein consistently detected in all of these latency programs as well as in all forms of EBV-positive malignancies. The basis for EBNA1's ubiquity is made apparent by the critical roles that it plays within the viral life cycle.

EBNA1 is required for the replication and maintenance of EBV's extrachromosomal genome in host cells. EBNA1 mediates replication and maintenance by binding site specifically to the latent origin of replication, oriP (19). Two distinct loci in oriP are related to these separate functions; the maintenance of the EBV genome occurs at the family-of-repeats locus (FR), and replication initiates at the dyad symmetry locus (DS) (1, 32, 55). An auxiliary origin of replication downstream of oriP, Rep*, is also composed of DNA-binding sites for EBNA1, which harbor similarities to sites in DS (51). Besides these activities, EBNA1 possesses a transcriptional transactivation function. When bound to FR, EBNA1 can positively regulate Cp, a downstream latent promoter (2, 37, 40, 47). Indeed, FR in the presence of EBNA1 can act as an enhancer element for heterologous promoters (17, 42). EBNA1 also binds site specifically to the site III locus within the viral promoter Qp (45). The site III locus consists of two binding sites for EBNA1, which may be involved in autoregulating the transcription of EBNA1 in a manner that is dependent upon latency type (45, 57).

All of EBNA1's known functions depend upon its ability to bind DNA in a site-specific manner. EBNA1 is a stable homodimer (15) that recognizes a DNA-binding site 16 bp in length (3, 5). Previously, a consensus sequence for EBNA1's binding to DNA was determined from multiple palindromic sequences that differed from the palindromic site in FR (FR site 2) of EBV (3). This consensus sequence differs substantially, however, from those sequences bound by EBNA1 in Rep* (27, 51). In an earlier study, EBNA1 was found not to bind human DNA with an affinity comparable to that of its binding of a pair of high-affinity sites, such as those in FR (23). However, Rep* sites are of lower affinity than sites in FR yet do have a biological function (51). We therefore reassessed EBNA1's ability to bind to the human genome in vivo by using chromatin immunoprecipitation (ChIP) and a microarray of DNA sequences from human promoters. Many of these sites were further defined by electrophoretic mobility shift assay (EMSA) and used to generate a position-weighted matrix (PWM) for EBNA1. Using this PWM, we identified DNA-binding sites for EBNA1 in the human genome and in several viral genomes, including EBV, Kaposi's sarcoma-associated herpesvirus (KSHV), and the related species cercopithecine herpesvirus 15 (CeHV-15) (also known as herpesvirus papio 15), which infects rhesus macaques. In particular, a homologue of Rep* was identified in CeHV-15 and was confirmed to function as an origin of DNA synthesis in the EBV-positive cell line Raji. These studies allowed us to refine the set of sequences to which EBNA1 binds.

MATERIALS AND METHODS

Plasmids.

To generate p3723 to p3726, which were used for short-term replication assays, oligonucleotides for Rep* of either EBV or CeHV-15 were introduced into the multiple-cloning site (MCS) of p3488 (30). These oligonucleotides contain a pair of EBNA1-binding sites, their endogenous 5-bp spacers, an EcoRV restriction site, and NheI cohesive ends for a final length of 48 bp. The structures and sequences of these plasmids were confirmed by restriction digestion and sequencing and are available online (http://www.bioinfo.wisc.edu/bxaf4/bm/pcs.php).

Cell lines.

All cell lines, 721 (24), Raji (41), and BJAB (35), were grown in RPMI 1640 medium supplemented with 10% fetal bovine serum, 200 U of penicillin, and 200 μg of streptomycin sulfate per ml (R10F medium).

EBNA1 ChIP-chip.

The EBV-positive lymphoblastoid cell line 721 was used to assay for in vivo EBNA1-DNA interactions according to the ChIP protocol for EBNA1 reported previously (51). Immunoprecipitated DNA fragments along with total input DNA fragments were nonspecifically amplified, labeled, and hybridized to human promoter arrays (NimbleGen, Madison, WI), a process called ChIP-chip. Chromatin samples for ChIP-chip were amplified by two rounds of linker-mediated PCR according to protocols provided by NimbleGen. Amplified chromatin samples from three independent ChIP experiments were pooled for submission to NimbleGen, which performed the labeling and hybridization of the amplified chromatin to arrays (CHAR0150-HP2). The microarray signal obtained by the total input DNA served as a normalization control for the microarray signal obtained by the EBNA1-specific ChIP. The average chromatin fragment size was 500 bp, with a range of 200 bp to 1.7 kb. ChIP-chip results were analyzed using SignalMap software, version 1.8 (NimbleGen, Madison, WI). Criteria for selecting promoters as candidates for containing DNA-binding sites for EBNA1 were a minimum intensity of the signal peak of fourfold above the background level, with a span of at least 500 bp of adjacent probes whose signal intensities were twofold above the background level. A list of the promoters identified by this ChIP-chip experiment is reported in Table S1 in the supplemental material; genes within 10 kb of these promoters are reported in Table S2 in the supplemental material.

EMSA.

EMSA analysis was performed as described previously (51). Modifications to this protocol were made when short oligonucleotides were used, as described below. Individual sites predicted by Multiple Em for Motif Elicitation (MEME)/Motif Alignment and Search Tool (MAST) (see below) were tested by EMSA using short, complementary oligonucleotides (purchased from Integrated DNA Technologies, Coralville, IA). Fixed sequences flanked the 16-bp sites of interest to a final length of 30 bp, for example, 5′-CCCTTTTGGTAGTATGTGCTGCCAAACCCC-3′ and 3′-GGGAAAACCATCATACACGACGGTTTGGGG-5′, where the central 16 bp (in boldface type) is a site of interest. The flanking sequences were selected to generate no additional predicted DNA-binding sites for EBNA1. Complementary oligonucleotides were resuspended in 1× EMSA buffer (20 mM HEPES [pH 7.6], 40 mM KCl, 1 mM EDTA, 1 mM MgCl₂, and 10% glycerol) at a concentration of 25 μM. Samples were placed into a boiling water bath for 10 min and then allowed to cool slowly to room temperature for annealing. All dilutions were carried out in the presence of 1× EMSA buffer to stabilize the annealed products. Clean-up of radiolabeled DNAs was performed according to the protocol for the QIAquick nucleotide removal kit, with elution in 1× EMSA buffer at 5 fmol per μl (Qiagen, Valencia, CA). Eight percent polyacrylamide gels were used for electrophoresis and run at 250 to 300 V at 4°C for 2 to 4½ hours. A complete list of oligonucleotides and primers for all EMSAs can be found in the supplemental material. EMSAs using short oligonucleotides have been quantified and compared to Rep* site 1 and are summarized in Fig. S1 in the supplemental material.

Two protein derivatives of EBNA1 that contain its DNA-binding and dimerization domain (EBNA1-DBD) were used in EMSAs: EBNA1-DBD Softag1 and a His-tagged protein. The Softag1 protein was purified as reported previously (51). His-tagged (His₆) protein was purified using a nickel affinity column, Ni-nitrilotriacetic acid (Qiagen, Valencia, CA). These protein derivatives contain the same amino acid sequence as the DomNeg1 derivative reported previously (25), except that the His-tagged protein does not contain the nuclear localization signal, and they both have added epitope tags. DNAs tested with both EBNA1-DBD derivatives have generated similar results. Fluctuations in the percentage of DNA shifted with equivalent amounts of EBNA1-DBD occur with different batches of purified protein and time of use since purification due to differences in activity or loss of activity over time.

Generation of a PWM and searching sequences.

To increase the number of identified DNA-binding sites for EBNA1, a method similar to phylogenetic shadowing (38) was used. Briefly, the nonredundant sites from FR, DS, site III, and Rep* were combined with those of the related virus CeHV-12 and the palindromic binding sites described previously by Ambinder et al. (3). The sequences chosen from CeHV-12 were from the oriP region (31). The sequences chosen from the study of palindromic base substitutions had at least 10% of the apparent binding affinity for EBNA1 as FR site 2 (3). These 40 sequences were uploaded in FASTA format to the MEME program website to generate a PWM (http://meme.sdsc.edu/meme/cgi-bin/meme.cgi) (6, 8). MEME parameters were one motif per sequence, 16 bp per motif, and a maximum of one motif per output. PWMs were then used to search target sequences, again by uploading FASTA-formatted files, this time to the MAST program website (http://meme.sdsc.edu/meme/search.html) (7).

Position P values of motifs matched to target sequences were used in evaluating potential binding sites for EBNA1. A position P value is a measure of the probability that a random sequence would match the motif (represented by the PWM) as well as the sequence of interest (as described previously by Bailey and Gribskov [7] and on the MAST input-field website [http://meme.sdsc.edu/meme/mast-input.html#ev]). The accession numbers for various viral target sequences are reported in the supplemental material.

Multiple sites were then identified based on the direct binding of EBNA1-DBD by EMSA. These confirmed binding sequences were subsequently added to the list of sites used to generate the PWM. This procedure allowed a reassessment of target sequences, which further affected the position P values associated with predicted sites. The most recently confirmed site to be bound by EBNA1 (the unique site in Rep* of CeHV-15) has not yet been included in the PWM for EBNA1. Rep* site 2, which is consistently associated with the highest position P value of the known binding sites for EBNA1 in MAST searches specific for EBV, has been used as a cutoff for each iteration of generating PWMs. The current set of DNA-binding sequences for EBNA1 used to generate its PWM, along with their associated position P values, is reported in Table S3 in the supplemental material and summarized in Fig. 3B and Table 1.

FIG. 3.

Identification of additional sequences bound by EBNA1 affects its PWM. PWMs are represented graphically as WebLogos, which were generated at the website of the University of California—Berkeley (http://weblogo.berkeley.edu/) (12, 46). MEME was used to perform the multiple alignment of EBNA1's binding sequences and generate its PWM; MAST was used for subsequent searches for predicted EBNA1 binding sites within target sequences, both of which are available on the San Diego Supercomputer website (http://meme.sdsc.edu/meme/intro.html) (6, 7, 8). (A) This PWM for EBNA1 was generated using the nonredundant sequences of FR, DS, site III, and Rep*, the previously known DNA-binding sites for EBNA1 in EBV. These sequences, when unweighted, correspond to the degenerate consensus sequence GRWWRYVYRYVCTDYY. (B) This PWM for EBNA1 was generated using the following 73 sequences: (i) the previously known DNA-binding sites for EBNA1 in EBV as listed above, (ii) those sites confirmed by EMSA analysis, and (iii) those sites predicted from MAST searches associated with low position P values that aligned with signal maxima of the selected promoters. The list of sequences used to generate the 73-site WebLogo in B including their order and orientation is given in Table S3 in the supplemental material.

TABLE 1.

Position P values^a associated with predicted sites for EBNA1

Sequence searched	No. of sites with position P value ranking ofb:						Total no. of sites predicted	Sequence size (bp)a
	10−10	10−9	10−8	10−7	10−6	10−5
Sites in the PWM	1	6	7	29	24	6	73	NAd
Selected human promoters	0	1	12	56	121	306	496	1,344,387
Randomized promotersc	0 ± 0	0 ± 0	0.2 ± 0.4	2.5 ± 2.1	22.0 ± 4.2	188.0 ± 13.9	212.7 ± 16.4	1,344,387
EBV	7	10	5	3	4	27	56	171,823
Randomized EBVc	0 ± 0	0 ± 0	0 ± 0	0.2 ± 0.4	2.0 ± 1.7	20.4 ± 3.9	22.6 ± 3.6	171,823

^aA position P value is a measure of the probability that a random sequence would match the motif (represented by the PWM) as well as the sequence of interest (7).

^bPosition P values are ranked by powers of 10 and grouped by the sequence searched in the leftmost column. Position P values reported are below 1 × 10⁻⁵ and range down to 2.6 × 10⁻¹⁰.

^cRandomized sequences represent 10 iterations of sequence shuffling, followed by searches for DNA-binding sites for EBNA1. Averages and standard deviations are reported.

^dNA, not applicable.

WebLogos.

PWMs were represented graphically as WebLogos, where nucleotide height is proportional to the frequency of occurrence at a given position and where overall height is inversely proportional to the randomness at a given position (12, 46). WebLogos were generated on the University of California—Berkeley website (http://weblogo.berkeley.edu/). The order of sequences and their relative orientation, the forward or the reverse complement sequence, were based on the alignment generated through the MEME program during PWM generation. It should be noted that more than one alignment is possible depending upon the order and orientation of the sequences when uploaded to the MEME website. This variability likely results from multiple equivalent alignment possibilities, which do not significantly alter the overall profile. Variations in these alignments will affect the exact position P values generated in MAST searches as well as the WebLogo. The exact order and orientation of DNA-binding sites for EBNA1 used to generate the 73-site PWM for EBNA1 in current use are given in Table S3 in the supplemental material.

Gene expression array.

The lymphoblastoid cell line 721 was transduced with a retroviral vector encoding Bcl-XL (Bcl-XL internal ribosome entry site [IRES] monomeric red fluorescent protein [mRFP], from p3381) and sorted for efficient mRFP expression on FACS-Vantage SE with the FACS-DIVA option (Becton Dickinson, San Jose, CA). These sorted cells were subsequently transduced with either an empty retroviral vector (IRES green fluorescent protein [GFP], from p3051) or DomNeg1 retroviral vector (DomNeg1 IRES GFP, from p2972) and were sorted for efficient expression of GFP 4 days posttransduction. Total RNA was isolated from sorted cells and converted into cDNA. This cDNA (enriched for polyadenylated RNAs) was used as a template for labeled cRNA synthesis. In cooperation with the University of Wisconsin Biotech facility, labeled cRNA was hybridized to Affymetrix U133 2.0 Plus arrays (Affymetrix, Santa Clara, CA). The Affymetrix GCOS software was used to determine expressed transcripts and then to compare levels of detected transcripts between control (empty retroviral vector) and experimental (DomNeg1 retroviral vector) samples. Transcripts were then selected as candidates for regulation by EBNA1 using a threshold change of a 50% increase or decrease. The results of this expression array can be found on the NCBI website. Genes identified by this expression array whose promoters overlap with those identified by the ChIP-chip experiment are highlighted in Table S2 in the supplemental material.

Luciferase assays.

BJAB, an EBV-negative B-cell line, was transfected by electroporation with luciferase-reporter constructs to determine if EBNA1 directly affects expression from candidate promoters. Human promoter sequences were cloned into an MCS upstream of a luciferase reporter (pGL2 Basic; Stratagene, La Jolla, CA). The promoterless vector (MCS only) was used as a normalization control. The Zs-Green expression vector (pZsGreen1-C1; Clontech, Mountain View, CA) was used to determine electroporation efficiency; pcDNA3 (Invitrogen, Carlsbad, CA) was used as carrier DNA. The FR-TK luciferase reporter construct p985 (36) served as a positive control. FR-TK contains the FR element upstream of the thymidine kinase (TK) promoter; FR in the presence of EBNA1 enhances expression from TK. EBNA1 was expressed from vector p1553; ΔUR1 was expressed from vector p3015. ΔUR1 is a protein derivative of EBNA1 containing a deletion that renders it transcriptionally defective (2, 26). ΔUR1 still possesses the ability to retain plasmids in the nucleus, serving as a control for the apparent effects that EBNA1 may have on transcriptional transactivation due to the trafficking of plasmids into the nucleus (29).

Promoter reporter constructs, pZsGreen1-C1, pcDNA3, and either p1533 or p3015, were electroporated into BJAB cells. A total of 5 × 10⁶ cells was electroporated in 500 μl R10F medium and 25 mM HEPES with a custom-built electroporator at 1,500 V, three capacitor banks with 1,540 mF of capacitance, R-adjust set to maximum, and a rise time set at 10 o'clock, as described previously (28). Two days posttransfection, cells were lysed in 1× cell culture lysis buffer (Promega, Madison, WI) and assayed for light emission on a Monolight 3010 luminometer using a luciferase assay system (Promega, Madison, WI).

Short-term replication assay.

Short-term replication assays were performed similarly to those described previously by Lindner et al. (30). Electroporation conditions were altered to 1 × 10⁷ Raji cells in a 500-μl volume of R10F medium. Extrachromosomal DNAs were recovered by Hirt extraction. Briefly, the plasmids tested in this assay contain FR with one of the following elements downstream to test for activity as an origin of DNA synthesis: (i) MCS (p3488), a negative control (30); (ii) wild-type DS (wtDS) (p3487), a positive control (30); (iii) one copy of Rep* from EBV (p3723); (iv) four copies of Rep* from EBV (p3724); (v) one copy of Rep* from CeHV-15 (p3725); and (vi) five copies of Rep* from CeHV-15 (p3726). The enhanced GFP expression plasmid (p2134) was cotransfected at 2 μg per 1 × 10⁷ cells to normalize for transfection efficiency; test plasmids were transfected at 10 μg per 1 × 10⁷ cells. Three rounds of independent electroporations were performed; each set of samples was split for replicate Southern analysis. Replicate averages were analyzed and compared by performing a Wilcoxon rank-sum test using Mstat 5.10 (N. Drinkwater, McArdle Laboratory for Cancer Research, University of Wisconsin [available at http://mcardle.oncology.wisc.edu/mstat/]).

Microarray data accession number.

Array results for the ChIP-chip experiment, as well as the gene expression experiment, are available on the NCBI website under GEO accession number GSE13315.

RESULTS

Identifying sites bound by EBNA1 in the human genome.

To assay for the binding of EBNA1 to human DNA sequences in vivo, ChIPs for EBNA1 were carried out using nuclear extracts from the EBV-positive lymphoblastoid cell line 721 (24). Immunoprecipitated DNA fragments along with total input DNA fragments were amplified, labeled, and hybridized to arrays that are representative of approximately 27,000 human promoters (NimbleGen, Madison, WI), a process called ChIP-chip. The criteria described in Materials and Methods led to the identification of 247 human promoters as candidates for containing DNA-binding sites for EBNA1 (reported in Table S1 in the supplemental material). The consensus sequence for EBNA1's binding DNA, GRTAGCNNNNGCTAYC (degenerate bases follow IUPAC standardized nucleotide sequences, where R is purine and Y is pyrimidine), described previously by Ambinder et al. (3) occurs in only 2 of the 247 promoter sequences identified in our search. Therefore, an alternate initial screen for DNA-binding sites for EBNA1 was performed using a consensus sequence of increased degeneracy. Combining the 13 nonredundant sites from FR, DS, site III, and Rep*, all of the previously known binding sites in EBV, resulted in the degenerate consensus sequence of GRWWRYVYRYVCTDYY (where W is A or T, V is not T, and D is not C). Fifty-four of the 247 candidate promoters contained sequences that fit this degenerate consensus sequence; the DNA-binding sites for EBNA1 in 46 of these candidate promoters aligned to the peaks of signal intensity in the ChIP-chip data or signal maxima. For each candidate promoter, we define the signal maximum as the peak of signal intensity generated by the EBNA1-ChIP sample normalized to the signal generated by total input DNA for the entire promoter region.

We performed EMSAs to determine whether a subset of the remaining 193 promoters lacking predicted sites contained bona fide DNA-binding sites for EBNA1. Regions of DNA within these promoters that corresponded to the signal maxima were amplified by PCR and incubated with a purified protein derivative of EBNA1 that contains its DNA-binding and dimerization domain (EBNA1-DBD). EMSA analysis revealed that DNA fragments corresponding to these tested promoters shifted in the presence of EBNA1-DBD, indicating that their sequences contain direct DNA-binding sites for EBNA1, whether or not they were predicted to do so based on the degenerate consensus sequence (Fig. 1).

FIG. 1.

ChIP of EBNA1 identifies both predicted and unpredicted binding sites. (A) Shown are the data of a ChIP-chip experiment for a promoter predicted to contain a DNA-binding site for EBNA1 based on the degenerate consensus sequence GRWWRYVYRYVCTDYY. The y axis represents the signal of enriched DNA immunoprecipitated with EBNA1 normalized to the total input DNA on a log₂ scale. This promoter corresponds to chromosome 19, positions 22409571 to 22417404 of the contig reported under GenBank accession number NT_011109.15 (NCBI build 36.2), and is associated with the RRAS and SR-A1 genes. Tick marks along the x axis represent 1,000 bp. The predicted site is located at position 22415154 (represented by the gray circle), which aligns with the signal maximum. (B) Shown is a second promoter identified by the ChIP-chip experiment to bind EBNA1 but lacking a DNA-binding site predicted as described above (A). This promoter corresponds to chromosome 3, positions 53864940 to 53869725 of the contig reported under GenBank accession number NT_022517.17 (NCBI build 36.2), and is associated with the SELK and ACTR8 genes. Tick marks along the x axis represent 500 bp. Aligned below A and B are the regions (approximately 500 to 700 bp) from the RRAS/SR-A1 promoter or the SELK/ACTR8 promoter used in C for EMSA analysis. A DNA fragment containing the two Rep* sites was used as a positive control, yielding two shifted signals in the presence of EBNA1-DBD; a DNA fragment from Raji ori was used as a negative control (50). The DNA fragments are of various sizes, so the unbound DNAs do not move similarly in the gel. Both promoters contain regions (1 and 3) shifted in the presence of purified EBNA1-DBD. Sites identified within these promoters are reported in Table 2 and in Table S3 in the supplemental material.

Developing a PWM to search for additional DNA-binding sites for EBNA1.

We generated a PWM to facilitate the identification of additional DNA-binding sites for EBNA1. A PWM represents a degenerate DNA sequence that incorporates the frequency with which each nucleotide occurs at a given position among a set of sequences, for example, those sequences recognized and bound by a protein of interest. However, a PWM is only as accurately predictive as the set of sequences used for its generation are generally representative. Because there were only 13 distinct sequences known to be bound by EBNA1 in EBV, most of which were similar, we needed to expand the list of sequences identified as binding sites for EBNA1 to improve the PWM. To do this, we used a method similar to the phylogenetic shadowing described previously by Ostrin et al. (38). Briefly, we expanded the number of sequences in the PWM to include binding sites for the homologous EBNA1 within the oriP region (31) of CeHV-12 (also called herpesvirus papio 12), a closely related lymphocryptovirus that naturally infects baboons. We also included the palindromic binding sequences described previously by Ambinder et al. (3), whose binding affinities were at least 10% that of EBNA1 for FR site 2. These 40 combined sites were aligned, and a PWM was generated using the MEME program (6, 8).

The PWM for EBNA1 was then used to search for potential binding sites in the 247 promoters selected by the ChIP-chip experiment via the MAST program (7). Sequences that match the motif of the PWM are given associated position P values; a position P value is a measure of the probability that a random sequence would match the motif as well as the sequence of interest (7). Position P values associated with the predicted sites in the selected 247 promoters are summarized in Table 1. The sites predicted by MAST that aligned with signal maxima were tested by EMSA using overlapping, labeled DNA probes and purified EBNA1-DBD. Multiple sites both were predicted to be bound (Fig. 2A) and were bound by EBNA1-DBD (Fig. 2B). Multiple predicted DNA-binding sites for EBNA1 were also confirmed by EMSAs using short oligonucleotides with 16 bp of test sequence flanked on both sides by 7 bp of constant sequence for each assay (Fig. 2C; see Fig. S1 in the supplemental material).

FIG. 2.

Newly developed PWMs predict DNA-binding sites for EBNA1. (A) Data from a ChIP-chip experiment identifying the promoter associated with the FAM100B and QRICH2 genes as containing binding sites for EBNA1 are shown. Four binding sites for EBNA1 were predicted with its PWM. The promoter corresponds to chromosome 17, positions 8183565 to 8188619 of the contig reported under GenBank accession number NT_010641.15 (NCBI build 36.2). Tick marks along the x axis represent 1,000 bp. As described in the legend of Fig. 1, the y axis represents the signal of enriched DNA immunoprecipitated with EBNA1 normalized to the total input DNA on a log₂ scale. The predicted sites occur at positions 8185422, 8185496, 8185540, and 8185819 (represented by the gray circles), which again align with the signal maximum. PCR-amplified regions of this promoter are aligned beneath its ChIP-chip data. (B) An EMSA was performed using the corresponding PCR-amplified DNA fragments (approximately 200 to 700 bp) for the FAM100B/QRICH2 promoter. A DNA fragment containing the two Rep* sites was used as a positive control, yielding two shifted signals in the presence of EBNA1-DBD. The DNA fragments are of various sizes, so the unbound DNAs do not move similarly in the gel. (C) EMSAs of 30-bp oligonucleotides containing predicted DNA-binding sites for EBNA1 based on the PWM for EBNA1 are shown for the promoters whose associated genes are indicated. Rep* site 1 was used as a positive control; a site predicted in Raji ori at position 148959 that was found to be a false positive was used as a negative control. All DNAs are 30 bp in length and therefore comigrate in the gel. The sites identified in these EMSAs are reported in Table 2 and in Table S3 in the supplemental material; the relative binding affinities for short oligonucleotides are summarized in Fig. S1 in the supplemental material.

A wide range of promoters was selected for analysis by EMSA on the basis of either signal maxima or associated position P values from EBNA1-PWM searches in order to assess the potential of selecting false positives within the selected 247 promoters. However, all of the 23 promoters tested were confirmed to contain direct DNA-binding sites for EBNA1 based on EMSA analysis (Tables 2 and 3). This finding indicates that those promoters that have signal maxima or associated position P values based on the PWM for EBNA1 within this range are likely to contain direct DNA-binding sites for EBNA1. This range encompasses 89% (219/247) of the candidate promoters based on their signal maxima and 90% (222/247) of the candidate promoters based on their associated position P values. Several of these promoters, comparing data from Tables 2 and 3, lack the degenerate sequence GRWWRYVYRYVCTDYY, demonstrating the predictive value of this newly generated PWM for EBNA1. Additionally, position P values associated with confirmed sequences tend to correlate inversely with their relative binding affinities for EBNA1-DBD (see Fig. S1 in the supplemental material). This correlation further separates the PWM from the degenerate consensus sequence, the latter of which does not distinguish the low-affinity Rep* site 2 from the high-affinity FR sites.

TABLE 2.

All 10 of the tested human promoters with a predicted degenerate site^a confirmed to contain a DNA-binding site(s) for EBNA1 based on EMSA analysis

Chromosome	Fig.b	Gene(s) within 10 kb of promoter regionc (bidirectional)	Signal max (log2)	Predicted position P valued
19	1, 2	RRAS, SR-A1	4.20	6.1 × 10−9
11	2	RNF169, XRRA1	4.02	5.2 × 10−8
8	4	ENY2, NUDCD1	3.44	1.4 × 10−6
10		CCDC6	4.32	1.2 × 10−8
2		GTDC1	3.93	1.2 × 10−7
19		KIAA1957, SHC2	3.72	6.1 × 10−8
20		NAPB	3.35	7.0 × 10−8
1		IL6R	3.20	3.4 × 10−7
10		MKI67	3.16	2.0 × 10−7
18		MYO5B	2.16	4.9 × 10−7

^aThe degenerate site was GRWWRYVYRYVCTDYY.

^bAs shown in the figures indicated.

^cSequences for these promoters were obtained based on reported NimbleGen ChIP-chip data and downloaded from NCBI (build 36.2).

^dWhere multiple sites are predicted to align to the signal maximum, only the lowest associated position P value is reported.

TABLE 3.

All 13 of the tested human promoters without a predicted degenerate site^a are confirmed to contain a DNA-binding site(s) for EBNA1 based on EMSA analysis

Chromosome	Fig.b	Gene(s) within 10 kb of promoter regionc (bidirectional)	Signal max (log2)	Predicted position P valued
3	1	SELK, ACTR8	4.64	2.7 × 10−6
17	2, 4	FAM100B, QRICH2	3.73	6.1 × 10−7
13	2	SPATA13, C1QTNF9	3.39	6.1 × 10−8
21	2	CBR1, SETD4	2.85	1.0 × 10−7
8	2	FBXO43, POLR2K, SPAG1	2.56	7.2 × 10−6
1	4	AP4B1, DCLRE1 B	3.98	2.0 × 10−7
7	4	GIMAP5, GIMAP3P, hmm49583, hmm49817	2.92	4.4 × 10−6
7		NFE2L3	3.95	2.9 × 10−6
9		RMI1, LOC729388, HNRPK, NIPSNAP3B	3.71	2.9 × 10−6
X		hmm46563, hmm28428	3.47	2.9 × 10−6
5		LOC389289, LOC153684	3.32	1.7 × 10−5
3		MED12L, P2RY14, GPR87	3.02	4.4 × 10−5
8		ARHGEF10	2.79	7.5 × 10−7

^aThe degenerate site was GRWWRYVYRYVCTDYY.

^bAs shown in the figures indicated.

^cSequences for these promoters were obtained based on reported NimbleGen ChIP-chip data and downloaded from NCBI (build 36.2).

^dWhere multiple sites are predicted to align to the signal maximum, only the lowest associated position P value is reported.

Multiple iterations of confirming, recalculating the PWM, and searching again extended the experimentally recognized DNA-binding sites for EBNA1 to a total of 73: (i) the 13 nonredundant DNA-binding sites for EBNA1 in EBV, including FR, DS, site III, and Rep*; (ii) 33 sites confirmed by EMSA analysis; and (iii) 27 sites associated with low position P values generated in MAST searches that align solely with signal maxima of the selected promoters. This PWM for EBNA1 does not include the untested sites that were initially used to improve previous iterations of the PWM. PWMs for EBNA1 based upon either the 13 previously known DNA-binding sites for EBNA1 in EBV (Fig. 3A) or the experimentally recognized 73 sites (Fig. 3B) are represented graphically as WebLogos, where nucleotide height is proportional to the frequency of occurrence at a given position and where overall height is inversely proportional to the randomness at a given position (12, 46). The list of these 73 sites including their associated position P values is given in Table S3 in the supplemental material and is summarized in Table 1. The position P values of additional predicted sites in viral genomes are given in Table S4 in the supplemental material.

EBNA1 does not alter the expression of a reporter gene placed downstream of tested cellular promoters to which it binds.

We measured the levels of cellular mRNAs to determine if EBNA1's binding to their promoters perturbed their accumulation. To do so, a dominant negative derivative of EBNA1 was introduced into 721 cells previously engineered to express Bcl-XL efficiently. The forced expression of Bcl-XL overcomes the apoptosis induced by the loss of EBV when this dominant negative derivative of EBNA1, DomNeg1, is expressed (25; data not shown). We compared the levels of cellular mRNAs in 721 cells in the presence of Bcl-XL and either DomNeg1 or an empty vector by expression array (Affymetrix, Santa Clara, CA). To identify changes in gene expression in the presence of DomNeg1, we used a threshold change of either a 50% increase or decrease. By this criterion, a total of 5,903 transcripts as defined by the expression array probe set were altered; 103 of these transcripts' promoters were bound by EBNA1 in the ChIP-chip experiment, corresponding to 83 genes highlighted in Table S2 in the supplemental material. The detected differences in accumulation could result directly or indirectly from the inhibition of EBNA1 by DomNeg1.

These results indicate that approximately 11% of the transcripts probed had altered levels when DomNeg1 was expressed in 721 cells. Approximately 1% of the probed promoters had sites bound by EBNA1, and of these bound promoters, 14% of their associated transcripts probed were altered when DomNeg1 was expressed in 721 cells. These changes, 11% and 14%, are sufficiently similar, so we cannot conclude that the expression from promoters bound by EBNA1 is preferentially altered in these experiments.

We used luciferase reporter assays in the presence of EBNA1 but in the absence of the EBV genome to determine if EBNA1 directly affected these promoters. Four genes, two of which were inhibited (DCLRE1B and NUDCD1) and two of which were stimulated (FAM100B and GIMAP5) by the expression of DomNeg1 in 721 cells, were analyzed. Approximately 3 kb of sequence upstream of the transcriptional start site up to the beginning of the protein-coding sequence was isolated. The DNA-binding site(s) for EBNA1 within all four of these promoters was present in these isolated DNAs, as verified by EMSA (data not shown). These DNAs were cloned upstream of luciferase (Fig. 4A), and the resulting vectors were introduced into the EBV-negative B-cell line BJAB (35) and assayed for a dose-dependent response to the presence of EBNA1 (Fig. 4B). A derivative of EBNA1 that is transcriptionally defective but still possesses the ability to retain plasmids in the nucleus, ΔUR1, was used as a control (2, 26, 29).

FIG. 4.

Analysis of four cellular promoters bound by EBNA1 in luciferase reporter assays. (A) EBV-negative BJAB cells were electroporated with vectors encoding luciferase downstream of either the FR-TK promoter, a promoter identified by ChIP-chip, or an MCS (normalization control). (B) Increasing amounts (0, 1, 3, or 10 μg) of vectors expressing either EBNA1 or ΔUR1 were cotransfected with these luciferase vectors. Two days after electroporation, cell lysates were analyzed for luciferase activity. The FR-TK-luciferase construct served as a positive control. The transcriptionally defective derivative of EBNA1, ΔUR1, serves as a control for any effects that EBNA1 may have on transcriptional transactivation due to the trafficking of plasmids into the nucleus (29). Results are averages from at least two independent experiments, plus or minus the standard deviation, except for the FAM100B construct, which was tested once. Chromosomal coordinates (Chr) are reported along the construct diagrams and are based on NCBI build 36.3.

In the absence of either EBNA1 or ΔUR1, the four vectors were active relative to the promoterless vector in BJAB cells, yielding a 26- to 82-fold increase in luciferase activity. This level of activity is similar to the basal activity of FR-TK (27-fold increase), a previously described vector that can be regulated by EBNA1 (36). However, in contrast to FR-TK, we found no dose-dependent effect, either positive or negative, in the presence of EBNA1 for any of the four cellular reporters tested. Thus, although EBNA1 binds these promoters in vivo, it appears not to regulate them in these assays in BJAB cells.

We then compared the results of our assays of gene expression with two recent studies identifying cellular genes potentially regulated by EBNA1 in epithelial cell lines (52) and Hodgkin cell lines (14). These previously published studies identified transcripts whose levels were affected by the expression of EBNA1, corresponding to a total of 551 genes; 242 of these genes' transcripts also had altered levels of expression in our gene expression assays. However, only three of the promoters of these genes with altered expression were bound by EBNA1 in our ChIP-chip assay; two additional promoters were bound by EBNA1 in our ChIP-chip assay, whose transcript levels were not altered in the gene expression assay. This limited overlap of genes identified by both ChIP-chip and gene expression assays may reflect the genes being indirect targets of EBNA1.

Predicting sites in viral genomes bound by EBNA1.

We also used the PWM for EBNA1 to search the sequences of EBV, the related lymphocryptoviruses CeHV-12 and -15, the rhadinoviruses KSHV and CeHV-17 (also gammaherpesviruses), and the papillomaviruses bovine papillomavirus type 1 (BPV-1) and human papillomavirus type 16 for sites likely bound by EBNA1. Predicted DNA-binding sites for EBNA1 were found in the closely related lymphocryptoviruses CeHV-12 and -15, which have comparable FR and DS loci based on site clustering, spacing, overall separation of these loci, and function (31, 39, 43). The loci of CeHV-15 were also aligned using ClustalX (see Fig. S2 in the supplemental material) (48). It was determined previously that EBNA1 of EBV can maintain CeHV-12 plasmid replicons (containing CeHV's oriP) and, conversely, that the EBNA1 homologue of CeHV-12 can maintain EBV plasmid replicons, with preference for their homologous oriP sequences (21, 39, 56). In CeHV-15, site III was predicted, consistent with a previous report (44).

The rhadinoviruses KSHV and CeHV-17 are more distantly related to EBV than are CeHV-12 and -15. KSHV and CeHV-17 contain few, nonclustered, predicted sites for binding EBNA1. Two predicted sites in KSHV were experimentally confirmed (Fig. 5A), and two were eliminated (data not shown). Only two sites were predicted for BPV-1, and no sites were predicted for human papillomavirus type 16.

FIG. 5.

DNA-binding sites for EBNA1 are predicted and confirmed in viral genomes. (A) An EMSA of predicted DNA-binding sites for EBNA1 within KSHV is shown. A site predicted in Raji ori at position 148959 and found to be a false positive was used as a negative control. All DNAs are 30 bp in length and comigrate in the gel. The relative binding affinities for short oligonucleotides are summarized in Fig. S1 in the supplemental material. (B) An EMSA was performed using PCR-amplified DNA fragments corresponding to specific regions in EBV (DNA fragments were between 400 and 500 bp in length). A DNA fragment containing the two Rep* sites was used as a positive control, yielding two shifted signals in the presence of EBNA1-DBD; a DNA fragment from Raji ori was used as a negative control (50). The DNA fragments are of various sizes, so the unbound DNAs do not move similarly in the gel. The sites predicted previously by Horner et al. (23) correspond to positions 54350 and 56173 in EBV and are contained within the DNA fragments in lanes 7 to 9 and 10 to 12, respectively. Newly predicted sites occur at positions 56412 and 129878 and are contained within the DNA fragments in lanes 13 to 15 and 16 to 18, respectively.

Two potential DNA-binding sequences for EBNA1 were predicted previously by Horner et al. (23) in the BamHI U region of EBV based on an alternate PWM derived from the in vitro palindromic binding sites described previously by Ambinder et al. (3). The presence of EBNA1-binding sites was verified by EMSA analysis of the BamHI U region after restriction digestion (23). These sites correspond to positions 54350 and 56173 (positions reported here are based on the wild-type sequence of EBV [GenBank accession number NC_007605]; positions reported previously by Horner et al. [23] are relative to those of strain B95-8 [GenBank accession number V01555]). Another binding site was predicted at position 56412 using the 73-site PWM for EBNA1. Positions 56173 and 56412 both occur within the restriction digestion fragment of the BamHI U region tested formerly (23). EMSA analysis of the BamHI U region revealed that the fragment containing site 56412 was shifted approximately 7- to 13-fold more in the presence of EBNA1-DBD than the fragment containing site 56173 (Fig. 5B). The fragment containing site 54350 was also shifted. These sites did not align with any sites predicted in CeHV-15 (data not shown). A site at position 129878 predicted in EBV was also confirmed by EMSA analysis (Fig. 5B). The function of these sites is currently unknown.

Short-term replication of a Rep* homologue predicted for CeHV-15, rhRep*.

A Rep*-like locus was also predicted for CeHV-15 by the PWM for EBNA1, again based on spacing between binding sites, overall distance relative to the DS locus, and ClustalX alignment (Fig. 6A; see Fig. S2 in the supplemental material). In contrast, when CeHV-15 was searched using the consensus sequence developed previously by Ambinder et al. (3), only 13 FR-like sites were predicted, even though one of the sites in this Rep*-like locus is identical to FR site 5 of EBV. When the degenerate consensus sequence GRWWRYVYRYVCTDYY was used in this search, two of the binding-sites predicted, one within the DS-like locus and one within the Rep*-like locus, were lost, and as a result, no origin based on a 21-bp center-to-center spacing was predicted. Pairs of binding sites for EBNA1 separated by 21 bp center to center are present in DS and Rep* of EBV, and their spacing is essential for their function (9, 51). CeHV-15 naturally infects rhesus macaques, and thus, we will refer to this newly identified homologue as rhRep*.

FIG. 6.

A Rep* homologue, rhRep*, is identified by the PWM for EBNA1 in CeHV-15 and is confirmed to function as an origin of DNA synthesis. (A) Alignment of the two homologous Rep*s of EBV and CeHV-15 was performed using ClustalX (48). Binding sites for EBNA1 are shaded in gray, with emphasis on the 21-bp center-to-center spacing. This homologue is termed rhRep*, after the host that CeHV-15 naturally infects, rhesus macaques. Asterisks below these sequences indicate identical nucleotides in the alignment. Positions within each virus are indicated above and below these sequences. (B) EMSA analysis of the two homologous Rep*s was performed using complementary oligonucleotides. DNAs are each 48 bp in length and therefore comigrate in the gel. (C) rhRep* was tested as an origin of DNA synthesis in a short-term replication assay. Briefly, 1 × 10⁷ Raji cells were electroporated with 10 μg of the plasmid to be tested along with 2 μg GFP expression plasmid, which was used to normalize for transfection efficiency. Cells were harvested 4 days posttransfection. The DNA was isolated by Hirt extraction, digested with DpnI and MluI, and used to perform replicate Southern assays per experiment. The plasmids tested all contain FR along with varied test sequences as potential origins. MCS (p3488) was used as a negative control; wtDS (p3487) was used as a positive control (30). (D) Efficiencies of DNA synthesis as determined by a short-term replication assay are summarized graphically for each of the plasmids tested. Values are adjusted for transfection efficiency as determined by GFP expression (p2134) 2 days after electroporation; wtDS is set to 100%. Averages of three independent experiments are reported along each sample bar with standard deviations as indicated. The difference in replication efficiency was determined to be statistically significant between samples 1× rhRep* and 1× Rep*, as well as between samples 1× rhRep* and 5× rhRep* (P = 0.025 each), using the Wilcoxon rank-sum test statistic.

We tested if rhRep* could function as an origin of DNA synthesis in the presence of EBNA1 by performing short-term replication assays. First, we confirmed that these sites are bound by EBNA1 in vitro by EMSA using complementary oligonucleotides containing both of the predicted sites; rhRep* bound EBNA1 more efficiently than did EBV's Rep*, requiring only one-tenth to one-fifth as much protein for its half-maximal binding (Fig. 6B). rhRep* was inserted either once (p3725) or multiple times (p3726) into an MCS downstream of FR and introduced by electroporation into Raji cells, an EBV-positive cell line (41). Likewise, Rep* was inserted either once (p3723) or multiple times (p3724) into an MCS downstream of FR to compare the efficiencies of DNA synthesis for these homologous loci. As shown in Fig. 6C, the plasmids containing rhRep* were synthesized 4 days posttransfection more efficiently than parallel ones containing EBV's Rep* (Fig. 6D). A plasmid containing multiple copies of rhRep* was more efficiently synthesized than a plasmid containing only one copy of rhRep* (P = 0.025). This observation is compatible with previous work correlating the efficiency of DNA synthesis with increasing copy numbers of Rep* (27). The finding that a plasmid containing one copy of rhRep* was more efficiently synthesized than a plasmid containing one copy of Rep* (P = 0.025) may reflect its containing higher-affinity binding sites for EBNA1 (30).

DISCUSSION

EBNA1 is central to EBV's latent infection. It is required for the maintenance of the viral genome in proliferating cells; it is required for recruiting the origin recognition complex to DS and to the auxiliary origin, Rep*; and finally, it is required for transcription from the viral Cp promoter used during latent infection (20). All of these activities depend upon EBNA1 binding site specifically to DNA. The sequence requirements for this binding by EBNA1 were appreciated to be poorly understood after the identification of Rep*, providing an impetus for reassessing what sequences can be bound by EBNA1 and if EBNA1 binds to the human genome. This possibility was tested by performing ChIPs of EBNA1 from nuclear extracts of the EBV-transformed lymphoblastoid clone 721. The DNA enriched by EBNA1 immunoprecipitations was then labeled and used in a microarray of human promoters to identify sequences as candidates for binding EBNA1 in vivo.

We have identified DNA sequences bound by EBNA1 in cells with an iterative approach. First, DNA sequences bound by EBNA1 were confirmed with EMSAs. These sequences were then used along with previously known sequences to construct a PWM, a representation in which the nucleotides at a given position within a consensus sequence are weighted by the frequency with which they occur. The PWM was then used to predict additional sequences bound among the candidate promoters, which were then tested by EMSAs. Those sites that were confirmed were used to refine the PWM, which in turn was used to search for additional binding sites yet again. This iterative process yielded the PWM represented by the WebLogo in Fig. 3B, which has predicted formerly unknown binding sites for EBNA1 in the genomes of EBV, KSHV, and CeHV-15 (Fig. 5 and 6; see Fig. S2 in the supplemental material).

These experiments have led to an improved definition of the sequences that EBNA1 binds and provided an independent means of validating the findings derived from the crystal structure of its DNA-binding and dimerization domain. DNA-specific contacts occur through lysine residue 477 at positions ±7 to ±8 in the crystal structure (10, 11). In addition, structural homology between EBNA1 and the E2 protein of BPV-1 (11) and comparative mutational studies (13, 16) indicate that the recognition helix of EBNA1 will make DNA-specific contacts similar to that of the recognition helix of E2, which occur at positions ±4 to ±6 (22). The base frequencies at each position represented by the 73-site PWM for EBNA1 (Fig. 3B) are consistent with positions ±4 to ±8 contacting the protein, with perhaps less specificity at position ±6.

Position P values associated with predicted sites can be used to rank their likelihood of being confirmed to bind EBNA1 experimentally (Table 1) as well as to estimate their relative binding affinity for EBNA1-DBD (see Fig. S1 in the supplemental material). The previously known binding sites for EBNA1 in EBV have position P values that rank as follows: those in FR occur between 10⁻¹⁰ and 10⁻⁸, those in DS occur between 10⁻⁹ and 10⁻⁶, those in site III occur at 10⁻⁷, and those in Rep* occur between 10⁻⁶ and 10⁻⁵. These rankings tend to correlate with prior analyses of the relative binding affinities for EBNA1 for these previously described EBV loci (3, 30, 51). For example, the relative affinities of EBNA1 for FR, DS, and Rep* have been measured in competitive EMSAs to be 12 to 6 to 1 (30). The sequences of EBV and of the 247 human promoters identified by ChIP-chip were randomized and predicted to have fewer binding sites for EBNA1 with higher position P values than their corresponding unshuffled sequences (Table 1). This analysis is consistent with both the sequences of EBV and the 247 human promoters identified by the ChIP-chip experiment being enriched for bona fide DNA-binding sites for EBNA1 relative to a DNA of random sequence.

While EBNA1 clearly binds sites within cellular promoters in EBV-positive B cells, we did not observe an EBNA1-dependent regulation from promoters tested in gene reporter assays (Fig. 4). This finding is consistent with previous studies that indicate that six or seven binding sites are required to detect EBNA1's stimulating reporter genes in assays akin to those shown in Fig. 4 (54). It is not clear, however, if these luciferase assays performed in BJAB cells accurately reflect the ability of EBNA1 to regulate genes, particularly out of context of the endogenous chromatin structure. The binding of EBNA1 to cellular promoters could affect them subtly by enhancing or excluding the binding of other factors in ways not detected by our reporter assays. The high levels of DNA templates in transfected cells could also mask effects of EBNA1's regulation of the reporter gene. Future studies focused on EBNA1-dependent effects independent of EBV may help to elucidate direct gene targets of EBNA1, allowing a better assessment of any regulatory role of EBNA1 for genes associated with the promoters identified by the ChIP-chip experiment. One benefit of our computing a robust PWM is that the entire human genome, rather than the 4% of its sequences represented in the promoter microarrays that we used, can now be searched for clusters of EBNA1-binding sites. Such clusters of sites could inhibit transcriptional elongation or even serve as sites for tethering EBV genomes via EBNA1's linking activity (34).

That a homologue of the auxiliary origin of DNA synthesis, Rep*, was predicted for the closely related lymphocryptovirus CeHV-15 was surprising. However, the Rep* of CeHV-15, rhRep*, aligns to Rep* of EBV, and the spacing between binding sites for EBNA1 is conserved. Both DS and Rep* in EBV have pairs of DNA-binding sites spaced 21 bp center to center, which is essential for their function (9, 51). rhRep* is bound by EBNA1 and functions as an origin of DNA synthesis (Fig. 6). rhRep* represents a functional set of EBNA1 DNA-binding sites first predicted by the PWM developed for EBNA1 of EBV. rhRep* functions as an origin of DNA synthesis more efficiently than Rep*, suggesting that rhRep* may contribute more to the synthesis of CeHV-15 than Rep* contributes to the synthesis of EBV. This hypothesis is predicated upon EBNA1 of CeHV-15 binding DNA and functioning comparably to EBNA1 of EBV. The identification of rhRep* implies that there has been evolutionary conservation of this locus among viruses closely related to EBV.

The identification of sites in KSHV DNA that can be bound by EBNA1 may help to explain phenotypes observed in pleural effusion lymphomas (PELs). We have found that forcing the loss of EBV from PELs dually infected with KSHV and EBV by expressing a dominant negative derivative of EBNA1 inhibits their ability to form colonies under limiting dilution (33). Only those cells that retain EBV can form colonies (33). Surprisingly, when these derivatives of EBNA1 are expressed in PELs singly infected with KSHV, these cells also fail to form colonies under limiting dilution (33). This phenotype of inhibited colony formation was not detected in five of five EBV-negative, KSHV-negative hematopoietic cell lines assayed (25, 33). We hypothesize that EBNA1's binding site specifically to the KSHV genome alters KSHV's support of proliferation of singly infected PELs. If this hypothesis is correct, it would identify a new mechanism for EBNA1's DNA-binding function, which we are now testing.