|Home | About | Journals | Submit | Contact Us | Français|
Epstein-Barr Virus (EBV), which is associated with multiple human tumors, persists as a minichromosome in the nucleus of B-lymphocytes and induces malignancies through incompletely understood mechanisms. Here, we present a large-scale functional genomic analysis of EBV. Our experimentally generated nucleosome positioning maps and viral protein binding data were integrated with over 700 publicly available high-throughput sequencing data sets for human lymphoblastoid cell lines mapped to the EBV genome. We found that viral lytic genes are coexpressed with cellular cancer-associated pathways, suggesting that the lytic cycle may play an unexpected role in virus-mediated oncogenesis. Host regulators of viral oncogene expression and chromosome structure were identified and validated, revealing a role for the B-cell-specific protein Pax5 in viral gene regulation and the cohesin complex in regulating higher order chromatin structure. Our findings provide a deeper understanding of latent viral persistence in oncogenesis and establish a valuable viral genomics resource for future exploration.
Viruses co-evolve with their hosts to establish stable and co-regulated genomes and gene expression programs (Iyer et al., 2006). DNA tumor viruses are distinguished by their ability to provide a selective growth advantage to host cells and typically establish long-term persistent intracellular infections (Moore and Chang, 2010). Among the most extensively characterized human tumor viruses is Epstein-Barr Virus (EBV), which has been implicated as a causative agent in multiple B-cell lymphomas, gastric carcinomas, and nasopharyngeal carcinomas (Rickinson and Kieff, 2007; Young and Rickinson, 2004). EBV is estimated to be responsible for ~1% of all human cancers and may also contribute to other disorders, including multiple sclerosis (Ascherio and Munger, 2010; Parkin, 2006). While EBV oncogenes and regulatory pathways have been characterized individually, there have been few studies that examine the network of virus-host interactions at a genomic scale.
Methods emerging from systems biology and functional genomics provide powerful approaches for elucidating these viral-host interaction networks (Aderem et al., 2011). Chronic infection by DNA tumor viruses, such as EBV, is particularly well suited for systems-level interrogation. EBV genomes persist as multicopy DNA episomes in the nucleus of human B-lymphocytes (Lieberman, 2006; Lindner and Sugden, 2007), which can be assayed by modern high-throughput sequencing methods. EBV infection of primary human B-lymphocytes leads to the efficient establishment of continuously proliferating genetically stable human lymphoblastoid cell lines (LCLs). Additionally, the EBV expression program in LCLs allows for the assay of the full repertoire of viral latency genes, which are able to drive cellular proliferation and survival in vitro and in vivo in immunocompromised hosts (Thorley-Lawson and Gross, 2004). LCLs have also been extensively characterized by human genetic studies, including large-scale consortium projects such as ENCODE and HapMap(ENCODE Consortium, 2007; International HapMap Consortium, 2010), which have created vast repositories of genotype, gene expression, and cellular phenotype data that can be leveraged to better understand host and viral regulatory networks. The contribution of EBV to LCL genome biology and the comprehensive mapping of functional elements of the EBV genome have not been evaluated in these prior studies.
Previous genomics studies of human viruses have been limited in scope, exploring novel virus discovery (Feng et al., 2008), environmental niche characterization (Breitbart et al., 2003; Reyes et al., 2010; Tadmor et al., 2011), in vitro protein interactions (Calderwood et al., 2007; Dyer et al., 2008; Pinney et al., 2009), and small-scale functional genomics. Many of these studies have been done in herpesviruses, such as a RNA-seq analysis of a Burkitt lymphoma cell line (Lin et al., 2010; Xu et al., 2010), a low density qPCR primer array for chromatin immunoprecipitation (ChIP) on a limited set of host factors (e.g. CTCF and a small set of histone modifications) (Tempera et al., 2010), and ChIP of the viral factors EBNA1(Dresang et al., 2009; Lu et al., 2010) and BZLF1(Bergbauer et al., 2010). There has also been recent exploration of the KSHV epigenome (Günther and Grundhoff, 2010; Stedman et al., 2008; Toth et al., 2010), though it is unclear if these epigenetic controls are conserved between gammaherpesviruses.
Here, we present a large-scale functional genomic analysis of EBV, which provides insights into viral pathogenesis and B-cell biology. We integrated our own experimentally generated nucleosome positioning maps and viral protein binding studies with over 700 publicly available high-throughput sequencing data sets for human LCLs that we have mapped to the EBV genome. Many of the cell lines we examined were created by the HapMap Project, and one of these was extensively assayed by the ENCODE Consortium (ENCODE Consortium, 2007; International HapMap Consortium, 2010). Using this massive repository of combined host and virus data, we create a comprehensive atlas of interactions between host factors and the EBV genome. We characterized genome-wide binding profiles of over 60 human transcription factors, which cluster into specific regulatory regions, suggesting combinatorial control of viral gene expression. We discovered host genes that are coexpressed with viral genes and cluster into B-cell proliferation pathways. We demonstrated that the host B-cell specificity factor Pax5 plays an unexpected role in regulating viral gene expression and chromatin organization. Finally, we characterized the Cohesin-mediated spatial conformation of the viral episome and demonstrate its relevance in gene regulation through knockdown of its structural components. Our study represents the exploration of functional elements and regulators of Epstein-Barr virus on a large scale. All raw and processed data are publicly available at http://ebv.wistar.upenn.edu.
To generate a comprehensive atlas of functional elements for the EBV genome, we analyzed many existing large-scale functional genomics data sets for EBV positive LCLs (Table S1). We examined 319 RNA-seq experiments from five separate studies (Cheung et al., 2010; ENCODE Consortium, 2007; Kasowski et al., 2010; Montgomery et al., 2010; Pickrell et al., 2010), covering LCLs from 143 donors, multiple RNA size ranges, cellular compartments and both poly(A)+ and poly(A)− transcripts. We explored DNA-binding proteins and histone modifications and variants by incorporating more than 300 ChIP-seq experiments from the ENCODE project (ENCODE Consortium, 2007). We also integrated cytokine peptide levels (Choy et al., 2008) and related independent functional genomic studies (Lee et al., 2011; Lu et al., 2010; Ramagopalan et al., 2010).
To ensure that we characterized reads of viral origin and excluded reads from homologous human regions, we subtracted reads mapping to the human genome prior to alignment against the EBV genome (Figure 1). This subtraction did not affect mappability of the EBV genome since only a small number of highly repetitive EBV elements are unmappable with reads longer than 24 nucleotides (Figures 1B and S1A–D). Alignments to EBV accounted for a significant number of reads in nearly all assays (Fig. 1C, and Table S2). In total, we aligned over 166 million reads to the EBV genome, which accounted for 1.2% of all mappable reads. As a negative control, we aligned reads from experiments performed in uninfected cell lines and primary tissue and found negligible alignment to EBV (Figure S1E and F).
To explore the overall organization of the functional EBV genome, we summarized viral gene expression, transcription factor binding, histone modifications, and binding sites of the chromatin insulator CTCF (Figure 2A). An initial overview of the data provided several striking observations.
We identified many candidate regulatory domains of the viral genome that were bound by transcriptional regulators and enriched in histone modifications. We analyzed binding profiles for over 60 human transcription factors, including factors involved in B cell development and viral response, such as Pax5, Irf3, Ebf1, and NFKB. We found that at least 26 transcription factors (TFs) cumulatively have at least 109 reproducible significant binding sites across the viral genome (Figure 2A, S2A, and Table S3). Furthermore, transcription factor binding sites cluster into regulatory loci that are reminiscent of “hotspot” regions identified in other model organisms(Gerstein et al., 2010; Moorman et al., 2006; Roy et al., 2010). Example clusters occurred at promoters for actively transcribed latent transcripts (Cp, RPMS1), as well as at the left lytic origin of replication (OriLytL). The divergent promoters at OriLytL are partitioned into two major clusters of binding factors (Figure 2B). The leftward transcript for BHLF1 has the highest RNA polymerase II peak in the viral genome over its initiation site, which also overlaps with known polymerase cofactors TBP and TAF1. The rightward promoter for BHRF1 (and associated viral miRNAs) contains a peak enriched with several factors, including p300, Gcn5, Bcl3, Pbx3, Egr1, Brca1, cFos and Chd2. We ensured that this locus did not have enrichment for IgG or input DNA sequencing controls and that the overlap was significant (Figures S2B and S2C). The prominent binding of the insulator protein CTCF separates these independent binding sites and their respective divergent promoters, consistent with prior findings (Tempera et al., 2010) (Figure S2D–F).
Histone modifications and chromatin boundary factors, like CTCF, are known to contribute to EBV latency regulation but have never been assayed at high resolution(Tempera et al., 2010). The highest measured histone modifications aggregated to a peak at the RPMS1 promoter region, which is the transcription initiation site for a cluster of EBV miRNAs and non-coding RNAs (Figure 2C). CTCF binds downstream of this promoter and appears to serve as a boundary for high-level promoter proximal histone modifications H3K27ac and H3K4me3 (Figure 2C). We also mapped nucleosome positions in two cell lines with different viral latency programs. Nucleosomes strongly occupied the transcriptionally silent Cp promoter in type I Burkitt lymphoma cell lines (which fail to express EBNA2 family genes), while nucleosomes did not occupy the transcriptionally active Cp in type III LCLs (which do express EBNA2 family genes) (Figure 2D). These findings are consistent with reports that active genes have nucleosome free promoter regions (Zhou et al., 2005).
The EBV genome is a complex patchwork of densely packed, overlapping, and extensively alternatively spliced genes. As expected in LCLs, the coding and transcribed non-coding domains of the EBV genome were strongly enriched for RNA-seq reads mapping to the latency transcripts for the EBNA genes, latent membrane proteins, RPMS1, as well as the non-coding RNAs, and miRNAs (Table S4). Less expectedly, we detected very high read counts covering the BHLF1 transcript, which is typically associated with early stages of lytic cycle gene activation and is transcribed from the lytic origin of replication (OriLyt). We also detected other immediate early transcripts such as BZLF1 and BRLF1, though these were expressed at much lower levels. We were also able to verify a recently reported viral-encoded snoRNA (Hutzinger et al., 2009).
We characterized all RNA splice junctions by analysis of paired end sequencing assays (Table S5). We found many known isoforms and confirmed several recently reported isoforms (Austin et al., 1988; Kelly et al., 2009; Lin et al., 2010). Using deep sequencing from hundreds experiments across multiple labs, we were able to detect dozens of isoforms, including alternative splicing in BZLF1, the BART locus, and multiple 5′ sites that spliced to acceptors in BHRF1 and BHLF1 (Figure 2E). The extensive usage of alternative donor and acceptor splice sites is remarkable and may be relevant for virus regulation and function.
RNA editing occurs in numerous host transcripts and was recently shown to also occur in viral RNAs (Iizasa et al., 2010; Li et al., 2011). Analysis of the many RNA-seq experiments revealed that the BHLF1 RNA transcript contains a guanine nucleotide that differs from the expected adenine residue encoded in the template DNA. This is consistent with classic ADAR-mediated deamination editing (Iizasa et al., 2010). Specifically, ~10% of RNA-seq reads (as averaged across 170 experiments with >10× coverage) mapping to this locus contain the alternative base whereas only 1 read in 1 DNA sequencing experiment (out of n=26 with >5× coverage) contained this alteration (Figures 2F and S2G–H). BHLF1 RNA has been implicated in the initiation of viral DNA replication, and RNA-editing may contribute to this process (Rennekamp and Lieberman, 2011).
In proliferating LCLs, the EBV genome persists predominantly as a type III latent infection, where a limited set of viral genes and miRNAs are expressed and the genome is replicated exclusively by host-cell replication machinery. However, all LCLs contain subpopulations of cells undergoing various degrees of lytic cycle gene expression and replication, the extent of which varies among cell populations and culture conditions (Davies et al., 2010; Glaser et al., 1989).
We analyzed the EBV gene expression profiles of over 300 LCLs and, after quality filtering, we clustered 201 viral expression profiles (Table S6 and Supplemental Material). The two main clusters correlated with canonical type III latent and lytic cycle patterns (Figures 3A, and S3). This clustering pattern was stable and consistent across multiple independent labs using independent LCL subclones (Figure 3B and S4). The gene expression clusters correlated with the average EBV load, as measured by the percentage of RNA-seq reads mapping to EBV (Figure 3A, top bar and Figure S3) and EBV episome copy number as assayed by an independent lab in independent subclones (Figure 3C)(Choy et al., 2008). Since lytic reactivation results in productive replication of the EBV episome, this correlation is an independent validation of lytic activity. However, we also observed disproportionately lower expression of EBV structural and packaging genes associated with infectious virion production (Table S4). This suggests that incomplete or abortive lytic gene reactivation occurs frequently in LCLs.
To better understand the dynamic intracellular environment during viral reactivation, we examined host gene expression and protein levels that correlate with lytic cycle. We correlated known cellular proteins, metabolic molecules, and surface markers with latent/lytic state. The cells lines with increased viral lytic cycle tended to have significantly higher ATP, more TNFα, more CCL5, less IL12, and more than double the levels of IL10 (Figures 4A and S4)(Choy et al., 2008). While these molecules are known B-cell regulatory factors, their role in lytic reactivation has been less well characterized.
We next used RNA-seq data to simultaneously assay host and viral gene expression and identify cellular gene transcripts that correlate with viral lytic phenotypes (Tables S7, S8, and Figures S3). LCL RNA-seq was quantified using reads per kilobase of exon per million mapped sequence reads (RPKM) to compute a correlation between EBV lytic gene BHLF1 and human host genes. Many viral genes were expressed at surprisingly high levels relative to human transcripts (Figures 4B, C).
We found that EBV lytic gene expression correlated with many human transcripts, for example, the WNT5B transcript (Figure 4D and S4). Searching pathway databases (Subramanian et al., 2005), we found that several families of human genes were significantly correlated with EBV lytic gene expression, including pathways involved in B cell chronic lymphocytic leukemia, interferon-alpha (INFα), WNT, and B cell receptor signaling (Figure 4E; Supplemental Methods; Table S9). An aggregate analysis of all cellular genes bound by the viral transcription factor EBNA1, which is upregulated during lytic reactivation (Figure S4D), revealed a statistically significant increase in virally targeted genes during lytic viral expression, suggesting direct viral regulation of host genes (Figure 4F). We also found that cellular genes identified in two-hybrid protein interaction assays with EBV proteins as bait (Calderwood et al., 2007) were correspondingly upregulated with increased lytic gene expression (Figure 4G). Furthermore, we found a small subset of host genes that may be coregulated by multiple means. For instance, the WNT5B gene is bound by EBNA1, is a member of the WNT pathway, and is commonly dysregulated in human cancers, including chronic lymphocytic leukemia (Lu et al., 2004; Lu et al., 2010).
We found that Pax5 binds with high occupancy to the viral terminal repeats (Figure 5), which are known to regulate LMP1 through an unknown mechanism (Repic et al., 2010). Pax5 is a B-cell specific factor that plays an essential role in B-cell development and contributes to B-cell tropism of EBV latency through initial activation of the latency transcription program at Wp (Tierney et al., 2000; Tierney et al., 2007). ChIP-seq analysis reveals two major peaks of Pax5 in at least one, though possibly all, of the viral terminal repeats (Figure 5A). Pax5 binding was validated by real-time PCR with primers specific for terminal repeat (TR) DNA, and a Pax5 consensus motif was identified within the centers of the ChIP-seq peaks (Figures 5A, B, S5A). We also confirmed that Pax5 binds the TR in an alternative EBV strain (Mutu) and in both type I (MutuI) and type III (Mutu-LCL) latently infected cells (Figure 5B). Surprisingly, Pax5 binding at the TR did not overlap with RNA polymerase II initiation factors, suggesting that its function at the TR is more complex than proximal promoter transcription repression. The Pax5 site in the TR is distinct from the known binding sites for Pu.1 and Sp1 in the LMP1 promoter (Sjöblom et al., 1995), which are identified by ChIP-seq along with binding sites for Ebf1 and Tcf12, and are the known sites of EBNA2 transactivation.
To explore the potential function of Pax5 in regulation of EBV gene expression, we depleted Pax5 protein from LCLs using lentivirus-delivered shRNA. We identified two different shRNA targeting sequences that effectively deplete Pax5 protein at five days post lentivirus infection and selection (Figure 5C). We first assayed the effect of Pax5 on EBV lytic activation by monitoring BZLF1 protein expression by Western blot (Fig. 5D) and EBV genome copy number by qPCR (Fig. 5E). We found that Pax5 depletion modestly (<2 fold) increased BZLF1 protein expression (Fig. 5D) and viral genome copy number (Fig. 5E). To assess the effect of Pax5 depletion on EBV genome configuration, we assayed both LCL and Raji cells for circularized and linear viral episomes by pulse-field gel electrophoresis (PFGE) (Fig. 5F). PFGE analysis indicated that Pax5 depletion had only marginal effects on EBV genome copy number in LCL (Fig. S5B). In contrast, a percentage of viral episomes were converted to linear and sublinear genomes in Raji cells (Fig. 5F). Since Raji are defective for viral lytic replication, these findings suggest that Pax5 may be important for maintaining the circular episomal form of the genome during latency.
To assess the effect of Pax5 depletion on viral transcription, we measured EBV gene expression for LMP1, LMP2 isoform 1 (LMP2-1), LMP2 isoform 2 (LMP2-2), EBNA1, EBNA2, as well as the lytic immediate early gene BZLF1. We found that Pax5 depletion led to the activation of LMP1 and both isoforms of LMP2 while decreasing EBNA1 and EBNA2 transcript accumulation (Figure 5G). Pax5 depletion also increased BZLF1 transcription, consistent with Western blotting results (Fig. 5D). These findings indicate that Pax5 contributes to the regulation of EBV transcription during latent infection.
To further explore viral chromosome architecture, we analyzed ChIP-seq binding profiles for the chromatin structural factors CTCF, SMC3, and Rad21 and observed a pattern of colocalization (Figure 6A and S6). Most striking was the strong ChIP signal at the convergence of the LMP1 termination site and LMP2 first intron (Figure 6B). This positioning was highly reminiscent of the CTCF-cohesin site identified within the latency transcript of the KSHV genome(Stedman et al., 2008). Interestingly, Rad21 and SMC3 also colocalized at the origin of latent replication (OriP), which is known to be a central regulatory region for episome maintenance and has been shown to regulate LMP1 through a long-distance enhancer mechanism(Gahn and Sugden, 1995) (Figure 6B).
As CTCF and cohesin have been implicated in long-distance DNA looping interactions, we used the chromatin conformation capture (3C) method to test whether the CTCF-cohesin sites physically linked OriP with the LMP1/LMP2 locus (Supplemental Methods; Table S10). Using anchor primers at the LMP1 locus and MseI digestion of the EBV genome, we found a strong 3C linkage formed between the anchor and positions centered around OriP (episome coordinates 6–9 kb) but not at several proximal control regions or other regions across the EBV genome (Figures 6C and 6D). These findings were corroborated by DNA sequencing of the PCR amplified junctions and by using different primer sets for real time qPCR.
To test whether cohesin subunits have any effect on EBV gene expression, we depleted SMC1 and Rad21 proteins using shRNA. LCLs were infected with lentivirus expressing Rad21 or SMC1 targeting shRNA and assayed by Western blot for knockdown efficiency (Figure 6F). Cohesin depletion led to a complete loss of the 3C loop formed between the OriP and LMP1/LMP2 control regions (Fig. 6E). qPCR analysis of EBV gene expression revealed that Rad21 or SMC1 depletion resulted in a general increase of latency transcripts, including LMP1 and LMP2 (Figure 6G). Cohesin depletion also led to a modest increase in BZLF1 expression. These results support the model that EBV chromatin forms higher order structures that include loop formation between OriP and the LMP1/LMP2 locus and help maintain viral latent cycle gene expression (Figure 6H).
Our work integrates large genomic data sets into a comprehensive atlas of functional elements of a human pathogen. We used a strategy of large-scale computational analysis, hypothesis generation, and functional validation to dissect the complex factors implicated in the regulation of both virus and host genes. Our meta-analysis approach also reveals how large-scale projects such as ENCODE and HapMap can have far-reaching scientific impact in a field outside their original scope. We discovered examples of coordinated mechanisms of viral gene control through the combined analysis of nucleosome occupancy, histone modification patterning, combinatorial transcription factor binding, and long range DNA looping. Several of these mechanisms are reminiscent of those in the human genome and illustrate the functional similarity of a virus and its host.
We provide insights into host-virus interactions through combined genome-wide expression analyses. We aggregated hundreds of RNA-seq experiments and observed many cell lines displaying surprisingly high levels of early lytic cycle genes relative to late lytic genes. The correlation of this gene expression program with EBV copy number suggests that abortive lytic DNA replication may be an important mechanism for viral genome maintenance and host cell population fitness. Furthermore, analyses of virus and host transcriptomes revealed unexpected coordination between viral lytic reactivation and cellular pathways involved in B-cell expansion and tumor promotion. These findings are consistent with studies showing that viral lytic cycle gene expression correlates with EBV carcinogenesis in humans (Dardari et al., 2000; Hanto et al., 1983) and contributes to lymphomagenesis in a humanized mouse model (Ma et al., 2011).
We also examine host-virus interactions at individual regulatory loci and provide insights into the relationship between chromatin organization, gene regulation, and how developmental regulatory factors link these processes. The B-cell identity factor Pax5 was found to bind within the viral terminal repeats and modulate transcription of viral oncogenes from within the DNA loop connecting CTCF-cohesin peaks at LMP1 and OriP loci. These binding sites are reminiscent of a similar coordination of Pax5 and CTCF sites identified within the IgG locus, which function to direct class switching during B-cell maturation (Ebert et al., 2011). Pax5 mutations have been implicated in B-cell lymphomagenesis, which may result in dysregulation of LMP expression in certain EBV malignancies, including Hodgkins’s disease (McCune et al., 2006; Pasqualucci et al., 2001). Additionally, down-regulation of Pax5 during typical germinal center reaction may derepress LMP1 and LMP2, providing a mechanism for virus-mediated rescue from apoptosis of somatically hypermutated B cells that harbor tumorigenic translocations so commonly seen in EBV-associated malignancies(Bechtel et al., 2005; Klein and Dalla-Favera, 2008; Mancao et al., 2005; Roughan and Thorley-Lawson, 2009). These observations suggest that breakdown of host regulatory pathways for controlling EBV gene expression may contribute to tumorigenesis through dysregulation of viral oncogenes.
EBV was discovered as the first human tumor virus nearly fifty years ago and was among the first genomes sequenced. It is thus fitting that it serves as the prototype for functional genomic and epigenomic elements in a human tumor virus, providing a foundation for future work in viral genomics and the emerging field of systems virology (Peng et al., 2009). We anticipate that system-level and data-driven approaches will ultimately lead to more comprehensive models of viral persistence and its role in human cellular development and oncogenesis.
Data was aggregated from a variety of sources (Table S1) and aligned to the EBV genome (Genbank accession NC_007605, March 2010) using the bowtie program allowing for one mismatch. We ignore all reads that aligned to the human genome, which ensured that all EBV reads are uniquely EBV and not actually from similar host sequence. To subtract human reads, we mapped all reads to hg19 (including all “Un” and “random” chromosomes), using the same parameters as when aligning to the viral genome. The subtraction of human-aligning reads resulted in small regions of the EBV-genome becoming unmappable (Figure S1).
Transcription factor ChIP-seq peaks with significantly more reads than IgG and input control sequencing experiments were identified. We modeled reads in the control experiments by a Poisson distribution and a p-value of 1.72e-7 was used as a cutoff for peak calling, which represents a genome-wide familywise error rate of 0.01. Validation of a subset of ChIP-seq peaks was performed as previously described (Tempera et al., 2010). Briefly, cells were fixed in 1% formaldehyde for 30 minutes and DNA was sonicated to between 200 and 350 basepairs. ChIP DNA was amplified on a ABI Prism 7000 using SYBR green chemistry.
TopHat and Cufflinks were used to align reads and quantify known transcripts (Trapnell et. al., 2010). Transcript abundances were normalized by Reads Per Kilobase per Million (RPKM). Human transcripts were taken from RefSeq annotations for hg19. Multimapping reads were allowed to map to 30 regions in the transcriptome before being ignored. RPKM values were normalized to account for multi-mapping reads. Gene abundance was estimated as the sum of all transcript isoforms. The per-million normalization was done relative only to reads mapping to EBV or human, since this ensured independence of transcript quantification in EBV and Human. Use of absolute transcript quantities (normalized to total alignable reads in both EBV and human) is used only when noted. RNA-seq experiments with low read-count (less than 1 million alignable human reads) samples were removed and technical replicates were averaged. From the original 294 RNA-seq assays, we obtained 201 reliable and independent samples.
Chromatin Conformation Capture (3C) was performed as previously described (Hagège et al. 2007 and Tempera et. al. 2011) with minor modifications. Briefly, cells were put through a 70μm filter to obtain single cells. 10M cells were fixed in 1% formaldehyde for 30 minutes and the reaction was quenched using 0.125M glycine. Cells were centrifuged, resuspended in lysis buffer, and lysed for 10 min. Nuclei were collected and digested with 500 U of MseI restriction enzyme overnight at 37C. The digestion was halted by incubation at 65C in 1.6% SDS on a 1200RPM shaker. The sample was diluted 10-fold, followed by ligation reaction containing 100 U T4 DNA ligase for 4 hours at 16C and 45 min at room temperature. The sample was digested by 300 μgr of proteinase K at 65C overnight, followed by RNase treatment for 1hr at 37C. DNA was phenol-chloroform extracted and ethanol precipitated and analyzed using PCR, qPCR, and sequencing. As a control, the EBV bacmid was MseI digested and ligated, thus creating all possible ligation products at background concentrations.
Nucleis were isolated from 5 × 107 Mutu and Mutu-LCL cells using a Dounce homogenizer in 4 ml Lysis Buffer (0.3 M sucrose, 2mM magnesium acetate, 3 mM CaCl2, 1% Triton X-100, and 10 mM HEPES (pH7.9)). The lysate was centrifuged through a glycerol cushion (25% glycerol, 5 mM magnesium acetate, 0.1 mM EDTA, and 10 mM HEPES (pH 7.4)) at 1000g for 15 min at 4°C. The nuclei were incubated with micrococal nuclease I (500 U/ml) at 37 °C for 20 min in 200 μl digestion buffer (25 mM KCl, 4 mM MgCl2, 1 mM CaCl2, 50 mM Tris (pH 7.4) and 12.5% glycerol). The reaction was stopped by equal volume of Stop Buffer (2% SDS, 0.2 M NaCl, 10 mM EDTA, 10 mM EGTA, 50 mM Tris (pH 8.0), and proteinase K (100ug/ml)) for 2 hours at 50°C. MNase I resistant DNA was extracted by phenolchloroform and ethanol precipitation. The ~150 bp mononucleosomal DNA was isolated from 1.5% agarose gel and purified by QIAquick Gel extraction kit (QIAGEN) following the manufacturer’s protocol. Purified DNA was then subject to Solexa sequencing using manufacturers recommendations (Illumina, Inc.).
Lentiviral shRNAs were obtained from the Sigma-Aldrich MISSION shRNA library. 4 million Mutu-LCL cells were passed through a 40μm filter, spundown, and resuspended in 2ml lentivirus suspension. Cells were spun at 500g for 90 minutes at 25C. Cell were cultured at ~500K cells/ml for 5 days with media changes in 1μg/ml of puromycin. Knockdown was confirmed by Western blot and qRT-PCR.
To resolve large viral genomic DNA-fragments, pulse field gel electrophoresis was performed as described previously (Dheekollu and Lieberman, 2011). Briefly, DNA migrated for a duration of 23 hours at 14C with the pulse ramping linearly every 60–120s through 120C using a Bio-Rad CHEF Mapper.
This work was supported by funds from NIH awards CA085678, CA093606, and DE017336 to PML, NIH award HG006798 to CL, and a K99AI099153 award from the National Institute Of Allergy And Infectious Diseases to IT. We thank Dr. Louise Showe and Priyankara Wikramasinghe from the Wistar Institute Genomics and Bioinformatics facilities and acknowledge the support of the Wistar Institute Cancer Center Core grant P30 CA10815. We also thank Chris Wawak and Joanne Edington for technical support.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.