RNA Sequencing Analyses of Gene Expression during Epstein-Barr Virus Infection of Primary B Lymphocytes

Epstein-Barr virus (EBV) immortalization of resting B lymphocytes (RBLs) is a useful model system to study EBV oncogenesis. By incorporating transcriptome sequencing (RNA-seq), chromatin immune precipitation followed by deep sequencing (ChIP-seq), chromatin interaction analysis followed by paired-end tag sequencing (ChIA-PET), and genome-wide clustered regularly interspaced short palindromic repeat (CRISPR) screen, we identified key pathways that EBV usurps to enable B cell growth and transformation. Multiple layers of regulation could be achieved by cooperations between multiple EBV transcription factors binding to the same enhancers. EBV manipulated the expression of most cell genes essential for lymphoblastoid cell line (LCL) growth and survival. In addition to proteins, long noncoding RNAs (lncRNAs) regulated by EBV also contributed to LCL growth and survival. The data presented in this paper not only allowed us to further define the molecular pathogenesis of EBV but also serve as a useful resource to the EBV research community.

ABSTRACT

Epstein-Barr virus (EBV) infection of human primary resting B lymphocytes (RBLs) leads to the establishment of lymphoblastoid cell lines (LCLs) that can grow indefinitely in vitro. EBV transforms RBLs through the expression of viral latency genes, and these genes alter host transcription programs. To globally measure the transcriptome changes during EBV transformation, primary human resting B lymphocytes (RBLs) were infected with B95.8 EBV for 0, 2, 4, 7, 14, 21, and 28 days, and poly(A) plus RNAs were analyzed by transcriptome sequencing (RNA-seq). Analyses of variance (ANOVAs) found 3,669 protein-coding genes that were differentially expressed (false-discovery rate [FDR] < 0.01). Ninety-four percent of LCL genes that are essential for LCL growth and survival were differentially expressed. Pathway analyses identified a significant enrichment of pathways involved in cell proliferation, DNA repair, metabolism, and antiviral responses. RNA-seq also identified long noncoding RNAs (lncRNAs) differentially expressed during EBV infection. Clustered regularly interspaced short palindromic repeat (CRISPR) interference (CRISPRi) and CRISPR activation (CRISPRa) found that CYTOR and NORAD lncRNAs were important for LCL growth. During EBV infection, type III EBV latency genes were expressed rapidly after infection. Immediately after LCL establishment, EBV lytic genes were also expressed in LCLs, and ∼4% of the LCLs express gp350. Chromatin immune precipitation followed by deep sequencing (ChIP-seq) and POLR2A chromatin interaction analysis followed by paired-end tag sequencing (ChIA-PET) data linked EBV enhancers to 90% of EBV-regulated genes. Many genes were linked to enhancers occupied by multiple EBNAs or NF-κB subunits. Incorporating these assays, we generated a comprehensive EBV regulome in LCLs.

IMPORTANCE Epstein-Barr virus (EBV) immortalization of resting B lymphocytes (RBLs) is a useful model system to study EBV oncogenesis. By incorporating transcriptome sequencing (RNA-seq), chromatin immune precipitation followed by deep sequencing (ChIP-seq), chromatin interaction analysis followed by paired-end tag sequencing (ChIA-PET), and genome-wide clustered regularly interspaced short palindromic repeat (CRISPR) screen, we identified key pathways that EBV usurps to enable B cell growth and transformation. Multiple layers of regulation could be achieved by cooperations between multiple EBV transcription factors binding to the same enhancers. EBV manipulated the expression of most cell genes essential for lymphoblastoid cell line (LCL) growth and survival. In addition to proteins, long noncoding RNAs (lncRNAs) regulated by EBV also contributed to LCL growth and survival. The data presented in this paper not only allowed us to further define the molecular pathogenesis of EBV but also serve as a useful resource to the EBV research community.

INTRODUCTION

Epstein-Barr virus (EBV), the first human DNA tumor virus discovered more than 50 years ago, causes ∼200,000 cases of cancers annually (1, 2). EBV-associated cancers include Burkitt’s lymphoma, Hodgkin’s lymphoma, posttransplant lymphoproliferative disease (PTLD), AIDS lymphomas, nasopharyngeal carcinoma, and ∼10% of gastric cancers. EBV establishes different latency programs in these cancers (3). Even though most primary EBV infections are asymptomatic, EBV infection can cause infectious mononucleosis when primary infection is delayed to late adolescence or early adulthood.

In vitro, EBV transforms primary human resting B lymphocytes (RBLs) into lymphoblastoid cell lines (LCLs). LCLs express type III EBV latency genes, including EBV nuclear antigens (EBNAs) 1, 2, leader protein (LP), 3A, 3B, and 3C, latent membrane proteins (LMPs) 1 and 2, nonpolyadenylate small RNAs, and microRNAs (miRNAs) (3). In vivo, cells expressing the type III EBV latency program are efficiently removed by the normal immune system. However, when the normal immune system is impaired, such as in transplant recipients undergoing immune suppressive treatment or in AIDS, type III latency program-expressing B cells can develop into lymphoproliferative diseases or lymphomas. Thus, LCLs serve as a useful model system to study the roles of EBV in oncogenesis. Understanding the molecular mechanisms of EBV-mediated growth transformation will not only provide insight into EBV pathogenesis but also identify potential therapeutics.

Genetic studies have demonstrated that EBNALP, 2, 3A, and 3C and LMP1 are essential for LCL growth and survival (3). When EBV infects RBLs, EBNALP and EBNA2 are first expressed (4, 5). EBNA2 is the major EBV transcription activator that activates the expression of both viral and host genes (3). EBNA2 is tethered to host DNA through its interactions with host transcription factors (TFs), and its C-terminal transactivation domain recruits basal transcription machinery (6–9). EBNA2 activates key oncogenes, including MYC (8, 10–14). EBNALP strongly coactivates EBNA2 by dislodging transcription repressors and coactivating EP300 (15–19). EBNA3A and EBNA3C can repress p16^INK4A, p14^ARF, and BIM to prevent senescence and apoptosis (13, 14, 20–25). EBNA3A and EBNA3C can also activate viral or host gene expression (24–29). LMP1 activates NF-κB to promote LCL survival (30–32). For EBV to transform RBLs, the virus has to overcome several check points, including DNA damage response (DDR) and metabolic check points. EBV-encoded proteins help the virus overcome these barriers to achieve immortal growth (33, 34).

Microarray-based assays identified many EBV-regulated genes (5, 24, 25, 28, 33, 35–38), and the genes essential for LCL growth and survival were confirmed by genome-wide clustered regularly interspaced short palindromic repeat (CRISPR) screen (39). However, recently developed transcriptome sequencing (RNA-seq) analyses can provide more accurate assessment of low abundance transcripts. RNA-seq can also simultaneously measure the abundance of long noncoding RNAs (lncRNAs) and EBV transcripts. We therefore performed a time course transcriptome study of EBV infection of RBLs.

RESULTS

EBV infection of RBLs causes dramatic changes in protein-coding RNAs.

RNA-seq was used to systemically identify the dynamic host transcriptomic changes following EBV infection of RBLs. RBLs from three healthy donors were infected with B95.8 EBV. Total RNAs were first purified from cells at 0, 2, 4, 7, 14, 21, and 28 days after EBV infection. Poly(A) RNAs were then further enriched. A directional RNA library preparation kit was used to prepare the sequencing libraries. Gene expression was quantified with mapped sequencing reads to human and EBV genomes. Variance-stabilizing normalized gene read counts were used to identify cell genes differentially expressed following viral infection (see Materials and Methods). The expression of 3,669 genes was significantly altered following EBV infection (q < 0.01, at least 1 time point had more than 20 reads). K-means clustering of gene expression identified 8 clusters (Fig. 1A; see also Data Set S1 in the supplemental material) with unique temporal patterns across time points.

FIG 1.

EBV-regulated host genes. (A) RBLs were infected with B95.8 EBV for 0, 2, 4, 7, 14, 21, and 28 days. Total RNAs were prepared from these cells. Poly(A) plus RNAs were sequenced using Illumina NextSeq. Differentially expressed genes were identified using a two-way ANOVA with an adjusted P value of <0.01 using Bonferroni’s correction. Examples of differentially expressed genes are listed on the right. (B) Expression of IRF2 following EBV infection. (C) Genome browser view of the AURKB loci. Normalized RNA-seq signals are at the top (green), ChIP-seq signals for EBNA2, LP, 3A, and 3C and NF-κB subunits are in the middle. POLR2A ChIA-PET links are in red. AURKB gene is indicated by one of the orange boxes. (D) Expression of IRF4 following EBV infection. (E) Genome browser view of the AICDA locus. Normalized RNA-seq signals are at the top (green), ChIP-seq signals for EBNA2, LP, 3A, and 3C and NF-κB subunits are in the middle (blue). POLR2A ChIA-PET links are in red. AICDA gene is indicated by the orange box. (F) Expression of genes essential for LCL survival from CRISPR screen during EBV infection of RBLs.

Cluster 1 had 840 genes. These genes were immediately upregulated 2 days postinfection and maintained at a slightly lower level afterwards. The expression levels slowly decreased afterwards. This cluster included E2F1, CHEK1, CDK4, CDK2, HDAC1/2, MCM2/3/4/6/8/10, ATIC, GART, etc.

Cluster 2 had 825 genes. These genes were upregulated 2 days after infection, peaked at day 4, and were maintained at this level afterwards. This cluster included CCNA2, CCNB1/2, CDK6, CDC2/3/5/7/8, AURKA/B, MYB, BATF, IRF4, etc. IRF4 mRNA expression level changes following EBV infection of RBLs were shown as an example for this group (Fig. 1B). AURKB is an aurora kinase family member of serine/threonine kinases required for the control of mitosis (40). AURKB is induced by EBNA3C (41). Chromatin immune precipitation followed by deep sequencing (ChIP-seq) identified many binding sites for EBNAs (including EBNA3C) and NF-κB subunits near AURKB. POLR2A chromatin interaction analysis followed by paired-end tag sequencing (ChIA-PET) linked enhancers to their direct target genes via DNA looping. Extensive interactions were present at the AURKB loci (Fig. 1C). Interestingly, even though there were extensive loops formed within this locus, AURKB and PFAS were highly induced by EBV, while VAMP2 and CTC1 were highly repressed (Fig. 1C). The extensive loops at this locus were associated with opposite transcription effects.

Cluster 3 had 401 genes. The expression levels of these genes increased slowly after EBV infection, and reached and maintained high levels 2 weeks after infection. This cluster included CCND2, IGF1, TRAF1, IRF2, AICDA, etc. IRF2 mRNA expression was used as an example for this group (Fig. 1D). AICDA is important for somatic hypermutation, class switch, and chromosome translocation (42). EBNA3C can induce AICDA expression (43). Multiple EBNA and NF-κB enhancers upstream of AICDA were linked to the AICDA promoter by POLR2A ChIA-PET (Fig. 1E).

Cluster 4 had 105 genes. These genes were first repressed 2 days after infection. Then, the expression slowly increased and reached high levels 2 weeks postinfection. Their expression levels were maintained at high levels afterwards. This cluster included CFLAR, PIM2, etc.

Cluster 5 had 366 genes. These genes were immediately upregulated 2 days postinfection. They were then downregulated at day 4 and further downregulated to levels slightly higher than preinfection. This cluster included MCM7, PIK3AP1, etc.

Cluster 6 had 166 genes. These genes were also upregulated 2 days postinfection. However, their expression levels decreased 4 days postinfection and were repressed 1 week after infection and remained at low levels afterwards. This cluster included TRIM28, JADE3, KAT2A, IRAK1, etc.

Cluster 7 had 481 genes. These genes were repressed starting from day 2 postinfection, and the repression progressed throughout the rest of the time course. This cluster included BCL6, SPI1, EBF1, SIN3A, etc.

Cluster 8 had 440 genes. These genes were immediately repressed following EBV infection at day 2. However, their expression levels recovered and reached steady levels afterwards. This cluster included BRD2, SP100, OTUD1, CYLD, BCL3, SIN3B, etc.

A genome-wide clustered regularly interspaced short palindromic repeats (CRISPR) screen identified 492 genes essential for LCL growth and survival (39). Four hundred sixty-one of these genes were differentially expressed during transformation, and these genes can be clustered into 6 clusters with unique expression patterns (Fig. 1F; see also Data Set S2) (P < 0.05). Clusters 1 and 2 had 194 genes whose expression slowly increased or remained high throughout the time course. These genes included CCND2, CFLAR, IRF2, POUAF1, RBPJ, FANCA/C/F/G/L, BATF, CDK6, IRF4, and XRCC2/6. Clusters 3 and 4 had 154 genes. These genes were first greatly upregulated and then slowly decreased to lower levels. These genes included MTOR, PRMT1, and XRCC3. Surprisingly, 113 genes in clusters 5 and 6 were downregulated throughout the time course. These genes included BRD4, EBF1, MEF2C, EP300, and ZNF217.

Pathways significantly enriched in EBV-induced genes.

Gene ontology pathway analyses found that EBV-induced genes were enriched in many biological processes (Table 1). These pathways are critical for cell proliferation and survival.

TABLE 1.

Pathways enriched in genes upregulated during EBV infection of RBLs

Term	P value
Mitotic nuclear division	2.29E−17
DNA replication	2.98E−15
G1/S transition of mitotic cell cycle	8.80E−15
Positive regulation of ubiquitin-protein ligase activity involved in regulation of mitotic cell cycle transition	1.31E−10
Type I interferon signaling pathway	1.43E−10
Tumor necrosis factor-mediated signaling pathway	9.87E−10
Defense response to virus	1.64E−09
DNA repair	7.49E−08
Regulation of cellular amino acid metabolic process	1.03E−07
Nucleotide-excision repair, DNA incision, 5′ to lesion	8.67E−06

EBV infection of RBLs causes continuous B cell proliferation. G₁/S transition of the mitotic cell cycle pathway was significantly enriched in EBV-induced genes. The upregulated genes in this pathway included CDC34, MCM2/3/8/10, RPA1/2/3, CDC45, ORC1/2/3/6, and CDK4/6. During G₁/S transition, D-type cyclins bind to CDK4/6 and translocate into the nucleus with CDK4/6. CDK4/6 then phosphorylates RB, resulting in the release of E2F sequestered by RB. E2F activates the expression of genes essential for cell cycle progression (44). CDK6 was scored to be essential for LCL growth and survival in the genome-wide CRISPR screen, and EBNA3C can modulate CDK6 (39, 45, 46). RNA-seq found rapid CDK6 upregulation upon EBV infection (Fig. 2A). ChIP-seq found multiple EBNA and NF-κB binding sites near the CDK6 locus. These binding sites looped to the CDK6 promoter as indicated by POLR2A ChIA-PET (Fig. 2A), suggesting that EBV TFs and EBV-activated NF-κB can regulate the expression of CDK6 to ensure cell cycle progression.

FIG 2.

EBV-regulated pathways. Genome browser views of CDK6 (A), FANCI (B), and OAS1, 2, and 3 (C). Normalized RNA-seq signals are at the top (green), ChIP-seq signals for EBNA2, LP, 3A, and 3C and NF-κB subunits are in the middle (blue). POLR2A ChIA-PET links are in red. Genes are indicated in orange boxes.

DNA replication and repair pathways were significantly enriched in EBV-induced genes. These genes included ORC1/2/3, MCM2/3/4/6/8/10, RFC1/2/3/4/5, CHEK1, XRCC2, FANCL, FANCI, etc. DNA replication errors in proliferating cells can cause genome instability and induce DNA damage response (DDR) (47). DDR can induce CHEK1 that is essential for LCL growth to protect genome integrity during DNA replication and EBV can modulate its expression (39, 48–50). A genome-wide CRISPR screen also identified Fanconi anemia (FA) family members FANCA/C/F/G/L and XRCC2/3/6 as essential for LCL growth (39). FA family members are activated by DNA damage sensors ATM or ATR and are involved in DNA repair (e.g., double-stranded breaks) (51). XRCC2/3 and XRCC6 are important for homologous recombination (HR) and nonhomologous end joining (NHEJ) (52). EBV infection significantly induced their expression (P < 0.01) (Data Set S2). For example, shown in Fig. 2B, FANCI expression was induced rapidly following EBV infection as shown by RNA-seq. CHIP-seq and POLR2A ChIA-PET found multiple EBNA enhancers linked to the FANCI promoter.

The interferon signaling pathway was also highly enriched in EBV-induced genes. Genes in this pathway included OAS1/2/3, IFITM1/2/3, ISG15, MX1/2, IRF2/4/5/7, etc. OAS proteins can synthesize 2′,5′-linked phosphodiester bonds, enabling ATP polymerization (53). OAS1/2/3 are on the same chromosome and neighbor each other. They were all induced rapidly after EBV infection and their levels remained high throughout the time course. ChIP-seq and POLR2A ChIA-PET identified multiple EBNA or NF-κB-driven enhancers linked to the promoter, providing the molecular basis for their coregulation (Fig. 2C).

Pathways significantly enriched in EBV-repressed genes.

Pathway analyses revealed EBV-repressed genes were enriched in many pathways (Table 2). These pathways included viral transcription, negative regulation of transcription, negative regulation of protein kinase B signaling, interferon gamma-mediated signaling pathway, antigen processing and presentation via MHC class II, positive regulation of the apoptotic process, cell cycle arrest, etc.

TABLE 2.

Pathways enriched in genes downregulated during EBV infection of RBLs

Term	P value
SRP-dependent cotranslational protein targeting to membrane	1.83E−14
Viral transcription	2.78E−13
Negative regulation of transcription, DNA templated	1.11E−10
mRNA processing	2.86E−08
Negative regulation of protein kinase B signaling	7.52E−07
Positive regulation of transcription, DNA templated	1.53E−06
Interferon gamma-mediated signaling pathway	2.72E−05
Antigen processing and presentation of peptide antigen via MHC class II	9.16E−05
Positive regulation of apoptotic process	1.11E−04
Cell cycle arrest	4.05E−04

EBV infection of RBLs causes significant changes in noncoding RNAs.

RNA-seq reads were also mapped to long noncoding RNA (lncRNA) or antisense RNA. Fifty-seven noncoding RNAs were significantly upregulated (Fig. 3A; see also Data Set S3). Thirty-one of them started to increase 4 days after EBV infection. These RNAs included CYTOR (Fig. 3B). CYTOR is required for breast cancer cell proliferation, cell migration, and cytoskeleton organization (54). To test if CYTOR is important for LCL growth and survival, CRISPR interference (CRISPRi) was used to downregulate CYTOR expression. Two guide RNAs (gRNAs) were used to target the dCas9-KRAB repressor to the CYTOR transcription start site (TSS). CRISPRi repressed CYTOR expression by 50% to 70% (Fig. 3C). LCLs with 70% CYTOR repression grew significantly slower than control gRNA-expressing cells (Fig. 3D). Twenty-six noncoding RNAs were immediately upregulated upon EBV infection. Ten noncoding RNAs were first upregulated and then downregulated. Six noncoding RNAs were first repressed and then increased. Ninety-one noncoding RNAs were downregulated slowly throughout the time course. These RNAs included NORAD (Fig. 3E). NORAD can bind to proteins involved in DNA replication and repair (55). CRISPR activation (CRISPRa) was used to determine the effect of NORAD on LCL growth and survival. Two gRNAs were used to tether the dCas9-VP64 activator and MS2-p65-HSF1 to the NORAD TSS. Both gRNAs significantly upregulated NORAD (Fig. 3F). CRISPR activation of NORAD significantly reduced LCL growth (Fig. 3G). Thirty-three noncoding RNAs went down immediately after EBV infection, slightly recovered 1 and 2 weeks later, and went down again 3 and 4 weeks after infection. These RNAs included MALAT and TSIX (Fig. 3A).

FIG 3.

EBV-regulated noncoding RNAs. (A) Long noncoding RNAs and antisense RNAs differentially expressed during EBV infection. (B) Genome browser view of the CYTOR locus. Normalized RNA-seq signals are at the top (green), ChIP-seq signals for EBNA2, LP, 3A, and 3C and NF-κB subunits are in the middle (blue). POLR2A ChIA-PET links are in red. CYTOR gene is indicated in orange box. (C) CRISPRi knockdown of CYTOR. LCLs stably expressing dCas9-KRAB repressor fusion proteins were infected with lentiviruses expressing gRNAs. qRT-PCR was used to detect CYTOR expression. The level of control gRNA treated cells was set to 1. ***, P < 0.001, *, P < 0.05. (D) CellTiter-Glo was used to measure the cell growth following CYTOR knockdown. **, P < 0.01. (E) Genome browser view of the NORAD locus. Normalized RNA-seq signals are at the top (green), ChIP-seq signals for EBNA2, LP, 3A, and 3C and NF-κB subunits are in the middle (blue). NORAD gene is indicated in orange box. (F) CRISPRa activation of NORAD. LCLs stably expressing dCas9-VP64 activator fusion proteins were infected with lentiviruses expressing gRNAs. qRT-PCR was used to detect NORAD expression. The level of control gRNA treated cells was set to 1. **, P < 0.01. (G) CellTiter-Glo was used to measure the cell growth following NORAD activation. **, P < 0.01.

EBV gene expression.

RNA-seq reads were also mapped to the EBV genome to identify virus genes expressed during B cell transformation (Fig. 4A; see also Data Set S4). Two days after EBV infection, EBV type III latency genes EBNA2, LP, 3A, 3B, 3C, and 1 and LMP2B were already expressed at very high levels (very similar to LCL levels). LMP1 and LMP2A also started to increase early on (i.e., 2 days after infection); however, their expression continued to increase, and they reached their highest expression levels 3 weeks after infection, which was in agreement with previously published protein expression results (5). Two weeks after EBV infection, EBV lytic cycle genes were also expressed at high levels and reached peak at 3 weeks. EBV glycoprotein gp350 was detected in ∼4% of freshly established LCLs by fluorescence-activated cell sorting (FACS) (Fig. 4B). In contrast, GM12878 LCLs growing in vitro for prolonged periods of time expressed much less gp350 (<1%).

FIG 4.

EBV gene expression following RBL infection. (A) RNA-seq reads were first mapped to the EBV genome. Normalized EBV gene expression levels are shown. (B) LCLs were first stained with Cy5-conjugated antibody against gp350 or IgG control antibody. gp350 expression levels were determine by FACS analyses.

EBNA-regulated cell genes.

ChIP-seq analyses found that most EBNA2 genomic binding sites are located at enhancers. We linked all of the EBNA2 binding sites to their direct target genes using POLR2A ChIA-PET. EBNA2 enhancer sites were linked to 2,240 EBV-regulated genes. EBNA3A and 3C also mostly bind to enhancers (56, 57). EBNA3A enhancers were linked to 1,478 EBV-regulated genes. EBNA3C enhancers were linked to 1,212 EBV-regulated genes. Only ∼60% of EBNALP binding sites are enhancers (17). EBNALP enhancers were linked to 2,671 EBV-regulated genes. Both BATF and CCND2 were rapidly upregulated upon EBV infection and maintained at high levels throughout the time course. Multiple EBNA2, LP, 3A, and 3C ChIP-seq peaks were linked to BATF and CCND2 TSS (Fig. 5A and B; see also Data Sets S5 and S6). CCNB1 was also induced and upregulated. While the CCNB1 promoter was linked to prominent EBNALP and EBNA3A ChIP-seq peaks, minor peaks for EBNA2 and EBNA3C were also visible and were linked to CCNB1 TSS (Fig. 5C). FOSB was rapidly downregulated upon EBV infection. Multiple EBNA2 peaks were linked to FOSB TSS (Fig. 5D).

FIG 5.

EBNA-regulated host genes. Genome browser views of BATF (A), CCNB1 (B), CCND2 (C), and FOSB (D). Normalized RNA-seq signals are at the top (green), ChIP-seq signals for EBNA2, LP, 3A, and 3C and NF-κB subunits are in the middle (blue). POLR2A ChIA-PET links are in red. Genes are indicated in orange boxes. Effects of EBNA2 (E), 3A (F), and 3C (G) inactivation on cell gene expression. Conditional LCLs were grown under permissive or nonpermissive conditions. RNAs were prepared from these cells. qRT-PCR was used to quantitate the expression levels of the indicated genes. ACTB was used as a loading control. EBNA minus conditions were set to 1. (H) Average distances between EBNA ChIP-seq peaks and genes linked by POLR2A ChIA-PET. ***, P < 0.001, **, P < 0.01.

Reverse transcription-quantitative PCR (qRT-PCR) was used to validate the effects of EBNA inactivation on the expression of BATF, CCNB1/2, CCND2, and FOSB. LCLs expressing conditional EBNA2, EBNA3A, or EBNA3C were grown under permissive or nonpermissive conditions for 3 days or 14 days. Total RNA was extracted from these cells. RNAs were first reverse transcribed into cDNA. qPCRs were then used to determine their expression. The expression levels of EBNA under the “on” condition were set to 1. EBNA2 inactivation decreased the expression of BATF, CCND2, and CCNB1/2 and increased FOSB expression (P < 0.01) (Fig. 5E). EBNA3A and 3C inactivation greatly reduced CCNB1/2 expression (P < 0.01) (Fig. 5F and G).

The average distances between different binding sites for each EBNA and their target TSSs were compared to determine if the looping patterns differed between EBNAs (Fig. 5H). EBNA2, 3A, and 3C loops were similar in bridging remote enhancers and their target genes. EBNALP loops were closer to promoters, supporting the previous notion (17).

NF-κB-activated cell genes.

LMP1 signaling activates both canonical and noncanonical NF-κB pathways. Five NF-κB subunits form heterodimers or homodimers to bind DNA (32). ChIP-seq analyses identified thousands of NF-κB subunit binding sites. These binding sites are also predominantly enhancer sites. RELA peaks were linked to 3,064 genes. RELB peaks were linked to 3,080 genes. REL peaks were linked to 3,060 genes. NFKB1 peaks were linked to 2,906 genes. NFKB2 peaks were linked to 3,162 genes. Since NF-κB subunits can form heterodimers or homodimers on DNA, collectively, these NF-κB enhancers may regulate the expression of a large number of genes.

Early versus late events during EBV transformation.

Immediately after EBV infection, B cells proliferate rapidly (58). The cell growth slows down afterwards. To determine the difference between these two different growth phases, EBV-regulated genes were divided into early induced (before day 7), late induced (after day 7), early repressed, and late repressed. Pathways affected in these two phases were identified for EBV-induced genes. Early induced genes were enriched with pathways, including DNA replication and cell cycle (see Data Set S7). Late induced genes were enriched with metabolism-related pathways (see Data Set S8). To determine if one or many EBNAs were involved in these changes, these genes were then linked to EBNA enhancers with POLR2A ChIA-PET. The linkages between each EBNA and EBV-regulated genes were then compared (Fig. 6). Early induced genes were linked to 2,761 EBNA enhancers, while late induced genes were linked to 4,278 EBNA enhancers. Late induced genes had more links to EBNA2, 3A, and 3C than early induced genes, while EBNALP had similar numbers of links to the early or late induced genes (Fig. 6). EBNA2 and EBNALP had many interactions that did not overlap other EBNAs, while EBNA3A and 3C were more frequently overlapping other EBNAs. The linkage patterns were similar for early and late repressed genes.

FIG 6.

Genes induced after day 7 by EBNAs. EBV-regulated genes were grouped into early induced (before day 7), late induced (after day 7), early repressed, and late repressed. Each gene was linked to an EBNA ChIP-seq peak by POLR2A ChIA-PET. The overlap between different EBNAs and their late induced targets is shown in the Venn diagram.

DISCUSSION

EBV readily transforms RBLs into LCLs in vitro. Since LCLs share many common features with posttransplant lymphoproliferative disease (PTLD) and AIDS lymphomas and can be manipulated easily, LCLs serve as an ideal model system to study the mechanisms through which EBV causes cancers.

Microarray was used extensively to study the effects of EBV infection and EBV latency genes on the host transcriptome (5, 24, 25, 28, 33, 35–38). However, microarray studies rely on hybridization-based quantitation; thus, transcripts with low abundancy cannot be detected with high confidence. Most studies used very stringent cutoffs to filter out low-confidence genes. For RNA-seq, the quantitation of transcripts is based on the numbers of the sequencing reads mapped to the gene. Therefore, low-abundance transcripts can be detected with high confidence. Here, we used RNA-seq to evaluate the effect of EBV infection of RBLs on host and viral transcription in a time course study.

Previous studies have shown that many EBV-induced genes are transcription factors; therefore, it is difficult to determine if the transcriptional changes are caused by secondary effect. To overcome this problem, a shorter time course was used (35). This prevented detection of transcripts with long half-lives. In this paper, we incorporated LCL EBNA and NF-κB subunit ChIP-seq and POLR2A ChIA-PET data into our RNA-seq analyses. This allowed us to link EBV enhancers to genes expressed differentially during transformation. Incorporating these data sets allowed the characterization of the EBV regulome.

Genes identified from analyses of cell lines conditionally expressing EBNAs or activating NF-κB only partially overlap (5, 24, 25, 27, 28, 33, 35–38). ChIP-seq data frequently find co-occupied enhancers by multiple EBNAs and NF-κB. ChIA-PET linked these enhancers to EBV-regulated genes. This suggests that EBV regulation of these genes is more complex and layered. Since these data were generated from hundreds of millions of cells, it is not known if they are on the DNA at the same time in the same cell. It is also possible that these occupancies are cell cycle dependent, as EBNA3C can regulate LMP1 in a cell cycle-specific manner (59). Single-cell-based analyses are needed to address these questions. Some of these EBV enhancers have same effect on the neighboring genes linked to the same enhancer. It is also interesting to note that sometimes an enhancer has opposite effects on the expression of its associated gene. It is possible that the promoters of these associated genes have different basal transcription factor occupancies, and the expression levels are determined by these factors.

A genome-wide CRISPR screen identified 492 cell genes essential for LCL growth and survival. Approximately 94% (461) of these genes were differentially expressed during EBV transformation. Approximately 76% of these genes were upregulated during transformation. Interestingly, ∼24% of these genes were downregulated during transformation. High-level expression of these genes might not favor continuous cell growth. However, these genes are likely to be important to maintain LCL growth at lower expression levels.

EBV induced the expression of genes that drive cell cycle progression, genes that disable host check points, and genes for host metabolic enzymes to provide energy and building blocks for cell growth. EBV infection also induced host antiviral responses. Interestingly, the host antiviral response was maintained at a high level across the entire time course. Therefore, these proteins may be beneficial for LCL growth.

EBV also regulated the expression of lncRNAs. lncRNAs frequently affect the expression of their neighboring genes through looping. It would be interesting to further characterize their roles in LCL growth and survival.

Newly established LCLs are prone to lytic replication, which is evident from our analyses of the viral transcriptome. In contrast, LCLs in long-term culture have much less DNA replication. DNA methylation is known to regulate viral gene expression (60). It is possible that long-term culture can cause DNA methylation to shut off lytic cycles.

MATERIALS AND METHODS

Purification of RBLs.

Discarded, deidentified human peripheral blood mononuclear cells (PBMCs) were collected from three platelet donors from Dana Farber Cancer Institute. CD19⁺ RBLs were negatively selected using EasySep human B cell enrichment kit and human B cell enrichment cocktail (StemCell Technologies). Purified cells were cultured in RPMI 1640 medium (Life Technologies) supplemented with 10% fetal calf serum (FCS) in a humidified incubator at 37°C with 5% CO₂.

Purification and titration of B95.8 EBV.

B95.8 cells expressing EBV immediate early gene ZTA fused to a modified estrogen receptor hormone binding domain were used to produce EBV. 4-Hydroxytamoxifen was used to induce ZTA nuclear translocation. Viral supernatant was harvested 5 days after ZTA induction and stored at −80°C. The titer of EBV was determined by infecting RBLs with increasing amounts of viral supernatant. Two days after infection, EBNALP expression was determined by immunofluorescence staining. The amount of EBV that turned >90% of RBLs positive for EBNALP staining was used for subsequent infection experiments.

EBV infection of RBLs.

Purified RBLs were first spun down and resuspended in medium containing virus supernatant. Cells were incubated at 37°C with 5% CO₂ for 4 h. The cells were spun down again and resuspended in fresh medium at 1 × 10⁶/ml. The cells were harvested for RNA preparation at later time points.

RNA and RNA-seq library preparation.

Total RNAs were prepared using an RNeasy Mini kit (Qiagen) according to the manufacturer’s instructions. RNA quality was evaluated using a bioanalyzer. RNA integrity numbers (RINs) were >9. Poly(A) RNAs were enriched using a NEBNext Poly(A) mRNA Magnetic Isolation Module (New England BioLabs). RNA-seq libraries were prepared using a NEBNext Ultra II Directional RNA Library Prep kit (New England BioLabs) according to the manufacturer’s instructions.

Illumina sequencing.

RNA-seq libraries were sequenced using NextSeq 500, and ∼400 million reads were obtained. On average, each sample had ∼20 million reads.

RNA-seq data analysis.

Single-end RNA-seq reads from different sequencing lanes were pooled for the same donor at any given time point. Raw read counts for gene expression were quantified using Salmon v0.8.2 (61). The cDNA sequences supplied for Salmon include all human GENCODE v28 (GRCh37) genes (62) and EBV genes (63). The gene expression count matrix was then normalized by library size and regularized log transformed using DESeq2 (64) for variance stabilizing. Two-way analysis of variance (ANOVA) tests treating donor information as random effects were performed for each gene in order to evaluate significant genes that are differentially expressed in at least one of the time points. To increase the specificity of differential gene selection, P values were adjusted for multiple testing comparisons using Bonferroni’s correction. A gene was considered differentially expressed at an adjusted P value of less than 0.01. Genes with fewer than 20 reads in all time points were excluded from downstream analysis.

Gene clusters and pathway analysis.

Significant genes were categorized into multiple groups: protein coding, lncRNA and antisense, EBV genes, EBV super enhancer targeted genes, and genes essential for LCL survival. Functional gene clusters within each group were extracted by k-means clustering using R package pheatmap. The total numbers of clusters were set to enhance the identification of temporal patterns of gene expression. For each cluster of protein coding genes, Gene Ontology (GO) term pathway analysis was performed using the DAVID functional annotation Tool (65).

ChIP-seq and ChIA-PET data analyses.

ChIP-seq data were analyzed according to a previously described method (34). ChIP-seq peaks were called using a ChIP-seq processing pipeline (SPP) (65). The ChIP-seq signals were first normalized to 10 million mapped reads. Read depth was further normalized to per base pair per peak. ChIP-seq signals (depth per peak per base pair) were also normalized by subtracting the input signal.

ChIA-PET data were downloaded from the Washington University epi genome browser.

CRISPRi and CRISPRa.

CRISPRi and CRISPRa single guide RNAs (sgRNAs) were designed with online tools from Benchling (San Francisco, CA). sgRNAs) for CRISPRi were cloned into lentiGuide-Puro (52963; Addgene). sgRNAs used for CRISPRa were cloned into lenti-sgRNA(MS2)_zeo (61427; Addgene). Lentiviruses were prepared by transfecting HEK293T cells as described previously (39). Briefly, HEK293T cells were transfected with pCMV-VSVG (8454; Addgene), psPAX2 (12260; Addgene), and plentiGuide-Puro or lenti-sgRNA(MS2)_zeo plasmids expressing sgRNAs with TransIT-LT1 transfection reagent (Mirus) according to the manufacturer’s instructions. Viruses were harvested at 48 and 72 h posttransfection. Virus supernatant was filtered through a 0.45-μm syringe filter and added to LCLs stably expressing a dCas9-KRAB fusion protein for CRISPRi or LCLs stably expressing a dCas9-VP64 fusion protein and MS2-P65-HSF1 activation helper protein for CRISPRa. At 48 h after infection, transduced LCLs were selected with puromycin for CRISPRi and Zeocin for CRISPRa.

qRT-PCR.

LCLs were grown under permissive or nonpermissive conditions for EBNA2, 3A, and 3C. Total RNAs were extracted using an RNeasy Mini kit (Qiagen). An iScript reverse transcription supermix kit (Bio-Rad) was used to generate cDNA, and a Power SYBR green master mix kit (Thermo Fisher) was used to quantitate the expression level. GAPDH (glyceraldehyde-3-phosphate dehydrogenase) was used as a loading control. Primers used in this study are listed in Table 3.

TABLE 3.

Primers used in this study

Primer name	Sequence (5′→3′)
qRTNORAD-F1	TGCTGTCGGAAGAGAGAAATG
qRTNORAD-R1	CCTTCCATAAACGGCCAGTAA
qRTCYTOR-F1	CTGGATGGTCGCTGCTTTTT
qRTCYTOR-R1	GATCTGAAGACAGGCACGGG
qRTCCNB1-F	AATAAGGCGAAGATCAACATGGC
qRTCCNB1-R	TTTGTTACCAATGTCCCCAAGAG
qRTCCNB2-F	CCGACGGTGTCCAGTGATTT
qRTCCNB2-R	TGTTGTTTTGGTGGGTTGAACT
qRTBATF-F	TATTGCCGCCCAGAAGAGC
qRTBATF-R	GCTTGATCTCCTTGCGTAGAG
qRTFOSB-F	GCTGCAAGATCCCCTACGAAG
qRTFOSB-R	ACGAAGAAGTGTACGAAGGGTT
qRTCCND2-F	TTTGCCATGTACCCACCGTC
qRTCCND2-R	AGGGCATCACAAGTGAGCG
NORADasg1-R	AAACCCCCGCTTGCTCTTCGCAAC
NORADasg1-F	CACCGTTGCGAAGAGCAAGCGGGG
NORADasg2-R	AAACCCACGCCGGAACATTGCAGAC
NORADasg2-F	CACCGTCTGCAATGTTCCGGCGTGG
CYTORI-sg1-R	AAACTCCTGATCCTCTCAGAGAGAC
CYTORI-sg1-F	CACCGTCTCTCTGAGAGGATCAGGA
CYTORI-sg2-R	AAACCTGTGTGGGCGACCCCTGGGC
CYTORI-sg2-F	CACCGCCCAGGGGTCGCCCACACAG

Statistical analysis.

Statistical significance of differences between means from at least 3 experiments was determined using paired Student’s t tests.

Data availability.

RNA-seq data were deposited in the Gene Expression Omnibus (GEO) under accession number GSE125974. Other data were obtained from GEO: GM12878 LCL POLR2A ChIP-PET, GSE127053; EBNA2 ChIP-seq, GSE29498; EBNALP ChIP-seq, GSE49338; EBNA3A ChIP-seq, GSM1429820; EBNA3C ChIP-seq, GSE52632; and NF-κB ChIP-seq, GSE55105.