An EBNA3A-Mutated Epstein-Barr Virus Retains the Capacity for Lymphomagenesis in a Cord Blood-Humanized Mouse Model

The EBV protein EBNA3A is expressed in latently infected B cells and is important for efficient EBV-induced transformation of B cells in vitro. In this study, we used a cord blood-humanized mouse model to compare the phenotypes of an EBNA3A hypomorph mutant virus (Δ3A) and wild-type EBV. The Δ3A virus caused lymphomas with delayed onset compared to the onset of those caused by WT EBV, although the tumors occurred at a similar rate. The WT EBV and EBNA3A mutant tumors expressed similar levels of the EBV protein EBNA2 and cellular protein p16, but in some cases, Δ3A tumors had less LMP1. Our analysis suggested that Δ3A-infected tumors have elevated T cell infiltrates and decreased expression of the CLEC2D receptor, which may point to potential novel roles of EBNA3A in T cell and NK cell responses to EBV-infected tumors.

ABSTRACT

Epstein-Barr virus (EBV) causes B cell lymphomas and transforms B cells in vitro. The EBV protein EBNA3A collaborates with EBNA3C to repress p16 expression and is required for efficient transformation in vitro. An EBNA3A deletion mutant EBV strain was recently reported to establish latency in humanized mice but not cause tumors. Here, we compare the phenotypes of an EBNA3A mutant EBV (Δ3A) and wild-type (WT) EBV in a cord blood-humanized (CBH) mouse model. The hypomorphic Δ3A mutant, in which a stop codon is inserted downstream from the first ATG and the open reading frame is disrupted by a 1-bp insertion, expresses very small amounts of EBNA3A using an alternative ATG at residue 15. Δ3A caused B cell lymphomas at rates similar to their induction by WT EBV but with delayed onset. Δ3A and WT tumors expressed equivalent levels of EBNA2 and p16, but Δ3A tumors in some cases had reduced LMP1. Like the WT EBV tumors, Δ3A lymphomas were oligoclonal/monoclonal, with typically one dominant IGHV gene being expressed. Transcriptome sequencing (RNA-seq) analysis revealed small but consistent gene expression differences involving multiple cellular genes in the WT EBV- versus Δ3A-infected tumors and increased expression of genes associated with T cells, suggesting increased T cell infiltration of tumors. Consistent with an impact of EBNA3A on immune function, we found that the expression of CLEC2D, a receptor that has previously been shown to influence responses of T and NK cells, was markedly diminished in cells infected with EBNA3A mutant virus. Together, these studies suggest that EBNA3A contributes to efficient EBV-induced lymphomagenesis in CBH mice.

IMPORTANCE The EBV protein EBNA3A is expressed in latently infected B cells and is important for efficient EBV-induced transformation of B cells in vitro. In this study, we used a cord blood-humanized mouse model to compare the phenotypes of an EBNA3A hypomorph mutant virus (Δ3A) and wild-type EBV. The Δ3A virus caused lymphomas with delayed onset compared to the onset of those caused by WT EBV, although the tumors occurred at a similar rate. The WT EBV and EBNA3A mutant tumors expressed similar levels of the EBV protein EBNA2 and cellular protein p16, but in some cases, Δ3A tumors had less LMP1. Our analysis suggested that Δ3A-infected tumors have elevated T cell infiltrates and decreased expression of the CLEC2D receptor, which may point to potential novel roles of EBNA3A in T cell and NK cell responses to EBV-infected tumors.

INTRODUCTION

Epstein-Barr virus (EBV) is a human herpesvirus that infects the majority (>90%) of the adult population. EBV causes infectious mononucleosis and contributes to human lymphomas, including posttransplant lymphoproliferative disease (PTLD), diffuse large B cell lymphoma (DLBCL), Burkitt lymphoma, and Hodgkin lymphoma (1). EBV-associated malignancies in humans are largely composed of latently infected cells, and latent EBV infection is sufficient to transform B cells into long-lived lymphoblastoid cell lines (LCLs) in vitro. LCLs express all 8 EBV latency proteins and form lymphomas when grown as xenografts in immunodeficient mice.

EBV can enter several different types of latency, which differ in regard to the number of viral genes expressed, but only type III latency (in which all 9 latency proteins are expressed) transforms B cells in vitro (2). In contrast, EBV-positive lymphomas in humans often have more restricted (and less immunogenic) forms of viral latency (i.e., types I and II) that are nontransforming in vitro. EBV-induced transformation of B cells in vitro is largely driven by EBNA2 (which mimics constitutively active Notch signaling and interacts with the cellular RBPJ-κ protein) (3, 4) and LMP1 (which mimics the effect of constitutively active ligand-independent CD40 signaling) (5–9). EBNA2 transcriptionally activates the promoters of all viral genes expressed in type III latency and activates cellular genes required for LCL proliferation, including c-Myc and cyclin D2 (10–13). LMP1 induces the NF-κB pathway, which promotes survival and proliferation of normal B cells and is activated in many human B cell lymphomas (including DLBCLs) (5–9, 14).

The EBV EBNA3 latency proteins (EBNA3A, -3B, and -3C) are also expressed in cells with type III latency and transcriptionally regulate both viral and cellular genes (15). Like EBNA2, the EBNA3 proteins, which share homology, bind to the cellular DNA binding protein RBPJ-κ (16). EBNA3 proteins can repress and activate cellular gene transcription through chromatin remodeling, which leads to the addition of activating or repressive histone marks (15, 17–20). EBNA3 proteins regulate overlapping but distinct sets of genes in vitro and, in some cases, can inhibit effects of EBNA2 by competing for RBPJ-κ (18, 21–23). Chromatin immunoprecipitation sequencing (ChIP-seq) data show a large array of overlaps at binding sites for EBNA3 proteins and EBNA2 across the human genome (17, 24, 25). While EBNA3C is considered absolutely required for transformation of B cells in vitro, EBV mutants that lack EBNA3A are capable of transforming B cells into LCLs in vitro, but with a greatly reduced efficiency (18, 19, 26, 27). LCLs derived from EBV with an EBNA3A deletion mutation proliferate at reduced rates and have enhanced apoptosis in comparison to the results for WT virus-derived LCLs but express similar levels of EBV latency genes (19). Interestingly, approximately 25% of the cellular genes found to be regulated by loss of EBNA3A in LCLs were previously shown to be regulated by EBNA2 (19). Although in some cases, EBNA3A inhibits expression of genes activated by EBNA2, in other cases, EBNA3A and EBNA2 collaboratively activate target genes (19).

EBNA3A collaborates with EBNA3C to repress the CDKN2A gene-encoded tumor suppressors p16 and p14 in LCLs, thereby inhibiting oncogene-induced senescence (18). EBNA3A has also been shown to repress other cyclin-dependent kinase inhibitors (CDKIs), including p15INK4b (28). EBNA3A and EBNA3C also collaboratively repress expression of the proapoptotic protein BIM in Burkitt lymphoma (BL) cell lines (22). EBNA3A has been shown to bind to a superenhancer element that regulates BIM in Mutu III-BL and BL31 cells. This element in GM12878 bears the repressive histone modification H3K27me3, which suggests that this function of EBNA3A may be important in transformation of B cells into LCLs in vitro (29). EBNA3A is also required for transcription of the transcript encoding the antiapoptotic BFL1 protein, as well as the mitochondrial localization of the antiapoptotic MCL1 protein (30).

A recent study using a humanized mouse model in which irradiated newborn HLA-A2 transgenic NSG mice were engrafted with human fetal liver CD34⁺ hematopoietic progenitor cells and infected with EBV 10 to 12 weeks later reported that an EBNA3A deletion mutant EBV can stably establish latent infection but cannot induce lymphomas (31). Cells infected with the EBNA3A deletion mutant had increased p16 expression and proliferated more slowly than wild-type virus and also had less LMP1 expression (31).

Here, we have used a cord blood-humanized (CBH) mouse model (in which purified nucleated cord blood cells are briefly infected with EBV in vitro and then injected intraperitoneally into NSG mice) to study the roles of EBNA3A in promoting tumorigenesis and regulating viral and cellular gene expression. We previously showed that two different mutant EBVs that are nontransforming for B cells in vitro (missing either the LMP1 gene or the EBNA3C gene) still cause lymphomas in vivo in the CBH model (32, 33). Although the EBNA3A hypomorph mutant virus (Δ3A) used in our studies (which makes a very small amount of EBNA3A derived from a downstream ATG at residue 15) was highly deficient in transforming cord blood B cells in vitro, it induced oligoclonal/monoclonal lymphomas in CBH mice with a frequency similar to their induction by the wild-type (WT) virus, although the tumors occurred at later time points. Δ3A virus-induced lymphomas appeared histologically similar to lymphomas induced by WT virus and expressed similar levels of EBNA2 and p16, although LMP1 expression was sometimes but not consistently decreased. RNA-seq analysis identified a set of cellular genes that were reproducibly altered in the mutant-induced versus the WT-induced tumors (including some genes previously shown to be regulated by loss of EBNA3A expression in LCLs) and suggested enhanced T cell infiltration in the EBV Δ3A-infected tumors (confirmed by immunohistochemistry [IHC]). Interestingly, Δ3A-infected tumors also had decreased expression of the cellular protein CLEC2D, which has previously been shown to influence the responses of T and NK cells (34–37). Together, these results suggest that EBNA3A may influence B cell lymphomagenesis by affecting both the host cell gene expression and the immune responses.

RESULTS

Construction of an EBNA3A hypomorph mutant virus defective for transforming B cells in vitro.

To examine the roles of EBNA3A in EBV induction of lymphomas in the CBH mouse model, the EBNA3A gene in the B95.8 EBV bacmid p2089 was altered by inserting two stop codons near the start of the EBNA3A protein, as described both in Materials and Methods and in Supplemental Methods S1 in the supplemental material. Infectious virus particles were produced by transfecting 293 cells stably infected with either the WT, the mutant, or revertant bacmids with BZLF1 (Z), BRLF1 (R), and BALF4 (Gp110) expression vectors and then determining the titers of virus, as described in Materials and Methods (38). Although the Δ3A mutant contained the inserted first and second stop codons in the EBNA3A gene as expected, it also had an unexpected nucleotide deletion that alters the EBNA3A coding frame (Fig. 1A). To determine if these mutations eliminate EBNA3A expression, we inserted identical mutations into an EBNA3A vector and examined the expression of EBNA3A in EBV-negative BJAB cells transfected with wild-type or mutant proteins (Fig. 1B). Surprisingly, a small amount of EBNA3A was expressed from the mutant construct; this slightly smaller form of EBNA3A is likely initiated from an in-frame ATG codon at residue 15, since it was eliminated when this residue was mutated. To determine which ATG is normally used to initiate EBNA3A translation, we also individually mutated each ATG codon (alone or in combination) in the EBNA3A expression vector. As shown by the results in Fig. 1B, mutation of either the upstream ATG alone or the downstream ATG motif alone did not significantly affect EBNA3A expression, although no protein was made when both ATG motifs were eliminated. These results suggest that EBNA3A can be derived using either the upstream or downstream ATG motif, although the protein is initiated from the upstream ATG codon when it is present.

FIG 1.

EBNA3A mutant (Δ3A) makes a small amount of EBNA3A but is not capable of transforming CBMCs in vitro. (A) Schema of the mutations of the EBNA3A mutants used in this study. (B) Using site-directed mutagenesis, wild-type and mutant EBNA3A expression vectors (cloned into the SG5 vector) were generated that (i) knocked out the first ATG (1st ATG mutant) (ii) or the second ATG (2nd ATG mutant), (iii) contained the specific mutations of the Δ3A mutant used in this study (Δ3A mutant), (iv) knocked out both the first and second ATGs (1st and 2nd ATG mutant), or (v) contained the specific mutations of the Δ3A mutant used in this study as well as knockout of the second ATG. BJAB cells were Nucleofected with each construct, and immunoblot analysis performed to compare expression levels of EBNA3A and actin as indicated. (C) Total human cord blood mononuclear cells (CBMCs) were infected with either WT virus (10,000 [10K] infectious units), or Δ3A mutant (10,000 and 40,000 [40K] infectious units), and 10 wells for each condition were examined for LCL outgrowth at 4 weeks after infection.

Given that the EBV Δ3A mutant virus can potentially make a small amount of EBNA3A protein, we next asked if it can transform cord blood B cells in vitro. Cord blood mononuclear cells (CBMCs) were isolated using the Ficoll gradient technique and infected with either WT or Δ3A virus as described in Materials and Methods. Infection with the Δ3A virus did not result in any transformed B cell lines (even when using a higher dose of this virus), while infection with the WT EBV resulted in transformed B cell lines being generated efficiently, as expected (Fig. 1C). Consistent with this result, the Δ3A mutant virus was recently shown to have a phenotype similar to that of another independently generated EBNA3A deletion mutant EBV in newly infected B cells (30). Together, these results reveal that EBNA3A can potentially be derived from either of two different in-frame ATG codons, although the upstream ATG codon is normally used, and confirm that the Δ3A virus generated for these studies is indeed highly defective for B cell transformation in vitro. Of note, the rare in vitro LCLs obtained using another EBNA3A deletion mutant virus exhibited reduced proliferation rates and elevated levels of apoptosis in comparison to the results for lines obtained with wild-type virus (19).

The EBNA3A mutant virus causes lymphomas in CBH mice with a delayed onset.

We next asked if Δ3A induces lymphomas in cord blood-humanized (CBH) mice. As shown by the results in Fig. 2A, while there was a tendency for Δ3A virus to induce fewer lymphomas than the wild-type (or revertant) virus did, this difference was not statistically significant. However, the lymphomas that developed in the Δ3A virus-infected mice occurred at later time points; an example of this is shown by the results in Fig. 2B These results indicate that the Δ3A mutant, while severely attenuated for the ability to transform cord blood B cells in vitro, is only partially impaired for the ability to cause lymphomas in the CBH model.

FIG 2.

Δ3A mutant virus induces lymphomas in a cord blood-humanized mouse model with delayed onset. (A) The frequencies of tumors in CBH mice using various doses of both WT and EBNA3A mutant (Δ3A) virus are shown (number of mice that develop tumors/number of mice infected). Left, experiments were performed using 10,000 infectious virus units; middle and right, experiments were performed using 30,000 infectious virus units (titers were determined on either Raji or Akata cells and are expressed in green Raji units or green Akata units [GRU and GAU, respectively, i.e., the amount of virus required for 1 green fluorescent protein-positive cell in each cell line], as described in Materials and Methods). NS, not significant. (B) The rates and incidences of tumors infected with WT (Revertant) versus Δ3A mutant viruses at different time points after infection are shown. A log-rank statistical test was performed to determine statistical significance for the Kaplan-Meier curve analysis.

Lymphomas induced by Δ3A and WT virus are phenotypically similar.

Formalin-fixed, paraffin-embedded (FFPE) tissue samples from animals bearing tumors from either WT or Δ3A virus were examined by hematoxylin and eosin (H&E) staining (Fig. 3). Both viruses induced highly aggressive activated diffuse large B cell lymphomas that stained positive for the B cell marker CD20 and for interferon (IFN) regulatory factor 4 (IRF4; a marker for the activated DLBCL subtype) and invaded multiple organs, including pancreas, liver, biliary tract, kidney, mesentery, abdominal lymph node, and (rarely) spleen. IHC staining of FFPE tissues showed that both WT- and Δ3A mutant-induced lymphomas expressed EBNA2, consistent with type III latency (Fig. 3). Interestingly, Δ3A and wild-type virus-infected tumors expressed similar levels of p16, in contrast to tumors infected with the EBNA3C deletion mutant virus, which, as we previously reported, had elevated p16INK4a expression (and served as a positive control for elevated p16INK4a expression) (Fig. 4) (32). Together, these results suggest that reduced EBNA3A expression does not significantly affect the phenotype of the EBV-induced tumors in CBH mice, although the tumors occur later.

FIG 3.

EBNA3A mutant virus causes aggressive activated diffuse large B cell lymphomas with type III latency. H&E staining of WT and Δ3A mutant tumors is shown as indicated. Immunohistochemistry (IHC) analysis of WT and Δ3A mutant tumors using antibodies against CD20 (B cell marker), EBNA2 (EBV type III latency protein), and IRF4 (marker of activated DLBCL) was performed as indicated. Levels of magnification are indicated.

FIG 4.

Δ3A mutant tumors express levels of p16 similar to the levels in WT tumors. IHC analysis was performed using an antibody that recognizes p16 on WT, Δ3A, and EBNA3C (Δ3C) deletion mutant tumors as indicated.

Lymphomas infected with Δ3A express less LMP1.

Since B cells infected with an EBNA3A deletion mutant EBV were previously shown to have similar levels of EBNA2 but less LMP1 than WT virus-infected cells in another humanized mouse model (31), we isolated protein extracts from two WT- and two Δ3A-infected tumors (each tumor arising in a distinct animal) and performed immunoblotting to quantitate EBNA2 and LMP1 expression. As shown by the results in Fig. 5, the two Δ3A lymphomas examined expressed lower levels of LMP1 than the two WT-EBV-infected lymphomas did (Fig. 5), although there was variation in LMP1 expression from tumor to tumor. Given that we previously showed that an EBV LMP1 deletion mutant causes lymphomas in CBH mice (33) (since CD4 T cell-derived CD40L appears to substitute for LMP1 in this model), the decreased LMP1 expression observed in Δ3A-infected lymphomas is unlikely to inhibit tumor cell growth.

FIG 5.

Δ3A mutant tumors express lower levels of EBNA3A and LMP1 than WT tumors. Tumor lysates were subjected to immunoblot analysis using antibodies against EBNA2 (EBV latency protein), LMP1 (EBV latency protein), EBNA3A, and actin. Each tumor sample was obtained from a distinct animal. Vertical bar indicates splicing of lanes in the blot.

RNA-seq analysis shows similar levels of lytic EBV gene expression.

RNA-seq analysis was performed on tumor tissues collected from WT- and Δ3A-induced lymphomas to compare the levels of EBV-specific transcripts and cellular genes. In both cases, viral gene expression patterns were consistent with type III latency and only very low levels of lytic transcripts were observed (Fig. 6 and Table S1). We further examined the results of the RNA-seq analysis to determine if there were differences in the EBV latent gene expression levels in the WT- versus Δ3A-induced lymphomas. Reads per kilobase per million (RPKM) values were calculated from the RNA-seq data set and plotted in Fig. S1. We observed that several EBV latent genes were expressed at lower levels in the Δ3A-induced lymphomas, including EBNA2/EBNALP, EBNA3B, EBNA1, and LMP2A, while one of the two Δ3A-infected lymphomas examined expressed a lower level of LMP1 transcript than the three WT EBV tumors. These results could either reflect a decreased number of EBV-infected cells in the Δ3A-induced lymphomas (as suggested by the concomitant decrease in EBER expression [Fig. S1 and Table S1]) or suggest that EBNA3A contributes to enhanced latent gene expression in Δ3A-induced lymphomas. Of note, since we did not observe a corresponding decrease in EBNA2 protein expression in the Δ3A-induced lymphomas by immunoblot analysis but did observe a decrease in LMP1 protein expression in one of the two Δ3A-induced lymphomas (Fig. 5), EBNA3A may be affecting LMP1 expression through transcriptional and/or posttranscriptional mechanisms.

FIG 6.

EBV genome mapping of RNA-seq reads in WT virus- and Δ3A virus-induced tumors. RNA-seq reads were mapped to the EBV genomes of 3 different tumors infected with WT virus and two different tumors infected with Δ3A virus. The locations of various EBV genes are indicated. Top, the transcripts shown map to the leftward strand (blue); bottom, the transcripts shown map to the rightward strand (red).

Δ3A lymphomas are derived from oligoclonal expansions of infected B cells.

Analysis of the B cell receptor (BCR) sequences (derived from RNA-seq data) of WT EBV lymphomas in CBH mice previously showed that most tumors are monoclonal or oligoclonal, indicating that they are derived from an expansion of a limited number of B cells (32). Consistent with the previous study, the majority of lymphomas infected with either the WT or Δ3A virus in this study had a single dominant BCR clone (Fig. 7), although one of the three WT-infected lymphomas examined was polyclonal. The B cell population in the spleens of mock-infected animals (injected with cells from the same cord blood donor used to derive the EBV-infected tumors) was more heterogenous, with a variety of different IGHV genes being expressed. Of note, IGHV4-34, which is selected for in some human lymphoma types (39), was relatively highly expressed in both of the mock-infected animals and was the predominant clone in WT tumor 1. Interestingly, B cell malignancies that arise from B1-type B cells commonly express IGHV4-34, suggesting that these tumors may arise from B1 B cells, which are greatly enriched in cord blood (40, 41).

FIG 7.

WT and Δ3A mutant virus-induced lymphomas are usually derived from a restricted B cell population. RNA-seq analysis was performed using RNA isolated from 3 different tumors infected with WT EBV, two different tumors infected with the Δ3A mutant virus, or splenic tissue of mock-infected CBH mice (all mice humanized with the same cord blood). The relative frequency of IGH transcripts containing various different IGHV genes is shown for each tumor and for mock-infected splenic B cells.

Gene expression analysis suggests alterations in several cellular pathways, including BCR signaling.

To compare cellular gene expression levels, RNA-seq reads were aligned to the mouse and human genomes and mouse genomes were removed from the analysis. Several hundred genes were found to be differentially expressed between the WT and Δ3A lymphomas (Fig. 8 and Table S2). Comparison of the genes differentially regulated in our data set to a previous microarray data set comparing cellular gene expression in LCLs derived from WT (B95.8) versus EBNA3A knockout virus (19) revealed a relatively small overlap; genes found to be altered in a similar manner in both data sets are summarized in Tables 1 and 2. Interestingly, a number of these overlapping genes are known to affect BCR signaling, including the genes for the negative BCR regulators ATP2B4 (decreased by EBNA3A in both CBH mice and LCLs), ID3 (increased by EBNA3A in both LCLs and CBH mice), and MSC (increased by EBNA3A). Another interesting consistent gene difference is increased expression of the HHEX gene in cells with mutated EBNA3A, since HHEX has been reported to repress p16INK4a expression and may substitute in this regard for loss of EBNA3A expression (42).

FIG 8.

Comparison of cellular transcripts in lymphomas infected with WT EBV versus the Δ3A mutant virus reveals differentially expressed genes. RNA was isolated from tumors infected with WT- or Δ3A virus-induced lymphomas, and RNA-seq performed. Mouse cell transcripts were removed from further analysis, and the levels of human genes in each tumor type were compared as described in Materials and Methods. The top 100 differentially expressed cellular genes in the RNA-seq analysis are shown.

TABLE 1.

Upregulated genes in EBNA3A mutant EBV-infected lymphomas that match previously reported EBNA3A targets^a

Gene product	RPKM in lymphomas infected with:		Fold increase
	WT	Δ3A
RPS6KA5 (MSK1)	0.83	2.05	2.48
WNT5A	0.093	1.59	17.03
MAL	0.34	2.21	6.44
EHD3	0.94	6.53	6.93
FNBP1L	0.15	1.55	10.62
LGR4	0.0084	0.091	10.73
HHEX	0.33	1.39	4.24
ATP2B4	2.96	6.96	2.35
RGS13	0.54	3.49	6.49

^aGenes upregulated in RNA-seq results that are also upregulated in previously published microarray data (19). Genes of interest are in boldface.

TABLE 2.

Downregulated genes in EBNA3A mutant EBV-infected lymphomas that match previously reported EBNA3A targets^a

Gene	RPKM in lymphomas infected with:		Fold decrease
	WT	Δ3A
ITGAM	1.19	0.27	4.35
BIN1	2.68	1.27	2.11
ZNF395	2.06	0.95	2.17
TPCN1	1.63	0.86	1.89
PFKFB4	0.86	0.30	2.89
PLEKHA1	0.31	0.084	3.73
IMPA2	0.92	0.13	7.22
F13A1	4.32	1.03	4.20
ID3	2.65	0.45	5.87
CLEC2D	31.43	11.48	2.74
CHST12	7.74	3.83	2.02
EPB41L2	4.60	1.31	3.51
SLC16A7	5.73	2.57	2.23
MSC	15.80	4.37	3.62
FCER1G	12.70	0.38	33.76
CLECL1	12.43	1.54	8.09
PIK3R6	4.34	1.39	3.13

^aGenes downregulated in our RNA-seq results that are also decreased in previously published microarray data (19). Genes of interest are in boldface.

Δ3A lymphomas have a gene signature pattern suggestive of reduced chromatin silencing and increased T cell infiltration.

To identify signaling pathways affected by loss of EBNA3A, we analyzed RNA-seq data using Gene Set Enrichment Analysis (GSEA), including the Gene Ontology (GO) Biological Process gene sets. Gene sets that were downregulated in the lymphomas infected with the Δ3A virus included “GO Chromatin Silencing” and “GO Negative Regulation of Gene Expression on Epigenetic” (Fig. 9A and B).These signatures are consistent with the known ability of EBNA3A to induce heterochromatin on certain target gene promoters, such as CDKN2A. Interestingly, the most downregulated gene set in Δ3A lymphomas was the chromosome 19p13 gene set (Fig. 9C), suggesting that lymphomas infected with Δ3A may contain a deletion in chromosome 19. However, slides with specimens from the WT and Δ3A lymphomas examined by fluorescent in situ hybridization (FISH) analysis (using a probe recognizing a region of 19p13) did not detect any deletion (Fig. 9D). Therefore, EBNA3A may broadly activate transcription (perhaps via a superenhancer) of genes borne in this region of chromosome 19.

FIG 9.

Gene set enrichment analysis (GSEA) suggests decreased gene silencing and expression from chromosome 19 in EBNA3A mutant tumors. (A to C) GSEA plots for the “GO Chromatin Silencing” (A), “GO Negative Regulation of Gene Expression on Epigenetic” (B), and “CHR19P13” (C) gene sets are shown comparing the Δ3A-infected tumors to the WT-infected tumors as indicated. NES, normalized enrichment score. (D) FISH analysis of WT- and EBNA3A mutant-induced tumors using probes for chromosome 19 regions 19p and 19q. Average number of signals per cell is shown for each probe. A ratio comparing the average numbers of signals per cell for the 19p probe and 19q probe was determined. A range of ratios of 0.8 to 1.2 is considered normal (no deletion).

Gene sets that were upregulated in Δ3A lymphomas included “GO T Cell Receptor Complex,” “Reactome Downstream TCR Signaling,” “Reactome TCR Signaling,” and “GO Antigen Receptor Mediated Signaling Pathway” (Fig. 10), suggesting that reduction of EBNA3A is associated with increased T cell infiltration of tumors and elevated BCR signaling. IHC analysis showed that Δ3A-infected tumors contained an increased number of both CD4⁺ and CD8⁺ T cells compared to the numbers in the WT virus-infected tumors (Fig. 11).

FIG 10.

GSEA analysis suggests increased T cell infiltration and decreased BCR signaling in EBNA3A mutant tumors. (A to D) GSEA analysis plots for the “GO T Cell Receptor Complex” (A), “Reactome Downstream TCR Signaling” (B), “GO Antigen Receptor Mediated Signaling Pathway” (C), and “Reactome TCR Signaling” (D) gene sets are shown comparing the Δ3A-infected tumors to the WT-infected tumors as indicated.

FIG 11.

EBNA3A mutant tumors have elevated T cell infiltrate compared to WT tumors. IHC was performed using antibodies against CD20 (B cell marker), CD4 (helper T cell marker), and CD8 (cytotoxic T cell marker) on lymphomas infected with either the Δ3A or WT virus as indicated.

EBNA3A promotes CLEC2D expression.

RNA transcripts for CLEC2D, a ligand for the NKR-P1A receptor that is expressed by all NK cells and a fraction of T cells, were consistently decreased in the Δ3A lymphomas. Immunoblot analysis on primary tumor tissues (identical to samples used in the analysis whose results are shown Fig. 5) obtained from Δ3A-infected versus WT-infected mice (Fig. 12A) confirmed that Δ3A-infected tumors express significantly less CLEC2D protein than their WT counterparts. Furthermore, immunoblot analysis of LCLs (derived from two different donors) infected with wild-type EBV or another (full deletion) EBNA3A mutant virus (30) showed that EBNA3A likewise promotes the expression of CLEC2D in LCLs in vitro (Fig. 12B).

FIG 12.

EBNA3A enhances the expression of CLEC2D in B cells. (A) CBH mouse lymphoma lysates infected with the WT or viruses were subjected to immunoblot analysis using antibodies against CLEC2D, EBNA3A, and CD19 (used as a loading control for number of B cells) as indicated. (B) Whole-cell lysates obtained from two different sets of LCLs infected with WT (B95.8) EBV or an alternative EBNA3A mutant EBV (in which the entire gene has been deleted) (30) were subjected to immunoblot analysis using antibodies against EBNA3A, CLEC2D, and actin as indicated.

Prior studies have suggested that the expression of CLEC2D may facilitate tumor escape from NK cells (34–37). To determine if loss of EBNA3A expression is sufficient to increase the NK cell response to EBV-infected LCLs in vitro, NKL cells (an immortalized NK cell line) were coincubated with either WT- or EBNA3A deletion mutant EBV-transformed LCLs (the same cell lines as used in the analysis whose results are shown in Fig. 12B) and the NK cell production of IFN-γ and cell surface expression of CD107a (an indicator of recent degranulation) were determined by flow cytometry. NK cell responses were similar in the presence and absence of EBNA3A expression in LCLs (Fig. 13). Thus, although CLEC2D expression is markedly diminished by EBNA3A, this change may not be sufficient to promote NK cell responses to EBV-transformed LCLs.

FIG 13.

NK cells are activated similarly to LCLs derived from WT and EBNA3A deletion mutant viruses in vitro. NKL cells were coincubated with EBNA3A-sufficient or -deficient LCLs, and effector cytokine responses and levels of cytotoxicity by the responding NK cells were determined using intracellular expression of IFN-γ (A) and surface expression of CD107a, a marker of granule release during cytotoxic responses (B). Geometric mean fluorescence intensities (gMFI) are graphed for each sample. Two unique donors were used in this study, and C5 and D3 indicate the unique donor.

EBNA3A mutant tumors do not contain confirmable cellular frameshift mutations.

To determine if loss of EBNA3A expression selects for compensatory cellular-gene frameshift mutation(s), we compared RNA-seq reads of cellular-gene exons expressed in lymphomas (obtained using the same donor cells) infected with wild-type versus Δ3A EBV, as described in Materials and Methods. Using this approach, we identified 4 genes containing a potential frameshift mutation in one or more Δ3A tumors but not the wild-type tumors (Table 3). However, Sanger sequencing of PCR-amplified gene regions surrounding each potentially mutated gene sequence revealed only wild-type sequences. Although this result does not exclude a very low frequency of these mutations within the tumors, more likely it reflects an error in the original RNA-seq library amplification or sequencing, particularly since the majority of potential frameshift mutations identified occurred within stretches of nucleotide repeats, MBD4 being an example (Fig. S2).

TABLE 3.

Cellular mutation analysis of EBNA3A mutant EBV-infected tumors^a

Gene	Chromosome	Position	Residue(s) in indicated sequence		Type of alteration	Sample	Frequency in samples infected with:
			Reference	Alteration			Δ3A	WT
ALG2	chr9	99218509	AT	A	Frameshift deletion	Δ3A 2	0.20	0
ASPM	chr1	197133436	A	AT	Frameshift insertion	Δ3A 2	0.18	0
MBD4	chr3	129436704	CT	C	Frameshift deletion	Δ3A 1	0.19	0
SMG1	chr16	18829449	TG	T	Frameshift deletion	Δ3A 2	0.10	0

^aMutations present in at least one of the Δ3A tumors that were absent in all the WT tumors.

DISCUSSION

In this study, we have used a cord blood-humanized (CBH) mouse model to examine the in vivo roles of EBNA3A in regulating viral gene expression and promoting tumorigenesis. EBNA3A is an EBV latency gene that decreases the levels of p16INK4a expression (in collaboration with EBNA3C) and is important for the ability of EBV to transform B cells in vitro (18). Despite this, EBNA3A knockout LCLs have been generated (although they exhibit reduced proliferation and increased apoptosis), indicating that EBNA3A is not absolutely required for transformation in vitro (19). We show that an EBNA3A hypomorph mutant (Δ3A) that makes very little EBNA3A protein (and is missing the first 14 amino acid residues) is highly defective for transforming cord blood B cells in vitro but induces lymphomas in CBH mice at delayed time points. WT and Δ3A tumors were phenotypically similar in CBH mice (both resembling activated human DLBCLs) and expressed similar levels of p16, although LMP1 expression was decreased in some (but not all) Δ3A lymphomas. As previously noted for lymphomas infected with an EBV EBNA3C deletion mutant in CBH mice, lymphomas infected with the Δ3A virus had an RNA-seq signature suggestive of enhanced T cell infiltration and were more heavily infiltrated with T cells on IHC analysis. Comparison of human gene exon sequences obtained from lymphomas infected with WT versus Δ3A viruses did not reveal compensating cellular frameshift mutations selected for in Δ3A lymphomas. Therefore, we hypothesize that factors derived from a supportive tumor microenvironment (such as CD40L-expressing CD4 T cells) can partially compensate for the loss of EBNA3A expression in the CBH model.

The finding here that Δ3A lymphomas in CBH mice in some cases express less LMP1 is consistent with a previous study using a different EBNA3A mutant in another humanized model (31). In that study, although the EBNA3A deletion mutant EBV was unable to induce lymphomas, it was able to establish persistent viral latency, and mutant virus-infected cells had less LMP1 expression than WT virus-infected cells (31). LMP1 expression is activated by EBNA2 in cells with type III latency, and a previous study found that many of the differentially expressed genes in LCLs infected with WT versus EBNA3A deletion mutant EBV are EBNA2 target genes (19). Furthermore, although EBNA3A was found to decrease the expression of some EBNA2 target genes in LCLs (perhaps by competing for interaction with RBPJ-κ), in other cases, it cooperated with EBNA2 to activate target genes (19). Although LCLs derived from the WT versus the EBNA3A deletion mutant EBV in vitro expressed similar levels of LMP1 (19), the results of the two humanized mouse studies suggest that EBNA3A and EBNA2 cooperatively activate LMP1 protein expression. The exact mechanism(s) for this effect (and whether it is primarily transcriptional and/or posttranscriptional) remains to be determined. The different results obtained in vitro versus in vivo likely reflect the fact that LMP1 expression is more essential for the growth of LCLs in vitro (and is thus selected for) than it is for persistent EBV infection in humanized mice (in which CD40 ligand-expressing T cells can at least partially substitute for LMP1).

The ability of the Δ3A mutant virus to induce lymphomas in the CBH mouse model while not being able to transform cord blood B cells efficiently in vitro (at least under the conditions used in this study) is likely at least partially due to the supportive tumor microenvironment in this model, which contains a high number of CD40L-expressing CD4 T cells that can substitute for loss of LMP1 expression (33). Likewise, we previously showed that an EBNA3C deletion mutant virus can induce lymphomas with delayed onset in the CBH model, while in another humanized mouse model, an EBNA3C mutant established persistent viral latency but not lymphomas (31, 32). Although we cannot totally exclude the possibility that a small amount of residual EBNA3A expression derived from the downstream ATG codon (residue 15) contributes to the ability of our EBNA3A mutant to induce lymphomas in CBH mice, we think this is unlikely since the mutant is defective for transforming cord blood B cells in vitro. In addition, the EBNA3A mutant virus used in this study affected a number of cellular genes (including the NK cell inhibitor CLEC2D) in a manner similar to that of a mutant virus with EBNA3A fully deleted in LCLs.

Interestingly, the rhesus lymphocryptovirus EBNA3A protein also contains a downstream in-frame ATG codon in a similar region of the protein, and deletion of both ATG codons is required to inhibit all protein expression in cells transfected with a rhesus lymphocryptovirus EBNA3A expression vector (43). As yet unknown is whether the downstream ATG codon in EBNA3A is used preferentially under certain conditions and whether the shorter form of the protein (missing the first 14 amino acids) behaves any differently from the longer full-length form. Of note, the human type 2 EBV EBNA3A protein does not preserve the ATG codon at residue 15, suggesting that the upstream ATG codon may be the only initiating codon used under normal circumstances.

Unexpectedly, we did not observe increased p16 expression in lymphomas infected with the Δ3A virus in CBH mice, although we previously found increased p16 expression in lymphomas infected with an EBNA3C deletion mutant. EBNA3C may be more important than EBNA3A for repressing p16 expression in the CBH model (32). Interestingly, the HHEX cellular gene, which encodes a homeobox protein that inhibits p16 expression (42), was overexpressed both in Δ3A lymphoma cells in CBH mice and in LCLs derived from an EBNA3A mutant virus in vitro (19). Overexpression of HHEX may be selected for in the absence of EBNA3A expression in EBV-infected B cells as an alternative mechanism for inhibiting p16 expression.

We cannot exclude the possibility that the decreased ability of the Δ3A mutant to transform B cells in vitro is due to the use of cord blood in our assays, since previous studies have generated LCLs with EBNA3A deletion mutant EBVs using peripheral blood B cells (PBLs), and EBNA-LP is required for transformation of cord blood but not PBLs in vitro (19, 30, 44). It is possible that EBNA3A has similar roles in transformation of cord blood and PBLs, and further studies would need to be performed to determine this. Since EBNA3A has been shown to protect infected cells from terminal plasma cell differentiation in vitro (21), it is also possible that infected cord blood cells are terminally differentiating and hindering our ability to generate LCLs.

Consistent with what we have previously observed in EBV-infected CBH mice, with the exception of one WT virus tumor, the WT and EBNA3A mutant tumors contained monoclonal/oligoclonal B cells and, thus, were derived from expansion of a small population of EBV-infected cells (Fig. 7). In contrast, mock-infected animals (which received the same cord blood as the EBV-infected cells) had polyclonal expansions of B cells in their spleens (Fig. 7). Interestingly, the dominant IGHV gene expressed in both of the mock-infected animals, as well as one WT virus-infected lymphoma, was IGHV4-34. The antibodies encoded by the IGVH4-34 gene are intrinsically autoreactive because they recognize conserved carbohydrate self-epitopes expressed at high levels on red blood cells and other cell types and do not require somatic mutation or the associated light chains for this effect (39, 45–49). IGVH4-34 usage is associated not only with autoimmune diseases such as systemic lupus erythematosus (49) but also with B cell lymphomas (including Burkitt lymphomas) and chronic lymphocytic leukemia (45, 50–52), where it is thought to promote tonic BCR signaling (39, 52). IGHV4-34 gene usage may be selected for in both mock- and EBV-infected cells in CBH mice to induce BCR signaling and survival (53).

Signaling pathways identified by RNA-seq data analysis indicated several differences between WT and Δ3A tumors. WT virus-infected cells had higher signatures of “GO Chromatin Silencing” and “GO Negative Regulation of Gene Expression on Epigenetic,” consistent with the known ability of EBNA3 proteins to inhibit target gene expression via epigenetic mechanism(s). WT cells also had a highly significant increase in the expression of genes located in chromosome 19 region p13. Since we did not detect a deletion of this region in the Δ3A lymphomas (Fig. 9), we speculate that EBNA3A may activate the expression of cellular genes in this region, perhaps by interacting with a superenhancer. Gene sets enriched in lymphomas infected with the Δ3A virus included gene sets related to T cell receptor signaling (“GO T cell Receptor Complex,” “Reactome Downstream TCR Signaling,” and “Reactome TCR Signaling”), reminiscent of the gene signature observed with an EBV EBNA3C deletion mutant in CBH mice (32). EBNA3A and EBNA3C may thus collaboratively inhibit T cell infiltration of tumors. However, we cannot preclude the possibility that the increased T cell infiltration in the EBNA3A mutant tumors may be due to increased exposure to EBV-infected B cells. The “GO Antigen Receptor Mediated Signaling Pathway” signature was also upregulated in Δ3A tumors, suggesting that EBNA3A may dampen BCR signaling. Consistent with this, WT lymphomas expressed higher levels of two different genes (MSC and ID3 genes) encoding proteins that inhibit the function of TCF3, a cellular transcription factor required for efficient BCR signaling (54). Increased MSC expression also occurred in WT EBV-infected LCLs compared to its expression in LCLs infected with an EBNA3A deletion mutant virus (19).

We also found that expression of the cellular CLEC2D protein was decreased by reduction of EBNA3A expression in EBV-infected CBH mice and in EBNA3A-deficient-EBV-infected LCLs in vitro. However, we did not find that loss of EBNA3A expression in EBV-infected LCLs was sufficient to increase the activity of NK cells in response to EBV in vitro. We speculate that EBV-infected B cells with type III latency (in which all 8 viral latency proteins are expressed) have multiple redundant mechanisms for inhibiting NK cell function. Interestingly, while CLEC2D can inhibit NK cells, CLEC2D expression on antigen-presenting cells has been shown to activate T cells (35). Thus, it is also possible that EBNA3A promotes CLEC2D expression as a mechanism for ensuring that T cells eventually eliminate cells with type III latency (thereby preventing the virus from inducing B cell lymphomas and killing the host). In addition, CLEC2D is a signaling receptor that has been shown to activate B cells, and it is possible that EBNA3A increases the expression of CLEC2D to promote survival of the EBV-infected B cells (55, 56).

Together, these results provide insight into previously unrecognized roles of EBNA3A in vivo that are not easy to study in vitro. While EBNA3A is not required for the establishment of EBV-induced lymphomas in CBH mice, it nevertheless has important in vivo functions, such as decreasing T cell infiltration in tumors and promoting CLEC2D expression.

MATERIALS AND METHODS

Ethics statement.

All animal experiments were approved by the University of Wisconsin—Madison Institutional Animal Care and Use Committee (IACUC) and conducted in accordance with the NIH Guide for the Care and Use of Laboratory Animals (protocols number M005197 and M005214) (57). We anesthetized mice using isoflurane and euthanized animals by performing cervical dislocations on anesthetized mice (38).

EBVs and construction of mutant viruses.

Mutant viruses were constructed as described in Supplemental Methods S1 in the supplemental material and in our previous paper (32). All primers used for construction and confirmation of the EBNA3A mutant and the corresponding revertant (ΔEBNA3A-Revert) are provided in Table 4.

TABLE 4.

Primers for constructing EBNA3A mutants

Primer	Sequence	Purpose
EBNA-3A-stop P1	GGTGTTGGTGAGTCACACTTTTGTTGCAGACAAAATGGACTAGGATAGGCCGGGTCCCCCGGCCCTAGGATGACGACGATAAGTAGGG	ΔEBNA3A mutant
EBNA-3A-stop P2	CTTCTTCCATGTTGTCATCCAGGGCCGGGGGACCCGGCCTATCCTAGTCCATTTTGTCTGCAACAACAACCAATTAACCAATTCTGATTAG	ΔEBNA3A mutant
EBNA-3A-Rev.WT-Pr1	GGTGTTGGTGAGTCACACTTTTGTTGCAGACAAAATGGACAAGGACAGGCCGGGTCCCCCGGCCCTAGGATGACGACGATAAGTAGGG	ΔEBNA3A revertant
EBNA-3A-Rev.WT-Pr2	CTTCTTCCATGTTGTCATCCAGGGCCGGGGGACCCGGCCTGTCCTTGTCCATTTTGTCTGCAACAACAACCAATTAACCAATTCTGATTAG	ΔEBNA3A revertant
EBNA3A-Stp-Chk-Fwd	TCCGGTGGTGACGTTAATTG	PCR and sequencing
EBNA3A-Stp-Chk-Rev	ACTTAGGCCATAGGGATGCT	PCR and sequencing
Cam-Ff-Kan-Fwd	CGGGCGTATTTTTTGAGTTATCGAGATTTTCAGGAGCTAAGGAAGCTAAAATGAGCCATATTCAACGGGAAAC	Swap chloramphenicol with kanamycin in F-factor
Cam-Ff-Kan-Rev	CAGGCGTAGCAACCAGGCGTTTAAGGGCACCAATAACTGCCTTAAAAAAATTAGAAAAACTCATCGAGCATC	Swap chloramphenicol with kanamycin in F-factor
2089-CamR-Fwd	CGGGCGTATTTTTTGAGTTATCG	Swap kanamycin with chloramphenicol in F-factor
2089-CamR-Rev	CAGGCGTAGCAACCAGGCG	Swap kanamycin with chloramphenicol in F-factor

Cell lines and production of infectious EBV.

HEK293 cells latently infected with WT EBV or EBNA3A mutant virus were maintained in Dulbecco modified Eagle medium (DMEM) supplemented with 10% fetal bovine serum (FBS), 1% penicillin-streptomycin, and 100 μg of hygromycin B. The NKL cell line was grown in T cell medium (RPMI medium, 10% bovine calf serum, 5% fetal bovine serum, 1% penicillin-streptomycin, 1% Glutamax) supplemented with 50 U/ml interleukin-2 (IL-2). Infectious viral particles were produced from 293 cell lines stably infected with WT or mutant virus following transfection with EBV BZLF1, BRLF1, and gp110 expression vectors as previously described (38). The titer of EBV was determined on Raji cells by using the green Raji cell assay or on EBV-negative Akata cells as previously described (38).

Creation of cord blood-humanized NOD/LtSz-scid/IL2Rγ^null mice.

CBH mice were generated as previously described (32, 33, 38, 58); more details are given in Supplemental Methods S1.

Infection of CBMCs with EBV strains.

Amounts of 1 × 10̂6 cord blood mononuclear cells (CBMCs) were infected with either 10,000 or 40,000 infectious units of virus for 1 h at 37°C. After infection, cells were plated at 100,000 total CBMCs per well (10 wells each). Infected cells were maintained for up to 4 weeks postinfection in RPMI medium containing 20% FBS and 1% penicillin-streptomycin. The number of wells with outgrowth of LCLs was determined 4 weeks postinfection.

Analysis of EBV infection and tumors.

Analysis of tumor tissue was performed as previously described (32); for more details see Supplemental Methods S1. Tissues underwent IHC staining using the antibodies listed in Table 5, as previously described (32, 38, 58).

TABLE 5.

Antibodies used for immunohistochemistry and immunoblotting

Target of antibody	Clone	Manufacturer	Dilution
CD20 (mouse)	H1	BD Pharmingen	1:500
CD20 (rabbit)	BV11	Abcam, Inc.	1:100
CD3	Polyclonal	DakoCytomation	1:200
LMP1 (IHC)	CS 1-4	Abcam, Inc.	1:10
LMP1 (immunoblotting)	CS 1-4	Abcam, Inc.	1:500
EBNA2	PE2	Abcam, Inc.	1:100
EBNA3A	Polyclonal	Exalpha	1:250
IgG	Polyclonal	Cell Marque	1:400
IgM	Polyclonal	Cell Marque	1:400
CD4	EPR68355	Abcam	1:1,000
CD8	D8A8Y	Cell Signaling Technology	1:200
p16	G175-405	BD Pharmingen	1:50
IRF4	MUM1p	Santa Cruz Biotechnology	1:50
Actin	AC-15	Sigma	1:5,000

Immunoblot analysis of tumor protein extracts.

Immunoblot analysis was performed as previously described (32); for more details see Supplemental Methods S1. The primary antibodies used are described in Table 5.

Plasmids and cloning.

Plasmid DNA was prepared using the Qiagen maxiprep kit according to the manufacturer’s instructions. Plasmid pSG5 was obtained from Stratagene. pSG5-EBNA3A has been previously described (59). EBNA3A mutants in the pSG5-EBNA3A vector were constructed using the Stratagene QuikChange II XL site-directed mutagenesis protocol as previously described (60). Primers are listed in Table 6. All constructs were verified by sequencing.

TABLE 6.

Primers used for site-directed mutagenesis

Primer	Sequence	Purpose
First ATG mutant Fwda	GCGGCCGTCTCCTTTAAGACAAATCGGACAAGGACAG	Mutating the 1st ATG
First ATG mutant Reva	CTGTCCTTGTCCGATTTGTCTTAAAGGAGACGGCCG	Mutating the 1st ATG
Second ATG mutant Fwd	ATGGGACTTCTTCTTCCGAGTTCTCATCCAGGGCCG	Mutating the 2nd ATG
Second ATG mutant Rev	CGGCCCTGGATGACAACTCGGAAGAAGAAGTTCCAT	Mutating the 2nd ATG
EBNA3A mutant nucleotide Fwd	GGGGGACCCGGCCTATCCTAGTCCATTTTGTCTTAAAG	Mutating nucleotides in Bac mutantb
EBNA3A mutant nucleotide Rev	CTTTAAGACAAAATGGACTAGGATAGGCCGGGTCCCCC	Mutating nucleotides in Bac mutant
EBNA3A mutant deletion Fwd	GGGGACCCGGCCTATCTAGTCCATTTTGTCTT	Adding deletion in Bac mutant
EBNA3A mutant deletion Fwd	AAGACAAAATGGACTAGATAGGCCGGGTCCCC	Adding deletion in Bac mutant

^aPrimers were used to generate the EBNA3A site-directed mutant used to determine if EBNA3A can be expressed from the second ATG.

^bBACmid (Bac) mutant.

Nucleofection of BJABs.

BJAB cells were transfected by Nucleofection (nucleofected), using the Amaxa Nucleofector 2b device (Lonza) and program M-013 (with buffer V), in 12-well dishes with 150 ng of vector control, pSG5-EBNA3A mutant, and pSG5-EBNA3A mutant as previously described (60). Immunoblot analysis was performed as previously described (32, 60).

RNA-seq analysis of tumor tissue.

RNA-seq was performed as previously described (32); details are in Supplemental Methods S1.

EBV genome analysis.

EBV transcripts were analyzed as previously described (32); details are in Supplemental Methods S1. EBV gene expression was determined using the approach described in Faure et al. (61). Briefly, reads mapping to the unique coding sequence (UCDS) region of each EBV gene were summed (UCDS regions used were the same as those listed in Table S3 of reference 61). For nested transcripts, an estimated count of the overlapping portion of the larger transcript B was subtracted from the raw count of the UCDS region of the smaller transcript A using the following formula: Counts(UCDSA) = RawCounts(UCDSA) − {RawCounts(UCDSB) × [Length(UCDSA)/Length(UCDSB)]}. Where possible, nonoverlapping regions were identified and separated by at least 152 bp. These counts were then normalized by total reads mapped to EBV and human genomes (in millions) and reported as reads per kilobase per million (RPKM) estimates of EBV transcript abundance.

B cell receptor immunoglobulin sequencing.

BCR sequencing analysis was performed as previously described (32); details are in Supplemental Methods S1.

Mutational analysis of RNA-seq data.

To identify potential mutations in the RNA-seq results of the EBNA3A tumor samples, reads were mapped to both human and mouse genomes (GRCh38 and GRCm38) using STAR (version 2.6.1d). Subsequently, NGS-Disambiguate (version 1.0) (62) was used to filter for reads preferentially mapping to the human genome. Mutect2 and Funcotator (GATK, version 4.1.2.0) (63) were then used to call potential frameshift mutations in the EBNA3A mutant tumor samples relative to the sequences of EBNA3A wild-type controls. Calls were then filtered to generate higher-confidence candidates using the following criteria: mutant allele present in at least 10% of calls in either EBNA3A mutant tumor, mutant allele absent in all wild-type controls, and the presence of at least 2 independent reads supporting the mutant allele in at least one of the EBNA3A mutant tumors.

NK cell activity assays.

NKL cells were harvested and labeled with CellTrace violet (Thermo Fisher Scientific) according to the manufacturer’s recommendations. Labeled effector cells were coincubated 1:1 (50,000 cells each) with treated LCLs in 96-well round-bottom plates at a final volume of 200 μl overnight. Six hours prior to harvest, GolgiPlug (brefeldin A; BD) was added to wells that would be evaluated for cytokine effector expression. Cells evaluated for surface CD107a (BioLegend clone H4A3) expression were not treated with GolgiPlug. Cord blood cells alone were also stimulated at this time with phorbol myristate acetate (PMA) and ionomycin with brefeldin A (BioLegend) for 6 h. Cells were harvested, Fc receptors were blocked using 20% human AB serum in phosphate-buffered saline (PBS) for 20 min, cells were pelleted by centrifugation, surface staining was performed (PD-1 clone EH12.2H7, CD69 clone FN50, and KLRG-1 clone 14C2A07, all from BioLegend), and cells were incubated for 30 min at 4°C. Cells were spun down again and prepared for intracellular staining using the BD Cytofix/Cytoperm staining kit and following the manufacturer’s recommendations for IFN-γ (clone 4S.B3; BioLegend) staining. Cells were resuspended with PBS and acquired on an LSRII flow cytometer (BD), and data were analyzed using FlowJo software version 9.9.5 (BD).

Statistical analysis.

All bar graphs were constructed in Microsoft Excel, and standard errors are shown by error bars in graphs. Kaplan-Meier analysis and dot plots were constructed in MSTAT statistical software. The Fisher exact test and log-rank analysis were performed using MSTAT statistical software (https://mcardle.wisc.edu/mstat/). A P value of <0.05 was considered significant in all tests used.

Data accessibility.

The RNA-seq data reported in this paper are available in the BioProject database under BioProject accession number PRJNA603504. All other data are contained within the manuscript and the supplemental material.