A New Model of Epstein-Barr Virus Infection Reveals an Important Role for Early Lytic Viral Protein Expression in the Development of Lymphomas

Immunodeficient nonobese diabetic/severe combined immunodeficient (NOD/LtSz-scid/IL2Rγ^null [NSG]) mice were purchased from Jackson Labs (catalogue no. 005557) and used at 6 to 10 weeks of age. Human fetal thymus and liver tissues (gestational age, 17 to 20 weeks) were obtained from Advanced Bioscience Resource (Alameda, CA). Mice were humanized by following the procedure described previously (31). In brief, the recipient mice were conditioned with sublethal (2 to 3 Gy) whole-body irradiation and implanted with fetal thymus and liver fragments under the recipient kidney capsule after irradiation. The mice also received an intravenous injection of purified CD34⁺ cells isolated from the same fetal liver by the magnetically activated cell sorter (MACS) separation system (Miltenyi Biotec, Auburn, CA). The purity of the injected CD34⁺ cells was at least 80 to 90%. At 10 weeks after immune reconstitution, the levels of human hematopoietic cells in mice were determined by multicolor flow cytometric (FCM) analysis using various combinations of the following antibodies: pan-anti-CD45 (clone HI30), -CD4 (clone RPA-T4 or OKT4), -CD8α (clone RPA-T8), -CD19 (HIB19), and -CD3 (SP34-2). Antibodies directly conjugated to fluorescent dyes were purchased from commercial sources. For flow cytometric analysis, blood was collected from mice 10 weeks after implantation of human cells. Samples were subjected to ACK lysis to remove red blood cells and further purified by density gradient centrifugation using Histopaque (Sigma). Fluorescence-activated cell sorting (FACS) analysis was performed on a FACSCalibur (Becton Dickinson, Mountain View, CA). Dead cells were excluded from the analysis.

In one experiment, humanized mice were made without cotransplanting fetal thymic tissue. In brief, NSG mice were irradiated with 2 Gy 24 to 48 h after birth, and purified fetal liver CD34⁺ cells (0.5 million) were injected intrahepatically into each mouse. Immune reconstitution was documented as above.

Wild-type (WT) EBV (B95-8 strain), expressing the green fluorescent protein (GFP) and a hygromycin B resistance gene, and the BZLF1-deleted mutant virus (Z-KO) were constructed using bacterial artificial chromosome technology as described previously (7, 8) and were a gift from Henri-Jacques Delecluse. WT and Z-KO 293 cells were maintained in Dulbecco's modified Eagle medium (DMEM) containing 10% fetal bovine serum (FBS), 1% penicillin-streptomycin, and hygromycin B (100 μg/ml; Roche). Raji, an EBV-positive Burkitt lymphoma cell line, was maintained in RPMI 1640 medium with 10% FBS and 1% penicillin-streptomycin. Plasmid DNA was purified through columns as described by the manufacturer (Qiagen). The EBV pSG5-BZLF1 and pSG5-gp110 expression vectors were gifts from Diane Hayward (34) and Henri-Jacques Delecluse (26), respectively.

293 cells latently infected with the WT or Z-KO viruses were converted to the lytic form of EBV infection by transfecting them with BZLF1 and gp110 expression vectors by using the FuGENE 6 transfection reagent (Roche) as per the manufacturer's protocol. Supernatants were harvested at 72 h posttransfection and filtered through a 0.45-um-pore-size filter. The virus was concentrated by centrifuging at 18,000 rpm for 3 h using an SW27 rotor, resuspended in phosphate-buffered saline (PBS) overnight at 4°C, and then stored at −80°C. To determine the titer of the EBV stock, Raji cells were infected with serial 10-fold dilutions of virus. After 48 h, cells were treated with 50 ng/ml phorbol-12-myristate-3-acetate (PMA; Sigma) and 3 mM sodium butyrate (Sigma), and the GFP-expressing Raji cells were counted 24 h later by fluorescence microscopy. The amount of virus required to form one GFP-positive Raji cell was defined as one green Raji unit (GRU).

Mice were injected intraperitoneally (i.p.) with 2,750 GRUs of WT or Z-KO EBV in 250 μl PBS or mock infected with PBS alone. In most experiments, half of the mice derived from a particular donor (8 to 12 mice per donor) were infected with the WT virus, and the other half were infected with the Z-KO virus. In one experiment, half of the mice reconstituted with a particular donor were infected with the Z-KO virus, and the other half were mock infected (and thus more animals were infected with Z-KO virus than with the control WT virus). One WT EBV-infected animal and one Z-KO virus-infected animal were sacrificed at days 3 and 20 postinfection, and the rest of the animals were sacrificed at 60 to 65 days postinfection unless they exhibited clinical signs, including weight loss, disturbed gait, or hair loss. Organs, including spleen, lymph node, lung, liver, pancreas, kidney, heart, implanted human fetal thymus, head, bone marrow, muscle, stomach, bowel, and salivary gland, were collected and fixed with 10% formalin and then paraffin embedded.

EBER in situ hybridization studies were performed using the PNA ISH detection kit (DakoCytomation) according to the manufacturer's protocol. Briefly, tissues were deparaffinized and rehydrated and then treated with 1:10 diluted proteinase K at room temperature for 20 min. After tissues were washed, EBER probe was added on top of the tissue and covered with a cover slide. Hybridization was performed at 55°C for 90 min. After hybridization, tissues were washed with stringent washing solution at 55°C for 30 min. Anti-fluorescein isothiocyanate (anti-FITC)/AP was added and incubated at room temperature for 30 min, and substrate was then added to develop the color. Tissues were counterstained with nuclear fast red. To compare the number of EBV-infected cells in animals infected with the WT versus that in animals infected with the Z-KO virus, we calculated the number of EBER-positive cells in each mouse by counting the positive cells in slides that included sections containing the kidney, liver, implanted fetal thymus, spleen, and lung from each animal. Although other tissues or organs, such as lymph nodes, also contained some EBER-positive cells, since these tissues were not consistently present on the slides of all animals, they were excluded in the comparison.

Plasma samples (approximately 100 μl) were collected from jugular vein bleeds after euthanasia of animals and frozen. DNA was extracted using the QIAamp DNA minikit and eluted in 100 μl nuclease-free water. Five microliters of the DNA was applied in duplicate quantitative PCR (Q-PCR) targeting the EBV BamH1W segment [EBV W1, 5′-GCAGCCGCCCAGTCTCT-3′; EBV W2, 5′-ACAGACAGTGCACAGGAGCCT-3′; and EBVWprobe, 5′-(6-carboxyfluorescein)AAAAGCTGGCGCCCTTGCCTG(6- carboxytetramethylrhodamine)-3′] as previously described (33). Human ApoB DNA was quantitated in a separate well as a marker of the efficacy of extraction and amplification. The EBV PCR assay has a sensitivity of 6 copies per PCR.

Formalin-fixed, paraffin-embedded tissue sections and cells were deparaffinized and hydrated and then treated with 10 mM citrate buffer (0.05% Tween 20, pH 6.0) for 20 min in a water bath at 98°C. Endogenous peroxidase activity was blocked with 0.3% hydrogen peroxidase solution, and nonspecific labeling was blocked in a 5% goat serum blocking solution. Sections were incubated with the first antibody for 1 h at room temperature. The Super Sensitive polymer-horseradish peroxidase immunohistochemistry detection system (BioGenex Inc.) was used by following the manufacturer's instruction. Colors were developed with the diaminobenzidine tetrachloride (DAB) substrate kit (Vector Laboratories Inc.) by following the manufacturer's instruction. In the cases of double staining, a combination of the DAB substrate kit (nickel solution was added to DAB) and the Vector VIP substrate kit (Vector Laboratories Inc.) was used by following the manufacturer's instruction. Antibodies (and dilutions) used were as follows: anti-CD3 (polyclonal; DakoCytomation; 1:200), anti-CD20 (clone H1, BD Pharmingen; 1:600), anti-CD27 (clone M-T271, BD Pharmingen; 1:300), anti-LMP1 (CS.1-4, DakoCytomation; 1:800), anti-EBNA1 (EB14, a gift from Richard Burgess, University of Wisconsin; 1:2,000), anti-EBNA2 (PE2, Abcam; 1:100), anti-BZLF1 (BZ1, Santa Cruz; 1:200), anti-BMRF1 (G3-E31, Vector Laboratories; 1:200), anti-GP350/220 (OT6, gift from Jaap Middeldorp, Vrije Universiteit University Medical Center, Netherlands; 1:2,000), anti-CD4 (BC/1F6, Biocare Medical; 1:25), and anti-CD8 (SP16, Biocare Medical; 1:50). In addition, to confirm type IIB latency status (EBNA2⁺/LMP1⁻), we repeated LMP1 staining using a different anti-LMP1 antibody (OTC21C, a gift from Jaap Middeldorp; 1:200) in some animals. EBV 293 cells were converted to the lytic form of EBV infection by transfecting them with the SG5-BZLF1 vector and then fixed with 2% agar in 1% formalin and paraffinized for immunohistochemistry detection to serve as a positive control for lytic viral protein immunohistochemical (IHC) assays.

hNSG mice were infected with 2,750 GRUs of WT or Z-KO EBV suspended in PBS or mock treated with PBS alone. Twenty to 35 days postinfection, animals were sacrificed and splenocytes were harvested. The splenocytes were depleted of B cells by magnetic sorting using anti-CD19 beads (Miltenyi Biotech) and labeled with 2.5 μM carboxyfluorescein succinimidyl ester (CFSE; Molecular Probes, Eugene, OR). Autologous uninfected B cells, or lymphoblastoid cell lines (LCLs) made by infecting autologous B cells with WT EBV or Z-KO EBV, were γ-irradiated (7,500 rads) and cultured with the CFSE-labeled splenic cells at a ratio of 1:50. After 7 to 8 days of culture, the cells were fluorescently stained for expression of CD3 to identify T cells, treated with DAPI (4′,6-diamidino-2-phenylindole) to allow for the exclusion of dead cells, and analyzed by flow cytometry. The percentage of T cells that had undergone cell division was determined by gating on DAPI-negative CD3-positive cells and assessing the fraction that showed diminished CFSE fluorescence intensity.

Splenocytes from WT or Z-KO EBV-infected or mock-treated mice were magnetically depleted of B cells and labeled with CFSE as described above. The splenic cells were incubated in culture medium at 37°C at a 3:1 ratio with autologous uninfected B cells or LCLs made by infecting the B cells with WT EBV or Z-KO EBV. In parallel, uninfected B cells or WT or Z-KO LCLs were incubated alone (without splenic effector cells) to assess the amount of spontaneous cell killing. After 4 h, the cultures were stained with DAPI to identify dead or dying cells and analyzed by flow cytometry. Percent target cell killing was assessed for each culture by determining the fraction of the total CFSE-negative population that showed elevated DAPI staining. Specific killing was calculated by subtracting the amount of spontaneous cell death of target cells alone from the amount of cell death observed in the presence of splenic effector cells.

WT or Z-KO EBV-infected thymus-engrafted hNSG(thy) mice (four animals each) and two mock-infected hNSG(thy) mice were subcutaneously injected into the right hind footpad with 12.5 μg of commercially available EBV antigens (Meridian Life Science, Inc., Memphis, TN) in a total volume of 25 μl 1 month postinfection. To control for swelling caused by the injection itself, PBS alone was injected into the left hind footpad. The hind footpad thickness was measured 24 h postinjection by using a dial thickness gauge. The preinjection measurement was subtracted from the postinjection measurement to obtain specific swelling values. DTH reactivity is shown as the change in thickness of the hind footpad expressed in units of 10⁻⁴ inches.

To generate an in vivo model that would allow analysis of EBV pathogenesis and lymphoma formation in the context of a self-educated human immune system, we injected irradiated NSG mice intravenously (i.v.) with human fetal CD34⁺ cells (purified from fetal liver) and cotransplanted a small piece of the fetal thymus (with liver tissue) under the kidney capsule, as previously described (31). Reconstitution levels of human cells in the NSG mice were determined using multicolor flow cytometric analysis at 10 weeks posttransplantation. At the time point of EBV infection (10 to 12 weeks after engraftment), human-derived CD45 cells comprised around 35% of the total leukocyte population, and both T cells (including CD4- and CD8-positive cells) and B cells were present (Fig. 1). In addition, the engrafted human thymic tissue remained viable under the mouse kidney capsule for the duration of the experiments.

To directly test the role of lytic infection in the pathogenesis of EBV infection, we compared a lytically active control virus to a deletion mutant strain (Z-KO) lacking BZLF1, a transcription factor required for lytic viral gene expression (20, 32). The Z-KO virus contains a kanamycin resistance cassette (inserted between nucleotides 102389 to 103388 using B95.8 coordinates) that specifically disrupts BZLF1 function, and the phenotype of the mutant virus can be rescued in trans by the expression of the BZLF1 gene product (7, 8). Immune reconstituted mice were infected i.p. with equal titers of the control or Z-KO virus and sacrificed at various time points after infection or sooner if they showed signs of clinical illness. Multiple different organs were examined for the presence of EBV-infected cells using the EBER in situ hybridization assay. EBERs are expressed in all latently EBV-infected cells regardless of their latency type (20). While the number of EBER-positive cells was somewhat variable, every EBV-infected animal, regardless of whether infected by the control or the Z-KO virus, was found to have at least some EBV-infected (EBER-positive) cells at the time of autopsy, except for the animals sacrificed at day 3 postinfection. No EBER-positive cells were found in the tissues from uninfected hNSG(thy) animals (data not shown).

To determine if horizontal viral transmission is required for the ability of EBV to infect certain cell types or to reach certain sites in the body, we compared the sites of EBER-positive cells in tumor-free animals infected with the control or Z-KO virus. EBER-positive cells were most frequently observed in the liver, spleen, kidney, lung, and transplanted human thymic tissue (Table 1). In some animals, EBER-positive cells were found in the bone marrow, nasal lymphoid tissues, mesentery, and muscle tissue (Fig. 2 and data not shown). In the transplanted fetal thymus, EBV-positive cells were most commonly found in the medulla area, often near Hassall's corpuscles (Fig. 3A). In the spleen, EBV-positive cells were usually located within CD20-rich lymphoid zones (Fig. 3B). No clear differences in the sites and numbers of EBER-positive cells were observed in animals infected with the control virus or the Z-KO virus (Table 1). Although Z-KO-infected animals were somewhat more likely than control virus-infected animals to have EBV-infected B cells in the thymus and spleen, the differences were not statistically significant. Since the Z-KO virus is not horizontally transmitted, these results indicate that EBV-infected B cells home to many different organs in this model and that transmission of EBV from cell to cell via the lytic form of viral infection is not required for the establishment of long-term viral latency.

Although EBV is primarily found in B cells and epithelial cells in healthy humans, it can infect T cells and monocytes/macrophages in vitro (35) and is found in some T cell and NK cell lymphomas (17). To determine if EBV can infect cell types other than B cells in the hNSG(thy) mouse model, we costained multiple tissues using an anti-EBNA1 antibody and antibodies specific for human B cells (CD20), T cells (CD3), monocytes (CD14 and CD68), hematopoietic stem cells (CD34), and epithelial cells (anticytokeratin). All EBNA1-positive cells costained with CD20 in every organ, including the transplanted human thymus (Fig. 4A). Some EBNA1-positive cells also costained with anti-CD27, a marker for memory B cells (Fig. 4B). Thus, few if any T cells or monocytes are infected with either the control or Z-KO EBV in this model.

To define the type(s) of viral latency established by the control and Z-KO EBV, we performed an IHC assay in EBER-positive areas from multiple different types of organs in tumor-free animals, using anti-EBNA1, anti-EBNA2, and anti-LMP1 antibodies. As summarized in Table 2, type I and type IIB latency were the most common forms of EBV infection in both the control virus-infected and Z-KO-infected mice. As we were unable to perform EBER in situ hybridization assays on the same cells used to perform EBNA IHC assays (due to loss of EBNA1 IHC staining under the EBER staining conditions and vice versa), we could not determine how many cells had type 0 latency. However, since we generally found more EBER-positive cells than EBNA1-positive cells in the tumor-free animals (data not shown), cells with type 0 latency may have also been present. Some animals had a mixture of various different types of latency depending upon the location. An example of cells with type I latency is shown in Fig. 5A, and examples of cells with type IIB latency are shown in Fig. 5B to D. Latent membrane protein 1 (LMP1)-positive cells were rare and/or nonexistent in most tumor-free hNSG(thy) animals infected with either the control or Z-KO EBV. Rare LMP1⁺/EBNA2⁻ cells (type IIA latency) were found in the kidney of a control EBV-infected mouse (Fig. 6A). In tumor-free animals, cells with type III latency were observed most frequently at the day 20 time point (Fig. 6B and C). Thus, in animals that successfully limit their EBV infection, type III latency is transient, presumably because the host immune response eliminates cells with type III latency.

Although many mice had asymptomatic infection, some EBV-infected mice eventually developed B cell lymphomas in this model. Importantly, the lytic replication-defective Z-KO virus produced significantly fewer lymphomas than the control virus. Six of the 11 mice infected with the control virus developed lymphomas, versus only 2 of the 14 Z-KO virus-infected animals (P < 0.05) (Fig. 7A). All EBV-positive tumors were CD20⁺ DLBCL-like lymphomas. A number of the EBV-induced lymphomas in this model were small and discovered only after review of a series of slide sections obtained from multiple different organs. Interestingly, in contrast to the EBV-induced lymphomas found in humanized mouse models that do not include engrafted human thymic tissue, which have all been reported to have type III latency (4, 38, 43), we found that some of the lymphomas in this hNSG(thy) model had type I (Fig. 7B) and type IIB (Fig. 7C) latency, although others did have type III latency (Fig. 7D) (summarized in Table 3 ). No uninfected mice developed lymphomas (data not shown). These results suggest that one or more lytic viral proteins enhance the development of B cell lymphomas in the hNSG(thy) mouse model and show that this model provides for the development of EBV-positive lymphomas with type I and type IIB latency.

To examine the effect of lytic viral infection upon the level of EBV DNA in the plasma, we performed viral load assays in a subset of mock- and Z-KO and control virus-infected animals. Somewhat surprisingly, the only animal that had detectable viral DNA in the plasma was a control virus-infected animal with a large type III lymphoma (Table 4). Although it is possible that EBV plasma viral loads might have been higher if measured at earlier time points after control virus infection, these results suggest that the development of an effective anti-EBV T cell response in this model acts to eliminate cells with lytic viral infection.

We also performed an IHC assay using antibodies directed against three different classes of lytic viral proteins (the immediate-early BZLF1 protein, the early BMRF1 protein, and the late gp350/220 protein) in all animals infected with the control EBV strain. Whereas no lytically infected cells were observed by IHC assay in any tissues from tumor-free mice (including the one animal examined at 3 days postinfection), rare cells expressing the BZLF1 and BMRF1 proteins, but not the gp350/220 protein, were found within B cell lymphomas (Fig. 8A); as expected, no BZLF1- or BMRF1-positive cells were found in any of the tumors infected with the Z-KO virus (data not shown). All BZLF1-positive cells costained with CD20 (data not shown), indicating that they were B cells. Notably, the absence of late viral antigen staining in the lymphoma cells suggests that lytic infection is either abortive or that cells expressing late viral proteins are rapidly eliminated. Interestingly, EBV infection of mice that were reconstituted with human CD34⁺ stem cells in the absence of cotransplanted thymic tissue resulted in the development of tumors showing greater evidence of lytically infected cells (Fig. 8B). Thus, the self-educated T cells of the thymus-engrafted mice may contribute to the elimination of lytically infected cells.

EBV-positive cells in tumor-free animals were almost always surrounded by CD3⁺ T cells (Fig. 9A, suggesting that T cells actively interact with EBV-infected B cells prior to tumor development. To investigate this directly, we assessed the responses of primary T cells from infected and uninfected mice to EBV-infected or uninfected B cells. Splenocytes from the control or Z-KO virus-infected or mock-infected mice were depleted of B cells and then labeled with CFSE and cultured with uninfected B cells or B cells that were infected in vitro with the control or Z-KO EBV (i.e., LCLs). After 7 to 9 days, the cultures were stained with anti-CD3 and DAPI and analyzed by flow cytometry for the percentage of live T cells that showed diminished CFSE fluorescence (indicating that cell division had taken place). As shown in Fig. 9B, T cells obtained from mice infected with either the Z-KO virus or the control virus showed significantly higher proliferation when exposed to EBV-infected B cells than when exposed to uninfected B cells. Additionally, T cells derived from mock-infected animals had minimal proliferative responses to the EBV-infected LCLs (Fig. 9B). Thus, primary T cells from EBV-infected mice proliferated specifically in response to EBV-infected LCLs, whereas T cells from uninfected mice did not.

To further investigate the potential for immune-mediated control of EBV-infected cells in this model, cell killing assays were performed using splenic effector cells (e.g., T cells and NK cells) and uninfected B cells or LCLs infected with the control or Z-KO virus as target cells. As shown in Fig. 9C, in a 4-h assay, splenic effector cells from a mock-infected animal showed no specific killing of EBV-infected B cells (either control virus or Z-KO virus infected). In contrast, splenocytes from each of the two Z-KO virus-infected animals examined showed cytotoxic effector activity to both control virus- and Z-KO virus-infected LCLs, as did splenocytes from one of the two control virus-infected animals examined. There was no specific killing of uninfected B cells by effector cells from any of the animals, confirming that the cytotoxic response was specific to infected LCLs (Fig. 9C). These results suggest that in this model, primary effector cells from EBV-infected animals can produce a functional killing response against EBV-infected B cells.

As another measure of the host immune response against EBV, we examined in vivo delayed-type hypersensitivity (DTH) responses to an antigen preparation from EBV-infected cells. DTH responses are localized inflammatory responses initiated by the activation of antigen-specific lymphocytes that result in multicellular infiltration and edema. Mice that were either mock infected or infected with the control or Z-KO virus were inoculated in the footpads with a preparation of protein antigens from a lytically induced EBV-positive cell line or PBS, and the amount of swelling in the footpads was measured 24 h later (Fig. 9D). Both control EBV- and Z-KO virus-infected animals had significantly more swelling in response to the EBV antigen than the mock-infected animal. The DTH responses were not significantly different between control EBV- and Z-KO EBV-infected animals. This assay provides powerful confirmation that mice infected with either the control or ZKO virus strain develop in vivo immune responses against proteins from EBV-infected cells.

Finally, we stained tumors with anti-CD3 antibody (which recognizes all T cells), as well as anti-CD8 and anti-CD4 antibodies. All of the EBV-positive lymphomas, including those with type III latency (in which the immune response is obviously inadequate), were infiltrated with both CD4⁺ and CD8⁺ T cells in this model (Fig. 10A). Interestingly, the CD4 cell infiltration of tumors with type IIB latency appeared particularly robust (Fig. 10B). Levels of T cell infiltration of the tumors containing the control EBV versus the Z-KO virus were not obviously different. These results indicate that host T cells interact with various different types of EBV-positive tumors in this model, even when the tumors are not effectively eliminated.