Cotargeting EBV lytic as well as latent cycle antigens increases T-cell potency against lymphoma

Although effective, autologous T-cell immunotherapies are often hindered by their narrow target antigen-specific repertoire. For example, the success of T cells expressing chimeric antigen receptors (CARs)^1,2 or transgenic T-cell receptors^3,4 is limited by their recognition of single epitopes on single tumor antigens, introducing the risk of immune evasion.^4-6 Given the extensive heterogeneity of tumor antigen expression, it may be beneficial to manufacture therapeutic products with a broader array of antigen/epitope specificities.

Epstein-Barr virus (EBV) is present in a range of malignancies, all associated with the viral latent cycle.^7,8 The overall incidence of EBV-positive (EBV⁺) cases of lymphoma is ∼30% to 40%^9,10 but, within lymphoma, the incidence varies widely, not only with the type of lymphoma but also with geographical region and patient age.¹¹ EBV⁺ lymphomas express viral latency proteins that provide unique and safe targets for EBV-specific T cells (EBVSTs) that have been evaluated as therapy in multiple clinical trials.^12-15 EBVSTs have proved safe and highly effective for the treatment of the EBV-associated lymphomas occurring after allogeneic transplantation,^14,16 but the tumors that develop in immunocompetent individuals are much less immunogenic, expressing only 4 of the 9 viral latent cycle proteins expressed in lymphomas occurring after allogeneic transplantation.^9,10 These 4 viral proteins, EBV nuclear antigen (EBNA)-1, latent membrane protein (LMP)-1, LMP-2, and BamHI A rightward reading frame (BARF), are known as type 2 (T2) latency antigens.^9,10 T2 latency tumors are also more immunosuppressive than lymphomas occurring after allogeneic transplantation, and EBVSTs targeting T2 latency proteins are difficult to reactivate from patients with lymphoma.^17,18 

Since the first clinical studies, EBVST manufacturing strategies have been shortened by changing the antigen-presenting cells (APC) from autologous EBV-transformed B-cell lines (LCLs) that take ∼6 weeks to establish,^14,16 to peptide library (pepmix)-pulsed autologous peripheral blood mononuclear cells (PBMCs) or irradiated pepmix-pulsed autologous activated T-cells (ATCs) combined with HLA-negative costimulatory cells.¹⁹ This modification extended the applicability of EBVSTs to patients with rapidly progressive disease and those lacking sufficient B cells to establish an autologous LCL.²⁰ However, in our trial of autologous, pepmix-activated, T2-antigen–targeted EBVSTs (T2-EBVSTs) for the treatment of EBV⁺ lymphoma, we were unable to generate EBVSTs from approximately one-third of patients,¹⁸ and the frequency of T2-antigen–specific T cells in many of the EBVSTs that passed the potency release criteria was low, as was their ability to recognize all 4 T2 antigens. Although pepmix-activated EBVSTs demonstrated increased recognition of T2 antigens compared with EBVSTs activated using autologous dendritic cells and LCLs overexpressing LMP1 and LMP2 from an adenovirus vector, the results of the clinical trial of T2-antigen–stimulated EBVSTs were not as impressive.¹² LCL-activated EBVSTs recognize a wide range of viral proteins, including viral lytic cycle antigens expressed by 1% to 5% of LCLs.^21-24 This raised the question of whether T cells specific for lytic cycle proteins played an important role in tumor elimination.

Lytic cycle transcripts have been detected in several EBV-associated malignancies, including gastric cancer,²⁵ nasopharyngeal carcinoma,^26,27 Burkitt's lymphoma,²⁸ and peripheral T-cell lymphoma²⁹ but have not been shown in EBV-associated Hodgkin lymphoma (HL). Lytic cycle gene expression is tightly regulated but may be driven by cellular stress, such as produced by hypoxia, acidity, and nutrient deficiency, all common in the tumor microenvironment (TME), and several lytic cycle proteins have oncogenic activity and could contribute to tumor growth.^30-33 Other observations support targeting both latent and lytic cycle antigens including: (1) the high immunogenicity of lytic cycle antigens,^34,35 (2) the heterogeneity of T2 antigen expression in malignancies,^36,37 and (3) recognition of additional target antigens should increase the production of type 1 T helper (TH1) cell cytokines and chemokines at the tumor site, reactivating the local environment and increasing epitope spreading.³⁸ 

Here, we first determined whether lytic cycle gene expression could be detected in HL biopsies and then whether we could reactivate and expand T-cells for both lytic and latent cycle antigens (broad repertoire [BR]-EBVSTs) in a single culture, first from EBV seropositive healthy donors, and then from patients with HL. Finally, we examined the antitumor efficacy of BR-EBVSTs in a murine model of lymphoma. BR-EBVSTs have now been incorporated into clinical trials targeting refractory/relapsed EBV⁺ lymphoma (www.clinicaltrials.gov identifiers: #NCT01555892 and #NCT04664179).

We obtained blood samples from EBV-seropositive healthy volunteers and patients with EBV⁺ lymphoma with informed consent in accordance with the Declaration of Helsinki on Baylor College of Medicine institutional review board–approved protocols (H7634, H7666, and H29617). We isolated PBMCs using Lymphoprep gradients (Axis Shield, Oslo, Norway) and used them to generate EBVSTs, and ATCs.

PBMCs were stimulated with CD3 (OKT3 hybridoma cell line American Type Culture collection no. CRL 8001, Manassas, VA) and CD28 (Becton Dickinson BD, Franklin Lakes, NJ) antibodies and expanded in interleukin-2 (IL-2; National Institutes of Health, Bethesda, MA), as previously described.^18,20 Before use as APCs, ATCs were restimulated with CD3/CD28 antibodies to upregulate costimulatory molecules, pulsed with a mastermix of EBV T2 and or BR antigen pepmixes and irradiated at 30 Gray using an RS2000 X-ray irradiator (RadSource, Suwanee, GA).²⁰ 

Overlapping peptide libraries (15mers overlapping by 11 amino acids) spanning the complete protein sequences of EBV T2 antigens (EBNA-1, LMP-1, and LMP-2, BARF-1) and EBV lytic antigens BZLF-1, BRLF-1, and BMRF-1 (JPT technologies, Berlin, Germany) and BMLF-1, BMRF-2, BALF-2, BNLF-2a, and BNLF-2b (Genemed Synthesis, San Antonio, TX). Lyophilized pepmixes were reconstituted at 200 μg/mL in dimethyl sulfoxide (Sigma-Aldrich, St. Louis, MO) and stored at −80°C.

The K562 costimulatory cell line (K562cs) was derived from the HLA class 1– and 2–negative chronic erythroid leukemia cell line, by genetic modification with CD80, CD83, CD86, and 4-1BB ligand, and was a gift from Carl June (University of Pennsylvania, Perelman School of Medicine). LCLs were generated from healthy donor PBMCs by infection with concentrated virus from the B95-8 strain of EBV in the presence of 1 μg/mL of cyclosporin A. In some experiments, K562cs was replaced with the ULCL, an HLA-negative LCL established in our Center using CRISPR–-Cas 9 gene editing.¹⁸ 

PBMCs were pulsed with the T2-EBV or lytic pepmixes cocktail or a BR-pepmix cocktail (including T2-EBV and lytic pepmixes) and expanded in the presence of IL-15 and IL-7 (R&D Systems, Minneapolis, MN). On day 9 after stimulation (S1), the cells received a second stimulation (S2) with irradiated pepmix-pulsed autologous ATCs and irradiated K562cs or ULCLs at an EBVST:ATCs:K562cs/ULCL ratio of 1:1:5 with IL-7 and IL-15.^18,20 Cells were split as needed during culture and analyzed for phenotype, specificity, and function on day 16, unless otherwise indicated. Although EBVST lines may not have been 100% EBV specific, we refer to them as EBVSTs.¹⁸ In our clinical trial, the potency release criterion of 20 spot-forming cells (SFCs) per 100 000 cells appears low, but the enzyme-linked immunospot (ELISpot) assay underestimates the true frequency of antigen-specific T cells by at least 1 log,^39,40 hence, this would translate to ∼200 000 antigen-specific T cells in an infusion dose of 1 × 10⁸ EBVSTs.

T-cells were phenotyped using CD3, CD4, CD8, CD45RA, CD45RO, CCR7, and CD62L antibodies (BioLegend, San Diego, CA; and BD Biosciences, Franklin Lakes, NJ). Stained cells were acquired using the Gallios Flow Cytometer or the BD FACSCanto II, and results were analyzed using Kaluza or FlowJo analysis software (FlowJo, LLC, Ashland, OR).

Virus-specific T cells (VSTs, 2 × 10⁵ cells) were plated in 96-well plates and stimulated with 100 ng of EBV pepmix or control pepmix in the presence of CD28 and CD49d (1 μg/mL) at 37°C for an hour after which brefeldin A and Monensin (BioLegend) were added. After a 16-hour incubation and surface staining with CD3, CD4, and CD8 antibodies, cells were stained with the fixable live-dead stain eBioscience Fixable Viability Dye efluor 780 (Thermo Fisher, Waltham, MA), fixed and permeabilized, followed by staining with interferon gamma (IFN-γ) and tumor necrosis factor α (TNF-α) antibodies (Becton Dickinson). Cells were then analyzed via flow cytometry.

The frequency of EBV antigen–specific T-cells within the EBVST population was measured via IFN-γ ELISpot assays that measured the number of IFN-γ -secreting cells (SFCs) per 10⁵ cells. EBVSTs were plated in duplicate at 1 × 10⁵ per well coated with anti-human IFN-γ monoclonal antibody 1-D1K (Mabtech, Cincinnati, OH) and stimulated with 100 ng pepmixes or medium alone. After the 16- to 18-hour assay, we quantified the frequency of antigen-responsive cells. The number of SFCs was reported after the subtraction of negative control values. The quantification or analysis of SFCs was done either at Zellnet Consulting (Fort Lee, NJ) or in the laboratory using Mabtech IRIS ELISpot reader (Mabtech). Although the ELISpot assay underestimates the true frequency of specific cells, it is the simplest HLA and epitope agnostic method available for analyzing the specificity of multiantigen-specific T cells.³⁹ 

The cytolytic specificity of VSTs was evaluated in a standard 4-hour ⁵¹Cr chromium release assay. ⁵¹Cr sodium chromate–labeled autologous ATCs alone or pulsed with pepmixes were used as target cells for EBVSTs at an effector-to-target ratio of 40:1, 20:1, 10:1, and 5:1. Target cells were cultured in EBVST medium or 1% Triton X-100 (Sigma-Aldrich) to achieve spontaneous and maximum release respectively. After 4 to 6 hours of coculture, the supernatant was harvested and ⁵¹Cr was measured using a γ counter. The percent specific lysis was calculated from the mean of triplicates as ([experimental release − spontaneous release]/[maximum release − spontaneous release]) × 100.

We subcutaneously engrafted 4- to 6-week-old NOD.Cg-Prkdcscid Il2rgtm1Wjl/SzJ (NSG) female mice (The Jackson Laboratory, Bar Harbor, ME) with LCLs suspended in Matrigel (Corning, Tewksbury, MA). Tumor cells were rendered bioluminescent using a retrovirus vector expressing green fluorescent protein–firefly luciferase. Tumor volume was determined using an external caliper (Thermo Fisher, Waltham, MA) by measuring the greatest longitudinal diameter (length) and the greatest transverse diameter (width) and calculated by the formula: tumor volume = ½ (length) × (width).² Once the tumor was established, we injected 1 × 10⁶ autologous (to the LCL donor) EBVSTs intravenously. Tumor bioluminescence was monitored using an IVIS Imaging system (Caliper Life Sciences, Hopkinton, MA) at the indicated time points. Living Image software (PerkinElmer, Waltham, MA) was used to visualize and calculate total luminescence covering the region of interest drawn covering the tumor area.

For analysis of human cytokines in mouse serum on days 3 and 8 after EBVST injection, blood was collected from the submandibular vein into EDTA tubes, then spun down to collect plasma, which was then frozen in 30-μL aliquots at −80°C. Controls included mice without tumors and/or bearing tumors but no EBVSTs. All procedures were done in compliance with Institutional Animal Care and Use Committee approval at Baylor College of Medicine approved protocol number AN5551. Human cytokine concentration in murine plasma was evaluated using a MILLIPLEX MAP Human Cytokine/Chemokine Magnetic Bead Panel Immunology Multiplex assay (Millipore Sigma, Darmstadt, Germany) and analyzed using a Luminex instrument and software.

RNA was isolated from 6 EBV⁺ and 2 EBV-negative HL biopsy samples using a Picopure RNA isolation kit (Thermo Fisher, Waltham MA). Poly-A messenger RNA was selected using an Oligotex mRNA mini kit (Qiagen, Hilden, Germany). Complementary DNA was generated using a High Capacity cDNA Reverse Transcription kit (Thermo Fisher) followed by RNase H digestion, which was then sent to the University of North Carolina at Chapel Hill Vironomics core for comprehensive whole EBV gene profiling by quantitative polymerase chain reaction using an automated system with 288 primers covering all EBV latent and lytic cycle gene transcripts and an integrated LightCycler 480 instrument.⁴¹ The Namalwa cell line was used as a positive control. Glyceraldehyde-3-phosphate dehydrogenase, actin, and beta2 microglobulin (B2M) housekeeping genes were used as standards. The list of primers used is available in supplemental Table 1.

We used GraphPad Prism 7 (GraphPad Software, Inc, La Jolla, CA) for data visualization and statistical analysis using tests as indicated in figure legends. Data are plotted as mean ± standard error of the mean, unless otherwise indicated. Significance is denoted by P < .05 (ns, nonsignificant P = 0.12, ∗P = .033, ∗∗P = .002, and ∗∗∗P < .001, unless otherwise indicated).

All animal procedures were done in compliance with Institutional Animal Care and Use Committee approval at Baylor College of Medicine approved protocol number AN5551.

EBVSTs were generated from patients with EBV⁺ lymphoma enrolled in our clinical trial (www.clinicaltrials.gov identifier: #NCT01555892) by pulsing patient PBMCs with the EBV T2 antigen pepmixes, EBNA-1, LMP-1, LMP-2, and BARF-1. Of 72 EBVST lines initiated, 27% failed the manufacturing release criterion of >20 IFN-γ SFCs per 10⁵ cells in response to pepmix stimulation, as measured in ELISpot assays, and/or failure to grow, as reported in our recent publication.¹⁸ An additional 42% of lines contained ≤100 IFN-γ–secreting cells (SFCs) per 10⁵ cells. Only 4.3% of lines recognized all 4 EBV T2 antigens with >20 IFN-γ SFCs per antigen per 10⁵ cells, whereas 67% recognized only 1 or 2 antigens (Figure 1A). This observation highlighted the narrow target antigen specificity of patient EBVSTs and a need to broaden the target antigen repertoire to achieve clinically potent EBVSTs for these tumors that heterogeneously express the viral target antigens.³⁷ 

To determine whether lytic cycle antigens could be clinically useful target antigens for EBV T2 lymphomas, we evaluated the presence of lytic cycle transcripts in EBV⁺ HL biopsy samples. We detected transcripts from both EBV latent (EBNA-1 and LMP-1) and lytic genes including BFRF1, BORF2, BALF4, BNLF2a, and BMLF1 (Figure 2A) in EBV⁺ but not in EBV-negative HL biopsies. EBNA2, which is expressed only in type 3 latency tumors, was detected in only 1 HL sample and the Namalwa positive control. Although lytic cycle transcripts do not guarantee the expression of the encoded protein,^42-44 we detected the EBV early lytic antigens BamHI Z replication activator, encoded by the immediate early transcript BZLF1, and EA-D encoded by the early BMRF1 transcript in EBV⁺ HL biopsy samples (supplemental Figure 1A).

To determine whether T cells specific for lytic cycle antigens produced in response to EBVSTs activated using Ad-LMP1/2-transduced LCLs might have contributed to the better clinical outcome in our previous clinical trial,¹² we compared the specificity of EBVSTs generated by stimulating PBMCs with LCLs or T2 pepmixes. LCL-activated EBVSTs showed a high frequency of T cells specific to lytic cycle antigens (supplemental Figure 2A), although, notably, we sometimes detected a usually lower frequency of lytic antigen–specific T cells in T2-EBVSTs, likely because of their high initial frequency in PBMCs.

To determine whether we could generate EBVSTs with specificity for both lytic and latent antigens in a single culture, we first compared the proliferation, antigen specificity, cytolytic activity, and phenotypic characteristics of EBVSTs generated from PBMCs of healthy EBV-seropositive donors by stimulation with T2-, lytic- and BR-pepmixes individually. PBMCs were stimulated with the 4 EBV T2 antigen pepmixes and 8 lytic antigen pepmixes alone or combined in the presence of IL-7 and IL-15 to generate T2-EBVSTs, lytic-EBVSTs, and BR-EBVSTs, respectively (Figure 3A). On day 9, the cells were restimulated with an irradiated antigen-presenting complex comprising pepmix-pulsed autologous ATCs as APCs and an HLA-negative costimulatory cell line, then harvested on day 16 for characterization. BR-EBVSTs demonstrated a higher total fold expansion than T2- or lytic-EBVSTs (253-fold ± 30-fold vs 161-fold ± 28-fold vs 178-fold ± 25-fold; mean ± standard error of the mean; BR- vs T2- vs lytic-EBVSTs) over the 16 days of culture (Figure 3B). T2- and lytic-EBVSTs contained a higher frequency of CD3⁻ CD56⁺ natural killer cell compared with BR-EBVSTs (Figure 3C). Lytic-EBVSTs had a decreased frequency of CD45RO⁺ CD62L⁺ central-memory T cells (T_CM) compared with T2-EBVSTs (36.9% ± 5.1% vs 45.34% ± 3.7%; P = .023, lytic- vs T2-EBVSTs; Figure 3D) and a higher frequency of CD45RO⁺ CD62L⁻ effector memory T cells (56.1% ± 5.2% vs 49.1% ± 3.8%; P = .039, BR- vs T2-EBVSTs; Figure 3E). These phenotypic differences were reflected in BR-EBSVTs that showed a decreased frequency of T_CM in (31.3% ± 2.7% vs 45.34% ± 3.7%; P = .016, BR- vs T2-EBVSTs; Figure 3D-E).

When we evaluated antigen specificity, T2-EBVSTs were specific for T2-latent antigens, lytic-EBVSTs were specific for lytic antigens, and BR-EBVSTs were specific for both, suggesting the successful expansion of BR-EBVSTs recognizing both latent and lytic antigens in a single culture (Figure 3F), although the results are donor dependent. Notably, BR-EBVSTs contained a higher frequency of latent cycle antigen-specific T cells than T2-EBVSTs in 4 of 7 donors, allaying fears that dominant lytic cycle antigen–specific T cells would overwhelm T2-antigen–specific T cells. The same was true for the frequency of lytic antigen–specific T cells in BR-EBVSTs compared with lytic EBVSTs (Figure 3F). Overall, this resulted in a trend of increased killing of autologous ATCs pulsed with lytic and latent pepmix by BR-EBVSTs compared with lytic- or T2-EBVSTs at an effector-to-target ratio of 20:1 in a 4-hour ⁵¹Cr release assay (Figure 3G).

The increased frequency of antigen-specific T cells in BR-EBVSTs was supported by tetramer analysis of HLA A:24:02–restricted T cells specific for LMP2 (latent), and BRLF1 and BMLF1 (lytic) antigens in both donors evaluated. We detected 0.48% compared with 1.20% and 1.94% compared with 3.95% tetramer-positive cells in T2-EBSVTs compared with BR-EBVSTs in donors 1 and 2, respectively (supplemental Figure 3A). BR-EBVSTs also showed an increased frequency of IFN-γ– and TNF-α–producing polyfunctional T cells compared with T2-EBVSTs when stimulated with T2 antigens as measured in intracellular cytokine secretion assays (20.52% ± 4.7% vs 14.18% ± 4.1% IFN-γ⁺ TNF-α⁺ CD3⁺ T cells, BR- vs T2-EBVSTs; supplemental Figure 3B).

Our findings demonstrate the successful generation and antigen-specific functionality of BR-EBVSTs and confirm the feasibility of generating EBVSTs from pooled T2 latent and lytic pepmixes without compromising the overall proliferation, antigen specificity, polyfunctionality, or cytolytic activity of the resulting BR-EBVSTs.

To ensure that we could generate BR-EBVSTs from patients with EBV⁺ lymphoma, we compared BR- and T2-EBVSTs from 8 patients. BR-EBVSTs showed increased expansion from day 0 to day 16 (656-fold ± 170.4-fold expansion on day 16 vs 390.3 ± 91.64; BR- compared with T2-EBVSTs; Figure 4A). As observed in healthy donors (Figure 3), T cells specific for both latent and lytic antigens expanded in BR-EBVSTs from 7 of 8 patients (only 4 T2-antigen–specific SFCs detected in BR-EBVSTs for patient 2), although the specificity for latent antigens was decreased in BR-EBVSTs for patients 2 and 6 compared with T2-EBVSTs (26 vs 4 SFCs in patient 2, and 1139 vs 688 SFCs in patient 6, T2- vs BR-EBVSTs; Figure 4B). However, in patients 1, 3, 5, 7, and 8, T2-latent antigen specificity was increased in BR-EBVSTs compared with T2-EBVSTs (Figure 4B). Patient BR-EBVSTs showed antigen-specific cytotoxicity as demonstrated by the killing of EBV pepmix–pulsed autologous ATCs in a 4-hour ⁵¹Cr release assays and did not kill autologous, EBV-antigen–negative target cells (Figure 4C). These observations demonstrated the feasibility of targeting both latent and lytic antigens from most patients with EBV⁺ lymphoma.

To compare the antitumor activity of T2-, lytic- and BR-EBVSTs, we transferred EBVSTs from 2 donors into NSG mice bearing autologous LCL tumors in which 1.32% and 1.46% expressed the early lytic cycle proteins encoded by BZLF1 (BamHI Z replication activator) and BRLF1 (supplemental Figure 4A-B). Ten days after subcutaneous tumor inoculation, we adoptively transferred autologous T2-, lytic- or BR-EBVSTs and measured antitumor efficacy and the presence of human cytokines in serum (Figure 5). Tumors responded more rapidly in mice receiving BR-EBVSTs, demonstrating decreased tumor bioluminescence by day 25 compared with mice receiving control, T2-, or lytic-EBVSTs (Figure 5A-B). This corresponded with a smaller tumor volume of 478.7 ± 35.40 mm³ by day 17 in mice receiving BR-EBVSTs, compared with 1446.8 ± 40.71 mm³ (P < .001) and 1096.8 ± 115.73 mm³ (P < .001) in those receiving T2-EBVSTs and lytic-EBVSTs, respectively (Figure 5C). Although tumors eventually regressed completely in all EBVST-receiving mice, we observed tumor bioluminescence in the upper-limb regions of groups receiving T2- and lytic-EBVSTs on days 11 and 25 after T-cell injections (Figure 5A,D). This metastatic spread was not observed in mice receiving BR-EBVSTs.

Analysis of human cytokines in serum isolated from murine blood on days 3 and 8 after T-cell injections revealed higher concentrations of granulocyte-macrophage colony-stimulating factor (GM-CSF), and IFN-γ (Figure 5E-F) but lower levels of IL-10 and TNF-α (Figure 5G-H) in mice receiving BR-EBVSTs compared with those receiving T2-EBVSTs. The group receiving lytic-EBVSTs demonstrated a cytokine secretion profile similar to that of BR-EBVSTs although at lower concentrations (Figure 5E-H).

BR-EBVSTs also demonstrated more rapid tumor control than T2-EBVSTs in a second donor (supplemental Figure 5A-C) and produced higher levels of GM-CSF, granzyme-B, perforin, IFN-γ, and soluble 41BB (supplemental Figure 5D). These observations demonstrate the enhanced in vivo activity and enriched proinflammatory TH1 cytokine secretion of BR-EBVSTs compared with T2-EBVSTs.

EBV⁺ lymphoma cells demonstrate a heterogeneous pattern of viral antigen expression, leading to a risk for immune evasion that is compounded by therapeutic T cells with a narrow target antigen repertoire. This work describes our identification of lytic cycle transcripts in HL and our strategy to broaden the target antigen repertoire of EBVSTs. We sought and identified lytic cycle transcripts in EBV⁺ HL biopsies, suggesting that lytic cycle antigens could be targeted together with latent-cycle antigens. BR-EBVSTs generated from both patient and healthy donor PBMCs showed higher fold expansion, maintained specificity to T2 antigens, and contained a higher frequency of polyfunctional T cells than T2-EBVSTs. In our autologous EBV⁺ lymphoma xenograft murine model, mice receiving BR-EBVSTs showed more rapid tumor control, increased concentrations of human proinflammatory cytokines, and decreased concentrations of TH2 cytokines. These findings supported the implementation of BR-EBVSTs for treating EBV⁺ lymphoma.

Although we demonstrated the expression of lytic cycle transcripts and antigens in bulk HL biopsies, this does not confirm the expression of lytic cycle antigens in the malignant Reed-Sternberg cells that comprise only ∼1% of all the cells within the tumor environment, which is often also populated by normal B cells. However, we detected EBNA2, a type 3 latency antigen, in only 1 of 8 biopsies, indicating that the EBV-infected cells detected were likely tumor cells. However, even if expressed in normal EBV-infected B cells, lytic cycle antigen stimulation of lytic EBVSTs would induce cytokine and chemokine production, resulting in reactivation of the tumor environment, and leading to epitope spreading.

EBV lytic cycle proteins are generally more abundant than latent cycle proteins,²⁸ and lytic cycle antigens dominate the hierarchy of T-cell specificity.^22,34 Thus, targeting lytic cycle antigens could enhance not only killing but also cytokine secretion in the immunosuppressive tumor milieu, resulting in the reactivation of endogenous tumor-reactive T cells. Even if the frequency of tumor cells expressing lytic antigens is low or the lytic cycle is abortive,^29,30,46 antigens released by dying tumor cells can be taken up and crosspresented by neighboring tumor cells, an advantage especially pertinent to B-cell lymphoma in which cells antigen crosspresentation is favored. This interpretation was supported by the higher levels of human TH1 cytokines such as IFN-γ, GM-CSF, 41BB, perforin, and granzyme B, and more rapid tumor control by BR-EBVSTs. These actions would favor antigen spreading in tumors containing tumor-infiltrating lymphocytes and reverse the immunosuppressive phenotype of myeloid cells and regulatory T cells in the TME.^47-49 

Although HLA antigens are often downregulated in lymphoma cells, the increased production of IFN-γ produced by BR-EBVSTs should maximize HLA upregulation,^38,50-52 whereas GM-CSF, if present in the right cytokine milieu, supports the TH1 polarization of myeloid cells.^53,54 Elevated levels of cytokine secretion by BR-EBVSTs may be attributed to their more differentiated state and. hence, increased effector functions compared with T2 antigen-specific T cells. Our data demonstrate that lytic EBVSTs reside predominantly within the effector memory T-cell subset, whereas more latent EBVSTs are found in the T_CM fraction. This finding can be explained by the reported differentiation state of circulating EBVSTs, with latent antigen-specific T cells being in the central memory pool, and lytic EBVSTs more commonly having an effector memory phenotype, reflecting persistent virus replication in the oropharynx.^55-57 Concerns that enhanced cytokine secretion by BR-EBVSTs might produce cytokine release syndrome similar to that observed after infusions of T cells modified with CARs^58,59 are diminished by our observation that VSTs have not produced cytokine release syndrome even while undergoing extensive in vivo expansion in patients with bulky tumors.^12,16,60,61 Although we do not address the role of higher concentration of proinflammatory TNF-α in mice receiving T2-EBVSTs, studies have implicated high levels of IL-10 and TNF-α as an adverse prognostic factor in EBV-related malignancies,^62-64 and these cytokines may derive from the more rapidly growing LCL tumors in our murine model.

We observed an increased frequency of T2-antigen–specific T cells in BR-EBVSTs compared with T2-EBVSTs in 6 of 8 patients and most healthy donors. This could be explained by the increased production of immunostimulatory cytokines by lytic-antigen–specific T cells during culture initiation, which supported the reactivation and expansion of low-frequency T cells rendered anergic in the TME.^34,57,65 We speculate that T2-EBVSTs supported by the proinflammatory cytokine milieu created by the lytic-EBVSTs in BR-EBVSTs will be able to expand and kill latent antigen-expressing tumors in patients. It was surprising that lytic-EBVSTs were able to eliminate tumors in our mouse model, because only a low frequency of cells expressed EBV lytic cycle early proteins. This may be explained by the crosspresentation of lytic cycle antigens by latently infected tumor cells, or that low numbers of T2-antigen–specific T cells were present in the infused EBVSTs. If needed, we can optimize the generation of BR-EBSVTs to facilitate the specific selection of T2-latent–specific T cells in BR-EBVSTs by excluding lytic pepmixes in the antigen-presenting complex during the second stimulation, thus favoring the expansion of T2-EBVST–specific T cells.

In the future, the clinical potency of BR-EBVSTs could be enhanced by combining BR-EBVSTs with agents that induce the EBV lytic cycle.^66,67 Recent studies have explored the induction of EBV lytic replication with or without antiviral drugs like ganciclovir for the treatment of EBV-associated tumors. Perrine et al, used a histone deacetylase inhibitor in combination with ganciclovir and demonstrated significant tumor responses without significant adverse toxicities in 10 of 15 patients with highly refractory EBV⁺ malignancies.⁶⁸ Histone deacetylase inhibitor–mediated EBV reactivation should boost adoptively transferred BR-EBVSTs, and their antitumor activity by providing a vaccine effect.

In summary, BR-EBVSTs targeting both T2-latent and -lytic antigens broaden the target antigen repertoire of EBVSTs and provide a clinically feasible platform for targeting EBV-associated malignancies. BR-EBVSTs derived from memory-enriched PBMCs have now been incorporated into 4 phase 1/2 clinical trials; autologous EBVSTs to treat EBV⁺ relapsed or refractory lymphoma (www.clinicaltrials.gov identifiers: #NCT01555892 and #NCT04664179); allogeneic, “off-the-shelf” CD30.CAR-EBVSTs in patients with relapsed or refractory CD30⁺ lymphomas (www.clinicaltrials.gov identifier: #NCT04288726); and as part of multivirus-specific T cells for the treatment of virus infections after hematopoietic stem cell transplant (www.clinicaltrials.gov identifier: #NCT04013802). If effective for lymphoma, BR-EBVSTs could be explored for treating more intractable EBV-associated malignancies, such as nasopharyngeal carcinoma and gastric cancer, in which the coexistence of EBV latent and lytic phase^25,28 has been reported.

The authors acknowledge the members of the good manufacturing practices and good laboratory practices groups at the Center for Cell and Gene Therapy, Baylor College of Medicine, Houston, TX, for their help with organizing patient samples used for this study. The authors also extend their gratitude to Carl E. Allen and Lymphoma Tissue Bank, Lymphoma and Histiocytosis Programs, Baylor College of Medicine, and Texas Children’s Hospital for providing HL biopsies.

This work was supported by CPRIT RP160283 Baylor College of Medicine Comprehensive Cancer Training Program, National Institutes of Health (NIH)-National Cancer Institute (NCI) grant P50 CA126752, NIH-National Heart, Lung, and Blood Institute (HHSN268201600015I), American Society of Gene and Cell Therapy Career Development Award 2019, Alex’s Lemonade Foundation Reach Award, A SCOR 7001-19 from the Leukemia and Lymphoma Society, and a sponsored reseach agreement from Tessa Therapeutics, Singapore, and NIH-NCI grant P01-CA019014 (D.P.D.).

Contribution: S.S. and C.M.R. conceptualized the project, designed the research and experimental methodologies, secured funding, and wrote the original draft of the manuscript; S.S. conducted the experimental investigation and designs, performed the majority of in vitro and in vivo studies, and conducted the data analysis and generated the figures; N.U.M. assisted with the in vitro studies; N.U.M., T.S., and L.A.R. assisted with the in vivo studies; D.P.D., at the University of North Carolina Vironomics Core, conducted the comprehensive EBV gene profiling assay and data analysis; and S.S., N.U.M., T.S., L.A.R., D.P.D., and C.M.R. contributed to reviewing and editing the manuscript.

Conflict-of-interest disclosure: C.M.R. has equity in Allovir and Marker Therapeutics; has served on advisory boards for Tessa Therapeutics and Marker; received research support from Tessa Therapeutics; and declares conflicts with Abintus, Allogene, Memgen, Turnstone Biologics, Coya Therapeutics, TScan Therapeutics, Onkimmune, and Poseida Therapeutics. The remaining authors declare no competing financial interests.

Correspondence: Cliona M. Rooney, Center for Cell and Gene Therapy, Baylor College of Medicine, Texas Children’s Hospital, 1102 Bates Ave, Feigin Tower, FC1770, Houston, TX 77030; email: crooney@bcm.edu.