A multidimensional workflow profiling of allogeneic virus-specific T cell therapies reveals potency-linked signatures

Abstract

Allogeneic virus-specific T cell (VST) therapies offer distinct advantages, including scalability, rapid deployment, and manufacturing consistency, and have demonstrated efficacy in multiple clinical trials. However, identifying VST products with high therapeutic potential remains a major hurdle. Here, we present a multidimensional analytical platform that integrates in vitro and in vivo anti-viral reactivity, T cell receptor (TCR) repertoire analysis, gene expression profiling, immunophenotyping, and functional validation in a humanized mouse model. Epstein-Barr virus (EBV)-specific T cells expanded from HLA-diverse healthy donors consistently enriched for TCRs targeting EBV-encoded antigens. Transcriptomic and high-dimensional flow cytometric analyses revealed a distinct effector-associated signature. Importantly, this integrative approach uncovered correlative biomarkers of T cell potency and effector function, validated in an in vivo model of EBV-driven B cell lymphoma. These findings establish a scalable framework for the characterization of allogeneic T cell products and may inform the development of predictive metrics for in vivo efficacy.

Allogeneic T cell therapies could be used in therapeutic applications because of their potential for ‘off-the-shelf’ access and standardised production. Here the authors have developed a multidimensional workflow profiling platform for EBV-specific T cell therapy and show that correlative biomarkers of T cell potency and effector function are associated with therapeutic effectiveness in xenogeneic mouse EBV-LCL models.

Introduction

Adoptive cell therapies (ACT) represent a transformative modality for treating cancer, infectious diseases, and autoimmune disorders^1–3. Among these, CD19-targeting chimeric antigen receptor (CAR) T-cell therapy has revolutionized the management of hematologic malignancies and is now globally approved^4–6. Despite this success, autologous T-cell approaches face critical limitations, including product-to-product variability driven by the intrinsic heterogeneity of patient-derived T cells⁶. This variability affects phenotypic and functional properties, ultimately impacting persistence and therapeutic efficacy in vivo. These limitations may underlie the reduced success of ACT in solid tumours, where durable T-cell function is essential for efficacy^1,7.

Allogeneic ACT offers a promising alternative by enabling the selection of T cells from healthy donors with defined and consistent functional attributes^8–10. These off-the-shelf T cells can be banked and delivered on demand, bypassing the manufacturing delays associated with autologous approaches^9,10. While concerns about graft-versus-host disease (GvHD) persist, clinical studies with partially HLA-matched allogeneic virus-specific T cells (VSTs) have demonstrated favorable safety profiles, with minimal alloreactivity^11–16. Rejection of allogeneic T cells remains a challenge, particularly in immunocompetent hosts; however, strategies such as HLA knockdown and gene editing are being pursued to enhance persistence^9,10. In immunocompromised post-transplantation settings, these therapies have shown marked efficacy, particularly against viruses that establish lifelong latency, including Epstein-Barr virus (EBV), cytomegalovirus (CMV), BK polyomavirus, and adenovirus^17,18.

EBV infects 90–95% of the global population and typically establishes latency in B cells following primary infection^19,20. While most infections are asymptomatic in early childhood, delayed exposure in developed countries often results in infectious mononucleosis during adolescence²¹. EBV latency is tightly regulated and linked to B cell differentiation states, with immune control largely mediated by cytotoxic CD8⁺ T cells²². Groundbreaking studies by Heslop and colleagues demonstrated that donor-derived, virus-specific T cells can effectively treat EBV-associated post-transplant lymphoproliferative disorders (PTLD)^23,24. Further studies by Haque et al. showed that partially HLA-matched, third-party EBV-specific T cells can be safely used with minimal GvHD risk—only one out of 57 hematopoietic stem cell transplant recipients experienced GvHD reactivation after treatment^25,26.

In this study, we describe a multi-omics platform to evaluate the phenotypic and functional attributes of allogeneic EBV-specific T cells, incorporating T cell receptor repertoire analysis, transcriptional profiling, and high-content flow cytometry. This comprehensive approach defines key correlates of potency in cytotoxic T lymphocytes and provides a foundational framework for linking product characteristics to clinical outcomes in EBV-associated diseases.

Results

Expansion and functional characterization of allogeneic EBV-specific T cells

Our EBV-specific allogeneic T-cell immunotherapy is designed to contain a selective enrichment of CD8⁺ T cells specific for the EBV-encoded latent cycle proteins, latent membrane protein (LMP) 1 and 2, and EBV Nuclear Antigen (EBNA) 1. To establish broad human leukocyte antigen (HLA) coverage, a repository of 21 batches of allogeneic EBV-specific T cells was generated from the peripheral blood mononuclear cells (PBMC) of donors recruited through the National Marrow Donor Program (Minneapolis, MN). T cells were expanded using a single stimulation with the AdE1-LMPoly vector, as previously described^27,28, cultured for 17 or 20 days in the presence of interleukin-2 (IL-2), and then cryopreserved in aliquots of 3 × 10⁷/cells per vial. To provide an increased yield for certain common HLA types, some donors were used to generate multiple batches. Fifteen of the batches contained EBV-specific CD8⁺ T cells restricted through at least two HLA-alleles (Fig. 1A). The other six contained T cells restricted to a single HLA. Fifteen batches contained CD8⁺ T cells specific for LMP2, 10 batches were specific for LMP1, and 12 were specific for EBNA1.

Fig. 1. Functional characterization of EBV-specific T-cell therapy.

A HLA-restriction and antigen-specificity of each EBV-specific T-cell therapy product were determined using a CD107a release assay. Red boxes correspond to the HLA-restriction of each batch of T-cell products. Blue boxes correspond to the antigens recognized by each batch of T-cell products. Grey shaded boxes indicate the T-cell products manufactured from PBMCs of an individual donor. B T-cell cytokine profile was assessed using multiparametric intracellular cytokine analysis following stimulation with HLA-matched EBV peptides. Representative analysis is shown for TIG.E.2 stimulated with a LMP1-encoded HLA-B*57:01 restricted epitope, TIG.J.1 stimulated with a LMP2-encoded HLA-A*23:01 restricted epitope, and TIG.D.1 stimulated with an EBNA1-encoded HLA-B*08:01 restricted epitope. C T-cell division was assessed using a cell-trace violet proliferation assay following stimulation with HLA-matched LCL. Representative analysis is shown for TIG.E.2 stimulated with HLA-B*57:01⁺ LCL, TIG.J.1 stimulated with a HLA-A*23:01⁺ LCL, and TIG.D.1 stimulated with a HLA-B*08:01⁺ LCL. D Kinetics of T-cell cytotoxicity was assessed against HLA-matched LCL using xCELLigence. Representative cytotoxicity is shown for TIG.E.2 against a HLA-B*57:01⁺ LCL, TIG.J.1 against a HLA-A*23:01⁺ LCL and TIG.D.1 against a HLA-B*08:01⁺ LCL.

To assess the key functional attributes of EBV-specific T cells that are likely to be associated with cellular potency, antigen-specific T cells in each batch were assessed for polyfunctional cytokine profile using a multiparametric intracellular cytokine assay (Fig. 1B, Supplementary Fig. 1A), proliferative potential using a cell trace violet proliferation assay (Fig. 1C, Supplementary Fig. 1B) and cytotoxic potential using an xCELLigence real time cytotoxicity assay (Fig. 1D). Most EBV-specific T cells displayed a polyfunctional profile typically associated with virus-specific CD8⁺ T-cell populations, characterised by the co-expression of TNF, IFN-γ and degranulation (CD107a expression), with a subset of cells co-expressing IL-2 (Fig. 2A, Supplementary Table 1). All EBV T-cell therapy batches displayed HLA-restricted cytotoxic and proliferative potential (Fig. 2B) against EBV-transformed lymphoblastoid cell lines (LCL). Correlative analysis revealed a significant association between cytotoxic potential and the proportion of CD107a⁺ CD8⁺ T cells (Fig. 2C). Although less significant, a correlation was also observed between cytotoxic potential and IFN-γ⁺ or TNF⁺ CD8⁺ T cells. A significant correlation was not observed between IL-2⁺ CD8⁺ T cells and cytotoxic potential, nor between proliferation and any of the other functional attributes.

Fig. 2. Correlative analysis of EBV-specific T-cell product cytokine profile, effector function and proliferative potential.

A Data represents the proportion of EBV-specific T cells in each batch producing IFN-γ, TNF, IL−2 or degranulating (CD107a) following stimulation with HLA-matched peptide epitopes. Rows represent each product assessed, and columns represent the HLA restriction of the peptides used for stimulation. The column labelled “All” represents the response to a peptide pool covering all peptide epitopes targeted in EBV-specific T cells. Batches generated from the same starting material and boxed. B Data represents the cytotoxic potential (AUC) and proliferative potential (proliferation index) of HLA-restricted T cells in each product. Both cytotoxic and proliferative potential were calculated following incubation with LCL HLA-matched to a single allele. Batches generated from the same starting material and separated by dashed lines. C Panels represent the correlation between HLA-restricted AUC or proliferation index with the proportion of HLA-restricted peptide-specific CD8⁺ T cells producing IFN-γ, TNF, IL-2 or degranulating (CD107a). Coefficient of determination (r²) was performed using a two-tailed Pearson in GraphPad software and considered statistically significant if p < 0.05.

TCR clonotypic landscape of in vitro expanded allogeneic EBV-specific T cells

To assess TCR identity and clonal diversity in the EBV-specific T-cell therapy batches, we used a cytokine capture (IFN-γ⁺TNF⁺) approach to define EBV-specific CD8⁺ T cells. To induce cytokine secretion, T cells were stimulated with a pool of peptide epitopes encoded in the AdE1-LMPpoly vector. TRBV CDR3 sequencing was then performed using the ImmunoSEQ platform on DNA isolated from IFN-γ⁺TNF⁺ CD8⁺ T cells and total T cells from each batch, and from PBMC CD8⁺ T cells prior to expansion. The frequency of EBV-specific TRBV CDR3 was then compared between each batch and PBMC CD8⁺ T cells to assess clonal expansion. A summary of the TRBV CDR3 analysis is provided in Supplementary Table 2. All batches displayed a dramatic increase in the proportion of EBV-specific CD8⁺ T cells relative to the starting PBMC (Fig. 3A). The EBV-specific T cells increased from a median of 0.4% (range 0.1–1.3%) in the PBMC CD8⁺ T cells to a median of 63.4% (range 40.5–90.6%) in the final product. We also noted a strong correlation between the TRBV CDR3 frequencies in the starting PBMC CD8⁺ T cells with the proportion in the product CD8⁺ T cells (Fig. 3B). Although we noted a consistent increase in the absolute proportion of TRBV CDR3 sequences in the EBV T-cell therapy batches, there were clear differences in the repertoire diversity between batches. Both Shannon Diversity (Fig. 3C) and Pielou Evenness (Fig. 3D) analyses demonstrate that while most batches contain a diverse TCR profile, TIG.K.1 and TIG.K.2, which were generated from the same starting material, contain limited diversity associated with a single dominant TRBV CDR3 sequence.

Fig. 3. TCR clonotypic composition of EBV-specific T-cell therapies.

A The ImmunoSeq platform was used to assess TRBV CDR3 usage in IFN-γ⁺TNF⁺ T cells isolated from (n = 17) stimulated with LMP1&2 and EBNA1-encoded HLA class I peptide epitopes for 4 hours. The frequency of EBV-specific TRBV CDR3 identified in the IFN-γ⁺TNF⁺ T cells was then determined in PBMC CD8 + T cells prior to expansion (PBMC) and in the final product CD8⁺ (product). Colours in paired PBMC and product samples correspond to the same TRBV CDR3 sequence. B Correlation between TRBV CDR3 frequency in starting PBMC and final product (n = 495 paired samples). Correlations were determined using a two-tailed Pearson and considered statistically significant if p < 0.05. C, D EBV-specific TRBV CDR3 populations in EBV-specific T-cell therapy products were assessed for diversity (C) and evenness (D). Batches generated from the same starting material are identified with the same shade.

To investigate the relationship between T-cell identity in different batches, we compared TRBV CDR3 usage between all batches sequenced in the study. We used a circos plot to display the sharing of TRBV CDR3 sequences from batches made from the same starting material and batches made from different starting materials (Fig. 4A). The outer band in the circos plot represents the batches made from the same starting material, and the inner band represents individual products. The number of TRBV CDR3 sequences shared between products is represented by the thickness of the ribbons. We observed a high level of concordance between batches generated from the same starting PBMC, whereby the frequency of dominant TRBV CDR3 sequences was maintained between batches (Fig. 4A). These observations also revealed sharing between unrelated batches, likely associated with shared HLA-restriction and EBV epitope reactivity. To further study the relationship between batches, we employed GLIPH analysis to define motifs within the whole library of EBV-specific TCR. In this analysis we identified several patterns associated with HLA and epitope specificity (Fig. 4B). These analyses revealed that the TRBV CDR3 sequences were dominated by a single motif, QGG, which was previously reported to be associated with the HLA-A*02:01 restricted LMP2-encoded epitope FLYALALLL²⁹ and was present in 11 batches restricted through HLA-A*02:01. We also noted two additional motifs that were shared in unrelated batches, YAG that was present in 6 batches with shared HLA-A*02:01 restriction and RTGA that was present in five batches with shared HLA-B*40:01 restriction. These observations further emphasize the consistent nature of the EBV-specific T-cell therapy batches.

Fig. 4. Diversity and TRBV CDR3 sharing in EBV-specific T-cell therapy products.

A Circos plot showing concordance of functional TCR identity in batches generated from the same donor material. The outer color bars represent individual donors. Batches generated from the same donor are grouped by donor letters. Ribbon thickness correlates with the number of shared TCR clones. B GLIPH TCR motif analysis was used to identify shared CDR3 motifs in all batches. The top left panel represents an overlay of all motifs of 14-amino acids in length. The bottom three left panels represent 13, 14 and 15-mers containing a QGG motif and present in 11 HLA-A*02:01-restricted batches. The right panel represents two motifs, YAG and RTGA, present in 6 HLA-A*02:01 and 5 HLA-B*40:01 restricted batches, respectively.

Phenotypic profiling of in vitro expanded allogeneic EBV-specific T cells

To assess key phenotypic attributes of EBV-specific T-cell therapy batches, we established a 25-color flow cytometry panel to study the differentiation status and trafficking potential of 21 EBV-specific T-cell therapy batches (Supplementary Table 3). When available, batches were co-stained with EBV-specific HLA-matched MHC-multimers (Supplementary Table 4). Samples were gated on CD45⁺CD3⁺CD8⁺ viable cells using FlowJo software (Supplementary Fig. 2), then Uniform Manifold Approximation and Projection (UMAP) and FlowSOM clustering analysis were performed on concatenated samples using OMIQ software. Marker expression patterns following UMAP analysis are shown in Supplementary Fig. 3. Clustering analysis of all cells is shown for both CD8⁺ T cells and MHC-Multimer specific cells in Fig. 5A (individual CD8⁺ T-cell clustering profiles for each batch are shown in Supplementary Fig. 4). Overlay analysis of MHC-multimer specific T cells on CD8⁺ T cells for each batch is also shown (Fig. 5B, Supplementary Fig. 5). While we saw uniform expression of some key differentiation makers, particularly CD49d, clustering was defined by the differential expression of other key differentiation makers, including CD27, CD28, CD57, CD62L and KLRG1 (Fig. 5C). Assessment of the proportion of both total CD8⁺ T cells and MHC-Multimer specific T cells demonstrated a consistent profile between batches generated from the same donor (Fig. 5D, E), which was consistent with our observations when assessing the functional profile and TRBV CDR3 repertoire. We did note some differences in our analysis, whereby TIG.A.1, TIG.B.1, TIG.B.2, TIG.G.1 and TIG.H.1 contain a high proportion of MHC-multimer specific T cells in cluster 1, which displayed high levels of CD28 but low levels of CD57 and CD27, in contrast to the remaining batches which were predominantly in cluster 4 and 5, that contain cells co-expressing CD28, CD57 and/or CD27.

Fig. 5. Multidimensional flow cytometric analysis of EBV-specific T-cell therapy batches.

T-cell batches (n = 21) were assessed for the expression of T-cell-associated markers and epitope specificity using MHC-multimer on a Cytek Aurora. Samples were gated on CD45⁺CD3⁺CD8⁺ viable cells using Flowjo software, then Uniform Manifold Approximation and Projection (UMAP) and FlowSOM clustering analysis were performed using OMIQ. A Clustering analysis of concatenated samples in total CD8⁺ T cells and MHC-Multimer⁺ CD8⁺ T cells. B UMAP from four representative T-cell product batches overlaid with MHC-Multimer⁺ CD8⁺ T cells. C Bubble plot highlighting the expression of markers (by low to high shades of red) and percentages (%) of cells expressing these markers (as sizes of circles) for the clusters C1-C6. D Proportion of CD8⁺ T cells from each product in each of the clusters. E Proportion of MHC-Multimer⁺CD8⁺ T cells from each product in each of the clusters.

Gene expression profiling of in vitro expanded allogeneic EBV-specific T cells

To broaden our assessment of the functional and phenotypic profile of EBV-specific T-cell therapy batches, we employed a custom-designed gene expression panel using the NanoString nCounter gene expression platform. Gene expression profiling was performed on unstimulated CD8⁺ T cells, enriched CD8⁺ T cells stimulated with EBV peptide epitopes, EBV-specific T cells enriched using cytokine capture (IFNγ⁺/TNF⁺) following stimulation with EBV peptide epitopes and EBV-specific T cells enriched with HLA-peptide multimers specific for EBV epitopes (Supplementary Table 4). The T-cell therapy batches displayed a high degree of correlation in their gene expression profiles with batches generated from the same PBMC, clustering close by in a Spearman correlation plot of all batch pairs (Fig. 6; Supplementary Data 1). The Spearman correlation was also consistent with the phenotypic profiling analysis, whereby TIG.A.1, TIG.B.1, TIG.B.2, TIG.H.1 and TIG.G.1 clustered together. This pattern was consistent in both unstimulated (Fig. 6A) and stimulated (Fig. 6B) CD8⁺ T cells, and following cytokine capture (Fig. 6C) or HLA-peptide multimer enrichment (Fig. 6D). Stimulation was also associated with strengthening of the correlation between lots, with an enriched cluster of lots in the centre of the Spearman plot. This is likely a consequence of these lots having a consistent activation profile. Clustering patterns were predominantly driven by differences in gene expression of key effector molecules. In unstimulated CD8⁺ T cells (Fig. 7A), the TIG.A.1, TIG.B.1, TIG.B.2, TIG.H.1 and TIG.G.1 cluster was associated with reduced expression of granzymes (GZMA, GZMK, GZMH), perforin (PRF), granulysin (GNLY) and Natural Killer Cell Granule Protein 7 (NKG7), all of which are involved in cytotoxicity. Conversely, this cluster was associated with the increased expression of genes involved in glycolysis, including ALDOA, PGAM1, LDHB and TPI1A. This cluster in stimulated CD8⁺ T cells (Fig. 7B) was associated with increased expression of the interleukins -32 (IL32), -13 (IL13), -5 (IL5) and -8 (CXCL8). The signature following cytokine capture was consistent with the profile seem in stimulated CD8⁺ T cells (Fig. 7C). The profile in HLA-peptide multimer enriched T cells was consistent with the unstimulated CD8⁺ T cells, characterised by lower expression of effector molecules and higher expression of glycolysis-related genes in the TIG.A.1, TIG.B.1, TIG.B.2, TIG.H.1 and TIG.G.1 cluster (Fig. 7D). The observations of our phenotyping and gene expression analysis suggested that our T cells could be grouped into two subsets, the first subset displaying a central memory like phenotype (higher CD28 expression) that was associated with lower effector potential and higher glycolysis, while the other subset could be characterised with an effector memory like phenotype (higher CD57 expression) associated with increase effector gene expression and lower glycolysis.

Fig. 6. Correlation heatmaps for normalised expression of EBV-specific T-cell therapies.

A Unstimulated CD8⁺ T cells; B enriched CD8⁺ T cells stimulated with EBV peptide epitopes, C EBV-specific T cells enriched using cytokine capture (IFNγ⁺/TNF⁺) following stimulation with EBV peptide epitopes; D EBV-specific T cells enriched with HLA-peptide multimers specific for EBV epitopes. Each panel shows the pairwise Pearson correlation between T cell products. Expression values were normalised to the average of six housekeeping genes (ACTB, B2M, GAPDH, HPRT1, LDHA, RPLP0), log₂-transformed, and filtered using group-specific gene variance thresholds (unstimulated CD8⁺ T cells: >0.0015; enriched CD8⁺ T cells stimulated with EBV peptide epitopes: >0.005; EBV-specific T cells enriched using cytokine capture (IFNγ⁺/TNF⁺) following stimulation with EBV peptide epitopes: >0.00; EBV-specific T cells enriched with HLA-peptide multimers specific for EBV epitopes: >0.0025). Gene expression was then standardised by row-wise z-scoring prior to correlation computation. Heatmaps display hierarchical clustering of samples using complete linkage and show correlation coefficients ranging from low (dark blue) to high (deep red) based on a custom diverging colour palette.

Fig. 7. Unsupervised clustering of gene expression in EBV-specific T cell therapies using normalised gene expression levels.

A Unstimulated CD8⁺ T cells; B enriched CD8⁺ T cells stimulated with EBV peptide epitopes; C EBV-specific T cells enriched using cytokine capture (IFNγ⁺/TNF⁺) following stimulation with EBV peptide epitopes; D EBV-specific T cells enriched with HLA-peptide multimers specific for EBV epitopes. Gene expression was first normalised to the mean of six housekeeping genes (ACTB, B2M, GAPDH, HPRT1, LDHA, RPLP0) per sample, log2-transformed, and filtered using a group-specific variance threshold (No Stim: 0.0015, Stim: 0.005, Capture: 0.007, Multimer: 0.0025). Gene expressions were then standardised (row-wise z-score) and clustered using hierarchical clustering with Euclidean distance and complete linkage. Colour intensity represents relative gene expression (z-score) within each panel, ranging from -2 (dark blue) to +2 (deep red). For each condition, the top 20 differentially expressed genes are shown.

We next explored the relationship between the attributes of the EBV-specific T-cell therapies. T cell products were given a qualitative score (0, 1 or 2) for Degranulation, Cytotoxic Potential, Division Index, Shannon Diversity and Pielou Evennes; or a categorical score (1 or 2) for Cell Phenotype and Gene Signature (Supplementary Table 5). Some T cell products did not have complete datasets and were therefore not included in this analysis. Ring diagrams demonstrate the relationship between these parameters (Fig. 8). This analysis demonstrated that a high degranulation score was consistently associated with an effector memory-like phenotype and a high effector gene signature. While the majority of products demonstrated a high score in at least one parameter, TIG.G.1 did not, and scored poorly overall. To provide further insight into the association between gene expression and functional attributes, we used principal component analysis (PCA). PCA analysis using the top-correlating genes with these features (see Methods) showed the batches spread out across the plane of top-two PC dimensions, with batches showing high proliferation index—e.g. TIG.J.1, TIG.J.2, TIG.E.2 and TIG.E.3—clustered in the same half of the PCA, but away from those with lower proliferation index—e.g., TIG.K.1, TIG.K.2 and TIG.K.3 (Fig. 9A). T-cell therapies manufactured from the same PBMC further clustered close by (e.g., TIG.E.1, TIG.E.2 and TIG.E.3), reiterating their uniformity in proliferation and supported by similarities in their gene-expression profiles. ITGAM, TBX21, CXCR3, and CD226 were some of the genes associated with the high proliferating batches of EBV-specific T cell therapies. Likewise, batches with high CD107a clustered in the same half of the PCA plot, with sister batches clustering close by, and these were associated with high expression of genes such as GZMK, NKG7, EOMES, and XCL (Fig. 9B). While TIG.G.1 had a gene expression profile similar to the TIG.B.1, TIG.B.2 and TIG.A.1, it clustered independently in the correlation with CD107a and Division Index and was placed at ‘extreme bottom left corners’ of these PCA plots. The clustering in both settings was associated with differential expression of the FLNA gene that encodes Filamin A. Gene expression profiles also correlated with Shannon TCR diversities of the batches. In unstimulated cells, low TCR diversity was associated with expression of CXCR6, GZMA, CD96, XCL1, CCL4 and CCR1 (Fig. 9C). In stimulated CD8⁺ T cells batches with high TCR diversity, such as TIG.B.1 and TIG.B.2, were associated with expression of cytokines and chemokines, including IL3, CXCL8, CSF2, IL21, and CCL3 (Fig. 9D).

Fig. 8. Ring diagram profiles of EBV-specific T cells.

EBV-specific T-cell therapy batches were scored for each parameter as follows: Degranulation 0: Low ( < 33%) 1: Intermediate (33%–66%) 2: High ( > 66%); Cytotoxic potential 0: Low (25% Percentile <14.76) 1: Intermediate (14-76-35.58) 2: High (75% Percentile) > 35.58; Division index 0: Low (25% Percentile) < 1.040; 1: Intermediate (1.04–2.9); 2: High (75% Percentile) > 2.900; Shannon diversity 0: Low (25% Percentile) < 1.636 1: Intermediate (1.636–3.146) 2: High (75% Percentile) > 3.416; Pielou evenness 0: Low (25% Percentile) < 0.637 1: Intermediate (0.637-0.8370) 2: High (75% Percentile) > 0.8370; Phenotype 1: Central Memory Like ( > Cluster 1) 2: Effector Memory Like ( > Cluster 4); Gene Signature 1: Low Effector 2: High Effector. Ring Diagrams were generated using Flourish Studio software and represent the scores for nine different T-cell batches generated from different starting materials.

Fig. 9. Associating NanoString gene expression profiles with functional attributes.

The EBV T-cell therapy batches were ‘laid out’ on principal-component planes based on the expression of genes, and functional attributes were mapped onto the EBV T-cell therapy batches as colours (low = blue, high = red) and sizes of dots. Four functional attributes are shown here: A exogenous proliferation index, B CD107a marker expression, C TCR diversity (Shannon diversity) of unstimulated CD8⁺ T cells, and D TCR diversity (Shannon) of antigen-stimulated CD8⁺ T cells. The batches clustered on these planes, with those with higher and lower measures for the four attributes clustering separately. Importantly, distinct sets of genes were associated with batches with high (vis-à-vis low) measures for these attributes.

Pre-clinical assessment of in vitro expanded allogeneic EBV-specific T cells

We next sought to assess the impact the unique transcriptional and functional profile of EBV-specific T cells has upon their potential potency. We selected four batches, TIG.B.2, TIG E.3, TIG.J.2, and TIG.G.1, that all contained T cells restricted via HLAA*02:01. These batches were selected due to their different parameter scores and unique clustering patterns in the PCA. To assess the potential potency of different batches, we used a xenogeneic EBV-LCL tumour model. The EBV-LCL selected was matched to all four T cell therapy batches by HLA-A*02:01 restriction alone. NOD/Rag1γC knockout mice were injected subcutaneously in the flank with 2 × 10⁷ EBV-LCL (Fig. 10A). Mice received 2 × 10⁷T cells intravenously on days 6 and 12 after EBV-LCL inoculation. Routine monitoring of these animals showed no evidence of adverse effects of T-cell therapy, including weight loss or graft-vs-host disease. After 6 weeks, mice were sacrificed and organs assessed for EBV DNA and human cell subsets (Fig. 10A). Although all four T cell therapy products contained similar antigen-specificity, including the presence of the shared FLY-specific TRBV CDR3 motif (QGG, see Fig. 4B), there was discordance in their ability to control the outgrowth of EBV-transformed B cells in vivo. While mice treated with TIG.B.2, TIG.E.3 and TIG.J.2 showed significant reduction in EBV DNA in the spleen (Fig. 10B), liver (Fig. 10C) and brain (Fig. 10D), we saw no significant differences in the EBV DNA load from mice treated with T cell therapy batch TIG.G.1 when compared to the control animals who were untreated, although there was trend towards a reduction in the EBV DNA titre (Fig. 10B–D). This was consistent with the number of human B cells in the spleen and liver in these mice (Supplementary Figs. 6 and 7). Mice treated with TIG.B.2, TIG.E.3 and TIG.J.2 also displayed significant reconstitution of human CD3⁺CD8⁺ T cells (Fig. 10E, Supplementary Fig. 8), human CD3⁺CD4⁺ T cells (Fig. 10F, Supplementary Fig. 8) and EBV-specific cytokine-secreting cells (Fig. 10G) that were not evident in mice treated with TIG.G.1. These observations demonstrate that, despite the similar EBV-specific antigenic reactivity of T cells, multidimensional profiling can reveal differences that impact potential potency in vivo.

Fig. 10. Disparate control of EBV-transformed B cells by different T-cell batches.

A NOD/Rag1 γC knockout mice were injected subcutaneously in the flank with EBV-LCL. On days 6 and 12 after tumour inoculation received intravenous injections of EBV-specific T cells. Mice were sacrificed 6 weeks later, organs harvested and assessed for the presence of EBV using qPCR and human cell subsets by flow cytometry. Part of this panel was created in BioRender. Khanna, R. (2025) https://BioRender.com/adcsk4v. B–D Box and whisker plots represent the EBV DNA copy (median with minimum to maximum range, dots represent individual mice) from 12 mice across two experiments in the spleen (B), liver (C), and brain (D). E, F Box and whisker plots represent the number of CD3⁺CD8⁺ T cells (E) and CD3⁺CD4⁺ T cells (F) in the spleens of 12 mice per group across two experiments. G Splenocytes were assessed for the presence of EBV-specific T cells following stimulation with the EBV peptide epitope pool using a multiparametric intracellular cytokine assay. Box and whisker plots represent the frequency of cells producing any cytokine from six mice per group in one experiment, following subtraction of the spontaneous cytokine produced by cells incubated without peptide. Statistical significance was determined using a Kruskal-Wallis test with Dunn’s multiple comparisons.

Discussion

Understanding potency attributes of allogeneic anti-viral T-cell therapies will be critical as these therapies transition through early phase clinical studies and into approved cell therapy drug products. While the primary mechanism of T-cell function is mediated through TCR recognition of peptide-MHC complexes on the surface of target cells, T cells have a range of cellular processes likely to influence drug product efficacy^30–32. In addition, the clonal composition of the cellular product may also have an impact on cellular processes³³. It is therefore critical that we develop robust platforms to delineate attributes associated with potency. In this study, we integrate critical functional, phenotypic and gene expression attributes of T cells to demonstrate both the consistent nature of these attributes in an off-the-shelf allogeneic EBV-specific T-cell therapy and some of the unique attributes that could potentially be leveraged for beneficial clinical outcomes.

Cytotoxic T cells are functionally dependent upon their capacity to kill target cells, predominantly via the perforin/granzyme pathway, and cellular efficacy is likely to be dependent upon the efficiency of this process³⁴. We previously reported a potential association between T cell functional attributes and improved outcomes in a Phase I clinical trial of autologous EBV-specific T-cell therapy^27,35. Whether derived from unrelated healthy donors or derived from the same donor, all the EBV-specific T cell batches generated as part of this study demonstrated the ability to degranulate and kill target cells. It is now established in both pre-clinical and clinical studies that T-cell survival and proliferative potential are also key mediators of cellular efficacy^36,37. Some studies have suggested that maintenance of a stem-like T-cell profile drives in vivo survival and proliferation, and promotes improved outcomes compared to terminal effector T cells^38,39. Efficacy in these settings is likely still dependent on cytotoxic mechanisms following activation or differentiation^40–42. It is therefore evident that T-cell products may benefit from maintaining a balance between effector function and proliferative potential. All EBV-specific T-cell therapy batches tested in this study showed evidence of proliferation in response to secondary antigen exposure. We did note that some batches displayed relatively higher proliferative potential that did not correlate with effector function. Importantly, this phenotype was maintained in batches manufactured from the same PBMC, indicating that it is inherent to the T cells derived from the donor starting material and not simply process variability. It remains to be determined if this discordant proliferative potential will impact clinical efficacy.

One attribute that is variable in all polyclonal T-cell therapies is clonal diversity^43–45. This may be driven by the recognition of multiple T-cell epitopes via distinct HLA molecules, or through the selection of multiple clones that recognize the same peptide-MHC complex. While it has been reported that diversity in antigen/epitope recognition is associated with efficacy²⁷, the role of clonal diversity is not fully elucidated. However, it has been assumed that increased clonal diversity is likely to promote improved outcomes through the recognition of multiple peptide-MHC complexes on the target cell and the increased likelihood of recognition of epitope variants⁴⁶. We noted variation in the clonotypic composition of the EBV-specific T-cell therapy products, with evidence that batches generated from some starting material are dominated by a limited number of TCRs. This was particularly evident in batches targeting a single peptide-MHC complex.

The emerging use of HLA-matched allogeneic T-cell therapies provides an opportunity to leverage favorable functional attributes of the drug product in order to optimise therapeutic benefit. It also avoids the need to generate therapies from patients who are often undergoing other treatments that can negatively impact the immune system, and allows multiple uses of donors with desirable features. An expectation on the use of allogeneic T cells will be that inter-batch variability will have minimal impact on overall batch functionality to ensure product consistency, particularly when generated from the same starting material. Our TRBV CDR3 analysis confirmed the consistent nature of our process used to manufacture EBV-specific T-cell therapies, with uniform clonal composition in products generated from the same starting material. In addition, both in our flow cytometry and gene expression analysis, we also demonstrated a consistent cellular profile, including the following antigen engagement. This profile was associated with the retention of CD28 expression, important for T-cell co-stimulation and the expression of CD49d, which plays an important role in trafficking to the brain and enhances memory T-cell activation. In addition, we noted a uniform activation signature in EBV-specific T-cell therapy batches, characterised by the upregulation of IFN-γ and other cytokine genes known to be associated with T-cell efficacy.

Although we observed uniformity in the functional and phenotypic attributes of the EBV-specific T-cell therapy batches, combinatorial analysis of multiple datasets allowed us to demonstrate some unique features of different product batches. While all batches retain effector cells capable of degranulation and lysis of targets, we did observe a small subset of batches with lower effector molecule expression, which was associated with a higher proportion of EBV-specific T cells in these products that did not degranulate. This subset also displayed increased expression of CD28 and higher glycolysis-associated gene expression.

Our combinatorial analysis incorporating gene expression with proliferation, degranulation and TRBV CDR3 profile provided further evidence of product differentiation. To validate these molecular signatures and functional profiles, we selected four batches of EBV-specific T cell therapies and assessed their therapeutic potential in vivo using a humanised EBV lymphoma model. These studies clearly showed that despite the similar EBV-specific antigenic reactivity of these four different batches of T cell therapies, there was a clear difference in their ability to control the outgrowth of EBV-driven lymphomas in vivo. One of these cell therapy products, TIG.G.1, failed to control the outgrowth of EBV lymphoma, while animals treated with TIG.B.2, TIG.E.3 and TIG.J.2 showed a significant reduction in EBV DNA in the spleen, liver and brain. Furthermore, these three T cell products showed significant reconstitution of human CD3⁺CD8⁺ and EBV-specific cytokine-secreting T cells, which was not evident in mice treated with TIG.G.1. TIG.G.1 showed a comparable cytotoxic gene signature to TIG.B.2, suggesting the reduced cytotoxic potential was not solely responsible for poor immune control. However, we did observe other differences between TIG.G.1 and other T cell products. TIG.G.1 scored the lowest in our parameter analysis, and it clustered separately from the other products in the PCA of degranulation and proliferation with gene expression. In both plots, this differential clustering was associated with expression levels of the FLNA gene. Filamin A, which is encoded by FLNA, plays a number of important roles in T cells, including in TCR signalling and cytotoxicity. Recent observations have suggested that reduced expression of filamin A may impart cytotoxic T cells with increased softness that helps them evade perforin-mediated autolysis⁴⁷. In addition to poor control of EBV-transformed B cells, the TIG.G.1 T cells also failed to persist following adoptive transfer. While further investigation is required to study the mechanisms underlying poor immune control and persistence of TIG.G.1T cells, it is probable that the dysregulated gene expression, such as retention of higher FLNA expression, is impacting the efficacy of T cell therapy.

This study demonstrates that an integrated multi-omics approach provides an effective strategy to assess the key attributes of T-cell therapies and can predict potential therapeutic efficacy in vivo. Our observations suggest that while differences in typical functional (division and degranulation) and phenotypic attributes correlate with gene expression profiles, these differences alone do not completely define T cell potency. Other molecular pathways that potentially impact survival and cellular signalling likely play a key role. A more complete understanding of the impact these pathways have on T-cell potency should enhance future approaches to select optimal T-cell therapies for clinical use. We envisage that the data presented here will provide an ongoing approach to not only assess the correlates of potency and efficacy of EBV-specific T-cell therapy products, but also provide a platform for studies of other allogeneic T-cell therapies currently under development.

Methods

Ethics statement

This study complied with all relevant ethical regulations. The study was approved by the QIMR Berghofer Medical Research Institute Human Research Ethics Committee and conducted in accordance with the principles of the Declaration of Helsinki. All donors enrolled in the study provided informed written consent.

EBV-specific T-cell therapy manufacturing

Peripheral blood mononuclear cells (PBMC) used for manufacturing EBV-specific T-cell therapy were isolated following leukapheresis of EBV-seropositive donors. Donors were selected from the National Marrow Donor Program (NMDP) based upon human leukocyte antigen (HLA) type (Supplementary Table 6). The AdE1-LMPpoly vector used to stimulate the expansion of LMP1&2 and LMP2 and EBNA1-specific T cells has been described previously. To generate EBV-specific T-cell therapy, PBMC were mixed at a ratio of 7:3 with autologous irradiated PBMC previously infected for 1 hour with AdE1-LMPpoly at a multiplicity of infection (MOI) of 10:1. PBMC were cultured at 37°C/6.5%CO₂/95%RH in G-Rex-10 flasks in RPMI1640 supplemented with 5% human AB serum (RPMI-5% AB) at a cell density of 2 × 10⁶cells/cm². Cultures were supplemented with 120 IU/mL of recombinant interleukin (IL)-2 on day 2 and 5. On day 8, cells were enumerated and reseeded at a cell density of 2–3 × 10⁶ cells/cm² in RPMI-5% AB supplemented with 120 IU/mL IL-2 in G-Rex-10 or G-Rex-100 flasks. Cells were reseeded at 2–3 × 10⁶ cells/cm² every 3 days until harvest. On the day of harvest, cells were washed, enumerated, and then cryopreserved in single-use vials at 3 × 10⁷ cells in 10% DMSO/5% human serum albumin.

Multiparametric intracellular cytokine assay

EBV-specific T-cell therapy batches were thawed in RPMI1640 supplemented with 10% fetal bovine serum (RPMI-10% FBS) and 120 IU/mL and allowed to recover overnight at 37 °C/6.5%C0₂/95%RH. T cells were stimulated with a custom peptide pool containing all epitopes encoded in the AdE1-LMPoly vector (all peptides) or with individual HLA-matched epitopes (Supplementary Table 7), and then cultured in the presence of anti-CD107a–FITC (BD Biosciences, clone: Ber-ACT8, 1:100 dilution), GolgiPlug (BD Biosciences, Catalog Number 555029) and GolgiStop (BD Biosciences, Catalog Number 554724). After 4 hours, cells were washed and incubated with anti-CD4–Pacific Blue (BD Biosciences, clone: RPA-T4, 1:25 dilution), anti-CD8–PerCP-Cy5.5 (eBioscience, clone: RPA-T8, 1:25 dilution) and Live-Dead Near IR (ThermoFisher, 1:250 dilution). Cells were then fixed and permeabilised using Cytofix/Cytoperm (BD Biosciences, Kit Catalog Number 554714), washed with Perm/Wash (BD Biosciences, Kit Catalog Number 554714) and incubated with anti–IFN-γ–Alexa Fluor 700 (clone: B27, 1:50 dilution), anti-IL-2-PE (clone: MQ1-17H12, 1:50 dilution), and anti-TNF–APC (clone: Mab11, 1:200 dilution) (all from BD Biosciences). Cells were acquired using a BD LSRFortessa with FACSDiva software (BD Biosciences), and post-acquisition analysis was performed using FlowJo software (BD Biosciences). To assess the correlation between cytokine expression and other functional parameters, we performed a simple linear regression using GraphPad Prism and used Pearson to calculate the coefficient of determination (r²).

T-cell proliferation assay

EBV-specific T-cell therapy batches were thawed in RPMI-10% FBS and allowed to recover overnight at 37 °C/6.5%C0₂/95%RH. T cells were labelled with Cell Trace Violet (ThermoFisher, Catalog Number, C34557) according to the manufacturer’s instructions, then incubated at 37 °C/6.5%CO₂/95%RH in triplicate at a T cell: target ratio of 1:1 with irradiated EBV-transformed LCL matched to the EBV-specific T-cell therapy batch by a single HLA allele. As a control, T cells were also incubated with autologous LCL where available, or with near HLA-matched allogeneic LCL (Supplementary Table 8). On Day 7, T cells were harvested and then stained with anti-CD8–PerCP-Cy5.5 (clone: RPA-T8, 1:400 dilution), anti-CD3 APC (BD Biosciences, clone: SKL, 1:50 dilution), anti-CD4 FITC (clone: RPA-T4, 1:25 dilution) and anti-CD19 PECy5 (BD Biosciences, clone: HB19, 1:25 dilution). Cells were acquired using a BD LSRFortessa with FACSDiva software, and post-acquisition analysis was performed using FlowJo software (BD Biosciences). The division index for each batch was calculated as the Number of Divisions/the number of CD8⁺ T cells at the start of culture.

Real-time T-cell cytotoxicity assay

EBV-specific T-cell therapy batches were thawed in RPMI-10% FBS containing 120 IU/mL IL-2 and allowed to recover overnight at 37 °C/6.5%CO₂/95%RH. Prior to assay set-up, viable T cells were isolated using a Ficoll-Paque (GE Healthcare, Catalog Number GE17144003) density gradient. The xCELLigence RTCA MP Instrument (ACEA Biosciences) was then used to measure the cytotoxic potential of EBV-specific T cell therapy batches against HLA-matched LCL pulsed with or without AdE1-LMPpoly encoded peptides. Two complete HLA-mismatched targets were used as controls (Supplementary Table 9). All EBV-transformed LCL targets were generated in-house from healthy volunteers. LCL targets were seeded at a density of 1 × 10⁵ cells per well of an E-plate (Aligent, Catalog Number 5232368001) coated with anti-CD40 antibody (Miltenyi Biotec, Catalog Number 130-108-041) to ensure cell-tethering to the plate. The plates were then equilibrated in the incubator for 30 minutes and then transferred to the instrument inside a 37 °C incubator for 5 hours prior to the addition of T cells. T cells were added to the appropriate wells in triplicate at an effector:target ratio of 1:1. The E-plates were kept in the laminar flow cabinet for 30 minutes before being transferred to the RTCA instrument. Data acquisition was resumed, and the cell index value was measured for 60 hours. Area under the curve (AUC) for each of the conditions is acquired from the xCELLigence RTCA Software Pro.

Polychromatic T-cell phenotype analysis

EBV-specific T-cell therapy batches were thawed and immediately assessed for cellular phenotype using a custom-designed antibody panel (Supplementary Table 3). T cells were first stained with Live-Dead NIR at room temperature for 15 minutes, washed, then, when available, stained with MHC-multimers for 10 minutes (Supplementary Table 4). Anti-S1PR1 was then added and cells incubated for 10 minutes, followed by anti-CCR1, CCR2, CCR4, CCR5, CXCR6, CX3CR1, and KLRG1 for a further 10 minutes at room temperature. The remaining antibodies were added and cells incubated for a further 30 minutes at 4 °C. The cells were then washed, fixed with BD Cytofix fixation buffer (BD Biosciences, Catalog Number 554655), washed again and resuspended in PBS containing 2% foetal bovine serum. Cells were acquired on a Cytek Aurora laser spectral flow cytometer using SpectroFlo software as per manufacturer instructions. Unmixed FCS files were exported for analysis. Batches were run in three separate experiments. For each experiment, a reference sample of T cells was used to control for batch effects. Data analysis was initially performed using FlowJo software to gate on CD8⁺ T cells in each batch (Lymphocytes/Single Cells/CD14⁻/CD19⁻/CD45⁺/CD3⁺/CD4⁻/CD8⁺). Gated CD8⁺ T cells were exported for subsequent analysis using OMIQ (Omiq Inc., Santa Clara, CA). OMIQ was used to perform UMAP on concatenated files (UMAP settings: Neighbours = 50, Minimum Distance = 0, Components = 2, Metric = Euclidean, Learning Rate = 1, Epochs = 200, Random Seed=6828, Embedding Initialization=Spectral; 13 features = CD27, CD56, CD16, CCR5, CD49f, CD28, CCR4, CD57, CCR1, CD62L, CCR2, CXCR6, KLRG1). FlowSOM was then performed on the UMAP algorithm (number of metaclusters = 6). Categorical filters were then created for each metacluster, and single marker gating was performed.

TRBV CDR3 sequencing analysis

EBV-specific T-cell therapy batches were thawed in RPMI-10% FBS and allowed to recover overnight at 37 °C/6.5%C0₂/95%RH in RPMI-10% FBS containing 120 IU/mL IL-2. IFN-γ and TNF-producing CD8⁺ EBV-specific T cells were isolated using the following method. The cells were activated by pulsing 10% of the cell suspension (stimulators) with a pool of EBV epitopes encoded in the AdE1-LMPoly vector for 1 hour, then the stimulators were washed twice with RPMI-10% FBS and incubated with the remaining EBV-specific T-cells (responders). After four hours, IFN-γ-FITC (Miltenyi Biotech, Catalog Number 130-090-433) and TNF-APC (Miltenyi Biotech, Catalog Number 130-091-267) capture assays were performed according to the manufacturer’s instructions. Cells were stained with anti-CD8–PerCP–Cy5.5 (clone: RPA-T8, 1:400 dilution), anti-CD4–Pacific Blue (clone: RPA-T4, 1:25 dilution), anti-CD3–Pecy7 (clone: SKL, 1:50 dilution), and LIVE/DEAD Fixable Near-IR. CD3⁺CD8⁺CD4^-IFN-γ⁺TNF⁺ viable cells were sorted using a BD FACSAriaIII. To isolate total CD8⁺ T cells from both EBV-specific T-cell therapy batches and starting PBMC, cells were incubated with anti-CD8–PerCP–Cy5.5, anti-CD4–Pacific Blue, anti-CD3–Pecy7, and LIVE/DEAD Fixable Near-IR. CD3⁺CD8⁺CD4^– viable lymphocytes were sorted using a BD FACSAriaIII (Becton Dickinson). DNA was isolated from the sorted cells using Qiagen DNeasy Kit (Qiagen, Catalog Number 69506) and then sent to Adaptive Biotechnologies for TRBV CDR3 deep sequencing analysis using the immunoSEQ platform. A summary of the TRBV CDR3 sequencing data is provided (Supplementary Table 2). Enrichment of EBV-specific TRBV CDR3 was assessed using the ImmunoSEQ Analyzer. Only clonotypes with a minimum of 5 template reads were included in the analysis. Significance was determined using a 2-sided binomial test with Benjamini-Hochberg multiple-comparisons correction, where α = 0.01. To assess the correlation between CDR3 sequences in PBMC and batches, a 2-tailed Pearson correlation was used. An intersection dataframe for N = 17 unstimulated batches was computed between TCR sequencing CD8 data and significantly expanded CDR3 data. Among the common CDR3 sequences that were identified, the CDR3 sequences were replicated depending on the number of templates. Random sampling without replacement was performed 10,000 times for a sample size of 1,000 CDR3 sequences each time. For each sampling, Shannon Diversity and Pielou Evenness were calculated. All the 10,000 samples were visualized using Box plots and Distribution histograms by taking the average of the samples. Motif analysis was also performed on GLIPH, which returned significant motif lists and TCR convergence groups, and WebLogo was used to design sequence logos for those significant motifs.

T-cell gene expression analysis

T cells were prepared and sorted from T-cell therapy batches for gene expression analysis as outlined above for TRBV CDR3 sequencing analysis. In addition to sorting stimulated IFN-γ and TNF-producing EBV-specific CD8⁺ T cells and non-stimulated total CD8⁺ T cells, stimulated total CD8⁺ T cells and EBV-specific MHC-Multimer⁺ T cells were also sorted from each batch. RNA was extracted from each sample and stored at −70 °C until gene expression analysis was performed. Gene expression analysis was conducted using the NanoString nCounter gene expression platform (NanoString Technologies) according to manufacturer’s instructions. A custom code set consisting of 326 genes involved in T-cell biology, immune regulation, and immune cellular markers was used. The samples were scanned at maximum scan resolution on the nCounter Digital Analyzer and gene expression data normalized by housekeeping gene expression. Differentially expressed genes were calculated using t tests with Benjamini–Hochberg multiple test correction (FDR < 0.05), and represented as a heatmap using the pheatmap R package. Top 20 gene lists that correlated with EBV-specific T-cell therapy batches features namely, proliferation index, CD107a marker expression, and Shannon TCR diversity of unstimulated and stimulated cells were calculated using Spearman correlation. These gene lists were used to ‘lay out’ the product batches in two-dimensional principal-component analysis plots for the batch features, using the Factoextra R package. Principal Component Analysis was used to cluster the TIG products based on their Nanostring gene expression profiles. Log2-normalized gene expression profiles were used to derive the top two PCA components. The PCA plot shows the relative placement of the TIG products on a two-dimensional plane representing the two PCA components and the key genes that contribute to the components. Product features, namely, proliferation, CD107a, and TCR diversities, were overlaid on the TIG products on the PCA plot.

In vivo mouse adoptive T-cell therapy model

All animal studies were approved by the QIMR Berghofer Medical Research Institute Animal Ethics Committee under P2337 A1708-614 M (for breeding) and P3734 A2111-638 M (for experimental studies) and performed in strict accordance with the Australian Code for the Care and Use of Animals for Scientific Purposes. All animals were housed at the QIMR Berghofer Medical Research Institute Animal Facility in OptiMICE caging on a 12-hour light–dark cycle at 25 °C. Dried granule food was sterilised by irradiation. The mice had free access to food and acidified water. Six-eight-week-old female NOD/Rag1 γC knockout mice were injected subcutaneously in the flank with 2 × 10⁷ HLA-A2⁺ EBV-LCL. On day 6, tumours were measured, and mice were randomized into different groups (6 animals in each group) to ensure equal tumour volumes before T-cell injections. Test group animals were treated with 2 × 10⁷ T cells from batches TIG.B.2, TIG.E.3, TIG.G.1 and TIG.J.2. Control mice were untreated. Mice received a second dose of the matching T cell batch 6 days later. Mice were then monitored for six weeks for tumour growth and well-being. Tumours did not exceed the approved maximal tumour burden of 1 cm³ at the site of injection. Mice were then euthanized by CO₂ asphyxiation, and organs (spleen, liver and brain) and blood were harvested. Organs were then assessed for the presence of EBV DNA using qPCR as previously described, and for human immune cell subsets by flow cytometry and intracellular cytokine analysis. Blood was assessed for the presence of human T-cell subsets. The experiments were performed twice, and statistical analysis was performed using all treated mice (n = 12 per group). Statistical comparisons were performed using a Kruskal–Wallis test with Dunn’s multiple comparisons test. Data was considered statistically significant if p < 0.05.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

C.S. and R.K. designed the study. V.D. supervised and analysed data related to in vivo assessment of T-cell therapy, S.S. led the bioinformatic analysis of data, L.L.T., M.S., A.P., T.L., G.A., J.R., S.R., L.B., P.C., P.K., A.M., and P.M. conducted experimental studies and analysed the data. All authors contributed to the drafting and review of the manuscript.

Peer review

Peer review information

Nature Communications thanks Alan K.S. Chiang, Lars Rogge and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Data availability

TCR sequencing data generated in this study are available on the following link: 10.5281/zenodo.17905556. All other data are available in the article and its Supplementary files or from the corresponding author upon request. Source data are provided with this paper.

Competing interests

C.S. and R.K. are listed as inventors on patents and patent applications describing the autologous and allogeneic EBV T-cell therapy. C.S. and R.K. have received research and consultancy funding from Atara Biotherapeutics. The remaining authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-025-67924-w.