Stem cell memory EBV-specific T cells control EBV tumor growth and persist in vivo

Abstract

Adoptive T cell therapy (ACT), the therapeutic transfer of defined T cell immunity to patients, offers great potential in the fight against different human diseases including difficult-to-treat viral infections, but persistence and longevity of the cells are areas of concern. Very-early-differentiated stem cell memory T cells (T_SCMs) have superior self-renewal, engraftment, persistence, and anticancer efficacy, but their potential for antiviral ACT remains unknown. Here, we developed a clinically scalable protocol for expanding Epstein-Barr virus (EBV)–specific T_SCM-enriched T cells with high proportions of CD4⁺ T cells and broad EBV antigen coverage. These cells showed tumor control in a xenograft model of EBV-induced lymphoma and were superior to previous ACT protocols in terms of tumor infiltration, in vivo proliferation, persistence, proportion of functional CD4⁺ T cells, and diversity of EBV antigen specificity. Thus, our protocol may pave the way for the next generation of potent unmodified antigen-specific cell therapies for EBV-associated diseases, including tumors, and other indications.

Early-differentiated EBV-specific T cells offer a great potential in the treatment of EBV complications upon adoptive transfer.

INTRODUCTION

T cell therapies are promising for treatment of hemato-oncological diseases (1, 2), difficult-to-treat viral infections, and autoimmune diseases (3, 4). The efficacy of these therapies depends on T cell activation by antigens and in vivo persistence for sustained impact (5). Activated T cells can differentiate to stem cell memory T cells (T_SCMs), central memory T cells (T_CMs), transitional memory T cells (T_TMs), effector memory T cells (T_EMs), and terminally differentiated, short-lived effector T cells (T_EMRAs) (6). During T cell differentiation, effector functions increase, but self-renewal capacity declines (7). The superior proliferation and persistence of T_SCM have been demonstrated after adoptive transfer of genetically modified lymphocytes (8), chimeric antigen receptor (CAR) T cells, engineered T cell receptor (TCR)–T cells, and tumor-infiltrating lymphocytes (TILs) (9–11). Long-lasting antigen-specific T_SCMs were also identified after yellow fever and bacillus Calmette-Guerin vaccination (12, 13), and CD8⁺ T_SCMs support T cell responses to chronic lymphocytic choriomeningitis virus infection (14) and are associated with improved prognosis in chronic HIV-1 infection (15).

T_SCMs might also offer exciting avenues to improve adoptive therapy with virus-specific T cells (VSTs) against viral infections that are important causes of morbidity and mortality of immune-deficient transplant recipients. Adoptive transfer of VSTs can restore virus-specific immunity and prevent or cure such viral infections (16, 17). This includes transfer of Epstein-Barr virus (EBV)–specific cytotoxic T cell lines (CTLs) that prolongs overall survival in patients with EBV-driven posttransplant lymphoproliferative disease (PTLD), other EBV-associated lymphomas, and possibly even immunopathologies due to inefficient EBV-specific immune control. However, current data for EBV-driven PTLD using third-party EBV-CTLs indicate complete remission or sustained partial remission in 54 to 75% of solid organ transplant recipients and 46 to 68% of hematopoietic cell transplant recipients, while observing decreased responses as the involvement of multiple sites increases (18–21). Conversely, the treatment of type II latency EBV lymphomas is hindered by the limited presentation of EBV antigens by tumor cells, resulting in response rates ranging from 30 to 60% in various studies when treating active disease with EBV-CTLs (22, 23). For other EBV-related diseases such as nasopharyngeal carcinoma, patient response rates typically range from 20 to 37% (24–26). Despite promising advancements in managing patients with poor prognoses in EBV-related diseases, these findings highlight the need for further improvements (27). Limited long-term response due to poor persistence and exhaustion of the transferred T cells due to the immune suppressive microenvironment, rejection, or ongoing systemic immune suppression might contribute to insufficient response rates. Most clinical studies used EBV-CTLs generated by long-term expansion with continuous restimulation with EBV antigen–expressing lymphoblastoid cell lines (LCLs) (18, 20, 27–32), driving the cells to late differentiation stages and exhaustion (33). Alternatively, VSTs are often generated by rapid expansion using a single stimulation with a viral peptide mixture, but the differentiation state and persistence for EBV lymphomas are understudied (34–37).

Here, we established a facile, robust, and clinically applicable protocol for rapid expansion of EBV-CTLs from healthy donor peripheral blood mononuclear cells (PBMCs) with a high fraction of EBV-specific T_SCMs for the first use of T_SCM VSTs to treat viral complication. These cells mediate EBV control in vitro and in vivo and are superior to previous VST protocols in terms of tumor infiltration, in vivo proliferation, persistence, proportion of functional CD4⁺ T cells, and diversity of EBV antigen specificity.

RESULTS

Rapid expansion in presence of interleukin-4/interleukin-7 and TWS-119 yields high T_SCM proportions

To generate EBV-CTLs with high proportions of T_SCMs, we modified the rapid expansion approach (37). We stimulated PBMCs of healthy EBV-seropositive donors with the PepTivator EBV Consensus peptide pool (further referred as “EBV pepmix”) (Fig. 1A) that covers 43 latent and lytic peptides derived from 15 EBV proteins. We determined the impact of the cytokines interleukin-7 (IL-7), IL-15, and IL-21, which promote T cell growth but limit differentiation (34, 38, 39); potassium-rich medium promoting T cell stemness preservation (40); and the glycogen synthase kinase–3β inhibitor TWS-119, which induces Wnt/β-catenin signaling limiting cell differentiation and promoting T_SCM generation (41). We determined the proportions of T_SCMs (CD45RA⁺CD45RO⁻CD62L⁺CD27⁺) in expanded lines using flow cytometry (figs. S1 and S2).

Fig. 1. Establishing rapid T_SCM-enriched EBV-CTL ex vivo expansion protocol.

(A) Adapted rapid expansion approach. Isolated PBMCs were stimulated with EBV pepmix in complete medium. After overnight pulsing, pepmix was diluted 5× with complete medium following 10-day incubation. (B) T_SCM proportions after culturing EBV-CTLs in the presence of different conditions [different cytokine combinations, in elevated potassium concentration (K⁺) or with the edition of TWS-119] as detected by flow cytometry; n = 5, medians with range. (C) PD-1, LAG-3, TIGIT and TIM-3, exhaustion marker expression of expanded CTLs; n = 5, medians with range. (D) Expansion folds, PBMCs versus after rapid expansion (left), and counts per 1 × 10⁶ cells (right) of spot-forming cells (SFCs) after culturing under different conditions; interferon-γ (IFN-γ) ELISpot with EBV pepmix stimulation, n = 5, medians with range. (E) Short-term cytotoxicity against autologous EBV-LCLs, medians with range. (F) Long-term cytotoxicity: 4-week EBV-LCL outgrowth control by expanded T cells; n = 5; medians of controlling effector:target (E:T) were shown. In (B) to (F), all samples were analyzed by Friedman test between each other, α = 0.05, *P < 0.05 and **P < 0.005.

IL-4/IL-7 promoted T_SCMs, whereas IL-21 had limited impact (Fig. 1B and fig. S3A). Cytokine combinations with IL-15 enriched for natural killer (NK) and NK T cells and the T_SCM proportions were insufficient to consider NK cell depletion for the use in the clinical setting (fig. S4, A to C). Potassium-rich medium and TWS-119 also promoted T_SCMs (Fig. 1B and fig. S3B). Overall, IL-4/IL-7/TWS-119 yielded the highest T_SCM proportion (~30%) with comparable CD4⁺ and CD8⁺ T cell subsets (Fig. 1B and fig. S3, C and D). This condition also showed the lowest levels of T cell exhaustion markers (Fig. 1C). The different conditions had no impact on overall and antigen-specific T cell expansion as measured by enzyme-linked immunospot (ELISpot) assays, major histocompatibility complex (MHC) class I multimer staining, and total cell counts (Fig. 1D and fig. S3, E and F). The optimized protocol yielded EBV-CTLs with reduced short-term cytotoxicity against EBV-transformed LCLs (possibly due to delayed activation of early-differentiated T cells; Fig. 1E, gating strategy in fig. S5) but comparable long-term LCL control (Fig. 1F). Thus, rapid expansion in the presence of IL-4, IL-7, and TWS-119 yielded EBV-specific CTLs with favorable properties for VST therapy including high proportion of T_SCMs without the loss of EBV specificity, high CD4/CD8 ratio, low exhaustion, and efficient long-term in vitro cytotoxicity.

Expanded T_SCM are EBV-specific and proliferate in response to restimulation

Our subsequent objective was to elucidate the specificity of the early-differentiated T_SCM subset within the expanded EBV-CTLs. Since we lack precise information regarding the exact peptide sequences in the commercial EBV pepmix used for expansion, we conducted staining of the expanded cell lines with the most common MHC class I and MHC class II tetramers. Four to seven tetramers per donor were selected based on donor human leukocyte antigen (HLA) typing and the HLA profiles known to present the EBV pepmix antigens, as indicated in the manufacturer’s data sheet. Tetramer staining revealed that, in some instances, up to 10.7% of T_SCMs were positively stained within the CD8⁺ or CD4⁺ populations (fig. S8A). Notably, memory phenotypes of tetramer-bound populations were quite diverse (fig. S8B), and there was a correlation between the tetramer⁺ population frequencies and their differentiation status: Less prevalent specificities were less differentiated (more T_SCM phenotype), and more prevalent populations, on the contrary, often exhibited later, transitional memory phenotype (fig. S8C). This suggests that the T_SCM population may serve as a reservoir for rare VSTs.

To get more insight into T_SCM antigen specificity within the EBV-CTLs, we sorted various memory T cell subsets after rapid expansion (fig. S6) and challenged them for 7 days with a combination of irradiated autologous EBV-transformed LCLs, overlapping peptide pools covering single lytic and latent immunodominant EBV proteins (refer to Materials and Methods for the full antigen list) and EBV pepmix (further referred as “LCL/peptide mix”). All memory populations including the T_SCMs showed proliferation of CD4⁺ and particularly CD8⁺ T cells in these cocultures (Fig. 2A, top plots, gating strategy in fig. S7). The T_SCM population further showed the ability to produce interferon-γ (IFN-γ) upon antigen restimulation (Fig. 2A, bottom). Antigen specificity of the T_SCM population was further corroborated by the results from MHC class I multimer staining (Fig. 2B) and IFN-γ ELISpot responses to overlapping peptide pools covering single lytic and latent EBV proteins (Fig. 2C and fig. S8C) and to the EBV pepmix, autologous EBV-LCL, and the LCL/peptide mix, respectively (Fig. 2D).

Fig. 2. Assessment of EBV specificity in sorted T_SCMs after rapid EBV-CTL expansion.

All asessments were performed after 7-day in vitro challenge of sorted populations with LCL/peptide mix (irradiated autologous EBV-LCLs mixed with EBV pepmix and overlapping peptide pools derived from single immunodominant EBV antigens). (A) Proliferation of sorted and CellTrace Violet (CTV)–stained memory T cell populations (flow cytometry, n = 4, medians with range) and cytokine production by proliferating T_SCMs (IFN-γ expression upon restimulation with EBV pepmix, representative plot). (B) Proliferation of antigen-specific T_SCMs (stained with a respective MHC class I tetramer, representative plots). Proliferating cells in (A) and (B) are cells that proliferated at least once or more. Specificity of sorted T_SCMs versus sorted bulk remaining cells (all cells other than T_SCMs) to various overlapping peptide pools of single EBV antigens (C) and to EBV pepmix, autologous LCL, and LCL/peptide mix [autologous LCL (auto LCL) mixed with EBV pepmix and overlapping peptide pools derived from single EBV antigens] (D); IFN-γ ELISpot, n = 3 (different donors), medians with range, multiple Wilcoxon tests (no significance). T_SCM, CD45RA⁺CD45RO⁻CD62L⁺CD27⁺; T_CM, CD45RA⁻CD45RO⁺CD62L⁻CD27+; T_TM, CD45RA⁻CD45RO⁺CD62L⁺CD27⁺; T_EM, CD45RA⁻CD45RO⁺CD62L⁻CD27⁻; n/c, incubation with no antigen challenge.

T_SCM-enriched EBV-CTLs exhibit a favorable phenotype and broad specificity compared to conventional protocols

Our protocol was largely based on a previously published and clinically used protocol that involves the rapid expansion of multi-VSTs targeting cytomegalovirus (CMV), adenovirus (AdV), EBV, human herpesvirus 6 (HHV-6), and BK virus (BKV) (37). We compared the EBV specificity and other T cell characteristics using either multi-VST conditions or the developed T_SCM-enriched protocol (CTL-R). Donors seropositive for at least EBV and CMV were selected for the expansion.

The multi-VST protocol yielded larger expansion rates of total cells (Fig. 3A). This was probably due to the additional viral stimulants. The proportions of T cells, NK cells, CD4⁺ and CD8⁺ cells (Fig. 3B), the expansion rates of antigen-specific T cells measured by IFN-γ ELISpot (Fig. 3C), and MHC class I tetramer staining (Fig. 3D) were comparable in both protocols. Notably, intracellular cytokine staining revealed higher production of IFN-γ, tumor necrosis factor–α, and IL-2 (Fig. 3E and fig. S9, A and B) in both CD4⁺ and CD8⁺ T cells in the CTL-R protocol.

Fig. 3. Comparison of rapidly expanded EBV-CTLs (EBV CTL-R) and multi-VST rapidly expanded CTLs.

(A) Expansion rates of total cells. n = 3, medians with range, Wilcoxon matched-pairs signed-rank test (no significance). (B) General phenotypical characteristics of expanded T cells: CD56⁺, CD3⁺, CD4⁺, and CD8⁺ T cell proportions and flow cytometer. (C) Expansion folds of EBV-specific T cells, IFN-γ ELISpot after restimulation with EBV pepmix, autologous LCLs, and LCL/peptide mix (autologous irradiated EBV-LCLs mixed with EBV pepmix and overlapping peptide pools derived from single immunodominant EBV antigens). (D) Proportions of different single EBV antigen–specific T cells measured by respective MHC class I multimer staining using flow cytometry. n = 6, Wilcoxon matched-pairs signed-rank test. (E) Intracellular cytokine staining of IFN-γ⁺ in CD4⁺ (left) and CD8⁺ (right) after stimulations with EBV pepmix, LCLs, and LCL/peptide mix using flow cytometry. (F) Frequencies of EBV pepmix–, autologous LCL–, LCL/peptide mix–specific T cells and (G) frequencies of single protein–specific T cells in the expanded products and after 7-day in vitro challenge with LCL/peptide mixes (“challenged”), IFN-γ ELISpot after restimulation with respective antigens; the dotted lines indicates the threshold (spot calculations below the line were considered not substantially different from control). Zero values are not shown due to the restriction of the logarithmic scale. (H) Memory phenotypes and (I) exhaustion marker expression using flow cytometry. For (B), (C), and (E) to (I), n = 3, medians with range, α = 0.05, two-way analysis of variance (ANOVA), P < 0.05, **P < 0.005, and ***P < 0.001. For (G), different latent single protein-derived peptide pools were used in expanded and challenged cells due to the shortage of EBNA3A and EBNA3B pools on the day of expansion harvest. T_EMRA, CD45RA⁺CD45RO⁻CD62L⁻CD27⁻ T cells.

This was further confirmed after a 7-day in vitro challenge of multi-VST and CTL-R T cells with LCL/peptide mixes: CTL-R showed higher response rates to EBV pepmix, LCL/peptide mix, and the majority of single EBV proteins, both latent and lytic (Fig. 3, F and G). Furthermore, CTL-R exhibited a less differentiated phenotype, with fewer T_EMs and more T_SCMs in the products (Fig. 3H), and a lower expression of exhaustion markers, specifically PD-1 (programmed cell death protein 1) among CD4⁺ cells and TIGIT (T-cell immunoreceptor with immunoglobulin and immunoreceptor tyrosine-based inhibition motif domain) among CD8⁺ (Fig. 3I). Together, the CTL-R protocol enhances the diversity of EBV specificity and improves phenotypical characteristics, thereby promising greater potential applicability to various EBV-associated diseases.

Conventionally used clinical protocol to treat EBV-associated diseases uses EBV-transformed LCLs as antigen-presenting cells for expanding EBV CTLs in 4- to 5-week-long cocultures (CTL-L) (Fig. 4A) (18, 20, 27–31). We compared the established protocol and the T_SCM-enriching CTL-R protocol to evaluate the differences regarding specificity and phenotypes. Both protocols showed comparable purity profiles (fig. S10). CTL-L showed a higher T cell expansion with a lower proportion of CD4⁺ T cells than CTL-R (Fig. 4, B and C). Overall, EBV specificity was comparable (Fig. 4, D and E), but CTL-R had broader antigen specificity for both latent and lytic peptides (Fig. 4F). Memory phenotypes differed substantially with higher proportions of earlier differentiation stages (T_SCMs and T_CMs) in CTL-R and later differentiation stages (T_TMs and T_EMs) in CTL-L (Fig. 4G) and lower levels of exhaustion markers in CTL-R (Fig. 4H). Thus, the standard protocol yields more cells, but the newly developed protocol yields a broader antigen specificity and more favorable memory and exhaustion phenotypes.

Fig. 4. Comparison of EBV CTL-R and EBV CTL-L.

(A) Schematic of two expansion methods. (B) Expansion rates of total cells. n = 7, medians with range. (C) CD4⁺ and CD8⁺ T cell proportions in expanded cells. Two-way ANOVA, n = 6, means with SD. (D) Frequencies of EBV-specific T cells in the expanded products, IFN-γ ELISpot after restimulation with EBV pepmix; n = 6, medians with range, Wilcoxon matched-pairs signed-rank test. (E) Pair-wise comparisons of proportions of different single EBV antigen–specific T cells measured by respective MHC class I multimer staining using flow cytometry. n = 11, Wilcoxon matched-pairs signed-rank test. (F) Frequencies of single protein–specific T cells in the expanded products (left graph, latent; right graph, lytic), IFN-γ ELISpot after restimulation with peptide pools derived from single EBV proteins; the dotted lines indicate the threshold (spot calculations below the line were considered not significantly different from control). n = 6, medians with range, Wilcoxon matched-pairs signed-rank test. Zero values are not shown because of the restriction of the logarithmic scale. (G) Memory phenotypes and (H) exhaustion marker expression using flow cytometry. n = 6, means with SD, two-way ANOVA. For (C) to (H), α = 0.05, *P < 0.05, **P < 0.005, ***P < 0.001, and ****P < 0.0001. IrrEBV-LCLs, irradiated EBV-LCLs.

T_SCM-enriched EBV-CTLs broaden the spectrum of antigen-specific responses upon in vitro challenge with autologous EBV-LCLs

Next, we assessed the development of antigen-specific responses in CTL-R and CTL-L upon 7-day challenge with the LCL/peptide mix. While the CTL-R showed a broader antigen specificity after challenge, CTL-L showed a marked decrease in response diversity (Fig. 5, A and B). In addition, whereas the overall CTL-L responses decreased after further antigen exposure, the overall specificity of CTL-R significantly increased (Fig. 5, C to F). The increased specificity of CTL-R is probably due to the expansion of rare EBV-specific T cells from the T_SCM reservoir, whereas the decreased response of CTL-L is likely a result of exhaustion of already late differentiated T cells that underwent multiple antigen exposures during initial expansion.

Fig. 5. Development of EBV CTL-L–and EBV CTL-R–specific responses after antigen rechallenge.

(A) ELISpot response of CTL-R after expansion (left column) versus after 7-day in vitro antigen challenge (right column) per each donor. (B) ELISpot response of CTL-L after expansion (left column) versus after 7-day in vitro antigen challenge (right column) per each donor. (C to F) Comparisons of overall EBV specifities of expanded CTL-R, challenged CTL-R, expanded CTL-L, and challenged CTL-L against the EBV pepmix, autologous LCLs, and LCL/peptide mix; n = 4, α = 0.05, two-way ANOVA, *P < 0.05, **P < 0.005, and ***P < 0.0005.

T_SCM-enriched EBV-CTLs control tumor growth, proliferate, persist, and release proinflammatory cytokines in vivo

To test the in vivo function of T_SCM-enriched EBV-CTL, we used a well-characterized mouse model of EBV-driven B cell lymphoma (42–44). A total of 2 × 10⁶ luciferase-expressing EBV-LCLs were injected subcutaneously, followed by adoptive transfer of 1 × 10⁷ autologous CTL-L or CTL-R 3 days later intravenously into nonobese diabetic–scid gamma_c^−/− (NSG) mice supplemented with high doses of human IL-2 to support T cells in the NSG system (Fig. 6A). Tumor growth dynamics revealed that both CTL-L and CTL-R controlled tumor growth equally well over 4 weeks (Fig. 6, B to D).

Fig. 6. Tumor growth control by CTL-L and CTL-R in vivo.

(A) Schematic of the in vivo experiments. A total of 2 × 10⁶ tumor cells (luciferase-expressing EBV-LCLs) per mouse were injected into NSG mice subcutaneously, and, on day 3, 1 × 10⁷ autologous EBV CTL-L or EBV CTL-R per mouse were infused intravenously. Groups of tumor-only mice were kept as a negative control. All mice were supplemented with 1 × 10⁵ U per hIL-2 3×/week starting from the day of adoptive T cell transfer. Mice were euthanized after ~4 weeks, and organs were collected. Pooled data from three independent experiments with three different donors (n = 13 mice per group) are shown in further plots unless there was no sample available. Data from CTL-L–injected mice are marked in black circles, those from CTL-R are marked in pink squares, and tumor-only mice are marked in green-blue triangles. Tumor growth dynamic measured by calipering ~3×/week (B) and tumor luminescence measured 1 to 2×/week (C). (D) IVIS images showing tumor luminescence in individual mice. Pooled data from two experiments, n = 8 per treatment group. Images were taken on days 7, 14, 21, and 28 after tumor seeding. For (B) and (C), means with SD; mixed-effects analysis [(B) and (C)], α = 0.05, *P < 0.05, ***P < 0.001, and ****P < 0.0001. L, CTL-L-injected mice; R, CTL-R-injected mice; sc, subcutaneously; iv, intravenously; ip, intraperitoneally.

Three of 13 mice (23%) receiving CTL-R but no mice receiving CTL-L lost weight at late time points (fig. S11A), together with increased levels of white blood cells (WBCs) and higher serum IFN-γ and tumor necrosis factor–α levels in the CTL-R group (Fig. 7, A and D). An allorecognition assay against HLA-mismatched PHA blasts showed no difference in alloreactivity between CTL-R and CTL-L (fig. S11B). Spleen weights and splenocyte counts were also higher for CTL-R than CTL-L and tumor-only groups (Fig. 7B and fig. S12A). The spleen, peripheral blood, and bone marrow contained more human CD45⁺ (hCD45⁺) cells, and most of these cells were CD3⁺, indicating substantial in vivo expansion of CTL-R T cells (Fig. 7C and fig. S12, B and C). In absolute counts, presence of CD19⁺ cells, indicative of residual LCL, was slightly higher in the PBMCs and spleens of CTL-R recipients (fig. S12C) but not after normalization to the absolute CD19⁺ cell counts of CTL infusion products (mean CD19⁺ cells infused per mouse: 9354 for CTL-L and 103,761 for CTL-R). After this normalization, on the contrary, presence of CD19⁺ cells appeared to be remarkably increased in CTL-L compared to CTL-R (fig. S12D), especially in PBMCs, suggesting presence of circulating tumor cells that were controlled better by CTL-R. CD8⁺ T cells expanded initially more in mice receiving CTL-R but CD8⁺/CD4⁺ T cell ratios returned to preinfusion levels at later time points, whereas CD8⁺ T cells dominated throughout in mice receiving CTL-L group (Fig. 7C and fig. S12, E and F). Thus, both long-term and rapidly expanded EBV CTLs efficiently control tumor growth, but T_SCM-enriched CTLs generate more CD4⁺ and CD8⁺ T cells and persist in vivo.

Fig. 7. Expansion of CTL-R in vivo.

(A) In vivo WBC expansion dynamic using flow cytometry of weekly bleedings. (B) Splenocyte total counts after euthanasia. (C) Proportions of hCD45⁺ cells, CD3⁺ and CD19⁺ among hCD45⁺ cells, and CD4⁺ and CD8⁺ among human CD3⁺ in the PBMC and spleen using flow cytometry. (D) Multiplex analysis of human cytokines in the murine sera collected after euthanasia. For (A) to (C), means with SD; mixed-effects analysis [(A) and (D)], one-way ANOVA (B), and multiple unpaired t test (C), α = 0.05, *P < 0.05, **P < 0.005, ***P < 0.001, and ****P < 0.0001.

T_SCM-enriched EBV CTLs infiltrate tumors and have broad antigen specificity

To determine the specificity of EBV-CTLs expanded in vivo, we measured tumor infiltration of CD8⁺ and CD4⁺ T cells and Granzyme B (GrB) expression by immunohistochemistry (Fig. 8, A and B, and fig. S13A). There was a trend towards higher infiltration of both CD8⁺ and CD4⁺ T cells and higher expression of GrB in the CTL-R group compared to CTL-L. CD4 and GrB costaining showed that expression of GrB among CD4⁺ varied in both CTL-R and CTL-L groups, although low counts of CD4⁺GrB⁺ TILs in both groups indicated that the majority of detected GrB was likely localized in CD8⁺ cells (fig. S13, B and C). Flow cytometry revealed a significantly higher CD3⁺ T cell infiltration and particularly a higher infiltration of CD4⁺ T cells into tumors of mice receiving transfer of CTL-R compared to CTL-L (Fig. 8, C and D).

Fig. 8. Specificity of expanding EBV-CTLs in vivo.

All samples were isolated after euthanasia. Immunohistochemistry analysis of tumor infiltrating lymphocytes: (A) representative micrographs of CD8 staining in tumors of CTL-L, CTL-R, and tumor-only samples, respectively (10×); CD8⁺ cells are stained in brown, and nuclei are stained in blue; (B) proportions of tumor infiltrating CD8⁺ and CD4⁺ lymphocytes by treatment group (pooled data from available samples from two experiments, 10 mice per group). (C) Proportions of CD3⁺ versus CD19⁺ among hCD45⁺ cells in TILs and (D) proportions of CD4⁺ versus CD8⁺ among CD3⁺ T cells, measured by flow cytometry. (E) A shift of CD8⁺ and CD4⁺ T cell memory phenotypes in TILs versus the spleen (pooled data of available samples). (F) Proportions of EBV-specific MHC class I multimer–stained CD8⁺ T cells. (G) In vitro short-term cytotoxicity of bulk, separated CD4⁺ and CD8⁺ splenocytes against autologous LCLs (pooled data from all available samples). (H) Presence of specific response of bulk CTL-L and CTL-R cultures before injection versus CTL-R bulk, CD4⁺ and CD8⁺ splenocytes after euthanasia to stimulation with single EBV protein antigen pools measured by ELISpot (based on mean data from all available samples collected from one experiment per one donor). For (B) to (G), means with SD; one-way ANOVA test [(B) and (G)], multiple unpaired t test [(C) to (E)], and two-way ANOVA (F) were used. *P < 0.05, **P < 0.005, and ***P < 0.001.

CTL-R-derived TILs showed a less differentiated phenotype with more T_CM compared to CTL-L (Fig. 8E), although T_SCM TILs were rare in both groups. Tumor-infiltrating T cells of both groups were more differentiated with a predominance of T_EM compared to T cells not derived from tumors, with high proportions of T_TM in the spleen and other organs (Fig. 8E and fig. S14, A to C). A higher proportion of CTL-R–derived antigen–specific TILs was detected via specific EBV peptide–loaded MHC class I multimers compared to the CTL-L group and CD8⁺ T cells in the spleen from both groups (Fig. 8F).

Both CD4⁺ and CD8⁺ splenocytes derived from CTL-R showed short-term in vitro cytotoxicity against EBV-LCLs (there was insufficient expansion in the CTL-L group to perform similar experiments) (Fig. 8G). ELISpot assays with peptides from 14 different EBV proteins revealed a remarkable increase in antigen coverage in CTL-R–derived splenocytes, consistent with in vitro data, and particularly CD4⁺ T cells (Fig. 8H and fig. S15). Thus, transferred EBV-specific T_SCM-enriched CTLs (CTL-R) showed robust proliferation and longevity and reconstituted a wide antigen diversity of different T cell compartments.

DISCUSSION

Adoptive T cell therapy (ACT) therapeutically transfers specific T cell immunity to patients. Persistence of memory T cell subsets in the recipients is often critical for long-term efficacy but difficult to attain. Very-early-differentiated T_SCM cells, characterized by high self-renewal, engraftment, and persistence, can reconstitute all types of effector and memory T cell subsets (45) and show encouraging results in T cell therapy (8, 10, 11).

T_SCMs might be also a promising avenue for treatment of viral infections such as EBV in transplant patients. Antiviral ACT is an active field with several ongoing phase 3 trials (clinicaltrials.gov: NCT03394365, NCT04832607, NCT04832607, and NCT04832607). Although current EBV PTLD treatments through ACT have overall response rates around 70% (27) and with prophylactic efficacy approaching 100% (30) in certain indications such as in patients with EBV-driven PTLD receiving third-party CTLs (18) or patients with latency type II lymphoma, EBV continues to pose a substantial challenge, with failure to respond observed in up to 60% of patients (22, 23).

To test T_SCMs for ACT against EBV infections, T cell priming and several steps of cell sorting can be used, but the complexity of this method makes it difficult to translate to the clinical setting (46). Moreover, CD4⁺ T cells are depleted, which might impair sustaining adoptive immunity (47). In addition, particularly in the context of EBV diseases, lytic antigens were found to be the immunodominant targets of CD4⁺ T cells (48).

Here, we developed a method to enrich both CD4⁺ and CD8⁺ T_SCM EBV-CTL from PBMCs with minimal cell handling steps. We did not use the common stimulation with autologous EBV-LCLs (18), which requires 4 to 8 weeks and yields predominantly late-stage T_EMs (30, 49). Continuous restimulation of T cells can promote T cell exhaustion (33) defined as a reduced functional capacity (50). Instead, we adapted a rapid expansion protocol that is safe and effective in transplant patients (37) and yields a higher proportion of T_CMs (34), which are superior to T_EMs in antiviral activity and persistence (51, 52). Because of its minimal cell handling steps, this protocol can be easily transferred to the clinic. The overall yield is lower than for long-term stimulation, but this is not a limiting factor because sufficient starting numbers of PBMCs can be obtained with standard blood donations.

We systematically tested various cytokines and other conditions to maximize the proportion of T_SCMs and found that a combination of IL-4 and IL-7 with induction of the Wnt/β-catenin pathway using TWS-119 triggered efficient enrichment or EBV-specific T_SCMs with broad coverage of antigens including lytic antigens. T cells specific for lytic antigens can be relevant for treatment of diseases such as nasopharyngeal carcinoma (53–55), EBV⁺ lymphomas (56), and gastric adenocarcinoma (57) in a therapeutic setting as well as prophylaxis of EBV-associated B cell malignancies and nasopharyngeal carcinoma, as the lytic phase of EBV contributes to oncogenesis (58–60). Furthermore, there is convincing evidence for the presence of lytic antigens such as BZLF1 (BamHI Z fragment leftward open reading frame 1) protein in PTLD cases following childhood solid organ transplant (61, 62), which indicates that lytic antigen–specific T cells might be of importance also for PTLD prophylaxis or treatment. While LCL-induced EBV-CTLs are highly polycloncal as LCLs express the entire range of latent antigens, specificities for lytic antigens are difficult to obtain with this method (63). In our protocol, rapidly expanded EBV-CTLs (EBV CTL-R) were expanded with the EBV peptide mix comprising a wide range of lytic and latent antigens. As a result, they demonstrated a robust ability to reconstitute broad multiantigen specificity in response to in vitro and in vivo antigen challenges. A wide diversity of antigen specificity not only broadens the scope of cell therapy but also reduces the risk of relapse due to antigen escape (64). Thus, the relevance of lytic antigen specificities in a therapeutic product may vary depending on the type of EBV-associated malignancy, underscoring the need for a differentiated approach in different clinical contexts. Moreover, lytic EBV antigen–specific CD8⁺ T cells require early memory differentiation with maintained CD27 expression for their protective function (65), and this is ensured by our rapid expansion protocol.

A key aspect of an ACT product is the balance between CD4⁺ and CD8⁺ T cell populations (66, 67), and a high CD4⁺ T cell proportion is associated with better responses to anti-EBV ACT for treatment of PTLD (31). In contrast to the widely used long-term stimulation with low yields of CD4⁺ T cells, our protocol enriched CD4⁺ and CD8⁺ EBV-specific T cells in all memory populations. The high CD4⁺ T cell proportion was maintained after adoptive transfer and was reflected in tumor infiltration, and CD4⁺ T cells recovered a broad antigen-specific profile in vivo. These EBV-specific CD4⁺ T cells may contribute to control of various EBV diseases.

Limitations of this study include lack of long-term data beyond 4 weeks in mice. High human T cell engraftment in murine organs may lead to xeno-GvHD (xenograft-versus-host disease) (68), signs of which (e.g., weight loss) were observed in some mice in the CTL-R group. Nevertheless, the efficient CTL-R infiltration into tumors, the in vitro cytotoxicity, and the specific responses of splenocytes indicated high specificity of the expanded T cells. Moreover, it is essential to emphasize that all factors associated with the mouse model may not readily translate into potential challenges in a clinical setting. This is further substantiated by the observation that CTL-R did not exhibit heightened alloreactivity in comparison to CTL-L in in vitro studies.

Our protocol lays the foundations for the first use of virus-specific T_SCMs in ACT to treat viral complications; however, several challenges still lie ahead. One of these challenges is the tracking of infused T cells in patients. Unlike CAR T cells, which have genetically integrated markers enabling single-cell tracing, VSTs are typically expanded without genetic manipulations. While the recently published method for polyclonal antigen-specific T cell-targeted genome editing (TarGET) allows for precise genome marking of EBV CTL-R, upscaling for ACT manufacturing presents considerable difficulties (69). One available method for tracking these cells would involve TCR sequencing to track the infused TCR clones and monitor their expansion (62). However, as demonstrated in our previous TarGET study, despite thousands of epitope-TCR pairs available in databases, assigning EBV epitopes to only a small proportion of found TCRs in specific donors is currently possible. Furthermore, most studies predominantly focus on MHC class I epitopes (70, 71), whereas MHC class II epitopes are also relevant for EBV (72, 73). Large-scale studies dedicated to the analysis of EBV TCRs, along with the prediction and validation of TCR-epitope pairs, are needed to address this challenge effectively. One potential approach to tackling this issue involves conducting comprehensive TCR sequencing analysis of sorted MHC class I–and MHC class II–bound tetramer populations. In addition, the TarGET method might also be used to generate immunosuppression-resistant EBV CTL-R based on a previously published approach (74). Last, while early memory phenotype and low exhaustion increase the persistence potential of the infused T cells, it is important to consider using serial ACT product infusions to mitigate potential rejection by the host’s immune system in case of third party adoptive transfer (18).

In conclusion, we demonstrate that our protocol yields promising EBV-specific T_SCM-enriched CTLs with favorable properties for adoptive transfer, namely, early-differentiated memory composition, low exhaustion, high tumor infiltration, efficient CD4⁺ and CD8⁺ T cell–mediated cytotoxicity, long-term persistence, and broad antigen specificity. This may pave the way for the next generation of unmodified antigen-specific cell therapies against viral infections. The safety and efficacy as well as the clonal diversity of these VSTs remain to be investigated in an upcoming clinical trial.

MATERIALS AND METHODS

Experimental design

Blood was obtained after informed consent from healthy donors in accordance with the Declaration of Helsinki. The study was approved by the local ethic committee (Ethikkommission Nordwest- und Zentralschweiz, Project ID PB_2018-00081). Seropositive donors were typed for HLA class I and class II alleles. Human PBMCs were isolated from EDTA blood of healthy donors (75). The donors for whose common HLA alleles (e.g., A*02:01, A*03:01, A*11:01, B*07:02, B*08:01, B*35:01, DRB1*01:01, and DRB1*01:07) we could find corresponding MHC class I and MHC class II multimers (see table S1 for the list of multimers used) were chosen for the majority of experiments.

Long-term EBV-CTL expansion with LCL restimulations and rapid expansion protocols were adapted from previously described protocols (28, 34). All T cells were expanded in CTL medium (CTL-M).

For rapid expansion, PBMCs were cultured in a G-Rex bioreactor (Wilson Wolf). A total of 3 × 10⁶ PBMCs per well of a 24-well G-Rex plate or 1.5 × 10⁷ per well of a 6-well G-Rex plate were cultured. On day 0, cells were pulsed overnight in CTL-M (or CTL-M with high K⁺ when applicable) containing the EBV pepmix and supplemented with cytokines (and TWS-119 when applicable). Afterward, the pepmix (and TWS-119 if applicable) was diluted 5× with CTL-M supplemented only with cytokines. Cell culture went on up to days 10 to 12 without further supplementation. See table S2 for cytokine, pepmix, cytokine, and other reagent concentrations used for in vitro experiments).

For long-term EBV-CTL expansion, PBMCs were stimulated with autologous EBV-LCLs at effector:target (E:T) = 40:1 for 10 days (2 × 10⁶ PBMCs per well of a 24-well cell culture plate) without cytokine supplementation. Afterward, T cells were restimulated weekly at E:T = 4:1 and supplemented with IL-2 (20 U/ml) 3×/week until days 28 to 35 (28).

Animal experiments were conducted according to the license approved by the veterinary office of the canton of Zürich, Switzerland (ZH049/20). NSG [NOD.Cg-Prkdc^scid Il2rg^tm1Wjl/SzJ (#005557)] or NSG-A2 [NOD.Cg-Mcph1^{Tg(HLA-A2.1)1Enge} Prkdc^scid Il2rg^tm1Wjl/SzJ (#009617)] mice were purchased from the Jackson Laboratory and bred and housed under specific pathogen–free conditions at the Laboratory Animal Services Center of the University of Zürich. Experiments were initiated at 6 to 12 weeks of age. The mouse models were adapted from previous studies (42, 43). LCL tumor cells were injected subcutaneously into the left flank under isoflurane narcosis. A total of 2 × 10⁶ tumor cells were resuspended in phosphate-buffered saline (PBS) and right before injection mixed in a 1:1 (v/v) ratio with Corning Matrigel Growth Factor Reduced Basement Membrane Matrix. Three days after tumor injection, 1 × 10⁷ T cells were adoptively transferred by tail vein injection. T cell expansion was supported by intraperitoneal injection of 10⁵ IU of recombinant human IL-2 (3×/week, starting from the day of adoptive T cell transfer in all groups; Peprotech), or as stated otherwise. Tumor size was monitored by calipering (3×/week), and bioluminescent imaging for tumor cells was transformed with a luciferase encoding recombinant EBV strain (a gift of W. Hammerschmidt, Helmholtz Institute Munich, Germany; 2×/week). General health was monitored by weighing and health parameter scoring 3×/week or daily, according to the animal license. Peripheral blood composition and expansion of adoptively transferred T cells were monitored by weekly tail vein bleeding and flow cytometric analysis (see table S3 for the flow panel and fig. S11F for gating strategy) on BD LSRFortessa. WBC counts were determined from full blood with an automatic cell counter (DxH 500, Beckman Coulter). For bioluminescent imaging, mice were injected with 5 μl/g of body weight of VivoGlo Luciferin (15 mg/ml; Promega) and imaged 10 min after injection in an IVIS machine (PerkinElmer) under isoflurane narcosis. Animals were euthanized when they met predefined criteria stated in the animal license or when the control group met the end-point criteria.

Antigen peptides

The EBV pepmix was used for stimulation during EBV-CTL rapid expansion, in vitro challenge, and IFN-γ ELISpot. It contains 43 peptides of 8 to 20 amino acids in length (32 peptides are restricted for MHC class I and 11 peptides for MHC class II). The sequences of these peptides were derived from the 15 different lytic and latent EBV proteins (LMP2a, BRLF1, BMLF1, LMP1, EBNA3C, EBNA3A, EBNA3B, BALF2, BMRF1, BZLF1, BNRF1, EBNA1, gp350, BXLF2, and EBNA2). The peptides are restricted for the following HLA molecules: HLA-A*02, HLA-A*03, HLA-A*11, HLA-A*24, HLA-A*26, HLA-B*07, HLA-B*08, HLA-B*15, HLA-B*18, HLA-B*27, HLA-B*35, HLA-B*40, HLA-B*4402, and HLA-DR. Further details can be found on the manufacture’s web site (www.miltenyibiotec.com/CH-en/products/peptivator-ebv-consensus.html#130-103-462).

Overlapping peptide pools (15 nucleotide oligomers with 11–amino acid overlap) derived from various single EBV antigens (latent: EBNA-LP, EBNA2, EBNA3A, EBNA3B, EBNA3C, and LMP1; lytic: BARF1, BMLF1, BMRF1, BRLF1, BZLF1, and GP350/GP340) (PepMix Collection EBV, JPT Peptide Technologies) were used for T cell stimulation during in vitro challenge, and IFN-γ ELISpot. For multi-VST expansion, the following overlapping 15–nucleotide oligomer peptide pools overlapping by 11 amino acids were used almost entirely on the basis of the previously published protocol (37): EBV-LMP2, EBV-BZLF1, EBV-EBNA1, AdV-Penton, AdV-Hexon, CMV-pp65, CMV-IE-1, BKV-VP1, BKV-large T, HHV6-U54, and HHV6-U90 (all JPT Peptide Technologies). Concentrations used are provided in table S2.

Cell culture medium

All media were sterile-filtered when supplemented with the human serum (HS). T cell lines were expanded in CTL-M medium composed of RPMI 1640 (Sigma-Aldrich) with penicillin-streptomycin (PenStrep) (Gibco) and 5% HS. For CTL-M with elevated K⁺ concentration, powdered NaCl-free RPMI 1640 (Gibco) was reconstituted as previously described (40). Briefly, NaCl (63.4 mM), NaHCO₃ (23.8 mM), and additional KCl (40 mM) had to be added for complete reconstitution. Then, PenStrep and 5% HS were added. LCL cell culture medium (LCM-10) used to generate and culture EBV-LCLs comprised RPMI 1640 (Gibco) with 100× PenStrep (Gibco), 5% HS, 50× GlutaMAX (Gibco), 100× Minimum Essential Medium Non-Essential Amino Acids (Gibco), and 100× sodium pyruvate (Gibco). Freezing medium comprised RPMI 1640 (Sigma-Aldrich) with PenStrep (Gibco), 10% HS, and 20% dimethyl sulfoxide (Sigma-Aldrich).

Generation of multi-VST lines

To generate multi-VSTs, the previously published rapid expansion approach was adopted (37). PBMCs were pulsed for 30 min at 37°C with the following peptide pools: EBV-LMP2, EBV-BZLF1, EBV-EBNA1, AdV-Penton, AdV-Hexon, CMV-pp65, CMV-IE-1, BKV-VP1, BKV-large T, HHV6-U54, and HHV6-U90 (all JPT Peptide Technologies, concentrations provided in table S2). Then, cells were resuspended in 30 ml of CTL-M supplemented with IL-4 (400 U/ml) and IL-7 (10 ng/ml) and cultured for 12 days in a G-Rex10 device (Wilson Wolf).

EBV-LCL generation and culture

EBV-transformed LCLs were generated and cultured in LCM-10 medium according to previously published protocols (76). PBMCs were incubated with recombinant B95-8 or B95-8-fLuc EBV strains (both gifts from W. Hammerschmidt, Helmholtz Center Munich, Germany), were cultured in LCM-10 medium, and were treated with cyclosporin A (2 μg/ml; Sigma-Aldrich) and CpG ODN 2006 (2 μg/ml; InvivoGen) weekly until the transformation. Nonirradiated LCLs were always cultured in LCM-10 medium (including cytotoxicity and outgrowth assays).

Generation of PHA blasts

PBMCs were incubated in CTL-M medium containing IL-2 (100 U/ml; PROLEUKIN) and PHA-P (5 μg/ml; Invivogen) for 7 days. Afterward, formation of blasts was assessed microscopically, cells were harvested and frozen down in freezing medium.

In vitro challenge with autologous EBV-LCLs

After fluorescence-assisted cell sorting (FACS) (staining described below) of EBV CTL-R, sorted cells were allowed to recover for 3 days in CTL-M supplemented with IL-4 and IL-7. Afterward, autologous LCLs were irradiated, and T cells were stained with CellTrace Violet (CTV). Irradiated LCLs were mixed with EBV pepmix and overlapping peptide pools of single immunodominant antigens (JPT Peptide Technologies) (LCL/peptide mix) and added to T cells at a ratio 1:1. Following 7-day culture, cells were harvested and analyzed by flow cytometry and/or ELISpot.

Short-term (killing and allorecognition assay) and long-term (outgrowth assay) in vitro cytotoxicity

Short-term 6-hour killing assay and long-term 4-week outgrowth assay were adopted as previously published (77). Briefly, for killing assay, EBV-CTLs were incubated with target EBV-LCLs at an E:T ratio = 30:1 for 6 hours. In allorecognition assay, HLA-mismatched PHA blasts were used as target cells. Afterward, cells were stained for viability (Zombie Aqua, BioLegend), apoptosis (CellEvent Blue, Thermo Fisher Scientific), and CD3 and CD19 surface markers (see the panels below). Cytotoxicity was calculated according to the following formula: 100 − ([V_test / V_control] × 100), where V = % viable (CellEvent⁻ Zombie Aqua⁻) CD19⁺ cells.

For outgrowth assay (long-term cytotoxicity assay) (77), T cells were incubated with EBV-LCLs at different E:T ratios (effector cells always at concentration 1 × 10⁴ cells per well with serial 2× dilutions of target cell concentrations) in triplicates in 96-well U-bottom cell culture plates for 4 weeks. The readout was the lowest E:T ratio controlling the outgrowth of LCLs, which was determined microscopically and confirmed by flow cytometry.

IFN-γ ELISpot

EBV-responsive T cells were identified by ELISpot assay as previously published (77). For PBMCs, 3 × 10⁵ T cells per well were added to the anti–IFN-γ–coated (1 mg/ml; clone 1-D1K, Mabtech) ELISpot MultiScreen_HTS 96-well filter plates (Merck). Concentrations for expanded CTLs were titrated where necessary between 800 and 1 × 10⁵ cells per well. T cells were stimulated overnight with EBV peptides or EBV-LCLs (see peptide concentrations in table S2), and E:T ratio for stimulation with LCLs was 1:1. Assays were developed using IFN-γ antibody (1 mg/ml; 7-B6-1-Biotin, Mabtech), streptavidin-alkaline phosphatase (Mabtech), and BCIP/NBT (Merck). The spots were counted on the dried plates with AID ELISpot Reader.

Intracellular cytokine staining

Intracellular cytokine staining for flow cytometry detection (78) was done as previously published. The panel is provided in table S8, and the gating strategy is summarized in fig. S16.

V-PLEX

Human cytokine presence in murine blood sera was analyzed using V-PLEX human proinflammatory panel-1 (MSD) and detected by MESO QuickPlex SQ 120 (MSD) system according to the manufacturer’s instructions.

Immunomagnetic cell sorting

CD4⁺ and CD8⁺ T cells were isolated using the MACS CD4⁺/CD8⁺ isolation kit (Miltenyi Biotec) according to the manufacturer’s instructions.

Immunohistochemistry

Tumors were fixed in a 4% paraformaldehyde solution; further sample preparation and immunohistochemistry staining were done commercially by the Pathology Department of the University Hospital of Basel. Slides were acquired on an automated slide scanning bright-field microscope (Vectra), and positive cells were quantified using inForm automated image analysis software (Akoya Biosciences).

Flow cytometry and FACS-based cell sorting

Flow cytometry panels are described in detail in tables S3 to S8, and the gating strategies are outlined in figs. S2, S5 to S7, S11F, and S16. If applicable, red blood cells were lysed using ACK (ammonium-chloride-potassium) lysis buffer until the pellet appeared no longer red. If applicable, then whole-cell staining for proliferation tracing and viability staining were performed in PBS according to the manufacturer’s instructions. Surface staining with antibodies and MHC class I multimers (if applicable) was performed in FACS buffer (5% fetal bovine serum and 0.1% NaN₃ in PBS). For intracellular staining, cells were fixed with fixation buffer (BioLegend, 420801) and stained for intracellular markers in the permeabilization buffer (BioLegend, 421002) according to the manufacturer’s instructions. For combined intracellular/intranuclear staining, cells were fixed and permeabilized using transcription factor buffer set (BD Biosciences, #562574) according to the manufacturer’s instructions.

Spectral flow cytometry was performed on Cytek Aurora. FACS was performed with BD FACSMelody. Weekly bleedings of mice were analyzed with BD LSRFortessa. Data were analyzed using FlowJo software. FlowSOM algorithm was used to define memory T cell populations: T_SCMs as CD45RA⁺CD45RO⁻CD62L⁺CD27⁺, T_CMs as CD45RA⁻CD45RO⁺CD62L⁺CD27⁺, T_TMs as CD45RA⁻CD45RO⁺CD62L⁻CD27⁺, T_EMs as CD45RA⁻CD45RO⁺CD62L⁻CD27⁻, and T_EMRAs as CD45RA⁺CD45RO⁻CD62L⁻CD27⁻.

Statistical analysis

Analyses were conducted using Prism software (GraphPad). Data of individual donors are shown as representative experiments or medians with SDs. Combined data of different donors are given as medians with range.

Supplementary Materials

This PDF file includes:

Figs. S1 to S16

Tables S1 to S8