iPSC-Derived Neoantigen-Specific CTL Therapy for Ewing Sarcoma

Ewing sarcoma is the second most common bone and soft-tissue malignant tumor in adolescents and young adults (1). Ewing sarcoma can be diagnosed by confirming the presence of the fusion gene EWS/FLI1, which acts as a transcriptional activator with tumorigenesis (1–3). Around 70% of Ewing sarcomas are localized at diagnosis, with 30% being metastatic (4–6). Metastatic Ewing sarcoma has a poor prognosis (5-year disease-free survival rate, 25%; refs. 4–6). Although multidisciplinary treatment combining surgery, radiotherapy, and chemotherapy is effective in localized Ewing sarcoma, metastatic Ewing sarcoma is extremely treatment refractory (4–8). Effective salvage therapy is not yet established (4–8), and thus, new approaches are needed.

In CTL therapy, CTLs are generated from a patient, expanded ex vivo, and returned to the patient. This is reportedly successful against selected tumors such as metastatic melanoma (9, 10). However, such CTLs often are ineffective (exhausted) due to chronic exposure to tumor antigens (11). To address this issue, we previously developed a system that reprograms peripheral blood–derived antigen-specific CTLs into induced pluripotent stem cells (T-iPSC) and redifferentiates T-iPSCs into functionally rejuvenated CTLs (rejT; ref. 12). Human immunodeficiency virus 1–specific rejTs produced with this method had the same antigen specificity as the original CTLs, only with higher proliferative capacity; they were functionally rejuvenated. Furthermore, their telomeres were longer than those of the original CTLs (12). Epstein–Barr virus (EBV)–specific rejTs have been shown to have a strong and long-term antitumor effect against EBV-associated lymphomas in vivo (13, 14), surviving as stem cell and central memory T cells (15). This approach is not restricted to hematologic malignancy. Human papilloma virus type 16 (HPV16)–specific rejTs have been demonstrated to more strongly suppress cancer proliferation and had better survival than original CTLs, demonstrating that HPV16-rejT therapy is a promising treatment even against epithelial cancer (16).

Fusion genes that include Ewing sarcoma breakpoint region 1 (EWS), a member of the FET family genes, are oncogenes of the Ewing sarcoma family of tumors, which includes Ewing sarcoma, primitive neuroectodermal tumors, biphenotypic sarcoma, and Askin tumors. EWS/Friend leukemia integration 1 (FLI1), a gene of the erythroblast transcription specific family of transcription factors is the most frequently encountered fusion gene, occurring in approximately 90% of cases (7,8). Although the breakpoint sequences of each phenotype are different, 60% of EWS/FLI1⁺ sarcomas harbor the same breakpoint sequence, in which EWS exon 7 is fused to FLI1 exon 6. The altered amino acid sequence at the break/fusion point of the EWS/FLI1 protein encoded by the fusion gene is a neoantigen presented by HLA molecules. It evokes an immune response that selectively targets EWS/FLI1⁺ sarcomas. The EWS/FLI1 neoantigen is not found in normal cells, making it a good target for CTL therapy (17). In 2012, Evans and colleagues modified the HLA-A*02:01–restricted peptide sequence of the EWS/FLI1 fusion protein to stimulate generation of EWS/FLI1-specific CTLs (EWS/FLI1-CTL) in healthy subjects. This successfully yielded EWS/FLI1-CTLs (18). Using the modified peptide, we generated iPSC-derived functionally rejuvenated EWS/FLI1-CTLs (EWS/FLI1-rejT). EWS/FLI1-rejTs proliferated more in vitro than did original EWS/FLI1-CTLs, were of stem cell memory phenotype, and both in vitro and in a mouse Ewing sarcoma xenograft model were effective in killing Ewing sarcoma. Thus, EWS/FLI1-rejTs could be a potential therapy for Ewing sarcoma.

HLA-A*02:01⁺ Ewing sarcoma cell line TC-71 with type 1 fusion (EWS exon 7 fused to FLI1 exon 6), the kind gift in 2017 of Dr. Melinda Merchant (Memorial Sloan Kettering Cancer Center, Manhattan, NY), and HLA-A*24:02⁺ Ewing sarcoma cell line RD-ES with type 2 fusion (EWS exon 7 fused to FLI1 exon 5) were cultured in RPMI1640 medium (Sigma-Aldrich) supplemented with 10% FBS (Gibco Life Technology) and 1% penicillin/streptomycin-glutamine (PSG; Thermo Fisher Scientific). To evaluate the antitumor effect of EWS/FLI1-rejTs on Ewing sarcoma, TC-71 cells were transduced with a retrovirus encoding both firefly luciferase (FFluc) and GFP to permit monitoring of tumor volume. GFP-expressing cells were sorted by flow cytometry and maintained in RPMI supplemented with 10% FBS and 1% PSG. HLA-A*02:01⁺ Ewing sarcoma cells (line A673) with type 1 fusion were cultured in DMEM high glucose medium (DMEM; Sigma-Aldrich) supplemented with 10% FBS and 1% PSG. All cell lines were obtained from ATCC holdings, with purchase of the RD-ES and A-673 cell lines in 2018 and 2020, respectively. All cell lines were validated as Mycoplasma free (MycoAlert Mycoplasma Detection Kit, Lonza). Cell line authentication was not performed, but in each cell line, the EWS/FLI1 type 1 fusion or type 2 fusion was confirmed. Low-passage cells (<10 generations) were used for experiments. For rejT differentiation from T-iPSCs, C3H10T1/2 feeder cells (RIKEN BRC) and C3H10T1/2 feeder cells that expressed the human homologues of two Notch Delta ligands delta-like1 (DL1)/delta-like4 (DL4) were used. C3H10T1/2 cells were transduced with a retrovirus encoding both DL1 and DL4. These feeder cells were cultured in minimum essential medium α (MEM-α; Thermo Fisher Scientific) supplemented with 10% FBS and 1% PSG. The HEK 293T cell line (RIKEN BRC) was maintained in DMEM supplemented with 10% FBS and 1% PSG and was used for the retrovirus production.

The retroviral supernatant was generated from 293T cells transfected with the retroviral constructs Peg-Pam, encoding gag-pol, and RDF, encoding RD114 envelope, using GeneJuice transfection reagent (Sigma-Aldrich). Supernatants were collected 48 hours after transfection (19). Cell lines were inoculated into 6-well culture plates (3 × 10⁵ cells/well) and cultured in corresponding complete medium overnight. Cultured medium was replaced by the retroviral supernatant (2 mL/well) and 1 μg of polybrene (Santa Cruz Biotechnology) was added. Transduced cells were trypsinized with 0.1% trypsine (Sigma-Aldrich) at 37°C for 5 minutes when the cells became confluent, and FFluc/GFP-transduced TC-71 cells were selected for GFP-expressing cells by flow cytometry (20). DL1- and DL4-transduced C3H10T1/2 feeder cells were selected with hygromycin (200 μg/mL; Thermo Fisher Scientific) and G418 disulfate salt (300 μg/mL; Sigma-Aldrich).

We recruited HLA-A*02:01⁺ donors because the epitope which is specific to the neoantigen encoded by EWS/FLI1 is HLA-A*02:01 restricted (18). We chose healthy adults, rather than adolescents, with Ewing sarcoma as donors to avoid the potential burden of taking blood from ill persons (100 mL blood/initial step in EWS/FLI1-CTLs generation). The experimental protocol was approved by the Research Ethics Committee for the Faculty of Medicine, Juntendo University (Bunkyo-ku, Tokyo, Japan), and was in accordance with the Declaration of Helsinki. Peripheral blood mononuclear cells (PBMC) from a single HLA-A*02:01⁺ healthy donor were obtained with written informed consent.

EWS/FLI1-specific CTLs were generated as reported previously (21–24). To generate dendritic cells (DC), PBMCs were obtained from the same donor, plated for 3 hours, then adherent cells were cultured for 5 days in DC medium (CellGenix) supplemented with IL4 (1,000 U/mL) and GM-CSF (800 U/mL; both Miltenyi Biotec). Immature DCs were then cultured for another 2 days with a cytokine cocktail containing IL1β (10 ng/mL), IL6 (100 ng/mL), TNFα (10 ng/mL; all Miltenyi Biotec), and prostaglandin E2 (PGE2, 1 μg/mL; Cayman Chemical) for maturation. PBMCs were cocultured with DCs (PBMCs 20 × 10⁶, DCs 1 × 10⁶) loaded with A*02:01/YLNPSVDSV (18) custom-synthesized peptide (Mimotopes) at a concentration of 100 ng/20 × 10⁶ PBMCs in the presence of IL4 (400 U/mL) and IL7 (10 ng/mL; both Miltenyi Biotec). On day 16, cocultured cells were harvested and antigen specificity against YLNPSVDSV was determined by staining with phycoerythrin (PE)-conjugated HLA-A*02:01/YLNPSVDSV MHC pentamer (ProImmune). The generated T cells stained with the pentamer were separated by fluorescence-positive cell sorting using BD FACSAria II equipment (BD Biosciences) and were single-cell cloned (EWS/FLI1-CTL clone).

CTLs were single cell sorted by flow cytometry and cultured in round-bottom 96-well plate. Each cell was mixed with 1.0 × 10⁵ feeder cells (50 Gy-irradiated allogenic PBMCs), stimulated with phytohemagglutinin (PHA)-L (5μg/mL; Sigma-Aldrich), and cultured in NS-A2 CTL medium (Nissui) supplemented with 1% PSG in the presence of IL2 (100 IU/mL; Miltenyi Biotec) for 2 weeks (12). Each single cell–derived clone was stimulated and expanded every 2 weeks. After several fortnightly stimulations, each CTL clone was stained for the PE-conjugated HLA-A*02:01/YLNPSVDSV major MHC pentamer and its antigen specificity was determined.

The remaining sorted pentamer-positive T cells were bulk cultured without single-cell sorting, aliquoted, and frozen (−80°C, 10 × 10⁶ T cells/mL) in TC protector freezing medium (KAC) for further studies (EWS/FLI1-bulk CTLs). Generated EWS/FLI1-specific CTL clones were transduced with reprogramming factors (Oct3/4, SOX2, Klf4, and c-Myc) and SV40 large T antigen via Sendai virus vectors as described previously (13–16). Established T-iPSCs were maintained in culture with essential 8 medium (Gibco).

T-iPSCs were differentiated into EWS/FLI1-specific rejTs as described previously (13–16). Briefly, small clumps of T-iPSCs were transferred onto C3H10T1/2 cells to induce differentiation of human iPSCs into hematopoietic cells. T-iPSCs and C3H10T1/2 cells were cocultured for 2 weeks in Iscove's modified Dulbecco's medium (Sigma-Aldrich) supplemented with 15% FBS (HyClone, GE Healthcare) and a cocktail of human insulin (10 mg/mL), human transferrin (5.5 mg/mL), sodium selenite (5 ng/mL), 2 mmol/L l-glutamine (all Thermo Fisher Scientific), 0.45 mmol/L α-monothioglycerol (Wako Pure Chemicals), and ascorbic acid (50 mg/mL) in the presence of VEGF (20 ng/mL; Miltenyi Biotec). Hematopoietic cells harvested from iPSC sac contents were transplanted onto DL1/4-expressing C3H10T1/2 feeder cells to undergo T-lineage differentiation. Hematopoietic cells and feeder cells were cocultured in MEM-α (Thermo Fisher Scientific) supplemented with both 20% FBS (HyClone, GE Healthcare) and 1% PSG (Thermo Fisher Scientific) in the presence of recombinant human stem cell factor (20 ng/mL; Miltenyi Biotec), Fms-related tyrosine kinase 3 ligand (10 ng/mL; Miltenyi Biotec), and IL7 (10 ng/mL). Four weeks after coculture began, the floating T-lineage cells were collected and stimulated with PHA-L (5μg/mL; Sigma-Aldrich) and 2.5 × 10⁶/mL of irradiated PBMCs in NS-A2 CTL medium (Nissui) supplemented with 1% PSG in the presence of IL7 (10 ng/mL) and IL15 (10 ng/mL; Miltenyi Biotec). Two weeks after stimulation, EWS/FLI1-rejTs were aliquoted and frozen (−80°C, 10 × 10⁶ T cells/mL) in TC protector freezing medium for further studies.

Used to examine T-cell subsets and antigen specificities via flow cytometry (as described below), were PE-conjugated HLA-A*02:01/YLNPSVDSV MHC pentamer, allophycocyanin (APC)/cyanin 7 (Cy7)–conjugated mouse anti-human CD3, clone HIT3a, APC-conjugated mouse anti-human CD4, clone OKT4, Pacific Blue-conjugated mouse anti-human CD8α, clone RPA-T8, (all BioLegend). Used to define memory phenotype subsets via flow cytometry (as described below) were Alexa Fluor 700–conjugated mouse anti-human CD3, clone UCHT1, V500-conjugated mouse anti-human CD8, clone RPA-T8 (both BD Biosciences); Pacific Blue–conjugated mouse anti-human CD45RA, clone HI100, PE-conjugated mouse anti-human CD62L, clone DREG-56, APC/Cy7-conjugated mouse anti-human CD95, clone DX2, FITC-conjugated mouse anti-human CXCR3, clone G025H7, APC/Cy7-conjugated mouse anti-human KLRG1, clone 2F1/KLRG1 (all BioLegend); APC-conjugated mouse anti-human CD27, clone L128, FITC-conjugated mouse anti-human CD28, clone CD28.2, and Brilliant Violet 421–conjugated mouse anti-human CD127, clone HIL-7R-M21 (all BD Biosciences). Used to examine expression of exhaustion markers via flow cytometry (as described below) were APC-conjugated mouse anti-human lymphocyte-activation gene-3 (LAG-3), clone 7H2C65, APC-conjugated mouse anti-human programmed death-1 (PD-1), clone EH12.2H7, and APC-conjugated mouse anti-human T-cell immunoglobulin mucin-3 (TIM-3), clone F38-2E (all BioLegend). Used to examine expression of HLA-02, PD-L1 (programmed death ligand-1), and PD-L2 (programmed death ligand-2) via flow cytometry (as described below), were APC-conjugated mouse anti-HLA-A02, clone BB7.2, APC-conjugated mouse anti-human PD-L1, clone 29E.2A3, and APC-conjugated mouse anti-human PD-L2, clone MIH18 (all BioLegend).

Flow cytometry was performed on BD FACSAria II or BD LSRFortessa equipment (BD Biosciences). The acquired data were analyzed using FlowJo software 10.5.3 (Tree Star). To verify specificity against EWS/FLI1, EWS/FLI1-CTL clone, EWS/FLI1-bulk CTLs, and EWS/FLI1-rejTs were analyzed. From EWS/FLI1-CTL clone and EWS/FLI1-rejTs, memory phenotype and expression of exhaustion markers were evaluated, as well as the expression of HLA-A2, PD-L1, and PD-L2 on Ewing sarcoma cell lines (TC-71, A673).

We also studied the persistence of EWS/FLI1-rejTs in vivo. Mouse peripheral blood samples were collected into heparinized tubes (Terumo) by cheek puncture under anesthesia. Red blood cells were removed using ammonium chloride potassium erythrocyte lysing buffer (NH₄Cl 8,024 mg/L, KHCO₃ 1,001 mg/L, EDTA Na₂·2H₂O 3.722 mg/L) according to the manufacturer's instructions (Lonza). The single-cell suspensions were incubated with antibody cocktail for 30 minutes at 4°C in the dark. The cells were then washed once with 6 mL PBS. In all analyses, propidium iodide (Thermo Fisher Scientific) was used to gate live cells. The fluorescence-minus-one method was used to interpret flow cytometry data for all antibody combinations. Representative gating is shown in Supplementary Fig. S1.

EWS/FLI1-CTL clones and EWS/FLI1-rejTs (1 × 10⁶ cells each) were separately plated in triplicate on 24-well culture plates with 1 mL NS-A2 supplemented with 1% PSG. Every 2 weeks, cells were counted and stimulated with PHA-L (5 μg/mL) mixed with 2.5 × 10⁶/mL of irradiated PBMCs in the presence of IL7 (10 ng/mL) and IL15 (10 ng/mL) for 10 weeks.

RNA was extracted from TC-71, A673, and RD-ES cells using TRIzol RNA isolation reagents (Thermo Fisher Scientific). First-strand cDNA synthesis was performed using 5 μg of RNA and the ReverTra Ace qPCR RT Master Mix with gDNA Remover (Toyobo). RT-PCR analyses assayed the presence of EWS/FLI1 type 1 fusion or type 2 fusion using PrimeSTAR GXL Premix (Takara Bio). The human EWS/FLI1 primer sequences were 5′-CCAAGTCAATATAGCCAACAG -3′ and 5′-GGCCAGAATTCATGTTATTGC -3′ (3). PCR cycle conditions were 30 cycles of 98°C for 10 seconds, 60°C for 15 seconds, and 68°C for 30 seconds. Purified PCR products for sequencing were obtained from gels using a NucleoSpin Gel and PCR Clean-up (Takara Bio). DNA sequencing was performed by FASMAC using BigDye Terminator v3.1 (Thermo Fisher Scientific). Both forward and reverse strands were sequenced. The reactions were run on an Applied Biosystems 3730xl DNA Analyzer (Thermo Fisher Scientific). The sequencing data were analyzed with SnapGene version 4.3.0 software (GSL Biotech) and are shown in Supplementary Table S1.

The cytotoxicity of EWS/FLI1-CTLs and EWS/FLI1-rejTs against TC-71 and RD-ES, as analyzed by ⁵¹Cr release assays, was determined at different effector-to-target ratios (E:T; 20:1, 10:1). Target cells were labeled with ⁵¹Cr (PerkinElmer) for 1 hour, and effector cells were cocultured with the target cells for 5 hours. After 5 hours of incubation, supernatant was collected and radioactivity was analyzed with 1450 MicroBeta TriLux (PerkinElmer). Percentage of cytotoxicity was calculated as [(experimental release − spontaneous release)/(maximum release − spontaneous release)] × 100 (%).

Cytotoxicity assays using xCELLigence Real-Time Cell Analyzers (RTCA; ACEA Biosciences) were performed using cocultured effector cells with target cells as described previously (16, 25–27). Briefly, 2 × 10⁴ target cells (TC-71) or 3 × 10⁴ cells (A673) were seeded per well, and effector T cells (E:T ratio, 1:1) were added after target-cell adhesion and proliferation was confirmed by identification of a continuous increase in cell index (CI) of target cells according to the manufacturer's instructions (ACEA Biosciences; ref. 25). In each experiment, epitope-mismatched rejTs were used as control rejTs. The electrical impedance by adherent cells in each well was evaluated and recorded automatically, as CI, by RTCA every 15 minutes for up to 45 hours, since tumor cells in control wells became confluent within that time period. Data were processed using the xCELLigence RTCA software package (version 2.0) with results reported and graphed as CI values (abscissa; time in coculture, ordinate) normalized to 1 at the time effector T cells were added. Detergent (0.2% TritonX-100, Thermo Fisher Scientific) was added to some wells to determine values with complete cell lysis (CI_max). Percentages of lysis were calculated as [(CI_{no effector} − CI_effector)/(CI_{no effector} − CI_max)] × 100 (%).

All in vivo studies were approved by the Animal Research Committees of Juntendo University School of Medicine (Bunkyo-ku, Tokyo, Japan). Six-week-old female NOD/Shi-scid, IL2R γKO Jic (NOG) mice (In-Vivo Science) were engrafted intravenously with FFluc/GFP TC-71 cells (1 × 10⁵ cells/mouse, n = 22). Tumor proliferation was monitored using an in vivo imaging system (IVIS; Caliper Life Sciences). Four days after tumor inoculation, mice were divided into four groups (three treatment groups and one no-treatment group) and were treated intravenously with EWS/FLI1-bulk CTLs (n = 6), EWS/FLI1-CTL clone cells (n = 2), or EWS/FLI1-rejTs (n = 7; 5 × 10⁶ cells, three times, once a week). Because EWS/FLI1-CTL clones did not proliferate well, we could treat only 2 mice, and we therefore also included mice treated with EWS/FLI1-bulk CTLs.

Six-week-old female NOG mice were inoculated intraperitoneally with FFluc/GFP TC-71 cells (1 × 10⁵ cells/mouse, n = 17). Four days later, mice were divided into four groups (three treatment groups and one no-treatment group) and were treated intraperitoneally with EWS/FLI1-bulk CTLs (n = 5), EWS/FLI1-CTL clone cells (n = 1), or EWS/FLI1-rejTs (n = 6; 5 × 10⁶ cells, three times, once a week). Because EWS/FLI1-CTL clones did not proliferate well, we could treat only one mouse.

Firefly d-luciferin substrate (OZ Biosciences) was intraperitoneally injected into mice 15 minutes before imaging. Tumor proliferation was monitored by bioluminescence using IVIS. Living Image software version 4.7.2 (PerkinElmer) was used for bioluminescence analyses. Signal intensity was measured as total photon/second/cm²/steradian (p/s/cm²/sr) as described previously (13, 15, 16, 19). Body weight of all mice was tracked throughout. A mouse from each group was sacrificed 20 days after the start of treatment when the mouse with no treatment reached endpoint. Endpoint was defined by morbidity or 20% weight loss based on institutional guidelines. The sacrificed mice were photographed, and their tissues were used for histopathologic and IHC studies.

Mouse tissue samples were fixed in phosphate-buffered aqueous 4% paraformaldehyde solution (Wako Pure Chemicals). The sections of paraffin-embedded tissues, picked up on glass slides, were stained with hematoxylin and eosin for histopathologic examination. Anti-human CD3 rabbit monoclonal antibody (SP7; 1:50 dilution; ab16669, Abcam) was used for immunostaining (15, 28–30). Tissue sections were deparaffinized with xylene and ethanol and then rehydrated with H₂O. To recover antigen, slides were heated in a microwave oven at 95°C for 20 minutes in ethylenediaminetetraacetic acid aqueous buffer solution (pH 8.0). After cooling and washing with buffer solution, the slides were treated using a Dako autostainer (Dakocytomation). Endogenous peroxidase activity was blocked by 5 minutes' incubation in 3% H₂O₂. The slides were incubated with antibody at room temperature in a humidified chamber for 30 minutes and then incubated with the Dako REAL EnVision Detection System horseradish peroxidase–conjugated anti-rabbit secondary antibody (K4003; Dakocytomation) at room temperature for 30 minutes. The immunoreaction was visualized by treatment with diaminobenzidine chromogen (Dako) for 5 minutes. Immunohistopathologic images were obtained using an Olympus BX53 microscope equipped with (Olympus) and analyzed using Olympus cellSens ver 1.16 Standard software.

All data are presented as mean ± SD or SEM. Results were analyzed using Student t test (two-tailed), unpaired Student t test (two-tailed), or ordinary one-way ANOVA as stated, with a P < 0.05 indicating a significant difference. The impact of EWS/FLI1-rejTs on overall survival rate was calculated using Kaplan–Meier analysis with log-rank testing. Excel (Microsoft) and Prism 8.0 (GraphPad Software) software was used for all statistical analyses.

To generate neoantigen-specific CTLs targeting the product of the EWS/FLI1 fusion gene, we used peptides corresponding to the modified break/fusion point epitope YLNPSVDSV (Fig. 1A and B; ref. 18). To confirm antigen specificity, CTLs were stained with an MHC pentamer. The frequency of EWS/FLI1-specific CTLs 8 days after peptide pulse was extremely low (only 0.003% of total T lymphocytes; Fig. 1C). Even after flow-cytometry sorting of MHC-pentamer–reactive CTLs for enrichment, the specificity of bulk cultured CTLs (EWS/FLI1-bulk CTLs) was still low (1.6%; Fig. 1C). Single-cell cloning established EWS/FLI1-CTL clones with 100% antigen specificity (Fig. 1C). Reprograming these clones with the four Yamanaka factors and SV 40 large T antigen (Sendai virus vector) established T-iPSCs. These were redifferentiated into EWS/FLI1-rejTs. EWS/FLI1-rejTs retained the same antigen specificity, with 99% pentamer reactivity, as that of the EWS/FLI1-CTL clone (Fig. 1C). EWS/FLI1-rejTs proliferated even after repeated stimulation, whereas the EWS/FLI1-CTL clone ceased to proliferate after a fourth stimulation (P = 0.0017; Fig. 1D).

We defined memory phenotype subsets of EWS/FLI1-rejTs and EWS/FLI1-CTL clones based on expression of CD45RA, CD62L, CD95, CD27, and CD28. The percentage of CD3⁺CD8⁺CD45RA⁺ and CD3⁺CD8⁺CD62L⁺ cells (naïve and stem cell memory T cells, respectively) was 36.0% in EWS/FLI1-rejT cells, significantly higher than that in EWS/FLI-clone cells (1.54%, P = 0.0002; Fig. 2A). The proportion of stem cell memory T cells (CD3⁺CD8⁺CD45RA⁺CD62L⁺CD95⁺CD27⁺CD28⁺; Fig. 2A) and central memory T cells (CD3⁺CD8⁺CD45RA⁻CD62L⁺CD95⁺CD27⁺CD28⁺) in EWS/FLI1-rejT cells was significantly higher than in EWS/FLI1-CTLs clones (SCM; 6.00% vs. 0.77%, P = 0.017, CM; 0.87% vs. 0.26%, P = 0.041; Fig. 2B). CD3⁺CD8⁺KLRG1^lowCD62L^highIL-7Rα^highCXCR3^high cells have been defined as central memory T cells having self-renewal potential (Fig. 2C; ref. 31). By this criterion, EWS/FLI1-rejTs included higher numbers of central memory T-cell populations than do EWS/FLI1-CTL clones (18.66% vs. 3.06%, P = 0.0029; Fig. 2D), results consistent with the proliferative ability and robust tumor-suppressive effect of EWS/FLI1-rejTs.

To compare expression of exhaustion markers between EWS/FLI1-CTL clones and EWS/FLI1-rejT cells, we examined the expression of LAG-3, PD-1, and TIM-3. Although EWS/FLI1-CTL clones did not express PD-1 (0.72%), the expression of LAG-3 and TIM-3 was high (80.4% and 74.0%, respectively; Fig. 2E). On the other hand, the expression of LAG-3 and TIM-3 in EWS/FLI-rejT cells was lower than those in EWS/FLI1-CTL clones (36.6% and 26.5%; Fig. 2E). That a larger proportion of EWS/FLI1-rejT cells was of stem cell memory phenotype and a smaller proportion was of an exhausted phenotype (with expression of LAG-3 and TIM-3) than among EWS/FLI1-CTL clones indicated phenotypic rejuvenation.

To check whether HLA expression was downregulated in the Ewing sarcoma cell lines used, we examined expression of HLA-A02 by flow cytometry. Both TC-71 and A673 cells expressed HLA-A02 (Fig. 3A). We also checked PD-L1 and PD-L2 expression on TC-71 and A673 cells by flow cytometry to assess PD-1/PD-L1 and PD-1/PD-L2 interaction in Ewing sarcoma. PD-L1 was expressed by approximately 45% of TC-71 cells, but PD-L2 was not expressed. A673 cells expressed neither PD-L1 nor PD-L2 (Fig. 3A). From these results, it was expected that PD-1–negative EWS/FLI1-rejTs (Fig. 2E) would be effective against Ewing sarcoma, even though Ewing sarcoma expresses PD-L1.

We used RT-PCR to determine the break/fusion point sequences in EWS/FLI1 in TC-71, A673, and RD-ES cells. The RT-PCR product length in TC-71 and A673 cells were each 100 bp, whereas the RT-PCR product length in RD-ES cells was 166 bp (Fig. 3B; ref. 3). The obtained sequences were compared with EWS/FLI1-fusion sequences (Fig. 3C). Fusion of exon 7 (EWS) with exon 6 (FLI1; type 1) was found in TC-71 and A673 cells, whereas fusion of exon 7 (EWS) with exon 5 (FLI1; type 2) was found in RD-ES cells (Fig. 3C).

To examine the antigen-specific cytotoxicity of EWS/FLI1-rejTs, we performed a ⁵¹Cr release assay. EWS/FLI1-rejTs showed stronger cytotoxicity than EWS/FLI1-CTL clones against TC-71 (46.6% and 37.7%; 37.8% and 19.4%, respectively; Fig. 3D). On the other hand, cytotoxicity of EWS/FLI1-rejTs against RD-ES cells was 13.6% and 10.1% and that of EWS/FLI1-CTL clones was 4.3% and 4.5% for the two indicated E:Ts, respectively (Fig. 3D). To compare the cytotoxicity of EWS/FLI1-rejTs with that of EWS/FLI1-CTL clones over a long period of time, we examined cytotoxicity against Ewing sarcoma using a real-time cell analyzer. TC-71 cells were cocultured with EWS/FLI1-rejTs, EWS/FLI1-CTL clones, or control rejTs (not EWS/FLI1-specific). The proliferation of TC-71 cells was suppressed by EWS/FLI1-rejTs, and the cytotoxicity of EWS/FLI1-rejTs was 75.7% after 10 hours of coculture with TC-71 cells, which was higher than that of cells of the original clone (22.2%, P < 0.0001; Fig. 4A), and continued over 40 hours (91.3%; 40 hours; Fig. 4A). On the other hand, the EWS/FLI1-CTL clones suppressed the proliferation of TC-71 cells to some extent (51.3%; 40 hours, P = 0.008; Fig. 4A). We also examined cytotoxicity against A673 using a real-time cell analyzer. A673 cells were cocultured with EWS/FLI1-rejTs or control rejTs. The cytotoxicity of EWS/FLI1-rejTs was 75.4% at 10 hours (Fig. 4B). On the other hand, control T rejTs were not able to suppress the proliferation of A673 cells (−0.02%, 10 hours; Fig. 4B). EWS/FLI1-rejTs also suppressed the proliferation of A673 cells for more than 40 hours. The cytotoxicity of EWS/FLI1-rejTs was 87.4% at 40 hours, whereas control rejTs could not suppress tumor proliferation (22.8%, P = 0.002; Fig. 4B). This confirmed more potent and continuous antigen-specific cytotoxicity against Ewing sarcoma in EWS/FLI1-rejTs than in the original CTLs. We demonstrated the cytotoxicity of EWS/FLI1-rejTs against PD-L1⁺ TC-71 cells, which was not affected by PD-L1 expression. EWS/FLI1-rejTs generated using the modified peptide sequence recognized the fusion of exon 7 (EWS) to exon 6 (FLI1) in TC-71 and A673 cells and exhibited antigen-specific cytotoxicity against Ewing sarcoma.

To observe the antitumor effect of EWS/FLI1-rejTs in vivo, TC-71 cells labeled with FFluc/GFP were intravenously inoculated into NOG mice (1 × 10⁵ cells/mouse). Four days after tumor inoculation, mice were divided into a control group (no treatment) and three treatment groups: mice treated with EWS/FLI1-rejTs, EWS/FLI1-bulk CTLs, or EWS/FLI1-CTL clones. Antigen specificity of EWS/FLI1-bulk CTLs was around 2% after pentamer-reactive sorting (Fig. 1C). Although pentamer-reactive EWS/FLI1-bulk CTLs were repeatedly sorted and expanded, pentamer reactivity did not increase (2.07%–1.91%; Fig. 5A). EWS/FLI1-rejTs clearly suppressed TC-71 cell proliferation (Fig. 5B). On the other hand, TC-71 cells progressively spread around the liver and throughout the bodies of untreated mice. They also proliferated in mice treated with EWS/FLI-bulk CTLs (Fig. 5B). In the 2 mice that could be studied, EWS/FLI1-CTL clones suppressed tumor growth more efficiently than did EWS/FLI1-bulk CTLs (Fig. 5B). By day 20, bioluminescence had progressively increased in the untreated mice (no treatment, 1.38 × 10⁷ p/s/cm²/sr; range, 1.90 × 10⁶ to 3.42 × 10⁷; Fig. 5C). No significant difference in tumor signal was observed between untreated mice and EWS/FLI1-bulk CTL–treated mice 20 days after first treatment (no treatment; 1.90 × 10⁶ to 3.42 × 10⁷ p/s/cm²/sr; EWS/FLI1-bulk CTLs; 7.78 × 10⁶ to 2.18 × 10⁷ p/s/cm²/sr, P = 0.99; Fig. 5C). In contrast, tumor signal in mice treated with EWS/FLI1-rejTs was significantly low, with efficient suppression of tumor growth on day 20 (EWS/FLI1-rejTs; 3.18 × 10⁵ to 9.26 × 10⁶ p/s/cm²/sr; EWS/FLI1-rejTs vs. no treatment: P = 0.037; EWS/FLI1-rejTs vs. EWS/FLI1-bulk CTLs: P = 0.071; Fig. 5C). EWS/FLI1-rejTs significantly prolonged survival (27–35 days) compared with that of untreated mice (P = 0.0013, 20–27 days) and EWS/FLI1-bulk CTL-treated mice (P = 0.0312, 19–29 days; Fig. 5D). Although body weight of EWS/FLI1-bulk CTL-treated mice decreased slightly from day 14 to day 22, no other significant change was noted (Fig. 5E).

On day 20, the frequencies of pentamer-reactive human CD3⁺ cells in mouse peripheral blood were determined by flow cytometry. Pentamer-reactive CD3⁺ cells were detected (0.15%) only in EWS/FLI1-rejT–treated mice (Fig. 6A). We also performed histopathologic examination of the liver and lungs from each group. Human CD3⁺ cells were found within a tumor in the liver of the EWS/FLI1-rejT–treated mouse (Fig. 6B). Human CD3⁺ cells were almost absent from tumors in the liver of the mouse treated with EWS/FLI1-bulk CTLs (Fig. 6B). EWS/FLI1-rejTs seemed numerous in lung capillaries of the rejT-treated mouse, whereas only a few EWS/FLI1-CTLs were noted in lung capillaries of the bulk CTL-treated mouse (Fig. 6B).

When EWS/FLI1-rejTs were administered intravenously, many rejTs were observed in lung capillaries. This prompted us to compare intraperitoneal with intravenous treatment. TC-71 cells labeled with FFluc/GFP were intraperitoneally engrafted into NOG mice (1 × 10⁵ cells/mouse), and bioluminescence was monitored to observe tumor growth. Four days after tumor inoculation, mice were divided into a control group (no treatment) and three treatment groups: mice treated with EWS/FLI1-rejTs, EWS/FLI1-CTL clone cells, and of mice treated with EWS/FLI1-bulk CTLs. EWS/FLI1-rejTs clearly suppressed tumor proliferation in rejT-treated mice (Fig. 7A). In contrast, bioluminescence rapidly spread in mice without treatment. Tumor cells also progressively proliferated in mice treated with EWS/FLI-bulk CTLs and EWS/FLI1-CTL clones (Fig. 7A). On day 22, bioluminescence in untreated mice, EWS/FLI1-bulk CTL–treated, and EWS/FLI1-CTL clone–treated mice was clearly elevated, showing that EWS/FLI1-bulk CTLs and EWS/FLI1-CTL clones failed to suppress tumors (no treatment, 2.79 × 10⁷ p/s/cm²/sr; range, 1.55 × 10⁷ to 4.90 × 10⁷; EWS/FLI1-bulk CTLs, 3.01 × 10⁷ p/s/cm²/sr; range, 1.06 × 10⁷ to 4.30 × 10⁷, P = 0.96; Fig. 7B). On the other hand, EWS/FLI1-rejTs suppressed tumor proliferation (EWS/FLI1-rejTs, 9.40 × 10⁵ p/s/cm²/sr; range, 2.05 × 10⁴ to 3.40 × 10⁶; P = 0.025: EWS/FLI1-rejTs vs. EWS/FLI1-bulk CTLs; P = 0.001: EWS/FLI1-rejTs vs. no treatment; Fig. 7B). EWS/FLI1-bulk CTLs did not prolong survival (60 days observation) compared with untreated control mice (P = 0.20). EWS/FLI1-rejTs, however, significantly prolonged survival (18–60 days) compared with untreated mice (P = 0.01, 13–25 days) and EWS/FLI1-bulk CTL–treated mice (P = 0.02, 22–35 days; Fig. 7C). No significant change was noted (Fig. 7D). On day 25, the abdomen was grossly enlarged, with multiple tumors, in the untreated mouse and EWS-FLI1-bulk CTL–treated mouse (Fig. 7E). In contrast, macroscopically visible tumors could not be identified in the EWS/FLI1-rejT–treated mouse (Fig. 7E). Histopathologic examination using light microscopy found no tumor in the liver or lungs of EWS/FLI1-rejT–treated mice (Fig. 7F). In contrast, deposits of small round cells, characteristic of Ewing sarcoma, were observed in the mouse treated with EWS/FLI1-bulk CTLs (Fig. 7F). These findings indicate that EWS/FLI1-rejTs could more robustly suppress TC-71 cell proliferation when delivered intraperitoneally than intravenously in mice and that survival of EWS/FLI1-rejT–treated mice was prolonged.

When mutant peptides are encoded by somatic mutations, the peptides are specifically targeted by CTLs as a tumor-specific antigen. Because these antigens are not present in normal cells, they are called neoantigens. Neoantigen-specific CTLs have high specificity, which constitutes a safety benefit (9, 10, 17, 32–34). Gros and colleagues found that neoantigen-specific CTLs could be generated from PD1-expressing CD8⁺ T cells in peripheral blood of patients with gastrointestinal cancer. These CTLs exerted neoantigen-specific cytotoxicity (33). Tran and colleagues generated CTLs targeting KRAS neoantigens from lymphocytes that infiltrated metastatic colorectal cancer. These neoantigen-specific CTLs successfully reduced the size of lung metastases (32).

We attempted to apply this method to target the neoantigen encoded by the EWS/FLI1 fusion gene. We successfully developed an EWS/FLI1-rejT therapy for Ewing sarcoma. Evans and colleagues successfully generated EWS/FLI1-specific CTLs from healthy donors by modifying peptides. They evaluated these cells' antitumor effect in vitro and in vivo. However, the frequency of EWS/FLI1-specific CTLs was low (∼3%) and that this vitiated antitumor effects (18). The frequency of EWS/FLI1-specific CTLs generated from PBMCs of a healthy donor was only 0.003% in our study even after peptide pulsing. However, single-cell cloning technology made it possible to establish a 100% EWS/FLI1-specific CTL clone. Because the EWS/FLI1-specific CTL clone did not proliferate well after repeated stimulation, we exploited iPSC technology. Because T-iPSCs can proliferate without limit, like embryonic stem cells, we could generate large numbers of EWS/FLI1-rejTs with antigen specificity identical to that of the original CTL clone. When using our T-cell differentiation protocol, the rejT number obtained from 1 well of master cell bank iPSCs/6-well plate (∼2 × 10⁶ cells) was approximately 100 × 10⁶ cells. If these generated rejTs were to be aliquoted, banked, and restimulated, each frozen stock sample would yield >100 × 10⁶ rejTs. The generated EWS/FLI1-rejTs displayed a stem cell memory phenotype and a central memory phenotype with self-renewal potential. They robustly proliferated, yielding more potent cytotoxicity in vitro than the original CTL clone.

With intravenous administration, many EWS/FLI1-rejTs were demonstrable in lung capillaries. This may reflect only mechanical filtration; fully activated EWS/FLI1-rejTs are slightly bigger (∼10 μm, long axis) than original CTLs (<7μm) and mouse T lymphocytes (<6 μm; refs. 35, 36). To deliver EWS/FLI1-rejTs more directly to tumor, we examined tumor-suppressive effects using an intraperitoneal-injection mouse model. EWS/FLI1-rejTs efficiently suppressed tumor proliferation and significantly prolonged the survival of EWS/FLI1-rejT–treated mice.

Unfortunately, no effective molecule-targeted drug for EWS/FLI1 has been established (8). Rejuvenated CTL therapy that targets a break/fusion antigen using iPSC technology offers a promising treatment in other tumors with fusion genes, as well as in Ewing sarcoma. The epitope thus far studied is HLA-A*02:01 restricted, with approximately 30% of Europeans and Americans expressing HLA-A*02:01 (37, 38). Generation of HLA-A24–restricted EWS/FLI1 type1-rejTs will expand availability of this treatment. Binding affinity prediction software has identified a suitable epitope, and we hope soon to generate HLA-A24–restricted EWS/FLI1 type1-rejTs.

Generating patient-derived rejTs is expensive and takes 5 months—too long for patients with extensive disease to wait. Patient-derived rejTs, however, have the advantage, in that immune rejection by patient CD8⁺ T cells and natural killer cells cannot occur. Use of allogeneic rejTs entails generation time much shorter than that with patient-derived rejTs, although immune rejection likely poses a problem. Use of HLA class I–edited allogeneic rejTs should evade rejection and thereby clearly demonstrate the superiority of allogeneic rejT therapy targeting Ewing sarcoma.

Our results suggest that functionally rejuvenated EWS/FLI1-rejTs will be very valuable as a novel treatment for Ewing sarcoma. The potential advantage of EWS/FLI1-rejT therapy is infinite generation of therapeutic CTLs from T-iPSCs. Editing of HLA class I antigen in T-iPSCs using CRISPR/Cas9 technology will make true “off-the-shelf” therapy available. Even without that refinement, EWS/FLI1-rejT therapy appears promising for recurrent and metastatic Ewing sarcoma, as well as for inoperable Ewing sarcoma in the localized stage.

M. Ishii reports grants from JPJS/KAKENHI during the conduct of the study. J. Ando reports grants from JSPS/KAKENHI outside the submitted work. Y. Suehara reports personal fees from Bayer outside the submitted work. M. Nakanishi reports personal fees from TOKIWA-Bio, Inc. during the conduct of the study. H. Nakauchi reports personal fees and other support from Century Therapeutcs and personal fees from Argonaut Genomics during the conduct of the study, as well as personal fees from QihanBio Therapeutics and personal fees and other support from Megakaryon Corp. outside the submitted work. M. Ando reports grants from JSPS/KAKENHI during the conduct of the study, as well as grants from Century Therapeutics and Daiichi Sankyo Company, Limited outside the submitted work. No disclosures were reported by the other authors.

M. Ishii: Conceptualization, resources, data curation, investigation, writing–original draft. J. Ando: Resources, investigation, methodology. S. Yamazaki: Methodology. T. Toyota: Resources, investigation. K. Ohara: Resources. Y. Furukawa: Writing–review and editing. Y. Suehara: Resources. M. Nakanishi: Resources. K. Nakashima: Investigation. K. Ohshima: Investigation. H. Nakauchi: Conceptualization, supervision, writing–review and editing. M. Ando: Conceptualization, resources, funding acquisition, investigation, methodology, writing–review and editing.

The authors thank A.S. Knisely for critical reading of the article. They thank Azusa Fujita, Yumiko Ishii, and Tamami Sakanishi for FACS operation and the members of the Laboratory of Radioisotope Research, Center for Biomedical Research Resources, Research Support Center, Juntendo University Graduate School of Medicine, for technical assistance. The authors thank Dr. Gianpietro Dotti and Dr. Nobuhiro Nishio for provision of retroviral FFluc/GFP plasmid and also thank Dr. Melinda Merchant for kindly providing TC-71 cells. These studies were supported by a grant from JSPS/KAKENHI (19J40304, 19K07781) and a Grant-in-Aid for Special Research in Subsidies for ordinary expenses of private schools from The Promotion and Mutual Aid Corporation for Private Schools of Japan (G3001, G1912).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.