This study was conducted in accordance with the Declaration of Helsinki and was approved by the Ethics Committee at Juntendo University School of Medicine. Peripheral blood samples were obtained from healthy adult volunteers carrying HLA-A*02:01 or HLA-A*24:02, and donors received the BNT162b2 SARS-CoV-2 mRNA vaccine (BioNTech/Pfizer, Mainz, Germany/New York, NY) according to standard protocols. Peripheral blood mononuclear cells (PBMCs) were collected from unvaccinated individuals (defined as more than 1 year after their last vaccination) and vaccinated individuals (defined as vaccinations within 3 months after vaccination).

The acute lymphoid leukemia cell line NALM-6 (RRID: CVCL_0092), Burkitt lymphoma cell line Raji (RRID: CVCL_0511), and pancreatic cancer cell line PANC-1 (RRID: CVCL_0480) were purchased from RIKEN BioResource Center (Tsukuba, Japan). The non-small cell lung cancer cell line A549 (RRID: CVCL_0023) was purchased from ATCC (Manassas, VA). Lymphoblastoid cell lines (LCLs) were derived from PBMCs in our laboratory using EBV isolated from B95-8 cell lines in the presence of cyclosporin A. NALM-6, Raji, PANC-1, and LCLs were cultured under standard conditions at 37 °C and 5% CO₂ using Roswell Park Memorial Institute 1640 medium (Gibco, Carlsbad, CA) supplemented with 10% FBS (Gibco), 100 U/mL penicillin, 100 μg/mL streptomycin, and 2 mM L-glutamine (Gibco). A549 cells were cultured under standard conditions in Dulbecco’s modified Eagle medium (DMEM; Gibco) containing 10% FBS and antibiotics as described above.

To generate a lentiviral Cs-Ubc-CD19-4-1BB-CD3ζ-F2A-mCherry construct, CD19-4-1BB-CD3ζ was amplified by PCR from the retroviral MSCV-CD19-4-1BB-CD3ζ-IRES-GFP plasmid (Harada et al., 2022) and fused with a lentiviral Cs-Ubc-iC9-F2A-mCherry construct (Ando et al., 2015) by replacing iC9 with CD19-4-1BB-CD3ζ using InFusion cloning (Takara Bio) (Figure 3A). After Cs-Ubc-CD19-4-1BB-CD3ζ-F2A-mCherry was generated, 293T cells were transfected with the lentiviral construct together with the packaging plasmids pMDLg/pRRE and pCMV-VSVG-RSV-Rev, using polyethylenimine (Polysciences, Warrington, PA) as the transfection reagent. Virus supernatants were collected 48–72 h after application of the forskolin-containing medium (Ando et al., 2014).

PBMCs obtained from a healthy donor harboring HLA-A*02:01 were activated with T Cell TransAct (Miltenyi Biotec, Bergisch Gladbach, Germany) in the presence of recombinant human IL-2 (100 U/mL). After 3 days, activated T-cells were transferred onto 24-well plates coated with RetroNectin (Takara Bio, Shiga, Japan) and CD19-CAR lentiviral supernatants were added to each well. CAR-T cells were cultured in NS-A2 medium (Nissui, Tokyo, Japan) supplemented with 10 ng/mL each of IL-7 (Miltenyi Biotec) and IL-15 (Miltenyi Biotec).

PBMCs from an HLA-A*02:01-positive healthy donor were cocultured with autologous DCs pulsed with the SARS-CoV-2 spike protein_269-277 peptide (YLQPRTFLL; S269), whereas PBMCs from an HLA-A*24:02-positive donor were cocultured with DCs pulsed with the spike protein_448-456 peptide (NYNYLYRLF; S448) (Mimotopes, Victoria, Australia) in the presence of IL-4 (400 IU/mL) and IL-7 (10 ng/mL) as described previously (Gerdemann et al., 2012; Rooney et al., 2014). On day 9 of culture, T-cells were restimulated with peptide-pulsed DCs. On day 16, CTLs were stained with HLA-A*02:01/S269 tetramer or HLA-A*24:02/S448 tetramer (MBL, Nagoya, Japan). Tetramer-positive CTLs were cloned by a limiting dilution to obtain SARS-CoV-2-specific CTL clones (Ando et al., 2020; Ando et al., 2015; Nishimura et al., 2013).

A SARS-CoV-2-specific CTL clone recognizing the HLA-A*02:01-restricted S269 (COVID19-CTL) was reprogrammed using Sendai virus vectors encoding OCT3/4, SOX2, KLF4, c-MYC, NANOG, and LIN28 (Honda et al., 2020). Transduced cells were transferred onto plates coated with iMatrix-511 (Nippi, Tokyo, Japan) and cultured in NS-A2 medium. Growth medium was gradually replaced with StemFit AK03N (Ajinomoto Healthy Supply, Tokyo, Japan). After 14 days, emerging iPSC colonies were picked up into new culture wells and passaged in CTS Essential 8 medium (Thermo Fisher Scientific, Waltham, MA, United States) several times until cell numbers sufficed for downstream experiments (Ando et al., 2020). The established T-iPSC line derived from COVID19-CTL was designated as COVID19-iPSCs.

COVID19-iPSCs (1 × 10⁵ cells) were transduced with CD19-CAR lentiviral vector on vitronectin (Thermo Fisher Scientific)-coated plates (multiplicity of infection of 20) and cultured in CTS Essential 8 medium (Thermo Fisher Scientific). On day 7, mCherry + iPSCs were isolated using a MoFlo Astrios EQ cell sorter (Beckman Coulter, Brea, CA). The CD19-CAR⁺ iPSC (CD19CAR-COVID19-iPSCs) population was subsequently enriched by repeated sorting to increase CAR positivity.

Two types of rejTs were generated, 1919-CARrejTs derived from CD19-CAR-COVID19-iPSCs, and COVID19-rejTs derived from COVID19-iPSCs. Small clumps of T-iPSCs were plated on C3H10T1/2 feeder cells in IMDM supplemented with gamma-irradiated 15% FBS (HyClone, GE Healthcare UK, Little Chalfont, UK) and a cocktail of 10 mg/mL insulin, 5.5 mg/mL transferrin, 5 ng/mL sodium selenite, 2 mM L-glutamine (Thermo Fisher Scientific), 0.45 mM α-monothioglycerol (Sigma-Aldrich, St. Louis, MO), and 50 mg/mL ascorbic acid (Takeda Pharmaceutical, Tokyo, Japan) in the presence of 20 ng/mL vascular endothelial growth factor (VEGF; Miltenyi Biotec). On day 14 of coculture, sac-like structures that contained hematopoietic progenitor cells were extracted and transferred onto C3H10T1/2 feeder cells expressing Delta-like ligand 1 (DLL1) and DLL4 in minimum essential medium (MEM) α (Thermo Fisher Scientific) with FBS (HyClone) in the presence of 20 ng/mL human stem cell factor, 10 ng/mL human Fms-related tyrosine kinase 3 ligand (both Miltenyi Biotec), and 10 ng/mL IL-7. On day 28 of coculture, T-lineage cells were harvested, stimulated by irradiated PBMCs in NS-A2 in the presence of 5 mg/mL phytohemagglutinin (Sigma-Aldrich) and of IL-7 and IL-15 (10 ng/mL each). SARS-CoV-2 specificity and CD19-CAR transgene expression were confirmed by staining with an HLA-A*02:01/S269 tetramer and by detecting mCherry fluorescence, respectively.

The cytotoxic activity of 1919-CARrejTs, COVID19-rejTs, CD19-CAR-Ts was evaluated in a standard 6-h ⁵¹Cr release assay at different effector-to-target (E:T) ratios (40:1, 20:1, 10:1, and 5:1) (Rooney et al., 2014). Target cells used in these assays included LCLs, NALM-6, and Raji cells as specified in each figure (Figures 1–3). Where indicated, targets were pulsed with S269 peptide prior to coculture. Target cells were labeled with sodium chromate (⁵¹Cr) and cocultured with T cells, with or without S269 loading. After incubation, 50 μL of supernatant was transferred from each well to LumaPlate-96 wells (PerkinElmer, Waltham, MA) and dried overnight. Released ⁵¹Cr was quantitated using a TopCount microplate scintillation counter (PerkinElmer), with % specific lysis = {([experimental ⁵¹Cr release] – [spontaneous ⁵¹Cr release])/([maximal ⁵¹Cr release] – [spontaneous ⁵¹Cr release])} × 100. Each condition was assayed in technical triplicate wells per experiment. Unless otherwise stated, ⁵¹Cr-release assays were repeated in three independent experiments performed on different days using independently prepared effector and target cells.

Samples were processed using LSRFortessa and FACSCalibur cytometers (BD Biosciences, San Jose, CA). The results were analyzed using FlowJo software 10.10.0 (Tree Star, Ashland, OR). After cells were incubated with the appropriate concentration of fluorescence-conjugated monoclonal antibody cocktail for 30 min at 4 °C, they were washed with phosphate-buffered saline. Chromophore-conjugated monoclonal antibodies directed against surface antigens were used; they included APC mouse anti-human CD4 antibody (BioLegend, 317416), APC mouse anti-human PD-1 antibody (BioLegend, 329908), APC/cyanine7 mouse anti-human CD3 antibody (BioLegend, 300318), APC/cyanine7 mouse anti-human TIM3 antibody (BioLegend, 345026), fluorescein isothiocyanate (FITC) mouse anti-human CD19 antibody (BioLegend, 302206), PE mouse anti-human TIGIT antibody (BioLegend, 372704), BV421 mouse anti-human LAG3 antibody (BioLegend, 369314), Pacific Blue mouse anti-human CD8 antibody (BioLegend, 301023). CAR expression was evaluated based on mCherry fluorescence in CAR-transduced cells. PE-conjugated HLA-A*02:01/S269 tetramer and HLA-A*24:02/S448 tetramer (MBL) were used to detect antigen-specificity.

CD19-CAR-Ts and 1919 CARrejTs were first cocultured with CD19⁺ NALM-6 tumor cells for an initial challenge and then subjected to sequential rechallenges. The initial coculture was established with 0.4 × 10⁶ effector cells and 0.1 × 10⁶ NALM-6 cells. For each subsequent rechallenge, NALM-6 cells were supplemented at 0.1 × 10⁶ cells. At each rechallenge, effectors were re-exposed to tumor cells under peptide-loaded or non-loaded conditions. Effector and tumor cells were cocultured in NS-A2 medium. Immediately before each rechallenge and at the final time point, tumor to effector ratios were quantified by flow cytometry based on CD19 (tumor) and CD3 (T cells) expression, according to the protocol shown in Figure 4A.

Proliferation of 1919-CARrejTs was quantified by carboxyfluorescein succinimidyl ester (CFSE) dilution. Effector cells were labeled with 5 μM CFSE (CellTrace™ CFSE Cell Proliferation Kit, Thermo Fisher Scientific) for 20 min at 37 °C and were then washed twice before coculture. Labeled effectors were cocultured with target cells for 72 h at a 4:1 E:T ratio under the following conditions: (i) PANC-1 ± S269 peptide (HLA-matched, CD19⁻), (ii) A549 ± S269 peptide (HLA-mismatched, CD19⁻), (iii) NALM-6 ± S269 peptide (HLA-matched, CD19⁺), or (iv) a no-tumor control containing effector cells only. After incubation, cells were harvested, stained with anti-CD3 antibody (BioLegend), and analyzed for CFSE dilution of viable CD3⁺ T cells detected by flow cytometry.

All data are presented as mean ± SD as stated in the figure legends. Results were analyzed by an unpaired Student’s t-test as stated in the text. All statistical analyses were performed using Excel (Microsoft, Redmond, WA) and Prism (GraphPad, San Diego, CA) programs. Values of p < 0.05 were considered significant.

SARS-CoV-2 spike protein-specific CD8⁺ T cells were not detectable or detectable at low frequencies in PBMCs from healthy donors carrying HLA-A*02:01 or HLA-A*24:02, accounting for 0.000% and 0.015% of total CD3⁺ T cells, respectively. Following SARS-CoV-2 mRNA vaccination, the frequencies within the CD3⁺ lymphocyte gate clearly increased to 0.186% and 0.674%, respectively (Figure 1A). To generate antigen-specific CTLs in vitro, PBMCs were obtained from healthy donors more than 1 year after their last vaccination and cocultured with autologous DCs pulsed with the corresponding SARS-CoV-2 spike peptides (S269 for HLA-A*02:01 and S448 for HLA-A*24:02). Tetramer staining confirmed the expansion of SARS-CoV-2 spike-specific CTLs following DC stimulation (0.154% and 0.089% of CD3⁺ T cells for HLA-A*02:01 and HLA-A*24:02, respectively) and subsequently single-cell cloned (83.8% and 96.9%, respectively) (Figure 1B). To examine cytotoxic functions, we performed 6-h ⁵¹Cr -release assays. SARS-CoV-2 spike-specific CTLs exhibited cytotoxicity against HLA-matched, peptide-loaded LCLs but not against HLA-mismatched LCLs (HLA-A*02:01 restricted CTLs: 78.7% ± 3.0% versus 4.0% ± 0.6%, p < 0.001, and HLA-A*24:02 restricted CTLs: 82.8% ± 1.0% versus 3.2% ± 1.3%, p < 0.001) at an E:T ratio of 40:1 (Figure 1C). These results confirmed that TCR stimulation by SARS-CoV-2 spike-protein vaccination increased proliferation of antigen-specific CTLs and that the generated CTLs exhibited cytotoxic activity against SARS-CoV-2 spike-peptide pulsed LCLs.

SARS-CoV-2 specific CTLs were cloned, reprogrammed into iPSCs, and differentiated into rejuvenated CTLs (COVID-19 rejTs), as outlined in the schematic overview (Figure 2A). In 6-h ⁵¹Cr -release assays, COVID19-rejTs mediated antigen-specific cytotoxicity. They showed robust killing against HLA-matched, spike peptide loaded LCLs and NALM-6 targets, with specific lysis of 70.4% ± 1.1% and 63.1% ± 7.7%, respectively, at an E:T ratio of 40:1, whereas activated T-cells from HLA-matched healthy donors showed minimal lysis of 12.0% ± 1.2% and 1.6% ± 3.6%, respectively (Figure 2B). When tested against HLA-matched but peptide-unloaded NALM-6 targets, only low levels of unspecific lysis was detected compared to activated T cells (23.4% ± 2.8% versus 1.2% ± 2.6%, at an E:T ratio of 40:1). On the other hand, COVID19-rejTs showed minimal cytotoxicity against HLA-mismatched, spike peptide-loaded Raji cells, with specific lysis comparable to activated T cells (1.4% ± 0.7% versus 6.9% ± 8.5%, at an E:T ratio of 40:1). (Figure 2B).

We next compared the lysis of COVID19-rejTs with that of original COVID19-CTL clone against HLA-matched, peptide-loaded LCLs. COVID19-rejTs demonstrated antigen-specific cytotoxicity comparable to that of the original CTL clones against HLA-matched, peptide-loaded LCLs (75.1% ± 4.0% versus 78.7% ± 3.0%, at an E:T ratio of 40:1), whereas neither effector cells mediated detectable lysis against HLA-mismatched LCLs, even when peptide-pulsed (7.4% ± 0.3% versus 4.0% ± 0.6%) (Figure 2C). These results confirm that cytotoxicity of COVID19-rejTs is retained in an HLA-restricted antigen-specific manner, even after iPSC reprogramming and redifferentiation.

To generate dual-antigen recognition rejTs, CD19-CAR-2A-mCherry (Figure 3A) was introduced into COVID19-iPSCs originated from HLA-A*02:01-restricted COVID19-CTLs, then differentiated into T cells, as outlined in the schematic overview (Figure 3B). The resulting product, referred to as 1919-CARrejTs, represent CD19-CAR-expressing rejTs that retain spike protein-specific TCR. Next, we evaluated CD19-CAR transgene expression and spike protein specificity in 1919-CARrejTs by flow cytometry based on mCherry detection and HLA-A*02:01/S269 tetramer staining. Both CAR transgene expression (mCherry positivity) and tetramer positivity were nearly 100%, confirming that 1919-CARrejTs uniformly expressed CD19-CAR and maintained spike protein-antigen specificity (Figure 3C). In parallel, we generated conventional CD19-CAR-T cells from peripheral blood using the same CAR construct. In these cells, CAR expression was detected in 53.2% of CD3⁺ T cells, reflecting transduction efficiency, whereas spike protein-specific TCRs were nearly undetectable (0.35%) (Figure 3C).

To compare T-cell functions of 1919-CARrejTs with that of conventional CD19-CAR-T cells, expression levels of cytotoxic molecules and exhaustion markers were evaluated by flow cytometric analysis. 1919-CARrejTs showed higher expression levels of intracellular cytotoxic molecules (granzyme B and perforin) than conventional CD19-CAR-T cells, with mean MFI values of 7771.7 ± 446.6 versus 3330.3 ± 211.2 (p < 0.001), and 10162 ± 1273 versus 2129.3 ± 241.4 (p < 0.001), respectively (Figure 3D). Furthermore, expression levels of T-cell exhaustion markers LAG-3, TIM-3, and TIGIT, were decreased in 1919-CARrejTs compared with conventional CD19-CAR-T cells (Figure 3E). These results demonstrate that 1919-CARrejTs exhibited more cytotoxic activity and less exhaustion compared to conventional CD19-CAR-T cells.

Six-hour ⁵¹Cr -release assays revealed that 1919-CARrejTs exhibit stronger cytotoxicity against LCL targets (CD19⁺, HLA-matched, spike peptide-loaded) compared to conventional CD19-CAR-T cells, with specific lysis of 58.5% ± 1.2% versus 43.7% ± 3.3% (p =0.018) at an E:T ratio of 40:1, respectively (Figure 3F, left). Similar results were observed when comparing 1919-CARrejTs and conventional CD19-CAR-T using NALM-6 (CD19⁺, HLA-matched, spike peptide-loaded), exhibiting specific lysis of 77.5% ± 4.6% versus 62.5% ± 5.6% (p =0.023) at an E:T ratio of 40:1, respectively (Figure 3F, middle). When Raji cells (CD19⁺, HLA-mismatched, spike peptide-loaded) were used, the difference in cytotoxicity was lost, with specific lysis of 53.5% ± 4.1% versus 51.4% ± 1.9% (p =0.478) at an E:T ratio of 20:1, respectively (Figure 3F, right). These results are consistent with our previous findings that dual-antigen recognition amplifies immediate cytotoxicity when the targets have antigens recognizable by both CAR and TCR (Harada et al., 2022).

To assess cytotoxicity of 1919-CARrejTs under repeated tumor exposure, we utilized a sequential rechallenge model in which effectors were cocultured with NALM-6 and then re-exposed at prespecified intervals. Four arms were tested under identical schedules: CD19-CAR-T cells ± spike protein peptide and 1919-CARrejTs ± spike protein peptide. E:T ratios were measured by flow cytometry (CD3 for effector cells; CD19 for target tumor cells (Figure 4B; Supplementary Figure S3). In this setting, conventional CD19-CAR-T cells showed progressive tumor outgrowth upon repeated tumor exposure. Tumor burden increased from 8.76% at day 1%–51.0% by day 19 (end of the seventh challenge). In contrast, 1919-CARrejTs showed sustained tumor suppression, with tumor fractions remaining below 5% throughout all rechallenge cycles. End-of-experiment analysis within the CD3⁺ gate showed lower TIM-3 expression in 1919-CARrejTs than in conventional CD19-CAR-T cells with TCR stimulation, with mean MFI values of 388.3 ± 12.0 versus 1723.8 ± 195.3 (p < 0.001), respectively. Similar results were observed also in the absence of TCR stimulation, with mean MFI values of 384.5 ± 6.5 versus 1762.2 ± 87.0 (p < 0.001). In contrast, expression levels of PD-1 and LAG-3 were similar across all arms (Figure 4C). In summary, our data demonstrate that 1919-CARrejTs not only exhibited stronger cytotoxic activity than conventional CD19-CAR-T cells, but also maintain effective tumor suppression even after multiple rounds of tumor rechallenge. Also, 1919-CARrejTs eradicated tumor cells independently of TCR stimulation with spike protein peptide.

We performed CFSE dilution assays of 1919 CARrejTs under three coculture conditions to distinguish the effects of TCR- and CAR-mediated cues on proliferation: (i) PANC-1 cells ± spike protein peptide (HLA-matched, CD19⁻; TCR-dependent), (ii) A549 cells ± spike protein peptide (HLA-mismatched, CD19⁻; negative control), and (iii) NALM-6 cells (CD19⁺; CAR-only stimulation) (Figure 4D). 1919-CARrejTs showed efficient proliferation when cocultured with spike protein peptide-loaded PANC-1, while proliferation levels were low when cocultured with spike protein peptide-unloaded PANC-1. As expected, 1919 CARrejTs with A549 cells induced low proliferation levels, regardless of spike protein peptide presence. Finally, 1919-CARrejTs showed limited proliferation upon CAR stimulation alone, whereas peptide loading of NALM-6 induced additional proliferation, consistent with an additive contribution of TCR signaling under concurrent CAR stimulation. These results suggest that native TCR stimulation augments CAR-driven proliferative cues. Together with the results from the sequential rechallenge model, these findings suggest that CARrejTs maintain cytotoxic activity over longer periods of time compared to conventional CAR-T cells, with TCR signaling serving as an auxiliary cue that supports their expansion and persistence.