This study was approved by the Institutional Review Board of Kyoto Prefectural University of Medicine (Approval Numbers: ERB-C-669 and ERB-C-1406) and the recombinant DNA experiments were approved by the safety committee of the recombinant DNA experiment of Kyoto Prefectural University of Medicine (Approval Numbers 2019-111 and 2019-112). All experiments involving human participants were performed in accordance with the Declaration of Helsinki guidelines. All animal experiments and procedures were approved by the Kyoto Prefectural University of Medicine Institutional Review Board (Permit No.: M2020-13).

Blood samples from healthy donors were obtained with a written informed consent, and PBMCs were isolated from the whole blood samples by density gradient centrifugation using Lymphocyte Separation Medium 1077 (FUJIFILM Wako Pure Chemical Corporation, Osaka, Japan), followed by multiple washes with Dulbecco’s phosphate-buffered saline (D-PBS; Nakarai Tesque, Kyoto, Japan). The number of live cells was determined by standard trypan blue staining and using an automated cell counter model R1 (Olympus, Tokyo, Japan). The human lymphoblastic leukemia cell line (REH) was purchased from the American Type Culture Collection (Manassas, VA). REH-expressing firefly luciferase (FFLuc) and green fluorescent protein (GFP) (REH-FFLuc-GFP) were obtained by introducing PB-based pIRII-FFLuc-puroR-GFP (18) in REH cells and subsequent fluorescent-activated cell sorting. REH and REH-FFLuc-GFP cells were cultured in Roswell Park Memorial Institute-1640 medium (Nacalai Tesque) supplemented with 10% fetal bovine serum (Thermo Fisher Scientific, Inc. Waltham, MA) and maintained in a humidified incubator at 37°C in a 5% CO₂ atmosphere.

The PB transposase plasmid, pCMV-piggyBac (24), which contained ~2.4 kb of transposase elements with identical 13 base pair (bp) terminal inverted repeats and additional asymmetric 19 bp internal repeats (25, 26), was artificially synthesized (Mediridge Co., Ltd, Tokyo, Japan). CAR construct for CD19− CAR-T cells which encodes the CD19 specific scFv, followed by a short hinge, the transmembrane and signaling domain of the costimulatory molecule CD28, and the ζ signaling domain of the TCR complex, was kindly provided from Dr. Cliona M. Rooney (Baylor College of Medicine) and was subcloned into pIRII transposon vector backbone (11) (pIRII-CD19-28z) as described previously (27). For the generation of antigen-presenting feeder cells for the stimulation of CD19-CAR-transduced T cells, we used a plasmid containing sequences encoding the extracellular, transmembrane, and 20 amino-acid-long intracellular portion of CD19 protein (tCD19) driven by CMV promoter, followed by CD80 and 4-1BBL (CD137L) with TA2 and P2A self-cleaving sites that enabled independent gene expression. The tCD19-CD80-4-1BBL sequence was artificially synthesized (Fasmac Inc., Kanagawa, Japan) and cloned into a pIRII PB transposon vector (pIRII-tCD19-CD80-4-1BBL) (Supplementary Figure S1) (17, 18).

CD45RA⁺ and CD45RA⁻ PBMCs were isolated by magnetic selection from the whole PBMCs using CD45RA MicroBeads, human (Miltenyi Biotec, Bergisch Gladbach, Germany) (Supplementary Figure S2A). The CD19-CAR transgene was then transduced into these cells using the PB transposon system, as described previously (17, 18). Briefly, pCMV-piggyBac (7.5 μg per 100 µl of electroporation buffer) and pIRII-CD19-28z (7.5 μg per 100 µl) (Supplementary Figure S1) were introduced into about 4 × 10⁶ CD45RA⁺ or CD45RA⁻ PBMCs, respectively, using the P3 Primary Cell 4D-Nucleofector™ X kit (Lonza, Program; FI-115) or MaxCyte ATX^® (MaxCyte Inc) with the optimized protocol for introduction of DNA plasmid into resting T cells (Protocol; RTC 14-3). Concurrently, an antigen-presenting feeder plasmid (pIRII-tCD19-CD80-41BBL; 15 μg per 100 µl) (Supplementary Figure S1) was introduced into approximately 1 × 10⁶ whole PBMCs by electroporation. After electroporation, the CAR-T cells and feeder cells were cultured in complete culture medium consisting of ALyS™705 Medium (Cell Science & Technology Institute) supplemented with 5% artificial serum (Animal-free; Cell Science & Technology Institute), IL-7 (10 ng/ml; Miltenyi Biotec), and IL-15 (5 ng/ml; Miltenyi Biotec). The feeder cells were irradiated with ultraviolet light for inactivation 24 h after electroporation and co-cultured with CAR-T cells for 14 days, as described previously (17, 18). CAR-T cells redirected to the EPHB4 receptor were manufactured using the PB transposon system as described previously (17) for its use as the control CAR-T cells in the in vivo stress test (Supplementary Figure S1).

Expression of CD19-CAR molecules on T-cell surface was measured by flow cytometry using the recombinant human CD19 Fc chimera protein (R&D Systems, Minneapolis, MN, USA) and goat anti-human immunoglobulin (Ig)-G Fc fragment specific antibody conjugated to fluorescein isothiocyanate (FITC) (Merck Millipore, Burlington, MA). Allophycocyanin (APC) or phycoerythrin (PE)-conjugated anti-CD3 antibody, FITC-conjugated anti-CD19 antibody, PE-conjugated anti-CD56 antibody, FITC-conjugated anti-CD15 antibody, APC-conjugated anti-CD14 antibody, APC-conjugated anti-CD8 antibody, PE-conjugated CD4 antibody, PE-conjugated anti-CD45RA antibody, and APC-conjugated anti-CCR7 antibody (all from BioLegend, San Diego, CA, USA) were used to characterize the phenotypes of CAR-T cells. APC-conjugated anti-programmed cell death protein-1 (PD-1) antibody, APC-conjugated anti-T cell immunoglobulin mucin-3 (TIM-3) antibody, Alexa Fluor 647-conjugated anti-CD223 (LAG-3) antibody, and Peridinin-Chlorophyll-Protein (PerCP)/Cyanine5.5-conjugated anti-CD57 antibody were used as the exhaustion and senescence markers of CAR-T cells (all from BioLegend). FITC-conjugated anti-CD19 antibody was also used to determine the phenotype of REH cells. All flow cytometry data were acquired using BD Accuri™ C6 Plus or BD FACSCalibur™ (BD Biosciences, Franklin Lakes, NJ) and analyzed using the FlowJo™ software (BD Biosciences).

After 14 days of culture, 1 × 10⁵ CAR-positive T cells were isolated using a Cell Sorter SH800 (SONY, Tokyo, Japan), and total DNA was then extracted using a QIAamp DNA Mini Kit (QIAGEN, Hilden, Germany). Quantitative PCR was carried out using the total DNA from 1 × 10³ CAR-positive T cells (equivalent to 1 µl of DNA extract) and custom primer/Taqman probe set specific for the CD19-CAR transgene at the junction of CD28 cytoplasmic domain and CD3ζ by a 7500 Real-Time PCR System (Applied Biosystems, Foster City, CA, USA). To measure DNA copy number for absolute quantification, pIRII-CD19-CAR plasmid was used (Supplementary Figure S3A). The relevant primers and Taqman probe are shown (Supplementary Table S1).

The GFP plasmid was introduced into the whole PBMCs by electroporation using the same protocol as that used for CAR-T manufacturing. After electroporation, the GFP-introduced PBMCs were cultured in complete culture medium consisting of the same components used for CAR-T manufacturing. PBMCs without electroporation were cultured in the same medium and was used as the control. After 48 h of incubation, the expression of PD-1, TIM-3, and LAG-3 on T cell surface was analyzed by flow cytometry by gating GFP-positive and CD3-positive T cells.

We co-cultured 1 × 10⁵ REH cells and 1 × 10⁵ CD19 CAR-T cells derived from CD45RA⁺ or CD45RA⁻ PBMC (RA⁺ CAR or RA⁻ CAR, respectively) in 24-well cell culture plates. Three days later, the CD19 CAR-T cells were collected, counted, and treated and reconstituted with fresh REH cells at a ratio of 1:1. Cell counting and treatment with fresh REH cells were repeated every three days for a total of three iterations. The killing effect of these CD19 CAR-T cells was evaluated by counting the number of residual REH cells by flow cytometry. The mean fluorescence intensity (MFI), exhaustion-related markers, cytokine production, and proliferation of the CAR-T cells were analyzed by flow cytometry.

The levels of interferon (IFN)-γ, tumor necrosis factor (TNF), and IL-6 were measured using a Cytometric Bead Array (CBA) Kit (BD Biosciences). Briefly, CAR-T cells were co-cultured with tumor cells at a ratio of 1:1. After 3 days of co-culture, the cell culture supernatant was collected and cytokine levels were determined and analyzed. Data were acquired with a BD Accuri C6 Plus (BD Biosciences) and analyzed with FCAP Array v.3.0 (BD Biosciences).

We examined the proliferation of CAR-T-cells using CytoTell™ Red 650 (AAT Bioquest Inc) dye dilution after serial stimulation with tumor cells. Briefly, CAR-T cells were incubated with CytoTell™ Red 650 dye at 37°C for 30 min, and the dye working solution was removed. Then, the CAR-T cells were co-cultured with tumor cells at a ratio of 1:1 or without tumor cells (no stimulation). After 3 days of co-culture, the proliferation of CAR-T cells was analyzed by flow cytometry, and the experiments were repeated for serial three rounds of antigen stimulation.

We cultured 1 × 10⁶ RA⁺ CAR-T or RA⁻ CAR-T cells in the presence or absence of IL-2 (final concentration; 100 IU/ml) in 24-well cell culture plates. IL-2 was supplemented weekly, and the numbers of live cells were determined every 7 days.

Total RNA was isolated from RA⁺ CAR-T cells and RA⁻ CAR-T cells after 14 days culture (these CAR-T cells were not sorted for CAR-positive population), with or without antigen stimulation, by co-culturing these cells with REH cells at an effector:target ratio of 1:1 for 3 days (RA⁺/Stimulation⁺, RA⁻/Stimulation⁺, RA⁺/Stimulation⁻, and RA⁻/Stimulation⁻, respectively), using RNeasy Mini Kit (Qiagen, Venlo, Netherlands). The concentration of total RNA was measured using NanoDrop 2000 (Thermo Fisher Scientific). Library preparation and high-throughput sequencing were performed using Eurofins Genomics (Ebersberg, Germany). In brief, mRNAs were enriched and their strand-specific library was prepared. Sequencing was performed using a NextSeq 500/550 system (Illumina, San Diego, CA, USA) and NextSeq 500/550 Mid Output Kit v2.5 150 cycles (Illumina). Adapter sequences and low-quality reads were removed using fastp version 0.21.0 (28). Filtered reads were aligned to the human reference genome (GRCh38.p13) using STAR version 2.7.6a (29). Counts per gene and transcripts per million were calculated using RSEM version 1.3.3 (30). Calculation of counts per million and differential expression analysis were performed using edgeR version 3.32.0 R package (31) and R v4.0.3 environment (https://www.R-project.org/). Pathway analysis was performed using R package for the Reactome Pathway Analysis (32). Differential gene expression profiles between RA⁺ and RA⁻ CAR were also analyzed and visualized using Morpheus (https://software.broadinstitute.org/morpheus) and specific gene signatures related to T cell activation, exhaustion, and differentiation (33).

Female 8-week-old NOD.Cg-Prkdc^scidIl2rg^tm1Wjl /SzJ (NSG) mice were purchased from Jackson Laboratory (Bar Harbor, ME, USA), and housed at the Kyoto Prefectural University of Medicine for more than a week before starting the experiment. Food and water were provided ad libitum. REH-FFLuc-GFP cells suspended in D-PBS were infused into mice via the tail vein. Six days later, RA⁺ CAR-T, RA⁻ CAR-T, or irrelevant CAR-T cells, which redirected the EPHB4 receptor (17) as a control, were infused via the tail vein, and tumor burdens were monitored using the IVIS Lumina Series III system (PerkinElmer, Inc.). The regions of interest on the displayed images were quantified in photons per second (ph/s) using Living Image v2 (PerkinElmer, Inc.) as described previously (17). Bone marrow (BM) cells were obtained by sequential BM aspiration from tibias at several time points. The BM cells were stained with a PE-conjugated anti-human CD3 antibody and APC-conjugated anti-human PD-1 antibody (BioLegend), and the long-term persistence of human T cells was evaluated by flow cytometry. The mice were euthanized at predefined endpoints, under conditions that met the euthanasia criteria given by the Center for Comparative Medicine at the Kyoto Prefectural University of Medicine.

Statistical comparisons between two groups were determined by two-tailed parametric or non-parametric (Mann–Whitney U-test) tests for unpaired data or by two-tailed paired Student’s t-test for matched samples. All data are presented as mean ± standard deviation. The log-rank test was used to compare survival curves obtained using the Kaplan–Meier method. A P-value of <0.05 was considered statistically significant. All the statistical analyses were performed using the GraphPad Prism 9 software.

First, we isolated CD45RA⁺ or CD45RA⁻ subpopulations from whole PBMCs using human CD45RA-targeted magnetic separation. We observed two peaks of CD45RA high and negative populations in the lymphocyte fraction (Supplementary Figure S2A). By magnetic bead sorting, CD45RA⁺ and CD45RA⁻ PBMCs were efficiently separated with >98% of CD45RA positive cells in the CD45RA⁺ fraction and >93% of CD45RA negative cells in the CD45RA⁻ fraction (Supplementary Figure S2A). Both RA⁺ and RA⁻ PBMCs consisted of about 60% CD3-positive cells, and the rest of CD3-negative cells were positive for CD56 in RA⁺ PBMCs, while CD56, CD14 and CD15 were negative in RA⁻ PBMCs (Supplementary Figure S2B). Moreover, electroporation did not greatly influence the CD3 positivity of both RA⁺ and RA⁻ PBMCs (Supplementary Figure S2C). Indeed, RA⁺ and RA⁻ PBMCs both contained approximately 40% CD3-negative cells, which included NK cells and other myeloid cells, but this percentage of these cells decreased at 24 h after electroporation and more decreased after 14 days. Based on this observation, CD3-negative T cells were likely destroyed during electroporation since the optimized protocol for introducing the DNA plasmid required a relatively high voltage. We then evaluated the transient gene transfer efficiency of CD19 CAR transgene into CD45RA⁺ or CD45RA⁻ PBMCs 24 h after electroporation. When the CD19 CAR transgene plasmid was introduced into unstimulated, magnetically sorted CD45RA⁺ or CD45RA⁻ PBMCs by electroporation, CD45RA⁺ PBMC subpopulation exhibited higher transient gene transfer efficiency than CD45RA⁻ PBMC subpopulation 24 h after electroporation (Figure 1A). Furthermore, CD19 CAR-T cells derived from CD45RA⁺ PBMCs (RA⁺ CAR-T) exhibited higher expansion capacity 14 days after culture compared with CD19 CAR-T cells derived from CD45RA⁻ PBMCs (RA⁻ CAR-T) (Figure 1B). After 14 days of expansion, we determined the CAR positivity, phenotype, and exhaustion-related marker PD-1 expression on these CAR-T cells by flow cytometry. RA⁺ CAR-T cells consisted of about 95% CD3-positive cells and a small number of CD3-negative/CD56-positive NK cells, while RA⁻ CAR-T cells consisted of about 85% CD3-positive cells and about 13% CD3-negative/CD56-positive cells, suggesting that residual NK cells could affect the immune function of these populations (Supplementary Figure S2D). Compared with RA⁻ CAR-T, RA⁺ CAR-T cells exhibited higher CAR positivity, lower expression of exhaustion-related marker PD-1 and T-cell senescence marker CD57 (34, 35), and enriched naïve/stem cell memory fraction, which were associated with the longevity of CAR-T cells (Figures 1C, D). The copy number of the integrated CAR transgene was calculated by qPCR, and these CAR-T cells had ~20 copies of the CAR transgene (16.5 ± 1.1 in RA⁺ CAR-T cells, 5.9 ± 0.3 in RA⁻ CAR-T cells, and 20.2 ± 1.3 in CD19 CAR-T cells derived from whole PBMCs (Supplementary Figure S3B)), which was consistent with the previous report (11). Furthermore, RA⁺ CAR-T cells were markedly CD8-dominant, whereas the CD4/8 ratio of both RA⁺ and RA⁻ PBMCs, the starting material, was CD4-dominant. These results suggest that CD8-positive stem cell-like T cells are primarily amplified in RA⁺ CAR-T cells (Supplementary Figure S1B). By contrast, RA⁻ CAR-T cells are CD4-dominant in the final product, which may be detrimental to CAR-T cell function as RA⁻ CAR-T cells may harbor more regulatory CD4 CAR-T cells. Other activation/exhaustion-related markers such as TIM-3 and LAG-3 were highly expressed in both the types of CAR-T cells (Figure 1D), and this finding is consistent with that of previous studies on PB-CAR-T cells (14, 16–18, 36). As most PB-CAR-T cells were engineered by electroporation, we hypothesized that the high expression of TIM-3 and LAG-3 would be induced by the stimulation of electroporation. However, PBMCs 48 h after electroporation barely expressed PD-1, TIM-3, or LAG-3 on T cells (Supplementary Figure S3). Therefore, the expression of TIM-3 and LAG-3 but not PD-1 on CAR-T cells was induced during the incubation period and not by the stimulation of electroporation.

To further investigate the effect of RA⁺ CAR-T or RA⁻ CAR-T cells at the molecular level and to identify the pathways involved in the favorable phenotype of RA⁺ CAR-T cells, we performed genome-wide transcriptional profiling by focusing on immunogenic gene signatures. A total of 29 genes were identified with higher expression and 78 genes with lower expression in RA⁺ CAR-T cells compared with CD45RA⁻ CAR-T cells (Figure 1E). Reactome pathway analysis showed that the differential gene expression profiles observed in RA⁺ CAR-T cells were related to the non-canonical NF-κB pathway and co-stimulation by the CD28 family pathway; PD-1 signaling pathway was significantly downregulated in RA⁺ CAR-T cells (Figure 1F). Transcriptome profiling showed that RA⁺ CAR-T cells exhibited activated but less exhausted profiles characterized by the upregulation of T-cell activation-like markers including transcription factor 7 (TCF7), lymphoid enhancer-binding factor 1 (LEF1), CCR7, and IL7R and the downregulation of canonical exhaustion-related markers including PD-1, T-cell immunoreceptor with Ig and ITIM domains (TIGIT), and Eomesodermin (EOMES) in RA⁺ CAR-T cells (33) (Figure 1G, left), even after antigen stimulation (Figure 1G, right). LAG3 expression was also downregulated in RA⁺ CAR-T cells, but flow cytometry results did not show statistical differences between RA⁺ and RA⁻ cells (Figure 1D). These gene expression analysis data suggest that RA⁺ CAR-T cells have an abundant naïve/memory phenotype even when stimulated by antigen-positive tumor cells, corroborating the phenotype of RA⁺ CAR-T cells assessed by flow cytometry.

To evaluate the antileukemic activity of RA⁺ CAR-T and RA⁻ CAR-T cells, we performed the tumor re-challenge assay in which fresh REH cells were added to CAR-T cells every three days. Both the types of CAR-T cells achieved complete killing of REH cells even after multiple rounds of tumor re-challenge (Figure 2A). Interestingly, the expression of CAR molecules on the cell surface after antigen stimulation in RA⁺ CAR-T cells was controlled at a relatively lower level than that in RA⁻ CAR-T cells (Figure 2B), which suggested less tonic CAR signaling and exhaustion of RA⁺ CAR-T cells compared with RA⁻ CAR-T cells (37, 38). PD-1 expression in RA⁺ CAR-T cells was lower than that in RA⁻ CAR-T cells during multiple rounds of antigen stimulation (Figure 2C). In contrast, the expression of LAG-3 was similar in RA⁺ CAR-T and RA⁻ CAR-T cells, whereas TIM-3 expression in RA⁺ CAR-T cells was higher than that in RA⁻ CAR-T cells; these expressions gradually decreased during multiple rounds of antigen stimulation in RA⁺ CAR-T and RA⁻ CAR-T cells, and relatively high expression of TIM-3 and LAG-3 did not impair the killing efficacy of CAR-T cells (Figure 2C). We also evaluated the production of inflammatory cytokines in the cell culture supernatant in response to serial co-culture with tumor cells. The secretion of IFN-γ was higher in RA⁺ CAR-T cells than in RA⁻ CAR-T cells, although transcriptome profiling showed that IFNG expression has been already downregulated in RA⁺ CAR-T cells on day 3 of co-culture with tumor cells (Figure 1G). Interestingly, the levels of TNF and IL-6 were significantly lower in RA⁺ CAR-T cells than in RA⁻ CAR-T cells (Figure 2D). Furthermore, the proliferation of RA⁺ CAR-T and RA⁻ CAR-T cells during serial stimulation was evaluated by CytoTell™ dye dilution with tumor cells. Although RA⁻ CAR-T cells proliferated faster than RA⁺ CAR-T cells without stimulation, there was no significant difference between the proliferation of RA⁺ CAR-T cells and that of RA⁻ CAR-T cells after serial tumor stimulation (Figure 2E). Notably, the proliferation of both RA⁺ and RA⁻ CAR-T cells did not occur spontaneously, but was dependent on antigen stimulation or cytokine supplementation, as confirmed by an IL-2-dependent proliferation assay (Supplementary Figure S4), indicating no disordered proliferative potential in these cells.

To evaluate the in vivo antitumor efficacy of RA⁺ CAR-T and RA⁻ CAR-T cells, we performed the in vivo stress test in which CAR-T cell dosage was lowered to the functional limits, so that these CAR-T cells were maintained and expanded in vivo to achieve antitumor efficacy. Because CAR expression in RA⁺ CAR-T and RA⁻ CAR-T cells was not exactly the same, we injected 1 × 10⁵ CAR-positive T cells in each group. RA⁺ CAR-T cells induced greater tumor reduction and prolonged median survival than those of RA⁻ CAR-T cells (Figures 3A–C). On day15, bone marrow in the RA⁺ CAR group exhibited abundant human CD3 positive T cells with lower expression of PD-1 and a relatively smaller number of REH cells than that in the RA^- CAR group (Figure 3D). Furthermore, in two of the long-lived mice in the RA⁺ CAR group, human CD3⁺ T cells were expanded even after 50 days of treatment as confirmed by sequential bone marrow studies (Figure 3E), which indicated antigen-induced proliferation and long-term functionality of RA⁺ CAR-T cells in vivo. The gating strategy of the bone marrow study was shown in Supplementary Figure S5.

To evaluate whether the selection of CD45RA⁺ PBMCs as a starting material would facilitate highly effective PB-CAR-T cell manufacturing, we performed the in vivo stress test in RA⁺ CAR-T and PB-CD19 CAR-T from unsorted (both CD45RA⁺ and CD45RA⁻) PBMCs (unsorted CAR-T). We infused 1 × 10⁶ REH-FFLuc cells into NSG mice via the tail vein, and six days later, 1 × 10⁵ CAR-positive T cells were infused into each group. RA⁺ CAR-T cells achieved greater tumor reduction and prolonged median survival than unsorted CAR-T cells (Figures 4A–C). Sequential bone marrow studies in the RA⁺ CAR group showed abundant human CD3 positive T cells with lower expression of PD-1 on day 15 (Figure 4D), lower expression of PD-1 in human CD3 positive T cells, and a relatively smaller number of REH cells than the unsorted CAR group (Figure 4E).