Scalable generation of T lymphocytes from human pluripotent stem cells in suspended organoids Authors: Qi Lei, Yudi Miao, Feng Zheng, Ruoqi Cheng, Yating Mo, Weifeng Lai, Zhengyang Zhou, Lejun Lv, Yuanyuan He, Huangfan Xie, Chenyi Dai, Yingfeng Zhang, Zixin Zhao, Mengxin Liu, Lemei Jia, Jun Xu, Cheng Li, Hongkui Deng, Chengyan Wang Published: 2026 ABSTRACT SummaryHuman pluripotent stem cells (hPSCs) are a promising cell source for producing T cells for regenerative medicine and immunotherapy. However, it is challenging to develop a scalable suspension culture system for generating functional T lymphocytes from hPSCs. Here, we developed a Matrigel-based artificial thymic organoid (Gel-ATO) system by encapsulating hPSC-derived induced hematopoietic progenitor cells (iHPCs) with MS5-hDLL4 stromal cells. The resulting Gel-ATOs produced functional iT cells capable of antigen-specific cytotoxicity in vitro and tumor suppression in vivo. Building on this system, we further established a rotating suspension culture to robustly generate iT lymphocytes with nearly 20-fold enhancement of iT cells production compared to stationary suspension culture. Importantly, these rotating suspension culture-derived iT cells were functional, exhibiting cytokine secretion, proliferation, and cytotoxicity. By addressing the dual requirements of scalability and function, our work paves the way for future clinical production of T cells for T cell-based cell therapy.Keywords: human pluripotent stem cells, hPSCs, induced hematopoietic progenitors, iHPCs, induced T cells, iT cells, Matrigel-based artificial thymic organoids, Gel-ATOs, rotating suspension culture Introduction T cell therapy has demonstrated remarkable success in curing cancer and autoimmune diseases; however, its broader application is limited by the lack of scalable “off-the-shelf” T cell production. The development of “off-the-shelf” T cells requires a scalable culture platform for large-scale cell generation and universal approaches to generate genetically modified T cells to prevent immune rejection (Baker et al., 2023; June et al., 2018; Lei et al., 2025). Human pluripotent stem cells (hPSCs) provide a unique cell source to resolve this challenge because of their unlimited self-renewal capacity, differentiation potential into various functional therapeutic cells (Guan et al., 2022; Wang et al., 2024), and ease to be genetically modified (Shi et al., 2017). The derivation of functional T lymphocytes from hPSCs offers a promising solution to provide a robust supply of T cells for potential clinical translation. Over the past decades, the pioneering work is established with two-dimensional (2D) co-culture with OP9 stromal cells expressing Notch ligands DLL1 or DLL4 to promote T-lineage differentiation (de Pooter et al., 2003; Mohtashami et al., 2010). However, this system showed limited efficiency and a tendency to produce γδ T cells (Themeli et al., 2013). Subsequent 2D method utilizing immobilized DLL4 eliminated animal-derived components, which enhanced clinical applicability (Iriguchi et al., 2021; Jing et al., 2025; Michaels et al., 2022; Shukla et al., 2017). More recently, three-dimensional (3D) artificial thymic organoids (ATOs) based on MS5-hDLL4 stromal cells have emerged as a valuable platform (Montel-Hagen et al., 2019; Wang et al., 2022). By providing both DLL4-mediated Notch activation and a biomimetic 3D microenvironment, the ATO system supports robust differentiation from hPSCs to mature αβ T cells, such as anti-CD19 CAR-induced T cells (iT cells), while enabling mechanistic studies of T cell development (Chang et al., 2024; Wang et al., 2022). This system represents a substantial advancement in the field of hPSC-derived T cell generation. However, it is still challenging to develop a scalable suspension-based culture system for generating hPSC-derived T cells in this field. Current state-of-the-art methods, including 2D stromal cell co-culture (Kawai et al., 2021; Themeli et al., 2013), 2D DLL4-immobilized approach (Iriguchi et al., 2021; Jing et al., 2025), and 3D Transwell-based ATO system (Montel-Hagen et al., 2019; Wang et al., 2022), are difficult to adapt for scalable production in suspension culture, hindering their widespread clinical application. One promising strategy to produce large amounts of function T cells from hPSCs is to develop a scalable suspension culture system. Suspension culture systems offer critical advantages for scalable “off-the-shelf” cell therapy production because it could enable linear expansion from lab-scale bioreactors to industrial volumes, supporting large-dose manufacturing; ensure consistent cell quality through real-time monitoring of pH and nutrients; and facilitate good manufacturing practice (GMP)-compliant, reproducible workflows. While other immune cells—including natural killer (NK) cells (Knorr et al., 2013; Thangaraj et al., 2024), macrophages (Gutbier et al., 2020; Shen et al., 2024), and eosinophils (Lai et al., 2021; Zhu et al., 2025)—have been successfully generated from hPSCs using suspension culture systems, generating iT cells from hPSCs in such suspension system faces a unique obstacle. Unlike other lineages, the extended induction process for iT cells stringently requires the sustained and direct contact between induced hematopoietic progenitor cells (iHPCs) or lymphoid progenitors and DLL4-presenting stromal cells to support functional iT cells generation (Montel-Hagen et al., 2019). We proposed a hydrogel-based organoid culture system for generating iT cells from hPSCs, which would resolve this problem and present promising potential for scaling-up. In this study, we first developed a Matrigel-based artificial thymic organoid (Gel-ATO) system and further developed a suspended Gel-ATO system using alginate. Finally, we established an alginate combined VitroGel-RGD (an RGD peptide-functionalized hydrogel) organoid to generate iT cells that exhibited features of T cell function from hPSCs. To further enhance scalability, we developed a dynamic suspended Gel-ATO system with a rotating device. Our study establishes a stable and scalable platform for generating highly proliferative iT cells from hPSCs, paving the way for future clinical-grade production in cancer immunotherapy. Results Generating iHPCs from human chemically induced pluripotent stem cells (hCiPSCs) Building on our previously developed stepwise protocol for iHPC differentiation from hPSCs (Lai et al., 2021; Wang et al., 2012; Zhang et al., 2024; Zhu et al., 2025), we optimized the method to efficiently generate iHPCs from hCiPSCs for subsequent iT cell induction (Guan et al., 2022; Zhang et al., 2024) (Figure 1A). The hCiPSC line hCiPSC-3 was primarily used in this study. iT cell differentiation from hCiPSCs in Gel-ATOs (A) Schematic illustration of iT cell differentiation in Gel-ATO protocol within Gel-ATOs. (B) Bright-field images of Gel-ATOs at the indicated time points (T0, T2, T14, T28, and T35). Scale bars ranged from 100 to 500 µm. (C) Flow cytometric analysis of T-lineage surface marker expression. CD4, CD8, and CD3 were gated on CD7+CD5+ population. (D) Quantification of CD7+CD5+ early lymphoid progenitors and CD4+CD8+ double-positive (DP) cells per Gel-ATO, using different iHPC to MS5-hDLL4 ratios. Data are from three independent experiments (mean ± SD). (E) Yield of CD7+CD5+ lymphoid progenitors and CD4+CD8+ DP T cells per input iHPC at different MS5-hDLL4 to HPC ratio. Data are from three independent experiments (mean ± SD). In the first stage, differentiation of hCiPSCs into primitive streak (PS)-like cells was induced by combined WNT and fibroblast growth factor 2 (bFGF) signaling, yielding nearly 90% BRACHYURY+ PS-like cells by day 2 of differentiation (Figures S1A and S1B). Quantitative reverse-transcription PCR (RT-qPCR) confirmed that these PS-like cells highly expressed genes such as BRACHYURY, MIXL1, EOMES, and MESP1 (Figure S1D). For hemogenic endothelial (HE) cell induction, hCiPSC-derived PS-like cells were treated with a combination of BMP4, VEGF, bFGF, and the TGFβ inhibitors SB-431542 or E-616452 (BVFSB/BVFE-6) (Wang et al., 2012). Flow cytometry analysis revealed that arterial HE cells, expressing DLL4, CD34, CD144, and CD31, were efficiently generated, while lacking CD235a expression (Figure S1C). RNA sequencing (RNA-seq) data analysis demonstrated expression of hematopoiesis-related genes (TAL1, FLI1, and GATA2), and arterial endothelial-related genes (DLL4, GJA5, CXCR4, and EFNB2) (Figure S1E) (Calvanese et al., 2022). During the endothelial-to-hematopoietic transition (EHT), VEGF, SB-431542, and IL-7, combined with hematopoietic cytokines stem cell factor (SCF), Flt3 ligand (FL), and thrombopoietin (TPO), supported the generation of iHPCs from hCiPSCs. By approximately day 12, suspended iHPCs were generated (Figure S1A), and transitioning adherent HE cells were also observed. This differentiation protocol could generate nearly 150-fold iHPCs from a single hCiPSC (Figure S1F). iHPCs expressed CD34, CD43, and CD45, with only a minor population expressing CD144. A population of adherent cells co-expressed CD34 with CD144, CD43, and CD45 (Figure S1G). We isolated these suspended iHPCs and adherent HE cells, and induced them into iT cells in Transwell-based ATOs. Both populations exhibited comparable potential for generating CD7+CD5+CD4+CD8+ iT cells (Figure S1G). We further found that iHPCs derived from BVFSB and BVFE-6 conditions exhibited similar potential to generate CD7+CD5+CD4+CD8+ iT cells in Transwell-based ATOs (Figure S1H). All CD34+ iHPCs/iHE cells were sorted for iT cell induction in this study. Thus, we established an iHPC generation protocol through step-wisely inducing PS-like mesoderm, arterial HE cells, and iHPCs from hCiPSCs for generating iT cells. Development of Gel-ATO for generating iT cells from hCiPSCs Organoids have been developed by combining stem cells, supporting cells, and hydrogels, such as Matrigel, to induce functional lineages from pluripotent stem cells (Aisenbrey and Murphy, 2020; Kim et al., 2022; Liu et al., 2019). We investigated whether Matrigel could be used to generate a Gel-ATO that supports T cell differentiation. We first developed this system by mixing cord blood (CB)-derived CD34+ HSPCs, and MS5-hDLL4 cells to form Gel-ATOs. Differentiating iT cells were collected at various stages for RNA-seq. iT cells could be induced from CB-CD34+ HSPCs in the Gel-ATO system, with T cell-specific genes such as TCF7, BCL11B, LEF1, and GATA3 showing increased expression during differentiation (Yui and Rothenberg, 2014) (Figures S2A and S2B). Then, we explored whether hPSC-derived iHPCs could generate iT cells in this Matrigel-based Gel-ATO system. In this system, compact MS5-hDLL4-iHPC clusters were dispersed in Matrigel at T differentiation day 0 (T0), gradually expanding as stromal cells extended and interconnected to form an organoid (Figure 1B). iT cells emerged and expanded in Gel-ATOs and substantial numbers of iT cells were generated in organoids (Figure 1B). In a representative tracing experiment, CD7-expressing cells were observed in the first week, followed by nearly 70% CD7+CD5+ early thymic progenitor (ETP) cells in the second week. At week-3, over 80% cells were CD7+CD5+, and most of these were CD4+CD8+ cells. By week-4, approximately 90% of CD7+CD5+ cells had differentiated into CD4+CD8+ double-positive (DP) cells. Notably, CD8 expression intensity gradually increased from week-1 through week-4. Finally, some DP cells matured into CD4−CD8+ single-positive (SP) cells, and expression of the functional surface marker CD3 was detected from week-6 to week-7 (Figure 1C). We next evaluated the iT cell differentiation capacity of other human pluripotent stem cell lines. Both the chemically reprogrammed pluripotent stem cell line (hCiPSC-2) and the transcription factor-induced pluripotent stem cell line (hiPSC-7) also generated CD7+CD5+CD4+CD8+ iT cells. Their iT differentiation efficiencies were comparable to those of hCiPSC-3 (Figure S2C). These hPSC-derived iT cells generated in Gel-ATO system did not express the marker CD56 (Figure S2C). T cell receptor (TCR) overexpression has been reported to facilitate the development of mature iT cells that co-express CD3 and TCRαβ during differentiation (Maeda et al., 2020; Minagawa et al., 2018). We therefore induced iT cells from hCiPSCs expressing the New York esophageal squamous cell carcinoma 1 (NY-ESO-1) -specific TCR 1G4 (hCiPSC-1G4), including a lentiviral transduced cell line (Lenti-RFP-1G4) and a CRISPR-Cas9-edited GAPDH knockin cell line (GAPDH-KI-1G4), in the Gel-ATO system (Figures S3A–S3D). Comparison of hCiPSCs and hCiPSC-1G4 showed similar kinetics of CD7, CD5, CD4, and CD8 expression during iT cell differentiation (Figure S3E). CD3+TCRαβ+ iT cells derived from hCiPSC-1G4 cells were detected as early as week-2 of differentiation, whereas hCiPSCs lacking TCR ectopic expression generated CD3+TCRαβ+ cells at only later time points. These results indicate that ectopic TCR expression facilitates the assembly of the TCR-CD3 complex and the transport to the surface of T cell progenitors (Degirmencay et al., 2024). Using 1G4-expressing hCiPSCs, we compared the generation of iT cells in Transwell-based ATOs versus our gel-based ATOs. Tracing experiments revealed that iT cells generated from week-1 to week-4 progressing from CD7+CD5+CD4−CD8− to CD7+CD5+CD4+CD8+ expression. The efficiency and kinetics of generating TCRαβ+CD3+CD4+CD8+ DP iT cells were similar in both systems (Figures S4A and S4B). Moreover, analysis of iT cell yield showed that the generation of CD7+CD5+CD4+CD8+ iT cells was comparable between the two systems (Figure S4B). We next titrated the optimal ratio of CD34+ cells to MS5-hDLL4 for iT cell generation. Different ratios yielded similar efficiency and iT cells counts (Figure S4C). Quantification at early stages indicated that by T21, CD7+CD5+ ETPs or CD4+CD8+ DP cells has expanded most, after which their numbers gradually decreased as mature DP and SP cells were generated (Figure 1D). The number of iT cells generated at different MS5-hDLL4/iHPCs ratios showed similar fluctuation patterns, with the ratio of 1:5 proving relatively more optimal (Figures 1D and 1E). At lower ratios, such as 1:2, larger numbers of early stage iT cells were generated; however, MS5-hDLL4 stromal cells were unable to encapsulate them effectively. Therefore, the 1:5 ratio was applied in subsequent experiments. Assessment of T cell-associated functions in Gel-ATO-derived iT cells We then evaluated the functions of hCiPSC-derived iT cells when CD3+CD4+CD8β+ iT cells were generated at differentiation week-6. CD7+CD5+CD4+CD8β+ DP cells were sorted and activated with anti-CD3/CD28/CD2 stimulating antibodies. After 7-to-14-day activation, we analyzed these iT cells and found that these DP cells were induced into mature CD4−CD8+ SP cells co-expressing CD8α and CD8β (Figure 2A). Functional evaluation of hCiPSC-derived iT cells (A) Representative flow cytometric analysis of naive T cell markers expressed on iT cells and on CD8+ T cells sorted from cord blood (CB). (B) Carboxyfluorescein succinimidyl ester (CFSE)-labeled iT cells were tested after 1 week with anti-CD3/2/28 antibodies activation. Cytotoxic markers CD25 and 4-1BB were analyzed simultaneously. (C) Expansion folds of 1G4-iT cells following anti-CD3/2/28 antibodies stimulation, measured on day 8 and day 14. Data are from three technical replicates (mean ± SD) and are representative of over three independent experiments. (D) Cytotoxic activity of 1G4-iT cells against K562-NY-ESO-1-EGFP-luciferase (K562-NY) or K562-MART1-EGFP-luciferase (K562-MART) tumor cell lines. Statistical significance was determined by two-way ANOVA (p = 0.008). Data points represent three technical replicates (mean ± SD) and are representative of over three independent experiments. (E) Schematic of the in vivo cytotoxicity assay. NPG mice were engrafted with tumors and treated with 1G4-iT cells. Tumor burden was monitored by bioluminescent imaging at the indicated time points. Mice were sacrificed when tumor size reached the predefined limit in accordance with institutional animal welfare guidelines. (F) Quantification of tumor burden shown in (E), presented as mean total flux (photons/second). An unpaired Student’s t test was performed at day 31 (p = 0.0047). Data are from four independent mice (mean ± SEM). These iT cells were CD45RA+CD45RO−; most expressed CCR7, and a subset was CD27+. In parallel, we activated T lymphocytes isolated from CB, which exhibited phenotypes similar to mature iT cells (Figure 2A). Gel-ATO-derived iT cells were labeled with CFSE and activated with anti-CD3/CD28/CD2 antibodies for one week. These CFSE-labeled iT cells had proliferated, as indicated by a decrease in CFSE fluorescence and an increase in co-expression of CD25 and 4-1BB (Figure 2B). We further validated the function of 1G4-iT cells. Given the rapid and robust differentiation kinetics, we sorted CD7+CD5+CD4+CD8β+ DP cells at approximately week-5 of T cell differentiation. Upon activation, 1G4+CD4+CD8β+ iT cells expanded nearly 100-fold after 14 days (Figure 2C). Furthermore, we detected iT cell cytotoxicity after sorting the CD8+NKP44- cells and co-incubated them with K562-NY-ESO-1 or K562-MART1 target cells. These Gel-ATO-generated 1G4-iT cells exhibited antigen-specific cytotoxicity (Figure 2D). Finally, we evaluated the in vivo cytotoxicity of Gel-ATO-derived iT cells. Tumor-bearing mice were generated and were imaged for tumor bioluminescence and allocated into groups with similar average signal intensity between each group at day 0. Expanded iT cells (14 days) were then injected into the mice with medium containing IL-7, IL-2 and IL-15. Bioluminescence imaging showed that Gel-ATO-derived iT cells significantly suppressed tumor growth (Figures 2E and 2F). Thus, Gel-ATO-derived iT cells exhibited measurable T cell-mediated responses, expanding efficiently and secreting cytokines. Importantly, TCR-engineered iT cells also showed antigen-specific cytotoxic activity in vitro, and suppressed tumor growth in vivo. Transcriptional analysis of iT cell differentiation kinetics in Gel-ATOs To characterize T cell differentiation in Gel-ATOs, we sorted T-lineage precursors, DP cells and SP cells, and preformed bulk RNA-seq analysis. Transcriptional profiling revealed the sequential activation of lineage commitment programs similar as the T cell development (Hosokawa and Rothenberg, 2021; Rothenberg et al., 2008; Yui and Rothenberg, 2014). Commitment to the T cell lineage from hPSCs was evidenced by progressive increases transcriptional factor expression, including TCF7, LEF1, and BCL11B across developmental stages, culminating in CD4+CD8+ DP T cells by week-7 of differentiation (Figure 3A). At final stage, maturation phenotypes were evident: early T cell differentiation genes RAG1/RAG2 and PTCRA were downregulated; functional maturation genes such as CCR7, CD27, and KLF2 were upregulated (Figure 3A). These results indicate that our Gel-ATO system could support the maturation of DP into SP cells. RNA-seq analysis showing iT cells were similar to conventional T cells (A) Heatmap showing expression dynamics of key T-lineage genes during iT cell induction. (B) Principal component analysis (PCA) of iT cells and primary lymphocytes from CB, thymus (THY), and periphery blood mononuclear cells (PBMCs). (C) Top GO terms of biological processes enriched in DP versus SP cells. The p value was calculated using the one-sided hypergeometric test. (D) TCRβ CDR3 length of iT cells (blue) compared with thymus T cells (red). (E) Sequencing of TCRβ CDR3 region showing frequencies of Vβ family genes in iT cells (blue) and thymus T cells (red). Further, we clustered our hCiPSC-derived iT cells with thymus-, CB CD8+-, and peripheral blood mononuclear cells (PBMCs)-TCRαβ+-T cells using principal component analysis. The data indicated that Gel-ATO-derived iT cells could be clustered with thymus DP cells and thymus SP cells (Figure 3B). In addition, we performed Gene Ontology (GO) term and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis on the RNA-seq data. Several upregulated pathways are “lymphocyte mediated immunity,” “regulation of immune effector process” (Figure 3C). The gene expression profiling of lymphocytes or T cells suggests a differentiation of T lineage cells from hCiPSCs. We next performed TCR sequencing to examine the TCR repertoire by comparing hCiPSC-derived iT cells with human fetal thymus lymphocytes (Seet et al., 2017). Deep sequencing of the TCR Vβ complementary-determining region 3 (CDR3) showed similar CDR3 length distributions in iT cells and human fetal thymic T cells (Figure 3D). The highly diverse Vβ gene recombination in iT cells was comparable to that in thymic T cells (Figure 3E). These data indicate that Gel-ATO-derived iT cells are highly polyclonal. We further performed single-cell transcriptional analysis to characterize the lymphocyte heterogeneity generated in Gel-ATOs. CD7+ cells were isolated for single-cell RNA sequencing (scRNA-seq) and compared with embryonic development datasets (Figures 4A–4C) (Park et al., 2020). Gel-ATO-derived DP iT cells displayed characteristic T-lineage transcription factors (TCF7 and BCL11B) and surface markers (CD3, CD4, and CD8), while maintaining PTCRA expression, suggesting that they remained developmentally immature and had not yet fully completed positive selection (Figures 4B and 4D). Low-level expression of T cell-associated genes (CD3E, CD3D, CD8A, TRAC, and TCF7) was detected within innate lymphocyte populations. Cluster analysis segregated populations into three groups based on lineage-specific gene-expression patterns, consistent with in vivo developmental trajectories (Figures 4A–4E). DP iT cells were further subdivided into cycling and quiescent states, reflecting thymic progression, with proliferative DP cells transitioning to resting states upon positive selection and SP maturation (Zhang et al., 2022). Collectively, these findings demonstrate that Gel-ATO-derived DP iT cells exhibit transcriptional profiles closely mirroring those of embryonic thymus-developed counterparts. Importantly, iT-cell differentiation outcomes were reproducible across another hCiPSC line (hCiPSC-2; Figures S5A and S5C) and a human iPSC line (hiPSC-7; Figures S5B and S5D). Single-cell RNA sequencing analysis of iT cells and comparison with human fetal thymocytes (A) and (B) Uniform manifold approximation and projection (UMAP) visualization of lymphoid populations from (A), a published dataset of human fetal thymocytes, and (B) sorted CD4+CD8+ iT cells derived from Gel-ATOs in this study. (C) and (D) Dot plot showing the expression of T cell and NK cell signature genes across distinct lymphocyte populations. (E) UMAP showing the expression of key T cell-related genes. Generation of iT cells in suspended Gel-ATOs Matrigel-based Gel-ATOs supported the generation of iT cells from hCiPSCs by enveloping iT cells and MS5-hDLL4 expressing stromal cells. This prompted us to test whether the Gel-ATO system could be cultured in suspension. However, a suspended Matrigel-based Gel-ATO could not be established, primarily because Matrigel poorly forms spherical organoids in suspension and ends up adhering to culture plates. To develop a suspended Gel-ATO system, an appropriate hydrogel was required to co-encapsulate MS5-hDLL4 and iHPCs with suitable stiffness and permeability. Alginate forms organoids (A-ATO) readily in suspension, offering advantages such as a defined chemical composition and low cost (Kozlowski et al., 2021). We evaluated alginate concentrations and identified 0.5% alginate was optimal for iT cell generation (Figures 5A, 5C, and S6A). Generation of static suspended ATOs using alginate-based hydrogels (A) Flow cytometric analysis of T-lineage surface markers on cells differentiated in A-ATOs with varying alginate concentrations. CD4, CD8, and CD3 were gated on CD7+CD5+ population. (B) Quantification of CD7+CD5+ early lymphoid progenitors and CD4+CD8+ DP cells per A-ATO, cultured with different iHPC to MS5-hDLL4 ratios. Data are from three independent experiments. (C) Bright-field images of suspended A-ATOs. Scale bars, 500 μm. (D) Flow cytometric analysis of T-lineage surface markers on cells from A-ATOs. CD4, CD8, and CD3 were gated on CD7+CD5+ population. (E) Yield of CD7+CD5+ lymphoid progenitors and CD4+CD8+DP T cells per AV-ATO at different alginate (ALG) to VGD ratios. Data are from three independent experiments (mean ± SD). (F) Expansion fold of iT cells from AV-ATOs following a 7-day stimulation with anti-CD3/CD2/CD28 antibodies. Data are from three technical replicates (mean ± SD). Representative bright-field images of the expanded cells are shown. Scale bars, 200 μm. In the suspended A-ATO system, iT-cell differentiation kinetics was similar to those in the Matrigel-based Gel-ATOs system. Most cells were CD45+ in week-1, with a subset differentiating into CD7+ lymphoid precursor cells. By week-2, nearly 80% of the cells were CD7+CD5+ lymphocytes, and CD3 and TCR were detectable because of the continuous expression of the 1G4 TCR (Figure 5D). A-ATO-derived iT cells at T35 could proliferate (Figure S6C). Titration of the CD34+ iHPCs-to-MS5-hDLL4 ratio showed similar iT-cell differentiation efficiency and yield across ratios (Figure 5B). Quantitative analysis of CD7+CD5+ and CD4+CD8+ yields revealed a 10-fold lower cell output in the A-ATO system comparing with the Matrigel-based Gel-ATOs system (Figures 1D, 1E, and 5E). We next mixed alginate with Matrigel, but the combination did not improve yields over those of alginate alone (Figure S6B). We therefore combined alginate with VitroGel-RGD, a hydrogel containing the cell-adhesive peptide RGD, to promote cell-matrix interactions during 3D cell culture (Alginate Vitrogel-ATO, AV-ATO) (Moghaddam et al., 2025; Neves et al., 2020). Tracking iT-cell generation showed similar-expression kinetics of CD7, CD5, CD4, CD8, CD3, and TCRαβ between suspended A-ATO and AV-ATO (Figures 5D and S7A). The yields of CD7+CD5+ and CD4+CD8+ cells were enhanced with alginate plus VitroGel-RGD than with alginate alone (Figure 5E) and AV-ATO-derived iT cells could proliferate (Figure 5F). These data indicate that AV-ATO system supports enhanced production of functional iT cells from hCiPSCs. iT cells generated in a rotating suspended Gel-ATO system To increase scalability, we cultured suspended AV-ATOs in rotating flasks as rotating-suspended-Gel-ATO (R-S-ATO) (Figures S7B and S7C). Weekly tracing showed sequential induction of CD7+CD5+ ETP cells, CD7+CD5+CD4+CD8+ DP cells, and a subset of CD7+CD5+CD4−CD8+CD3+TCRαβ+ SP cells, similar to stationary suspended AV-ATOs (Figure 6A; Figure S7A). Scalable production of functional iT cells using rotating suspended ATOs (A) Flow cytometry analysis of T-lineage surface marker expression of iT cells from R-S-ATO. CD4, CD8 and CD3 were gated on CD7+CD5+ population. (B) Quantitative comparison of CD7+CD5+ lymphocytes and CD4+CD8+ DP cells per ATO between AV-ATO and R-S-ATO cultures. Data are from three independent experiments (mean ± SD). (C) Fold change of CD7+CD5+ lymphocyte count at day T28, corresponding to the data shown in (B) (mean ± SD). (D) Expansion folds of R-S-ATO-derived iT cells on day 5 and day 12 after anti-CD3/CD2/CD28 antibodies stimulation. Data are from three technical replicates (mean ± SD) and are representative of three independent experiments. (E) Representative flow cytometric analysis of naive T cell markers on iT cells derived from R-S-ATOs. (F) Flow cytometric analysis of cytokine secretion and activation markers (4-1BB, CD25) following stimulation with phorbol 12-myristate 13-acetate (PMA) and ionomycin. (G) Cytotoxic activity of iT cells against target cells at the indicated effector-to-target (E:T) ratios. Data are from three technical replicates (mean ± SD). (H) Cytotoxic activity of 1G4-iT cells generated in R-S-ATO against K562-NY or K562-MART tumor cell lines. Data are from four technical replicates (mean ± SD). An unpaired Student’s t test was performed (p = 0.0007). Intriguingly, although rotating and static suspension culture exhibited similar differentiation kinetics (Figure 6B), a larger percentage of lymphoid populations and nearly 20-fold enhancement of CD7+CD5+ lymphocytes were generated in R-S-ATOs compared to static suspension culture at T28 (Figure 6C). In one representative culture with a large-scale volume (450 mL), approximately 3 × 107 CD7+CD5+CD3+ iT cells were generated, corresponding to a maximum yield of approximately 66,000 CD7+CD5+CD3+ iT cells per milliliter (Table S1). iT cells generated in R-S-ATOs were activated and expanded nearly 40-folds (Figure 6D), and subsequently induced into mature SP cells (Figure 6E). R-S-ATO-derived SP cells-maintained expression of CD7, CD5, CD8α, CD8β, and CD45RA, and nearly half expressed CD27 (Figure 6E). Furthermore, R-S-ATO-derived SP cells upregulated CD25 and 4-1BB expression and secreted cytokines (Figure 6F). These cells exhibited potent anti-tumor activity against tumor cells (Figure 6G). To further evaluate an antigen-specific killing assay of R-S-ATO-derived iT cells, we sorted CD7+CD5+CD4−CD8+NKp44− SP cells and incubated them with K562-NY-ESO-1 antigen-positive target cells and K562-MART1 antigen-negative target cells, respectively. R-S-ATO-derived cells showed preferential cytotoxicity against antigen-positive targets, whereas killing of antigen-negative targets was quite lower (Figure 6H). In addition, we collected iT cells generated in R-S-ATO and AV-ATO cultures for RNA-seq analysis to investigate the potential mechanisms underlying the enhanced iT-cell production observed in R-S-ATO on day 21. This time point was selected to capture early transcriptional differences that might contribute to the subsequent divergence between the two culture conditions. Compared with AV-ATO, R-S-ATO culture upregulated pathways associated with CD8 TCR signaling, immune responses, and lymphocyte-mediated signaling, while downregulating pathways related to myeloid cell activation. Moreover, ITGAL and ITGB2, which encode the αL and β2 subunits of the force-regulated adhesion receptor LFA-1, respectively, were upregulated in R-S-ATO-derived iT cells (Roy et al., 2020; Valignat et al., 2013). KLF2, a transcription factor previously associated with cellular responses to fluid shear stress, was also elevated (Parmar et al., 2006). Together with the modest increase in PIEZO1 expression (Liu et al., 2024), a gene encoding a mechanosensitive ion channel (Figures S8A–S8C). Collectively, these transcriptional changes are consistent with altered expression of genes associated with cell adhesion and mechanosensing in R-S-ATO cultures; however, further studies are required to determine whether these changes are directly induced by rotational flow or mechanical forces. Taken together, our data support the development of a scalable R-S-ATO system for generating iT cells. The generated cells exhibited selected T-cell-associated response, although additional optimization and validation will be required before potential translational applications. Discussion In this study, we developed a Matrigel-based artificial thymic organoid Gel-ATO system to generate iT lymphocytes from hPSCs. The Gel-ATO-derived iT cells displayed high proliferative capacity and potent cytotoxicity, effectively eliminating tumor cells in vitro and in vivo. Notably, we further developed a suspended scalable rotating Gel-ATO system by utilizing alginate and VitroGel-RGD to stably produce large quantities of iT cells with T cell function properties for clinical application. Importantly, we successfully developed a scalable suspension rotating culture to generate iT cells from hPSCs. Suspension culture is the most scalable approach for advancing the clinical translation because it could be expanded from lab-scale to industrial volumes for manufacturing and also easily standardized. However, a challenge for scalable iT cell production lies in a fundamental biological requirement: the long induction process requires sustained and strong Notch1-DLL4 engagement between iHPCs/lymphoid progenitors and stromal cells, a condition that is difficult to maintain in monolayer induction systems (Montel-Hagen et al., 2019; Nandagopal et al., 2018). Thus, an effective suspension culture system should resolve this obstacle to co-encapsulate DLL4-presenting stromal cells and iHPC/iT precursors within the same microenvironment, preventing precursor cells from migrating away from DLL4-expressing stromal cells during the extended differentiation process. Although a 3D Transwell-based ATO system has already been established to generate iT cells from hPSCs (Chang et al., 2024; Montel-Hagen et al., 2019; Wang et al., 2022), its reliance on a liquid-air interface for organoid formation in Transwell fundamentally restricts scalability. Therefore, in order to overcome this problem, we aimed to develop a novel suspension culture system to stepwise induce T cells from hPSCs. Firstly, we utilized Matrigel to develop a system to generate iT cells from hPSCs in a Matrigel-based Gel-ATO system since Matrigel could encapsulate stromal cells and iT progenitors within ATOs. In this system, functional iT cells were generated and they exhibited antigen-specific cytotoxicity in vitro and tumor suppression in vivo. This approach makes it feasible to further develop a suspended culture system in the following through encapsulating DLL4-stromal cells and Notch1-expressing iHPCs/iT cells. Secondly, we defined alginate as a potential hydrogel with fine-tuned stiffness and viscoelasticity to co-encapsulate stromal and iT cells securely while maintaining sufficient permeability for nutrient diffusion and cell viability (Lai et al., 2025). In our study, we found that Matrigel-based Gel-ATO system cannot be applied to develop suspension culture conditions. We attributed this limitation to the soft mechanical properties of Matrigel, which produces a relatively loose 3D structure. We identified alginate as a suitable material for forming compact organoids in suspension culture that supports iT differentiation. Alginate is widely used for the 3D culture of stem cells and tumor organoids due to its low cell adhesiveness, which also facilitates the formation of suspended Gel-ATOs (Kozlowski et al., 2021). However, unmodified alginate hydrogels exhibited weaker mechanical properties and yielded fewer iT cells than Matrigel-based systems (Figures 1D and 5B). Based on the evidence that modifying hydrogels with RGD peptides improved cell viability and differentiation potential (Neves et al., 2020), we incorporated VitroGel-RGD into alginate hydrogels to generate suspended AV-ATOs, which indeed increased iT-cell production (Figure 5E). Finally, we succeeded in developing a rotating suspended Gel-ATO culture system. Compared with stationary suspension cultures, our rotating system further increased iT cell yields with nearly 20-fold, likely because of improved nutrient distribution and the beneficial effects of shear stress that mimic physiological blood flow (Adamo et al., 2009; Ito et al., 2018). This rotating bioreactor enables large-volume suspension cultures, establishing a foundation for future mass production of hPSC-derived iT cells. Notably, our rotating suspension culture system produced-iT cells are functional. While scalability is essential, the functional competence of hPSC-derived iT cells is paramount for their therapeutic utility. By leveraging a 3D ATO system, we successfully incorporated hydrogel-based encapsulation of DLL4-expressing supportive stromal cells and iHPCs/iT cells. Our system simultaneously addresses both scalability and functionality. To date, a suspension bioreactor culture system was reported recently by using soluble notch agonists to confer the phenotypic iT generation but no anti-tumor functions have been demonstrated (Mout et al., 2025). The suspended 3D approach containing supporting stromal cells is particularly valuable for T lymphocytes, as T cell differentiation and maturation rely on fine-regulated, spatial cues within the thymic niche—a microenvironment that is difficult to replicate in conventional 2D cultures (Ashby and Hogquist, 2024). The incorporation of supportive stromal cells within tunable hydrogel matrix could mimic thymic microenvironment to regulate T cell differentiation and maturation. Therefore, our hydrogel-based 3D suspension culture system represents a robust and promising platform for the generation of functional, mature iT cells, offering a feasible approach toward translation. In our Gel-ATO or R-S-ATO culture, hPSC-derived T cells have acquired most of functional features such as the expression of CD3 and TCRαβ, response to anti-CD3/CD28 antibodies to proliferate, and exhibit the activity of anti-tumor cells, but our current T induction system did not achieve in efficiently generating fully mature primary-equivalent SP T cells. In our study, only a portion of T cells are CD8+SP cells and some of these induced T cells express CCR7 and CD62L surface markers. In our future work, we will further optimize our system to improve mature SP generation of CCR7 expression, and generate full functional mature T lymphocytes. Although the scRNA-seq and TCR repertoire analyses support thymocyte-like differentiation and the generation of a diverse TCR repertoire and transcriptional resemblance to fetal thymic outputs do not necessarily establish fully physiological positive and negative selected or full matured T cells induced. The persistence of PTCRA expression in a subset of DP thymocytes suggests that they remained developmentally immature and had not yet fully completed positive selection. In addition, the emergence of an NK or innate lymphoid cell (ILC)-like population indicates product heterogeneity and suggests that the differentiation system may generate non-T-lineage or incompletely specified lymphoid populations. Future studies will therefore be focused on generating physiological thymic selection and mature T cell functions. Although our R-S-ATO possesses great scalable potential, the use of MS5-hDLL4 stromal cells in our study remains a major translational limitation. MS5 is of murine origin, this feeder-based component is not readily compatible with xeno-free, GMP-compliant manufacture system. Therefore, while our rotating suspension system supports the improved T-lineage induction and cell yield, and the scalable potential for future enlarged volume of culture to generate iT cells. It should be viewed as a proof-of-concept culture format rather than a clinically validated manufacturing platform. Future optimization will require replacement of MS5-hDLL4 with clinically compatible Notch-ligand presentation systems such as to present hDLL4 overexpressed in hPSC-derived stromal cells. In summary, we have developed a rotating hydrogel-based suspended ATO system that enabled the scalable production of hPSC-derived-iT cells. This work provides a robust platform for large-scale generation of functional iT cells, establishing a foundation for future clinical applications in cancer immunotherapy. Resource availability Lead contact Further information and requests for reagents and resources should be directed to and will be fulfilled by the lead contact, Wang Chengyan (chengyanw@pku.edu.cn). Materials availability This study did not generate new unique reagents. Data and code availability •The raw sequence data reported in this paper have been deposited in the Genome Sequence Archive (GSA) in National Genomics Data Center, China National Center for Bioinformation/Beijing Institute of Genomics, Chinese Academy of Sciences (GSA-Human: HRA013688) that are publicly accessible at https://ngdc.cncb.ac.cn/gsa-human/s/YA6K0CZw. This paper analyzes existing, publicly available data. These accession numbers for the datasets are listed in the key resources table. The raw sequence data reported in this paper have been deposited in the Genome Sequence Archive (GSA) in National Genomics Data Center, China National Center for Bioinformation/Beijing Institute of Genomics, Chinese Academy of Sciences (GSA-Human: HRA013688) that are publicly accessible at https://ngdc.cncb.ac.cn/gsa-human/s/YA6K0CZw. This paper analyzes existing, publicly available data. These accession numbers for the datasets are listed in the key resources table. •This paper does not report original code. This paper does not report original code. •Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request. Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request. Acknowledgments We thank Xiao Yang for gifting us the MS5-GFP cell line. We thank Professor Xin Lin from the Tsinghua University School of Medicine for kindly gifting us 1G4-plasmid. We thank the Flow Cytometry and Living Imaging Core at the National Center for Protein Sciences at Peking University. We thank Dr. Zhengyuan Zhang, Jingyang Guan, Jinlin Wang, Yao Fu and Gaofan Meng for the support of hCiPS cells. We thank Dr. Yan Shi and Lulu Liu for the suggestion and support to set up rotating suspended culture system. We thank Dr. Yangin Li for helpful suggestions for manuscript revision. This work was supported by Changping Laboratory funding (2021C-04-05), National Natural Science Foundation of China (32288102), and Beijing Science and Technology Major Project Z191100001519001 to H.D. Author contributions C.W., H.D., Q.L., Y. Miao, Y. Mo, and W.L., designed the study and interpreted the experiments; Q.L., F.Z., Y. Miao, Y. Mo, W.L., L.L., H.X., Z.Z., M.L., and Y.Z., performed the in vitro experiments; R.C., C.L., and Y.H., performed and analyzed the bulk RNA-seq and scRNA-seq data; Y. Miao, Z.Z., and L.J., designed, constructed to develop the CRISPR-Cas9 knockin cell line; C.W., Z.Z., and C.D., designed and performed in vivo experiments; C.W., H.D., Q.L., Y. Miao, and J.X., wrote the manuscript; H.D., C.W., conceived and supervised this project. All authors contributed feedback on the final manuscript. Declaration of interests C.W., H.D., W.L., and Y.Z. have a patent application related to this work. STAR★Methods Key resources table REAGENT or RESOURCESOURCEIDENTIFIERAntibodies7-AAD Viability Staining SolutionBioLegendCat#420404Goat anti-Human/Mouse Brachyury antibodyR&D SystemsCat#AF2085; RRID: AB_2200235FITC anti-human CD31BiolegendCat#303104; RRID: AB_314330PE anti-human KDRBioLegendCat#359904; RRID: AB_2562704APC/Cyanine7 anti-human CD34 AntibodyBioLegendCat#343514; RRID: AB_1877168BUV395 anti-human CD235a AntibodayBD BioscienceCat#563810; RRID: AB_2738434APC anti-human DLL4 antibodyBioLegendCat#346508; RRID: AB_11204071PE anti-human CD43 AntibodyBioLegendCat#343204; RRID: AB_2255209Brilliant Violet 421™ anti-human CD45 AntibodyBioLegendCat#304032; RRID: AB_2561357APC anti-human CD45 AntibodyBioLegendCat#304012; RRID: AB_314400FITC anti-human CD7 AntibodyBioLegendCat#343104; RRID: AB_1659216APC anti-human CD4 AntibodyBioLegendCat#300514; RRID: AB_314082CD8b Monoclonal Antibody (SIDI8BEE), PE-Cyanine7InvitrogenCat#25-5273-42; RRID: AB_11219680APC/Cyanine7 anti-human TCR α/β AntibodyBioLegendCat#306728; RRID: AB_2566601PE anti-human CD3 AntibodyBioLegendCat#300308; RRID: AB_314044Brilliant Violet 421™ anti-human CD3 AntibodyBioLegendCat#300434; RRID: AB_10962690FITC anti-human CD3 AntibodyBioLegendCat#317306; RRID: AB_571907Brilliant Violet 605™ anti-human CD5 AntibodyBioLegendCat#364020; RRID: AB_2565941PE anti-human CD5 AntibodyBioLegendCat#364014; RRID: AB_2565284APC/CY7 anti-human CD5 AntibodyBioLegendCat#364010; RRID: AB_2564506PE anti-human CD45RA AntibodyBioLegendCat#304108; RRID: AB_314412PE anti-human CD8a AntibodyBioLegendCat#300908; RRID: AB_314112APC/Cyanine7 anti-human CD27 AntibodyBioLegendCat#302815; RRID: AB_571976Brilliant Violet 605™ anti-human CD28 AntibodyBioLegendCat#302968; RRID: AB_2800755Brilliant Violet 650™ anti-human CD56 (NCAM) AntibodyBioLegendCat#318344; RRID: AB_2563838APC anti-human CD336 (NKp44) AntibodyBioLegendCat#325110; RRID: AB_2149432FITC anti-human CCR7BioLegendCat#353216; RRID: AB_10916386APC/CY7 anti-human CCR7BioLegendCat#353212; RRID: AB_10916390APC anti-human CD45ROBioLegendCat#304210; RRID: AB_314426APC anti-human TCRVβ 13.1BioLegendCat#362408; RRID: AB_2728349APC anti-human CD25 AntibodyBioLegendCat#302610; RRID: AB_314280PE/Cyanine7 anti-human CD137 (4-1BB) AntibodyBioLegendCat#309818; RRID: AB_2207741Alexa Fluor 488-AffiniPure donkey anti-mouse IgGJackson ImmunoResearch,Cat#715-545-150; RRID: AB_2340846Brilliant Violet 421™ Mouse IgG1, κ Isotype Ctrl AntibodyBioLegendCat#400157; RRID: AB_10897939PE/Cyanine7 Mouse IgG1, κ Isotype Ctrl AntibodyBioLegendCat#400125; RRID: AB_2861433PE Mouse IgG1, κ Isotype Ctrl AntibodyBioLegendCat#400111; RRID: AB_2847829APC/Cyanine7 Mouse IgG2a, κ Isotype Ctrl AntibodyBioLegendCat#400229; RRID: AB_326477Biological samplesCord blood Chinese PLA General Hospital N/ACord blood Chinese PLA General Hospital N/ACord blood Chinese PLA General Hospital N/AChemicals, peptides, and recombinant proteinsY-27632MCECat#HY-10071PolybreneMilliporeCat#TR-1003-GChroman 1MCECat#HY-15392EmricasanMCECat#HY-10396Recombinant human bone morphogenetic protein 4 (BMP-4)StemImmune LLCCat#HST-B4-1000CHIR-99021SelleckCat#S1263Recombinant human fibroblast growth factor 2 (bFGF)OriGeneCat#TP750002L-Ascorbic acid 2-phosphate sesquimagnesium salt hydrate (ascorbic acid)Sigma-AldrichCat#A89601-ThioglycerolSigma-AldrichCat#M6145Recombinant human vascular endothelial growth factor (VEGF)StemImmune LLCCat#HVG-VF5-1000SB-431542MCECat#HY-10431E-616452MCECat#HY-13012Recombinant human stem cell factor (SCF)StemImmune LLCCat#HHM-SF-1000Recombinant human fms-like tyrosine kinase 3 ligand (FLT3L)StemImmune LLCCat#HHM-FT-1000Recombinant human thrombopoietin (TPO)StemImmune LLCCat#HHM-TP-1000Recombinant human IL-7StemImmune LLCCat#HCT-I7-1000Recombinant human IL-2NovoproteinCat#C013N-acetyl-L-cysteineSigma-AldrichCat#A7250Minocycline hydrochlorideMCECat#HY-17412Recombinant human IL-21PeproTechCat#200-21-100Z-VAD-FMKMCECat#HY-16658BBovine serum albumin (BSA)Sigma-AldrichCat#A1470-100GD-Luciferin potassium saltBioMolCat#E011306Human lymphocyte separation mediumDRKEWECat#7111011ImmunoCult™ Human CD3/CD28/CD2 T cell ActivatorSTEMCELL TechnologiesCat#10970Vitronectin XFSTEMCELL TechnologiesCat#100-0763mTeSR PlusSTEMCELL TechnologiesCat#100-0276ReLeSRSTEMCELL TechnologiesCat#05873Dulbecco’s Modified Eagle Medium (DMEM)GibcoCat#C11965500BTFetal bovine serum (FBS)GeminiBioCat#900-108Penicillin/streptomycin (PS)GibcoCat#15140122Trypsin-EDTA (0.25%)GibcoCat#25200056RPMI 1640GibcoCat#22400089Non-essential amino acids (NEAAs)GibcoCat#11140050GlutaMAXGibcoCat#35050061B-27 Supplement (without vitamin A)GibcoCat#12587010Iscove’s Modified Dulbecco’s Medium (IMDM)GibcoCat#C12440500BTAccutaseMerck MilliporeCat#SCR005ImmunoCult™-XF T cell Expansion MediumSTEMCELL TechnologiesCat#10981Insulin-transferrin-selenium (ITS-G)GibcoCat#41400045Fixable Viability Stain 575VBD HorizonCat#565694BD Cytofix/Cytoperm™ Fixation/Permeabilization KitBD BiosciencesCat#554714BD Perm/Wash™ BufferBD BiosciencesCat#554723Human CD8 MicroBeadsMiltenyi BiotecCat#130-045-201Human CD34 MicroBeadsMiltenyi BiotecCat#130-046-703eBioscience™ Cell Stimulation Cocktail (PMA/Ionomycin)InvitrogenCat#00-4970-93Deposited dataRNA-seq dataThis paperGSA-Human: HRA013688Human thymus data for single cell RNA-seq analysis(Park et al., 2020)Development Cell Atlas: https://developmentcellatlas.ncl.ac.uk; ArrayExpress: E-MTAB-8581Human thymus data for Bulk RNA-seq analysis(Montel-Hagen et al., 2019)GSE116015Human PBMC data for Bulk RNA-seq analysis(Jing et al., 2025)GSE195667Experimental models: Cell linesHuman chemically induced pluripotent stem cell line CiPS-3(Guan et al., 2022)N/AHuman iPS cell line iPS-7Cauliscell Biotechnology200101MS5-GFP cell lineXiao Yang (Academy of Military Sciences)N/AK562 cell lineProcellCL-0130Experimental models: Organisms/strainsNOD.Cg-Prkdcscid Il2rgtm1Vst/Vst (NPG) miceBeijing Vitalstar BiotechnologyVS-AM-001OligonucleotidesCRISPR/Cas9 sgRNA for GAPDH:5′-GGGTCTACATGGCAACTGTG-3′GenScript Co.N/ARecombinant DNApCDH-hDLL4-T2A-Puro-MSCVThis paperN/ApHAGE PGK-GFP-IRES-LUC-WKe Lei Biological TechnologyCat#46793(plenti3)EF1a-GFP-2A-luc2-SV40-MART1-B2M-HLA0201This paperN/A(plenti3)EF1a-GFP-2A-luc2-SV40-NYESO1-B2M-HLA0201This paperN/ApHAGE-1G4-IRES-mRFP1Tsinghua UniversityN/ASoftware and algorithmsFlowJo version 10.8.2BDhttps://www.flowjo.com/solutions/flowjocellSensEvident Scientifichttps://lifescience.evidentscientific.com.cn/en/support/downloads/Living Image SoftwarePerkinElmerhttps://www.perkinelmer.com.cn/lab-products-and-services/resources/in-vivo-imaging-software-downloads.html?_ga=2.29909951.1041849854.1637318373-1604795059.1637318373#LivingImageCell Ranger software (version 5.0.1)10× Genomicshttps://www.10xgenomics.com/cn/support/software/cell-ranger/latestR software (version 4.2.2)R Core Team, R Foundation for Statistical Computing, Vienna, Austriahttps://www.r-project.org/Trimmomatic (version 0.39)(Bolger et al., 2014)http://www.usadellab.org/cms/?page=trimmomaticSTAR (version 2.7.10b)(Dobin et al., 2013)https://software.cqls.oregonstate.edu/updates/star-2.7.10b/featureCounts (version 2.0.6)(Liao et al., 2014)http://subread.sourceforge.net/Sva (version 3.50.0)(Leek et al., 2012)https://www.bioconductor.org/packages/release/bioc/vignettes/sva/inst/doc/sva.pdfSeurat (version 4.3.0.1)(Stuart et al., 2019)https://satijalab.org/seurat/DESeq2 (version 1.42.1)(Love et al., 2014)https://bioconductor.org/packages//release/bioc/html/DESeq2.htmlclusterProfiler (version 4.10.0)(Yu et al., 2012)https://guangchuangyu.github.io/software/clusterProfiler/MiXCR (version 4.7.0)(Bolotin et al., 2015)https://github.com/milaboratory/mixcrAdobe IllustratorAdobeRRID: SCR_010279Adobe PhotoshopAdobeRRID: SCR_014199GraphPad PrismGraphPad Softwarehttps://www.graphpad.com/scientific-software/prism/ Experimental model and subject details Mice Mouse experiments were conducted in accordance with a protocol approved by the Institutional Animal Care and Use Committee (IACUC) of Peking University (LSC-DengHK-15). All experiments used 6 to 8-week-old male NOD.Cg-Prkdcscid Il2rgtm1vst/Vst (NPG, stock number: VS-AM-001) mice purchased from Beijing Vitalstar Biotechnology. Cord blood cells for sorting CD34+ HSPCs This protocol was approved by the Ethics Committee of the Chinese PLA General Hospital (S2022-436-01). Cord blood samples were collected from consenting donor using procedures that were ethically approved at the Chinese PLA General Hospital. Human PSC cell lines and culture Human chemically induced pluripotent stem cell lines were established in our group (Guan et al., 2022) and two independent cell lines (hCiPSC-3 and hCiPSC-2) were used in this study. The human iPSC line iPS-7 was purchased from Cauliscell Biotechnology (Cat. 200101). hCiPS and hiPS cells were maintained in six-well plates coated with Geltrex (ThermoFisher Scientific, Cat. A1413302) or Vitronection (STEMCELL Technology, Cat. 100-0763) and were cultured using mTeSR™ Plus Medium (STEMCELL Technology, Cat. 100-0276) in 5% CO2 at 37°C. Method details Lentiviral vectors and transduction for generating hCiPS-1G4 cells The lentiviral vector encoding the HLA-A∗02 restricted NY-ESO-1 specific TCR 1G4 was kindly donated by Professor Xin Lin at Tsinghua University. The 1G4 gene fragment was cloned into pHAGE downstream of EF1α promoter with mRFP1 cassette. The pHAGE-1G4-RFP1 plasmid was packaged to produce lentiviral 1G4-RFP1 virus particles according to a previously described protocol. The hCiPSC-3 cells were transduced with the pHAGE-1G4-RFP1 pseudo-virus in the presence of 8 μg/mL polybrene, and medium was changed after 24 h. hCiPSC-3-1G4-RFP1+ cells were sorted more than three times before use in differentiation experiments. CRISPR/Cas9 construct for developing hCiPS-GAPDH-KI-1G4 cells A single guide RNA (sgRNA) targeting exon 9 of GAPDH was designed using CRISPOR (CRISPOR.org) and synthesized by GenScript (5′-GGGTCTACATGGCAACTGTG-3′). The homology-directed repair (HDR) donor vector was designed to insert the remaining GAPDH exon 9 sequence downstream of the cleavage site followed by a self-cleaving 2A peptide (P2A) and the 1G4 TCR and a PGK promoter driven puromycin-resistance gene between GAPDH homology arms (Figure S3A, synthesized by GENECFPS). This gene cassette ensured that the 1G4 TCR and puromycin-resistance gene were inserted in-frame at the stop codon of GAPDH exon 9. For CRISPR/Cas9-mediated knock-in, the dsDNA HDR donor was prepared by PCR using primers targeting the 5′ and 3′ ends of the left and right homology arms, forward 5′-GGTATCGTGGAAGGACTCATGG-3′ and reverse 5′-GCCAGGGGTCTGGACCTAG-3′. SpCas9 protein (IDT, Cat. 1081059), GAPDH sgRNA, and the HDR dsDNA donor were co-nucleofected into 5×104 hCiPSC-3 cells using the P3 Primary Cell 4D-Nucleofector® X Kit L (LONZA, Cat. V4XP-3024) according to the manufacturer’s protocol. Nucleofection was performed with 4D-Nucleofector® X Unit (LONZA, Cat. AAF-1003X) and the program used was CA137. The nucleofected single cells were then diluted and seeded into Vitronectin-coated six-well plates for the following clone picking. For engineered clone validation, hCiPSCs were selected with 0.5 ng/mL puromycin (MCE, Cat. HY-B1743A) for six days, starting at two days after nucleofection. After approximately seven days, single-cell-derived clones were hand-picked and seeded into 24-well plates. Genomic DNA from single-cell clones was genotyped using forward primer 5′-AAGCTCATTTCCTGGTATGTGGC-3′ and reverse primer 5′-AGTAACTGGTTGAGCACAGGGTA-3′. Cell mRNA was also extracted for RT-qPCR validation using primers targeting the 1G4 TCR sequence: forward 5′-CTGTGGCTTTACCTCGGTGT-3′ and reverse 5′-TGACCAGCACAGCATACAGG-3′. Hematopoietic differentiation from hCiPSCs Figure 1 shows a schematic overview of the differentiation protocol. hCiPSCs were first harvested after ReLeSR (STEMCELL Technology, Cat. 5873) for differentiation. The digested hCiPSCs were suspended in mTeSR™ Plus Basal Medium supplemented with 10 μM Y-27632 (MCE, Cat. HY-10071), and seeded at a density of 2-3×105 cells in Vitronectin-coated six-well plates. After being cultured for 2-3 days, on day 0 of hematopoietic differentiation, mesoderm induction was initiated with 1640/B27 differentiation medium containing 10 μM CHIR-99021 (Selleck, Cat. S1263), 20 ng/mL bFGF (Origene, Cat. TP750002), and 2 μM Y-27632. The 1640/B27 differentiation basal medium consisted of RPMI 1640 (Gibco, Cat. 22400-089) supplemented with 2% B27 Supplement (Gibco, Cat.12587-010), 100 μg/mL L-ascorbic acid 2-phosphate sesquimagnesium salt hydrate (ascorbic acid) (Sigma Aldrich, Cat. A8960), 1% GlutaMAX (Gibco, Cat. 35050-061), 1% penicillin-streptomycin (Gibco, Cat. 15140-122), 0.04 μL/mL 1-Thioglycerol (Sigma Aldrich, Cat. M6145-25ML), and 1% MEM non-essential amino acids (Gibco, Cat. 11140-050). On day 2, the medium was replaced with BVFE-6, containing 10 ng/mL BMP4 (B) (Stemimmune LLC, Cat. HST-B4-0100), 50 ng/mL VEGF (V) (Stemimmune LLC, Cat. HVG-VF5-0100), 50 ng/mL bFGF (F), and 2 μM E-616452 (E-6) (MCE, Cat. HY-13012), with medium changes every day. On day seven, when endothelial-to-hematopoietic transition (EHT) stage was initiated, the basal medium was replaced with IMDM/B27, consisting of IMDM (Gibco, Cat. C12440500CP) supplemented with 2% B27 Supplement, 100 μg/mL L-ascorbic acid 2-phosphate sesquimagnesium salt hydrate (ascorbic acid), 1% GlutaMAX, 1% penicillin-ptreptomycin, 0.04 μL/mL 1-Thioglycerol, 1% MEM non-essential amino acids, 24 μM N-acetyl-L-cysteine (MCE, Cat. HY-B0215), and 2 μM minocycline hydrochloride (MCE, Cat. HY-17412). The IMDM/B27 medium was supplemented with 20 ng/mL SCF (Stemimmune LLC, Cat. HHM-SF-0100), 20 ng/mL TPO (Stemimmune LLC, Cat. HHM-TP-1000), 20 ng/mL FLT3 Ligand (FL, Stemimmune LLC, Cat. HHM-FT-0100), 5 ng/mL VEGF, 10 μM SB-431542, and 10 ng/mL IL-7 (Stemimmune LLC, Cat. HCT-I7-1000). Isolation of cord blood CD34+ HSPCs Cord blood mononuclear cells were processed as follows: first, blood was diluted with sterile phosphate-buffered saline (PBS) (Sangon Biotech, Cat. E607008-0500), gently layered onto human lymphocyte separation medium (DRKEWE, Cat. 7111011) and centrifuged at 1800 rpm for 20 min. This process yielded distinct layers, the middle layer, containing mononuclear cells, was collected and diluted with sterile PBS. The collected cells were centrifuged again at 1800 rpm for 20 min, and the resulting pellets were retained. For CD34+ HSPCs isolation, mononucleated cells were incubated for 15 min protected from light with human CD34 microbeads (Miltenyi Biotech, Cat. 130-046-703) in PBS containing 0.5% bovine serum albumin (BSA) (Sigma Aldrich, Cat. A1933-100G). Cells were then washed twice with sterile PBS, and CD34+ cells were isolated using a magnetic-activated cell sorting (MACS) system with MiniMACS Separator (Miltenyi Biotech, Cat. 130-042-102). MS5-hDLL4 stromal cell line The MS5-GFP cell line was a kind gift from Professor Xiao Yang (Academy of Military Sciences) and was transduced with a lentiviral vector. The hDLL4 (OriGene, Cat. SC113239) gene was cloned into the pCDH-MCS-T2A-Puro-MSCV vector and lentiviral hDLL4-Puro virus was obtained in accordance with the previously described protocol, which was then transduced into MS5-GFP cells. GFP+DLL4+ MS5 cells were sorted and expanded. The resulting MS5-hDLL4 cell line was cultured in Dulbecco’s modified Eagle medium (DMEM) (Gibco, Cat. C11965500CP) containing 10% FBS (GEMINI, Cat. 900-108) and 1% Penicillin-Streptomycin with 5% CO2 at 37°C. Cells were passaged using Trypsin-EDTA (0.25%) (Gibco, Cat. 25200-056). T cell differentiation from hPSC-derived iHPCs For Matrigel-based Gel-ATO differentiation, on day 12, MS5-hDLL4 cells were trypsinized and co-cultured with magnetically sorted CD34+ iHPCs at specified ratios (typically 1:5 or 1:2 unless otherwise stated) in T culture medium (TCM). TCM was based on IMDM/B27 medium, as described above, and supplemented with 10 ng/mL SCF, 5 ng/mL Flt3 Ligand (FL) and 5 ng/mL IL-7. The cell mixture was seeded at 3×106 cells per well in Ultra-low-attachment six-well plates (Corning, Cat. 3471) and incubated overnight. In the following day, cell aggregates formed in the supernatant were collected and suspended in a 1:1 mixture of Matrigel (Corning, Cat. 354230) and TCM. For the Gel-ATO system, 2.5-3×106 cells were suspended in 100 μL of the Gel-Cell aggregate mixture and 4 μL as droplet was dispensed into each well of 24-well plate. After incubation at 37°C for 20 min to allow Matrigel polymerization, 1 mL of TCM was added to each well. Medium was changed every 3-4 days. For T-cell differentiation of CB CD34+ HPSCs, CB-derived CD34+ HSPCs were magnetically sorted and mixed with MS5-hDLL4 cells at a ratio of 1:100. The cell mixture was seeded at 3×106 cells per well in low-attachment six-well plates and incubated overnight. The following day, Gel-ATOs were generated using the procedure described above. For transwell-based ATO differentiation, iHPCs and MS5-hDLL4 cells were mixed at specified ratios, centrifuged, and resuspended in TCM (50 μL per 2.5-3×106 cells). The cell suspension was dispensed as 4 μL droplets onto transwell inserts in 12-well plates (Labselect, Cat. 14211), with 500 μL of TCM added to the lower chamber. Medium was changed every three days. For suspended A-ATO differentiation, MS5-hDLL4 cells and iHPCs were mixed at specified ratios and incubated overnight as described above. The following day, 3×106 cell aggregates were collected and resuspended in a mixture of 48 μL TCM and 48 μL 1% sodium alginate (Sigma Aldrich, Cat. W201502). Twenty-four droplets of 0.08 M CaCl2 (Sigma Aldrich, Cat. C5670-100G) (2 μL per each droplet) were dispensed into a 10 cm-dish. The cell suspension was then mixed with CaCl2 at 4 μL per each droplet and was incubated at room temperature for 10 min. Subsequently, 10 mL of 0.08 M CaCl2 was added to the dish, followed by incubation at 37°C for 10 minutes to complete alginate polymerization. The CaCl2 solution was removed, and the dish was washed once with 10 mL PBS before resuspending the suspended-Gel-ATOs in 10 mL TCM. Alginate and CaCl2 were resolved in 0.9% NaCl. By further improving, VitroGel-RGD (TheWell Bioscience Inc, Cat. TWG003) was titrated through mixing with sodium alginate with different ratios to encapsulate MS5-hDLL4 and iHPC clusters. Then, the suspended AV-ATO clusters were transferred into Ultra-low-attachment six-well plates. For rotating suspended AV-ATO culture, clusters were collected and cultured in 100 mL TCM in a Celstirs culture bottle (WHEATON, Cat. 356876) over a Micro-Stir magnetic rotating device (WHEATON, Cat. W900701-F) at the speed with 60 rpm. Half of the medium was changed every 3 days and clusters were collected at indicated time point for analysis. T cell expansion On T35, CD5+CD7+CD8+CD3+ lymphocytes were isolated by fluorescence-activated cell sorting (BD, FACSAria III). Cells were cultured in XF medium (STEMCELL Technology, Cat. 10981) supplemented with 10% fetal bovine serum (FBS), 1% GlutaMAX, 1% insulin-transferrin-selenium (ITS-G) (ThermoFisher Scientific, Cat. 41400045), 100 μg/mL L-ascorbic acid, 24 μM N-acetyl-L-cysteine, and 2 μM minocycline hydrochloride. The culture medium was further supplemented with 20 ng/mL IL-2 (Novoprotein, Cat. C013-1), 10 ng/mL IL-21 (Novoprotein, Cat. CC45), 10 ng/mL IL-7, 10 μM pan-caspase inhibitor Z-VAD-FMK (MCE, Cat. HY-16658B), and anti-CD3/CD28/CD2 antibodies (STEMCELL Technology, Cat. 10970) for 2 days of stimulation. Cells were seeded at a density of 5×105/mL in Ultra-low-attachment 24-well plates (Corning, Cat. 3473), with complete medium replaced every two days. Flow cytometry Before flow cytometry staining, all cell samples were prepared as single-cell suspensions. Cells in the early hematopoietic differentiation were digested with Accutase (Invitrogen, Cat. 00-4555-56). Suspension cells were collected directly. All Gel-ATO or ATO samples were mechanically dissociated and filtered through 40 μm nylon cell strainer (FALCON, Cat. 352340) to obtain single cells. For the detection of cell surface markers, samples were resuspended with 100 μL of PBS containing 0.5% BSA. Then, 0.2 μL of each indicated antibody was then added to each sample and incubated for 15 min in the dark at room temperature. Subsequently, 1 mL of PBS was added to wash once. The supernatant is removed by centrifugation, and cells were resuspended in 100 μL of PBS for later use. They were then filtered through 40 μm nylon cell strainers (Falcon, Cat. 352340) before analysis. For intracellular marker staining, Fixable Viability Stain 575V (BD Biosciences, Cat. 565694) was added for labeling dead cells for 5 min, and then washed with PBS. Then, the extracellular markers were stained as described above. After washing with PBS, cells were fixed and permeabilized using the BD Cytofix/Cytoperm™ Fixation/Permeabilization kit (BD Biosciences, Cat. 554714) at room temperature in the dark for 20 min. Samples were then washed once with 500 μL 1× BD Perm/Wash buffer (BD Biosciences, Cat. 554723), resuspended in 200 μL 1× BD Perm/Wash buffer, and were incubated with the intracellular marker antibodies at room temperature in the dark for 20 min. Then 500 μL 1× BD Perm/Wash buffer was added to wash twice. Cells were resuspended in 100 μL 1× BD Perm/Wash buffer for later use, and they were then filtered through a 40 μm nylon cell strainer before analysis. Monoclonal antibodies used for surface and intracellular staining of the following molecules were obtained from BioLegend and ThermoFisher: 4-1BB (4B4-1), CCR7 (G043H7), CD144 (BV9), CD235a (H1264), CD25 (BC96), CD27 (O323), CD28 (CD28.2), CD3 (UCHT1), CD31 (WM59), CD34 (581), CD4 (A161A1), CD43 (MEM-59), CD45 (HI30), CD45RA (HI100), CD45RO (UCHL1), CD5 (L17F12), CD56 (HCD56), CD62L (DREG-56), CD7 (CD7-6B7), CD8α (HIT8a), CD8β (SIDI8BEE), DLL4 (MHD4-46), IFNγ (4S.B3), Mouse IgG1, κ (MOPC-21), Mouse IgG2a, κ (MOPC-173), Mouse IgG2b, κ (MPC-11), IL-2 (MQ1-17H12), KDR (7D4-6), NKp44 (P44-8), TCR Vβ 13.1 (H131), TCRαβ (IP26), TNFα (Mab11). T cell cytokine expression assay CD8+ iT cell populations were purified using FACS, and plated at a density of 5×104 cells per well in 96-well U-bottom low-attachment plates (Corning, Cat. 7007) with 200 μL T cell expansion medium. Cell stimulation cocktail (eBioscience, Cat. 00-4970-93) and protein transport inhibitor cocktail (eBioscience, Cat. 00-4980-93) were added to each well and incubated for 6 hours. After washing, cells were collected and stained with Fixable Viability Stain 575V and antibodies against CD8β and CD3, then were fixed and permeabilized to stain with intracellular antibodies against IFNγ, IL-2 and TNFα as described in the flow cytometry protocol above. T cell CFSE proliferation assay Carboxyfluorescein succinimidyl ester (CFSE) (BioLegend, Cat. 423801) was diluted to a 5 μM working solution with PBS. iT cells were labeled with CFSE at 37°C for 20 minutes in the dark. Staining was quenched by adding 5 volumes of culture medium supplemented with 10% fetal bovine serum. Cells were resuspended in T cell expansion medium with or without stimulating anti-CD3/CD28/CD2 antibodies for 2 days. The stimulating antibodies were washed out and iT cells were further cultured in T cell expansion medium for 4 days, followed by staining with antibodies against CD25 and 4-1BB for flow cytometry analysis. In vitro T cell cytotoxic assay Effector K562-NY-ESO-1 cells were generated by transducing Lentivirus containing HLA-A∗02:01/NY-ESO-1 single chain trimer and firefly luciferase, followed with GFP (K562-NY-ESO-1-Luci). K562-MART1 cells were similarly generated with HLA-A∗02:01/MART1 single chain trimer. K562-MART1 GFP+ cells were used as target cells to assess iT cell cytotoxicity. Expanded 1G4-iT cells were sorted with CD8+CD4-CD3+NKP44-1G4-RFP1+ and resuspended in expansion medium. Five thousand effector cells were seeded in a 96-well plate (Biosharp, Cat. BS-MP-96W-CL), and 1G4-iT cells were added at the indicated effector-to-target (E/T) ratios for co-incubation. D-luciferin potassium salt (BioMol, Cat. BM8056) was added to each well at a final concentration of 0.14 mg/mL for every 4 h. The cytolytic efficiency of T cells were quantified by a standard luciferase-based bioluminescence assay using a multimode plate reader (PerkinElmer). The percentage of lysed cells was calculated using the following equation: % lysis = 100 × (spontaneous death relative light units [RLUs] – test RLUs)/(spontaneous death RLUs). In vivo T cell cytotoxic assay The tumor burden mice model was established with 6 to 8-week-old NOD.Cg-Prkdcscid Il2rgtm1vst/Vst (NPG, stock number: VS-AM-001) male mice purchased from Beijing Vitalstar Biotechnology, which were subcutaneously implanted with 5×104 K562-NY-ESO-1-Luci cells. Mice were imaged according to the schedule in Figure 2E until day 31. For tumor bioluminescence, mice were intraperitoneally injected with 150 mg/kg D-luciferin potassium salt, anesthetized with 2,2,2-tribromoethanol (TBE; Sigma‒Aldrich, Cat. T48402), and imaged 5 min later. Mice were sacrificed based on disease burden criteria. For testing in vivo cytotoxicity of iT cells differentiated from hCiPSCs, we induced hCiPSC-IG4-RFP cells to differentiate into 1G4-TCR iT cells. The sorted CD7+CD5+CD4+CD8+CD3+ iT cells were activated by anti-CD3/CD28/CD2 antibodies in T cell expansion medium for two days. Then, the stimulated iT cells were expanded without the antibodies and the expansion medium was changed every other day. After 14 days, 1G4-RFP+-iT cells were collected and transplanted into K562-NY-ESO-1 tumor burden mice. This time point was depicted as day 0. At day 0, each mouse was intraperitoneally injected with 2.5×106 1G4-iT cells suspended in 200 μL TCM containing 100 ng/mL IL-2, 100 ng/mL IL-7 and 100 ng/mL IL-15. Injection of TCM containing IL-2, IL-7 and IL-15 in control mice was also performed. After 4 days, 2.5×106 1G4-iT cells were resuspended in TCM containing cytokines and intraperitoneally injected into the mice. Bulk RNA-seq data processing Samples for bulk RNA-seq analysis include: hCiPSC-3-derived HE cells at D7 and D8 and they were MACS sorted with CD34 microbeads; CB-CD34+ HPSC-derived iT cells generated in Gel-ATO and these iT cells were sorted with CD7+CD5+, CD4+CD8+; hCiPSC-3-derived iT cells from different iT-differentiation days as indicated in Figure 3A, and were respectively sorted with FACS according to the markers of CD7, CD5, CD4, CD8; CB-derived CD8 T cells sorted with MACS as in vivo T control. hCiPSC-3-derived hematopoietic cells generated at Day21 in R-S-ATO and AV-ATO were respectively collected for RNA-seq, and the data were presented in Figure S8A-8C. Total RNA was isolated using a Direct-zol™ RNA MiniPrep kit (Zymo Research, Cat. R2051). RNA-seq libraries were constructed using the NEBNext® Ultra™ II RNA Library Prep Kit for Illumina® (NEB, Cat. E7775L). The fragmented paired-end libraries were sequenced using an Illumina NovaSeq 6000. All sequencing was performed at Novogene. Data could be downloaded from https://ngdc.cncb.ac.cn/gsa-human/s/YA6K0CZw. All raw reads were first trimmed by Trimmomatic (version 0.39) software to remove adapters and low-quality reads. Then, cleaned data were mapped to the human reference genome (hg38) using STAR (version 2.7.10b) with default parameters. The count matrix of gene expression in each sample was generated by featureCounts (version 2.0.6). The gene expression level was normalized by Fragments Per Kilobase Million (FPKM). The Pheatmap R package was finally used to visualize the dynamic gene expression of samples. ComBat function in sva (version 3.50.0) R package was used to remove the batch effect when integrating public data (GSE116015; GSE195667). Single-cell RNA-seq data processing Gel-ATO-derived cells were sorted with CD7 using flow cytometry sorter for single cell RNA sequencing. The single-cell 50 Library and Gel Bead Kit (10x Genomics, Cat. 1000006) and Chromium Single Cell A Chip Kit (10x Genomics, Cat. 120236) were used to load cell suspensions (300–600 living cells/mL according to Count Star) onto the Chromium single-cell controller (10x Genomics) for generation of single-cell gel beads in emulsion (GEMs), in accordance with the kit manufacturer’s protocols. Briefly, single cells were suspended in PBS containing 0.04% BSA. Approximately 6,000 cells were added to each channel, with an estimated target cell recovery of 3,000 cells. Captured cells were lysed, and the released RNA was barcoded through reverse transcription in individual GEMs. Reverse transcription was performed on a S1000TM Touch Thermal Cycler (BioRad) at 53°C for 45 min, followed by 85°C for 5 min; the resulting cDNA was stored at 4°C. The cDNA was amplified and subjected to quality assessment using an Agilent 4200 instrument; these assays were performed by CapitalBio Technology (Beijing). In accordance with the manufacturer’s instructions, single-cell RNA sequencing libraries were constructed using the Single Cell 50 Library and Gel Bead Kit. The libraries were sequenced on an Illumina NovaSeq6000 sequencer with a sequencing depth of R100,000 reads per cell using a 150 bp paired-end (PE150) approach. This RNA-seq sequencing was performed by CapitalBio Technology (Beijing). Data could be downloaded from https://ngdc.cncb.ac.cn/gsa-human/s/YA6K0CZw. The single-cell RNA-seq data were collected and mapped to human reference genome (hg38) using Cell Ranger 5.0.1 for all our samples. We performed preprocessing using Seurat 4.3.0.1. In detail, quality control was firstly performed to remove the cells with mitochondrial UMI counts >25%. The normalization was performed using function NormalizeData with default parameters. We performed the standard Seurat clustering pipeline using the following functions: FindVariableFeatures with 2000 genes, ScaleData, RunPCA, FindNeighbors with first 20 PCs, and FindClusters with resolution 0.2. The UMAP dimension reduction was performed with first 20 corrected PCs using function RunUMAP. The DEGs were analyzed using function FindAllMarkers based on normalized gene expression. Data generated by Jong-Eun Park et al. (Park et al., 2020) were downloaded from https://developmental.cellatlas.io/thymus-development and preprocessed as the same procedures as described above. It was used as an in vivo reference. Ontology annotation DEGs of different samples were found by the DESeq2 (version 1.42.1) R package following the standard DESeq2 workflow. And then, they were used for gene set enrichment analysis using clusterProfiler (version 4.10.0) R package. GO terms with p-value <0.05 were defined as significantly enriched. TCR repertoire analysis Total RNA was isolated and were analyzed by NGS for TCRα and β using the ImmuHub® TCR profiling system (ImmuQuad Biotech, Hangzhou, China). Briefly, a 5′ RACE unbiased amplification protocol was used. Primers corresponding to the constant (C) regions of each of the TCRα and β were designed to facilitate PCR amplification of cDNA sequences in a less biased manner. This protocol uses unique molecular identifiers (UMIs) introduced in the course of cDNA synthesis to control bottlenecks and to eliminate PCR and sequencing errors. Sequencing was performed on an Illumina NovaSeq® system with PE150 mode. The UMI attached to each raw sequence reads were applied to correct PCR and sequencing errors correction and PCR duplicates removing. Map V, D, J and C segments with IMGT® and then extract CDR3 regions and assemble clonotype for all clones. The resulting nucleotide and amino acid sequences of CDR3 of TCRα and β were determined and those with out-of-frame and stop codon sequences were removed from the identified TCRα and β repertoire. Amounts of each TCRα and β clonotype were defined by adding numbers of clones sharing the same nucleotide sequence of CDR3, and clonotype data were processed using custom scripts in R for further analysis. For the upstream processing of TCR repertoire sequencing data, MiXCR was employed. The core procedures included: (1) alignment of trimmed reads to the reference TCR germline gene database to identify variable (V), diversity (D), joining (J), and constant (C) gene segments; (2) assembly of full-length TCR CDR3 sequences by stitching aligned V, D, J, and C segments; and (3) quantification of TCR clonotypes to generate a standardized clonotype table with read counts and frequency information. For downstream bioinformatic analysis of the processed TCR repertoire data, immunarch was utilized. Key analyses included: (1) V/J gene segment usage profiling to detect biases in TCR gene rearrangement; (2) CDR3 sequence motif analysis to identify shared or signature amino acid patterns. Quantification and statistical analysis For analyses of the effects of MS5-hDLL4 to hematopoietic cell ratios on differentiation efficiency and the in vivo cytotoxic efficacy of iT cells against tumor cells, group comparisons were performed using two-way ANOVA. For other experiments, unpaired Student’s t-tests were used. All statistical analyses were two-sided, with p < 0.05 considered statistically significant. Statistical details and results are described in the figure legends. Published: September 3, 2026 Footnotes Supplemental information can be found online at https://doi.org/10.1016/j.stemcr.2026.103076. Contributor Information Hongkui Deng, 邓宏魁, Email: hongkui_deng@pku.edu.cn. Chengyan Wang, 王承艳, Email: chengyanw@pku.edu.cn. Supplemental information Document S1. Figures S1–S8 and Tables S1 Document S2. Article plus supplemental information References Associated Data Supplementary Materials Document S1. Figures S1–S8 and Tables S1 Document S2. Article plus supplemental information Data Availability Statement •The raw sequence data reported in this paper have been deposited in the Genome Sequence Archive (GSA) in National Genomics Data Center, China National Center for Bioinformation/Beijing Institute of Genomics, Chinese Academy of Sciences (GSA-Human: HRA013688) that are publicly accessible at https://ngdc.cncb.ac.cn/gsa-human/s/YA6K0CZw. This paper analyzes existing, publicly available data. These accession numbers for the datasets are listed in the key resources table. The raw sequence data reported in this paper have been deposited in the Genome Sequence Archive (GSA) in National Genomics Data Center, China National Center for Bioinformation/Beijing Institute of Genomics, Chinese Academy of Sciences (GSA-Human: HRA013688) that are publicly accessible at https://ngdc.cncb.ac.cn/gsa-human/s/YA6K0CZw. This paper analyzes existing, publicly available data. These accession numbers for the datasets are listed in the key resources table. •This paper does not report original code. This paper does not report original code. •Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request. Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.