To characterize the ProTcell product after a 7-day-culture on the DLL4 platform, we performed scRNAseq as well as CyTOF of the progenitors obtained from three different CB CD34⁺ and three mPB CD34⁺ starting materials ( Figure 1A ). Since both cell products are functionally similar [1], the six samples were integrated into a single scRNAseq dataset for analysis ( Supplementary Figure 1A ). Following filtering and normalization, the dataset contained 40048 cells divided into 14 clusters ( Supplementary Figure 1B ). We first analyzed the expression of CD7, an early marker of differentiating T cell progenitors, which was expressed in about 80% of the dataset’s cells (32661 cells, Supplementary Figure 1C ).

Flow cytometric analysis confirmed that the CD7⁺ cell subset dominates the proTcell product, comprising up to 96.97% ± 0.00% of the CB-derived and 83.78% ± 0.02% for mPB-derived product after 7 days of manufacturing ( Figure 1B ), while initial CD34⁺ cells were lacking CD7 expression ( Supplementary Figure 1D ). These findings led us to identify distinct subpopulations and to characterize their potential in the final lymphoid progenitor’s product by focusing analysis on cells expressing the CD7 marker. In the scRNAseq dataset, the CD7⁺ lymphoid progenitors were clustered in seven cell subsets ( Figure 1C ). GATA3, transcription factor 7 (TCF7), and B-cell lymphoma 11B (BCL11B), the main transcriptional drivers of the T cell lineage program, were broadly expressed by these CD7⁺ cells demonstrating their commitment to the T cell lineage ( Figures 1D, E ). Additionally, interleukin 7 receptor (IL7R) and CD3 subunits, characteristic of early T cell development, were broadly expressed at the transcriptomic level ( Figures 1D, F ). It should be noted that despite the expression of CD3 subunits at transcript level, the CD3 protein was not detectable at the cell surface of the ProTcell product ( Supplementary Figure 1E ), suggesting the immature T cell characteristic of these cells. Consistent with this observation, recombination-activating gene (RAG) 1 and RAG2 transcripts were not expressed ( Supplementary Figure 1F ), supporting the absence of RAG1/2 complex-mediated T cell receptor (TCR) B gene rearrangements as demonstrated in our previous study [2].

Differential gene expression profiles of selected markers within different clusters ( Figure 1G ; Supplementary Table 1 ), as well as the top 10 genes defining each cluster ( Supplementary Figure 2 ), are represented. The cells in cluster 0 correspond to earlier lymphoid commitment stage, based on their low expression of the lymphoid-associated transcription factors TCF7, BCL11B, and GATA3. In addition, these cells in cluster 0 display an enrichment of genes associated with cell cycle entry: several histone genes, cyclin-dependent kinase 1 (CDK1), proliferating cell nuclear antigen (PCNA), aurora kinase B (AURKB), and the proliferation marker Ki-67 (MKI67) (18). Cells belonging to clusters 1, 2 and 3 are instead characterized by increased expression of GATA3, BCL11B and TCF7, along with DLL4-regulated genes such as neurogenic locus notch homolog protein 1 (NOTCH1) and NOTCH regulated ankyrin repeat protein (NRARP) ( Figure 1G ; Supplementary Table 1 ). Notably, cells in cluster 3 exhibit higher expression of T cell commitment markers, including CD5, CD1E, and pre-T cell antigen receptor alpha (PTCRA), which encodes the alpha chain of the pre-TCR, suggesting they represent cells in a more advanced stage of lymphoid development ( Supplementary Figure 3A ). Supporting this hypothesis, cells in cluster 3 also demonstrate higher expression of several homing molecules—notably C-X-C chemokine receptor type 4 (CXRC4), selectin-L (SELL), and integrin alpha 4 (ITGA4)—than is expressed in other clusters ( Supplementary Figures 3B, C ).

With respect to cluster 5, differential expression analysis comparing this cluster to all others reveals that it is defined by several markers associated with proliferative activity ( Supplementary Table 1 ). Specifically, this cluster shows elevated expression of key cell cycle–associated genes, including: cytoskeleton organization: cytoplasmic FMR1 interacting protein 2 (CYFIP2), rho GTPase activating protein 45 (ARHGAP45), WAS/WASL interacting protein family member 1 (WIPF1), NCK Associated Protein 1 Like (NCKAP1L) and gene expression remodeling (SWI/SNF related BAF chromatin remodeling complex subunit ATPase 2 [SMARCA2]), CCR4-NOT transcription complex subunit 1 (CNOT1) and the E3 ubiquitin ligases thyroid hormone receptor interactor 12 (TRIP12) and tripartite motif containing 56 (TRIM56) all of which are characteristic of cells actively cycling. We therefore interpret this cluster as comprising predominantly cycling cells, potentially in transitional or mixed states, rather than reflecting a lineage-restricted identity.

We then used SingleR to compare these ex vivo-generated ProTcells with existing datasets of in vivo human thymocytes (19, 20) ( Figure 1H ) and their respective transcriptional profiles. The Cordes dataset comprised early CD4-CD8- double negative (DN) thymocytes from six healthy donors. Lymphoid progenitor cells exhibit a highly similar gene expression pattern to DN1 (CD34⁺ CD38⁻ CD1a⁻) and DN2 (CD34^+/- CD38⁺ CD1a⁻) cell subsets of human thymocytes ( Figure 1H ) and share, to a lesser degree, the pattern of gene expression found in more advanced stages of T cell development. Additionally, ProTcells showed a higher similarity score with cells in the “committed” cluster than with those from the Lavaert dataset generated from CD34⁺ human thymocytes, while showing minimal similarity to more differentiated cells.

Taken together, these results show that the ex vivo generation of lymphoid progenitors from human CD34⁺ cells faithfully recapitulates the early stages of T lymphoid development.

Given the presence of distinct clusters among ProTcells, we next further explored the molecular heterogeneity across these subsets. Despite expressing T cell lineage markers such as CD7, TCF7 and CD3D ( Figures 1D–F ), cells in clusters 4 and 6 were distinguished by higher expression of killer cell lectin-like receptor B1 (KLRB1, encoding for CD161), DNA-binding inhibitor 2 (ID2), nuclear factor interleukin 3 regulated (NFIL3) and the receptor tyrosine kinase (KIT) ( Figures 2A, C ) all of which are associated with ILC populations (21–23). Similarly, interleukin 1 receptor type 1 (IL1R1) and integrin beta 7 (ITGB7), prostaglandin D2 Receptor 2 (PTGDR2 or CRTH2) and CCR6, other common markers of ILC and ILC progenitors, exhibited higher expression in cells in these two clusters ( Supplementary Figure 3D, E ). Of note, GATA3 expression was higher in cluster 6 than in cluster 4 ( Figure 1D ). Conversely, cluster 4 showed stronger expression of granzyme (GZM) A, GZMB, perforin 1 (PRF1), and neural cell adhesion molecule 1 (NCAM1, encoding for CD56) ( Figures 2B, D ) than cluster 6. Since GATA3 is predominantly involved in the early stages of ILC differentiation (24, 25), while granzymes, perforin, and CD56 are associated with later stages of NK cell development, this finding suggests that cluster 4 may represent a more advanced stage of development than cluster 6.

To validate these findings, we performed CyTOF analysis, which confirmed that CD161 is expressed at the protein level in 53.4 ± 10.2% of CD7⁺ cells. This CD161⁺ cell subset contains cells that also express KIT ( Figure 2E ). This subset was, most significantly, distinct from cells exhibiting CD5 expression, which indicates commitment to the T cell lineage. In accordance with this observation, lymphoid enhancer binding factor 1 (LEF1), a T cell specific transcription factor, was shown to be expressed separately from KLRB1-positive subsets in the scRNAseq data ( Supplementary Figure 3D, F ).

This data strongly indicates that the ProTcell product is heterogenous, comprising subpopulations of lymphoid progenitors at different stages of development. These subpopulations include T cell progenitors as well as cells with a molecular profile characteristic of ILC/NK cell progenitors.

We next explored the lineage relationship between T cell and ILC/NK cell progenitors to determine whether they arise from a shared precursor population or if the DLL4 platform facilitates the expansion of two distinct lineage-biased precursors. To address this, we performed pseudotime analysis using Slingshot to delineate cell trajectories within the lymphoid cell clusters. The pseudotime analyses of CD7⁺ clusters identified three main trajectories: L1 trajectory describing the progressive transition from developing T cell subsets; L2 trajectory oriented toward clusters exhibiting an ILC/NK cell gene expression signature; and L3 trajectory oriented toward cluster 5. Of note the cluster 5 contained the cycling cells ( Figure 2F ). Differential gene expression analysis of the most significant contributors to these trajectories revealed distinct markers associated with each pathway. The L2 trajectory, leading to ILC-oriented cells (clusters 4 and 6), was enriched for markers such as GZMA, PRF1, and ITGB7 ( Supplementary Figure 4A ). In contrast, markers associated with the L1 trajectory contain components of NOTCH signaling pathway, a positive regulator of T cell development, such as NRARP and small nucleolar RNA host gene 3 (SNHG3) (26) ( Supplementary Figure 4A ). The L3 trajectory aligns with the cluster of cycling cells suggesting that this trajectory represents a lineage influenced by cell cycle dynamics and the cellular states captured along this trajectory are linked to progression through the cell cycle.

Together, these analyses suggest that the DLL4 platform allows for the expansion of a progenitor population capable of undergoing two main differentiation trajectories, one toward T cells and the other toward ILC/NK cells. Then we next aimed to functionally evaluate the lineage potential and plasticity of both T cell–oriented and NK cell–oriented subsets toward T or NK fates, in vitro and in vivo.

Our analysis of scRNAseq and CyTOF revealed a strong association between elevated CD161 expression and transcriptional commitment to the ILC/NK fate. These results are consistent with previous reports of CD161 expression during early stages of NK cell development (27–29). Furthermore, we have demonstrated the feasibility of producing a highly pure NK cell product from ProTcells using an optimized culture protocol (30). To further investigate whether NK cell potential in ex vivo-generated lymphoid progenitors segregated to the CD161⁺ ProTcell fraction, we used fluorescence activated cell sorting (FACS) to sort the CD7⁺ cells from the ProTcell product based on their negative or positive CD161 expression. The resulting populations were tested under either NK or T cell culture conditions ( Figure 3A ). After a 14-day-culture in NK-favoring conditions, both CD161⁺ and CD161^- subsets exhibited comparable capacities to differentiate into NK cells in vitro, as demonstrated by the presence of CD3^- CD56⁺ cells in similar frequencies and numbers ( Figures 3B, C ; Supplementary Figure 5A ). In parallel, the same sorted fractions were cultured under T cell differentiation conditions using an OP9-DLL1 (OP9 cells expressing human DLL1) co-culture system for 4 weeks. Both fractions generated T cells, characterized by CD4⁺ CD8⁺ double positive (DP), CD3⁺ TCRαβ⁺, and CD3⁺ TCRγδ⁺ cell populations ( Figure 3D ; Supplementary Figures 5B–D ). Notably, the CD161^- subset yielded a higher number of CD3⁺ cells ( Figure 3E ).

These results demonstrated that while CD161⁺ and CD161^- cell subsets display distinct gene expression pattern consistent with NK or T cell fate, respectively, both cell fractions retain lymphoid lineage plasticity. As such, both populations can differentiate in vitro into either T cells or NK cells, depending on the environmental cues directing their differentiation.

We then tested if cells from both CD161 subsets can colonize thymus in vivo upon adoptive transfer. We initially assessed the expression of surface markers involved in thymus homing and colonization using CyTOF. Both CD161⁺ and CD161^- subsets expressed key homing markers, including chemokine receptors (C-C chemokine receptor type [CCR] 9, CCR7, CXCR4), adhesion molecules (CD44, SELL, Selectin P Ligand [SELPLG]), and integrins (ITGA4, ITGB7) ( Figure 3F ), suggesting that both CD161⁺ and CD161^- cells have thymic homing ability. We next injected the same FACS-sorted CD7⁺ CD161^- or CD7⁺ CD161⁺ cell subsets, as used in T and NK cell differentiation assay, intra-hepatically into non-irradiated NOD.Cg-Prkdcscid Il2rgtm1Wjl/SzJ (NSG) neonate mice ( Figures 3G, H ). Six weeks post-transplantation, both fractions engrafted in the thymus, as evidenced by the presence of human CD45⁺ cells in recipient mice ( Figure 3G ). Both fractions also contributed to the development of TCRαβ⁺ CD3⁺ T cells ( Figure 3H ; Supplementary Figure 5E ).

Overall, these results demonstrate that both CD7⁺ CD161^- and CD7⁺ CD161⁺ subsets retained T and NK lineage plasticity with the ability to differentiate into NK or T cells in vitro depending on environmental signals. Their T cell fate has been further confirmed in vivo, where both populations exhibit thymic homing capacity and contribute to thymopoiesis.

The dual potential of ProTcells raises important questions about the molecular mechanisms governing lineage specification in early human lymphoid progenitors. The specification toward either the T or ILC lineage is tightly regulated by opposing transcription factors (31): commitment to one lineage requires the repression of the alternative gene program. In this context, we focused on BCL11B, a transcription factor traditionally viewed as a master regulator of T cell development and an inhibitor of NK/ILC differentiation (32–34). However, the role of BCL11B in regulating NK/ILC differentiation is controversial as other studies have shown that it is essential for the development and function of ILCs in mice (35, 36). Our data indicates that BCL11B is widely expressed in CD7⁺ lymphoid progenitors ( Figures 1D, E ), including in CD161⁺ ILC/NK-fated cells, albeit at lower level ( Supplementary Figure 4B ). This finding suggests that BCL11B expression does not entirely abrogate ILC/NK potential in human CD7⁺ lymphoid progenitors.

To further investigate the specific role of BCL11B in human lymphoid specification, we used a knock-in BCL11B-Enhanced green fluorescent proteins (EGFP) reporter human pluripotent stem cell (hPSC)-line with a bacterial artificial chromosome (BAC)-based monoallelic integration of the EGFP transgene within the endogenous exon 1 of BCL11B gene ( Supplementary Figure 6A ). Differentiating this cell line into T cell progenitors allowed us to assess the T, NK and myeloid potential of either BCL11B-EGFP⁺ or BCL11B-EGFP^neg hPSC-derived T cell progenitors ( Figure 4A ). We isolated CD34⁺ CD144⁺ CD43^- CD73^- CXCR4^- hematopoietic precursors from embryoid body-based differentiation cultures (37) and we cultured them on OP9-DLL4 for 14 days in T cell specification medium, enabling the emergence of CD7⁺ progenitors ( Figures 4A, B ). These CD45⁺ CD56^- CD7⁺ hematopoietic progenitors were FACS-sorted into BCL11B-EGFP⁺ and BCL11B-EGFP^neg fractions and further cultured on either OP9-DLL4 or OP9 to assess their T/NK lymphoid and myeloid potential, respectively.

In OP9-DLL4-based NK cell potential assay, both BCL11B-EGFP⁺ and BCL11B-EGFP^neg fractions generated CD45⁺ CD56⁺ NK lymphoid cells ( Figures 4C, D ). In contrast, only BCL11B-EGFP⁻ progenitors could generate CD45⁺ CD11b⁺ myeloid cells ( Figures 4E, F ), which were EGFP negative at the time of analysis ( Supplementary Figures 6B, C ). Thus, while the activation of BCL11B regulatory elements in CD7⁺ progenitors does not preclude NK cell differentiation, it marks the stage where myeloid potential is repressed.

When T cell potential was tested on OP9-DLL4 co-cultures, both BCL11B-EGFP⁺ and BCL11B-EGFP^neg progenitors upregulated CD5, a marker associated with T cell developmental progression ( Figures 4G, H ). BCL11B-EGFP expression was also dynamically regulated in T cell cultures, with both sorted progenitor subsets yielding a mixture of EGFP^neg and EGFP⁺ cells ( Supplementary Figure 6D ). Significantly, CD4⁺ CD8⁺ DP cells were rarely generated using the BCL11B-EGFP reporter line, likely due to the BCL11B heterozygosity caused by the monoallelic reporter integration that mimics BCL11B haploinsufficiency (38–40) (not shown).

As an overall conclusion, despite the expression of BCL11B, the ProTcell platform stands at the intersection of the NK/ILC and T cell lymphoid lineages. In summary, our experimental model with human hematopoietic progenitors shows that the BCL11B regulatory element is activated early in lymphoid commitment, specifically in cells harboring both T and NK potential that lost the ability to generate myeloid cells.

Umbilical cord blood (CB) samples were collected from donors at AP-HP, Saint Louis Hospital (Unité de Thérapie Cellulaire, CRB-Banque de Sang de Cordon, Paris, France) in accordance with ethically approved protocols and following the provision of informed consent. CD34⁺ HSPCs were magnetically enriched (purity > 90%) from CB as described previously (81). For specific experiments, enriched CD34⁺ cells were purchased from Lymphobank (Besançon, France) and Tebu-bio (Le Perray-en-Yvelines, France).

For ProTcells’ culture, human CD34⁺ HSPCs isolated either from CB or mPB were cultured for 7 days at a cell concentration of 5x10⁴ and 1x10⁵ cells/mL respectively with 5µg/mL human DLL4 fused to IgG Fc fragment (DLL4-Fc) protein and 2 µg/mL RetroNectin^® (Takara Bio Europe, Saint-Germain-en-Laye, France)-coated wells of culture plates/culture flasks containing α minimal essential medium (MEM) (Gibco, Life Technologies, Courtaboeuf, France) supplemented with 20% defined fetal bovine serum (FBS) (HyClone, Cytiva Europe GmbH, Velizy-Villacoublay, France) and human cytokines (100ng/mL IL7, 100ng/mL FMS-like tyrosine kinase 3 ligand [Flt3L], 100ng/mL stem cell factor [SCF] and 100ng/mL thrombopoietin [TPO]; all from PeproTech France, Neuilly-sur-Seine, France. Also used was 100ng/mL TNFα (R&D Systems, Minneapolis, MN). Medium was half-changed after 3 days of culture (13).

The in vitro NK cell potential of FACS-sorted CD7⁺ CD161^- and CD7⁺ CD161⁺ progenitors was assessed through an IL15-based feeder-free NK cell differentiation assay. The CD7⁺ CD161^- and CD7⁺ CD161⁺ progenitors sorted from CB ProTcells were cultivated in RPMI 1640 Medium, GlutaMAX™ Supplement (Gibco, Life Technologies, Courtaboeuf, France) supplemented with 10% defined FBS (HyClone, Cytiva Europe GmbH, Velizy-Villacoublay, France) and human cytokines (20ng/mL IL7, 50ng/mL Flt3L, 50ng/mL SCF, 20ng/mL IL15, 500 IU/mL IL2; all from PeproTech Inc., Rocky Hill, NJ). Medium was refreshed every 2–3 days. After 1 and 2 weeks of culture, cells were harvested and analyzed (using flow cytometry) for the presence of CD56⁺ NK cells.

The in vitro T lymphoid potential of FACS-sorted CD7⁺ CD161^- and CD7⁺ CD161⁺ progenitors was assessed by using an OP9-DLL1 co-culture system, as described previously (81). CD7⁺ CD161^- and CD7⁺ CD161⁺ progenitors sorted from CB ProTcells were co-cultured on OP9-DLL1 stromal cells for 4 weeks in αMEM (Gibco, Life Technologies, Courtaboeuf, France) supplemented with 20% defined FBS (HyClone, Cytiva Europe GmbH, Velizy-Villacoublay, France), and human cytokines (2ng/ml IL7, 5ng/ml Flt3L, 10ng/mL SCF and 10ng/mL TPO, all from PeproTech France, Neuilly-sur-Seine, France). Every week, a part of the co-cultured cells were harvested and analyzed (using flow cytometry) for the presence of CD4⁺ CD8⁺ and TCRαβ- or TCRγδ-expressing CD3⁺ T cells, and the other part of the co-cultured cells were reseeded on fresh OP9-DLL1 cells.

The BAC plasmid clone CH17-365C4 containing the BCL11B gene flanked by approximately 90kb upstream and 36kb downstream of the start codon was obtained from CHORI BACPAC (Children ‘s Hospital Oakland Research Institute, Oakland, CA, USA). To insert an EGFP reporter cassette into the BCL11B locus of the BAC plasmid, an EGFPpA-LoxP-PGK-Gb2-Kan/Neo-pA-LoxP vector was targeted to the start codon of the BCL11B gene using recombination-mediated genetic engineering. In brief, DH10b bacteria, containing the CH17-365C4BAC plasmid were grown under chloramphenicol antibiotic selection (Sigma-Aldrich). pSC101BAD Gba A[tet] (Genebridges, Heidelberg, Germany) was transfected into these BAC-containing bacteria and grown at 32°C under tetracyclin antibiotic resistance (Sigma-Aldrich). Bacterial cells containing both plasmids were then transfected with PCR amplified targeting construct (EGFPpA-LoxP-PGK-Gb2-Kan/Neo-pA-LoxP), using primers with overhanging 50bp homology arms, and grown overnight at 37°C under chloramphenicol and kanamycin resistance (Sigma-Aldrich). BAC DNA from selected bacteria was purified and verified for correct integration of the EGFP cassette using restriction analyis. The BCL11B-EGFP reporter construct was subsequently functionally validated by transfection in T cell leukemia Jurkat cells, which express levels of BCL11B within a biologically relevant range, similar to BCL11B expression during human embryonic stem cell line (hESC)-derived T cell differentiation. The erythromyeloid leukemia line K562 cell was used as negative control. After transfection and neomycin selection, the selection cassette was removed using Cre recombinase and expressed EGFP as expected.

A total of 2x10⁶ single cell adapted hESC were nucleofected with 5μg purified BAC reporter using Lonza Amaxa human stem cell nucleofector kit 2 (Lonza) with program F16. Cells were replated at high density (1x10⁵ cells/cm²) in single cell conditions on drug resistant DR4 mouse embryonic fibroblasts (MEF). A total of 10μM of the Rock inhibitor Y-27632 (Selleckchem) was added. G418 (50μg/ml) (Gibco) was added from day three onwards and retained for three weeks. Single colonies were picked, expanded and screened for transgene integration. Colonies showing transgene integration were transfected using pCAGGS-NLS-Cre-PGK-Puro and kept under puromycin (Sigma-Aldrich) selective pressure (300ng/ml) for three days to remove the neomycin selection cassette. Single colonies were expanded and screened by PCR. Genomic DNA was isolated using Genelute mammalian genomic DNA miniprep kit (Sigma-Aldrich) according to the manufacturer’s instructions. Colonies containing successfully floxed cells were clonally expanded using single cell deposition on the FACS Aria. (clone 3.8 and 3.10). To determine the integration site of the BAC, targeted locus amplification (TLA) sequencing was performed which revealed homologous integration of the EGFP reporter gene in the endogenous BCL11B locus in one allele (Cergentis). No random integration at other genomic sites was detected.

BCL11B-EGFP human embryonic stem cell line (hESC) was cultured on irradiated MEF, in Dulbecco’s Modified Eagle medium (DMEM)/F12 medium (Corning) supplemented with 20% of KnockOut™ Serum Replacement (Thermo Fisher Scientific), 1% penicillin-streptomycin (Lonza), 1% glutamine (Lonza), 0.1% 2-Mercaptoethanol (Sigma-Aldrich) and 0.7% of MEM Non-Essential Amino Acids Solution (Thermo Fisher Scientific). Basic fibroblast growth factor (bFGF) at 20ng/mL (Miltenyi Biotech) was added to the hESC medium shortly before use.

hESCs were expanded for two passages on MEF and one passage on Matrigel (Corning), for feeder-depletion. After 24–48 hours on Matrigel, cells were dissociated by TripLE treatment (Thermo Fisher Scientific) plus scraping and collected in “wash medium” (DMEM/Hams F-12 [Corining] 50/50 mix, 5% KnockOut™ Serum Replacement [Thermo Fisher Scientific] and 25mM HCl) supplemented with 10µg/mL DNAse I (Calbiochem, 260913). At day 0, cells were resuspended as embryoid bodies (EB) at a maximum density of 250000 cells/mL and plated into non-adherent 6-wells plates coated with 5% polyheme solution (Sigma). The differentiation medium was supplemented as previously described (82). The base medium for the first three days of mesoderm induction was Serum-Free Differentiation (SFD) medium (IMDM/Ham’s F-12 [Corning] 75/25 mix with 0.05% w/v bovine serum albumin (BSA) [Sigma-Aldrich]). On the first day of differentiation, SFD medium was supplemented with 2mM l-glutamine (Lonza), 1mM ascorbic acid (Sigma-Aldrich), 400µM 1-thioglycerol solution (MTG, Sigma-Aldrich), 150µg/mL transferrin (R&D System) and 10ng/mL bone morphogenetic protein 4 (BMP4, R&D System). Twenty-four hours later, 5ng/mL bFGF (R&D System) was added. At the second day of differentiation, 3µM CHIR99021 (Cayman Chemical Company) was added. On the fourth day, EBs were changed to StemPro-34 medium (Thermo Fisher Scientific) supplemented with penicillin–streptomycin, l-glutamine, ascorbic acid, 1-thioglycerol and transferrin, as above, with additional 5ng/mL bFGF and 15ng/mL vascular endothelial growth factor (VEGF, R&D). On day 6, 10ng/mL IL6, 25ng/mL insulin-like growth factor 1 (IGF1), 5ng/mL IL11, 50ng/mL SCF (all from Miltenyi Biotech) and 2U/mL erythropoietin (EPO) (Peprotech) were added.

For differentiating hematopoietic precursors into BCL11B-expressing hematopoietic progenitors, CD34⁺ CD144⁺ CD43^- CD73^- CXCR4^- cells were FACS-sorted from day 9 EB-culture and plated on confluent OP9-DLL4-coated 24-multiwell plates. Differentiating cells were cultured in αMEM (Thermo Fisher Scientific) with 2.2g/L sodium bicarbonate (Corning), supplemented with 20% FBS (HyClone), 1% penicillin - streptomycin (Lonza), 1% glutamine (Lonza) and 400µM MTG (Sigma-Aldrich). Base medium was enriched with 50ng/mL SCF (only for the first week), 5ng/mL Flt3L and 5ng/mL IL7 (all from Miltenyi Biotech) to engage T cell lineage specification. At day 14, both BCL11B-EGFP positive and negative fractions were FACS-sorted from CD45⁺ CD56^- CD7⁺ hematopoietic progenitors and further cultured on either OP9 or OP9-DLL4 to assess their myeloid and T/NK lymphoid potential. Myeloid differentiation was induced after 5–6 days on OP9 stroma, by using the above-mentioned αMEM base medium supplemented with 50ng/mL SCF, 50ng/mL macrophage colony stimulating factor (M-CSF), 10ng/mL IL3, 10ng/mL Flt3L and 30ng/mL granulocyte-macrophage colony-stimulating factor (GM-CSF) (all from Miltenyi Biotech). NK-specific differentiation was performed by leaving hematopoietic progenitors for an additional 10–14 days on OP9-DLL4 with the above described “T cell medium” supplemented with 10ng/mL IL15 (Miltenyi Biotech). For T cell specification, a medium with the same composition as above was used, and the culture was split every 4–5 days and replate onto new stroma.

For surface staining, cells were incubated with the appropriate antibodies, listed in Table 1 , for 15-30 min on ice, washed and then resuspended in staining buffer (phosphate buffered saline [PBS, EurobioScientific, France] supplemented with 0.5% w/v BSA [EurobioScientific, France], 2mM Ethylenediaminetetraacetic acid [EDTA, Invitrogen, Life Technologies, Courtaboeuf, France]).

All flow cytometry data were acquired with a MACSQuant^® analyzer (Miltenyi Biotech) or a Gallios flow cytometer (Beckman Coulter, Krefeld, Germany) and then analyzed using FlowJo software (version 10.10.0, TreeStar, Ashland, OR, United States) or Kaluza Analysis Software (version 2.3, Beckman Coulter, Krefeld, Germany). During flow cytometry analyses, all gatings were performed on live cells (determined by exclusion of the dye 7AAD [BD biosciences, Le Pont-de-Claix, France]). Results are shown in 2D dotplots and populations are represented by contour plot with outliers.

For surface staining, cells were incubated with the appropriate antibodies for 30 min on ice, washed and then resuspended to 2x10⁷ cells/mL in PBS (EurobioScientific, France) 2% FBS (HyClone, Cytiva Europe GmbH, Velizy-Villacoublay, France), supplemented with 50µg/mL Gentamicin (Gibco, Life Technologies, Courtaboeuf, France).

A FACSAria II SORP cell sorter (BD Biosciences) in the two-way high purity mode was used for this sorting. Cells were recovered in αMEM (Gibco, Life Technologies, Courtaboeuf, France) supplemented with 20% defined FBS (HyClone, Cytiva Europe GmbH, Velizy-Villacoublay, France) and 50µg/mL Gentamicin (Gibco, Life Technologies, Courtaboeuf, France).

A large-scale mass cytometry analysis was performed using phenotypic and functional markers, listed in Table 2 , allowing the identification of populations and subpopulations within samples.

Labelled antibodies were supplied by Fluidigm, Inc Canada, except for CD2, CD4, CD14, CD90 CD110, CD117, CD120a, CD120b, CD133, CD164, CD244 (BioLegend) and CD53 (BD Bioscience). The latter were labelled manually using Maxpar X8 Antibody Labeling Kit for Lanthanides and Maxpar MCP9 Antibody Labeling Kit for Cadmium (Fluidigm, Inc Canada), following manufacturer’s instructions.

Each antibody was functionally validated and titrated in a mix of cells containing peripheral blood mononuclear cell (PBMC), CD34⁺ HSPCs and ProTcells.

Cells stored in liquid nitrogen were thawed in pre-warmed RPMI 1640 Media + GlutaMAX™ (Gibco, Invitrogen, France) and washed in PBS (EurobioScientific, France). Cells were incubated in PBS with Cisplatine Cell-ID™ (Fluidigm, Inc Canada) at 2.5µM for 5 minutes at room temperature (RT) for viability staining. Cells were then washed using a staining solution (PBS 0.5% w/v BSA [EurobioScientific, France], 2mM EDTA [Invitrogen, Life Technologies, Courtaboeuf, France]) and resuspend in 50µL. Staining mix was added on cells during 30 minutes at RT. Following staining, cells were washed with staining solution and fixed with 300µL of staining solution 2% paraformaldehyde (Sigma-Aldrich, Lyon, France) for 15 minutes at RT.

For intracellular staining, cells were permeabilized using Foxp3/Transcription Factor Staining Buffer Kit (Tonbo, US). Briefly, pelleted cells were resuspended within 1mL permeabilization solution and incubated 15min at RT in dark. Cells were then washed 2 times using Flow Cytometry Perm Buffer form the above-mentioned kit.

Following staining, cells were washed and resuspended in 1mL Maxpar Fix and Perm Buffer (Fluidigm, Inc Canada) with 1:4000 of Iridium intercalator (pentamethylcyclopentadienyl‐Ir (III)‐dipyridophenazine, [Fluidigm, Inc Canada]) for incubation overnight at 4°C. Cells were then frozen at -80°C until analysis.

The day of analysis, cells were washed and resuspended in Maxpar Cell Acquisition Solution (Fluidigm, Inc Canada), a high-ionic-strength solution, mixed with 10% of EQ Beads (Fluidigm, Inc Canada) immediately before acquisition. Cell events were acquired on the HELIOS mass cytometer and CyTOF software version 6.7.1014 (Fluidigm, Inc Canada) at the Cytometry Platform of La Pitié-Salpétrière Hospital (CyPS). Mass cytometry standard files produced by the HELIOS system were normalized using the CyTOF Software v. 6.7.1014. This method normalizes the data to a global standard determined for each log of EQ beads.

Mass cytometry standard files analysis was done using OMIQ data analysis software (www.omiq.ai).

Events are gated according to 191Ir positivity, corresponding to stained cells. This gate enables removing events that do not correspond to cells of the sample. All 140Ce-positive events, corresponding to calibration beads, are removed from the analysis. A gate is then placed on events negative for 195Pt which has been used for viability staining. Additional gates are established to remove contaminants from Barium (138Ba), Lanthanum (139La), Osmium (189Os),

Uniform manifold approximation and projection (UMAP) algorithm (83) was used with “2 dimensions” parameters.

A total of 1x10⁶ ProTcells generated from CB HSPCs were intrahepatically transplanted into newborn NSG mice (between 1 and 4 days old). At 6-week-post-transplantation, the thymus was analyzed (using flow cytometry) for human cell engraftment and thymopoiesis. The study protocol was approved by the French Ministry of Higher Education and Research (reference: APAFIS 2010-2015090411495178v4, dated November 2nd, 2015 and APAFIS#29592–2020120216106476 v8, dated February 18th, 2021).

Libraries from CB and mPB ProTcell samples were generated with Chromium Next GEM Single Cell 3’ GEM, Library & Gel Bead Kit (v3.1), following manufacturer’s instruction. Quality control was performed by High Sensitivity DNA assay, analyzed by Agilent 2100 Expert software.

Sequencing was made using Illumina technology (Illumina NextSeq 500) with libraries embedded in NovaSeq S2 FlowCell.

For read alignment and unique molecular identifiers (UMI) quantification, CellRanger software v6.1.2 (https://www.10xgenomics.com/support/software/cell-ranger/latest/release-notes/cr-release-notes) from the Chromium Single Cell Software Suite by 10x Genomics was used on Human cellranger reference 2020-A (genome GRCh38, annotations GENCODE v32/Ensembl 98).

Gene counts for each cell were quantified using the Cell Ranger ‘count’ command with default parameters.

The resultant gene expression matrix was imported into the R statistical environment (v4.2.2) for further analyses. Cell filtering, data normalization and clustering were carried out using the R package Seurat v4.3.0 (84). For each cell, the percentage of mitochondrial genes, number of total genes expressed, and cell cycle scores (S and G1 phases) were calculated. Cells with a ratio of mitochondrial versus endogenous gene expression >0.2 were excluded as putative dying cells. Cells expressing <500 or >8,000 total genes were also discarded as putative poorly informative cells and multiplets, respectively. Putative doublets were identified and discarded using scDblFinder R package (v1.4.0) (85) by imputing doublet rates (dbr) equal to 0.05. Dbr was established in agreement with the number of loaded cells and following the 10X Genomics guidelines. Cell cycle scores were calculated using the ‘CellCycleScoring’ function that assigns to each cell a score based on the expression of the S and G2/M phase markers and stores the S and G2/M scores in the metadata along with the predicted classification of the cell cycle state of each cell. Counts were normalized using Seurat function ‘NormalizeData’ with default parameters. Expression data were then scaled using the ‘ScaleData’ function, regressing on the number of unique molecular identifier, the percentage of mitochondrial gene expression and the difference between S and G2M scores. By using the most variable genes, dimensionality reduction was then performed with principal component analysis (PCA) by calculating 50 principal components (PCs) and selecting the top 30 PCs.

PCA embeddings were corrected for sample batch by applying the Harmony algorithm (v0.1.0) (86), implemented by ‘RunHarmony’ function using the first 30 PCA dimensions and default theta (theta = 3). UMAP dimensionality reduction (87) was performed on the calculated PCs to obtain a two-dimensional representation for data visualization.

Cell clusters were identified using the Louvain algorithm at resolution r = 0.6, implemented by the ‘FindCluster’ function of Seurat. To find the differentially expressed genes from each cluster, the ‘FindAllMarkers’ function (iteratively comparing one cluster against all the others) from the Seurat package was used with the following parameters: adjusted P values <0.05, average log FC >0.25, and percentage of cells with expression >0.1. Dot plot was constructed using ‘DotPlot’ function in Seurat. Density plots were constructed using the ‘Plot_Density_Custom’ function from R packages scCustomize (v2.0.1) (88) and the Bioconductor/R Nebulosa package (89).

Single cell pseudotime trajectories were computed with the R package Slingshot (v2.6.0) (90), which exploits previously computed cell clusters and focuses on the transcriptional lineage that describes the progressive transition from cluster 0 to cluster 6 by identifying genes whose expression changes along that trajectory. The differential expressed genes along the pseudotime tracjectory were analysed using the Bioconductor/R package tradeseq (v1.12) (91).

Statistical analyses of results were conducted using GraphPad Prism (Between v9.4.0.673 and 10.4.0).