NOTCH Activation at the Hematovascular Mesoderm Stage Facilitates Efficient Generation of T Cells with High Proliferation Potential from Human Pluripotent Stem Cells

Adoptive T cell therapies show promise in the treatment of several types of blood cancers. Recent clinical trials demonstrated remarkable clinical outcomes in relapse and refractory lymphoma patients treated with chimeric Ag receptor (CAR)–redirected T cells (1, 2). However, complicated logistics and impaired T cell function in patients with cancers or infection increase the costs and limit the utility of autologous T cell therapies (3). Human pluripotent stem cells (hPSCs) offer the potential to serve as a versatile and scalable source of off-the-shelf T cells for immunotherapies, which could be coupled with genetic engineering technologies to meet specific clinical needs (3). In addition, generation of human induced pluripotent stem cells (hiPSCs) from Ag-specific CTLs and their redifferentiation into functional CTLs provide opportunity to “rejuvenate” and enable scalable production of CTLs (4, 5). Although multiple reports demonstrated T cell generation from hPSCs (6, 7) and the feasibility of hiPSC-based CAR T cell therapies (8), increasing scalability of T cell production from hiPSCs is critical for advancing these technologies to clinic.

Previously, we identified major stages of hematopoietic differentiation from hPSCs and showed the critical role of NOTCH signaling for specification of definitive hematopoiesis and T cells from hPSCs (9–12). Following differentiation in coculture with OP9 or in defined conditions, hPSCs undergo stepwise progression toward APLNR⁺PDGFRα⁺ primitive posterior mesoderm with hemangioblast potential (day 3 of differentiation); KDR^hiCD31⁻ hematovascular mesodermal progenitors (HVMPs) with definitive hematopoietic potential (day 4 of differentiation); VE-cadherin (VEC)⁺CD43⁻CD73⁻ hemogenic endothelium (HE), VEC⁺CD43^loCD235⁺CD73⁻ angiohematopoietic progenitors (AHPs), and VEC⁺CD43⁻CD73⁺ non-HE (days 4–5 of differentiation); and CD43⁺ hematopoietic progenitors (HPs; days 6–8 of differentiation) that include CD235⁺CD41⁺CD45⁻ erythromegakaryocytic progenitors (E-MkPs) and CD235/41⁻CD45^+/− multipotent HPs (MHPs) (Fig. 1) with a lin⁻CD34⁺CD90⁺CD38⁻CD45RA⁻ hematopoietic stem progenitor cell phenotype (9, 11–13).

In the present studies, we focused on identifying the stage of hematopoietic development at which NOTCH activation allows for the highest efficacy of T cell production with robust expansion potential. We found that day 3 APLNR⁺PDGFRα⁺ primitive posterior mesodermal cells did not produce T cells, whereas all downstream subsets except VEC⁺CD43⁻CD73⁺ non-HE and CD235a⁺CD41a⁺CD45⁻ E-MkPs do produce T cells when cultured on OP9-DLL4. As determined by limiting dilution assay (LDA), the highest frequency of T cell precursors was detected from day 4 HVMPs. In addition, we found that T cells generated from HVMPs have the capacity to proliferate for six to seven weeks, in comparison with T cells generated from HE and MHPs, which could only be expanded for four to five weeks. T cell differentiation from hPSCs proceeded through a CD5⁺CD7⁺ progenitor stage that eventually transitions into CD8⁺CD4⁺ double positive cells. To confirm T cell development, we analyzed the genomic DNA for the presence of TCR rearrangements. In vitro–generated T cells were functionally active and proliferated upon stimulation with PMA and IL-2. Upon activation, the cells express CD25⁺CD69⁺ markers, IFN-γ, and cytolytic protein perforin.

These studies should improve our understanding of the early steps of lymphopoiesis from hPSCs and pave the way for developing robust T cell differentiation protocols from hPSCs for disease modeling and adoptive T cell therapies.

Irradiated mouse embryonic fibroblasts, human embryonic stem cell (hESC) line H1 (WA01), and fibroblast-derived DF-19-9-7T hiPSC line were obtained from WiCell (Madison, WI). hESCs and hiPSCs were maintained on mouse embryonic fibroblasts as described previously (14). Mouse OP9 stromal cells were provided by Dr. T. Nakano (Osaka University, Osaka, Japan). The OP9 cell line expressing human DLL4 (OP9-DLL4) was established by using lentivirus expressing human DLL4 under the EF1α promoter. Wild and engineered OP9 cells were cultured in αMEM media containing 20% FBS and passaged every 3–4 d on 0.1% gelatin–coated dishes. Overgrown OP9 cultures were prepared by prolonged culture of confluent OP9 monolayer for an additional 4–8 d (15). Wild-type OP9 and OP9-DLL4 cells used for hematopoietic and lymphoid differentiation were used for up to 50 passages. OP9 cells transduced with DLL1 (OP9-DLL1) were also generated and maintained as described above.

hESCs/hiPSCs were differentiated in coculture with OP9 stromal cells and depleted of OP9 cells using anti-mouse CD29 Abs (AbD Serotec) as described (16, 17). Indicated cell subsets were isolated at days 3, 4, 5, and 8.5 of hematopoietic differentiation in OP9 coculture and used for T lymphoid differentiation. At day 3, ALPNR⁺ cells were isolated by MACS using ALPNR–allophycocyanin Ab and anti-allophycocyanin magnetic beads (Miltenyi Biotec). For isolation of HVMPs on day 4 of differentiation, cells were isolated by MACS using KDR-PE Ab and anti-PE magnetic beads (Miltenyi Biotec). Following enrichment, KDR^hiCD31⁺ and KDR^hiCD31⁻ cells were isolated by FACS. For isolation of day 5 HE, AHP, and non-HE subsets, CD144⁺ cells were isolated by MACS using CD144-FITC Ab and anti-FITC magnetic beads. Positively selected cells were stained further with CD73–allophycocyanin, CD43-PE, and CD235a-PE Abs and then sorted into HE, AHP, and non-HE subsets using a FACSAria cell sorter. Day 8.5 CD43⁺ HPs were isolated with CD43-FITC Ab and anti-FITC magnetic bead by MACS. Positively selected cells were stained with CD235a/CD41a-PE Ab and CD45–allophycocyanin Ab and sorted into E-MkP and MHP subsets using a FACSAria cell sorter (BD Biosciences).

OP9-DLL4 cells were maintained in αMEM media containing 20% FBS on a 0.1% gelatin–coated 10-cm cell culture dish. For lymphoid differentiation, OP9-DLL4 monolayer cultures were prepared in six-well plates. Cells were isolated at different stages of differentiation and seeded on confluent OP9-DLL4 cells in αMEM, 20% FBS, IL-7 (5 ng/ml), Flt3L (5 ng/ml), and stem cell factor (10 ng/ml) at 37°C and 5% CO₂ for 3–4 wk with weekly passage. Every 6–7 d, cells were collected by vigorous pipetting, filtered through a 40-μm cell strainer, and transferred onto fresh OP9-DLL4 monolayer. For Ab staining, cells were prepared in PBS containing 2% FBS, 1 mM EDTA, and 0.1% sodium azide and stained with CD45–allophycocyanin, CD4–allophycocyanin, CD5-PE, CD7-FITC, and CD8-PE T cell–specific Abs. 7AAD (5 μg/ml) was added 10 min before flow cytometry to exclude dead cells. Expression of T lymphoid markers was evaluated following gating of CD45⁺ cells. Cell analysis was performed with the FACSCalibur flow cytometer (BD Biosciences) and MACSQuant (Miltenyi Biotec), and acquired data were analyzed by FlowJo software.

For LDA, floating cells were collected from day 14 cultures of various hemogenic subsets on OP9-DLL4 in T cell conditions. Row A of a 96-well plate received 500 cells per well, and each subsequent row afterward had half as many cells as the previous row (e.g., row B contained 250, row C contained 125, …, row H contained three to four cells). The wells were scored 2 wk later by eye and flow cytometry for CD5⁺CD7⁺-containing cells. Extreme limiting dilution analysis was conducted using the previously established algorithm (18).

hPSC-derived T cells were stimulated with Cell Activation Cocktail according to the manufacturer’s instructions (BioLegend). Briefly, T cells were resuspended in cell culture medium (1 × 10⁶ cells/ml) along with Cell Activation Cocktail (2 μl/ml; BioLegend), IL-2 (10 ng/ml; PeproTech), and brefeldin A (3 μg/ml; eBioscience) for 24 h, then surface-stained with CD25–allophycocyanin and CD69-FITC Abs. Evaluation of perforin and IFN-γ expression was performed using intracellular staining.

Abs used include the following: CD3 (SK7), CD4 (RPA-T4), CD5 (UCHT2), CD7 (M-T701), CD8 (HIT8a), CD31 (WM59), CD41a (HIP8), CD43 (1G10), CD45 (HI30, 2D1), CD73 (AD2), CD144 (557H1, 16B1), CD235a (GA-R2), KDR (89106), TCR α/β (T10B9.1A-BD), and perforin (δG9) from BD Biosciences; CD144 (16B1) and IFN-γ (4S.B3) from eBioscience; APJ (72133) and DLL4 (447506) from R&D Systems; and TCR γ/δ (5A6.E9) from Invitrogen.

For TCR rearrangement assay, DNA was isolated by FlexiGene DNA Kit (QIAGEN, Hilden, Germany). TCR β and TCR γ clonality detection was performed by PCR amplification kit (Invivoscribe, San Diego, CA) with AmpliTaq Gold DNA Polymerase (Applied Biosystems).

Day 4 HVMPs, day 4 HE, and day 5 HE were isolated from H1 hESC/OP9 cocultures and subsequently cocultured on OP9 or OP9-DLL4 for an additional 2 d. After 2 d of secondary coculture, OP9 cells were depleted by MACS using anti-mouse CD29 Ab (15). RNA was extracted with PureLink RNA Micro Scale Kit (Thermo Fisher Scientific) and reverse transcribed using Advantage RT-for-PCR Kit (Takara Bio). Quantitative PCR analysis was performed on all cDNA samples using Power SYBR Green PCR Master Mix (Life Technologies) and the following primers: NOTCH1, 5′-CAATGTGGATGCCGCAGTTGTG-3′(forward) and 5′-CAGCACCTTGGCGGTCTCGTA-3′ (reverse); HEY2, 5′-TTCAAGGCAGCTCGGTAACTGAC-3′ (forward) and 5′-CATACTGATGCACTGCTGGATGG-3′ (reverse); and HES1, 5′-TACCCCAGCCAGTGTCAAC-3′ (forward) and 5′-TCAGCTGGCTCAGACTTTCA-3′ (reverse). PCR was performed using the MasterCycler RealPlex thermal cycler (Eppendorf). Expression levels were calculated by minimal cycle threshold values normalized to GAPDH.

The significance of differences between the mean values was determined by one-way ANOVA followed by Tukey post hoc test as appropriate using GraphPad Prism software (GraphPad, San Diego, CA).

To identify the stage of development and cell population with the most robust potential, we isolated hemogenic populations obtained from H1 hESCs in coculture with OP9 on days 3, 4, 5, and 8.5 of differentiation and cultured them in T cell conditions on OP9-DLL4 (Fig. 1). After 3–4 wk of culture, the presence of T lymphoid cells was detected by flow cytometry. As we previously demonstrated, the most primitive APLNR⁺PDGFRα⁺ mesodermal precursor with hemogenic potential arises in OP9/hPSC coculture on day 3 of differentiation (11, 17). These cells have features of posterior primitive streak and hemangioblast potential, which reflects primitive hematopoiesis. We found that day 3 mesodermal cells failed to produce T cells, which is consistent with their primitive hematopoietic characteristics (Fig. 2A). In contrast, KDR^hiCD31⁻ HVMPs isolated on day 4 of differentiation efficiently generated T cells in OP9-DLL4 cultures (Fig. 2B, 2C), thereby confirming that day 4 HVMPs possess a definitive hematopoietic potential. Similarly, T cells were generated from day 4 and day 5 VEC⁺CD235a/43⁻CD73⁻ HE and VEC⁺CD43/CD235a⁺CD73⁻ AHP subsets (Fig. 2D, 2E), although AHPs produced significantly fewer CD4⁺CD8⁺ cells compared with HE. Consistent with their nonhemogenic nature, VEC⁺CD43/235a⁻CD73⁺ cells failed to produce T cells (Fig. 2D). Assessment of the T cell potential of CD43⁺ cells generated on day 8.5 of differentiation revealed that CD235a/CD41a⁻ CD45^+/− MHPs could efficiently generate T lymphoid cells. As expected, CD235a/CD41a⁺ CD45⁻ E-MkPs were essentially lacking T cell potential (Fig. 2F, 2G). Similar results were obtained with hemogenic subsets obtained from fibroblast-derived DF-19-9-7T hiPSCs (Supplemental Fig. 1A–C).

To determine which hPSC-derived hemogenic population possesses the most robust T cell potential, we assessed the frequency of lymphoid progenitors in hematopoietic cells obtained from cultures of various hemogenic subsets in T cell conditions on OP9-DLL4 after 14 d of culture using LDA. We found the highest frequency of T cell precursors from day 4 HVMPs (1 in 14 HVMPs). The frequencies of T cell progenitors were slightly lower in day 4 and 5 HE and day 8 MHPs (Fig. 3A). The most substantial differences in T cell progenitors were observed in AHP cultures, which revealed a very low 1 in 51 T cell progenitor frequency, thereby suggesting their limited T cell potential. We also evaluated the expansion potential of T cells generated from different subsets. We found that T cells generated from HVMPs have the capacity to proliferate for 6–7 wk, in comparison with HE and MHP subsets, which could only be expanded for 4–5 wk (Fig. 3B). Similar expansion potential was observed in T cell cultures generated from fibroblast DF-19-9-7T hiPSC-derived HVMPs (Supplemental Fig. 1D). Based on these studies, we concluded that the most efficient production of T cells with robust expansion potential could be achieved by culture of day 4 HVMPs in NOTCH-activating T cell conditions on OP9-DLL4.

To investigate differences in NOTCH signaling during initiation of T cell differentiation from HVMPs and HE cells, we compared changes in the expression of NOTCH-associated molecules NOTCH1, HEY2, and HES1 after 2 d of secondary coculture of these cell subsets on OP9-DLL4 versus wild-type OP9 (Supplemental Fig. 1E). We found that coculture on OP9-DLL4 versus wild-type OP9 induces greater upregulation of NOTCH1 and HEY2 expression in HVMPs as compared with day 4 and 5 HE. Fold changes in HES1 expression were higher in day 5 HE, but no difference was observed between HVMPs and day 4 or 5 HE. These findings suggest that the favorable NOTCH signaling activation may occur when T lymphoid cultures were initiated from HVMPs.

As determined by flow cytometry, CD4⁺CD8⁺ T cells generated from HVMPs included populations expressing TCR α/β and TCR γ/δ (Fig. 4A, 4B). Analysis of TCR gene arrangement by PCR revealed DNA rearrangement at the variable, joining, and diversity regions in TCR β and TCR γ loci (Fig. 4C). Finally, to examine whether T cells derived from HVMPs in vitro are functional, we stimulated these T cells using a PMA–ionomycin mixture and IL-2, which are known to activate T cells, and quantitated surface marker expression and cytokine production. After stimulation, CD25 and CD69 double positive T cells derived from HVMPs increased from nothing to 78% (Fig. 5A). These T cells also upregulated the expression of IFN-γ after stimulation (Fig. 5B). Perforin, which is secreted by both CTLs and NK cells, was also observed after stimulation (Fig. 5C). These data suggest that T cells generated by our system are functional.

NOTCH signaling is an essential factor in determining the T lymphoid fate of HPs (19, 20). During development, NOTCH signaling is critical for arterial specification and hematopoietic stem cell development (21–24). In the present studies, we assessed the stage of development at which NOTCH activation could induce the most efficient T cell generation from hPSCs. We found that NOTCH activation at the KDR^hiCD31⁻ mesodermal stage of development allows for the most efficient T cell production with robust expansion potential. Cultures of HE or already established MHPs on OP9-DLL4 produced fewer T cell progenitors and had more limited expansion potential. As we recently reported, NOTCH signaling is critical for specification of arterial HE, which is highly enriched in definitive lymphomyeloid progenitors (10). Thus, it is highly likely that robust T cell production from HVMPs can be explained by the synergistic role of NOTCH signaling in induction of arterial HE and T cell generation. Activation of NOTCH signaling at the HVMP stage promotes arterial HE formation with lymphoid potential, which subsequently produces T cells following continuous exposure to NOTCH signaling. When NOTCH signaling is activated at the MHP stage, its enhancing effect on arterial HE generation is skipped, and therefore T cell production is reduced.

Several studies describe generation of T cells from hESCs (25, 26) and hiPSCs obtained through reprogramming of peripheral blood T lymphocytes (6–8). Although some researchers generated hiPSCs with MART-1–specific TCR (6), others transduced hiPSCs with second-generation CD19 CAR (8). Other studies showed the generation of T cells from reprogrammed Ag-specific CD8⁺ T cells from HIV-1–infected patients (7), mucosal-associated invariant T cells from reprogrammed Vα7.2⁺ cord blood T cells (27), and NKT cells from reprogrammed peripheral blood CD4⁺ (28) and Vα24 NKT cells (29). The hESCs and hiPSCs were differentiated to blood either using coculture with stromal cells (6, 7, 26–29) or an embryoid body differentiation system in serum- and feeder-free conditions (8, 25). Blood cells generated in these conditions were subsequently differentiated into T cells using OP9-DLL1 or OP9-DLL4 cultures. In our study, we used OP9 feeder cells to obtain HPs and then used OP9-DLL4 cultures for lymphoid differentiation. We found that OP9-DLL4 cultures were more efficient than OP9-DLL1 in inducing lymphoid differentiation (Supplemental Fig. 2A). In addition, sorting OP9-DLL4 by high and low DLL4 expression demonstrated that OP9 cells expressing high levels of DLL4 are essential for achieving efficient T cell production (Supplemental Fig. 2B). Whereas most prior studies have used hiPSC-derived HPs for lymphoid differentiation, we have shown that more efficient T cell production could be achieved when coculture with OP9-DLL4 was initiated using definitive hemogenic progenitors isolated at earlier stages of hematopoietic development, such as the HVMP stage.

The promising clinical results with adoptive T cell therapies call for the search for novel universal T cell sources to simplify logistics and reduce the costs of these therapies. Because hPSCs can be expanded indefinitely and engineered to express CARs, they potentially can serve as an endless supply for off-the-shelf CAR T cells. Alternatively, hiPSCs can be generated from T cells with particular TCR and used to generate T cells expressing TCR of interest (6). However, several challenges for applying hPSC-based strategies for T cell therapies remain. It is important to develop chemically-defined conditions that allows for robust T cell production with optimal expansion potential. In addition, strategies to overcome potential graft-versus-host disease and rejection of infused cells have to be developed. One of the most attractive strategies to prevent graft-versus-host disease could be based on infusion of T cell progenitors rather than mature T cells. T cell progenitors undergo positive and negative selection in host thymus and become restricted to host MHC (30). Alternatively, the TCR locus can be deleted in hPSCs to allow for production of TCR-less T cells. Immune rejection of infused T cells can be mitigated by establishing hiPSC banks with most common HLA haplotypes or from HLA-homozygous donors. It is estimated that only 55 cell lines homozygous for highly conserved HLA haplotypes would provide a beneficial HLA match for 80% of the population in Japan (31), and only 150 homozygous lines would provide a match for more than 93% of the U.K. population (32). In conclusion, demonstration of the importance of NOTCH activation at the hematovascular mesodermal stage of hPSC differentiation to amplify T cell generation in current studies offers an optimized strategy for achieving robust T cell production from hPSCs and advancing their use for immunotherapy.

We thank Dr. Toru Nakano (Osaka University, Osaka, Japan) for providing OP9 cells, Mitch Probasco (Morgridge Institute for Research) for cell sorting, and Mathew Raymond (Wisconsin National Primate Research Center) for editorial assistance.

The authors have no financial conflicts of interest.