T cells develop from multi-potent hematopoietic progenitors in the thymus and provide adaptive protection against pathogens and cancer. However, the emergence of human T cell-competent blood progenitors, and their subsequent specification to the T lineage, has been challenging to capture in real time. Here, we leveraged a pluripotent stem cell differentiation system to understand the transcriptional dynamics and cell fate restriction events that underlie this critical developmental process. Time-resolved single cell RNA sequencing revealed that cell-cycle exit, downregulation of the multipotent hematopoietic program, and upregulation of >90 lineage-associated transcription factors all occur within a highly co-ordinated and narrow developmental window. Computational gene-regulatory network inference elucidated the transcriptional logic of T lineage specification, uncovering an important role for YBX1. We mapped the differentiation cell fate hierarchy using transcribed lineage barcoding and mathematical trajectory inference and discovered that mast and myeloid potential bifurcate from each other early in haematopoiesis, upstream of T lineage restriction. Collectively, our analyses provide a quantitative, time-resolved model of human T cell specification with relevance for regenerative medicine and developmental immunology.
Overall design: For time course and lineage tracing experiments, the following “Cord-blood-optimized” serum-free media was used for T cell differentiation; Iscove’s modified Dulbecco’s medium (IMDM) with GlutaMAX basal medium (Thermo Fisher Scientific, 31980030) supplemented with 20% BIT 9500 serum substitute (STEMCELL Technologies, 09500), 60 μM ascorbic acid (Sigma-Aldrich, A8960), 24 μM 2-mercaptoethanol (Sigma-Aldrich, M3148), 0.05% low-density lipoprotein (STEMCELL Technologies, 02698), stem cell factor (SCF) (23.9 ng/ml; R&D Systems, 255-SC), Flt3L (8.7 ng/ml; R&D Systems, 308-FK), IL-3 (5.3 ng/ml; R&D Systems, 203-IL), CXCL12 (9.7 ng/ml; R&D Systems, 350-NS), IL-7 (10 ng/ml; R&D Systems, 207-IL), and TNFα (4.9 ng/ml; R&D Systems, 210-TA). In parallel work that took place during the course of this study, we developed an improved “PSC-optimized” progenitor T cell differentiation media which supports more efficient survival. We therefore used this media for CRISPR perturbation experiments, and their associated controls, to maximise viable cell recovery. The improved media comprises IMDM with GlutaMAX basal media (Thermo Fisher Scientific, 31980030) with 4% B27 minus Vitamin A (Gibco, 12587010), 60 µM Ascorbic acid (Sigma-Aldrich, A8960), 24µM 2-Mercaptoethanol, SCF (12.37 ng/ml), Flt3L (8.61 ng/ml), IL-3 (0.97 ng/ml; R&D Systems, 203-IL), CXCL12 (97.39 ng/ml), IL-7 (65.25 ng/ml), and TNFα (0.07 ng/ml). Harvested cells from the EHT culture phase were pelleted, re-suspended in T cell differentiation media at a split ratio of 1:3 (unless otherwise specified), and transferred in 100uL volume per well on to 96 well plates freshly coated with DLL4 and VCAM1 as described above . Cells were incubated at 37°C, 5% CO2. The media was topped up with an additional 100uL of media per well (bringing the total volume of media to 200uL) after 3 or 4 days of progenitor T cell differentiation. Cultures proceeded for up to a maximum of 7 days. Timecourse dataset We performed a reverse differentiation time course such that new differentiations were initiated every 24 hours for 14 consecutive days. Thus, samples representing each developmental time point could be collected at the same time for parallel capture. For each differentiation time point, we thawed an aliquot of the same cryopreserved stock of iPS11-derived hemogenic endothelial cells produced as described above. For each timepoint, at least 5 independent wells were seeded for downstream pooling. EHT cultures and progenitor T cell differentiations were performed as described above. On the final day of the timecourse, all timepoint samples were harvested and an additional aliquot of the input population was thawed. For samples from the progenitor T cell differentiation stage, cells were harvested by pipetting, whereas for the EHT timepoints, non-adherent cells were harvested by pipetting before adherent cells were harvested using 30uL per well of 0.25%Trypsin EDTA (Thermo Fisher Scientific, 25200072). For the sample coresponding to the passage day, adherent and floating cells were sequenced separately. Each sample was pelleted by centrifugation at 300x g for 5 minutes. Replicate wells were pooled together and washed with PBS. Samples were Fc blocked (BD, 564220) and labelled with zombie-UV viability dye (BioLegend, 423108) on ice in the dark for 20 minutes. Next, each sample was labelled with 0.1 ug of a cell hashing antibody (TotalSeq-B anti-human Hashtag antibodies 1 through 8, BioLegend). Cells were labelled for 30 minutes on ice in the dark. Next, samples were washed 3 times with HF buffer and counted prior to pooling. Samples were combined in to two pools, one for the EHT time points and another for the progenitor T cell differentiation time points. To enrich for Zombie-UV-negative viable cells, the EHT time point pool was subject to Fluorescent Activated Cell Sorting (FACS) using a MoFlo Astrios (Beckman Coulter). Lineage barcoding dataset In vitro differentiations from cryopreserved CD34+ cells were initiated as described above, with the modification that 7500 cells were seeded on to one well of 96 well plate for transduction and sampling while additional “proxy” wells were seeded at the same density for flow cytometry analysis and estimating the cell count in the experimental well. Three days after seeding, cells were transduced with 3.75uL per well of the concentrated barcoding virus. Twenty-four hours later, wells the cells were fed with an additional 100uL per well of EHT media. Five days after seeding (48 hours after transduction), the well was mixed by gentle pipetting and half of the total volume was sampled for scRNA sequencing (Chromium Single Cell 3’ Reagents kit v3.1). One week following the initiation of the differentiation culture (4 days after transduction), the remaining cells were collected and passaged 1:2 on to plates coated with DLL4 and VCAM1 in EHT media as described above. Three days later, remaining cells were harvested, counted and a sub-sample of 16,500 cells were taken for scRNA sequencing.
Less...