Key gene controllers of blood and vessel cell development found using targeted CRISPR in mice

As F_SP HEM is further differentiated towards the haemato-endothelial fate compared to the other mesodermal populations, F_SP HEM should display an increased developmental potential to form blood and endothelial lineages. To examine this, we isolated DN, P_SP, DP priMes, and F_SP HEM cells by FACS and placed them into haemangioblast (HB) culture to support their differentiation to haematopoietic and endothelial progenitors as well as vascular smooth muscle cells (VSM)^6,25. After 2 days, we used CD41 and VE-cadherin (VE-cad) expression to distinguish haematopoietic (CD41⁺), endothelial (VE-cad⁺), pre-haematopoietic stem and progenitor cells (pre-HSPC, CD41⁺VE-cad⁺), and VSM cells (CD41^–VE-cad^–)²⁶ by flow cytometry. As there is no specific surface marker for VSM cells suitable for flow cytometry, we will refer to CD41^–VE-cad^– cells as VSM-enriched, acknowledging the inferred nature of this identity. The HB culture experiments confirmed that DP priMes indeed had less capacity to form haematopoietic and endothelial lineages compared to F_SP HEM, which generate a higher frequency of CD41+ and VE-cad⁺ cells and fewer VSM-enriched cells (Fig. 1d, e). Accordingly, and as expected, DN and P_SP had a further reduced capacity to form haematopoietic and endothelial lineages compared to DP priMes/ F_SP HEM (Fig. 1d, e). To examine the lineage potential of the two different mesodermal populations at higher resolution, we next established a high-dimensional flow cytometry panel using antibodies recognising 23 surface markers (Supplementary Table 1). We performed unsupervised clustering to define nine major clusters based on their expression patterns. F_SP HEM was more proficient at forming haematopoietic and endothelial lineages, including primitive (primitive erythrocyte progenitors (pEryP, CD41^+/- CD71⁺c-kit^–) and definitive blood subsets comprised of definitive haematopoietic progenitors (dHPC, CD41^+/−c-kit⁺CD71⁺CD43⁺), pre-haematopoietic stem and progenitor cells (pre-HSPC, CD41⁺VE-cad⁺Tie2⁺CD34⁺c-kit⁺) and macrophage progenitors (MacP, CD41^+/−c-kit^+/−CD11b/c⁺CD44⁺CD45⁺Cx3Cr1⁺) and endothelial type 1 (Tie2⁺FLK1⁺CD31⁺VE-cad⁺c-kit⁻) and type 2 (Tie2⁺FLK1⁺CD31⁺VE-cad⁺c-kit⁺) clusters, whereas DP priMes had a higher potential to form VSM-enriched cells (CD41⁻VE-cad⁻CD71⁻Epcam⁻) (Fig. 1f, g and Supplementary Fig. 2). To further assess the functional differences of DP priMes and F_SP HEM, we established a long-term haematopoietic assay to promote the maturation of haematopoietic progenitors. We compared (i) the blood-forming potential and (ii) the ability to generate mature blood lineages of these two mesodermal populations. Defined numbers of FACS-sorted DP priMes and F_SP HEM (18k, 10k, or 5k) isolated from EB cultures were plated and analysed after 9.5 days. Whereas F_SP HEM generated haematopoietic cells under all tested conditions, DP priMes showed reduced blood-forming potential, particularly at lower cell inputs: For 5k cells, 100% of F_SP HEM formed colonies, but only 8% of DP priMes (Supplementary Fig. 3a, b). Additionally, for blood lineages analysis, we scored 8 randomly selected wells for each condition and analysed the number of live cells and the frequency of leukocytes (CD45+) and erythroid (CD71 + ) cells, the two main blood lineages forming in the assay (Supplementary Fig. 3c–f). F_SP HEM consistently showed higher number of live cells and increased frequencies of CD45+ and CD71+ cells in all analysed conditions compared to DP priMes (Supplementary Fig. 3c-f). Collectively, these results demonstrate that mesodermal populations, defined by FLK1/PDGFRα expression, differ significantly in their blood-forming capacity. F_SP HEM is functionally distinct from DP priMes with superior potential to form mature haematopoietic cells due to increased potential to form haematopoietic progenitors. This highlights the importance of FLK1/PDGFRα expression in marking functionally distinct mesodermal subpopulations during early haematopoietic development.

Lastly, to further underline the relevance of our ESC model, we performed a scRNA-seq characterisation of both stages of our model employing SORT-seq. By relying on the lineage defining markers FLK1, CD41, VE-cad, and c-kit, we isolated single cells corresponding to 7 distinct populations from the EB and HB stage (Supplementary Fig. 4a). Our analysis confirmed the generation of distinct lineages characterised by appropriate signature gene expression patterns, including pHPCs (primarily primitive erythrocyte progenitors), dHPCs, VSM-enriched cells, and endothelial lineage (Supplementary Fig. 4b–d). Additionally, we integrated the transcriptional profiles of our in vitro-generated cells with a mouse gastrulation atlas²⁷, which includes cells isolated from E6.5–E8.5 embryos. A high proportion (39.8%) of the in vitro-generated cells could be directly projected onto the in vivo haemato-endothelial trajectory graph, evidencing the similarities between in vitro- and in vivo- generated cells (Fig. 1h). Taken together, our functional analysis, single-cell and bulk RNA-seq profiles confirm FLK1 and PDGFRα as appropriate markers to distinguish mesodermal populations and attest that the model faithfully recapitulates early mesodermal commitment and subsequent blood differentiation.

To understand those genes' impact on cellular differentiation, we further analysed the frequency of total FLK1+ cells (including both DP priMes and F_SP HEM) and calculated the ratio of F_SP HEM/ total FLK+ cells in cultures derived from the KOs (Fig. 3c and Supplementary Fig. 6e). Importantly, cultures generated from KOs of our four validated candidates generated similar frequencies of total FLK1+ cells compared to the control that resulted in a significant decrease in the ratio of F_SP HEM/ total FLK+ cells (Fig. 3b, c). This points to a specific role of these genes in regulating the transition of DP priMes towards F_SP HEM and not to a general defect of FLK1+ mesodermal commitment. In addition, no significant differences in the total number of live cells were detected between KO cultures or compared to controls, ruling out an effect on cell proliferation/viability (Fig. 3d). Surprisingly, the gene KOs for Ldb1, Smad1, and Six4 had an equally strong effect on arresting DP to further differentiate towards F_SP, as the known TF ETV2. We subsequently term these validated TFs F_SP HEM-driving TFs. Additionally, to understand if the identified TF factors lead to a delayed maturation of F_SP HEM differentiation, we extended the EB culturing time (Supplementary Fig. 6f–h). We could not detect any increase of F_SP HEM on EB day 6 in the KO cultures compared to day 5 and conclude that the identified F_SP HEM-driving TFs lead to disruption or block in differentiation, not a shift in timing (Supplementary Fig. 6f–h).

We next considered the enriched genes, analysing four hits and validating one: Zbtb7b (rank2). EB cultures of Zbtb7b KO cells generated F_SP HEM at a significantly higher frequency than controls, leading to an increased ratio of F_SP HEM/ total FLK+ cells (Fig. 3e–g). Zbtb7b KO cells showed decreased frequencies of earlier transitions, including DP priMes and P_SP (Supplementary Fig. 6i–k). Importantly, Zbtb7b KO cultures contained a comparable number of live counts and total FLK1+ cells as did controls (Fig. 3h and Supplementary Fig. 6l). These results suggest that Zbtb7b KO cells have an enhanced potential to commit to F_SP HEM at the expense of DP priMes/ P_SP, without affecting cell proliferation: Zbtb7b therefore functions as a repressor of F_SP HEM differentiation. Uniquely, EB cultures of Npm1 KO cells resulted in significantly fewer live cells, but F_SP HEM and DP priMes still formed, pointing to a general function of this gene in cell proliferation (Fig. 3e, h and Supplementary Fig. 6i). Due to the low number of surviving cells, significantly fewer F_SP HEM and DP priMes were generated, but a similar ratio of F_SP HEM/total FLK1+ cells were achieved compared to the Ctrl (Fig. 3e, g, h). These results agree with previous reports, which showed that Npm1 suppression inhibits ESC proliferation but does not have a general effect on cell differentiation of neural stem cells^29,30. As we aimed to identify factors that act specifically on the differentiation of F_SP HEM, we excluded NPM1 from further analysis.

Taking advantage of our scRNA-seq dataset, we then analysed the gene expression of the validated candidates along the mesodermal differentiation trajectory. In agreement with previous reports³¹, Etv2 showed a peak of expression at the transition from DP towards F_SP mesoderm; Smad1 gene expression peaked earlier than Etv2, at the beginning of DP mesoderm development but was broadly expressed in all populations; while Ldb1, Six4 and Zbtb7b were equally expressed in the different mesodermal populations, with Ldb1 showing the highest and Zbtb7b the lowest level of expression (Fig. 3i). These data show that our candidate genes (except for Etv2), would have been difficult to identify by employing a differential gene expression analysis between DP priMes and F_SP HEM.

In summary, by applying strict criteria to selecting genes, we were able to validate our hits with high efficiency. We confirmed the critical role of the known master regulator ETV2 in enabling F_SP HEM differentiation and identified additional factors, including SMAD1, LDB1, SIX4, and ZBTB7b, which act as drivers and repressors of F_SP HEM commitment.

The clusters did not form distinct, well- segregated populations, instead, distinct subpopulations are positioned next to each other, underpinning continuous developmental processes. Based on a differential gene expression analysis incorporating well-established mesodermal and haemato-endothelial genes, the developmental trajectory towards HEM could be defined (Fig. 5a, b and Supplementary Table 2) from C1-C2-C3-C4-C5-C6 (priMes1/2; Foxf1^hi mesoderm; pre-HEM and HEM 1/2, respectively). We denoted C1 (priMes 1) as the most naïve mesodermal cluster due to high expression of early mesodermal markers such as Eomes, Mesp1 and Mixl1 (Fig. 5b). By contrast, the most advanced cluster along the HEM trajectory C6 (HEM 2) was characterised by upregulation of known haematopoietic genes crucial in the process of endothelial-to-haematopoietic transition and blood development including Tal1, Gata2 and Runx1 (Fig. 5b, c). Following our previous nomenclature of DP priMes and F_SP HEM, we assigned the clusters C1-4 to DP priMes and C5 and C6 to F_SP HEM, based on the expression of FLK1 (encoded by KDR gene) and PDGFRa (Fig. 5b, c). Combining our two scRNA-seq datasets further supports this assignment (Supplementary Fig. 7b). Strikingly, several clusters were absent in, or specific to, individual KOs and cluster frequencies are distinct across conditions: for example, Etv2 KO and Ldb1 KO cells are stalled early during development, with only C1-C3 and C1-C5 being present, respectively (Fig. 5a, d, e). This highlights the early and critical role of ETV2 during HEM development and shows that LDB1 plays a subsequent role, allowing the cells to progress further. Interestingly, Etv2 starts to be upregulated in C4 (pre-HEM) and has its highest expression in C5 (HEM 1) (Fig. 5b, c). The early developmental block of Etv2 and Ldb1 KO cells agrees with the differentiation experiments where neither endothelial nor haematopoietic lineages develop in those KOs (Fig. 4a). In Six4 KO cell cultures, all clusters were present that were in the Ctrl samples, albeit with altered frequencies, but—uniquely among the F_SP-driving TFs—Six4 KOs were able to form cells belonging to the most advanced C6, albeit at highly reduced frequency (Fig. 5a, d, e). This is in agreement with our differentiation experiment, where Six4 KO cells were able to form all haematopoietic lineages (Fig. 4a). Concurrently, the early C1 (primitive mesoderm1) was expanded in Six4 KO cell cultures compared to the Ctrl (Fig. 5a, d, e). Smad1 KO cells were the most different to the Ctrl and the other candidate KOs, forming three clusters that were almost absent in the other cultures: C7 (neuroectoderm), C8 (mixed mesoderm) C9 (early endothelial progenitors). C7 was characterised by high expression of Nnat, Dll1 and Dll3, which are collectively expressed during neural cell development (Fig. 5b and Supplementary Table 2). C8 (mixed mesoderm) did not resemble any mesodermal population commonly found during development and was characterised by high expression of Tll1 (Tolloid-Like Protein 1) and Opticin (Optc gene) (Fig. 5b and Supplementary Table 2). Alongside, C9 (early endothelial progenitors) was also mainly found in Smad1 KO cell cultures and was defined by high expression of endothelial markers including Plxnd1, Rasip1, and Rhoj (Fig. 5a, b and Supplementary Table 2). Despite the different composition of mesodermal subpopulations, Smad1 KO cells were still able to form blood and endothelial lineages albeit at highly reduced frequency (Fig. 4a). Zbtb7b KO cells gave rise to a higher frequency of C5 (HEM1) and a lower number of cells in C3 (Foxf1^hi mesoderm) compared to the Ctrl (Fig. 5a–e). C5 was characterised by the highest expression of Etv2 and additional TFs associated with haematopoietic development, including Gata3 and Lmo2 (Fig. 5b, c). Therefore, C5 is clearly committed towards blood development (Fig. 5b, c). C3 was characterized by expression of early mesodermal markers like Hand1 and Mest and a high expression of the TF Foxf1 (Fig. 5b, c and Supplementary Table 2). To note, a previous study implicated the expression of Foxf1 in inhibiting haematopoietic lineage commitment during mesoderm specification³³. A reduced frequency in C3 or Foxf1 expressing mesodermal cells could therefore result in the observed phenotype of enhanced F_SP HEM commitment of Zbtb7b KO cells and increased commitment towards the haematopoietic and endothelial lineage (Fig. 4a). The intermediate C4 (pre- HEM ) revealed only two distinct marker genes (Lmo2 and Fgf3) and was present at low frequencies in all analysed conditions (ranging from 2 to 6% of the cells in the different genotypes) (Supplementary Table 2). Zbtb7b KO cells had the highest frequency of C4 among all analysed conditions (6.53%), pointing to an advanced commitment towards the HEM lineage (Supplementary Fig. 7c). Cells belonging to C0 did not express FLK1 but rather the pluripotency marker Nanog; accordingly, we denoted them as uncommitted cells which were equally present in all genotypes (between1- 3%) (Fig. 5a, b, d). Cells belonging to C0 were most likely caught during FACS sorting due to a less stringent gating strategy and were excluded from further analysis. To characterise all cells committed towards mesodermal fate we did not select for cells that only highly expressed FLK1, accepting the risk of sorting and sequencing FLK1neg cells.

To further analyse the developmental order of the clusters and to not solely rely on the expression analysis of known marker genes, we performed condition-wise pseudotime analysis (Fig. 5f and Supplementary Fig. 7d). We used C1 (priMes1) as the root, based on its naïve mesodermal transcriptomic profile. We confirmed C6 (HEM 2) as the most committed mesodermal population following the haemato-endothelial mesodermal developmental trajectory. Etv2 KO cells, Ldb1 KO cells and Six4 KO cells followed the same trajectory as the Ctrl cells, with the former two being stalled early during development. Intriguingly, Zbtb7b KO cells, which are largely composed of the same clusters as the Ctrl, formed the most advanced C6 (HEM 2) via a different trajectory, mainly through C5 (HEM 1), and included very few cells belonging to C3 (Foxf1^hi mesoderm) (Fig. 5f). This could indicate that Zbtb7b KO cells do not form C3 or pass through it faster than the Ctrl cells. Smad1 KO cells appeared to undergo an entirely different developmental trajectory compared to the Ctrl and formed a trajectory of the Smad1 KO specific clusters via C1-C7-C8-C9 (Fig. 5f). We additionally, projected the in vitro generated cells of the individual conditions onto the in vivo haemato-endothelial trajectory graph (Fig. 1e and Supplementary Fig. 7e). In agreement with our UMAP analysis, Etv2, Ldb1 and Smad1 KOs cells display an early developmental block, whereas Six4 KO cells can be projected further onto the trajectory (Supplementary Fig. 7e). Smad1 KO cells had the lowest frequency of all analysed conditions (16.5% vs.20.4 % for Ctrl) which agrees with the alternative trajectory that Smad1 KO cells follow (Fig. 5f and Supplementary Fig. 7e). Collectively, our scRNA-seq data of the candidate gene KOs reveals that each TF impacts distinctly on the mesodermal developmental trajectory and distinct mesodermal subpopulations form during EB cultures, which are differently impacted by candidate gene KO.

Mouse ESCs (HA36CB1, 129×C57BL/6) were cultivated on 0.2%-gelatine-coated dishes in Dulbecco's Modified Eagle Medium (DMEM) supplemented with 15% foetal calf serum (FCS), 1x non-essential amino acids, 1 mM L-glutamine, LIF, and 0.001% β-mercaptoethanol.

The KO cell lines were generated by co-transfecting pX330-U6-Chimeric_BB-CBh-hSpCas9 (Addgene 42230) with two sgRNAs (sequences see Supplementary Table 3). Control samples were transfected with non-targeting sgRNAs (sgRNA1_GCGTATCTACCCTACCGCCG, sgRNA2_GCGGTTACCGCGAAAACCAT). pRR-Puro recombination reporter⁵² (Addgene 65853) was co-transfected, and after 36 h, cells were treated with 2 μg/ml of puromycin for another 36 h. KO clones were validated by Sanger sequencing. KO validation by western blot detection was performed as described previously⁵³ for the identified candidates Etv2, Ldb1, Six4, Smad1 and Zbtb7b (Supplementary Fig. 8a–e). Transfections were conducted using Lipofectamine 3000 reagent (Thermo Fisher Scientific) at a 2:1 Lipofectamine/DNA ratio in OptiMEM (Thermo Fisher Scientific).

Twenty-four hours prior to the initiation of EB cultures, ESCs were transferred into gelatinized plates containing Iscove's Modified Eagle Medium (IMDM) (Gibco). For the generation of EB cultures, ESCs were trypsinized then plated at 10,000 cells/ml in non-adherent 10 cm² petri dishes in EB medium (IMDM supplemented with 1% L-glutamine (Gibco), 10% FCS (Gibco), 0.6% transferrin (Roche, 10652), 50 µg/ml ascorbic acid (Sigma, A4544) and 0.03% monothioglycerol (Sigma, M6145). BMP4 was added from day 0 of EB culture, and bFGF, activin A and VEGF from day 2.5 (all Peprotech) all at 5 ng/ml. After 5 days in culture, EBs were harvested, and TrypLE Express Enzyme (1×) (Gibco, 12605036) was used to generate a single-cell suspension.

Prior to HB culture, mesodermal populations underwent FACS sorting (for either DN, P_SP, DP priMes, F_SP HEM or total FLK1+) and 0.5 × 10⁶ cells were cultured on 0.2%-gelatine-coated plates in IMDM supplemented with 1% penicillin–streptomycin, 1% L-glutamine, 15% FCS (Gibco), 0.06% transferrin (Roche), 0.03% monothioglycerol (Sigma, M6145), 50 mg/μL of ascorbic acid (Sigma, A4544), 5 ng/ml VEGF (Peprotech) and 10 ng/mL IL-6 (Peprotech). After 2.5 days in culture, cells were harvested for flow cytometry analysis. TrypLE Express Enzyme (1×) (Gibco, 12605036) was used to generate a single-cell suspension.

Defined numbers (18,000, 10,000, or 5000 cells) of FACS-sorted F_SP HEM or DP priMes were plated on 0.2%-gelatine-coated 96-well plates. For each cell number a minimum of 24 wells were seeded. The cells were first cultured for 2.5 days in HB culture media (to induce the differentiation to haematopoietic progenitors) and subsequently in conditions supporting the maturation of haematopoietic cells in IMDM supplemented with 1% L-glutamine, 10% FCS (Gibco), 0.6% transferrin (Roche), 0.03% monothioglycerol (Sigma, M6145), 50 mg/μL of ascorbic acid (Sigma, A4544), 100 ng/ml SCF (Peprotech), 1 ng/ml IL-3 (Peprotech), 10 μg/ml G-CSF (Peprotech), 5 μg/ml IL-11 (Peprotech), 5 μg/ml IL-6 (Peprotech), 5 μg/ml TPO (Peprotech), 10 μg/ml M-CSF (Peprotech) and 2000 U/ml erythropoietin (R&D Systems). After 9.5 days in culture wells were scored for the presence of haematopoietic cells and cells were harvested for flow cytometry analysis. TrypLE Express Enzyme (1×) (Gibco, 12605036) was used to generate a single-cell suspension.

No statistical method was used to pre-determine sample size. All experiments were confirmed by replication at least once. For each generated KO line, three independent clones were generated. The exact handling of replicates is depicted in figure panels. Randomization/Blinding is not relevant for the experiments performed in this study. Samples were allocated to either wild type control or mutant and processed in parallel through identical analysis pipelines.

Further information on research design is available in the linked to this article. Nature Portfolio Reporting Summary