A subpopulation of subaortic mesenchyme in the AGM co-expresses NG2 and Runx1

We examined the expression of PC/vSMC markers in the dorsal aorta of E10.5 and E11 mouse embryos. Wholemount immunostaining and immunohistochemistry on frozen sections were performed using PC/vSMC markers NG2 or SMA with CD31, an endothelial and HSPC marker (TableS1). Imaging analysis showed that NG2+SMA+CD31- vSMCs surround NG2-SMA-CD31+ endothelial cells (Figs.1a, S1a, b), confirming previous reports27. Further to its expression in hematopoietic and hemogenic endothelial cells, Runx1 was also detected in the sub-aortic mesenchyme22,23. Therefore, we hypothesised that at least some of these cells also express NG2. We first confirmed that both intra-aortic hematopoietic cell clusters (IAHCs) (Fig.1b, stars) and hemogenic endothelial cells (Fig.1b, arrowheads) are Runx1+; we also identified a subpopulation of NG2+ PC/vSMCs, mainly located in the ventral aspect of the dorsal aorta, that also express Runx1 (Fig.1b, arrows). Other Runx1+ cells in the perivascular area do not express NG2 (Fig.1b). Finally, we confirmed our recent study28 that some cells around the notochord express NG2 in the trunk (Fig.S1a, circle). However, these peri-notochord cells do not express SMA, CD31 (Fig.S1a, circle) nor Runx1 (Fig.S1ef). To confirm the presence of NG2+Runx1+ cells in the E11 AGM, we used Runx1-IRES-GFP mouse embryos29. In these GFP knock-in mice, GFP intensity correlates with Runx1 expression level. Flow cytometric analysis showed the presence of a distinct population of NG2+Runx1(GFP)+ cells in the AGM (Fig.1c). These cells first appear at E10, in line with the presence of Runx1 in mesenchymal cells30 and importantly, their frequency peaks at E10.5 (Fig.1d). Together, these data show that in the AGM, a subset of the sub-aortic mesenchyme expresses both NG2 and Runx1 and that the highest frequency of these cells coincides with the onset of HSC generation at E10.5.

a Three-dimensional (3D) wholemount immunostaining with SMA, CD31 and NG2 of E10.5 (3138 somite pairs (sp)) WT dorsal aorta; b NG2 and Runx1 expression on single plane wholemount WT E10.5 sections. NG2+Runx1+vSMCs (arrows), hemogenic endothelial cells (arrowheads) and intra-aortic hematopoietic clusters (IAHCs, stars) (TableS1); c Representative example of flow cytometric analysis of NG2+Runx1(GFP)+ (green box) in E10.5 Runx1-IRES-GFP AGM and E10.5 WT control. d Percentages of NG2+Runx1(GFP)+ cells in E9 (21-25sp) body (n=6), E10/E10.5/E11 AGMs (n=8/7/7), N=5, Kruskal-Wallis and Dunns post-hoc test. e Representative examples of wholemount 3D-images showing SMA, CD31 and NG2 in E10.5 cKO dorsal aortae; f SMA, Runx1 and CD31 immunofluorescence of E11 WT and cKO transversal frozen sections; n=WT/cKO: 2/2, N=2. g cKit and CD31 wholemount 3D-images in E10.5 WT and cKO AGM; h Number of intra-aortic hematopoietic clusters (IAHCs) in E10.5 AGM; n=WT/KO: 5/4, N=4. Number of colony forming unit-culture (CFU-C) in i E10.5 (31-38sp) AGM; n=WT/HET/KO: 14/10/5 embryos; N=7 and j E11 (4352sp) AGM; n=WT/HET/KO: 22/8/19 embryos; N=11; one-way ANOVA and Tukeys post-hoc test (TableS2). k Percentages of donor cell chimerism 4-months post-transplantation of 6 E11 WT (NG2+/+;Runx1fl/+or NG2+/+;Runx1fl/fl), 7 HET (NG2-Cre;Runx1fl/+) and 6 cKO AGMs (NG2-Cre;Runx1fl/fl) into sub-lethally adult irradiated recipients (1xAGM cells transplanted/recipient; N=4). Each dot represents one recipient. Mice are reconstituted when 5% donor cells are found in the host peripheral blood (dashed line); one-tailed Z score test for two population proportions (TablesS3 and S4). For wholemount staining in a, b, e, g: WT/cKO (N=6/4): SMA (n=9/7), CD31 (n=10/7), cKit (n=3/2), NG2 (n=3/1) and WT Runx1 (n=4) in 3 distinct combinations (TableS1). D = dorsal, V = ventral. N = number of independent experiments; n = number of biological samples (embryos). All data are presented as mean valuesSEM. Source data for d, h, i, j and k are provided as a Source Data file.

Runx1 deletion in endothelial cells impairs HSC emergence in the AGM24,25,26. However, the effect of Runx1 deletion in PC/vSMCs on hematopoiesis in vivo is still unknown. To address this, we examined conditional knock-out (cKO) NG2-Cre;Runx1fl/fl mouse embryos. In previous studies, the NG2-Cre mouse strain revealed a role for pericytes in supporting both fetal liver and adult bone marrow HSC maintenance31,32. Our data shows that E10.5 and E11 cKO embryos do not exhibit visible vascular abnormalities. This was confirmed by the normal expression of CD31, SMA and NG2 (Figs.1e, f, S1c, d) in the AGM. In contrast, SMA+Runx1+ PC/vSMCs with low expression of Runx1 were reduced in the cKO dorsal aorta compared to WT littermate controls (Figs.1f, S1eg). CD31+Runx1+ endothelial cell number and frequency was also decreased (Fig.S1g). Furthermore, CD31+cKit+ IAHC numbers were significantly reduced by three-fold (p=0.02) (Fig.1g, h). Hematopoietic progenitor (HP) assays were performed to test if hematopoietic function was affected. All HP numbers were significantly reduced in cKO AGMs at both E10.5 (Fig.1i, TableS2) and E11 (Fig.1j, TableS2). To test whether definitive HSCs were also affected, we performed HSC assays in vivo. At 1- and 4-months post-transplantation of AGM cells into sub-lethally irradiated mice, chimerism and multilineage reconstitution were examined by flow cytometry in the peripheral blood. Compared to the WT littermate control group, in which 66.7% (4 out of 6) recipients were reconstituted, only 14.3% (1 in 7, p=0.025) and 16.7% (1 in 6, p=0.040) mice injected with heterozygous or homozygous cKO AGMs, respectively, were reconstituted over the long term (Fig.1k, TablesS34). These findings indicate that the absence of Runx1 in aortic NG2+ cells impairs HSC generation and/or maintenance and HP development in the AGM.

To test whether NG2+ cells contribute directly to hematopoietic lineages, we isolated NG2+ and NG2+Runx1(GFP)+cells from E11 WT and Runx1-IRES-GFP AGMs, respectively, and seeded them in methylcellulose. In parallel, NG2- or NG2-cKit+ cells were sorted as controls. HPs were exclusively found in the NG2- cell fractions. Neither NG2+ cells (Fig.S2a) nor NG2+Runx1(GFP)+cells (Fig.S2b) gave rise to hematopoietic cell colonies in vitro (TableS5). To further assess whether NG2+ cells are hematopoietic precursors, we crossed NG2-Cre mice with a knock-in reporter mouse line in which tdTomato is preceded by a transcriptional stop flanked by two loxP sites under the Rosa26 promoter. In these mice, NG2+ cells and their progeny are tdTomato+. E11 AGM-derived tdTomato+ and tdTomato- cells were sorted and seeded in methylcellulose. HPs were only found in the tdTomato- cell fraction (Fig.S2c, TableS5) reinforcing the observation that NG2+ cells and their progeny do not contribute to hematopoietic lineages at this stage. Flow cytometric analysis confirmed the presence of tdTomato in a subset of NG2+ cells in the E11 AGM (Fig.S2d), validating our mouse model, while no overlap was found between tdTomato and CD45, a hematopoietic cell marker (Fig.S2e). We next performed immunohistochemistry on NG2-Cre;tdTomatofl/+ frozen sections and confirmed the expression of tdTomato in a subset of SMA+ cells (Fig.S2f) in the E11 AGM. CD31+ cells did not express tdTomato (Fig.S2g). Further analysis revealed that cells expressing hematopoietic markers F4/80 and CD45 do not co-express NG2 nor SMA (Fig.S2h). Together, these data indicate that NG2+ cells do not contribute to the AGM HSPC pool and suggest that NG2+Runx1+ PC/vSMCs act as a supportive niche to maintain hematopoietic activity in the AGM.

In the early developing embryo, HSPCs reside in other intra-embryonic and extra-embryonic hematopoietic organs such as the head, fetal liver (FL), placenta and yolk-sac (YS). Flow cytometric analysis of these organs harvested from Runx1-IRES-GFP mouse embryos also confirmed the presence of NG2+Runx1(GFP)+ cells (Fig.S3a, b). We next performed in vitro HP functional assays with cells harvested from all organs and genotypes of NG2-Cre;Runx1fl at both E10 and E11 developmental stages. No significant differences were found when comparing the total CFU-C numbers between genotypes in most organs (Fig.S3c, d, TablesS67). A significant increase of total number of CFU-C was observed in E10 AGM in both heterozygous and cKO mouse embryos (Fig.S3c). When analyzed individually, a significant increase in the number of erythroid colonies was detected in the cKO compared to WT littermate (p=0.0149) (TableS6). Likewise, a 2.8-fold increase in the number of erythroid colonies was detected in the E11 cKO head compared to the WT littermate (p=0.01), while the total number of CFU-C in the E11 head remained unchanged (TableS7). Moreover, we found a significant decrease in both CFU-GM (p=0.016) and CFU-GEMM (p=0.039), between WT and cKO YS (TableS7), possibly due to the defect found in the E11 AGM.

To test whether HSC activity increases in the FL due to the possible migration of AGM HSCs, E11 FL cells from all genotypes were transplanted into sub-lethally irradiated recipient mice. Neither the donor chimerism nor the percentage of reconstituted mice by donor cells showed changes between the groups (Fig.S3e). Compared to NG2+/+;Runx1fl/+ WT littermates, in which 70% of recipients (7 out of 10) were reconstituted, mice injected with NG2-Cre:Runx1fl/+ heterozygous or NG2-Cre:Runx1fl/fl cKO E11 FL showed similar reconstitution over the long term, with 67% (2 out of 3, p=0.348) and 60% (3 out of 5, p=0.421) reconstituted mice, respectively (Fig.S3e, TablesS34). Since the deletion of Runx1 in NG2+ cells only affects HSPCs in the AGM, immunohistochemistry on WT embryonic head and placenta was performed to localise NG2+Runx1+ cells. The rare NG2+Runx1+ double positive cells identified did not seem to be perivascular (Fig.S3f, stars). In line with this observation, we found that Runx1 and SMA do not overlap when NG2 and SMA were expressed in PC/vSMCs (Fig.S3f, arrowheads). Instead, the head contains few NG2+SMA- that are F4/80+, suggesting that NG2+Runx1+ cells are macrophages (Fig.S3f, arrowhead). Overall, our data shows that the deletion of Runx1 in NG2+ cells only affects selective HSPC subsets in non-AGM hematopoietic organs in the E11 mouse embryo.

To better understand the role of Runx1 in the HSC-generating microenvironment, single-cell RNA-sequencing (scRNA-seq) on NG2+/+;Runx1fl/+ E11 AGM was performed. We used graph-based clustering and known marker distribution to define and investigate the gene expression profiles of various populations that reside in the E11 AGM and identified eight populations of interest (Fig.2a, b). The co-expression of Cspg4 (NG2) and Acta2 (SMA) in the PC/vSMC population was confirmed (Fig.2c). This population is also enriched in Rgs5, Pdgfrb and Pdgfra in line with our previous work28, and a subset of these cells express Runx1 (Fig.2c, d), confirming our imaging and flow cytometric analysis. The expression of Mcam (CD146 or S-ENDO1), a pericyte/vSMC precursor marker recently identified in a subset of NG2+ cells in the E11 AGM21 and upregulated in AGM hematopoiesis supportive stromal cell lines19, was detected in a subset of PC/vSMCs, partially overlapping with Runx1+ cells (Fig.2c, d). However, Mcam was mainly enriched in endothelial cells (ECs) and also in subpopulations of hemogenic endothelial cells, including those entering endothelial-to-hematopoietic transition (HEC/EHT), IAHCs and SNS cells (Fig.2d), confirming published work including ours28,33. Immunostainings with CD146 and CD31 on E11 WT AGM frozen sections further validated our sequencing analysis at the protein level: both CD31+ endothelial cells (Fig.2e, f, arrows) and SMA+ PC/vSMCs (Fig.2f, stars) are CD146+. Importantly, Pecam-1 (CD31) expression in PC/vSMCs was low/negative in our scRNA-seq data (Fig.2c, d), in line with our immunohistochemistry, confocal imaging, and our recent published work28. Other genes expressed by hematopoietic and hemogenic/endothelial cells such as Adgre1 (F4/80), Mrc1 (CD206), Cdh5 (VE CADHERIN), Tek (TIE-2), CD34, CD93, Pdgfb, Sox7, Sox17, Sox18, Gfi1b, and Itga2b (CD41) were not expressed in PC/vSMCs (Fig.2d). These genes were used to distinguish populations of macrophages (MPs), IAHCs, HEC/EHT, and ECs (Fig.2ad). Erythroid cellsand erythroid progenitors (Ery/EryP; Gypa/CD235+), SNS (Gata3+) and skeletal muscle progenitors (SkMP; MyoD1+, Cdh15+) were also identified (Fig.2ad). Kit was expressed in all IAHCs and in a subset of PC/vSMCs, while Kit expression in HEC/EHTs was low (Fig.2d). Altogether, these data show that we successfully captured multiple cell types that comprise the E11 AGM, including a population of Runx1+ PC/vSMCs which constitutes 19.7% of all NG2+Acta2+ PC/vSMCs cells. Furthermore, the transcriptome of Cspg4+Runx1+ non-hematopoietic non-endothelial PC/vSMCs was found to partially overlap with that of the Cspg4+Mcam+ PC/vSMC precursors previously described21.

a t-SNE plot highlighting eight populations of interest identified in the E11 WT AGM. Each dot represents one cell and colours represent cell clusters as indicated. The number of cells in each population is shown in brackets. MP (macrophages); Ery/EryP (erythroid/progenitors); IAHC (intra-aortic hematopoietic clusters); HEC/EHT (hemogenic endothelial cells including those that enter endothelial-to-hematopoietic transition); EC (endothelial cells); SNS (sympathetic nervous system); SkMP (skeletal muscle progenitors), PC/vSMC (pericytes/vascular smooth muscle cells, NG2+Acta2+). Other cells (OC) are coloured in grey. b t-SNE plot highlighting the eight populations identified after excluding all other (grey) cells. c Zoom into PC/vSMC cluster (black rectangle) further show the presence or the absence of selected genes that characterise this population and confirms the presence of Runx1 in a subset of cells. d Violin plots showing distribution of expression for selected genes that contributed to the identification of cell clusters. Immunohistochemistry on frozen E11 WT sections stained with eCD146/CD31/DAPI and fCD146/SMA/DAPI, n=2 samples tested, N=2 independent experiments. Arrows: vascular cells, asterisks: perivascular cells. DA: dorsal aorta, CV: cardinal veins, NC: notochord. Source data for e (first column,20X) is provided as a Source Data file.

Our scRNA-seq analysis revealed that not all Cspg4+Runx1+ cells in the E11 AGM express Acta2 (Fig.3a, b). We therefore investigated if Cspg4+Runx1+Acta2 cells are PC/vSMCs which had not yet acquired Acta2 expression. Differential expression analysis of Acta2+ versus Acta2 cells within the NG2+Runx1+ cell population in the WT AGM revealed that markers of sclerotome-derived vSMCs such as Sox9, Pax1, Pax9 and Col2a134 are among the highest upregulated genes in Cspg4+Runx1+Acta2 cells (Fig.3c). In contrast, Cspg4+Runx1+Acta2+ cells are enriched in genes that identify more mature pericytes such as Acta2, CD248, Mcam, Rgs5 or Pdgfrb (Fig.3c), some of which are potential Runx1 target genes (star). Pdgfra and Ptn genes were recently associated with Runx1+ subaortic (non-smooth muscle) mesenchymal cells with possible role in hematopoiesis in the E10.5 AGM of the mouse embryo35. Our scRNA-seq analysis show that, in E11 AGM, Pdgfra and Ptn are also expressed in Cspg4+Runx1+ cells with no significant difference between Acta2+ and Acta2 (Fig.3c). Further analysis showed that Gene Ontology (GO) biological processes significantly enriched in Cspg4+Runx1+Acta2+ cells include smooth muscle cell chemotaxis and migration, collagen-activated signalling pathway, neural crest cell differentiation and regulation of BMP signalling (Fig.3d), previously shown by our laboratory to control in vivo HSPC generation in the mouse AGM28. In Cspg4+Runx1+Acta2 cells, significantly enriched GO biological processes include mesenchymal stem cell differentiation and cartilage and bone development (Fig.3e), consistent with the sclerotome origin of these cells.

a t-SNE plots showing the distribution of Runx1 and Acta2 expression in NG2+Runx1+ cells in the WT E11 AGM after excluding all other (grey) cells found in the Fig.2a. b Zoom into NG2+Runx1+ cluster (black rectangle) shows the presence or the absence of Acta2. c Heatmap showing the expression of Cspg4 and Runx1 and 15 selected genes out of 25 top significantly upregulated genes in WT NG2+Runx1+Acta2+ cells (upper half) and NG2+Runx1+Acta2- cells (bottom half) at single cell level; *Runx1 potential target genes. Pdgfra and Ptn genes were next added to inform their expression in both populations. Barplot of fold enrichment for selected GO biological processes significantly overrepresented in genes significantly upregulated in both dWT NG2+Runx1+Acta2+ and eNG2+Runx1+Acta2- cells. f t-SNE of WT E11 AGM cells, overlaid with principal pseudotime curve inferred by Slingshot, predicting a lineage from NG2+Runx1+Acta2- cells to NG2+Runx1+Acta2+ cells. g WT NG2+Runx1+cells arranged in pseudotime (x-axis) based on the inferred curve. Y-axis represents log normalised gene expression.

Indeed, PC/vSMCs in the AGM have been shown to originate from the sclerotome and display markers of this compartment at least during the early phases of mural cell recruitment36. A recent study showed that the maturation of sclerotome-derived vSMCs in the mouse AGM depends on a transcriptional switch from a sclerotome signature with the repression of Pax1, Scx and Sox9, and activation of Acta2 and other vSMC genes34.

To test whether NG2+Runx1+Acta2 cells follow a maturation trajectory towards Cspg4+Runx1+Acta2+ vSMCs, we performed cell lineage inference with Slingshot, a trajectory inference method for scRNA-seq data that can incorporate knowledge of developmental markers. Having defined a cluster of Cspg4+Runx1+Acta2 cells as an origin, Slingshot infers a cell lineage and constructs a pseudotime curve representing that lineage (Fig.3f, arrow). Gene expression along pseudotime shows that sclerotome markers such as Sox9, Pax1 and Pax9 are gradually downregulated while markers of mural cells such as Acta2, Rgs5, Pdgfr, Cnn1, Mcam and CD248, are gradually upregulated in an inferred transition from Cspg4+Runx1+Acta2 to Cspg4+Runx1+Acta2+ cells (Fig.3g). Our scRNA-seq analysis shows that Cspg4+Runx1+ AGM cells display a sclerotome-derived vSMC transcriptomic profile.

We next explored the impact of Runx1 deletion in the hematopoietic niche and its possible effect on PC/vSMCs by performing scRNA-seq of NG2-Cre;Runx1fl/fl cKO E11 AGMs (Fig.4a, b). Cell populations were defined in a similar way to the WT AGM by using graph-based clustering and known marker distribution. This comparison revealed changes in the proportions of the different cell types between WT and cKO AGM, including a significant reduction in the proportion of cells associated with clusters 2 (Ery/EryP), 3 (IAHC), 6 (SNS), and 7 (SkMP) (Fig.4c).

a t-SNE plot showing eight populations of interest found in the E11 cKO AGM. Each dot represents one cell and colours represent cell clusters as indicated. MP(macrophages); Ery/EryP(erythroid/progenitors); IAHC (intra-aortic hematopoietic clusters); HEC/EHT (hemogenic endothelial cells including those that enterendothelial-to-hematopoietic transition),EC (endothelial cells); SNS (sympathetic nervous system); SkMP(skeletal muscle progenitors); PC/vSMC (pericytes/vascular smooth muscle cells, NG2+Acta2+). Other cells (OC)are coloured in grey. The number of cells in each cluster is shown in brackets. b t-SNE plot highlighting the eight populations identified after excluding all other (grey) cells. c Percentage of single live cells found in each E11 AGM sample (cell number/total cells) defined by scRNA-seq in WT (full bars) and cKO (empty bars) AGMs. Colours and numbers correspond to each population defined in a; chi-squared two-tailed test was used for comparison. d Barplot of fold enrichment for selected GO biological processes significantly overrepresented in genes significantly downregulated in cKO PC/vSMCs compared to their WT counterparts. Heatmap of ligand-receptor interactions inferred by NicheNet from e WT and f cKO E11 AGM cells. Colour represents the interaction potential score between the 10 top-ranked ligands expressed in ECs and their inferred targets expressed in PC/vSMCs. Ligands and receptors are ordered by hierarchical clustering. g Scatter plots of AUC vs log10(FDR) showing downregulated genes associated with selected GO terms in cKO PC/vSMCs. Red dots represent significantly downregulated genes (FDR<0.05); dashed line shows FDR = 0.05. Gene labels with red borders represent potential Runx1 target genes.

Changes in gene expression between WT and cKO Cspg4+Runx1+ cells were first investigated. We found that genes significantly downregulated in cKO Cspg4+Runx1+ cells were mainly associated with biological processes including translation, oxidative phosphorylation, cellular response to stress and mitochondria-related function (Fig.S4ac). As deletion of Runx1 may have also affected Runx1 PC/vSMCs, the transcriptome of all cKO Cspg4+Acta2+ PC/vSMCs with their WT counterpart was compared. Genes significantly downregulated in cKO Cspg4+Acta2+ PC/vSMCs had significant enrichment of biological processes including translation, smooth muscle cell differentiation, cytoskeleton, vasculogenesis and cell communication (Fig.4d). One pathway essential to vasculogenesis is PDGF-B/PDGFR; we therefore applied NicheNet on our WT scRNA-seq data to predict ligand-receptor interaction between ECs and PC/vSMCs, focusing on PDGF-B-related genes. The highest scoring predicted interaction was between Pdgfb, a growth factor released by ECs, and Nrp1 (Fig.4e), a receptor known to control the differentiation/recruitment of mesenchymal stem cells and the stimulation of smooth muscle cell migration37,38.

The interaction between Pdgfb and Pdgfrb was also amongst the highest scoring interactions in both WT (Fig.4e) and cKO (Fig.4f). Additional interactions involve Edn, Tgfb or Bmp pathways, previously associated with a role in AGM hematopoiesis39,40. Interestingly, in cKO ECs, Pdgfa, another gene from the PDGF family, was no longer in the top 10 ranking ligands (Fig.4f) possibly due to the downregulation of Pdgfra in cKO PC/vSMCs (Fig.4f). Other genes including Des, Angpt1, Gsk3b, Tcf21, Col1a1, Pcna, Ccnd3 and Mcm7, potential Runx1 downstream target genes41, were also significantly downregulated (Fig.4g, red boxes). The reduction inCol1a1 expression suggests changes in the gene profile of the extracellular matrix (ECM). Indeed, additional ECM related genes were significantly downregulated in the cKO PC/vSMCs, such as Sparcl1, Col3a1 and Col5a1 (Fig.4g). Collectively, these data show that the genetic programme of PC/vSMCs in cKO AGM is modified upon Runx1 deletion and this involves changes in molecules that constitute the ECM of the aortic wall.

Endothelial cells share the same basement membrane with PC/vSMCs42. This, coupled with the transcriptomic changes in the cKO PC/vSMCs described above, suggest that the genetic programme of the adjacent ECs may have also been altered by Runx1 deficiency in NG2+ cells. Although the number of endothelial cells in the NG2-Cre:Runx1fl/fl cKO did not significantly change (Fig.4a) and the formation of the dorsal aorta appeared to be unaffected (Fig.1e), we investigated transcriptomic changes in ECs that could affect their function in vivo. As before, we performed differential expression analysis, followed by overrepresentation analysis on genes significantly downregulated in cKO ECs (Fig.5a). Multiple GO biological processes were significantly overrepresented in these genes, with many related to EC development and angiogenesis; proliferation, migration and differentiation; response to hypoxia and fluid shear stress; as well as smooth muscle cell or mesenchymal cell development and hematopoiesis (Fig.5a). Interestingly, we found that Sox18 was the most downregulated gene in cKO ECs (Fig.5b). Col4a1, the most abundant extracellular matrix associated gene, known to co-localise with Sox18 in ECs in the mouse embryo43, was also found within the top 25 downregulated genes (Fig.5b). Sox18 and Col4a1 were the most downregulated genes associated with the blood vessel development GO term, while other gene expression including Cdh5, Pecam1, Sox17, Pdgfb, MCam and Notch were also affected.

a Barplot of fold enrichment for selected GO biological processes significantly overrepresented in genes significantly downregulated in cKO ECs compared to their WT counterparts. b Scatter plots of AUC vs log10(FDR) showing downregulated genes associated with selected GO terms including blood vessel development and mesenchymal cells and vSMCs in cKO ECs. Red dots represent significantly downregulated genes (FDR<0.05); dashed line shows FDR=0.05. Sox18 and Ctnnb1 expression in WT ECs in both scRNA-seq (c, EC zoom and t-SNE plots) and bulk RNA-seq post-sort (d, TPM). e Scatter plots of AUC vs log10(FDR) showing downregulated genes associated with selected GO terms including the basement membrane and extracellular matrix in cKO ECs. Red dots represent significantly downregulated genes (FDR<0.05); dashed line shows FDR=0.05. Selected genes that were altered in cKO ECs ine are shown in WT ECs in both scRNA-seq (f, ECand HEC/EHT zoom and t-SNE plots) and bulk RNA-seq post-sort (g, TPM). TPM: transcript per Million mapped reads values.

Genes associated with cell adhesion, regulation of smooth muscle cell proliferation and differentiation, along with mesenchyme development such as Sox18 and Ctnnb1 were also significantly downregulated in cKO EC (Fig.5b, arrow). We confirmed that both Sox18 and Ctnnb1 are expressed by ECs in our single cell datasets (Fig.5c) and next validated their expression in NG2-PDGFR-ckit-CD45-CD31+ Runx1- purified ECs from E11 Runx1-IRES-GFP AGMs (Figs.5d, S5a).

Some significantly downregulated genes associated with blood vessel development such as Loxl2, Hspg2, Col4a2, Col15a1 and Col18a1 (Fig. 5b)are also known to be associated to the ECM. Further analysis of endothelial extracellular matrix encoding genes previously described44 revealed that most of these genes were also significantly downregulated in cKO ECs (Fig.5e). The expression of these genes in WT ECs at single-cell level (Fig.5f) was confirmed post-sort at population-level (Fig.5g) with most genes being highly expressed in ECs only. One of them was Sparc (Fig.5e blue arrow, Fig.5g), a central ECM secreted Ca2+-binding glycoprotein that interacts with many other ECM proteins including Col1 and Col445,46. Among the SPARC family, Sparcl1 (Sparc-like 1), known to bind to Col147, was also found to be significantly downregulated in cKO ECs (Fig.5b, c). Together, these analyses show that Runx1 deficiency in NG2+cells leads to significant transcriptomic changes in endothelial cells including extracellular matrix related genes. We did not detect transcriptional changes in the NG2-Cre;Runx1fl/fl cKO HEC/EHT cell cluster, although this observation is inconclusive due to the low number of cells captured.

Transcriptomic changes in vascular and perivascular cells may have also affected IAHCs. As hematopoietic cells are highly heterogeneous and progenitors were significantly affected (Fig.1), we first explored WT IAHCs in more detail. Previous studies showed that IAHCs are composed of both Runx1+ and Runx1- cells28,48,49 and we were able to confirm this by flow cytometry in Runx1-IRES-GFP AGMs (Fig.S5a). We also confirmed the expression of Runx1 in HEC/EHT and its absence in ECs by flow cytometry in Runx1-IRES-GFP E11 AGMs (Fig.S5a), in line with published work50. To validate their cell identity, we next purified and sequenced 243 Runx1 (GFP)+ and 27 Runx1 (GFP)- IAHCs (NG2-PDGFR- CD31+ckit+) as well as 5822 EC (Runx1-) and 248 HEC/EHT (Runx1+) cells from NG2-PDGFR- ckit- CD45- CD31+ E11 Runx1-IRES-GFP AGMs (Fig.S5a) and performed bulk RNA sequencing (RNA-seq). The purity of the sort was first confirmed (Fig.S5b). While CD45 antibody was not used to isolate IAHCs (Fig.S5a), our bulk RNA-seq data (Fig.S5b) show that not all IAHC cells express Ptprc (CD45) in line with previous studies48,49,51, and seems to be present only when Runx1 is expressed. Next, the identity of all sorted cell populations based on the expression of genes known to be expressed in these cells15,35,48 was confirmed (Fig.S5c). Interestingly, the transcriptomic profile of Runx1 (GFP)+ and Runx1(GFP)- sorted IAHCs appears to be distinct. While CD34, Gata2, Lmo2, Etv6, and Eglf7 are expressed in both Runx1(GFP)+ and Runx1(GFP)- IAHCs at various levels, Adgrg1, Gfi1, Myb and CD44 are mainly found in Runx1(GFP)+ IAHCs (Fig.S5c). Instead, as they also express Tek, Kdr, Eng, Esam and Gata2 (Fig.S5c, d), Runx1(GFP)- IAHCs are at the transcription level, closer to type-1 pre-HSCs52,53 or to recently described CD31+ckithighGata2medium IAHCs that are Runx1-Ptprc-48 with possible (micro)-niche role54. Our analyses confirm the heterogeneity of Runx1(GFP)+/-CD31+C-KIT+IAHCs in the E11 AGM at both protein and transcriptomic levels, and indicate that most IAHC cells captured in our full/unsorted AGM scRNA-seq are Runx1-.

To explore transcriptomic changes between WT and cKO IAHCs, differential expression analysis followed by overrepresentation analysis on genes significantly downregulated in cKO IAHCs was carried out (Fig.6a). Several GO biological processes were significantly overrepresented in these downregulated genes, including ribosome assembly processes, regulation of translation, RNA transport and localisation, and others such as response to DNA damage, gene expression and cellular processes (Fig.6a). In line with this, we found that the top 25 significantly downregulated genes in cKO IAHCs were mostly ribosomal protein coding genes from both Rps and Rpl families. Other genes in the top 25 are known to be required for transcriptional or translational initiation such as Btf3, Pabpc1 and Bclaf1 (Fig.6b). Interestingly, one of the top significantly downregulated genes in the cKO was Sox18 (Fig.6b, arrow), previously reported to be expressed in both IAHCs and ECs in the mouse AGM55 and confirmed here by our WT scRNA-seq data (Fig.2d). Furthermore, Sox18 has been transiently detected during early hematopoiesis in a model of embryonic stem cell differentiation in vitro, controlling early HP proliferation and maturation56. In line with this, further GO analysis revealed that Sox18 is associated with cellular processes including cell maturation, cell differentiation and regulation of stem cell proliferation (Fig.6b). The latter two GO terms are also associated with other genes significantly downregulated in cKO IAHCs such as Hmgb2, encoding a chromatin-associated non-histone protein involved in transcription and chromatin remodelling (Fig.6b). This transcriptomic analysis shows that the deletion of Runx1 in NG2+ PC/vSMCs within the AGM niche significantly alters the genetic programme of IAHCs.

a Barplot of fold enrichment for selected GO biological processes significantly overrepresented in genes significantly downregulated in cKO HSPCs compared to WT HSPCs. b Scatter plot of AUC (representing strength of downregulation) vs log10(FDR), showing the top 25 significantly downregulated genes (red circles) in cKO HSPCs. Scatter plots of AUC vs log10(FDR) highlighting downregulated genes associated with Gene Ontology (GO) biological processes. Red dots found above the dashed line (corresponding to FDR=0.05) represent significantly downregulated genes (FDR<0.05).

Despite the decrease in HPs and HSCs in cKO AGM, NG2-Cre;Runx1fl/fl mice are born with no obvious defects and develop into adulthood. Because of this, we sought to explore the effect of Runx1 deletion in NG2+ PC/vSMC on adult HSPCs. The presence of these progenitors in the adult bone marrow (BM) of mutant mice was analyzed by flow cytometry and compared to WT mice. No significant differences were found in either Lin-Sca1+cKit+ (LSK) (Fig.7a, b) nor LSK CD150+CD48-(SLAM) cell frequencies (Fig.7c, d) between cKO mice and WT controls. We performed HP assays and found that the frequencies of hematopoietic cell colonies were similar in all mutants and WT littermates (Fig.7e, TableS8). To assess the capacity of these cells to reconstitute hematopoiesis in vivo, 5105 bone marrow cells harvested from all genotypes were transplanted into sub-lethally irradiated WT mice recipients. Compared to the control group in which 62.1% (18 out of 29) mice were reconstituted, mice injected with NG2-Cre;Runx1fl/+ or NG2-Cre;Runx1fl/fl BM cells showed a significant reduction in the long-term reconstitution potential, with only 27.3% (3 out of 11, p=0.024) and 20% (4 out of 20, p=0.002) of transplanted mice being reconstituted respectively (Fig.7f, TableS3). In addition, the percentage of donor chimerism was significantly reduced in the cKO group. On average, the donor chimerism with WT cells was 33% compared to the 16% and 9% observed when BM cells from NG2-Cre;Runx1fl/+heterozygous and NG2-Cre;Runx1fl/fl cKO (p=0.002) were injected respectively (Fig.7f, TableS4). The remaining HSCs in the mutant NG2-Cre;Runx1fl/+ and NG2-Cre;Runx1fl/fl adult BM are multilineage, showing similar contributions of donor cells to myeloid or lymphoid cell compartments (Fig.7g), and self-renew (Fig.7h). Interestingly, no NG2+Runx1(GFP)+ cells were detected in adult Runx1-IRES-GFP BM hematopoietic niches (Fig.7i), suggesting that they are exclusive to the embryo and that the BM hematopoietic defect found in adults is developmentally driven.

a, bRepresentative plots and percentages of Lin-Sca1+cKit+ (LSK) and c, dLSK CD150+CD48-(SLAM) bone marrow (BM) cells by flow cytometry of WT/ NG2+/+;Runx1fl/+,NG2+/+;Runx1fl/fl (n=9), HET NG2-Cre;Runx1fl/+ (n=4) and cKO NG2-Cre;Runx1fl/fl (n=4) adult mice is shown. e Colony-forming unit-culture (CFU-C) numbers per 104 adult BM cells; n=WT/HET/cKO: 13/7/8 mice. N=7 independent experiments. Data are meanSEM (TableS8). f Hematopoietic stem cell repopulating potential and donor chimerism of WT and mutant BM cells in vivo. 5105 BM donor WT, HET and cKO cells were injected into 29, 11 and 20 Ly5.1 HET recipients, respectively, with 18, 3 and 4 found to be reconstituted respectively (Table S3, p=0.024 (WT/HET) and p=0.002 (WT/cKO) by Z score test for 2 population proportions). Mice are reconstituted when 5% donor cells are found in the host peripheral blood; p=0.002 (WT/cKO) by Kruskal-Wallis and Dunns post-hoc test (TableS4). g Histograms showing the contribution of CD45.2+CD45.1- donor cells to myeloid cells (CD11b+Gr1+/-), B cells (CD19+) and T cells (CD4/8+) in all reconstituted host mice from (f). (n=WT/HET/cKO=18/3/4), p=0.019 (WT/HET) for B cells by one-way ANOVA and Tukeys post-hoc test. h BM cells from selected reconstituted primary recipients (found in f) were transplanted into multiple irradiated secondary recipients. Mice are reconstituted when 5% donor cells are found in the host peripheral blood (TableS34). i Representative flow cytometric analysis plot of NG2 in Runx1-IRES-GFP adult BM (n=6). All data are presented as Mean values+/-SEM. N=number of independent experiments; n = number of biological samples. Source data for b, d, e, f, g and h are provided as a Source Data file.

Follow this link:
Runx1+ vascular smooth muscle cells are essential for hematopoietic stem and progenitor cell development in vivo - Nature.com