Runx1+ vascular smooth muscle cells are essential for hematopoietic stem and progenitor cell development in vivo

Hematopoietic stem cells (HSCs) produce all essential cellular components of the blood. Stromal cell lines supporting HSCs follow a vascular smooth muscle cell (vSMC) differentiation pathway, suggesting that some hematopoiesis-supporting cells originate from vSMC precursors. These pericyte-like precursors were recently identified in the aorta-gonad-mesonephros (AGM) region; however, their role in the hematopoietic development in vivo remains unknown. Here, we identify a subpopulation of NG2+Runx1+ perivascular cells that display a sclerotome-derived vSMC transcriptomic profile. We show that deleting Runx1 in NG2+ cells impairs the hematopoietic development in vivo and causes transcriptional changes in pericytes/vSMCs, endothelial cells and hematopoietic cells in the murine AGM. Importantly, this deletion leads also to a significant reduction of HSC reconstitution potential in the bone marrow in vivo. This defect is developmental, as NG2+Runx1+ cells were not detected in the adult bone marrow, demonstrating the existence of a specialised pericyte population in the HSC-generating niche, unique to the embryo.

ventral tissue 10 , immune system 11 and neural crest cells 12 , mesenchymal stromal cells have been shown to regulate HSC emergence and/or HSC maintenance [13][14][15][16][17] .Some of these mesenchymal cells support HSCs in vitro better than others 18 suggesting that cultured mesenchymal cell lines are heterogeneous 19 .
A previous study found that hematopoietic supportive mesenchymal stromal cells follow a vascular smooth muscle cell (vSMC) differentiation pathway in culture 20 .Indeed, alongside mesenchymal markers found in E11 AGM stromal cell lines, markers that identify early/intermediate vSMC lineage such as αSMA and SMactinin have also been described 20 .These findings strongly suggest that at least some hematopoietic supportive cultured cells originate from vSMC precursors in vivo.Roostalu et al. recently identified such vSMC precursors that resemble pericytes (PC) in the AGM.These cells express NG2 and CD146 and their emergence in the AGM at E10.5 coincides with the timeframe when HSCs are first generated 21 .However, whether PC/vSMCs play a role in HSC emergence in vivo has not yet been documented.
In vivo, AGM PC/vSMCs are highly heterogeneous.Previous work showed that a subset of sub-aortic mesenchymal cells express Runx1 [22][23][24] , a key transcription factor required for HSC generation in the mouse embryo [24][25][26] .It is not known whether Runx1 in PC/vSMCs plays a role in HSC development.
In this work, we hypothesised that PC/vSMCs expressing Runx1 are involved in AGM hematopoietic stem/progenitor cell (HSPC) development in vivo.Using single-cell RNA-sequencing (scRNA-seq), and 3-dimensional imaging in transgenic mice, we show a perivascular and hematopoietic functional and developmental relationship in the AGM at both cellular and molecular levels.

A subpopulation of subaortic mesenchyme in the AGM coexpresses 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 αSMA with CD31, an endothelial and HSPC marker (Table S1).Imaging analysis showed that NG2 + αSMA + CD31 - vSMCs surround NG2 -αSMA -CD31 + endothelial cells (Figs. 1a, S1a, b), confirming previous reports 27 .Further to its expression in hematopoietic and hemogenic endothelial cells, Runx1 was also detected in the sub-aortic mesenchyme 22,23 .Therefore, we hypothesised that at least some of these cells also express NG2.We first confirmed that both intraaortic 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 study 28 that some cells around the notochord express NG2 in the trunk (Fig. S1a, circle).However, these perinotochord cells do not express αSMA, CD31 (Fig. S1a, circle) nor Runx1 (Fig. S1e-f).To confirm the presence of NG2 + Runx1 + cells in the E11 AGM, we used Runx1-IRES-GFP mouse embryos 29 .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 cells 30 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.

Runx1 deficiency in NG2 + cells impairs AGM hematopoiesis
Runx1 deletion in endothelial cells impairs HSC emergence in the AGM [24][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;Runx1 fl/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 maintenance 31,32 .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, αSMA and NG2 (Figs. 1e, f, S1c, d) in the AGM.In contrast, αSMA + Runx1 + PC/vSMCs with low expression of Runx1 were reduced in the cKO dorsal aorta compared to WT littermate controls (Figs.1f, S1e-g).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, Table S2) and E11 (Fig. 1j, Table S2).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, Tables S3-4).These findings indicate that the absence of Runx1 in aortic NG2 + cells impairs HSC generation and/ or maintenance and HP development in the AGM.
NG2 + Runx1 + cells do not contribute to hematopoietic lineages 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 (Table S5).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, Table S5) 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;tdTomato fl/+ frozen sections and confirmed the expression of tdTomato in a subset of αSMA + 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 αSMA (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.
Deletion of Runx1 in NG2 + cells affects selective hematopoietic progenitor types in embryonic organs other than 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, Tables S6-7).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) (Table S6).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 (Table S7).Moreover, we found a significant decrease in both CFU-GM (p = 0.016) and CFU-GEMM (p = 0.039), between WT and cKO YS (Table S7), 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 +/+ ;Runx1 fl/+ WT littermates, in which 70% of recipients (7 out of 10) were reconstituted, mice injected with NG2-Cre:Runx1 fl/+ heterozygous or NG2-Cre:Runx1 fl/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, Tables S3-4).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 αSMA do not overlap when NG2 and αSMA were expressed in PC/vSMCs (Fig. S3f, arrowheads).Instead, the head contains few NG2 + αSMA -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.
Cspg4 + Runx1 + AGM cells display a sclerotome-derived vSMC transcriptomic profile 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 Col2a1 34 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 embryo 35 .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 AGM 28 .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.
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 recruitment 36 .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 genes 34 .
The genetic programme of PC/vSMCs is altered in NG2-Cre;Runx1 fl/fl cKO E11 AGM 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;Runx1 fl/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).
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 hematopoiesis 39,40 .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 genes 41 , were also significantly downregulated (Fig. 4g, red boxes).The reduction in Col1a1 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.
Deletion of Runx1 in NG2 + cells leads to changes in the endothelial cell gene profile Endothelial cells share the same basement membrane with PC/ vSMCs 42 .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:Runx1 fl/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 embryo 43 , 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.
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 -PDGFRβ -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 described 44 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 Ca 2+ -binding glycoprotein that interacts with many other ECM proteins including Col1 and Col4 45,46 .Among the SPARC family, Sparcl1 (Sparclike 1), known to bind to Col1 47 , 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;Runx1 fl/fl cKO HEC/EHT cell cluster, although this observation is inconclusive due to the low number of cells captured.

The genetic programme of cKO IAHCs is different from their WT counterpart
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 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.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.cKO was Sox18 (Fig. 6b, arrow), previously reported to be expressed in both IAHCs and ECs in the mouse AGM 55 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 maturation 56 .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 chromatinassociated 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.The reconstitution potential of bone marrow cKO HSCs is significantly reduced Despite the decrease in HPs and HSCs in cKO AGM, NG2-Cre;Runx1 fl/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, Table S8).To assess the capacity of these cells to reconstitute hematopoiesis in vivo, 5 × 10 5 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;Runx1 fl/+ or NG2-Cre;Runx1 fl/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, Table S3).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;Runx1 fl/+ heterozygous and NG2-Cre;Runx1 fl/fl cKO (p = 0.002) were injected respectively (Fig. 7f, Table S4).The remaining HSCs in the mutant NG2-Cre;Runx1 fl/+ and NG2-Cre;Runx1 fl/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.

Discussion
In this study, we examined whether PC/vSMCs control AGM hematopoietic development in the mouse embryo in vivo.Previous work showed that pericytes support HSC maintenance in both the adult bone marrow and the fetal liver.We identified a subset of sub-aortic mesenchymal cells that co-express NG2 and Runx1 and display a sclerotome-derived vSMC transcriptomic profile.Analysis of scRNAseq data and immunostaining observations showed that a subset of these cells also express CD146, a marker of PC/vSMC precursors in the AGM, recently described 21 .Adult CD146 + PCs were shown to support human and murine hematopoiesis 57,58 .More recently, Runx1b:RFP + AGM mesenchymal cells were shown to support hematopoiesis in vitro 35 .In this study, authors performed aggregation cultures of hematopoietic cells with sorted stromal cells  from E10.5 AGM in the presence or the absence of Runx1b:RFP + cells, and found that the CFU potential was decreased in the absence of Runx1b:RFP + cells.While this shows that at this stage Runx1b:RFP + mesenchymal cells support hematopoiesis in vitro 35 , the direct identification of the precise mesenchymal cell population expressing Runx1 with a functional role in the mouse hematopoietic AGM niche in vivo had not yet been elucidated.Similarly, scRNA-seq showed that the Runx1b:RFP + mesenchymal cell population is sub-divided into high Pdgfra + and Acta2 + cell subsets 35 .Whether only one of these individual populations or both support mouse HSPC development in vivo or in vitro remains unclear.
We hypothesised that NG2 + Runx1 + PC/vSMCs control in vivo HSPC development in the mouse embryo.Indeed, conditional deletion of Runx1 in NG2 + cells significantly reduced the number of HSPCs in the AGM together with the number of reconstituted mice by donor cKO AGM cells in vivo.Importantly, neither antibody-selected NG2 + Runx1(GFP) + cells nor lineage-traced NG2-derived daughter cells in the AGM gave rise to CFUs, demonstrating a functional role for Runx1 as a positive hematopoietic niche regulator in the PC/vSMCs perivascular layer (Fig. S6).
We found that the number of mice that reconstituted hematopoiesis upon transplantation with NG2-Cre;Runx1 fl/fl cKO BM cells and the level of donor chimerism was significantly decreased.Interestingly, the cKO mice are born with no apparent hematopoietic defect.This is not a surprise.Yokomizo et al. recently showed that the embryonic hematopoietic system can be maintained with minimal involvement of HSCs 59 .Furthermore, Dignum et al., demonstrated that in vitro, multipotent progenitors and HSCs arise independently from hemogenic endothelial cells in the mouse embryo 60 .In line with these studies, we showed that the HP numbers in the placenta, FL and adult BM are not affected in NG2-Cre;Runx1 fl/fl cKO mice.This suggests that an HSC-independent production of these progenitors is sufficient for the mice to survive in addition to the remaining cKO BM HSCs that are multilineage and self-renew.Furthermore, this also suggests that the development of HSCindependent progenitors does not require NG2 + Runx1 + niche cells.However, we cannot rule out that the presence of HSCs in the cKO BM may be due to the low Cre recombination efficiency of the NG2-Cre mice used, as in average, only 20% of NG2 + cells expressed tdTomato in the AGM.S8).f Hematopoietic stem cell repopulating potential and donor chimerism of WT and mutant BM cells in vivo.5 × 10 5 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 Dunn's post-hoc test (Table S4).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 Tukey's 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 (Table S3-4).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.
We next discovered that changes in the in vivo PC/vSMC genetic programme, modified both IAHC and EC transcriptomic profiles in the E11 AGM.Although we did not capture all IAHCs by scRNA-seq, we compared the transcriptional profile of mainly Runx1 -IAHCs between cKO E11 AGMs and its WT littermate controls.We found that one of the top 25 significantly downregulated genes in the cKO IAHCs was Sox18.Ectopic expression of SOX18 was previously linked to the self-renewal potential of HPs in the embryonic system in vitro 56 , suggesting a role for NG2 + Runx1 + niche cells in HSPC maintenance.
In addition to SOX18, two other SOX transcription family members, SOX7 and SOX17, were reported to be involved in both blood vessel development and hematopoiesis 61 .We investigated their expression in our datasets and found that all three transcription factors were highly expressed in IAHCs and ECs as previously shown 43,55,56 .In ECs, Sox17 is known to play a role in the maintenance of clusters containing HSCs 55 .We discovered that in the NG2-Cre;Runx1 fl/fl cKO AGM, both Sox18 and Sox17 were significantly downregulated in ECs, specifically driving the enrichment of blood vessel development-related processes.Indeed, we observed a significant reduction of IAHC in the NG2-Cre;Runx1 fl/fl cKO AGM alluding to the importance of the fine regulation that PC/vSMCs and ECs play in controlling one another as well as HSPCs.Together with changes in PC/vSMCs affecting ECs, an opposite effect was also observed.We found that genes associated with the regulation of smooth muscle cell proliferation and differentiation, or mesenchymal development were significantly downregulated in cKO ECs.One of these genes, Pdgfb, detected in a significantly smaller proportion of cKO ECs (42% in cKO compared to 59% in the WT, p = 0.0001) but also with significantly lower expression (FDR = 0.02), was shown to mediate the vSMC phenotype switch from a contractile to a synthetic/proliferative state, by regulating Sox9 62 or Yap1 63 gene expression.In our cKO scRNA-seq analysis, Yap1 was downregulated in both ECs and PC/vSMCs, although the difference in PC/vSMCs was not significant.However, major proliferation-associated genes were significantly downregulated in the cKO PC/vSMCs, suggesting a premature switch of these cells towards a more mature phenotype 21 .Whether PC/vSMC proliferation was exclusively due to Pdgfb decrease is not clear.Some of these proliferation-associated genes significantly downregulated in PC/vSMCs, such as Pcna, Ccnd7 and Mcm7, are known downstream target genes of Runx1 and thus, changes in the proliferation status of PC/vSMCs may be intrinsic.The upregulation of genes that characterise a more mature/contractile smooth muscle cell phenotype such as Cnn1 or Tagln, in cKO PC/vSMCs, was not significant.This could be due to the presence of CD146 in these cells, known to prevent vSMC differentiation 21 , and expression of which remained unchanged in the cKO PC/vSMCs.
Our scRNA-seq analysis also revealed that Sparc is significantly downregulated in cKO ECs in the E11 AGM and we found that many collagen genes, known to interact with Sparc, were downregulated in ECs and/or PC/vSMC.Reduction in collagen concentration and compromised collagen fibril formation has been found in SPARC-KO mice, suggesting a critical role for SPARC in the ECM assembly 64 .In addition, the depletion of SPARC in flies results in the absence of Col4 in the basal lamina during development, leading to embryonic death 65 .Interestingly, our ongoing preliminary data suggests that SPARC plays a role in HSC maintenance or maturation ex vivo.When added in explant culture, SPARC is able to rescue the NG2Cre;Runx1 fl/fl cKO AGM hematopoietic defect upon transplantation into irradiated recipient mice (data not shown).
In conclusion, we have described and characterised the regulatory potential of NG2 + Runx1 + perivascular cells in the HSC generating niche in vivo.Our study shows the crucial role of pericytes/vSMCs in maintaining the AGM niche, and thus the regulation of HSC generation.These findings offer a deeper molecular understanding of the complex regulation of the AGM niche providing novel and important observations that may lead to ways to resolve the complexity of HSC generation.

Animals and embryo generation
All experiments were performed under a Project Licence granted by the Home Office (UK), approved by the University of Edinburgh Ethical Review Committee and conducted in accordance with local guidelines.Mice were bred and housed at the Centre for Regenerative Medicine, Edinburgh, UK under a 12-h light/12-h dark cycle mimicking circadian rhythm and fed a chow diet and water ad libitum.NG2Cre transgenic mice (Jackson Laboratory, B6.Cg-Tg(Cspg4-cre/Esr1*)BAkik/J) were backcrossed into the C57BL/6J background and maintained as heterozygous.Runx1 fl/fl mice (from Professor Nancy Speck, available in Jackson Laboratory), R26-TdTomato mice (Jackson Laboratory, Gt(ROSA)26Sor tm14(CAG-tdTomato)Hze ) and Runx1-IRES-GFP mice (from James Downing, Runx1 tm4Dow ) are on a C57BL/6J background and were maintained as homozygous.C57BL/6J (Ly5.2) WT mice (CD45.2+CD45.1-)and Ly5.1 homozygous (CD45.2-CD45.1+)and heterozygous (CD45.2+CD45.1+)inbred mice were provided by our animal facilities.All mice used in this study were males and females mixed between 3 and 6 months of age.The day of vaginal plug detection was designated as developmental day (E)0.5.

Genotyping
DNA was extracted from the remaining embryonic organs that were not used in experiments.Tissue was placed in a solution containing 100 µl of Extraction solution (Sigma, E7526) and 25 µl of Tissue preparation solution (Sigma, T3073).Samples were incubated in a shaking block at 55 °C for 10 min, followed by DNA denaturation for 3 min at 95 °C.After denaturation, 100 µl of Neutralisation solution B (Sigma, N3910) was added to each sample.Primers per mouse line were used as follow: NG2Cre: mNG2cre forward: GCG GTC TGG CAG TAA AAA CTA TC, mNG2Cre reverse: GTG AA ACA GCA TTG CTG TCA CTT; Runx1fl: mEx4_intF563_F: CCC ACT GTG TGC ATT CCA GAT TGG, mEx4_R837_R: GAC GGT GAT GGT CAG AGT GAA GC and CAC CAT AGC TTC TGG GTG CAG, Runx1-IRES-GFP: AML1-GFP USA F: GTC CAG GAG CGC ACC ATC TTC TTC, AML1-GFP USA R: GTA CAG CTC GTC CAT GCC GAG AGT; R26-TdTomato: ROSA WT_F: AAG GGA GCT GCA GTG GAG TA, ROSA WT_R: CCG AAA ATC TGT GGG AAG TC, tdTomato_F: CTG TTC CTG TAC GGC ATG G, WPRE_R: GGC ATT AAA GCA GCG TAT CC.
In vitro Colony Forming Unit-Culture assays (CFU-C) Except for the fetal liver (FL) which was only mechanically disrupted into single-cell suspension, other embryonic organs were dissected and dissociated with collagenase type I (Sigma, C0130, 0.12% v/v) in PBS for 45 min (AGM, head) and 75 min (placenta, yolk-sac/YS) at 37 °C, mechanically dissociated into single-cell suspensions and washed with 10% FCS in PBS.Adult bone marrow (BM) cells were flushed from wild type, heterozygous and knock-out femur/tibia bones and lysed with ammonium chloride solution (Stem Cell Technologies 07850) for 12 min at room temperature (RT).AGM (1 embryo equivalent (ee)/dish for E10 and E10.5 and 0.33ee/dish for E11), FL (0.5ee/dish for E10.5 and 0.05ee/dish for E11), head (0.5ee/dish for E10 and 0.33ee/dish for E11), fetal placenta, YS (0.5ee/dish for E10 and 0.33ee/dish for E11) and 10 4 BM cells/dish were plated in methylcellulose (MethoCult GF M3434, Stem Cell Technologies Inc.) supplemented with 1% PS in 35 mm Petri dishes (Falcon 1008).Dishes were then incubated for 10-12 days at 37 °C, 5% CO 2 in a humidified glass chamber and the different CFU-C types were identified based on their morphology and then counted.CFU-C types are defined as: blast forming unit erythroid progenitors (BFU-E), colony forming unit-granulocyte progenitors (CFU-G), macrophage progenitors (CFU-M), granulocyte and macrophage progenitors (CFU-GM), and the most immature granulocyte, erythrocyte, monocyte and megakaryocyte progenitors (CFU-GEMM).

Counting of Intra-aortic hematopoietic clusters
Wholemount immunostaining and three-dimensional imaging of E10.5 WT and cKO mouse embryos stained for cKit and CD31 were used to count IAHCs.Groups of more than one cKit + CD31 + or CD31+ cell attached to the ventral wall of the dorsal aorta were considered as one cluster.The total number of clusters were counted and compared between WT and cKO embryos.

Single-cell RNA sequencing and analysis
Cell preparation.E11 AGM cells were dissected and dissociated using collagenase type 1 as described above.Single-cell suspensions were washed, and the pellets were resuspended in PBS 10%FCS and 1% PS.Library preparation and sequencing.Cells were processed through the 10× Genomics Chromium Single Cell Platform using the Single Cell 3' Library and Gel Bead Kit v3 (10X Genomics) as per the manufacturer's instructions.In brief, cells were counted and assessed for viability using a Bio-Rad TC20 then loaded at a concentration for the recovery of 7000 cells and processed through the 10X Chromium Controller.Single-cell libraries were obtained according to manufacturer protocol.RNA concentration was obtained using Quibit RNA HS (Thermo-Fisher) and the quality of the libraries was verified using the LabChip GX24 (Perkin Elmer).Libraries were sequenced on an Illumina Novaseq 6000 S1 lane.The sequencing run met Illumina's quality metrics.Data were demultiplexed using 10X Genomics Cell Ranger 'mkfastq' command.Data analysis.Alignment of scRNA-seq data and barcode counting was performed using Cell Ranger 'count' (v3.1.0)with reference dataset mm10/GRCm38-3.0.0.The unfiltered UMI count matrix from Cell Ranger was used as input for downstream analysis following the OSCA Bioconductor workflow 67 .EmptyDrops 68 was used to distinguish cells from droplets predicted to contain only ambient RNA, calling 17,419 (WT) and 19,047 (cKO) cells at the default FDR of 0.1%.Quality metrics (library size, number of expressed features and percentage of mitochondrial reads) were computed for the called cells using the scater Bioconductor package (v1.14.6) 69 .Cells with any quality metric more extreme than 3 median absolute deviations from the median were filtered out; 15,220 (WT) and 17,280 (cKO) cells were retained for downstream analysis.Count matrices for both samples were log-normalised; size factors for all cells were computed by deconvolution 70 .Technical noise was modelled using a Poisson-based trend, serving as a lower bound for the variance of endogenous genes.Dimensionality reduction for denoising was performed with principal component analysis (PCA) using the modelled trend.Shared-nearest neighbour graphs were constructed using the PCA-reduced datasets and used as input to the Walktrap community finding algorithm to obtain cell clusters.These clusters were used as an initial approximation to known cell types; we refined this approximation based on the distribution and expression levels of literature-derived marker genes.For each sample, doublet scores were computed for each cell by performing an in silico simulation of doublet cells from cell expression profiles 71 ; no doublet-driven clusters were detected.Data from both samples were integrated using a fast version of the mutual nearest neighbours method from the batchelor Bioconductor package (v1.2.4) 72 .Projection of all cells into a shared t-SNE space confirmed the consistency of the cell-type annotations.Differential expression was computed by Wilcoxon rank sum test between appropriate groups using the scran Bioconductor package (v.1.18.7) 70 .Genes with FDR < 0.05 were considered to be significantly differentially expressed.Enrichment analysis with the PANTHER classification system 73 was used to identify GO biological processes overrepresented among significantly differentially expressed genes.Fold enrichment of GO biological processes was computed by Fisher's Exact Test with FDR correction; processes with FDR < 0.05 were considered to be significantly enriched.We used AmiGO (v.2.5.13) 74 to perform additional data mining of GO terms.Prediction of ligand-receptor interactions was computed using the nichenetr (v.1.0.0)R package.We considered genes to be expressed in either WT ECs or PC/vSMCs if they had detectable expression in at least 25% of cells of that cell type.The top 10 ligands ('prioritised ligands') expressed in ECs were chosen based on Pearson correlation coefficient; target genes expressed in PC/ vSMCs were inferred using these ligands.Ligand-receptor network inference was used to predict and score potential interactions.Cell lineage inference was performed on WT NG2 + Runx1 + cells using the Slingshot Bioconductor package (v.2.2.1).We defined a discrete cluster of NG2 + Runx1 + Acta2 -cells as the cluster of origin.For pseudotime plots, cells were ordered according to the principal curve defined by Slingshot and log normalised expression of each cell plotted.Trend lines and standard error bounds were computed by loess local polynomial regression fitting.

Statistical tests
GraphPad Prism 7 was used to perform all statistical tests.Normal distribution was assessed using the Shapiro-Wilk test.Based on data distribution, either one-way ANOVA (parametric) or Kruskal-Wallis (non-parametric) with Tukey's and Dunn's post-hoc test, respectively, were used.When required, unpaired t test was used.

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

Fig. 2 |
Fig. 2 | AGM PC/vSMCs have a distinct transcriptome from endothelial and hematopoietic cells.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 e CD146/CD31/DAPI and f CD146/αSMA/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.

Fig. 3 |
Fig. 3 | NG2 + Runx1 + AGM cells display a sclerotome-derived vSMC transcriptomic profile.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 Cspg4and 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 d WT NG2 + Runx1 + Acta2 + and e NG2 + 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.

Fig. 4 |
Fig.4| The genetic programme of PC/vSMCs is altered in NG2-Cre;Runx1 fl/fl cKO E11 AGM. 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 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.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

a
Endothelial cell morphogenesis Cellular response to fluid shear stress Regulation of cellular response to hypoxia Negative regulation of stem cell differentiation Establishment of endothelial barrier Vascular associated smooth muscle cell differentiation Endothelial cell differentiation Epithelial to mesenchymal transition Positive regulation of telomerase activity Response to transforming growth factor beta Endothelial cell migration Vasculogenesis Erythrocyte homeostasis Regulation of membrane permeability Regulation of cell shape Angiogenesis Regulation of smooth muscle cell migration Erythrocyte differentiation Myeloid cell homeostasis Cellular response to tumor necrosis factor Regulation of phosphatase activity Cellular response to interferon-gamma Blood vessel development Regulation of actin cytoskeleton organisation Regulation of cell-matrix adhesion Regulation of intrinsic apoptotic signaling pathway Establishement or maintenance of cell polarity Stem cell population maintenance Regulation of smooth muscle cell proliferation Myeloid cell differentiation Response to oxydative stress Mesenchyme development Regulation of leukocyte migration Cellular response to cytokine stimulus Regulation of protein serine/threonine kinase activity Circulatory system development Regulation of hemopoiesis Hematopoietic or lymphoid organ development Immune system development Regulation of MAPK pathway Regulation of cell communication /doi.org/10.1038/s41467-024-44913-zNature Communications | (2024) 15:1653

Fig. 5 |
Fig. 5 | Deletion of Runx1 in NG2 + cells results in changes in the extracellular matrix gene profile.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 in e are shown in WT ECs in both scRNA-seq (f, EC and HEC/EHT zoom and t-SNE plots) and bulk RNA-seq postsort (g, TPM).TPM: transcript per Million mapped reads values.

Fig. 6 |
Fig. 6 | The genetic programme of cKO IAHCs is different from that of WT 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).