Dorsal aorta polarization and haematopoietic stem cell emergence

The dorsal aorta (DA), the first and largest blood vessel to be formed during embryonic development, has two main crucial functions. First, it receives and circulates blood cells from the heart to the whole developing embryo, providing oxygen and nutrients, and assuring waste removal (reviewed by Ribatti et al., 2015; Rossant and Howard, 2002). Second, it is also crucial for the generation of haematopoietic stem cells (HSCs), which give rise to all blood cell types during the life of the organism (Orkin and Zon, 2008). Formation of the DA is a complex, multi-step process that is globally conserved between species with very few discrepancies. The ontogeny of the DA is instrumental to our understanding of blood vessel generation and also in deciphering the cellular and molecular processes by which HSCs are formed. In this Review, we highlight the different mechanisms operating to shape the DA in the most commonly used vertebrate models: zebrafish, chicken, mouse and human embryos. We describe how DA generation is tightly controlled in time and space by the surrounding tissues, and how these tissues shape its dorso-ventral polarization. Concomitantly with DA building, we discuss how this polarization influences HSC generation, found in the ventral side of the vessel, and we provide an overview of the molecular signals leading to the formation of the first HSCs.

The mesoderm (i.e. the germ layer that gives rise – among other derivatives – to the vascular and blood systems) arises during gastrulation as a sheet of cells. It then divides rostro-caudally into several medio-lateral components that are the somites, the intermediate mesoderm and the lateral plate mesoderm, which is itself divided into two layers: the somatopleural mesoderm (dorsally) and splanchnopleural mesoderm (ventrally) (Fig. 1A) (reviewed by Prummel et al., 2020; Wymeersch et al., 2021). In the quail embryo proper, the first endothelial cells (ECs) arise in the splanchnopleural mesoderm in contact with the endoderm (Coffin and Poole, 1988; Pardanaud et al., 1987) as cells upregulate the expression of the transcription factors GATA2 and TAL1, and expression of the tyrosine kinase receptor kinase insert domain receptor (KDR) (also known as vascular endothelial growth factor receptor 2, VEGFR2) (Drake and Fleming, 2000; Minko et al., 2003). Commitment of mesoderm into the endothelial lineage is thought to be controlled by soluble signals emitted by the adjacent endoderm. Although not totally decoded, the cascade of events involves the Hedgehog (Hh) family members expressed by the endoderm, fibroblast growth factor 2 (FGF2) and VEGF (Dyer et al., 2001; Goldie et al., 2008; Poole et al., 2001; Sato, 2013). These interactions have mainly been demonstrated by associations between endoderm and mesoderm. In birds, culturing the early mesoderm with the embryonic endoderm results in an appearance of angioblasts, which are immature migrating endothelial cells (ECs) not yet connected into a vascular network; however, in the absence of endoderm, no angioblasts form (Poole et al., 2001). In the mouse, Indian hedgehog (Ihh) is necessary and sufficient to re-specify epiblast into endothelial and haematopoietic cells in embryo explant cultures lacking endoderm (Dyer et al., 2001). FGF2 is capable of inducing angioblasts in mesoderm explants alone (Poole et al., 2001). In addition, isolated epiblast cells treated with FGF2 are capable of developing ECs in vitro (Flamme and Risau, 1992). Immediately after the expression of Hhs and FGF2 in the endoderm, this tissue initiates expression of VEGF (Dumont et al., 1995). Nascent angioblasts emerge at the endoderm-mesoderm interface, upregulate KDR and become extremely sensitive, in a dose-dependent manner, to VEGFA levels. This step is crucial for the formation of the endothelial precursors. VEGFA is a polypeptide growth factor for ECs (Ferrara et al., 2003). The absence of even one allele of the Vegfa gene leads to embryonic lethality in the mouse with impaired vessel formation (Carmeliet et al., 1996; Ferrara et al., 1996). EC development is delayed, and vessel sprouting, remodelling and survival are impaired. Two high-affinity VEGFA receptors, belonging to the receptor tyrosine kinase family, have been identified, fms-related receptor tyrosine kinase 1 (FLT1) and KDR. Deficiency in either VEGF receptor also leads to abnormal embryonic vascular development. A null mutation in Kdr results in a lack of vasculature and very few ECs, suggesting that this gene functions in the differentiation and/or proliferation of ECs (Shalaby et al., 1995). In contrast, mice deficient in Flt1 have excess ECs that are not organized into normal tubular networks (Fong et al., 1995). These findings have been recapitulated in the avian embryo using soluble Flt1 as a loss-of-function approach (Fong et al., 1995) and transfection of VEGF as gain-of-function approach (Flamme et al., 1995), even in the absence of endoderm.

In sharp contrast with the mouse embryo, disruption of the VEGF signalling pathways in zebrafish does not display the same severe phenotype on vascular emergence, indicating a different molecular regulation and factor hierarchy for angioblast commitment (Gering and Patient, 2005; Habeck et al., 2002; Lawson et al., 2002). Defective aorta formation or anomalies of intersomitic vessels are observed, but ECs are specified and the global vessel patterning remains unchanged (Jin et al., 2005). Taken together, two different signalling pathways are acting to specify angioblasts. In amniotes (reptiles, birds and mammals), angioblasts are determined following a Shh-FGF-VEGF axis originating from the endoderm. In anamniotes, angioblasts emerge following determination mechanisms yet to be defined, where the VEGF/KDR axis is dispensable (Habeck et al., 2002; Lawson et al., 2003, 2002; Sumanas and Lin, 2006). Of note, the downstream genes are globally conserved between the two vertebrate groups.

Once specified from the mesoderm, angioblasts undergo vascular formation. Vascular development occurs through two successive and distinct mechanisms during ontogeny. First, in vasculogenesis, angioblasts emerge from pre-existing avascular zones in the mesoderm. They form a primitive vascular network wherein the vessels are of uniform size (reviewed by Risau and Flamme, 1995). Subsequently, this primitive vascular network undergoes remodelling and sprouts new vessels from pre-existing vessels to form a network with vessels of different sizes – a mechanism called angiogenesis. During angiogenesis, mature ECs can revert to a more immature stage (e.g. angioblasts) to contribute to new vessels (reviewed by Carmeliet, 2000). In the next section, we review the various aspects of aorta formation in the different vertebrate species studied so far and highlight the different strategies retained during evolution.

In zebrafish embryos, a continuous series of cells located in the posterior lateral mesoderm migrates towards the midline to form a temporary structure called the inner cell mass (ICM), which is composed of primitive red blood cells and angioblasts (Fig. 1B) (Fouquet et al., 1997; Gering and Patient, 2005; Gering et al., 1998). Angioblasts fated to form the DA reach the midline first, followed by another cohort of primitive haematopoietic red blood cells and then by a population of angioblasts fated to give rise to the central vein (Jin et al., 2005; reviewed by Swift and Weinstein, 2009).

Work in zebrafish and frogs has shown that this convergence is mediated by the Shh-Vegfa signalling axis (Cleaver and Krieg, 1998). Shh is expressed in the floor plate of the neural tube and the notochord, and Vegfa is transiently produced by the hypochord, an endodermally derived structure that lies immediately ventral to the notochord in the amphibian and fish embryos (Fig. 1B) (Cleaver and Krieg, 1998). Of note, Vegfa isoforms differentially regulate blood and vessels, as the short isoform is required for vascular formation, whereas the mid/long isoforms are necessary for HSC production in the aorta (Leung et al., 2013). At the 10-somite stage (ss) (14 h post-fertilisation, hpf), dorsal angioblasts migrate to the midline and aggregate to form a vascular cord immediately underneath the hypochord and the notochord. Shortly afterwards, another generation of angioblasts and primitive red blood cells reaches the midline and positions ventral to the DA at close contact with the endoderm. The DA lumenizes by hollowing the axial EC cords, a mechanism that is under the control of an EC-derived secreted factor: EGF-like domain 7 (Egfl7) (Parker et al., 2004; Xu and Cleaver, 2011).

Our understanding of the avian circulatory system was propelled by the discovery of QH1, a monoclonal antibody that specifically recognizes the endothelial and haematopoietic cells of the quail embryo (Pardanaud et al., 1987). This has been instrumental in precisely describing the early steps of vascular formation, in identifying the sites where angioblasts emerge and in following the formation of the vascular network. Using the development of in toto QH1 staining, DA formation was shown to be a progressive and continuous developmental process occurring along the anterior-to-posterior axis of the embryo. The first angioblasts are detected as scattered QH1⁺ cells located from the head fold to the first somite at 1-2 ss (∼22 hpf, based on the quail developmental table; Ainsworth et al., 2010). At 5 ss (∼25 hpf), discontinuous angioblastic cords in close contact with the endoderm become visible from the head to the last formed somite. At 7 ss (∼30 hpf), the angioblasts form a conspicuous cord of interconnected cells that begin to lumenize. The paired DA and the segmented arteries are formed by 12 ss, while the capillary plexus prefacing the vitelline arteries are being formed (Coffin and Poole, 1988; Pardanaud et al., 1987; Poole and Coffin, 1988). Complementary to whole-mount observations, elegant quail-into-chick transplantation experiments have demonstrated that the splanchnic mesoderm, a lateral plate mesoderm derivative, gives rise to ECs of the native DA (Poole and Coffin, 1989). The two aortas, which are located underneath the lateral plate mesoderm and line the endoderm layer, then progressively migrate to the midline of the embryo to eventually fuse into one single vessel starting around ∼60 hpf in the anterior part of the embryo (Garriock et al., 2010; Jaffredo et al., 2010; Wiegreffe et al., 2007). This process appears to be conserved between vertebrate amniotes.

Several key molecular players are crucial during DA formation and fusion. Similar to zebrafish, HH and VEGF signalling pathways from the notochord and endoderm are attractive signals for EC migration (Garriock et al., 2010). Furthermore, expression of CHORDIN [a bone morphogenetic protein (BMP) antagonist and a major vascular inhibitor] by the chick notochord is downregulated, while at the same time, SHH and VEGF expression is sustained during the fusion process (Bressan et al., 2009; Garriock et al., 2010). Conversely, BMP (BMP1, BMP2 and BMP4) transcripts are detected within the lateral and ventral structures (the lateral plate and endoderm) in the early chicken embryo (Lempereur et al., 2018) and promote EC differentiation and EC assembly (Reese et al., 2004). This fine orchestrated combination of positive and negative regulators coming from the notochord and the endoderm acts as an anterior-to-posterior wave of downregulation of vascular inhibitors, coupled with persisting positive vascular factors, enabling DA fusion/formation.

This aorta formation appears similar, albeit shorter in time, in the mouse embryo: angioblasts are visible in close contact with the endoderm as early as 1-2 ss [embryonic day (E)8]. At 3-5 ss, these angioblasts have formed two continuous cords that progressively lumenize. Lumenization is complete by the 6-8 ss (E8.25) (Strilic et al., 2009) and fusion starts around E9.5, also progressing along the anterior-to-posterior axis of the embryo (Drake and Fleming, 2000; Strilic et al., 2009). ECs start expressing Kdr and the transcription factor Tal1, one of the earliest genes expressed by the angioblasts, around E8. The fusion of the two aortic anlagens then follows, which is when ECs express the endothelium- and haematopoietic-specific markers CD34, platelet endothelium cell adhesion molecules (PECAMs) and the vascular endothelium-specific cell-cell adhesion molecule VE-cadherin. TEK, the receptor for angiopoietin 1, is the last marker to appear (Drake and Fleming, 2000).

In the human embryo, the first probable signs of angioblast formation are reported around 19 days post-conception (dpc, Carnegie stage 7), defined as cells close to the endoderm expressing angiotensin-converting enzyme (ACE, also known as CD143) (Sinka et al., 2012). However, these cells do not express CD34 (Sinka et al., 2012). At this stage, the neural groove has not yet closed and the embryo is flat, resembling an avian embryo at the same equivalent developmental stage. The first conspicuous angioblastic cords are detected at 5-6 ss, in close contact with the embryonic endoderm. These cells express KDR mRNA and, to a lesser extent, the CD34 antigen. The paired aortae are formed by 15 ss. They are separated by the notochord, closely associated with the endoderm, while a continuum of vascular structures, lateral to the aortae, line the endoderm (Cortes et al., 1999; Labastie et al., 1998). At around 28 dpc, the notochord begins to separate from the endoderm and the paired aortae undergo fusion, which is completed by 25-26 dpc. Aortic haematopoiesis then occurs between 27 and 42 dpc (Tavian et al., 1999, 1996). Based on gene expression analysis and experimental approaches, it is clear that both chick and mouse are following the same rules for angioblast formation and the first steps of aorta genesis. The human embryo, albeit not investigated at this stage, is likely to follow similar, if not identical, regulation.

In zebrafish, as detailed above, the DA forms a unique vessel from angioblasts migrating from the lateral mesoderm to the midline. Recent data indicate that tissue rearrangement, in particular the separation of the notochord from the endoderm, creates a ventral midline cavity that facilitates angioblast migration and the subsequent aggregation of angioblasts and axial vessel formation in between the notochord and the endoderm layer (Paulissen et al., 2022). Once formed, the DA faces two different embryonic structures, the notochord (dorsally) and the posterior cardinal vein (ventrally) (Fig. 2A). The DA is likely receiving a gradient of Hh signals from the floor plate of the neural tube and the notochord, with Hh signals decreasing from the dorsal to ventral wall, the latter of which is also exposed to higher BMP signalling (Gering and Patient, 2005; Wilkinson et al., 2009). The Hh gradient is required for the continued expression of tbx20, a member of the T-box transcription factor family, in the roof of the aorta, i.e. close to the source of Hh (Gering and Patient, 2005), which is therefore polarizing the DA in a dorsal-ventral manner. Of note, expansion of the Hh signal results in the expansion of the tbx20 signal to the aortic floor (Wilkinson et al., 2009).

In amniotes, the DA is formed in sequential steps that require fusion of the nascent paired aortae (Fig. 2B-D). During its formation, the two vessels are facing changing environments. At first, the paired aortae are located laterally, lining the endoderm, underneath the intermediate mesoderm and somite, separated by the notochord. During their migration to the midline, they are progressively facing the notochord to ultimately fuse into one single vessel, with the notochord placed on the dorsal side of the DA. Interestingly, experimental approaches on the avian embryo have shown that the notochord plays a key role in aorta fusion by preventing maturation of the aortic ECs, a mechanism mediated by noggin and chordin, two potent BMP inhibitors (Reese et al., 2004). In addition, notochord-derived noggin and chordin also impede angioblast commitment and vascular structure formation in the mesoderm immediately adjacent to the notochord, creating a central EC-free zone in the embryo (Bressan et al., 2009). As the embryo develops, the notochord progressively separates from the endoderm and, at the same time, the BMP antagonistic effect of the notochord diminishes, allowing paired aortae to converge and ultimately fuse (Garriock et al., 2010). At the end of fusion, the DA faces the somites on its dorso-lateral side, the gonad/mesonephros on the ventro-lateral side and the mesenchyme on the ventral side (Fig. 2B).

Another important aspect of DA polarization comes from the cellular origin of the ECs. Indeed, in the avian system, the DA is built in two successive steps involving ECs originating from two different mesodermal populations. Quail-into-chick transplantation experiments nicely demonstrated that different mesoderms can generate ECs (Pardanaud et al., 1996). The paraxial (somitic) mesoderm provides ECs for the vascularization of the trunk and limbs, while the splanchnopleural mesoderm provides ECs for the vascularization of the viscera and aorta. A precise and dynamic image of the aortic evolution was obtained using a modified version of the same experimental approach, although differing in two crucial aspects (Pouget et al., 2006). First, the grafted material involved the whole pre-somitic mesoderm (a length of ∼10 future somites) plus the last segmented somite; second, the host embryos were examined at different timepoints. Quail somitic cells invaded the roof of the chick DA around 15 h after grafting i.e. before the fusion of the paired aortae. At 24 h, the aortic floor was derived exclusively from the chicken host (i.e. splanchnopleural), but subsequently became colonized by quail-derived ECs (i.e. derived from the grafted somites). As a result, the endothelium of the definitive aorta at E4-4.5 was entirely composed of somite-derived cells (Pouget et al., 2006). The DA is therefore built in two steps. First, ECs of splanchnopleural origin (ventral) form the two primitive paired aortae (Richard et al., 2013) that will progressively fuse. Before fusion, ECs from the somite (dorsal) replace the roof of the DA. At the time of fusion, the DA roof is made of ECs from the somite, whereas the floor is of splanchnic origin, endowed with haemogenic potential. As haematopoiesis proceeds, haemogenic ECs of the floor undergo endothelial-to-haematopoietic transition (EHT), while somite-derived ECs replace the haemogenic ECs that have been used in haematopoietic production, by the end of which the whole aortic endothelium has been replaced by somite-derived ECs (Pouget et al., 2006) (see Movie 1 for the building of the dorsal aorta exemplified in the avian embryo). This dorso-ventral polarity, in term of cell origin of the DA, is strikingly related to the other main function of the DA, which is to generate HSCs, as the generation of HSCs occurs only at the ventral side of the DA (discussed in detail in the next section) (for a review, see Weijts et al., 2021).

Is the chick embryo unique in its aorta construction and patterning? EC generation by mouse somites has been described using a modification of the Kdr knock-in mouse model displaying the same pattern as described for the avian model (Ema et al., 2006). When mouse somites were grafted into chick embryo hosts, they assembled into a perfectly patterned and functional vascular network (Ambler et al., 2001; Yvernogeau et al., 2012), and colonized the chick aortic roof; however, their contribution to the floor was unclear (Yvernogeau et al., 2012). Whether significant differences in aorta assembly and/or differences in molecular control of dorso-ventral patterning are responsible will have to be specified.

Lineage-tracing experiments have shown that ECs forming the paired aortae at E8.5-E9 are derived from Hoxb6⁺ mesoderm, i.e. from the lateral plate (Lowe et al., 2000) and not from the somites, suggesting a splanchnopleural origin of the paired DA (Lowe et al., 2000; Wasteson et al., 2008). This suggests that, as observed in the avian system, the nascent DA originates from the splanchnic mesoderm. After DA fusion (E.9.5 onwards), Hoxb6-derived ECs are observed in a more mosaic pattern, dispersed in the DA, suggesting that some ECs have been replaced (Wasteson et al., 2008). Based on these observations, ECs from the mouse DA appear to follow the same two-step process formation as the one observed in the avian model.

Interestingly, zebrafish somites also contribute, at least to some extent, to aorta remodelling. Using different fate-mapping strategies targeting the somite, one study showed that somite-derived ECs reach the aorta and activate the HSC program (Nguyen et al., 2014). For example, using the Kaede photoconvertible or somite-specific mesogenin (msgn) promoter, they identified the middle part of the somite as the source of ECs that contribute to the DA. However, these somite-derived ECs never contribute to HSC generation within the aorta. These observations have recently been confirmed and extended using a combination of molecular and genetic approaches (lineage tracing) showing that the ECs from the dermomyotome – the dorso-lateral aspect of the somite – incorporate the developing DA, acting as a supportive niche for haematopoiesis (Sahai-Hernandez et al., 2020 preprint). Whether all ECs of the DA are replaced by somite-derived ECs is yet to be determined.

The tight association of the DA with dorso-ventral polarity is also crucial for generating the first definitive HSCs. HSCs are characterized by two main properties: they are able to self-renew to maintain the pool of HSCs during the life of the individual and they are multipotent, capable of producing all haematopoietic cell types (reviewed by Orkin and Zon, 2008). The first observation of haematopoietic clusters (‘haemoblasts’) attached to the endothelial layer of the aorta was reported more than one century ago in the avian embryo (Dantschakoff, 1907; Jordan, 1917). These groups of haematopoietic cells were later designated as intra-aorta haematopoietic cluster cells (IAHCs) and are a common feature of most vertebrate species (Dieterlen-Lievre et al., 2006; Dzierzak and Bigas, 2018; Weijts et al., 2021), the fish being an exception, as haematopoietic cells emerge as a single cell (reviewed by Weijts et al., 2021) (see Box 1 and Fig. 2).

In the avian system, IAHCs protruding into the lumen of the DA are tightly attached to the endothelium and share a number of markers with ECs. Chick IAHCs are restricted to the aortic floor and are visible soon after fusion of the two bilateral aortic anlagen (Jaffredo et al., 2005). IAHCs are derived from HECs that are undergoing EHT (Kissa and Herbomel, 2010). The transcription factor Runx1 is crucial for EHT because, in its absence, no IAHCs and no HSCs are formed (Cai et al., 2000; Chen et al., 2009; Lacaud et al., 2002; Wang et al., 1996). During EHT, IAHCs are quickly distinguished from ECs by the expression of transcription factors MYB and RUNX1, the pan-leukocytic CD45 antigen, and by the disappearance of KDR and CDH5 (encoding VE-cadherin) (Bollerot et al., 2005; Jaffredo et al., 2005; Jaffredo et al., 1998). Transplantation assays performed in mouse and chick clearly indicate that IAHCs do contain the first HSCs, which are capable of long-term multilineage haematopoietic reconstitution (De Bruijn et al., 2000; Medvinsky and Dzierzak, 1996; Muller et al., 1994; North et al., 1999, 2002; Yokomizo and Dzierzak, 2010; Yvernogeau and Robin, 2017).

In the avian embryo, the haematopoietic clusters appear soon after the two distinct original aortic anlagen fuse to form the DA. Interestingly, RUNX1 expression is not constitutive but is initiated in the lateral part of the paired aortae and progressively reaches the most ventral part of the aorta concomitantly with the fusion of the vessel (Arraf et al., 2017; Richard et al., 2013). RUNX1 expression is under the control of the formation of the sub-aortic mesenchyme that originates from the proximal splanchnic mesoderm. Contact between the sub-aortic mesenchyme triggers the haematopoietic programme in DA ECs.

Despite few discrepancies in the repartition/generation of haematopoietic cells within the DA observed among species, a common feature is conserved: HECs/HSCs are generated solely by the floor of the aorta (Weijts et al., 2021; Yvernogeau et al., 2020). Whether the haemogenic EC competence/potential is acquired through intrinsic (cell origin) or extrinsic (environment origin) factors, or a combination of both, is a fascinating, yet unanswered question.

One of the key elements involved in DA polarity is the tissue underneath the aorta. Mesenchymal cell lines derived from the aorta-gonad-mesonephros (AGM) region carry a potent HSC supportive capacity, leading to the identification of several HSC regulators (Durand and Dzierzak, 2005; Ohneda et al., 2000; Oostendorp et al., 2005; Renstrom et al., 2009). Furthermore, tissues ventral to the aorta have a positive effect on HSC emergence, whereas those of the dorsal aspect decrease HSC production (Peeters et al., 2009; Souilhol et al., 2016; Taoudi and Medvinsky, 2007). This HSC-supportive role has been proposed based on the expression of several key molecules, in particular BMP4, a known potent inducer of ventral mesoderm and blood in Xenopus (Huber and Zon, 1998). BMP4 is expressed in the sub-aortic tissue in the human (Marshall et al., 2000), mouse (Durand et al., 2007), chick (Lempereur et al., 2018) and zebrafish embryos (Wilkinson et al., 2009) (Fig. 3). Inhibition of BMP signalling decreases the number of aorta-associated haematopoietic stem and progenitor cells (HSPCs) (Durand et al., 2007). Furthermore, transforming growth factor β1 (TGFβ1, which includes BMP signalling), is expressed by human IAHCs, suggesting a role in embryonic haematopoiesis (Marshall et al., 2000). Active TGFβ signalling promotes EC differentiation at the expense of haematopoietic cells in an ES cell differentiation model (Vargel et al., 2016) and triggers haemogenic EC commitment and EHT in developing zebrafish, chicken and mouse (Lempereur et al., 2018; Monteiro et al., 2016). This indicates a rather stage-specific role for TGFβ signalling in EC development and HEC commitment. Finally, a recent study highlighted the existence of Runx1⁺ cells located in the ventral mesenchyme of the aorta using transgenic reporter mice that express the Runx1b isoform. These Runx1⁺ mesenchymal cells support haematopoiesis in an aggregate culture system (Fadlullah et al., 2022).

Shh expression in the notochord is highly conserved between vertebrate embryos. In the zebrafish, shh expression from the midline tissues induces VEGF expression in the somites, which ultimately induces EC specification and migration to form the DA (Lawson et al., 2002). Shh is involved in different stages of DA formation, e.g. in the medial migration of endothelial progenitors of the DA, in arterial gene expression and in haematopoietic development (Gering and Patient, 2005). In the avian and mouse models, Shh is not only expressed by the notochord but also by the endoderm (Vokes et al., 2004). As the receptors for Shh [patched 1 and patched 2 (Ptch1 and Ptch2) and smoothened (Smo)] are expressed by ECs of the developing aorta, this signalling pathway appears to be important for EC generation and for the correct migration of ECs during aorta formation (Vokes et al., 2004). Pharmacological inhibition of Hh, e.g. using cyclopamine, induces a downregulation of BMP4, VEGF and KDR expression, demonstrating that Hh signalling acts upstream of these EC-related genes (Moran et al., 2011; Nagase et al., 2006).

Signalling involving BMP, which is part of the TGFβ superfamily, is a crucial pathway involved in processes such as gastrulation and specification of the mesoderm (Mishina et al., 1995; Winnier et al., 1995). BMP4 is expressed ventrally in the mesenchyme of the DA (Durand and Dzierzak, 2005; Wilkinson et al., 2009) and inhibition of its expression, using morpholino knockdown experiments in zebrafish, leads to a complete loss of HSC formation within the aorta (Wilkinson et al., 2009). In mammals, chemical inhibition of BMP signalling in explant aorta cultures reduces the number of functional HSCs, while addition of BMP has the opposite effect (Durand and Dzierzak, 2005), demonstrating a conserved mechanism between vertebrates for HSC specification within the aorta. These observations suggest that BMP4 is required during the first step of HSC specification. Of note, the polarised expression of BMP4, to the ventral side of the aorta, is a conserved feature in human and chick embryos (Lempereur et al., 2018; Marshall et al., 2000).

In the mouse aorta, sub-dissection of the dorsal and the ventral aortic region followed by RNA-sequencing identified Bmper as a BMP inhibitor, which is gradually expressed from the dorsal to the ventral part of the aorta (McGarvey et al., 2017). Immunostaining analysis indicated expression of Bmper in perivascular cells and in the sub-aortic mesenchyme of the aorta. Although Bmp4 appears to be expressed constantly during aorta haematopoiesis (from E9.5 to E10.5), BMPER expression increases during this time window. Bmper is therefore crucial to balance the BMP pathway to provide a sufficient level of signals within the aorta ECs for proper HSC specification (McGarvey et al., 2017).

Notch is a cell-cell interaction signalling pathway involved in cell fate determination processes. The Notch pathway is composed of four Notch receptors, Notch1 to Notch4, two Jagged ligands, Jag1 and Jag2, three Delta-like ligands, Dll1, Dll3 and Dll4, and the nuclear transcription factor RBPjk (for a review, see Kopan and Ilagan, 2009). Notch is crucial with Dll4 for arterial specification, but it is also required, along with Jag1 and Jag2, for haemogenic endothelium commitment and IAHC formation within the aorta (Robert-Moreno et al., 2008). One elegant study indicated that the difference between arterial and haematopoietic fates relies on the activation level of Notch signalling (Gama-Norton et al., 2015). Expression pattern studies on E9.5 and E10.5 mouse embryos demonstrated that Notch1 and Notch4 receptors are expressed in the aortic endothelium but also in IAHCs (Robert-Moreno et al., 2005). Furthermore, the ligands Dll4, Jag1 and Jag2 are expressed both by the ECs of the aorta and by IAHCs, whereas Dll1 and Dll3 are not (Robert-Moreno et al., 2005, 2008). Functional studies show that if Notch2 is dispensable for HSC specification, Notch1 is required from ECs to trigger Hes1 transcription. In turn, HES1 protein represses Gata2 in the emerging HSCs (Guiu et al., 2013; Hadland et al., 2004; Kumano et al., 2003). Mouse embryos mutant for RBPjk lack transcription factors crucial for haemogenic commitment and IAHC formation, such as Runx1, Gata2 and Scl/Tal1 (Khandekar et al., 2007; Minegishi et al., 1999; Robert-Moreno et al., 2005).

In the zebrafish, notch1a and notch1b are both required to specify HSCs, whereas notch3 is required earlier, within the developing somite, and functions downstream of wnt16 through two Notch ligands, dlc and dld (delta-like c and d, respectively) (Clements et al., 2011; Kim et al., 2013). During DA formation, HSC precursors express the junctional adhesion molecule (JAM) jam1a, which mediates Notch signal transduction. These cells migrate axially across the ventral somite, where jam2a, dlc and dld are expressed. These Jam1a-Jam2a interactions facilitate the transduction of requisite Notch signals from the somite to the precursors of HSCs (Kobayashi et al., 2014) (Fig. 3). Altogether, this work indicates that the specification of HE occurs well before the DA is formed in the zebrafish and illustrates well the contribution of the zebrafish model to the fine molecular dissection of haematopoiesis.

In birds, aortic ECs express both JAG2 and DLL4 at the onset of haematopoiesis. This is followed by a downregulation of JAG2 when ECs start expressing RUNX1 and initiate EHT, highlighting the finely tuned regulation necessary for the haematopoietic fate (Richard et al., 2013).

Gata3 is a transcription factor expressed by various tissues, including the kidneys (Grote et al., 2006) and the sympathetic nervous system (Pandolfi et al., 1995). In mouse and chick embryos, Gata3 is spatially and temporally expressed around the aorta at a time of haematopoiesis (Jaffredo et al., 2005; Mascarenhas et al., 2009). It is also expressed by tissues near HSCs on the dorsal aorta, such as the sub-aortic mesenchyme and neurons that are derived from the migrating neural crest cells. Gata3 is required for the generation of the sympathetic nervous system that produces signalling molecules, such as the catecholamines (Lim et al., 2000; Moriguchi et al., 2006; Tsarovina et al., 2004, 2010). In the Gata3 knockout mouse, fewer HSCs are generated during embryonic development (Fitch et al., 2012). Furthermore, addition of catecholamines in Gata3^−/− AGM explant culture rescues HSC defects, demonstrating a role of Gata3 in HSC specification via catecholamines produced by the sympathetic nervous system (Fitch et al., 2012). Of note, a recent study demonstrated that Gata3 is also expressed in a specific, quiescent subset of HECs of the mouse embryo and is required for EHT and to obtain proper HSC production (Zaidan et al., 2022).

Platelet-derived growth factors (PDGFs) are regulatory factors that control cell growth and proliferation, acting through protein tyrosine kinase receptors, PDGFRα and PDGFRβ (reviewed by Andrae et al., 2008). PDGF signalling is involved in EC differentiation (Rolny et al., 2006) and also functions in HSC development. In the zebrafish, pdgfb promotes HSC production by increasing inflammatory molecules such as il6 and/or il6r (Lim et al., 2017). Pdgf signalling is also required for proper migration of trunk neural crest cells that eventually contact the DA, a requirement for HSC production (Damm and Clements, 2017). In the mouse, Pdgfrb is expressed around the aorta in perivascular stromal cells, creating a niche that supports HSC generation (Sá da Bandeira et al., 2022). This study also proposes, based on lineage tracing, a contribution of Pdgfrb⁺ precursors to DA ECs and HSCs. Recently, mesoderm-derived Pdgfra⁺ stromal cells were also shown to transiently contribute to the haemogenic endothelium of the DA (from E10.5 to E11.5) (Chandrakanthan et al., 2022).

Inflammatory mediators are also involved in HSC generation within the embryonic aorta. In the mouse, interleukin 1 (IL1) is an important regulator of aorta HSC development by limiting the differentiation of HSCs along the myeloid lineage (Orelio et al., 2008). Il1 and its receptor Il1r1 are expressed at the mid-gestation stage of the mouse embryo, around the AGM, with Il1r being concentrated at the ventral side of the aorta in ECs and mesenchymal cells (Orelio et al., 2008). Exogenous addition of Il1 to AGM explant cultures promotes HSC differentiation through a myeloid lineage, indicating that Il1 and Il1r are important regulators of proper generation of HSC in the aorta (Orelio et al., 2008).

Il3 is expressed by non-haematopoietic and non-ECs of the AGM, while its receptors are expressed by the ECs and haemogenic ECs of the aorta (Robin et al., 2006). A role of Il3 in HSC proliferation and survival was also proposed, as exogenous addition of IL3 on AGM explant cultures haploinsufficient for Runx1 can rescue the number of HSCs generated (Robin et al., 2006).

In zebrafish, pro-inflammatory signals via the tumour necrosis factor (TNF) pathway play an important role in HSC generation (Espin-Palazon et al., 2014). TNFα is a powerful pro-inflammatory cytokine, which exerts its effect after interacting through specific receptors (TNFRs). TNFα, signalling through TNFR2, is an activator of the transcription factor nuclear factor-κB (NF-kB). In this aspect, tnfa produced by neutrophils acts through tnfr2 receptors expressed on ECs of the DA. This activates Notch signalling and promotes HSC specification (Espin-Palazon et al., 2014).

In the mouse and fish embryos, interferon γ (IFNγ) or IFNα signalling is crucial for HSC generation in the aorta (Li et al., 2014). Embryos lacking IFNs show a decreased number of HSCs. Although multiple cellular sources of inflammatory cytokines exist in the developing embryos, the myeloid population might contribute to the local inflammatory response, impacting the generation of aorta-derived HSCs (Li et al., 2014).

Other regulators or signalling pathways act to produce HSCs from the aorta. Using microarray transcriptomic approaches applied to different AGM regions collected at E9 and E11, i.e. before and during the acquisition of a HSC potential, respectively, p57Kip2 and insulin-like growth factor 2 (IGF2) were identified as important regulators (Mascarenhas et al., 2009). p57Kip2, a cyclin-dependant kinase inhibitor implicated in controlling cell proliferation (Passegue et al., 2005; Santaguida et al., 2009), was detected preferentially in the middle region of the aorta, in the ventral mesenchyme and in the nephrogenic chords. IGF2 has a wider expression pattern, but is more highly expressed around the aorta (Mascarenhas et al., 2009). Another study used a combination of ex vivo and in vivo approaches, in which different tissues of the aorta were co-cultured [the ventral and/or dorsal part of the aorta with or without urogenital ridges (UGR)] to evaluate their influence in HSC generation (Souilhol et al., 2016). They reported reciprocal dorso-ventral inductive interactions and lateral input from the UGR, which drove HSC development. More precisely, they demonstrated that Shh, stem cell factor (SCF) and BMP inhibitory signals were produced in the dorsal region of the AGM, the ventral and lateral UGR regions, and ventral tissue, respectively. These signals are integrated to orchestrate the proper induction and/or instruction of DA HECs to become HSCs (Souilhol et al., 2016).

Deciphering the molecular aspects of the human embryo aorta niche is a major challenge due to the availability of early human embryos. The first emergence of IAHCs occurs at early stages (around day 27; Ivanovs et al., 2017), which limits their availability. However, a recent study performed laser capture microdissection on transverse cryosections of human embryos to isolate the different layers of cells that surround the ventral and dorsal sides of the aorta, and to isolate their respective identities. This approach highlights the cardiac epidermal growth factor (EGF) and its major receptor, endothelin 1 (expressed and secreted by ECs), as potent enhancers of HSC generation in human embryo aorta (Crosse et al., 2020).

A tomography-sequencing (tomo-seq) technique was developed to go deeper into the spatial information (Junker et al., 2014). This approach was used on embryo slices and/or trunks of zebrafish, chick, mouse and human embryos to compare gene expression profiles of the ventral mesenchyme at two different developmental timepoints (Yvernogeau et al., 2020). This study highlighted adrenomedullin (ADM), a hypotensive and vasodilator agent, and its receptor RAMP2 as new regulators of HSC emergence in the aorta. Additionally, sushi, Von Willebrand factor type A, EGF and pentraxin domain containing 1 (SVEP1), a secreted extracellular factor that is crucial for proper lymphangiogenesis (Karpanen et al., 2017), is also a regulator of IAHC cellularity and HSC production in the aorta (Yvernogeau et al., 2020) (Fig. 3). Recently, a combination of scRNA-Seq, spatial transcriptomics and immunofluorescence approaches in human embryo identified a six-gene signature (RUNX1⁺HOXA9⁺MLLT3⁺MECOM⁺HLF⁺SPINK2⁺) that discriminates AGM-derived nascent human HSCs from lineage-restricted progenitors (Calvanese et al., 2022).

It is now extensively acknowledged that HSCs have an endothelial origin; a trait conserved among species, with the RUNX1 transcription factor playing a key role in this molecular control. However, how HSCs become specified within the aorta remains an unanswered question. Does it obey a stochastic mechanism or is it something more subtle that escapes our scrutiny? Another pending question is the multiple natures of the HECs within the aorta. Do several types of HECs generate different haematopoietic fates even in the same part of the DA and, if so, how are these different fates determined? We are at the very beginning of the fine dissection of this phenomenon and these questions are crucial if one wants to produce HSCs ex vivo from non-haematopoietic sources.

The powerful, high-throughput technologies, such as sc-RNA-sequencing, together with emerging approaches of spatial transcriptomics and proteomics, will allow us to go one step further. One of the biggest challenges will be to integrate all these ‘omics’ datasets generated during aorta formation in different species to identify how all these fine-tuned molecular regulators are integrated both in time and space.

The emergence and fast development of new technologies, such as the ‘gastruloid’ or ‘organ-on-chip’ (for reviews, see Low et al., 2021; van den Brink and van Oudenaarden, 2021) systems are promising tools to go one step further in dissecting the molecular regulation controlling HSC emergence. A recent study using a modified protocol to generate mouse gastruloids demonstrated the possibility of obtaining blood vessels and associated haematopoiesis that phenotypically and functionally resemble that described during mouse aorta development (Rossi et al., 2022). Transferring and adapting this technology to human PSCs will certainly improve our knowledge in decoding the mechanisms leading to early aorta and blood vessel formation, a process that is difficult (even impossible) to tackle in the human embryo. The organ-on-chip technologies could provide insights into the normal functioning of organs, such as blood vessels, and could be an efficient way to test drug safety and efficacy, a pre-requisite before progressing to clinical trials. We are convinced that the different approaches that research groups worldwide are undertaking will feed our knowledge and converge to help produce HSCs for curative purposes.

We thank all members of the Jaffredo laboratory, past and present, for their contributions. We also apologize for omitting research due to space constraints. We thank Pierre-Yves Canto for providing the human cryosections. Sophie Gournet is acknowledged for the design of Movie 1 and help in figure formatting.